fl. B. 3eJii>;ioBii’i, H. M.

Htjiom
BBICIIIAH MATEMATHKA JJJIH HAMHHAIOIIi;HX
(DH3HKOB H TEXHHKOB
ITa^areJiiiCTBo «HayKa» MocKBa
 )

Ya.B.Zeldovich and I.M.Yaglom

( Mosdy Physicists and Engineers

Mir Publishers Moscow

Translated from the Russian
by Eugene Yankovsky

First published 1987

Revised from the 1982 Russian edition

R(l mzJiuucKOM H3bine

Printed in the Union of Soviet Socialist Republics

© H3AaTejibCTBO «HayKa», 1982

© English translation, Mir Publishers, 1987
Foreword

+
To h?e’ JOTit u~<£ t retiTp L^ui^a^piiys- Thsey "xSlten ran ~erf±i^roy
icist and a mathematician to write an themost modern ideas that have sprung
entirely new type of book for future sci- from the depths of abstract mathema-
entists and engineers. This calls for a tics or even to devise their own new
few words about its purpose and scien- mathematical methods. But the ulti-
tific ideology. mate result of the investigations of a
Many physicists, as is known, are physicist must always be effective. It
displeased with the presentation of the must be expressed as a number or for-
basics of mathematical analysis that mula relating to the quantities ob-
was introduced by mathematicians in served. The physicist feels it is unimpor-
the first half of the 20th century. The tant to justify and establish the logi-
reasons are understandable. The devel- cal structure of the foundations: he has
opment of mathematical research con- faith in mathematics and does not doubt
nected with the elaboration of the log- that its methods contain no internal
ical foundations of our science left contradictions. Theoretical physicists
its imprint on the style of presentation are not satisfied with the existing text-
even of the very first chapters of text- books and popular scientific literature
books for beginners. Exact definitions on mathematics written by mathemati-
of such concepts as a real number, lim- cians themselves (especially by those
it, and continuity were the result of a who are far removed from physics)
prolonged and very nontrivial logical they feel it necessary to make known to
analysis of theories that were already the beginner their own concepts of how
created and were intuitively clear (on a scientist should use the mathematical
the scientific level of rigor). These de- apparatus and in what simple way it is
finitions,which are not at all simple for possible to grasp, “to a first approxima-
the beginner, came to be used in the tion,” the methods he will have need
wrong context. Textbooks presented of. It seems to me that their tremendous
them before any explanation was given mathematical experience gives them
of the theory and its applications, there- this right.
by complicating an understanding of Of course, such a view of mathemat-
things that were intuitively clear. ics is somewhat one-sided, but surely
It must be borne in mind that theo- there is no harm in that. Those who
retical physicists make extensive use go on deeper into mathematics will be
in their research of all, or nearly all, able better to grasp the meaning of the
the methods of classical mathematics. foundations of mathematics since they
6 Foreword

will already have an intuitive under- or her work by mastering its methods
standing of the essence of the matter without going into a full logical sub-
and will have developed the techniques stantiation of them, allowing the stu-
of problem solving. As I see it, many dent to view mathematics as a section
courses in mathematics are spoiled by of natural sciences and to solve as many
the desire to start the teaching of a concrete problems as possible.
new theory by confronting the student The previous book had a number of
with the foundations and the formal technical drawbacks. Much has been
language. As a mathematician who has done in this book to eliminate the short-
worked with physicists for many years, comings and improve the presentation.
I take the liberty to express the opinion Considerable new material has been
that it would be advisable not only for added, and the text has been carefully
future physicists and engineers but also edited. All this work was done jointly
for future professional mathematicians by the two authors: the physicist Acad-
(and their teachers) to learn to under- emician Ya. B. Zeldovich and the
stand better the style of mathematical mathematician I. M. Yaglom, Doctor
thinking of the natural scientist and to of Physico-Mathematical Sciences.
grasp the problems that confront him. It seems to me that the book achieves
This text is the result of a great effort its objective. I would confidently
to improve the book Higher Mathemat- recommend it to the thinking young
ics forBeginners and Its Application to reader who wishes to master the methods
Physics by the well-known Soviet phys- of higher mathematics informally and
icist Academician Ya. B. Zeldovich. to apply them to problems in physics
The purpose of this book is to enable and technology.
the future physicist (chemist, engineer,
etc.) to use higher mathematics in his Academician S. P. Novikov
Note to the Reader

There are probably not many people The traditional methods of teaching
who have never heard of Pushkin or higher mathematics have complicated
Leo Tolstoy. But very many grown-up matters. It would seem simple to under-
people, one-half or even nine-tenths, stand “Crime and Punishment” or
would not be able to explain what a “War and Peace” and to grasp the spher-
derivative or an integral is. Yet without icity of the earth or the atomic structure
these concepts it is impossible to de- of matter, yet difficult to master
scribe or investigate the variables or differentiation or integration. Students
functions that characterize the depen- of the first half of the 20th century were
dence of quantities on one another. The terrified by the theory of limits and the
laws of nature are formulated in the language of infinitesimals, which is so
language of higher mathematics, the unlike ordinary arithmetic and school
language of derivatives and integrals. algebra. When higher mathematics was
It’s our view, though we are not im- introduced in the traditional way in the
partial, that higher mathematics is as first year of study at institutions of high-
beautiful as the poetry of Pushkin and er learning, it proved to be the main
as profound as the prose of Dostoevsky subject that caused students to drop
or Tolstoy. But is higher mathematics out. The only way to change the atti-
seen by the general public as a cultural tude of the student toward this subject is
asset? Has it really entered the public to change the way it is taught. Higher
consciousness as such? Now that mo- mathematics must be transformed from
dern science has greatly extended the a dry and. difficult subject into a set
boundaries of the visible world by dis- of clear and natural concepts that open
covering the secrets of the structure of the way to the study of physics, chemi-
the atom and the many mysteries of stry, and engineering disciplines.
the starry sky, the concepts of deriva- The first problem confronting the au-
tive and integral have become necessa- thors of this book was to give the stu-
ry elements of general culture. And in dent an understandable introduction to
everyday life, too, it is quite useful higher mathematics unencumbered by
to understand the rate of change of a an unwieldy apparatus or logical sub-
quantity (the derivative) and the over- tleties. We believe that anyone acquaint-
all effect of the action of a factor (the ed with the basics of arithmetic and
integral). This expands one’s outlook who had above-average marks in algeb-
and can be used in a great variety of ra in secondary school will find it easy
situations. (and, we hope, interesting) to read this
8 Note to the Reader

book. We seek for a friendly student who bound up with his development of the
has no doubts but believes and wants to foundations of theoretical (we can also
take up this book with a view to learn- say mathematical) physics. He saw
ing something new, undeterred by the the concepts of higher mathematics as
fact that at times the authors offer a (literal!) translation of the basic con-
simple examples in place of “very sien- cepts of mechanics into mathematical
tific” theory. language. 1
The book contains many concrete The first chapters of this book give
examples (perhaps even too many for an idea both of the essence of higher
some). There are calculations and prob- mathematics and of its possible appli-
lems connected with natural phenom- cations. So we can imagine a reader who
ena. We hope that these examples will wishes to restrict himself to a selective
make. iL easier, tor. tba sindnoT to. bis, stjuiy ai Chastens, 1 fsi 7- Hmt It was, mat
studies. If we had expected the reader to for him that we wrote this book. To
be given to arguing, we would have writ- grasp the basics of higher mathematics
ten it in a different way. Perhaps for it is necessary to have a good under-
such a reader the standard presentation standing of how this apparatus is used.
with all the usual reservations would Knowledge not used is not real knowl-
have been more suitable, but we have edge. It is grasped with difficulty and
no wish to lose a hundred possible easily forgotten. This book is intended
friends just because of a single one who for people especially interested in phys-
likes to argue. ics and engineering. For this reason
This book is intended for beginners, the applications are largely taken from
that is, for high-school students in these areas. It would be possible today
the upper grades, students of trade to compile a textbook of higher mathe-
schools and vocational schools, and stu- matics for a biologist (or future biolo-
dents in the first years of college. We gist), geologist, or economist, but that
also have in mind anyone who by him- would be an entirely different book and
selfwishes to become better acquainted other people must write it.
with higher mathematics, say, people The second part of this book deals
who finished school some time ago. with a number of topics from physics
The place of mathematics in life and and engineering. In these areas the pow-
in science is determined by the fact er of higher mathematics shows itself
that it permits one to translate an eve- to the full. A great many highly impor-
ryday intuitive approach to reality based tant physical phenomena can be de-
on purely qualitative and hence ap- scribed with a degree of completeness that
proximate descriptions into the lan- is unattainable without the use of a de-
p^i-ayp. rvLoxart t ions^jxcL formulae ri y aJtiy uiteprAL. By, way, of .illu-
from which quantitative conclusions stration we can mention the theory of
can be drawn. So true it is that the radioactivity (Chapter 8). Another ex-
scientific level of any discipline is de- ample is the phenomenon of resonance ,
termined by the extent to which it uses which in this book appears in two differ-
mathematics. But the real language of ent forms, as one of the effects appear-
science is not at all elementary algebra
or geometry. Higher mathematics
1
plays a far greater role. It studies varia- It is also no accident that the other in-

bles and processes for which algebra or ventor of higher mathematics, Leibniz, worked
out evolutionary conceptions that were
geometry do not suffice. It also inves- closely linked to his mathematical ideas and
tigates other more complicated divisions were considerably ahead of their time. These
of mathematics which are only touched conceptions touched on biology, geology, and
on in this book. It is no accident that the physics; for instance, Leibniz conceived of
the concept of energy, a quantity that trans-
development of differential and integral forms from one form to another but never vani-
calculus by Newton was intimately shes in physical phenomena.
 Note to the Reader 9>

ing in the theory of mechanical oscil- with many seemingly unnecessary de-
lations (Chapter 10), and as one of the which was so popular in the 18th
tails,
properties of electric circuits (Chapter and 19th centuries and which has al-
13). Although some of the sections of most been forgotten. We think the tele-
physics are elaborated in Part 2 in great graph style is not at all suitable for
detail and others are only outlined or the beginner. The extra information,
even omitted altogether, the amount of which in some cases can be ignored, is
information about physics presented not burdensome, whereas insufficient in-
here is so great that as a whole the book formation or its extreme concentration
can be viewed not only as a text in in a book for the inexperienced reader
mathematics but also as an important is quite impermissible.
addition to the ordinary physics It is also necessary tobear in mind
textbook. that this book contains considerable
Chapters 8 to 13 (Part 2 of the book) material for persons interested in an
are devoted to physical problems and in-depth study of the basics of higher
are almost independent of each other. mathematics. During the first reading-
They can be read any order. The only
in much of this material can be ignored.
exception is Chapter 10, which is de- Say, all the small print, including the
voted to oscillations. To understand story about the origin of higher mathe-
this chapter it is necessary to read Sec- matics, and the starred sections. The
tions 9.1 to 9.6 of the chapter devoted same goes for the entire third part of
to mechanics. Although Part 1 of the the book (“Some Additional Topics”)
book, called “Elements of Higher Math- and the Conclusion (“What Next?”).
ematics,” contains many examples And, finally, we have already empha-
from physics and mechanics, various sized the possibility of a selective ac-
sections of Part 2, called “Higher Math quaintance with the second (physics) part
Applied to Problems in Physics and of the book (Chapters 8 to 13).
Engineering,” will be of interest to all The large first chapter (“Functions
categories of readers. and Graphs”) mainly contains material
The large size of this book would familiar from school. A quick run-
seem to contradict the word “beginners” through should suffice for most readers.
in the title. But don’t be afraid of the Chapters 2 and 3 present the founda-
number of pages. We
are well aware of tion ideas of higher mathematics (diffe-
all the shortcomings (and also of the rential and integral calculus), while
merits) of the detailed and frankly rath- Chapters 4 and 5 elaborate the basic
er wordy system of presentation that techniques of higher mathematics suffi-
we have chosen. The book contains a cient for elementary applications. Chap-
great many examples and calculations. ter 6 is devoted to certain more ad-
Some problems are elaborated several vanced areas of mathematical analysis
times and from different angles. Much (series and differential equations), and
space is devoted to the history of the Chapter 7 is concerned with purely
theories presented. All the information mathematical applications of the con-
contained in this book could, of course, cepts of a derivative and an integral
be compressed and presented on far (physical applications are considered
fewer pages. But then it would require in Part 2 of the book). Elements of the
a greater effort to read the book, and history of higher mathematics are found
anything omitted or incorrectly under- in the concluding sections of Chapters
stood would hamper further progress 3 and 6. These are separated from the
(or even make it impossible to proceed). main text by three asterisks. And
Of course, there are situations in life finally the Conclusion (“What Next?”),
when the best source of information is which discusses some of the newer divi-
a brief telegram, but there is also charm sions of mathematical science now
in the art of unhurried presentation widely used in applications.
 10 Note to the Reader

The third part of the book (notably only in the second half of this century. 2
the Conclusion) is of a slightly differ- Although the Conclusion and, in fact,
ent character than the first and sec- the entire third part of the book can
ond. This last part is devoted to two easily be omitted (it is not needed for
trends in higher mathematics that con- an understanding of the main part of
siderably expand (and generalize) the the book), we believe that most readers
classical differential and integral cal- will wish, at least cursorily, to familia-
culus of Newton and Leibniz. The first rize themselves with these, since they
is complex numbers and functions of a reveal prospects for the future and ex-
complex variable ,
which have long plain certain ideas that are important
played a major role in physics and engi- to and characteristic of modern science.
neering. It is indeed remarkable that In short, this book can be used in a
many facts connected with ordinary variety of ways. The reader interested
real functions of real arguments become largely in mathematics will naturally
clearer when we move into the com- devote most of his attention to Part 1.
plex domain. The reader will find some He will undoubtedly find it useful, how-
examples of this in Chapters 14, 15, ever, to familiarize himself with var-
and 17. The second trend involves ious sections of Part 2, say, Chapter
generalized functions (or distributions) 8 or the first portions of Chapter 9 and
/ (#), and first of all the Dirac delta func- some of Chapter 10, or with Chapter
tion, which is always equal to zero 13. Anyone who wishes to deepen his
except when x equals zero, in which knowledge in mathematics is advised
case it is equal to infinity (!). Such func- to turn to the concluding sections of
tions also play a big role in physics Chapter 10 and/or the third part of the
and engineering. In fact, their origin book. The reader interested in the func-
is much more closely connected with tions of a complex variable and the del-
physics than with mathematics. We ta functions should not ignore Chapter
^consider it most desirable that future 17, which is devoted to the applications
physicists and engineers become ac- of the corresponding theories. Addition-
quainted with these remarkable func- al information about mathematics can
tions as early as possible. be derived from the Conclusion, but
Chapter 14 to 17, which are devoted for an in-depth grasp of the topics
to functions of a complex variable and tuched on in the Conclusion it is necessary
to generalized functions, are intended to turn to other literature. Finally, a
for the especially interested reader. reader interested in physics more than
They are written in more of an outline in mathematics need only run through
:form than the rest of the book (al- the first part of the book, ignoring the
though they can be understood by a suf- text in small print and omitting most of
ficiently persevering beginner). They the starred sections (but not Section
'can be read in greater or lesser detail, 6.7, which shows how differential equa-
as one wishes. tions can be used to analyze phenomena
The Conclusion (“What Next?”) is in the natural sciences). Such a reader
an even more general survey. We do need not dwell long on Chapter 7,
not expect the reader to grasp the ma- which is purely mathematical, but he
terial fully. It is only intended to arouse should give careful study to Chapter 6
the reader’s curiosity and encourage because its applied significance is great.
further work. It would be naive to
think that the topics embraced by this 2
Here we draw particular attention to
hook can complete the mathematical the theory of groups, which mathematicians
education of a physicist or engineer. developed in the 19th century and which
at that time was considered far removed
In the Conclusion we touch on certain
from any physics. At present the physics of
other trends in science that found wide elementary particles owes its achievements
application in physics and technology largely to the theory of groups.
 Note to the Reader 11

The reader interested in physics will book. They will save the reader the
probably make a careful study of Part trouble of turning to other textbooks
2 (and perhaps also Part 3, which is when doing the exercises that complete
likewise connected with profound physi- each section of the book. The physics
cal theories). He can study Chapters 8 examples are based on the now accepted
to 13 in any order, though in some re- international system of units (SI). The
spects it is best not to allow for too big reader insufficiently acquainted with
a break in time between Chapter 10 this system will find Appendix 5 help-
and Chapter 13. ful. Finally, the concluding Appendix
The list of literature given at the 6 presents the Greek alphabet, which is
end of the book includes certain other widely used in this book.
publications that outline the basics of Like any book on mathematics, this
higher mathematics, also textbooks that one presupposes an active reader (one
cover more material, and books that with pencil and paper) willing to work
deal with topics that go far beyond the the many exercises that are given here.
material presented here. As a natural Better still if you have a pocket cal-
continuation of this book we recom- culator or a programmable calculator.
mend [151.
3
On
the other hand, the Then you will be able to verify and sub-
books [19], [21] are far removed from stantiate the general theorems with cal-
the main content of this book: they are culations and make freer use of the
devoted to branches in mathematical method of “arithmetical experimenta-
science that relate to trends mentioned tion,” that is, approximate computa-
in the Conclusion. tions that reveal the content of the con-
In the literature on physics we espe- cepts introduced and the meaning of
cially recommend an excellent three-vol- the formulas. The many specific exam-
ume course [341 (there are also prob- ples will help you to evaluate the
lems and exercises relating to the whole exactitude of the approximate formu-
course). In books [22-30, 37] higher las scattered throughout the book, and
mathematics is not used. you will better appreciate the wisdom
In Appendices 1 to 4 at the end of of our forerunners who created methods
the book the reader will find short ta- and concepts without modern compu-
bles of derivatives, integrals, and numer- ter techniques. At the same time you
ical series (which are discussed in Part will feel the new power which we chil-
1) and also extracts from numerical dren of the 20th century derive from
tables, which, in a sense, round out the computers.
3
Students of technical colleges who have
studied our book in the first year are advised, Ya. B. Zeldovich
in the second year, to turn to [15]. F. M . Yaglom
 Teaching Notes

This book is intended not only for stu- physics and technical subjects (the the-
dents but also for teachers who, while ory of oscillations, for instance). Most
familiar with the subject matter, might important of all, of course, is the fact
be interested in the methods that we that this book will help them not only
recommend for teaching higher mathe- to grasp the substance of differential
matics to future physicists and techni- and integral calculus but also to realize
cians. Our textbook is not designed for how and why calculus arose and why
any particular category of students: this key subject of mathematics is
it can be used in teaching mathematics needed.
or physics in high school, technic- The Selected Readings at the back of
al school, vocational school, or col- the book are intended for both students
lege. High-school teachers will find and teachers. Some of the books listed
it helpful in refreshing their ideas, [4-6, 12-14] demonstrate possible ways
obtained at college, about the relation- to achieve an integrated approach to
ships and interconnections between the teaching of higher mathematics and
mathematics and physics. They can use its applications.
it in elective classes or in physics and This book consists of three parts, the-
mathematics study circles; some parts first of which (Chapters 1 to 7) chiefly
of it can be conveyed to students di- reminds readers of facts learned at
rectly in the classroom. The college school. Here we set forth the ideas and
or university instructor can use the ma- methods of higher mathematics. Part 2
terial in practical classes in the calcu- (Chapters 8 to 13) deals with applica-
lus course (higher mathematics) or in tions of higher mathematics in physics
the general physics course. Our book and engineering; it is a fundamental
will also be useful to instructors during part of the book, unlike Part 3 (Chapters
work on exercises in differential and 14 to 17), which is supplementary and
integral calculus with first-year stu- can, in principle, be skipped. The in-
dents of technical colleges or in the separable connection between the first
physics department, even if a different part of the book, devoted to mathema-
trend predominates in the lecture course tics, and the second part, dealing with
in higher mathematics. Work on top- physics, is underlined by the fact that,
ics in Part 2 of the book (possibly Part for instance, the section on Fourier se-
3 as well) will help students to get a bal- ries is included in Part 2 (in the chapter
anced idea of mathematics and facili- on oscillations), while the process of
tate assimilation of certain parts of water flow is studied in Chapter 6 de-
Teaching Notes 13

"voted to series and differential equa- traditional to open the calculus course
tions. with the theory of limits based on a re-
In writing this text we did not spe- fined delta-epsilon technique using pro-
cifically take account the possi-
into found concepts whose value was at
ble use by present-day students of per- first only stated but not explained. All
sonal computers. Naturally, these this created a situation in which the
would further simplify the teacher’s ex- concepts of the derivative and integral,
planations. For one thing, they make intuitively fairly clear and simple,
“numerical experiments,” such as the seemed, to the uninitiated, to be some-
one in Section 4.7, much easier and fas- thing very profound and aroused a kind
ter (and also, therefore, more natural). of mystical awe (in the case of first-year
They enable students to employ numer- students there was also the real fear of
ical (approximate) differentiation and not passing the exam). Indeed, many
integration on a much broader scale. mathematicians, it is to be regretted, pro-
Computers should, naturally, also be moted that feeling by stressing the logi-
taken into account when teaching stu- cal subtlety of the concept of limits and
dents the applications of mathematics. the need to use as a basis one of the va-
The present volume is linked up with riations of the theory of real numbers
a book by one of the authors (Zeldo- (though it is common knowledge that
vich). came out in English in 1973
It substantiations of this kind, of neces-
( Higher Mathematics for Beginners and sity fairly complicated, appeared only
Its Application to Physics Mir Publi- ,
in the second half of the 19th century,
shers, Moscow). 1 One of the reasons for that is, two centuries after the birth
writting it was that in those years dif- of higher mathematics). The students’
ferential and integral calculus was not logical criteria were developed on the
even mentioned in the Soviet Union’s basis of theorems that seemed clear to
high schools: the analytical section them without any proof. For instance,
of the school mathematics course was the fact that a variable can have only
limited to the fundamentals of algebra one limit 2 was traditionally established
and of trigonometry, then still a sep- without any regard for the intuitive
arate subject. As for the college course, idea of a limit, which clearly tells us
this taught the fundamentals of cal- that if a variable moves back and forth
culus according to a system far re- between two “limits”, it does not tend
moved from the needs of physicists and to either of them.
technicians. We remind our reader that The situation today is indisputably
the main college textbook in higher better. A course in the rudiments of
mathematics in the Soviet Union differential and integral calculus is
1 +
^hrmu^iroTit^hib UT?3bb"W'dsH Urnnivi ln?"b .
/
ahJiigvtorr^ hir^ii "SvhnnJis hib
well-known book, which had been trans- Soviet Union. And among the new col-
lated into Russian and then later re- lege and university textbooks there are
vised by Academician N. Luzin. Revised some based on an approach similar to
editions of this simple book went ours. For example, the author of the
through a large number of printings and textbook [91,which has gone through
were highly popular among engineering a number of editions, doesnot attempt
students. However, the harsh criticism to prove a single theorem on limits, pre-
leveled at the textbook by “pure” math- ferring, instead, to replace strict logic
ematicians, who accused it of being 2
Compare this with a theorem from a pre-
slipshod, led to its complete replace- sent-day American school textbook on
ment by textbooks of an altogether dif- geometry: E.E. Moise and F.L. Downes, Jr.,
ferent trend. In the 1950s it became Geometry, 2nd ed. Addison-Wesley, Menlo
f

Park, Cal., 1971: p. 51, Theorem 2-3: “every

1
The first Russian edition came out in segment has exactly one midpoint.” This is
1960, while the fifth and last, from which the obviously intended only for future mathema-
English translation was done, came out in 1970. ticians.
14 Teaching Notes

with a frank appeal to common sense. nored. Since it is connected with the*
And even the textbooks which tradition- training of future mathematicians, the-
ally open with an exposition of the growing importance of mathematics’
theory of limits are read in an entirely compels us to pay due attention to this
different way by students who learned point. But it would be absurd to main-
the fundamentals of the derivative and tain that teaching shaped by aim (2)
the integral while still in high should be greatly expanded; after all,
school. mathematicians will never comprise a
Nevertheless, we feel it necessary to significant part of the population. From
detail the motivation behind our meth- this point of view the trend in schools in
od, especially since this method un- many countries to teach the new math,,
doubtedly does not meet with general ap- understood as the fundamentals of sym-
proval. Although elements of higher bolic logic and of the general theory
mathematics have won a firm place in of sets and the doctrine of abstractly
high-school mathematics, the ques- introduced functions and so-called bi-
tion of how these elements should be nary relations, strikes us as harmful.
presented is far from settled. An “en- Nor can aim (3) be the main guideline
gineering” approach or a “natural sci- in planning large-scale mathematics
ence” approach is quite possible here, instruction; the training of future prog-
but some take a fundamentally differ- rammers and other computer experts
ent view. Nor can it be said that the is important, yet for the most part it is
principles underlying the mathematics only a special case.
course in technical and vocational Our book is meant for future engi-
schools are clear and definite, though we neers and natural scientists who want
believe that a more rigorous treatment to know mathematics for aim (4). In
than the one applied in this book is our opinion, this undoubtedly large
hardly needed for these schools. category of students should, as quickly
Clearly, before starting to plan a sys- as possible, be taught the information
tem of teaching mathematics we must they need, without cluttering it up
give thought to the reasons why we are with superfluous logical subtleties or
teaching it. The answer to “how” can striving for maximum scope or a com-
only be arrived at after we answer the pletely rigorous approach. For them,
question “why.” There can be the follow- after all, mathematics will always be a
ing reasons for studying mathematics: tool, a language with which to describe
(1) to pass an exam; phenomena, and not something in which
(2) to develop one’s capacity for abs- they are interested for itself. Moreover,
tract and logical thinking; we feel that, as was the case in the 19th
(3) to learn to use computers; century and will probably be the case in
(4) to apply it in studying engineer- the 21st century too, the main stress
ing, physics, biology, the microcosm should be put on the elements of mathe-
of atoms and elementary particles, or matical analysis, differential and integ-
the macrocosm of stars and galaxies. ral calculus, and the simplest differ-
Tnese aims h'ffrer grad fry, auli "firey mftrdi equations.
necessitate completely different instruc- The approach to the main concepts
tional methods and systems. Indisput- of analysis should proceed from the idea
ably practical though it is, the first that Nature is arranged fairly simply,
of the above-listed aims can of course that “God may be subtle, but He is not
be dismissed at once. It is clear that malicious,” as Einstein put it. Hyper-
the examiner must meet the needs of trophy of the theorems of existence can
the students rather than the other way only discourage students who are en-
around (although, unfortunately, this dowed with a healthy feeling for phys-
is not always the case). The second aim ics. It is clear that a moving particle
cannot and should not be wholly ig- has a velocity, and that a curved line
 Teaching Notes 15

has a tangent; to demonstrate the exis- exactly, in present-day textbooks, tho

tence of velocity is just as unnecessary fairly archaic reasoning of Newton,
as it is to demonstrate that a flat figure with its long-forgotten terms and sym-
has an area. Just take some paper or bols, or of Leibniz, with his characte-
cardboard, cut out a figure and weight ristic “metaphysics of infinitesimals.’^
it, divide its mass by the mass per square However, the ideas of a course for be-
meter of the material, and you have ginners could, it seems to us, be based
the area. In other words, in the initial on familiar classical graphic ideas in-
stage of the study of analysis the exis- stead of on the modern theories of real
tence theorems of the derivative and numbers or on strict definitions devel-
integral are not needed (in fact, they oped over the centuries. Of course,
are harmful): both of these concepts there comes a time when the student
have a natural physical meaning, and has to be told about possible complica-
hence they exist. 3 It would be absurd to tions, but not about exotic “continuous-
start learning grammar by studying the functions that do not have any deriv-
exceptions instead of the general rules, atives anywhere,” 4 but simply about
or to start a language course by study- the possibilities of a gap or a break in
ing the grammar instead of first build- the curve. However, we believe such,
ing up a vocabulary. Yet that is the warnings should be given not before
same, unfortunately, as the fairly wide- but after the student has intuitively
spread system of learning the “mathe- assimilated the basic ideas, just as it
matical language” in which the concept would be absurd to begin the study of a
(and properties) of the limit precedes new mechanism by learning the instruc-
any substantive application of this tions about a possible breakdown.
concept. Furthermore, we are inclined to believe*
A fundamental law of biology is that that a far more instructive (and in-
“ontogeny recapitulates philogeny,” formative) approach than the tradition-
that is, the development of an individua- al nihilistic is the modern “constru-
al repeats, in one degree or another, ctive” approach to peculiarities of the*
<?lv ^ "gou x

instance, at a certain period of its that every continuous function (even

growth the human embryo has gill if not smooth) has a derivative
it’s
slits, because life originated in the sea which, however, may prove to be a del-
and the ancestors of mammals, includ- ta function instead of an ordinary func-
ing man, were fish. We feel that teach- tion; we devote much attention to-
ers, too, should remember this law of this approach in Chapter 16 of the book.
biology. The future physicist or engi- We also take advantage of every oppor-
neer should first study mathematics in tunity to stress such important general
the form in which it was created by the ideas as linearity and the principle of
great scientists of the 17th and 18th superposition which are sometimes im-
,

centuries instead of receiving it immedi- permissibly left in the shade in books

ately in the form that had been worked
4
out by the second half of the 20th Here we are inclined to concur with the*
French mathematician Charles Hermite, one
century after a long period (Revolution. of the most prominent analysts of the second
This, it goes without saying, should half of the 19th century, who said, in a letter
not be understood as a call to reproduce to his friend T.J. Stieltjes, the Dutch analyst,
that “I turn away with fright and horror from
3
When students of the famous William this lamentable evil of functions which do 1

Thomson, Lord Kelvin (1824-1907), tried to not have derivatives.” A book by the French-
determine the derivative by the Cauchy pro- mathematician B. Mandelbrot ( Fractals ,

cedure, he grew irritated. “Oh, forget Totgent- W.H. Freeman, San Francisco, 1977), which
ner”, he said. “The derivative is just the takes a diametrically opposite stand, is highly
velocity.” (Totgentner was a professor of interesting but clearly intended for far more
“pure” mathematics who in Kelvin’s time sophisticated readers than those for whom
taught a course in calculus at Cambridge.) our elementary textbook is designed.
16 Teaching Notes

intended for physicists and engineers ber of the Academy of Sciences, out-
nr students training for these fields. standing scientist, technician and engi-
Mathematics is developing by leaps neer whose work and achievements cov-
and bounds, and so is the body of know- ered an exceptionally broad range,
ledge needed by those who use it. author of the first Russian book on nu-

wa, today, than, w

Physicists and engineers have to be
heinrn,.
Some of the new scientific disciplines
merical methods in mathematics, trans-
lator. of.
ciples of
Nawtoms. Mstih esn ati csiL,
Natural Philosophy and an au-
,
?w? r
are named in the Conclusion entitled thority on celestial mechanics, wrote
“What Next?”. Computational mathe- the following: “That on which Newton
matics, based on the enormous poten- based the whole of the modern theory
tials of the electronic computer, is hard- of the Universe and his incontrovert-
ly touched on in this book, but it too ible proofs of the structure of the system
is acquiring tremendous importance. of the world cannot be regarded as in-
We think it quite possible to acquaint sufficiently rigorous
for a 16-year-old
beginners more quickly with the initi- high-school student.” Krylov point-
al sections of higher mathematics so ed out that beginners often accept
that they can apply their knowledge the refined rigor of mathematical proofs
immediately and have more time to as “a triumph of science over common
learn new substantive mathematical sense.” In his well-known Autobio-
theories. graphical Notes Albert Einstein, one of
The pedagogical approach taken by the greatest physicists of all times,
the authors of many familiar calculus wrote:
textbooks is reminiscent of the debates
conducted by medieval Scholastics. At the age of 12-16 I familiarized
myself with the elements of mathemat-
Such authors regard the student as an
ics together with the principles of differ-
experienced opponent eager to find weak ential and integral calculus. In doing so
points in the views of the teacher and I had the good fortune of hitting upon
they believe that the work of the teach- books which were not too particular in
their logical rigor, but which made up
er is all possible objections.
to refute
for this by permitting the main thoughts
We, on the contrary, regard the student to stand out clearly and synoptically.
as a friend and ally who is ready to be- This occupation was, on the whole, truly
lieve the teacher or the textbook and fascinating; climaxes were reached whose
impression could easily compete with
whose primary concern is to be able as
that of elementary geometry— the ba-
quickly as possible to employ, in the sic idea of analytical geometry, the in-
study of nature and technology, the finite series, the concepts of the differen-
new methods learned 1 1 is chiefly through
.
tial and the integral....

an analysis of many examples and

applications that the student comes to A similar view was held by the distin-
understand the substance of the new guished Soviet theoretical physicist
tools and gains the intuition that fore- Lev Landau (1908-1968), Lenin Prize
warns when the tools refuse to function. and Nobel Prize winner, Member of
The strictly logical approach bypasses the USSR Academy of Sciences, a man
the question of the origin, significance, who combined outstanding intuition in
and benefit of the concepts and physics with a high level of mathemat-
theorems under study. This book focuses ical technique. When asked for his
mainly on “rough” mathematical ideas opinion about a mathematics syllabus
and their connections with natural phe- for a Moscow physics institute, he
nomena. wrote:
We find support for our positions in
Unfortunately, your syllabus suffers
the writings of many prominent scien-
from the same shortcomings that are
tists. The famous Russian naval archi- common to the mathematics syllabuses
tect Aleksei Krylov (1863-1945), Mem- that turn half of the study of mathe-
 :

Teaching Notes 17

matics by physicists into a tiring waste dean geometry as a model of a strictly

of time. Important as mathematics is to
logical approach, of strict logical dedu-
physicists, what they need, as we know,
isa calculating analytical mathematics; ction. But here is what he says further
however, for reasons I cannot under- (Scientific American September 1964,
,

stand, mathematicians force logical exer- p. 43)

5

cises on us as an obligatory part of the

course.... I think it is high time to teach But emphasis on this [axiomatic, log-
physicists what they themselves think ical] aspect of mathematics is totally
they need instead of trying to save their misleading if it suggests that construc-
souls against their own wishes. I have no tion, imaginative induction and combi-
desire to discuss the idea worthy of medie- nation and the elusive mental process
val scholasticism that people learn to called intuition play a secondary role in
think logically by studying things they productive mathematical activity or ge-
do not need. I most emphatically main- nuine understanding. In mathematical
tain that all theorems of existence, education, it is true, the deductive meth-
highly rigorous proofs, etc., should be od starting from seemingly dogmatic
completely expelled from the mathe- axioms provides a shortcut for covering
matics studied by physicists. a large territory. But the constructive
Socratic method that proceeds from the
These differences of opinion between particular to the general and eschews
mathematicians and physicists led to dogmatic compulsion leads the way more
a paradoxical situation in which three (!) surely to independent productive think-
ing.
Nobel Prize winners in physics and
chemistry (Walther Nernst, Svante All the above provides a sufficiently
Arrhenius, and Hendrik An toon Lorentz) fullexplanation of the basic guidelines
found it necessary to write elementary followed in our textbook. Its principles
textbooks in higher mathematics “for and underlying propositions coincide
natural scientists and physicians,” as with those typical of the textbook
the title of the book by Arrhenius says, Higher Mathematics for Beginners and
or “for natural scientists, with special Its Application to Physics', however,
attention paid to the needs of chemists,” the details of the exposition differ sub-
as Nernst pointed out to his readers. stantially from the old one, even in the
The best known of these books is the sections that were part of the above-
Course in Differential and Integral Cal- named book. There are many new top-
culus, With Applications in Natural ics here; this refers both to a number of
Science by H. A. Lorentz, which went particular points (the second differen-
through several editions between 1901 tial of a function of one variable or the
and 1926 and was still highly popular total differential of a function of two
in the 1930s as a textbook for future en- variables; the idea of inertial refer-
gineers and also for self-instruction. ence frames) and to entire large areas of
The ideas in the present book are fairly science (Fourier series, the theory of
similar to those set forth in all the functions of a complex variable, or the
above-named books. design of a laser).
It is worth noting that many “pure” When using this textbook, the teach-
mathematicians who do not special- er can freely draw on material from
ize in abstract algebra or mathematical the physics sections to illustrate mathe-
logic and have a taste for applications matical concepts and theorems or, con-
think along the same lines as the phy- versely, can include separate mathemat-
sicists mentioned in the above para- ics topics in physics lessons. For ex-
graph. For instance, the famous Ger- ample, in the text we illustrate the con-
man (and later American) mathemati- cept of the derivative by the connec-
cian Richard Courant (1888-1972), found- tion between the path traversed and
er of New York University’s Courant
5
Institute of Mathematical Sciences, Another excellent book, G. Polya’s
Mathematics and Plausible Reasoning 2 vols,
wrote in 1964 that mathematicians had (Princeton Univ.
,

Press, Princeton, N.Y.,

for a very long time accepted Eucli- 1954), follows the same approach.
2-0946
18 Teaching Notes

the speed, or between the quantity of ing instruction, part of this material
heat needed to heat a body and the can be omitted of course; however, it
body’s heat capacity; but nothing pre- seems to us that some use of physical
vents the teacher from referring here considerations in the exercises in mathe-
to, say, the connection between the elec- matics (and partly in the lectures as
tromotive force and the current inten- well) is very useful, and that a well-
sity for a coil (or for a capacitor, where thought-out coordination of the elemen-
the relationship is the reverse; cf. Sec- tary courses in mathematics and physics
tion 13.1). It goes without saying that is absolutely essential.
the number of such examples can be in- In conclusion we would like to call
creased indefinitely. We believe that the attention of the teacher (and stu-
integrated teaching of calculus and its dent) to textbook [11], which is close
applications is very important for fu- to our ideas in spirit but is more mathe-
ture technicians and physicists because matical in content. The modern “com-
a mathematical technique acquires deep puterization era” makes it necessary
meaning only in its application. The to mention another excellent (but more
physics sections of the book are doubly complicated) book [20], which is
beneficial: they not only give the stu- aimed at acquainting the reader with the
dent important information in physics applications of computers in mathe-
but shed additional light on the mathe- matics and mathematical physics.
matical constructions in the initial
chapters and help the student to gain a
new understanding of how natural and Ya B. Zeldouich
.

necessary these constructions are. Dur- I. M. Yaglom

 Contents

Part 1. Elements of Higher Mathematics

CHAPTER 1. FUNCTIONS AND GRAPHS 23

1.1 The Functional Relationship 23
1.2 Coordinates. Distances and Angles Expressed in Terms of Coordinates 26
1.3 Graphical Representation of Functions. The Equation of the Straight Line 30
1.4 Inverse Proportionality and the Hyperbola. The Parabola 34
1.5 Higher-Order Parabolas and Hyperbolas. The Semicubical Parabola 42
1.6 The Inverse of a Function. Graphs of Inverse Functions 47
1.7 Transforming Graphs of Functions 50
1.8 Parametric Representation of a Curve 59
1.9* Some Additional Topics from Analytic Geometry 61

CHAPTER 2. WHAT IS A DERIVATIVE? 67

2.1 Motion, Distance, and Velocity 67
2.2* Specific Heat Capacity of an Object. Thermal Expansion 70
2.3 The Derivative of a Function. Simple Examples of Calculating Derivatives 72
2.4 Properties of Derivatives. Approximating the Values of a Function by
Means of a Derivative 75
2.5 A Tangent to a Curve 80
2.6 Increase and Decrease of Functions. Maxima and Minima 85
2.7 The Second Derivative of a Function. Convexity and Concavity of a Curve.
Points of Inflection 89

CHAPTER 3. WHAT IS AN INTEGRAL? 92

3.1 Determining Distance from the Rate of Motion. The Area Bounded by
a Curve 92
3.2 The Definite Integral 96
3.3 The Relationship Between the Integral and the Derivative 101
3.4 The Indefinite Integral 104
3.5 Properties of Integrals 109
3.6 Examples and Applications 112

CHAPTER 4. CALCULATION OF DERIVATIVES 127

4.1 The Differential 127
4.2 Derivatives of a Sum and of a Product of Functions 132
4.3 The Composite Function. The Derivative of the Fraction of Two Functions 134
4.4 The Inverse Function. Parametric Representation of a Function 135
4.5 The Power Function 138
4.6 Derivatives of Algebraic Functions 139
4.7 The Exponential Function 140
4.8 The Number e 142
4.9 Logarithms 145
2*
20 Contents

4.10 Trigonometric Functions 148

4.11 Inverse Trigonometric Functions 151
4.12 Differentiating Functions Dependent on a Parameter and Functions of
Several Variables. Partial Derivatives 153
4.13 The Derivative of an Implicit Function 158

CHAPTER 5. INTEGRATION TECHNIQUES 162

5.1 Statement of the Problem 162
5.2 Elementary Integrals 162
5.3 General Properties of Integrals 164
5 4 _ In teflK&tj/iTl- JlV JJ a 166 _ „ „
5.5 Change of Variables in Integration 168
5.6 Change of Variable in a Definite Integral 170
5.7 Integrating Functions Dependent on a Parameter 173

CHAPTER 6. SERIES. SIMPLE DIFFERENTIAL EQUATIONS 177

6.1 A Series Representation of aFunction 177
6.2 Computing the Values of Functions by Means of Series 184
6.3 Cases Where Series Expansions Cannot be Applied. The Geometric Progres-
sion 187
_
6.4 The Binomial Theorem for Integral and Fractional Exponents 192
6.5 The Order of Increase and Decrease of Functions. L’ Hospital’s Rule 194
6.6 First-Order Differential Equations. The Case of Variables Separable 198
6.7* The Differential Equation for Water Flow from a Vessel 202

CHAPTER 7. INVESTIGATION OF FUNCTIONS. SOME PROBLEMS FROM

GEOMETRY 214
7.1 Smooth Maxima and Minima 214
7.2 Other Types of Maxima and Minima. Salient Points and Discontinuities.
The Left and Right Derivatives of a Function 221
7.3 Investigating Maxima and Minima of Functions Dependent on a Parameter 227
7.4* Convex Functions and Algebraic Inequalities 234
7.5 Computing Areas 241
7.6* Estimating Sums and Products 245
7.7* More on Natural Logarithms 250
7.8 Average, or Mean, Values 253
7.9 Arc Length 260
7.10 Curvature and the Osculating Circle 265
7.11 Solid Geometry Applications of Integral Calculus 271
7.12 Curve Sketching 275

Part 2. Higher Math Applied to Problems of Physics and

Engineering
CHAPTER 8. RADIOACTIVE DECAY AND NUCLEAR FISSION 281
8.1 The Basic Characteristics of Radioactive Decay 281
8.2 Measuring the Mean Lifetime of Radioactive Atoms 283
8.3 SeriesDisintegration (Radioactive Family) 289
8.4 Investigating the Solution for a Radioactive Family (Series) 291
8.5 The Chain Reaction in the Fission of Uranium 294
8.6 Multiplication of Neutrons in a Large System 295
8.7 Escape of Neutrons 297
8.8 Critical Mass 298
8.9 Subcritical and Supercritical Mass for a Constant Source of Neutrons 299
8.10 The Critical Mass 301

CHAPTER 9. MECHANICS 303

9.1 Force, Work, and Power 303
9.2 Energy 308
9.3 Equilibrium and Stability 312
9.4 Newton’s Second Law 316
Contents 21

9.5 Impulse 317

9.6 Kinetic Energy 320
9.7 Inertial and Noninertial Reference Frames 321
9.8* The^ Galilean Transformations. Energy in a Moving Reference Frame 324
9.9* The^Path of a Projectile. The Safety Parabola 327
9.10 The“Motion of a Body in Outer Space 331
9.11 Jet Propulsion and Tsiolkovsky’s Formula 334
9.12 The Mass, Center of Gravity, and Moment of Inertia of a Rod 337
9.13*Centers of Gravity of a String and of a Plate 342
9.14 The Motion of a Body in a Medium that Resists this Motion with a Force
Dependent Solely on the Velocity 346
9.15*The Motion of a Body in a Fluid 350

CHAPTER 10. OSCILLATIONS 354

10.1 Motion Under the Action of an Elastic Force 354
10.2 The Case of a Force Proportional to Deviation. Harmonic Oscillations 357
10.3 Pendulums 360
10.4 Oscillation Energy. Damped Oscillations 363
10.5 Forced Oscillations and Resonance 366
10.6 On Exact and Approximate Solutions of Physical Problems 368
10.7 Combining Oscillations. Beats 372
10.8 The Vibrations of a String 375
10.9 Harmonic Analysis. Fourier Series 380

CHAPTER 11. THE THERMAL MOTION OF MOLECULES. THE DISTRI-

BUTION OF AIR DENSITY IN THE ATMOSPHERE 387
11.1 The Condition for Equilibrium in the Atmosphere 387
11.2 The Relationship Between Density and Pressure 388
11.3 Density Distribution 389
11.4 The Molecular Kinetic Theory of Density Distribution 391
11.5 The Brownian Movement and Kinetic-Energy Distribution of Molecules 393
11.6 Rates of Chemical Reactions 395
11.7 Evaporation. The Emission Current of a Cathode 396

CHAPTER 12. ABSORPTION AND EMISSION OF LIGHT. LASERS 399

12.1 Absorption of Light: Statement of the Problem and a Rough Estimate 399
12.2 The Absorption Equation and Its Solution 400
12.3 The Relationship Between Exact and Approximate Absorption Calcula-
tions 400
12.4 The Cross Section 402
Effective
12.5 Attenuation of a Charged-Particle Flux of Alpha and Beta Rays 403
12.6* Absorption and Emission of Light by a Hot Gas 405
12.7* Radiation in Thermodynamic Equilibrium 407
12.8* Emission Probability and the Conditions for Thermodynamic Equilibrium 410
12.9* Lasers 414

CHAPTER 13. ELECTRIC CIRCUITS AND OSCILLATORY PHENOMENA IN

THEM 418
13.1 Basic Concepts and Units of Measurement 418
13.2 Discharge of a Capacitor Through a Resistor 423
13.3 Oscillations in a Capacitance Circuit with Spark Gap 425
13.4 The Energy of a Capacitor 428
13.5 Inductance Circuit 431
13.6 Breaking an Inductance Circuit 433
13.7 The Energy of Inductance 435
13.8 The Oscillatory Circuit 439
13.9 Damped Oscillations 441
13.10*The Case of a Large Resistance 443
13.11 Alternating Current 444
13.12 Average Quantities. Power and Phase Shift 447
13.13 An Alternating-Current Oscillatory Circuit. Series Resonance 449
13.14 Inductance and Capacitance in Parallel. Parallel Resonance 451
22 Contents

13.15 General Properties of Resonance in a Linear System 452

13.16*Displacement Current and the Electromagnetic Theory of Light 453
13.l7*Nonlinear Resistance and the Tunnel Diode 454

Part 3. Some Additional Topics

CHAPTER 14. COMPLEX NUMBERS 458
14.1 Basic Properties of Complex Numbers 458
14.2 Raising a Number to an Imaginary Power and the Number e 463
14.3 Trigonometric Functions and the Logarithm 466
14.4* Trigonometric Functions of a Purely Imaginary Independent Variable.
Hyperbolic Functions 469

CHAPTER 15. FUNCTIONS THE PHYSICIST NEEDS 475

15.1 Analytic Functions of a Real Variable 475
15.2 The Derivative ^of a Function of a Complex Variable 478

CHAPTER 16. DIRAC’S REMARKABLE DELTA FUNCTION 486

16.1 Various Ways of Defining a Function 486
16.2 Dirac and His Function 487
16.3 Discontinuous Functions and Their Derivatives 489
16.4 Representing the 'Delta Function by Formulas 493

CHAPTER 17. APPLYING FUNCTIONS OF A COMPLEX VARIABLE AND

THE DELTA FUNCTION 496
17.1 Complex Numbers and Mechanical Oscillations 496
17.2 Integrals in the Complex Plane 500
17.3 Analytic Functions of a Complex Variable and Liquid Flow 507
17.4 Application of the Delta Function 512

Conclusion. What Next? 516

Selected Readings 530
Appendices 532
1. Derivatives 532
2. Integrals of Some Functions 532
3. Series Expansions 534
4. Numerical Tables 534
5. The International System, or SI 536
6. Greek Alphabet 536
Hints, Answers, and Solutions 537
Name Index 555
Subject Index 557
 )

Part

Elements of Higher Mathematics

Chapter 1 Functions and Graphs

This chapter is of an introductory na- (3) The volume V of a cone with a

ture. It contains material that the majo- given altitude h is a function of the
rity of the readers already know and is radius r of the cone’s base:
devoted to functional relationships be-
tween quantities. In it the reader will F = y:rt r 2 h. (1.1.1)
find some specific functional relation-
ships that we will often use in our nar- On the other hand if we assume that r
rative. With other important functional then formula (1.1.1) expresses
is fixed,

relationships the reader will get acquain- the volume V of the cone as the func-
(5) of its altitude h.
tion
ted in the other chapters.
(4) The distance z traversed by a free-
ly falling body depends on the time t
1.1 The Functional Relationship that elapses from the beginning of fall.
Nature and technology abound in This relationship is expressed by the
functional relationships that is, rela-,
formula
tionships between various quantities.
It is natural then that mathematics
pays so much
attention to such rela-
where g ~ 9.8 m/s 2 is the acceleration
tionships. A
functional relationship be-
of gravity.
tween one quantity ( y and another The current i depends on the re-
(x) signifies that to every value of x
sistance R of a conductor: for a given
there corresponds a definite value of y.
potential difference u we have
The quantity x in this case is called the
independent variable and y is the func-
,

tion of this variable. We

also some-
times say that x is the argument of the This list could be extended without
function y. end.
Here are a few examples taken from In mathematics, functional relation-
geometry and physics. ships are ordinarily defined by formu-
(1) The area S of a square is a func- las, for example,
tion of the length a of the square’s = 2x + S, y^x 2 + 5,
side: S a2= y
y = 3x* — x — x,
.

2
(2) The volume V of a sphere is a (1.1.4)
function of the sphere’s radius R : x
V= (4/3) nR\ y= ^j> y=V 3 *+ 7 .
24 1 Functions and Graphs

(Here y is everywhere a function of ar- current i. In this experiment the cur-

gument x ). A formula enables us to rent depends only on the resistance, the
compute the values of the function for quantity u in (1.1.3) being regarded as a
each given value of the independent constant coefficient.
variable. Sometimes this method re- Sometimes the formula gives only
quires using tables or other technical de- values of y for x varying within certain
vices. For instance, in the last example
in (1.1.4) we must extract the square
limits. Say, the formula y x en-= Y
ables finding the values of y only for
root of a number to find the value of
nonnegative values of x that is, for
,
the function. This can be done via ta-
x ^ 0. If y —
log 2 {x —
2), then y
bles of square roots or, even simpler,
exists only for x 2 0, — >
that is,
with a pocket calculator.
In physics and technology, function-
for x >
2. In the formula y (x = —
\)!{x + 1) in (1.1.4) we must as-
al relationships are usually recorded
by measuring devices; say, the pointer
sume that x 1, —
while in the
last formula in (1.1.4) we must assume
of the speedometer in your car shows,
that 3x + 7^
0, that is, 7/3 x^ —
together with the hand of your watch, - 2.33.
the functional relationship of the car’s
In formulas that appear in physical,
speed and time. But in this case, too,
engineering, geometric, and other prob-
the relationship can often be represent-
lems, the restrictions on the possible
ed by a simple formula, which yields
values of the independent variable
sufficiently accurate results. For instan-
sometimes follow from the very meaning
ce, if the potential difference in an elec-
of the problem considered. In the exa-
tric circuit remains constant, the ex-
perimentally observed (via the scale
mples at the beginning of this section,
say, when calculating the area of a
of the ammeter) relationship between
square and the volume of a cone, we
the electric current i and the resistance
R of the circuit is described by formu- assumed that the side of the square and
the base radius of the cone and its al-
la (1.1.3) fairly well. Ifwe have a for-
titude are positive quantities. Often
mula that relates the values of quanti-
the formulas “tell” us about the possible
ties that appear in physics and technol-
restrictions on the values of the inde-
ogy, we say that we have established
a law for the phenomenon in question.
pendent variable. For instance, the
formula for the cone’s volume yields a
In relation to formula (1.1.3) the law is
negative value if we take h negative,
Ohm's law.
It is characteristic that in most cases which result is of course unnatural;
in many formulas of the theory of rela-
in nature and technology the quanti-
tivity, the speed v of a moving body
ty of interest (the function) depends on
appears in expressions of the type
several other quantities. In Example
(5) the current depends on two quanti- Y —
c
2
v 2 where c the speed
,
of light,
ties: the potential difference u and the from which it follows that v < c; and
resistance R of the conductor. The so on.
cone’s volume in Example (3) depends In the first three formulas of (1.1.4)
on altitude h and base radius r. Assum- and, obviously, in many other formu-
ing all quantities except one to be las, no restrictions are imposed on the
given and constant, we study the de- independent variable: x can be large or
pendence of the function on a single small, positive or negative. The same
variable. In this book we will confine is true of many physical quantities;
ourselves mainly to functions of one say, electric current can conveniently
variable. For example, taking a given be assumed positive when the electrons
storage battery with a definite potential move in a specified direction and nega-
difference u, we will vary the resistance tive when they move in the opposite
R of the conductor and measure the direction.
1.1 The Functional Relationship 2&

Knowing the formula that states the Experiments show that for a given
dependence of y on x we can easily
,
conductor given material, a given
(of a
construct a table of values of y for sev- cross section, and a given length) thu
eral arbitrarily chosen values of x. electrical resistance depends on the
By way of an illustration, we will temperature of the conductor. For each
set up a table for the function y = value of the temperature T, the conduc-
3x 3 —x2 —
x The upper row contains
. tor has a definite resistance i?, so that
the values of x that we choose, the we can speak of a relationship in which
lower row, under each value of x, con- resistance R is a function of temperature
tains the appropriate value of y: T. Carrying out experiments, we can
find the values of R for various T
and thus find the dependence (relation-
x —3 —2—10 1 2 3
ship or function) R versus T, or R =
y = 3x*—x 2 — x —87 —26 —3 0 1 18 69
R ( T ). Here, the results of experi-
ments are given in the table below (va-
Using this formula, we can make a lues of R for distinct values of T ):
more detailed table specifying, say,
the values x = 0, 0.1, 0.2, .... Thus a T (degrees 0 25 50 75 100
formula is stronger, so to say, than Celsius)
any table. The formula not only con- R (ohms) 112.0 118.4 124.6 130.3 135.2:

tains the information necessary to com-

pile the given table but also enables If we are interested in the values of
one to find the values of the function R which do not
for other temperatures
for values of the independent variable appear in the table, additional mea-
not contained in the table. On the other surements are required because there is
hand, a table is convenient in that it no exact formula defining the depen-
immediately gives the value of y for dence of R on T. Practically speaking,
any given value of x, provided that the however, we can always offer an approx-
needed x is given in the table. A table imate formula that is in good agree-
is also more pictorial than a complicat- ment with the experimental data at
ed formula, by using which it is often temperatures at which the measure-
difficult to estimate the values that the ments were made. Let us take, for ex-
function assumes. However, a simple ample, the formula
formula, with certain experience on
the part of the researcher, provides
R = 112.0 + 0.272T — 0.0004T 2

a faster means for determining the be- (1.1.5>

havior of the function than the inex-
pressive sequence of numbers in a table.
and compile the following table using
this formula:
In natural sciences and technology
it often happens that the theory of the
phenomenon of interest to us is lacking T (degrees 0 25 50 75 100
Celsius)
and the physicist (or chemist, biologist,
R (ohms) 112.0 118.55 124.6 130.15 135.2
engineer) is only able to supply exper-
imentally obtained facts the depen- —
dence of the quantity of interest upon We see that the formula yields val-
the quantity that was given in the ex- ues of Rthat are very close to the ex-
periment. This is what happens, say, perimental values for those tempera-
in studies of the relationship between tures at which the measurements were
the resistance of a conductor and the made, and so one is justified in assum-
temperature of the conductor. Here the ing that at intermediate temperatu-
functional relationship can only be giv- res (say, at T =
10, 80, or 90 °C) the*
en in the form of a table containing formula will likewise give a correct
the experimental data. description of the functional depen-
 ) ?

26 1 Functions and Graphs

-denceofi? on T. Mathematicians say in H

this case that the established dependen-
ce of R on T does not yield large errors
— Afx = 2,y=4)
i

in interpolation which is a process in

,
i

which we go from the known values of

R
I

to new, intermediate, values (the J

1
Latin word interpolare means to refur-
I

bish, alter). The dependence of resis- A-

0 1 B CO
tance R on temperature T found via
formula (1.1.5) is called an empirical Figure 1.2.1

dependence or an empirical formula

,
.

{The adjective “empirical,” from the axis of ordinates and the point of inter-
,

Greek word empeiria meaning experi- section of the two lines (point 0 in
ence, means relying on experience or Figure 1.2.1) the origin of coordinates ,

observation, based on observation or or simply the origin. It is customary to

experience, or capable of being veri- picture the plane with the x and y
fied or disproved I:y observation or axes not flat on a table but vertically in
experiment.) However, an empirical front of the reader, just as if it were
formula, of course, requires verifica- drawn on the wall opposite you. The
tion, since the error introduced by using arrow of the x axis is from left to right
it may be considerable (clearly, the and the arrow of the y axis is pointed
reliability of an empirical formula is upward. These arrows show that, say,
the higher the thicker the grid formed the segment OB of the x axis is taken
by those values of the variable from to be positive if B lies to the right of
which the empirical formula was con- point 0 (as in Figure 1.2.1) and nega-
structed). An empirical formula be- tive if B lies to the left of point 0\ simi-
comes quite unreliable when we use it larly, the segment OC of the y axis is
outside the range of the independent taken positive if C lies above 0 (as in
variable for which it was constructed Figure 1.2.1) and negative if C lies
(such continuation of the range of the below 0.
empirical formula is known as extra- A definite pair of values of x and y ,

polation from the Latin extra meaning

,
say, x =2 and y —
4, is represented
outside); the errors in this case may be in the coordinate plane by a single
very large. For instance, we cannot ap- point. To construct this point we plot on
ply formula (1.1.5) outside the range the axes of abscissas and ordinates the
•of investigated interval, say, at
the segments OB =
2 and OC 4 (Fig- =
T = —200°C or at T = 500 °G, since ure 1.2.1), with both segments being pos-
there are no grounds to expect that for itive because they are plotted from left
all T R(T) will be expressed by a to right and upward, respectively, and
1
quadratic trinomial in 7 or formula
, the x and y are positive. If we erect
(1.1.5). perpendiculars at points B and C on
the x and y axes, the point at which
they intersect is the sought point
1.2 Coordinates. Distances and Angles
Expressed in Terms of Coordinates
A(x = 2, y = 4) or, briefly, A (2,4).
Conversely, to find the coordinates of
Rectangular Cartesian coordinates an arbitrary point A it is sufficient to
are used as a pictorial way of represent- drop perpendiculars and AC on AB
ing functional relationships by means the coordinate axes and “read” the val-
of drawings (graphs). Draw two per- ues of x and y corresponding to points
pendicular straight lines in a plane. B and C (these values are equal to the
Gall the horizontal line the x axis (also lengths of segments OB and OC taken
knownasthea#£so/a&sc£ssas), the ver- with the appropriate signs). The coor-
tical line the y axis (also known as the dinate axes divide the plane of the draw-
1.2 Coordinates 27

y -

>C
D • .B
6 • -/ •A
,
r

F 0 1 X
G* •M
H • •L
/P
Figure 1.2.2

Figure 1.2.4

ing into parts or quadrants,

four
I, II, III, and IV, with the signs in resorting to additional lines and desig-
each quadrant being as follows: (+, nations.
+), or positive abscissas and positive Thus, we can conclude that fixing two
ordinates, in quadrant I, and ( +), — , numbers, values of x and y, determines
(— ,
— ), and (+, )
in quadrants— the position of a point in the plane.
II, III, and IV, respectively (Figure Therefore, all geometric quantities re-
1.2.2). The axis of abscissas, Ox ,
cor- ferring to this point can also be ex-
responds to the value y 0, while the = pressed in terms of the coordinates of the
axis of ordinates, Oy, corresponds to point.
the value x =
0. The origin 0 has coor- Let us find, for instance, the distance
dinates (0, 0). Figure 1.2.3 shows a few r from the origin to the point A with
examples of points for which the cor- coordinates x and y, that is, the length
responding values of the coordinates x r of the line segment OA joining the ori-
and y are given in parentheses. gin 0 and the point A (Figure 1.2.5),
An important piece of practical ad- and the angle a between the straight
vice: get into the habit of evaluating line OA and the axis of abscissas. From
approximately the values of the coordi- point A we drop the perpendiculars AB
nates of points in the plane and of find- and AC
on the axes of coordinates. The
ing the points from the known values length of OB is equal to x and | | ,

of the coordinates. For practical work it that of segment is equal to the AB

is often convenient to use squared pa- length of OC or y |. From the right
, |

per (it usually has a grid of millimeter triangle OAB, by the Pythagorean the-
squares) to locate points and plot orem, we have
curves. For instance, we advise the reader
(OA) 2 = r2 = (OB) 2 + (AB) 2
to try and determine the coordinates of
points A to N
in Figure 1.2.4 without
= x 2
+ y
2
,

or
y, ‘
A(2A)
?
r = Yx 2 i-(/ 2 . (1-2.1)
Dl-2.3) i

By the definition of the tangent,

i
<f
I
i

| i

<‘- 2 2>
-
i
1

)K('0J)\
0
1

1
<

1 1
For example suppose that x 2 and =
U-3,0) i
0 i !
x y = 3 (see Figure 1.2.5). Then

r = Y 13 ~ 3.6, ~ ~ 56°.
1

FH-n 1

A a— arctan
5(3,- 2)
Note that the angle a is reckoned from
Figure 1.2.3 the positive direction of the x axis coun -
28 1 Functions and Graphs

—
quadrant) in these two quadrants the
tangent has the same sign. To deter-
mine a completely, we must take into
account the signs of x and y which make
,

it possible to determine the quadrant

in which the particular point lies, or to
use more complete formulas than
(1.2.2), which are also more complicat-
Figure 1.2.5 ed:

terclockwise For this reason, if, say, x — cos a= sina^y-, r = Yx2jr y 2

.

— 2 and y = 2 (point in Figure

^4
y ,
.

1.2.6), the angle a obtuse: than a —

is (1.2.2a)
2/ — 2 = — 1 and a = 135°. When a point
lies below the x axis, the angle a It is easy to solve the inverse problem:
proves to be greater than 180°. In Figure suppose we are given a point A at
1.2.6 we have two instances of this a given distance r from the origin O
case: point# (2, —2) for which a 315° = with the line segment OA forming an
,

and point C ( 3, —3) for— which a = angle a with the x axis (the positive
225°. Since such large angles often direction of the x axis is assumed as
prove to be inconvenient, it is customa- usual). It is required to find the coor-
ry to reckon them from the same posi- dinates of point A. Looking at Figure
tive direction of the x axis but clock- 1.2.5, we see that
wise, taking a negative. For point B
we then have a =
—45°, and for point i-rcosa, sin a (1.2.3)
C we have a —
—135°. In other words,
we can assume that for any point the
These formulas are valid, without ex-
angle lies either within the range from
ception, for arbitrary positive and neg-
0 to 360° (without the value of 360°,
= ative angles a and yield the proper
since it is more convenient to take a
signs of x and y in any quadrant.
0° instead of a =
360°) or within
Clearly, the position of a point A
the range from —180° to 180°, with the
= —
180° and a = in a plane can be fixed by specifying
values a 180° be-
the Cartesian coordinates x and y\
longing to points on the negative “half”
however, instead of these two numbers
of the x axis (here both values of a
we can use the distance r and the angle
have an equal status).
a. These two numbers, r and a, are
In this respect formula (1.2.2) is
incomplete since it does not enable us
known as polar coordinates of point A y
point O is called the pole of polar coor-
to determine whether a point lies in the
dinates in the plane, and the ray Ox
1 or III quadrant (or in the II or IV
is called the polar axis (theOy axis does
not take part in the definition of polar
coordinates). Thus, formulas (1.2.2a)
and (1.2.3) define the transition from
rectangular Cartesian coordinates to
polar coordinates and vice versa.
Let us now examine problems involv-
ing two points, A x and A 2 We
denote
.

the coordinates of the first point by

xx and y x and the coordinates of the
,

second point by x 2 and y 2 (Figure 1.2.7).

We wish to find the distance r l2 between
Figure 1.2.6 these points and the angle a l2 between
 29
12 Coordinates

Figure 1.2.7

line segment and the x axis. 1 1

A A2r
-

The reader should assure himself

It is convenient to draw through A 1 that the formulas (1.2.4), (1.2.5), and
a straight line parallel to the x axis and (1.2.5a) hold true for arbitrary signs
through A 2 & line parallel to the y of all four quantities x l9 y l7 x 2 y2 ,

axis (in Figure 1.2.7 they are shown as and for any relationships between the
dashed lines and their point of inter-
B
coordinates: x1 2 or >x
x x2 y { < , { >
section is B). In the triangle A X A 2 <y For example, in Fig-
y 2 or y x 2.
the line segment A X B is equal to ure 1.2.8 we have the case x 1 0 and <
|
x2 — Xi |
and the segment A 2 B is
x2 > 0, with A 1 = A 1 (—2, 1) and
equal to j yz — lh I
• The construction A 2 = A 2 (3, 3). The length of segment
of triangle A A 2B
X
is similar to the A B is equal to the sum of the
X
absolute
construction of triangle OAB in Figure values |
x1 = 2 and x 2 = 3, which
| \

1.2.5. is strictly in accord with the general

By the Pythagorean theorem, formula A X B = x 2 — xx = 3 (—2) =
—
r i2 = V (?2 ~ *i) + 2
(5/2 — J/i)
2
- (1-2.4) 5. Consequently, the expressions for
r l2 and than a l2 are also correct.
The angle a 12 is found from the condi-
tion
Exercises
tan a 12 — » d-2-5)
1.2.1. Plot the points (1, 1), (—1, 1),
or (cf. (1.2.2a)) (— 1, —1), and (1, —1).
— #1 .2.2. Plot the points (1, 5), (5, 1), (—1, 5),
x2 X1 — _ X2
1

C0S “12 _ (-5, (-5, -1), (1, -5), and (5,-1).

y^
1),
r l2 x ^J r ^y 2 y^2
’
1.2.3. Plot the points (0, 4), (0, —4),
v*-y* (4, and (—4, 0).
0),
sin a ia = . 1.2.4. Find the distance from the origin
r i2
V (z 2 — a?i)
2
-h(i/2 — !/i) and the angle a for the points (1, 1), (2, —2),
(1.2.5a) (—3, —
3), and (—4, 4).
1.2.5. Find the distance between the follow-
The subscripts on the letters are known
1,1
ing pairs of points: A x (i, 1) and ^l 2 (l, 1), —
as indices(the Latin word index means to A^i, 1) and A 2 (— 1, —1), A x (2, 4) and
They are read: “x sub one” for x 11
indicate). A 2 ( 4, 2), and ^(—2, -4) and i4 2 (— 4, —2).
A sub two” for A 2 The same letter with . 1.2.6. Write out the coordinates of the
different indices is used in place of a variety vertices of a square with side a if the diago-
of letters to emphasize that we are dealing nals of the square coincide with the x and y
with similar (yet different) quantities. For axes.
instance, x 1 and x 2 are quantities on the x 1.2.7. Write out the coordinates of the
axis (both are abscissas), but they refer to vertices of a regular hexagon with side a if
distinct points. Quantities denoted by deffer- one of the diagonals coincides with the x
ent letters but the same index refer to one axis and the center lies at the origin.
and the same point: A 1 denotes a certain 1.2.8. (a) Write out the coordinates of the
point, x 1 denotes the abscissa of that point, vertices of an equilateral triangle with side
and y 1 denotes the ordinate of that same point. a, with the base on the x axis, and with the
We sometimes use double-index notation: vertex of the subtended angle on the y axis;
which is read “r sub one two” and not
r 12 , (b) the same if the base lies on the x axis and
“r sub twelve”, is the distance between the the vertex of one of the angles lies at the
first point, A lt and the second, ^4 2 . origin.
 30 1 Functions and Graphs

1.2.9. Given a point A x with coordinates

x 1 and y x Write out the coordinates of point
.

A 2 symmetric to A x about the x axis; the

same for A z symmetric to A x about the y
axis; the same for A 4 symmetric to A x with
respect to the origin.

Graphical Representation
1.3
of Functions. The Equation
of the Straight Line

In Section 1.2 it was shown that each

pair of values x, y is associated with a
definite point in the plane. If it is giv- Figure 1.3.1
en that y is a definite function of x,
then this means that to every value of
x there corresponds a definite value of the entire graph of the function: for
y. Therefore, if we are given a range of
any x, the corresponding point (x, y)
values of x we can find the various lies on the straight line that connects
,

corresponding values of y and these ,

any two points of the graph.
pairs of values, (x, y), will yield many How do we prove that, for any func-
points in the plane. If we increase the tion of the form y kx — +
b (for
number of distinct values of x by tak- arbitrary k and b), all points of the
ing them closer and closer together, graph lie on a single straight line?
then finally the points will merge into In other words, how can we determine,
a solid curve. This curve is called the without construction, merely by com-
graph of the function. puting from the values of coordinates,
Actually, only a few points suffice that three points, A x ,
A 2 and A 3l
to plot a graph, the intermediate points lie on a single line? It is clear that if

and the whole graph (curve) of the func- the angle a l2 between the segment
tion being obtained by joining the A X A 2 and the x axis is equal to the
points with a smooth curve. However, angle a 13 between the segment A X A 3
in order to avoid crude errors, we must and the x axis, this means that the line
have a general picture of the form of segments A XA 2 and A X A 3 belong to one
curves representing various functions. straight line, and if this is not so, then
We begin with a few of the more typical the segments belong to different
and important functions. straight lines. In Figure 1.3.2 we have
We consider the linear relationship the case where a 13 is greater than a l2
,

or linear function , and point A 3 lies above the extension

of A 4 A 2 the same figure shows that if
y = kx + b. (1.3.1)
,

a 13 were equal to a 12 then A 3 would be

,

Suppose, y 2x
say, 1.= We con-
+
struct a few points for which x and y
are given in the table below:

X
y 1312
0
5
3
7

Now plot these points on the graph in

Figure 1.3.1. It is immediately seen
that these points lie on one straight
line. In this case, we draw the straight
line (whence the term “linear relation-
ship” or “linear functions”) and obtain Figure 1.3.2
1.3 Graphical Representation of Functions 31

on the straight line that is the exten- of points of the graph, in particular for
sion of A t A 2 . the points A(x x y x )
,
and C(x 3 y 3 )+
,

From the expression of the tangent of we also get

angle a 12 (1.2.5), it follows that at
y »— yi —.j — ya— y\
<*12 —
a i 3 we have the following rela- x3 — X1
c
x2 —X 1
tionship between the coordinates of
points A x A 2 and A 3 , ,
: This means that for any three points
—y — Vi of the graph, A{x ly y x ), B(x 2 y 2 ),
V2 i Vs u g 2)/
,

x2 —x x3 —x \ and C(x 3 y 3 ), the relation (1.3.2) is

,
1 1
valid, which means that any three
Without using trigonometry, we can points of the graph lie on a single straight
say that condition (1.3.2) is a condition line and, hence, all points of the graph
of similarity of two right triangles of the function y kx =
b belong to +
A xA 2 B and A X A 3 C (see Figure 1.3. 2). The one straight line. Thus, the graph of
similarity of the triangles indicates that the function y kx —
b is a straight +
the angles at the vertex A x are equal. line, which we will often call, for tho
The relation (1.3.2) is also applic- sake of brevity, “the straight line y —
able in the case where point A x lies be- kx +6” (we will also call it “tho
tween points A 2 and A 3 (Figure 1.3.3); straight line (1.3.1)”).
if the three points lie on one line, then The equation y kx =
b is called +
from the similarity of the triangles the equation of the straight line. The
A X A 2 B and A X A 3 C follows the propor- coefficient k determines the angle betwe-
tion (1.3.2). In the example given in en the straight line and the x axis. Sub-
Figure 1.3.3, x 3 xx 0 and y 3 — < — stituting x =
0 into the equation yields
yx < 0, but their ratio is positive and
y =b, which means that one of the
equal to the ratio of two positive quan- points of the straight line is the point.
tities, x2 — x t and y 2 — y1 .
(0, b). This point lies on the y axis at
Now verify that condition
let us a distance b above the origin if b is
(1.3.2) remains valid for any triple of positive or at a distance b below the ori-
points of the linear function (1.3.1). gin if b is negative. Thus, b is the ordi-
Consider two points, A(x x y^) and , nate of the point of intersection of the
B(x 2 y 2 ), whose
,
coordinates obey straight line and the y axis (sometimes
Eq. (1.3.1). In this case y 1 kxx = + b is called the initial ordinate of tho
b and y 2 kx 2 5, =
with the re- + straight line), and b is the length
sult that y 2 y± kx 2 —
b (kx x = + — + of the line segment cut off by tho
| \

b) = k (x 2 —x x ), whence straight line on the axis of ordinates

(in Figure 1.3.1, b =
1), which is called
— Xi =
y% y i
Jc
X2 the y intercept.
To construct a straight line corre-
The ratio proves to be independent of
sponding to a given equation one need
x 1 and x 2 Hence, for any other pair
.
not compute the coordinates of a largo
number of points and plot them on tho
graph: clear that the construction
it is
of two
points fully determines tho
straight line passing through these two
points. For instance, we can always
take two points, y b for x =
0 and =
y = bk for x 1, =
and draw tho
line. For the second point we could also
take the point of intersection of the
straight line and the x axis, that is, tho
point with x —
x 0 and y 0. From tho =
condition y kx 0— & 0 we find+ —
<32 1 Functions and Graphs

the x axis, which means that it bisects

the angle between the x and y axes.
The “angle with the x axis” here stands
for the angle with the positive direc-
tion of the x axis (the one with the ar-
rowhead). An extension of the straight
line lying in the third quadrant forms
an angle a = —
135° with the x axis.
The entire straight line bisects the
angle between the x and y axes in the
first and third quadrants.

(3) For k = —
1, the part of the
straight line lying in the second quadrant
forms an angle a =
135° with the x
axis, while the extension of the line
in the fourth quadrant forms an angle
that Xq = — b/k, which is known as the a — — 45°. The entire straight line
x intercept. bisects the angle between the x and y
It is useful to do some drilling in the axes in the second and fourth quadrants.
construction of graphs so as to be able (4) If k
|
<
1, the straight line is
|

to glance at an equation and picture sloping, that is, is closer to the x axis
roughly the variation and the position than to the y axis, and the smaller the
of the curve in question. |
k |, the closer the line is to the x
This is easy to do when we have a axis. If k | 1, >
the straight line
|

linear function whose graph is a is steep, it is closer to the y axis than

^rtaight line. The line depends only on to the x axis, and the greater the k |, |

two quantities, k and b, of the equation. the closer the line is to the y axis.
Thus, not so many variants have to The quantity k is called the slope
be examined: k can be positive or neg- of the line; it is fixed by the value of
ative, k can be large or small in ab- angle oc between the straight line and
solute value (greater than 1 or less the x axis.
than 1), and 6 can be positive or neg- Now that this is clear, let us inves-
ative or even zero. Let us see how to tigate the general case of a straight
carry out such an investigation. line with 6 different from zero. Sup-
We start with the case 6 0, or = pose we have two straight lines: y =
the equation y =
kx. The straight line kx (say, with k = 0.5; see Figure
here will clearly pass through the ori- 1.3.5), that is, 6 0 in Eq. (1.3.1),=
gin, that is, through the point with and the straight line with the same
x = 0 and y =
0. Figure 1.3.4 depicts slope k but with 6 =£ 0 (say, y =
several straight lines with different 0.5a; + 2 in Figure 1.3.5). For the
k' s, whose values stand at the end- sake of convenience we introduce the
points of each straight line. Check the
correctness of each line and you will
feel sure of the following general con-
clusions:
(1) If k is positive, the straight line
lies in the firstand third quadrants,
while if k is negative, the straight line
lies inTthe second and fourth quadrants.
(2) By the foregoing, if k 1, the —
line lies in the first and third quadrants.
The part of the straight line in the first
^quadrant forms an angle a 45° with =
1.3 Graphical Representation of Functions 33

9\

Bfqs)

0 £
Figure 1.3.6

notation y 0 =
0.5x and y 2 0.5x = +
12 For each given
2. x the quan- ,

tity y 2 is two units greater than y 0 .

tive variation of the independent vari-
To summarize, then, the points of able. Indeed, for any two points on the
line y 2 are obtained from the points of A(x1?
straight line, y-f) and B{x 2 y 2 ),
,

line y 0 with the same x by an eleva- we have

tion of two units. The straight line y 2
therefore parallel to y 0 and lies two
y te) — y (si) „ fcg 2 +6 — (toi+6) __
is
— ^
units above it. Quite obviously, this
#2 *^2 — #1
rule holds true for any b (if b is nega- We have already calculated
this ratio
tive, the line lies below the origin and when we proved that a linear function
below the corresponding straight line on a graph is depicted by a straight
y = kx). line. In the general case of an arbit-
Nowthat we see how straight lines rary function y (x) the similar quan-
with equations y —
kx are located for tity, [y (x 2 ) —
y (.x 1 )\l(x 2 xfj is the —
distinct ft’s, we can readily picture the tangent of the angle between the seg-
general positions of straight lines y = ment connecting points (x u y (^)) and
kx +b with arbitrary ft and b. Exer- {x 2 , y (x 2 )) and the x axis (Figure 1.3.7a).
cises that will help you to drill this A linear function is distinguished by
material are given at the end of this the fact that this ratio is the same for
section. In the particular case of ft 0 = any two points: it depends neither on
the equation is y b (it is assumed= x 2 nor on x x (Figure 1.3.76). For this
that y =
b for all values of x), which reason, all points of a linear function
is associated with a horizontal straight belong to a single straight line.
line with a slant (slope) of zero (Fig- Note, in addition, that a straight line
ure 1.3.6). parallel to the y axis (such a straight
We can imagine a man walking from line does not represent any function)
left to right in the direction of increas- is written as x —
a (it is assumed that
ing values of x. If ft 0, then he walks> x = a with y arbitrary; see Figure 1.3.8,
uphill (a positive slope), while if ft 0, < where a is negative). It is natural to
the man walks downhill (a negative assume that the ft of such a straight line
slope); if ft =
0 (zero slope), the man is infinite (ft =
oo), since in this case
walks along a horizontal path.
The slope ft indicates the ratio of the
variation of the function to the respec-

1,2 The subscripts here

are used somewhat
differently from our earlier practice: y 0 refers
to the entire line and not to the ordinate of
a point— it is the ordinate of an arbitrary
point on a line with given k and 6 0. In =
the same manner, y 2 is the ordinate of an
arbitrary point on a line with given k and
6 = 2. In other words, neither y0 nor y2
are numbers; they are functions: y 0 = y 0 (x)
and y2 = y 2 (x). Figure 1.3.8

946
 34 1 Functions and Graphs

1.4 Inverse Proportionality and

the Hyperbola. The Parabola
Let us recall the idea of direct propor-
tionality of two quantities. The formu-
la

y = kx (1.4.1)
(geometrically, as we already know,
thisformula represents a straight line)
means that y is proportional to x: an
increase in x by a certain factor leads
to an increase in y by the same factor.
the ratio k (y 2 =
yi)/(x 2 —24 ), where — In other words, for any two points
points (#!, i/x) and (x 2 / 2) belong to
,
1
A(xi, y x ) and B(x 2 y 2 ) whose coor-
,

the straight line, has the form c/0 (the dinates satisfy (1.4.1) we always have
straight line x =
a is one with an in-
yJyi =
x 2 /x 1 (Figure 1.4.1a). It is also
finitely great slope). commonly said that formula (1.4.1)
Now we can easily find the equation expresses the fact of direct proportion-
of a straight line passing through a giv- ality between y and x (with the propor-
en point M
(£ 0 y 0 ) and having a given
, tionality factor, or coefficient, k). The
slope k. Such an equation will have the more general equation
form of (1.3.1), with k known, but b
has yet to be found, and the values x 0 y = kx + b (1.4.1a)

and y 0 must satisfy this equation (Fig- characterizes the proportionality (or
ure 1.3.9). This suggests the form of direct proportionality) between the in-
the sought equation: crement x 2 —
x x of the independent vari-
—y = k (X — X 0 ), (1.3.3)
able and the corresponding increment
y 0
y2 —
yi of the function (dependent vari-
or able): for any two increments x 2 —
xx
y = kx + (y 0 — kx 0 ). (1.3.3a) and x 4 — x 3 of the independent vari-
able of the function (1.4.1a), the cor-
Indeed, the slope of the straight line
responding increments of the function
(1.3.3a) or (1.3.3) is k; on the other
= (dependent variable) will be propor-
hand, if we substitutes x 0 and y y0 = tional to the increments of the inde-
into both sides of Eq. (1.3.3), we ar-
= pendent variable (Figure 1.4.1&):
rive at an identity, 0 0.
1/2 — yi _ y4 Us
Exercises x2 —X 1 x\ —x 3

1.3.1. Determine whether the point tri-

ples lie on a straight line: A 1 (0, 0), A 2 (2, 3),
A3 (4, 6); A1 (0, 0), A 2 (2, 3), A 3 (-2, -3);
and A x ( 2, —3), A 2 (4, —6), A 3 (— 2, 3).
1.3.2. Construct the straight lines y 3x, —
y =3x +
2, y 3x 1, = —
y 2 x, y — — =
2 —0.5a:, and y x 3. = —
1.3.3. Find the equation of (a) a straight
line that passes through point A (1, 1) and
—
has a slope equal to —1, and (b) a straight
line that passes through point B (2, 3) and
has a slope equal to 2.
1.3.4. Prove that the equation of each
straight line that has a nonzero slope (this
slope must not be infinitely large), that is, a
straight line not parallel to either axis, can be
written in the form x!a y/b +
1 (the inter- =
cept form). What geometrical meaning have
the quantities a and b in this equation?
 35
1.4 Inverse Proportionality and the Hyperbolas

1.4.2 depicts a “unit” hyperbola

and below we give the values of y for

some values of x for this hyperbola:

x —1 —0.1 —0.01 —0.001 0.001 0.01 0.1 1

y —1 — 10 —100 —1000 1000 100 10 1

The hyperbola has the peculiarity that

for small positive x the value of y is
very large, while for small negative x
the value of y is also negative and very
large in absolute value.
This property of curve (1.4.2a) (or the
Another often encountered relation- general curve (1.4.2)) is often expressed
ship between y and x is by stating that at x = 0 we have
y
— 00 ,
where the plus or minus sign
is chosen depending on the side from
which we approach x 0. The = exact
Such a dependence is known as in- meaning of “y ±00 at x =
= 0” is
verse proportionality (with coefficient that for a sufficiently small x the value
k) between y and x: if points A(xx z/
x) ,
of y become arbitrarily large (in abso-
and B(x 2 y 2 ) satisfy Eq. (1.4.2), then
, lute value), with it being positive or
x 2 is greater than x x by a factor by negative depending on the sign of x
which y 2 is less than y x or x 2 /x 1 = yjy 2
,
(the symbol 00 is read “infinity” and
(since y2 -r- yx — klx 2 — k!xx — isof course not a number). The smaller
xi -T- #2)* the value of x we choose (in absolute
Note that direct and inverse pro- value, of course), the higher (or the
portionalities are reciprocal if y is di-
rectly (inversely) proportional to x
:
lower in the case of x 0) the cor- <
,
responding point lies on the hyperbola:
then x is directly (inversely) propor- both of hyperbola (1.4.2a)
branches
tional to y, and vice versa. However, “go to infinity” (in the positive or
off
while the proportionality factor in negative direction) along the y axis
direct proportionality (the factor that (upward or downward) and approach
connects x with y) is the inverse of the the y axis without limit (but never in-
proportionality factor in the (direct) tersect it), since the abscissa can be
proportionality relation that connects arbitrarily small. Similarly, as x in-
y with x (if y = kx, then a; = (1 Ik) z/),in creases without limit (in absolute
the inverse proportionality between y value), that is, when x “tends to 00
”
+
and x the coefficients of (inverse) pro- or “tends to —
00 (we remind the read-
”
portionality connecting y with x and x er once more that the symbol 00 does
with y are the same (if y klx, then — not correspond to a number), the value
x = k/y). of y =
l/x becomes arbitrarily small
The curve corresponding to Eq. (1.4.2) (in absolute value): the hyperbola “goes
is known as the hyperbola. 1 * Figure off to infinity along the x axis”, appro-

We
are speaking here only of equilateral
aching this axis without limit but never
hyperbolas since by “hyperbola” mathemati-
,
intersecting it (the two branches ap-
cians mean a somewhat more general curve than proach the x axis from different di-
the one specified by Eq. (1.4.1). However, rections). This property of the hyper-
nowhere in our narrative do we use such
bola is expressed briefly by saying that
general hyperbolas, and so the adjective
“equilateral* will always be dropped. the straight lines x =
0 (the y axis)
3*
 —

36 1 Functions and Graphs

tion along the y axis (or by a “shrink-

ing” transformation toward the x axis),
since each point of the curve (1.4.2)
is closer to the x axis than the respec-
tive point of the curve (1.4.2a) by a
factor of k (in this case the ratio y -r-
y0 = AP -f- A 0P = k is less than
unity). 1 - 4

Figure 1.4.3 shows the curve # 1/x =

(the dashed curve), the curve y 21 x =
(the solid curve), and the curve cor-
responding to the function y ( H2)lx =
which corresponds to a negative value
of k equal to —
1/2 (the dotted curve);
this last function is negative for posi-
tive x and positive for negative x .

The direct and inverse proportion-

alities are often encountered in physical
laws. For instance, Ohm's law (1.1.3)
and y =
0 (the x axis) are the asymp- states that current i in a conductor
totes to the hyperbola (from the Greek changes in direct proportion to the po-
word asymptotos meaning “not meet- tential difference u and in inverse pro-
ing”). portion to the resistance R of the con-
As seen from Figure 1.4.2, the hyper- ductor. For a given i?, the current i
bola (1.4.2a) consists of two parts, or is directly proportional to u (with a
branches, corresponding to x 0 and> proportionality coefficient equal to 1 IR)\
x < 0; these branches are separated, on the other hand, for a given u the cur-
that is, they do not intersect. rent i is inversely proportional to i?,
The arbitrary curve (1.4.2) can be or i =
k/R, where potential difference
obtained from the curve (1.4.2a) by u plays the role of k. A similar simple
a simple transformation. Suppose that relationship s =
ut exists between the
A 0 (^o> yo) is a point of the curve distance s traveled in uniform motion
(1.4.2a), that is, y 0 =
1 lx 0 (see Fig- with a velocity v in the course of a time
ure 1.4.3., in which the curve (1.4.2a\ interval t. It shows that t s/v., or =
is shown by a dashed curve). In this that the time of traversal is directly
case the same value x —
x 0 has corre- proportional to distance s and inverse-
sponding to it on curve (1.4.2), where ly proportional to velocity u. Fi-
we assume that k is positive, a point A nally, in Boyle’s law, pressure is in in-
with coordinates x 0 and y k/x 0 = = verse proportion to volume of gas.
k(ilx 0 ), that is, A(x x0 =y ,
= Examples abound. The formulas s =
ky 0 ). This point was obtained from vt and t = s/u show that, in uni-
point 4 0 by a “stretching” transforma- form motion, the distance s traveled is
tion along the y axis (or to put it dif- proportional to the time t of traversal
ferently, by a “stretching” transfor- with the coefficient of proportionality
mation away from the x axis) with a equal to v, while t is proportional to s
transformation coefficient equal to k, with coefficient llv, that is, if, say,
that a fc-fold increase in all vertical
is, v —
20 m/s, then the above-noted coef-
—
dimensions it is farther away from ficients of proportionality are 20 m/s
the x axis than point A
0 by a factor of k. and 0.05 s/m, respectively. The situ-
Here we assume that k >
1; when k ation with Boyle’s law pV constant =
is positive but less than unity, the
curve (1.4.2) is obtained from the curve 1 » 4 On the connection between curves
(1.4.2a) by a “shrinking” transforma- (1.4.2) and (1.4.2a) see Exercise 1.4.3.
1.4 Inverse Proportionality and the Hyperbola 37

is different. Here, if the temperature

is 0 °C and the volume V and pressure p
are measured in m 3
and Pa (pascal;
see Appendix 5: 1 Pa = 1 N/m 2 =
1.450377 10~ 4 lb (wt)/in. 2 ), it can
X
be proved (see Section 11.2) that pV =
0.8737 N*m, and we see that p is
inversely proportional to V, and V
is inversely proportional to p and in ,

both cases the coefficients of proportion-

ality are equal to 0.8737 N*m.
These examples proved a good illus-
tration the difference between the
of
coefficients in direct proportionality
y =kx and inverse proportionality y =
kjx. If x is measured in units of e 1
and y in units of e 2 then k has the di- ,

mensions of e 2le i, whereby there is no

way in which k can serve as the coef-
ficient k' in x =
k'y, since the latter
coefficient must be measured in units of
eje 2 The situation is opposite with k ly
.

which has the dimensions of e1 e 2 This . Figure 1.4.4

implies that may (and actually ^
does) coincide with the coefficient k[
in x = k'Jy. both in sign and absolute value. This
Let us now turn to the quadratic means that the curve is symmetric
function y = ax 2 bx + + c. We start about the y axis .

with the simplest case: The curve (1.4.3a) is shown in Fig-

ure 1.4.4 by a solid line. It is called
y = ax 2 , (1.4.3)
a parabola and this term will be used
,

where coefficient a may be arbitrary. for all curves (1.4.3) with arbitrary a
For the sake of simplicity we take (and for curves of a broader class, as
a — 1, that is, we consider the curve
1 5 *
we will see below). For every positive
value of a, the curve (1.4.3) possesses
y = x2 . (1.4.3a)
the same properties 1 to 3 as the curve
What general properties does this (1.4.3a). Indeed, the transition from
function have? Eq. (1.4.3a) to Eq. (1.4.3) in all re-
1. It is always true that y > 0, both spects is similar to the transition from
for x > 0 and for x < 0. Only for x = 0 the “unit” hyperbola (1.4.2a) to the
do we have y = 0. This means that “general” hyperbola (1.4.2) (with a pos-
the entire curve lies above the x axis and itive coefficient k ): curve (1.4.3) is ob-
touches the x axis only at the origin. tained from curve (1.4.3a) by stretch-
2. y has a minimum (smallest value) ing all the dimensions along the y
at x = 0. The minimum is equal to 0. axis a-fold (see Figure 1.4.4, where we
On the graph, the minimum is the low- have depicted parabolas y = x2 and
est point of the curve. y = 2x2 ).
3. Associated with two values of x What will happen if a <
0? Consid-
identical in absolute value but with er an example with a = —
2, that is,
opposite signs are values of y identical the curve y = —2a: 2 The dotted line
.

(the one below the x axis) in Fig-

1,5 In a certain sense this case can be ure 1.4.4 depicts this curve. The prop-
thought of as general (see Section 1.8). erties of this curve are:
38 1 Functions and Graphs

y i
y’(x-3) z y i
\
y=ffa) y=ffjc-n)

^-X z
\
\
\ /\
!

/
/
/ (x0 ,y0 )J
/ yf
'do * n, yd
/ \ 1

\
\
N_
/ \ !
J s**
’

/
/^\
1
'
I

ri x 0 Xq JOq + fl X
Figure 1.4.5
Figure 1.4.6

1. y <0
arbitrary x =^= 0. The for istic point x — x 0 (a kind of notch, so
whole curve below the x axis and lies
to say). For example, at this point the
touches the x axis at the origin. function may have, say, a salient point
2. The function has a maximum value
or a maximum or merely assume some
at x =
0. The maximum is equal to definite value y 0 (Figure 1.4.6). Then
0. On the graph the maximum is the that same value y 0 or the same salient
uppermost point of the curve. point or the maximum will appear on
3. The curve is symmetric about the y the graph the new function y ±
of =
axis, just as in the case of a positive.
f (x — ri) whenthe independent var-
Now let us consider a more general iable in the function / is equal to the
equation old value x Q that is, at x ,
n x0 — = .

y = a (x — n) 2
. (1.4.4) This means that now the coordinates
of the notch are x x0 n, y 0 = + =
We take a 1 and n =
3. The respec- = It is clear then that any notch,
/ (£ 0 ).
tive curve is depicted in Figure 1.4.5. as it were, moves together with the
This is the same parabola as y x2 = whole graph to the right: point (# 0 y 0 ) ,

but it has been displaced rightward on the first graph has corresponding
three units along the x axis. to it on the second graph the point (x 0 +
This simple fact is not usually real- n, y 0 ) (see Figure 1.4.6 and the solid
ized as readily as it should be. If a and dashed parabolas in Figure 1.4.5,
function y —
f (x) is given and we com- where for the notch we can take the
pare it with another function y = point x 0 =
0, y 0 f (x 0 ) 02 0).= = =
f (x ri), —
the graph of the second All this is very simple and elementary,
function is shifted rightward from that but it is extremely important and the
of the first by n units (assuming n is student should not merely learn it but
positive). It is assumed here that in fully comprehend the meaning of it.
both cases / is one and the same func- The first urge of most students is to
tion. In our example, the symbol / say that when we replace y x2 with —
denotes a squaring of the independent
y = (x — 2
3) the curve is displaced to
variable, that is, the quantity inside the left because we subtract 3 from the
the parentheses: value of x. It is well worth your time
/ {x) = x2 , f (— x) = {—x) 2 = x2 ,
to make a detailed analysis of the exam-
which demonstrate this common
f(x- 2) = (s - 2) 2
,
ples,
error.
/ (x — n) = (x — ri)
2
, Now we can state the general rules:
x 2
= {x
2 2 = x 4
and so on. 1. = a (x ri) 2 has the
The curve y —
vertical line x — n for its axis of sym-
f (. ) ) ,

But why is the graph shifted to the metry.

right? We
will go into this in more de- 2. This curve, for a 0, lies above >
tail. Suppose the graph of the function the x axis and has a minimum y 0 =
y =
f (x) has some kind of Character- at x =
n, while for a 0 it lies below <
1.4 Inverse Proportionality and the Hyperbola 39

lie atthe intersection of the respective

curve and its axis of symmetry.
Removing the brackets in the expres-
sion y a (x =
n) 2 —
yields +m
y = ax 2 — 2 anx + an + m2
. (1.4.5a)

On the right-hand side of (1.4.5a) we

have a polynomial of degree two, which
in its most general form has the nota-
tion

y = ax2 + bx + c. (1.4.6)

This formula can be transformed as

follows:

y= a (
x2 +^ x +i)
~ a (*2 +ir x+ -&)+ a
(i-^)
=a (
x
+iY+( -£)- c

Hence,

+ c = a (* + i ) + (e - £)
2
ax * + bx ,

the x axis and has a maximum y 0 =

at x =
n. which implies that curve (1.4.6) is also
Finally, there is yet another modi- a parabola with an axis of symmetry

fication of this equation that does not x = —b/2a and a minimum point or
alter the shape of the curve. Let us
maximum point (-— 6/2a, c — 6 2 /4a).
consider the function Using the graph of a parabola, we
can investigate the solution of a quad-
y = a (x — n) + 2
m. (1.4 5) ratic equation and the various cases that
arise in this connection. We can ap-
This curve (also a parabola) clearly proach the solution of the quadratic
differsfrom the preceding one (without equation
m) solely in the vertical displacement
by the quantity m. The position of the ax2 + bx +c= 0
axis of symmetry of the curve remains
unchanged; for a >
0 the function has this way: consider the curve
a minimum at x = n and the value of
the function at the minimum is equal y — ax + bx + c = a
2

to y =m(the minimum, together with

the whole curve, was shifted by the
amount m), while for a 0 the point <
(x =n, y =
m) is the point of maxi- and find the points of intersection of
mum of the curve. this curve with the x axis. At these
Two examples will suffice (Figure points we have y 0, and so the val- =
1.4.7): y (x 3)= —
2
2 +
and y = ues of x corresponding to the points
—(x — 3)
2
+ 2. The axes
of sym- of intersection are the roots of the
metry in both parabolas are depicted quadratic equation.
in Figure 1.4.7 by dotted straight lines: But we know that the curve y =
the minimum point in Figure 1.4.7a ax2 bx+ c is a +
parabola. We
and the maximum point in Figure 1.4.76 also know that this parabola has an axis
 b

40 1 Functions and Graphs

instant of tangency, will merge. That

is why, when c b 2 /Aa 0 we speak— = ,

not of one root but of two equal (co-

incident) roots of the equation.
The case a <
0 is considered in a sim-
ilar manner. The respective curve has
a maximum and the limbs point down.
The reader is advised to draw the curves
and to verify that (a) when a 0 and <
c — 6 2 /4a <
0 simultaneously, there are
no real roots, (b) when a 0 and c < —
Figure 1.4.8 b 2 /Aa 0 >
there are two real roots,
,

and (c) when a 0 and c <

6 2 /4a 0, — =
of symmetry, the vertical line x = there are two equal roots (tangency).
— bI2a and that for a
,
0 the pa- > The ordinary formula for the roots
rabola has a minimum on the
point of a quadratic equation is
axis of symmetry and the altitude
(ordinate) of this minimum is y c = — x i,2—
— b zb V — 4 ac 2

2a *

6 2 /4a (if we glance at the right-

hand side of the last formula, we see The equation has two real roots when
that it has the customary form a (x — we are able to take a square root of
n) 2 +
m). For a 0, >
the limbs of b 2 — 4 ac, that is, when b 2 — 4 ac 0. If >
the parabola point upward. we write this as
It is clear that if the minimum lies
above the x axis, the parabola does b2 — 4ac = — 4 a (c — >0,
not intersect the x axis at any point
(Figure 1.4.8, curve 1). This means that we see that this condition is satisfied
when a 0 and c> 6 2 /4a— >
0 sim- when
ultaneously, the quadratic equation
has no roots. 1 6 But if the minimum
- ( 1) a> 0 ,
c ——< 7 0 ,
and ( 2) a< 0 ,

lies below the x axis and the limbs of

the parabola point upward, the para-
bola will definitely intersect the x axis
at two points; these points will be sym- These are the two cases of the existence
metric with respect to the straight line of two roots, which were obtained ear-
x = n = —
6/2a, the axis of symmetry lier from a consideration of the curves
of the parabola (curve 2 in Figure 1.4.8). corresponding to the function y =
This means that when a >
0 and c — ax2 bx + c. +
b 2 !ka <
0 simultaneously, the equa-
We can approach the question of solving
tion has two roots x x and x 2 as shown
quadratic equations from a somewhat differ-
in Figure 1.4.8. ent angle. Let us divide all terms in the equa-
Finally, there may be an intermedi- tion ax 2 bx +
c + =
0 (where, of course,
ate case where the parabola touches the a 0) by a. The result is

x axis (curve 3 in Figure 1.4.8). This x2 + px + q = 0, or x 2 = —px — q ,

(1.4.7)
case occurs when c —
b 2 /4a 0. If= where p = bia and q = c/a. Let us plot in the
we gradually move curve 2 upward, it xy- plane two functions:
will finally coincide with curve 5, the
Vt = *2 (1.4.7a)
two roots x 1 and x 2 will come closer to
each other and, ultimately, at the (this is the “unit” parabola (1.4.3a)) and

y2 = —px — q (1.4.7b)
1
Here by “root” we mean a real root of
* 6

the equation; for the time being, until Chap- (a straight line). Clearly, for the values of x
ters 14 and 15, we will simply ignore complex- that satisfy Eq. (1.4.7) we have y x y2 = ,

valued roots. that is, these values of x correspond to the

1.4 Inverse Proportionality and the Hyperbola 41

Figure 1.4.10

Figure 1.4.9

and c in the equation of the parabo-

points intersection of parabola (1.4.7a)
of la (1.4.6).
with straight line (1.4.7b). But parabola
(1.4.7a) can be plotted with great accuracy
An exact definition of convexity is
(using, say, graph paper), so that to solve this: take two points A{x x y x ) and,

any quadratic equation it suffices to draw the B{x 2 y 2 ) on a curve and draw a line
,

straight line (1.4.7b) (say, finding any two through them (Figure 1.4.10). If the
points of the straight line and using a ruler)
on the same graph paper. The roots can then
portion of the curve between the two
be “read off” the paper. The straight line points lies below the straight line (be-
(1.4.7b) may intersect the parabola (1.4.7a) low the chord AB
of the curve), we say
at two points (as line 1 in Figure 1.4.9 does) that the curve is convex down while ,
or touch it at a single point (as line 2 in Fig-
if the portion of the curve between the
ure 1.4.9; two coincident points of intersec-
tion with the parabola) or even have no two points lies above the straight
points in common with parabola (1.4.7a) (as line (above chord AB
), we say that the
line 3 in Figure 1.4.9). All this corresponds to curve convex up.
is
cases where Eq. (1.4.7) has two distinct roots
or two coincident roots (i.e. one root) or no
The convexity of a parabola is readily
roots at all. seen in a drawing, but we can also
The aforesaid also enables us to conclude define it algebraically. Take arbitrary
the following. We know that the equation values xx and x 2 of the abscissa x. They
+ +
ax 2 =bx c 0 has a single root if and
are associated with points on the pa-
—
only if c= b 2 /4a 0 whereby the equation
x2+ + =px q
,

0 (the case with a 1) has = rabola, A(x i, y x ) and B(x 2 yz), where ,

a unique root if and only if yx= ax{ bx ± +

c and y2+ ax\ = +
bx + 2 c.

We wish to
find the coordinates of
point M
lying at the midpoint of the
Thus, (1.4.8) is the condition for tangency of
the straight line (1.4.7b) and parabola (1.4.7a);
line segment AB
(Figure 1.4.10). It
in other words, the straight line y kx b = + may be demonstrated geometrically
touches parabola y =
x 2 if and only if 1 7 * that if AM — MB
the coordinates of
,

point Af(x 0 y 0 ) are arithmetic means

h*+4b = 0. (1.4.8a)
,

of the coordinates of A and B :

In Chapter 2 we derive the same condition

(1.4.8a) for tangency of a straight line and a
parabola using another method. = and „0 «a+L’.

Observe, finally, that, depending on

2 (This follows from the fact that in
the sign of the coefficient a of x in the
Figure 1.4.10 the length of the mean-
equation of the parabola (1.4.6), the
curve is convex down (a 0) or convex > line 0 of Mx
the trapezoid ABx 2 x x is
equal to one-half of the sum of the
up (a 0). <
This property does not
lengths of the bases Ax x and Bx 2 of thu
depend on the values and signs of b
trapezoid.) Now let us find the coordi-
1«7
It is clear that if in (1.4.8) we substitute nates of the point N(X, Y) lying on
k for —p and b for —q ,
we arrive at (1.4.8a). the parabola for the same valun
 . k

42 1 Functions and Graphs

Exercises

1.4.1. Plot the following curves: y — 3/ar,

y = — 0.5 lx, and y = 1/x + 3.

1.4.2. Plot the following curves: y =
x — 2x + 2 and y = 2x + Ax.
2 2

1.4.3. Show that the hyperbola y = k/x,

where k positive, can be obtained from the
is
hyperbola y = 1/x by (a) a stretching transfor-
mation from the origin O (a homothetic trans-
formation) with a coefficient (for k Y
1 <
it is more appropriate to speak of a shrinking
transformation), which means that if M' and
M
are points belonging to the curves y k/x =
and y =
1/x and lying on the straight line
that passes through the origin 0 then OM' ,
~
OM = Yk\ and
(b) a stretching transforma-
tion along the x axis (or away from the y axis)
with a coefficient k (here, too, for k 1 it is <
more appropriate to speak of a shrinking
transformation).
1.4.4. Prove algebraically that for k 0
the hyperbola y —
k/x is convex down when
= fa + x 2 )l 2 of the abscissa. Here x I> 0 and convex up when x 0. <

Y = aX 2 +bX + c 1.5 Higher-Order Parabolas and

Hyperbolas. The Semicubical Parabola
= a (^l±^) 2 + &£l±^ + c.
The curve given by the graph of the
function
The reader can assure himself that
axn
y = (1.5.1)

—W
Y-j -«(a±^) -(a4 + «T')
2
(where n is a positive integer) is often
1
called a parabola of order n. For in-
stance, Figure 1.5.1 depicts a third-
order parabola (or a cubical parabola)

y = x* (1.5.2)
{since the terms involving b and c
1 and a fourth-order parabola
cancel out). But the quantity
^ j
y = £4 - (1.5.3)
is positive arbitrary xx and
for
x 2 Consequently, for a
. 0 we have > We see that a fourth-order parabola
Y < y 0 or the point on the parabola
, (1.5.3) resembles an ordinary (second-
lies below the corresponding point on order) parabola: it has a symmetry axis
the straight line (i.e. having the same (the y axis), at point (0, 0) it has a
abscissa ), X
that is, the parabola is minimum (the only one), and at the
convex down. On the other hand, for origin 0 it touches the x axis that does
a <0 we have y 0 Y, and the pa- < not intersect it (just as the parabola
rabola is convex up. V =z 2 )-
The hyperbola y k/x (where we = The cubical parabola y x? po- =
nssume that k is positive) consists of ssesses quite different properties. It has
two branches. The reader can clearly neither maxima nor minima: an increase
see that the branch of the hyperbola in x always brings about an increase
corresponding to positive x's is convex in y, that is, as a point moves from
<1 own . w
hile__tbe .second. _hr anrJi_ nf the
7
Jftft_to ri^ht^^on Abe^cjiryn^ . it always^ ~
hyperbola is convex up (see Fig rises. It is said that the function (1.5.2)
ure 1.4.11 and Exercise 1.4.4). increases for all values of x (what we
 x x

1.5 Higher-Order Parabolas and Hyperbolas 43

metric with respect to the origin (9,

which evenly
divides the line seg-
ment BB l into two parts (Figure 1.5.1a).
Figure 1.5.2 depicts a curve defined
by the formula
y = + x. (1.5.4)
Figure 1.5.1
This curve also has the distinguishing
feature that on any portion of it an
increase in x brings about an increase
mean by this is that the function in-
in y and the curve constantly rises
creases with the independent variable
from left to right, just as the function
and that this behavior does not change).
Curve (1.5.2) has no axis of sym- y —x3 does. The curve (1.5.4) has
neither maxima nor minima. Quite ob-
metry: the values of y in this case are
viously, such a curve cuts the axis of
negative for negative x and positive
abscissas only once, at point 0 .
for positive x. Instead it has a center
Figure 1.5.3 shows a curve construct-
of symmetry the origin 0 (0, 0). In-
,
ed on the basis of the formula
deed, if we take the curve (1.5.3), we
can see that to every two values of the y = o? — x. (1.5.5)
independent variable that differ in sign
As is evident from the graph, this curve
but not in absolute value, x 0 and —
0 ,
has two portions where y increases with
there corresponds only one value y 0 = x: for negative x <Z —0.58 and for
of the function (or two coinci-
xl
dent values), that is, to each point
positive x >
0.58. Between them, on

A(x 0 y o) 011 curve (1.5.3) there cor- the interval 0.58 x— < <
0.58, the
,

responds a point —
A ± ( x 0 y 0 ) sym-
,
function is decreasing: y decreases as x
grows.
metric about the y axis (Figure 1.5.16).
Now if we take the function (1.5.2),
we see that to every two values of the
independent variable that differ in
sign but not in absolute value, x 0 and
— x 0l there correspond two values of
the function that differ in sign but not
in absolute value, namely, y0 =
x\
and — y0 = —
x s0 so that to each point
,

B(x o, y 0 ) on the parabola (1.5.2) there

corresponds a point B x (— 0 , —
y 0 ) sym- Figure 1.5.3
 a

44 1 Functions and Graphs

order three. The cubic equation that

we get by equating the respective
polynomial to zero has one (real) solu-
tion (or root), x 0.48, in the case of ~
the upper curve, and three roots, x x 1, =
x2 =
2, and x 3 3, in the case of the =
lower curve. It is easy to see that a cu-
bic equation always has at least one real
root: to be sure of this, the reader is ad-
vised to examine the behavior of the
curve y ax 3 =
bx 2 cx d for very + + +
large (in absolute value) positive and
negative values of x.
The graph of the cubical parabola (1.5.2) can
be used to solve (approximately) an arbitrary
equation of degree three. Let us write the
general equation of degree three (the cubic
Figure 1.5.4
equation) as
The function (1.5.5) has a maximum xs + 3 ax 2 + bx + = c 0 (1.5.6)
at (x —0.58, ~ y ~ 0.38).
1 8
-
In
(we will find it convenient to denote the coef-
this context, the word “maximum” ficient ofx 2 by 3 a rather than by a). We trans-
does not mean that in the given case form Eq. (1.5.6) thus:

y
~
0.38 is the greatest possible value (x 3 + 3 ax 2 + 3a x + a + 2 3
)
(b — 3a 2
)
x
of y given by the expression (1.5.5), =
since it is clear that for large positive
+ (c~ a3 ) 0,

values of x the quantity y will assume or

So what is so
arbitrarily large values. (x + a) — (3a — (x + a)
3 2
b)

conspicuous about the maximum point — [(a —

3
— a (3a — =c)
2
6)] 0,

{x —0.58, y ~
0.38) of the function or, finally,
(1.5.5)?
X3 — KX — B = 0, (1.5.6a)
As the graph
of this function shows,
at this point y is greater than at ad- where X = x + a, K = 3 2 —b and B =
ab — c — 2a
,

3
jacent points. The point of maximum .

Clearly, solving Eq. (1.5.6) is the same as

separates the portion of the curve where solving Eq. (1.5.6a), which can also be re-
the function is growing (to the left of written as
the maximum) from the portion where X3 = KX + B. (1.5.7)
the function is decreasing (to the right
Hence, if we can sketch the cubical parabola
of the maximum). This is what is
called the local (or relative) maximum: y =1°wiin gre'ai accuracy qsay; using gra^n
paper), we need only construct on the same
the value of y at this point is greater graph the straight line y 1 B and = KX +
than the values of y at other points, but find the points (or point) of intersection of the
only for values of x that are not too straight line with the cubical parabola— the
result is the solution, or the approximate val-
far from x max (in our case x max ~ ues of the roots of Eq. (1.5.6a) (and hence of
—
0.58). Similarly, at point (x ~ Eq. (1.5.6)).
+0.58, y ~—
0.38) the function has For example, in the case of the equation
a local (relative) minimum. x3 — +
6z 2 llx— = 4 0 (see Figure 1.5.4)
we have a = — =
2, b
= —
11, c 4, whence
In Figure 1.5.4 we have two more exam-
# = 3a — 6 = 3x4 — 11 =
2 and B = 1
ples of curves describing polynomials of ab — — 2a =-2Xll+4 — 2X (—8) =
c
3

—2. Thus, to find the solutions to our equation

1 * 8
In Section 2.6 we will learn how to we need only find the points of intersection of
find the maxima and minima of a function.
the = X
parabola y 3
and the straight line
For example, in the case of formula (1.5.5)
y1= X— 2. Figure 1.5.5, yields the approx-
we will find that £ max — —UY 3 ~ 0.57735,
'

imate solution: X ~ —1.5, whence x =

i/max = 2/(3/3) ~ 0.3849. X — a ~ —1.5 + 2 ~ 0.5.
1.5 Higher-Order Parabolas and Hyperbolas 45

and a hyperbola of degree three

y=fs- (i.5.io)

As one can easily understand, curve

(1.5.10) resembles an ordinary hyper-
bola (1.4.2a): both consist of two
branches (symmetric about the origin O),
and at x = 0 the value of y is +oo
or —
oo (compare the results of Sec-
tion 1 .4 concerning the hyperbola (1 .4.2a)
with this result). On the other hand, the
figure 1.5.5 hyperbola (1.5.9) differs from an ordi-
nary hyperbola, since here y is positive
Similarly, in the case of the equation for x both positive and negative. This
x3 — 6x +H x — 6 = 0 (see Figure 1.5.4)
2
curve also consists of two branches,
we have a — —2, b — 11, = —6, K = c
3a — 6 =
2
and B — ab — c — 2a = 0. Here,
1,
2 which are symmetric about the
obviously, the points of intersection of the straight line x = 0 (the y axis) rather
cubical parabola y — X s
with the straight
line y = X correspond to values of X that
than about O. For the function y l/x 2 =
2
are —1, 0, and +1, which implies that the
we can symbolically write y = +oo
roots x = X — a = X + 2 of the initial at x = 0 (this means that for small
equation are equal to 1, 2, and 3 (see Fig- values of x the quantity y may become
ure 1.5.5). as large as desired).
The graph of the function
The curve that represents the func-
tion y = ±x>t 2 -=±V^ 3
, (1.5.11)

y= (1-5.8) or, more precisely, the curve with the

equation
where n a positive integer, is some-
is
times called a hyperbola of order n 1 + y
2 = x3 (1.5.11a)
or a hyperbola of degree n. For example, is known as a semicubical parabola (Fig-
Figure 1.5.6 depicts a hyperbola of ure Since no values of y cor-
1.5.7).
degree two
respond negative values of x (for
to
negative x’s the right-hand side of
(1-5-9)
(1.5.11a) is negative, which is impos-

Figure 1.5.6
 x

46 1 Functions and Graphs

sible since y 2
is never negative), the
entire curve lies in the right half-
plane. Since to each value of x there
correspond two values of y that differ
in sign but not in absolute value, pre-
cisely, +]/ x? and — y^
the curve 3
,

is symmetric with respect to the x axis:

to each point on the curve, A(x Q y 0 ), ,

there corresponds a point symmetric

about the axis of abscissas, A x (^ 0 y 0 ); ,
—
below, in Section 2.5, we prove with
sufficient rigor that the semicubical
parabola at the origin 0 touches the x
axis.

These examples illustrate the behavior of

all power functions that is, functions of the
,

type

y =xn , (1.5.12)

where the exponent n may be positive or non-

positive, greater or smaller than unity in
absolute value, an integral or a fractional
number.
If n> 1, the curves representing (1.5.12)
behave like the quadratic parabola (1.4.3a)
(we will consider only nonnegative values of x A
since the very definition of the quantity xn (b)
with x negative and n a noninteger presents
certain difficulties; for instance, x 1 ! 2 = V Figure 1.5.8
does not exist for x negative). All curves of
this type touch the x axis at the origin 0 and
pass through point Q (1, 1), and the higher the greater the absolute value of n, that is,
the value of n, the closer the corresponding the greater the positive number m 7z, the =—
curve (1.5.12) is to the x axis in the vicinity of faster these curves tend to merge with the x
the origin and the steeper is the curve in the axis (a complete merge never happens, how-
vicinity of point Q. In the interval 0 n 1, < < ever) and the slower they approach the y axis
on the other hand, curves (1.5.12) touch the y (see Figure 1.5.86, where we have depicted
axis, and the smaller the value of n the clos- , curves (1.5.12) with the following values of
er they are to the y axis; these curves also n: —1/10, —1/3, -1/2, -1, —2, —3, -10,
pass through point Q (1, 1), and the smaller that is, m =
1/10, 1/3, 1/2, 1, 2, 3, 10 in
the value of n the sharper is their turn away Eq. (1.5.12a) and the “limit line” y x° 1 = =
,

from the origin to this point (see Figure 1.5.8a, corresponding to n 0).
—
which shows the different curves, which cor- Note also that through each point (x, y) M
respond to different positive values of n in of the first quadrant (i.e. points with x and y
(1.5.12), and the straight line y x, which = positive) for which x =^= 1 there passes only
corresponds to n 1 =
and separates the one curve (1.5.12); ii x — 1 and y — 1 have
curves (1.5.12) with n 1 from < those with the same sign, that xis,1 and y
if >
1 or >
n > 1). x < 1 and y 1 < the value of n for this
—
,

The curves(1.5.12) with n negative behave curve is positive, while if x 1 and y 1 —

quite differently. These curves, specified by have different signs, that is, if one of the
the equation numbers x, y is greater than unity and the
other is smaller than unity, then n 0 for <
= *-”»= A-, m>;0, (1.5.12a)
this curve. As for the points (1, y) with N
y
y> 0, not a single curve of (1.5.12)
passes through such a point that differs N
also pass through point Q (1, 1), but, in con- from Q, while all curves specified by (1.5.12)
trast to the case where n is positive, they do pass through point Q.
not enter the square CAQB with the diagonal As for the behavior of the curves specified
OQ and lie completely outside it. The x and by (1.5.12) at great absolute values of n see ,

y axes are the asymptotes to these curves, and Exercise 1.5.3. (Mathematicians love to speak
 x

1.6 The Inverse of a Function. Graphs of Inverse Functions 47

of asymptotic behavior, or simply asymptot-

ics, of the functions (1.5.12) when n oo
or 72 — OO .)
9 * 1

Exercises

1.5.1. Construct the curves given by the

equations x4 1 (b) y
(a) y x4 1 = + ,
= — + ,

and (c) y x2 = x4 + .

1.5.2. Construct the curves given by the

equations (a) y
x~ 3
1 and (b) y
= — + =
— x~ 2
-f- 4.
What shape have the curves
1.5.3.
(a)y x 2n (b) y= z 2n+1 (c) y
,
x~ 2n and = ,
= ,

(d) y (a: =
2n+1 )~ 1 for very large values of the
and the function by x contrary to tra-
positive integer n? dition):

y--=x + a, x~y — a\
1.6 The Inverse of a Function.
Graphs of Inverse Functions
y~ 3a: + 2, x =jy-
1

y = i — x, x — 1 — y] (1.6.1>
By fixing a quantity y as a function of
another quantity x we mean that to y = 2
, *= ± V~y
each x there is assigned a definite val-
y = x + i, X= fy
3
ue of y. But this dependence can be
inverted, namely, we can fix y and then It is easy to grasp how the graphs of
find x. For example, the law of motion
the functions / and g are connected.
of a train traveling from one station
Suppose that we have the graph of tho
to another can be fixed by specifying
the position z of the train at every mo-
function y =
/ (x) (Figure 1.6.1). For
this graph to represent the function
ment t where z can be, say, the distance =
,
x g (t), we must view it at a dif-
traveled from the initial station. This ferent “angle”, precisely, the y axis
fact, of course, can be stated mathe-
must be considered the axis of ab-
matically as z / (£), and the engine- = scissas (the independent variable) and
driver uses the function in his time- the x axis the axis of ordinates. If wo
table. But a passanger is more inter-
wish the axis of the independent vari-
ested in the “inverse” timetable, the able to remain horizontal and the
dependence of the time t x at which the axis representing the values of the func-
train will be at a specified station de-
tion vertical, we must rotate the graph
termined by the coordinate z x The de- .
together with the coordinate axes
pendence t =
g (z) is called the inverse through 180° about bisector of the first
of z =
f ( t ) or, in other words, g is a and third quadrant angles (the straight
function that is inverse to f. Note that, line l). After such a rotation the y axis be-
of course, / is the inverse of g.
comes horizontal and the x axis vertical.
Here are some examples, where the The result of such a rotation is depicted
left column "lists the function y = in Figure 1.6.1 by a dashed curve,
y(x) and the right column the func- and the continuations of the x and y
tion x = x (y) (in the last instance the
axes are also depicted by dashed seg-
independent argument is denoted by y ments to stress the fact that the axes
have changed places.
x 9
*
The word “asymptotics” is of the same In Figure 1.6.2 the two parts of Fig-
origin as “asymptote” since the statement that ure 1.6.1 (the solid curve and the
the function y i/x has two asymptotes =
y
— 0 and x =
0 characterizes the behavior
dashed) have been separated; in Fig-
of this function for small and large absolute ure 1.6.26 the abscissas are denoted by
values of x. x and the ordinates by y (as is custo-
 x )

48 1 Functions and Graphs

Figure 1.6.4
mary). We that while in Fig-
see
ure 1.6.2a the value a =
OA of the inde- y — x 2
is two-valued: to each posi-
pendent variable corresponds to 6 = ,

tive value of the independent variable x

OB of the function (where b f (a)), = there correspond two values of the
in Figure 1.6.26, depicting the graph
function, y =
]/ x and y = —Y
x. On
of g, we have a —
g (6). A rotation the other hand, there are no values of y
through 180° about the bisector l maps corresponding to negative values of x
the segments OA a and OB = b of = since the graph of the function y =
,

Figure 1.6.2a into the segments OA x —

a and OB 1 6 =
of Figure 1.6.26. ±Y x ^ es entirely in the right half-
plane corresponding to x positive. There-
Thus the graphs of two functions each
fore, the statement that to each va-
of which is the inverse of the other are
lue of x there corresponds a value of
symmetric about the bisector l of the
the inverse function y will be incor-
first and
third quadrant angles. For exam-
rect in the given case, since there are
ple, Figure 1.6.3 depicts graphs for a
pair of such functions: y 3x 2 = + values of x (x <
0) for which no val-

and y =
x/3 —
2/3. In particular, if
ues of y exist, while for other values of
the graph of a function is symmetric
x (x > 0) we have even an excess of
values of y (two values of y for each
about the bisector l the function coin-
cides with its inverse. Such are, say,
value of x 0). >
the functions y 1 =
x and y —1 lx = The situation with the curve of a func-
(see Figure 1.4.2). Indeed, y 1 x = — tion / ( x depicted in Figure 1.6.5 is
even more complicated (the dashed curve
implies that x 1 =
y, while y — =
corresponds to the graph of the in-
Mx implies that x — My.
verse function g). Here to a value of,
Let us now turn to Figure 1.6.4,
which depicts the graphs of the func- say, x —
a there correspond four val-
tions y = x 2 (the solid curve) and y =
y (the dashed curve). The reader
can see that the function y = ±Y x,
which is the inverse of the function

-2
9»-L x
3 * 3

cc

Figure 1.6.3 Figure 1.6.5

 — x

1.6 The Inverse of a Function Graphs

. of Inverse Functions 49

ues of g x ), namely, 6, 6 1? b 2
(.
and ,
corresponds only one value of x For .

b3 This, however, is not surprising

. example, there is no difficulty in deter-
there is no rule by which a function mining the inverse of y 3x 2, since = +
that is the inverse of a given function this (linear) function increases mono-
/ (x) must be single-valued. If we take tonically. The same is true for the
the example of the train discussed above, dependence of the distance traveled by
the points (stations) through which the a train on the time, 2 / (£), if the =
train passes (the position of each point train moves without stops and in one
is defined by the distance z from the direction. But if the direction in which
initial point 0) are determined unique- the function changes is reversed (say,
ly by the value of t (time), and it the train starts moving in the opposite
is this dependence z =
z (t) that is direction), for example, the function
given in the timetable for the engine- first decreases, as y x 2 , and then, =
driver. But if a given suburbian train after passing through the minimum
does several runs a day, then it passes point, begins to increase, each value
a certain point on its trip back and of the function is passed a second time,
forth several times in the course of the and therefore we cannot guarantee that
day, which implies that the inverse it is a certain value of x that corre-
dependence of time on distance, t = sponds to a given value of y, since there
t (2 ), is multiple-valued —
to one value can be several values of x correspond-
of z (the distance of a point from ing to a single value of y (two values
the initial point) there correspond sev- in the case of the function y x2 ). =
eral moments in time when the train Finally, if we take the function y a =
is exactly at a point distant 2 from 0. (see Figure 1.3.6), there is no way in
So where does the difference lie betwe- which we can define the inverse, since
''JBifcF'i&n.tviUii Lr 0^ ^secPr Tg- T,
i t -ni, n
ndu ^nence,
/ ,

ure 1.6.3; here the inverse function is cannot be reconstructed from y\ in

single-valued) and a function like y = other words, the function y a has —
x2 (see Figure 1.6.4; the inverse func- no inverse.
tion is two-valued) and, all the more, In the case of other functions the si-
a function like the one depicted in tuation is not as bad as one would think.
Figure 1.6.5 (the inverse function is If we wish to construct a single-valued
multiple-valued)? The answer is ob- function that is the inverse of, say,
vious. The inversion, so to say, of a y =x 2 we need only confine ourselves
,

function, that is, finding the inverse to one of the monotonic parts of this
of a given function, consists in recon- function, say consider only positive
structing the values of abscissas from values of the independent variable x
the respective values of the ordinates, (in Figure 1.6.4 this corresponds to the
say, reconstructing the time t from part of the curve y x2 lying in the =
the distance 2 traveled by a train. We first quadrant). This mono tonic sec-
are lucky if to each value of the func- tion of the function has a single-valued
tion y there corresponds exactly one inverse (the graph of the single-valued
value of the independent variable x
that is, if the inverse function is single-
, inverse function y x correspond-
ing to this section is depicted by the
= Y
valued, but such good “behavior” rare- dashed curve lying in the first quad-
ly happens. However, an inverse func- rant of Figure 1.6.4; it is this func-
tion is always single-valued if the ini- tion that is known as the arithmetic
tial (single-valued) function y f (x)— value of the square root of x and is de-
is monotonic that is, it either always 1 * 10
,
noted by ]/’a:). Similarly, if only
increases or always decreases. Here, as
the independent variable x changes, 1 * 10
One must bear in mind, however,
the function runs through new values
of y and to each such value of y there
that the quantity +
Vx, where by }/" we
mean the arithmetic value of the root is not
 50 1 Functions and Graphs

a point M (x 0 , y 0 ) on the graph is not a 1.7 Transforming Graphs of Functions

maximum or a minimum of a func-
tion y =
f (x), like point M in Fig- Above we repeatedly encountered
problem transforming graphs, that
of
the
ure 1.6.5, then in the neighbourhood of
point M
a monotonic interval can be
1 11
is, the transformation of one graph
into another graph that is similar to
specified. To this interval there cor-
*

responds a single-valued branch of the the initial graph. We will now return
inverse function g (see the parts of the to this problem having in mind a more
solid curve and dashed curve from point systematic approach.
0 to point B in Figure 1.6.5). In gen- The simplest case is the translation
eral, a multiple-valued function, like of a curve. Suppose that we have the
the function y g (x) depicted in Fig-= graph of a function y = f (x) and we
ure 1.6.5 by a dashed curve, usually would like to know the shape of the
splits into separate single-valued curve obtained through shifting this
branches (in our case these are the graph b units upward. The answer is
branches depicted in Figure 1.6.5 by arcs almost obvious: it is clear that if we
OB,BC CD and 29 i?). The difficulty here consider two curves, y — f {x) and y =
, ,

may arise only when we wish to choose f (x) +

6, then to each point A(x Q y 0 ) ,

the “principal” branch, but from the of the first graph there corresponds
viewpoint of mathematics this ques- a point A x (x 0 y ± ) of the second graph,
,

tion is irrelevant. which lies b units above point A (see

Figure 1.7.1, with b positive). Putting
it differently, we can say that if we

Exercises have two functions, y — f(x) and

1.6.1. What are the inverses of (a) y = y — b =
f (x), then the graph of the
2x + 4, (b) y — x2 —
2, (c) y xz =
3x 2 + + second is b units higher than the graph
3x, and xA
(d) y =
2z 2 ? + of the first function, that is, replacing y
1.6.2. Employ Figures 1.5.2, 1.5.3, 1.5.6a, with y —
b in the equation of a curve
and 1.5.66 to construct the graphs of the is equivalent to shifting the curve b
functions that are the inverse of (a) y —
xB +x (b) y xs =
x, (c) y x~ 2 —
i/x 2 — = units upward.
and (d) y
,

x~ s =1/a:
3 =
Separate, where pos-.
,
It could seem that there is no sense
sible, the intervals on which the functions are in formulating the same statement in
monotonic, which permits arriving at single- two ways that differ very
different
valued inverse functions.
1.6.3. Find the functions that are the
little from each other, namely, lift-
inverses of (a) an arbitrary linear function ing a curve means performing the fol-
y = ax +
6, (b) an arbitrary quadratic func- lowing transformation of the equation
tion y ax 2~ bx c,+ +
and (c) the func- of the curve: go over from y = f (x) ta
tion y {ax = b)/(cx +d ). + y = f (x) b +or from y = / {x) to
1.6.4. Suppose that / (x) and g {x) are two
functions each of which is the inverse of the y— b = f (x). However, the second for-
other, withdefined on the interval a
/ c mulation, which involves replacing y
x c=
6 and g on the interval where with y —
b is more convenient than
/ (a) a and / (6) =
p and the function y = the first when the curve is specified
f (x) increases monotonically on the interval
a < c=
x 6. Prove that (a) / (g (x)) x, = implicitly, that is, by an equation of
g (/ (a:)) x, and (b) the functions F {x) —
f (/ (x)) and G (x) =
g (g (x)) constitute ano-
ther pair of such functions. What are the do-
mains of these two functions?

the complete inverse of y x 2 only the two- = ,

valued function y x is the exact = +V

inverse of y x2 = .

1 11
In this connection one speaks of the
-

function y =
/ (x) as being locally monotonic
at point M, that is, locally monotonic in
a certain neighbourhood of point M.
 1.7 Transforming Graphs of Functions 51

the type F x y) 0, which is not (, ,

=
resolved for y. For instance, the equa-
tion of a circle S of radius 1 with cen-
ter at the origin is conveniently writ-
ten as

x2 + y
2 = l,

or
Figure 1.7.3
F (x, y) = *2 + y
2 -1= 0 (1.7.1)

(compare with formula (1.2.1) for the

equation of the curve with y b means +
that we shift the curve downward b
distance OA with A = A(x, y)). Re- , units (the dashed curve in Figure 1.7.1).
placing y with y b, we get — Similarly, the curve specified by the
- 1 = 0, equation F (x a, y) +
0 is obtained =
x 2
+ (y-b) 2
from the curve F(x y) 0 by shift- , =
which the equation of a unit circle cen-
is ing the latter a units to the left .
tered at the point Q 1 (0, 6), an equa- For instance, the centers of the cir-
tion obtained from S by shifting S up- cles x2 + (y + b) — 1 = 2
0 and (x +
ward by b units (Figure 1.7.2). a)
2
+ y — = 0 are
2
1 the points
Similarly (as was discussed in great Q3 (0, — b) and (—-a, <? 4 0) (see Fi-
detail in Section 1.4), replacing in the gure 1.7.2). These translations may all
equation F(x y) 0 the independent ,
= be combined; for instance, the curve
variable x with x a is equivalent to — F(x +
1, y 2) — =
0 is obtained from
shifting the curve a units to the the curve F(x y) 0 by shifting the=
right. For instance, the equation (x — ,

latter one unit to the left and two

a)
2
+ y
2 — 1=0 defines a circle of units upward (Figure 1.7.3).
radius 1 centered at point Q 2 ( a 0), , Example. Let us consider the linear-
that is, an equation obtained from S fractional function
by shifting S rightward a units (see
Figure 1.7.2). c *°- t1 - 7 2)
-

It is clear that the quantities a and

b may be negative; for instance, trans-
Many experimentally established re-
lationships obey, either exactly or ap-
forming the equation y / (x) into the =
equation y f (x) =
b (where, as usu- — proximately, this function, whereby it
is important to know how to simplify
al, b is positive) or replacing y in the
the function and construct its graph.
It is clear that if the numerator and
denominator of the fraction on the
right-hand side of (1.7.2) are such that
a -7- c b — —
d, or ad be, the func- =
tion specified by (1.7.2) is a constant, or
y =k, where k ale =
bid whereby = ;

the only interesting case is where ad

be. We
state that in this case the
graph of the function (1.7.2) is a hyper-
bola and, hence, the function (1.7.2)
represents inverse proportionality of
two quantities.
We leave it to the reader to examine
the general case (see Exercise 1.7.6) and
consider the particular case
19 — Gx (1.7.2a)
y 2x —5 *

4 *
 5

52 1 Functions and Graphs

the and third quadrant angles is

first

y0 = The equation y x = 10# corre-

x.
sponds to a straight line that is more
steeply slanted: for a given x the ordi-
nate is 10 times greater (see Fig-
ure 1.3.4).
The law by which y 0 — f (x) is trans-
formed into y x = cf (x) may also be
described thus: in the equation of the
curve z/ 0 = / (x ) replace y 0 by yjc,
that is, write yjc = / (x). Then the
dependence of y x on x (the new y) is char-
Figure 1.7.4 acterized by the fact that the curve
z/i (x) elongated c-fold vertically as
is

In the right-hand side we isolate the compared to the curve y 0 (#) (the old z/).
integral part: Again the need for two formulations
whose equivalence is quite obvious (in-
_ (15 — 6*) + 4 15 — 6x 4
y ~~
2x—5 “ 2x —5 + 2x — deed, it is clear that the equations y =
cf (a?) and ylc =
are the same
/ (x)
equations) follows from the fact that
it is more convenient to replace y with

The last equation can be rewritten ylc if we are forced to “stretch” c-fold
thus: away from the x axis a curve whose equa-
tion F (x, y) =
0 is not resolved for y.
!' + =^7T’
3 d' 7 - 2b> J ust substitute ylc for y in the new equa-
tion, and we have the sought result:
from which we see that the curve (1.7.2a), F(x ,
ylc) = 0.

or (1.7.2b), is the graph of inverse pro- Let us again turn to the unit circle S
portionality (with coefficient 2) be- whose equation is x2 z/
2
1 0, + — =
tween y +
2 and x —
5/2 (Figure 1.7.4). that is, a unit circle centered at the
This completes the solution of the origin. How would you write the equa-
problem. tion of the curve obtained by stretching
Now let us see how the equation of the circle along the vertical axis 3-fold
a curve must be changed so that all (Figure 1.7.5; the new curve is denoted
vertical dimensions (along the y axis) by z/i (#))? By the rule that we have
are increased c-fold. 112 Obviously, in just stated, in the equation of the cir-
place of the equation y —
f {x) we must
take a new equation y =
cf (x) (we will
write y 0 =
f (x) and y x =
cf {x), since
these are two different curves). Then for
the same x the quantity y t will be c
times greater than before, that is to
say, c times z/ 0 and the curve will be
,

stretched in the vertical direction en-

fold.
As an example, recall the equations
of straight lines passing through the
origin. The equation of the bisector of

1,12
For the sake of simplicity, we from
now on assume that c is greater than unity;
the case with c less than unity (and even
negative) we discuss below. Figure 1.7.5
 r )

1.7 Transforming Graphs of Functions 53

cle we must replace y with y! 3. This yields

*2 2-1==0 ll 7,3)
+(i )
-

(here we write y 1 instead of y to distin-

guish between curve (1.7.3) and circle
(1.7.1)).
The curve obtained as a result of
Figure 1.7.6
stretching the unit circle away from
the horizontal (or vertical) diame-
ter Exercise 1.7.5) is an example
(see (see Figure 1.3.4). The first straight

of an ellipse
1 13
We have therefore
.
line slants at an angle of 45° to the x
transformed circle (1.7.1) (in Fig- axis, the second straight line is less

ure 1.7,5 this cicrle is denoted by y 0 (#)) steep.

into an ellipse, (1.7.3). Another illustration: x\ + y 2 —1=
In the given case Eqs. (1.7.1) and 0 and {xJ2) 2 + y 2 1=0. The — first

(1.7.3) can easily be solved: y 0 = equation corresponds to a circle S of

Yi — x and y x = 3 1
2
x which Y — 2
,
radius 1 centered at the origin, while
the second to the curve obtained from
clearlyshows that y ± = 3 y 0 for equal x .

that circle via 2-fold stretching along

But the rule, or statement, by which
the x axis (Figure 1.7.6). It is clear that
replacing y with ylc leads to a c-fold
the new curve is an ellipse. The proof
stretching of the curve along the ver-
of this is almost obvious. If we solve
tical axis holds true also for curves de-
the equations of the first and second
fined by
complicated aequation
curves,
F(x, y) = 0, that
an equation that can- is,

not be resolved algebraically for y say ,

y = / (*o) and y — f fo/c), (1.7.5)

* + y log y = 0. (1.7.4). for x ,

we obtain
The curve that results from stretching = <P (y
and xjc = cp (y),
curve (1.7.4) 3-fold along the vertical =
axis is described by the following equa-
i.e. x1 = ccp (y) cx 0l (1.7.5a)

tion where cp is the inverse of function /

(see Section 1.6). This corresponds to
x + J y log (|-) = 0. (1.7.4a) the initial statement that substitu-
tion oixlciov x stretches the curve along
The statement concerning the re- the x axis c-fold. The important thing
placement of y with ylc is readily ex- here is that / is one and the same func-
tended to the ^-coordinate. When we tion in formulas involving x 0 and x lf
replace x 0 with xlc in the equation of (1.7.5). Therefore cp is also the same
the curve, F ( x 0 y) 0, that is, when
,
— in the formulas involving x 0 and x u
we go over to the equation F(xjc,y) 0, = (1.7.5a).
the initial curve stretches along the Here is another example:
x axis c-fold, which is to say, for
equal y the value of x1 is c times the y = 10*° and y = 10^/2 . (1.7.6)
value of Xq.
Webegin with examples instead of The inverse of a power function is a
a proof: y x 0 and y =
x x H0 0Axx = = logarithmic function, whereby

may
13 The
3 *
stretching coefficient, or ratio,
be equal to 1 (of course, such an identity
= iog y and V 2 = log

transformation does not change the shape of i.e. x1 = 2 log y. (1.7.6a)

a curve); accordingly with this a circle is
usually considered as a particular case of an Thus, the graph of function y 10 x/2 =
ellipse. is obtained from the graph of function
 —x — x )

54 1 Functions and Graphs

y =10* by a 2-fold stretching of the

latter along the x axis.
What do we do if in the equation y =
/ {x) the quantity kx is substituted
for x? To take advantage of the above-
stated rule, let us recall that multipli-
cation by k is the same as division by
ilk, since kx =
xl(ilk). Hence, ilk plays
the role of c in the earlier formulas,
and if, say k =
1/2, then c ilk 2.= =
This means that the substitution of
0.5#! for x 0 is the same as replacing x 0
with xJ2 and leads to a stretching of
the curve along the x axis by a factor
of 2. If k 3, =
then c ilk =
1/3, =
and the replacement of x with 3x is the
substitution of xl( 1/3) for x .

What do all these substitutions mean two substitutions:

geometrically? In the case where c 1 >
the result can be stated thus: when,
in the equation of the curve, y is re-
placed with ylc the curve is stretched
,

vertically by a factor of c, while in

= — , SO that i/2 -= — !/i) .
( -f-
the substitution of xlc for x the curve
is stretched horizontally by a factor The first operation, substitution of
of c . yjb for y 0 where b 0, has already
, >
When 0 < <
nothing really
c 1, been discussed to a change— it leads
changes in our reasoning, with the excep- in the dimensions by a fac-
vertical
tion that the word “stretching” cannot tor of b. What remains to be clarified is
be used, since stretching 2-fold means the meaning of the change in sign of y,
increasing the dimensions by a factor that is, the meaning of the substitution
of 2, while stretching “1/3-fold” means of —y for y. Obviously, the points of
multiplying the dimensions by 1/3, the curve y 0 / {%) and y x =
/ (x — —
which means not multiplying but di- corresponding to each other lie sym-
viding by 3, or “shrinking” the curve metrically about the x axis (Fig-
(in a certain direction) 3-fold. Hence, ure 1.7.7). Therefore, the curve y x =
when we substitute ylc for y, the verti- — / (#) is obtained from the curve y 0 —
cal dimensions change by a factor of c , / (x) as the mirror reflection of the lat-
which results in the curve “shrinking” ter with respect to the x axis, and the
in the vertical direction; for instance, same can be said of the curves
at c =
0.5, a transition from curve F(x y 0 ) = 0 and F (
, ,
—y = x) 0, where
y = curve y
f (x) to cf (x) =
0.5/ (x) = the second equation is obtained from
means an actual reduction in height the first by reversing the sign of y.
by one half. The same goes for the sub- Reasoning in a similar manner, we
stitution x x!c\ for 0 c < <
1 this can say that since points (x, y) and
substitution amounts to a “shrinking” ( x y) are symmetric with respect to
,

of the curve along the x axis. the y axis, the replacement of x with
Now let us establish the meaning of — in the equation of a curve (the
the substitutions y -*~ylc and x -+xlc transition from curve F (x, y 0 ) 0 to =
when c is negative Such a replacement
. curve F ( x y 2 ) 0) replaces the ini-
,
=
can be conducted in two stages. We tial curve with a curve symmetric to
introduce c b = — = —
( 1) b, where b the initial one about the y axis. Fi-
is positive, and carry out subsequently nally, the transition from equation
 y x

1.7 Transforming Graphs of Functions 55

substitution of —
x for x or the substi-
tution of —
y for y or under the simul-
taneuos substitutions of x for x and —
— for y. A
firm grasp of these rules
will make it possible, after you have
built a curve y / (x) or F ( = y) 0, ,
=
to picture the graphs of all functions
of the type

^ -/(*-=*) or *•(£=?,

with arbitrary values of the constants a,

Figure 1.7.8 b c, and d
,
.

Here are two more simple (but im-

F (x , y 0) —
0 to equation F (—x, z/
3) — portant) examples. In Figure 1.7.9 we
= 0 reduces to successive alternations have two curves:
of the sign of y (symmetry with respect = =
y sin x and y sin 3#, (1.7.7)
to the x axis) and of the sign of x (sym-
metry with respect to the y axis). But where, of course, x is measured in ra-
two successive reflections, from the x dians. The second curve has been com-
axis and the y axis, are equivalent to pressed horizontally by a factor of
a reflection with respect to the origin three.
0(0, 0) (see Figure 1.7.7). Thus, si-
— The relation y = sin x is a periodic
multaneous substitution of x for x function: at x — 2n 6.3
radians
and of —y for y in the equation of a (which corresponds to an angle of 360°)
curve is equivalent to a symmetry trans- the sine has the same value as for
formation with respect to 0. x =
0. (Adding 2n to any angle leaves
Notwithstanding the simplicity of the the value of the sine unchanged.) Thus,
above reasoning, beginners (for which the graph of the first function, y =
this book is intended) often make mis- sin x transforms into itself under a
,

takes concerning the various transforma- translation along the x axis (in either
tions we have just discussed. To clarify direction) by 2 jx, which simply means
matters we give an example: that sin a: is a periodic function with
F (x, y)
a period of 23X. The function y sin 3x —
= (* - 3) + 2
(y - 5)
2 -4= 0.
is also periodic, but the period here is
less by a factor of 3. If x varies by
This is the equation of a circle of ra- 2n/3 ^2.1 rad, 3x (the angle whose
dius 2 centered at a point with coordi- sine is laid off on the axis of ordinates)
nates x 3 and y — 5. Figure 1.7.8 = varies by 23X, and sin 3x returns to the
shows the initial curve and also the same value:
following curves:
sin 3x= sin 3 •

F (.
r , -if)
= (x - 3) + 2
(-y - 5) - 4 = 2
0 Use this example to think over the
general assertion that the substitution
F (—x, y) of kx for a; in a periodic function re-
= (~x - 3) + 2
(y- 5)
2 -4= 0,

F (—x, —y)
= (-* - 3) 2 + (-y - 5)
2 -4=0
The letter F in all four cases denotes
one and the same function. See what
happens to the curve (circle) under the
 56 1 Functions and Graphs

suits in the decrease in the period by

a factor of k and an increase in the fre-
quency by the same factor (the frequen-
cy is the number of periods per unit
length).
The second example deals with loga-
rithmic (and exponential) functions.
Take two curves:
Vi = lo g a x and y2 = log hx ,
(1.7.8)

or, which is the same.

x = a v1 and x = b y *, (1.7.8a)

where it is assumed that both a and b

are greater than unity. It can be estab-
lished, by a method similar to that
employed in the analysis of (1.7.6) and
(1.7.6a) (see Section 1.4.9) that the metric with respect to origin 0 (Fig-
second curve in (1.7.8) is obtained from ure 1.7.105), the function is said to be
the first by stretching the latter along odd. These names originate in the fact
the y axis by a factor c log & a = = that parabolas of even order, y =
1
(log a &)" , where c is the modulus of ax2m are even functions, while para-
,

base a with respect to base b. Similarly, bolas of odd order, y =

ax2m+1 are odd ,

any two exponential functions, y ax = functions (cf. Section 1.5).

and y x
=
b , are related in the same Finally, we note another important
manner: the second is obtained from fact. It is necessary to understand (this
the first by stretching the first curve is often a neglected aspect) that in ap-

along the x axis by a factor c = log 5 a. plied problems the procedure in which
Two simple but highly important all ordinates y or abscissas x are changed

concepts are related to what we have in a given ratio (and nothing more)
just said. A curve given by the equa- is related to a change in the units of

tion F (x, y) =
0 that retains its form measurement of y or x so that the graphs
,

under the substitution of x for x — obtained as a result of such a change of

(say, the parabola y ax 2 ) is sym- = scale represent one and the same de-
metric about the y axis. A curve whose pendence of y on x to within, so to say,
equation retains its form under the a change in scale (or the system of
substitution of —x for x and y for y — units).
1 * 14

(say, the hyperbola y Hx or the cu- = Suppose that on the axis of abscis-
bical parabola y =
x 3 ) is symmetric sas, Ox, we plot time and on the axis

with respect to the origin. Finally, a of ordinates, Oy, we plot distance (Fig-
curve whose equation F (x, y) 0 maps — ure 1.7.11); the function y = / (x)
into itself under the substitution of y — (or, as we will often write, z = f (t))
for y (say, the semicubical parabola fixes the law of motion. But since y
y
2 =x 3 see Figure 1.5.7) is symmetric
;
and x are measured in different units,
with respect to the x axis. All these no comparison of the two quantities
statements follow directly from the is possible (indeed, there is no way
geometrical meaning of the respective that we can compare 1 cm with 1 s or
substitutions. say which of the two quantities is great-
A function y = f {x) whose graph is
1 14*
However, if x and y are measured in
symmetric with respect to the y axis
the same units, the substitution of cy for y
(Figure 1.7.10a) is said to be even, has a physical meaning, since the functions
while if the graph of a function is sym- y= / (x) and y —
cf (x) are different.
 x , , ,

1.7 Transforming Graphs of Functions 57

a new system of units along the y x axis*

(changing the scale on this axis); it is-
in this sense that, as mentioned in
footnote 1.5, the parabola y x 2 can —
be considered as the general case of
a parabola.
In the same way, any linear-fraction-
al function y ( kx -f- 1)1 (mx n) can — +
be reduced, by changing the position of
the origin on the x and y axes and the
Figure 1.7.11
scale on the y axis, to the “unit” hyper-
bola” Y=
l/X (see Exercise 1.7.6).
er). The
statement, say, that for a Similarly, any cubic function (poly-
given point ( M
y ) the segment
,
OM nomial) y ax s bx 2 =
cx d can + + +
lies at an angle of 45° to the axis of be reduced to one of the three func-
abscissas (see Figure 1.7.11) is mean- tions: y x3 = y x +
x, or y z, = 3 — =
ingless, that is, if we change the scale a? (no further reduction is possible).
of the y axis (go over from centimeters The list of examples can be continued.
to meters), point M
is replaced with
a new point M' corresponding to the Lei us discuss a more complicated example-
involving cubic functions (polynomials):
same moment in time, but now Z_POM'
is not 45°, that is /_POM' 45°, while y =+ bx + cx d
ax 3 2 -j-

in all other respects M' is equivalent =

a (x + px + qx +
3 2
(1.7.9)- r),

to point M. Thus, in physical problems

where p = 6/a, q = c!a and r = d!a (of
the ellipses in Figures 1.7.5 and 1.7.6
,

course, a ^ 0). In Section 1.5 we saw that the-

do not differ from the circles depicted substitution X — x + p/3 transforms (1.7.9)
in the same figures, and if we are inter- into
ested in a particular curve (ellipse), y = a(X* + PX + <?), (1.7.9a)
then it is wise to select units of or Y = aX* + BX = a (X 3 + kX),
measurement such that the ellipse
where Y = y — aQ, B = aP and k —
is simply a circle.
B/a (= P), with P 2
= — =
q p / 3 and Q
Similarly, any parabola y ax = 2
+ (2/27) p
3
r + —
(1/3) pq (see formulas (1 .5.6)
bx + c can be reduced, as we have and (1.5.6a)). At k 0 (i.e. if B 0) the = =
seen earlier, to a parabola y x ax\ = function (1.7.9a) has the form Y aX 3 or = ,

by a translation, or shift (i.e. by a tran- yi = *i> (1.7.10a)

sition from coordinates x and y to a set where x1 =
X (the X-coordinate does not
of new coordinates, x 1 = x p and + change) and y x Y/a. If k >> 0 (i.e. B and =
z/i = y +
q, where p and q must be a are of the same sign), we put x 1 X/Y k, =
chosen appropriately). But physically i.e. X = Ykx ±. Then (1.7.9a) becomes
such a translation means nothing else
but a change in position of the origin
Y =a ( k 3 f 2 x\ + fc
3/2
#!) = ak 3 2
/ (x 3 +x x)

from which the physical quantities x that is, we have the law
and y are reckoned, say the initial =
moment t = 0 and the initial position £/i *i +x i (1.7.10b)

z = 0, and so does not influence in any where y 1 —

Y/akY Finally, if k 0 (a and <
way physical process. After this
the B are of opposite sign), a substitution th at is -

we can transform y x =
ax\ into the similar to the on e used above, x x = X/Y I
k |,

“unit” parabola Y = ±X
2
(or even to that is, X= Y |
k |
x± ,
transforms (1.7.9a)
Y = X
2
if we select the positive di-
into

rection along the y axis 1,15 ) by choosing Y = a(\k3/ 2 \x 3 + k\k\V 2 x 1 )t

1 15-
While in the case of time, t. we have past are not equivalent, for distance, z, this-
a natural direction of growth, which corre- is not the case, and we are free to choose*
sponds to the simple fact that the future and the any direction for the increase in z.
 ' — h ^

.58
1 Functions and Graphs

that is,
into such a form that the bases acquire
Vi — A— (1.7.10c) a convenient form, since the transi-
tion from one base to another means
where y l = — Y/akY |
k |. Thus, for a physi-
actually changing only
cist any scientific function of the type (1.7.9) the units of
that links two quantities of different dimen- measurement of the quantities (cf. the
sions, y and x, is equivalent to one of the above material on logarithmic and ex-
three functions (1.7.10a)-(l. 7.10c) (see Fig-
ponential curves). Therefore, each time
Tires 1.5.1a, 1.5.2, and 1.5.3 and Exercises 1.7.8
-and 1.7.9). we encounter the logarithmic func-
tion y =
log c x we can freely assume ,

Let us use the following example to the base to be equal to 10 or to e ^

illustrate the aforesaid. It is well known 2.272 (see Section 1.4.9) or even to
that the time dependence of the alti- two; 1 16 when exponential functions are
-

tude h of a stone thrown upward from considered, y dx the base d is usu- = ,

.a tower of height w with an initial ally assumed equal to e .

velocity v is given by the formula

Exercises
h=w + vt — jrt 2
(1.7.11) i/2
,

1.7.1. Construct the curves —— |-

where g ~ 9.8 m/s is the acceleration

2

of gravity. If, however, for the zero i-o, <i±a + i£=a .„d
level we take not the ground level but
the greatest altitude reached by the
-— —— 1=0 (it is natural here

stone, h 0 and for time zero we take

,
to employ the
fact that x2 2
1 0 +y — =
is the equation of a unit circle centered
the moment f 0 when this altitude is at the origin).
reached, then in the new coordinates, 1.7.2. Construct the curve y sin x by =
= h—h
Kdescribing 0 and tx = t—t 0 Eq. (1.7.11) taking, for example, eight values of x in the
the motion
,

of the stone
interval from —
ji to +ji by 0.25ji each. Employ-
ing this graph, sketch the graphs of the func-
simplifies considerably: tions (a) y 2 sin x, (b)= y sin 0.5-z, =
(c) y—
3 sin 3x (d) y cos x, (e) y
,
=
cos x = +
(1.7.11a) sin x (= y 2 sincos 2x+y (xcos x ji /4 )), (f) — 2

(— 1/2 (1/2) +
), (g) y sin 2 x =
But the coefficient gl 2 in Eq. (= 1/2 —
(1/2) cos 2x). [Hint. All these curves
(1.7.11a) has nothing to do with the may be obtained via translation, stretching,
physics of the process and is related
and shrinking the sinusoid y sin x\ to solve =
Exercises 1.7. 2d, f, g, employ the identity
^solely to the choice of the system of cos x sin (x =
ji/2).] +
units. Equation (1.7.11a) can be simpli- 1.7.3. Plot the following curves: (a) y =
fied still further by dropping the minus ± Yz 2
-l, (b) y = ±Y * + 2
1, (c) y = 2 ±
sign on the right-hand side; for this we Y —{x l) 2 1, and (d) 4 y —
2
4y x* 0. + — =
need only agree that h is measured from [Hint. Construct curve (a) using twenty values
point h = h 0 not upward but down- of x in the interval from 5 to with each — +5
value differing from the adjacent one by 0.5
ward. Next, we can go over to a “na- (and each being a multiple of 0.5). Transform
tural” system of units in which the accel- the equations of the curves (a) and (b) to
eration of gravity is equal to 2 (time y
2 —
x2 1 + =
0 and x 2 y
2
1 0, re- — + =
spectively, and note that (b) is obtained from
can be measured in seconds, as usual,
while distance can be measured in units 1 * 16
For instance, in the theory of informa-
*of gl 2, that is, approximately, 4.9 m). tion, where the amount of information is ex-
Then, in the new coordinates, H= pressed by logarithmic functions of the various
(2/g) (h — /i
0 ), T = = — tx t t0, and
variables, all logarithms are assumed binary
which no importance, of course, and is
is of
,

the law (1.7.11) or (1.7.11a) assumes connected solely with the system of units
the “canonical” form T2 H= . employed (with the fact that information is
In a similar way we can always trans- measured in bits). The interested reader can
form logarithmic and exponential func- refer to the book of A.M. Yaglom and I.M. Ya-
glom, Probability and Information Reidel, ,
tions often encountered in scientific laws Dordrecht, 1983.
 ) ) ))

59
1.8 Parametric Representation of a Curve

(a) by interchanging x and y. To construct

curve transform the equation of this curve
(d),
into 4 (y 1/2)
2 .r
+2
1 —
0 and then, by — =
translating and shrinking curve (b), you will
arrive at the result.]
1.7.4. Write the equations of the curves
obtained from the curve y x3 x (see = —
Figure 1.5.3) by (a) shrinking along the axis
Oy with a ratio 1/3 (i.e. diminishing all
vertical dimensions 3-fold), (b) substituting
xf2 for x and (c) substituting —y/2 for y
,
.

Sketch the three curves.

1.7.5. Prove that stretching (or shrinking)
the unit circle x 2 y
2 1 + — =
0 to any straight
line with any ratio results in an ellipse.
[Hint. This problem may be formulated as
follows: stretching (or shrinking) from a circle
(x —
a) 2 (y +
b) 2 r2 — — =
0 away (or toward) Figure 1.8.1
the x axis, that is, substituting ky for y,
+T*aDsffirms_tbe„ciTTla.inta an., ellinsg.l , _
"U, J: xk~of~ahscksas„a nd_ sr. _th a_ _
1.7.6. Prove that the graph of an arbitrary axis of ordinates in one drawing, and
linear-fractional function y (ax b)f = +
t on the axis of abscissas and y on the
(cx +
d) with c =#= 0 and A ad be =#= 0 = —
may be obtained from the graph of inverse axis of ordinates in the other (Fig-
proportionality y k/x, where k = A/c 2 by ure 1.8.1). = — ,

shifting the latter to the left by d/c units (or by Let us investigate the trajectory of
die units to the right if die
| |
0) and upward < point M. To each value of t there cor-
by afe units.
1.7.7. Prove that every function may be respond a value of x ( t and a value of
represented by the sum of an odd and even y (t). What curve will point M=
function. [Hint. Employ the identity M (x (t) 9 y (t)) =
describe as t M (
t

varies? To answer this question, we can

/(*)=
4
- (/(a:)
+ /(- * ))
eliminate t from the two equations
x =
x (t) and y y (t). This yields =
+ \{1(x)-f{-x).\ an expression that will involve only y
1.7.8. Prove that no two curves in and x that is, an equation of the type
,

(1.7.10a)-(1.7.10b) can be
substitutions of the
reduced to each
y y (x) or F (x y y)
= 0. Then we =
other by applying
= — construct the curve in the usual fashion
form x' ax p and y' by +
q (where, + by specifying various x's and find-
of course, a 0 and b =#= 0), which are equi-
=#=
valent to changing the position of the origin ing the corresponding y's. For instance,
for x and y (the translation of the origin) in our example,
and the units of measurement for x and y.
1.7.9. Find the rule that enables us x2 + y
2 = cos 2 t + sin 2 t — 1,
to establish from the coefficients a, b, c and that is,
d whether (1.7.9) can be reduced to
(1.7.10a) or (1.7.10b) or (1.7.10c).
y=±]/l — x2 ,
or x 2 + y — 1 — 0,
2

1.8 Parametric Representation so that the curve (1.8.1) is a circle

of a Curve of unit radius in the xy - plane (Fig-

ure 1.8.2a).
Let us now study the motion of a ma- However, it often happens that even
terial point, M—M (x, y). Each of the comparatively simple expressions for
coordinates, x and y, changes in the x ( t and y ( t lead to such complica-
course of time t, and the motion of ted expressions when trying to eli-
point M
is specified by two functions minate t that it makes no sense to tackle
x =
x (t) and y y (£), say, = them. For instance, if
x = cos t, y = sin t. (1.8.1) x = x = (t) a x £4 + b-J* + cx t2

These relations can be depicted grap- + dit + y — V (t) —

hically as two curves by plotting t +b z+ 2t
“1“ d 2t
-4- £2 >
 t ) t

60 1 Functions and Graphs

1 17
called a parameter -In physical prob-
lems, a parameter often has a definite
physical meaning. For instance, in the
example with which we started this
section, t may be the time variable,
and in this case both x ( t and y ( )
are functions of time. Of course, one
can be interested only in the shape of
the trajectory described by point M(x y) ,

but it is also interesting to know the

Figure 1.8.2 velocity with which the point moves
at a certain moment of time and the
position of the point at various moments
in time. To find these two quanti-
then to eliminate t we would have to
ties, we must retain the values of the
solve a quadratic equation, and this
lead to expressions that in view of parameter and for each point M=
their unwieldiness can simply not be M(t) write the number t (it goes
written. Yet it is possible to construct without saying that only a finite num-
the curve in the xy-plane without ber of points can be plotted on a draw-
eliminating t ing, e.g. see Figure 1.8.26). In this
it suffices to specify
:

various values of t and find x and y way, a “parametrized” curve yields

for each of them. To illustrate this more information than the trajectory of
point, we
construct the following table point M alone does (compare Fig-
of values of x (£) and y (£) corresponding
ures 1.8.2a and 1.8.26).
to the simple example (1.8.1):
Let us take the following curves:
(1) x = cos y = sin
t, t;

(2) x = cos y = —sin

t, t;
0 jt/4 jt/2 331/4 jt

X = cos t 1 0.7 0 —0.7 —1 (3) x = cos 3t, y = sin 3£;

y= sin t 0 0.7 0.7 0
y = cos 3
1
(4) x = sin 3 1, 1.

In each case, if we eliminate t, we

arrive at the equation of a unit circle
5n/4 3n/2
t

x = cos t -0,7 0
7jt/4
0.7
2ji
1
in the xy- plane, or x 2 y
2
1=0. + —
y= sin t -0.7 -1 -0.7 0 So in what respect are all these four
cases different?
In the first three cases, at t 0 the =
not necessary to take t
It is clearly point lies on the axis of abscissas,
greater than 2n because the values while in the fourth case it lies on the
of x and y repeat. Using this table, axis of ordinates. In cases (1) and (3)
we can plot the points of the curve. the point (one can imagine a heavy
In so doing we employ only the values material point) is moving along a circle
of x and y. Those values of t for which counterclockwise, while in cases (2)
the x's and i/’s have been computed and (4) it is moving clockwise. In
are no longer needed for plotting the absolute value, the angular velocity of
points. “The Moor has done his duty; rotation (or the rate of rotation) or
the Moor can go”, as the German pro- 1 17
-
The word “parameter” was invented
verb has it. by mathematicians of Ancient Greece for just
This method of representing a curve this purpose and has no exact translation.
or,what is the same thing, of specifying Webster s Third New International Dictionary
gives the following definition: “An arbitrary
a functional relationship y y (x), by = constant characterizing by each of its values
two functions, x (t) and y ( ), is called some member of a system (as of expressions,
parametric representation, and t is curves, surfaces, functions).”
 t x

1.9* Some Additional Topics from Analytic Geometry 61

x = TA = TP — A P = rt —
t t
rsint,
y = OT = BP = Q B - Q P t t

= r — r cos t.

The final result is

x = r (t — sin t), y = r (1 — cos t)

(1.8.2)
(the solid curve in Figure 1.8.3 depicts
the cycloid).
even the linear speed v (since the | |
1.8.1.
radius of the circle is equal to unity) Exercises
is 1 rad/s (i.e. u 1 m/s if the
| |
=
Construct the curves given by the
radius of the circle is 1 m) in cases(l) =
= following equations: (a) x cos t, y sin 2 1 =
and (2), and 3 rad/s (or v 3 m/s) | |
and (b) x cos t, y =
sin 3*. [Hint. Since =
in cases (3) and (4). sin 3 1 varies rapidly, you must take close-
Here is another example. Let us lying values of t, say, t 0 0 1 0 2 ., = ,
.
,
.
,
. .

or, if you have no trigonometric tables with a

build the curve that a point on the
rim of a bicycle wheel describes when
column for radians, t 0, 5°, 10°, 15°, .] = . .

1.8.2. Construct the curve x cos (5f 1), = +

the cyclist is moving along a straight y = sin (5* 1). +
line, the point A on the circumference 1.8.3. Construct the curve x = cos t, y =
of a circle of radius r as the circle rolls
cos ( + ji/4).
=
along a straight line. The curve is called
1.8.4. Construct the curve x cos t, y =
cos t.

a cycloid. 11 * 1.8.5. Express the parametric dependence

Let us take a horizontal straight (1.8.2) explicitly, in the form of a function
line along which our “wheel” is moving * = 9 (*/)•
1.8.6. A circle s of radius r is rolling
as the x axis, and let us assume that at without slipping along a circle S of radius R.
the beginning of the motion, t 0, = Write the parametric equations of the curve
the center of the wheel, Q Q 0 lies — ,
described by a point A on the moving circle if
(a) s lies outside S and (b) s lies inside S or
on the y axis at the point (0, r ). We
will also assume that the wheel rolls
(for R < r) S lies inside s. The curve (a) is
called an epicycloid ,
and the curve (b) a
with a constant velocity. For instance, hypocycloid.)
let us assume that in unit time the Consider the following particular cases: in
wheel rotates through a unit angle. case (a) r = R, r = R/ 2, and r = 2R; in
Then, in the course of time t, the wheel
case (b) r = R/2, r = R/3, and r = R/4.
(At r =
R the epicycloid is called a cardiod
will travel a distance rt in the positive (from the Greek kardia meaning “heart”), at
direction of the x axis (note that the r —Rl 3 and r =
Rf 4 the hypocycloids are
wheel does not slip in the process of called the Steiner 9 curve and the astroid M
(from the Latin word astrum meaning “star”).
motion, so that, rotating through an What is the appearance of the hypocycloid at
angle t it travels a distance equal to
, r = R/ 2?)
the length of the arc BA t in Figure 1.8.3,
which amounts to rt). The center (or
1.9* Some Additional Topics
axis) of the wheel Q 0 will occupy the ,
from Analytic Geometry
position Q t ( rt r), and the point A 0
,

will occupy A t (We wish to find the

. Above (in Section 1.2) we considered
coordinates of point A%. a number of geometrical problems
From the right triangle QtA t P (see solved by the coordinate method, or by
Figure 1.8.3), in which ALA t Q t P t = methods of analytic geometry.
and QtA t =
r, we get A t P r sin t = Suppose that A 1 (xu y x ), A 2 (x 2 y 2 ),
and Q t P =
r cos t. Since BO PT = — and A3 y 3 ) are three points on
( 3,
,

rt and Q t B Q 0O = r, we get = the coordinate plane (Figure 1.9.1).

18 1 19 Jacob
1 •
From the Greek kykloeides meaning Steiner (1796-1863), an out-
circular. standing German geometer.
 (

62 1 Functions and Graphs

The comparatively complicated for-

mula (1.9.1) (and also the formulas
(1.9.2a) and (1.9.2b)) are rarely
employed. All these formulas hold true
for any values (arbitrary both in size
and sign) of the coordinates of the
points considered and the differences
of these coordinates provided that we
agree to reckon a from A 1 A 2 to A X A%
and that a can be both positive and
negative.
How does one find the angle Z-A 2 A X A 3 = Since tan a is meaningless when a =
a? In Section 1.2 we found that n/2 (the tangent of a tends to*

a-^n/2 and said to be-

tan a 12 :=— —— and tana 13 -= ——— ,
infinity as
come infinite, the denominator in the
is

expression for tan a must be zero at

where a 12 and a 13 are the angles formed
by the x axis and the segments A 1 A 2 a = n/2), formula (1.9.1) yields the
following condition necessary and suffi-
and A X A 3 (see formula (1.2.5)). But
cient for the straight lines A X A 2 and
Figure 1.9.1 clearly shows that
A r A 3 to be perpendicular to each other :
a = Z-A A A = a 13 — a l2
2 3
— x )(x — Xj)
.
1
{x 3 1 2
Since, according to a well-known formu-
la from trigonometry, + — yi)(y —
( 2/ 3 2 2/ 1 )
= 0 . (1.9.3)
tan — tan y
tan (P — y) 1 + tan P tan y
ft
The following conclusions can be made
on theof (1.9.1), (1.9.2a), and
basis
we have (1.9.2b). Note that the expressions in
tan q — tan q
tan a = tan (a 13 —-a 12 = l3 12 the denominators in the right members
)
1 + tan a tan a l3 12 of (1.9.2a) and (1.9.2b) are the dis-
y% yi y2 yi tances A 1A 2 and A XA 3 (cf. (1.2.4)):
x3 —x t x2 —x : i

1
y3 y i y2 y1 V (z — Zi) +2
2
(2/ 2 -2/i)
2
= AA 2>
|

x3 —x x2 —x
l/(x — *i) +
2 =
or, finally,
± x
3
2
(2/ 3 —y i)
A A l 3.

—^ — —^ x —x From the school course of geometry

tan a= (^2 1) (?/ 2 1) ( 3 i)
the reader should know that
(x 3 x 1 )(x 2 a:!)-!- (y
3 — y±) (y — 2 1/
1)
*

(1.9.1)
SaAiA*a 3 = X A A3 X
t
|'sina|,
Thus, knowing the coordinates of points
Ai, A 2 ,and A 3 we can find the angle , where S AAtAtAs is the area of triangle
a = /LA A A 2 1 3. A x A 2A 3 . This*. yields the following for-
Combining (1.9.1) with the well- mula for the^area:
known formulas sin a [tan 2 a/(l = +
tan 2 a)] -1 / 2 and cos a (1 tan 2 ^" 1 / 2 = + , A A2A3
j
= ~2 l(®2 *l) (2/3 2/l)
we arrive at the following formulas:
sin a=
— (x — Xi)(y — y
3 2 f )|. (1.9.4)

(
x2 ~x i) (y 3 - y - (x — Xl (y — yj)_
x) 3 ) 2
On the other hand, S AAiA A , —
Y (^2 — ^i) + 2
(2/2 — Y x3— *i) + (y —yi
^/i)
2 2 2 (V2)AjAt X ft, where h is the distance
( 3 )

(1.9.2a) 1 20
In formulas (1.9.4) and (1.9.5) you 7 can
cosa =
-

drop the vertical bars and assume that

(x — x (x — x
2 (y — y{) y — y
x) 3 x ) -f- 2 ( 3 x) S Aai a 2a * an(l Ai A* possess a sense of
Y x2 — ^i) + — yi)2 Y x3— x + (y — yd
2
(i/ 2 ( i)
2
3
2 direction, so to say, that is, their sign deter-
mines on which side of ^4 2 point A 3 lies.
(1.9.2b) We will not dwell on this any longer.
 ' * y P

1.9* Some Additional Topics from Analytic Geometry 63

in the second system, where, obviously,

rx = r and cp x = cp — a. Therefore (see
Eqs. (1.2.3)), x = r cos cp, y —
r sin cp
and xx = rx cos cp x — r cos (cp — a) r
yx = r sin (cp — a), that is,

xx — r cos (cp — a) = r (cos cp cos a

+ sin cp sin a)
= r cos cp cos a + r sin cp sin a
= x cos a + y sin a,
yx —r sin — a) = r sin
(cp cp cos a
— cos sin a) cp

from point A 3 to the straight line A XA 2 .

= — r cos sin a + r sin cp cp cos a
In view of (1.9.4), we then have the = — x sin a + y cos a.
following formula for the distance
h = h AsiAl A 2 from point A 3 to straight On the other hand, from the same-
line A A2 :
Figure 1.9.2 it follows that
1

hAs AiA 2
x’ = x1 + a, V
r
= Vi + 6*
,

— yi) — (^3~x )(y — y where a = O' and b = PO are the*

= 1
Xt) (y 3 1 2 1) I

coordinates of the old origin 0 in tlm

V — + (^ —
(*2 l)
2
2 */ l )
2
new system of coordinates x'O'y'.
(1.9.5) Thus, the final result is 1 21 *

Wenote also that the s stem of x = x cos a + y sin a + a,

coordinates in a plane for geometrical y' — — x sin a + y cos a + b. (1.9.6)
problems is rather arbitrary: to specify
a system of coordinates we must fix These formulas are fundamental to
the point 0 (the origin) and the two geometry. The method of cooridnates
axes Ox and Oy (perpendicular to each makes it possible to characterize every
other), which, however, in all other point in a plane by a pair of numbers,.
respect can be chosen at our will. x and y, the coordinates of the point;
if we have two points, then two pairs
For example, suppose that xOy and
x'O'y are two different coordinate sys- of coordinates are needed; and a set
tems (Figure 1.9.2). To establish the of points can be replaced with a set of
relationship between the “old” coordi- pairs of numbers. For instance, a curve 1

nates (x, y) and the “new” coordinates in a plane has corresponding to it

(x
'
y') of one and the same point M, a one-parameter set of pairs of numbers,
,

we introduce an “intermediate” coordi- x (t) and y ( t ), that depend on parame-

nate system x 1 Oy 1 whose origin coin- ter t (cf. Section 1.8). The transition
cides with the origin of the old system 1,21
As is customary in analytic geometry
while the directions of the xx and y x we restrict our discussion only to right-handed
axes coincide with the directions of the coordinate systems, that is, systems in which
x and y' axes, respectively. To relate a rotation of the positive semiaxis Ox through
the coordinates (x, y) and (x ly yx ) of
an angle of 90° that maps the semiaxis into
the positive semiaxis Oy must be performed
a point, it is convenient to consider in the positive direction, or counterclockwise.
two systems of polar coordinates with The reader can easily show that if xOy is
r r
the same pole 0 and the polar axes a right-handed coordinate system and x O'y
a left-handed one, then
Ox and Ox± respectively (see Fig-
,

ure 1.9.2). If ZLxOx 1 a, then point — x' — x cos a + y sin a + a,

M(r, cp) (i.e. point with the polar M y' = x sin a — cos a + b.
coordinates in the first system of units (What is the meaning here of angle a and line
being r and cp) has coordinates rx and cp x segments a and 6?)
«84 1 Functions and Graphs

related to the “geometry” of the curve

(with the size of the circle; see Exer-
cise 1.9.6).
To demonstrate the aforesaid, let us
turn to formula (1.2.4) for the
distance r 12 between two points. If
A 1 (#!, y x ) and A 2 (x 2l y 2 ) are two points,
Figure 1.9.3 then, in view of (1.2.4), the square of
the distance between them, r 2 2 is equal ,

from one system of coordinates to to (x 2 —

xx ) 2 (y 2 yx)
2
+
Let us in- — .

f
another in no way affects the geomet- troduce new coordinates, x and y '
.

ric properties of curves, and for this Then, according to (1.9.6),

reason in geometry the only relation-
ships between coordinates of points that
x[ = x cos a + y
x x sin a + a,
have any meaning are those that retain y[ = —x sin a + x y x cos a+ 6*
their form under any transformation
and, similarly, we can also find the new
of the type, that is, under
(1.9.6)
coordinates x'2 and y'2 of point A 2 in
;a from coordinate system xOy
transfer
terms of the old coordinates x 2 and
to any other coordinate system x'O'y'.
y 2 We then have
.

And these are the relations that were

studied above. — x[ — (x — %) cos a
x'2 2

The can be worded in

aforesaid + (y — sin 2 2/i) cc,

•a different manner. The equation

y = y —
'2 = —(% — ^1) sin a
y'i 2
f(x) or F ( x y) of a certain curve T
,

depends, obviously, not only on the

+ (y — yi) COS a
2

shape of the curve but also on the and, respectively,

•choice of the coordinate system; in
other words, the equation describes not (x 2 — x[)
2
+ {y 2 — y[) 2

the curve alone (Figure 1.9.3a) but = — x cos a + (y — y

[(x 2 x) 2 x) sin a] 2
-a much more complicated object: the + [—(^2 ^ %i) sin a
O/2 — z/i)cos>l
curve T and the coordinate axes (the
axis of abscissas and the axis of ordi-
+ 2

= (x — x (cos a + sin a)
2 2 2
2 x)
nates; and not only the axes but also the
unit segments on the axes that fix the + 0/2 — yi (sin a + cos a)
)
2 2 2

units of measurement for x and y; see = (*2 — Vi + — )* ( */2 £/ l )

2
,

Figure 1.9.36). But of geometric mean-

ing (and we
are interested here in geo-
which proof of the independence
is

metry and not in physics 1 22 ) are only *

of r 12 choice of the system of
of the

those expressions connected with the coordinates.

equation of a curve that retain their Clearly, two straight lines y =
form under a substitution of variables k xx +
b x and y k 2x b 2 (Figure = +
via formulas (1.9.6). For instance, in
the equation of a circle of radius r
centered at point Q (a, 6),
(x — a) +(y — b) — r or
2 2 2
,

(x — a) + (y — b) — r = 0,
2 2 2
(1.9.7)

the quantities a and b characterize

(only) the position of the circle in the
coordinate plane, with the number r [

1 *
22 In physics the situation is somewhat
different, in this connection see Section 9.8.
 x

1.9* Some Additional Topics from Analytic Geometry 65

1.9.4) are parallel if and only if where Z~BB X P and /LxQl are angles with
their slopes are the same, that is, if mutually perpendicular sides). But
= k 2 Let us now consider two strai-
. tanZjr(?Z = tana = k by the very defi- 7

ght lines y = k x x and y = k 2 x which ,

nition of the slope of a straight line,
pass through the origin 0 and are paral- which means that tan Z^BB X P k =
lel to our straight lines (which are not and, hence,
pKi'aVnJi \v 'na'dn T/ftrer, ‘m 'geirei^Jr,
the dashed lines in Figure 1.9.4). Since cos Z. BBaP = cos a = — — —
there are two points A x (1, k x ) and
V + tan l 2
a
A 2 (1, k 2 ) that belong to the new straight _ 1

lines (i.e. the ones that pass through

” j/T+ik 5 ‘

0 ), the angle a, equal to the angle From triangle BB X P Figure 1.9.5)

(see
Z_A x OA 2 between our straight lines
we obtain
(it is unimportant whether these are
= ~6
the new lines or the old lines), is deter- B P = BB cos a |6 1-
1

mined via formula (1.9.1):

l i
/l+A 2

(* 2 -0) (1—0) — — — We have found that the distance

tana = (k\

(1-0) (1 0)-*-(fc 2 — —
0) (1
— 0)
0)
d = B P between
X parallel lines l and Z,
0) (*!
is given by the formula
(since here the points O (0, 0), A x (1, k x ),
and A2 k 2 ) play the role of points —J
_
fti— ft I

(1.9.10)
Al (xt ,
y t ),
(1,
A2
(x 2 , y 2 ), and 3 {x 3 , y 3 ) A V i-t-k 2

in formula (1.9.1)). Thus, the final result If l is a straight line given by the
is equation y kx b and = (x 0 , +
y 0) M
is an arbitrary point (see Figure 1.9.5),
tan a = -~rh (1.9.8) then another straight line Z, passing
+ k ki i
• v '
2
through point is given by the equa-M
In particular, two straight lines y = tion y y0 k —(x x 0 ), or y= kx — = +
kxx +
b x and y k 2x b 2 are per- = + by, where by — y 0 kx 0 (see for- —
pendicular to each other if and only if mula (1.3.3)). Hence, in view of (1.9.10)
kxk 2 -f- 1 = 0, or kxk 2 = — 1. (1.9.9)
the distance h from point to straight M
line l is given by the formula
(Why?)
= = i/
0 — to — 6 0
Now, suppose that y kx + b and h |

V l+F
|

(1.9.11)
y = kx +
two parallel straight
b x are
lines l and l x whose slopes are the same, (note that in the numerator on the
k ,
and which intersect the y axis right-hand side of (1.9.11) we have
at points B (0, b) and B x (0, b x ) (Fig- the absolute value of the result of sub-
ure 1.9.5).
—
In this case, obviously, stituting the coordinates of point M
BB X = on the other hand, the
| &i b |
, into the left-hand side of the equation
perpendicular B X P dropped from point of the straight line, y kx 6 0). — — =
B x onto l forms an angle a with the y
axis that is equal to the angle between l
Exercises
(or l x ) and the x axis (see Figure 1.9.5,
1.9.1. Given a point and a straight line Z:M
(a)M~M (1, -1), y = x i; (b) + M=
M —1),
(0, y 4; (c) = M=M
(—2, 0),
y = — —
— M M
2; and (d) (0, 0) =
(the
origin), y (3/4) x =1/4. Draw a line p
that passes through M
and is perpendicular
to L
1.9.2. Given three points A x A2 A3 , ,
:

A x (0,
J" 0),
A 2 (4, 3), A 3 (-6, 8); A ± (4, -4),
A 2 ( 4,0), A 3 (-l, 1); and A x (1, 2),
Figure 1.9.5 A 2 ( 2, 1), A 3 (- 1, 1). (a) Find A A]A 2 A q ,

5-0946
 66 1 Functions and Graphs

(b) calculate S AAlA2A3 , and (c) find ft Al A2A3 ,

. (1.9.6), that is, to express the old coordinates
1.9.3. Given two parallel straight lines l x and y in terms of the new coordinates x'
and l 4 (a) y x 2 y — x 2 +and = — and1.9.7.
y'.]
—
:

— , ;

(b) 3x 4y 1 0, 3x— = Ay 0. Find the + = Prove that under a transition (1.9.6)

distance d between the lines. to another system of coordinates the following
1.9.4. Given the point and the straight M conditions and formulas retain their form:
line l of Exercise 1.9.1. Find the distance h (a) the condition k 4 = k 2 that two straight
between M and l. lines are parallel, (b) the condition ( 1 9 9 ) . .

1.9.5. Suppose that Ax = At (x ly y x ), that two straight lines are perpendicular to

A2 — A2 (#25 1/ 2)1 A 3 —
A 3 (^ 3 y s ), and A 4
,
— each other, (c) the formula (1.9.8) for the
A4 (x 4 , y 4 ) are four points in the xz/-plane. angle between two straight lines, (d) the
Prove that A X A 2 A SA 4 if and only if
± formula (1.9.10) for the distance between two
(x 2 —x ± ) (x 4
2 — X + (y — y
± ) (y 4
3) V 3) °* — = parallel straight lines, (e) the formula ( 1 9 11 )
. .

1.9.6. Establish how the equation (1.9.7) for the distance from a point to a straight line,
of a circle transforms under a transformation (f) the formula (1.9.4) for the area of a trian-
to new coordinates x' and y' via (1.9.6). Ver- gle, (g) the formula (1.9.1) for an angle in a
ify that the quantities a and b in (1.9.7) triangle, (h) the formula (1.9.5) for the dis-
change (i.e. that the new equation will be of the tance from a point to a straight line, and
same form but with different a and b) while r (i) the conditions for two straight lines being
remains the same. [Hint. To solve this problem perpendicular to each other (see Exercise 1.9.5)
it proves convenient to “reverse” formulas [Hint. See the hint to Exercise 1 .9.6.1
Chapter 2 What is a Derivative?

2.1 Motion, Distance, and Velocity abscissas (the t axis) and the distance
from point M to point 0 on the axis of
Let us examine the translational mo- ordinates (the z axis).
tion of an object, or body, along a In uniform motion, with constant
straight line. Denote the distance of some velocity v, the distance s covered by the
point M of the body to a specific body in time t is directly proportional
pointOonthe line (the coordinate of to t, with the coefficient of proportion-
points /byzrWe will consider tne ’dis- ality ’being v 'jnere s =
vt). "Denote
tance in one direction to be positive and by z 0 the coordinate of the body at
in the opposite direction negative. For time t = 0. The distance s covered in
example, suppose that the straight line time t will then be equal to the dif-
along which our body is moving is ference z (t) —
z 0 Thus. .

vertical. Points above 0 will corre-

spond to positive values of z, and those z (t)[= Zq + vt. (2.1.1)
below 0 to negative values of z. It is
often convenient to assume that the Hence, in uniform motion, the
body is small, so that we can simply dependence of the coordinate z on time t
speak of a material point or parti- is given by a linear function. The
cle, M, and of the distance of this point graph of the function z = z (t) in this
from a definite point 0 on the straight case is a straight line in the £z-plane
line, or the origin of our coordinate (Figure 2.1.1).
system. In the case of nonuniform motion, the
Problems on the motion of bodies function z =
z (t) is expressed by more
with constant velocities v lead to sim- involved formulas and the coorrespond-
ple arithmetic and algebraic computa- ing graph is some kind of curve.
tions based on the fact that distance is Let us analyze the following basic
velocity (speed) multiplied by time, problem, given a function z —
z (t), or
that is, the elementary formula s =vt, the dependence of the coordinate of
where s is distance and t is time. How- the body on the time variable, find
ever, in nature as a rule we deal with the rate of motion (or velocity) v of
motions whose velocities vary with the body. In the general case of nonuni-
time. Studies of such motions lead to form motion, the velocity does not
two important physical concepts, dis- remain constant, it changes in the course
tance and velocity as functions of time,
,
of time. This means that the veloci-
and already at this stage there appear ty v is also a function of the time vari-
the two basic concepts of higher mathe- able t, or v =
v ( t ), and our problem
matics, the concept of the derivative consists in expressing this function in
and the concept of the integral. Start- terms of the known function z z (t).=
ing with this chapter, we consider
these concepts in great detail.
The process of motion of a particle M
consists in the fact that the z-coordinate
of this point changes with time. The
motion of the body (or point M) is
determined by the dependence of z
on £, that is, is characterized by the
function z = z (t). Knowing this func-
tion, we can find the position of the
body at any moment in time. The func-
tion z (£) may be represented graphical-
ly by laying off time on the axis of Figure 2.1.1
 — )

68 2 What is a Derivative?

Everything is simple in the particu-

lar case of uniform motion. The veloci-
ty is defined as the distance covered in
unit time. Let us find, for instance, the
distance traversed in one second from
time ti seconds to time t x 1 seconds. +
This distance (equal numerically to
the velocity) is equal to the difference
between the coordinates z (t x 1) and +
z (tj):

+ 1) — z
Figure 2.1.2
z (t x (f x ) = [z 0 + v (f x +1)7
— + VtJ =[z 0 V.
In the general case of nonuniform
Instead of this we can take an arbitrary motion, the right-hand side of (2.1.2a)
instant of time between t x and t 2 and yields the average velocity y av in the
divide the distance traveled z 2 z x by — interval At between t and t A£ (the +
the magnitude of the interval t 2 tx — :
size of the interval is therefore A£) is

Z2
t2
—
—h
Z1
__
"
(gp + Vt 2)

h—h
(
Z
Q + ^l) _ (9
'
\ 9\} ^ = 5- (2.1.3)

It is precisely because the velocity is We speak here of the average velocity

constant that we are able to take any because the velocity itself can change
interval t2 —h to compute the veloci- over the interval A*.
ty, and the answer will be independent Let us consider an example in which
of both the time t x and the magnitude z (t is given by a quadratic function

of the time interval, \t 2 tx The — |.

instead of the linear function shown in
situation is different in the case of Figure 2.1.1:
a variable rate of motion. =
Before going over to the more general
z (t) z0 + bt + ct
2
. (2.1.4)

case, it will be convenient to change Figure 2.1.2 illustrates a possible graph

the notation. will write t xWe t and — corresponding to a function of the
4-j Ai. so- t.haJL the, di ff erprixo-
7
form (2.1.4) Let us compute the. averr
;

t2 —
t x (the time interval) is denoted age velocity v ay over the interval At
by A* (see Figure 2. 1.1). 21 Similarly, using the formula (2.1.3). In the case at
we write A z to denote the difference hand we have
z (£ 2 )
“z (£ x ) = z (t + At) —z (t) = z (
= z + bt + ct
t) 0
2
y

= A z. z + At) = z + b +
(t 0 (t At)

In this notation, formula can c At) — Zq

(t bt
2
-f- -f- b At
(2.1.2)
be rewritten thus: + +
ct
2
2ctAt + c (A t)
2

— (2.1.2a)
and, hence,

2-1
Note that A is not a factor but a sym-
Az = z (t + A£) — z (t) = bAt
bol, and A* is not the product of A by t just ,
+ 2ctAt + c (At) 2
.

as sin a is not the product of sin by a, and so A

cannot be canceled from the numerator and
From this we get
denominator on the right-hand sides of A7
(2.1.2a) and (2.1.3), in the same way as we v& v = = b -f- 2ct -f- cAt, (2.1.5)
cannot cancel sia from the numerator and
denominator in the fraction sin cc/sin |3. A is Compare the results of (2.1.2a) and
the capital Greek letter delta; At is read (2.1.5) for the average velocity when
“delta and A z is read “delta 2 ”; these are
the motion obeys the law (2.1.1) and
also spoken of as the increment, or change,
in time and the increment, or change, in the law (2.1.4). The second example
path (or distance). differs from the first in that here the
 2.1 Motion Distance and Velocity
, ,
69

average velocity depends both on the We have computed the instantaneous veloc-
ity at time t on the basis of the average veloc-
time t and on the time interval A t.
How can we find the instantaneous
ity over the interval from t t© t At. + Now
let us try to compute it by choosing the inter-
velocity at time t which depends solely val in a somewhat different way. find the We
on this moment in time? average velocity in the interval from t x t = —
The velocity changes gradually, and so (3/4) At to t 2 t = +
(1/4) At. As before, the
duration of the interval is At. From formu-
the smaller the time interval over which la (2.1.4) we get
the distance traveled is measured, the
smaller the change in velocity and, hence
z{h) = z<>+b 1- At)
the closer will the average velocity b
be to the instantaneous value. Formu- 2
la (2.1.5) for u aY contains two terms that + <(*-! a*) ,

do not depend on the size of A t and one

=z ^+-£-A«)+ c (h
term that is proportional to A t. For
very small At the product cAt will also
z ( h) 0 -{-b
—^-A i)

be small, whence the last term on the and

right-hand side of (2.1.5) can be ignored — z (h)~ bAt-\-2ctAt 1
- -^-c{At) 2
z(t 2 ) ‘.

and the formula for v aY will transfer in-

to a formula for the instantaneous
Whence it follows that
velocity:
= b + 2 c £. (
2 1 6)
. .
»av = -i- cAt.

For instance, suppose z 0 = 1, b = 2, (

2 . 1 . 8)
and c =
—1 in (2.1.4). We wish to
find the velocity at t = 0. Equation Comparing (2.1.5) and (2.1.8), we see that
the average velocities over the interval from t
(2.1.5) then yields to t +
At and over the interval from t —
— 1) +
(1/4) At differ by c At [1 —
u RY = 2 + 2 X ( X 0 + — 1) (
(3/4) A* to t
(—1/2)] =
(3/2) c At. But if we want to find
X At = 2 — At. the instantaneous velocity, we have to take a
very small time interval At. Then the differ-
We can construct the following table: ence between the two expressions for the
average velocity will vanish and we again ob-
tain for the instantaneous velocity iq nS f =
to (from 0 to At s) 1 0.5 0.1 0.05 0.01 b 2ct. However, the value of the average
u aY 1 1.5 1.9 1.95 1.99 velocity over the interval from t (3/4) At to —
t + (1/4) At given by formula (2.1.8) is closer
to the instantaneous velocity (2.1.6) than the
value of the average velocity over the interval
Thus, the smaller the size of At, the
closer v aY is to 2, and it is natural to
from t to t +
A* given by formula (2.1.5).
This implies that (with greater precision) the
assume that v lnst = 2. choice of the interval from t to t At is less +
The attentive reader will have most convenient: here v aY is a worse approximation
for i?i ns t. It is even better to select the interval
likely recognized the expressions (2.1.4)
in such a way that the value of t we are in-
and (2.1.6) from the school course of terested in lies at the center of the interval;
physics to be the formulas for uniform- in this case, for the quadratic dependence
ly accelerated motion: (2.1.4) between z and t the value of vaY will ,

coincide with iq ns t (see Exercise 2.1.2). We

2 {t) = z +v +
0 0t — 2

,
v(t) =v 0 +at.
will dwell on this fact again in Section 6.1.

(2.1.7) We have considered the concept of

instantaneous velocity for two specific
All that is needed is to substitute for b cases: uniform and uniformly accelera-
in (2.1.4) the initial velocity v 0 (i.e. the ted motion. In Section 2.3 we give
velocity at time t 0) and for c in = a more exact definition of instanta-
(2.1.4) the “half-acceleration” a/2 (since neous velocity for an arbitrary law of
a is the acceleration of the body). motion.
 ) )

70 2 What is a Derivative?

Exercises initial temperature (it is unimportant

2 . 1 1 . The following dependence of the
.
which initial temperature specifically)
distance (z) traveled by a material particle M up to temperature T is the analog of
along a straight line on time (t) was observed: the distance z of Section 2.1, and it is
(a) z =t* + t, and (b) z = (1 + t*)~K What clear that this quantity depends on T,
isthe average velocity vaY of each motion over
the time interval from t to A tl What is the or Q — Q (T). Heating the 1 kg of the
instantaneous velocity iq ns t at time tl substance from temperature T 1 to tem-

on
2.1.2. For the quadratic dependence of z
given by formula using
perature T2 requires Q Tx
( ,
T2 )
=
the
t

formula
(2.1.4), find vaY
Q (T 2 ) — Q
(7\) joules of heat, while
heating the same amount of sub-
stance from T to (T AT) °C requires +
Q (T + AT) -Q (
T = AQ joules of
heat. Therefore,
average specific
the
and compare the result with the value of iq ns t heat capacity c av over the temperature
at time t.
interval from T to T AT is natural- +
ly defined as
2.2* Specific Heat Capacity of an
Object. Thermal Expansion Q(T + AT)-Q(T) AQ
AT ~ AT ’

Note that a procedure similar the to

one discussed in the previous section is
and the instantaneous specific heat ca-
often encountered in many physical
pacity c inst (the word “instantaneous”
problems. We will demonstrate this
refers in this case not to a definite mo-
using two simple examples and the know-
ment in time but to a definite tempera-
ledge that a school physics course
ture defined as a value of the
T) is
supplies.
average specific heat capacity c RY over
The reader will recall that the specific
a very small temperature interval A T,
heat capacity of a substance is the quan-
and the narrower the interval the closer
tity of heat (in joules) required to change
c aY will be to c inst Note that in the .

the temperature of 1 kg of the sub-

majority of cases the value of 1 °G
stance (water, iron, gold, etc.) by 1 °C.
for AT is sufficiently small to yield an
But for different initial temperatures =
c (T). Here the ex-
exact value of c
the quantity of heat required to change
pression “sufficiently small” means that
the temperature of 1 kg of substance by
the value of c obtained in this manner
1 °C is different, so that the specific
will for all practical purposes coincide
heat capacity function of tempera-
is a
with the value at which we arrive by
ture T, or c =
For instance, to
c (T).
selecting a smaller (and even consider-
heat 1 kg of iron taken at a tempera-
ably smaller) interval AT of tempera-
ture of 0 °C up to 1 °C we must supply
ture variation.
440.857 J of heat, while to heat the
Example 1 The quantity of heat
.

same amount of iron taken at 50 °C up

to 51 °C we must supply 470.583 J.
Q = Q (T) required to heat 1 kg of
iron from 0 °C to T °G is given by the
Then how does one define the heat capac-
following empirical formula (which
ity of a body corresponding to a fixed
quite sufficiently represents the process
temperature T ?
we are interested in, at least for
The content of Section 2.1 suggests
the following method for defining the
T < 200 °C):

quantity we are interested in, c c (T) — Q T - 440.857 T + 0.29725 T\

(
(the specific heat capacity at tempera-
ture T). Here the quantity of heat Q (
2 2 2)
. .

(in joules) that must be delivered to the

specified amount (1 kg) of the substance In full accordance with the material
in order to heat the substance at the of Section 2.1 this yields (note that
 ) )

2.2 * Specific Heat Capacity of an Object Thermal Expansion 71

the function (2.2.2) is quadratic, just However, this elongation refers not to a
as (2.1.4) is) 1-cm rod but to a rod of length L, so
(7+A7 1
2 P er unit ro(* we ^ ave
= 440.857 (7+A7 )+0.29725
7
A <9 )
(on the avera £ e )
at
440.8577 + 0. 29725 T
7 2
J_
AL
440.857 * (2.2.5)
AT L AT
(2.2.5)
The value of k av given by formula
7
0.59457 + 0.29725A7
7

+ (2.2.3)
2 2
the approximate value
is of -

and, hence
the (instantaneous) value of the coeffi-
Cinst - 440.857 + 0.5945 T
7

. (2.2.4) cient of linear expansion k k (T) of =

the substance we are interested in, and
We see that c (0) = 449.857 J/kg X
the smaller the temperature interval
°C, as we noted earlier, while, say,
- - AT we select, the closer fc av will be to
c (100) 440.857 + 0.5945 X 100
^inst*
500.307 J/kg-°C.
Example 2 For platinum, within a
A similar situation arises in all prob-
broad temperature
.

range, the function

lems where the quantity we are inte-
rested in is the rate of variation of an-
L =L ( T ) we are discussing here obeys
fairly well the following empirical
other quantity. For instance, the coef-
formula:
ficient k of (thermal) linear expansion
is usually defined as the number that L{T) 1 =
8.806 X 10~ Q T+ +
shows by what amount a thin rod made + 1.95 X 10- 9 T 2 (2.2.6)
of the substance of interest (the length
of the rod is assumed equal to 1 cm) (note that the dependence of L on T
changes its length when its tempera- is again quadratic). wish to express We
the coefficient of linear expansion
ture is increased by 1 °C. Here too the
number k is not a constant: it depends k =
k ( T of platinum as a function of
T and find the values k (0) and k (1000).
on the initial temperature T of the rod,
Answer Just as we did above (check
whereby, strictly speaking, our defini- .

all the calculations yourself), we find

tion, which involves heating the rod
that
from T to (T +
1) °G, is not quite cor-
rect, since in the process of heating the
quantity (or function) k k (T =
varies. 8.806 X10“ 6 + 3.90 Xl0- 9 f (0 0 * ^
Then how we
determine
are to
— 1_|_ 8.806 Xl0" 6 r+1. 95 XlO- 9 T 2 ’ '
*
'

&in S t? or the quantity k corresponding from which it follows that k (0) =

to a definite temperature 7 ? Let us take
1

8.806 X 10~ 6 (1/°C) and k (1000) =

a rod 1 cm long at 0 °C and gradually 12.57 X lO" 6 (1/°G).
heat it. The length of the rod, L, will
change in the process and, hence, is a (2.2.5)
Exercises
function of the rod’s temperature T,
or L = L (T). If we heat the rod from 2.2.1. It known that the amount of heat
is

temperature T to temperature T + Q T
mond
( )
(joules) required to heat a 1-kg dia-
from 0 to T °G can be expressed fairly
AT, where the temperature incre- (!)

ment AT assumed small, the length

is 2-2
The approximate nature of formula
of the rod will increase by manifests itself in the fact that we
relate the elongation AL (more precisely the
AL = L {T + AT) —L (T), specific elongation AL/ AT) to the length
L =L (T) of the rod at the initial tempera-
so that the following average elonga- ture, although in reality the rod’s length
tion of the rod corresponds to an in- changes continuously in the entire heating
crease in temperature by 1 °G: process. (Why should we divide AL/ AT by

L{T + AT) — L(T) AL

L (T) and not by L (T AT)1) This inaccu- +
_
~ '
racy in formula (2.2.5) for k is unimportant.
AT AT (Why?)
 ) - —

72 2 What is a Derivative?

well by the following empirical formula (which it is assumed that A z (the numerator) is
is valid in the 0 to 800 °G temperature range): not just any distance interval but pre-
Q T (
= 0.3965T + 2.081 X KHZ’* cisely that distance that corresponds to
—5.024 x ?
io- r 3
.
the time interval A t. This was obvious
in formula (2.3.1) from the way the ra-
Find the average specific heat capacity c av of tio was written; the numerator is the
diamond over the interval from T to
(T +
AT) °G and the “instantaneous” specific difference z (t 2 ) —
z (^) of the values of
heat capacity c =
c (T) as a function of the the function z =
z (t) at t 2 and at t x \

temperature T. What are the values of c (0), the right-hand side of (2.3.2) is simply
c (100), and c (500) for diamond? an abbreviation of the ratio (2.3.1).
The quantity of heat Q (T) required
2.2.2.
for heating 1 kg of water from 0 to T °G is Thus, the quantity that interests us,
fairly well represented by the following empi- the instantaneous velocity y inst ==
rical formula: v(t x ) at time t u is the limit of the
Q (T) = 41 86.68T + 8373.36 X 10
~b
T2 ratio Az/At as A t tends to zero (where

+ 1256 x io r 6 3
.
At = t2 — t x ). This statement can be
written as
What is the specific heat capacity c (T) of
water? Find the values of c (10) and c (90). v(t i )= lim
2.2.3. Prove that (for every substance and At^O Jf.
1X1
at any temperature T) the coefficient of volume
expansion (also known
as the hulk expansion Here u t is precisely the instantaneous
( x)
coefficient ), that the “instantaneous rate” of
is,
increase in volume of substance resulting from velocity and lim stands for “limit”.
an increase in temperature per unit volume, The particular kind of limit we have in
is three times the coefficient of linear expansion. mind is indicated underneath lim
when A t approaches zero, and the arrow
2.3 The Derivative of a Function. stands for “approaches”. The quantity
Simple Examples of Calculating to the right of lim is the one whose li-
Derivatives mit is being sought. The notation is
read as follows: “the limit of A z over
In Section 2.1 we considered the prob- At with At approaching zero,” where the
lem of instantaneous velocity and exam- word “over” replaces the phrase “divid-
ined ratios of the form [z ( t 2 ) — ed by the corresponding value of.”
z (^i)l /(^2 *i), —
where the values What meaning do we attribute to the
t x and t 2 must be considered very close- terms “limit” and “approaching the li-
lying. The expression “close-lying” is mit”? The calculations carried out in
not exact, that is, mathematically it is Section 2.1 served as an illustration of
not rigorous. The exact formulation is these notions. We
saw that for small
this. It is necessary to find the limit to intervals At the value of i; av in Example
which the ratio 2 in Section 2.1 differed from the value of
z( * 2) ~* (fl) i?inst by a quantity proportional to
r2 — ti
(2.3.1) At. For small At this quantity is small,
too: the smaller the A t, the smaller the
tends as t 2 tends to tx . Using the desig- quantity, and for infinitely small val-
nations A t and A z, we can rewrite this ues of At (i.e. so small that they can
ratio as be assumed negligible against the back-
ground of b and 2 ct in (2.1.5)) the
= (2.3.2) quantity also becomes infinitely small.
It is precisely for this reason that we
and the condition t2 tx now takes the
can neglect the term with At in the ex-
form A t -* 0. pression for z; av when At is small.
In (2.3.2), the quantities At and A z
Thus, the ratio
are related: any time interval A t —
t2 — t± can be selected, but after Az z(t 2 ) — z(ti)
‘’tne’ denominator* Ai nas neen selected, At t2 — 1 1
 , ) t

2.3 The Derivative of a Function 7 $

or, as we predominantly wrote in Sec- of the independent variable as the in-

tion 2.1 by replacing t1 with t and t2 crement of the independent variable*
with t + At, tends to zero is of prime importance in*
higher mathematics and its numerous
Az _ ^ (£ + At) — z (
t)
applications. We have already seen,
At “ At
(2.3.3a)
for example, that such an important
tends to a definite limit when At = concept as the (instantaneous) velocity
t2 —
tx tends to zero.
2 8
-
The cor- of motion is found with the aid of the*
responding limit is the instantaneous limit of such a ratio. (In Section
velocity v which is also a function of 2.2 we reduced to a similar pro-
t: cedure the problems of calculating the
specific heat capacity and the coeffi-
lim
A*-0 ^= v(t). (2.3.4) cient of linear expansion; think over
these examples.) This is why the limit
Why is that, when computing the of this ratio has a special name: the de-
velocity from the given formula z (t), rivative of the function or, simply, the
we have to carry out so many calcula- derivative This name is due to the fact
.

tions and find Az for distinct At and that if z is a function of t, or z = z ( ),

Az then the limit (2.3.4) depends both on

only then find the limit lim -r— ?
=
the type of function z z (t) and on the
Az-0
Couldn’t we simply take the value value of t at which this limit is calcu-
At =
0 from the very start? No, because lated; in other words, this limit is also
5
in this case we would simply get Az 0, = a function of t, a new function derived
'

since Az = z(t + At) — z(t ), and from the old function z (t). It is natural
if At =
then Az 0. By this thought-
0, — then to speak of this function as the de-
less mode of operations we would rivative, where the word derivative
get Az/At 0/0, =
which means we stresses the dependence of this new func-
would get nothing definite. tion on the basic function z = z (t).
When computing velocity, the whole We
have special notations for the de-
idea is to take small At and small Az rivative. One notation (differential no-
that correspond to At In this way, we .
tation) is

obtain a very definite ratio Az! At each dz Az

'

time. When At is reduced, (tends to ze- Hi

f
= Jim
l A<-0 A7,
ro), then Az diminishes in approximate
proportion to At, and so the ratio re- Here the quantity dz/dt (it is read
mains approximately constant, that is, u
d z d t”) is not a fraction but is simply
the ratio Az/At approaches a definite an abbreviated way of writing the lim-
limit when At tends to zero. 2,4 The it on the right. The quantity dz/dt
magnitude of this limit is the instanta- is written in the form of a fraction to-
neous velocity v ( t for the case where remind us that it is obtained from the*
t is time and z is distance.
fraction Az/At by a passage to the
The limit of the ratio of the incre- limit. 2,5
ment of the function to the increment A
different notation for derivatives is-
the prime notation, v z' (£), or, for =
2- 3
We
could have also started from the
example, for the function y =
initial formula (2.3.3) for the ratio and assu- y (x),
med that * 2 an d t x tend to a definite value of t
— = y' (*) = -37=
,

so that At t
2 t —
± tends to zero (in partic-
y' lim
Ax
ular, see the text in small type at the end Ax-*0
of Section 6.1).
2 4 -
In some “unnatural” laws of motion 2
Below (in Section 4.1) we will see that-
-

z = z (t) this limit may not exist; the re- the expression dz/dt (or dy/dx) can also be
spective restrictions will be considered later. interpreted as a fraction; for the time beings
We advise the reader not to think, at least however, the reader must interpret this nota-
for the time being, about such exceptions. tion as simply a symbol.
 t

74 2 What is a Derivative?

Next we remove the brackets in the nu-

merator,
A z= (t + to) 2 —= t
2
t
2
+ 2tto
+ (to)
2 - t
2 = 2tto + (to)
2
,

and get
Az _
~
2tAt + (At) 2
2t + to. (2.3.7)
At At

Since the term on the right-hand

first
Figure 2.3.1 side of is independent of At,
(2.3.7)
where only one term is left that does
Occasionally, in place of the func- not decrease as we send At to zero,
tion symbol, one gives the expression and we can write
of the function: if z at 2 b, then = + dz d(t 2 ) Az
instead of dzldt we can write directly
dt dt A*
d (at +
b)/dt or (at 2
2,
b) + f
.
a<-*0

Thus (and this is extremely impor-

tant), the derivative of a function is de-
-= lim
At + Q
(2t + At) = 2t. (2.3.6a)

fined as the limit of the ratio of the incre-

Let us consider another example:
ment of the function to the increment of
the independent variable as the latter s = t\ (2.3.8)
tends to zero :
Here
—f—
dx
dy
— ,.
lim
Ax-* o
~
Ay
^x
. (2.3.5) Az — (t-\- At) 3 — = t
3
t
3 -{-
3 t 2 to +3 1 (to) 2
(to) 3 — = 3t to + 3 2
St (to) 2 + (At) 3 ,
The instantaneous velocity of motion
-^ = 3* + 3*A* + (A*) 2
2
of a body is equal to the derivative of ,

the body’s coordinate with respect to

and
time. By analogy with this, when x
is not time and y is not distance, we dz __ d (t s )

nevertheless speak of the derivative ~dt

~ dt
dyldx as the rate of change (or variation)
of the function y under the variation of
= Hm [3t 2ji-StAt Jr(At) 2 ]==St 2 , (2.3.8a)
A^O
the indepedent variables. For instance,
using the notations of Figure 2.3.1, The limit was readily found in these
we can say that the ratio Ay! Ax is the examples because At canceled out when
“average rate” of increase of the function we computed the ratio Az! At. Let us
y on the interval within which AB consider a more complicated example:
the independent variable x varies,
while the limit of this ratio, with (2.3.9)
point B approaching point A, expresses
the rate of growth of y at point A on In this case we have
the x axis. 1 i_
Let us now find algebraically the de- Az t

rivative of the function

t~
~E At '

-2 w t
2
(2.3.6) Can we disregard the quantity A£
in the first fraction, in the expression
(we have performed this operation in
Section 2.1). Form the ratio
1 /(t +
A^), when we pass to the limit?
No, we cannot, because we have not
Az _ (Z-f-Az) 2 — t
2
yet canceled out the quantity A£ in
~A T~ At *
the denominator. By substituting 1 It
 AJ

2.4 Properties of Derivatives 75

for II (t + At) when A t small, we in-

is But this, as can easily be seen, is
troduce a small error (if At
is small) in sure to be the case if At in absolute
one of the summands of the numerator value is made smaller than 0.02 (and
of the fraction Az/ At. But in this frac- as At is made smaller than 0.02
tion both numerator and denominator we will never encounter a value of the
are small if At is small. For this reason difference greater than 0.0025).
we cannot allow for a small error in the The same goes for other functions as
numerator (it transforms the numera- well: the approach to a limit as At ->0
tor into zero). signifies the opportunity of choos-
Here is the proper way to do this: ing At such that any degree of close-
—U+ ness to the limit is attainable (this
A _ _ i
1_ __ t
limit is then never lost).
t + At t
'
t(t+At)
Finding the derivative in the special
At Az 1
case of z t =
is particularly simple:
t (t + At) * At t(t + At) quite obviously, Az A£and Az/At 1, = =
Now we can find the limit (the deriva- that is, the ratio Az/At is equal to
tive) by dropping At in the denomina- unity for arbitrary (large or small)
tor: At and hence in the limit as well. Thus,
d
lz=
dt
__d
m_
dt
Ai-0 L
r
t(t
l

+ At)
i
J f
2 *
if z _ ( , then -g-=-£=l. (2.3.10)

(2.3.9a) Finally, if we consider a constant,

z = can also be regarded as a spe-
c, it
These examples illustrate a very im-
portant property, the fundamental prop-
cial case of a function the graph of —
this function is a straight line parallel
erty of limits (this property is com- to the x axis (see Figure 1.3.6). In this
monly taken as the definition of a limit). case clearly Az —
0 for all At, whereby
As At is made smaller and smaller, the the following rule holds true:
difference between the value of the ra-
tio Az/At and the limit of this ratio if 2 -e, then •£=-£ = 0. (2.3.11)
(this limit is the derivative), lim =
—dz
,
may be made as small as we please, 2.4 Properties of Derivatives.
Approximating the Values
which is to say, less than any given of aFunction by Means
number. What is essential here is that
of a Derivative
the difference then remains smaller than
the given number in the further process Here are some general properties of de-
of variation of the ratio Az/At. rivatives.
An example will serve to illustrate If a function is multiplied by a constant
this point. For z lit we have = factor ,
then the derivative is multiplied
dz Az
by the same factor For instance,
.
1
=“ 1
~dt~ W' ~A t (t + At) •

if z = 3* 2 , then ~dt dt
Let us take, say, t = 2. Then dzldt =
—0.25. Can we choose At such that ~3 —3 X =6 f.
Az/At differs from the limit, or —0.25,
by less than 0.0025? What we mean is In general,
that At will have to be chosen such that
if z(t) = ay(t), then -§-=«+.
Az _
/
\~
1
~ ““4 1 \

At 2 (2 + t) + 2A t J (2.4.1)
lies between— 0.25 + 0.0025 =—0.2475 It is also obvious that the derivative
and —0.25 — 0.0025 = —0.2525. of the sum of two functions is equal to
 ) ) ) t) , t ?

76 2 What is a Derivative

the
tions:
sum of the derivatives of the two func-
whence ~=at at dt , and so on.
It is now easy to find the derivative
if z(t) =x ( t) +y (t), of a polynomial . We already know that

dx dy dC rN dt a d(t 2 ) _9
then (2.4.2) ~dT 1,
f
dt ~dt dt ~df' dt

The last rule can easily be generalized dm : 3t 2 .

(2.4.4)
dt
so as to include the sum of three, four,
and, in general, of any number of func- This yields
tions.
d (a -)- bt -{- c£ 2 -|- efi) da dt dt 2
Using the above two rules, we find “ nr
. i
0 c
~di ~dt dt
that the derivative of a sum of several
functions taken with constant (but, gen- -\-e^- = b^2ct+3et 2
.
(2.4.5)
erally speaking, different) coefficients
is equal to the sum of the derivatives of
For instance, in Section 2.1 we consid-
these functions with the same coefficients :

ered the function z (

t) = z0 + bt +
if z t
( )
= ax t
( ) + by (
t + cu t
( )
ct
2
(see formula (2.1.4)). In view of
(2.3.5), where we must put a z 0 and =
dz dx dy du =
ir = a ^r + b ^r +c ~dF
then
,
, .

e 0, we have

(2.4.3) V (t) = -~ = b + 2ct,

Each of these rules is readily proved which is the result we arrived at in
starting directly from the definition Section 2.1 without turning to the
of the derivative: these rules hold true general properties of derivatives.
for the increment Az z (t At) = + — The technique for finding derivatives
z (t of the function z z (t) (for = (also called the differentiation of func-
any At), whereby they are valid for the tions) is given in detail at the beginning
ratio Az! At and for the limit of the ra- of Chapter 4.
tio, or the derivative dz/dt. For in- Running ahead a bit, we may point
stance, if z ( t =
ay ( ), with a constant, out that finding derivatives of functions
then given by formulas is a relatively simple
Az = z (t + At) —z t
( )
= ay (
job, much easier, say, than the solu-
tion of algebraic equations. The for-
+ At) — ay t
( )
= a [y (t + At) mulas for the derivatives of functions
—y (01 = a &y are often even simpler (and never more
and, hence, complicated) than the formulas defining
the functions. For instance, if the func-
Az _a Ay dy tion is a polynomial, its derivative is
At
for all A t, i.e. =a -
At dt dt also a polynomial, and the new polyno-

= x(t) -\-y
If z (£) then mial is simpler than the initial polyno-
(£),
mial in the sense that its degree is low-
Az = z + At) — z(t (
t er the above example of a third-
(see
degree polynomial, whose derivative
— [% (t-)rAt)-\-y (£+A£)] — [x t)-\-y ( (£)]
proved to be a quadratic function of the
= [x(t + At) — x (^] independent variable; a similar situa-
+ [y + At)—y(t)] = Ax + Ay
{t tion arises for polynomials of other
degrees). If the function is an algebraic
and fraction, then the derivative is also a
Az Ax -{-Ay Ax Ay fraction. If the function contains roots
—
T~
,

”A At At '

"A 1 ’ of fractional powers, then the derivative

 x ) )

2.4 Properties of Derivatives 77

also contains them. The derivatives considerably if one employs a pocket

of trigonometric functions are also tri- calculator.
gonometric functions, and in some ca- The derivative dzldt is defined as the
ses (the logarithmic function or inverse limit of the ratio of the increments
trigonometric functions, for instance) AzIAt as At ->0. When At is not equal
the derivative proves to be a simpler to the ratio of the increments
zero,
function (an algebraic fraction for the Az/At is not equal in general to the de-
logarithm and the arctangent). rivative dzldt but this ratio is ap- ,

Finding derivatives does not require proximately equal to dzldt and the appro-
any kind of special ingenuity or imagi- ximation is the better the smaller
nation. The problem is always solved At is. Therefore, we can write the ap-
in a neat fashion through the use of proximate relationship
the simple rules given above. As we
Az
have already said, other rules and ex- /x-'
dz
— —z ,
(t),
At dt
amples of application will be given in
Chapter 4.
Az ~ At = z' ( t) At. (2.4.6)
So far, all the functions we have con-
sidered are defined by formulas. One
should not think, however, that this is From this we can find the approximate

absolutely necessary for the existence of value of the function z (t At): +

a derivative. For example, we can re-
gard the dependence of distance cov- z [t 4- At) = z + Az ~ z + -^-At
(
t {t)

ered upon time as having been found

from experiments, in the form of very
= z(i)+z'(f) At. (2.4.7)

extensive tables. It is clearly possible, Note that in (2.4.7) the first equality
using these tables, to compute the in- sign is exact in accord with the defini-
stantaneous velocity (i.e. the derivative) tion of Az and the second one denotes
by forming the ratios 4r = — ~~~
/
approximate equality.
Let us now return to the designations
for various A t and observing how the ra- t2 = t At and tx — t used earlier.
tios behave when we send At to zero. We can then rewrite Eq. (2.4.7) as
Of course, here we cannot speak of the
limit of the ratio Az/At with At ap- z (£j) ~z (t x ) + z'(^) (
t2 — tj.
proaching zero, since for a function spec- (2.4.7a)
ified by a table there is no way in which
wa cam makn At. as. small as. desired Thus,, given a small difference to-t^,
{At cannot be made smaller than the that is, t 2 is close to t u the function
step interval of the table), whereby we z (t 2 ) can be expressed by an approxi-
^cannot find the exact value of the deriv- mate formula involving the value of
ative. In the majority of cases, howe- the function z ( t and its derivative
ver, tables yield a sufficiently good esti- z'(t) at t =
t v Note that this formula

mate for the derivative dzldt, which is linear in t 2 (to the first power).
•exists for all “good,” or “well-beha- Formula (2.4.7) (we will return to
ved” (or “smooth”), functions, regardless this formula in Section 4.1) is extreme-
of the origin of the functions and the ly important, whereby we discuss it in
way in which the functions are specified. detail. Let us take an example. Sup-
We note also that the “arithmetic cal- pose v =
z3 and for the sake of clarity
,

culation of the derivative,” or finding we assume that we are speaking of a

the values of the ratio AzIAt (or, for cube of volume v and edge x. The deriv-
the function y = / (#), the ratio Ay! Ax) ative dv/dx, as we know, is equal to
for a number of decreasing values 3x2 , with the result that (2.4.7) is
of At (or Ax) and observing the beha-
vior of the values of the ratio is simplified
,
v ( + Ax) = (x + A#) 3 ~x+ 3
3x2 Ax,
 78 2 What is a Derivative?

that is, edge

is not precisely 1 but 1.1 mm mm
(which, of course, is quite a realistic
Ay = v (x + Aar) — v (x) ~ 2
3x Ax,
assumption), the first term on the right-
(2.4.8) hand side of (2.4.8a) will be equal to
while the exact expression for v (x + 0.0033 instead of the initial value of
Ax) is, in view of the well-known Tf^Uh, anh a'ii hrgfrsTn wiJihje-
formula, come unreliable. So is there any rea-
son to perform unnecessary work and
clutter up the calculations with unnec-
+ Ax) = (x + Ax
v (x )
3
essary figures?
= x + 3x Ax + 3x (Ax) +
3 2 2
(Aa;)
3
,
A
similar picture arises at other va-
lues of Ax (values that can still be con-
whence sidered small). For instance, at
Ax = 0.005 m =
1/2 cm) we have
Av = 3x2 Ax + 3x (Ax) 2 + (Ax) 3 .
(

(2.4.8a)
= 3xl X 0.005 = 0.015,
3x Ax2 2

3x (Ax) = 3 X 1 X (0.005)
2 2

Suppose that the edge is 1 m long. Then = 0.000075,

the cube’s volume is, obviously, 1 m 3
.

But what can we say about the vol- (A*) = (0.005) = 0.000000125.
3 3

ume if it is found that in measuring the

Thus, the final result is
edge length we introduced an error,
that the edge is somewhat longer than y (1.005) = (1.005)
3 = 1.015075125,
1 m, say, by 1 or 5 or even by 1 or mm where, of course, we can also confine our-
2 cm? selves (and even more, this is necessary,
If our error amounts to 1 mm, then
as a rule) to the approximation v
actually x =
1.001 m, and the initial
1.015.
value of the volume v 1 must be = The results of numerical calculations
increased by Av given by formula associated with this example are listed
(2.4.8a). Let us estimate the contribu-
in detail in the table below, from which
tion of each term on the right-hand side
the reader can see that even at Ax —
of (2.4.8a):
0.02 m—
2 cm, or with low require-
3* 2 A;r = 3xl 2
X 0.001 - 0.003, ments concerning the accuracy of the
3x (Ax) 3 = 3 X 1 X (0.001)
2 = result and at Ax 0.05 —
5 cm, m=
= 0.000003,
replacing (x Ax) 3 with + x + 3
3x2 Ax
is quite admissible:
(A*) 3 - (0.001)
3 - 0.000000001;
= 1 Ax 1+0=
= (1++ A#)
x 1 1.001 1.005 1.01
here 3
v 1 1.003003 1.015075 1.0303
v (1.001) = 1.003003001. 1+3AZ 1 1.003 1.015 1.03

Az; - 0.003003001. (2.4.9)

x = + A#
1 1.02 1.05 1.1 1.5 2
But is there any sense in writing out i> = (l + A;r)3 1.061? 1.1576 1 .3310 3.375 8
all the digits in y? The third term on the 1+3A* 1.06 1.15 1.30 2.50 4

right-hand side of (2.4.8a) is smaller

than the first term by a factor of
Another example is y = Yx (say,
y is the side of a square plate with
3 000 000 (three million!), whereby re-
taining the third term in this case can-
area or mass 2 6 x). find the values
*
We
not be justified. The second term is al- 2
The mass m of a homogeneous square
- 6

so smaller than the first term, by a fac- plate whose side is y is equal to pz/ 2 where p is ,

a constant (the plate’s density); the function

tor of 1000, so that the precision it
m — 2
p y differs from x
2
y by an =
unim-
provides for the final result is illusory, portant factor p (in this connection see Sec-
since if the error in measuring the cube’s tion 1.7).
 1

2.4 Properties of Derivatives 79>

of the function y for x close to 4. In pendent variable x where the derivative ,

this case y (4) = ]/ 4 = 2. The deriv- dyldx maybe considered as the-

still

ative z/' (x) = 1/2]/ x (see Exercise

rate of variation of function y at a given
value of independent variable x (see
whereby z/' (4)
2.4.2), = 1/2]/ 4 = 1/4,
Figure 2.3.1 and the text referring to
and the approximate formula (2.4.7)
this figure).
is of the form
The cases we have examined (where
we denote by t the independent vari-
y (x) == ]/ 4 + A;r ~ 2 -f- 0.25A;r.
able and by z the value of the function
and assume that t is the time and z-
We again compare the approximate is the distance covered) confirm this
and exact values of y (x): conclusion. In the first example, z —
3
t ,
when t varies from 1 to 2 (at
x— 4 + Ax 4 4.1 4.5 5 At —
1), the derivative z'(t) 3 1 2 var- —
y=Y 4+ A* 2 2.02485 2.1213 2.24 ies from 3 to 12 (which is to say, by a
2+0.25A;r 2 2.025 2.125 2.25 factor of 4). In the second example,
z = ]/£, when t varies from 4 to 9*
.

(At =
5), the derivative z ( t ) 1/2}/ =
x— 4 + Ax 6 7 8 9 varies from 0.25 to 0.167 (or roughly by
y= Y 4 + A:r 2.45 2.65 2.83 3 30%). Therefore, in the latter instance
2 + 0.25A;r 2.50 2.75 3.0 3.25
formula (2.4.7) or (2.4.10) yields a good
result for larger values of At; for exam-
Let us now return to our basic exam- ple^! At =
4, which constitutes a 100%
pie concerning distance, time, and ve- increase against the initial value of t ,

locity, that is, we assume that t is the the error introduced by this formula
time, z (t) is the distance covered in constitutes only 6% of the true value
time f, z (t) =
dz/dt is the instantane- of the function, while at At 5, which =
ous velocity, and A z is the increment constitutes a 125% increase against
in distance, that is, the distance cove- the initial value of t, the error consti-
red during the small time interval A t. tutes 1/12 ^
8% of the true value. On
The formula the other hand, if we turn to the func-
tion z t
3 —
we see that (2.4.7) be-
Az ~ z' (
t) At (2.4.10) comes invalid already at At
,

0.5, when =
the error constitutes 25% of the true
then signifies that the distance covered
is equal to the product of the instanta-
result (at At =
l the error constitutes
a 100% of the true result). Of course,
neous velocity and the time interval.
what we have said here is valid for
But the instantaneous velocity itself
both positive and negative values of
varies with time. Therefore, the appro-
(2.4.10) At Exercise 2.4.5).
(see
ximation expression (2.4.10) is true
only when the instantaneous velocity
Adetailed discussion of the range of
application of formula (2.4.7), of the
does not perceptibly change during
question of estimating the error intro-
the time At. Hence, the faster z' (t)
duced by this formula, and of the vari-
varies, the smaller At must be taken in
ous refinements is given in Chapter 6.
and conversely, the slower
;
Anticipating the first sections in
z (t) varies, the larger At may be taken.
Chapter 6, we note that for small differ-
That is to say, the magnitude of the
increment At for which formula
ences t 2 —
t ± the function z (t 2 ) can
be represented as
(2.4.10) still yields a small error de-
pends on the rate of change of the deriv-
ative over the interval At. Of course, z (t^) = z (tj a (t 2 — £x ) +
the same holds for an arbitrary func- b (t 2 — tj)
2
+ c (t 2 — £i)
3

tion y —
f (x) of an arbitrary inde- (2.4.11)
 1 x

80 2 What is a Derivative?

and this relationship can be considered Exercises

«exact if we assume that the sum on the 2.4.1. Find the derivatives of the follo-
right-hand side consists of an infinite wing functions: (a) y = z 4 (b) y — 4a: — 3

+ 2x —
,

number of summands (the exact mean- 1, (c) y = (2x + l) 2

(d) y = lx 1
2
,

y = a (x +
,

ing of this assertion will be discussed (e) y ax

i/x), and (f) = 2
+ b/x 2
.

2.4.2. Prove that the derivative of y =

in Chapter 6). Since we assume that
t2 tj—is small, each term in (2.4.11)
V is equal to l/2y x. [Hint. Multiply
the numerator and denominator ^f
is smaller than the preceding one, since —
(V* + Ax y x)l Ax by the sum yx + Ax
it contains a higher power of the small
difference t 2 —
t v The first two terms of
+Y2.4.3. Prove
x. ]
that the derivative of y =
the general formula (2.4.11) coincide y equal 3 is
to (3/2) x. [Hint. y
Use the
with the right-hand side of the approxi- method employed in Exercise 2.4.2.]
mate formula (2.4.7a), since a z' (t-f) = 2.4.4. Find
2
(1.2)
2
(1.1)2, (1.05)
using formula (2.4.7). Compare the
2
, ,
and
= dz (tj/dt. The difference between the (1.01)
resultswith the exact values.
approximate formula and the exact one 2.4.5. Using the expression for the deri-
lies in the fact that the latter contains vative of the function z (t) 2 20 1 5 12 = + — ,

-a term proportional to ( t 2 tf)

2
and — find z (1.1), z (1.05), and 2 (0.98). Compare the
results with the exact values. [Hint. In the
terms with higher powers of this small last case take t 1 and At — —0.02.] =
quantity.
The common terminology here is as 2.5 A Tangent to a Curve
follows: t 2 —
t x and a ( t 2 —
£ x ) are said

to be (provided that t 2 —
A t is = Using the concept of a derivative, we
small) quantities of the first order of can solve an important geometric pro-
.smallness (here and in what follows we blem: to find the tangent line to a cur-
assume that t 2 tends to t 1 and A t tends ve given by the equation y f (x). —
to zero); the quantities ( t 2 tff and — The coordinates of the point A of tan-
b (t 2 £x )—2
are said to be quantities of gency are given: x = x0 and y =
the second order of smallness ( t 2 tff ;
— Uo = f (*o)-
and c (t 2 £x )
3
—
are of the third order of From the viewpoint of analytic (co-
smallness etc. With this in mind, we
.!
;
ordinate) geometry introduced by Des-
>can say that (2.4.7a) is exact to within cartes (see Chapter 1), to find the tan-
first-order terms, while formula (2.4.7a) gent line means to find the equation of
is exact to within second-order terms the line. It is clear that the tangent
(the word “smallness” is usually line is one of the straight lines passing
dropped). through the point of tangency. But the
If we employ (2.4.7a) to derive an equation of any straight line passing
approximate expression for the deriv- through a given point A (x 0 y 0 ) can ,

ative, we will be forced to divide both be written as y y0 k (x —

x 0 ) (see = —
sides of (2.4.7a) by t 2 - f
lt which — formula (1.3.3)). In order to find the
reduces the order of smallness by one. equation of the tangent line, it remains
The equality to determine the quantity &, the slope
of the tangent line (its “steepness”).

Z ' (t ~ To do this, we first find the slope of the

)
h— 1 line passing through two given points
A and B of the curve at hand (Figure
is approximate, and the error introdu- 2.5.1). We
call this line a secant line .

ced by it is not of the second but of the When these two points of the curve ap-
first order of smallness, that is, the er- proach each other, the line approaches
ror is proportional to t 2 h; neverthe- — the tangent line.
less, as t 2 that is, as A£ t2 tx = — In Figure 2.5.1 we see two secant
->0, the error tends to zero. This lines through points A and B and
fact follows from the definition of the through points A and C, with C lying
derivative given above. closer to A than B. The closer point B
 2S A Tangent to a Curve 81

Figure 2.5.1

Figure 2.5.2

or C is to A, the closer the secant line

AB or AC is to the tangent line. There- tangent line is equal to the limit of
fore, the slope of the tangent line is k s as xx tends to x 0 :

equal to the limit approached by the

slope of the secant line as the distance be-
tween the two points of intersection of
k = lim k s = lim -
Xl
L
X()
<

xi-*xn xi -*xn
the secant and the curve tends to zero
(or the points of intersection approach We denote the difference x1 — x0
each other). This fact can be taken as by Ax; then x1 = x 0 + Ax and A/ =
the definition of the tangent line; of
/ (*l) — / (*o) =/(*<> + A *) — / (*<>)•
course, it would be more precise to say In this notation, the slopes ks and
that point B tends to point A or that k of the secant and tangent lines,
the two points B and C tend to A respectively, are given by the for-
and the secant line BC tends to the mulas
tangent line at the chosen point A.
A/ A/_
The slope k s of a secant line can easi-
Ax
k= lim
Ax *

ly be expressed in terms of the coordi- Ax-»0

nates of the intersection points. For

Thus, the slope of the tangent line is
one of the points of intersection of the
equal to the derivative of the function
secant and the curve we take the point
f (x) at point x0 :

A y 0 ) at which we desire to draw a

(x 0 ,

tangent line to the curve; we denote by k=zJ

x x and y x the coordinates of the second t= f'(Xo). (2.5.1)

point of intersection, B. Since both We know that the derivative /' {x)
points lie on the curve whose equation of a function / {x) is itself a function of
is y f [x)= (“belong to this curve” is x. Since we sought the slope of the tan-
usual phrase, it follows that y 0 = gent line at a fixed point A (x 0 i/ 0 ), ,

and
f (x 0 ) y 1 = f (^i). As can be we assumed, in computing the limit of
seen from Figure 2.5.2 the slope of the Afi Ax, that x =
x 0 is fixed. That is
secant line is why in the final formula we have /' (x 0 ),
which is the value of the derivative
k s == tan a yi — yo f( x i) — f( x o)
f(x) at x =
x 0 On the other hand, the
Xi —X 0 X1 —X P
.

function /'(#), obviously, expresses the

slope of a tangent line at a variable
(see Section 1.3). point (x, y) or (x, / (x)) of this curve.
In order to obtain the slope of the Let us consider the example of a
tangent line at the point (x 0 y 0 ), we ,
parabola y a;
2
=that is, / (x)
,
x2 = .

must take point B closer and closer to We set up the equation of the tangent
A, which means that x 1 must approach line at the point x 0 2, y 0 =
/ (x 0 ) = =
z 0 Consequently, the slope k of the
. 4.

6-0946
 x

82 2 What is a Derivative?

y = x
2
bola can be expressed as
b = —(hi 2) 2 (cf. (1.4.8a)).
Here is another example. Take the
semicubical parabola y — ]/ x3 (see
Section 1.5). Here y (3/2)]/a; (see
f
=
Exercise 2.4.3), whence k f (x 0 ) = =
(3/2 )Y 0 . In particular, at the ori-
gin the slope of the tangent line is
k =
/'( 0) (3/2) "j/0 =
0; whence, at =
point 0 the semicubical parabola touches
the x axis, as depicted in Figure
1.5.7.

Without the aid of derivatives, it is rather

difficult todraw a tangent line to a curve giv-
Figure 2.5.3
en by the equation y = f (x): you have to
compute a large number of points of the curve,
then, using a French curve, draw the curve
We know the derivative of x 2 : through these points and, by eye, apply a ruler
to the curve at the given point and pay special
dx*
dx
— 2x attention to see that you do not intersect the
curve near the point of tangency. Using deriv-
atives, we find the equation of the tangent
(cf. (2.3.6a)). Consequently, at the line, then from this equation we find two
point of interest the slope of the tan- points lying on the straight line given by this
gent line is equation, and then we draw the straight line
(tangent line) with a ruler (through the two
k — / (^Tq) points). For one of the two points it is natu-
ral to take the point of tangency itself,
while the equation of the tangent line A (# 0 y 0 ). The second point C may be taken
,

is (Figure 2.5.3) on the straight line at a good distance from A

(the farther the better); we can then more
y —y = 0 k (x —x 0 ), i.e. accurately determine the slope and the posi-
tion of the tangent as the straight line passing
y —4= 4 (x — 2), through the two points A and C.
For example, above we found the equation
or of a straight line tangent to the parabola

= — y = x 2 at the point x 0 =
2, y 0 =
4. It has the
y 4.
form y = —
Ax 4. Let us find the coordinates
bi"rvvo~ pmlris" uu hint* 'rurer a x —° ^"wdmfiu
1

* i
In general, the slope of the tangent
that y = 4. This the point of tangency
is
line to the parabola y — x 2 at point A (2, 4); the coordinates need not have been
(x 0 ,
a:®) is equal to 2x 0 (since here computed since the tangent must pass through
dyldx 2a;);=hence, the equation of it. For the second point, C, we choose the
point of intersection of the tangent line and
the tangent line is the y axis. Putting x 0, we find y =
4, so
= —
y — x\ = 2x (x —x or y = 2x x — a%.
that C =
C (0, —4) (see Figure 2.5.3).
0 0), 0
Note the curious fact that for x 0, y — —
(2.5.2) — y o the point C of intersection of the tangent
line with the y axis lies below the x axis fust as
We have thus arrived at the result estab- much as the point of tangency itself lies above
lished in Section 1.4 by a different the x axis. This is not accidental. The rule
holds true for all tangents to quadratic para-
method, namely, that the straight line
y = kx b touches the parabola y =
+ bolas with an equation y ax 2 (with a 0). — >
Indeed, if the tangent is drawn to the point
x2 if and only if the coefficients k A(xq, y 0 =
ax%), its equation is
and b in the equation of the straight
— y —y 0 = 2 ax 0 (x —x 0) (2.5.2a)
line can be represented as 2x 0 and x\,
which depend on the parameter x 0 (cf. (2.5.2)), and for x =
0 we get y — y 0 ==
in other words, the condition for a
;
—2 ax\,or y y0 = —
—y 0 2 ax% y0 =
2 y0 — = .

Thus, the tangent passes through the points

straight line y kx = + b touching a para- A x0 y 0 axl) and C (0, y
,
= y0 —axl). = — =
 2.5 A Tangent to a Curxe 83

Whenplotting a curve by points it has been moved upward and the result
is hard to construct the curve if there is depicted by the dashed curve) by an
are few points. Using derivatives, you arbitrary segment 6, the curve y'(x)
can draw the tangents to the curve at does not change in any way because
these points beforehand, and then the in translation in the vertical direction
curve itself can be drawn with greater all the slopes remain the same (to be
ease and accuracy. precise, we must speak of the slopes of
Pictorially, clear that the tan-
it is the tangent lines to this curve at all
gent is horizontal
at the points of points: compare the upper and lower
maximum and minimum ; in future curves in Figure 2.5.4a and, in partic-
we will return to this aspect many ular, the tangents to these curves at
times. The equation of a horizontal stra- points A and B ). This result is in ac-
ight line is y =
constant, and the slope cord with the property of derivatives:
of the horizontal straight line is k 0. = the graphs of the functions y =
f (x)
Hence, we arrive at the following and y 1 =
/ (x) +
b (the graphs of these
theorem: the derivative of a function functions are obtained through trans-
y =
/ (x) ( the graph of which is a curve) lation upward or downward) have equ-
is zero at the points of minimum and al derivatives at corresponding points.
maximum of the curve . Another mathematical game is this:
Using this theorem, one can find draw freehand the graph of the deriva-
the ^-coordinates of the points of max- tive and then give a rough construction
imum and minimum of the curve. The of the graph of the function. Here you
respective {/-coordinates can then be have to specify in arbitrary fashion one
easily found by substituting the estab- point (x 0 y (^ 0 )) on the curve and then
,

lished values of x into the equation draw the graph up or down (depending
of the curve. It is also obvious that know- on the sign of the derivative).
ing the coordinates of the points of In conclusion we must dwell on one
maximum and minimum we can draw important point closely related, as we
the curve more accurately.
itself will see below, with the question dis-
It is a useful exercise to draw a curve cussed concerning the connection be-
y (x) freehand and then, at least ap- tween the slope of a tangent line to a
proximately but rapidly, draw the curve curve and the derivative (above we ig-
(
7/ \ju) ,
'iiuVii/g kli/b ‘oYgi-L 'vft ’’(j
“oujuL nan%4x \ki4b ’ptm/u ^purpub^f IWifc
.

the points where y'(x) vanishes. This in physical problems, the quantities
is illustrated in Figure 2.5.4a (the x and f x) are usually dimensional
(,

graph of y (x)) and Figure 2.5.46 (the (for example, x or t is time and y =
graph of the derivative y'{x)). / {x) or z =
f ( t) is distance). But
For the derivative y' (x), the points this means that dyldx is dimensional,
at which y (x) vanishes are of no inter- too. Clearly, when z is measured in cen-
est. If the curve y (x) is moved upward timeters and t in seconds, dzldt = v (£)
or downward (in Figure 2.5.4a the curve is the velocity, with dimensions cm/s.
If, say, y is expressed in kilograms and
x in months (say, y = f (x) represents
the weight of a baby, or its mass, as a
function of time), the derivative
dyldx ( = lim Ay/ Ax) expresses the
rate of weight gain by the baby and has
the dimensions kg/month. If y is the
current in a conductor and x is the con-
ductor’s resistance, then the derivative
y' =dyldx is measured in A/£2. If x
isthe edge of a cube and y its volume,
Figure 2.5.4 then dyldx has the dimensions of
 84 2 What is a Derivative?

cm 3 /cm = cm Examples of this type

2
. l cm in the drawing, we get
abound. In general, if the independent
variable x is measured in units of e 1
and the function y = f (x) in units
of e 2 then the derivative dyldx has the
,
If l = 5, we have tan a = (1/5) dyldx =
dimensions of eje (1/5) y.
But, as we know, the trigonometric In the general case, if one x unit in
function tana is dimensionless (being the drawing is laid off to a scale of l x
equal to the ratio of the lengths of two cm and one y unit is laid off to a scale
line segments). Therefore we cannot of Z2 cm, then
simply write dyldx — tana since the
left and right members have different tan a = -£-
11
-t*-.
dx
(2.5.3)
v

dimensions. Only in the rare instances

when the two quantities, x and y have,
The Z x and l 2 in this for-
scale factors
the same dimensions (say, when we are mula improve the situation they —
studying the speed of a yacht as a func- make the formula proper from the stand-
tion of the speed of wind) or both are point of dimensions. Thus, in the exam-
dimensionless (say, the function y = ple with a baby’s weight, Z x has the di-
sin x where x is measured in radi-
,
mensions of cm/month (1 cm in the dra-
ans) does the relationship dyldx = wing per month of age) and Z 2 has the
tana have a meaning. dimensions of cm/kg (1 cm on the graph
How can we get round this difficul- per kilogram of weight), so that
ty? Note that above we assumed the (Z 2 /Z x ) dyldx is dimensionless: in the
scales along the x and y axes to be the formula all the dimensions cancel out
same, that is, that the unit of measure- and the formula becomes meaningful
ment for x and the unit of measurement (and correct).
for y are depicted by the same segments; All this should be borne in mind when
this assumption seemed so natural that comparing the derivative and the slope
we did’nt even spell it out. But in the of a curve representing the basic func-
case where x and y have different di- tion (i.e. the slope of the tangent line
mensions (and this is the general case) to the graph at the point of interest).
the above assumption is not only unnat-
ural but really has no meaning, since
there is no way in which we can assume Exercises
that 1 s = 1 cm. Therefore, in reality 2.5.1. Construct the graph of the function
the scales along the two axes are differ- y — x2 +
1 within the range from x l.n = —
ent, and because of this we cannot write to x =
2.5 and draw the tangent lines at the
dyldx = tana. points x =
—1, x 0, x = 1, and =
x 2. =
2.5.2. Construct the graph of the function
Suppose that y is the distance trav- y = x s 2
3x ,— 1 — <
x <z 3.5; draw the tan-
eled andx the time.
is We
construct the gent lines at x = —
1, 0, 3. Find the points
graph of the position of a body depen- with horizontal tangent lines.
= 2.5.3. On the curve y x3 x = — +
1 find
ding on the time, y y (x). On the axis points having horizontal tangents. Construct
of ordinates we lay off y using the scale: the curve for x —2 < <
2. [Hint. In Exer-
1 meter of distance equals 1 cm in cises 2.5.1 to 2.5.3 it .is advisable to use graph
the drawing. On the axis of abscissas we paper and a large scale.]
lay off time using the scale: 1 second of
time equals 1 cm in the drawing. Then
the velocity v expressed in meters per
second and equal to the derivative
dyldx will indeed be equal to tan a,
the tangent of the angle formed by the
tangent line and the x axis. But if we
choose a different scale for x say, 1 s =
 2.6 Increase and Decrease of Functions. Maxima and Minima 85

(point D ). But how must we define such

concepts with sufficient mathematical
rigor and how to determine, without
looking at the graph, the behavior of a
function at, say, point x — x 0 (drawing
a graph always involves errors and
therefore constitutes a nonreliable
method)?
Without a knowledge of derivatives,
2.5.4. Construct (freehand) the curve y' (x) we have to seek the answer to these
for the function y (x) depicted in Figure 2.5.5.
questions numerically, that is, we must
2.5.5. The graph of a derivative y'(x) is
depicted in Figure 2.5.6. Construct (freehand) take the temperature T at a given time
the graph of the function y (x) passing through t and then take it at some following
the point (5, 0). At what angle will y (x) in-
2.5.5, time and see if it has increased or
tersect the axis of ordinates? At what angle
decreased. This is clearly not a reliab-
will y (x) intersect the axis of abscissas at
= le approach: even if T (£ x ) is greater
point x 5? [Hint. In Exercises 2.5.4 and
first copy Figures 2.5.5 and 2.5.6 on a than T (£), it still might be that at time
fresh sheet of paper and then construct the t the temperature fell, then soon af-
respective graphs; it is advisable to construct
terward (after t but prior to t x ) it
these graphs on the same sheet of paper with
the initial graphs, say, strictly below the re- reached a minimum, and only then be-
spective graph.] gan to grow and by t± had risen above
2.5.6. Set up the equations of the tangent T re-
lines to the cubical parabola y =
x 3 at the
using derivatives we get an exact so-
points with (a) x — 0.5 and (b) x =
1. Find
the points of intersection of the tangent lines lution: we have to find the derivative
with the x and y axes. dyidx. If dyidx = y' {x) is positive for
2.5.7. Find the general rule that enables a given x, then y (x) is an increasing
determining the points of intersection with the function, that is, if x increases by a
x and y axes of the straight lines tangent to the
curves (a) y =
ax 2 and (b) y = bx 3 at point small quantity Ax, the value of y
(*o» V (*o))- increases by a small amount Ay ~
y' {x) Ax (as was clarified earlier,
the smaller the value of Ax the more
2.6 Increase and Decrease of Functions.
Maxima and Minima exact the equation). We consider
Ax > 0 (say, time x increases). If
Observing a curve that depicts the be- >0, Ax >>
y'(x) 0, then, of course,
haviour of a function y = / (x) (say, Ay > that0, the is, value of y in-
the dependence of temperature T (t) creases with x (say, the temperature in-
on time t ), one can easily see the points creases in time). The numerical value of
at which the function grows or increa-
,
the (positive) derivative shows how
ses (say, point A in Figure 2.6.1), the fast y (the temperature) rises: in the
points at which the function decreases example with the temperature we find
(say, point B), the points of maximum, that if T'(t) —
10, then in the vicinity
where growth of the function is re- of t the temperature increases 10 times
placed with its decrease (point C ), and faster than time (for example, by 10 °C
points of minimum where decrease of
,
each second), while if T'(t) 0.1, =
the function is replaced with its growth then the temperature increases 10 times
slower than time (however, com-
pare this with what was said in Section
2.5 about the dimensions of the deriv-
ative and its connection with the choice
of units for the independent variable
and the function). But if y'{x) 0, <
Ax >0, then Ay <
0; for instance, if
Figure 2.6.1 T'(t) <0, the temperature T (t +
86 2 What is a Derivative?

-f- A t) at t +
At will be lower than T ( t ) for further serious applications of
at the given moment in time. Thus, mathematics.
a positive derivative indicates that the There are functions that have the
function is increasing and a negative der- same sign of the derivative for any val-
ivative that the function is decreasing. ues of the variable: such is the property
The expressions “increasing function” of the linear function y kx 6, = +
and “decreasing function” are applied since here the derivative dyldx k =
to any functions y x ) and not only to(. is a constant. Later on we will see that
those that depend on time (functions in the case of the exponential function
of time); here the independent variable y = a x the derivative has a constant
may be an arbitrary quantity (with or sign (although it is not constant in
without dimensions). An increasing ( de- magnitude) for arbitrary x However, .

creasing) function is one in which y

i a derivative need not have a constant
increases as the independent variable sign; the sign of the derivative of a
x increases (decreases). The derivative given function may be different for
dyldx is what indicates the rate of different values of the independent
growth that is, the ratio
,
of the varia- variable.
tion in y to the corresponding (small) Let us imagine a function y (x) whose
variation in x. A
negative rate of growth derivative y'{x) is positive for x x0 <
means a decrease in y as x increases,
—
and negative for x x0 in short, > ;

and if dyldx < 0, then |

dyldx |
y'(x)> 0, x <x 0 ,
and y’(x) < 0,
— dyldx the rate at which the func-
is x >x 0 .

tion decreases. What can we say about such a func-

The expression “the quantity y has tion? We begin with x x 0 As x < .

a large negative derivative with respect increases to x 0 the function y (x) will
,

to x ” means that y decreases rapidly as increase; however, as x continues to

x increases, while a positive dyldx increase the derivative becomes nega-
means that y increases with x. In this way tive and y falls. The conclusion is that
the derivative of y points to the ten- for x = x 0 the function y (x) has a
dencies in the variation in y and enables maximum.
us to know what to expect from further Consider the contrary case: y' (x)<0
variations in the independent variable. for x x0 and < y'(x) 0 for >
This constitutes the main reason for x >x 0 Reasoning as before, we con-
.

studying the derivative of a function. clude that in this case y {x) has a
Physicists and mathematicians, espe- minimum at x x0 = .

cially those in the making (who have If a function y (x) is defined by a for-
just learned what a derivative is), fre- mula associated with a smooth curve, so
quently put it to use in everyday life that y'{x) also varies smoothly with x,
like this: “the derivative of my mood then the different signs of y'{x) for
with respect to time is positive” in x <x 0 and x >
x 0 in both cases signify
=
place of “my mood is definitely im- that the derivative vanishes at x
proving.” x0 :

Solve this joke problem: what sign

does the derivative of my mood have V’ (*0 ) = -^ sL = 0. (2.6.1)
with respect to the distance from the
dentist’s chair? My mood deteriorates, Thus, as already noted in Section
“decreases,” becomes “negative” as the 2.5, by equating the derivative to zero we
distance decreases; hence, the deriva- can find those values of the independent
tive positive.
is variable for which the function has a max-
serious editor may complain
The imum or a minimum 21 Now we can re- .

about abuse of the English language, s' 7

More precisely, the values at which
but actually this free-style use of the functionmay have (but also may not have)
mathematical concepts is good practice a maximum or a minimum. For example, the
 y

2.6 Increase and Decrease of Functions. Maxima and Minima 87

fine the general theorem discussed in

Section 2.5: at the points where a func-
tion attains maxima the derivative
changes its sign from plus to minus ,

while at the points where a function at-

tains minima the derivative changes its
sign from minus to plus .

Here are some examples. First we

turn to the function y 3x 3 x2 x = — —
discussed in Section 1.1 (see the table
for this function in Section 1.1). Judg-
ing by this table, one might think that
the function increases for all values
of x since every increase in x by uni- Figure 2.6.2
ty caused an increase in y.
Let us calculate the derivative, and y ( —0.18) confirms the fact that
however: when x ~ — 0.24, y reaches a (local) max-
imum, the adjacent values of y being
y'(x) = 9x 2 — 2x — 1.
smaller. The graph of the function y =
Taking x = 0, we find that y'( 0) = 3x3 — x2 —x is given in Figure
— 1 < 0, which means that when
2 6 2
. . .

This example shows once more that

x =
0, the function is decreasing. This
the word “maximum” should not be un-
refutes the supposition (obtained from
derstood as meaning the largest of all
a glance at the table) that the function
possible values of y. Indeed, at the
is an everywhere increasing function.
We
equate y'(x) to zero. Solving the
maximum point y (—0.24) 0.14, ~
equation y'(x) = 9x 2 — 2x — 1=0 while y 1 for x= 1, y =
+18 for =
yields two roots: x1 ~ —0.24 and x =2, y =
269 for x 10, and so on, —
x2 ~
0.46. y rapidly increasing without bound as x
increases without bound. In what way
Now we form a detailed table inclu-
^—
does the maximum point x max 0.24,
ding the maximum and minimum
points just found: i/max 0.14— that we found differ
from all other points of the curve?
The difference is that for close-lying
x —3 —2
—1 —0.3 —0.24 —0.18 values of x both larger than x mSLX and
, ,

y —87 —26 —3 0.129 0.140 0.131 less than x max the value of y is less
than y mRX {=y (x m&x )) This peculiarity of
.

Xmax is clearly seen in the table (e.g.

x 0 0.40 0.46 0.52 1 2 —
compare y ( 0.30), y (—0.24), and
y 0 —0.372 —0.381 —0.370 1 18 y (—0.18)). The same arguments can be
applied to the minimum x mln 0.46, ~
We see that, true enough, on the por- Umm —
—-0.381: for large (in absolute
tion from x —0.24 to x = 0.46 the value) negative x’s the value of y de-
function y falls from +0.14 to —0.38. creases without bound and becomes less
A comparison of the value y (—0.24) than y mn (and, in general, y can be
with the adjacent values y (—0.30) made than any negative number),
less
but x min value z/ min
differs in that the
function y =
x z has a derivative, y' 3a;
2 = (= ( (^mm)) is less than the values of y
which vanishes at x =
0; however, at this
,

for x close to x mn The condition of a

.

point the function (see its graph in Figure vanishing derivative enables finding
1.5.1a) has neither a maximum nor a min-
imum. (Other exceptions to our rule that just such {local) maxima and minima.
refer to curves that are not smooth are dis- For the second example we take the
cussed in Section 7.2.) function y x3 = —
x discussed in Sec-
 x 3 “

88 2 What is a Derivative?

tion 1.5. The derivative of this func-

tion is y' = 3x — 1, whereby y'(x) = 0
2

at x — ± l/)/"3. Let us see how

the sign of the function y'(x) — 3x — 1 2

varies in the vicinity of points

x = 1/]/ 3 ~
0.577 and x — — 1/]/ 3 ~
— 0.577. We will readily find that
at x — — l/y
the function y (^)hasa
(local) maximum and at x=lI]/3 the
function y )
has a (local) minimum
(,

(see Figure 1.5.3).

Now we can give a complete explana-
tion why the graphs of the functions
y =
x3 +
x and y x3 x (Figures = —
1.5.2 and 1.5.3) differ. Let us consider
Figure 2.6.3
the general equation

y = x3 + cx , (2.6.2) which is clearly impossible, and it is

where c is an arbitrary number. In clearly unjustified to expect that the

this case y' = 3x2 4- c. It is clear that point is a maximum (and not a mini-
for c >
0 the derivative y is positive mum) or a minimum (and not a maxi-
mum). All these peculiarities of the
everywhere, which means that y in-
creases for all values of a;, that is, there curves specified by Eq. (2.6.2) are
are neither maxima nor minima. On shown in Figure 2.6.3. see that there We
the other hand, for c 0 the deriva- < are two distinct cases: c 0 and <
tive y' vanishes at x = ±Y — c/3 = c >0; the “intermediate” case corre-

db x 0 and in the interval from sponds to the “limit” curve y #3 , =

x
,

= — oo to
0 x — —x
the derivative
for which c = 0.
y' is positive (the function is increas-
Let us return to Exercise 2.2.1 devoted to
ing), for x0 x— < <
x 0 the derivative the specific heat capacity of diamond. There
is negative (the function is decreas- we saw that there exists an empirical relation-
ing), and for x 0 <
x <C oo the deriva- ship between the temperature T and the quan-
tity of heat Q =
Q (T) (in joules) required to
tive is again positive (the function is
heat 1 kg of diamond from 0 to T °C:
again increasing); here the symbols
— oo and oo denote, as usual, very large Q (T) = 0.3965 J + 2.081 X 10~ 3 r 2

negative and positive numbers. Thus, — 5.024 X io- 7 r 3 ,

at x = —
x 0 the function has a maxi- which implies that the specific heat capacity
mum and at x = x0 it has a minimum. of diamond, c — c (T) (in J/kg*°C), is ex-
If we make c smaller and smaller (in pressed by the formula

absolute value), the maximum and min- c (T) = 0.3965 + 4.162 X 10" 3 r
imum approach each other: at c 0 = — 15.072 X io- r 7 2

the points dzx 0 ^/3 merge = ± V— (see the solution to Exercise 2.2.1). To analyze
into a single “critical” point x 0, = the behavior of the specific heat capacity of
the point where the derivative of y diamond in the temperature interval in which
the empirical formula is valid, we need the
vanishes. In this case (for the curve
expression for the derivative of c
=
:

y x3 ) the critical point is neither a

maximum nor a minimum, since if we c'(T) = 4.162 X 10- 3 — 30.144 X 10 7
T.

consider “neighboring” curves cor- It is clear that c' (T) is positive for T <
responding to negative values of c (4.162/30.144) X 10 4 T0 1380 °G,= is ^
that are small in absolute value, we
zero at T =
T0l and is negative for T T0 > .

But, as noted earlier, the above empirical

might decide that our point is simul- formula is valid only up to 800 °C. All the
taneously a maximum and a minimum, above formulas imply that c' (T) vanishes at
 ) T )

2.7 The Second Derivative of a Function . Convexity and Concavity of a Curve 89

T ~1380 °C, that c ( T vanishes at T ~ maximum or a minimum) x x 0 of the=

2850 °C, and that the quantity of heat Q ( )
vanishes at T ~
4320 °C. Of course, all these
curve y — y (x) representing the func-
results are meaningless and prove that empiri- tion, or points at which
cal formulas cannot be employed outside their
range of application. y'(*o) = 0. (2.7.1)

Determining maxima and minima How to distinguish a maximum from a

arithmetically (by computing and com-
minimum condition (2.7.1) is sati-
if
paring the values of the function for
sfied? We know that condition (2.7.1)
different values of the independent var-
holds true both for (local) maxima and
iable) is many times more arduous and
(local) minima, the difference being in
less exact. However, if you employ a
pocket calculator, then an “arithmetic
the sign of y'(x) for x <x 0 and for
experiment” enables finding a maximum
x >x 0 .

But how is it possible to determine

or a minimum by meaningfully compar- the sign of y'(x) for x close to x 0 with-
ing the values of the function at differ-
out computing y' directly for other val-
ent points (a calculator must be used,
ues of x? Let us start with the case of a
of course, in a proper manner), and our
maximum of the function y {x), when
problem is considerably simplified. But
even in this case the use of derivatives
y'(x) >0 for x <x 0 and y'(x) <0
makes the calculations more transparent
for x >x 0 . We
see that here the deri-
vative y'{x) is itself a decreasing func-
and simple. Higher mathematics is not
tion: as x increases, the derivative,
only a remarkable achievement of the
mind. Practical computational prob-
which was at first positive (for x x 0 ), <
lems are resolved much more easily by
vanishes (when x x 0 ) and, continuing =
the methods of higher mathematics.
to fall, becomes negative when x >
x 0 But we already know how to dis-
.

Exercises tinguish a decreasing function from an

2.6.1. Find the values of x for which the increasing function: its derivative is
following functions have a maximum or a negative. Hence, at a value x = x a
minimum. In each case determine whether the at which y has a maximum the deriva-
minimum or maximum is involved. For func- tive y' (x 0 ) vanishes and the derivative
tions involving constants give the answer for
various values of these constants (in particu- of the derivative is negative This quan- .

lar, with different signs): (a) y = ax 2 (b) y = tity, derivative of a derivative*

the
x -f- 1 lx, (c) y = ax + b/x, =
(d)
,

y x3 — which by the ordinary rules can be writ-

3x + 100, (e) y= xs +— + +
px 2 qx r,
ten as a “double-decker” fraction,
(f) y = x* + ax 2 + 6, and (g) y + ax 2 b/x 2 .

2.6.2. The temperature dependence of the

volume of 1 g (or cm 3 ) of water is given by the
d
following empirical formula: v (t) = 1 + dy’ (lte)
8.38 X 10" 6 ( t — 4)
2
. At what temperature dx dx ’

will the volume be minimal?

2.6.3. Solve Exercise 2.6.2 with the follow-
ing refinements (suggested by various scien- is called the second derivative It ha& .

of the above empirical formula: 2 2

tists) the notation y"{x) or d y!dx and is
(a) v (t) = 1 — 61.045 X 10" 6 * + = —
77.183 X
read “d two y over d x squared.”
10-7*2 _ 37 34 x io-9*3 an(j (b) v (t) 1
=
57.577 X 10- 6 * +
?

75.601 X 10" * 7 2 — 35.07 X It is clear that if z z ( t is a func-

X10“ 9 * 3 . tion that specifies the dependence of
distance on time (see Section 2.1),
2.7 The Second Derivative then z (t)
f
=
v is the velocity of the mo-
of a Function. Convexity and tion and z"(t) =
dvldt is the rate of
Concavity of a Curve. change of velocity, or the acceleration
Points of Inflection (see also Chapter 9). If y has the dimen-
Let us once more study the behavior of sions of distance (cm) and x is time
2
a function near its “critical” points (s), d^yldx has the dimensions of accel-
(i.e. points at which it might have a eration (cm/s 2 ); if y is measured in units.
 ) )

90 2 What is a Derivative?

of e 1 and x in units of e 2 the derivative ,

To summarize, then, if at a certain val-
<l y!dx 2 has the dimensions of ex /e\, as
2
ue of x the function y (x) is such that
indicated by the notation d 2 y/(dx) 2 .

It is advisable to bear in mind the di-

y'(x) =0, y"(x)< 0, (2.7.2a)

mensions of the second derivative when then at this point the function has a
dealing with physical problems that in-
maximum and if at a certain value of
,
volve y"{x).
x the function y (x) is such that
It is the meaning of the second deriva-
tive, acceleration, in our distance-time y\x) = 0, y”(x)> 0, (2.7.2b)
problem that makes the idea of a sec-
then at this point the function has
ond derivative so important to physics,
a minimum. (The reader is advised
.since by Newton’s second law the accel-
to return to the exercises accompanying
eration is the basic characteristic of
Section 2.6 and apply the second-deri-
motion. In Chapter 9 we discuss all
vative criterion, which enables distin-
these matters in greater detail.
guishing between a maximum and a
Let us revert to the examples given
above. If y (x) 3x3 x2 =
x, then — — minimum.)
y'{x) =
9a;
2
2x 1— and, hence,— We
see that the sign of the derivative
of a function y (,x has a simple geo-
y”(x) = (y'(x))' = 18* - 2. metrical meaning: the fact that the de-
rivative is positive means that the func-
At x ~ — 0.24 we have y' = 0 and tion is increasing that is, its graph is
,

y" ~ — and, indeed, point

15 ."3 Hirodreh upward ’fi" we nrove droirg iife ~x
J

x ~ —0.24, y ~ 0.14 a maximum of is axis from left to right, while if y'(x) <0,

y {x). At x ~ 0.46 we have y' = 0 and

the function is decreasing that is, ,

y" ~ 6.3 > 0, that point x ~ 0.46, is, its graph is directed downward. The

y ~ — 0.38 is a minimum of the func-

sign of the second derivative y"{x)
tion. also has a simple geometrical meaning.

y = x — x
Similarly, then
if
3 If the second derivative of a function is
y' = 3x — and y" = 6*. Therefore,
2
1
,

positive, or y"{x) >

0, this means that

at x = — 1/V^ 3 ~ — 0.577 the second

the first derivative y'(x) is increasing
= or, in other words, the angle a (where
derivative y" is negative, and at x
tan a k= =
y'(x)) formed by the tan-
+ 1//3 positive:
it at the first
is
gent line to the curve and the x axis
point the function has a maximum and
increases. But in this case, as we move
at the second a minimum. (This result
from left to right along the x axis, the
could also be predicted geometrically,
tangent line rotates counterclockwise
.since at x 1 = —
and at x 0 the = (point A in Figure 2.7.1). And this, ob-
function y = x —
x (x 3
x ) — — — 3
viously, means that our curve is convex
vanishes and at the intermediate points downward (cf. Section 1.4); sometimes it
it is positive, so that at x 0.577 ~ — is said that the curve is concave upward.
there can be only a maximum and not a
minimum.) In exactly the same way,
in the example with the specific heat
capacity of diamond (see Section 2.6)
the second derivative c "( T) ca 3.0144 X —
10 -6 is always negative, which would
mean that the curve representing
c (T has a maximum if our formulas
could be applied at T 0 where c'{T 0 ) =0
{actually the specific heat capaci-
ty of diamond, just as that of other sol-
ids, increases with T and tend to a
-constant value at high temperatures).
2.7 The Second Derivative of a Function. Convexity and Concavity of a Curve 91

(point B in Figure 2.7.1) are known as

points of inflection 2 9 .

For instance, when the graph of a func-

tion y =
f (x) is intersected by a hori-
zontal tangent at point x x 0 (Figure =
2.7.2), this point is neither a maximum
nor a minimum of the function, while
the derivative at the point is zero (the
tangent line is horizontal). Thus, we see
Figure 2.7.2
that the cases where condition (2.7.1)
does not work, that is, the point where
the derivative vanishes is neither a max-
Similarly, the fact that y"{x)<zO
means that the first derivative y'(x) of imum nor a minimum, are due to the
fact that in addition to (2.7.1) the fol-
the function y f(x ) =
decreases, and
lowing condition holds true at this
so does the angle a. Thus, at the points
where y"(x) <
0, the tangent line ro-
point:
=
tates clockwise as we move from left to y" (x 0 ) 0, (2.7.3)
right along the x axis, that is, the curve
is covvex upward or concave downward
which characterizes a point of inflec-
,
tion. (On the other hand, compare what
{point C in Figure 2.7.1). The points at
which y"{x) =
0 can be generally char-
we have just said with the footnote
2.8, where the function yv = x 4 satisfies
cei Txdu. v&'thwcv* seceriAi
both conditions, (2.7.1) and (2.7.3),
derivative changes its sign, that is,
at point x = 0 and nevertheless has a
points at which concavity is replaced
with convexity or the other way round. 2 8 -
maximum at this point.)
In Section 7.4 we will once more en-
At such points the tangent to the
counter the concept of convexity of a
graph of the function goes from one side
function.
of the curve to the other, that is, the
tangent intersects the curve. Points at Exercises
which the tangent intersects the curve
2.7.1. Find the second derivatives of the
following functions: y x2 y x3 y =x4 — —
2 - 8
The
rare instances of the type of the and y ax 2 bx — c. + +
, , ,

origin for the curve y =

x 4 (see Figure 1.5.16; 2.7.2. Find the acceleration of a point mo-
here y' 4a: 3 =
y" 4 X 3a; 2 = =
12a: 2 so that , ving along a straight line according to the
y" =0 at x
,

=
0, while the curve does not following law: z at 2 bt = +
c. What can be +
change the direction of convexity at this point) said about the acceleration?
refer exclusively to the points at which the 2.7.3. Specify the ranges of convexity and
first, second, and third derivatives vanish concavity for the following functions: (a) y =
simultaneously (or the second derivative x s (b) y
, x3 px 2 = qx +
c, and (c) y+ + =
has a salient point or a discontinuity). (The
third derivative is the derivative of the second
x + i/x.

derivative and is written as y"' or d?y/dx 3 2 - 9

Note that at a point
of inflection the
and has the dimensions of eje^ where e 1 and tangent the curve smoothly,
intersects
e 2 are the units of measurement of y and x. touching it without a break, contrary to the
Below we will encounter third and higher-order case where the y axis intersects the parabola
derivatives.) y =*2 .
 t) t )

Chapter 3 What is an Integral?

3.1 Determining Distance z (6), and that at the beginning of the

3 1
from the Rate of Motion. interval, z (a). *

The Area Bounded by a Curve Now let us compute z (a, b). In the
most elementary case, when the velocity
The problem determining the instan-
of is constant, or
taneous rate of motion, or velocity, v t
( )
— constant = y0 ,
(3.1.2)
v t from a given dependence of the
(i

position of a body upon time, the the distance covered will obviously
z = z ( t ), led us to the concept of the be equal simply to the product of the
derivative: time of motion into the velocity:
2 (a, b) = (b — a) u 0 . (3.1.3)

Taking advantage of the graph of ve-

locity versus time, we find that a con-
The inverse problem consists in deter- stant velocity is associated with a hor-
mining the position of a body, z — izontal straight line (Figure 3.1.1).
z ( ), that is, the distance covered by The distance covered is clearly equal
a body in a given interval of time £, to the hatched area because the area of
when we know the instantaneous veloci- a rectangle is equal to the product of the

ty as a function of the time, v v ( ). = base (b —

a) by the altitude (v 0 ). What

This problem brings us to the second do we do in the general case where the
most important concept of higher instantaneous velocity is not a constant
mathematics, that of the integral. quantity?
Let us agree on some convenient no- Let us make a detailed study of a nu-
tation. Weconsider the distance trav- merical example. Suppose the velocity
eled during time from t x to t 2 So as to .
is given by the formula 3 2 v t
2
We *
= .

avoid subscripts, let us denote the be- seek the distance covered during the
ginning of the time interval by a time interval from t a 1 to t = = =
and the end of the interval by 6, so b = 2.

that t x =
a and t 2 =
b. We denote the
We
partition the entire interval from
distance covered from time a to time a to b into ten subintervals and set up
b by z (a, b ). Remember that when the the table of velocity:
two quantities, a and 6, stand under
the function symbol z in parentheses, t 1.0 1.1 1.2 1.3 1.4 1.5
V 1.0 1.21 1.44 1.69 1.96 2.25
then z (a, b) is the distance covered dur-
ing the interval of time from t a =
to t =6, whereas z ( t ) with the single
t 1.6 1.7 1.8 1.9 2.0
quantity t in parentheses is the posi- V 2.56 2.89 3.24 3.61 4.0
tion (coordinate) of the body at a speci-
fied time t. These quantities are related
in a simple manner: Why is it difficult to compute the
distance at a velocity v (
t given by a
z (a, b) = z (b) — z (a), or 3 1
For the sake of simplicity we consider
-

z(b) = z (a) + z (a, b) (3.1.1) only motion in one direction in such pictorial
considerations, although possibly, the veloci-
ty varies in time.
(here we assume that b > a). We see 3 2
If the velocity v is expressed in units of
-

that the distance covered during time cm/s and time t in units of s, then, in order
from a to b is equal to the difference to maintain the requirements of dimensionali-

between the coordinate at the end of

ty, we write v =
kt 2 where the factor k has ,

the dimensions of cm/s 3 Here we assume that .

the time interval under consideration, k = 1 cm/s 3

.
 ,

3.1 Determining Distance from the Rate of Motion 93

zr slightly less than the actual value

and so the result has a deficit, too.
mm Let us now compute the distance
somewhat differently, namely, in each
lfl§l subinterval At we will take the value of
~0 *£
L 2 5'

velocity at the end of the subinterval.

Figure 3.1.1 For the first subinterval At from 1 to
1.1 this velocity is equal to 1.21 cm/s,
for the last one from 1.9 to 2.0 s it is
formula? Clearly because the velocity 4 cm/s. We
then get the following esti-
is variable (for a constant velocity, the mate for the entire distance traveled:
answer is trivial). In the case at hand,
the velocity changes four times over z (1, 2) ~ 0.121 + 0.144
the time interval from t 1 to t 2. = = = 2.485 cm.
However, interval is parti-
after this + . . . + 0.400 (3.1.5)

tioned into ten parts (subintervals),

the velocity varies less over each subin- This calculation clearly yields the
terval of duration 0.1 second (only 10 distance z (1, 2) with an excess.
to 20 percent). Therefore, in the subin- Hence, the true value lies between
tervals the velocity can roughly be tak- 2.185 and 2.485 cm. The difference
en to be constant, and we can compute between these numbers amounts to about
the distance covered during such a sub- 15 % Rounding off the boundary val-
.

interval of time as the product of ues for z yields 2.18 z (1,2) <
2.49. <
that subinterval by the velocity. In These computations can be illustrat-
what follows, we denote the subinter- ed by means of a graph. We
construct
vals into which the interval from t a = the graph (Figure 3.1.2) with time laid
to t =divided by A t (the time
b is off on the axis of abscissas and velocity

increment); each such subinterval

for on the axis of ordinates. In the figure
At we will assume the distance to be we divide the time interval into five in-
simply proportional to A t, with the tervals instead of ten (as we did in the
velocity assumed constant over At. table), which makes each part (step)
To compute the distance covered dur- stand out more clearly. Each term in
ing each subinterval At, equal to 0.1 s, the first sum (3.1.4) is the area of a nar-
we utilize the initial velocity in the giv- row rectangle with the corresponding
en subinterval: 1 cm/s in At from 1 to subinterval At as the base and the ve-
1.1 1.21 cm/s in At from 1.1 to 1.2 s,
s, locity at the beginning of the subinter-
and so on; finally, 3.61 cm/s in the last val as the altitude. Thus, the sum is
subinterval At extending from 1.9 to the area under the polygonal (step-like)
2.0 s. The total distance covered during line (the hatched area in Figure 3.1. 2a).
the time interval from t 1 to t 2 = —
is then computed to be

z (1, 2) ~ 0.1 + 0.121 + 0.144

+ . . . + 0.361 - 2.185 cm.

(3.1.4)

It is quite clear that we have reduced

the actual distance covered because the
velocity here increases with time and
so the velocity at the beginning of each
subinterval is less than the average ve-
locity. Each of the ten terms into which
the entire distance was partitioned is
 94 3 What is an Integral?

Figure 3.1.3

The second sum (3.1.5), in which the velocities in each At, we compute the
velocity in each subinterval is taken at distance to be
the end of the subinterval, corresponds
to the hatched area in Figure 3.1.26.
2 (1, 2) ~ 0.05 + 0.05 x 1.1025
How can we make a more accurate + . . . + 0.05 x 3.8025 = 2.25875.
computation of the distance covered
(3.1.4a)
during a given time (from t = a = Is
to t = b = 2 s in our example)? Clear- Using the terminal velocities in each
ly, the distance between the lower and subinterval, we get the distance
upper estimates, that is, the difference
between 2.18 and 2.49 in our case, de-
2 (1,2) ~ 0.05 X 1.1025 + 0.05
pends on the variation of velocity with- X 1.21 + ... + 0.05 X 4 = 2.40875.
in the limits of each subinterval A t.
(3.1.5a)
For this difference to become smaller,
we must simply see to it that the veloci- The difference between 2.25875 and
ty varies less rapidly within each subin- 2.40875 now amounts to about 7%.
terval; but since the velocity is a quan- The range within which z (1, 2) lies
tity over which we have no power, we has narrowed down. Rounding off these
have to partition the time interval in-
to smaller subintervals. This approach
figures, we get 2.26 < z (1, 2) < 2.41.
As we reduce the subintervals At,
suits us perfectly since it brings us clos- the result approaches the true value of
er to our goal: if all the subintervals the distance covered. It will be comput-
are made smaller (while the total num- ed later on and proves to be equal to
ber of the subintervals increases pro-
portionately), the two estimates for z (1, 2) = 2~ ~ 2.333. As we reduce
the distance z (a, b) obtained through At, the difference between the initial
the method discussed above move and terminal velocities in each subin-
closer to each other. terval At decreases, and so also does the
For instance, if we split up the l-to-2-s relative error in each summand. It is
interval into 20 subintervals (instead because we are dealing here with the rel-
of 10) of 0.05 s duration each, using the ative error we can be sure that although
method discussed above and the initial the number of terms (that is, the num-
 )

11 Determining Distance from the Rate of Motion 95

Such methods are convenient and quite*

justified when the velocity is not known ex-
actly, that is, is given in the form of a table or
graph obtained empirically (in an experi-
ment) rather than by a formula. We will not
dwell on these approximate methods and will,
attempt to express the distance by a formula
when the velocity is given by a formula. Note*
also that in some cases when, say, the velocity
is expressed by the formula v ( t (1 z)
-1
= — ,

with t varying from 0 to 1 s, and increases

without limit as t -> 1 s, the concept of dis-
tance becomes meaningless (the distance tends
to infinity). For the time being we will not
consider such cases (but see Section 5.2).
We can also make more precise the numer-
ical method used above to determine the dis-
ber of errors) increases, nevertheless the tance from the velocity. We will assume that
precision with which we estimate the in each subinterval At the velocity is con-
entire sum of the distances, that is, stant and equal to the arithmetic mean (half-
the precision with which we estimate sum) of the initial and terminal velocities in
the given subinterval. In this approach, with
z (1, 2) is sure to increase.
a partition into ten subintervals, the velocity
To illustrate the above reasoning, in the first subinterval from 1 to 1.1 s is
in Figure 3.1.3 we depict the construc- taken equal to (1 +
1.21)/2 =
1.105 cm/s
tions connected with the estimate of (see the table at the beginning of this section)
2
= and the distance covered during this subin-
z (1 2) from the given velocity v
,
t
terval of time is 0.1105 cm, the distance cov-
at A t =
0.5, 0.2, and 0.1 s. Geometri- ered during
the second subinterval is
cally, it is obvious that as we increase 0.1 (1.21 1.44)/2+ =
0.1325 cm, and so on.
the number of subintervals A t and reduce Adding all the distances obtained in this
manner, we get the distance covered during
the duration of each one, the dimen- — =
the time interval from a 1 s to b 2 s:
sions of each step in Figure 3.1.3 become 2 (1, 2) = 0.1105 + 0.1325 +
...
smaller, whereby the step-line comes . . . =
2.335 cm. If we divide the interval into-
closer and closer to the curve v versus t .
20 subintervals, we get (using the same half-
We thus conclude that the distance
sum of velocities) z (1, 2) =
2.33375 cm.
These values are much closer to the true value
covered during time from t a to = 2.33333 cm than those computed on the basis of
t = b, given an arbitrary dependence the initial and terminal values of velocity for
of the instantaneous velocity on time, the same number of subintervals: for ten sub-
v = v (£), is equal to the area bounded
intervals, the error is equal to 0.07% instead
of 15% in the earlier method, and for 20 sub-
by the curve v =
v ( t ), the vertical lines intervals the error is only 0.02% instead
t — a and t =
6, and the t axis (Fig- of 7%.
ure 3.1.4). The new method for finding the distance can
also be displayed vividly on a graph. The pro-
This conclusion yields a method for
duct of the half-sum of the velocities at the
practical computation of the distance: beginning and end of an interval by the mag-
we can construct the graph on graph pa- nitude of the interval of time is equal to the
per and determine the hatched area ei- trapezoid ABCD (Figure 3.1.5). With bases
ther by counting the squares, or, for AB and DC and altitude AD the area is ,

instance, by cutting out the area of pa- ab±dc_ v_

ad=
per, weighing the sheet, and comparing
its weight with the weight of a rectangu-
lar or square piece of the same paper of
For this reason, determining the distance on
3 3 the basis of the half-sum of the velocities is
known area. *

known as the trapezoid method For the shape.

of the curve shown in Figure 3.1.5, the area

3 * 3
A pocket calculator, however, makes of the trapezoid is somewhat greater than the
the method of constructing sums of the type area bounded by the straight lines BA , AD ,

(3.1.4) and (3.1.5) (or (3.1.4a) and (3.1.5a)) DC and the portion BC of the curve u v (t). =
much easier to carry out than in the case of The difference between the area of the trape-
primitive weighing of paper or cardboard fig- zoid and the area bounded by the arc of the
ures. curve is equal to the area of the crescent-like
 —

<96 3 What is an Integral?

a 6

Jtf AtL t

Figure 3.2.1

body and the equivalent problem of

finding the area under a curve led to —
a consideration of sums of a special
type with a large number of small
terms (summands). A
rigorous approach
to such problems leads to the concept of
the integral.
Figure 3.1.5
The distance z =
z (a, b) found from
figure formed by the chord BC and the por- a given velocity v =
v (t) is called the
tion BC of the curve shown hatched in Fig-
ure 3.1.5). The sum of such areas yields the
all definite integral of the function v (t)
error, that is, the difference between the true (velocity) with respect to the variable
value of the distance and the value computed t (time) taken from a to b (the same
by the trapezoid method. A comparison with quantity is sometimes called the integral
Figures 3.1.2 and 3.1.3 shows vividlv that the
error in the trapezoid method should be less of function v ( t )).
than that when formulas (3.1.4) and (3.1.5) Wenow give a mathematical defini-
are employed (the step method or, as mathe- tion of the integral that corresponds to
maticians call it, the rectangular method. the ideas illustrated by the numerical
example of Section 3.1. The definition
Of course, when one compares the
will remain valid when we consider
distance with the area under the graph
physical or mathematical quantities of
representing the v versus t dependence,
a nature different from velocity and
it is important to take into account the
distance.
scale used (compare to what was said
Suppose that we have a function
in this connection in Section 1.7).
Suppose that 1 cm along the axis of ab-
v = v ( t ). To find its integral from a
to b we partition the interval into a large
scissas on the graph corresponds to a
time interval of T seconds and 1 cm on
number of subintervals, n. denote We
the values of the independent variable
the axis of ordinates corresponds to a
t at the endpoints of the subintervals
velocity of V cm/s. Then, if the motion
by h,
*o> t\,*n-i where, obvious-
•
is at a constant velocity v 0 during a
>

time from t =
a to t =
6, the distance
ly? t0 =
a and t n =
b (Figure 3.2.1).

a), and — The lengths A£ of the small subinter-

covered is equal to v 0 (b
vals of time are equal to the difference
the area of the (hatched) rectangle on
between adjacent values of t. 3A Thus,
the graph (Figure 3.1.1) is equal to =
for an arbitrary * (with l 1,2, ., n) .

—
.

S=(v0 /V) ((b a) IT) cm 2 .

A* z — h — h- 1*
Thus, here we have z (a, b) SVT. = The subscripts on the quantities t and
This relationship between the distance A t are merely number labels, or in-
covered and the area on the graph of dices, as they are sometimes called
velocity bounded by the curve v (t), (see footnote 1.1).
the axis of abscissas, and two vertical
lines is preserved in the case of a 3 4 If the interval from a to b is specialh
-

variable velocity v, that is, in the case partitioned into n equal parts, then eacl
= subinterval At equals (b — a)/n. Itisnotobli
of an arbitrary function v v (t).
gatory, however, that the subintervals b(
equal; the only thing we require is that eacl
3.2 The Definite Integral subinterval be small. The reader will se(
the truth of this if he or she thinks througl
In the preceding section, two problems the distance-velocity example of Section 3.1
finding the distance traversed by a see also Exercise 3.2.3.
 a 1 )

S.2 The Definite Integral 97

The approximate value of the integ- Unlike Af, the symbol dt signifies
ral z b ) is given by the formula
(<
,
that in order to obtain the exact value
l—n of the integral it is necessary to pass to
z(a, b)~ At . (3.2.1) the limit as all subintervals At tend to
t
1= zero (just as the derivative dz/dt is
1=71 obtained from the ratio Az/Af if we send
2
z=i
signifies that the expression standing Az and At to zero and pass to the limit).
The formulas (3.2.1) and (3.2.2), in
to the right of the summation symbol 3 •
5
which the At are small but finite, only
be taken for all values of l from 1 to w
yield approximate values of the integ-
and then all these expressions are to be
Az/At only yields
ral, just as the ratio
added together. For example, if n — an approximate value for dz/dt for finite
1 = 10 Az and At.
10, then 2 v (*z_ i)
—v (t 0) At x + When the subintervals At become
z=i
smaller and smaller, it is immaterial
y(^) A^ 2 + . .
. +y (t )At 9 l0 . In the example whether we take the value of the func-
tion v (t) at the beginning, at the end,
of Section 3.1 (see the table in that
or in the middle of the subintervals,
section), t 0 = 1, t x = 1.1, t2 = 1.2,
which is to say that it is immaterial
1 = 10 whether we proceed from (3.2.1) or
. . ., and z (1, 2) ~ 2 = 2.185. from (3.2.2) or from some other expres-
i=i
sion for the “integral sum” similar to
In
the approximate expression the one in the right-hand sides of
(3.2.1),the value of the function v ( t
(3.2.1) and (3.2.2) (say, we could have
in each subinterval was taken at the
added the terms v (t m ) A£, where t m
beginning of the subinterval, at the is the middle of the corresponding sub-
point different approximate A interval M). For this reason formula
expression is obtained if we take the
(3.2.3) simply has v (£), which is a val-
value of the function at the endpoint ue of the function in the subinterval
of each subinterval:
At without any indication of the value
1=71 of t being taken inside (or at the begin-
z (a, b) ~2v (t t
) (3.2.2) ning or end) of the subinterval.
i=i
Another way in which the integral
In the example of Section 3.1, this sum (3.2.3) differs from the sums (3.2.1)
for n =
10 was equal to 2.485. and (3.2.2), which yield approximate
The definite integral of a function values of the integral, lies in the fact
v ( t ) taken from a to b is the limit ap- that as the At becomes smaller and smal-
proached by the sums (3.2.1) and (3.2.2) ler and the number of subintervals in-
as all subintervals A t x tend to zero. creases, we no longer label them. For
The integral is written this reason, we indicate on the integral
b only the limits of variation of t (range
z(a, b)= of t) from a to 6. The quantity a is
v(t)dt (3.2.3)
j placed at the bottom of the integral sign
a
and is termed the lower limit of integra-
(read: z (a, b) equals the definite inte- tion, while b is placed at the top and is
gral of v (t) from a to 6, dt). The inte- called the upper limit of integration. 3 6 *

gral sign is merely an elongated S

^ 3 6
* In this section the word “limit” is used
(the first letter of the word summa , in two meanings. First, the integral is the
which is the Latin for sum). limit of a sum in the same sense that the
3
derivative is the limit of a ratio. Here, limit
* 5
.The symbol
sigma. It corresponds to S in the Latin alpha-
2 is the capital Greek letter corresponds to the sign lim. Second, we speak
of the limits of variation of t from a to b,
bet, the first letter of the term sum. the limits of integration a and b. The meaning
7-0946
 )

98 3 What is an Integral?

The range t from a to & is called the

of Naturally, not just any sum of a large
domain of integration. The function number n of small terms tends to a
v (£) in the expression of the integral definite limit as n oo. Say, the sum
the integrand t being the
is called
variable of integration.
, of n terms each of which is n is UY
Thus, the integral is defined as the
equal to Y n and tends to infinity as
n -> oo, which means that this sum has
limit approached by the sum of pro-
ducts of the values of the function multi-
no We shall try to explain
finite value.
why our case the limit must exist.
in
plied by the difference of the values of
Let us partition the interval from a
the independent variable when all
to b into n equal subintervals, the
differences of the independent variable
tend to zero: 3 * 7
length of each being ( b —
a)/n. If for the
sake of simplicity we take the velocity
l—n v as being constant, we get a sum of n
lim 2 v ( t i )Afj
terms each being equal to vAt =
A ‘i~°
i=i v (b — a)/n. The total sum (that is,

l—n b the distance equal to

covered) is

= lim y\ v At t = \v(t)dt . (3.2.4)

nvAt = nv (b — a)/n =a ), which
v (b —
A*/-0 means that it is independent of n .
1 1=1A
J
a
What is important here is that each sep-
Although the first and second sums in arate term diminishes in exactly the
(3.2.4) are different for a finite number same proportion (in proportion to 1 hi)
l of small subintervals At h their limits as does the number of terms n increase.
coincide when all subintervals A t de- It is also clear that for the case of a
crease without limit (tend to zero). variable velocity and the partition of
These limits yield the value of the in- the interval from a to b into n equal
tegral. small segments the same is true, that is,
As At tends to zero, each separate the decrease in the length of each subin-
summand tends to zero, but on the oth- terval is inversely proportional to the
er hand the number of terms in the sum growth in the number of such subinter-
increases and approaches infinity. The vals. For instance, when each small
sum itself tends to a very definite lim- subinterval At is divided into two halves,
it, which is the solution of the prob- A x t and A 2 t, the terms in the integral
lem and is termed the integral. If the sum corresponding to these halves prove
function is the instantaneous velocity, to be approximately half the initial,
this limit, the integral of the function, “large”, summand, but the total number
is equal to the distance covered. If the of terms doubles, so that the order of
integrand represents the ordinates of magnitude of the entire sum does not
the points of the graph y f (x), the — change. 3 8 The reader that is not con-
*

b vinced by all this reasoning should do the

integral / (#) dx is equal to the area exercises at the end of this section.
£ This explanation is useful if we are
a
bounded by our graph, the x axis, and dealing with the mathematical defini-
the vertical lines x = a and x — b. tion of the integral as the limit of a
certain sum. In physical problems, how-
here is different. The attentive reader will
readily see which of the two meanings is used 3* 8
Note that in the example with n sum-
in any given case.
3 7
-
To he more precise, we should have mands each of which is equal to 1 /Y n, a 2-
written fold increase in the number of terms is accom-
l=n panied by a decrease in the magnitude of each
lim A ti (and term only by a factor of Y%- In the example of
y) v ( ti
Ati, Ats , ..., A tn 0 1=1 the sum of terms each of which is l/ra 2 a 2-fold ,

increase in the number of terms leads to a 4-

similarly for the second sum), but such nota- fold decrease in the magnitude of each term,
tion is too unwieldy. so that the new sum is only half the old one.
 S.2 The Definite Integral 9&

ever, the existence of the integral,

that is, the limit of the “integral sums”
considered, is as a rule quite obvious.
For instance, there is no doubt that a
body traveling with a finite (constant
or variable) velocity will, over a finite
time interval, traverse a certain (finite)
path, that is, it will travel a certain
distance. In what follows we will ex-
plain how by employing the concept
of the integral one can calculate the whose lateral sides are the line segment
area of curvilinear figures (this question AB the axis of abscissas and the
of
was discussed in brief above). Here portion CD of the curve y y(x) re- =
too there can be no doubt that the prob- duces to computing the integral
lem has a solution, that is, a figure b
must have an area, and so the approp-
S= y {x) dx (3.2.5)
riate integral exists. j
Since the variable of integration can a

assume the values a and b it is clear, To explain this, recall Figure 3.1.2*
that the limits of integration have di- Imagine that the values of an (arbit-
mensions and their dimensions are those rary) independent variable x are laid
of the variable of integration (in the off on the axis of abscissas, the values
distance-velocity example, the limits of a function y (#) on the axis of ordi-
of integration have the dimensions of nates, and both x and y are assumed to
time). (As for the dimensions of the in- be the distances of a variable point
tegral, see the end of this section.) M(x y) from the axes of coordinates,
,

Clearly, the value of a definite integ- (see Section 1.2) and in no way are relat-
ral depends on the values of the func- ed to the physical concepts of time and
tion under the integral sign solely distance. The sum of the areas of the»
within the domain of integration. Val- hatched rectangles in Figure 3.1.2a;
ues of the function outside the domain l=n
of integration have no effect on the mag- is equal to Zj y ( x i-i) A#*; the same-
nitude of the integral and are of no in- i—i

terest to us. For instance, this distance

sum in Figure 3.1.2& is equal to*
l—n
covered, z (a, b), depends of course only
on the value of the velocity v = 2
i=i
y (x i) Ax t . In the limit, as Ax t
v (t) inside the domain of integra- these sums are, by definition, equal to
tion and does not depend on the ve- the integral, and the sum of the areas
locity before t =
a or after t b. = of the rectangles tends to the area bound-
It was pointed out in Section 3.1 ed by the curve y x ), since the smaller
(,

that the distance can be determined by the subintervals Ax t the closer to tho
computing the area on the graph in curve is the polygonal (step-like) line-
which the velocity is a function of bounding the rectangles.
time. The problem of finding the area S In conclusion we note that the def-
bounded above by a curve with a spe- inite integral depends on the integrand
cified equation y —
y (x), below by the and the limits of integration, but it
axis of abscissas (the x axis), and on the is independent of the designation of
sides by the vertical lines x a and= the variable of integration. Judge for
x =b (Figure 3.2.2), that is, of calcu- yourself. Suppose that we have thfr
lating the area of the curvilinear trape- integrand in the form v ( t) 3 12 5. = +
zoid ABCD, whose bases are the paral- Substituting x for t we get v (#) ,
=
lel line segments AC
and BD of the 3z 2 + 5. When we compute the integ-
straight lines x = a and x = b and ral, it is immaterial how the variable?

7*
 ) 1

100 3 What is an Integral?

b
of integrationis designated, the only
important thing being over what range distance z (a, b) = v (t) dt is mea-
j
it varies and what are the values of a
the function. For this reason, sured in units of (cm/s)-s cm. If =
the abscissa x and the ordinate y are
b
measured in centimeters, the area
Z (a, b) = v (t
dt = b
j
a S = j*
y (x) dx is measured in square
a
centimeters (cm* cm cm 2 ). A transfer =
to new units of measurement, say, to
e\ in x and to e' in y, results in the
2
or even multiplication of integral (3.2.5) by
b a factor of where ex = k and
z(a, b) = ^
v ( o) do e2 = The
k 2 e'2 . integral increases in
a the process by kt k 2 but a factor of ,

it expresses the same quantity as be-

^the last form is never used, of course). fore (only in new units, e[e 2 ). For in-
Any letter will do for the variable of tance, in going over from centimeters
—
integration the result will be the o millimeters, we increase the numer-
same. b

Avariable which (like the variable ical value of the area S = \ y (x) dx
of integration) does not appear in the
final result is called a dummy variable. by a factor of 10 X 10 100, while =
The variable of integration under the in going over from cm/s to m/min we
integral sign can be changed to any are forced to multiply the value of the
letter without disrupting the validity distance (which before was measured
of the formulas. An ordinary (not dum- b

my) variable can be replaced by a dif- in centimeters) z (a, b) = v t

( )
dt by
ferent letter only in all parts of a for- j
mula; for instance, in the formula a factor of (0.01 1/60) X 1/60 0.01 =
{x +l)
2
#2 2x =1 one cannot + + (the new value, of course, will be ex-
write (x l)
2
£
2
+2£ 1, but in = + + pressed in meters).
integrals we can write z (a, b) =
b b

^
v (i
r) dt or z (a, b) = j
v {x) dx. Exercises
a a
Finally, let us turn to the question 3.2.1. Consider the case v kt s (uni- = +
formly accelerated motion). Find the distance
of dimensions By definition, .
covered during the time interval from a to b
b l=n by dividing this interval into m equal subin-
tervals; take advantage of the fact that the
V
[y(x)dx-~ lim
Ax 1z -0
2 y ( x i) &x t
. (3.2.6)
terms of the sum form an arithmetic progres-
a
.
1 = .

sion. Find the limit of the sum as m oo.

Compare the resulting expression with the
If x is measured in units of e 1 and y area, equal to the distance covered, of a tra-
in units of e 2 then each term in the, pezoid in the ^-plane.
sum on the right-hand side of (3.2.6) 3.2.2. Consider the case v t
2
and for =
this case find the distance covered in the time
lias the dimensions of e x e 2 (the factor =
interval from t 1 to t == 2, in other words,
Axx has the dimensions of el9 and 2
y fo) has thejdimensions of e 2 ), whereby find the integral j* t
2
dt. To do this, partition
the limit of the sum (the integral) has
l
the same dimensions. For instance, if the interval from 1 to 2 into m equal parts
the velocity v (£) is measured in units
of cm/s and t in units of s, then the and compute the sum 2 —or the sum
 ) )

3.3 The Relationship Between the Integral and the Derivative 101

and right members, regarding u as

2 Compare these two sums. \ Hint.
the (independent) variable and a as a
To calculate these sums, employ the for- constant, with the result that z (a) —
mula l
2
+ 22 + 32 + ... + n2 = on the right-hand side is also a con-
n(n + l)(2n + l)
] stant and d (—z ( a))/du 0. This yields =
dz ( a u)ldu,
=
dz ( u)ldu But we know .

that the derivative of the coordinate of

3.2.3. Evaluate the integral
]
3
x dx. To a body with respect to time is nothing
1
but the instantaneous velocity of that
end divide the interval from 1 to 2 into
this
m small subintervals Ax whose lengths form a
body, so that dz ( u)ldu v ( u is the =
geometric progression ,
and pass to the limit velocity at time u and, hence
as m -* oo.
dz{a ' u)
-
=v(u). (3.3.3)
du
3.3 The Relationship Between
Substituting here the expression (3.3.1)
the Integral and the Derivative
for distance z {a, u) in the form of an
In the preceding sections we consid- integral, we get
ered, separately, the concepts of the u
derivative and the integral. In the
»(t)dt)=v(u). (3.3.4)
present section we examine (using the -hi]
a
distance-velocity example) the rela-
tionship between these two concepts. This equation is the most important
This step will unite the calculus of general property of the definite integral.
derivatives and the calculus of inte- In the given form, this equation is a
grals into a single science that abounds general mathematical theorem. Its va-
with applications, the differential and lidity is independent of whether v {t)
integral calculus or, as it is sometimes , is velocity (and the integral is distance)
called, mathematical analysis . or v (t is some other quite different
We assume as given and known the quantity. For any function, say y x ), (;

instantaneous velocity as a function of we have

time v =
v (£). We regard as constant b
the time t =
a at the beginning of
x) dx -=y( b)’ (3.3.4a)
the motion (a train leaves a certain -fb(\y( )

station at a known time t a), and = a

consider the distance covered in the where we have replaced the upper
time interval from a to b as a function (variable) limit of integration with
of the terminal instant b. To emphasize b.
this, we will denote the terminal (un- The theorem is stated thus: the de-
defined) moment not by b but say, by rivative of a definite integral with re-
the letter u ,
since it is customary (but spect to the upper limit is equal to the
not mandatory) to denote variable value of the integrand at the upper limit.
quantities by the last letters of the Eng- Because the theorem is so important
lish alphabet. The result is (it is called the Newton- Leibniz theo-
u rem? ), we give
2
a different derivation
of it based on a consideration of area.
z(a ,
u) — j
v (t) dt . (3.3.1)
We will compute the derivative by
a
first principles, that is, as the limit
We know that of the ratio of the increment of the
z (a, u) = z (u) — z (a) (3.3.2) 3 9
In some old textbooks on differential
*

and integral calculus this theorem is called the

(in the example with the train it is
fundamental theorem of higher mathematics.
natural to assume that z (a) is zero). This name sounds rather high-flown, but it is
Let us take the derivative of the left quite justified.
102 3 What is an Integral?

The definite integral of a known func-

tion y (x) or v (t) is a function of the
limits of integration a and 6, that is,
a function of two variables, a and b .

Formula (3.3.4) or (3.3.4a) yields the

value of the derivative of this function
at the upper limit of integration, the
variable b. Below, in Section 3.5, we
Figure 3.3.1
will find the value of its derivative at
the lower limit of integration, a. 3 10 -

function to the increment of the in-

The definition of an integral as the
dependent variable. We consider
limit of a sum, which was given in the
b
preceding section, explains the role
I (a, b) == y (x) dx. the integral concept plays in the solu-
j
a tion of physical or geometric problems:
in computing the distance traveled when
This integral is the area bounded from
a body is moving with a (variable)
above by the curve y (#), from below
velocity that is known, in determining
by the x axis, on the left by the verti-
the area bounded by a curvilinear tra-
cal line x =
a, and on the right by the
pezoid, and in other problems that lead
vertical line x = b (see Figure 3.2.2).
to the construction of “integral sums”
How do we find the increment A I
(below we will encounter many prob-
of the integral I caused by a small
lems of this type). But this definition
increment A b of the upper limit &?
does not yield a convenient general
By the definition of an increment, method for computing an integral or
AI =I (a, b A b) +
I (a, b). The - finding the value of the integral in the
area equal to the integral I (a, b A b) + form of a formula, as a function of the
differs from the area equal to the in-
limits of integration. 311
tegral I (a, b) in that the right vertical
is displaced rightward by a small
A method for finding such a formula
follows from the Newton-Leibniz theo-
A b (compare Figure 3.3.1 with Fig- rem proved above, a theorem that con-
ure 3.2.2). Consequently, the increment
cerns the derivative of an integral.
A / is the difference between two areas: Here, besides the property of the de-
that with the base from a to b A& + rivative of an integral, we make use
and that with the base from a to b.
of yet a second property of the definite
A / is clearly the area of the thin strip integral: the definite integral is equal
that is hatched in Figure 3.3.1. The
to zero when the upper and lower limits
base of this strip on the x axis is a line
of integration coincide :

segment of length A 6.
The desired derivative is equal to a
the limit
z (a, a) = v ( t) dt = 0. (3.3.5)
dl(a, b) A/ j
lim a
db A6-*0 A b
*

3 10
Quite obviously, as A b tends to zero, -
Here, of course, we are speaking of the
the area of the strip approaches partial derivatives of the function I ( a b) ,
=
b
y (b) A 6, and the ratio A// A 6 approaches y (x) dx of two variables, a and b (see
the quantity y(b). We have thus j
a
once again given a pictorial proof of
Section 4.13).
the Newton-Leibniz theorem: 3 n
-
Only in rare cases and with great diffi-
b culty is one able to sum an arbitrary number
of small summands present in the definition of
y(x)dx)=y (5). the integral (e.g. see Exercises 3.2.2 and
( j
a 3.2.3).
 a 5

3.3 The Relationship Between the Integral and the Derivative 103

This property is obvious because the

distance is equal to zero if the time in
transit is zero.
•
CCf £2 &3 3?5 ^

The formula which yields the value

of the integral as a function of the lim- Figure 3.3.2

its of integration will be derived in

this fashion in Section 3.4. that is, the increment Ax becomes
—
z

Let us now study the question of the smaller. But

as the difference x x
integral of a derivative. In the integ- x z-1 = Ax z decreases, that is, as xx
ral approaches x z _1? the difference between
b / (x x ) and / {x l _ 1 ) also diminishes. For
this reason, we have just as much right
v (
t) dt (3.3.6)
j (the degree of accuracy is the same)
<x
to put / (x z-1 ) in the right-hand side of
the integrand may be an arbitrary func- (3.3.9) as we do / (x z ), as was done
tion,and of course there is no universal above, where we selected the latter
method that enables finding the integ- case.
ral for an arbitrary function v ( t ). Let us now write formulas like (3.3.9)
But suppose we are lucky and have for all subintervals into which the
found a function F such that the in- domain of integration, that is, the
tegrand v ( t ) is the derivative of this interval from a to b is partitioned. ,

function F :
Suppose that the interval is partitioned
into five subintervals (Figure 3.3.2) so
= (3.3.7)
that x 0 — a and x 5 = b. Then
In such a favorable case everything F (xj) — F (x ~ / 0) (xj) (a:, —x 0 ),
turns out wonderful: here we can easily F (x 2 ) — F (x ~ / (x x) 2) (x 2 — Xj),
find the exact value of the integral.
To write it out, we recall the approxi-
F (x s ) — F (x ~ / (x 2) 3) (x 3 —x 2 ),

mate expression for the increment of (3.3.10)

function (cf. (2.4.6)): F (x 4 ) — F (x ~ / (x 3) 4) (x 4 —x 3 ),

AF ~ F' (
t) to = v (
t) M, F (x 5 ) — F (x ~ f (x 4) 8) (x 5 —x 4 ).

or in the more common notation, which Now add them together. In the left
b members, all values of the function
refers to the integral / (x) dx ,
with F for intermediate values of x cancel
^ out leaving only
/ (x) = dF (x)!dx,
F (x 5 ) -F (x„) =F (b) —F (a).
AF ~ F' (x) Ax — / (x) Ax. (3.3.8)
On the right-hand side we have pre-
The quantity in the right member of cisely those sums with the aid of which
the equation is precisely one of the we approximately expressed the in-
summands whose sum is equal to the tegral in Section 3.1 (expressing the
integral. And so, by putting, say, distance z ( b) for a given velocity
,

Ax z
= x t
— x z _x and, hence, AF = v ( t )). Thus,
F (xi) — F (x z-1 ), we can write, ap- 1 =
proximately,
F {b) — F (a) ~ 2 / (*i) (*i —x i-\)
AF = F (x,) — F (x,_ x ) ~ / (x,) (x, 1=1
—x i-i)
= / (
x i) Ax,. (3.3.9) b

As already mentioned, (3.3.9) is ap-

= 2 / ( x i) Ax, ~ j
/ (x) dx,
a
proximate and its accuracy increases
as the difference between x and x z-1 x , where / (x) = dFIdx.
 b. b

104 8 What is an Integral?

The smaller each increment Ax, or 3.4 The Indefinite Integral

the difference x x —- x , the more exact M In the preceding sections we introduced
the expression (3.3.9) for the increment
the concept of a definite integral as the
A F. But as all the differences Ax x = limit of a sum of a large number of
xx —
x x _x decrease (and the number of
small terms. In Section 3.3 we elu-
subintervals Ax grows without limit), x
cidated the principal property of the
the sums tend to the integral. There-
definite integral: the derivative of a
fore, the equation
definite integral with respect to the
upper limit is equal to the integrand,

F ( b)
— F(a = j / )
b

(a:) dx for f(x) = ~ that is,

—
b
d*
dt , then
(3.3.11)
if z (a, b)
j
a
v t
( )
^
=v(b). (3.4.1)
is now exact.
We now wish to take advantage of this
The last statement seems to be quite con- property to compute a definite integral.
vincing, but it can be justified by even a more We seek a function of b whose de-
exact computation. The thing is that the rivative is the known function u (b).
error introduced by a formula similar to
(3.3.8) or (3.3.9) is, generally speaking, quad-
We denote this function by F ( ). Then,
ratic: it is approximately proportional to by definition,
the second power of the length Ax of the in-
terval, that is, has the form c (Ax) 2 where = (3.4.2)
the quantity c =
c (x) depends on x but weak-
,

ly and with a high degree of accuracy can be

assumed to be a constant (all this will be This equation does not fully define
explained Chapter 6). Thus, for Ax
in = the function F(b). know that the We
( —
a)/n each expression in (3.3.10) (the
,
addition of any constant to a function
number of such expressions must be taken to does not change the derivative of that
be n instead of five) introduces an error pro-
— function. Hence, if F(b) satisfies (3.4.2),
portional to (Ax ) 2 2
(1 In ) (b= a) 2 The total .

—
error obtained through summation of all such so will the function G(b) F (b) C +
approximate equations will be proportional for any constant C.
to n [(1/w 2 ) (b a) 2 ] —(1 In) (b = a) 2 from — , The function F(b) which satisfies
which it follows that it tends to zero as n->
oo
Eq. (3.4.2) is called the indefinite
integral of the function v ( b ). Similarly,
if
Formula (3.3.11) establishes a re-
lationship between
the derivative. From this formula it
and the integral -^ = /(*), (3.4.2a)

follows that if we are able to find a it said that F(x) is the indefinite
is
function F whose derivative is equal integral of the function / (x ). This
to the integrand /, the problem of eval- term reflects two properties of the func-
uating the integral is solved, for all tion F(x). First, the function F has
that remains is to substitute a the same derivative with respect to
and b for the independent variable of b as the (definite) integral z (a, b) y
this function and find the difference b

F (b) —F (a). or
j
/ (x) dx (cf. (3.4.1), (3.4.2), and
Since formula (3.3.11) is so impor- a
tant, in the sections that follow we will (3.4.2a)); for thisreason F is also called
give a different derivation of it on the an integral. On
the other hand, con-
basis of a more detailed consideration dition (3.4.2) or (3.4.2a) does not de-
of the properties of the integral and fine this function completely, since we
the function F . can always add a constant quantity C
 a ) b ) )a ) .

3.4 The Indefinite Integral 105

to it, whence the modifying adjective nite integral using (3.4.5), taking G
“indefinite.” instead of F :

Any solution to Eq. (3.4.2) or (3.4.2a) z (i b) = G (b) — G (a)

from a solution F ( b (or
,
can differ
F(x)) solely by a constant Indeed, if .
= lF(b) + C]- \F(a) + C]
there is another solution to (3.4.2), = F(b) - F (a).
G (6), then for their difference F — G
we have The is the same as above.
result
Indistance-time problem it is-
the
±lF(b)-G(b)] = v(b)-v(b) = 0. convenient to denote the indefinite
integral by the same letter z as we used
But only the derivative of a constant for the definite integral. For a given
isequal to zero for arbitrary values of integrand v ( t ), the definite integral
the argument, that is, only a body at depends on the upper and lower limits*
rest has a zero velocity all the time. of integration, which is to say, it is
According to (3.4.1), the definite in- a function oftwo variables: z = z (a, b).
tegral z b) is (i
,
also one of the solu- The indefinite integral is a function of
tion to (3.4.2). Hence, z (a, b) can one variable, which we denote by t .

likewise be represented as Thus, the indefinite integral z (t) i&

a function that satisfies the equation
z {a, b) =F b + C, (3.4.3)
=J
2T = '>V)-
(
z '( t ) (3-4.6)
where F (b) is a solution to (3.4.2)
Using this function, we can find the-
and C is a constant, and it only remains
definite integral z a b) of the function,
to determine the constant. To do this, ( ,

v t from the formula

we take advantage of the second prop- (

erty of the definite integral: it is b

equal to zero when the upper limit z(a ,

b)— \ v (t) dt — z(b) — z (a)
coincides with the lower limit, J

* (a, a) = 0 (3.4.4) (3.4.7)

The following compact notation is*
(see (3.3.5)). Substituting a for b in used to state the difference between
(3.4.3) and using (3.4.4), we get the values of one and the same function
0 = F(a) + C, or C = ~F(a ). From for two different values of the variable:-
this we finally have
z (a, b) = F{b) —F z(t)\ a
b
= z(b)-z(a). (3.4.8)
(a), (3.4.5)
Here, on the left is the function of a,
or dummy variable followed t, which is
b
by a vertical line at the top of which,
j
f(x)dx = F (b) — F (a), (3.4.5a) is the value of the variable at which.

a we desire to take the function with,

a plus sign, and below, the value for
where F'(x)=f(x).
which we want to take the function
It is to be noted that the “indetermi- with a minus sign.
nacy” of the function F(b) does not in Substituting, under the integral sign
any way hamper computation, with in (3.4.7), v ( t expressed in terms of
z (t) in accord with (3.4.6) and putting;
its aid, of a definite integral from for-
mula (3.4.5) or (3.4.5a). Indeed, in expression (3.4.8) into the right-hand
place of F ( b ) take some other solution side of (3.4.7) we get the identity
to Eq. (3.4.2), say G (6), which differs b

from F(b) by a constant: G ( ) = \

z' (t) dt =z (t)
6
, (3.4.9>
C. We will evaluate the defi-
)

F(b) + J |
a
 ^

106 3 What is an Integral?

or, in other notation, Thus, using the indefinite integral,

we obtain in a few lines the exact result
b which we laboriously approached in
F' (x) dx = F (x)
*
(3.4.9a)
Section 3.1 by numerical computations.
a
The definite integral is the limit of
a sum of the form
It will be seen that the positions of
andb on the left and on the right are v (to) (h — t 0) + V (tj) (f 2 — tj) + ...
the same, which is a good mnemonic as each term tends to zero and the num-
-device for memorizing (3.4.9) or
ber of terms increases correspondingly.
(3.4.9a). In an approximation of this sum, one
We have studied in sufficient detail has to partition the domain of inte-
the concept of the indefinite integral. gration into a number of subintervals,
It is now time to examine some exam- find the approximate value of the dis-
ples. Let us consider the problem of the tance vkt in each subinterval, and
distance covered during the time in- then add these values. To obtain good
terval from a to b at a velocity equal accuracy requires numerous arithmetic
to v (t) t
2 —
The sought distance is
.
operations. But if we know the indef-
'equal to the definite integral inite integral z (t), that is, if we know
b the function whose derivative is equal
z a ,
b) = t
2 dt. to the integrand u (;t ), then any definite
(
j b
a
integral v (£) dt is obtained immedi-
In this problem, the indefinite integral j
a
.z ( t ) is obtained by solving the equation
ately (3.4.9). The pos-
from formula
sibility finding functions with a
of
iiGL= y (*_)_= i*. (3.4.10) given derivative (indefinite integrals)
has “suddenly” put at our disposal
a powerful method for computing sums
But we know that d(t*)/dt = 3 1 2 hence
,
(definite integrals).
3
d(t /3)/dt (1/3) (3 F) = = t\ Conse-
The indefinite integral is sometimes
quently, the equation is satisfied by
called a primitive function (antideriv-
the function
ative). This term is understood as the
inverse of a derivative: we are speak-
z(*)=4 +C, (3.4.11)
ing of a function whose derivative is
known. The term is used in textbooks
where C is an arbitrary constant that
where the problem of finding a func-
can be taken to be zero, since we know
tion from the known derivative of the
that this does not affect in any way
function is solved before definite inte-
the difference of two values of function
grals are considered, that is, where in-
jz for two values of the independent
tegral calculus is started from the con-
variable.
cept of an indefinite integral rather
Substituting (3.4.10) and (3.4.11)
into yields
than that of a definite integral. We
(3.4.9)
do not use this approach here.
s b 63 a3
An indefinite integral can always
t
= be expressed in terms of a definite in-
3~ a 1 F *

tegral:

The special case of a = 1 and b = 2

t

yields z (t) =C+ v (x) dx . (3.4.12)

j
a0

T = T“ 2 333 ' - Applying the rule concerning the de-

rivative of a definite integral with
 ) —

3.4 The Indefinite Integral 107

respect to the upper limit, it is easy In addition, the results of Exercises

to verify that z ( t given by Eq. (3.4.12) 2.4.2 and 2.4.3 enable us to write
satisfies (3.4.6) for arbitrary constants
C and a 0 . $-j^- = 21/i. jl/tdt=^Y¥.
In all problems, the answer always
involves the difference z (b) z (a), — To verify any these formulas, it
of
and this is independent of C and a 0 . suffices to find derivative of the
the
Therefore (3.4.12) can be written more right-hand side. If in doing so we get
compactly: the function under the integral sign, the
formula is correct.
t
Methodsfor finding indefinite inte-
z (t) = J
v (x) dx. grals of a variety of functions are con-
sidered in detail in Chapter 5. Thanks
This is frequently abridged still further
to the relationship which exists be-
to tween an integral and a derivative, we
z(t)= v(t)dt. are able to find the integrals of a large
(3.4.13)
j number of functions.
The problem of integration is techni-
This notation is widely used and we
cally a much more complicated job
will employ it, but one should bear
than the problem of finding derivatives.
in mind that, properly speaking, it
312 It may be compared The complexity is due, for one thing,
is not correct.
to the fact that in the integration of
with those grammatically loose expres-
rational (i.e. not containing radicals)
sions that occur in everyday speech
algebraic expressions we get answers
and are clear to all (except children and
involving logarithms and inverse tri-
pedants), for example, expressions like
gonometric functions. In the integra-
“to eat another plate” or “I’m ready
tion of algebraic expressions with rad-
to throw the whole thing over”. No-
icals the result is sometimes expressible
tations and expressions that are not
only with the aid of new, nonelement-
altogether precise but are sufficiently
ary, functions that cannot be expressed
clear are frequently used in mathemat-
in terms of a finite number of opera-
ics, they do not usually create any
tions involving algebraic, power, and
difficulties for anyone.
trigonometric functions (see Section 5.5).
The notation (3.4.13) violates the
However, the difficulties of express-
rule by which the dummy variable of
ing integrals by formulas should not
integration does not appear in the re-
eclipse the fundamental simplicity and
sult. Therefore,when using the abridg-
clarity of the integral concept. If it
ed notation (3.4.13) one should always
is impossible (or difficult) to evaluate
bear in mind that this is a conventional
an integral by formula (3.4.5a), it is
contraction of the exact expression
always possible to evaluate an inte-
(3.4.12), in which C = z (a 0 ).
gral by means of cumbersome, yet
The familiar formulas for derivatives
fundamentally very simple, computa-
yield a table of indefinite integrals:
tions.

dt = t, t dt= — t
2
dt =~ ,
of
In the last few years, with the advent
programmable pocket calculators, it
j j ,
j
dt_ has become very easy to calculate the
1_
*
2 “ t
-
sum of 10, 20, or even 50 terms
this requires 10 to 30 minutes depend-
3 12
Note that in the left-hand side of
-
ing on the complexity of the integrand.
(3.4.13)we have a function z ( t ), while in the In many cases, therefore, a direct cal-
right-hand side we have an indefinite integral,
culation of an integral is preferable
that is, a family of functions F (t) C differ-
ing from one another by a constant (since in
+ over trying to obtain complex formu-
F (t) -f- C the constant C is arbitrary). las. In the diary of the famous physi-
 108 3 What is an Integral?

Enrico Fermi there is a calcula-

cist
tion of a rather simple definite integral.
Notwithstanding the fact that the cor-
responding indefinite integral could
be expressed in terms of the inverse sine
and the natural logarithm, Fermi pre-
ferred to find the integral numerically,
not employing formula (3.4.5a).
Many books are devoted to fast
and exact methods of numerical
integration. The reader wishing to
familiarize himself with these methods
can turn to the book [15]. vertical line x —x 0 and the axis of abscissas.
Express the area in terms of the coordinates
x 0 and y 0 of the end of the arc OA of the pa-
Exercises rabola.
(b) The same for a parabola passing
3.4.1. Evaluate the following integrals
through the origin and having a horizontal tan-
1 l.l 2
gent at point B (x 0 y 0 ) (Figure 3 A. 2b). [Hint.
,
2 dt,
(a) t (b) t
2 dt , (c) j* dt/t 2 , and The answer may be obtained at once by using
^ J the result of Exercise 3.4.3 (a). Still and all,
o l i
3 take the hard way and do all the operations in
their requisite order, without resorting to a
(d) \dtiV~t- clever trick. Seek the equation of the parabola
1 in the form y —kx 2 + px +
q, where k, p,
3.4.2. (a) Using an integral, find the area and q are found from the condition of passage
of a right triangle having a base b and an al- through the points (# 0 y 0 ) and the origin
,

titude h. [Hint. Put the origin of coordinates (0, 0) and from the condition that the tangent
at the vertex of an acute angle of the triangle at the point (s 0 y 0 ) is horizontal. Express the
,

and the right angle on the x axis (Fig- area in terms of x 0 and y 0 If you are not able
.

ure 3.4.1a). Find the equation of the hypote- to use the result of Exercise 3.4.3 (a), then
nuse in this system of coordinates and find the first try the above-described procedure and
area as an integral.] the result will suggest the relationship between
(b) Find the area of the same triangle by Exercises 3.4.3 (a) and 3.4.3 (b).
placing the right angle at the origin and an 3.4.4. Write the expression for the area of
acute angle of the triangle at the point (6, 0) a semicircle of radius r (Figure 3.4.3) in the
(Figure 3.4. 16). [Hint. In integrating, make form of a definite integral. [Hint. Use the
use of the obvious property of the integral of equation of a circle, x 2 +y
2 =
2
r .]
1
a sum of two terms, (/ + g) dx — / dx +
j j 3.4.5. Evaluate the integral dx/{l-{-x 2 )
^
g dx.] o
using the trapezoid formula (see Section 3.1),
Remark. Do not be indignant that a lot of putting first n =
5 and then n 10. Carry —
effort is put into finding the familiar answer out the computations to the fourth decimal
S a = (1/2) bh because the method of inte- place.
gration will be used later on in cases where Remark. The exact value of the integral is
elementary methods of finding areas do not jt/4 ~
0.785 398 (see Exercise 5.2.4). An
suffice. approximate evaluation of the integral enables
3.4.3. (a) Find the area under the parabo- us to obtain an approximate value for the
la y = ax 2 (Figure 3.4.2a) bounded by the number ji.

Figure 3.4.1 Figure 3.4.3

 ) ) ,

3.5 Properties of Integrals 109

simple physical meaning of a distance

covered (if v v (t is the velocity) =
or an area (if v v ( t is the equation =
of curve). But what is the sign of
a
the integral of a negative function,
that is, in the case of v (t) 0? <
For the time being, we leave the
3 y\ !h condition b a. >
In the expression
for the sum (which in the case of pas-
\

sage to the limit becomes an integral),

the factor At in each term is positive,

0 X 0
— x X
the factor v (t) is negative, which means
0 that summand is negative, the
each
sum negative, hence the integral is
is
negative, too. To summarize, if v (£)
isj^negative and a t <Z b (so that
b
<
(a) (6) (c)
b > a), then
j
v (
t
)
dt < 0. In the
Figure 3.4.5 a
case the meaning of the
of motion,
3.4.6. Construct the graph of the function answer simple: a negative v (t) sig-
is
X nifies that the motion occurs in a di-
F (x) = y (x) dx. The function y (x) is giv- rection opposite to the positive direc-
J
a tion, or the direction of increasing z
<en graphically (Figure 3.4.4). For values of a coordinate. The distance traversed in
take a 0, a =
4, and a 8 — = .
the negative direction will always be
3.4.7. Construct the graph of the function
x regarded as negative. In this case,
F(x) = (x) dx where the (possible) graphs
z decreases, z (b) z (a). The general <
J y formula
,

o
of the function y(x) are depicted in Fig- b
ure 3.4.5.
3.4.8. Construct the curves F(x) = z(b) = z (a) Jrz(a, b) = z (a) + v dt ,

X
j
a
(*

\ 9 (x) dx where the functions 9

,
(x) are given b

b or v t
( )
dt = z(b) — z (a)
by the curves which appear in Figures H .6 and ^
a
H.7 at the end of this book. Compare F (x)
with the curves given in Figures 2.5.5 and
2.5.6. remains valid, only in the second form
both sides are negative.
In the case of a velocity that changes
3.5 Properties of Integrals
sign it may happen that, for one thing,
,

b
Above we considered the simplest case
of a definite integral having a positive
j
v (
t) dt = 0, although b >a and
integrand and with the upper limit of a
integration greater than the lower lim- b -=£ a. This occurs if part of the time
it: between a and b the body moves in
b
one direction, and during another time
interval it moves in the opposite di-
z (a, b) = j
v ( t) dt, v (t) >0 and 6>a.
rection, with the result that by time b
a
it returns to the position it occupied at
The here is clearly positive,
integral time a.
since it is equal to the limit of a sum Let us examine the problem of the
of positive terms. The integral has the area under a curve. For b a and >
 110 3 What is an Integral?

ir=ir(t)

less than the lower. In this case we will

no longer speak of the distance, time*
and velocity (cf. Section 3.1) and will
regard the definition of the integral
as a sum (see Section 3.2). Referring
to Figure 3.5.3, we again partition the
Figure 3.5.1 interval between a and b into n parts
by the intermediate values £ x t 2 ... , ,

v (t) >
the integral is equal to the
0, .
., £ n _ x and convince ourselves that
.

area bounded by the curve v u (£), = all the At are now negative. It is now
the t axis, and the vertical lines t a = easy to see that
and t =
b (Figure 3.5.1a). For b a
b
> b a

and v (£) < 0, we know that

j
v dt <Z 0.
j
v (t) dt= — j
v (
t
)
dt ,
(3.5.1)
a a b
In this case the curve lies below the
axis of abscissas (Figure 3.5.16). Thus, since in any partition of the interval
in order to preserve the law by which from a to b the corresponding sums
the area is equal to an integral, it will differ as to signs of the subintervals
is necessary to regard the area as neg- At in all terms.
ative if the curve v v (t) lies below = An
essential property of the integral
the axis of abscissas. consists in the fact that the domain
If we take a function that changes of integration may be divided into parts:
sign, say, v ( t ) =
sin t (Figure 3.5.2), the distance covered in the time inter-
then, as can easily be seen, the area val between a (beginning) and b (end)
under such a curve over an interval may be represented as the sum of the
equal to the period of the function, distances traversed between time a to
that is, from t 0 to t =
2ji, will = c (an intermediate point in time) and
be equal (by our definition) to zero. This between c and b (Figure 3.5.4a):
means that the area under the first
half-wave, which we assume to be pos- b c b
itive, and
negative area under
the v (t) dt= v (t) dt+ v (t) dt •
the second half-wave cancel each other j j j
a a c
3 * 13
exactly.
The definite integral also generalizes (3.5.2)
to the case when the upper limit is
With
the aid of (3.5.1) we can extend
* 13
3
our problem is to find how much
If formula (3.5.2) to the case where c
paint is required to paint over the hatched
lies outside the interval from a to &
portions in Figure 3.5.2, such a definition of
area is no good. In this case we have to parti- instead of inside the interval. Suppose
tion the entire integral into parts over which v
does not change sign (in our case there will be
two parts, from 0 to it and from n to 2jt),
then compute the integral over each portion
H
6
— — — — — —at
tn. tn-1
1 H- 1 I
^3
I I
tf
1
to

separately, and finally add the absolute values

of the integrals of the separate parts. Figure 3.5.3
 )

8.5 Properties of Integrals 11E

a c 6 t a & c t

(*) (6)

Figure 3.5.4

that c b > >a (Figure 3.5.4&). Then,

obviously,
Figure 3.5.5
c o c

j
v (t) dt = ^
v (
t) dt + ^
v {t) dt.
Similarly,
6 c
(3.5.3)

|
v t
( )
dt — z(b) — z(a), f v (
t) dt — z (c)
Transpose the last term to the left and d a
take advantage of (3.5.1). The result b
is

c c c b
— z(a), j
v(t)dt = z(b — z(c),
)

j
v dt — j
v dt = j
v (t) dt + j
v (t) dt
whence
b c b
u
v (t) dt= v (t) dt+ i; (^)
= j
z; dt. (3.5.4) j
a
j
a
j
a

The fundamental law by which the*

In this way we get an equation that
derivative of an integral is equal to*
coincides exactly with (3.5.2).
the integrand refers to the derivative*
We can likewise consider different with respect to the upper limit. If tha
arrangements of numbers a, b and c , ,
definite integral is regarded as a func-
points on the number axis (there are
tion of the lower limit of integration,,
six such variants in all). The reader
with the upper limit held constant, then
will find no difficulty in considering
we get the answer with opposite sign:.
them and in convincing himself that
formula (3.5.2) proves to be valid in
all these cases, that is to say, irrespec-
tive of the mutual arrangement of the
numbers (points) a, 6, and c.
j r -=i( a
i* <'>*)=
a

Actually, we have derived all these (3.5.5)

properties of definite integrals from the
The minus sign in this formula is easy
definition of an integral as the limit
to understand if we regard the integral
of a sum. These properties likewise
as an area: a positive increment A a
follow from the expression for a definite
clearly diminishes the area under the*
integral in terms of an indefinite inte-
curve (Figure 3.5.5). 314 We can form-
gral. Indeed, suppose that the indefinite
ally obtain the same result by inter-
integral v (
t) dt equals z (£). Then changing the limits of integration (this-
^
will introduce a minus sign) and by
b

j
v(t)dt=z(b) —z (a) and
3 14
a The area bounded by the vertical lines.
*

a b t =a+ Aa and t = b, the curve v = v (t),

and the t axis is smaller than the one bounded
v (
t) dt =z (a) — z(b)= — jv t
( )
dt. by the lines t = a and t = b, the curve v —
b a v ( t ), and the t axis.
112 3 What is an Integral?

musing the familiar law on the deriva- 3.6. Examples and Applications
tive with respect to the upper limit:
In Chapter 2 and in the preceding
sections of Chapter 3 we examined the
d_
da (i
dt )=7^(-\ v ^ dt )
relationship between distance and ve-
a b locity and the relationship between
= —v the equation of a curve and the area
(a)
under the curve. These relationships
In connection with the question of are the concrete problems on the basis
the sign of the integral, we note an of which differential and integral cal-
example that frequently disconcerts culus took shape in the history of ma-
;the beginner. Let us consider thematics. But the derivative and the
dx integral are, of course, applicable not
(3.5.6) only to the foregoing problems but to
i x2
an extraordinarily broad range of phe-
This equation follows from the earlier
nomena in the most diverse fields of
;found value of the derivative science, technology, and everyday life.
d( i/x) _ 1 Actually, the derivative, the integral,
dx ~ x* * and the Newton-Leibniz theorem (which
establishes the relationship between
Is the sign of the integral correct here?
these two concepts) constitute the most
‘Can the integral of a positive function,
suitable language describing the laws of
i/x 2 be negative? Does not this sign
,
nature.
^contradict the assertions made above?
Anyone beginning the study of a
Any dismay is due to the fact that foreign language has to repeat all
formula (3.5.6) is not written in proper manner of simple phrases just to get
fashion. If we write it as
used to the new medium: “there is a
table in the room,” “a glass is on the
table,” “there is a cat on the floor,”
“there is a mouse near the cat.” The
then we cannot say that the sign of the same goes for the student of higher math-
integral is always negative, since this
ematics. He, or she, has to repeat the
also depends on the sign and value of
relationships between the derivative
•quantity C.
and the integral in a multitude of sim-
Actually, all statements concerning ilar examples. One must first learn
the sign of the integral have referred
a foreign language before attempting
to the definite integral. Let us take to express ideas in it. And that is our
task now: we have to learn to express
familiar relationships and to formulate
problems in the language of higher
mathematics before trying to solve
(6— a) 315
1
a b
= ab
problems and obtain new results.

3 15 way, “Mathemati-
-
Goethe put it this
’When b > a (and both a and b are pos- cians are a species of Frenchmen; if something
is said to them, they translate it into their own
itive), the integral is positive, as it
language and presto! it is something entirely
should be, that is, formula (3.5.6) is
different.’’ A well-known statement in a simi-
correct and leads to a correct result lar vein was made by the distinguished Ameri-
for the definite integral. can physicist and mathematician Josiah
Looking ahead, we may note that the Willard Gibbs (1839-1903), father of modern
thermodynamics. During a discussion of what
integral l dx/x 2
involves other, no should be given preference to in curricula —
v languages, especially Latin and Greek, or
longer fictitious but real, difficulties, —
mathematics Gibbs said that mathematics is
which will be considered in Section 5.2. a language too. That was his only public
 ) 1 )

3.6 Examples and Applications 113

Here are a few typical examples that Thus, acceleration the second deriva- is
illustrate content of the present
the with respect to time
tive of the coordinate .

chapter. The majority of these will Note the exponents (the “twos”) in
be developed further in subsequent the expression d*z!dt 2 the dimensions :

chapters. of acceleration are those of zft 2 (the

1. Acceleration and uniformly ac- symbol d is not a quantity).
celerated motion; the work performed If we know the acceleration as a
by a force. Earlier, in Chapters 2 and function of time, the instantaneous
3, we constantly considered the dis- velocity v =
v (t) may be written in
placement of a body and the velocity the form of an integral of a ( t =
of motion (as the derivative of the dv Idt :

body’s coordinate with respect to time). t

But after the instantaneous velocity v =v + a (0 dt. (3.6.3)

v (t =
dz ( t)/dt has been found and we
(*) o
j
to
know the relationship between the
instantaneous velocity and time, we (Here v 0 =
v (t 0 ) is the instantaneous
can ask how the velocity varies with velocity at the initial moment of time
time. In brief this question has been t =t0; to verify this we need only
studied in Section 2.7. substitute t 0 for t on both sides of Eq.
The derivative of velocity with re- (3.6.3).) For one thing, if the accelera-
spect to time is called the acceleration tion is constant (a constant, the =
and is denoted by a: case of uniformly accelerated motion),
from (3.6.3) it follows that
~ff~ =a (0 (3-6.1) t

Since the dimensions of velocity are v(t) = v0 + a dt = v0 + a(t— 0 ),

j
cm/s or m/s, the dimensions of acce-
leration are cm/s 2 or m/s 2 . (3.6.4)
We write the velocity in the form of or
the derivative of the coordinate with v(t)= at + b, (3.6.4a)
respect to time, v =
dzldt and substi-
where b = v —
,

tute this into Eq. (3.6.1). The result is 0 at 0 that is, the ve- ,

locity of uniformly accelerated motion

is a linear function of time.
Knowing the velocity v v (t) of a =
statement the more remarkable!) on the
(all motion, we can establish how the coor-
general aspects of teaching. Almost the same dinate varies with time, that is, find
thing was stated much earlier by the famous
Galileo Galilei (1564-1642), who declared
the law of motion z z (t): =
that “the laws of Nature are recorded in the t

greatest of all books, which is always open to

our gaze (I refer to the Universe), but you
z(t) =z + 0 v(t)dt, (3.6.5)
j
cannot understand it unless you first learn to to

understand the language in which it is written,

and it is written in the language of mathema- or, in view of (3.6.3),
tics.” (And, rather surprisingly, he goes on to t t
say that “its letters fi.e. the letters of mathe-
—
matics Ya.Z. and I.Ya.] are triangles, cir- z (t) = Zq-\- -f’
j* a (t) dt dt
j J
cles, and other geometric figures without to to
which its words cannot be understood.” Thus
t t
Galileo regarded geometry and not differential
and integral calculus as the language of Natu- —z 0 -\-v0 (t — +J t 0)
£J
a t
( ) dt dt ,
re. But we must bear in mind that in Galilei’s j
time that is all there was to mathematics the — to to

geometry of the ancient Greeks. Naturally, (3.6.5a)

Galileo could not foresee the creation of higher
mathematics by Leibniz and Newton, who where z0 =
z (i 0) is the initial posi-
came later.) tion (coordinate) at time t 0 (substitute
8-0946
 g g

114 3 What is an Integral ?

t 0 for t on both sides of (3.6.5)). Of

course, in general form, formula (3.6.5a)
looks unwieldy. But if a constant, =
then the velocity v v (t) is given by — $
a simple formula (3.6.4a) or (3.6.4),
Figure 3.6.1
and Eq. (3.5.5a) enables us to describe
the motion of a body in closed form:
Figure 3.6.1). However, this fact can
t be employed only when the force is
z (
t) =z 0 -f-
j* (at + 6) dt — z 0
-\- a ^
" 1 constant, a situation that almost al-
to
ways is unrealistic. If the force F var-
ies in the course of motion, then the
+ b(t—t = ±t + bt + c, 0)
2
(3.6.6) whole path, or distance, from z z0 =
to another (variable) position of the
body, or coordinate, z has to be par-
where c —(a/2) f0= bt 0 z0 — + titioned into small intervals of length
(check to see whether the derivative
A z. The products FAz, obtained on the
of the right-hand side of (3.6.6) with
assumption that the force in each in-
respect to t coincides with the right-
terval remains constant (since the A z
hand side of (3.6.4a)). Thus, the de-
are small, F does not have time to
pendence of the coordinate z of a body
vary appreciably), must then be added.
in uniformly accelerated motion on time
In this way we arrive at the “integral
is quadratic. For example, in the case of
sum” 2
FAz, in which in order to ob-
motion under the action of gravity
tain an expression for A we must pass
(free fall) we have a where = —
g ~9.8 m/s 2 (the minus sign being
,
to the limit as all the A z tend to zero.

due to the fact that the upward direct- The lastcircumstance makes it possible
ion is taken to be positive). Assuming to assume that v =
dzldt is constant on
a =— in (3.6.4), (3.6.4a), and each subinterval, with the result that
the length A z can be expressed by the
(3.6.6), we get the well-known formu-
las
product of v by the time At which it
*
takes to traverse A z. Hence, the sum

v(t) = v — 0 gdt = v 0 —g(t — t0 ) S Ft* is replaced with the sum

to
j
2 FvAt, which enables representing
the work A either in the form of an
= v0 —gt t, integral with coordinate z as the vari-
where =t— t0 is the new time able of integration or in the form of an
variable reckoned from t = t0 , and integral with time t as the variable of
integration:
t

—g(t — t 0 )]dt
z t
z(t) = z0 + [v 0 /r
\ %/ W--J* Fv dLv. \£. J>, ;
to
Zo to
tl

=z + o (Vo — gti) dti Now we recall that the force F is

j equal to the mass of the body, m, times
0
the acceleration a dvldt (Newton’s =
= z 0 + y0 *i—frfj. second law). This enables us to rewrite
(3.6.7) thus:
In connection with acceleration we *

recall that the work performed by a

A=m ~j-vdt. (3.6.8)
force is equal to force F times distance s ^
to
(here we assume that the distance or
displacement coincides with the di- Since, obviously, the function v (dvldt)
2
rection in which the force acts; see is the first derivative of v /2 (see Sec-
 3.6 Examples and Applications 115

tion 2.3), we finally obtain

A=m (i T1 ' < 3 6 9'

- -

where v = v (z) and v 0 = v ( z 0 ): the

work A done by a force on a body is
equal to the increment mv2 I 2 — mv\!2
of the kinetic energy mv2 / 2 of the body
(of course, both the work A and the
increment of the kinetic energy may
prove to be negative simultaneously).
For a further detailed investigation
of the questions touched on here, see
Chapters 9 and 10.
2. Stretching of a wire. Consider a the free end B) into small sections of
rod one meter long, 0.4 cm in diameter, length Az each; for each such section
and made of copper. This piece of wire at a distance z from B the force may ,

is hung by one of its ends. The ques- be assumed constant and equal to the
tion is: how much will the rod stretch weight of the part of the rod from point
under its own weight? The stretching of B upward (the4 length of this part is
the wire is governed by Hooke's law 313 z): F —
p gSz, where p is the density
which is valid for not too large forces of copper, g the acceleration of grav-
(and elongations) and states that the ity, and Sz =
V the voluble of the
deformation (and stretching) l is di- part of the rod under consideration.
rectly proportional to the applied force Thus, to this small section of the wire
F and the length of the wire L and is , ,
of length A z there corresponds an
inversely proportional to the cross-sec- elongation
tional area S:
1 FL AZ = — as .££. zAz, (3.6.11)
l (3.6.10)
E S ’

and the total elongation of the entire

where £ is a constant factor depending wire is obtained by summation (or
on the material of the wire and known integration) of all such “elementary”
as Young's modulus . elongations:
Naturally, we cannot replace L with 100
the length of the rod, 1 (100 cm), m
and F with the weight of the rod, which
0
can easily be found from the volume of
2
the rod and the density of copper, pg (100)
(3.6.11a)
since the forces acting on different por- E 2

tions of the rod differ from point to Substituting the numerical values
point: while for point A at which the =980 cm/s 2 p = 8.9 g/cm3 E =
g , ,

rod is suspended this force is indeed 9.8 X 10 13 g/cm-s 2 we obtain

,

equal to the weight of the entire rod,

for the other (free) end B this force is 980X8.9X10
2X9.8X10 13 -J4.5X10’ 6 cm
simply zero, since this point is under
no load (Figure 3.6.2). For this reason =
(note that the factor 100 2 10 4 in
we must partition the entire wire the numerator has the dimensions of
(with coordinate z being reckoned from
cm 2 ).
3. Water flow from a vessel. Picture
3 16
-
Robert Hooke (1635-1703), an out-
a vessel of arbitrary shape (Figure 3.6.3)
standing English scientist, chemist, and mathe-
matician and a contemporary (and in many with water flowing out. The mass of
cases a scientific opponent) of Sir Isaac Newton. liquid in the vessel at a given instant
8*
116 3 What is an Integral?

have
t

m (t) = j
W (t) dr ,

to

(3.6.13a)
t

M t
( )
=M (t 0 ) -jw(r) dr.
to

Ordinarily, this is simply written as

Figure 3.6.3 t

m(t) — j
W (
t) dt 9

of time is equal to M
which is a func- ,
to

tion of time: M=M

(it )• The liquid
t
(3.6.13b)
collects in another vessel; the mass of
liquid here is (£). m
denote the We M(t) = M (t0 - )
W(t)dt,
Jf
mass of liquid flowing out of the first U
vessel in unit time by (£); W— W but remember that the t under the
this quantity is known as the flow rate integral sign has a different meaning
and has the dimensions of g/s. The from the independent variable t in
quantities m, M, and are connected W M
(£) and m ( t ), which coincides with
by the following relations: the upper limit of integration. In this
respect, the notation (3.6.13) and
- W(t), -^=W(t). (3.6.12) (3.6.13a) is more exact than (3.6.13b).
The formulas given above correspond
These differential, that is, involving to an experiment in which and flow M
derivatives, relations can be written rate W
are measured at distinct mo-
in the form of integrals. Suppose that ments of time. Often, however, the
at a certain initial moment t 0 the mass problem is stated differently: the flow
of the liquid in the first vessel was rate W
depends in some known fashion
M (t 0 ) =
Af 0 while the second vessel
,
on the pressure, that is, the height of
was empty, or (t 0 ) 0. Then m = a column of liquid, h In turn, given .

a definite shape of vessel, the function

h is a function of M. Thus, we know
m (t = W (t) dt , the flow rate if as a function of the
t)
j quantity of liquid in the vessel, W=
(3.6.13)
W M ( ). Then (3.6.12) takes the form
u ™L= -W(M). (3.6.14)
M (t t ) =M (t 0 ) -j W(t)dt.
Equations of such a type, which con-
to
nect the unknown function (in our case
Thus, the mass of liquid at a definite the function M =
(£)) with its der- M
time t x is expressed in terms of an in- ivative are called differential equations ;
tegral in which the variable of inte- thus, Eq. (3.6.14) is a differential equa-
gration t runs through all values from
t0 to t.

Itwill be convenient in (3.6.13)

to replace t x with t and rename the
variable of integration (using the fact
that it is a dummy variable) and call
it, say, tau (which is the Greek letter
corresponding to the Latin t). We then Figure 3.6.4
S.6 Examples and Applications 117

tion for the unknown function (t), M

that is, a function that will be found
when we solve the equation. The topic
of differential equations, as well as
the solution to this specific problem,
will be considered in Chapter 6.
4. Capacitor. denote the charge We
accumulated on a capacitor C (Fig-
ure 3.6.4) (the quantity of electricity)
by q In the SI system of units (see
.

Appendix 5), q is measured in coulombs

(abbreviated C). The electric current
; flowing
in the wire is the quantity Figure 3.6.5
of electricity flowing in unit time and
is measured in amperes (abbreviated A).
One ampere is a current of one coulomb Problems involving capacitors are dis-
per second: 1 A 1 C/s; thus, the = cussed in detail in Part 2 of this book.
quantity of electricity has the dimen- 5. Atmospheric pressure. Imagine a
sions of A* s, since in the SI system of vertical column of air with constant
units the ampere is a base unit. cross-sectional area S cm 2 The density .

The charge on a capacitor 3 17 and the of the air is p g/cm

3
and depends on
current are connected by the equation the altitude h above the earth’s surface.
The volume of a thin layer contained be-
tween h and h + Ah is equal to SAk
(Figure 3.6.5). Inside this thin layer*
(the positive direction of the current is
the density p {h) may be regarded as
indicated by the arrow in Figure 3.6.4).
constant, which is precisely why we*
If the variation in current flow with
took a thin layer. In the given instance,
time, ; / (£), =
is given or has been
Ah may be pictured as 1 meter or
found via experiment, then we can 10 meters or even (with a slightly lower
write the integral relation
accuracy) as 100 meters, since air den-
ti
sity varies roughly by 12 to 14% per
q (t t ) = q (t0 + )
j
/ (t) dt. (3.6.15a) kilometer of altitude.
to Since the volume of the layer of air
If the capacitance C of the capacitor Ah thick is S Ah, the mass of air con-
isgiven (note that the capacitance and tained in this layer, by the definition
the capacitor are designated in Figure of density, is Am = pSAh. To find
3.6.4 by the same letter), then the the mass of air in a column extending
voltage drop across the capacitor may from h x to h2 ,
we must construct the
be expressed in terms of q in the fol- sum of expressions of the type pSAh r

lowing manner: cp c qlC, and the = with S constant and p varying with
voltage drop across the resistance R is altitude h The sum is extended over
.

(pi?
=E — 0 cpc =E — 0 q!C, where E0 all Ah into which the column
layers
is the battery voltage. By Ohm’s law, of air partitioned. Decreasing all
is
the current flowing through resistance Ah without limit (that is, sending them
R is / i?" 1 (E 0=?/C). Using (3.6.15), — to zero), we arrive at the integral
we arrive at the differential equation
ht

7T=Tr( E°-i)- < 3A16 > m (h^ h2) = S j p» dh. (3.6.17)

hi
3 .i7 \y e use the capaci-
term charge on
tor to mean the quantity
of positive electricity
on the left-hand plate of capacitor C in Fig- The mass of air in a column from the
ure 3.6.4 expressed in coulombs. earth’s surface (h 0) to an altitude=
118 3 What is an Integral?

In is

m (0, h) =S p (h)dh. (3.6.17a)

j
0

The mass of air above a given altitude

h is
oo

m=S ^
p(h)dh, (3.6.17b)
h

"where the symbol oo in the upper limit Figure 3.6.6

takes the place of a very large number
H such that any subsequent increase
problem we will employ a system of
in this quantity does not substantially
coordinates in space specified by the
-change the integral.
x axis and the y axis in the horizontal
The pressure P at some altitude h plane xOy and by the vertical z axis
multiplied by the area S is equal to
(the third coordinate axis in space).
the force with which the entire column
The volume AF of a thin layer bounded
of air above h is attracted to the earth.
by the planes corresponding to the
The force of gravity is equal to the
mass multiplied by the acceleration
values z and z +
Az on the z axis and
318 parallel to the xOy plane is approximate-
of gravity g, whence
ly equal to the product of the cross-
oo
sectional area S (z) by the thickness
P (h) — gp{h)dh . (3.6.18) of the layer, A z. Thus, if we know the
j
h area of a cross section of the solid

we get the
cut by a horizontal plane, S S (z),=
Using formula (3.5.5),
then the volume of the solid can be
differential equation
computed by the formula

1 wifi). (3.6.19) Zi

F=jS(z)dz, (3.6.20)
This formula could have been written z0
straight off by considering the equi-
where z 0 and z 1 are the smallest and
librium of a thin layer dh acted upon
greatest values of the z coordinate in
from below by the pressure P (h) and
our solid.
from above by the pressure P (h + dh);
Let us apply this formula to a reg-
the resultant of these two forces bal-
ular quadrangular pyramid with its
ances the attraction to the earth of
the mass of air in the layer Ah.
We will continue on this subject in
Chapter 11.
6. Volume. Let us employ integral
•calculus to compute the volume of
bodies (solids). Let us slice a solid of
total volume V (Figure 3.6.6) by planes
z —
constant into thin layers. In this

3 18 In the majority
-
of cases we can assume
that the acceleration of gravity g does not
alter over the altitudes h considered in this
problem and that the earth has a flat surface,
although both assumptions do not in any
way change our arguments. Figure
 .

3.6 Examples and Applications 119

vertex at the origin of coordinates and

with its axis of symmetry directed along
the z axis (Figure 3.6.7). Let the al-
titude of the pyramid be h and the base
(at the top of the figure) a square with
side a. From geometry we recall that
a cross section of a pyramid cut by a
horizontal plane at an altitude z is a
square, the side b of which is to a as
z is to h that is, b ,
b (z) a (zlh). = =
Hence, the cross-sectional area is
S (z) = b2 = (a /h
2 2
)
z
2
. The volume of
the pyramid is

h Figure 3.6.8
a2
V= z 2 dz.
H2 j S (x) = (1/2) MP X PQ— (1/2) (1 x2) —
o Whence, in view of formula (3.6.20), where
now we must write x instead of z, the volume
Let us take advantage of the result of of the “hoof” is
Section 3.4:
1 1

v=
Y — x2 ) dx — (1 — z 2 ) dx
z2 dz — z 2 dz = h3 j
-1
(1
J
-1
* \ O

We -
ume
then get an expression for the vol-
of a pyramid: ~K -i
J
dx
S
-i
x* dx
]

Vv = ——h
h2 3
3 =
3
a 2h

We found that the volume of a pyramid

is equal to one third the product of the =t[<»-<-*»-(-t f)]
area of the base by the altitude of the
pyramid To derive this formula in
.

solid geometry without the aid of in-

-U«-4K
tegrals is a rather complicated job. ( here we employed the obvious properties
b b
Wewill return to the question of the
\

use of integral calculus for the com- of the definite integral; cfdx—c / dx ,

J J
putation of volumes of solids in Chap- a a
b
ter 7.
where c is a number, and
J
(/ + g) dx =
Here is another example, and it is much a
more striking than the preceding examples. b b
We wish to find the volume of a cylindrical
“hoof ”, that is, the part cut off from a (right f d*+ g dx'j .
j j
circular) cylinder by a plane that
intersects the a a
cylinder along the diameter of the cyl- AB Note that the volume of this “round” solid
inder’s base and forms an angle of 45° with is and in no way is con-
a rational fraction
the base (Figure 3.6.8). For the sake of simplic- nected with the number n.
ity we take the radius of the cylinder equal
to unity (cf. Exercise 3.6.3). The cross section
of the “hoof” cut by a plane perpendicular to Exercises
diameter AB
and distant x from the OM =
center 0 of the base is a right isosceles tri- 3.6.1. Derive the formula for the volume
angle (the right triangle with acute MPQ of an arbitrary pyramid using the] properties
angle Z OMP = 45°) in which the length of of parallel sections.

side MP is V OP — OM = Vi — x But
2 2 2
.
3.6.2. Find the volume of a (right circular)
cone of altitude h and with a base that is a
then PQ “ PM = ]/T — x and S A MpQ — 2
circle of radius R. Will the result change if
 120 3 What is an Integral?

the (circular) cone is oblique, that is, if its surmountable, at least via approximate
vertex does not lie exactly over the center of procedures if not by precise methods.
the base
If the last three sections of this chap-
3.6.3. Find the volume of a “hoof” cut off
from a cylinder of radius R by a plane pas- ter have appeared to be difficult, the
sing through the diameter of the cylinder’s reader will do well to reread Chapters
base and forming a known angle a with the 2 and 3.
base.
3.6.4.Use the formula for the volume The rise of “higher mathematics”, that is,
of cylindrical “hoof” and evaluate the inte- differential and integral calculus, was a turn-
1
ing point in the history of civilization: it
gral y Vi — y 2 dy. gave man a powerful tool for analyzing pro-
j cesses of many different kinds, for developing a
o
fundamental explanation of physical phenome-
na, and for constructing a scientific picture of
* * * the world. Actually, it was the thinkers of an-
cient Greece, primarily the genius of Archi-
In Chapters 2 and 3 we considered the medes of Syracuse (287-212 B.G.), who skill-
fully solved the first problems of differential
concepts of the derivative and the in- and integral calculus. Archimedes successful-
tegral, some of their simpler proper- ly applied mathematics in designing machi-
ties, and the relationship between the nes and devices (the weapons of war which he-
integral and the derivative; in other invented struck terror into the Romans laying
siege to his native Syracuse). Archimedes
words, we introduced the framework in knew, for example, how to draw tangents to
which the concepts and theorems of curves and how to calculate an area bounded
the so-called higher mathematics must by curves. He solved the problem of drawing a
be developed. The techniques involved tangent to what came to be called the Archi-
medes spiral, that is, the line described by a
in practical computation of derivatives snail crawling steadily along the spoke of a
and integrals of various functions will wheel turning at a constant rate, and also the
be discussed in Chapters 4 and 5. Only roblem of squaring the parabola, that is,
nding the area of a segment of a parabola.
the simplest examples were illustrated
Today problems of this kind can be solved'
in Chapters 2 and 3. without difficulty by any college student (or
A word of warning to the reader: even a student of higher school), but in
do not get into the habit of measuring those remote times they were within the pow-
ers of only a giant like Archimedes. There
the significance (and even the difficulty)
was no general method for solving such prob-
of any section of mathematics by the lems. Each one required great effort.
number of formulas, their complexity, For that matter, the primitive level of tech-
and unwieldiness. Actually, the most nology in ancient Greece did not require the
solution of problems that called for great in-
important thing and the most difficult
ventiveness but were devoid of applications
thing is the mathematical statement outside mathematics: as a rule, the thinkers of
(or formulation) of a problem in the ancient times regarded mathematics not as an
form of an algebraic equation or an effective method of solving practical problems
but only as a theoretical science whose perfec-
integral or even a differential equa-
tion reflected the profound harmony of the
tion, and not the transformations that world, yet only explained the world in a pu-
follow. That is where our attention rely philosophical sense. (Archimedes was
should be focused. practically the only exception in this respect
Almost every physicist knows that because he closely linked mathematics with
mechanics and physics. But he was a genius; in
those pieces of research that he did not this, as in many other respects, he was far
succeed in completing (and which other ahead of his time.)
workers did complete) were left undone The static nature of life in ancient Greece,,
because he confined himself to a gen- where hardly any machines were known and
life centered around city squares, palaces, and
eral reflection and did not find the temples decorated with beautiful statues as
courage to write down the equations immobile as the temples and city squares*
and formulate the problem mathemat- themselves, gave rise to metaphysical think-
ing (rigidity) that was so characteristic of
ically. The computational difficulties
the mathematics of antiquity: it was not
in a properly posed problem with a customary to regard processes in flux. They
clear-cut physical content are always limited themselves to unchanging “states,” &
 -

3 What is an Integral? 121

reflection oxwhich were, for the famous geo- whose discoveries played such a big role ins
meter Euclid of Alexandria (early third centu- building a scientific picture of the world, have-
ry B.C.), merely chains of congruent triangles, any such method. Kepler was indisputably the-
which occurred frequently in his arguments. leading master of integration in his day
With the flowering of Italian cities in the (though it was not called that yet). In 1614,
15th and 16th centuries and the appearance of when Kepler married a second time, he had
the first factories, which heralded the imminent occasion to buy a good deal of wine for the
rise of machine-based production, the static wedding reception, and he saw how difficult
metaphysical thinking of the ancients became it was to estimate the cubical contents of wine-
inconceivable. In this period it could only casks of different shapes from the base radius
hinder much needed scientific progress. The and the height. The problem intrigued him.
great Galileo Galilei (1564-1642) was the first The following year, 1615, he published his-
to proclaim publicly that mathematics provid- Nova Stereometria Doliorum Vinariorum (The
ed the key to the mysteries of the Universe New Science of Measuring Volumes of Wine
(see footnote 3.15). Under Galileo’s influence, Casks), “with addenda to Archimedean stereo-
his pupils— Evangelista Torricelli (1608-1647) metry,” as he also informed the reader. Here he-
who discovered the principle of the barome- collected a large number of problems in de-
ter, and the geometrician Bonaventura Ca termining the volume of bodies limited by
valieri (1598-1647) who continued the work of curved surfaces and displayed great ingenuity
Archimedes, for whom their teacher had such in developing formulas for such volumes, which
great affection— solved many specific pro- are today established by integral calculus (see*
blems which today lie within the province of Section 7.10).
higher mathematics. For one, Cavalieri evol- Rene Descartes (1596-1650) was truly the-
ved a method of calculating volumes that is forerunner of higher mathematics. Soldier,
very similar to those described above. Galileo diplomat, natural scientist, and abstract think-
was the first to see that the problem of deter- er, he was the father of analytic geometry (the
mining the path of an object from its velocity method of coordinates in geometry; see Chap-
practically coincides with the problem that ter 1) and a profound philosopher who decla-
had so interested Archimedes, that of deter- red that the world was knowable. He stated
mining the area of curved figures. Continuing the basic principles of dialectics and the place
Galileo’s work, Torricelli established that the occupied in life by various processes whose*
inverse problem, that of finding the velocity study, he believed, constituted the chief aim
from the path was akin to the problem of dra-
,
of mathematics. Descartes substantially im-
wing a tangent to a curve But at that time
. proved the symbols and language of mathemat-
there did not yet exist general methods of solv- ics, giving them their present-day appear-
ing such problems; there was no common al- ance. This greatly stimulated further prog-
gorithm to enable one “to calculate without ress and democratization of mathematical!
thinking.” knowledge.
Nor did Johann Kepler (1571-1630), that Another French scientist, Pierre Fermat
outstanding astronomer and mathematician (1601-1665), a lawyer by profession and a pro-
 T22 3 What is an Integral?

Ttene Descartes
Pierre Fermat

[found amateur mathematician, was a con-

temporary and, in a way, a rival of Descartes. lem “in the old way.” (True, but you had to
Working independently of Descartes, Fermat have the brains of Huygens to do that.)
^somewhat earlier elaborated (though it was Two outstanding thinkers of the 17th cen-
published later) a system of using, in geo- tury, the Englishman Sir Isaac Newton (1643-
imetry, the methods of coordinates and alge- 1727) and the German Gottfried Leibniz (1646-
braic computations now known as analytic 1716), are rightly regarded as the real found-
^geometry. Fermat’s exposition of the subject ers of higher mathematics. Both indisputab-
was perhaps closer to the modern one than ly rank among the most profound scientists
that of Descartes; and the fact that the writ- that the world has ever known. To them we
ings of Descartes have come to hold a much owe a coherent exposition of the new calculus,
more significant place in the history of science a chain of formulas for finding the derivative
is connected primarily with his active teaching of any given algebraic function without any
and his more advanced system of notation. difficulty and a complete understanding of the
{The situation was much the same in the devel- connection between the derivative and the in-
opment of differential and integral calculus.) tegral and the significance of that connection,
’Simultaneously, Fermat occupied himself with which provides a general algorithm for cal-
problems that are now part of mathematical culating integrals by turning to a list of deriv-
analysis. He evolved the concept of the differ- atives (see Chapters 4 and 5).
ential (see Section 4.1 below) and the overall Leibniz was a truly versatile scientist: he
idea that the maxima and minima of a (smooth) delved into philosophy, philology, history,
function must lie at the points where the (first) psychology (in which he was one of the pio-
derivative of the function vanishes, and also neers of penetration into the sphere of the
carried out ingenious calculations of the values subconscious), biology (he was one of the pre-
of some integrals (e.g., the method of deducing cursors of the theory of evolution), geology,
formula (5.2.1) indicated in Exercise 3.2.3). mining, mathematics, and mechanics (he evol-
Descartes, too, made contributions to the field ved the concept of “living force,” that is, kinet-
*of analysis. ic energy). At the same time he was active in
Christian Huygens (1629-1695) of the Neth- politics and diplomacy (he strove to reconcile
erlands (a junior contemporary of Descartes the German principalities because he foresaw
and Fermat) created the wave theory of their eventual union into a single state, and
light and was noted for his work in the appli- dreamed of uniting the Catholic and Protestant
cation of mathematics to mechanics and phys- churches). He was also an organizer of scien-
ics (e.g., he produced the strict mathematical tific academies; for one, he was the founder
theory of pendulum clocks (see Section 7.9)). and first president of the Prussian Academy of
In his investigations, however, Huygens used Sciences in Berlin, and he had several talks
the archaic methods of thinkers of anti- with Peter the Great of Russia about found-
quity, Archimedes for one, because he believed ing a Russian academy of sciences, which
that a newer method would yield no advan- was later established in full conformity with
tages inasmuch as he could solve any prob- his suggestions and wishes. The mathematical
3 What is an Integral? 123

Christian Huygens Isaac Newton

analysis of Leibniz [to whom we owe, among ter 4); use of this method provides a recipe for
other things, the modern terms “derivative” calculating derivatives. Leibniz used the mod-
(Ableitung in German) and “integral,” 3 19 ] was
-

ern symbol / (x) dx to designate the inte-

j
of a form much like that adopted in our book.
Leibniz designated the (first) derivative of a
gral of the function y = f (x).
Leibniz was a tireless teacher, and that
function y =
/ (x) by the symbol dyidx\ he plus his felicitous system of notation and terms
understood the “differentials” dy and dx as
resulted in the universal acceptance of higher
the “limiting” values of the increments A y
mathematics in the form he developed it.
and Ax at which we arrive by an unbounded
Leibniz is also considered a classic in the
reduction of Ax (neither the word “limit”
field of philosophy (here he did much to
nor its concept existed in the mathematics of
extend and to amplify the work done by
Leibniz). This approach corresponded to in-
Descartes). He also deserves credit for many
troducing the concept of the derivative as the
profound ideas that partially became reality
tangent of the angle formed by the tangent
only in the 19th and 20th centuries: e.g., the
line to the graph of the function and the x
idea of a “geometrical calculus,” from which
axis: to a small Ax there corresponds a small
the modern vector calculus later arose; his
triangle MNP
(see Figure 7.9.1), where
attempts to “algorithmize thinking,” in which,
tan a x =
NP/PM =
Ay/Ax\ when Ax is re-
as Leibniz wrote, the two sides in a contro-
duced without limit, the increments Ax and Ay
versy would no longer have to conduct lengthy
are replaced by the differentials dx and dy ,
debates inasmuch as one of them could always
and the secant MN
of length ds (the differen-
say to the other, “Well, let us verify which of
tial of the arc length; see Section 7.9) is re-
placed by the appropriate tangent line (Leibniz
us is right, let us calculate, my dear sir.”
The rough outlines of this kind of “proposition-
-called such a triangle the characteristic tri-
al calculus,” found in the papers of Leibniz,
angle).
closely resemble the mathematical logic of
Leibniz emphasized in every possible way
the 19th and 20th centuries.
the algorithmic aspect of the new calculus
Newton arrived at the same ideas as Leib-
and the system of rules that automatically
niz quite independently and even somewhat
guarantee a correct result when seeking deriva-
earlier. He was perhaps more of a physicist
tives. He also worked out a method of handl-
and astronomer than a mathematician; his
ing differentials (dealt with here in Chap-
contributions to the birth of physics and to
the rise of a new method in natural science
3 * 19
At first Leibniz simply spoke of the cannot be exaggerated. Joseph Louis Lagrange
sum (and also “summational calculus” instead (1736-1813), the distinguished 18th and 19th
of “integral calculus”); he enthusiastically century mathematician and investigator in
adopted the terms integral (from the Latin the field of mechanics (we will discuss his
integer meaning whole) and integral calculus
,
work later on) once said of Newton: “He is
proposed by his pupils, the brothers Jakob the most fortunate of men, for the system of
and Johann Bernoulli. the world can be discovered only once.” The
 124 3 What is an Integral?

These are obviously the problems of finding

the derivative from the given function, and of
finding the function from its derivative, that
is, the problem of calculating an indefinite
integral. The fact that these two problems (in
their geometrical presentation, the drawing of
a tangent to a given curve and the calculation
of the area hounded by the curve) are inverses
of each other had evidently been discovered
first by Newton’s teacher, Isaac Harrow? (1630-
1677), who subsequently resigned his post as
Lucasian Professor of Mathematics at Cam-
bridge (a rare case in the history of science!) in
favor of his brilliant pupil because he felt
that Newton was more worthy of the post
than he. However, it is not at all by accident
that the connection between derivatives and
integrals is named after Newton and Leibniz
instead of after Barrow, for only these two
great scientists realized the full extent of the
mutually inverse nature of the operations of
differentiation and integration, which had
been discovered by Barrow (and also, inde-
pendently, by Leibniz), and the possibility
Gottfried Leibniz of making this fact the foundation for a broad
calculation of derivatives and integrals. It is
same idea conveyed in the famous “Epitaph
is to the far-sightedness of Newton and Leibniz
for Sir IsaacNewton” written by his contem- that mathematical analysis owes the rapid
porary Alexander Pope (1688-1744): progress that began immediately after the
first publications and statements by these two
Nature, and Nature’s laws lay hid in reat scientists and which truly marked the
night: God said, Let Newton be\ and all awn of a new era, the era of great scientific
was light. discoveries, the era of a sweeping assault on
Nature’s secrets to promote human welfare.
Newton,of course, identified the derivative Priority in applying differential and inte-
with velocity: he regarded its properties as gral calculus to fathom Nature’s secrets un-
the physical properties of velocity. Yet the doubtedly goes to Isaac Newton, who put
formal mathematical theory of derivatives (and forward the general idea that the laws of Na-
also integrals), founded on a variation of the ture must have the form of differential equations-
theory of limits, was not alien to him either. linking the functions that describe the pheno-
However, in the absence of a definition of the menon under study. We owe to Newton the
term limit, the theory did not make his con- formulation of the basic equations of motion
structions more flawless than the (logically not (for more details see Chapters 9 and 10),
indisputable) manipulations of Leibniz with which he then applied to deduce the law of
infinitesimals, but only made them more un- gravitation and to study the motion of cele-
wieldy; also the fact that Leibniz clearly stial bodies and the fairly complicated pro-
exerted a greater influence on European mathe- blems of celestial mechanics (see Lagrange’s
matics than Newton may be connected with above-cited assessment of Newton). In effect,
this. Newton’s work gave birth to a theoretical
Newton gave the name of fluxion to the de- physics based on the use of a profound mathe-
rivative, and the initial function from which matical apparatus, which physics, in its sub-
the derivative was calculated was called the stantive part, has gone far beyond the limits
fluent (from the Latin fluere to flow), empha-
, of the problems which Newton grappled with
sizing that the quantities under consideration and, in its mathematical part, rests on the
were variable; the fluxion arose as the rate of whole body of modern science, many times
change of the fluent, while the fluent was re- surpassing the differential and integral calcu-
stored from the fluxion as the path from the lus created by Newton and Leibniz.
speed. Newton began his exposition of the Newton designated a fluxion (derivative)
analysis with two basic problems, to which by a dot placed above the letter symbolizing
all the others are reduced.
the initial function. For example, he wrote
1. Knowing the length of the path travers-
ed, find the speed of motion over a fixed time the derivative of the function y == f (x) as y.
interval. (This notation is retained today in mechanics,
2. Knowing the speed of motion, find the but we will not use it.) Newton proposed des-
ength of the path traversed over a fixed time ignating the operation of a transition from a
nterval. fluxion to a fluent by a dot placed underneath:
3 What is an Integral? 125

thus, if y = x 2 and y 1 = 2x, then, inthissys- German or French scientists. It seems to us,
however, that the disagreement between New-
tem of notation, y = y and y = y. The
1 l
ton and Leibniz was not accidental and deser-
symmetry of these designations makes them ves a more detailed examination of its causes.
attractive. The theorem of the connection The fact that Newton and Leibniz simul-
between the derivative and the integral can taneously and, beyond dispute, independently
discovered differential and integral calculus
be written as follows: y = y, where the expres-
best of all demonstrates the timeliness and
sion y can be understood in two ways: as the inevitability of the great scientific revolution
fluxion of the function y and as the fluent of of the 17th century. Moreover, the different
(even opposite) psychic make-up and scientif-
the function y. But today this appealing nota-
ic programs of these two outstanding scien-
tion has only historical significance (besides,
tists, which shaped the pathways by which
Newton himself was not particularly consistent
they arrived at their discoveries, make a
in using it). The designation y ' for the deriva-
comparison of their investigations highly edi-
tive of a function was introduced by the
fying. 3 21 Leibniz, the philosopher, was large-
-

French mathematician Augustin Cauchy (1789-

ly guided by what he called the “metaphysics
1857), whom we will be hearing more about
of infinitesimals” (differentials; incidentally,
later.
this concept, which came from Fermat, appea-
One of the regrettable episodes in the hi-
red in Newton’s works under another name
story of science was the bitter controversy be-
before it did in those of Leibniz). Leibniz was
tween Newton and Leibniz, conducted chiefly
carried onwards by the language he had cre-
by their admirers and pupils and not by the
ated, by the developed symbols and terminol-
two outstanding scientists themselves. 3 * 20 It
was a controversy that brought neither honour ogy characteristic of the new calculus. (We
remind our readers once again of Leibniz’s
nor advantage to either side. Leibniz, accused
great interest in the science of language: he
(quite groundlessly, as we know today) of di-
was one of the founders of what is now known
rectly borrowing his ideas from Newton, lost
as comparative-historical linguistics.)
the patronage of the dukes of Hanover, who
Moreover, the definition (later included in
had long supported him, and died in poverty
all textbooks) of infinitesimals as variables
and oblivion. His death was not reported in
whose limit is equal to zero was undoubtedly
a single German (to say nothing of an English)
alien to Leibniz. It corresponded much more
publication; and even the Berlin (Prussian)
closely to the scientific thinking of Newton.
Academy, which he founded, took no notice of
Leibniz regarded infinitesimals as “special
his death. Only the French Academy, of
numbers,” the rules for using which completely
which Leibniz had been an active member,
differed from those governing operations car-
paid tribute to him in a eulogy. On the other
hand, the refusal to recognize Leibniz (and 3 21 These
-
differences in approach and
also the refusal to recognize his differential
viewpoint make, we believe, the whole contro-
and integral calculus) substantially hindered
versy over priority quite meaningless, since
the progress of English science and completely
separated English pure and applied mathemat-
now it can be said that the inventions of New-
ton and Leibniz were entirely different (e.g.,
ics from continental mathematics: English
see A.R. Hall, Philosophers at War: The Quar-
university graduates were not familiar with the
rel Between Newton and Leibniz, Cambridge
terms “derivative” and “integral” or with the
University Press, Cambridge, 1980).
notation introduced by Leibniz, and hence
Note that analytic geometry was created
were unable to read books and treatises by
simultaneously and independently by Descartes
3 20
*
The Royal Society of London appoin- and Fermat; moreover, these two scien-
ted a special committee to discuss the contro- tists also belonged to different psychological
versy over priority (the committee sided with types. In contrast to Descartes, who thought
Newton). This committee, naturally, did not largely in terms of physics (and also geomet-
include scientists whose works it discussed. ry), Fermat can be regarded as an algorith-
More than that, the first phrase of the com- mist similar to Leibniz. This is reflected in his
mittee’s report (published under the title far more systematic treatment of the subject
Commercium Epistolicium in 1712) stated that of analytic geometry. (The “algorithmicity” of
only the absence of the scientists involved in Fermat’s thinking was most fully manifested
the controversy could ensure impartiality. But, in his outstanding achievements in the theory
alas, a copy of the draft of the report written of numbers, a theory that did not interest
in the hand so well known to historians of Descartes in the least; incidentally, Fermat
science shows with all certainty that the re- had a most profound grasp of physics —he
port (including the first phrase) was written discovered a remarkable principle that the
completely by the then President of the Royal path of a ray of light from one point to an-
Society, Sir Isaac Newton. On the other hand, other, through one or more media, is such that
the behavior of Leibniz in the controversy the time taken is a minimum. Fermat’s prin-
with Newton can also hardly be considered ciple played an outstanding role in the further
irreproachable. development of physics.
126 3 What is an Integral?

ried out with ordinary (real) numbers 3 22 New-

.
-
Newton as a mathematician), because the very
ton, the physicist, however, proceeded from idea of action-at-a-distance (Newtonian gra-
the substantive meaning of the new concepts vitation) contradicted his general scientific and
and their role in natural science: he regarded religious views.
the derivative as speed, and the reason why he The subsequent evolution of the scientific
did not carry through to the end the theory ideas of Newton and Leibniz in European
of limits which he had set forth in outline science is not without interest. Newton emerged
was because he saw no particular need for it. the indisputable victor in the debate about
T
in his view,’tne pnysicai meaning 61 ail \ne priority, ’nut ’tne outwarcT lorms oi’dTner-
concepts that he introduced fully justified ential and integral calculus have come down
them, and he felt no need for formal mathemat- to us entirely from Leibniz. As for scientific
ical theories here. ideology in the broad sense of the word, this
We can assume that the profound mutual was influenced for many centuries to a far
antipathy that arose between Newton and greater degree by Newton than by Leibniz.
Leibniz occurred at the time of their first This is illustrated not only by the above-cited
(and, apparently, only) meeting. This was lines from Pope and Lagrange. The entire
when Leibniz came to England to demon- history of European science in the 18th, 19th,
strate his version of a “mathematical” (to be and first half of the 20 th centuries was inspi-
more exact, arithmetical) machine [the first to red by Newtonian mechanics as a true model
put forward this idea was the outstanding of a scientific theory in the finest meaning of
17th-century French mathematician and phys- the term. And our book, too, owes much more
icist Blaise Pascal (1623-1662) 3 23 the very
-
: to the views of Newton than to the world-
idea of such a machine ran counter to New- outlook of Leibniz.
ton’s type of intellect]. Still more remote from But in the second half of the present cen-
Newton was Leibniz’s dream (and prevision) tury Leibniz’s dream of “thinking machines”
of “logical machines” (actually, the prototypes has suddenly come true. What is more, to-
of today’s computers) for mechanizing mental day’s “computerization of knowledge” has once
processes. On the other hand, Leibniz comple- again put the algorithmic ideas of Leibniz in
tely rejected Newtonian mechanics (altho- the forefront of scientific thinking, not to men-
ugh he had an exceptionally high opinion of tion the significance of diverse particular
achievements of the great German, from the
3 - 22
Interestingly, these ideas of Leibniz elements of mathematical logic he created to
received a fully logical substantiation only in his serious interest in combinatorics something
,

1960 in what is known as nonstandard analysis quite unusual for the 17th century.
introduced by the American mathematician Summing up, we can say that while New-
Abraham Robinson (1918-1974), which has ton and Leibniz approached higher mathemat-
already found definite application. ics from different starting points, the subse-
3 23 In
-
some respects Leibniz’s machine sur- quent history of science has confirmed the
passed Pascal’s arithmetical machine: it not unquestionable value of both avenues of
only could perform arithmetic operations on thought which they represent, the value of
numbers but could, for instance, extract both Newton’s approaches and the scientific
square roots. ideas of Leibniz.
 Chapter 4 Calculation of Derivatives

Convenient and pictorial modes

notation for various relationships and
of
lim
Ax
= 4—
dx >
that is, for the de-
Aoc-*0
simple rules that permit carrying out rivative.
computations mechanically without er- For this reason we will frequently
rors are of great significance both for write formulas which are only valid in
teaching and for the development of the limit, when the increments tend
mathematics as such. The place oc- to zero. For small Ax these formulas-
cupied by Descartes and Leibniz in the will be “almost” valid, that is, exact
history of mathematics is due not only equalities are replaced with approx-
to the mathematical discoveries they imate ones, and for large Ax they may
made but also to the new notation prove to be totally incorrect. To em-
developed by these scholars; in partic- phasize this we will use the symbols dy
ular, the language of the calculus of and dx instead of Ay and Ax. We have*
differentials developed by Leibniz. to work out rules for handling the quan-
In Chapter 2 we analyzed the mean- tities dy and dx which are called differ-
,

ing of the derivative concept and gave entials 41 so that the basic equation
,

simple examples of how to calculate

derivatives. In Chapter 4 we present the
j$r -**<*> < 4 -<-‘>
general rules for finding the derivat-
ives of various functions: polynomials, holds true.
rational functions involving ratios of this equation was already
Note that
polynomials, algebraic functions in- discussed in Chapter 2. There, however,
volving the unknown under the radical it was not a true equation but only a
sign (fractional powers), the exponential symbolic way of writing the derivat-
function, the logarithmic function, trig- ive; it meant that the notation y' (x)
onometric functions, and so forth. We and the notation dy/dx coincide. Now,
will find the general rules for the deriva- however, we wish to consider (4.1.1)
tives of various combinations of function as a meaningful equation that connects
whose derivatives are known: a sum of the derivative y' (x) and the differen-
functions, a product of functions, a tials dy and dx of the function y and the
ratio"of functions, a composite function. independent variable x, namely, the
A table of derivatives of a number of derivative y' (x) must be equal to
functions that summarizes the work the ratio dy/dx of the differentials.
carried out in Sections 4.1 to 4.13 is Before, we wrote an approximate
given in Appendix 1 at the end of expression for the increment of the
the book. function:

y (x + Ax) —y (x) = Ay ~ 9

y Ax.
4.1 The Differential
(4.1.2)
From the definition of a derivative
we have the following rule (algorithm) We can say that Eq. (4.1.2) becomes-
for calculating the derivative: specify exact in the limit, as Ax -> 0. Here
a value of x and its increment Ax, the words “becomes exact in the limit”
and
find / {x) / (x + Ax), find the in- do not mean, of course, only that
=
crement Af = f (xAx)+ —
/ (x), form
at Ax 0 the left- and right-hand
the ratio A// Ax, and then pass to the sides of (4.1.2) coincide (are equal to-
limit as Ax -> 0. However, the for- zero) —
they stress that for very small
mula which yields the general expres-
4 1
In Latin this word means “difference,”'
Af/ Ax for arbitrary Ax (not
-

sion for
or increment. In this way, Ax and Ay are incre-
tending to zero) is, as a rule, more com- ments, while dx and dy are differentials, that
plicated than the formula for the limit , is, “almost” increments.
 a

128 4 Calculation of Derivatives

values of Ax the left- and right-hand ues of Ax we can ignore the term a
sides of (4.1.2) are “almost’’ equal in the on the right-hand side of (4.1.3). How-
sense that the difference between them ever, the words about the limit as
is much smaller than the members Ax ->0 and about ignoring a term in
themselves. Indeed, the difference be- an equation become redundant if we
tween the left- and right-hand sides of replace the language of increments
(4.1.2) for small Ax is of the order of with the language of differentials; as
(Ax) 2 or even higher (see the end of a result Eqs. (4.1.1), (4.1.2a), and
Section 2.4), so that it is indeed (for (4.1.2b) are common (exact) equali-
y' = ^= 0, of course) much smaller than
5 ties.
A y and y' Ax, which are of the first order
of smallness: even if we divide both All this has been discussed in great detail
in Section 2.4 in connection with the question
sides of (4.1.2) by a very small Ax, of approximately calculating the values of
we will nevertheless arrive at an “al- functions (p. 77 on). When we calculate, say,
most” exact equality. Let us now agree the values of the function y x 2 or y x =
xz , =
to replace (4.1.2) with the exact equa- we are justified to assume that if Ax is small,
tion
then Ay ~
2x Ax or A y t ~
3x 2 Ax, since the
error introduced by such an assumption is
dy = y' (x) dx . (4.1.2a) negligible. For instance, the
equality Ay ~
approximate
2x Ax means that the incre-
Let us explain this further. We as- ment Ay of the area (y =
x 2 ) of the square
sume that the differential of the inde-
ABCD with a side AB=x
(Figure 4.1.1) caused
by the increment of the length of the side,
pendent variable dx is simply the, , Ax, can be replaced with the sum of the areas
increment Ax (by definition). Then of the two rectangles BB^C
and DD^c^C,
we define the differential dy of the func- ignoring the area of the very small square
tion y —
y(x) by the exact equation
Cc 1 C 1 c 2 equal to (Ax) 2 if X BB =
Ax is small.
If we examine the function y —
x z and put,
(4.1.2a),
4 2
which can be augmented
-

say, x —
2 and y 1 (x)
1
=
8, we can construct
in the following manner: the following table of increments Ay l =
yi (x -J- Ax) —
y 1 (x) and differentials dy x =
dy = y' (x) dx — y
r
(x) Ax. (4.1.2b) yi (x) Ax for this function and the errors
o/Ay 1 x— (A y —
dy-f)/ Ay ± introduced by the
Thus, the differential dy of a function approximate formula (4.1.2):
y =
y (x) differs from the increment
Ay, since for Ay the equality (4.1.2)
Ax 1 0.5 0.2 0.1
related to (4.1.2b) is satisfied approxi-
A^i 19 7.625 2.648 1.261
mately, Ay ~
y' (x) Ax, while the exact
dy i 12 6 2.4 1.2
equality for Ay / (x Ax) / (x) = + — cr/Ajfi 37% 21% 9% 5%
is of the form
= Ax 0.05 0.01 0. 001
Ay y’(x) Ax + <r, (4.1.3)
A 2/i 0.615 0.1206 0. 012006
where quantity of a higher order
cr is a dy i 0.6 0.12 0. 012
of smallness than Ax. It is this fact o/Ai/i 2.5% 0.5% 0. 05%
that we have in mind when we speak
of (4.1.2), or Ay/ Ax y' (x), as being ~
“exact in the limit”: the ratio Ay/ Ax This table vividly demonstrates the validity
generally differs from y'(x), but of the statement that formula (4.1.2) is exact
in the limit.
lim ~~ = y'(x), since for small val- More information concerning the remain-
Ax-*0 der a in the approximate formula (4.1.2) can
be found in Chapter 6. There we will prove
4 2
Note that while the approximate equa-
-
that (T can be represented in the form of a se-
tion (4.1.2) is a theorem that is, a statement ,
ries a (Ax) 2 b (Ax)
z
+ ., +
whose coeffi-. .

that requires proof (i.e. we need to prove that cients are independent of Ax. But for small
- on the right-hand side of (4.1.3) isof a high- increments Ax all the powers, (Ax) 2 (Ax) z , ,

er order of smallness than Ay and Ax), etc., are much smaller than Ax, whereby
Eqs. (4.1.2a) and (4.1.2b) are simply defini- a =
a (Ax) 2 b (Ax) z + is much + . .

=
| | |
. |

tions of dy. smaller than the differential dy | |

4.7. The Differential 129

Df cz

Figure 4.1.1
Figure 4.1.2

/Ay of a function in the sense that for

|
y' (x) Ax |
(which is proportional to |
Ax |).
Ax small the second term on the right-
Of course, the last statement becomes invalid
= hand side of (4.1.3a) is much smaller
when if (x) 0, that is, when the differential
dy vanishes and hence no small positive a | |
than the first.
can be smaller. However, even in this case we The reader will recall that the de-
can employ (4.1.2) and ignore the small error rivative y'(x) is the tangent of the
cr. We advise the reader to ignore, for the
angle a formed by the tangent line
time being, cases where a function y y (x) = =
has no derivative (and differential) since these to the curve y y (x) and the x axis
are really exotic (e.g. see Section 7.2). (see Section 2.5). Therefore, if PQ =
Ax (Figure 4.1.2), then X NQ =
The idea of a differential and the tan a X Ax =
y' (x) Ax dy,= while
notation dyldx for a derivative were the increment of the function, A y =
introduced by Leibniz, who did not
y (x + Ax) —
y (x) is depicted by seg-
give much thought to the content of ment MiQx- We see that for small val-
the idea of a differential, since he was ues of Ax the differential dy differs
more interested in the rules of opera- little from the increment A y of the
tion on differentials, rules that consti- function, while for Ax large this, in
tute the essence of Leibniz’s calculus general, is not the case. However, the
of differentials. However, an exact representation of the differential of
meaning can easily be attributed to a function in the form of a graph is
the idea of a differential; moreover, this needed for a correct understanding of
idea can be illustrated by a drawing. the role that differentials play in ap-
We can rewrite Eq. (4.1.3) thus: proximate calculations, a role that has
= been described somewhat differently
Ay dy + cr. (4.1.3a)
in Section 2.4 (without the use of the
This leads us to the following state- word “differential”). In this chapter we
ment: the differential of a function is the need only the technique for handling
principal linear part of the increment differentials, the symbolic calculus of
of the function In other words, the . differentials; we will nowhere use the
differential dy =
y' (x) [Ax (= y' (x) dx) exact meaning of the differential il-
is a linear function of the increment lustrated by Figure 4.1.2, and the reader
/Ax, while the increment A y generally may simply forget about it.
depends on lAx in a complicated man- The concept of a differential can be
ner (is not a linear function of Ax). 4 3 -
given a simple and graphic mechan-
On the other hand, dy constitutes the ical interpretation. Let us suppose that
principal linear part of the increment the abscissa x in Figure 4.1.2 is the
time variable (it is for this reason we
4 - 3
The differentially = y'(x) Azof
a func- have labeled the x axis by ( t )), while
tion depends on two numbers (variables),
the ordinate y (or z) is the distance
x and Ax, or is a function of two variables (it
is linear in variable Ax but is generally covered. We are interested in how the
a complicated function of variable x). distance covered varies with time,
9-0946
 1 Z x Z

130 4 Calculation of Derivatives

y —
y (x), or in the customary nota- Let us consider an elementary exam-
tion of Chapters 2 and 3, z z ( Z ). The = ple and then compare the increment tech-
idea of a velocity varying constantly nique with the technique of differentials.
under forces is not very simple; there- We take the function y x2 Original- = .

fore, in studying motion in the neigh- ly, we did as follows:

borhood of a fixed moment in time (the
position of a moving body at this mo-
Ay = (x + Ax) — x 2 2

ment is depicted by point on the M = 2x Ax + (A#) 2

,
(4.1.4)

graph of motion), it is convenient to whence

assume that starting from this moment
the velocity ceases to change (this as-
^ 2x+ Ax and y' = lim
A
= 2x.
Ax-*0
sumption is equivalent to the hypothe- Using differentials, we write
sisthat at that moment we “turn off” the
forces acting on the body, with the dy = (x + dx) 2 — x2 = 2xdx, (4.1.4a)
result that subsequent motion is en- where the term (dx) in the right 2
mem-
tirely by inertia, that is, with a con- ber was dropped immediately.
stant velocity; see Chapter 9). Then, The meaning of such operations
starting with that moment x the ve- , with differentials is quite clear now.
locity remains constant and equal to Of course, Ay = (x + dx) 2 x2 = —
instantaneous velocity v (#) dyldx at = 2 xdx +
(dx) 2 (see (4.1.4); we remind
time x (or dz/dt at time t), and the the reader that dx and Ax is the same
distance covered during time Ax will thing). When we go over from Ay to
be equal to v (#) Ax y' Ax =
dy. — dy we leave only the principal linear
,

(In Figure 4.1.2 the uniform motion part of Ay, that is, drop all terms that
of the body corresponds to the straight involve Ax (or dx) in second or higher
line Z, while the graph of the real (non- powers and retain the terms that are
uniform) motion corresponds to the proportional to the first power of dx .

curve T.) For small Ax (or At ), this These rules of the calculus of differen-
hypothetical path 1 (= NQ
dy or dz) tials considerably simplify all calcula-
differs from the true path (= Ay tions after one gets accustomed to them,
or A z) by an extremely small quantity and in subsequent sections we will
NM = X o whose order of smallness widely employ them.
is higher than that of PQ (= Ax or Finally, let us discuss a remarkable
A Z). 4 - 4
property of differentials, a property that
The rules for using differentials must Leibniz knew and rated highly. Let
be such that the ratio of differentials us take a function, say y x 2 Then = .

is equal to the derivative, or (4.1.1)

is valid. To achieve this, in formulas
y' = 2x and dy = 2x Ax. (4.1.5)

we have to drop all terms proportional Now suppose that x is not an indepen-
2
to (dx) and higher powers of dx. dent variable but an auxiliary variable
dependent on an independent variable
4-4
was in this mechanical form that
It t (this can be the time variable). Will
Newton's differential, which he named “fluent”
(4.1.5) retain its meaning? Of course
appeared; Newton arrived at this concept
earlier than Leibniz, who did not begin to
not, since if x = t2 then the deriva- ,

speak about differentials until 1675, whereas tive of y = x 2 = Z 4 is equal to 4 3 and

Newton’s variable quantities (or “fluents,” not to 2x =
2 2 In the second formula
as he preferred to say) were in evidence as
early as 1666. Even earlier than that, approx-
in (4.1.5), Ax is
.

now equal to (t + AZ) 2 —

imately in the late 1630s, Fermat began to t
2 = 2tAt+ (AZ) 2 ,
whence
use differentials, but without giving them any
special name or sign; it was on these concepts
2xAx = 2 [2ZAZ + 2
( Az) 2 ]
that he based the well-known theorem of max- = 4Z AZ + 2Z
3 2
(AZ) 2 ,

imum and minimum points. However, the

credit for the extensive calculus of differen-
which from the expression dy =
differs
tials undoubtedly goes to Leibniz. 4Z 3 AZ on the left-hand side of the second
 ,

4.1 The Differential 131

formula in (4.1.5). But the main rela-

tionship (4.1.2a) retains its validity in
the new conditions: here dx 2tAt, =
so that 2 xdx =
2 1 2 (2tAt) At 3 At dy. = =
Is this fact accidental? No, it is not.
Formula (4.1.2a) retains its meaning
in the case where x is the independent
variable and in the case where x is
an (intermediate) function of another
independent variable, t (cf. Exercises
4.1.1 and 4.1.2).
In view of this it is often more con- interpreted as the ratio d 2 y -f- (dx) 2 of the sec-
venient to work with differentials than ond differential d 2 y of y to the square of the
(first) differential (increment) of the indepen-
with derivatives and increments, since
dent variable x or (dx) 2 The second differ- .

there is no need to remember whether ential d 2 y of the function y y (x) may also =
x is the independent variable or an be depicted on a graph, which offers possibil-
auxiliary function. We will see many ities for using the second differential in cal-
culating values of a function. For the time
examples of this fact below.
being we will write x Q instead of x and x in-
The simplest way to substantiate the above-
stead of x +
Ax. We wish to find the para-
bola n,
formulated property of differentials is to use
formula (4.1.3), where a is a quantity of Y= a (x 0)
2—x b (x x0 ) c,+ (4.1.8) — +
higher order of smallness than that of dx that is the closest to the graph T of the func-
(and dy =
A dx). In view of this formula we tion y —
y (x), that is, a parabola for
which
—
have the difference (x) Y
y (x) for small x —
Ay =B dt + a, (4.1.6)
x0 = Ax
as small as possible.
is
4 5 The problem -

of finding such a parabola (the problem of the

where B dt = dy, and a is small. On the “best quadratic approximation” (4.1.8) of
we
other hand, y (x)) coincides with one of the problems
will discuss in Chapter 6; there we will find
Ax = C dt + P, (4.1.6a) that the coefficients in the equation of para-
n are a =
(1/2) y" (x 0 ), b =
where C =
dx and p is small, and if the
dt
bola (4.1.8)
=
y (x 0 ). In this case the
,
y' (x 0 ), and c differ-
derivative dy/dx at the given point (at t t0 = Y —
y\(x) for small Ax is a quantity
or at x =
x (t 0 ) =
x 0 ) is equal to D, then
ence (x)
of third order of smallness (or even higher) with
Ay =D Ax + y, (4.1.6b) respect to Ax x x0 =
If we — . depict the
curve y y (x), the parabola Y
— —Y (x), and
where y is small. But from (4.1.6a) and (4.1.6b)
the tangent line l ,
it follows that D dx (= DC
At) coincides with
the differential of y = Yx — kx + s, (4.1.9)
y (t), since (a) it is
linear in At and (b) the difference to the curve T (where k y' (x 0 ) and s == =
—D (x —
kx in one drawing (Figure 4.1.3),
Ay dx = (D Ax + y) — D dx y 0) 0)
then the straight line x constant will in- =
= [D (dx + P) + y] — D dx = £>p + y tersect T, l and II at the points
,
and N ,

iVj, with (see Figure 4.1.3)

is a small quantity (a quantity with an order
QiMi = y (x)—y (x )—Ay, QiN y' (z ) A x
0 0
dy> = =
of smallness higher than that of At) because
P
and y are such quantities and dx is, of D QiNt = y y" (x 0) (Az) 2 -j-y' (x 0 ) (Ax)
course, simply (dy/dx) dx.
The second derivative y" (x) of a function
y =
y (x) was denoted earlier (see Section 2.7)
d*y+dy.
as
4-5
This parabola (the osculating parabola
described
1 of curve T; cf. Section 7.9) can be
is follows. We draw a
parabola througn
ana
^
Such notation can be justified by a line of hree close-lying points ( x 0 y 0 ), (x\, yih ,

x 2l y 2 ) of curve T, that is, we select

reasoning similar to the one used in connection a, o,
with the notation y' dy/dx. = nd c in (4.1.8) in such a manner that —
We call the quantity y" (x) ( Ax ) 2
L

r (x n ) =y0 Y (x x )
, yi, and Y\{x 2 ) =
(= y" (x) {dx) 2) the second differential of (x)
y :hen°, as ? and i s tend to x 0 , the parabofa
and denote it by d 2 y. In this case the notation vill tend to n (so that II is
the limit of 1^
(4.1.7) for the second derivative y" (x) can be s Xi and x 9 tend to x 0 ).
 ) x )

132 4 Calculation of Derivatives

i.e. Using differentials, we can write

QxN = dy and NN = — d r
2
y
dy ~ y (x + dx) — y (x)

= +[Af + Bg (x + dx)]
(x dx)
(in the case at hand d y is negative). 2 — [Af + (*)]
Bg (x)
Figure 4.1.3 provides a persuasive illus-
tration of the advantage of the approxi-
= A (x + dx) — f[f (a:)]

mation Y y0 ,dy =2
+
(1/2) d y (= y (^ 0 ) + + + B [g (x + dx) — g (#)]
y' (x 0 ) Ax + (1/2) y" (x 0 ) Ax 2 )' for the func-
Adf + Bdg = Af'dx + Bg'dx
tion y over the approximationTT == z/ 0
{x + ,

dy (= y (x 0 ) +
y' (x 0 ) Ax): in the second case whence
point 1 M
in Figure 4.1.3 is replaced with
point TV, while in the first it is replaced with y'= J =A f'+ B z'- 4 2 1>
N x a point that lies closer to
, x than N. M £r (
- -

The second differential d 2 y of a function

y =y (x) can be given, just as in the case
The reader can easily arrive at this
with the first differential, a certain mechanical formula if increments and limits are
interpretation. We again assume that the used instead of differentials.
axis of abscissas in Figure 4.1.3 is the time
In particular, for the sum and differ-
axis t and the axis of ordinates (the z axis) is
the distance axis. Now suppose that starting = —
ence of two functions (A B 1 and
from a certain time (corresponding to point M A = 1, B = — 1) we have
of T) the body (whose motion we are studying)
moves with uniform acceleration', this assump-
tion is equivalent to the hypothesis that
+ gy = r + g',\ if-gy =r -
(/
g'-

(4.2.2)
the forces acting on the body cease to vary at
that point in time, since a body is uniformly Let us now find the derivative of a
accelerated if the forces on it are constant (see
product of two functions, g {x) and
Chapter 9). In this case parabola II becomes
the graph of motion and the (hypothetical) h ). We put f (x)
(.
g (x) h (x). Then =
path traversed by the body in time Ax (or A*)
will be equal to dy
2
2
(1/2) d y (or dz + + df (a;) / (x + dx) —/ (x)
(1/2) d z), which in Figure 4.1.3 is depicted
by segment Q±NV — g (x + dx) h (x + dx)

— g {x)h (a:).

But
Exercises
g {x + dx) ~ g (x) + dg,
dy dy dx
4.1.1. Prove that
dt dx dt
h (x + dx) h {x) + dh.

(a )y = x 2 and x = Y t ,
and (b) y = Yx and Whence
x= t
2
. df ~ [g (a:) + dg] [h (x) + dh]
4.1.2. Suppose that y = x2 ,
with x = — g {x)h {x) g {x) dh =
V*-j-l ,
an d t — u2 - Prove that + h (x) dg dgdh. +
and dh —
dy =
^ dx= dy
dt
dt =
du
du .
Note
h' {x) dx,
g' (a;) h (a:) (dx)
that

'
2
g' (a;) dx
whence
dg ==
dhdg =
The quantity dhdg .

is proportional to (dx) 2 and so, accord-

4.2 Derivatives of a Sum and ing to the rules for handling differen-
of a Product of Functions tials, we ignore the product dhdg in

As another example of how to use the the expression for df. Finally we get
language of differentials we take the df = (x) dh +h (x) dg. (4.2.3)
g
sum of two functions, / (a:) and g (x),
taken with (arbitrary) constant coef- Dividing all the terms in (4.2.3) by
ficients A and B :
dx, we get

df __ d ( gh dh dg
l7h—£-
= 8
5 h (4.2.4)
y Af (x) + Bg (x). dx dx dx [

dx
 £ , -

4.2 Derivatives of a Sum and of a Product of Functions 133

r
The expression is remembered in the fol- in the limit the product h'g and we
lowing manner: the derivative of the sent Ax to zero.
product gh is equal to the sum of the Using increments and passage to the
derivative taken on the assumption limit,we obtain the same result as that
that h is a function of x while g is held obtained with the aid of differentials,
constant (the term g (dh/dx)) and the but it takes more time. This is not
derivative taken on the assumption surprising since in the case of differen-
that h is held constant and only g de- tials we dropped dhdg mechanically, on
pends on x (the term h (dg/dx)). Here, the of an earlier acquired rule
basis
naturally, the constant value of g according to which we have to reject
in the term g (dh/dx) is taken for the terms involving (dr) 2 (dr) 3 etc., hence,
, ,

x for which the value of the derivative any products of two, three, or more dif-
is being sought. The same goes for h ferentials. When carrying out the com-
in the second term. Below we will see putations with the aid of increments,
that a similar procedure can be applied we actually, in the very process, proved
in other situations, 4 6 for instance in this rule once again for the case of a
g xK
the case of y / (x) = ( product of functions.
How would we have handled this The succession of operations using
in the old way? Simple algebra yields increments is needed to justify the
the exact equation rules and to understand them. But once
they are understood, the use of differ-
A/ = (g a g) (h + Ah) — gh
+ entials is faster and more efficient. It
= gAh + hAg + AhAg. would be silly every time to start from
rock bottom in solving a specific prob-
Dividing both sides by Ax, we get lem and to write out in full that the
derivative is the limit of a ratio, and
-g~fM-£+*w£+£&*«. so on.
=
Example. Suppose that / (x)
(4.2.5) (3x 2 +
5) (2x — 4). Find /' (x) and,
Note that the last term for the sake in particular, /' (0).

of convenience we multiplied and di- Here

vided by Ax.
g = 3x + 5, i| = 3x2* + 0 = 6.r;
2
Up to now
equations are exact
all the
and hold true for all values of Ax.
h= 2x— 4, i£-= 2-0 = 2.
Now pass to the limit as Ax ->0. Then ’
dx

Ah Therefore
lim
Ax
= /' lim
Ax
= h',
Ax->0 Ax-* 0 =( 3 * 2 + 5 2 + 2 *
) ( 4) 6 *
A
lim = 18* 2x -24*+10,
Ax-*0 Ax
and in particular
and, in view of (4.2.5),

= /' (0) = if = 10 .

K -h *4*2-6} I

In passing to the limit, the last term The rule for finding the derivative
on the right-hand side of (4.2.5) va- of product generalizes to the case
a
nished since the first two factors yield of many factors. For example, for the
product of four functions / (x), g (x),
4 6
It is clear that the formula for the
* h (x), and k (x) we get
derivative of the sum of functions (if F =
/ +g, then F' /' = +
g') also obeys this
rule, since if / is assumed constant, the deriva-
tive F' is simply g', while if g is constant,
the derivative F' is /'.
 * )

134 4 Calculation of Derivatives

Thus, here we once more encounter Finding derivativesis simplified if

the same rule, according to which the we reduce functions to combinations

all
derivative of a product of several func- of the simplest possible functions: sep-
tions can be found as the sum of the arately, each of the functions 2 =
derivatives computed on the assump- 1 ly and y x2 =
5 is simpler than +
tion that each time only one function 2 (x =
2
5)
_1
+
By reducing compli- .

varies while the other remains constant. cated functions to combinations of

The formulas we have obtained en- simpler functions, we are able to get by
able making further conclusions. It with the rules for finding the deriva-
is clear that if f (x) g (x) h (x) = + tives of these simple functions.
and f'{x) g'(x) =
h'(x), then + Let us find the differential of the com-
posite function 2 [y (#)]. Regarding 2
/" (*) = (/' (*))' = (§’ (*) + h' (*))' as a function of y, we write
= g"(x) +h"(x). (4.2.8)

The formula for the second deriva- dz^^-dy (4.3.1)

tive of a product of functions is somewhat
more complicated: if f (x) = g x ) h (x)
(,
y
(compare this with what was said in Sec-
then /' {%) = g (x) h' (x) g' (x) h (x) tion 4.1 about the differential of a
and /" (x) = 0g (
x ) h‘ (x) + g' (x) h (*))' = composite function). But y is a func-
(g (a:) h" (x) + g’ (x)h‘ (x)) + (g' (x) x tion of #, whereby
h' (x) + g" ix ) h (#)), that is,
dy=z! dx (4.3.2)
f" (*) = g {x)h" (x) + 2 g’ (x) h’ (x) al
-

+ g"(x)h(x). (4.2.9) Substituting (4.3.2) into (4.3.1), we

obtain
Exercises

4.2.1. Use the rules found above to cal-

culate the derivatives of the following fun-
dz = ^^ dx -
<
4 3 3)
- -

ctions: (a) y = = ax +
x* (= A 2
), (b) y
5
Dividing both sides by dx we get a ,

bx + cx + dx + ex + y = Yx
4 3 and 2
/, (c) rule(known as the chain rule) for deter-
using the identity Y x Y x = x. [Hint. Use the mining the derivative of a composite
following equality: y'y + yy' = x = 1.]
r
4 7
function -

4.2.2. Express the third derivative /"' (x) =

:

of the function f (x) — g x h (x) in

(x))' dzdy_
(f"
terms of the functions g and h and their deriv-
( )

dx
dz_
= dy dx
(4.3.3a)
atives.
4.2.3. Suppose that / (x) = g (x)h (x)k (x).
The form of the formula is in full
Find f (x).
accord with what was stated about the
possibility of handling differentials as
4.3 The Composite Function.
ordinary algebraic quantities: we can
The Derivative of the Fraction
cancel out dy in the product ( dz/dy
of Two Functions
(dy/dx).
Let be given as a function of y, say
z Recall that 2 is given as a function
2 =
1 ly, and y as a function of x say ,
of y, and so dz/dy is also a function of
y =
x2 5. It is clear that to each x y. But since y itself is a function of x ,

there corresponds a definite y and it follows that by substituting y =

o.inft5b + 6U '\5u\hi 7
y wnsjpmdfb a* i
y
(
^ )
for* dzlrtiy,
definite 2 ,
we see that each x is associat- we get dz/dy as a function of x and,
ed with a definite 2 that is, 2 is a func-
,
hence, also dz/dx as a function of x.
tion of x. It is always possible, by sub-
4 7 The prime notation z* instead of
stituting the expression of y in terms -

dz/dx can lead to confusion. When we write z',

of x, to write directly formulas for
it is not clear whether we mean dz/dx or
2 (#); in the given example, 2 = dz/dy. (However, we could have written
_1
(x 2 + 5) . or zfj.)
4.4 The Inverse Function. Parametric Representation oj a Function 135

8
Let us carry out computations for a or 4 -

case that will be needed later on. Sup-

pose <4 - 3 ' 8)

1
z (4.3.4)
y(x) Exercises
We know that if z = l/y, then dz/dy = 4.3.1. Find the derivative of 2 = (ax + b) 2

— 1/i/
2
and so as that of the composite function 2 = y with 2
,

dy y — ax + b. Remove the parentheses in

= dy = +
1
dx (ax b) 2 and find the same derivative di-
dz 2
y dx rectly.
and 4.3.2. Find the derivatives of the fol-
lowing functions:
dz 1 dy
2
(4.3.5) (a) Z= (b) Z==
dx y dx
ax+6 ’
(ax+bf
'

For example, if y x = 2
+ 5 and z = x3 5a: 2
2
=
5)~\ then + (c) 2 = 1
(d) y =
1 !y (x
+ 1 1/a:
’
x+1
dz 1 d(x 2 -\- 5)
~~
2x
y=
— X 1
dx~ (x 2 + 5) 2 dx (z 2 + 5)
2 • (e)
+2 X2

Here is another simple example: 4.3.3. Suppose that y f (x)/g (x). Ex- —
press the second derivative d 2 y/dx 2 of the func-
y = x 3/2 = ]/^
3
. This formula can be tion y in terms of the functions / and g and
rewritten y = u with u as Y *
= x3 .
their first and second derivatives.

According to (4.3.3a) we have

2
dy _
“
dy du 1 3a: 4.4 The Inverse Function. Parametric
dx da cte Representation of a Function

£ V*. Specifying y as a function of x signifies

2
that to each x there corresponds a def-
Earlier we arrived at the same result inite value of y. Hence, conversely,
by a more complicated method (see we can often say that with each definite
Exercise 2.4.3). y there is associated an x. For instance,
The chain rule (4.3.3a) for forming if y 8x3 9, = then x3 (y
— = +
the derivative of a composite function
holds true for a more complicated rela-
9)/8 and x =\/^=j/ *±2. Thus,

tionship as well. Suppose that z = the specification of y (x) also yields the
2 (y), y = y (*)> * = X (*)> and f = functional relation x (y). This relation-
t w ). Then ship is called the inverse function as ,
(
we have already said in Section 1.6;
dz dz dy dx dt (A o
' • • on the other hand, y (x) is the inverse
dw dy dx dt dw ( /
4 9
of * (y).
-

(4.3.5), we can easily find the

Using In many cases the inverse function
derivative of a fraction (quotient, ra- has a more simple form than the initial
tio) of two functions, / = hlg . To do function; for instance, the function
we write / in the form of a prod- —
this,
= h Then
y= Yx 1 contains a cube root,
uct: / (1 Ig). while the inverse function x y
3
1 = +
is a polynomial, which is simpler
/'
= My)' + /.'y. <4 3 - 7>
-

4 8
1 1 is clear that formula
But (1/g)' = -(1/g 2 ) g’ (see (4.3.5).
-
(4.3.8) corre-
sponds to the general principle formulated in
Substituting this result into (4.3.7), Section 4.2: (h/g) h'/g h (1/g)'.
f
= +
we get the sought result 4 9 In the general case the function x
- —
x (y) that is the inverse of the given function
y —y (x) is multiple-valued (see Sections
1.6 and 4.11).
 x ) ) 1 1 x

136 4 Calculation of Derivatives

than y (. ). In this case, it is simpler and If a function is represented paramet-

easier to find the derivative of the in- rically (see Section 1.8), this represen-
verse function, dxldy than it is to find ,
tation may be regarded as a special
the derivative of the initial function, case of a composite function. Indeed, if
dyldx The problem now is to express
. it is given that
the derivative of the initial function
in terms of the derivative of the in-
x = f {t), y = g (it
), (4.4.5)

verse. For y x (.
we have then the first of these equations may be
regarded as an equation whose solu-
d y = -%-dx=y'{x)dx. (4.4.1) tion yields t t (x). Substituting this =
t x into the second equation, we get
(,
This enables finding the derivative
x (y) of the inverse function x (y):
y =
g (t) g {t (x)). Hence, =
dy dy dt
dx 1
dx dx
(y) dy ~ y' (x)
(4.4.2) dt

(Of course, (4.4.2) once more illustrates

But to use this formula one need not
express as a function of x (if we were
t
that differentials can be treated as
= to do this we would get rid of the para-
numbers, or dxldy 1/ {dyldx).) On the
meter, but this is not always possible).
right is an expression in the form of a
function of x But if we know x .
= It suffices to know x = / (£), which is
the inverse of t (. ). Thus,
x (y), this expression can be repre-
sented as a function of z/, namely, as dt _ 1

WA (* (*/))•
few examples will serve as illus-
dx ~ dx
dt
7

trations the foregoing. The first

of and so
example (a linear function) is too sim- dy
ple. We
start with the second example: dy _ dt
(4.4.6)
dx dx
y=x 2
> %= v' x
( ) = 2x <

4 4 3)
~dt

jl = _L_
dy y' {x)
=± 2x
<
- -
This formula
showing that
is yet another instance
we can handle differ-
entials ordinary algebraic quan
like
Substituting into (4.4.3) the inverse tities: the quantity dt in the right mem-
function x = l/y, we get ber of (4.4.6) is canceled out.
Here is an example Suppose that .

dx _ dVy = 1 _ 1

dy ~~
~~
dy 2x 2 Yv
(4.4.4)
x — t
2 — t, y = t
2
+ t. (4A7)

In Chapter 2 we obtained the same re- Then

sult (the derivative of y = Yx a
£-*+*•
more roundabout fashion.
)
§
Here is a third example: dy 2^-f-l
dx 2 —
y — x3 + \, jjL = y’( x = Sx )
2
,

In constructing the graph of a

dx _ ~ 3x* 1 1
function represented parametrically,
dy y' (x)
9
x = x y ( t ), we set up a
(t) and y =
dx _ d (Yy — l) ~~
~
1 _
~~
1 table where specify the values of x we
dy dy 3x 2 3 ^_!)2 and y corresponding to the chosen va-
lues of t, this set of x and y specifying
=4o/-i)- 2/3 a point M=M
(x, y) on the graph of
the function. It is also convenient to
(cf. Section 4.5). calculate for a given t the derivatives
 ) ) )

4.4 The Inverse Function. Parametric Representation of a Function 137

dxldt and dy!dt\ then the ratio function y y = (x) specified parametrically*
(dy/dt)/(dx/dt) will fix the slope of the
x =x (t) and y = y (t):

tangent to the graph at point (.x y). M ,

d2 y _ d / dy \
where x=x(t).
Formulas (4.3.5) also make it
(4.4.2) and dx 2 dx \ dx ) ’

possible to arrive at formulas for the second dy

derivative of a function x x (y) that is = _ y'(t)
•

the inverse of another function y y (x). Of = dx x' (t)

course, following the line of reasoning that

led us to (4.4.2), we would like to think that Hence,
x"(y) =
t/y"(x)— but this is not so if only
because of dimensionality considerations. It d y' (
t 1 y" (t)x' (t)-y' (t)x" (t)
1 |

is clear that if x and y are expressed in units d2 y

' 1
dt x (t) (x (,t))»

of e 1 and e 2 the derivatives y'(x)

,
dy/dx — dx 2
\

x (t)
)’

x’ (t)
and x'(y) =
dx/dy have the dimensions of
eje x and eje 2 respectively, so that the expres-
,

sions on both sides in (4.4.2) have the same that is,

dimensions. However, the second derivatives
y"(x) =
d 2 yldx 2 and x"(y) d 2 x!dy 2 have = d2 y __
y" (t)
f
x (t)-y' (t) x" (t)

the dimensions of eje\ and eje\, in view of dx 2 (x' (t)) 3

which x"(y) (whose dimensions are those of
eje\) can in no way be equal to i/y"(x)
(whose dimensions are e\/e 2 ). (If x , y, and t are measured in units of e 1 , e 2r

The true formula for the second derivative and e, then the derivatives dxldt and dy/dt have
of x (y) can be written as follows:
the dimensions of eje and e 2 /e, so that the
fraction on the right-hand side of (4.4.6) has

d2x d dx \__ d 1 1
the dimensions of (eje)!(eje) eje^ that is, =
dy 2
“ dy
I
\ dy ) dy ~ )
\ y’ (*) 1
dimensions of the right-hand side of (4.4.6).
On the other hand, the second derivatives
x" ( t and y" (t) have the dimensions of eje %

(y’ (x))
2
~dx^ V
(y’
{X))

(x)) 2
^ and eje 2 so that the products in the numera-
,

tor of the right-hand side of (4.4.9) all have

the same dimensions, those of (eje 2 ) eje (< =
(
eje 2 ) (e 2 /e) =
(e-^ej/e 3 with the dimensions of ,

the entire right-hand side being those of

«•<*>
[V(t )1
1
(eyejte 3 (eje) 3 =
eje f, that is, coinciding
with the dimensions of the second derivative
(y (x)) 2
2
d y/dx .)2

»'(*)(!/»' (*)) For instance, in the case of function (4.4.7>

(y' (*)) 2 we have

d 2x _ d2y
Thus, finally,
dt 2 ’
dt 2
~ ’

x " (v)=
y
"
^ (4 4.8)’
d2y 2 (2t — 1) — (2£ + l) 2
{y)
(y' (*)) a
• K
dx 2 ~ (2t — l) 3

4
(It is x and y are measured in
clear that if ~ (2t — l) 3 *

units of e x and e 2 respectively, then the right-

,

hand side of (4.4.8) has the dimensions of

(eje\)l(ejej 3 =
eje §, that is, dimensions that
coincide with those of the left-hand side
Exercises
of x" ( y ).)
For instance, if y =
x2 x ,
= y y, then
Find derivatives of the follow-
y' = 2x, y" 2, = and, hence,
4.4.1. first

ing function: (a) y = y^2x-J 1 , (b) y =

r
x"(y)=
d2 Wj) 2 1
]/ »
an d (c) y = l/y x.
(2x) 3 Ax 3
= y = bsint.
dy*
x
4.4.2. acost, Find the de-
1 rivative dy/dx.
4y 3 / 2 4.4.3. Find the second derivative d 2 yidxt
f x 1

Similarly it is easy to use (4.4.6) and de-

of the following functions; (a) y=y ~~
q
'

~\ r
m

rive a formula for the second derivative of a (b) y = l/y~x, and (c) x = acost, y = bint.
 1 -

138 4 Calculation of Derivatives

4.5 The Power Function cise, (4.3.5); here y = 1 If and / = x k ),

we find that
Let us consider the power function,
n
y = x , (4.5.1)
dy 1
2
df 1
dx f dx x 2k
where n is a constant number. For n
n
a positive integer, x is the product of
72 identical factors, y x-x-x x, — . . .
Recalling that k = — 72, we find that
for 72 negative
n factors
and, hence (via formula (4.2.7)), dy dxn
nx,n-i
dx dx
-^=l-xn ~ +l-xn ~ + i l
. . , + which coincides with (4.5.2).
n terms
Thus, the formula for the derivative
of a power is applicable to any ration-
whence al exponent 72. It can also be extended
411
dy to the case of an irrational exponent.
dx
= nxn- (4.5.2) Formula (4.5.2) is of greatest impor-
tance. It implies, for one, that for r
We will show that thisformula holds small in absolute value we can write
true for arbitrary 72, whether fractional n
or negative. 410 (1 + r) 1 + nr. (4.5.4)

For fractional n we write n = m/p, Indeed, if x = 1, Ax = and y =

r,
where m and p are integers: xn then Ay — (x+ Ax) n — xn —
—
,

n
m p
y = x l (4.5.3a) (1 + r) 1. On the other hand,
,

dy = y
r
(.
x) Ax = 72- 1 r = nr.
71 "" 1
- For-
or mula (4.5.4) follows from the fact that

y
v = xm
. (4.5.3b) (for Ax =
r small) the increment Ay
is close to the differential dy.
The expression y v on the left-hand side It is also expedient to write (4.5.2)
of (4.5.3b) is a composite function of in another form:
x since y depends on x Therefore, cal-
,
.

culating the derivatives of both sides if y = cxn ,

then -^|*=72-|-. (4.5.5)
and employing (4.3.3a), we obtain
One has to get the feeling of this result.

w^
For positive n the power function has the
p~ 1 =
or py % obvious property that for x
,

0, y is also
equal to 0. For a given n >» 0, the curve y
=
=
whence cxn can be drawn through any point (£ 0 y 0 ), ,

m
just choose c =
yjx ft. Let the curve pass
through the origin and through the point
xm
-1 ~1
dy m x 771 m
= ~/T (x m IV)V- m m p
“ 3

(x 0 y 0 )• We wish to find the average value of

~dx “ 7~ 1
,

the derivative over the section of the curve

between the origin and point ( x 0 y 0 ). Accord-
Noting that mlp — n, we finally ing to the definition of an average (see Sec-
,

have tion 7.8),

x0

For a negative exponent we write

n = —k, where k is a positive inte-
ger, so that y — xn =
x~h ilx
h = .
4 n The
function y
-
x n where n is an = ,
Using again (4.3.3a) (to be more pre- irrationalnumber, is defined by the condition
xn =
lim xr where r is a rational number
,
4 - 10 The reader will immediately see that and lim r =
n, whereby if formula (4.5.2)
it is valid at n = 0, when y = constant and is valid for all rational exponents n, it will
= o. necessarily be valid for irrational exponents.
 . x . x x

4.6 Derivatives of Algebraic Functions 139

Exercises

4.5.1. Find the derivatives of the follow-

ing functions: (a) y x5 Sx^-^-x^-^-lx 2 = — —
2x +—
5, (b) y (a
3
x L 2 (c ) V = ;
+ + =
(x 2 x +r l) (3x 2
4 l) 10
,
(d) y =
>

— ,

(e) y x =\1, and (f) y

/ 2
f ’z
2 — = .

4.5.2. Obtain formula (4.5.2) for n a po-

sitive integer with the aid of the binomial theo-
rem. [Hint. Use the binomial theorem (Sec-
tion 6.4) to calculate Ax (x Ax) n x n .\ = + —
4.5.3. Find the values of y (9) and y (11)
if y (10) 5 and (a) yoc = x, (b) y oc 1 / x Y ,

and (c) y OC x 2 where stands for proportion- ,

al. Solve the problem mentally without any

computations. Compare the answers with the
exact values.

4.6 Derivatives of Algebraic Functions

The collection of rules in Sections 4.1-

4.5 enable finding the derivative of any
function involving addition, subtrac-
whence, using (3.4.9), we get tion, multiplication, division, and rais-
y(xo) — y{Q) _ yM __ y0 ing to an (arbitrary) power including
Xq Xq Xq fractional powers, i.e. extraction of
Indeed, as x varies from 0 to x 0 the value of y roots, of the independent variable x.
grows from 0 to z/
0 Hence, the average rate of
.
,

An example will illustrate how to do

'g'ny^hi ^5x 'y '(ire. "average drerivdiiv^; r» ; +
hiib ‘m + hib hretf l ’prailinrdA
n
^mh
equal to yjx 0 ,
a result obvious without inte- the derivative of the function
grals.
As evident from (4.5.5), the value of the
derivative at point (x Q y 0 ) is n times the aver-
,
/ {x) =x 3
/"x 2 — 1

age value of the derivative (n is the exponent).

Figure 4.5.1 depicts a number of curves with
It is best to write the answer at once,
different n (n =
1/2, 1, 2, 5) passing through that is, without introducing any new
one and the same point (
N
x 0 y 0 ) and, hence,
having the same average derivative on the
, designations (such as / 2 1 y).
3
— =
interval from 0 to x 0 It is clearly evident that
.
The derivative is taken, as it were, sep-
the larger the n, the greater the derivative at arately with respect to each side involv-
point N
(the more steeply the curve rises). ing x, and we say roughly the follow-
ing (the letters “a,” “b,” “c,” “d” show
Let us return to formula (4.5.5), to what portions in the expression of
dyldx= n (ylx), whence dy = n ylx)dx ( ,
the derivative the words refer): the de-
and so for small increments Ax we rivative of (a) with respect to x in
have front of the radical sign plus (b) the

Ay ~ 72 -|- Ax (4.5.6)
derivative with respect to 2
1 Y —
multiplied by (c) the derivative
For Ax =
O.Olx, that is, for a 1% Y 2 — 1 with respect to x2 — 1 mul-
variation in the independent variable, tiplied by (d) the derivative of x — 12

formula (4.5.6) can be assumed exact. with respect to x:

Then (a)

Ay ~ n - X 0.01.r,
r
dx
or Ay ~
n x 0.01 y. (c)
Thus, when the independent variable (b) (d)
1 1

varies by 1%, the power function 1 y**— 1

(4.5.1) varies by 72 %. + X X 3 x* —l 2x. (4.6.1)
 x X 1 1

140 4 Calculation of Derivatives

used to this efficient

It is well to get —2
x2
= xj/( 2x + 3) 2
4.6.16. y .

approach (without indulging in a lot of (i + x V i+x

2
)
2

writing) by applying the following

principles:
4.6.17. y (x 3 — 1 ) Y X — + x f/1 1 l.

2
(a) The rule for differentiating a com- xj/_(2x 3)
4.6.18. y = — 1)* _ 4.6.19. y ••=

posite function (Section 4.3, formulas (*

(4.3.3a) and (4.3.6)).
lEi »- *’+f+‘ yi+r.
(b)the expression is made up of
If
several functions, then its derivative
/Hr-
c+i
4 6 - 20 -
-

xY_ z 2 —
is equal to the sum of the derivatives 4.6.21. y =
__ 4.6.22. y =
computed on the assumption that each
x3 + i

time only one of the functions is taken

(4.2.7)
to be variable and the rest are held con-
(4.3.8)
stant (see Section 4.2, formulas (4.2.4),
Y • 4 - 6 - 23 - y= x ^*2 -
1
1 X

and also and V Y~~x 4.6.24. y — x + -t-V'

,

).
(4.2.1), (4.2.2), (
v
y x >

The formula for the derivative of a 4.6.25. y 1 3

power is conveniently used in the form
(1+ a :) /

4.7 The Exponential Function

Consider the function y ax with = ,
was done above
as (see portion “c” in
a >
=
1. The graphs of this function for
(4.6.1)).
a 2 and a =
3 are shown in Figure
To acquire the necessary facility in
4.7.1a. x When 1 for arbit- = =
0, y
handling these rules, a good 10 to 20
rary a (all graphs pass through the
exercises devoted to the pure techniques
point (0, 1)).
(without regard for the physical or
For all x the function y is positive
,
geometrical content of the problems
and grows with increasing x so that the
involving finding derivatives) are def-
initely needed.

Exercises

Find the derivatives of the following func-

tions: 4.6.1. y = x s (x 2 — l) 2 . 4.6.2. y —
x3 Y x2 + x. 4.6.3. y =x b
Yx2 — 1 (a:
3 — 2x) 1
/
5
.

4.6.4. y=(x - 1
jjrr-)
'
Y — 2. 4.6.5. y =
V
y x

x2 Y Yx + x- y — (^Yx + -at-
j
Y~x

= y-- _
x2 +x + i
4.6.7. y -. 4.6.8.
1— a;
2 X2 — 4- 1
4.6.9. y = (x-l)(x + 3) 4.6.10. y =
x+i
3a: — 1 x
-
Y^+ 2- 4.6.11. y
b
Y x2 —
4.6.12. y = . 4.6.13. y— Y- x 2 -\-xY x
Yx+i
-

4.6.14. y = Y 2
+ Vx • 4.6. 15. y = Figure 4.7.1
4.7 The Exponential Function 141

derivative positive asis everywhere Thus, we have

well. When
increased by a constant x is
quantity c we get y (x c) a x+c
, + — =
a c ax ba x —
by (#), with b =ac = ,

that is, the quantity y is multiplied by (where, of course, the constant on the
a constant quantity. Thus, if x is var- right-hand side has been determined
ied in successive fashion, by indentical only approximately, albeit reliably).
steps (in arithmetic progression ), Hence,

x == x 0 £0, + c, x 0 +2c, . . ., x 0 +nc, ..., ^-(10

x
)
~ 10x x 2.3. (4.7.1)

then y will assume the values We found the derivative of 10* in

n
experimental fashion, so to speak, by
*/o> t>y 0l 62y 0 ,
. . ., b y0 ,
. . ..
means of an arithmetic experiment.
For any other exponential function it is
It will be recalled that such a law of
now easy to reduce the problem to the
growth is called a geometric progression. preceding one. Using the concept of a
Let us find the derivative of the ex- logarithm, we write
ponential function for a = 10. 4 -
12
We '

can write a = 10 log a

,
ax = 10x losa . (4.7.2)

d( 10*) _ 10*-t-ds_lQ*:
_|QX 10*«-1 By the rule for finding a derivative of a
dx dx dx composite function, we get

where dx (instead of Ax) emphasizes the ~ 10x log a

X 2.3 log a
passage to the limit as Ax 0.
413
What is the quantity (10 dx — 1 )/dx?
= ax X 2.3 log a. (4.7.3)
It is the limit of the ratio (10** — 1)/ The remarkable peculiarity of the
Ax as Ax
Let us find this limit 0. exponential function is that its deriva-
numerically, arithmetically. Using a tive is directly proportional to the func-
four-place table of logarithms or a pock- tion itself :

et calculator with a y x key, we get

da x
ka x , (4.7.4)
dx
10 01 ~ 1.2589, ~ 2.589 ;
where the proportionality factor k de-
pends on base a. Therein lies the chief
100-01 ~ 1.0233, 1H__—L /v/ 2.33 ; property of a geometric progression:
the greater the quantity, the faster the
IQO.ooi ~ 1.0023,
/jnO.OOl ~ - a

growth. 4 14
om 2.3.
If 0 a < < 1, the graphs of the ex-
ponential function will be as shown in
4 * 12 =
10 is taken simply to facilitate
a
computation
Figure 4.7. ib. When x increases in arith-
involving a table of base-10
logarithms. The reader that owns a pocket metic progression, y diminishes in
calculator with a y x key can write out the geometric progression. Formula (4.7.3)
values of, say, 2 0 1 2 0 01 ... or 3 0 1 3 0 01 -
,
-
,
-
,
-
,
. . .
is still applicable, but now log a is
and the ratios corresponding^ to these val-
ues: (2
0 - 1 — 1)/0.1, etc.
negative (since a 1) and, hence, the<
4 ’ 13 It is this fact that makes us write derivative, which is proportional to
the function, is of opposite sign. In
d(10*)
— = l0*+d* — 10^, an exact equa tion) instead ^ 4 14 This
property of geometric progres-
~dx ~dx
.

sions, their exceedingly fast rate of build-up,

, d(10)*
“ 10
x+Ax — 10* .

a PP rox i ma i e
is a favorite topic in many popular-science
811 books on mathematics, for instance, in
dx ~kx
Ya.I. Perelman’s Mathematics Can Be Fun
equation). (Mir Publishers, Moscow, 1985).
142 4 Calculation of Derivatives

Part 2 we will give some instances of a Let us find the number e solely on
quantity diminishing in time in such a the basis of formula (4.8.1). By the gen-
manner that the rate of decrease is pro- eral properties of exponential func-
portional to the quantity at a given in- tions, e° —
1 Let us consider the func-
.

stant: tion y e
x = We
have y (0) . 1 and =
y = 1 (see (4.8.1)).
= — cy, c > 0. (4.7.5)
(0)
We take a small Ax r and com- =
From the foregoing evident that
pute the increment of y x
e when pass- =
it is
ing from x 0 to x r. We =
know that =
in this case the solution of the problem we can write Ay~y'(x)Ax, which
the exponential function
is
means that Ay ~ 1 X Ax — r and
y =y 0 a
x
,
a< 1. (4.7.6) y (x) = y (0) + Ay whence
,

Exercises
1 + r. (4.8.2)

This equation should be interpreted

Find the derivatives of the following func-
in the same way as we interpreted
tions:
4.7.1. y = 2*. 4.7.2. y = 5* +1 4.7.3. y
.
= y (x + Ax) ~ y(x) + y'(x) Ax that
,

is,in (4.8.2) all terms of the second and

(1/2)*. 4.7 .4. y = 10^. 4.7.5. y=2
4.7.6. y = 2 x+1 f x .
higher orders, or terms proportional
3
to r 2 r ,
etc., are ignored, while only
,

4.8 The Number e the first order term (proportional to r )

is retained. What is important here is
Let us find a base a for which the deriv- that the coefficient of r in (4.8.2) is ex-
ative of an exponential function y = actly unity— this follows from the prop-
a x is of the simplest form, namely, erty of the number e formulated in (4.8.1)
such that the coefficient in the expres- (this is simply the definition of e).
sion for the derivative, k in (4.7.4), We write the small number r as a
is equal to unity, so that it need not be fraction with a large denominator,
written at all. We
denote this base by = r<C 1,
r 1 In: if then n >>1. 416 Then
e. Thus, by definition,
from (4.8.2) we get e 1/n 1 + 1 ln r ~
dex . ^ .., whence e
. (1 4- 1 /n)
n ~ .
x
(4.8.1) Since formula (4.8.2), which i
dx
This number nores terms proportional to r 2 r 3 etc
is easily found by formu- , ,

la (4.7.3):
is the more exact the smaller r is, tl
last expression for e is the more exa
2.3 log e ~ 1, log e ~ 1/2.3 ^ 0.43, the larger n is. have thus arrive We
whence, referring to a loga- table of at another rigorous definition of e :

rithms, we get e 2.7. This practical ~ 1

approach, however, does not follow the e— lim (l-\
n
^ (4.8.
historical development of mathematics n ^°°
and fundamentally unsatisfactory.
is
We made use of numbers taken from logarithmic tables. Even less suitable a
logarithmic tables and did not stop to school logarithmic tables for finding a value
e that can easily be remembered (there
think how they were computed. 4 15 *

really no sense in this!): e 2.7182818284590 ~

4 - 15
(15 decimal places!). Note that e cannot
The accuracy with which e can be expressed by a periodic decimal fraction.
determined in this manner is low; it is highly On questions concerning the formulas (ai
improbable that, using a four-place table of methods) that enable finding e with a ve]
logarithms for determining the derivative of high accuracy, see Sections 6.1 and 6.2.
10*, you could find the correct value of the 4 16
The notation r<$; 1 means that t]
-

second digit after the decimal point in e. number r is very small compared to unit
10 dx
Actually, — i
^ 2.302585, but of course
similarly, the notation n 1 means that n >
dx a very large number (n is considerably great
this value was obtained without the use of than unity).
 *

4.8 The Number e 143&

(this
n
is read: e is the limit of (1 + many times will production increase^
1 !n) as to infinity). 417
n tends over a period of 50 years if the annual
Similarly, for a fixed k and a small growth rate is 2%? We have to compute-
r (such that kr <C 1) we get e hr ~ + kr
1 ,
1.02 50 The use of the number e con-
.

which yields, if we put r= 1 In sists in our setting, approximately^

(where n is very large), eh ~ (1 + 1.02 =
e 0 02 (see formula (4.8.2), whence-
-

k/ri)
n
,
or, to be more precise, 1.02 50 ~
£>0.02x50 e =
2.72. ~
The*
n general formula is
e
k = lim (l + "~) *
(4.8.3a)
(i + r) m ~em \ r< 1. (4.8.4>
One should not fear such words as
“limit” or “infinity”. Actually (1 + To apply this formula r must-be-
1/100)
100 ~ 2.705, which is only small (there is no other requirement),
"Jiigrtliy Aiftreml t ^rrorn + ti?e u
while^ m . ancL mv. ne&cL not . Tf .

of e. We advise the reader to find (1 + mr is also small, then emr 1 -f mr r ~ .

1/8)
8
for himself. and we obtain the earlier formula
We
have found that e r cz. 1 r for + m ~1+
small r, and this is the more exact the (1 + r) mr (4.8.5>
smaller r is. 418 Let us check this using (see formula (4.5.4)). However, we can-
numbers. We set up the following ta- not use formula (4.8.5) if mr is large,
ble:
419
whereas expression (4.8.4) remains val-
r -0.05 -0.4 -0.3 -0.2 -0.1 -0.01 0
id (provided that r 1). For the exa-
<
<
1 +r 0.5 0.6 0.7 0.8 0.9 0.99 1 mple given above, the exact value of
e
r
0.6065 0.6703 0.7408 0.8187 0.9048 0.9901 1 1.02 50 (with three decimal places) i&
2.693, formula (4.8.4) yields 1.02 50 ~
r 0.01 0.1 0.2 0.3 0.4 0.5 e
1 ~2.72, and formula (4.8.5) yields
1 +r 1.01 1.1 1.2 1.3 1.4 1.5 (1 + 0.2)
50 ~ + 50 X 0.02 = 2. Com-
1
r
e 1 .0101 1. 1052 1 .2214 1.3499 1.4918 1.6487 putations using e yielded an error of
about 1% whereas formula (4.8.5) yield-
We see that even when r = ±3, the ed an error of about 25%.
error introduced by (4.8.2)
does not
exceed 6%. A physicist or engineer In general form an estimate of the preci-
sion of formula (4.8.4) is given in Section 6.1
must remember not only that e (see Exercise 6.1.4). It will be found there-
2.72 but also that e 2 7.4, e
3 ~ ~ that the relative error introduced by this for-
20, e
4 ~ 55, e b cx 150.
Values of mula is of the order of mr2 so that formu- ,

e and e~ x
x
are given in Table A4.1 la (4.8.4) can be employed when mr 2 1
and formula (4.8.5), as we already know, only
at the end of the book. when 1. But since (alas) the condition
The number e greatly simplifies the mr <C 1 for r small is much stronger than
solution of problems involving geomet- mr 2 < 1, there are values of m and r for

ric progressions and compound interest. which the approximation (4.8.4) is valid but
(4.8.5) is not. For instance, in the above-con-
Consider the following example. How =
0.02 and m =
sidered case of r 50 we have
4 * 17
mr — 1 and mr 2 = 0.02, which means that
Of course, from the viewpoint of a pe- mr 2 is small and mr is not. In the same manner,
dantic mathematician the statement that from for r = 0.001 and m — 10 000 we have mr — 10-
e
i/n r^, 1
-f- 1 In follows e (1 1 /n) n ~ + and there is no way in which condition mr <c 1
is not rigorous, since both sides of the approx- can be satisfied, while the condition mr 2 < 1
imate equality (valid only for n 1) cannot » can be assumed valid (cf. Exercise 4.8.2).
always be raised to a large power n. Thus,
our reasoning does not prove the definition
(4.8.3), but just explains it. Nevertheless,
In accordance with the original defi-
we hope the reader will find the reasoning nition of the number e by formula
convincing. (4.8.1), the derivatives of exponential
4 18 Detailed tables have been
*
compiled for functions are especially simple when the*
the function y ex — .

4 19 The values of e x are taken

*
from a base is equal to e These derivatives are
.

four-place table. conveniently expressed in terms of the*

 •

144 4 Calculation of Derivatives

function itself. Here are a number of

formulas:
dy
y = eA x

— rp x dy ^ eX
yy
7
’
dx
=c dx
Ce x = y\
dy dekx
y = Ce
hx
,
dx
Q dx

= Ce hx d (kx)
ky,
dx
de m (x) dm
y=--e
m (
x
\
dy
dx dx
=• e
m W (x)

dx
Figure 4.8.1

—y 77
dm ix)
1

dx The simplest exponential function is

the one whose curve intersects the y
m^
y = f(x) e ,
-=/'(*) «"(*>
axis at (0, 1) at an angle of 45° (strictly
speaking, the tangent to the curve at
+ / (x) e m <* W (x) = !/ (
-y^p +
m' (x)
)
. this point intersects the x axis at an
angle of 45° (Figure 4.8.16). From
The exponential function of x to (4.8.1) it follows that this curve is the
.base £ is commonly written Often, one corresponding to the function e x
however, another designation for e x (recall the geometric meaning of the
is used: y = exp ^(read: the exponential derivative).
of x). The law e x is called the exponen- Conclusion. To summarize, we can
tial law and the function e x is termed give four distinct definitions of the num-
the exponential function (in the narrow ber e\ (1) from the condition (e )'
x =
sense). The designation with exp e
x
(2) from the condition
,
e
r
1 ~ +
is especially convenient when x proves r for r<l, (3) as the limit of
n
to be a complicated and unwieldy ex- (1 +
1 ln) as n -> oo, and (4) as the
pression. For example, it is more con- base in the exponential function y e
x =
(the graph of this function intersects
venient to write exp than
the y axis at an angle of 45°).
/7 * 2 + 24 * \3
K *3 + 5 ' The number e has other remarkable
<e .

definitions and properties; some of

The number e can also be defined geo-
these will be discussed in Chapter 6.
metrically. Two exponential functions,
To fix this important material in his
y =
a x and y x 6
X
are connected via = ,
mind, the reader is advised to
a simple relation:
put aside the book and see how from
b
x — a
hx
,
where k = log a b (4.8.6) each definition there follow the other
three regarded as properties of number
(this fact has already been used in for- e.

mula (4.7.2)). From (4.8.6) it follows Note, finally, that the equation in
that the graph of y x is obtained from y =e
x
x and y are, of
the quantities
the graph of y by shrinking the latter course, dimensionless numbers, since
A-fold along the x axis (Figure 4.8.1a; e
x
with x m
100 or=
x = 5 h has no
cf. Section 1.7). meaning. Consequently, in the laws of
It is clear that all graphs of the func- science that involve the exponential
tion y =
a x corresponding to different law connecting two dimensional quan-
T
v\hu>5b nfx a* "puss *hm)iq>ix an ft ’pun/*, +
iJin5b, T
y onih ?&,, wa ohvAjyb hmvb j
y
{0, 1) (since a0 — 1 for arbitrary a); l 2e
x/l
\ where l x and Z 2 are the units
however, the graphs intersect the y of measurement of x and y. The transi-
axis at this point at different angles. tion to a new unit for x, or Z' = cl x ,
4.9 Logarithms 145

always leads to a transition from the

old function y e
x
=
to a new function
y =e
cx
(equivalent
'

to the old one),

where x —
xlc is the same quantity x
but measured in new units Z'. Similar-
ly, the substitution 1'2 l^d for the =
unit for y leads to a transition from for-
mula y = e x to y' = dex (where y'
is measured in units ofZ'). Correspond-
ingly, functions of the type y
all =
ae kx ,
a and k may be arbitra-
where
ry, are called exponential functions,
since in the majority of cases a transi- Figure 4.9.1

tion to new a and k means only a

change in the units of measurement, so a raised to any power yields a positive
that only the fact that the two quanti- number, there are no logarithms of neg-
ties (x and y) are connected by an ex- ative numbers. 4 20
ponential law and the signs of a and As seen from Figure 4.9.1, the deriv-
k (k positive corresponds to an increas- ative of the function y = log a ^
ing function y e
hx
and k negative = (for a > positive for all values of
1) is
to a decreasing function y e~\
h x
)
= \
x, since y increases with x (for a 1). >
have physical meaning. Note also that the rate of growth of
y =
loga & with x decreases, that is, the
derivative decreases as x increases (be-
Exercises
low we will give a rigorous proof of
4.8.1. Find the derivatives of the following this).
functions: (a) y e~ x (b) y —
5e x e 3x ,
= — ,
Let us derive a formula connecting
(c) y e =
(d) y=e v
x
,
and (e) y=e , the logarithms of one and the same
4.8.2. The interest rate on a sum on a
bank account is 0.1% every month. How much number to different bases. Suppose that
will the sum
in the bank account increase in a
= h.
thousand years? [Hint. Use formula (4.8.4).] t = \oga h, af (4.9.2)

Taking the logarithms of both sides of

4.9 Logarithms (4.9.2) to base 6, we get

By definition,
the logarithm of a quanti- / log 5 a = log 6 h, whence f = .

ty h to a base a is the exponent / of

the power to which a (the logarithmic Taking (4.9.2) into account, we get
base) must be raised in order to obtain
the given number h : 1 °S.* = -Kg7- < 4 9 3>
- -

/ = 1 og a h means that h = a f
. (4.9.1)
This, can_ ha cawcitten. as,
In this sense the logarithmic function
log& x — k log a x, where k=log b a.
y =
log a x is the inverse (see Section
(4.9.3a)
1.6) of the exponential function y =
ax .
The number k == log & a is known as the
The curve representing the relation- modulus logarithms to base b with
of
ship y log a x (with — 1) is depict- a> respect to logarithms to base a. From
ed in Figure 4.9.1. Note that at x 1 = (4.9.3a) it follows that the graph for
(irrespective of the value of a) we have
y = log 6 x is obtained from the graph
y = 0, for x > 1 we have y > 0, and for y —
log a x by a /c-fold stretching
for x < y 1 we have
0. The entire < along the y axis.
curve is located to the right of the axis
of ordinates: since a positive number 4 20
*
See, however, Section 14.3.

1 0-0946
 146 4 Calculation of Derivatives

Logarithms to base e are called nat- logarithms of both sides; since In e r =

ural logarithms and are denoted by In x . r, we arrive at an extremely im-
The “naturalness” of this system of portant formula:
logarithms is due to fact that in some ^
respects this system appears to be the
In (1 + r) r (4.9.5)

simplest; the rate of

for instance, (the approximate equation has the
growth (the derivative) the simplest is same meaning as (4.8.2): it is valid for
in the case of logarithms if y In x. = r<c 1). Combining (4.9.4) and (4.9.5),
These remarkable properties of natural we get
logarithms, which we will repeatedly
4
encounter below, led to a situation in d In a: = In (1
\
+
1
— =—X )
) X
which the two creators of the theory
of logarithms, the Scottish amateur (employing dx instead of Ax makes it
mathematician John Napier (1550- possible to assume that the last equa-
tion is exact), whence
1617) and Swiss watch and instrument
maker Joost Burgi (1552-1632), inde- d In x _ 1
(4.9.6)
pendently and almost simultaneously dx x
discovered precisely this type of loga-
rithms (or very similar system of loga-
Whenx varies in geometric progres-
sion, In x varies in arithmetic progres-
rithms). The base- 10 logarithms are
sion, that is, if x = a, ab, ab ab
2 3
known as common logarithms and were , , . . . ,

by Henry Briggs then lnx = In a, In a c, In a + 2c, +

considered
c = In b.
first
(1561-1630), an English astronomer,
In a 3c, + ., where . .

For this reason, the larger x is,

geometer, and numerical-table maker,
the slower In x grows and the small-
prompted by Napier, who was his friend
21 er is the derivative, which is reflected
and whom he greatly admired. -

be in formula (4.9.6).
Natural logarithms can character-
The derivative of a natural logarithm
ized by the property that the graph of
= can also be found by using the fact that
y In x intersects the axis of abscis-
the logarithmic function and the expo-
sas at point (1, 0) at an angle of 45°,
nential function are inverse functions
since the curve is obtained from that of
e =
by reflection withx (with respect to each other). We can
the inverse, y ,
write
respect to the bisector of the coordinate
angle (see Section 1.6), and the curve — In#, — ey d(eV)
= ev
y x x'
y e=x
intersects the axis of ordinates
,
dy ,

at point (0, 1) at an angle of 45° (see dy _

Figure 4.8.15). dx
1
eV
= 1
x
#

Let us now find the derivative of a

Now, using (4.9.3a), we can calculate
natural logarithm. We consider d In x=
— the derivative of a logarithm to any
In (x + dx) In x. Take advantage
=
of the familiar formula In a — In b = base a. Suppose that y log a x Then .

In ( alb ). Then
(see (4.9.3) and (4.9.3a))

In £ dy d In x
— In ~ 1 1

<7 In a: = In (l + -f-) . In a
9
dx In a dx In a x

(4.9.4) (4.9.7)

We already know (see formula (4.8.2)) Replacing a with e and b and h with a,
that e r 1 ~ + r for r small. Take the we get In a =
(log a c)”\ we can rewrite
(4.9.7) in the following manner:
4 21 In the beginning, common logarithms
-

d l0gq X lOggg
were known as Briggs logarithms and natural = X *
\
(4 g 7 a )
\
logarithms as Napierian logarithms. dX
 4.9 Logarithms 147

The simplest one of the formulas ative of fh computed on the assumption

(4.9.6), (4.9.7), and (4.9.7a) is (4.9.6), that only h is a variable while /
which is valid for logarithms to base is held constant, and the second term,
e For rough mental calculations, it is
. h (x) f (x)^*)- 1 f'(x), is the derivative
advisable to memorize the following of fh computed on the assumption that
facts: In 2 0.69, In 3 1.1, ~and ~ only / is a variable while h is held con-
In 10 ~ ~
short table
2.3 1/0.434. A stant. This confirms the general prin-
of natural logarithms is given in Table ciple expressed in Section 4.2.
A4.2 at the end of the book. Note, finally, that in y log a x =
If some function f (x) is under the (just as in y =
a x ), the variables x
sign of the logarithm, the derivative and y and the constant a are dimension-
is found by the rule for differentiating less. If # is a dimensional quantity
composite functions (see Section 4.3): (say, length), then the number x is
specified only to within a constant fac-
d In / (x) __ 1 df (x) ,4 g gx
\ /
tor depending on the chosen unit for
dx f (x) dx '

x. Correspondinly, the quantity lo g a x

The derivative (In / (x))' of the loga- (for any fixed base a) is determined on-
rithm of a function f (x) is often called ly to within a constant term (to within
the logarithmic derivative of / (x)\ ac- an additive constant as mathematicians
,

cording to (4.9.8), it is equal to the ra- usually say in such cases); in that respect
4 22
tio of the derivative f (x) to the func- it is similar to the indefinite integral. -

tion itself, / (;r). Here is an example If both x and y are dimensional quan-
that illustrates the usefulness of this tities, then the right way to approach
concept. this problem is to write y= l 2 log a ( xll x ),
Formula
(4.9.8) enables one to find where the factors l x and Z2 are
the derivatives of expressions of the units of measurement for x and y,
form / (^) /l 3C) that is, such that contain
(
,
since a logarithm can be taken only of
the variable both in the base and in the a dimensionless quantity, say of num-
exponent. Suppose that ber 100 obtained as a result of taking
the fraction, say (distance AB)!(x\mi
y = f (x)
h W. (4.9.9)
Z x ), where l x could be one meter. The
Taking logs (the logarithms can be taken value of a logarithmic function is also
to any base; we choose natural loga- a dimensionless quantity, whereby the
rithms), we obtain dimensional quantity must be equal
to Z 2 log a (z/Zi), that is, log a xll x ) units
In y — h x ) In(, / (x). (4.9.10) Z2. These simple considerations must
(,

Let us now take the derivatives of be allowed for since the logarithmic
both sides in (4.9.10) and have regard function (just as the exponential func-
for the fact that In y is a composite tion) is often encountered in the laws
function of x (just as In f(x) is): of science.
It must be
also noted that, in view of
'^ h '
(*) ln f( x ) + h W-jy]- (4.9.3) and
(4.9.3a), the transition from
Y y *

one base a to another base b is reduced

y' =y [V (x) In f (x) + h (x) -j~-] ,
to multiplying all logarithms to the old
base a by a constant, k= lo g b a that is, ,

or, allowing for (4.9.9), to a change in the unit for y log a x. =

This is why, in the majority of cases, the
y
f
= f (x)
h ( X)
h' (x) In / (x) base of a logarithm does not play an !

+ h{x)f {xfW-'f (x). (4.9.11) 4 22

As for the deep connection between
•

Consider this formula. On

the right logarithms and integrals (which partially
explains this analogy), see formula (5.2.2)
we have a sum of two terms: the first
(which sometimes is used to define logarithms)
h
term, / (x) Wti (x) In / ( x ), is the deriv- and the subsequent discussion in Section 7.7.
10 *
 x _ d

148 4 Calculation of Derivatives

important role, which makes it possible

almost always to employ natural loga-
rithms as being the simplest.

Exercises

4.9.1. What is the value of log 3 15? [Hint.

Use formula (4.9.3)].
4.9.2. Derive the formula for the deriva-
tive of (a) the product of two functions, using Figure 4.10.1
the formula log ( uv ) log u —
log z;, and +
(b) the quotient of two functions, using the by any units of length (centimeters,
formula log (u/v) log u log v.= — inches, or meters). The sine is equal to
4.9.3. Find the derivative of each of the
— In the length of the line of sines (in
following functions: (a) y (a: +3), (b) = z/

In 2x, (c) y ln(z 2 =

1), + (d) y = ln(x+i/x), centimeters) divided by the length of
the radius (in centimeters,) and only
(e) y = ln(3*»-*+l), (f) y = In when =
,
r 1 cm is it numerically equal
Vx
<g )y =ln 3 —
=T, (h)
V *+i
y = x\nx, (i) y = to the length of the line of sines. The
lines of sines and cosines are shown in
x 3
In ( 1), (j) y — xx , and (k) y = Figure 4.10.1.
y'xV**- 1 .
Recall the form of the graphs of sine
and cosine as functions of the angle
(Figure 4.10.2). The period of the sine,
4.10 Trigonometric Functions like that of the cosine, is equal to
In this section we will find the deriva- 2n ~ 6.28 and corresponds to a complete
tives of trigonometric functions. revolution of the radius of the circle.
Trigonometric functions are defined Let us find the derivatives of the sine
as ratios of line segments, ratios of the and cosine geometrically. In Figure
sides of right triangles or ratios of cer- 4.10.3, the endpoint of the radius drawn
tain line segments in a circle (the line at an angle cp is A, and the endpoint of
of sines, the line of cosines, etc.) to the radius drawn at a close angle
the circle’s radius. For this reason trig- cp+ dtp is B. The length of arc A B
is therefore equal to dcp. Draw from A
onometric functions are dimensionless,
and the independent variable in such a perpendicular AC to the line of sines
functions is also dimensionless, an an- BB of the angle cp -f- dcp. As can be
r

gle. The natural measure of angles in seen from Figure 4.10.3,

higher mathematics is the radian a ,
= sin
^4^1' cp, BB' — sin (cp -f-
central angle subtended in a circle by an BC = sin (cp -f-
dq>) — sin cp
arc whose length is equal to the radius
of the circle. A
central angle of t radians
— d sin cp

is subtended in a circle of radius 1 by (here, of course, the last equation is

an arc whose length is t The importance . “exact in the limit”, as dcp ->0). More-
of trigonometric functions to the de- over,
scription of physical processes will be = cos OB' — cos
OA' cp, (cp -f- dcp),
discussed in Part 2 (see Chapter 10).
In what follows we will always con- A'B' = AC = cos — cos cp (cp -f- dcp)

sider a circle of unit radius. Then we = — cos cp.

will briefly say that the sine is equal to

the length of the line of sines in that
circle, the angle is equal to the arc
length, and so on. The reader must bear
in mind, however, that both trigonomet-
ric functions and angles are dimen-
sionless quantities and are not measured
 p t

4.10 Trigonometric Functions 149

Therefore, strictly speaking, the above rea-

soning is not rigorous and the relationships
are incorrect. However, they are true to within
infinitesimals of first order with respect to the
small quantity Acp. Hence, while for incre-
ments we have only the approximate equations
A sin cp ~ Acp X cos cp and A cos cp ~ —Acp X
sin cp, the corresponding equations for differ-
entials are exact, which means that Eqs.
(4.10.1) are exact, too.
The derivation of the formulas for the de-
rivatives of trigonometric functions we have
just described is equivalent to the derivation
based on mechanical analogies (for one, a deri-
vation that incorporates the idea of a vector).
We imagine a point (t) whose move- M= M
ment in the plane is described by two para-
metric equations:
Since the angle dcp is small, the arc * = T (*)> y = ^ (t), (4. 10. 2>
length AB
does not differ much from
where t is the time variable (see Section 1.8).
the length of the chord and the an- AB In this case the rate of variation in the coordi-
gle ABC
formed by the chord and AB nates x and y of point will be given by the M
the vertical line BOB' is equal to cp
4 23
.
*
derivatives x' dx/dt and y' =
dy/dt. The =
From a consideration of triangle ABC vector MM!/ At v av =
(Ax/At, Ay/ At), where =
= AB Ax/ A and Ay/ At are the coordinates of the
we find that BC coscp and
vector (see Figure 4.10.4), characterizes the
AC = AB sin cp. Thus, average velocity over the line segment X
MM
d sin = cos cp dip,
of the path (here 1 (t At)). VectorM =M +
cp
v =
(dx/dt, dy/dt) characterizes the instan-
—d cos cp = sin cp d( taneous velocity at point of the path: it is M
directed along the tangent to the trajectory at
and, hence, point M
and its length is equal to
As ds
lim —rr — -jr
.
d sin (p
= cos cp,
d cos cp
— sin cp.
.

1X1 at ,
where As ~ MM| ± |
is the
dcp d<p At-* 0
increment of the distance s over time A*.
4 10 . 1 )
(
.
Now suppose the point M moves uniform-
Such a simple and pictorial way of calcu- ly in a circle of radius 1; it travels, say, 1 cm
lating the derivative of the sine or cosine is in 1 s. Then Eqs. (4.10.2) assume the form
quite convincing for a beginner. On the other x — cos t, y = sin t, (4.10.3)
hand, an experienced and knowledgeable reader
(not the one for whom this book is intended) and the velocity v of the motion is given by a
will notice the difficulties inherent in such an vector with coordinates d (cos t)/dt, d (sin t)/dt.
approach. For a small but finite (i.e. not in- Clearly, in this case vector v will be of unit
finitesimal) angle Acp the length of chord AB lengthf (since byf hypothesis the increment
is less than the arc length: it can be proved

that
V
(Acp ) 3
AB = Acp
12
"
r
,

Furthermore, actually the angle CBA is cp +

Acp/2 instead of cp, so that
(Acp) 3
BC = AB cos (cp-f- = Acp
j ( 12

+ j
(cos cp — ^ sin cp+ ...
j
.

4 23-
The tangent line t to the circle at
point A forms with the vertical line A A' an
angle A' At equal to cp (Z A' At z A' — AO
as angles with mutually perpendicular sides);
A ABC differs very little from A A' At since
chord AB
has practically the same direction
as tangent t. Figure 4.10.4
 M

150 4 Calculation of Derivatives

As of distance is equal to the increment At The second term on the right-hand

of time) and directed along the tangent to
the circle (at every point; see Figure 4.10.4).
side of (4.10.4)must first be trans-
The coordinates of vector v, which is perpen- formed by using the familiar formula
dicular to the radius r of the circle, OM = 1 —
cos 2a — 2 sin 2 a whereby 1 — ,

are obviously —
sin t and cos t (compare the cos Acp =
2 sin 2 (Acp/2). In this formula,
right triangles and AOM
whose sides VMP we substitute Acp/2 for sin (Acp/2)
are mutually perpendicular in Figure 4.10.4;
they have the same hypotenuses OM = V = (this is justified since Acp is small).
1 and equal acute angles A AOM = A MVP = Hence,
t;the minus sign in front of sin t here specifies
the direction of the line segment MP). Thus, 1 — cos Acp ^ 2 (Acp/2) 2 _ Acp
we again arrive at formula (4.10.1): Acp Acp ~ 2
*

d sin
d cos
dt
t
— sin t,
It
t
COS t.
Hence, in the limit, as Acp -> 0, the
second term vanishes: lim
1
~^ os A(P ==
A<p-0 A(P
Finally, we will give another meth-
od for calculating the derivatives of 0, whence lim
A(P
:= =
A<p-0 ^q>
sin cp and cos cp, which does not require
cosqp. Similarly,
a drawing and is more formal. According
to the general formulas, A sin cp = A cos cp = cos (cp 4- Acp) — cos cp
sin (cp Acp) + sin cp* Recall — the
= cos cos Acp — sin sin Acp — cos
cp cp
formula for the sine of the sum of two q)

angles, = — cos (1 — cos Acp) — sin sin Acp

cp cp

sin (oc
-f- (3)
= sin a cos (3
(here we have used the formula for the
+ cos a sin (3,
cosine of the sum of two angles), and
therefore
and apply it to sin (cp Acp) to get
-f-
A cos cp
— cos 1 — cos Acp
sin (cp -f Acp) — sin cp cos Acp Acp
= cp
Acp

+ cos cp sin Acp,

— sin q>
sin Acp
Acp
whence
A sin cp == sin cp cos Acp
Since we already know that
— cos Acp
4 cos cp sin Acp — sin cp lim
A<p-**0
1

Acp
=o,
A<p-0
lim _ 4 A(P
A(P
1,

— cos cp sin Acp

the final result
— sin cp (1 — • cos Acp).
is

d cos A cos
Let us form the ratio of the increments
cp
~ j.
cp

dv A <p™ A(P
as follows:
= — cos x 0— sin cp cp x 1 = — sin cp.
A sin cp sin Acp
cos cp
Acp Acp Equations (4.10.1) are valid for ar-

— 1 — cos Acp bitrary angles and not only for those in

sin cp (4.10.4) the first quadrant. It is also useful to
Acp
verify, glancing at the graphs of the
Now we have to send Acp to zero and functions sin x and cos x, that Eqs.
thus pass to the limit. We know that (4.10.1) give the proper signs of the de-
for angles a or Acp tending to zero, the rivatives for any x.
sine is equal to the arc length, sin a ~ Let us check the validity of Eqs.
a, or sin Acp Acp as Acp ->0. In (4.10.1) for small angles. For small cp,
other words, it is obvious geometrically that
sin cp q)~ and cos cp 1, where the ~
sin Acp
lim approximate equality sign just as
Acp
A<p-*0 in many other cases, means equality to
4.11 Inverse Trigonometric Functions 151

within terms of the first order of By a completely analogous device we

smallness. The first equation in (4.10.1) find that
for small 9 yields d (sin (p)/dcp 1 and ~ d cot x
the second yields d (cos cp)/rfcp cp, ~— dx
1

sm 4 x
(4.10.6)
which means that d (cos 9)^9 0 at =
(p= 0. The fact that the derivative is The derivatives of a tangent and a cotan-
zero signifies that the cosine has a max- gent can also be found directly. Note that
imum at cp =
0, while the fact that
— tan p0 sin a sin 6
= tan a
=
,
£-
d (sin 9)/^ 1 at 9 0 signifies that cos a cos p
for small 9 the sine increases approxi- sin a cos — sin ft
cos a _ sin (a — ft)
mately like 9. ~ cos
ft

a cos P ” cos a cos |3

’

If we know derivatives of the

the
functions y sin x= and y cos x, = whence
it is easy to find the derivatives of all
A tan 9 = tan (9 + A9) — tan 9
other trigonometric functions by using
the formulas interrelating them. For sin A9
instance, tan x = — By the formu- cos (9 + A9) cos 9 (4.10.7)

la for finding the derivative of a quo- Bearing in mind that lim ^ _ 1, we

A<p-*-0 A9
tient we get
get, from (4.10.7),

d tan x d (sin s/cos x) d tan 9 A tan 9

dx dx c?9 ~ ay~>0 A9
__ cos x cos x
— sin x — sin x)
(
= lim
sinA(P- 1
COS 2 X
’

Av-*0 ^9 cos (9 + A9) + cos 9

whence = IX 1
*

2
x-\- sin 2 x cos 2 9 cos 2 9
d tan x _ cos __
1
,
gx
^
dx cos 2 x cos 2 x * '

Exercises

From Figure 4.10.5 (the graph of Find the derivatives of the following func-
tan x) we can see that the function tions:
~
sin (2x + 4.10.2. =
y =
tan x has a positive derivative for 4.10.1.
cos (x 1).
y
— 4.10.3. y
3).
cos = (x 2 — x +y 1).
arbitrary x Near the points of discon-
.
4.10.4. y =
sin 2 x. 4.10.5. y = sin 3x cos 2 x .

tinuity (x =
Jt/2, x 3ji/2, .) the= . . 4.10.6. y =
(sin 2x) x 4.10.7. . y —
x tan x.
derivative increases without bound, and 4.10.8. y e =
tan 2x
4.10.9. y . = cot (s/2).
both the function and its derivative
become infinite at these points. Both
of these conclusions are in full agree- 4.11 Inverse Trigonometric Functions
ment with formula (4.10.5).
New and very interesting results are
obtained when we
consider the inverse
trigonometric functions We remind .

the reader of how these functions are

defined. The function

y
— Arcsin x (4.11.1)

is an angle y whose sine is equal to x:

sin y = x. (4.11.2)

These two equations denote the same

thing. Similarly, the function

y = Arctan x (4.11.3)
 :

152 4 Calculation of Derivatives

Figure 4.11.2
Figure 4.11.1

principal value of the function Arcsin x

and is denoted by y = arcsin x. If we
is an angle y whose tangent is equal to confine ourselves to a consideration of
x
y =
arcsin x, then to each x there cor-
tan y = x. (4.11.4) responds only one value of y. Thus,
the function y arcsin x = is single-
The definitions are similar for the func- valued.
tions y =
Arccos x (i.e. x cos y) = The principal value of the arctan-
and y == Arccot (i.e. x =
cot y). Note gent function is defined in similar
that the functions y =Arcsine and fashion:
y = Arccos x are meaningful only for
— < < 1 x 1 (cf. (4.11.2)). The functions
— ~C arctan
y = Arctan x and y = Arccot x are
meaningful for all values of x. (since the function x = tan is mono-
Let us consider in more detail the tonic on the segment — Jt/2 <y y < jt/2,
function y Arcsin x = For instance,
. Figure 4.11.2).
let x = 1/2. Then y = Arcsin (1/2). Let us find the derivative of y =
We can take y = nl 6 since sin (n/6) = arcsin x. We take advantage of the
1/2; however, we can also take y = fact that the arcsine is the inverse of
5jt/6, since sin (5 k/ 6) is also equal to the sine, y =
arcsin x and x sin y. —
1/2. We can likewise take y = 13jt/6, We then have
y = 17k/ 6, and so on. We see that one
value of x is associated with an infin- x (u) ==—— = cos y,
dy
itude of values of y. All that this means (4 11 5)
dy
is that Arcsin x is multiple-valued y
a
;/
(x)
v
v

'
— —r~—
dx
= -77T
1

x (y)
- 1

cos y
.

(see 49) and even infinitely valued

p. .

These properties of y Arcsin x are = But we consider x the independent vari-

seen in the graph of Figure 4.11.1. able and not y, so dyldx should be ex-
Take the portion of the curve repre- pressed in terms of x and not in terms
senting the function x sin y on = of y, as in (4.11.5).
which the function is monotonic; usu- From sin 2 y -f co s 2 y = 1 it follows
ally the portion of the curve y = that cos y = ± ]/ 1 — sin 2 y. Since
Arcsin x on which this condition is we are considering the principal value
satisfied is such that n/2^. y k /2 — ^ . of the arcsine function, it follows that
This part of the curve is called the — jx/2^ jx/ 2, cos y^
0, and so
 ) ) 2

a Parameter 153
4.12 Differentiating F unctions Dependent on

we take the plus sign in f ront of the Exercises

radical sign: cos y = ]/ 1 — sin 2

y. Find the derivatives of the following func-

Since sin y x = by (4.1 1.2), it follows tions:

4.11.1. y =
arccos x (this function is de-
that cos y ~\f \ x2 = — . Substituting fined by the condition ji). 4.11.2. y 0< =
this into (4.11.5) yields arccot x (here 0 y jx) 4.11.3. y< < arcsin 2x. —
4.11.4. y arctan (3x —1). 4.11.5 y
.

+ =
——
dy 1
or
d arcsin x — —x— 1
arctan (;r
2
— z). 4.11.6. y
— e arctan
Y \—x Y
,

dx 2 ’ dx 1
2
l
’

(4.11.6)
4.12 Differentiating Functions
This formula may be used only for Dependent on a Parameter and
the principal value. If we want it to Functions of Several Variables.
be valid for other values, we must choose Partial Derivatives
the appropriate sign in front of the
radical Y —x
(this radical can be
1
2 Often in physical problems we are forced
to find the derivative of a function
considered double-valued). Indeed, for
one and the same value of x, the deriv- y (x) which, aside from depending on
ative has different signs: on various the independent variable x, depends
portions of the curve at points A and on one or several parameters, orfquan-
C of the graph of the function y = tities that characterize the system being
Arcsin x (Figure 4.11.1) the deriva- investigated, say, masses or linear di-
tive is positive and at points B and D mensions, and that are assumed con-
it is negative. stant, fixed, for the system considered
Let us now find the derivative but in general may have different val-
ar an X ues. Clearly, if the function y —
-
^ - If y~ arctan x ,
then x = tan y, x t) depends on the independent
/ ( ,

whence, by the foregoing, we find that variable x and the parameter t its ,

derivative y' =
dyldx will also depend
dx on the same parameter t. For instance,,
*'(») = 1

dy cos 2 y ’

(4.11.7)
if y =
sin (Xx) (X is the parameter) r
1 then y' x dyldx = X cos (tar) and
=
*<*>=-£=
dx x' (y)
cos 2 y.
y" (x) =
d2yldx 2
(.

X 2 sin (tar); = — if

y = e~ aX2 (a being the parameter),

From trigonometry we have then y' (x) dyldx =
2axe aX
~
and =
y" (x) = d?yldx 2
2a (2ax 2 =
l)e~ axZ —
tan 2 v
y + =—
cos —
1% (verify this).
Note that since the function y =
1
2 9
,
y

f (x,t and its derivatives y' =

and therefore
dyldx d2 y!dx 2 etc. depend not only
, ,

1
1 +x 2 .
on x but on parameter t which may ,

cos 2 y assume different values, it would be

interesting to know to what extent y
Using formula (4.11.7), we finally get and y' and y'\ all taken at a fixed val-
ue of x, depend on parameter t if the
latter is slightly varied. The answer
(i-ll.o)
dx dx l-j-z 2 *
lies in the value of the rate of change of

y with t, that is, in (dy/dt) x= constant-

This formula holds true for any other For instance, in the case of the two func-
branch of the arctangent (see Figure tions we considered above, yx =
4.11.2) since any other branch is ob- sin (tar) and y2 e~ aX '\ the rates =
tained from the principal one by a paral- dyJdX and dyjda are, respectively,
lel translation, and this does not affect x sin (tar) and —
x 2e~ aXZ (x is fixed but
the magnitude of the derivative. X and a vary).
154 4 Calculation of Derivatives

Obviously, if we are dealing with a z = F (x, y) we must fix both inde-

function y f =the notation y\
(x, t), pendent variables, x = x 0 and y = y0 ,

y", etc. is ambiguous since it does not that is, (dF(x, y)/dy) x = Xo , y=yo means
allow us to know whether we are speak- that we have the partial derivative of
ing of (dyldx) t= constant or of an entire- F {x, y) with respect to x calculated
ly different, quantity (dy/dt ) x== constant .
on the assumption that y — y 0 and
To explain the notation used here, we taken at the point x = x 0 (in other
turn to the general case of a function of words, the derivative of cp (x)=F(x,y 0 )
two (independent) variables z F (x, y), ,
= with respect to x, which is equal to
since the above is a particular case of
lim \ . Note also that if
such a function. True, the function was Ax->0 A* 1

denoted by y and the independent vari- the value z of the function F {x, y)
ables by x and t but nothing prevents ,
ismeasured in units of E and the varia-
us from choosing other letters. bles x and y are measured in units of e l
If we assume that y y 0 is constant = and e 2 respectively, then the partial
,

and x varies, we will arrive at a func- derivatives dzldx and dzldy have diffe-
tion z =
F (x, y 0 ) cp (x) of one vari-= rent dimensions, E/e 1 =/= EIe 2 *

able x The derivative cp' (x) of such a

. The partial derivatives of a function
function can, naturally, be written as z =F (x, y) enable estimating the in-
(dF (x, y)/dx) y=Vo= constant- Since such crement Az =F (x -f- Ax, y -f- A y) —
notation is cumbersome and not too F (x y) of
,
the function caused by
useful, we use OF (x, y)ldy, where the small variations in x and y, or incre-
symbol d (the “round dee”) is used ments Ax and A y that are small. Ob-
instead of d to emphasize the fact that viously,
the derivative is taken at y constant ,

that is, the function F (x, y) is consi- Az = F (x + Ax, y + Ay)

dered as a function of one variable x = [F (x + Ax , y + Ay)
{it depends, however, also on parameter
y). The derivative — F (x,y + Ay)]
dF (g, y) d£{x , y) + IF (x, y + Ay) — F {x, y)].
dx dx Inconstant

lim
F(x+_ As, y) — F(x, y) But the A xz F (x -|- Ax,
difference =
Ax-+0 Ax y -f- Ay) —F
y -f- Ay) is the incre-
(x,
ment of a function of a single variable,
is said to be th e partial derivative of
F (x, y -f- Ay) ^-f-Aj/.=constant *P (*^)>
F (x, y) with respect to the independent due to an
increase in
I

the indepen-
variable x: similarly, variable x by Ax; according to
dent
dF {x, y) __ dF(x, y) formula (4.1.2), A xz = (x Ax) —
cp -|-

= a^t^L ‘Ax.
I

J
^Uy a'y |x=confetafit
cp'(x) ^ cp' \x)Ax
__ L (£» y+ A y) ~ F( x
^ ' y)
Similarly, the difference F (x,
Ay y -|-
Ano Ay) — F (x, y) = A 2z is the incre-

the partial derivative of F (x, y)

ment of function
the (y)
=
is
with respect to y. For example, if
F (x, y) Inconstant of a single vari-
= = able y due to an increase in the inde-
y =
y (x, X) sin (>ix), then dyldx
pendent variable y by Ay: according
X cos (Xx) and dyldX x cos (Xx), — dF{ *
while if y y (x, a) =
e~ aX \ we = A 2z ~ I|)'(y) A y =
to (4.1.2), -
y
Ay.
have dyldx = —
2axe~ aX2 and dylda = y
And since we can always assume that
—x2 e~ aX2 .
^+ A ^
Of course, in calculating the values for Ay small we have
dx
of the partial derivatives dF (x, y)ldx dF(x y),
the final expression for the
-and dF (x, y)!dy of a function dx ’
4.12 Differentiating Functions Dependent on a Parameter 155

increment Az = AjZ + A az of the func-

tion z =F (x, y) is

V- (
4 - 12J )

It is clear the accuracy of the ap-

that
proximation (4.12.1) is the higher the ,

smaller the increments Ax and A y of the

independent variables x and y of function
z. In other words, for small Ax and A y
{which here are assumed to be of
the same order of smallness), the differ-
ence Az — Ax +
|^
Ayj will be of
a higher order of smallness, and, as Ax
and Ay tend to zero, the ratio of the
right-hand sides of (4.12.1) (where
left- to Figure 4.12.1
it is assumed, of course, that the
right-hand side does not vanish) tends
to unity. For instance, if z xz/
3 = ,

= The meaning of (4.12.1) can be

then, obviously dzldx y
3
and dzfdy = explained geometrically. A function
3xz/
2
and small increments Ax and
Ay lead
,

to an increase in z by approxi-
z = F (x, y) of two variables x and y
is defined at all points of a plane domain
mately y
3
Az + 3 xy 2
Ay = —Ax + ® (the domain of the function) where
(
each point is fixed by its two coor
PMAlthough the partial derivatives
dinates,
and N
x and y Points Af(x, y)
(x -f Ax, y -f- Ay)
.

are the
dzldx and dzldy of a function z = opposite vertices in the rectangle
F (x, y) are measured in different n = MPNQ depicted in Figure 4.12.1.
units, this in no way influences formula Movement along the straight lines MP
{4.12.1). If x, y and z are measured in ,
and QN changes only the variable x,
units of e l9 e 2and E, then the deriva-
,
so that if we confine ourselves to these
tives dzldx and dzldy have the dimen- two straight lines, the function z
sions of E/e x and E/e 2 but the incre- ,
can be considered a function of only
ments Ax and Ay have dimensions that one variable x (these are the functions
coincide with those of xand z/, that is, of (p (x) — F (x, y) j/= co nstant & n d cp x (x)
|
=
e 1 and e 2 so that the two terms on the
,
F (x, y +
Ay) y +A,y=constant)?
|
the
right-hand side of (4.12.1) have the increment (due to the transition from
same dimensions, of E, which coincide M to P and from Q to N) can be found
with the dimensions of the left-hand via (4.1.2). In the same fashion, if we
side, Az. If we go over to a new unit move only along the sides and MQ PM
for x, that is, e[ ke x the partial de- = ,
of the rectangle, the function z trans-
rivative dzldx will get multiplied by k, forms into a function of only one
but the numerical value of Ax will be- variable y (these are the functions
come k times smaller, so that the product ^(y)=F(x, y)\ x=constant and % (y) =
(dzldx) Ax will remain unchanged. F (x Ax, y) jc+Aa:=constant-
-(- \
each
But if we go over to a new unit for z, vertex of the rectangle in Figure
or E' — KE
then the partial deriva-
,
4.12.1 we have written the (approxi-
tives dzldx and dzldy will be divided mate) values of the function z.
by K
and, therefore, the increment Az The situation with a function with a
given by (4.12.1) will be times smal- K number of variables greater than two is
ler (as was to be expected). similar. For instance, if u =
f (x, z/, z ,£),
 156 4 Calculation of Derivatives

then the increment of the function (13.4.3)). Therefore, if we start with

that is caused by the increments in all the function W = W {C, cp), we get
four independent variables, Ax, Ay, A z,
At, is
I?
OC, <p
4- cp
Lt C
=
which is the"J rate 2 =
of variation of the energy as the capa-

A “-£ A *+-| l

r
*+£-**+%-*' citance varies (and the voltage is kept
constant), while if we start with the
(4.12.1a) function (C, q), we will find W=W
dW =— W
Tc = ~
2
1

Won q ,
.

du _ df (x, y, z ,
t) that wlth the
where, say, q ~c <
dx dx y=constant,
z=constant,
charge the capacitor being constant);
i=constant, thus,

etc. The accuracy of formula (4.12.1a) dW I

_“
— dW I

will be the greater, the smaller the dC |(p dC Iq'

quantities Ax, Ay, A z, At are. It is clear that the partial derivatives.

Note also that the function z = dz/dx and dz/dy of a function z F (x, y) =
F (x, y) actually connects three vari- (with respect to x and y) are defined for all x
ables, x, y, and z, and each variable and y from the domain of the function F, O
that is, are themselves functions of x and y.
can be considered a function of the oth- Hence, we can define the partial derivatives of
er two. For instance, y is a function partial derivatives, or what is known as par-
of x and z, since by fixing the values tial derivatives of second order:
x =
x 0 and z =
z 0 of the variables d / dz \ d2z d l dz \ d2z
x and z we can, using the equation dx \ dx } dx 2 ’
dy \ dx )
~ dx dy ’

z0 =F (# 0 y), find the value of y (it

,
d / dz \ d 2z d dz \ _
d2z
could happen that there are more than dx dy dy dx '
dy
I
dy
~~
dy 2
\ / \ j
one value, but this would only mean
that z = f (x, y) specifies ^ as a mul- For instance, in the case of the function z =
xy z which we have already met before, we-
tiple-valued function of x and z). This ,

= —
have d 2 z/dx 2 d (y z )/dx 0, d 2 z/dxdy =
fact enables one to determine such par- d (y s )/dy =
3 y 2 d 2 z/dydx d (3 xy 2 )/dx == 3 y 2 r =
=
,

tial derivatives as (dy/dz) x== const ant and and d 2 z/dy 2 d (3 xy 2 )/dy 6xy. Thus, we see =
(dy/dx) z==co nst ant and the like. In phys- that d 2 z/dydx d 2 z/dxdy = .

ical problems the choice of variables We can easily see that this is not accidental
but is a general law: d 2 z/dxdy d 2 zldydx. In- —
on which a given quantity z depends deed, according to the definition of a deriva-
can be very different. Two different tive (or formula (4.1.2)),
functions z =F (x, y) and z = G (x, u) d 2z d dz \
I
where y and u are different physical dx dy ~~ dy \ dx J
parameters on which z depends, en-
able finding two different values of
dz(x , y + ky) dz(x , y)

the partial derivative dz/dx, which

would be appropriately denoted by
{dz/dx) y— C0VLS tant and {dz/dx) u — constant where the approximate equality may be as-
precise as desired (we need only select Ay small
(or, briefly, by and g|j. It is enough). On the other hand, it is obvious,
that
clear that the two derivatives have the
same dimensions, those of E/e, where
dz(x, y + Ay)
dx
E and e are the units of measurement
of z and x, but numerically they are ^ + y + Ay)— z(x, y + Ay) ^ i 2 .3a>

not necessarily the same. For instance, Ax

the energy IF of a capacitor is (l/2)Ccp
2 = dz{x, y) _ z {x + Ax, y) — z(x, y) 12 3hV
%
(1/2 )q /C, where C is the capaci- Tx “ Ax ’

tance, cp the voltage across the capacitor, where the approximate equalities become
and q the quantity of electricity, or exact equalities as Ax 0. Substituting
charge, on the capacitor (see formula (4.12.3a) and (4.12.3b) into (4.12.2), we get
 x x x 0

4.12 Differentiating Functions Dependent on a Parameter 157

d 2 zj
~ 1 r 2 (x + Ax, y+Ay)— z(s, y+Ay) point
lel
P (the perpendiculars are paral-
( , y)
axis) we lay off a segment
to the z PM
dx dyj ’Ay L A#
of length z (this segment is laid off upward or
z(x + Ax, y)—z(x, y) 1 downward depending on whether z is posi-
Ax J tive or negative). Suppose that the surface 2
is smooth in the neighborhood of a point
_ Z (x + Ax, y'+ Ay) - z (x, y + Ay) M (x o, y 0 *o)> with z o
,
F(x 0 y 0 ), that is, = ,

Ax Ay it has no salient points or discontinuities or

z (ff Ax y) ,
z (
x , y) “cusps” (like the vertex, or apex, of a conical
(4 12 4b.) surface) or similar “singular” points. In this
Ax Ay
case the tangents to all the curves that pass
(note that in the numerator on the right-hand through point M
and belong to 2 form a plane
we have the sum (with alternating signs)
.•side o known as the tangent plane (the plane
of the values of function F at the vertices of tangent to a surface at a certain point). Clear-
rectangle II depicted in Figure 4.12.1, and in ly, the tangent plane o is completely defined
the denominator, the area of H). Similarly by the tangents t x and t 2 to the curves along
which the planes y constant (=y 0 ) and =
d2z d / dz \
x =
constant (=^ 0 ) cut 2, and the slopes of t x
dy dx dx \ dy ) and t 2 in relation to the zy-plane are given by
the values of the partial derivatives dzldx
dz(x-{-Ax, y) dz(x, y)
and dz/dy. (Why?) It is easy to see that the
dy dy
^ equation of plane o in the xyz coordinate
Ax system has the form
and
z — z °= — J/o). (4.12.5)
dz (x-{- Ax, y)
dy where, say, dzldx is (dz/dx) x=:xo y y ,
= .

z(x+ Ax, y-{-Ay) — z(x-\-kx, y)

The right-hand side of (4.12.1) {dzldx) Ax +
1
T~ (dz/dy) Ay is called the total differential of
Ay function z
,

=
F (x, y) and is denoted by dz .

dz (x, y)
~ z (
x i/
+ Ai/)— z(x, y) If z =
x (i.e. F (x, y) x) then, obviously, = y

Ay the partial derivatives dzldx and dz/dy of this

simplest function of two variables (which
whence actually depends only on x) are 1 and 0, respec-
y+Ay) — z(x+Ax,
tively, and therefore the total differential of
J*z J__ f
z (x-\-Ax, y)
this function is dx 1 X Ai Ay — + OX =
dy dx Ax L Ay Ax. Similarly, it can be proved that dy Ay, =
z(x, y + Ay) — z (x, y) 1
where on the left-hand side we have the total
differential of the function G ( y) y of two ,
=
Ay J variables (actually of one variable) and on
_ z (ar-f- Ax, y + Ay) — z (s + Ax, y) the right-hand side we have the increment of
y. Since Ax dx and Ay =
dy we can re- =
Ax Ay ,

write the formula for the total differential

£ (f> y + Ay) — z (£, y) of z =
F ( x y) as ,

Ax Ay '
(4.12.4b)
dzr-^^-dx+^-dy. (4.12.6)
Since the approximate equalities (4.12.4a) and dx dy
(4.12.4b) may be made as accurate as desired,
we conclude that d 2 z/dxdy d 2 z/dydx. = The equation of the plane a tangent to
surface 2, (4.12.5), clarifies the geometrical
Partial derivatives of higher orders are
defined similarly; the value of the derivative
meaning of the total differential. This equation
implies that the Z-coordinate (below we will
is determined by how many times we diff-
see why the letter “Z” is more convenient than
erentiate with respect to each variable and
not by the order in which this is done. It is the letter “z”) of point x (x, y, N
Z) in plane a,
quite easy to generalize the facts for the case a point corresponding to the value x 0 Ax +
of a function of three or more variables, but
of the ^-coordinate and to the value y 0 Ay +
we will not do this here. of the y-coordinate, is equal to z 0 dz (com- + —
pare with (4.12.5), where the increments x
If in space we introduce a Cartesian system
of coordinates (see Figure 4.12.2) consisting
x 0 and y —
y 0 of the x- and y-coordinates are
of three coordinate axes, x, y, and z, then the
now denoted by Ax and Ay). On the other
'‘graph” of the function y F (x, y) is a surface = hand, the z-coordinate
N
of the respective
(2). This surface is formed by all the points
point ( y, z) on 21 is equal to z 0
,
Az, where + —
M(x, y, z) obtained as a result of the follow-
Az is the increment of z, or simply z
Thus, the approximate formula (4.12.1),
z0 .

ing process. On each straight line that is

perpendicular to the zy-plane and erected at z ~ z0 + dz, (4.12.7)
 158 4 Calculation of Derivatives

If equation can be solved for

the
x or then we revert to the ordinary
y,
representation of the function in ex-
plicit form. But sometimes such a solu-
tion leads to complicated formulas and
at other times it cannot be found at all.
For instance, the equation of a circle
in the form
X 2
+y 2_ l = 0 (4.13.2)

is simpler than the following expression

derived from it:

y(x)= ±YT^ 2
. (4.13.3)

If the left-hand side of (4.13.1) is an

arbitrary polynomial involving x and
suggests that, for small Ax and A y, the point
y to a power exceeding the fourth, then
N(x, y, z) of 2, where x x0 — +
Ax and in general this equation cannot be
y */o
— +A y, will lie sufficiently close to
solved for x or iovy. Also, for example,
point N
1 (x, y, z), where 2 z0 = +
Az and
Z z0 — +dz, that belongs to plane a tangent unsolvable is the simple-looking equa-
to 2 at point (;r
0
M
y 0 z 0 ) (see Figure 4.12.2).
, , tion
The error introduced by the approximate for-
mula (4.12.7) is a quantity of second order of F (x, y) = x sin x -f- y sin y — = ji 0.
smallness if compared with dx and dy (or
(4.13.4)
Ax and Ay), which are assumed to be quanti-
ties of equal order of smallness.
However, even in those cases where
Quite similarly, for a function of, say,
there is no solution in the form of a
four variables, u f (x, y, 2= f), the total ,

differential du is the expression on the right- formula that specifies directly a proce-
hand side of (4.12.1a) in which the increments dure for computing y for a given x,
Ax Ay, Az, At of the independent variables
,
it still remains a fact that y is a definite
can be replaced with dx dy, dz, dt. Thus,
we can rewrite the approximate formula
,
function of x For every x it is possible,
.

(3.12.1a) as by solving the equation numerically,

to find a corresponding y and to con-
Au ~ du.
struct a curve (4.13.1) in the xy-plane.
It may be that the curve will not exist
Exercises for all x (in the case of a circle (4.13.2),
for example, it exists only for x be-
Find the first partial derivatives of the
following functions:
tween —
1 and -j-1) or that the function

4.12.1. z =
x 2 -h y 2 4.12.2. z l/(x 2 . = + is multiple-valued, that is, for a given
y
2
4.12.3.
). z 2 2
= Vx +y
4.12.4. z .
= x there may be more than one value of
e -(x*+y*). 4.12.5. z x2yz =
4.12.6. z .
= y (in the case of the circle, for instance,
xeV +
ye *. 4.12.7. z In (x = +
y). 4.12.8. z = there are two values in accord with the
cos (xy).
dz sign in front of the square root sign).
4.12.9. Find the second partial derivatives
of the functions in Exercises 4.12.1-4.12.8 However, these complications do not
and check whether d 2 z/dxdy d 2 z/dydx. = detract from the basic fact, which is
that the equation F (x y) 0 defines ,
=
y as a function of x.
4.13 The Derivative of an Implicit How can we find the derivative
Function dyldxt And can this be done without
solving the equation, that is, without
To define a function implicitly means
expressing y {x) in explicit fashion?
to define it via an expression of the form
The procedure which enables finding
F (.x , y) = 0. (4.13.1) y’ from Eq. (4.13.1) was developed by
 )

4.13 The Derivative of an Implicit Function 159

Newton. Let the variables x and y Let us find the derivative in the case&
satisfy Eq. (4.13.1) and let x -f- Ax of (4.13.4):
and y +
Ay be adjacent values of the
fdF(x
variables that still satisfy Eq. (4.13.1), r
V,
dx
, y)
= sin x + x cos x
. .

,
’
where the word “adjacent” emphasizes
the smallness of Ax and Ay. Since — = sin y + y cos y,
z — F (x, y) = 0 and z -f Az =
F (x Ax, +y Ay) 0, we + = dy _
~ sin x + x cos x
have Az F (x Ax y= Ay) + , + — dx sin y+y cosy
#

F {x y) — 0* But, in view of the basic

->

formula (4.12.1), we have Thus, the expression of the derivative*

dF of an implicit function involves both
Az ~ dx
(£, y)
A# dFJx,
dy
y)
A y, quantities, x and y. To find it numeri-
cally, we must find y numerically for
which means that a given x. But if we did not have for-
mula (4.13.7), then to find the deriva-
-W Ax + ^7 a ^-° ( 4 13 5 )
- -
tive would require finding, numerical-
ly, two values y 2 and y x for two adjacent
and, so
values x 2 and x 1 and finding the ratio
dF (x, ij
)
(y 2 —
yi)l (x 2 xi) an d the limit of—
Ay dx this ratio.4 24 Here, the closer x 2 is to
-

(4.13.6)
Az dF (x, y) ’
x l9 the more exactly we would need to
dy compute y 2 and y ±1 and this is often very
where the approximate equality in difficult to do. On the other hand,
(4.13.5) and, therefore, in (4.13.6), is the use of (4.13.7) presents no difficul-
the more exact the smaller the incre- ties.

ments Ax and Ay are. If F(x y) = 0 leads to a nonsingle-

,

Rasstinxg + hPv, Bmit* ffJiisL valued function ii(x). that is, when therfr
will also mean that Ay 0), we get
are two or more values of y for one
the derivative on the left, and the ap- value of x (several branches of the
proximate equality in (4.3.6) becomes curve), then (4.13.7) yields the values of
an exact equality. Finally, we have the derivative at appropriate points of
the curve F (x, y) = 0 when a given x
dF (x, y)
and different y's are substituted. The
dy dx
(4.13.7) reader is advised to verify this by using
dx dF(x , y)
the equation of the circle, (4.13.2),
dy
for which the derivative is given by
Note the minus sign in (4.13.7) and also formula (4.13.8).
the fact that we cannot simply cancel Let us assume that we have a func-
the dF (x, y) in the numerator and the tion of two variables, z F (x, y ), for =
denominator. which we do not require that it be zero.
We will demonstrate the application Instead we will consider this function
of (4.13.7) using as an example along a curve y in the xy - plane fixed
Eq. (4.13.2). We have F (x, y) = x 2 -\- parametrically by the equations x =
y
2 - 1. Then cp
(
t and y \|) (£),=where parame-
dF ter t can be the time variable, for in-
Or, y) _ dF (x, y)
= 2 y, stance. The value of z will then depend
dx dy
dy __ 2x
4 24
dx
(4.13.8) Additional difficulties arise when
-

2y
Eq. (4.3.1) specifies a multiple-valued function.
easy to see that this result coin-
It is y =y (x), since then one must “match”
the values y x (x x ) and y 2 (x 2 ); otherwise for
cides with that obtained if we compute a small difference x 2 —
x 1 the difference y 2 —
the derivative of (4.13.3). y r may prove to be large.
 — . t x

160 4 Calculation of Derivatives

on t, too. Let us find the rate of varia- plied to a function with a higher num-
tion of the derivative dzldt
z, or ber of variables. Suppose that T is the
We introduce a small increment At temperature at a point (x, y, z) in M
into the variable t. The variables x space and that the temperature varies
and y will then increase by Ax
~ with time £; we can say that the temper-
<p' (t) At = (dxldt) At and Ay ~ ature is a function of four variables:
nj/ (t) At = (dy/dt) At. On the other T = T We wish to find
(x, y, z, t).

hand, by (4.12.1), we have the rate with which the temperature

dx
changes when we go over from point
A z~-||- Ax-f-
dz
dy
A ^2
dx dt
At M [x, y, z) (
= M
x 0 y 0 z 0 ) to a neigh-
, ,

dz
boring point
Afj ( dx, y dy, + +
z -f- dz) (this transition occurs over a
dy dt ’

small time interval dt we assume). ,

whence Ignoring the changes in temperature

Ay in time, we arrive at a stationary
Az
At
^
”
dz
dx
Ax
At
, dz
dy At ’
(4.13.9) (or steady-state or time-independent)
temperature distribution T — T (x, y, z)
‘

and the approximation is the more exact (~ T (x, y, z, t 0 )). Here the rate of
the smaller the increment At is (and, temperature variation caused by tran-
hence, the smaller the Ax and Ay are). sition from point to point u in M M
We now send At to zero in (4.13.9). accord with formula (4.13.9) modified
The ratios Az/At Ax! At and Ay! At , in the appropriate manner, is
will transform into the respective deriv-
atives and the equality in (4.13.9)
v con — "r
dy
|
dT dz
will become exact. The final result is qx Qy .
qz dt

dz _ dz dx ,
dz dy
~~ (4.13.10)
dt dx dt ^ dy dt This change in temperature, as we have
already said, is caused by a shift in
In particular, if the curve y is speci-
= space from one point, (x, y, z), to M
fied explicitly, by an equation y another point, 1 (x -f
dx, M dy M + ,

/ (x) (i. e. if parameter y is the abscis- z 4- dz), along a certain curve x =

sa x), then (4.13.10) transforms into =
x (£), y y ( ), =z z (0, say, a
dz dz dy dz transition from a point with a lower
,

(4.13.11)
dx dy dx '
dx temperature to a point with a higher
temperature; in accord with this rate
For instance, if z = xy the rate of ,
of change of temperature, u con associat-
growth of z along the circle x~acost,
it
v — a sm . .

t
.

is -j—
dz
dt
— dz
-r
dx
dx
-
dt
— ,

^
—
dz
dy
dy
~77~
dt
~
ed with such a shift is called the con-
vective rate. Now let us turn to the study
—a sin t) x (a cos t) + = 2
a cos 2 — of the temperature variation in time at
y (

a 2 sin 2 t —
a 2 cos 2 1, while along
t

the
a certain point M
(x 0 y 0 z 0 ). Here the , ,

dz temperature depends only on time t,

straight line y = kx we have — that is T = T (t) (= T (x 0 , y 0 z0
, ,
t)).

The rate of variation of temperature

xk + y = 2 kx.
dz dy -----

dy dx associated with this function (the fact

Formulas (4.13.10) and (4.13.11) spec- that the temperature field also depends
ify the increment of the z-coordinate on time) is called the local rate :
of a point M
of the surface 2 (whose
equation is z F (x, y ); see Figure = dT
4.12.2) caused by a shift along a curve dt •

T lying on 2, a curve whose projection

in the xy - plane is curve y with equa- Finally (compare this with formula
tions x = and y = y (t) (or y =
x (t) (4.13.11), the total rate of temperature
t (x)). The same formulas can be ap- variation can be found by adding the
 :

4.13 The Derivative of an Implicit Function 161

local rate to the convective rate ing the other limit of integration
(b or a) fixed (or constant). in such Even
l
hot-
_ ^con-i-^loc _
.

—
dT _
—
dT dx
elementary questions as finding the de-
dt QX dt
rivative of the product of several func-
d d
tions, y — h (x) g (a:), or as finding the
d d
| JL JL |
Jj §1 _L
dy dt ‘
dz dt '
dt
derivative of y =
h (x) g x > (see Section <

(the derivative dTIdt is sometimes 4.9), the concept of a partial derivative

called the total derivative of T = was actually present. Indeed, as was

mentioned earlier, y' in these cases is
T (x, y, z, t), where x = x (f), y = y t
( ), composed of terms obtained when tak-
z = z (t)). ing the derivatives with respect to x
In finding the derivative of an impli- in h (x) and with respect to x in g (x),
cit function (or the derivative of a func- which clearly involves partial deriva-
tion z =
z (t) specified in the form of a tives. The corresponding general rule
composite function of two variables, can be written as follows: if y =
z =F (x, y), where x — x (t) and y = F h (r)], then, in accord with
y (£)) we had to introduce a new con- (4.13.10),
cept, that of the partial derivative (see
Section 4.12). This notion is of great > _ jy _ dF_ dg d£ dh
*
importance to the theory of functions dx dg dx dh dx
of several variables, on which we have
only briefly touched here. Actually,
Exercises
we have already, latently, made use of
the concept of a partial derivative.
For instance, the definite integral I = 4.13.1. Find the derivative dy/dx of a
function defined by Eq. (4.13.4) at the follow-
b
ing points: (a) z =
Jt/2, y ji/2, and =
f(x)dx = /(a, b) is a function (b) x = —ax/2, y —
ji/2.
4.13.2. Find the derivative dy/dx at the
a
two variables: above we found the point x = =
y 1 of a function defined by the
of
— equation x z +
3x +
y* 8 — = 0.
partial derivative di/da (= / (a)) and
dl/db (= / (6)) with respect to one of
4.13.3. Let z —
Y' x 2 -\- y 2 Find (a) dz/dt
.

with x =eht cos (Zi)fand

y =
eM sin (It), and
the limits of integration (a or b) assum- (b) dz/dx with y x2= .

11-0946
 t

Chapter 5 Integration Techniques

5.1 Statement of the Problem 5 * 1

which it is possible to solve problems of
type (2). For certain simple types of
In Chapter 3 we introduced the concept
functions v it has been possible, with ,
and noted the close con-
of the integral
the aid of algebraic identity transform-
nection between two different (at first
ations, to find the general rules of
glance) problems. The problems are
solving problems of type (2), say, when
(1) sum of a large number
finding the
v is a sum of functions whose integrals
of small summands when the terms can are known (see Section 5.3).
be represented as v ( t ) At (or v ( t ) dt);
However, this is not possible
(2) finding the function z (t) whose de- for all the elementary functions, so
rivative is equal to a given function
that integration is more difficult than
v (t):
differentiation (finding derivatives). Nev-
ertheless, the formulas found for cer-
= (5.1.1) tain integrals in the second statement
of the problem are very important. If
Most
of the problems that arise in phys-
for a given v (t) it is possible to find
ics, engineering, chemistry, as well
the integral (the indefinite integral or
as in mathematics, are problems of type
the antiderivative) z (t), then all prob-
(1), that is, they involve finding the lems in the first statement, all sums,
sum of a large number of small sum- that is, all definite integrals
mands. This statement of the problem b
is more pictorial and suggests a simple,
v (£) dt ,
are then expressed by
though approximate, way ofcomputing
the quantity of interest, namely, by
simple formulas via the function z (t):
directly adding the (small) summands
b
of which we spoke earlier. However,
this “direct” method of solving prob- v (
)
dt —z (
b) — z (a).
j
lems of type (1) does not make it poss- a

ible to express the answer by any gen- Such a result is more complete, more
eral formula, and higher mathematics exact, and more valuable than the re-
appeared when the relationship was es- sult of every separate numerical com-
tablished between problems of types putation of a sum, that is, of the defi-
(1) and (2), which opened the way for b
general algorithms for solving problems nite integral v (t) dt between definite
of type (1). j
a
The statement of problems of type limits a and b. For this reason we aim
(2) ismore artificial, but it has its ad- primarily to solve the problem of in-
vantages. The finding of derivatives tegration in its second statement.
proved to be a very simple matter,
which reduced to four or five formulas
(the derivatives of x
n
e
x
In x, sin x 5.2 Elementary Integrals
, , ,

cos x) and two or three rules. It is there- Let us write down the formulas for
fore easy to find the derivatives of a derivatives that have been found in
large number of functions. Every time Chapter 4 and the corresponding (in-
the derivative dzldt =
v of some func- definite) integrals:
tion is found, we can record the fact
= nx dx — x°
"1
that for this v the integral z is known (
xn) 77-1
,
n x 71 + C;
j
(see Section 5.2). In this way we can
build up a range of particular cases in -JL (
e
kx
)
= kekx , k e
hx
dx = ehx + C ;
j
6 1-
Before reading this chapter, the reader
advised to reread Chapter 3.
-^-(Inx)
dx x 1
——
x 1
, [
—dx~lnx
x + C;
J
 k ) , 1 1 —a —

5.2 Elementary Integrals 163

(sin kx = k cos kx. is meaningless, but In ( x) is meaning*

ful. Since
k cos kx dx = sin kx -f C\ d In ( x) __
j dx

(cos kx) = — k sin kx, for x C 0 we have dxlx = In (— x) -f-

j

— sin kx dx = cos kx + C; C. Both formulas for

J
dxlx can be
j
combined into one:
(tan x)
dx cos 2 X r
-^==ln \x\+C. (5.2.3)
j
dx = t&Rx-\-C\
1

cos2 X This formula may be used for any do-

main of integration that does not con-
&-( cota;) = tain x = 0.
The
integral of the exponential func-
dx = cot x + C\ tion can be written thus:
Insin 2 x

±. {&vcsinx)=yj=, e
hx
dx = ±ehx +C. (5.2.4)
j

J
f
—y i— xi
dx = a rcsi n x 4- C\ we
Similarly,
get
for the sine and cosine

(arctan x) = -1+Z 2
(•

\
sinkxdx = —— 51 4

coskx +C , (5.2.5)

cos kx dx = -y sin kx + C.
1+X da^arctanz + C.
(5.2.6)
2 j
We conclude with an example that
Let us perform a few manipulations. shows the necessity of a careful exam-
In the first integral, we denote n 1 — ination of the function and the danger of
by m, that is, n m 1, and re- = + a purely formal approach. We will
write it thus: M b

evaluate the integral I = dxlx*. The

j
xm dx= 1 ^ T
xm+l + C. (5.2.1)
indefinite integral is given
a
J
by the for-
It is clear that the formula is valid for
mula dxlx 2 = — tlx + C, whence
all m except m = — 1; for m— — [

the denominator vanishes and xm+1 =

x° =
1, so that we arrive at an
expression unsuitable for computa-
tions: 1/0 +
C. However, it is precisely b —
in the case where 1, that is, for m= — ^ab~ m (5.2.7)
c
(
1/x ) dx that we have the formula 5 - 2
Since the integrand is positive, the
j
result must be positive, too, provided
~dx = lnx+C. (5.2.2)
that b >
a. The answer is indeed posi-
j tive if b >
a and both a and b have the
This formula holds true only for posi-
same sign. However, for the integral
l
tive values of x since In x is meaning- ,
dxlx 2 formula (5.2.7) yields a clear-
ful only for x > 0. For x < 0, In x
j*

-
5 - 2
In Section 7.6 we will discuss the
ly absurd result I —2. The reason —
marked difference between formulas (5.2.1) for this is that the integrand becomes
and (5.2.2) for integrals of power functions. infinitely large inside the domain of

n*
164 5 Integration Techniques

integration x
0 ), and at this
(at = Exercises
point the — Hx
experiences an infinite
Evaluate the following definite integrals:
jump (this function is the indefinite in-
2 l
L 2
tegral of 1 /x ).
To find the true reason for all this, 3.2.1. a:
5
dx. 5.2.2. -4- d*
j j x2
we # must exclude a small neighbourhood o i
of the “singular point” x 0 from the = n 1
domain of integration —1 < x < 1, 5.2.3. sin x dx .
5.2.4.
dx
that is, we must takeinterval the ^ J
ex c C
x e 2 where e x and e 2 are small
,
jt/4
0

positive numbers, and consider the sum da;

5.2.5. e*dx. 5.2.6.
j j

5.3 General Properties of Integrals

T
It is natural to assume that integral / ~±fi'^ctnm& 4.1“ 16 i.o we Gstdbilsnetf tne
is obtained from K if we send e x and formulas for the derivative of a sum
e2 to zero. But formula (5.2.7) yields
(and difference) of functions, the deriv-
ative of a product (and ratio) of func-
~~ 81 ~~ &2 tions, and the derivatives of a com-
K= 1
jl
r
i

e2 e2
posite function and of the inverse of a
6i
function. To each of these properties
there corresponds a definite property
When ex and e2 tend to zero, K tends
referring to integrals. This section and
to oo.
Sections 5.4 and 5.5 are devoted to
In other cases an integral whose in-
finding the “integral” analogs of the
tegrand becomes infinitely large within
properties of derivatives considered.
the domain of integration may assume
For integrals we have the following
a finite value. For instance,
formula:
i

dx/Y x = 2. [Af(x) + Bg(x)\dx

j j
0

Indeed, the corresponding indefinite

=A f(x)dx +B j g(x)dx. (5.3.1)
j
integral is given by the formula
To prove it, we need only take the de-
*~' ndx rivative of the expression on the right.
1 fr= J If the formula is true, then we will ar-
rive at the integrand. Indeed,
"= xUi + C = 2yx + C,
’Tj2

whence [-4
j
/ (*) dx + B j g {x) dx J
=A / (*) dx
~ B g (*) dx ~\
[j \ [ j
D '
= Af(x) + Bg (x).
Which tends to 2 as e -> 0. We have thus proved the validity of
Consideration of this kind must al- (5.3.1).It shows that the integral of a
ways be carried out if the integrand be- sum of several terms splits up into the
comes infinitely large; here, however, sum of the integrals of the separate sum-
we will not discuss this important top- mands, and any constant factors can be
ic any further (see Section 6.3). taken outside the integrrl sign .
 . i b +

5.3 General Properties of Integrals 165

As an example to illustrate the above (5.3.3) we get dz — a dx or dx ,

— a~ x dz %

statement, we consider the integral of Thus,

a polynomial :
ndz_^±
(ax + b) n dx = Z
C
z
n
dz

J
a x
( 0
k
+a +a x " + +a
2
k 2
. .
. k _2 x2 j j a a J
~n +
4-C= -
ax
(< + b) n+1 C. (5.3.4)
-\-a k _ l x-\-a k ) dx = a x dx 0
k a (t2 + 1) a (n -j- 1)

j
The correctness of the result is easily
*“ 1 ~2
+a i
x dx + a 2 xh dx-\- ... seen if we compute the derivative of
^ [
the Tight member:
+ ak -2 i
4- f x dx + ak \
dx d r (ax-^b) n+1 .

r ~\ d f (ax +b) n+1 ~|

#0 a2
dx L a (n + l) '

J dx L a(n-\-i) J
x h+l\J!±. xh yh — i
+ 72+1
*+1 k— 1
a + 1
{ax + b) n + (ax + b)
+ ^_ x + ^L x 2 + a ^ + c>
(72 )

3
(5.3.2)
^±^a = (ax + b)\
Whence, for one thing,
m 2. In similar fashion, in the inte-

\
n
h
(a 0 x + + a 2 xh ~ 2 + a k .2 x 2 gral
j
C
l
—— ax
fix

|
r-
u
we can make the change

of variable z — ax + b, dx — a' 1
dz , and
+ ak_ iX + ak )
dx = -j+- m h+l — nh+1 ( )
hence
dx
+ -j- (m k —n + h
m k~ — n h~ )+.
l l — 1
j'4=4‘"i*i c
) ( .
i az +6 a
In \ax-\-b\
+
fl
g~
2
(m — + 3 re )
3 ak ~ 1
(
m2 — re
2
) a
= C. (5.3.5)

+ — re).
When we are dealing with such
simple examples in practical situations,
For instance,
the transformations are ordinarily car-
(2a:
2 — 6a; + 1) dx — -|-a:
3 — 3a: 2 + £ + C, ried out without introducing separate
designations for the new intermediate
j
2 variables. For example, instead of
(2a:
2 — 6a: + 1) dx = ~ (8 — 1) (5.3.4) one writes
j
l
n
(ao:+ b) dx= (aa: + b) n d(ax-\- b)
— 3(4— l) + (2-l)=— 3 + j
1
(ax
j

+ b) n+ + C. '

It possible, under the integral

is (72 + 1) CL

sign, to make a change of variable and We now give a more complicated

pass to a new and more convenient example of an integral that can be re-
variable. We
will discuss this question duced to the integrals we already know
in detail in Section 5.5, while here we via algebraic transformations.
examine a number of simple examples. dx
1. Find
|
(ax + b)
n
dx with n
, ^— 1.
Note that
5 (x — a) (x—'b )
'

For the new variable, call it z, we

take the expression in the parentheses: 1 1 CL b tr q
x —a x —b {x — a) (x — b) ’ \ /

ax + b =z. (5.3.3)
wvhinhi ~fi ^nJfiuws hnti
A

In so doing, we also have to pass from

— a) (x — b) _
1 1 / 1 1 \

differential dx to differential dz From . (x a — \ x —a x —b ) *

 x a
b — — j

166 5 Integration Techniques

whence are found via the method of undetermined

coefficients (see the hint to Exercise 5.3.7);

T7 — tt
— a)(x = —— r - (— —
— da;
Cx dx
\x — a
f f 7 to find it proves convenient to
7( x
+
.
b) )
a b x 92
.

J b) J
,
l
J

a —b [In \x — a — In — \
| a; 6] |
-}- 67 denote ;r
2
+ l by z.J

1
In
x — +c.
a —b x — (5.3.7)
5.4 Integration by Parts
(Formula (5.3.7) can be used in any do- Let f (x) and g (x) be two distinct fun-
main integration that does not in-
of ctions of the variable x. The rule for
clude the points x a and x b.) An — = finding the derivative of a product
integral of any algebraic fraction, that yields
is, the ratio P n (x)/Q m (.x ) of polynomi-
als of degrees n and can be ex-
pressed as a combination of elementary
m , i m=s-§-+i£. (5.4.1)

functions (algebraic functions, loga- This enables us to write

rithms, and arctangents). Several
simple examples of this type are given t ‘=
in the exercises below.
- J
>-§-*=+ j
f%r d *- (5-4.2)
Exercises
We can see the validity of (5.4.2) by
Find the following (indefinite) integrals: taking the derivatives of the left and
right members to get the true equa-
5.3.1. (3x* — 2x+l )dx. 5.3.2. (4r 4 — tion (5.4.1).
j (
Let us rewrite (5.4.2) as
3:r 3 + # — x) 2 dx. 5.3.3. # (z — i)* dx.
j

**+ 2*+ 3 \fik dx==te-\ *i£ dx ’

5.3.4. f
X
5.3.5. C *4*.
X—
J J 1
or, which is the same,

Hint. Employ the fact that

\fdg = fg—^gdf.
5.3.6. dx.
^ £ (5.4.3)
ax-\-b a . be— ad "1
f xdx
cx+d~ c
1

c(cx + d)' \
'
J (x—2) (x—3)' What is the meaning of this formula?
When evaluating an integral, there is
L
Hint. Employ the identity 7
{x — 2)^(x — 3)rr == unfortunately no rule that expresses
the integral of a product of two func-
——— —
1
-z [
~~n the numbers A and B are tions in terms of the integrals of each
x 2
1

x—3 ;

of the factors. However, if in the prod-

found via the method of undetermined co- uct of two functions fw the integral of
efficients, that is, by equating the coefficients one of the factors is known, or
of identical powers of x after clearing
w dx = g, u> = -|J, (5.4.4)
j
fractions.J
then it becomes possible to express the

integral j* fw dx in terms of an inte-

gral involving the derivative dfldx .

-i+ 2
+ -[
Hint. Use
dx. the identi- we rewrite (5.4.3) in
j x Using (5.4.4),
x-\- 2 x+2 A B the form
x +
ty '

x{x* + l) z2 + fwdx = f —
l
wdx} wdx^j-
x
dx.
Cx j ( j j ( j
<r 2 4- 1 ,
where the ~
coefficients A ,
B and C
,
(5.4.5)
 ) .

5.4 Integration by Parts 167

Since dx = it follows that the

Knowing this, we need not perform
g,
integration by parts n times, but can
last integral in (5.4.5) is simply^ g (df/ write down directly the coefficients of
the polynomial Q n (#).
dx) dx . Sometimes it is simpler than

the original integral fw dx or reduces Let us take the same Example 2. Find
^

to a known integral. If f is a power \^x 2 e x dx. We write the equation x e xdx

2 =
j
function or a polynomial, dfldx is Q2 (x) e x + C with the (still) unknown coef-
also power function or a polynomial
a ficients of the polynomial Q2 (^):

whose power is that of / minus unity.

Formula (5.4.3) or (5.4.5) is called x*e*dx = (a x*+aix-{-a
i 0
)ex+C. (5.4.6)
j
the formula for integration by parts .

From this formula follows a similar Taking the derivatives of both sides of (5.4.6),
relationship for definite integrals: we get

b b x2e x — (2 a 2 x +a x) e
x
+ (a 2 x 2 + a ±x +a x
0) e .

j
/ (x) dg (x) = / (x) g (X
£
— j
g (X) df {x) or
a a x 2e x = [x 2 a 2 +x (2 a 2 +a + x) (a ± +a x
0) ] e .
b

= lf(b)g(b)-f (a) g (a) - ]

j
g (x) df (x) Equate the coefficients of identical powers of x
in the polynomials on the right and left to
get a 2 = 1, 2 a 2 a x = 0, a x +«o 0, whence + =
(5.4.3a) = —2 and a 0 2. =
Finally as before,
we get
some examples illustrat-
Here are
ing the method of integration by parts. x*exdx = {x* — 2x+2)e* + C.
j
1. Find xe dx.
x
Put = x.f Then By a similar technique we can find the in-
j
w — dg/dx — e x
and e
x
dx = dg g—
tegrals of the functions P n (x) cos kx and
,
P n (x) sin kx, with P n (x) a polynomial.
e
x
dx = e
x
,
and df=dx. By formula In both cases the answer is of the form
[
Q n (x) cos kx +R
n (x) sin kx, where Q n (x)
(5.4.3) and R
n (x) are polynomials of degree n
(or less than n). Examples of this kind are
xe dx
x
= xex — e
x
dx = xe x — e x 4- C given in the exercises below.
^ j
= ex (x — 1) + C.
Exercises
2. Find
j
x2 e x dx = x 2 Then
. Put f .

Find the following integrals:

w = dgldx = e x ,
e dx = dg, g=
x
e dx =
x
j 5 . 4 ... x cos x dx. In x dx.
and df = 2xdx By formula
x 5.4.2.
e , .
(5.4.3), j J

j
x 2 e x dx — x 2 e x —2 xe x dx ,
5.4.3.
J
x 2 sin 2x dx. 5.4.4.
j
x z e~x dx.
j
whence, using the result of the first
example, we get
5.4.5.
J
{x 2 -\-x + 1) cos x dx. 5.4.6. j
(2a; 2 + l)X
x 2 e x dx = x ex — 2
2xe x + 2ex + C
cos Sx i
dx. [Hint. Exercise 5.4.6. is|studied
j in detail in Hints, Answers, and Solutions
= {x 2 -2x+2) e
x
+C. at the end of the book; Examples 5.4.3 to
To find Pn (x)e kx dx, where Pn (x) 5.4.5 are handled similarly to Example 5.4.6]
|
is polynomial of degree n, we have
a
5.4.7. arc sin x dx. 5.4.8. arctan x dx.
to perform integration by parts n j
times. We then get Q n (x) e kx where ,

5.4.9. e 2x sin 3x dx. 5.4.10. ex cos 2x dx.

Q n (x) is a polynomial of degree n .
j
 . a ~ —

168 5 Integration Techniques

5.5 Change of Variables in Integration

Tm the. ijitajgyd* - + ha.
^J —x
We have already mentioned the gen- V ( a2 2 3
)

eral method of changing the variables appropriate change of variable is x =

in integration, or integration by sub- a cos t that is, t arccos {x!a)\ then =
stitution (of a new unknown for the initial dx = —
,

a sin t dt a 2 x2 a 2 (1 ,
— = —
unknown). If in the integral j* F (x) dx cos 2 £) =
a 2 sin 2 t, and we obtain

the integrand F (x) can be represented P dx P —a 2

sin t dt

in the form J Y (a 2 —x 2 3
)
J y (a 2 sin 2 t) 3

F(x) = f (cp (x)) 9' (x),5

a sin t dt
_L f dt

\ a3 sin 3 t a2 J sin 2 1
where cp (x) is a function of variable x ,

we can denote cp {x) by a single letter z = -4-coU-j-C = '

•cot ^arccos + C.
a2
1

and take z as the new variable. Since

cp' (x) dx = dz we arrive at the inte- ,
This expression can be simplified still
further (see Exercise 5.5.10).
gral j/ (z) dz, which may prove to be
An almost similar substitution can
tabulated (that is, found it may be
be applied to the “almost tabular”
among the integrals of Section 5.2) or
even be simpler than the initial inte- integral [
J
— a —x 2 2 \
\
which by the sub-
y
gral. This general method is widely
stitution x= az, dx = a dz is reduced
used in integral calculus.
For instance, the integration of many to the known result \
—
J j/T —z 2
functions involving radicals and trig-
onometric functions may be reduced, by arcsin z+ C . Indeed, if x = asint r
j
means of an appropriate change of dx = a cos tdt, and ]/ 2 — x = acost 2
f
variable, to the integration of polyno-
then
mials or algebraic fractions with integral
dx
= acos *^ = t+C
powers. Let us consider several exam- f - f = [dt
ples. J Y a?— x* J a cost J

Find j* xY x -f 1 dx. We make a change = arcsin — + C,

of variable: z = Yx
—
-|- 1
, # + = 1
since = arcsin (x/a)
z
2
,
whence x = z
2
1 and dx 2z dz = .
t
/t
‘
dx
Passing to the new variable in the in- 7 can be evalu-
i y a 2 -j-x 2
tegral, we obtain
ated in a similar manner, only in-
xY x+ idx= j(z 2 — i)z2zdz stead of trigonometric functions we
j must employ the so-called hyperbolic
=2
j
(z 4 —z 2
)
dz =2 — 2 -^- +C sine sinh t (1/2) ( e
hy perbolic cosine cosh t
* f
=
e ) and the
(1/2) (e -j-
—
— l

l
e~ ) (for more details on these func-
tions see Section 14.4). We can easily
see (verify this!) that cosh t
2
sinh 2 t= —
In the integral
j^f^ thea pp r °p-
1, (sinh t)' cosh t and =
(cosh t)' ,
=
riate change of variable x = at
is , sinh t. Therefore, if we put x =
that is, t=x/a: then dx = ;
adt and we a sinh t, then dx a cosh t dt and =
have
]/a2 x2 +
a cosh = t. Thus, our integral
dx
f
_ P a dt
~ r a dt takes the form
J a2 +x 2
J a2 +a 2t2
J a2 (l-H 2 )
dx _ a cos h t dt
=—
a
i
J
TXTF
l +
2 £
11=7 — arctan + C
a
t
f

J y'a 2 + x2
~
f
J a cosh 1

=—
a
arctan — + C.
a = dt=t-\-C — arsinh
j
5.5 Change of Variables in Integration 169

where arsinh z is the inverse hyperbolic

sine defined by the condition: if t —
arsinh sinh t.
z, then z =
We leave to the reader the task of
bringing this example to the final re-
sult (see Exercise 5.5.11).
Finally, we give an example of an
integral that cannot be represented in
terms of a finite number of elementary
Figure 5.5.1
functions. The integral is

no more complicated than, say, those

/ (
x) = e~ xZ dx. (5.5.1) involving trigonometric functions.
j
Note that often, when using various-
The proof that it cannot be expressed techniques, one obtains distinct ex-
interms of a finite number of elementary pressions for one and the same integral
functions (power functions, including (see Exercise 5.5.6). This should not
functions with fractional powers, the dismay the reader. If the computa-
exponential function, the logarithm, the tions are correct, such expressions should
trigonometric functions, and the in- differ by a constant only. The re-
verse trigonometric functions) is ex- sults are identical when evaluating a
tremely complicated and we will not definite integral (for any limits of in-
give it here. tegration).
Theintegral (5.5.1) constitutes a We see that computing an integral
function whose properties can be stud- is much more complicated than find-
ied. The definition of / (x) implies ing derivatives: here one must employ
that various ingenious techniques (integra-
tion by parts, change of variable), and
often it is not clear at first glance what
technique to apply (e.g. how to choose
Since /' (x) e~ is = xZ
positive for ar- the new variable). Moreover, in some
bitrary x, it follows that / (x) is an cases (as in the case with integral (5.5.1))
increasing function. The derivative of there is not a single technique that
/ (x) has a maximum at x 0, where = enables finding the integral in closed
it is equal to unity, which means that form, that is, expressing it in terms of
a finite number of elementary functions,
/ (x) has a maximum angle of the tan-
gent line with the x axis at x 0 = and there is not a single general meth-
(the angle is 45°). For large absolute od that could establish whether a giv-
values of x (positive or negative), the en integral can be expressed in such
5 3
derivative df/dx is very small, whereby a form. Integration techniques can-
-

the function is almost constant for such

5 3 Often we encounter the case where the
values. The graph of the function *

® x2 formula for a definite integral j* dx

(x) = j
e~ dx is depicted in Fig-
a
/ (x)

o
with known limits of integration a and b
ure 5.5.1 (for the sake of definiteness,
exists, but the general formula for the corre-
the lower limit of integration has been sponding indefinite integral cannot be found.
chosen to be equal to zero). The fact oo

that <D (oo) ~

0.885 is equal to ]/ jt/2 For instance, as noted earlier,
j
e~ x2 dx ==•
is given without proof in this book. o
Extensive tables have been compiled
°o m
for the function <D (x), and so com- y ji/2, where \
e~ x2 dx = lim l e~ x2 dx ,

J M-+oo J
putations involving integral (5.5.1), are 0 0
 +

170 5 Integration Techniques

not, in principle, be reduced to algo- P dx

5.5.11. Find the integral
rithms, while with differentiation tech- J V +a2 a:
2 *

niques (i.e. finding the derivatives) this dx

5.5.12. Find the integral
is possible. Often it is easier to use (x 2 +a 2 2
)
*

special tables when one has to find in-

tegrals of functions. Such tables of in-
5.6 Change of Variable in a Definite
tegrals contain lists of (indefinite) in-
Integral
tegrals classified according to the type
of integrand and other characteristics We consider an example. Let it be
that enables finding the sought inte- required to calculate the definite
gral easily, as well as certain types of Q
definite integrals (with limits of in- We
tegration 0 and oo, 0 and jt, oo and — integral
J
(ax + ft)
2
dx. can do as
v
oo, etc.). 5 4 When an integral cannot
-

follows. First calculate the indefinite

be evaluated, one has to resort to (ap-
proximate) numerical integration. integral
j
(ax + b) 2 dx and then form
the difference of its values for x =q
Exercises and x — p.
Find the following integrals: To compute
^
(ax + b) 2
dx we make
a change of variable using the for-
5.5.1. j* cos (3# — 5) dx. 5.5.2. sin (2a:
z = ax Then dz = adx
j mula + b. and

[\^3x — 2 —— —
i) dx. 5.5.3.

{Hint, Make
J

the
dx. 5.5.4.

substitution
[
J x+V
Y x=z.]
x'
j
(ax+b)*dx = ±
\
z*dz =z
+ ++. (

Therefore
x dx
[Bint. Make the substitu-
J(a* + i,)*<te =
J <2S±S|»
tion a:
2 —5= z.] 5.5.6. sin 3 x cos x dx. [Hint. V
^
_ (a? + &)
3 — (ap + b) 3

Make the substitution sina: =z or cos x = u.] 3a

C0S
.5.5.7. f
sin 4 x
^- dx. [Hint. Make the substi- However, it is possible to do other-
J
wise. Let us determine how z will vary
tution sin x = z.] 5 .5.8. tan x dx. when x varies from p to q. Since z and x
dx
j
are related as z ax 6, it follows = +
f
5 - 5 9- -
that as x varies from p to g, the new
J
variable z will vary from ap -f- b to
C dx
5.5.10. Prove that aq +
b Hence,.

J Y —£(a 2 2 3
)
q aq+b

a2 Ya — 2 t
+ C.
j
(ax + b)
2
dx = — |
z2 dz
V ap+b
while the corresponding indefinite integ- _ I
aq+b __(aq-\-b) 3 — {ap + b) 3
ral cannot be calculated. (One method of eva- 3 a \ap+b 3a
luating a definite integral that does not rely
on formulas for the corresponding indefinite When evaluating integrals, it is con-
integral is discussed in Section 17.3.)
5 4-
See, for instance, I.N. Bronsthein and venient to do just that way, that is,
K.A. Semendyayew, A Guide-Book in Mathe- when making a change of variable, to
matics, Deutsch, Frankfurt, 1971; H.B. Dwight, find the new limits of integration at the
Mathematical Tables McGraw-Hill, New ,
same time. This will obviate a return
York, 1941; and I.S. Gradshteyn and I.M. Ry-
zhik, Tables of Integrals Series and Products to the old variable in the expression for
, ,

4th ed., Academic Press, New York, 1966. the indefinite integral.
 2

. 5.6 Change of Variable in a Definite Integral 171

Let us consider some examples. We will show that there is a simple

1
dx
relationship between the functions ®
Calculate Note from and
1. [ -, 0
,1 ( —x )
3
. (p.

In the expression for cp(a) make the

6
the start that the function (2 x )~ 3 — change of variable kx2 = From this t
2
.

assumes positive values as x varies from we find that ]/ kx=t, x=t/\/ k, and
0 to 1, and therefore the integral is dx=(l/Yk) dt. It is clear that
positive. At the same time, the de-
nominator in this domain of integra- t —0 at x = 0 and t = a Y^ x = a.
tion does not vanish, so that the inte-
And so we get

grand is finite throughout the domain. aV~k aV~k

Make the change of variable 2 x — — cp ( a) e
-< 2 dt
e~ t2 dt
y, with dx —dy. Then y =
2 at = :

j Vic Vkl )

x =
0 and y 1 at x —
1, so that = ~

-V MaY k) -

Consequently, for an arbitrary value

0 2 of the independent variable x ,

On the right-hand side of (5.6.1) the

limits of integration are given for y .
<p(*) = JLO (xVk). (5.6.2)

The reader may wonder about the mi-

nus sign in the last equation. Indeed, If we have
a table of values of the func-
on the right and left we have integrals tion ®
it is possible
(#), to find the
of positive functions, so why is the values of cp (x) for any value of k and ,

right-hand side of (5.6.1) positive? The vice versa, knowing the values of cp {x)
point is that the lower limit of inte- which correspond to a definite value of
gration on the right is greater than the k, we can find, via (5.6.2), the value of
upper limit. Since an integral changes ® (z) for any z (in (5.6.2) we are
sign upon interchanging the limits of forced to put z = xY &). 5-5

integration, (5.6.1) may be written as 3. In Chapter 3

that the defi- we saw
follows: nite integral if the has dimensions
1 2 integrand and the limits of integration
f dx P dy
have. It is often convenient, however,
J (2-z)3 ” J
~y*'
to reduce the integral to a dimensionless
0 1
form by taking all factors having di-
Now, the right-hand integral, the
in mensions outside the integral sign.
upper limit exceeds the lower one and What remains to be done is to consider
it is clear that the integral on the right
is positive. The computation can now
be completed with ease: 5 *
5
Note that more customary are not tables
X
2

- A
x2
dy__ L — 2 of the values of 0(;z)= e~ dx, of which
[
J y*
~~
2y* 8
+ 2 “ 8
*
j
0
l we spoke at the end of Section 5.5, but

2. In Section 5.5 we considered the

those of the function (x) = r xZ/2 dx
(p ,
X
‘Y ^ i
function ® (x) = j
e~ x2
dx .One often has which is closely related (see for- to O (x)
o mula (5.6.2), where in this case we must put
to deal with the function cp (a) = k =1/2). The function cp(a;) plays a very
a important role in the more advanced section
of mathematics, the theory of probability
e~ kxZ
,

dx , where i: is a constant, of which we speak in brief in the Conclu-

j
o sion of this book.
 172 5 Integration Techniques

(transform) numerical (dimension- a and, finally,

Let us see how to trans-
less) integral.
form an integral to dimensionless form. 1

Suppose we have an integral

j
/ ( x) dx = / max (b — a) <p (z) dz,
b j
0
/ (x) dx, where, for the sake of sim-
j (5.6.6)
a
plicity, we assume f (x) to be non- 1

negative. Denote by/max the greatest

where <p(z)dz is a dimensionless
value of the function / (x) on the domain j
0
of integration and take it outside the
6 quantity.
integral sign: 5 -

If / (x) varies but slightly on the

domain of integration, then / (^)// max —1
(but, of course, / (#)// max ^l, since
/ (x) < /ma and
x) f
so (p (z) ~ 1 (but
1 1

cp (z) ^ and I = j
1) cp (z) dz ~1 j
dz =i
(5.6.3) 0 0
(but /^l). Then the dimensionless
1

It is obvious that in the last integral factor I = cp

(
2 ) dz is of the order of
j
the integrand f {x)/f m ax is dimension- 0

less, since / (#) and / max have the same unity and the value of integral (5.6.3)
dimensions. Let us pass to dimension- is determined mainly by the product
less limits of integration. To do this, /max (^ a) •

make the change of variable Let us consider a simple example:

the free fall of a body in the course
of time t 0 The distance of free fall
z— X
b Z-a ,
or x=a -b z (b — a)- (5.6.4)
.

to

is equal to |
v (
t) dt. The acceleration
We Since
see that z is dimensionless. 0
dx = (b — x
a) dz and z a 0 at = = of the falling body is g, and, there-
and z 1 at x =
b, it follows that the = fore, the rate of fall, or velocity,
integral on the right-hand side of (5.6.3) is v (t) =
gt (we assume that u(0) 0 =
takes the form and that the maximum velocity u max is
attained at time t q, or z^max gt 0 ). =
Note that the maximum here is not
= — a) f dz. due to any decrease in velocity after
J / max /max
a
J
0 t =t
0 but
simply to the fact that times
exceeding t 0 and corresponding to veloc-
(5.6.5) ities greater than gt 0 lie outside the
Set / {a -f- z (b — a))//ma x = <p (z). Then domain of integration 0 t t0 We
=
^ ^ .

=
from (5.6.5) we have introduce the notation z t/t 0 , cp (z)

v/v max = gt/gt 0 = t/t 0 = z. The result

b 1
is
f
dx= (b — a) \
cp (2
)
dz.
J /max J

6 * 6
We 0 and
have assumed that
/ (x) ^
/max ¥= In the case of a function that
0.
changes sign or is negative, for /max we must
take the greatest of the absolute values of
function /, or / max | |
.
 ) x

5.7 Integrating Functions Dependent on a Parameter 173

We see that in the given case the di- condition the result will be simply
1
incorrect; we will return to this con-
inensionless factor 1= qp {z) dz (whose dition later).
j
o The definition (3.2.4) of a definite
order of magnitude is generally equal integral reduces an integral to a sum
to unity but whose magnitude is of a large number of small summands.
always smaller than unity) is equal If the integrand f (x) / (x, s) (the =
to 0.5, which agrees fairly well with function u ( t in the notation of for-
our rough estimate. mula (3.2.4) or the function y ) in the (.

notation formula (3.2.6)) depends

of
on a parameter s, all the summands
5.7 Integrating Functions Dependent will also depend on this parameter.
on a Parameter Since the sum
of a finite number of func-
In Section 4.12 we discussed briefly tions our case, functions of para-
(in

the problem of differentiating func- meter s) can be differentiated and in-

tions dependent on a parameter. Here tegrated term by term, the function
we wish to investigate the question of
b
integration of such functions. The defi-
b /(x, s)dx — I(s) (5.7.4)
j
nite integral I = l
/ (x) c£x is a num- a

a
her, but if the integrand depends on a can also be differentiated and inte-
parameter s that is, f f (x, s ), the
,
= grated with respect to s under the in-
number will depend on this parameter, tegral sign:
too, that is, we will have I I (s ). =
For instance, it is clear that
+i r (
s) = i j (*’ s) dx= df{
~ dx '
1 j .[ d's

j
(kx + k~ l
)
dx =
^Y kx 2
-
fy
x
j J
^
a a

-i (5.7.5a)

X T 6

1 k j
I (s) ds = j £ (
/ (x, s) dx~^ ds
o o a
j' sin Ax dx = -jj- sin y dy b x
j

—
0 o

1 1^ 1 — COS A jr n o\
= j JT
j
/ (a:, s) cfejda:, (5.7.5b)

= x (- C os!/)
/ \

| Q
= ,
(5.7.2) a a

1
where under the integral sign on the
(s+l)xMz = (s + l) = 1 right-hand side of (5.7.5a) is the
j
0 derivative of / (x, s) with respect to s
for s> — 1. (5.7.3)
calculated on the assumption that x
is kept constant, what is known as the

We see that integral (5.7.1) depends partial derivative of / (x, s) with re-
on parameter k, integral (5.7.2) de- spect to s (see Section 4.12). (To prove
pends on number A, while integral the validity of the rules (5.7.5a) and
(5.7.3)seems to depend on s but this (5.7.5b) differentiation and inte-
of
dependence is fictitious, since I (5) 1
,

= gration under
the integral sign, it
is independent of s, that is, I (5 ) is suffices to replace the integral I (s) with
a function of s that equals unity iden- the approximating “integral sum,” of
tically. Example (5.7.3) is also in- the type used in formulas (3.2.4) and
structive due to the condition that s (3.2.6), to differentiate or integrate this
must be greater than 1 (without this
— sum term by term, and then pass to
 x )

174 5 Integration Techniques

the limit as the number of summands

grows without limit). For instance, we can state that I (A,)
= dx
j

a is a normally convergent integral and*

•si-1 <**+n*>= j
-&±p<ix
-l say, V (X) = dx =
l J
l

— (x — hr ) dx 2 00

j
-l
J
P cosTuz
x
^X '

and 1

If integral (5.7.4) not normally

is
P i i P
convergent, the direct application of
sin^5ia:da:jd^= sin Xx dh'jdx rules (5.7.5a) and (5.7.5b) may lead
j ( j j ( j
a 0 0 a to mistakes (here we may even encoun-
ter a case where the integral I (5) is
COS %x \X=P/x
<2# finite, while, say, its derivative,
-i(- ^ / h—(x/ constructed according to (5.7.5a),.
b

COS P
**>•
I' (
s)=
j
— •
dx , is infinite, or the
j a

other way round).

Difficulties may arise when the
integral (5.7.4) is “improper,” that is, For instance, we are sure that
00
when either the function f (x) in the
domain of integration becomes infinite = In (x + s) ~ = 00 (we assume
j
or the domain itself is infinitely large 0
oo M that s is positive). But if we find the
(here, say, [f(x)dx=lim f / (x) dx\ derivative of the integrand with respect
to 5, we arrive at the “finite” deriva-
cf. (7.5.5)). For instance, for sC 0 tive I' ( s ) of the “infinite” integral
the integral (5.7.3) becomes an im-
proper integral, since the function (s -f-
for such values of x become in-
s
1) x 7 J Q
ax
finite at x 0; for 0 s —1 this= ^ > a
ds
*
i

x+s~
* 1

dx
dx
J J
integral assumes a constant value equal 0 0
to unity; and for s —1 the integral <
itself becomes infinitely large. But if, 1 1 X=oo 1
oo dx
=
{x + s? x-\- s x=0 s *

say, I (s) / (x, s) dx is “normally

j
a
convergent,” that is, there exists a To clarify the meaning of this para-
function F (x) that is greater than doxical result, we proceed as follows.
oo
Let us “cut 08” the singularity in the
| / (
x ,
s) |
and such that j* F (x) dx is OO
dx
o .
,
that is, re-
finite, then we can deal with I (s) as S
6
if it were an ordinary, “proper,” in-
place the integral with the (proper)
tegral (for instance, formulas (5.7.5a)
N
and (5.7.5b) remain valid for this in- dx
2
tegral). For example, since |(sin tar)/# |^
J
,
where N is
oo
0

l/x 2 and
j
dxlx 2 = (— l/x) |“ =1 < oo, large. We then study the asymptotic
00 Obvi- N
behavior of IN (s as .
 ) , —

5.7 Integrating Functions Dependent on a Parameter 175

ously,
S1 — s2 *1 — s2 X=°° \

N dx- )• The
x=N X=1
IN (s)=l^T =\n(x+s)
\x= 0
result (5.7.6) can also be easily under-
= In (N + — In s = In (1 + N/s stood if we start by considering the
s) ) oo

whence defference I N (s 2 ) — IN (s 4 ) = l —x
+ s2
'
1N = “/v+7 T ’
OO
dx
r

which agrees with (5.7.4a) (since (s) J ^+ 5 i

0
is a proper integral): Here is one more striking example
N N dealing with a nonnormally convergent
7 »< s >- dx integral dependent on a parameter.
.(
0
i(iTi)*" J
0
Take the quantity
1 'x=N 1 1

x-\-s |x=0 N+s /( X)=j^fdx. (5.7.7)

If we now send N to infinity, we will

0

see that I N (s) tends to infinity and It can be proved that I (1)
=
oo
I'N (
s tends to —1/s, in complete
agreement with the result we arrived dx = jt/2; here we will not
\
above.
at o

The above-described procedure of cut- discuss the validity of this neat rela-
ting off the singularities often helps tionship. On the other hand, for X 0 >
understand the meaning of results re- we have
lated to nonnormally convergent in-
tegrals or even divergent integrals, re-
sults that at first glance seem para-
0 0
doxical. For instance, notwithstanding
oo

the fact that I (s) == = oo, the = j

-^dt, = /(li = -f ,

j 0
o
difference between the integrals I (s 2 ) while for ^<0 we have
and I (s T ) proves to be finite:

dx I (/.) = dx = HeLzIIU!) dx
j j
0 0

=-
j
— l
X]x)
x
dx =
-u X + S2 x + Si )
dx

OO
—s
$1 2

=J * 2 +(Sl + S2)* + SlS2 dx (at X =

0, obviously, I (0) 0). Thus, =
although a small variation in X changes
= lni (5.7.6)
the integrand but little, the inte-
*2 gral (5.7.7) for some reason experiences
a jump at X 0: =
the last integral is normally conver-
|

gent, since if, say, 0 we

Jt/2 if X > 0,

— si s2 Si — s /(*,)=! 0 if X = 0, (5.7.8)
have and
X2 + + X+
(
51 s2) s l s2 ' — k/ 2 if X<z0.
176 5 Integration Techniques

This is due to the fact that, when we

integrate an infinitely large
over
domain, even a small variation of the
integrand / ( x) may result in an appre-

ciable effect, since ft ( x) dx —

j

f (x) dx=
j
Ifi (x) — f(x)\ dx and here
j
we add small differences / 4 (x) —/ (a:)

“extended” over an infinitely large

domain. 5
In this case, too, cutting off singu-
larities in the integral (5.7.7) may
clarify the situation. Of course, the
N
I
—-— dx
Sill
depends the value of I N (A) is positive for
o A > 0, negative for A < 0, and zero
on A, continuous fashion, since
in a at A = 0. (Why?) If N is very great,
this function, I N (A), has a derivative: the function I N (A) very rapidly passes
from “large” positive values (close to
J'N (k) = Jj-
(1!^£) dx = ( cos kx dx jt/2) to values that are very close to
j
o o — jt/2, performing in the neighborhood
(while if we apply formula
oo
(5.7.5a) to
of point A =
0 a “rapid” jump (but, of
course, remaining continuous at this
/ (A), we get /' (A)
= j
cos Xx dx ,
whose point). But if N
is relatively small,
o the passage of the function I N (A) from
right-hand side not defined). 7 But
is
-

positive values to negative ones is much

it is clear (in view of the very definition smoother (Figure 5.7.1). In the limit,
of the improper integral (5.7.7)) that, when N oo, the transition from “es-
as oo, the value of I N (A) tends to sentially positive” to “essentially neg-
the values defined in (5.7.8); for any N ative” values of I N (A) occurs, so to
say, over an infinitely small interval
5 7 Compare this with the procedure,
- de- of values of A, that is, here the func-
scribed in Section 16.3, of finding the derivative tion I (A) =
lim I N (A) experiences a
of sgn x, a function that is closely related to N-* CO
/ (A) (we can easily see that J(A)= (jc/2) sgn A). jump (cf. Chapter 16).
 ) , )

Chapter 6 Series. Simple Differential Equations

6.1 A Series Representation count this change in y

f
(x) has the
of a Function form 6 -
1

X
Suppose that a function y (,x is de-
y (x) = y(a)+ j
y' (t)dt. (6.1.2)
fined exactly by a formula so complex
a
that it is awkward for calculating the
This formula is known as the -first or
values of y =
y (x). We
pose the problem ,

linear* 2 approximation for function

of constructing a simple and convenient
approximate expression for the func- y (*)•

tion y (x) over a small range of the

Applying (6.1.1) to the derivative
independent variable x say, for val-
y'(x), we get
,

ues of x close to a fixed number a . y'(x) ~ y‘ (a) + (x - a) y (a). (6.1.3)

The
definition of a derivative given
Before proceeding, let us recall that
in Chapter 2 may be written as fol-
y"(x) =
dy’ldx is the second deriv-
lows:
ative of function y (x) with respect to x ,

denoted also d2yldx2 It is obtained .

from y' just as y' is obtained from y.

y'(a)
V
= lim
x ~a ,
The third derivative y"' dy"ldx is =
x-+a
derived in a similar fashion. The fourth
derivative is denoted y IV the fifth y w ,

from which it follows that or y( 5) on. The derivative of

,
and so
order n or the rath derivative, which is
,

y(x) ~ y(a) + (x — a) y'(a), (6.1.1) obtained by taking the derivative of

the function y (x) n times in succession,
is denoted by or d n y/dx n (the sec- y^
where the approximate equality “exact is ond, third, and fourth derivatives are
in the limit,” that is, the smaller the denoted by y"' and?/ IV but not by
y\ ,

difference x a \
—
the greater the ac-
|
*/
(2 {3
\
4
y \ and z/( >. In the notation
n)
curacy. This formula shows that for y( the n is enclosed in parentheses to
,

values of x close to a the variation of distinguish it from an exponent.

the independent variable by a (small) Clearly, if x is measured in units of e i
quantity x —
a corresponds to a va- and y in units of e 2 then the dimensions ,

n
riation in the value of the function of the nth. derivative d y/dx n is eje™.
equal to y'(a)(x a), —
that is, the For instance, if y is distance (measured
formula fits the meaning of the deriv- in centimeters) and x is time (measured
ative as the rate of change of the func- in seconds), the second derivative (the
tion at point x =
a, or (dy/dx) x==a = acceleration) y" has the dimensions of
y’(a). cm/s 2 the dimensions of the increment
;

For large values of x a the |

— \
ac- Ay" of the acceleration are also cm/s 2 ,

curacy of (6.1.1) becomes poor: the which means that the rate of change
greater the difference x a |
— \
the 6 - 1
Formula (6.1.2) follows directly from
poorer the accuracy. Indeed, if we put X
Ay = y (x) y(a) —y’(a)(x =
a), — I*

=
\x
—
we are assuming that the rate of change
(3.4.5a), since
j
y' (t) dt y (
t I = y (x)
a
of the function everywhere between x
y (a). Formula (6.1.5) is verified in the same
and a the same and equal to the rate
is manner.
6 2
y' (a) of change of the function at point In the expression (6.1.1) for y, the quan-
*

a, while actually the rate y also changes ' tity x appears only in the first power, in other
words, y is a polynomial of the first degree in x.
in the interval between x and a. This relationship is termed linear because its
The exact formula that takes into, ac- graph is a straight line (see Section 1.3).
12-0946
 —

178 6 Series. Simple Differential Equations

of the acceleration y"' lim (Ay "/Ax) — applied not to the initial function y (x)
3
has the dimensions of cm/s and so on. ,
but to the function y"(x ):
let us return to the problem of
Now
the approximate expression of a func-
y"(x) ~ y" (a) + (x — a) y'"(a) (6.1.6)

tion. Formula (6.1.3) for the derivative and then substitute this expression
is nothing but formula (6.1.1) ap- into (6.1.5). From (6.1.5) and (6.1.6)
plied to y' instead of y. Now substitute we obtain
the approximate expression (6.1.3) for y'(x)
the derivative y' into the exact for- x
mula (6.1.2). The result is
X
^ y' (a) +J [y” (a) + (t — a) y'" (a ) ] dt,
a

y (x) c^.y (a) + J

[y
r

(&) + (£ a) y" (a)] dt

or
a
y' (x) ~ y' (a) -{-(x — a) y" (a)
= y (a) + — a) y'
(x (a) -f- -
- " V* a
( )
•

j 4-^f(«). (6.1.7)
(6.1.4)
Note that (6.1.7) is a formula of the
This formula is approximate but is more type (6.1.4) written for y' x ). (.

exact than (6.1.1). In deriving (6.1.4) Now let us substitute the expression
we took into account that the deriva- for y' (x) (6.1.7) into (6.1.2):
tive y'(x) is not constant, but the va-
V(x)
riation in y' (x) was considered only
X
approximately: formula (6.1.3), which
we made use of when deriving (6.1.4), <^y(a) + j
[y' (a) + y“ (a)(t — a)
assumes that y"(x) is constant (and a
equal to y"(a)), which is what gives
us the linear dependence of
y' on x. + dt
quadratic.
relationship
= y(a)+y' (a) (x — a)
For y (x) the is

Let us again check the dimensions in

(6.1.4) by employing the standard dis-
tance-time example, namely, if y is

measured in centimeters and x in sec- (6.1.8)

onds, the dimensions of y' and
then
The general law becomes all the more
y" are, respectively, cm/s (velocity)
obvious if we compare the expressions
and cm/s 2 (acceleration), while the quan- obtained above. In the crudest approx-
tities y (x) and y (a) have the dimen-
— imation we assume that for a small
sions of cm and the factors (x a)
difference x a |, —
and (l/2)(x a) 2
—
have, respectively, |

the dimensions of s and s'\ whereby y (x) —y (a) (6.1.9)

the dimensions of the left- and right-
(this fact obvious and does not re-
is
hand sides of (6.1.4) coincide (cm).
us make formula (6.1.4) still quire knowing
higher mathematics).
Let
more precise by taking into conside- We call this equation the zeroth appro-
We take ximation. Then expression (6.1.1) is
ration that y" is not constant.
called the first approximation expres-
advantage of the formula ,

sion (6.1.4) the second approximation ,

X
and expression (6.1.8) the third approx-
y' (x) = y' (a) + j
y" (t) dt, (6.1.5) imation. Listed, they are
a
y ~ y (a) (zeroth appoximation)
{x)
which obtained from (6.1.2) by sub-
+ (x — d) y' (a)
is
y(x) ~y (a)
stituting y' for y. Let us represent
y" (x) using a formula of the type (6.1.1) (first approximation)
 x ) c

6.1 A Series Representation of a Function 179

y(x)~y (a) + (x — a ) y' (a) We denote the right-hand side of

(6.1.10) (a polynomial of degree n in
variable x) by Pn
+£=f- »<«>
(x):

(second approximation)
Pn(%) = y {a) + Cx y (a) (x — a)

+ C 2y"(a)[x — a) 2

y(x)~y (a) + (x—a) y' (a)

+ c 3 y'"(a)(x — af
+ (X^fy" (a)+ i£^l y'" {a) + . . . + cn yW (a)(x — a) n . (6. 1. 11)

From this it readily follows thatP n (a) —

(third approximation).
y (a). Compute the first derivative
of (6.1.11):
It isnow easy to imagine the aspect
of the approximate formulas for y (x) Pn (#) = cx y'(a) + 2 c 2 y"(a)(x — a)
if the approximation process is con-
+ 3 c 3 y"'(a){x — a)
2

tinued: if we take into account that

y"' not constant, then y
is (a) will 1Y + . . . + nc n y (n) {a) (x — a) n
“1
.

be involved and the expression for 1.11a)

(6.
y (x) (the fourth approximation) will
contain a term with (x a)
4
Each — . This means that P’n (a) c x y’ (a). Now =
subsequent step in approximating y (x) let us require that not only the value of
yields an additional term with a higher the polynomiall(&AAi) at point x = 0
power of x —
a. One can expect that coincide with the value y ( a ) of the func-
the more powers of x a that enter — tion y {x) at the same point but also the ,

into the formula, the more exact the rate Pn [a) of the variation at point x = a
formula is. Of course, what we have be the same as the rate y' (a) of the varia-
said is valid only for small x a — tion of function y ). For this to hap- (.

—
\ \

while for large x a all the above

| |
pen, we must put
cx 1. Finding the =
formulas may prove invalid (see Sec- first and
higher-order derivatives of
tion 6.3). (6.1.11a) and replacing x with a in the
The general formula for the nth ap- resulting expressions yields
proximation has the form
Pn (a) = 2c 2
y" (a)
y (x)lcalyja) + Cjy'iaWXt— a) and P'n (a) = y" (a) at c2 .= = 1/2,

+ c y"(a)(x — a) + c y"\(a)\
2
2
z (x — a) 3
Pn (a) = 6 c3 y"' (a)

+ + c n ym (a) X - a) n
. . . ( , (6.1.10) and P'n (a) = y’" (a) at c 3 == 1/6,

which follows from dimensional con- p lY

(
a) — 24 C/
yv (a)
siderations. Indeed, if y is measured IV
in units of e 2 and x in units of e l9 then
pn
and r (a) j/iv (a) at C 4 == 1/24,

y(
k
(a )
)
= (dh y/dx h ) x=a has the dimen-
e 2 /e hi9 so that in the approxi-
sions of
rpin)
n (a) = 2-3-4 ... nyW (a
mate expression for y (x) the term with
( n)
y(
h)
must
contain a factor of di-
(a) and pn («) = ^> (a)
mensions that is, the factor
of e\,
at c n = 1/(2 !-4 ..n).
(x a)—k
Of course, these ideas do not
.
- .

enable us to calculate the numerical Thus, if we require that the values of

values of the coefficients c l9 c 2 c 3 , ,
. . . n successive derivatives of the polynomial
in (6.1.10). (We know that cx 1, — Pn (x) at x =
a coincide with the deri-
c2 —
1/2, c s =
1/6; but what are c 4 vatives of the same order of the function
and all the other coefficients?) These y (#) at the same point (this requirement
coefficients can be calculated by the means a closeness between P n (x) and
following method. y (^)), this means that we must put
12 *
 )

180 6 Series. Simple Differential Equations

c1 = 1, c2 = 1/2, c s = 1/(2- 3), c4 = X

1/(2- 3-4), . . ., cn = 1/(2 3 - . . . rc). y(x) = y(a)+ y' (t) d (t — x)

j
Thus, a
x

(£) = p(a) + y' (a) (x

— a) = y(a) + y’(t)(t-x) I*- ( t-x)dy’(t
j
a
+ (x — a )
2
X

+ •+ )“• f 6 - 1 - 12 * = y (a)+y' (a)(x—a)+

j
(x—t)y"(t)dt.

(6.1.14)
We have arrived at the approxima- If we agairijSubstitute d(t — x) for dt and in-
tion formula tegrate by parts, we get

y(x)~y (
a) +y ' (a) (x — a) X X

(x — t)y"(t)d(t—x) = - y"{t)d ^J?-

{

j j
+ j
Q?L (x-a)* +!^-(x-a)* a a
X
*- x
+ + 2 .f^ -(^~a)
n
. (6.1.13) = y"«)
(
a
t
n+ |a j
{fL=Jl dy »( t)
a

X
We
have a convenient designation for
x-a 2 1
the product of a succession of natural = V W , (

2
)

+2
,

j
(x—t) i y"'(t)dt,
numbers 1-2-3. .n. It is n\ and is . a
read factorial n. For example, 3! = whence
1-2-3 = 6, 4! = 1-2-3-4 = 24, 5! =
1-2-3-4-5 = 120, etc., while 1!= 1 and y (*) = y (a) + y ' (a) (x — a) + y
(x — a) 2

2! = 2. Using the factorial notation, X

we can rewrite (6.1.13) as + -J- j (*— t) 2 y'" (t) dt (6.1.14a)

y (*) ~ y(a)+±-^-(x— a) Performing the integration by parts n times,

we get an exact expression for y (x) consisting

+ i^ +
of n 2 terms:
(
x_a)»+... + J^-(*-ar.
y{x) = y(a) + y' (a) (x—a)
(6.1.13a)

Here a fuller proof of formulas (6.1.13)

is
+ £wL(—‘f+Tr"-* 1
'

and (6.1.13a), which also leads to results that

are stronger than the ones obtained above.
Let us return to the exact equation (6.1.2)
and substitute d(t x) for dt under the in- — X
tegral sign: (*- i)V n+1
+-^j- > (6.1.15)
X J
a
y(x) = y(a) + ^
y’
(
t)dt
Thisformula, in contrast to (6.1.13a), is
a exact, it was obtained as a result of
since
x transformation of the exact formula (6.1.2).
X
= y( a )~ j* y' (t)d(t x).
t"
The last term, r 7i = ~j" (x—t) n y( n+1 )(t)dt,
a j
a
on the right-hand side of (6.1.15) is known as
This substitution is justified since the upper
the remainder ; it is the difference between the
limit of integration x is regarded as a constant
— y (x) in (6.1.13) and the approximate express-
and therefore d(t x) dt. = ion for y (z) expressed by the right-hand
Let us now integrate by parts (see formu- side of the same formula.
la (5.4.3a)) to get
 )

6.1 A Series Representation of a Function 181

Formula (6.1.15) can be rewritten in a ourselves to a finite number n of the

more convenient form that does not contain
an integral: 6 * 3
first terms in the series. The statement
that for sufficiently small x a |
— |

y (*) = V (a) + y (a) (X

— a) '
( x ~ a) 2 formula (6.1.17) is “exact” means that
the sum on the right can be made as
ym (a) (x — a) n (x— a) n+1 close to y x ) as desired (what is required
(,

+ ... -\-
y<
n+1
> (c), is a sufficiently large number n of terms).
n\ (n + l)\
We can also say that y {x) is the limit
(6.1.16)
of the sum of the first n terms of the
where c is a number lying between x and a.
oo. 6 5
series as n -
.

The coefficients c 0 c l9 . . ., cn, . . .

In the general case of an arbitrary ,

in (6.1.17) are different for different func-

function y (x), no finite number of pow-
they also depend on point a.
ers oi x —
a with constant coefficients
tions;
As we have seen earlier, c = y(a)
0 9
can yield an absolutely exact formula
for the initial function. 6 4 In other -
c! =/(a), c2 = y"(a)/ 2!, c 3 = y »/3!,
etc., that is,
words, the right-hand side of (6.1.13)
or (6.1.13a) only approximately coin- y(x) = y (a) + y' (a) (x —a)+ y~- (x—a) z
cides with y (x), but is not equal to
y (x) exactly. An exact formula can be
obtained only by adding an infinite + ^(x-a »+... )

number — a:
of powers of x
-f
^t( X —a) n + .... (6.1.18)
y = c + (x — a) + c
(%) 0 2 (x — a)
2

A particular case where

+ c (x — a) + 3
3
. . .
a =0 is
of (6.1.18)

+ Cn (# — a) n + * . *. (6.1.17)
y(x) = y (0) + y' (0) x + xz
The expression on the right-handside
of (6.1.17) is called an infinite series
Ordinarily, we drop the word “infinite”
.
+ (6.1.19)

and simply say “series”. Of course, Using the summation sign, we can
we can also say that (6.1.17) is true write formulas (6.1.18) and (6.1.19)
“in the limit”, but this means some-
in compact form as follows:
thing different than it did above. The
77=00
crux of the problem is that we cannot
add directly an infinite number of V(x) = v(a)+ 2 (
x —a n ) >

terms; thus, in writing out the sum 77=1

on the right, we are forced to restrict (6.1.18a)

3
6
As for the validity of the transition
*

from (6.1.15) to (6.1.16), see the text in small y(x) = y(0)+ 2 (6.1.19a)
print at the end of Section 7.8, in particular, 77=1
formula (7.8.13), in which the integration
variable x must be replaced with x t z — — These two formulas well as (6.1.18) (as
and the function / (x) with y ( n+1 ) ( t (which and (6.1.19)) yield the expansion of
depends on the new variable x t z; we — — a function y (x) in a series of integral
remind the reader that d (x t) dt, — = — —
since x is regarded as constant). We then get powers of x a and x respectively. ,

X The series on the right-hand side of

(x-t ry<n+l) [t) dt = J jL Zjig t l ?(
n + i )(c)> 6 5
*
It may so happen that for different val-
j
a ues of x the accuracy with which the sum of the
first n terms approximates y (x) depends on n.
where c is a value of the independent variable (We will not dwell on the general questions
lying between x and a. of the applicability of formulas of type (6.1.17)
6 4
;
Except for the case when y (x) is a poly- and of the functions for which the expansions
nomial (see the beginning of Section 6.2). (6.1.18) and (6.1.19) prove to be invalid.)
 . 2

182 6 Series. Simple Differential Equations

(6.1.18) or (6.1.18a) is called a Taylor's x is close to a, while formula (6.1.21a)

series while the particular case of a
,
is good when x is close to zero.

Taylor’s series, present on the right- In Chapter 2 we gave a definition of a de-

hand side of or (6.1.19a),
(6.1.19) is rivative as the limit of the ratio of the incre-
known as Maclauriri' s series ment of the function to the increment of the
Examples of applying Taylor’s and independent variable. Now that we have ex-
pressed the function as a series, we state gen-
Maclaurin’s series to concrete func- erally the law according to which the ratio
tions are discussed in Section 6.2, but a Ay/ Ax approaches dy/dx as Ax tends to zero.
simple example of representing func- Let us take Taylor’s series (6.1.18) and
tions in the form of series will be con- denote x a by Ax. —
then shift y (a) to We
the left-hand side and divide both sides by
sidered here.
—
Ax denoting y (x) y (a) by Ay. The result
Suppose that y = e
x
. Then is
,

y = e
x
,
y" = e
x
, . . y^ = e
x
, . . .. - = V' (a)+~y"(a) Ax
-jjf

We take advantage of the formula

+{f( 4)(Ai)»+.... (6.1.22)
(6.1.19) for a Maclaurin’s series. In the
case at hand, For small Ax the second term on the right-
,

hand side, containing Ax, is greater than the

y( 0) = y' (0) = y" (0) = . . . = l. third. Discarding the latter, we conclude that
the difference of the ratio Ayl Ax from the
Substituting into (6.1.19), we arrive value of the derivative y' (a) at the endpoint a
at the expansion of the function y e
x
= of the interval extending from x a to x = =
in a Maclaurin’s series: a +Ax is proportional to the second deriva-
tive y" (a) and the size of the interval, Ax:

^l + *+X + -T+-'- + ^r -^y’ (a)+~y"(a) Ax.

6 . 1 . 20 )
( Note that here we compare the ratio of the
increment of the function on the range of
Let us now examine the formula ob- variation of x considered in the problem to the
tained from the Taylor expansion increment of x with the value of the derivative
(6.1.18) if we confine ourselves to, y'(a) at the endpoint a of this interval.
say, three terms in the series: The derivative may be evaluated differently
by assuming Ax/2) that
f (x [/ (x + — —
Ax/2)]/2 is its approximate value rather than
y\(x)\~ yl(a) + {x—a) y' y" (a).
[/ (x Ax) +
/ (x)]/2. —
In other words, to
calculate the derivative at point a, we take
the increment of the function as x varies from
Removing brackets on the right-hand —
side and arranging the result in increas-
a Ax/2 to a +
Ax/2 and divide it by Ax.
Let us now compare this new ratio with the
ing powers of x, we get derivative y' (a) evaluated at the midpoint
of the interval. We get

y Mb* [y'M) — ay'M) + -y-V'j(a)] A J/ = /(a+^)-/ (a~4f) ,

+ [y'da) — ay" (a)]]x + ^y' i a )\ x -

f («+4n) — /( a + ^r /'(<*) )

(
6 1 21 ). .

(a)+
On the right is a polynomial of de- It) f
Tl~) f (a) ’

gree two. Note that this polynomial does

not coincide with what we would have /
(a —^-) ^
if we took three terms in Maclaurin’s
series:

y\(x) ~'i/(0) + i/' (0)x+V-fp x2 (6.1.21a) .

2
6*2
This vill become clear if we recall that (Ax) 2
~ *
,
a' ,

*
.

formula (6.1.21) yields a good result if ' ‘

24
 a a 1

6.1 A Series Representation of a Function 183

We see that the new method of estimating while the more exact formula (the one that
/' (a) is much more
exact: the difference be- allows for all four terms) is
tween the ratio of the increments and the de-
rivative is proportional to (Ax) 2 and not to
1
„
W^
.
, / (a + 2A x) — 2/ (a+ Aa) + / (a)
Ax and, what is more, contains the small coef- ' (A*) 2
ficient 1/24. Thus, approximately,

f'(a) ~ f (a+Ax) — f (a)

(6.1.24)
—~f"'(a) Ax. (6.1.26)

Ax
To obtain /" (a), it is best to combine not
while the more exact formula follows from the values / (a), / (a +
Ax), and / (a + 2 Ax)
( 6 1 22 ),
. .
of the function / (x) calculated on the interval
—
f {a)
~ MMzIM Y f" (a) Ax.
between a and a
Ax ), / (a), and
+
Ax). (Here the value of
/ (a
Ax but the values
+
/ (a

the independent variable x a coincides with —

(6.1.24a) the midpoint of the range of variation of x.)
Indeed, it can be easily shown that
On the other hand,

+ Aa;/2) - /(a - Aa/2) o / (a + As) — 2/ (q)+/(q- Ax)

r {a) « /(a
. (6.1.25) t W-d .

(Ax)*

while the more exact formula follows from or, more precisely,

^ o3 f(a + Ax)~ 2/A 5 + /(a— Ax)

(6.1.23), (a)
x

r (a) „ f(a + *x/2)-f(a-Ax/2) f (a) (

^

-—/'"(a) (Ax)K (6.1.25a) ~i /IV(a)(A * )4 (6 1-27)

-

(see Exercise 6.1.6).

All these equations, (6.1.24), (6.1.24a), (6.1.25),
Geometrically, the differences between the
and (6.1.25a), are of course approximate, approximate formulas (6.1.24) and (6.1.25)
since, we Jija;ve J o^nred^higji.errord£r- rryaniities- p 1
t Tu
: :
ior Icne Aieriv Airve lie Tu
‘
* ia 1
in comparison to those that enter into these
first case we replaced the tangent t to the graph
formulas (with respect to Ax).
of the function y / (x) =
(Figure 6.1.1) at
In a similar manner we can obtain appro-
ximate formulas for calculating the second de-
point M
(a, / (a)) (the slope of the tangent
line at this point is equal to /' (a)) with the
rivative. By definition, straight line ± Z, whereMM 1 = x (a M =M +
+
Ax / (a ,
Ax)), while in the second case t
NN =
replaced with the straight line ± 2 lx ,
J
dx Ax *

N =N
where 1 1 (a Ax/2, f (a —
Ax/2)) and —
— N =N + 2 (a Ax/2, / (a Ax/2)). It is clear +
Replacing A/' with /' ( Ax) /' (a) and + 2
that the straight line l ± is much closer in di-
estimating /' (a +
Ax) and /' (a) via the ra- rection to the tangent line t than the straight
tios [/ ( 2 Ax) +
/ (a —
Ax)]/ Ax and [/ (a+ + line Z (see Figure 6.1.1). In a similar manner,
Ax) —
/ (a)]/ Ax we arrive at an approximate
,
formula (6.1.26) is obtained on the assump-
expression for the second derivative of the tion that the rate of change of derivative y'
function that incorporates [/ (a 2 Ax) + — (the rate with which the tangent to the curve
2/ (a +
Ax) +
/ (a)]/ (Ax)
2
To estimate this .

y =/ (x) rotates as point

moves along the M
ratio, we again turn to Taylor’s series: curve) is estimated through the differences in
variation of the derivative from point M
/(«+ 2Ax) -/(«) + !/' (a) (2Ax) to point x and M
from point x to point 2 (x-{- M M
2 Ax, f (x +
2 Ax)), while for formula (6.1.27)

+ |/" (a) (2Az) 2 -f-L f (a) (2Ax)\

f (o+ Ax) ~ f\(a)+ y /' (a) Ax

+ \f' (a) (Ax)*+y£ V" {a) (hX? '

Using the first three terms on the right-hand

sides of these formulas, we obtain

f (a-\- 2Ax) — 2/ (a-\-Ax)-\-f (a) _ 1

(Ax) 2
~ 3
! yh
 x

184 6 Series. Simple Differential Equations

the rate of change of derivative y' is estima- simpler than a direct computation of
ted through the differences in variation of the
the function.
derivative from point M' ( Ax f (x — —
M Since only in the case of a polynomial
,

Ax)) to M
and from to M
1 (make a rough
sketch). does a Taylor’s series terminate, or
contain a finite number of terms, it
follows that any function different from
Exercises
a polynomial will be represented by an
6.1.1. Expand
the third-degree polynomial infinite series (this statement will be
y =ax 3 + cxbx 2 + +
d in a series in powers proved somewhat later). The practical
of x —x0 where x 0 is an arbitrary fixed num-
,
value of such infinite series for com-
ber. Compare the sum of the first two, three,
and four terms of the expansion obtained with putational purposes is due to the pos-
the polynomial. sibility of confining ourselves to a few
6.1.2. Expand the function y xe x in = (commonly two or three) terms of the
a Maclaurin’s series. Verify that the expansion
series in order to obtain a sufficiently
can be obtained from the expansion of e x mul-
tiplying the latter by x. accurate result. This requires, of course,
6.1.3. Expand the function e x in a Tay- that the discarded terms of the series
lor’s series in powers of x 1. — be small.
6.1.4. Determine the accuracy of the ap-
Let us consider a few very simple ex-
proximate formula (1 r) + m~
e mr for 1 r<
and mr <
1. To do this, write the left- and amples of series expansions of functions.
right-hand members approximately as the We have already mentioned the (simple)
sum a br + +
cr 2 obtained by ignoring the
, example when the function
third and higher powers of r.
6.1.5. Given the interval Ax 1, 1/2, = V — b2 x 1 + .
-f- bn x
n
(6 2 1)
1/4,and 1/8 and using formulas (6.1.24) and
(6.1.25) (or (6.1.26) and (6.1.27)), respectively, is a polynomial of degree n. Its deriv-
find the approximate values of (a) the deriva- ative y'(x) is a polynomial of degree
tive of e x at x = 0 and (b) the second deriva-
tive of e x at x =
0. Compare the results with
n —1, the derivative y"(x) is a poly-

the exact values. nomial of degree n 2, and so on; —

n)
6.1.6. Prove formula (6.1.27). finally, its derivative y( (x) is a con-
n+1
stant, while z/( > (x) and all higher-
order derivatives are identically zero.
6.2 Computing the Values For this reason, for a polynomial the
of Functions by Means of Series
corresponding Taylor’s series termi-
Let us dwell briefly on the principles nates; in other words, it contains a
underlying the formulas of Section 6.1. finite number of terms. have ar- We
When we began the study of the deriv- rived at the same polynomial, but in
ative, or the rate of change of a func- the form of a Taylor expansion in pow-
tion, we assumed the function as known, ers of x — a:
that is, we proceed from the fact y = c0 + c l
(x — a) + c 2 (x — a) 2

we could compute the value of the n

function for any value of the indepen-
+ . . . + c n (x— a) . (6.2.1a)

dent variable. This is why, when we For polynomials of nth degree, the sum
considered derivatives, we found them of the first n +
1 terms in the Taylor’s

directly, empirically so to say, by series yields the exact expression valid

computing the values of the function for all values of x rather than only for
for close-lying values of the independent values close to a while formula (6.2.1)
,

variable. Later on we learned how to can be regarded as Maclaurin’s series

find derivatives by formulas and it (6.1.19) for our function y(x). (Why?)
turned out that setting up formulas for In the preceding section we considered
derivatives is a rather simple matter. the exponential function y ex and =
And so finding the values of a func- obtained formula (6.1.20),
tion by means of a formula involving
derivatives (all formulas of Section 6.1
were of this type) turns out to be even ( 6 . 2 . 2)
 n

6.2 Computing the Values of Functions by Means of Series 185-

TABLE 6.1

X e
x 1 *+*+ —
*2
1
+*+~r + V i +*+— + 3C2
—+
0C3 x4
-24-

0.10 1.1052 1.10 1.1050 1.1052 1 .1052

0.25 1.2840 1.25 1.2812 1.2838 1.2840
0.50 1.6487 1.50 1.6250 1.6458 1.6484
0.75 2.1170 1.75 2.0312 2.1015 2.1147
1.00 2.7183 2.00 2.5000 2.6667 2.7083
1.25 3.4903 2.25 3.0312 3.3568 3.4585
1.50 4.4817 2.50 3.6250 4.1876 4.3986
2.00 7.3891 3.00 5.0000 6.3333 7.0000

In particular, substituting x 1 and = increases ra-fold, and all we have to da

x = —
1, we get expressions for num- wait until the increase in the de-
is to
_1
bers e and e in the form of series: nominator overcomes the increase in
the numerator (that is, until n becomes
greater then x). As can be seen from
1 Jr 1 + 4- + 1T + • • * + ~T+ • • •»
Table 6.1, when x 2, the sum of =
(6.2.3) the first five terms in the series yields
an error of 5%. But if we add the sixth
± = l_i +'2±_i +
r !_JL term, x b l 120, then we get 7.3500, which
* 6 24 120
yields an error of only 0.5%.
1 1
+ (2n)! ( 2/2 + 1)!
... (6.2.3a) us construct formulas of the
Let
same type for trigonometric functions.
Formula enables us to com-
(6.2.2)
For instance, for y (x) =
sin x we have
x
pute e rapidly and to a high degree of y'
(#) r= cos x y" (x)
,
= — sin x ,

accuracy, as can be seen from Table 6.1

If at x =
0.1 we confine ourselves
w
y (x) = — cos x y
1Y {x) = sin x
,

only to the first two terms, the error The law for subsequent derivatives is
will not exceed 0.5% of the true value; obvious. Substituting x — 0, we get
at x =
0.5 the first three terms yield
= o, y' (0) =
an error of 1.4%; finally, at x 1.0 — y (0)
y"( 0 ) =0 y'" 0)
1,
= —1 ....
, ( ,
the first four terms yield an error of
1.8% of the function y e
x — . Consequently,
Such a high accuracy is plainly due
to the fact that the terms of the series sm x=x g- +i 20
— 5040^7
(6.2.4)
(6.2.2) fall off rapidly. Each subsequent
term in the series is less than the pre- In similar fashion we get the formula
ceding one primarily because the de- X4
cosx =
X*_
(6.2.5)
+
|

nominator of the ( l)st term is n 24“ 720 "i”

times the denominator of the preceding
nth term. If x <
1, then in addition Figure 6.2.1 shows the graphs of tho
we have that xn is the smaller the great- sine, cosine,and also the graphs of the
n
er n is, and x rapidly decreases as n polynomials if we take one, two, and
grows. But even if x 1, the increase > three terms in the corresponding se-
in the denominator in the distant terms ries. Accuracy improves visibly when
of the series as n grows will inevitably we take more and more terms of the
overcome the increase in the numera- series.
tor, since as we move from the nth term Tables 6.2 and 6.3 list the values of
to the (n +
l)st term the numerator the sine and cosine functions, respec-
increases ^-fold while the denominator tively. In the tables both cp and x are
186 6 Series. Simple Differential Equations

the tables that two or three terms of the

series suffice to obtain excellent accura-
cy in the interval from 0 to jt/4. Thus,
a power series offers a very convenient
practical method for computing the
values of trigonometric functions. Note
that in absolute value the nonzero terms
of the series for the sine and cosine are
exactly equal to the corresponding terms
x
of the series for the function e For .

this reason, everything we have said

pertaining to the falling off of terms
with high powers of x in formula (6.2.2)
x
for e refers also to the series (6.2.4)
and (6.2.5), and these series, just like
enable calculating the value of
(6.2.2),
the corresponding function for an ar-
bitrary x.
It must be emphasized, however, that
the fact that the terms in a series de-
angles, but cp is expressed in degrees, crease, as n -* oo, for an arbitrary x
while x the corresponding angle ex-
is and, the more so, the sequence of “par-
pressed in radians. It is evident from tial” (finite) sums tends to a definite

TABLE 6.2

*3 X3 X5
X <p° sin x X 3C_
IT + 120

0 0 0.0000 0.0000 0.0000

jt/20 9 0.1564 0.1564 0.1564
3T/10 18 0.3090 0.3142 0.3090 0.3090
Sji/20 27 0.4540 0.4712 0.4538 0.4540
4:n;/20 36 0.5878 0.6283 0.5869 0.5878
5jt/20 45 0.7071 0.7854 0.7046 0.7071
&it/20 54 0.8090 0.9425 0.8029 0.8091
7it/20 63 0.8910 1 .0996 0.8780 0.8914
83T/20 72 0.9510 1 .2566 0.9258 0.9519
Dit/20 81 0.9877 1 .4137 0.9427 0.9898
3T/2 90 1.0000 1.5708 0.9248 1.0045

TABLE 6.3

X <p° cos x i_ i-
X2
2
+
x4
24
1
X2
2
+
X4
24 ~
x6
720

0 0 1 .0000 1 .0000 1 .0000 1 .0000

n/20 9 0.9877 0.9877 0.9877 0.9877
n/10 18 0.9510 0.9506 0.9510 0.9510
3n/20 27 0.8910 0.8890 0.8911 0.8910
An 120 36 0.8090 0.8026 0.8091 0.8090
5ji/20 45 0.7071 0.6916 0.7075 0.7071
6ji/20 54 0.5878 0.5558 0.5887 0.5877
7jt/20 63 0.4540 0.3954 0.4563 0.4539
Sn/20 72 0.3090 0.2105 0.3144 0.3089
9ji/20 81 0.1564 0.0007 0.1672 0.1561
n/2 90 0.0000 -2.337 0.0200 -0.0009
 2

fi.3 Cases Where Series Expansions Cannot be Applied 187

6
number as n serve oo, can in no way In particular, 6 *

as a general rule: in certain respects the

series (6.2.1), (6.2.1a), (6.2.2), (6.2.4), j* e
<2
dt =x — xh
2dT
x1
F7
and (6.2.5) and more “safe” than the
0
general series (6.1.18) and (6.1.19). Sec-
x 2n+i
tion 6.3 is devoted to this question. + **. +(~D n n\ (272 + 1) ( 6 2 10 )
. .

Taylor’s and Maclaurin’s series find wide Exercises

application in higher mathematics; some of
these applications are discussed below (e.g. Express the coefficients
6.2.1. c0 ,
cx ,
see Section 6.6). Here we wish to discuss the c2 ,... of the series (polynomial) (6.2.1a)
problem of employing series in the integration in terms of the coefficients b 0 bx , ,
&2, . . .

of functions. If the function y f (x) can be — of polynomial (6.2.1).

expanded in a power series, 6.2.2. It is known that the sine-integral
function six = \
J^JUdx cannot be expressed
j (x) =a l) -\-a 1 x-{-a i x'i -\- . .
.
+a n xn + .
J *
by a formula explicitly. Represent si x as
( 6 2 . 6)
.
an infinite series.

then, generally speaking (with certain reserva-

tions discussed in Section 6.3; here we will
6.3 Cases Where Series Expansions
not touch on these restrictions), Cannot be Applied. The Geometric
Progression

dx In the preceding section we set up

/ (x)
j formulas in the form of series in powers
of x with constant coefficients for four
— ^
(a 0 Jra 1 x-\-a 2 x 2 -{- . .
.
+a n xn -{- . .
.) dx
functions: the polynomial, e
x
sin x, ,

and cos x. In all these cases it was

~ C + a 0x + x 2
+ x3
found that for arbitrary x each subse- ,

quent term in a series, with the pos-

+ ...+^*n+1 +.<*. (6.2.7) sible exception of the first few terms,
is less than the preceding one, and the
greater the number-label of the term,
This result is valid regardless of whether the the closer the term is to zero (while
indefinite integral (antiderivative) dx in the first, “safest,” case all terms
/ (x)
j starting from a certain term vanish).
can be expressed by an algebraic formula or
not. For instance, above we stated that the
In these examples, we can compute the
value of the function for any x by means
function ^e~ t2 dt cannot be expressed by a
of a series if a large enough number of
(finite)formula, but an infinite series (“for- terms of the series is taken so that the
mula”) can be employed to express it. Re- discarded terms have practically no
placing x with t
2 —
in (6.2.2), we get
effect on the result.
To summarize, then, we began with
the problem of approximating a func-
tion in a small range of the variable and
f27l constructed more and more exact formu-
6 2 8)
( . .
las by taking into account the first,

Formula (6.2.10) is applicable for any

6 - 6
-whence
finite x\ however, it is clear that the important
oo
b
e~ t2 dt = C+t — t* ,

1
t

result^ e
~l
dt =y nl2 cannot be derived
j 3 10 42
271+1 6
^
+ ...+(-iy 72 ! (272 + 1)
(6.2.9)
from (6.2.10)— quite
required to do this.
different methods are
 —

188 6 Series Simple Differential Equations

.

second, third, and higher derivatives. We thus arrive at the series

The accuracy of each formula,
y ~y (a),
(x) (0)
—^- = l + x-f-x + x 2 3

y (x) ~y (a) + (x — a)y (a), (I) + ... + xn +.... (6.3.1)

y(x) ~y(a) + (x—a)y (a) The example of the function (1 x)~ 2 —

is remarkable not only due to the unusu-
+ -^V(a), (II) ally simple form of the resulting power
series (all the coefficients are equal to 1).
etc., is the greater, the smaller the Here it is easy to give an exact formula
quantity x a |. Moreover, for a — for the sum of the first n terms of the
—
|

given x a formula (I) is more accu- series (6.3.1):

rate than formula (0), formula (II) is
more accurate than formula (I), and i + x+x + 2
(6-3.2)
so on. Hence, if we increase the number
of terms of a series, this permits increas- The validity of this formula, the formu-
ing the quantity x a while pre- |
— \
la for the sum of the n terms of a geo-
serving a given accuracy. metric progression is known to any se- ,

The question now arises as to whether condary-school student. (Just multiply

it is always possible to attain a given both sides of (6.3.2) into 1 x.) can — We
accuracy for any value of x a |
— \
rewrite formula (6.3.2) as follows:
merely by increasing the number of "1
terms in a series. We will use an im- 1 +x+x + 2
. . .
-f x 71

portant example to illustrate this point

and will show that this is not so. A pow-
er series constructed so as to yield a
good approximation in a small range
Comparing (6.3.1) and (6.3.3), we see

of x that is, for small

,
x —- a |, can
that x n /( 1 x) —
is the remainder in
the approximate equation
—
|

be useless in the region of large x a | |.

In other words, this series may have a

natural limit of applicability, the limit
~YZ~p — 1 +# + # + 2
. . .
-f #n_1 , (6.3.4)
of admissible increase in x a (the |
— |

limit not depending on the number of

that is, the quantity we neglect when
terms taken). In the examples of the
confining ourselves to the first n terms
preceding section this was not evident
in the series
because the series were specially
chosen, but such good behavior is an
exception rather than a general rule. 1 -{-x-\-x ZjrX 3 -\- . . .+x njr . . (6.3.1a)
Consider the following simple func-
tion: which is the right-hand side of (6.3.1).
If x lies between —1 and 1, then the
V= greater the n, the closer x
11
is to zero
and, consequently, if we take a suf-
Taking the derivatives in succession ficiently large number of terms, we dis-
yields card only a small quantity (remainder).
1 1-2 Series that possess this property are
=
V' (i x) 2 ’ y (1 —z )
3
known as convergent . The convergence
ym ~
n\ of the series (6.3.1a) at x = 1/2 is illus-
(1 — x)n+ i ’
*

trated by the table below; here 1/(1 —

x) = 2. The quantity A in the table is
Substituting x = 0, we get
the error (in %) introduced by the ap-
y (0) = 1, y'( 0) = 1, y "(0) - 2, . .
proximate formula (6.3.4), that is, the

y
W (0) = n\ ratio of the difference between the left-
 .

6.3 Cases Where Series Expansions Cannot he Applied 189

and right-hand sides of (6.3.4) to the

left-hand side.

n 1 2 3
1 +£ + 2 +;r . . .
+ 1 1.5 1.75
A 50% 25% 12.50/0

n 4 5 6
1 + x+x + 2
• • -
,
+ xn~ 1
1.875 1.9325 1.9637
A 6.25% 3.12% 1 .56%

Note that the closer x is to 1, the

more terms we have to take in order
to obtain a given accuracy.
The whole picture changes if we take
|
x |^1. For instance, if x 1, each > series does not tend, as n -> oo, to any
subsequent term in the series (6.3.1a) finite limit.
is greater than the preceding one. For-
Also note that any periodic (decimal)
mula (6.3.3) remains valid, but for sum of terms of a geometric
fraction is a
x > 1 the quantity x
n
increases with-
progression. For instance,
out bound together with n and for this
reason we cannot disregard the frac- l.(l) = l.llll..
tion xn /(l —
x). Here, (6.3.1) does not = 1 + 0 1 + 0 01 + 0 001 + ...
. . .

hold true. There is not even any qual-

1

itative similarity between the sum of = 1 + 0 1 +( 0 1) + 0 1) + ...

. .
2
(
.
3

the positive terms of (6.3.1a) and the

negative (since x is greater than 1) quan-
— — 0.1 __
1
— 0.9~ 10 ” 1
1

9
1

9
, 4_
‘

tity 1/(1 —
x). For instance, at x = 2
formula (6.3.1) becomes absurd: Thus, we have already encountered an
elementary series (the geometric pro-
—1 = l + 2 + 4 + 8 + 16 + . • •• gression) in arithmetic and algebra.
The function y 1/(1 x) (Fig-= —
For x > 1, the sum of the series (6.3.1a)
ure 6.3.1) has a discontinuity at x 1; =
increases without bound with n (this
if x is close to 1 but greater than 1,
is clearly seen from (6.3.3)). And if —
then 1/(1 x) is a large (in absolute
x ^— 1, in the right-hand side of
value) negative number, while if x is
(6.3.1) we have an alternating sum,
close to 1 but smaller than 1, then
which is not represented in any way by —
1/(1 x) is a large positive number.
the quantity 1/(1 x). For instance, — Thus, when x passes through the value
at x = —1 formula (6.3.1) yields x = 1, the value of the function 1/(1
—
1/2 = 1 — 1 + 1 —1+ — 1 x) jumps from large positive values to
large (in absolute value) negative num-
Series similar to those into which (6.3.1a) bers. A series cannot describe such
transforms when |
x | ^ 1 are called behavior, and so it is of no use here.
divergent . We note yet another circumstance.
(
L ,\ x —
the terms of an infinite geometric pro- x) becomes infinite (the closer x is to 1,
gression (6.3.1a) is equal to 1/(1 x) — the greater y is in absolute value), and
if |
x |
<
1, but if x 1, the in- | | ^ at x =1 the terms in (6.3.1a) cease to
finite geometric progression has no decrease as n -> oo. series is suit- A
finite sum (it diverges), in other words, able for computational purposes only
the sum of the first n terms of such a if its terms tend to zero rapidly, that
 1 x

190 6 Series. Simple Differential Equations

is, decrease in absolute value. 6 7 At *

This yields y( 0) = 0, y'(0) = lr
x =
1, the formula (6.3.1) is incorrect, y"( 0) =0, y'"
(0) = 2, y iv (0) = 0.
since the terms of the series on the right- y
v
(0) = 16, and so on. Whence,
hand side do not decrease. This means
that the series (6.3.1a) is unsuitable for tan a: = 0+l*a; + 0-;r 2 + - | -j x3 -\- 0-x*
computing the values of the function
y =
1/(1— x) atx 1 and for x = — <—
(the terms in (6.3.1a) do not diminish
either) although the function itself does Thus,
not have a discontinuity at x = —
and is equal to 1/[1 (—1)1 1/2. — = tan £ = + -i
a:
+^ x7
No matter how we choose the coeffi-
cients of a polynomial, the graph will
x9 + .... (6.3.5)

always be a solid continuous line: a

polynomial does not have discontinui- (the coefficients of x 7 and x 9 can be
ties. Therefore, if some function / (x) obtained in the same manner as the
has a discontinuity at x x 0 (x Q 1 = — x 3 and x b were ob-
coefficients of x, ,

in the example involving 1/(1 #)), — tained).

then for the value x x Q the series = What can be said of the range of ap-
constructed for / (x) is definitely unsuit- plicability of the series (6.3.5)? The
able for computations. Since the great- graph the tangent function (Fig-
of
er the absolute value of x the greater ure 4.10.5) tells us that this series is
,
n suitable for computational purposes
(in absolute value) each term c n x in
the series c 0 cx x c 2x 2
+
... is, it + + only for x <. jc/ 2, since at x
| |
jt/2 =
follows that for an arbitrary x that is the function’s behavior is as bad as that
greater in absolute value than x 0 the ,
of function 1/(1 x) was for x =1. —
series is likewise unsuitable for compu- But the appearance of the series
tation. (6.3.5) does not suggest the value of x
Thus, in the case of a discontinuity in for which the series cannot be employed,
a function / {x) we can indicate before- since the law by which the expansion
hand an x 0 such that for all x exceeding x 0 coefficients are constructed is not simple,
in absolute value the series represent- in contrast to the law for the coeffi-
ing / {x) will prove to be unsuitable cients in series (6.3.1a). 6 8 -

for computational purposes. Note that the presence of a disconti-

Consider another example. We wish nuity of a function is a sufficient con-
to constructMaclaurin’s series for the dition for the series to cease to con-
function y =
tan x. Applying the gen- verge, but it is not a necessary condition.
eral rule for finding derivatives, we ob- By way of an illustration let us consider
tain the function y 1/(1 x). Applying = +
formula (6.1.19) or simply replacing x
y = tan x sm x i
with —
in (6.3.1), we get
cos x y COS 2 X
9

2 sin x
y'"
2+4sin 2 x
x
~ 1 — x-\-x2 — x 3Jr • • *• (6.3.6)
y'w-cofz (*) cos 2 x
_]_

Take x = 2, for instance. Then

*/
IV
(*)
, v 16 sin a; + 8 sin 3 x
1 1 1
5
cos X
i+x \x=2 1 +2 “ 3 •

y
v (x)
16 + 88 sin 2 £+16 sin 4 x
cos 6 x 6 8
Note that the ratios of the successive
-

coefficients in the series (6.3.5), 1 1/3 3, =

6 7
Of course, if one or two or several of the
*
1/3 -4- 2/15 =
2.5, 2/15 -4- 17/315 2.4706, -
first terms increase, there is no harm done if 17/315 -4- 62/2835 ^ 2.4677, rapidly converge
the subsequent terms of the series fall off to the value (ji/2) 2 ~ 2A67A. However, there
rapidly, see the example involving e x for is no way in which we can prove this remar-
x =2 (Table 6.1). kable fact here.
 .

6.3 Cases Where Series Expansions Cannot be Applied 191

Figure 6.3.3

For the reader acquainted with complex

numbers (see Chapters 14 and 15) we note that
the convergence of a series for real-valued x
is affected by the behavior of the correspond-
ing function for complex values of the inde-
pendent variable. Here is an example (see also
Section 15.2). Replacing x with x 2 in (6.3.6),.
we get

The sum of the terms of the series _l_ = l_* 2+ *4_ x6+ .... (6 .3. 8) ,

1 —x+ x 2 —x + 3
. . (6.3.7)
The graph of the function y — 1/(1 +x 2
)
(Fig-
ure has no discontinuities for positive-
6.3.3)
however, oscillates rapidly as the num-
and negative x's and does not go to infinity,
ber n of terms increases, and does not but the series (6.3.8) is suitable for computa-
resemble 1/3 in any way: tions only if x 2 <
1 that is, for —1
,
x 1 < < .

The reason for this is that when x 1— ± V— =

7... =bi, that is, at x 2 = — 1 , the function y =
n 1 2 3 4 5 6
Sum of n terms
1/(1 +x 2 ) becomes infinite and therefore the
terms of the series do not decrease in absolute
of series (6.3.7) 1 -1 3 -5 11 -21 43 . .
value at x 2 = —
1. Hence, neither do they de-
crease in absolute value at x 2 =
1 (see p. 481).

The series is clearly unsuitable for

= Other examples of functions that
calculating at x 2 the value of the
= Why cannot be expanded in a Taylor’s se-
function y 1/(1 x). +
does this
-1 /* 2
=
occur, particularly since the function ries, say, the function e at x 0
itself, y =
1/(1 +
x), has no discon- (see p. 198), can also be cited. Fraction-
tinuity either for 2 or anywhere be- al powers of x and functions that con-

tween 0 and x =
2 (within these limits tain fractional^ powers of the indepen-
the function is smooth, well-behaved, dent variable, say, the function sin x Y >-

see Figure 6.3.2)? cannot be expanded in a Maclaurin’s

The reason why the series (6.3.7) series.
does not “work” at x =
2 is that y = Remaining within the scope of this
1/(1 + x) has a discontinuity at x = book for beginners, we find it rather
— 1. For this reason, at x 1 the = — difficult to go any further into the ques-
absolute values of the terms of the se- tion of the applicability of Taylor’s
ri as_ ( {T 3, 73 da noh diminish, asL n. cgswfL. and, MnrJaurjn’a, sari or, ta da this,, the.
7
7

But the absolute values of the terms in reader must turn to a textbook on the
(6.3.7) do not depend on the sign of x. theory of functions of a complex va-
Consequently, for x —
1 (and all the riable (cf. Chapter 15). However, we
more so for x >1) this series is not must say, to comfort the reader, that
suitable for computation. practically all functions that a physicist
Thus, even if we are interested in or engineer needs can be expanded in
the behavior of a series only for posi- Taylor’s series everywhere with the
tive values of x, we still have to take exception of a finite number of (singu-
into account all the values of x includ- ,
lar) points. As a rule we encounter func-
ing negative values as well, for which tions that can be expanded, at least in a
the function undergoing expansion has properly chosen finite interval,** into'
a discontinuity. Taylor’s series.
 )

192 6 Series. Simple Differential Equations

Exercises and so are all higher derivatives. The

6.3.1. Write Maclaurin’s series for the formulas (6.4.1) to (6.4.3) reflect this
functions (a) y = (x + 1)/(1 — x) , and circumstance. Indeed, at 72 = m 1 the +
<b) y In (1 x). = + factor 772 72 1 —
vanishes, for n + >
6.3.2. Write the Taylor expansion of y = +
there will be, some place in the
1
In x in powers of x — 772

m—
1.
What are the ranges of applicability of the sequence of the factors m, 1, . .

series obtained in Exercises 6.3.1 and G.3.2? a factor equal to zero and, the pro-
6.3.3. Suppose that / (x) and g (x) are duct will be equal to zero, too.
known functions. Find the first three terms For a positive integer tti, the product
of the series expansion in powers of x of the
in the numerator can be written in a
function / (a-) g (x). Construct the same series
by multiplying together the series for / (x) more convenient form. To this end we
and the series for g (x). Compare the results. multiply and then divide the prod-
uct 772(772 — 1). . .(772 — 72 + 1) by
6.4The Binomial Theorem for Integral
(772 — 72) (772 — — 72 1). . .3-2-1. The
resultis
and Fractional Exponents
Let us form Maclaurin’s series expan-
772 (772 — 1) . .
. (772 — + 72 1)

sion of a binomial a x to an arbi- + _ m(m — 1 ) . . . 3 2-1

-

_ m\
trary power m, that is, y = {a + x ) m .
(m—n) {m — n — i) ... 3 - 2*1 (m — n)\ *

Using the general rule, let us first

find the derivatives
Thus, for positive integral m we finally
have
y' = m (a +
y" = m (m — 1) (a + x)
m ~2
, . . .,
{a + x) m^ a m + _ (mm\— i)\
am
~i
x
rn\
(6.4.1) .
m
a -*x 2
n =m (m — 2! (to — 2)!
y 1) . . .

. . .(m — 72 + 1) (a + x)
m ~n
,
. . .
+ ... +
ml

and the values of the function and de- ml

rivatives at x = 0: + • • • + (to — 2)! 21

^(0) = am ,
y'
(0) = ma m ~ 1 , to !

y"{ 0) =m (m — 1) am 2
"
, . . .,
(to — 1)! 1!
(6.4.4)

(6.4.2) In formula (6.4.4) we have polyno-

y
n
(0) — m (m — 1). . .
mials of degree m on the right and on
the left. Thus, for the case of a positive
. . .(m — n + 1) am n
~
y .... integer m, we obtain an exact equa-
From this we get Maclaurin’s series tion that is valid for arbitrary powers of
x, that is, formula (6.4.4). This formula
m
x) = a
m
+ ^am- x
r
(a + i
is symmetric with respect to x and a:
the coefficients of the terms a m ~ n x n
and a nxm ~ n are the same. This is obvi-
ous since (x m does not depend +
a) on
(6.4.3)
. .

-j-
in ( in 1 ) ( in 2)
— . .
. ( n% — tl —J— 1 the order of the summands in the pa-
-t- . . .
m m + =
rentheses: ( x a) (a x) +
X a m n xn +
-
.... (6.4.3) Formula (6.4.4) is called the binomial
theorem (sometimes it is called New-
If the exponent
a positive inte- m is ton’s binomial theorem, see below) or the
m
ger, then (a a polynomial of + x) is binomial expansion. It can be obtained
degree m, so that in this case the series without resorting to derivatives and
is finite: the (m + l)st deriv- series. We have to take the prod-
ative of the function (a + x) m is zero, uct (a -f- x)(a -f- x)(a -F x). . .(a -
r x)
 aa 2

6.4 The Binomial Theorem for Integral and Fractional Exponents 193

m m with
consisting of cofactors, perform the In the expansion of (a x) +
multiplication, and collect like terms. arbitrary m, all the terms have the same
However, when is specified in the m sum of powers of a and x each subse- ,

general form by a symbol and not a quent term differing from the preceding
number, collecting like terms is ra- one by the factor xla and the coefficient.
ther difficult. On the whole, the deriva- A physicist would say that a and x
tion of the binomial expansion via me- in formula (6.4.3) must have the same
thods of higher mathematics, that is, dimensions, and so xla is dimension-
using Maclaurin’s series, is simpler. less. From the very beginning we could
We note that Newton obtained the take a outside the brackets:
general formula (6.4.3), that is, the ex-
m for the case of an + x)
m = am + x/a) m
pansion of (x a) + {( (1 ,

arbitrary exponent m. It would there- and expand (1 + x/a) m in powers of

fore be more appropriate to call formu- xla .

la (6.4.3) Newton’s binomial theorem It turns out that for all m (fraction-
instead of (6.4.4), which was known be- al, and negative) the
positive, series
fore Newton’s time 6 9 and which is a -

(6.4.3) is suitable only for xla | | < 1,

simple particular case of the relation-
ship connecting the coefficients in the
that is, for x <C a For xla
| |

the series in (6.4.3) is divergent. The

|
|. | | ^ 1

representations (6.2.1) and (6.2.1a) of positive integers are an exception m

a polymial, see Exercise 6.2.1. because in that case formula (6.4.3)
Let us return to the general formu- consists of a finite number of terms..
la (6.4.3). Suppose that is not a posi- m Putting a =
1 in (6.4.3), replacing x
tive integer. Since the powers n of the with r (where r <C 1), and confining \ |

variable x are positive integers, this ourselves to the first two terms on the
means that not a single factor m , right-hand side of (6.4.3), we arrive at
m 1, — ., n 1. in the nu-
. m— + the following approximate formula
merator of the coefficient of xn will
vanish and, hence (6.4.3) yields an (1 + r)
m ~1+ mr,
infinite series. For instance, for m= which was often used above,
—1 this series is of the form this for
2 3
mula is valid only for small r (a
1 _ 1 x . x x
more exact formula has the form (1
| |

+
a-\-x a a2 '
a3 a* '

(6.4.5) r)
m ~ + 1 mr +m
Note a 1 formula (6.4.5)
that at = Formula (6.4.6) offers a good meth-
passes into the familiar formula (6.3.6) od for taking roots. Here, the smaller
for the sum of the terms of an (infinite) the ratio xJa the fewer terms one has
| |

geometric progression. to take in (6.4.6) to attain a specified

From (6.4.5) we also find that accuracy.

a—x
4
i
\
=.-Lj
a ^
X
a2
-4-
^ —^ ^
X2
a3
I-
X2
a4
1- * * * *

Exercises
For m — 1/2 we have
1 x2 6.4.1. Using a se ries expansion, find
Ya + x—Va>- 8 aV Via and V as Vl+x
1.5 at a: =0.1 and
x =0.5 retaining two, three, and four terms
L *4 5 7 x5
in the expansion. Compare the results with
i

16 a 2 ~\f a 128 a s j/" $

|

256 a 4>
Y the tabular values.
21 6.4.2. Show that for x < 1 the approx-
(6.4.6)
ya'"'
| |

1024 imate formulas (a) f/"l +x~ 1 +x/n, and

In Europe formula (6.4.4), where m is
6 * 9
(b) 1 y +x ~ 1 -j-
are valid —n
and - a:
2
a positive integer, was first discovered by n 2n2
Niccolo Tartaglia (c. 1500-1557), but even that the smaller the value of x, the more accu-
before that the formula was known to Arabian rate formulas (a) and (b) are (formula (b) is
mathematicians. more accurate than formula (a)).
13-0946
 194 6 Series. Simple Differential Equations

6.4.3. Using the fo rmul as of th e pre ceding

2 X4
exercise, find y+.2, f/ 1.1, and f/ 1.05, com-
A
1 cos x = ~2X 24
"
.

• • • >

pare the values obtained with tabular values.

6.4.4. Find ]/*6 to three decimal places. e
x — l = x + -f + . . . . (6.5.1)
[Hint. Employ the fact that 6 4 2 and = +
=
2 and apply formula (6.4.6).] This for one, yields (sin^)/x =
6.4.5. Why is it impossible to expand 1 —£ 2
/6+.... Hence, (sin x)/x-+\
y— Yx by Maclaurin’s formula? Can the
as £->0, or lim-
sin;r
=1. Similarly y
function y — z 1000 be expanded using
Y this
x-*- 0
x
formula?
from (6.5.1) we find that

tan x
X
= 1+
ir
r2
+ ...-l as to
6.5 The Order of Increase and Decrease 1 — cos x _ 1 x2 1
of Functions. L’Hospital’s Rule x2 2 24 + •••->-
2
as x i

The
1 — cos a: x
u as x —^
series expansion of functions yields
x 2 24
^ * * * 1

a general method for reducing different

functions to the same form and enables
one to compare the functions. This meth-
+ ...-*-1 as x^
od of comparison is needed, for exam-
ple when we consider the ratio of two
More complicated relations can also
be found. For instance, from
functions, f (x)!g (,x ), for a value of the
independent variable x for which the
values of the two functions are close
sm x = x — 120
r5
x°
'

to zero. In the computation of deriva-

tanx = a; + -^- x z + ^x h {-
tives it was demonstrated that the ratio
of two almost-zero quantities can be
it follows that
a quite definite number. Such ratios
may prove to be not very small (but tan x — sin = y +a: a:
3
-g- + .

not very large, either). In certain cases,

this ratio may be equal to zero or in- and, hence,
finity (positive or negative), and such tan x — sin x 1
cases can be classified. few examples A z3 2
as x 0.

will suffice. For the sake of simplicity

of notation, we take examples in which A scale can be constructed of the
the value of x that interests us is equal order of decrease of various functions as
to zero. x tends to zero. If / {x) tends to zero
For small values of x, the functions as x -> 0, then we term the order of
sin x and tan x are also small. The decrease (or order of smallness) of / (x)
function ex and cos x are close to unity with respect to x the power of x that
and, hence, e x 1 and 1 —
cos x are — decreases just as rapidly as / (x). Pre-
small. Here the smaller the value of cisely, if we say that f {x) has fcth order
x the closer are the values of the of decrease with respect to x this means ,

—
| |

x
functions sin x, tan x, e 1 and ,
that it decreases as x k that is, the ratio ,

1 —
cos x to zero. / (x)lx
h
has for its limit, as x -> 0, a
Let us compare these functions with finite nonzero number.
x To do this, we write out their Mac-
. Thus, sin x tan x, ex 1 decrease by
,
—
laurin-series expansions: order one as x -> 0, while 1 cos x —
decreases by order two, and tan x —
^3 sin x decreases by order three with
sin£ = ;r g- + . . . ,
respect to x.
£3 In certain cases it is possible to de-
tan x — x + -g- -4- ... ,
termine the order of decrease without
 6.5 The Order of Increase and Decrease of Functions. UHospitaVs Rule 195

dent can master, while the ingenious

methods of ancient scholars were suit-
able only for the best minds of the
time.
Using series, it is possible not only
to find the ratio of a function to a power
of x, but also the ratio of one function
to another. Here are some examples:

<y*2 /v»3

Figure 6.5.1 ex -
-T+T-+-
sm x
a series expansion. For instance, from *—Xr+- s
,

the well-known Figure 6.5.1, where MP

and NA are lines of the sine and the i + T-+...
tangent corresponding to /_AOM x, = as 0,
we can easily see that tan x x > > 1
•r2

T+
sin x (the reasoning is based on the fact
that SaOMA
Since sin x
^sctr OMA
~ < Saona )*
ex —i *+—+•••
7*2

tan x for small x's (be-

cause in this case sin x ! tan x = cos x ~ 1 — cos x X2 X4
1), we conclude that sin x ~ x and ~2 "24* '

tan x ~
x when x is small, that is,
sin x and tan x have the first order of 1
+ 1T+...
decrease. Next, since 1 cos x — — X
~ X ”+ •••3
00 as x- -
0,
2sin 2 (x!2), and sin (x!2) is of the first ~2
24
order of decrease, we conclude that 3-2
1 —
cos x must be of the second order of ex — 1
+... Ya
*3/2

decrease. Finally, the function tan x — 2

sin x can be written as sin z/cos x — Y~x Yx

sin x =
(sin z/cos x) (1 —
cos x). Since -v 0 as x—>-0.
for small x's the function sin x is of the
first order of decrease, 1 cos x is— The coefficients of Maclaurin’s series
of the second order, and cos x 1, ~ are expressed in terms of derivatives.
it is clear that tan x sin x —
must It is therefore possible to state the re-
be of third order of decrease. However, sults obtained by means of series in
these concrete devices require a great the form of rules referring to deriva-
deal of ingenuity and so precisely for
this reason a general method that oper-
tives. If / (0) = g (0) = 0, then
ates
useful.
without failure is particularly
/(*) = m+ f(o) x
Such
ous
a relationship between ingeni-
solutions of individual problems
+ |/"(0)x + 2
. . .,

and general methods is in evidence every-

where: the properties of tangent lines e (*) = g (0) + f'(0) x
to a parabola, the area of a circle, the
volume of a pyramid, and the volume + | *"(0) X 2 + . . .

of a sphere were all familiar to the

can be simplified thus:
ancient Greeks, but only differential
and integral calculus provided us with
general and simple methods for solv- /(x) = /'(0)x + -if(0)x2 + ....
ing all problems of that type, methods
that any engineer, physicist, or stu- g (x) = g'(0) x +~ g"{0) x 2 + ... .

13 *
 - - ;

196 6 Series. Simple Differential Equations

From this we have tions can be replaced with the ratio of

their derivatives. This result is cal-

m
g{X)
f'_ (Q)*+4-/
—
g
(0)* 2 +--- led L’ Hospital' s rule (actually discov-
ered by Johann Bernoulli and given
g" (0) * 2 +...
g' (0) x-\ to L’Hospital in return for salary). 6 11 -

After studying series, it is more con-

f'(0)+-j-f'(0)x+--- venient not to bother remembering some
_ ^
/>( 0 )
'
special rules for finding derivatives
8 {0)
g'(0)+4-e"(0)*+... but, for small values of x, to use series
in which the function is expanded in
as x-^0 powers of x. Whenever there is a sum
of different powers of x, we leave only
and, hence,
the lowest-degree term when passing
if f (0) = g (0) = 0, to small values of x .

Just as we considered, for small x ,

then •

g(x)
-4 -7
^r as ^->0 (6.5.2)
v the order of decrease of functions equal
g (0)
to zero when x 0, we can examine =
(if, in addition, /'(0) g'(0), then = the behavior of functions when x in-
the ratio f{x)/g'(x) tends to the limit creases without bound that is, as x -> ,

of the ratio f"(x)/g"(x) as 610 ^->0 .

oo. Here it is appropriate to expand
Of course, rule (6.5.2) retains its the function in powers of Hx (since
meaning completely when f{a) = g(a)= 0, 1/x is small by hypothesis) via substi-
where a is an arbitrary number, and tution of t for l/x. If our function is
we are interested in the value of the a polynomial, then it is obvious that
ratio / (x)/g (.x ) when x is close to a: for large values of x only the highest-
degree term in x is of importance, since
= g (a) then all other terms are smaller: axn +
if / (a) 0,
f(x) f' (a)
bxn
-x
+ . . . = axn 1 + -y -i-
+ . .
)
>
g' (a)
as x-+ a (6.5.2a) ( .

g (*) here the expansion in the parentheses

tends to 1 as x oo. can speak of We
(if,in addition, f(a) g'(a), then the = the order of increase of a function, that
limits of the ratios /' {x)/g' (x) and is, that a function increases as x as x 2 , ,

f (x)!g {x) coincide as x ->• a, in view as x 3

etc. It is said that a function
,

of which / (x)lg (x) -> /" ( a)!g " (a) as / (x) that increases without bound in
^->a). Here, one must use Taylor’s absolute value as x oo has kth order
series instead of Maclaurin’s series (see of increase if the ratio f(x)/x h has for
Exercise 6.5.2). its limit, as x -*• oo a
,
finite nonzero num-
Thus, if the values of two functions ber. It is also clear that any polynomial
are close to zero, that is, if the two of degree n has an order of increase
functions vanish at a single value of equal to n.
the independent variable in the vicin- A fact of prime importance is that
ity of which these functions are con- the function e x increases faster than any
sidered, then the ratio of the func- power xn for x increasing without bound,
that is, e x has an infinitely large order
6.10 Since here, in view of the same of increase, or k =
oo. To prove this,
/'(*)__ f
use the series expansion (6.2.2) of ex
(0)
formula (6.5.2), lim g' ~ n we arrive
x -*• 0 (*) g (0)
which, as pointed out above, is valid
r /(*) no) —
at lim— t r
g"( oy
Similarly, if / (0)
x-+0 g
= 0, then n The same
/' (0) = r (0) = g (0) = g' (0) = g" (0)
6 -
term is used in similar situ-
ations, instance, the one mentioned in
for
lim
f(x) f( 0) and so on. footnote 6.10 and those examined in Exer-
g(x) w 0
x-*- 0 g ( ) cises 6.5.3 and 6.5.4.
 6.5 The Order of Increase and Decrease of Functions. UHospitaVs Rule 197

It is risky to use the series expansion of

e~x for large x to prove this because the
expansion is an alternating expansion,
so that the various terms compensate
each other to a certain extent. there- We
fore consider the reciprocal:

1 // = x- n /e~x = e*/x
n
.

According to (6.5.4), for arbitrary n

the quantity / _1 = ex /xn tends to oo
as x -> oo which simply means that
Figure 6.5.2 / 0 as x oo.
To summarize: in the limit, for large
any value We have absolute values of the independent vari-

— 11
for of x.
able in the exponent, the exponential
T—n!__
pX 1
_L j t
TL _i function e x depends more strongly on x
xn n
x x "- 1
than any constant power of x that ,

is, for any integer n the function e x

(rc + l)! + 0? + 2) !
+ • ’ • (6.5.3)
increases faster than x n and e~ x de-
,

creases faster than x~ n This is vividly, .

For a given n and a sufficiently large x
demonstrated in the table for xh and
,

x n
the fraction e !x will become as large x
e \
as desired due to terms with positive
powers of x on the right-hand side of
(6.5.3). Clearly, the same goes for the X 1 3 5 10
Xh
function e hx with any positive value 1 243 3125 10 5
of ft (or for the function a
x
e x ln = ex 2.72 20 150 2X10*
with a >
1): setting kx y, we find — X*
ex
e -x

x~ 5
0.37 12 21 5
that

eb* __
xn
]
*
n eh*
{kx) n
~ h ^n'
oo as oo,
X
X*
20
3X10 6
50
3Xl0 8
100
1Q10
y
ex 4xl0 8 5xl0 21 1Q43
i.e. as x->-oo. (6.5.4)
xb e -x
0.01 10” 13 icr 38
hx ~e*~~ x~ b
Thus, any exponential function e ,

where ft 0 >
(or a x
where a 1) grows, >
faster, as # ->• oo, than any power func-
tion £
n
(or cx
n
where both n and c are
,
What we have said about the expo-
positive); although for small values of nential function can be applied to the
x the graph, say, of y — x 1000 slopes logarithmic function y log a x, name- =
upward much faster than the graph of ly, the logarithmic function log a x
the function y x — e 0
001x
(Figure 6.5.2), (with an arbitrary base a 1) increases, >
with the growth of x the graph of func- as x -> oo slower than any (however
,

tion y 1 will finally intersect the graph small) power xn of the independent vari-
x x
of function y, since y x will overcome y. able while,
the function e~ ! for
As x -> oo, the exponential function x 0 >
and x 0 decreases faster than
with a negative exponent decreases faster any positive power of x:
~v
than any negative-power function x .

This assertion, for arbitrary n 0, > 0 as (6.5.6)

is written
*0 as x->0 and #>>0. (6.5.7)

/ = = xn e'x ->0 as x -> oo

To prove (6.5.6) it is sufficient to in-
(6.5.5) troduce the substitution log a x u, =
 — * ) x) ,

198 6 Series. Simple Differential Equations

x = a
u
xn
transforms — a
nu
which 6.5.5. Prove that the function e~1 / x grows
,

the ratio of interest to us into ula

nu
,

= in absolute value, for x 0 and 2 -^ 0 faster c ,

nu than any (integral) negative power of x.

1 i- % a
- /u while to prove (6.5.7) it is
,

sufficient to introduce the variable v =

lAr, which transforms the ratio into 6.6 First-Order Differential Equations.
er !v~ n The Case
v of Variables Separable
(compare with (6.5.5) and also
see Exercise 6.5.5).
If we turn to the function w = Above we encountered (e.g. see Sec-
tion 3.6) the concept of a differential
e -!/*% which is naturally assumed equal
= equation , that is, an equation that
to zero at x 0 (since for small; in
absolute value, x that is, large values ,
connects the (unknown) function y =
of 1 Ar 2 w e
-1 / x2
=
can be made as small y x and its derivatives:
(,

as desired), we can state that the func- F { , y, y’ , y\ . .

.) = 0, (6.6.1)
tion decreases, as x 0, faster than
any power of x that is, w!xn -> 0 for ,
where F is a known function depending
all n >
0. This implies, in turn, that on several variables. The laws of sci-
ence and the operation of various devices
it is impossible to expand this (seem-
ingly well-behaved) function in a or mechanisms can often be described
Maclaurin’s series (see the graph of using the language of differential equa-
this function in Figure 15.1.1). In- tions, so that in many cases a substan-
deed, if the formula (6.1.19) were valid tial part of studying the phenomenon

for this function, then a number of that interests us consists in analyzing

first coefficients on the right-hand side and solving an appropriate equation.
of this formula could vanish and the In Chapter 9 we will illustrate all that
series expansion would be w (x) = has just been said with examples taken
a hxh a k+1 3^ +1
+ ., with a k 4= 0, + . .
from mechanics, a science whose basis is
and the function would have, for small N
formed by ewton s second law the dif- ,

values of x a finite order of decrease ( k ferential equation (9.4.2).

,

with respect to x instead of an “infinite” The order of a differential equation

order of decrease (cf. Section 15.1). is the order of the highest derivative
which appears. Here we limit our dis-
cussion to first-order differential equa-
Exercises
tions:
6.5.1. Find the following limits:
In ( l+a) (! + «>-* F (x, y, y’) = 0, (6.6.2)
(a) lim ,
(b) lim jP, ,

x -*0 x x
x-*- 0
or, if we solve this equation for the de-
tan x —x e*- -1 — tan x
(c) lim (d) lim rivative y '

x-*-0
x2 ,

x-*0 x3

(e) lim
Px -1
lim
sin x- -x y' = f (a;, y)- (6.6.3)
„0 sin
:

a;
(f)
x-v 0 x — tan x ’

Prove the
6.5.2. variant (6.5.2a) of
The simplest case of an equation of
L’Hospital’s rule. type (6.6.3) is when / is independent of y,
6.5.3. Prove that if f(x) f'(x) f'(x) = = = that is, when the equation has the
. =
. f
n ~ l (x)
. 0 at a; 0, = and, similarly, = form
s(0) = = = (0) = 0,
g'(0) g"(0)=. .. g(
n -
then 1)

lim
1{x)
= m (where, of course, we
y' = f (*). (6.6.4)
*-0 ?(*) g (
x)
assume that the last ratio, which can vanish
This equation implies that y is the an-
or become infinite, exists). tiderivative of function / (x), or the
6.5.4. Prove that (a) if / (x) 0 as x oo indefinite integral of / (x):
and g (x) 0 as i-)- oo, the ratios / ( x)/g (x)
and f'(x)/g'(x) tend to the same limit as
x -> oo, and (b) if / (x)-> oo as x-+ a (where,
y = j
/ {x) dx. (6.6.5)
possibly, a oo and g (x) =oo as x a, -
r f(x) f'M Chapters 3 and 5 were devoted to the
x-*-a g(*) x—a 8 & solution of Eq. (6.6.4). Of course, this
 )

6.6 First-Order Differential Equations 199

equation the simplest, and its solu-

is rivative (or in the same complicated
tion follows from the very definition of way as Eq.
since finding the
(6.6.4),
the infinite integral, speaking of (gen- antiderivative may constitute a com-
eral) differential equations, it is use- plex problem). Precisely, if the func-
ful to bear in mind this first example tion / (x, y) is the product g (x) h (y)
that sheds light on the general situa- of a function of x and a function of y,
tion. For instance, already in this exam- then the variables x and y in the equa-
ple we see the ambiguity in the state- tion
ment of the problem, namely, the y
presence of an infinitude of solutions to I
-=g(x)h(y ,) ( 6 . 6 . 6)
a given equation. For instance, in the
case of Eq. (6.6.4) these solutions have
can be separated by carrying all terms
the form
dependent on x over to one side of the
equation and the terms dependent on
y = G (x) + C, (6.6.5a) y over to the other. Such equations are
called differential equations with
where G'(x) = f (x). The graphs of all variables separable .

the functional relations described by Let us divide both sides of Eq. (6.6.6)
y =y (x) are obtained from a single by h (y):
graph by parallel translation along the y
1 dy
axis (see Figure 6.6.1). To select a uni-
h(y) dx
= g(x). (6.6.6a)
que solution to the problem, we must
specify a value y =
y 0 for a given ini- and then multiply both sides of (6.6.6a)
tial value x =
x 0 of the independent by dx:
variable, the so-called initial condition
for differential equation (6.6.4); this
J± = g( x)dx. (6.6.6b)
initial condition determines the uni-
r
que solution Now, to find the solution to Eq. (6.6.6)
we need only integrate the left- and
right-hand sides of Eq. (6.6.6b):
y = lto+ j
f{x)dx (6.6.5b)
x0 = g{x)dx {6 - 6 * 7)
\w) \
~
-

to'Eq. (6.6.4), that is, it determines the

If these integrals can be computed, then
unique integral curve y = y {x) pas-
M we have the function y y (x) in im- =
0 =
sing through the given point
M o (x 0 , y 0 ).
plicit form:

Many textbooks are devoted to the H(y) = G(x) + C ,

(6.6.8)
theory of differential equations, and we
do not intend to repeat their content here. where H {y) = j
and G (x =
We will limit our discussion to equa-
tions of type (6.6.3) that can be solved g (x) dx are definite forms of these
j
as simply as Eq. (6.6.4) for the antide- integrals, and C is an arbitrary con-
stant.
The simplest example of an equation
with variables separable is Eq. (6.6.4),
which is obtained if we put g / =
and h =
1 in Eq. (6.6.6). Substituting
this into (6.6.7), we arrive at the solu-
tion (6.6.5) to Eq. (6.6.4). An equally
simple case is the one where / (x, y) =
/ (y) in (6.6.3) depends only on y, that
is, where (6.6.3) becomes
Figure 6.6.1 v
§
= / Of), (6.6.4a)
 200 6 Series. Simple Differential Equations

here, as we see, g = 1 and h = /. For

one, the equation

y' = Ay, (6.6.9)

which we will often encounter below

(e.g. see Chapter 8), also belongs to
this type of differential equations of
variables separable. The solution (6.6.7)
to Eq. (6.6.9) assumes the form
dyly = k dx, or In y = kx -f-
C, or,
j j
finally,

y = e hx+c = Ae hx , (
6 6 10 ). .

where we have put A = ec (since C

is an arbitrary number, we conclude
that A is an arbitrary (positive) num-
ber). Thus, we see once more that
the exponential function y Aehx is =
characterized by the proportionality
(expressed by (6.6.9)) between the deriv-
ative of the function and the function
proper.
Let us examine some simple examples
of differential equations with variables
separable.
1. Suppose that

{/' = or
^= xy. (6.6.11)

dy
Separation of variables yields
y
xdx y whence y^dy^ xdx ,
or
j j
Iny — x 2 /2-\~
C, or, finally,
Figure 6.6.2

y = exp (-y- + C) =^4exp

(-y] ,

Figure 6.6.26; here again |

A |
= e c ).
(
6 6 12 )
. .
3. if

where, as before, A The graphs = ec . y' =— x

,
or
dy
—x ? (6.6.15)
y dx y
of the function (6.6.12) are depicted
in Figure 6.6.2a.
2. Suppose that then we obtain \ydy = x dx. or
j
l2:= x /2 -f-C,
2 2 or, finally,
y
y =* or
dy_

dx
^ y_
x
(6.6.13) y
2 —x = 2
a, (6.6.16)

where we have denoted 2C by a

Then y
-1
dy = x~ l dx ,
or In | y |
= (Figure 6.6.2c).
j j
ln|x| + C, or, finally, Of course, all these examples were spe-
cially selected, since here not only the given
y = Ax (6.6.14) equation (6.6.3) is that of type (6.6.6) but the
 '

6.6 First-Order Differential Equations 201

Figure 6.6.3
we are speaking, of course, not only of equa-
tions with variables separable). Suppose that,
integrals in solution (6.6.7), too, can easily we have an equation of type (6.6.3) in which
be calculated, so that the general solutions the dependence of / (x, y) on x and y is quite
(6.6.12), (6.6.14), and (6.6.16) are easily found. simple (the meaning of this rather hazy state-
But in real life the situation is not all that ment will be elaborated on below) and wish
simple. In the general case, the right-hand to find the solution y y (x) to this equation —
side of Eq. (6.6.3), of course, need not sepa- that passes through a fixed point 0 (x 0 y0 ) M ,

rate into two factors each depending on x in the ary-plane. assume that the function We
or on y; besides, if the variables do separate, y (x) in the vicinity of point x x 0 at which =
this does not mean that the integrals in (6.6.7) y 0 (# 0 ) =
is expandable in a Taylor’s series
,

can be expressed in terms of elementary func-

y (x) #o "4~ a 1 (x jtq) a 2 (x Xq) 2
tions. The only path here is to resort to the
various methods of approximate (numerical) + a3 (x- x0 )s
+ . . . (6.6.17)
solution of differential equations. 12
Then 6 -

Note that Eq. (6.6.3) specifies the depen-

dence of the slope y of the solution curve y' (x) = a1+ 2 a 2 (x —x 0)
y = y (x) for this equation on the position + 3a 3 (x -x 0)
2
+ . . . (6.6.18)
of point ( x y) in the plane; in other words,
,

Eq. (6.6.3) specifies a field of directions in the Substituting (6.6.17) and (6.6.18) into the ini-
zy-plane, that is, it specifies at each point tial equation (6.6.3), we in many cases arrive
M= M (x, y) a direction with a given slope at the condition of equality of two power se-
k = y
f
=/ (x, y) (Figure 6.6.3). This fact ries: the series (6.6.18) for y ' and the series-
suggests a natural method for an approximate expressing the function / (x, y), where y is-
solution of Eq. (6.6.3). given by the series (6.6.17) (it is at this point
Suppose that we wish to find the curve that it is desirable for the function / to be
y — f (x) that passes through a given point “simple”, that is, such that / ( x a 0 a x (x + —
M 0 (x 0 ,yo) in the xy-plane and such that the *0 ) +
« 2 ix x,) 2 —
.) can be represented + . .
,

function y (x) satisfies condition (6.6.3). Let as a power series b 0 b x (x xQ ) b 2 (x + — + —

us move from point M0 along the direction x 0 ) 2 +...). Equating the coefficients of like'
specified at this point to an adjacent point powers of x of the two series, we can then often
M i ( x i, 2/i), where x x = +
x0 Ax 0 y x
, = y0+ use the system a x fe 2a 2 bx 3a 3 b2 = = —
Ay 0 and Ay 0 /Ax 0 = =
k0 / (x Q y 0 ). Then
0
etc. to find numerically the coefficients a x
, , ,

M
, ,
,
from point 1 we move along the direction a 2 ... of the series (6.6.17), that is, find
,

specified at this point to another adjacent point the solution to Eq. (6.6.3) (since the coefficient
M 2 (* 2 2/ 2)1
>
where x 2 = +
x1 Ax x y 2
, = yx+ a 0 is known beforehand). 6 13 -

A y u and AyJAx1 x = k = f (xu

y*), and so on.
Repeating this procedure many times, we ob- 6 12
The reader will have to take it for grant-
-

tain a broken curve (known as the Euler broken ed that a power series, like the one on the right-
line; see Figure 6.6.4), which (for small “steps” hand side of (6.6.17), can be differentiated
Ax 0 Ax x etc.) yields a rough picture of the
, ,
term by term. We have not proved that this
behavior of the integral curve (i.e. solution) can be done, but the supposition seems quite*
y —y (x) to Eq. (6.6.3) (in a certain sense the natural.
6 13
Euler broken lines “converge” to the integral -
Note that expansion (6.6.17), as we
curve when the lengths of all of its “chains” know, only valid in a certain neighbourhood
is
tend to zero). Computer calculations carried of point x =
x 0 whereby it often happens- ,

out via this method prove to be simple, and that, upon reaching a certain point x x± = ,

the method on the whole is convincing and we are forced to discard expansion (6.6.17)
reliable. and look for the solution in the form of a se-
Another method of solving differential equa- ries expansion in powers of x xlt and so on. —
tions is based on the use of power series (here, The examples we consider below are selected^
just as in the case with the Euler broken lines, so that this difficulty does not arise.
:202 6 Series. Simple Differential Equations

To illustrate this method, let us take two which implies that

simple examples, the first being of a purely
illustrative value since the respective equation
can be solved without resorting to series, the
<H-1.
A _
«2—
«0
2 .
a3 — —— 1>3
1
.

second being typical.

1. Let us return to the equation considered <l4=
T = '2T’
fl/Q
a5==
CL

T"
g
= 1-3-5 1
’

above, (6.6.11), or y' xy, and write —

6
— a0
’
*“*
6 2.4*6
y = aQ + ax x + a 2x 2
+ a 3x s
+ . . .. (6.6.19)
This enables finding the general solution to
Then Eq. (6.6.22):
jj' — a1 + 2a 2x + 3 a 3x 2 + 4a 4x
3
+ . . ..
T 7*3 7*5
-
>».7 \

(
6 6 20 )
. . ( T“ T3 i~l-3-5
t
'“l-3-5-7+'")

Substituting these two series expansions into

(6.6.11), we get
+a 0 (l

where the value

+ ^-+^ +
of a 0 can he determined from
^ + . .
. )
, (6.6.23)

ax + 2 a 2x + 3 a 3x2 + 4 a±x + 5 a x z
5
4
the initial conditions', for instance, if we know
+ 6a 6 ^ 5 +...== a x + a x + a x
0 ±
2
2
z
that z/
(0) 0, =
then a 0 0 (see Exer- =
cise 6.6.4).
-j~ a3 x4 -f- a^x b
-f- a 5£ 6 -{-.... (6.6.21)

Equating the coefficients of like powers of x Exercises

,

we get
6.6.1. Solve the following differential equa-
=
•Oj=0, a = a0
. «3 = — =°. a
tions (with variables separable): (a) y'
(b) y' = k x/y
y' = cos x sin y.
( ), (c)
y
y' = k (y/x), (d) y' = ®V»
2
,

(e)

-a ^ — *
’
_
a5
_ 5-
_A
— 0,
6.6.2. Prove that the straight lines (a)
— 0 and (b) y — ±x depicted in Fig-
2-4 y
ure 6.6.2a and Figure 6.6.26 are also integral
a #0 curves of Eqs. (6.6.11) and (6.6.15), respec-
a* = -5-= a

2-4-6 tively.
6.6.3. Point out the “singular” solution to
Eq. (6.6.9) similar to the “singular” solutions
where a 0 =y =y 0 (0). In view of these equal-
of Eqs. (6.6.11) and (6.6.15) considered in
ities we can write expansion (6.6.19) in the
Exercise 6.6.2.
following form:
6.6.4. Prove that the integral curve (6.6.23)
2
of Eq. (6.6.22) that passes through the origin
x 1 x4 can be written via the following formula:
M = a0 2 'IT X

— e**/ 2 ®(x), where ®(;r)= e~ l ^ dt.

from which follows that y 1
a^e*' !^ (sub-
it = y
j
stitute x 2 /2 for x in the right-hand side of 0

(6 . 2 2 )). by the method of series ex-

6.6.5. Solve
.

2. Suppose that pansion (a) Eq. (6.6.9), (b) Eq. (6.6.13), and
equation (a) in Exercise 6.6.1.
y' = xy + 1. (6.6.22)
(c)

"This is not a differential equation with vari- 6.7* The Differential Equation for
ables separable, and it can be proved that its Water Flow from a Vessel
solution cannot be written in the form of an
explicit dependence of y on x. Nevertheless, As one more example of a problem that
it is easy to express the solution to Eq. (6.6.22)
requires solving a differential equation,
in the form of a power series.
Let us again employ expansions (6.6.19) we now consider the flow of water from
and (6.6.20) and substitute them into (6.6.22). a vessel (see Section 3.6). will We
The result is assume that the vessel has an opening
near the (see Figure 3.6.3)
bottom
a! + 2a 2 ;r + 3 a 3 x2 + 4a x + 5a x 4
3
h
4
or that a small pipe is connected to
T" 6a 6 £ 5 + . . . = + 1+ a^ (IqX
2
the bottom, in view of which the
—|- a 2 xz -f- a 3 x4 -f- a^x + b
. . .,
water can flow out of the vessel, on
 t ) t

€.7* The Differential Equation for Water Flow from a Vessel 203

the other hand, we will assume that differ mathematically from the velocity-
the vessel can also receive an inflow distance problem, which, as we know
of water from some outside source. (see Chapter 3) is solved by computing
The statement of this problem is very a (definite) integral.
simple and pictorial. At the same time, For this problem to have a definite
the mathematical methods required to solution, we must know the amount
describe the flow of water are also F 0 water in the vessel at a definite
of
employed in more complicated and initial time t 0 The condition that .

interesting problems. V= F t 0 is called the initial

0 at t =
So let us imagine a vessel with water condition which specifies a unique solu-
,

flowing in or out. We
denote the volume tion to Eq. (6.7.1) (cf. Section 6.6).
of water in the vessel by F. This volume The amount (volume) of water receiv-
varies with time, which means that F ed by, or flowing out of, the vessel during
is a function of time. What meaning ti

has the quantity dV/dt ? time t

0 to tx is q (
t) dt. Whence the
It is quite clear that dV ~
V (t -f- dt) j
to
— V ( ) is the amount of water that amount of water in the vessel at time
has entered the vessel during time dt t± is
(if dV is negative, then it is the amount n
of water that has left the vessel during q(t)dt.
V(*i)==Fo+J (6.7.2)
time dt). Therefore, dV/dt q (t =
is
to
the rate of change of the amount of water
in the vessel. The quantity q (t) has This expression holds true for any
a special name, flow rate. If q 0, > time t x and, consequently, fully deter-
mines the desired dependence of F
then water is entering the vessel;
if q < 0, water is being discharged on t v At t x £0 =
the integral in (6.7.2) ,

from the vessel and the amount (or is zero and F (£ 0 ) = F Thus, solution .
0

mass) of water in the vessel diminishes. (6.7.2) does indeed satisfy the stated
In the particular case where the flow condition relative to the amount of water
rate is constant, that is, q does not at time t 0 (the initial condition).

1) =
depend on time, we have V (t Formula (6.7.2) can also be used
F ( t ) -f q. Here q is the amount of for t x £0 <
but its meaning is different
,

water entering the vessel in unit time from that when t x t0 For t x t0 > .
>
(the quantity may be negative as well). the quantity F (£ x ) is the amount of
In general, in the course of a unit of water that will be in the vessel at time
time (which can be a second, an hour, t x if at time t
0 the amount of water was

or even a year) the value of q may F 0 and the flow rate is given by the
change, whereby we define q as a function q = q (£). For t x tQ the <
derivative. The reader that has tackled quantity F (£ x ) is the amount of water
the relationship between the average that must be in the vessel at time t x
velocity and the instantaneous velocity, so that at a later time, by time £ 0 ,

the ratio of increments and the deriva- there should be an amount of water
tive (see Chapter 2) needs no further equal to F 0 with the flow rate fixed
,

explanations. by the function q = q (t).

We know that Instead of writing t x in (6.7.2) we

can simply write £, although such no-
tation is not rigorous because t is the
£•=««>• (6 - 7 - 1)
integration variable, but this is not
If we know the dependence of the flow
really confusing. Formula (6.7.2) takes
the form
rate on time, or the function q q ( ), =
the differential equation (6.7.1) in all t

respects is similar to Eq. (6.6.4), and V(t)=V0 +jq(t)dt. (6.7.3)

the problem of finding F does not to
 ) ) t t

204 6 Series. Simple Differential Equations

Remember only that q ( ) here is not

the value of q at the upper limit of
integration but the (variable) function
of the variable of integration, which
runs through all values from t 0 to t.
Formula (6.7.3), which yields the
solution to the water-discharge problem
if the flow rate q (£) is given and so is

the amount V 0 ~
V ( 0 ) at the initial
time t t0 =
can also be obtained by
,

somewhat different reasoning. From

(6.7.1), by virtue of the definition of
the indefinite integral, it follows that

V(t)= [q(t)dt. Figure 6.7.1

Suppose that the indefinite integral of water level z (Figure 6.7.1). For
of the has been found
function q (
t
example, when water is flowing out
in some way. Denote it by / (£). Then of a long thin pipe, we can assume
that
q(t)dt =I (t) + C,
j
q = — kz ,
(6.7.5)
where C is the constant of integration.
where coefficient & is a positive constant,
From this,
the minus sign meaning that the water
V (t) = I (t) + C. (6.7.4) is being discharged When the water.

flows out through an opening in a thin

To determine the constant of integra- wall, the law is quite different:
tion, us take advantage of the
let
initial condition, that is, let us require g= — aY z (6.7.5a)
that at t =
t 0 we have V V 0 Sub- = .

stituting t0 for t in (6.7.4), we get

(this law was established by a pupil
V0 = I (to) + C, whence C= V0 —I (t 0 ).
of Galileo Galilei, one Evangelista
Torricelli (1608-1647), an Italian phys-
Substituting this value of C into (6.7.4)
icist and mathematician). Also possible
yields
is a combination of a constant influx
V (t) = V +I 0 (t -/ (to), of water from above (q 0 ) and a discharge

which coincides with since

of water by law (6.7.5) or (6.7.5a):
(6.7.3),
t
q = g — kz
0 or q=- q 0 — Y cl z. (6.7.6)
q(t) dt =I (t) | i = I(t)-I(t0)
j In each of these cases, until the
u
problem has been solved, we do not
Formula (6.7.4) may be termed the know the dependence of water level on
general solution of Eq. (6.7.1). By time, z — z (t), and hence we do not
choosing one or another value of C, know the flow rate. For this reason,
we can, from formula (6.7.4), obtain the problem of determining V from the
a variety of particular solutions that equation
correspond to different initial condi-
tions. #-«<•> < 6 7 - 7)
-

Ordinarily, however, the flow rate

is not known as a function of time. cannot be reduced to the preceding
More often we know a physical law problem.
describing the flow rate as a function Here we stated the problem in the
of the water head, that is, of the height general case for an arbitrary relation of
6.7* The Differential Equation for Water Flow from a Vessel 205

the flow rate q = q (z) as a function Let us rewrite Eq. (6.7.8) in the
of the level z. Equation (6.7.7) involves form dz/dt = q(z)/S(z) and put
two unknown quantities: the amount S(z)/q (z) = /(z). Then we have
{volume) of water V and the water dz _ 1
(6.7.9)
level z. It is clear that these quantities
f (z)
•

are not independent. A definite water

level is associated with a very definite
We have already dealt with such equa-
tions (see Eq. (6.6.4a)). Let us rewrite
amount of water so that V is a function
— Eq. (6.7.9) as
of 2 or V ,
V (z).
It is clear that the form of this dt
(6.7.10)
function is determined by the shape dz
of the vessel. For a cylindrical vessel, This notation fits the fact that we
for example, of radius r 0 we have temporarily regard t as a function of z,
V (z) Jtr^z, =
where z is the water that is, we seek not the law z z (t) =
level reckoned from the base of the describing the change of water level
vessel. For a conical vessel depicted in with time but the inverse function
Figure 6.7.1, the formula for the volume t =
t (z), and after finding it we will
of a cone yields V (z) (1/3) S (z) z, = express z in terms of t.
where S (z) =
n [r (z)] 2 is the area of Let us multiply the left- and right-
a cross section of the vessel at level z, hand sides of (6.7.10) by dz and in-
and r (z) is the radius of the cross tegrate:
section at level z. From the similarity
z 7
of triangles, we find that r (z) r 0 zlh, = t

where r 0 is the radius of the upper base f(z)dz. (6.7.11)

J J
of the cone and h is the total height, to z0

so that V (z)
2
=
(1/3) ( nr 0 /h?) z
3
. From this it follows that
Substituting the function V V (z) =
into Eq. (6.7.7) yields t = t 0 -j- f(z)dz. (6.7.12)
j
dV (z) dV {z) dz z0

dt dz dt
= q(z). We have the solution to the problem:
on the right is a function of z, on the
The derivative dV/dz of the volume
left is the time t so that t is expressed
with respect to the height of water ,

as a function of variable z (the reader

level, z, is equal to the cross-sectional
will recall that t 0 is simply a number).
area S (z) at height z (see formula
Solution (6.7.12) also enables us, for
(3.6.20)), or dV (z)ldz = S (z). The final
each value of t to find the correspond-
equation is
,

ing value of z and satisfies the initial

S(z)^ = q(z). condition: z = z0 at t = t

0 (at the
(6.7.8)
initial time t = t0 the level of the
water in the vessel, z 0 ,
is given).
Clearly, this is a differential equation Here are two concrete examples of
with variables separable. Let us con- the functions / (z).
sider the solution of this equation for 1. Water flowing out of a cylindrical
different functions S S (z) and q = = vessel of radius r 0 through a thin pipe .

q (z).«.“
Here S (z) constant =
and q = =
6 14
— kz, so that Eq. (6.7.8) assumes the
Of course, other variants of water dis-
*

charge from a vessel are also possible (say,

form
a vessel with a shape that does not allow us
to write out the functions S S (z) explicit- = n <17=- kz ’ <
6 - ' 13)
ly) when an exact solution of the respective
equation is impossible and one is forced to or
resort to approximate methods (see also the
text at the end of Section 6.6 in small print).
-n rl — = kdt.
u z
 1 ,

206 6 Series. Simple Differential Equations

Integrating the left-hand side from It is quite clear that if the process
z 0 to z and the right-hand side from t 0 starts at the water level z being equal
to t, we get to z st , then z will not change: the-
— nr\ (In z — In z = — nr 0)
2
0
In —
zo
discharge water from the vessel
is completely compensated for by the-
of

influx of water, and the water level

- nr\ In JjL = k(t-t 0). (6.7. 14) remains constant. Now suppose that
the initial water level z 0 is higher
In this example it is easy to express z than z st so that the discharge (pro-
,

in terms of t. Indeed, portional to z) exceeds the influx and

the level z diminishes. We will look for
In z = In z — — 0 (t — 1 0 ), the difference z x z z st between z — —
and the steady-state level z st This
—
.

i.e. * =z 0 exp[--j^-(t f0
)]
. difference, obviously, satisfies the equa-
tion obtained by substituting z z t -f- =
We consider two instants of time, t and z st and q 0 kz st = :

t -f- AZ, and find the ratio z (t -[- At)/

z(t): .2
d (
zl + z St)
kzst —k (z t -f- z sl )
0
dt
z ( t -f- Af)
dz ±
i.e. nr:
dt
— kz i
.

(* + Af — — + £0 * *<>)]
But this equation differs from (6.7.13)
only in notation, so that its solution
nrl 1
has the form
We see that this ratio depends only on
At and is independent of t. Therefore, zi =z i
0>
exp
[
— (t — 0)
]
,
the water level z falls in equal ratio
during equal intervals of time; it con-
stantly diminishes exponentially, but where z[
0)
= z0 —z st . The final result
will never reach the value z 0, that = is

is, the water will never flow out of the

z = Zst + (z — Zst) eX P — 0 -(t — o] •»

vessel completely. 6 15 [
The problem changes little if there here z tends to the steady-state level
is a constant influx of water into the z s t of the water but never reaches it.
616
vessel, or q (z) == q 0 kz. —
Equation 2. Water flowing out of a conical
(6.7.13) then assumes the form vessel (altitude h and radius of base r 0 ;
see Figure 6.7.1) through a thin pipe .

(6.7.15) Here S (z) = jtr^ (zlh)

2
and q = —kz.
In this case we have the equation
Here it proves convenient to start by
finding the so-called steady -state mode ,
dz _ q (z) kzh 2 kh 2 1

S 2
that is, we look for a solution of this dt z
( ) nrfe nrl z ’

equation in the form z z st con- = = (6.7.16)

stant. Since z z st and dz/dt —
0, we get = which can be rewritten thus:
kz s i = 0, z si = qjk .

6 15
This paradox (the water will never
-

flow out of the vessel) stems from the fact that

8 16
Eq. (6.7.13) gives a relatively accurate picture ’
The
last statement remains valid for
of the process only when z p, where p is the > zo < see Exercise 6.7.3), which indicates
z st (

radius of the pipe through which the discharge that the steady state z z st of the water in =
occurs. the vessel is stable (see also Section 9.3).
 1 ? . .

207-
6.7 * The Differential Equation for Water Flow from a Vessel

* * *
Integration yields
While in Chapters 2 and 3 we dis-
nrl cussed the ideas underlying differential
kh 2 J«*— SMt-4) and integral calculus, Chapters 4 to 6
were devoted to the tools of higher
= — t t0 . mathematics, and each person who
wants to master mathematical analysis
Here we can easily express 2 in terms
to such an extent so as to have the pos-
of t :

sibility of employing the theorems and

2
Z Zn
^ 0)
concepts in practical applications must
2 become familiar with these tools. Of
or course, the main problem that con-
fronts a scientist or engineer is the
2 kh 2
==/ (6.7.17) problem of adequately describing a
problem interest in mathematical
of
This formula completely solves the terms often in the form of a
(most
problem. It is readily verified that differential equation); the analysis of

dz _
~~
kh 2 r
Jir§ L
Z
2
°
W
ftrg
,

°
1 - 1/2
J
the model that follows is purely mathe-
matical (solution of a differential equa-
tion) and can be entrusted to a mathe-
kh 2 matician or passed over to a programmer
for processing on a computer. (However,
the formulation of the model cannot be*
so that 2 does indeed satisfy the equa-
= reassigned to anyone; it must be carried
tion. It is also clear that 2 z 0 at
= out by a specialist who has a deep know-
t t0 . The expression (6.7.17) permits
ledge of the phenomenon in question.)'
finding the time when the vessel is
completely emptied:
On the basis of this the engineer may
decide there is no need to study the*
z =0 when t = + jj£- t0 (6.7.18) mathematical tools, but he, or she,
will be deeply mistaken. Understanding
an interesting qualitative
There is the essence of higher mathematics is
difference between these two examples. closely connected with mastering cer-
In the second case, the water completely tain mathematical tools: without these*
flows out of the vessel during a finite tools there can be no real understanding
time interval T (
nr20 /kh2 ) z\!2 (see = of the principles. To take an example

(6.7.18), while in the first case the water from another field, reading special
level 2 approaches the zero level z 0 = literature in a foreign language, even,
asymptotically (it never reaches this with the help of a dictionary (which
level; in otherwords, an infinitely long plays the role of a programmer, so to*
interval of time is needed for the water say), is impossible without knowing-
to flow out completely). the basic facts of the grammar of the
language (the “theory” of the language)
and without having a certain minimal
Exercises vocabulary (the basic of “engineering”
skills). Without knowing the grammar
6.7.1. Consider the solution to Eq. (6.7.15) of the language we are helpless, since*
that satisfies the initial condition z 0 z s t. < we do not know what words (and in
How will the water flow out of the vessel in
this case? what form) we must look up in the
6.7.2. Consider the case where the water dictionary, while not possessing even
flows out of a conical vessel through a thin
a minimal vocabulary means that the
pipe under the condition of a constant influx
of water. [Hint. The differential equation of
dictionary is useless, since the totally
unfamiliar text is a stumbling block.
this process is tl = q0 — kz.\ However, even a small vocabulary of
208 6 Series. Simple Differential Equations

words and expressions and a rudimen- properties of integrals (say, a fluent

tary knowledge of the grammar make use of the integration-by-substitution
possible the full employment of a dic- method) developed in the course of
tionary. In the same way, a physicist calculating integrals are necessary to
or engineer often (but not always) needs every student of science and engineer-
only a basic knowledge of higher math- ing, since most often the integral that
ematics and a smattering of technical you will encounter will not exactly
skills. coincide with an integral in a table
Chapters 4 and 5 give an overall picture but can almost always be reduced to
{rather sketchy, perforce) of differen- such an integral via simple transform-
tiation (finding derivatives) and integr- ations, including the change of vari-
ation techniques. Finding derivatives ables.
is quite simple, and therefore it is Chapter 6 begins with the topic of
desirable to develop this skill to a state seriesThe technique by which com-
.

when finding derivatives of simple func- plicated functions (for instance, ob-
tions is carried out automatically, so tained through experiments) can be
to say. Here there is no need to solve reduced to simple functions is one of
as many complicated problems as one the basic contributions of higher math-
can put his, or her, hands on, since a ematics to engineering practice every —
physicist or engineer seldom encounters physicist and engineer must be ac-
complicated functions. It is much more quainted with it. The key to the tech-
important to know how to solve simple nique lies in the theory of expansion of
problems at first glance, so we recom- functions into power series (Sections 6.1
mend solving all the exercises on cal- to 6.4); in other cases the expansion of
culating derivatives collected in this functions into trigonometric series, that
book and not calculating derivatives is, the representation of functions as
of more complicated functions. If solv- linear combinations of trigonometric
ing problems from this book proves to functions (see Section 10.9), constitutes
be insufficient for developing the neces- a good addition to the technique.
siir^ skills, raadar* cam fisumdman^
problem book on differential calculus our discussion to elements of the cor-
or even compile a list of arbitrary responding theory, which is now under-
functions and try to find their deriv- going rapid growth, for which there
atives. There is no need in trying to are several reasons; among them are
memorize the entire table of derivatives the rapid development of the sections
{see Appendix 1); it is much more ad- of pure mathematics that are widely
visable at the first stages in the reader’s used in the theory of function approxi-
studies to copy the formulas onto a mation (functional analysis, the theory
card and use the compiled table when of functions of a complex variable,
solving problems. The formulas that and other fields), practical reasons con-
are used more often will be memorized, nected with the constant need to sim-
and finally there will be no need of the plify complicated functions, and the
card. invention of the electronic computer,
Integration techniques developed in which opens up new possibilities in
Chapter 5 are much more complicated. simplifying functions and at the same
We believe there is no need to devote time poses requirements on the form
too much time to these techniques. Of of the functions used in computer cal-
course, it is useful to know how to culations.
evaluate simple integrals, but it is The second topic discussed in Chap-
much more important to know how to ter 6 is differential equations which ,

use VcWres 'ntt&gxuks . QA course, VW£&CfCittfSb Vlife Wi tfi

skills in integrating the simplest func- and engineers. Indeed, a mathematical

tions and the knowledge of the basic analysis of natural phenomena usually
 209
6 Series. Simple Differential Equations
• • •

starts with attempts to represent the tation, z = up, then z = uv + uv where

laws of nature as differential equations
— u, v,
if

and z are fluents and

•

u, v,
•

and
•

z
%

are flux-
these connect the various (variable) ions. The second problem a problem of the
is

quantities that describe the phenome- solution of differential equations.

Newton also devoted much attention to
non of interest (such representation is infinite series (discovery of the binomial se-
usually “abstract”, that is, it gives ries made a great impression on him
6 17
). The *

only an approximate picture of the whole of Newton’s theory of fluxions and flu-
real phenomenon). In Section 6.7 we ents (derivatives and integrals) was obviously
the fruit of his investigations into the theory
studied this general scheme using the
of infinite series. Nor is it accidental that the
example of water flowing out of a ves- basic theorems (see formulas (6.1.18) and
sel. Note that the content of Section 6.7 (6.1.19)) of the theory of series are linked with
and the examples of differential equa- the names of Newton’s pupils and associates:
Brook Taylor (1685-1731), who was Secretary
tions discussed in Section 6.6 are of an
to the Royal Society of London (the British
illustrative value only; they can be academy of sciences) when Newton was its
freely replaced with any other examples President, and Professor Colin Mafllaurin
and there is no need to memorize (1698-1746) of Edinburgh University, who
knew Newton personally and greatly admired
them (although the concept of a first- him. Incidentally, both formulas, (6.1.18)
order differential equation with vari- and (6.1.19), were first recorded by Taylor.
ables separable is so important that Maclaurin’s contribution was that he posed
it is advisable to master it). In this the question of the sphere of applicability of
the formulas and partially answered it. The
respect it is appropriate to note the
exact formula (6.1.16), the one containing a
difference between Chapter 4 and the finite number of terms, was established by
last sections of Chapter 6: while the Lagrange (we will return to him again later
reader must carefully read Chapter 4, on).
Leibniz, too, made important contribu-
Section 6.6 and especially Section 6.7
tions to the theory of series and to differential
can just be browsed through, so as to equations in the solution of various geometri-
understand the principal points. cal problems. Note that Leibniz was more in-
terested in geometry than in mechanics: he
associated the concept of the derivative with
Higher mathematics was created by New- a tangent and not speed. Significantly, his
ton and Leibniz in the 17th century, and even
fundamental memoir on the “new mathe-
in their works it was a highly developed dis-
matics” had an involved title: “A New Method
of Maxima and Minima, and Also of Tangents,
cipline. Of course, its development did not
for Which Method Neither Fractions Nor Ir-
stop there. Newton and Leibniz did not con-
rational Quantities are an Obstacle, and a
fine themselves to the basic definitions and
initial theorems. Their contribution was much
Special Kind of Calculus for This”.
greater. Newton clearly realized the impor-
The excellent school of Leibniz, headed by
tance of differential equations in analyzing the two brothers, Jakob Bernoulli (1654-1705)
natural phenomena. Take his famous “Prin- and Johann Bernoulli (1667-1748), made an
cipia” (1687), which contains Newtonian me-
outstanding contribution to the differential
chanics (the Newtonian world system), and and integral calculus. They were the founders
of the famous Swiss family of mathemati-
you will see it starts with the differential
equation of motion (see Eq. (9.4.2)). This equa- cians, which in the 17th and 18th centuries
tion is taken as an axiom, whereas all sub- produced eight first-class mathematicians.
sequent propositions of mechanics are theo- (The most prominent of the later members of
rems derived from this axiom (and also from the family was Johann’s son Daniel Bernoulli
the law of gravity that follows from experi- (1700-1782), who worked in Basel and St.
mental findings (Kepler’s laws) and axiom Petersburg and was one of the founders of hy-
(9.4.2)). In his mathematical investigations
drodynamics and the kinetic theory of gases.
Newton formulated “the main problems of We will have occasion again to discuss his
mathematical analysis” as follows: work.)
(1) from a given relationship between flu-
ents (initial functions) to determine the re-
lation between fluxions (derivatives); 6 * 17
For an account of how this outstand-
(2) from a given equation containing fluxions ing discovery was evidently made, see, for
to find the relationship between fluents. example, G. Polya, Mathematical Discovery
The first of these problems is obviously a On Understanding Learning and Teaching
(i
,

roblem of the differentiation of known com- Problem Solving ), Vol. 1, Wiley, New York,
ma tions of functions: thus, in Newton’s no- 1962, pp. 91-93.

14-0946
 210 6 Series. Simple Differential Equations

Jakob Bernoulli
Johann Bernoulli

Jakob Bernoulli received a theological received its present form, basic symbolism,
education, but his interest in mathematics and terminology. For example, originally
prevailed. He studied the mathematical lit- Leibniz had been inclined to speak of “differ-
erature by himself and came upon the works ential calculus” and “summational calculus”
n
of Leibniz, which made a profound impression as the two branches ofthe “new mathematics,
on him. In fact, they s© overwhelmed him but on a suggestion from Johann Bernoulli
that he gave up the secure life of a pastor and he finally chose the Latinized term “integral
for a number of years held the little-respected calculus” (instead of “summational”). Jakob
and low-paid post of a home tutor. It was only and Johann Bernoulli greatly advanced the
thanks to the good offices of Leibniz that in new calculus, obtaining, in particular, im-
1683 he was appointed a professor (at first, portant results in the theory of differential equa-
of physics, and later of mathematics, too) tions (see Section 6.6) and laying the founda-
at the University of Basel. tions of what is called calculus of variations ,

Johann Bernoulli’s father wanted him to which mentioned in passing in Section 7.2
is
go into commerce, hut thanks to his elder (see also the text above Example 5 in Sec-
brother, who had taught him mathematics and tion 7.1). 6 18 *

physics, his clearly expressed scientific in- The first printed textbook of differential
terests ran counter to his father’s wishes. Since and integral calculus, entitled “Infinitesimal
Switzerland in those years provided very Analysis for the Study of Curves” (note how
little opportunity for the study of mathe- it echoes the famous memoirs of Leibniz) ap-
matics, Johann Bernoulli was compelled to peared in 1696. Its author was Guillaume
obtain a medical education and for many years F. A. L'Hospital (1661-1704), a French noble-
earned his living as a physician. His doctoral man (Marquis de St. Mesme) who was a pupil
dissertation on the movement of muscles was of Leibniz and Johann Bernoulli. This excel-
interesting in that it was evidently the first lent textbook went through many editions
attempt to apply the methods of higher math- and was translated into many languages. The
ematics to physiology. On the recommenda- ideas L’Hospital set forth closely follow the
tion of the famous Christian Huygens, Johann lectures which he heard Johann Bernoulli
was appointed to a professorship in physics at deliver in Paris; they also follow Bernoulli’s
Groningen (the Netherlands) in 1695. Return- manuscript manual “Lectures on Differential
ing to Basel in 1705, he at first could obtain Calculus,” which was discovered in the li-
a university post only as a professor of Greek. brary of the University of Basel in the 20th
It was only after the death of his brother Ja- century and was published for the first time
kob that Johann Bernoulli was appointed pro- in 1922. Thus, while Bernoulli’s text remained
fessor of mathematics at the University of unknown, the lectures which were based
Basel and held this post to the end of his life. on it and were attended by only one student.
The Bernoulli brothers maintained a live-
ly correspondence with Leibniz, who time 6* 18
Variational calculus became an inde-
and again expressed his admiration for their pendent scientific discipline in the 18th cen-
successes. It was precisely in the course of this tury thanks to the works of Euler and La-
correspondence that mathematical analysis grange.
 6 Series. Simple Differential Equations 211

Jean Le Rond d’Alembert

Leonard Euler

lin. 6 19
Here he became head of the physics
L’Hospital, exerted a tremendous influence on and mathematics department of the Berlin
the entire subsequent development of higher (Prussian) Academy of Sciences, but he kept
mathematics (in particular, it spread the sym- up his close contacts with the St. Petersburg
bols and terminology of Leibniz). Among other Academy. Euler was the most prominent and
things, L’ Hospital’s textbook saw the first most productive mathematician of the 18th
publication of the method of calculating limits century. His work dealt with literally every
which Johann Bernoulli taught his pupil and sphere of mathematics and mathematical phys-
which is now with so little justification simply ics. We write about some of Euler’s findings
called “L’Hospital’s rule” (see Section 6.5). below (see Section 10.8 and Chapters 14 and
Another of Johann Bernoulli’s pupils was 15). Euler’s multivolume textbook of differ-
the Swiss mathematician Leonhard Euler ential and integral calculus was translated
(1707-1783). He was introduced to Bernoulli into other languages.
by his father, Pastor Paul Euler, who had Among Euler’s contemporaries was Jean
once studied mathematics under Jakob Ber- Le Rond d' Alembert (1717-1783), a distin-
noulli. Pastor Euler wanted his son to be a guished French mathematician and author of
pastor too, hut Johann Bernoulli convinced treatises on mechanics. His exceptional scien-
Paul Euler that the boy had outstanding tific versatility is particularly evident in his
mathematical ability. On Johann Bernoulli’s collaboration with Denis Diderot (1713-1784).
recommendation Leonhard Euler went to They produced the famous “Encyclopedic, ou
St. Petersburg in 1727, where an Academy of Dictionaire Raisonne des Sciences, des Arts
Sciences had recently been founded and where et des Metiers,” in which he wrote nearly all
two of his teacher’s sons (one of them was Da- the articles dealing with the natural sciences.
niel Bernoulli) were working. It was orig- D’Alembert was named after the church (Jean
inally planned that Euler would take the Le Rond; literally, Jean the Round) on the
vacant post of Professor of Physiology at the steps of which he was found as an abandoned
St. Petersburg Academy, and he made a thor- infant. Notwithstanding this handicap he be-
ough study of this science in order to follow came a prominent man of science. In Section
his teacher’s example and apply mathematical 10.8 we describe a most edifying and fruitful
methods in it. But when he arrived in St. scientific debate in which Leonhard Euler,
Petersburg, he found that the post of professor d’Alembert, and Daniel Bernoulli took part.
of mathematics was also vacant. So he gave Both Euler and d’Alembert strove for concrete
up physiology and devoted himself to mathe-
6 19
matics, physics, mechanics, and astronomy ‘
This refers to the “Biron period” in
(for one, the movements of the moon). Russian history, named after E. J. Biron, a
Leonhard Euler spent a large part of his disreputable favorite of Empress Anne (1693-
life in St. Petersburg. There was a break of 1740), whom she brought from Courland. The
25 years when, on invitation from King reign of Empress Anne, from 1730 to 1740,
Frederick II of Prussia, he moved from St. was marked by arbitrary arrests and execu-
Petersburg, in those years a place not very con- tions of innocent people and total disorgani-
ducive to calm scientific research, to Ber- zation of the machinery of government.

14 *
212 6 Series Simple Differential Equations
.

Joseph Louis Lagrange Augustin Louis Cauchy

results, displaying outstanding intuition in “Analytical Mechanics” (1788), which greatly

mathematics and physics and a lack of inter- furthered mathematics and physics. Lagrange
est in scholastic discussions of mathematical published an excellent, and in many respects
ideas. This is illustrated, for example, by revolutionary, textbook of mathematical
d’Alembert’s well-known call to the young: analysis (“The Theory of Analytic Functions,”
“Keep on working— complete understanding 1797) based on lectures that he delivered at
will come in time.” the Polytechnical School. He is also known for
When Euler
decided to return to St. his fundamental studies in algebra. In the
Petersburg after having headed the physico- breadth and range of his mathematical inter-
mathematical department of the Berlin Acad- ests and in the brilliance of his achievements,
emy from 1741 to 1766, the question of his Lagrange was second perhaps only to Euler
successor arose. Euler recommended the young among the mathematicians of the 18th cen-
mathematician Joseph Louis Lagrange (1736- tury.
1813), who was born into a French family that One of the pupils and ardent admirers of
had moved to Italy. Lagrange was only 30 Lagrange was Jean Baptiste Fourier (1768-
at the time. His candidature was enthusiasti- 1830), a professor at the Polytechnical School
cally supported by d’Alembert, who corre- whose name is linked with outstanding achieve-
sponded with King Frederick II. Lagrange was ments in the field of partial differential equa-
already a well-known scientist. He had begun tions (also called equations of mathematical
to teach in the Turin Artillery School in 1726, physics see Section 10.8), for one, in the theory
;

when he was 20, and the following year he was of the propagation of heat and in the theory of
one of the founding members of the Turin trigonometric series (see Section 10.9).
Academy of Sciences, in whose proceedings Another professor of the Polytechnical
he published many papers. In 1787, after the School was the famous Augustin Louis Cauchy
death of Frederick II, Lagrange moved to (1789-1857), who created the modern theory
France, where he played an outstanding role of limits (incidentally, his main ideas were
in the rise of the Parisian Polytechnical School borrowed from d’Alembert), which helped to
(Ecole Poly technique), an institution of high- clarify all the concepts of higher mathematics.
er learning of a new type, which trained re- Cauchy is rightly regarded as the father of the
search engineers. 6 20 It was during the Parisian
»
theory of functions of a complex variable (see
period that Lagrange compiled his two-volume Chapters 14, 15, and 17). He shares this honor,
incidentally, with Georg Friedrich Bernhard
6 * 20 The example
of the famous Polytech- Riemann (1826-1866), the German mathema-
nical School of Gaspard Monge (1746-1818) tician and one of the most prominent scientists
and Lagrange was undoubtedly taken into of the 19th century, who made outstanding
account by the founders of the Moscow Physico- contributions in literally every field of mathe-
Technical Institute in the Soviet Union and matics and mathematical physics.
the founders of the Massachusetts and Cali- While Cauchy was a representative of the
fornia institutes of technology in the United Polytechnical School of Paris, Riemann was
States. connected with another educational and scien-
6 Series. Simple Differential Equations 213

19th and 20th centuries— the University of

Gottingen in Germany. Here he attended lec-
tures by the famous mathematician, physicist,
astronomer, and geodesist Karl Friedrich Gauss
(1777-1855), who is usually acknowledged to
be the first mathematician of the 19th century.
At the University of Gottingen Riemann pre-
sented the first-ever course in the theory of
functions of a complex variable, as well as
a course in the theory of partial differential
equations. His lectures in this second course,
ublished after his death, were the first text-
E ook on the equations of mathematical phys-
ics.
David Hilbert (1862-1943) belongs to an
altogether different generation of Gottingen
scientists. He is regarded as the father of func-
tional analysis,
which deals with the functions
of functions, or functionals (see the text pre-
ceding Example 5 in Section 7.1). Today this
branch of mathematics includes generalized
functions (or distributions), such as Paul
Dirac’s delta function (see Chapters 16 and 17).
Today the ideas of mathematical analysis
Georg Riemann continue to undergo intensive development, and
computers are giving them a new and powerful
tific center that played a big role in the devel- impetus.
opment of mathematics and physics in the
 ) ) ,

Chapter 7 Investigation of Functions.

Some Problems from Geometry

In Chapters 1 to 6 we introduced the main con-

cepts of higher mathematics, the concepts of
say, f(x 0 ) =0
=x
and f"{x 0 ) 0, then <
at point x 0 the function / (#) has
a derivative and an integral, and suggested some
techniques for using them. The present chap- a maximum.
ter, which concludes Part 1, partially devel- The same conclusions can be drawn
ops the material of the previous chapters. from the considerations involving the
For instance, here we will dwell in detail on
idea of convexity of the curve repre-
the methods used to find the maximum and
minimum points of a function, of which we senting a function; we touched on this
spoke briefly in Section 2.6. We will also tell concept in Section 2.7 and will return
about the basic geometrical applications of to it Section 7.4. Indeed, the con-
in
differential and integral calculus. Some themes
in the chapter can be discussed in extracurric-
dition f(x 0 ) =
0 means that the tan-
ular studies and are written for a reader with gent to the graph of the function at
an inquiring mind (e.g. see Sections 7.4, point x =
x 0 is horizontal. From the
7.6, and 7.7).
Since this chapter was written as an addi-
fact that f"(x 0 ) 0 it follows that<
tion to Chapters 1 to 6, it is naturally of a
point x =
x 0 is a point of convexity ,
patchy nature. Loosely, the material can be that is, the curve representing y / (x) =
grouped as follows: Sections 7.1 to 7.3, Sec- close to x =
x 0 lies under the tangent
tion 7.5, Sections 7.9 to 7.11, Section 7.12, (see Figure 2.7.1). But these two facts
Sections 7.4 and 7.8, Section 7.6, and Section 7.7.
simply mean that the function / (#)
has a maximum at point# x 0 Similar = .

7.1 Smooth Maxima and Minima reasoning shows that if f (x ± ) = 0 and

The problem of finding the value of x0

f" i x i) >
0, at point x x x the func- =
tion / (x) has a minimum.
for which a given function y / (#) = These conclusions are also obtained
attains a maximum
or minimum is not by considering Taylor's series
solvable in general form by the tools
of elementary algebra.
f{x)=f (x 0) + /' (x 0 ) (x —x 0)

In Chapter 2 we established that at

points where a function has a maximum + 4-/"(*o)(*-*o) + 2
... (7.1.1)

or a minimum the derivative is equal to for the function / (#) at point x x0 = .

zero It was also shown there how,

0, then for x close to x 0
.
If f'(x 0 )
using the derivative y' to establish ,
the quantities (x # 0 ) 2 (x —
# 0 ) 3 , etc. ,
—
exactly what the function has at the may be neglected when compared with
given point x 0 (where /' (# 0 ) 0), a = x —
x 0 We therefore arrive at the
.

maximum, a minimum, or an inflection approximate expression

(neither a maximum nor a minimum).
To do we were forced to compute
this f{x)—f + /' (*o) x ~ x
(*o) ( 0)

f (x) — f(x ~ (x — x
the values of y' for values of x close to or (x /'
0) 0) 0).
.x 0 on the right and on the left of x 0 .

For instance, we found that the con- From this we see that if, say, f(x 0 ) >0,
ditions y' x for (, 0 >0
y 0 x<^ ,
= '
then f(x) — f(x 0 ) >0 (i.e. / (#) >
at x = x 0 and y (x) , 0 <0 for x > x / (# 0 )) for x >x 0 and / (x) — f(x < 0 0)
imply that at point x 0 the function (i.e. /(#)</ x x0 This
(# 0 )) for < .

y = y x attains a maximum. In Chap-

(.
means that x 0 the function
at x =
ter 2 we
investigated another
also has neither a maximum nor a minimum.
method finding the maxima or
for Similarly, there is no maximum or
minima of a function, a method that minimum when /' (# 0 ) <! 0, only in
invokes the second derivative y" (only the first case the function is increasing
its value at point x =
x 0 was required, at point # 0 while in the second it is
,

however), namely, we found that, if, decreasing.

7:1 Smooth Maxima and Minima 215

But if f{x o) = 0, then the term first nonzero derivative

is of odd order
with (x —
x 0 ) in (7.1.1) vanishes and (first,third, fifth, etc.), then the func-
we cannot ignore the term containing tion has neither a maximum nor a
(x —# 0 ) 2 Ignoring terms with (x
. ;r 0
)
3 — ,
minimum at this point, while if the
(x — 4
x 0 ) etc., small as compared with
,
first nonzero derivative is of even order
the term with (x —
x 0 ) 2 we get, from , (second, fourth, etc.), the function has
(7.1.1), either a maximum or a minimum depend-
ing on the sign of that derivative (a
/ (a:) ~ / (x + jp f" (x
g) 0) (x—x0 )
2
(7.1.2) minimum if the derivative is positive
and a maximum if it is negative). Here,
(the approximate formula (7.1.2) is as everywhere in this section, we con-
the more exact the smaller |
x — x0 |
sider only the maxima and minima that
is). From this we see that if f"(x 0 )>0, a function attains at points where it is
then f (x) > f (x 0 ), irrespective of represented by a smooth curve , that is,
whether x <C. x0 or x x0 >
that is, ,
at points where the function has a deriv-
/ (x 0 ) is less than any adjacent value ative (or even several of them, y'(x 0 ),
of / {x) and therefore / (a; 0 ) is a minimum y"(x 0 ), etc.). Such maxima and minima
of the f" (£ 0 ) <C 0, then
function. If are sometimes called smooth other cases ;

/ (x) <
/ (x 0 ) and / (x 0 ) is a maximum related to maxima and minima that are
of the function. not smooth will be considered in
It may, however, happen that both Section 7.2.
f{x 0 ) =0 and f"{x 0 ) 0. We then = Let us consider some examples.
have to take the next terms in Taylor’s Example 1 Suppose that .

series (7.1.1). If f"'{x 0 ) 0, then we

x 0) s — (a) y = A + B (x — a) 3
,
cannot neglect the term with (x
but we can freely ignore the terms with
,
(b) y = A + B {x - a) 4
, (7.1.3)

(x —
x o y, (x —
£ 0 ) 5 etc., which are ,
where B =7^= 0. Find the maxima and
small if compared with the term with minima of y .

(x —
£ 0 ) 3 Then we have
.
The derivatives of (7.1.3) are

—x (a) y' =3 B — a) y" = 6B (x — a),

(x
2

/ (x) ~ / (x + J f” (x
()) 0) (x 0)
3
, (7. 1.2a)
with z/'(a)= y "{a)=
,

0, and y"'=6B^=0.

and we have neither maximum

.(b) y = 4 B (x — a) y"= 12 B (x — 3
,
a nor
a)
2
y = 24 B (x — a), with y' (a) =
a minimum, just as we had neither
,

y "(a) = y"'(a)= 0, and y IV = 24B ^ 0.

a maximum nor a minimum at x = 0
for the function y = x (see Figure 1.5.1;
3
In the Case (a) the first nonzero deriv-
for this function y' (0) y" (0) 0). = = ative is the derivative of the third {odd)
But if /'(s 0 ) /"(s 0 ) -
/"'(s 0 ) 0 = - order, whereby here at point x a the =
and / IV (z 0 ) z 0, then in the vicinity
¥ function has neither a maximum nor
of point x 0 we have a minimum (the function has in general
no maxima or minima since y'(x) 0 =
H*) — 1 (*o> + ^4 /
IV
W { x — ^o)
4
-
only at x
point at x
=
a), but it has an inflection
=
a. In the case (b) the
(7.1.2b)
only nonzero derivative at point x a =
Here the sign of / (x) / (x 0 ) is the — is that of fourth {even) order and the
same for x >
x 0 and for x <. x 0 it is ;
function has a minimum for B 0 >
determined by the sign of / 1V (s 0 ). If
IV
(y
IV (a)
0) and a >
maximum for B
lY <T 0), which also follows from the
0 <
f (zo) is positive, we have a minimum {y
(just as in the case of the function formula for the function. (Why?)
y =
x 4
see Figure 1.5.1), and if / IV (x 0 )
;
Example 2. It is required to construct
is negative, we have a maximum . an open-at-the-top box of maximum
The attentive reader has probably volume using a square sheet of tin
already guessed that if for x x 0 the = with side 2 a by cutting out equal
 216 7 Investigation of Functions. Some Problems from Geometry

we have V"(al 3) == —8a <C 0. Conse-

quently, the function V (x) has a max-
imum at x = a/3.
To summarize, the maximum value
is obtained for x =
a/3, that is, we
have to cut out squares whose sides are
1/6 the side of the original square.
Let us compute V (x) for several x
close to a/3 and tabulate the results:

x 0.25a 0.30a 0.33a 0.40a 0.45a

V(x) 0.562a 3 0.588a 3 0.592a 3 0.576a3 0.540a*
squares at all corners of the sheet and
then bending the tin to form the sides
of the box (Figure 7.1.1). What is the We see that small variations in the
length of the side of the squares to be value of x near x =
a/3, that is, near
cut out? the value of x to which corresponds
Let the length of the side of the cut- the maximum of the function, bring
out squares be x. The volume of the box about very small changes in F, which
will depend on what kind of square we means that the function near the max-
cut out and therefore it is natural to imum varies very slowly.
denote it by V (#). Let us compute this This is also evident from Taylor’s
volume: formula (7.1.1). If f(x) 0 at the =
point of maximum or minimum of the
V (x) = (2a - 2a;)
2
x function, (7.1.1) takes the form
= 4 (a — a;)
2
x (7.1.4)
,

f(x) = f (x0 + y /" (x

) 0) (X —X 0)
2

which true because the base of the

is
rectangular box is a square with side
2a —
2a;, while the height of the box
+ {r(.x*){x-x (7.1.5)

is x. This series does not contain a term with

Now find the derivative of (7.1.4): x —
x 0 the smallest power of x
: x0 —
V'(x) — —8 (a — x) x + 4 (a — a:)
2
. on the right-hand side of (7.1.5) is
(x —x 0 ) 2 which is extremely small
,

Solve the equation V'(x) — 0: for x close to x 0 In our example, a

.

change in a; by 9% (from 0.33a to

—8 (a — x) x + 4 (a — x) = 2
0, 0.30a) causes a change in F by less
than 1%, while a change in x by 24%
or (a — x) (a — 3a;) = 0,
(from 0.33a to 0.25a) causes a change in
whence x x — a and x 2 = a/3 .
V by 5% Therefore, if we are interested
.

We note at once that the value xx — a in the maximal value of the function
no interest to us because then we and if we make a small error in finding
x 0 from the equation V' (x) — 0 (for
is of
wouldn’t have a box by cutting the
sheet in that fashion. There remains example, if we solved this equation
in an approximate fashion), then this
x = a/3. Then
has but slight effect on the maximal
tt /\
f (t)= /
4
— T=^r-°-
a a r on o
593a3 -
value of the function.
Note also that point x a/J in the =
Here V'(al 3) = 0, and since problem we
discussing here is
are
precisely the maximum point Common
V"(x) = 8x — 8 (a — x) — 8 (a — x) sense leads to such a conclusion. The
.

= 24a: — 16a, function that expresses the volume of

 7.1 Smooth Maxima and Minima 21T

Figure 7.1.2
expression for the power output
we get
the box is, obviously, positive for all
values of x between 0 and a at the ;

“endpoints” x 0 and x =
a it vanishes —
(in the first case the box has zero We must find the maximum of (7.1.6)
height and in the second it has a zero with R variable. We find the derivative-
base area). This implies that as x dW/dR:
is changed from 0 to a, the quantity
V (x) first grows and then diminishes. dW __ u2 2u2 R u2 — R)
(r
dJR (r+R) 2 (r+i?) 3 “ (r + i?)
3 *

This means that somewhere in the

middle the function becomes maximal, (7.1.7)*

and the only “suspect” is the point

x — a/3 (since only at this point does
This yields dWldR 0 at R =
r. =
V' (x) vanish), which means that this
It can be easily seen that R r —
corresponds to precisely a maximum
point is the maximum point. To make
this line of reasoning more graphic,
of the function W
(/?); this follows if

we depict the curve V V (x) in = only from the fact that for R r the <
derivative (7.1.7) is positive and for
Figure 7.1.2 (at point x a the graph = R > r it is negative. Common sense
touches the axis of abscissas since
also leads to the same result: for small
r (a) = 0).
R the energy output W
is low because
Example Suppose we have an elec-
3.
the resistance is low (in this case^
tric circuit that consists of an emf
almost the entire energy output “resides’ r
source (the emf equal to u\ compare
with Chapter 13), a resistor (the resis-
in r, while the presence of changes R
tance of which is r, including the internal
almost nothing; if R 0, then = 0), W=
while for R large, then, in view of
resistance of the emf source), and a
Ohm’s law, the current j will be low,
variable resistor (resistance R) (Fig-
ure 7.1.3). What will be the value of
which means W
will be low, too. Some-
where between small R and large R
R if the power output of this resistor, lies the value of R for which the power
W ,
is maximal? output W is maximal, and this can
Since the
resistance of the
total
happen only at R = r (see the graph
circuit is r +
R, the current j flowing
of W W
— (R) in Figure 7.1.4).
through the circuit is, by Ohm’s law,
Example 4. Using available boards,,
7= ul(r +
R). The power output W= we can build a fence of length l How
ju Rl where u R is the voltage across R ,
can we fence off a rectangular yard of
.

which by Ohm’s law is equal to jR.

maximal area using for one side the*
Substituting this value of u R into the
wall of an adjacent building (Figure
7.1.5)?
n i d —

t’2x
Figure 7.1.3 Figure 7.1.5
 ) .

218 7 Investigation of Functions . Some Problems from Geometry

Let two sides have length x. Then the

third side is Z — 2x. The area of the
yard is

S (x) = (l — 2x) x = —2x + 2

lx,

(7.1.8) Figure 7.1.6

whence S' (x) = —4x + Z. Clearly,

S' x )(.
= 0 only at x = Z/4; here £"(#) this case the area whose maximum we are seek-
ing depends not on one variable x but on an
= —4 < 0, so that at x —
Z/4 the
arbitrary function that specifies the shape of
function S (, x has a maximum. the boundary. It is possible in this case to
employ the methods of higher mathematics,
This result can also be obtained without re- too, by introducing an analog of the concept
isorting to higher mathematics. Indeed, sup- of a differential and by proving that the points
nt _ maxima^ a n d„ minima^ n £ _ tha J f u nc.ti.au_ n t
4

minu-
imum) of a second-degree polynomial (7.1.8) functions” considered (in mathematics such
of variable x. But for any second ‘degree poly- objects are known as functionals) correspond to
nomial the zeros of the “generalized” differential.
With the aid of such methods (which cannot be
.y — ax 2 + bx + c (7.1.9 discussed in a book as elementary as this)
it can be proved that the general solution to
we have
Dido’s problem is a semicircle (Figure 7.1.6)

:y==a
(
x2 + l x + ±} Example 5. A hiker walking from tent
A to camp-fire B wants to gather water
in river How can he do this by
traveling the shortest possible distance
(Figure 7.1.7)?
We have AA =
X a, BB = X and b ,

A XBX = c; the values of a, and c b,

are given and t AA ||
BB X _L A X B X Let .

Since (x + 6/2a) 2
nonnegative for all x's is the broken line AMB be the path taken
•equality occurring only at x bf2a we — — , by the hiker. Our aim is to find out
find that y has a maximum if a is negative and
this maximum is at x b/2a and y has a —— for what position of point M on the line
minimum a is positive and this minimum is
,
A X B X is this path the shortest. To
-at x = — b/2a.
if
In particular, the function determine the position M of it suffices
{7.1.8), with a = — 2, b = andl, c = 0, to specify the distance from M to point
.attains its maximum at x = Z/4.
The example we have just considered is
A 1? the
foot of the perpendicular dropped
one of the variants of Dido's problem. As le- from A
onto the straight line re-
gend has it, the mythical Elissa Dido, the presenting the river. Denote this dis-
daughter of the Tyrian king Mutton, after her
husband was slain by her brother, fled to Cy-
tance A X M
by x. Then
prus, and thence to the coast of Africa, where
she purchased from a local chieftain, Iarbas, AM =y + MB = y b + — x) 2
{c
z
,

a piece of land on which she built Carthage.

Iarbas agreed to sell a piece of land on the mock-
ing condition that it be no larger than the and the distance S (.r) traveled by the
area covered by an oxhide. But the cunning
Dido did not cover the tiny piece of land by
the oxhide, as Iarbas expected she should; in- \

stead, she cut the oxhide into thin strips and 1

“fenced” off a large piece of land, which was

made even larger by using the coastline. If 1
\ i a3
we assume the coastline to form a straight 1

line and require that the fenced-off area be 1

a rectangle, then the problem reduces to the

one we have just solved. However, the situa- AV
tion complicates if we assume that the bound-
ary of the area, the long strip made out of
the oxhide, can have an arbitrary shape. In Figure 7.1.7
 M M

219
7.1 Smooth Maxima and Minima

hiker will be that sin a =

sin p, or, since both a and
P are acute,
S(x) =am+mb = Y¥+7*
a = p. (7.1.12)
+ ]/fe + (c — 2
a:)
2
. (7.1.10)
Thus, the hiker must take the path of
This yields a ray of light bounced off a mirror ( the
c ~~ x equal to the angle
W_ x angle of incidence is
S'
Va +x 2
2 y'bz + —x ic )
2 *
of reflection ).
For a complete solution to the prob-
Equating S' (x) to zero yields lem it remains to demonstrate that
~x
-
=^
c
(7.1.11)
for such a position of point the dis- M
Yb + — x tance is indeed minimal (and not max-
.

Ya + 2 x2 2 (c )
2

imal). This can be done by computing

It is easy to solve this equation. the second derivative of (7.1.10).
Squaring both sides, we get But it is also possible to reason dif-
ferently. From the expression (7.1.10)
x2 (c—x) 2
a 2
+x 2 ~_ b 2
+ — (c x) 2
9 for S (x) we see that S (x) is pos-
itive for all x’s and increases with-
or x 2 b 2 -{- x 2 — x2
(c ) out bound together with the absolute
=a 2 (c—-x) 2
+x — x 2
(c 2 value of x, irrespective of whether
C
,
)

that is,
x >
0 or x 0. And since 5"(x) van-
ishes only for one value of x it is clear ,

x2 a2 that at this value of x the function

x 2
b 2 = a 2 (c — x) 2
,
(c—x) 2 ~~b 2 9
S {x) has a minimum. In general,
if in the interval we are interested in
whence
the first derivative of a function has
x
—X ±x ,
or x t = a ac
+b
only one root, obvious considerations
often permit dispensing with a formal
’
c

ac investigation by means of the second

X2 = —b
a derivative (see what was said in this
connection in Example 2).
Substituting the values of x 1 and x 2
The problem in Example 5 can be solved
into the original equation (7.1.11), we
in a purely geometrical manner without re-
see that the second root does not satisfy sorting to methods of higher mathematics.
the equation. This is an extraneous Referring to Figure 7.1,8, extend the seg-
root generated by the squaring pro- = AA
ment AAi to A' (A X A' X ) and join A'
with B. Then since A AA M —
AM — A'
cess.
71 Thus, x ad (a b). = + AA A'M. Therefore A'B = A’ + MB =
X
X

It however, to give a
is possible, AM + MB. For any other point D on the
pictorial geometrical representation segment A B we will have AD + DB =
1 1

that will enable us to obtain the answer A'D + DB and A'D + DB > A'B, since a
polygonal line is longer than any segment of
without solving the equation. Rewrite
the condition (7.1.11) thus:

AXM MB X
~~ (7.1.11a)
AM MB *

But A x MlAM Z^A X MA — sin a. cos

Similarly cos Z^B X MB =
MBJMB =
sin p. The condition (7.1.11a) means

7 1
*
The extraneous root x 2 of Eq. (7.1.11)
-can be dropped immediately because x/(c—x)
is positive and, hence, is not — alb. Figure 7.1.8
 x ) x

220 7 Investigation of Functions . Some Problems from Geometry

a straight line. Consequently, the desired point

M is the point of intersection of the straight
linesA'B and A X B^ whence follows (7.1.12).
(Why?)
Examples 4 and 5 show that certain prob-
lems involving the finding of maxima and min-
ima may be solved by the tools of elementary
mathematics, without resorting to differential
calculus (see the text to the solutions to the
corresponding problems printed in small type).
But, first, not all problems can be tackled
without appealing to higher mathematics and,
second, the solution by elementary means re-
quires a good deal of ingenuity, whereas high-
er mathematics offers a standard method of
solution of such problems.
Do not get the idea that higher mathemat-
ics does not require ingenuity! It will now
be used for still harder problems.
Let us now turn to the question of maxima
and minima of functions of two variables.
The definition dy = y'dx of the differen-
tial of a function y =
/ (x) (see Section 4.1)
shows that the points of (smooth) maxima or
minima of the function are characterized by
the fact that the differential vanishes at these
points. This is closely related to the geomet-
rical meaning of the differential (see Figure
4.1.2) and the fact that the tangent to the
curve representing the function at the point
where the function is at its maximum or min-
imum must be horizontal. It is equally easy
to see that at the point of a maximum or a
minimum of a function z F (x, y) of two =
variables, the differential of that function,

dz — dx
dy
+
dy must vanish; in other
dx
— other, g (y), a minimum. In this case the point
words, dF/dx dF/dy = 0. This is due to
(x 0 , y 0 ) constitutes a saddle point of the sur-
the fact that the plane tangent to the surface
specified by the function z F (x, y) is hori- = face z — F (x, y) and at this point F attains*
neither a maximum nor a minimum (Fig-
zontal at the point of maximum or min-
ure 7.1.9&).
imum (Figure 7.1.9a), 7 2 as well as to the *

fact that at the point (£q, y 0 ) where the

function z =
F (x, y attains a maximum
the functions f (x) =
F (x, y 0 ) and g (y) = Exercises
F ( 0 y) of one variable (x in the first
, 7.1.1. We want to build a box out of a rec-
case and y in the second) attain a maxi- tangular sheet of tin of sides a and b cutting-
mum, too, that is, f' (x) (—dF/dx) and g' (y) out equal squares at the corners. What must
(—dF/dy) must vanish at this point. However, the side of a square be so that the box is of
one must bear in mind that while the fact that maximum volume?
f(x) =
0 almost always means that / has a 7.1.2. Inscribe in an acute-angled triangle
maximum or a minimum (since cases where the with base a and altitude h a rectangle of the
second derivative vanishes are exceptional), largest possible area, two vertices of which lie
the conditions dF (;r 0 y Q )/dx dF (a: 0 , y 0 )/dy = on the base of the triangle and the other two
— 0 (or dz =
,

0) might mean that one of the on the sides of the triangle.

abovenoted functions of one variable, say 7.1.3. Determine the greatest possible area
/ (x), attains a maximum at (x 0 , y 0 ) and the of a rectangle inscribed in a circle of radius R.
7.1.4. For what radius of the base and for
7 2 what altitude will a closed cylindrical can of
The plane that is tangent to surface II
*

(represented by the equation z F (x, y)) = a given volume V have a minimum total sur-
at point M= M
(;r
0 y 0 ) contains the tangent
,
face area?
lines to the curves y y 0l z —F (x, y 0 ) = = 7.1.5. Two bodies are moving along the
/ (x) and x =x 0, z = F (
Q, y) = g (y) (to sides of a right angle with constant speeds v x
the curves y —
constant and x = constant); and v 2 (meters per second) in the direction of
of all the planes that pass through Af, the the vertex, from which, at the beginning, the
tangent plane lies closest to II. first was at distance a meters and the second, b
7.2 Other Types of Maxima and Minima. Salient Points and Discontinuities 221

meters. How many seconds after they started

will the distance between the bodies be at a
minimum?
7.1.6. Prove that the product of two posi-
tive numbers whose sum is a constant is the
greatest when the factors are equal.
7.1.7. A straight line l divides a plane into
two parts (medium I and medium II). A body
moves in medium I with a speed v± and in
medium II at a speed v2 What path must the .

point take so as to get, in minimum time, from

a given point A of medium I to a given point B
of medium II? (This problem poses the question
of the law of light refraction when light passes problem of maximum power does not
from medium I to medium II, and the speed
of light in these media is different; it is known
have a solution? From the physical
that the path the light takes between points A essence of the problem it follows that
and B is always such that the time taken is a the power output will be maximal at
minimum ( Fermat's principle see footnote
3.21)).
;
R = 0 (in this case W
= u 2 /r). Why
did we not get the value R = 0 from
the equation dWIdR = 0?
7.2 Other Types of Maxima and Minima. To see why, consider the graph of
Salient Points and Discontinuities.
The Left and Right Derivatives of
W = W (R) (Figure 7.2.1). It is evident
from the graph that if R could assume
a Function negative values, then for R 0 there —
Up now we have said that maxima
to would be no maximum. But from
and minima of a function occur at physical considerations it follows that
values of x for which the first derivative negative R have no meaning every —
vanishes. However, maxima and minima physical problem presupposes that R
can also arise for values of the argument is nonnegative. Thus, W
has a maximum
that do not make the first derivative at R = 0 because the range of the
vanish. independent variable is bounded. This
We revert to Example 3 of Section means that if the range of the indepen-
7.1,where an emf source, resistance r, dent variable (the domain of the func-
and variable resistance R are connected tion) is bounded, we must take into
in series (Figure 7.1.3). For what value consideration the boundary values of
of Rwill the power output of r be W the independent variable when testing
maximal? for maxima and minima.
In a manner quite similar to that of Let us touch once more on the exam-
Example 3, we get, from (7.1.6) ple of specific heat capacities discussed
in Section 2.6, only now we will take
u2r
W= 2
/ r
= ( r +B )2
* (7.2.1) water instead of diamond. The quantity
of heat (in joules) that is required to
The reader will recall that u and r are heat 1 kg of water (at atmospheric
assumed known and R variable and pressure) from 0 to T °G is given ap-
unknown. Equation (7.2.1) implies that proximately by the formula Q ( T ) =
dW ~~
2 u2r 4186.68T +
8373.36 X lO~ b T 2 1256 +
dR ( r+RY X 10- 6 3
r which implies that the spe-
,

with the result that the condition

cific heat capacity of water, c c (T ), = 7

at a temperature T is
dWIdR =0 becomes
—
(r+R)*
= U>
o c{T) = ^L = 4186.68
which cannot be satisfied for a single R. + 16746.72 x 10- f + 3768 x lO"
5 6 ?’ 2

Does this mean that the power can (see Exercise 2.2.2 and its solution).
increase without bound and that the Clearly, for T positive c T) increases
(
 )

222 7 Investigation of Functions. Some Problems from Geometry

with T. Does this mean that c ( T for maximum. The maximum is attained
water may be as high as desired? Of at x = b; indeed, we see that y —
a
course not, since water can exist in nullifies the radicand, while x cannot
the liquid state only in the temperature be greater than b because otherwise
interval from 0 to 100 °C, and the spe- the radicand would become negative.
cific heat capacity is maximal at the At x — b we have
a; for other y =
right endpoint of the interval and values of x the value of y is obtained
minimal at the left endpoint. through subtraction of a positive num-
Let us now return to the question ber 7 3 ]/ b x from a and
-
— is therefore
of the maxima and minima that a smaller than a.
function attains at its boundaries. A maximum (or minimum) may
When a maximum (minimum) is at- also occur at interior points where
tained at an endpoint x 0 of the domain the derivative does not vanish. This
of a function y f (x), the series— is the case when the curve has a salient
point {corner). Such points occur, in
1 0*0 — 1. (*o) = /' (xo) (x —x o) particular, when the curve consists

+ jf"(xo)( x ~ x + 2 of two parts described by different

o)
formulas for x 3> x 0 and for x x0 < .

Here is an instance of a physical

may begin with the term with ( x x 0 ). — problem of this nature.
Therefore, if a maximum (minimum)
Suppose a teakettle is being heated
of the function is obtained when x x0 = on an electric hot plate. Our problem
and we have departed somewhat from
is to determine the instant of time
x 0 (an endpoint of the domain of the
when the teakettle has the greatest
function), we may err considerably in
amount of heat. For the sake of sim-
determining y, since the error will be
plicity, we assume that the efficiency
proportional to x x 0 and not to— of the hot plate is 100%, which means
(x —
x 0 ) 2 as was the case in Section 7.1.
,
that all the heat is delivered to the
Hence, even a slight error in determin-
ing the value of the independent vari-
teakettle. We
put the teakettle on to
heat at time t 0, at which time it =
able that yields a maximum may lead
had q joules of heat. 7 4 By the definition -

to a sizable error in the value of the

of the unit amount of heat Q (the
function.
joule), the amount of heat released by
In the above examples the function
the hot plate is fRt, where j is the
f (x) existed for x x 0 as well, but C current in amperes, R the resistance
the values of the function for x <Z x 0
in ohms, t the time in seconds. Hence,
did not interest us because they were
the amount of heat in the teakettle
devoid of any physical or geometric
meaning. It may
happen, however,
by time t will be (in joules) Q q = +
that / (x) is simply meaningless for
fRt .

At the time t t 0 the water in the =

certain values of the independent vari-
teakettle begins to boil, and the amount
able. For example, there can be no
of heat accumulated by the teakettle
meaning in an even-degree root, say
a square root, of a negative number,
will be Q 0 q
2
j Rt 0 =
When the + •

water boils, it begins to turn into steam

with the result that the values of the 5
and boil away. 7 The formation of 1 g -

independent variable that make the

of steam requires approximately 2256.7
radicand vanish are the boundary val-
maxima minima
ues. In a test for
independent variable
such values of the
or 7.3 y —x
root.
^ i s understood to be the posi-
tive
must be considered sep aratel y. For 7 4
For zero we take the thermal energy
-

instance, if y = a b x then — Y — ,
of water at 0 °C.
7.5 Formation of steam starts at less than
y' = 1/2 —
Y b x, and although y
'

100 °G, that is, even before boiling starts,

does not vanish anywhere, y has a but we ignore this fact.
 t

7.2 Other Types of Maxima and Minima. Salient Points and Discontinuities 223;

joules of heat. Denoting by dm the

quantity of water that boils away in
time dt, we get
dm ~ (7
2
i?/2256.7) dt
~b
cz. 45 X 10 j
2
R dt.

And so in one second a total of dmldt ~

45 X 10~ 5 7 2 jR grams of water boil away.
This requires

dt
418.68 —
dt
drawing
~ 418.68 x 45 x ~5 Q (£) has a maximum*
that
10 pR when t =
although the derivative*
t 0l
~0.1884/' 2 i? J/s Q' ( ) does not vanish at this value of
t, that is, the tangent to the graph at
of heat because 1 g of water contains
point (t 0 Q (^ 0 )) is not horizontal (the
,
approximately 418.68 joules of heat
function Q — Q (t) has no derivative
at the temperature of boiling. There-
at t =
t 0 and the graph does not have
fore, by time * *0 the amount of > a tangent line at this point).
heat spent on transforming the boiled-
Thederivative of the function
away water into steam will be Q, ~ Q —Q has a discontinuity at t —
0.1884/' i? 2
(* — *0 ) joules of heat.
(t)

Indeed, if we consider only values of

t0 .

We see that the amount of heat (in

t less than f 0 we must assume that,
joules) in the teakettle is expressed by
two different formulas: for f <
Q'(t) 2
=
y R, while for t t 0 the deriv- >
f 0 (i.e.

=
ative is Q'(t) ^—
0.t884/ 2i?. The graph
prior to boiling) it is
q pRT, Q + of the derivative is shown in Figure-
while for * >
f 0 (after the water started
7.2.2 b.
boiling) it is
This example shows that a maximum
Q ~ q — 0.1884pR —
+ pRt 0 (* t0) (or minimum) may occur if the deriv-
ative is discontinuous, that is, if the
= q + pR (1.1884i _ 0.1884*). 0
curve representing the function has a
The graph of Q —Q (*) is shown in salient point (corner).
Figure 7.2.2a. It is clear from the From Figure 7.2.3 it is clear that a
minimum (or maximum) may occur
for those values x 0 of the independent
variable at which the derivative has
an infinite discontinuity (in the pre-
vious example the discontinuity was
finite). (We have depicted in Figure
7.2.3 the graph of the function
y =
x 2/3 = V #2 the inverse of this func-
;

tion,. or y =
or y 2 = x 3 is a x3 / 2 , ,

semicubical parabola see Section 1.5 ;

and, in particular, Figure 1.5.7.) A

point of this type is called a cusp the' ;

function y =
x 2 / 3 has a minimum at
this point. The graph of the derivative
of y —
x 2 / 3 is shown in Figure 7.2.4.
Here, as in the case of an ordinary
minimum, y' 0 for x < x 0 (in Figure <
7.2.4, x 0 —
0); the function falls off
as x approaches x 0 from the left. For
Figure 7.2.2 x > x 0 we have y r

0, and the func->

 ) —
:224 7 Investigation of Functions. Some Problems from Geometry

—— ^
corresponds to tx = t

and t2 = t At > t, whereas when

we take the fraction ——A t
,
we
have ttx = —
and t 2 = £. 7 6
t At < *

If
y y smooth= function, (t) is a
all these fractions yield the same limit,
which is equal to the derivative at the
given point. The situation changes
when we deal with a curve with a
salient point (see Figure 7.2.2). If

(Figure 7.2.4 by t0 we denote

the value of t at which
the salient point occurs, then, taking
tion increases with x after the value — 0
we get, for At pos-
.x =x 0 has been passed. But at x = x 0
itive and tending to zero, a definite
it becomes meaningless to speak of
quantity (in the example with
the
a derivative. The derivative becomes
teakettle this quantity is equal to
arbitrarily large for x close to x 0 and
greater than x 0 and arbitrarily large
— 0.1884/ 2 i?) called thej derivative on
the right of the function y (£) at point
in absolute value but negative for x
-close to x 0 and smaller than x 0
t =
t0 On the other hand, taking
.

The maxima and minima attained

.

—— ,
we get, for At pos-
for values of the independent variables
itive and tending to zero, another
when the derivative is discontinuous
limit the aforementioned
(equal in
are called cuspidal. Cuspidal maxima
example to j 2 R) called the derivative
and minima and the maxima and mini- on the left of the function.
ma attained at the endpoints (boundary Taking t2 and tx on different sides of
points) of the domain of the function
£0 we can obtain different values of the
can also be said to be nonsmooth.
,

ratio (7.2.4) as t 2 — 1 0 and as t x 10 ;

In connection with this consideration
it is easy to see that if is a corner A
of singular points on curves, primarily
or a cusp, then the chord where BC
salient points (see Figure 7.2.2 a), we
,

B and C lie on different sides of , A

<&an make precise our reasoning that
may have different directions (imagine
led us to the concept of the derivative.
such a chord in Figures 7.2.2a and
In Chapter 2 we considered only smooth
7.2.3), This means that the limit of
curves without specially stipulating
the chord, as B and C tend to A, may
this fact. The derivative y' ( t taken
also be different— it depends on how
at the point t is equal to the limit of
precisely the points B and C tend to
the ratio
point A. Thus, at a salient point of
y(t 2 ) — y(t i) in
2
a function the derivative of that func-
tion has no definite value, but the deriv-
atives on the left and on the right
as t2 and t1 tend to t (it is clear then
can be found unambiguously.
that the —
difference t 2 tx tends to
zero). We have specially emphasized
that this limit does not depend on how
7 « 6
For smooth curves the derivative was
t 2 and t x are chosen: they can both
be greater than t or both smaller than t calculated as the limit of the ratio +
or one greater and the other smaller
than t or one equal to t and the other — y^t— yJJ/Ai as At-* 0 (see the

.greater or smaller than t. Indeed, when text in small print at the end of Section 6.1);
we take At positive, the fraction here t x t = —
At/2 <C t and t 2 t At/2 >>£. = +
 t t t

7.2 Other Types of Maxima and Minima. Salient Points and Discontinuities 225

Figure 7.2.5

In Chapter 2, when we first began

studying derivatives, we simplified mat-
ters by not assuming all the time that
Figure 7.2.6
the derivative has a definite value
irrespective of the mode of approach b

of A t to zero (from the left or from the Now how does the integral dt
y ( t)
right) only for points at which the j
a
curve representing the function is behave when the function y(t) is not
smooth. As is evident from Figure 7.2.26, smooth? If the function has a salient
the curve of the derivative y'(t) has point, then no new problems arise
a discontinuity at the point where the when we compute the area bounded by
curve y ( ) has a salient point. Now if the curve y (t). In Section 3.2 we split
we replace the salient point on the up the definite integral (the area under
curve y ( t ) by an arc of small radius the curve) into rectangular strips with
that is tangent to the curve on the left an area y ( t n ) (tn+1 —
t n ) or
y ( t n+1 ) X
and on the right (what draftsmen call
(*n+i —t n ). In the limit, as the in-
conjugation ), the resulting smooth curve
will correspond to a continuously vary-
tervals, that is, the differences t
n+1 —
tn,
get smaller and smaller, it makes
ing derivative; however, on the range no difference whether one takes y (t n )
of t where the curve y ( t ) is replaced or y (tn+i) either in the case of a
by the arc the curve y'(t) changes di- smooth curve or in the case of a curve
rection sharply (compare Figures 7.2.5 with salient points.
and 7.2.2). If a curve y ( ) is discontinuous at a
the curve y ( ) has a discontinuity
If point t —
t0 but remains bounded,
at point t 0 (Figure 7.2.6a), then we then for the interval that contains the
can say that at t 0 the derivative y'(t)
is infinite (although actually the func-
discontinuity (t n < < t0
t n+1 ) the
quantities y (t n ) and y
n+1 ) remain (t
tion has no derivative at this point). distinct no matter how t n and t
Indeed, if the discontinuity is replaced n+1
approach one another. To summarize,
by a smooth variation of y from y x to then, in the expression of the integral
y 2 over a small interval from t 0 e — as a sum, the value of one of the sum-
to t 0 -f- e, then on this interval the mands in this case to a great extent
derivative is equal to (y 2 —
y^)l2z, depends on how the sum is taken:
which is a very large quantity, increas- by formula (3.2.1) or by formula (3.2.2).
ing as 8 decreases (Figure 7.2.66). However, as f n+1 —
t n tends to zero,
15-0946
 226 7 Investigation of Functions. Some Problems from Geometry

the summand tends to zero,

itself ima and minima of a function goes as
whereby the limit the sum, the
of follows. First we must find the deriv-
integral, has a definite value (indepen- ative of the function under investi-
dent of the way in which the sum was gation and determine the stationary
computed) also in the case where the points, that is, points at which the
integrand has a discontinuity in the variation of the function is the small-
domain of integration. Here, of course, est— points at which the derivative
we assume that the values of the func- vanishes. But in addition to these
tion to the left and to the right of the points, “suspects” (i.e. points at which
discontinuity and, hence, the size of the function may be maximal or minim-
the discontinuity are finite, the func- al) are the boundary points of the
tion does not behave as y l/;r does = domain of the function, salient points,
in the neighborhood of point x 0. = and points at which the derivative
The relationship between the integral becomes infinite and changes its sign
and the derivative is likewise preserved. (“cusps”). We must then check how
Referring to Figure 7.2.2, let us take the function changes at all these points.
the function Q'(t) (whose graph is If at a point where /' (x) 0 the =
shown in Figure 7.2.2 b) and denote graph is represented by a smooth curve,
it by / (£). Then the function Q (t), information on the type of the point
whose graph is shown in Figure 7.2.2 a, (maximum, minimum, inflection) can
is the integral Q (
t) = j/ (
t) dt. This be obtained through the second deriv-
ative of the function (or higher-order
example shows that a discontinuity in the derivatives if the first and second
integrand function / (t) Q'(t) leads to = derivatives vanish at this point simul-
a salient point in the integral Q ( t ) of taneously). In other cases an idea about
this function. The definite integral of the behavior of the function in a point
a function with a discontinuity of this being considered may be given by the
type can be found with the aid of the one-sided (i.e. left or right) derivatives;
indefinite integral by the general rule it is clear that at the boundary points
b of the domain of a function only one-
f(t)dt=Q(b)-Q(a). sided derivatives can exist. Finally,
j if we are interested in the absolute
a
maximum and/or minimum of a function
We may continue: consider Figure 7.2.6. over the range of variation of the
We can say that for a function tending to in- function, that is, the largest of all the
finity on an interval tending to zero (Fig-
7 -7
(local) maxima or the smallest of all
ure 7.2.66), the integral is a discontinuous
function (Figure 7.2.6a). However, in this the (local) minima, we must compare
case we must make precise the law by which the values of the function at all points
the function tends to infinity and the interval of local maxima (or minima).
to zero. We will not dwell on that here. Exam-
ples of this kind (a fuller consideration of which
requires additional refinement) lead to the Exercises
concept of the delta function (see Chapter 16).
7.2.1. Find the smallest value of the func-
tion y = x2 —
2x +3 as x varies from 2 to 10.
Thus, the final scheme for solving 7.2.2. A fisherman F is sitting on the bank
problems that involve finding the max- of a river and fishing. Going down the river,
at a distance h from the place where F is sit-
7-7
The expressions “an interval tending to ting and with speed u, there is a steamer S of
zero” and “a function tending
to infinity” point length l, and at time t =
t 0 the bow of the

to the fact that in essense we are speaking not steamer is positioned exactly opposite F.
of a single function but of a family of functions (We ignore the fact that S has width and as-
with the following property: as we go over sume that S moves along a straight line paral-
from one function in this family to another, lel to the bank.) For the distance between F
the “growth interval” becomes more and more and S it is natural to take the distance from S
narrow and the “degree of growth” higher and to the point on S closest to F. How does this
higher. For more details see Chapter 16. distance D vary with time? When will D —
 )

7.3. Investigating Maxima and Minima of Functions Depending on a Parameter 227

D t
( )
be minimal? Draw the graph of the func- We
take the radius of a as the unit of
tion D =D {t) for h = 300 m, 1—6 0 m,
length; the distance from the center OM
and = m/s.
7.2.3.
v 5
Find the cuspidal maxima _of the of circle 0 to point M is denoted by a.
following functions: (a) y = — 5) x 2 and
,
Obviously, the altitudes BK and DL
(b) = 1 — of AABC and AADC are, respective-
;/

ly, MB sin a and DM sin a, which

yields
7.3 Investigating Maxima and Minima
of Functions Depending on a Parameter S ABCD = S = S& AB c 4- S& AD c
In Sections 7.1 and 7.2 we separated
= y AC BM sin a + y AC DM sin a
• •

the cases of “internal” maxima and

minima, which are attained by a func-
tion in interior points of the domain
= y AC (BM -\-DM) sin a
of the function (the range of the in-
dependent variable on which the func- = Y AC-BD sin a = BD sin a,
tion depends), and maxima and minima
attained by the function at the end- since, obviously, AC = 2.

points of the domain of the function.

Moreover, the distance OP ( —b from
However, in studying functions one the center of a to the chord BD is
may encounter problems in which both OAfsina =
a sin a, which yields BP —
types of maxima (or minima) are pres- PD — \ 1 — a 2 sin 2 a (since OB =*
ent. Situations of this kind often occur OP — 1) and , hence, BD =
when the function we are interested 2]/ 1 —a 2
sin 2 a, that is,
in depends not only on the indepen-
dent variable but also on a parameter, S = BD sin a — 2 sin a V 1 —a 2
sin 2 a.
and a variation in this parameter within (7.3.1)
the limits allowed by the problem
of finding the maximum
The problem
changes the nature of the maxima (or
minima) or even the nature of the
of S becomesreally simple if we go
over from S to the square of the area
function itself. To illustrate what we
have just said there are several exam-
S2 =
4 sin 2 a (1 —
a 2 sin 2 a), since it
is clear that S and S 2 attain their
ples.
Example 1. Select a point on the M maxima (or minima) simultaneously,
and introduce a new variable, x —
diameter AC of a circle a. What must
sin 2 a. Then
the angle a be between the diameter
and a chord BD of a passing through S 2 - 4z (1 - a x).
2
(7.3.2)
point M
so that the quadrangle ABCD
Since in the #y-plane — S the graph
2
inscribed in a is of maximal area (y )

(Figure 7.3.1)?
of (7.3.2) is a parabola , we can do with-
out finding the derivatives; however,
since there is a general (and simple)
method for finding the maxima and
minima of a function through differen-
tiation, we use this method. Clearly,
if F(x) 4r = —
4 a¥, then F'(x) —
4 —8a 2x and F'(x) 0 at x l/2a a = = .

And since S (a) (and, hence, S 2 (a) =

F (x )) vanishes at a 0 and a 180° = =
and between these two values is posi-
tive, we conclude that it attains a max-
imum (see the text at the end of Exam-
ple 2 in Section 7.1) somewhere in this
Figure 7.3.1 interval; precisely, we should expect

15 *
 a a , a ) a a a

228 7 Investigation of Functions. Some Problems from Geometry

that the maximum of S (a) and that OP = a sin a a 2/2a ]/ 2/2 = Y = =

of F (x) is attained at x — sin 2 a — constant, and in this way the distance
l/2a 2 .
between BD
and the center 0 of circle a
But it clear that sin 2 a is equal
is is constant and is equal to the distance
to 1/2 2
only for l/2a 2 1, or 2
2
1, ^ ^ from 0 to a side of square T\ the chords
or 2/2, and in this case the BD and B 1 1 that pass through point D
solution of the problem is indeed sup- M which is exterior with respect to T
,

plied by the condition sin 2 a 1/2 a 2 = ,

(lies outside T) and lies on AC, touch

or sin a =
there are two+'|/^2/2a; the circle a x inscribed in T.
chords, BD
and BJ)^ that are sym- This result, which appears unexpected at
metric about AC
and form the approp- first glance, becomes crystal clear if we draw
— —
the graph of the function y F (x) 4r =
riate angle a with AC (at a = ]/ 2/2 2 2
4a x 2
(x= —1/2 2 ) 2 —
l/4a 2 a straight +
merge into one chord BD J_ AC). For line at a =
0 and a parabola with its vertex
,

such a value of x the function F x (.

at point Q ( 1/2 2 1/4 a 2 ) for a =^= 0 (Fig- ,

has an “internal” {smooth) maximum. ure 7 3 2 ). For

. . a> /
2/2 the vertex Q lies in-
side the domain of F (x), and the 0<a:<l,
For a < Y 2/2 the function F {x) varies function attains its maximum at this vertex,
monotonically with x varying between while for a <
j/2/2 the point Q lies outside
0 and 1 (since x = sin 2 a), whereby the domain, and the maximum is attained at
the greatest value is attained at the the boundary point x 1. Note also that for =
boundary point x 1, corresponding = a 1 —
> ^ 2/2 the value x not 1 of the independent

= to a maximum of
to a 90° (since x sin 2 a): here = variable corresponds
F (x) but, on the contrary, to a (local) min-
only one chord, BD which is per- , imum because for values of x close to unity
pendicular to yields a maximum AC ,
but less than unity (for values of a close to a
for S. We
note also that in view of right angle) the magnitude of F (x) (the area

(7.3.2) the value S S m ax is expressed = S (a)) will be larger than at point x 1 (at =
a = 90°).
differently for the two cases we are Of course, in solving this problem there
was no need to go over from function S (a) =
considering:
=
for a ^ Y 2/2 we have
= 4 a —a 2
sin 2 a F
F max F (1/2 2
)
(l/2a X X 2
)
2 sin ]/l to function (x).
The derivative S (a) with respect to a can be
(1 — 1/2) = 1/a 2 ,
which means that found by the rules of differentiation. The re-
In this Case ^max ==
Y -Fmax === 1/^j sult is

= F (1) = — — 2 cos a X— — 2a~ sin

2 2
dS a) m .o.O)
but if ^ 2/2, then /'max
— a Y and, hence,
a —
da
:
— /-
(1
:
. ( i
0 ON

4 (1 2
)
S max = y i a2 sin2 a

Y ^max = 2 ]/ 1 — a2 . Thus, the From (

7 3 3)
. . we see that the fact that S' (a)
=
vanishes is equivalent to cos a 0 or to
graph for 5 max = S (a) (in the aS-plane)
is represented, on the segment 0 ^
a^l, by a “broken” curve consisting
of two different arcs: the arc of the
a 2 -f (S/2) 2
“ellipse 1 on the segment —
0^ ^ Y =
a 2/2 and the arc of the
hyperbola S 1/a on the segment

Y 2/2 ^ a ^ connected 1 to the first

arc.
Note also that the square T inscribed
in a for which is the midline cuts AC
out of this midline a segment L 1 L 2 ,

with OLi = OL = Y 2/2;2 thus, if

point M lies inside T then the sought

chord BD is perpendicular to AC ,
but
if M lies outside T then the ,
first case
of the above two occurs. Here b ==
 ,

7.3. Investigating Maxima and Minima of Functions Depending on a Parameter 229

cal meaning when x greater than unity;

is
the maximum of F (x) at this
(or minimum)
point is a “boundary” one, since here F' (1) =£ 0.
In the latter, on the other hand, point a =
90° is an interior point for S (a); the maximum
(or local minimum) attained by S (a) at this
point is a smooth one. This should not be a cause
for surprise, since it is clear that the functions
S (a) and S 2 (a) (= F(x)) attain their maxi-
mum and minimum (since S 0) in the same >
point, but the nature of the maximum (min-
imum) may change. For instance, all three
functions y j/| * [, V\= y
2
z I, and = =
y2 —
yl x 2 —
(Figure 7.3.4) have the same
I

minimum point x =
0; however, y has at this
point a cuspidal minimum (both the left and
right derivatives become infinite at this point),
y 1 has a “corner” minimum (the left and right
derivatives are nonzero and do not coincide),
and y 2 has a smooth minimum at x 0 =
sin a = ± 1/a l/2. The first case corresponds (here y'2 (0) = 0). Indeed, if Y= F (x) =
to a = 90°, that a chord
is, BD 1. AC, y
2
(x), then dY/dx =
2 yy' and the conditions
while the second case is possible only if a ;> Y' =
0 and y' =
0 are not equivalent. Anoth-
2/2 (~0.71). Accordingly, we arrive at differ- er reason why the nature of the maximum
ent solutions to this problem depending on or minimum may change is the change in the
function itself (in addition to this, when we
whether a> j/2/2 or a < |/2/2. For a > went over from S to F we simultaneously
l/2/2, the function S (a) vanishes at a single
= — changed the independent variable from a to x =
point, a 90°, and since the function S sin 2 a): since if y / ( x ) and x = cp (t), we =
S (a) must have a maximum, there can be no have dy/dt =
(dy/dx) (dx/dt), and dy/dx 0 =
doubt that a =
90° corresponds to precisely is not equivalent to dy/dt 0. =
a maximum of the function. However, when a
is less than ]/ 2/2, the derivative S' (a) van- Figure 7.3.5 depicts the graph of
ishes at three points: at a 90° and at = the function a a (a), where a is =
sin a = ±l/]/2a. If we check
the sign of
defined by the condition S AB cd =
S" (a) (or check the sign of S' (a) in the neigh-
borhood of points a 90° and = a = S =
5 max in a certain sense this
;

+ arcsin (1/ j/2 a)), wewill see that in this case drawing summarizes the information
the maxima are attained precisely at points a = we obtained during solving the prob-
dt arcsin (1/ ]/ 2a), while at point a — 90° lem. The reader can clearly see that
the function has minimum
ure 7.3.3).
a local (Fig-
if a ^
]/ 2/2, the problem of finding
Here another aspect in which
is Fig- the quadrangle with the great-ABCD
ure 7.3.2 differs from Figure 7.3.3. In the for- est possible area has a unique solution,
mer, point x =
1 is a boundary point of func-
tion F (x), and this function has no geometri-
while if a > j/2/2, there are two
solutions, and they correspond to two

Figure 7.3.4 Figure 7.3.5

 S B

230 7 Investigation of Functions . Some Problems from Geometry

chords BD and B D1
1 symmetric about In the latter case the distance OP from
AC. (Note that such the center O to BD is equal, as we saw earlier,
a “splitting”, or
to a sin a, that is, it can vary from 0 (the case
branching, at a critical point of one where a =
0 and BD coincides with AC) to a
solution into two, an event correspond- (the case where a —
90°, or BD 1. AC). Ac-
ing to a change in the nature of the cordingly, the length d BD of chor d BD =
solution to the problem, is rather often can vary from 2 (when BD == A C) to 2|/’l a2 —
encountered in mathematics.) Also, the (when BD _L A C); the angle P varies from
very fact that at the critical point a = 180° (when BD =
AC) to 2 arccos a (when
BD _L AC), that is, remains constantly ob-
]/ 2/2 the point of maximum a = 90° tuse (since a is less than )/*2/2, we conclude
of the function S = S (a) transforms that arccos a is greater than 45°). The quan-
into a point of (local) minimum is tity sin p attains its maximum at P p in = m =
inherent in many problems and not
2 arccos a, that is, at A. AC, and it is BD
this chord BD
that ensures that the area of
only in this, rather randomly chosen, A BCD is at its maximum.
problem. Of course, the unnatural solution that we
are discussing here is much more difficult to
Of course, this problem, too, can be solved comprehend than the elegant solution involv-
without resorting to differential calculus, that ing finding the derivative of S (a) (or F (x)),
is, by purely geometrical methods. It is ob- since one must guess that the areas of S and s
vious that are proportional, and this is not an easy job.

Sabcd — Example 2. Let the product of the

~ S AMAJ3+ S A MBC+ S AMCD~^~ S AMDA distances from a (variable) point M
to two fixed points A and B be a What .

— ysin a X MA-MB + MB-MC

( is the greatest distance d = = MP
+MC-MD+MD-MA) d m ax from M
to the straight line A B,
and how to specify the point for M
t=-^-(AC -BD) sin a = BD sin a; which this distance d max is realized?
In many respects this problem is
on the other hand, similar to the previous one, and for
this reason we go into less detail. The
S AObd “ s = ~21 BD'OP — 1
-g- BD-a sin a
set of all points for which M MA MB —
•

= — a*BD sin a,
1 constant (=&) constitutes a curve 2
called the oval of Cassini (Figure 7.3.6)
that is, s =
(a/2) S, which means that s and in honor of the Italian astronomer
S attain their maxima simultaneously. But Giovanni Domenico Cassini (1625-1712).
since
(Being a follower of the heliocentric
theory of the solar system, Cassini
= — OB -OD sin Z BOD = —
1 1
s sin (3
Li h tried to correct the faults of Ptolemy’s
where P =
A BOD, it remains to he estab- system by substituting for Kepler’s
lished at what point sin P attains its maximum. ellipses as the paths of the planets curves
Clearly, if p can be equal to 90° (sin p 1), = like the one just mentioned; such curves
this will correspond to the absolute maximum
are also called Cassinians ). If we in-
of 5 (and, hence of S , too). But p z OBD = = troduce a plane system of coordinates
90° ifjhe length of the chord of a is equal BD
to Y
2, the length of the side of the square T
in such a manner that points A and B
inscribed in cr, or if the distance OP from the have coordinates (—1, 0) and (1, 0)
center of a to BD is equal_to half the side of T (in this way the distance A between
(the half, of course, is l/*2/2 units long), that the points is equal to two units of
is,when BD
touches the circle inscribed in
length), then, in view of the fact that
T, which is obviously impossible if point M for M=M x , y), obviously, MA =
belonging to AC lies inside the square T (with (;

midline AC) inscribed in cr. Thus, when point x l)

2
+y 2
and MB =
M belonging to AC lies outside T the prob- ,
]/ (

x
-f-

— 2 2 the equation of curve

lem is solved by drawing through the M y (
l) -f y ,

straight line BD tangent to (either BD or 2 has the form

#i£>i) for which P =
90°; but if lies inside M
T, the situation changes. y {x- r l)
2
+ y*xV (s-i )
2
+ y = a,
2
 x .

7.3. Investigating Maxima and Minima of Functions Depending on a Parameter 231

yA

Figure 7.3.6

which can be simplified thus: by the following equation: F(x , y) = 0,

2 2 2 — 2 (x
2 —y =2
a2 — with F{x, y) = (. x2 +y 2 2
)
— 2(x — y 2 2
(,
-f y ) ) 1
— a2 -f- 1 (cf. (7.3.4)). In view of
)

(7.3.4)
(4.13.7), we have
(Verify this.) dy dF/dx 2 2
2 (x -f- y ) 2x — kx
It clear that the oval of Cassini
is dx dF/dy 2 (x 2 +y 2
)
2y + 4y
degenerates into a pair of points, A (x 2 -\-y 2 — 1) x
and B
(in which case our problem has (x*+y 2 +i)y' (7.3.5)
no meaning), if a =
0. If a is very
whence dyldx = 0 if and only —
x 0
small, then either the distance or MA or x 2 -f y 2 = 1. But at x
if
— 0 the
the distance MB
is very small (other-
equation of the oval of Cassini, (7.3.4),
wise the product MA MB
cannot be
•

= — =
small), and in this case the oval of
yields y*
2
2 y2 a2 +
l,or (y 2 +1) 2
a from which it becomes completely
Cassini splits into two ovals surround- ,

ing points A and B. As a increases,

clear that for a 1 the oval of Cassini <
does not intersect the straight line
the two ovals of Cassini increase in
— x = 0 (the y axis); here dyldx = 0
size and, at a 1, finally touch at
only at points for which x 2 -f y 2 = 1,
the origin 0, so that the entire curve
or at points where the oval of Cassini
resembles an “eight” lying on its side.
intersects the unit circle S with diam-
This curve (in view of (7.3.4), its
eter AB (since S has the equation
equation is (x 2 -f- y 2 ) 2 2 (x 2
2 = —
was studied by Jakob Bernoulli and
y )),
+y —
!)•
2
If is clear that there
are four points at which 2 intersects
is called the lemniscate of Bernoulli .

5, and these points are pairwise sym-

When a is greater than unity, the oval
metric about the x axis and the y axis.
of Cassini is a single closed curve
enveloping the
They necessarily yield the solution
two “poles” of the
to our problem, since they lie farthest
curve, A and B. Finally, when a is
from the x axis.
very large, the distances and MA MB The points of intersection of 2 and
are also large and will differ little;
here the oval of Cassini
S can be found by solving the following
resembles a
system of equations:
circle of an extremely large radius
Ya centered at 0 .
x 2
+y = 12
and
Let us now find the derivative dyldx, 2 (x
2 —y = 2
)
(x 2 -f y
2 2
) -f 1 — a2
where the dependence of y on x is given = 2 -a 2
.
 2

232 7 Investigation of Functions. Some Problems from Geometry

or part of the axis of ordinates external

x 2
y
2 = 1 and to S (Figure 7.3.7): for a 2 the point ^
of maximum of y belongs to circle S,
*2 —y 2 = 1 — a 12,
2
and the closer a is to 2, the closer is
the pair of points above the x axis
which immediately yields to each other (the same, of course, is
x2 — 1 — a2 / 4 and y 2 = a 2 ! 4. true of the pair of points below the x
axis); at a =
2 the two points with
This readily implies that these points y >0 (and the two points with y 0) <
exist only when a is no greater than 2, merge, and for a greater than 2 the
since x 2 is necessarily positive. Whence, points of maximum of y belong to the
for a 2 the condition dyldx = 0 is
> y axis and not to circle S (The reader .

satisfied only at x 0, which means= can easily convince himself that the
that the points of the oval of Cassini graph of y =
y max (a) in the ay-plane
that lie farthest from the # axis (such consists of the segment of the straight
points are sure to exist) are those at line y =
al corresponding to 0 a ^ ^
which 2 intersects the y axis. 2 and the arc of the parabola y 2 =
Finally, for 1 a 2, the deriv-< < a —1 corresponding to the values of
ative dyldx on the upper half of the a greater than 2.)
oval of Cassini (in view of the fact that
2 is symmetric about the axis of abs- Of course, this problem can also be solved
by the methods of elementary geometry, that
cissas we can always consider only the is, without using differential calculus. It is
upper half of this curve) vanishes clear that
three times: at the point where 2 1
intersects the axis of ordinates and at S A mab ~~2 MA MB sin a = y sin a,
‘
CL

the two points where 2 and S intersect.

while on the other hand
If we study how dyldx changes sign
near these points (or find the sine of S A mab=J ab -MP = MP,
cPyldx2 ), we will see that y is at its
maximum exactly at the points where where MP is the distance from M to AB, and
2 and S intersect, while x = 0 corre- a =A A MB, whereby the maximum of MP
sponds to a (local) minimum of y (see corresponds to the maximum of sin a. There-
Figure 7.3.6). Thus, the set of all fore, our curve 2 contains points
if such M
points of maxima of y for different
that A A MB =
90° and sin a 1, then these =
points yield the maximum of MP, since all
values of a consists of circle S and the points M
of the circle S whose diameter is AB
with A A MB =
90° are indeed the points of
maximum of MP
(of course, different points
on S correspond to different values of para-
meter a). Completing A MAB
to parallelogram
MAM X B centered at 0 readily leads to

4 MO = 2MA + 2 MB — AB
2 2 2 2

= 2 (MA — MB) + AMA -MB - 2

4
= 4 {a — 1) + 2 (MA — MB) 2

>4 (a - 1),

where we still assume that AB = 2. Thus,

if a >> 2, then the equality MO = 1, which
means that M belongs to 2 and to circle S
with diameter AB and, hence, a = A A MB =
90°, is simply impossible (while at a — 2
only the points of intersection of 2 with the
axis of ordinates, points for which MA —
MB = 0, satisfy the condition 1). MO =
Hence, while for a 2 the maximum of thec
distance y |
= MP
from point
|
on curve 2 M
 7.3. Investigating Maxima and Minima of Functions Depending on a Parameter 233

to the axis of abscissas is attained at the points plane known as isotherms (curves of equal tem-
where 2 and S intersect, for a >
2 the situa- perature). Each value of T has its isotherm.
It is clear that ideal-gas isotherms given
tion is different.
The reader can easily see that if a > 2, by formula (7.3.6) for different values of (pa-
then for each point belonging to 2 the M rameter) T comprise a family of hyperbolas
=
angle a i_AMB is =
always acute (since pv constant (=RT) with asymptotes v 0 =
point M
lies outside circle S whose diameter is and p =
0; these isotherms have neither max-
A B)\ therefore, we need only find a point M ima nor minima. The situation complicates
for which angle a (and, hence, sin a) is max- considerably when we go over to the van der
imal. By the laws of cosines applied to Waals isotherms (7.3.7).
T
e will now inve- W
Z MAB, stigate such isotherms. The curve specified
by (7.3.7) depends on three parameters: the
2MA MB • cos a coefficients a and b in the van der Waals equa-
= 2 a cos = MA + MB — AB
a 2 2 2
tion of state (these coefficients are character-

= (MA — MB) + 2 MA MB — 4
2
•
istics of the gas being investigated) and the
temperature of the gas, T. Later we will see-
= (MA — MB) + 2a - 4,2
that the qualitative behavior of isotherm
(7.3.7) is determined by the ratio of T to the
which implies that the smaller the value of fraction alb.
cos a (and, hence, the greater the value of a), Let us use (7.3.7) and find how p depends
the smaller the quantity (MA MB) 2 whence — , on v,or the function p p (v). The answer is =
the smallest possible value of cos a (and,
hence, the greatest possible value of sin a) RT a_
will be achieved at 0, that is, MA — MB — v —b v2
*
(7.3.8)

when M
belongs to the y axis. Thus, for a 2 >
the maximum distance from to MP M From this it follows that
AB is attained for the points on the axis of
dp RT , 2a
ordinates. This completes the solution to the
problem.
P
dv (v — b) 2 v3

Example 3. Here is another example. It _ 1 [~ 2a (v — b) 2 ,

RT ]•
is more complicated than the previous two, ~~
(v — b) 2
L vs
(7.3.9)
has a serious physical meaning, and also deals
with finding the maxima and minima of a in view of which the possible maxima and
function. It is well known that the relation- minima of p p (v) (for a given temperature =
ship between the volume v, the pressure p T) are determined from the condition ,

and the absolute temperature T (in kelvins)

2a (v—b)*
of an ideal gas is given by Boyle’s law and
n vfaJ-L7Qssac 3 raw;"Vn\cii cai) 0, (7.3.10) RT =
oe ufiu'ed 'tnio
the ideal gas law which is equivalent to p ' 0. Unfortunately, —
pv — BT y (7.3.6) however, (7.3.10) a cubic equation in v is ,

and the solution of this equation is not that

where B is the molar gas constant of an ideal simple.
gas the adjective “molar” meaning that when
, Not being able to solve Eq. (7.3.7) di-
related to one mole of a gas, R is the same for rectly, we will study it qualitatively. Consid-
all gases; for instance, if v is measured in m 3 er the function / (z;) 2a (v b) 2 /v* = — — RT
and p in Pa, then R ~ 8.25 N -m/K. However, (which depends on the same three parameters a,
formula (7.3.6) does not agree very well with b, and T). We will try to establish how it va-
the properties of real gases, and so a com- ries, that is, find its maxima and minima. To
mon way to correct it is to write it in the this end we find the derivative of /:
form
df _ 2d [ 2(v-b) 3 (v — b) 2 ~|

(?+-£-) (”->>) = RT > (7.3.7)

dv L y3 i>
4
J
2a (v fr) ^
rj 2
which is known
van der Waals equationas the
of state. 1 8 The constants a and b are empirical
•

From the physics of the problem (see

(i.e. determined for each gas separately) pos-
Eq. (7.3.7)) it follows that v is always greater
itive constants (in the system of units used
than b (the absolute temperature T cannot be
above to define R the dimensions of these con- negative), whereby (7.3.11) may vanish only
stants are N -m4 and m 3 respectively). If the =
temperature of the gas under investigation
,
at v 3b. Further, for b v 3b the right- C <
hand side of (7.3.11) is negative, which im-
is kept constant, the formulas (7.3.6) and
correspond to certain curves in the vp-
plies that v =
3b corresponds to the maximum
(7.3.7) of / (v), the greatest possible value, equal to

7
8a/27b — RT. Now we must consider three
- 8 Johannes Diderik van der Waals (1837- different cases, corresponding to three tem-
1923), a Dutch physicist. perature intervals.
 ) p ) x

234 7 Investigation of Functions. Some Problems from Geometry

volume v which corresponds to the possibility

,

of only one state of matter occurring at T >

8a/27bR, and this is the gaseous state. On the
other hand, if T <
8a/27bR, to each value of
pressure p there correspond, as shown by Fig-
ure 7.3.8, three different values of volume
v, so that here matter can exist in different
states. The state with the smallest volume
v —y liq (he. the greatest density) is the liq-
uid state. The state with the greatest volume
v —vg aS (at the same temperature and pres-
sure as the liquid state) is the gaseous state.
Finally there is the intermediate state, which
proves to be unstable. If the substance we are
studying is placed in a vessel of a certain vol-
ume v (somewhere between zqiq and gas ) and z;

heated up to a temperature T lower than a cer-

tain temperature T c (see below), part of the
substance will be in the gaseous state and the
remainder will be in the liquid state. A
(a) 8a/27b — RT < 0, that is T > 8a/ definite state with the intermediate volume v
27bR, or T/(a/b)> 8/27 R. In this case / (v) 0 < does not exist.
for all v and this means that p' is negative for The intermediate case with T 8a/27bR =
nil values of v, whereby the function p (v
is = plays an exceptional role: the temperature T
constantly decreasing as v grows (the upper two in this case is called the critical temperature
van der Waals isotherms for carbon dioxide of the gas (it is usually denoted by T c ). The
in Figure 7.3.8; the temperature at each curve point of inflection C of the respective isotherm
is given in degrees Celsius). The respective corresponds to the following values of volume
•curves of p vs. v resemble hyperbolas, and the and pressure: v =
3b (—v c ) and p a/27b 2 =
asymptotes are the straight lines p 0 and = = Pc)? these volume and pressure are also
v = b.
(
called critical. At T T c if = p c (or v p> <
(b) 8a/27b — RT > 0,
that is T 8a/ < v c ), the substance is in the gaseous state,
,

27bR, or T/(a/b) >> 8/27 R (see the lower two and if p <
p c (or v >> v c ), the substance is in
curves in Figure 7.3.8). Since at v b the = the liquid state; thus, p c is the highest pres-
function / (v) is negative (it is unimportant sure at which the substance can exist in the
that there is no such state of matter with form of saturated vapor. Note also that point C
v = b) and at v —
3b we have / (v) 8a/27b= — of the critical isotherm (each substance has
RT> 0, between b and 3 b there must be a only one critical isotherm) corresponds to
value v 0 of v such that / (v0 ) =
0 and, hence Pc v c =a/96 =
(3/8) RT
C which is only three-
p' (vq) =
0. For this value the function p = ,

eighths of the quantity predicted by the ideal

p (v) is at a minimum (for smaller values of v gas law.
the derivative p' is negative and for larger va- A study of the van der Waals equa-
fuller
lues it is positive). This is not all, however. tion, consequences, and further modifi-
its
As v becomes greater and greater (v -> oo), cations can be found in textbooks on physics
the value of p diminishes (in view of (7.3.8)), and engineering thermodynamics.
and so somewhere at v vx = >
3b the function
p = p (v) must attain its maximum. We see
that while in case (a) the van der Waals iso-
therms have neither maxima nor minima and 7.4* Convex Functions and Algebraic
resemble hyperbolas (ideal gas isotherms), in Inequalities
case (b) their nature changes: here the func-
tion / ( v) (=p) has a (local) minimum and In Section 1.4 we defined the convexity
a (local) maximum. The asymptotes of the p
of a curve y — f (x) (or a function y —
vs. v curves in case (b) are the same as in case
(a). f (#)) in the following fashion: a function
(c) 8a/27b — RT =
0, that is, T 8a/ = y =f {x) is said to be convex on the
27 bR, or T/(a/b =
8/27 R. Since the greatest interval a x ^ ^
b of variation of x
possible value of f (v) here is zero, for all other
values of v this function is negative, which
(or curve y =
f (x) is said to be convex
means that p' = —
b) 2 is negative, too.
upward, or simply convex) if any chord
f (v)/(v
Thus, the function p p =
(v) is everywhere a AB of this curve (A A {x±, f (o^)) =
decreasing function, just as in case (a). At
=
and B =B {, a 2, xi/ (# 2 ))> ^ <
v 3 b it has a point
— 0).
of inflection (since x2 ^ b)below the corresponding
lies
P'(3b)
In the case (a), obviously, to each value of arc AB
of the curve y f (x) (Figure =
pressure p there corresponds a single value of 7.4.1a). On the other hand, if a chord
 ~

7.4* Convex Functions and Algebraic Inequalities 235

x = constant

(6)

Figure 7.4.1

AB lies above the corresponding arc AB

of the curve y =
/ (.z), the function /
is said to be concave (or the curve y = instance, from this convexity con-
/ (x) is said to be convex downward or dition it immediately follows that the
= —
,

simply concave) (Figure 7.4.1&). In other functions y In x, y =

x 2 and y ,
=
words, the function y = f (x) is convex —1 lx (for x 0) are >
convex since their
on the interval a ^ ^
x b if the point second derivatives are respectively,
N at which the straight line x = const- y
" = (1 lx)'
= -Ux\ y" = (- 2 x y =
ant [a xt ^ < < ^
x x2 b) intersects —2, and y" = l/x (
2 )' = —2 lx (Fig-
3

the chord AB lies below the point M ure 7.4.2; the reader will recall that
where the same straight line intersects the function y —
In x is defined only
the curve representing y = f (x)\ on for x >
0).
the other hand, if N
lies above M, the There exists a simple but important
function / is concave This definition
. Theorem 1 If y =
/ (x) is a function
already relates the concept of con- that convex on the interval from a to b
is
vexity with inequalities, since if the and x 1 and x 2 are two values of the
points M
and N
have coordinates independent variable within the interval
(x y) and (x Y), the convexity of y
, ,
— (i.e. two arbitrary numbers such that
f x ) implies y >Y, while the con-
(; a x x <L x 2 6), then
cavity of y —
f {x) implies y <7.
In Section 2.7 we established a / (^i) t~/ te)
^ j ^ ^
simple condition for the convexity
of a function y =
f (x), namely, that Proof . In Figure 7.4.3, OA — x 1 and
a function is convex if its second deriv- OB = x 2 Then
. AM = / and BN =
(xf)
ative y”
,
=
d*y!dx 2 is negative
,
For .
f (x 2 ). Moreover, if S is the middle of
 =

236 7 Investigation of Functions. Some Problems from Geometry

(b) y — —x 2
. Reasoning in the same
way as in (a), we get

*L+*2 xl + x2\ 2

2 ^ f
l 2 ) ’

or, in another form,

X\ +X 2 X1 +X 2
^ ^
Figure 7.4.3
f x\+X. 2
2 X 1 +X 2
7
l 2 2

the line segment AB ,

OS then = f a\+4+- -••+&%
rp, .

(x x -f-^ 2 )/2 and, hence, SP = f( ~^~ .

the expression 1 / ,

On the other hand, since the length which is the square root of the arith-
of the midline SQ of the trapezoid metic mean of k squares of the numbers
ABNM is equal to one-half the sum a 1? a 2 ., a is known as the
. . root-
AM
, ft ,

of the lengths of the bases and2?iV, mean- square of these numbers. Thus,
we can write SQ [/ (xj) -f / (x 2 )]I2. = the above result can be formulated as
But according to the definition of a follows: the root-mean-square of two
convex function, the midpoint Q of the distinct positive numbers is always greater
chord MN
lies below point P of the than the arithmetic mean of these num-
arc MN
hence ; bers.
(c) y = — 1 lx. Theorem 1 implies
f (
X1) + / (
x2 ) ^ 4 l
x l~\r x 2 t
’
that
2 2

which is what we
\ /

set out to prove. - 9

—— \x—2
(
x
-j-
^ fa +Lx
x2 I
<4
2 )/2
’

Examples.
= nr
1
(
1
^
+ *2P
1
\
2
In x. The convexity of this
!

y (a) 2lli x l+ x ’
2
function implies that
or, finally,
In xi ^^
+ In x ^ m 2 xt +x 2
+X
^^
*
2 Xt 2
2 2 (7.4.3)
*1
lfa + ifa 2

i.e. In Y x x < ln-i 2 ^* 2

,

The quotient 1 + laJ+V a*+

i
.
-
-+V a *
or. finally,
"’ hich ls
(
i/«.+i/n .+. ..+</«, )
V t (7.4.2) simply the arithmetic mean of the
numbers l/a l7 l/a 2 ,
. . ., lla k (k reci-
which demonstrates that the geometric procals of the positive numbers a x a 2 , ,

mean of two distinct positive numbers . a h ), is called the harmonic mean

is smaller than their arithmetic mean. of the numbers a x a 2 ., Thus, . .
, ,

the inequality (7.4.3) states that the

7 9
In proving Theorem 1 (and all subse-
-
harmonic mean of two distinct positive
quent theorems in this section), we confine numbers is smaller than the arithmetic
our discussion to the case where / fa) and
same sign. We advise the rea-
mean of these numbers.
f (x 2 ) are of the
der to consider the case of opposite signs on Theorem 1 can be generalized as
his own (instead of employing the property of follows:
the midline of a trapezoid, the reader should
use the following theorem: the length of the
Theorem 2
If a function y / (x) is =
convex interval from a to b
in the
segment of the midline of a trapezoid lying
between the diagonals is equal to ’one-half the and x1 and x 2 are two values of the
difference of the lengths of the bases). independent variable within the interval
 i x

7.4* Convex Functions and Algebraic Inequalities 237

AM = f (x x ), and BN = f (x 2 ). Accord-
ing to what we have just proved, the
coordinates of the point Q that divides
Figure 7.4.4
MN in the ratio q -f- p MQ QN =
are (px1 +
qx 2 pf (x^ qf (z 2 )); thus, +
^
,

{a ^ x 1 <T x 2 b) and p and q are two the length of SQ in Figure 7.4.5 is equal
arbitrary positive numbers whose sum is to pf (x x ) +
qf (x 2 ) and that of SP is
equal to unity then ,
/ (px i +
qx 2 ). But since y / (x) is =
pf fa) + Qf M< f (px + qx 2 ). a convex function, point Q lies below
point P, which means that pf (x x ) +
(7.4.4)
qf (^ 2 ) / (px 1< qx 2 ), which is what +
(For p = q = 1/2 this theorem trans- we set out to prove. 711
forms into Theorem 1.) Examples .

Proof. First of all, we note that if M (a) y = In x. In this case (7.4.4)

and N
are two points with coordinates
(xx y ± ) and (x 2 y 2 ) and Q is the point
, , yields
that divides segment in the ratio MN In x x + q In x 2 C In (px + qx
MQ -f- QN =
p, with p ? q -T- + = 1> p x 2 ),

then the coordinates of Q are (pxx + qx 2, whence

pi/i + g*/ 2 ). Indeed, if we denote by
x\x\ < px + qx > >
Xu X 2 and X and Y Y2 and Y the 1 2, p 0, q 0,

projections of
,

Q
x,
points M, N,
,

and p +q= i-

on the x and y axes (Figure 7.4.4), then (b) y = — 2

. We have
points and X Y
divide the segments
— px\ — qx\ < — (paq + ^ 2
XX1 2 and 1 2 in YY
the ratio q -r- p.
2) ,

This yields 710 or p^ + ^ >(P^i + ^ 2 2)

2
»

r
OX = OX + X X x x or p^ + ^ >P^ + ^
]/ 2 1 2>

= Xl + q {x — x 2 x) where p > 0, q > 0, and p + </

= 1.

= (1 — q) x + qx x 2 = px1 + qx 2 (c) y = — 1/#. Here we have

and
*1
7
*2
<
^ P^i
1

+ 7^2 '

oy = or 2 + y 2 y 1

= y2 + p —^ ( 2/1 2) a:i
P
£2
7
^ P^i +^ 2
’

= (1 — p) + P0i = py 1 + qy 2. 2/2 7 n
As the reader can easily see, the coordi-
*

nates of any point belonging to the segment

We return now to the graph of our MN
convex function y = f (x) (Figure 7.4.5).
can be represented in the form (px ± +
Let us assume that OA = x ly OB = x 2
qx 2 py x
, qy 2 ), +where the numbers p
and q (which are different for each such point)
+ —
,

are such that p >* 0, q^> 0, and p q 1.

. 10
7
in Figure 7.4.4 we depicted the case Thus, the inequality (7.4.4) states that the
where all four numbers x x x 2 y x and y 2 are , , ,
entire chord MN
lies below the curve y =
positive. The reader is advised to consider the / (x), which is equivalent to the function
alternative cases himself. being convex.
 -

238 7 Investigation of Functions . Some Problems from Geom e try

side MM 1 2 of this k-gon (M 1 Q 2 -f-

Q 2M 2 = 1/2 -T- 1/2), Q3 is the point

that divides in the ratio 2 -r- 1 MQ 3 2
(M 3 Q 3 2/3
-f- Q 3Q 2
1/3), Q, is the = ~
point that divides M^Q 3 in the ratio
3 -7- 1 (M 3 Q 4 -s- Q,Q 3 3/4 -f- 1/4), = . .

. .,
. and, finally, Q k is the point
that divides MkQk^ in the ratio
(& —
1) -M, that is, kQ k Q kQ k ^ M ~ =
(ft -
l)/& -r- 1/ft.
Point is known as the centroid
(the center of mass) of the k-gon X 2 MM ,

. . ,M k . In
the case of a triangle,
MMM1 2 3 (Figure 7.4.66) the centroid

Q 3 coincides with the point of intersection

of the medians of the triangle; indeed,
in this case Q 2 is the middle of side
MM x 2 the segment , 3 Q 2 is a median, M
and the point Q 3 that divides 3Q 2 M
in the ratio 3 Q 3 -f- Q 3 Q 2 2 -f- 1 is M —
the point of intersection of the medians
of the triangle.
Let us prove that if the coordinates
Figure 7.4.6
of the vertices M M x, 2, . . ., M k of a
k-gon are fo, y x ), (x z y 2 ), ,
. . (x h , y h ),
then the coordinates of the centroid Qk
where x x > 0, x2 > 0, p > 0, q > 0,
are (fo x2 + +
x h )/k, {y 1 + y 2 + • . . .

and p q + = 1.
. .
+ y h )/ky 13
.

Theorem Indeed, by the assumption we made

1 can also be generalized as
at the beginning of the proof of Theorem
follows:
2, the points Q 2 Q 3 (? 4 ., Q k have
Theorem 3 If y f (x) is a function = the following
,

coordinates:
, ,
. .

convex on the interval from a to b and

#i, x 2 x k is a set of k values of the
,
. .

independent variable within this interval

such that some are distinct then ,
Vz V
2 ’
2
— ;
’

n 2 x i+ x 2 1 2 yi+y 2 1

f (Sl) + / (^2)+ « • +/ (
x k)
l
l 3 2
"i"
,

3
31
3 2 +3
.

y3
\

)
f

k
0r n (
X1 + X + ^3 2 ^l + 2/2+J/3\ .

*!+*«+:•_+ £* 3 3
<
’ ’
l 1
/ (7. 4 . 5 )
( )
n /3 tj +*2+^3
“T 4
1
1
X4
^4 [ 4 3
;

(a particular case of Jensen's inequal

3 yi+yn+y3 „
ity 712 ). For k 2 Theorem 3 trans- = T 3
,

h T
1
J/4
\
/
>

forms into Theorem 1.

Proof, We
start by defining a con-
cept that is often used in geometrical 7 13
This implies, for one, that the centroid
*

and analytical problems. Suppose that of a &-gon is determined solely by the &-gon
MMM
1 2 3 ,M k is an arbitrary /c-gon
. .
and does not depend on the order in which
the vertices of the k-gon are numbered (con-
(Figure 7.4.6a), Q 2 is the midpoint of
trary to what we may assume from the defini-
tion of a centroid). In the case of a triangle
7 12
Johan Ludvig
-
William Valdemar this also follows from the fact that the centro-
Jensen (1859-1925), a Danish analyst, algeb- id of a triangle coincides with the point of
raist, and engineer. intersection of the medians.
 1 X x

7 A* Convex Functions and Algebraic Inequalities 239

(see Figure 7.4.7). But the centroid

of a convex fc-gon lies inside the k-gon
(this follows from the very definition
of a centroid); hence, point Q lies below
point P and
/(Sl)+/(*2)+---+/(*fe)
k

<Cf[
+
j

which is what we set out to prove.

This line of reasoning remains valid
when some (but not all) points x M ,

or M 2 , ., . . M
k coincide (not all of the*

n / zi+*2+z3+*4 yi+y 2 +y 3 +yi numbers x x x 2 ., x k are distinct)

, ,
. .

l 4 ’
4 and the fc-gon degenerates into a poly-
gon with a smaller number of vertices.
Examples.
T=1 f T Xh '
(a) y = In x. From Theorem 3 it
follows that
k — 2/1 + + 2/ 2
• • •
-\-yk-i 1
„ \
k— + ln #2+
,

k 1 A:
’ In xx . .
.
+ln x k
k
or
x1 + x + ...+oc k
^
, 2
j
xi +#2+ . .
. +^fe-i +#fe k
l A:

yi+^2+ • •
»+yfe-i+y/i \
’
or Y x^x 2 . . .x h < ^

A /
that is, the geometric mean ofk
positive
Let us now return to the convex numbers some which
are distinct is
function y = f x ). Suppose that x (. M smaller than
of
the respective arithmetic
M
,

M 2, . • ., h are k sequential points mean This is

. known as the theorem
on the graph of this function taken on geometric and arithmetic means .

within the interval from a to b (Figure (b) y

2 =—
In this case we get .

7.4.7). Due to the convexity of the

function, the fc-gon .M h is .
x i+ x l+ •••+*!
convex and
lies entirely under the k
curve y =
/ (x). If the abscissas of the ^1+^2+
X-j -\~
-}-Xkxh 2
M M ^ 2 • . .

points 2 Mu ., k are x2 , . .
,
. .
(
^
\ k j
x k then their ordinates are, obviously,
,
or
/ Ui> / (x 2 ), ., / (x k ). Hence, the
• •

coordinates of the centroid Q of the /c-gon x\ -f- #1 “h • • •

~h x \ xj+x 2 + 11 +£fe
MM 1 i. . .M h are 3 ,+ .
+ • • • +*fe
,
y k ^ k
r

(
/(xi) + / (»,)+••+/(«>) with the result that is, the root-mean-square of k positive
) numbers some of which are distinct is
that greater than the respective arithmetic
mean.
Os — =—
(c) y l/;r. In this case Theorem 3
o/^) / (si)+/ (^2)+ • • H~/ (^fe)
yields

and
SP = f( k ) xi + x +-..+**
2
1
 + M . = ^ l ) p ,

:240 7 Investigation of Functions. Some Problems from Geometry

that is, P2 Pi
P1 +PP1
,

P2
.

+ , , Q3 is a point
F
1 / J_ .
J_ + , J_\ of the line segment
2
MQ 3 2 such that

MQ Pa P1 +P 2
> X +X 1 2 +
*
• • • + xk 9
3 3 *
Q3 Q2 -

PI+P 2 +P 3 *
P1+P2 + P3’
whence Qk is a point of the line segment MQ k z
such that
^ x l~\~ x 2~\~ Xk

^Q
• • •
~f~

+ 1/^2+ -+l/ x h ^ k
^

M
1/^1

that
••

harmonic mean of k positive

the
,& Pl+P2 ^ 3+P4
is,
numbers some of which are distinct is _i_ +P +P Pi 2 3

P1 + P2 + P3 + P4
* '
’

smaller than the respective arithmetic

mean. finally, Qk = Q is a point of the seg-
Finally, we will prove a theorem that ment MkQk-i such that M Q ~ QQh-i
k
generalizes Theorems 2 and 3. — Pk -T- (Pi + +P2 • • • + Pk- 1) (ifPi =
Theorem 4 Suppose that y f (x) = p2 = . . . = pk = 1/ft, then () is the
is a function convex in an interval from centroid of the fc-gon 1 2 k ). MM . . . M
a to b and x x x 2 x^ are k values , , . .
Using the proposition from which we
of the independent variable some of started the proof of Theorem 2, we can
which are distinct and taken within the find the coordinates of the points
interval and p x p 2 , ,
. . ph are k positive Q 2? Q 3? Q 4 »
• • •» Q :

numbers whose sum equal unity

Then
is to
r\
V*
( Pl + P X Pi/(*i) + p (*
xl 2 2 2f 2)

1 +P P1 P +P 2 1 2

pj (*i) p 2 f te) + •
+ Phi fo) / P +P Pi x i~\~P x
1 Ps 2 2 2
4- ^3?
Vpi + P2 + P3 Pi + P P +P +P
3

< / {Pl x i + P2 X2 + •
•
+ Ph x h) (7.4.6)
P +P Pit X + P 1
2

2
1

{ 1 )
2 3

2/ (« 2 >

(the general case of Jensen s inequality). P +P + P P +P 1 2 3 1 2

For k =
2 Theorem 4 transforms into
n
rp
]cneoi‘em z,“SVh ne *ior p x —~p 2
1 i
= . . . pi-rP2-rP3 /

= Pk 1 Ik it = transforms into Theor- or

em 3.
Let us once more take the
Proof. <?3
+ p + p3 (Pl*l +* 2 ** 4- P 3* 3 )
graph of a convex function y = f (x) ( pi 2

and a convex fc-gon X 2 k MM . . .

1 /»i/ (*1) + Ptf (*2) + 3/ (®s)]);
inscribed into this graph. The vertices Pl+P2 +P 3
of the /c-gon have the following coordin-
ates: (x x / (x± )), (x 2 ,
, f {x 2 )), . .
0 / Pl^l + P X + + Pk- x k-l + Pk x k
2 2 •

(x k ,f (x k )) (Figure 7.4.8). Suppose that \ P +P + 1+Pfe -1 + 2 • • • Pfc

Q2 is a point of the side MM 1 2 of this Pit ( X i)+P 2 f fa)+ • »

+Pfe-i/ ( x k-i JrPk 1
MQ
•

fc-gon such that X 2 -f- QzM 2 = Pi +P + 2 • • +PA -1 + PA. )

or, in another form,

Q + P2 X + + Pk x
(Pl X l 2 • • • ki

Pit X + Pzf x + ---+Pkf xk)) ( l) i l) ( ,

since Pi + p + .+p k =l. Thus, in 2 ..

Figure 7.4.8,
SQ — p if x i) + 2 / x 2) + + Pkf ( (
• • •
i
x k) 1

OS -= P& + p 2 x 2 + + pk k
x . .
. ,

SP = f (Pi x + P 2 x 2 Jr ...+P/A). i
 »

7.5 Computing Areas 241

Since point Q lies below point P (the whence we can easily obtain
entire k-g on 1 2 ,M k lies under MM . .

the curve y / (x), and Q is an interior = Pl/ X l~\~ P 2 / x 2~\~

1
• • •
-\-Pk! x k
point of this P-gon), we have
< Pl x + P2X +
l 2 • • • + Pk xki
Pif ( x ) + Pif ( + Pfe/
x z) i~
Pk ^0
• •
\ • (^fe)
where Pi , p2 ,
and p t +
</ (Pi^i *+- P2 X 2 + • • • + Pk X k) > P2+ •+Pk = I-

which is what we set out to prove. 7 14 *

Exercises
Examples.
(a) y
— In x. In this case we get 7.4.1. Prove that the following functions
are convex:
y — —x > 1 and x >> 0, (b)
(a)
n for n
Pi In x* 4- In x 2 Jr + xk = x m for 0 < m < and x > 0, y =
p2 - • *
Pk In y 1 (c)

< In (p x — l/x k for k > 0 and x > 0, (d) y = — x log

1 i +p 2 x2 -f . .
. + Ph x k )
for x > 0 y =
— x log x — — x) logx
(e) (1
(1 — x) for 0 < x < (the base in (d) and
,

1 (e)
whence, taking antilogs, we have is any number greater than unity).
7.4.2. Write out the inequalities (7.4.1),
x\'xl* ...x^
k
<p X +p X + i i 2 2 . . . +Ph x k, (7.4.4), (7.4.5), and (7.4.6), where the func-
tion / (x) is (a) the function of Exercise 7.4.1a,
(b) the function of Exercise 7.4.1b, (c) the
where p u p 2 ,p >0 . . k and p x + function of Exercise 7.4.1c, (d) the function
Pz+ + Pk • (This is the gener-
• • = 1-
of Exercise 7. 4. Id.
7.4.3. (a) Prove that the geometric mean
alized theorem on geometric and arith-
of two(positive) numbers is the geometric
metic means.) mean of their arithmetic and harmonic means,
(b) y = — x 2 We . have
(b) Derive inequality (7.4.3) using the result
of (a) and inequality (7.4.2).
— PiX\ — p x\ — ... — pkXI 2

< (Pi x + Pzx + + Pkxk) l 2 • • •

2
i
7.5 Computing Areas
or
In Chapter 3 we showed that the value
b
VPi x\ + P2 x t+ • +Pk x l
of a definite integral dx yields
> Pi x + Pz x + l 2 • • • + Ph x k> ^
a
f(x)

the area of a figure bounded from above

where p t , p2 , . .

= 1-
., ph >0 and + by the curve y f (x), from below by =
P2 + • • • + Pk (This is the gen- the x axis, and on the sides by the
eralized theorem on the root-mean-square
vertical lines x a and x 6, or the = —
and the arithmetic mean.) vertical bases of the trapezoid (see
(c) y = — i/x. Here Theorem 4 yields Figure 7.5.1; for the sake of simplicity

_Pl P2__ Ph
we assume that / {x) 0 and a <C b). >
Thus, being able to evaluate definite
Xc^ Xfa

Pl X l
9
4" P2X 2 Pk X k
7 14
The reader can easily see that the
•

coordinates of any interior point of the /c-gon

MM1 2 .Mh can be represented in the form
. .

(Plfl + P 2*2 + + Ph xk Plf(x + • • •

1 l)
Pzf ( x 2 ) + •
+ Pkf x h))< where p u p
• •
( 2,
are positive and p + p +
— Pk
. •
± 2. . .

Pk 1- Thus, inequality (7.4.6) expresses

the fact that a polygon inscribed in the graph
of a convex function always lies below that
graph.

16-0946
 242 7 Investigation of Functions . Some Problems from Geometry

Figure 7.5.2

integrals enables us to use standard

techniques in computing various areas, from below by the x axis, and on the
whereas elementary mathematics only left and right by the straight lines
allows for calculating the areas of
rectilinear figures (polygons) and also
x = x 0 and x = A (here A x 0 Fig- > ;

ure 7.5.3). This area is

of the circle and some of its parts
A
(segment, sector).
Let us find the area of a figure bounded SA — ce~
x/ a
dx= —cae~ xt a \x 0
j
from above by the curve of the power x0
function y c x
n (c =
0 and n 0), > > = ca {e-*« — e- Ala). ,a
(7.5.4)
from below by the x axis, and on the
right by the vertical line x x 0 (we = If A great compared with £ 0 then
is ,

assume that x 0 >

0; in Figure 7.5.2,
e-xja e A/a t n w
iH b e seen f rom
n —
2 and c 0.25): = (7.5.4) that increasing A hardly at
all changes S A As A increases without
.

n CXn +l x0 bound, the value of e A ! a approaches

cx dx (7.5.1) zero without bound. And so we can
rc + 1 0 rc + 1
speak of the area of the figure in Fig-
Let us rewrite formula (7.5.1) as ure 7.5.3 as being unbounded on the
right: the unlimited figure obtained
cx>°' from the figure in Figure 7.5.3 as
n + i
A oo has a finite area
or, since ex™ = y (x = 0) z/
0 , oo

Soo ~ dx — cae' x»l a — ay0

xf a
S— (7.5.2)
ce~ , (7.5.5)
w _|pj" V o^o* j
3C 0

Since the quantities y and x have the where y 0 = y =

(# 0 ) ce~ ^ .
x a
dimensions of length, from (7.5.2) it In formula exponent(7.5.3), the
follows that S is indeed measured in must be a dimensionless number, which
units of area (the square of the unit means that the dimensions of a must
of length). (We see that the area S coincide with those of x, that is, a
is, as to order of magnitude, y 0 x 0 and , has the dimensions of length. The
differs from this product solely in the dimensions of y and S are those of
factor 1 l(n 1), which, +
as to order length and area respectively.
of magnitude, is close to unity for n It turns out that the area under one
not too large (compare with (5.6.6), arch of a sine curve (Figure 7.5.4) is
where y0 / max x0 b = a, and ,
= —
l/(n +
1) is a dimensionless factor of
the order of unity).
In the next example, we find they
are bounded from above by the exponen-
tial curve

y = ce“*/ a , e> 0, a > 0, (7.5.3) Figure 7.5.4

 n

7.5 Computing Areas 243

which means that the area under one

arch of a cycloid is three times the area
of the circle that generates the cycloid.

The results (7.5.2) and (7.5.5)-(7.5.7) can

also be formulated in a different manner.
Figure
The area of the hatched curvilinear triangle
7.5.5
OAM in Figure 7.5.2 is (n +
i)~ l OA -AM —
OA [(n +
l)- 1 AM], that is, coincides with
the area of the rectangle with base GA and
expressed very simply. Indeed, this
area,bounded from below by the line
altitude ( n +
i)~ x AM. In view of this it is
sometimes said that the effective altitude of
segment of the x axis from 0 to n, is our curvilinear traingle (or the curve y =
cx n
si
within the limits x =
0 to x —
x 0 ) is one-

— cosx — 2. (n+ l)th of the real altitude AM. Here the

*5=^ sin xdx = 1^ (7.5.6) effective altitude is understood to be the alti-
tude of the rectangle with the same base as that
o
of the curvilinear triangle and the same area.
Similarly, the effective altitudes of the curvi-
It is not difficult to find the area
linear trapezoids depicted in Figures 7.5.4
under one arch of a cycloid (Figure and 7.5.5 with bases n and 2R ji, respectively,
7.5.5; see also Section 1.8). If the curve are 2 /ji (^0.6366) and 1.57?, while the real
is given parametrically, x cp (t) and = altitudes, or the values of y max are 1 and 2 7?,
respectively. Finally, the effective length of
,

y = Section 1.8), the for-

(7) (see
the infinite figure depicted in Figure 7.5.3,
mula for the area of a curvilinear a figure bounded by the exponential curve
trapezoid assumes the form y = ce~ x l a the x axis, and the straight line

b
x — ,

x 0 is equal to a, since the area of this

,

p figure is that of the rectangle with base a and

S= ^ydx = \y(t)-§ dt, altitude y 0 .

The concept of effective length (or effec-

a a tive altitude) of a curve often proves to be use-
where a — x (a) and b x (P) are the = ful. For instance, it can be shown that the
(infinitely long) bell-shaped curve in Figu-
endpoints of the segment considered. At re 7.5.6, y —c_exp (—x 2 /a 2 ) restricts a finite
x — R(t sin t) and y— i?(l cos t), = — area S — caY in other words, the effective
with a 0 =
and P =2n, we have width of this curve, whose altitude is y maX =
2n y (0) =
c, is equal to aY ~ 1.77a (see [15],
Section 3.2).
*5=
j
R (1 — cos R (1 — cos t) t) dt
0 Let us determine the area S of an
2Jt
ellipse with semiaxes a and b (Fig-
= i? 2
^
(1 — cos t)
2
dt ure 7.5.7). Of course, this area can easily
o be found by methods of elementary
2jt
geometry (see the text below printed
= jR 2
j
(1 — 2 cos + cos t 2
1) dt in small type); however, we prefer using
o
a more standard method (although in
2jx 2jt the given case a more cumbersome one)
=R 2
j
dt — 27? 2
^
cos t dt
o o
2jx

+R 2 l+ gg.s _2* dt=R 2 t If

.

j
0

— 2 R sin 2
t ~ {t ->r ~ sin 2
= i X 2 n — 2 Z?
?

7?2
2 2
x 0 + 4x 2 it

+ 4x0 = 3ni? 2
, (
7.5.7)
Figure 7.5.6
16 *
 a

244 7 Investigation of Functions. Some Problems from Geometry

An ellipse with semiaxes a and b (a^> b)

is obtained from a circle of radius a by shrink-
ing the latter to the x axis with a ratio of
k =
bla (see Figure 7.5.7). It is easy to see
that such shrinking transforms a figure F
of area s into a figure F' of area s' ks. In- =
deed, a grid of small squares with side 6 and
area 6 2 (the sides of the squares being parallel
to the coordinate axes) is transformed by the
shrinking transformation into a grid of rec-
tangles with sides 6 and kb and area k6 2 each.
But a grid of equal squares (a measuring grid)
is used to measure areas: if a figure F is covered

Figure 7.5.7 by N squares of the grid, its area s is approx-

imately N6 2 On the other hand, the figure F',
.

obtained through shrinking figure F in the

above-described manner, will be covered by
involving integral calculus. Note that N rectangles of the transformed grid; there-
symmetry fore, s' ~
iV/c6 2 whereby s' ks (since 6 =
by virtue of it suffices to ,

can be chosen as small as desired, so that the

find the area iS^of that portion which approximate equalities s ~ N6 2 and s' ~
lies in the first quadrant and then Nkb 2 can be assumed as precise as desired).
multiply it by four: S 4 S x To com- = .
And since the area of a circle of radius a is
equal to na 2 the area of the ellipse with semi-
pute S x we find y from the equation of ,

axes a and b (the ellipse being obtained by

the ellipse x /a y !b
2
1 (see
2
Sec- + 2 2 = shrinking the circle with a ratio k bla) is =
tion 1.7): y (
bla ) a2 =
x 2 (here the Y — kna 2 =
(bla) na 2 nab. =
square root is understood to be the
positive root, since in the first quadrant
Note an important circumstance. In
positive). Thus,
Chapter 3 we already pointed out that
y is
the area (the integral) can be either
a positive or negative. This calls for a
S i
=— ^
]/* a 2 —x 2
dx . (7.5.8) certain amount of care when finding
o
areas. Suppose we want to know the
amount of paint needed to paint an
The integral (7.5.8) can easily be found area bounded by two arches of a sine
by making the change of variable curve, from x = 0 to x = 2ji, and the
x — a sin t. This yields x axis (see Figure 3.5.2) if unit area
a Jt/2 requires a grams of paint. As was shown
\y 2 —x 2
dx = [ a Y l —sin 2
1 a cost dt above, one integral cannot be used to

h o
compute the entire area. have to We
take separate integrals over the inter-
jt/2 n/2
vals from x 0 to x =
n and from =
----
j
a 2 cos 2 tdt =a 2
1
^~^QS — dt x =
Jt to x 2 ft. —
o o Generally, if the integrand y f (x) =
= 9 r t

+
.

—
sin2n^/2
Jo “T-
na 2
(7.5.9)
changes sign then to solve the problem
,

in paint consumption, so to say, we

must split the interval of integration
into parts in which / {x) preserves sign,
Using this, from (7.5.8) we get then evaluate the integral over the
separate parts, and finally sum the
c b na 2 nab
absolute values of the resulting in-
tegrals.

The area of the entire ellipse is S —

nab. If a b r, —
then we have S = = Exercises
nr 2 (the area of a circle) in complete
7.5.1. Find the area of a figure bounded by
accord with the fact that for a b — r — the x axis and a single arch of (a) y sin 2 x =
an ellipse becomes a circle. and (b) y —
cos 2 x. [Hint. Draw the graphs
 J 1

7.6 * Estimating Sums and Products 245

of both functions and take advantage of the

formulas sin 2 x 1/2 =
(1/2) cos 2x— and
cos 2 x 1 — —
sin 2 x ]
7.5.2. Find the area of a figure bounded
from above by the curve y x (1 =
x) and —
from below by the x axis.
7.5.3. Find the areas into which the para-
bola y = 2
(1/2) x divides the circle x
2
+ y
2 =
8.
7.5.4. Find the amount of paint needed
to cover the area of a figure bounded by (a)
the curve y x/(i = +
x 2 ), the x axis, and the
vertical lines x 1 —
and x = — 1 and (b)
the curve y x3 = +
2x 2 — —
x 2 and the x
,

axis. [Hint. First construct the graph of the

function y x3 2x 2— + x— — 2.)
7.5.5. Find the area of the ellipse x 2 /25 +
y
2
/4 = 1.

7.6* Estimating Sums and Products

In Sections 3.1 and 3.2 we introduced

the concept of a (definite) integral as with 0 >k>— 1 as an exercise. 715
the limit of a sequence of certain sums. The area s of the curvilinear triangle
By calculating these sums we can estim- OA n M n depicted in Figure 7.6.1 is,
ate the integral, or the area of the obviously,
appropriate curvilinear trapezoid (see
the method of rectangles and the trapezoid x h+i n nk+i
(7.6.2)
method described in Section 3.1). But k + 1 0 k + 1
*

the relationship that exists between

integrals (or areas) and (finite) sums On the other hand, the method of
can be exploited in the opposite direc- rectangles yields
tion, for estimating sums by means of
integrals. This simple idea (see also
i + 2 + ... + nh >s
k h

[15], Section 1.2) will be illustrated and 0* + 1* + + (n — l) < . .

.
fe
s,
by several instructive examples. that is,
We
start with the sum of powers of
h
natural numbers :
l + 2h + + nh . .
.

nh+i
SW = l* + 2 + 3 + fe ft
...-j- nk , (7.6.1) > z+l >1* + 2*+... + (ji-1)*.
where k 1. >—
Let us consider the or
function y —
x h Figure 7.6.1 depicts 72& +1

+ <sy
\
(
the case with k positive, while the k
reader is advised to consider the case
= + 2 +... + n
lft ft ft
<^-r + re
fe
.

(7.6.3)
This implies that for large values of n
we have
72* +1
S (k) = l
k -
4- 2h -

k+ 1 ’

(7.6.4)

7 15
For k
*
<
0 we must consider not the
curvilinear triangle OA n n but the curvilin- M
ear trapezoid A A n n similar to the one M M
Figure 7.6 depicted in Figure 7.6.2.
 ) +

246 7 Investigation of Functions. Some Problems from Geometry

in the sense that for n^> 1 the ratio that is (compare with (7.6.5)), a <Z
S^l yj j - differs but little from unity, Sn — 1 In> —
and cr Sn 1, whence

since a + — = In n + — < S n
n n 1

< + 1 (7.6.3a) <lnn+l — cr-fl. (7.6.8)

Thus,
(these inequalities are obtained if we
divide all members in (7.6.3) by ^n — 1 + y + yH- •••+“ = In + Yn»
k+1
n /(k + 1)).
(7.6.9)
where the number y n lies within the
From (7.6.3) follows a more precise esti-
interval
mate of the sum sty than (7.6.4):

= + 2'‘+...+«fc~ Y<Vn<l* (7.6.9a)

S)
ft)
l
lfc —L-rcH+i
rC —p 1
+ crefe,

Since, as it can easily be seen, y n

with 0 < < c 1. Indeed, it can be proved that is the difference between the area of
for k > 0 we have the square 2 M'2AA M
1 (equal to
unity)
and the sum of the areas hatched in
+ (7.6.4a)
Figure 7.6.2, it monotonically decreases
as n oo and tends to a limit that
in the sense that for large values of n the can be estimated if we find (approxim-
ratio [s n ( ft
— (i^~[ nk+1 + Y nh ) ]/” ft
is ately) the hatched area in Figure 7.6.2
for a fixed value of n that is not too
very small.
large; this limit, y ~ 0.577, is known
as Euler's constant Thus, for n 3> 1,
.

The behavior of the sum we have the following approximate for-

mula:
-l+ff- -!+... 1
(7.6.5)
5n =i+Y+-|+ ...+Y— inre +v
~ lnra-f 0.577. (7.6.10)
for large values of quite different. n is
The graph of the function y ilx, to = The next example is the sum
which we naturally turn, is the hyper-
bola (Figure 7.6.2). The area of the
P.6.11)
curvilinear trapezoid n n bound- AA MM The experience we have gained in the
ed by the x axis, the hyperbola y = process of studying the sums (7.6.1) and
llx, and the straight lines x 1 and =
(7.6.5) tells us that in this case it is
x —n, is given by the following for-
expedient to start by estimating the
mula:
area t of the curvilinear triangle bound-
n
ed by the curve y =
(In x)/x, the axis
<r-
j -^ = lntc I? = lnra, (7.6.6) of abscissas, and the straight line
1 x =n (Figure 7.6.3); we compare t

since In 1 0. On the other hand, =

by the method of rectangles (see Fig-
ure 7.6.2),

i
1 + T + T+ +
_i
(7.6.7)
2 Figure 7.6.3
 e

7.6* Estimating Sums and Products 247

with the sum T n . The area % is given no less than two, the maximum coin-
by the formula cides with the value of the function
n n at one of the endpoints of the specified

t= dx— In x d (In x)
interval. We
arrive at the following
^ ^ estimate for the area t:

In
1

n
u du =
ln
”,-4(lnn)* (7.6.12)
^
T< —+T+—+—
In 2 ,
1 ,
In 3 ,
In 4

-l 'Y' 0 ^
+ In (n — 1) / „ In rp 1 In n \
, ,

n — 1 l
',

jr
e n ) ’

{here we have introduced a new vari-

and
able, u = In x).
In 1 In 4 In n
Next we use a modified version of ...+
4 n
the method of rectangles. The area r 1

In 3
is replaced with the sum of areas of \

n — 1 rectangles with bases A 1 A 2i 3 J ’

A 2A 3 ,
A n ~\A n (see Figure 7.6.3),
. . that is,
while the altitude of rectangle with the
base Ai^Ai is in one case the greatest
In n
n
<T n < T In 3
3
’

value of the function y (In x)/x =

attained in the interval i 1 x i — ^ ^ Finally we get (see (7.6.12))
the smallest
and in the other is
value of the same function attained y ( 1 nre)
2 -^+~<r„
in the same interval. All this requires
a detailed study of the curve y = <|(l nre)2 + i|i; (7.6.13)

{In x)lx. Obviously, y (1) 0 and y ->0 =

in other words, T n cannot differ con-
as x —>• oo (cf. Section 6.5); let us 2
siderably from (1/2) (In ra) for all va-
find the maximum of the
<y \x) 'Since tiie equation
.
function
huwo nfx m hfu^ :

,
Tn =~ (In 2
+ 8n>
y' (*)= (i^ ^+ ln
*(+)
where, in any case, — 0.368
rc)

— _1
<
— In
6 n < (In 3)/3 ~ 0.367. Here,
1 In x 1 a:

S*
“ X2
~ ^
U since the
difference between t and its lower
has a unique solution, x = e (In x = 1 bound determined by the method of
only at x e), and y' = {x) >> 0 for rectangles, that is,
x <e and y' (x) <0 for x > e, the
graph of the function y = (In x) lx has
7 16
'-('.-T
1
)
(-¥-•)•
the shape depicted in Figure 7.6.3.
At x = e the function has its only
-

can only increase as n grows, (In 3)/3 —

maximum (y max — e ~ x 1/2.7), from — 6 n monotonically grows as n -* oo, but
x = 1 to x = e the function grows, always remains smaller than (In 3)/3 —
and then decreases monotonically. e
-1 ~ 0.735. From this it follows that

Thus, the maximum of the function (In 3)/3 — 6„ tends to a certain limit
y {x) on the interval from 2 to 3 is
as n—* oo, which means that 8 n also
equal to 1/e, while on all other inter- tends to a certain limit as n -> oo (the
vals from i 1 to i with i an —integer ,
limit lies between 0.37 and +0.37). —
This method can be applied for
7 - 16 The
fact that the value x e at which — estimating the value of n\ 1-2-3. = . .

y'(x) =
0 corresponds precisely to the maxi- n. We
write the number In (re!) =
mum of y =
(In x)/x can easily be established
In (1-2-3. . .re) as
without analyzing the sign of y' (x) for differ-
ent values of x that is, simply by common-
,

sense reasoning (compare with what was said

In (re!) = In 1 + In 2 + In 3

at the end of Example 2 in Section 7.1). + . . . + In re. (7.6.14)

 1 i .

248 7 Investigation of Functions. Some Problems from Geometry

= (In 1 + In 2 + In 3 + . .
. + In n)
— yin n = In (n\) —y In n. (7.6.15a)

On the other hand, the area cr n of the

curvilinear triangle is smaller than
the sum S n of the areas of n 2 trap- —
ezoids with midlines A t u i 2, 3, M =
. ., n
. —
1, and altitudes of unit length
(the altitudes are the line segments
with endpoints
1/2, 0) and ( —
(i + 1/2, running through
0), with i
Figure 7.6.4 the same values; the upper lateral side
is a segment of the line tangent to the
To estimate the sum on the right-hand curve y — In x at point ) plus the M {

side of (7.6.14), we consider the cur- area of the small trapezoid with the
vilinear triangle n n bounded by AA M altitude ^4^42/2 of length 1/2 and the
the x axis, the curve y In x, and the = M
midline A 5/a 5/4c (here A 5/4 = (5/4, 0)
straight line x n (n is a natural = and M
5 / 4 is the point on the curve
number greater than unity) (Figure
y — In x corresponding to A 5/4 ) plus the
7.6.4). The area o n of this triangle is, area of the rectangle A n _ 1/2 A n M n Mn- U2
obviously, (here A n _ 1/2 = (n — 1/2, 0) and
n n
Mn -i /2 = (n — 1/2, In n)). It is clear
on — \
In x dx — x In x |? — x d (In x) that
^
l i

n n
S n = -iln4 + Mn2
= x\nx — j x^- = x\nx — |i If l| dx

= (x\nx — x)
i

= n\\in — n -\-
i
+ 1- In 3+ + In (n — 1) f y In . .
.
1 • re

\\
= In 2 + In 3+ + In (n — 1) . .
.

(see Section 5.4). If we inscribe in our

curvilinear triangle M
A 1 A n n the right + In n In
4 — y In re

triangle A 1 A 2 M 2 and n 2 trape- —

zoids A 2A 3 3 2 A A±M 4 M
MM , 3 3, . . .
= In (re!) — y In + -j In 4 re

. . ., An^AnMnMn^ (here A x ==
AjA 2 , . . ., An are points on the x axis (7.6.15b)
with abscissas 1, 2, . .., n and
< on ^ Sn
,

M 2, . . ., M n are the corresponding Since sn ,

we can write
points on the curve y In x), we =
will find that a n is greater than In ( n\ ) — 4
—
the sum of the areas of the right
triangle and the trapezoids, the sum
being equal to < In (
n \
)
— ~ In n -f -j In ,

sn = y In 2 + ~ (In 2 + In 3) or

In >relnre + - -l n re — re+1
+ y (In 3+ In 4) (re!)
2

4- . .
.+4 [In (re — 1)4- In re]

= la 2 + In 3+ . .
.+ In (n — + j In n
1) and In (re!) < re In re
+ y In — re+1. re
 . . 1

7.6* Estimating Sums and Products 7 2491

figure bounded from above by the curve y =

x n e~ x and from below by the x axis for x 0’ >
(Figure 7.6.5; n 4).
18 To evaluate the = -

integral, we use integration by parts setting

e~ x dx dg and x n= /, which means that
—
g — —e~ x and df nxn l dx. Thus, we ob-
~
=
tain
oo oo

xne~ x dx = — x n e~ x nxn^~ 1 e~ x dx.

^
( ) | J° + J
0 0

In Section 6.5 it was established that

Figure 7.6.5 = xn /e x 0 as x — oo Since xn e~ x = O'
x n e~ x
that is for;r = 0, it follows that — xn e~x J° = 0; hence,. |

(
1 — In — + In
)
(
nn ]/ n e~ n ) < In (ft!)
oo

^
xn e~ x dx =n
oo

j
x n ~ 1 e~ x dx y
Le. In = nl n - ±..

~n
< + In 1 ( n n
Yne )
0 o

Applying integration by parts to 7 n _i,

In other words, —
we likewise get I n - 1 (n = 1) In - 2 ,
and so-
forth. For this reason
j//" -jt en n
y n e~ < n\ < n
£ft
n
]/ n e~
n
.

In = n (n — l)(u — 2) . . . 3-2-1 -/ 0 ,

(7.6.16) oo

This formula shows that for large n where I 0 = j* e~ x dx = — e~ x |

= —0+ = 1 1..

n n
the value of ft! is close to C ]/ nn e~ 0

where the number C lies between e ~ ,

Thus,
2.72 and ^475 e 2.43. Here, from oo

the fact that the difference a n —- s n xn e~ x dx

1n =
increases monotonically as ft oo, we j
0
can easily derive, as we did earlier,
= n (n — 1) (n — 2) . . . 3-2- = nL (7.6.19>
that n\iy rin^e' 71 tends to a certain
limit as n oo (this limit lies between For large values of n we can arrive at a
good approximation for I n if we find the va-
Y and
that is, between 2.43 e,
lue x =
^ ma x at which the integrand x n e~ X:
and 2.72). It can be proved that this attains its maximum value. This value is.
limit is equal to 2n Y ~ 2.507. Thus, found from the condition
we arrive at Stirling’s formula 1 17
(
= nxn~ e~x xn — e~x
xn e~ x )' 1
-f- ( )

n\ Y 2nn nn e~ u ,
(7.6.17) = (n — x) xn~ e~x = 0, 1

in which the approximate equality where £ maX — The magnitude of this max-11

y maX = (x e~ max =
imum, obviously, n x is
means that, as ft — oo, the ratio >•
nn e~n If we write our function in the form
.
)

n\iy 2nnn n e~ n tends to unity (and for y = eH where / (x) = n In x — x and ex-
x ), r

n large it is very close to unity). pand / (x) in a Taylor’s series (6.1.18) near the*
value x = £ ma x (=w), we can estimate the
effectivewidth of our (unlimited) figure, that
Stirling’s formula (7.6.17) can be substan-
is, the width of a rectangle whose height is.
tiated in another manner. Let us evaluate the
integral j/max and whose area is that of the figure con-
sidered, or I n This width proves to be ap-
.

oo
proximately equal to j/" 2nn (see [15], Sections.
/„ = ^xn e-*dx, (7.6.18) 3.2 and 3.3). He nce, we arrive at Stirling’s-

0
formula: n\ 2ji n n e~ ~Y
n n for n 1. >
where n is a positive integer. The integral I n 7 18
-
Note that for any (arbitrarily large)
can be thought of as the area of an (unlimited) value of n the quantity y x n e~ x tends ta =
zero as x -*• oo (see Section 6.5; this fact will
7 17
James Stirling (1692-1770), a Scottish
-
be used later); if this was not so, we could not
mathematician. speak of an area equal to I n .
 1 .

250 7 Investigation of Functions. Some Problems from Geometry

7.7* More on Natural Logarithms for calculations. Any number raised

to the zeroth power is equal to unity,
Let us examine the problem of natural
which means that if we substitute
In Chapter 3, the integra-
logarithms
tion
.

power functions led us to the

e = 0 into (7.7.4), we
will have 6
8
—
a? = 1 — 1 =0. Thus,
of
following results:
at 8 0, both =
the numerator and the denominator in
[x n dx = — xn+l -f C for n^= — 1, (7.7.4) vanish.
In this lies the answer to a question
(7.7.1) often posed by students that compare
the indefinite integrals (7.7.1) and
j^L = lnz + C, (7.7.2)
(7.7.2): the right-hand side of (7.7.1)
contains n -f 1 in the denominator,
where, of course, the constant C may be and at first glance it seems to tend to
any number and not the same in infinity asn-> —1, while the logarithm
Eqs. (7.7.1) and (7.7.2).
is finite. But this infinity in (7.7.1) is
But why does the case (7.7.2) differ fictitious, it disappears when we
since
-•so strongly from (7.7.1)? What is the
pass to a definite integral and, hence,
reason for this? At first glance there
it can be eliminated by appropriately
is a contradiction here. Imagine a
selecting the constant of integration C
sequence of power functions, or curves, (e.g. C = — + l) -1 = — l/e).
y = ar\ y =
(rc
y = y =
£ -0 9 £' 0 99
- -

, ,
How
does one go about simplifying
^r*"
1 01
-
,
and y = x* 11 All these curves .
the expression for an area that incor-
lie close to each other, and the curve e
porates a 8 and b ? How should we
'with n = —1 in no way differs quali-
calculate powers with small expon-
tatively from the adjacent curves, for
instance, from the curves with n = ents? Let us write the following identi-
— 0.99 and n = —
1.01. Hence, the
ties:

area under the curves y = x (within = e lna = e 8lna

n
a ,
ae .

fixed limits a x b) < <

must be a
We use the main property of the num-
.smooth function of n that is, it must
•change smoothly as we pass the value
,
ber e that is, that for small
,

r
r<l we
n = —1. But this area is expressed by have e ~\-\-r (to within terms of
the second and higher orders of small-
a definite integral, whose form at
n = — 1 is quite different from that at ness, that is, quantities of the order of
2 3
—
n = — 0.99 and n = — 1.01. r r
,
etc.). If n is close to
,

=— +
1, that
is, n 1 e, where the term 8
To
resolve this seeming contradiction,
is extremely small, we can apply the
we must show that for n close to — formulas we have just written:
the integral in (7.7.1) is close to the
logarithmic expression in (7.7.2). Let
us carry out the following calculations:
£ = — (bs — a E) =— elnb — e« lna
(e )

b
+ e In b) (1 — + 8 In a)]
S= f
— = In-
x
for n= — 1
[(1

j
a
a
= In b — In a = ln-^-. (7.7.5)

and a, 6>0; (7.7.3) This resolves the contradiction. In

6 the limit of small 8 the quantity 8 has
S= dx-=^- (b
z
—a 8
)
canceled out, and we have found that
j
a formula (7.7.1) leads to the same result
for ra = — 1 + e. (7.7.4)
as formula (7.7.2). The area under the
curve y == x n proves to be a smooth
If the exponent n is close to 1, then — function of the exponent n even when
z issmall and the expression (7.7.4) for n passes through the “singular” value
S assumes a form that is inconvenient n — —1
7.7* More on Natural Logarithms 251

We know how to calculate small The area of the strip does not change
powers x z for 8<1. From school in the
process.
mathematics we know that x° — 1. Now This result can easily be obtained
we know how much x£ differs from without applying geometrical methods.
unity if e is close to zero but still Let us introduce the variable z = kx ,

differs somewhat from zero. z = zlk, dx = (1 Ik) dz. For the new
We can approach the integral J — variable the limits of integration will

x
-1
dx from still another angle. Imag-
change, too; namely, x a becomes —
z — ka and x = b becomes z = kb.
ine a person that knows nothing about We have
logarithms. However, we assume that A
b kb Jl. dz kb
he or she knows what integral calculus
is and, confronted with an integral,
z
would like to study its properties. a ka ka
Thus, let us consider the integral kb

b = j
ka
^ = J(ka,kb).
J(a,b)= (7.7.6)
Jf-. (At the end of the chain of formulas we
a
employed the fact that the variable of
A remarkable property
integral of
integration is a dummy variable and
(7.7.6) is that the value of J (a, b) does
can be denoted by any letter.)
not change if a and b are multiplied by
If the function of two variables
a quantity k ( the same for a and b). As for
the area under the curve y x' 1 — = J (a, b) does not change when we mul-
llx, we can say that if we replace the
tiply both a and b by a constant k ,
limits of integration a and b in (7.7.6)
then this function depends only on the
with ka and kb the length of the strip ratio of the two variables:
whose area we are calculating will
increase fe-fold, but simultaneously its J (a, b) = F (t), where t = fo/a. (7.7.7)
length will decrease fc-fold (here we Indeed, suppose that /( 1, t) = F(t).
.assume that k>
1; see Figure 7.7.1). Then, since J(a b) J(ka, kb) for , =
all values of k we put k 1/a and ,
=
obtain

= /( 1, t)=F(t).

But we know that J is an integral.

Using general properties of an
the
integral, we obtain an important proper-
ty of F (£). Let us set b ac
m where = ,

m is an integer, and split the domain

of integration into parts. Then J (a, b) m
is equal to a sum of integrals:

J (a, b) = J (a, ac) -f J (

ac ac 2 )
,

+ / (ac 2
,
ac 3
) + . .
. +/ (acm_1 ,
b).

(7.7.8)
But all the integrals on the right-hand
side are the same, since for every one
(b) of them the ratio of the upper limit
Figure 7.7.1 of integration to the lower limit is c
252 7 Investigation of Functions. Some Problems from Geometry

(we recall that b = acm Thus, ). the logarithmic function) can be defined
m = F (A) =F(c m by means of an integral!
J {a, b) = J (a, ac ) ), In school we often encounter func-
= tions that are specified by means of algo-
J (a, ac) = J ac ac(
.
== F (c), 2
)
. . .

rithms, or rules for calculating their

and, hence, values. What does the notation / (x) =
m x 3 mean? Here we have the rule: multi-
F {c )
= mF (c). (7.7.9)
ply x by x and then by x once more, and
We see that the function F (t), which you get / (x). Starting with this rule,
t
we can study the properties of function
is defined as the integral dxlx with
j / (#); for instance, it is clear that if x
i is increased by a factor of two, the
the lower limit of integration equal to values of the function increase by a
unity, possesses the following remark- factor of eight. But a function may be
able property: when the independent specified by an integral, and its prop-
variable is raised to the mth power erties can be studied even if we do not
m
(i.e. c -> c ), the function is multiplied know how to find its values directly
by m. But this is the main property of (i.e. not by integration but by means of
the logarithm! If we denote the ratio various algebraic transformations). 7 20 -

F d)lF (c) (where c is fixed and d can

(< For instance, the function
vary) by / (d), then

(7.7.10)
0
that is, if d = c
m then = m\ in
, / (d) plays an important role in the theory
other words, / (d) is the power to
of probability, while the function
which we must raise the constant fixed
number c (the base) to obtain the — constant X
number d. This means simply that t ( (7.7.11)
J Y*l-x 2
/ (d) is the logarithm of number d (the
reader will recall the common definition in the theory of vibrations specifies the-
of a logarithm): 7 19 -
dependence time on the position of
of

/ (d)
= log c d.
a vibrating particle. Luckily, the in-
tegral on the right-hand side of (7.7.11)
Is there any need to consider the can be evaluated in finite form.
simple formula \
x~ x dx = In + C in

such detail? For many calculations

this is unnecessary, and in the majority and this dependence can be reversed,
of textbooks this is not done. Of course, that is, we can pass from the function
it is nice to know how to deal with the t = t (x) to the function x = x (t)~ The

problem when the exponent in the result is x — x 0 sin tot (with co = 1 lk)»
integral is close to unity but is not However, in more complicated cases
exactly unity and what the error in- in the theory of vibrations, we arrive
troduced is when we
substitute 1 for — at the integral
— +
1 b; the first part of this section
was devoted to these aspects. But t=[ — = — (7.7.12)
besides the purely technical aspects,
7 20
there is an aspect of principal import- -
Note that at the beginning of the
20th century the German mathematician
ance here (and the remaining part of
F. Klein suggested introducing in school math-
the section is devoted to this), namely, ematics the logarithm by means of the for-
that a function (in the case at hand, ce

7 * 19
See A. I. Markushevich Areas and mula x~ x dx = In a.
J
Logarithms ,
Mir Publishers, Moscow, 1981. 1
 ) ”

7.8 Average, or Mean Values ,

253

This integral cannot be evaluated di-

rectly, that is, it cannot be expressed in
terms of simple functions, but its prop-
erties can be studied and tables of
its values can be compiled.
Such a generalization of the concept
of a function and a study of the prop-
erties of the function specified by the

integral x~ x dx require of the reader

^
the involved line of reasoning given
above. Figure 7.8.1

Exercise is the graph of a function in the xy-

plane depicted in Figure 7.8.1. Then
7.7.1. Derive from the definition (7.7.6) b
and (7.7.7) of F (t) the main property of this

function: F (ab) — F (a) + F (

b for all posi- y x ) dx
(, is the area under the curve.
tive a and b.
a
The formula
b
7.8 Average, or Mean, Values
x ) dx
$ y (

The concept an integral can be used

of
to give an exact definition of the mean ,
y ( a *>)=
’
\- a
- »

or average of a quantity that is a func-

,
b

tion of another quantity. or (6 — a) y ~ y (x) dx, (7.8.1a)

If we have a quantity that assumes j
a
a set of several values, say the m values
#i, v 2 v3 ., v m then the natural means that y is the height of a rect-
—
. .
, , ,

angle with base b a whose area is

way defining the
of mean v of this
equal to the area under the curve. 7 21 -

quantity is to write
This means that the hatched area with
~ __ V\ H~ V2 +v s H~ • • • H~ the sign above the straight line
m = y (Figure 7.8.1) is exactly equal
But how should we define the mean value
y
to the hatched area with the sign “
—
of a function v (£) of a variable t that under the straight line y y. The —
assumes all values in the interval from graph the function y
of y {x) (if it =
a to b (a <Z t <Z b)? is not a straight line parallel to the
Let us assume that v (t) is the instant- x axis) must lie partly above the line
aneous velocity. How should we define = y and partly below, where y is the
y
the value v (a, b ), that is, the mean mean value determined by (7.8.1a).
velocity over the time interval from a Hence y greater than the smallest
is
to 6? The average, or mean, velocity value and smaller than the
of (x)
y
is defined as the ratio of the distance greatest value of y {x) on the interval
traveled to the time that has elapsed: a < <
x b.
b Let us consider the following exam-
v (
t) dt ples.
J
21
7 -
In Section 7.5 we called the quantity

This definition of the mean is mean-

y (for y (x)> 0 on the interval a x b) < <
the effective altitude, or height, of the cur-
ingful when the function is not only the b

rate of motion but any other function. vilinear trapezoid with area S = y dx.
£
For instance, suppose that y y (x) = a
 a )

254 7 Investigation of Functions. Some Problems from Geometry

Figure 7.8.2

Suppose that y (x) is the linear func-

y = kx
Figure
tion +
m. Then the integral 7.8.3

I ( 6), which is the area of the trape-

,

zoid in Figure 7.8.2 with altitude b a, — decelerated motion that is, when a body
,

bases y (a) and y ( b ), and midline moves under a constant force, say, the
force of gravity, when v gt v0 = + .

In calculating the distance traveled

I {a, b) = y{a) + y- {b)

(
b—a we use the properties of the mean of the
linear function:

= y!±*(b-a). (7.8.2) z(a, » = (>-.)[ ]

This formula can easily be obtained -(6-«)(i5±K+a c ).
by resorting to geometrical methods:
b
b
The reader must bear in mind, how-
I (a, b) = j
(kx-\-m) dx — {^--^mx^
|
ever, that for another, that is, non-
a linear, dependence, the formula (7.8.2a)
becomes invalid.
= ifi +«»_(-+«) Let us consider the example of the
quadratic function (parabola) y rx 2 — +
px +
q. Suppose, for the sake of defin-
iteness, that r is positive, and let us
with y b
( )
— kb-\-m, y (a) = ka + m, and consider an arch of the parabola, say
= k (a + 6)/2+ with a <Z x <Z b. Figure 7.8.3 then
y from which
shows that 7 22
readily follows (7.8.2). Thus, the linear
function, y
/ a +b \ ^
^
y(a)+y(b)
\ 2 / 2
- = ±£<5L = „(!±i), (7.8.2a) We now turn to the integral
b

that is, for the linear function the mean y (x) dx, that is, the area under the
value over a given interval is the arith-
metic mean of the values y (a) and y {b) parabola. This area is smaller than
at the endpoints of the interval Here is .
that of the trapezoid with bases A 0A
another way of stating this fact: the and BqB. On the other hand, if we
mean value of the linear function over
an interval is equal to the value of the 7 22
See Section 7.4. We recall that the
‘

function at the midpoint of the interval. parabola y rx 2 r = , >

0, is convex downward,

An important example of the linear and the parabola y rx 2 px q with = + +

arbitrary p and q is derived from the parabola
function is the time dependence of the y =
rx 2 via parallel translation (see Sec-
velocity in uniformly accelerated or tion 1.4).
7.8 Average or Mean Values
, ,
255*

This is also very simply obtained

draw through point C , y
) from formula (7.8.1a):
the tangent to the parabola, it inter- b
sects the vertical lines x a and x=b = C dx
j
at points A' and B' and completes a C (b a,)

the trapezoid A 0B0 B'A' with midline C (a, 6)

b —a b —a q
C0 C = y (~y-j ;
the area of this tra- 2. The mean value of the sum of two>
pezoid obviously smaller than the
is functions is equal sum
to the of the mean-,
area under the curve. Thus, in the values of the summands :

case of a parabola with r 0 we have > 2/i + — + #2-

J/2 £/l
b

(b — a) y (^) < j
y (*) dx
Indeed,
b
a

< — (6 a) .
y^ + y 2 = b~ra I
[yi W+ yz dx
a
b
Correspondingly, we have the following
inequalities for the mean value y over
= J»i ^ dx
a
the interval from a to 6:
b

*(*¥)<>< . + j
V* (x)dx = y + y2i .

a
For the quadratic function there is
also the exact formula, known as Simp- Here are some additional examples^
son s rule 7 23 (we give it without deriv- Let us find the mean value of the func-
ation; see Exercise 7.8.2), which is valid tion y = sin x on the interval from,
for

~y=h( ~¥)+i [
both signs of

w+jy
a

(nr)
r:

J
^ +4^>-
L
]
x — 0 to x = n:

^ (0. «> =

The mean value

it

j
i
sin x dx

4r=ro-
= -I-
of the function y
0.637.

=
(7.8.3)
sin x on the interval between x = 0*
This formula provides a good approxim- and x = 6 is
ation scheme for calculating the area b
under any smooth curve (see Exercise sin x dx
7.8.46).
Here are two more simple facts per-
l(0-i>)-
J
L
T=<r=
±=r :-b -
P.8.4).
taining to mean values. What will happen if we increase the
1. The mean value of a
constant over number 6 without bound, that is, if
any interval is that constant. This is we increase without bound the interval?
clear physically: if the instantaneous In (7.8.4) the numerator does not
velocity does not change, then the exceed two for arbitrary 6 (it is equal
mean, or average, velocity over an to 2 if cos 6 = —
1, that is, for 6 ax, =
interval is equal to the constant value 3n, 5 jt, 7n, etc.). The denominator in
of the instantaneous velocity.
(7.8.4) will increase without bound,
7 - 23
Thomas Simpson
(1710-1761), an Eng- and so we can say that over an infinitely j

and probabil-
lish algebraist, analyst, geometer, long interval (we arrive at such an
ist. Simpson’s rule (7.8.3) is valid even for interval if we send 6 to oo) the mean
a cubic function y Ax s =
Bx 2 Cx +f + + D value of the sine is zero (namely, the
(for arbitrary A, B, C, and D\ if A 0, we =
have the case of a quadratic function; see larger the interval, the closer to zero
Exercise 7.8.2c). is the mean value of sin x). By applying;
 ) ^
t256 7 Investigation of Functions. Some Problems from Geometry

a similar method, we can show that It is usually assumed that the student
the mean value
the function y of — studying higher mathematics has mas-
cos x over an infinite interval is also tered arithmetics and algebra and would
•equal to zero. Indeed, never allow an error of ten times the
b quantity determined or an error in the
cos x dx sign. Experience has shown, however,
j

y 0 = 0 sin x | § _ sin b
~ that is not the case. For this reason,
( ,
b)
6 —0 b ~b~ ’
calculations must always be carried
(7.8.5) out in such a manner so as to reduce
the probability of an undetected error.
.and if we
increase the number b without
hound, the denominator of (7.8.5) will We note also that the definition
increase without bound and the numer- (7.8.1) v of the function
of the mean
ator will not exceed unity. Hence, the v (
t perhaps, to the concept
is closer,

= of the weighted mean or weighted aver-

entire fraction tends to zero: y (0, oo)
age than to the concept of the mean
0. Literally in the same way we find
that the mean value of the function v = m~ x
(z>i + i>2 + • • • + v m) of a set

y =cos kx is also equal to zero over an ofnumbers understood as the arithmetic

infinitely long interval. mean. Suppose that we have to deter-
Let us find the mean value of the mine the average mass p of one stone
function y =
sin 2 x over the infinite in a pile of stones (or the average price
interval from x 0 to x =
oo. Using = of a book in a bookstore). Let us first
a familiar formula of trigonometry, find the total mass of the stones (or the
2
.sin x —
(1/2) (1 —
cos 2x ), we find that total price of the books) by setting up
sum P = p1n1 + p2n2 + +
-v- n
sin 2
— 0 11
^=-y— y cos2;r= y — 0=y.
A 1 1
the
phnk (respectively, Z = z
.

+
. .

z2n2
+ . . . + n± z k n k ),
is the where
If we now take advantage of the formula number stones (or books),
of “like”
.sin x + cos x = 1, we mean
2 2
get the that is, stones with the same mass p t
2
-value of cos x over the same interval: (or books with the same price z ). Then t

we divided P (or Z) by the total number

cos 2 x = 1 — sin 2 x = 1 — 1/2 = 1/2. of stones in the heap (or books in the

Often it is convenient to use means

store). The result is

instead of integrals. In essence, these

~ 4~Pk n k
__ Pl^l4~P2^2~E - • •

p N
quantities are equivalent: knowing the
b
or
integral / = y dx ,
we can easily find
zini Jr z 2 n 2 ~{- '~\~ z k n k
j • •

a
N
the mean y = H{b — a), while having
calculated the mean y we can easily We do the same every time we
wish to find the average of the
find I = (b — a) y.
quantities «/i, •» y n when • • these
The convenience of the mean lies
quantities are separated into k groups
in the fact that y is a quantity with of “like” objects, with w l9 n 2 ., nh ,
. .

the same dimensions as those of y objects in each group (that is, n x objects
and, obviously, is of the same order in the first group, n 2 in the second, and
of magnitude as y is on the interval so on; obviously, n2 + + • • • +
under investigation (compare with what
was said in connection with formula
^-N ).

Let us now return to our example

(5.6.6).Therefore, it is more difficult of a continuously changing velocity v.
to miss an error that is ten times the Suppose that we know the law v =
walue than to miss the same error
of y v (x) by which the (instantaneous) vel-
in the value of the integral. ocity v can be found at each point on
 J

7.8 Average or Mean Values

, , 257

the path of travel. We wish to find the

total time T
took an object to travel
it
with velocity v (x) from point x a —
to point x =
6. The time At it takes
the object to travel a small distance A#
will obviously be equal to the quotient
Ax/v where v is the velocity over Ax
, ,

which can be assumed constant since

it has no time to change appreciably
over Ax. (The more exact way of stating
this is to write At Ax/v, since even ~
over a small distance Ax the velocity
changes somewhat). The total time
of travel will be approximately expres-
sed by the sum

(7.8.6)'
V
yl v2 Vk

where Ax — x t t
— x t
_x (here i — 1, 2,
. ., k, and x 0
. a xx x 2 ., x k b = , , ,
. . =
are points that divide the entire distance
from a to 6 into k small intervals), and Figure 7.8.4
where v t
can be taken as being either
v (xi) or v (x _ i) (cf. Section 3.1).
t
The while the average reaction rate is ex-
same value of T is provided by the pressed by the formula
following integral:

F=T^\ F(6)dt - 7 ' 8 ' 7a >

*-]%!• <
7 8 - 6a >
-

tl
<

a
which the limit of the weighted mean
is One must not think that the average
(7.8.6) which the various Ax are
in t
reaction rate F coincides with the rate
taken with different weights l/v t the ; F taken at the average reaction
(0)
mean value of l/v is given by the ratio temperature, 0. For instance, suppose
of (7.8.6a) to the entire distance trav-
— that 0 0 = (£) is a linear function
eled, 6 a.
There are several aspects that are (Figure 7.§.4a). Then, obviously, 0 =
related to the fact that in the formulas (e x + 0 2 )/2 = 0( )
. As for the
of integral calculus we often employ reaction rate F we assume that it,
weighted means and that must not be
grows exponentially with temperature
ignored. Suppose we have a chemical =
0, that is, F e ke But because F
reaction whose ratethat is, F =F (0),
(0)
grows so rapidly (see Figure 7.8.46; note
.

the quantity of substance entering the

that since the temperature 0 increases
reaction per unit itme, depends on the
with time uniformly, F (£) is also an
temperature 0 at which the reaction
exponential function), the value of the
occurs; the temperature 0 changes in
integral (7.8.7a) almost entirely is de-
the course of the reaction, or 0 0 (t), = termined by|the (large) values of F at the
where t is time. The average (or mean)
right end of the interval of variation of
temperature of the reaction lasting
0 (or t) and depends but little on the
from time t — tx to time t = t2 will,
other values of F whereby we cannot ,
obviously, be
*2 write F = F(fi)=±[F (0 X ) + F (0 2 )].

0 («).*, (7.8.7) This fact completely explains the non-

6=-^^- tl homogeneity in the weights of F (0)

17-0946
258 7 Investigation ef Functions. Some Problems from Geometry

associated with different intervals A0 b

of variation of 0.
> j* m dx=m (b — a)
In connection with the question of mean a
values we recall two theorems that refer to
and, hence,
continuous and smooth functions.
1. Lagrange’s theorem The mean value of b
a function over an interval is equal to the value
of the function at some point within the interval
M > _a '

h
i
/ (*) dx=J(x)^i m.
.
j
In the notation of formula (7.8.1a), a

T(a,b) = f(c), a<c<b. (7.8.8) But this leads us to (7.8.8), since a continuous
Let us draw a rectangle whose area is equal function assumes on the interval from a to b
b (at a point c in this interval) a value A that is
intermediate between the greatest and smal-
to the integral j*
/ (x) dx, which is the area lest values (i.e. such that A m), in M> >
a particular the value
of a curvilinear trapezoid bounded from above b
by the curve / ( x ), from below by the x axis,
and from the sides by the straight lines x = a
and x — b. The base of the rectangle is the a
line segment from a to b, while the height of
the rectangle, or the effective altitude of the In a similar manner we can establish a
simple generalization of the above theorem:
trapezoid, is equal to the mean value / (a, b)
of the function. Obviously, the upper base of
if a function f (x) on the line segment a
b is continuous and another function g (x) does
x c c
the rectangle is certain to intersect the arc
not change its sign on this interval (i.e. is either
of the curve y =
/ (x) within the limits a and everywhere positive or everywhere negative),
b at a certain point (c, / ( c )) that lies some-
then
where between the endpoints of the arc (Fi-
gure 7.8.5a). Here / (x) was considered conti- b b
nuous, since if / (x) is discontinuous at some
point within the interval, the above state- j
/ x
( )g(x)dx = f(c) j
g(x) dx, (7.8.9)

ment may prove to be invalid (Figure 7.8.5&). a a

W ithout looking at the figure, we can reason
<
as follows. If and M
are the greatest m where as usual a <. c
,
b.
Indeed, suppose that, for the sake of de-
,

and smallest values of f (x) on the segment a c C c

x < b, that is, if M>/ (x) > m, then
finiteness, g (x) is nonnegative for a
Then, obviously,
x b.

b b

M — a)~ M dx^z
(b / (x) dx
b b b

J j f g (x) dx = f Mg (x) dx > f f (x)

g {x) dx
a a
a a a
b b
e
\ mg (x) dx — m i
?(x)dx. (7.8.10)
j
a a

where again M and m are the greatest and smal-

lest values of the function / (rr) on the interval

a <xc b. Dividing both parts of (7.8.10)

b

by
J
g (x) dx > 0, we get 7 • 24

7 24
Clearly, if a function g (x) that does
-

not change its sign on an interval a x

is not “pathological” (say, not like the func-
b < <
tion / (x) described at the beginning of Sec-
b

tion J14.1), then the equality

g (x) dx 0 =
J
^\a
for b a means that g (#)= 0, and in this
case the validity of (7.8.9) is obvious.
7.8 Average, or Mean Values ,
259

b b

M> f(x)g (x) dx

^
g (x) dx > m. (7.8.11)
j
a a

But a continuous function / (x) assumes on the

line segment a -C x b all the values that
lie between the greatest and smallest
values
of the function on this interval; hence, each
number that lies between and m, which M
means the fraction in (7.8.11), too, is equal
to the value / (c) at an intermediate point c
(lying between a and b ):

b b (b)

j/ (x) g ( x ) dx g (x) dx =/ (
c ). (7.8.12)
Figure 7.8.6
j
a a

From this it follows that

(It is clear that (7.8.12) is equivalent to (7.8.9)).
Example. an arbitrary continuous
If / (x) is
function and a and b have the same sign (i.e.
both are either nonnegative or nonpositive), $
y' (x) dx
then y'(a, &) =— —a
y(b) — y(a)
b b —a
*
f
xn dx =
^n+i — an+i
\ x n f(x)dx = f(c) /(c), But by the hypothesis of Rolle’s theorem,
(b) = y (a), whereby here y = 0. Now we
J '

a cl
y
only need to apply Lagrange’s theorem to
7 9 13 )
. .
(
y' x ): if the mean value of a function (in our
0
case the derivative of y (x)) is zero on the in-
where number c lies"" between a and b. terval from a to b, then there is a point c
2. Let us take a function y =
y (x) such? within the interval at which the function van-
that y (a) =
y (b). It is assumed that the func- ishes, y
f
(c) — 0.
tion is smooth. The above hypothesis implies
that the function cannot be monotone in the
interval from a to b. But if the function is not
monotone, it must have a maximum or mini- Exercises
mum somewhere between a and b. This means
that within the interval at a certain point x c, =
where a <C c <C b, the derivative of the function 7.8.1. Find the mean value of the func-
vanishes y' (c) : =
0. This statement is known as tion y =
x 2 on the interval from 0 to 2. Com-
Rollers theorem 7 25 . pare this mean value with the arithmetic
The geometric meaning of Rolle’s theorem mean of the values of the function at the end-
is clear from Figure 7.8.6, where we have points of the interval and with the value in
depicted two types of curves with maxima the middle of the interval.
and minima (the second curve has both a ma- 7.8.2. Verify Simpson’s rule (7.8.3) for (a)
ximum and a minimum between a and b). the function of Exercise 7.8.1, (b) the gener-
—
—+Ax +
Lagrange’s and Rolle’s theorems are in- al quadratic function y rx 2 px + q and
,

terconnected: Rolle’s theorem can be obtained (c) for the cubic function y Bx +
s 3

as a corollary of Lagrange’s theorem. Indeed, Cx -f- D.

let us write the following identity that follows 7.8.3. The force of gravity diminishes as
from the relationship between a derivative the distance from the earth’s center grows ac-
and an integral (see Section 3.3): cording to the law F A/r 2 Find the mean = .

value of the force of gravity over a path that

b starts at the earth’s surface (i? the radius of
the earth) and ends at a point that is R dis-
y(b) = y(a)+ ^
y' (x) dx.
tant from the earth’s surface (i.e. at a point
a that is 2 R distant from the earth’s center).
[Hint. To find the mean value, use the “energy
integral” F (x) dx where F is the force and x
,

7 - 25
Michel Rolle (1652-1719), a French the distance. This integral has an important
analyst, algebraist, and geometer. physical meaning.]

17 *
260 7 Investigation of Functions. Some Problems from Geometry

7.8.4. Compare the mean value found in

Exercise 7.8.3. with (a) the arithmetic mean
of the values of the force of gravity at the end-
points of the interval (which coincides with
the average when the quantity changes linearly
with distance), and (b) the mean value calcu-
lated via Simpson’s rule (7.8.3) (which coin-
cides with the true mean value for a quadra-
tic law of variation of the quantity with dis-
tance).
7.8.5. Find the average value of y xn =
over the interval from x 0 to x x0 — = .
Figure 7.9.1

7.8.6. Find the mean value of the function

x
y =i c e^ over the interval in which y varies whence
from y n to y = —m
express this mean value ;

in terms of n and m, eliminating c and k from

the answer. Investigate the resulting expres-
sion when m
is close to n, or n v, m = + E--/ '+(¥? f7 ' 9' 1)

v < n.
We pass to the limit in (7.9.1) as A#->0.
7.8.7. Find the mean values of the func-
tions y —
sin 2 x and y cos 2 x over the fol- = The ratio Ay /Ax becomes the deriv-
lowing intervals: (a) from x 0 to x n, = = ative y' —
f (x). Thus, we get
and (b) from x 0 to x =
jt/4. —
7.8.8. Determine the period of the func-
=
ds =yr+jf (x)) 2
dx.
tion y sin (co£ +
a), where co and a are
constants. Find the mean value of y 2 over one
The entire length of the arc is
period of y 2 .

s = s(a, 6)= /! + (/' (x))*dx. (7.9.2)

j
7.9 Arc Length

Let us pose the problem of finding the the curve is specified parametri-
If

arc length s of a curve y f (x) from — cally, that is, x x (t) and y y(t) = =
the point where x a to the point = (see Section 1.8), then, dividing the
where x =
b (Figure 7.9.1). replace We (approximate) equations
the length As of a small arc of the MN As c^Y A# 2 + Ay 2
curve y f (x) =
by the straight-line
segment MN
connecting and A. 7 26 M -
by At, passing to the limit as A£ 0,
(We consider only curves that have no and taking into account the fact that
discontinuities or cusps.) By the Pyth- Ax dx Ay dy
lim and lim we
agorean theorem, Af-0 At dt
At-0 At dt
obtain
As (Ax) 2 + (Ay) 2
ds , / / dx \ 2 / dy \ 2
~ V
= Ax j/"l + ,
dt [ dt J \ dt ) ’

or ds = Y i
x ') 2Jr (y')
2
dt. Integrating
7 - 26 The
difference between the arc length this from t —a to £ = we get
and the length of the line segment (the chord)
is of the order of Ax\ 3 and so can surely be
| ,
13

neglected when we pass to the limit (to differ- s(a, V(x')*+(y') 2

entials). For instance, the length of the arc
P)= j
dt, (7.9.2a)

of a circle of radius r subtending a small angle

a
A* is equal to rAt (we use radians to measure where s (a, (3) is the length of the arc
angles), while the length of the curve corres-
ponding to this arc is 2 r sin (At/2) = of the curve limited by the values a
2r [ Af/2 —
(1/6) (At/2)
3
...] rA* + = — and P of parameter t .

(1/24) r (A*)
3
(see +
Section 6.2), so
. . . Because of the radical under the in-
that the difference between the two lengths is
tegral sign in (7.9.2) and in (7.9.2a),
(l/24)r (Ai) 3 that is, is of the order of (As) 3
, ,

where As is a quantity of the order of the it is rarely possible to evaluate the

length of the arc (or chord). integral directly, that is, express s (a, b)
 .

7.V Arc Length 261

and s (a, |3) in terms of functions of the

limits of integration a and b (or a and P)
expressed explicitly.
Note also that from the same right
triangle 7 27 with sides Ax
-
Ay and ,

MN (~As) which we used to express

As in terms of Ax and Ay we conclude
that

Ax ~ As cos t|), Ay ~ As sin ty,

Figure 7.9.2

where is the angle between the x

axis and the line segment As Ax -> MN whence we obtain the well-known for-
0, obviously cp where cp is the mula for the circumference C = (nRl2) X
angle between the x axis and the tangent 4 = 2nR.
to the curve y = / (x) at point
1.
Thus, M . 2. Catenary curve. This is a curve
whose equation is
dx—ds coscp, dy — dssincp,
ds = Y dx2 + dy
, -
2
(7.9.2b)
y = -j (e*/«-j-e-*/a), (7.9.4)
,

where a constant. 7 28 The word

is a -

where the fact that we have used differ-

“catenary” comes from the Latin catena f
entials means that dx (=A;r), dy and ,
meaning “chain”; the curve has the
ds are small.
form of a freely hanging heavy, flexible,
Some examples
follow in which the
and inextensible cable (or chain) sus-
computations can be carried out with
pended from two fixed points (the
relative ease.
density of the cable is assumed the
The circumference of a circle. We same at each section of the cable). The
seek the
circumference of the circle
graph of the catenary curve is given
x2 +
y
2 =
R 2 or, to be precise, the in Figure 7.9.2 (for a = 2).
length s of one-fourth of the circum-
Let us find the length of arc of the
ference in the first quadrant and then
catenary from x--0 to x = x 0 From .

multiply it by 4.
From the equation of the circle we (7.9.4) we have y' = -
e
—— ,
and so
have

y=VR -* 2 2
,
YT+W=i/i+
e x/a_j_ e-x/a) 2 e x/a _j_ e -xfa
By formula (7.9.2),
(

4
^ 2
R and, hence
S=
I
0
V +R^ l dx = f
J V#
— dx
2 -* 2
'

s= j— + e
Xo
— dx =~(e x! a—e- x a !
1*0
)
'o
(7.9.3) 0

We introduce a new variable t thus: — e *o/a — e~ x °l a

x = R sin t. Then dx R cos t dt and = ( ) .

from (7.9.3) we get (see Section 5.6) 3. Cycloid. We

know that the (para-
Jt/2
metric) equations of the cycloid are of
the form
Rdt=^~,
j
0
x = a (t — sin t), y = a (1 — cos t),

7 28
7 27
-
Equation
(7.9.4) can also be written in

in the
This triangle played an important role
*

reasoning of Leibniz, who used it in

the form y —
a cosh (x/a), where by cosh t we
have denoted the hyperbolic cosine of (see
introducing differentials. Section 14.4, in particular Figure 14.4.1).
 '

262 7 Investigation of Functions . Some Problems from Geometry

where a is the radius of the circle gener-

ating the cycloid (see Section 1.8).
Whence, by (7.9.2a),

ds = Ya — cos + (a sin dt
2
(1 t)
2
t)
2

— a Y — 2 cos + cos + sin dt

1 £
2
1
2
t

- a Y 2 — 2 cos dt — a 4 sin 2
y t

= 2a sin y- dt ,

and the length of the arch of the cycloid

lying between the points corresponding
to the values a and (3 of parameter t is
P

5 = 2a sin -y dt — 4a — cos -|- P

j ^ j |

= 4a ^COS -y COS
y .
j

Since two successive “cusps” of the Figure 7.9.3

cycloid (see Figure 1.8.3) correspond
to values 0 and 2 n of parameter t, This formula shows us that in the given case
the length of one arch of the cycloid .
the arc length of the curve is close to that of
the vertical line segment with endpoints y (a)
is equal to 8a, which is four times the
and y (b) (Figure 7.9.3 b). Indeed, if the deriv-
diameter of the circle that generates ative y' is great, then the curve bends steep-
the cycloid. ly upward and so is similar to a vertical
straight line (for a vertical line the derivative
As pointed out earlier, in most cases it is is infinite).
difficult (or even imposs ible) to integrate the The formulas (7.9.5) and (7.9.6) yield
simple approximate formulas for the arc length.
function ]/~ 1 (y' (x))
2
+
in terms of elemen-
But these are very rough approximations,
tary functions due to the radical. For this
reason, approximate formulas for computing
which can be obtained without appealing to
arc length are of particular interest. (7.9.2). It is easy to obtain more exact for-

Suppose that (y' (x)) 2 is small compared mulas.

with unity: y' (x) 1.|
Then, neglecting
|
< Let y' (x)
|
1. |
<
Retaining two terms
( y' (x))
2
in (7.9.2), we get in the binomial expansion of 1 (y' (x)) 2 +
(see Section 6.4) and discarding the other terms,
b
we can write
s ~ VI <?* = &— a. (7.9.5)
j
a
/i+(7w= [l+to' (z)) 2 ] 1 / 2

The difference b —
a is the length of the =* 1 +-g-(2'' (
x )) 2 -
horizontal line segment with endpoints x a =
and x —
b. Formula (7.9.5)?Jshows that if y Formula (7.9.2) now yields
is small in absolute value (i.e. the curve di-
ffers little from the horizontal), then the length b

of the corresponding arc of this curve is also

close to the length of the horizontal line seg-
S ~ 1
[! + Y (/(*))*] dx
ment (Figure 7.9.3a). a
But if (y' (x)) 2 ~> 1, then in (7.9.2) we can b
ignore unity in comparison with (y' {x)) 2
The result is
.
= {b—a) + ~ (y'(x))*dx.
j
a
b b

V(y' ( x )) 2 dx= ? y' (x)dx=y(b) — y(a). But if |

y' (x) | > 1, then

W
[
a a 1

(7.9.6)
Y !+(»' 2 =y' (*)]/" 1 +
</ (*))
2 ‘
 b

7.9 Arc Length 263

To the radical on the right-hand side we can forms an angle of 45° with the axis of abscis-
apply the binomial theorem since (y' ( x ))~ 2 1. < sas).
Retaining only two terms in the expansion, By the first of the formulas (7.9.8),
we get b

s~ {b— a) + 1

~Y
(*

dx =—
3
— = 1.5 —
(b d) (b a).
J

(7.9.10)
~ V> (*)
t
1 +2 (y> (a))* ]
= y> (X)+ 2y' (x) *
The second formula in (7.9.8) yields the same
(7.9.7) result: s ^ 1.5 ( —
a). Comparing (7.9.9) and
(7.9.10), we see that the maximum error in
Substituting this into (7.9.2), we get
the approximate formulas is about 6%.
b When computing arc length, we must de-
vide the curve into portions over which either
|
y' |
cl
everywhere or 1 everywhere. / >
a Then the error will at least not exceed 6%.
And since (y' (x)) 2 assumes a value equal to
unity only at certain points of the curve, a
A a
y
'
{x)dx+
T]yk a
proper partition of the curve into portions
will reduce the error below 6% (and usually
substantially below 6%). Of course, there is
no sense in finding the length of rectilinear
= y(b)-y(a)+± portions (for one, portions over which y' (x) ^
a 1) by means of approximate formulas.
Let us take a look at some examples; al-
We now have the following approximate most always the computations are carried out
formulas: to two decimal places.
1. Find the arc length of the parabola
b
y =x 2 between the points with abscissas
s~ (6 — a) + — (^' (a:))
2
da: x = 0 and x — 2.
j We
find the derivative, y' 2x. It is =
a equal to 1 at x = 0.5, is greater than 1 for
if |
y' (x) | <1,
(7.9.8)
x >• 0.5 and is less than unity for x < 0.5.
Therefore, the arc length sx corresponding to
a variation of x from 0 to 0.5 can be found
from the first formula in (7.9.8) and the arc
length s 2 corresponding to a variation of x
a
from 0.5 to 2, by the second formula:
if lif'(*) I>1. 0.5

The integrals here are simpler than the in- sx ~ (0.5 — 0) + 0.5 j
4x 2 dx
tegral in (7.9.2) and it is easier to perform the 0
computations by these formulas than by (7.9.2).
3
But these formulas are approximate, whereas = 0.5 + 2 (0.5)
~ 0.58,
<7.9.2) is exact. 3
What errors result from their use? The 4
r
first formula in (7.9.8) is best for small y | | dx
r
while the second is best for large y |. Both
= J 2^
|

formulas yield the worst results at y' 1. | |

0.5
Therefore, to estimate the error here we will
examine the worst case, which is y' (x) 1. = = 3.75 + 0.25 (In 2— In 0.5) -4.10,
r
It is clear that if y (x) 1, then y (x) = =
x +c; the graph of this function is a straight whence, the sought-for arc length is
line.
By the exact formula (7.9.2), s = S! + S2 — 0.58+4.10 = 4.68.
b

— a) ~ 1.415 (b — a) On the other hand, by (7.9.2),

s= V^f-pT dx= \/~2 (b
j 2
a

(7.9.9) s = 1
^ + 4z
l
2 dx ,

0
(this immediately follows from the
result
Pythagorean theorem or elementary trigo- whence, by making the change of variable
nometry, since the straight line y x c = + 2x =
z and employing formula 33 from Ap-
 2 R * ) )

264 7 Investigation of Functions. Some Problems from Geometry

pendix 2 at the end of the book, we find the (here we have used the fact that 7? sin 30° =
exact value of s : 7?/2). The integrand is transformed as follows:
7? 7?
r
dx= — r\xy
i -

Y —
VY !-(:R
\ \-\-kx 2 *kX 2 -\-\ 2 2
x 2
i \
R
4- 4-
2
In (2* + VW + 1) 1J + C, '
(7.9.11) _ 1 __ r. / * in- 1 /2

, /~ ( x \
2 L \7?/J
and, consequently,
(7.9.14)
r
s= -|-[2V 17 + 4' ln ( 4 +lA 17 )]
— 4.65. We put x/R 2 t and expand (7.9.14) in
(,
=
a binomial series (see formula (6.4.3)) to get
(7.9.12)
1/2
(Any doubting reader can convince himself
of the truth of the formula (7.9.11) by taking
[«-m
,1 .3
= (l — 0" 1 2
/

35
the derivative of the right-hand side of this
formula.)
1 + T + T* f
2 "

16 128
*
4
+.
The error introduced by formulas (7.9.8) 2
came out to about 0.7%
2. Find the arc length of the exponential
=h4(t) +4(t)‘
curve y —
ex between the points with abscis-
sas x =
0 and x 1. —
In this case, y' =
ex the derivativejgrows ,

from 1 to e as x increases from 0 to 1. So we Substituting (7.9.15) into (7.9.13) and inte-

use the second formula in (7.9.8): grating, we find that

s _ ( A T |_
4
|
3 _i 5
\ 2 6-2 3 40-2 5 '
16-7-2 7

128-9-2 9 +•••)' (7.9.16)

~ 2.72 — 1 — 0.56-* ~ 2.04. | J
1 1 is clear that the terms of the ser ies ( 7 9 1
6) . .

The exact formula

for the arc length yields rapidly decrease and so we need only a few
the value (see Exercises 7.9.16 and 7.9.2) terms of the series to get s. Taking one term,
we have s ~
R/ 2, whence the circumference
Y + — e2 is C ~
6R. Taking two terms, we have s ~
= V 1 + — 1^2 +
l 1
s 62 0.5217?, that is, C 6.252 R. Three terms ~
V"e* + i+l yield s ~
0.5237? and C ca 6.2767?, and so on.
Y 2—1 We know that the circumference of a
~ 2 00 . . circle is C —
2ziR. Comparing this with the
VT+i results we have obtained, we can approximate
the value of the number n: 3, 3.126, 3.138, ....
The approximate formula introduces an error The more terms of the series (7.9.16) we
of 2%. take, the more exact the value of n we obtain.
The arc length may also be approximated The value of jt to four decimal places is jx ~
occasionally by a series expansion of the in- 3.1416.
tegrand in (7.9.2) in powers of x. By retaining
an appropriate number of terms of the ex-
pansion, we can obtain the arc length to any Exercises
degree of accuracy.
Let us consider an example. We wish to 7.9.1. Write the arc length in the form of
determine the circumference of a circle by an integtal for (a) the parabola y = x 2 from
seeking the arc length s of the circle corres- point (0, 0) to point (1, 1), (b) the exponen-
ponding to a central angle of 30°. The cir- tial curve y =
e* from the point with x 0 =
cumference is C —
125. Quite obviously we to the point with x 1, and (c) the ellipse =
will obtain the same kind of integral as in x 2 /a 2 + y /b = 2 2
1.

(7.9.3) but with a different upper limit: Complete Exercise 7.9.16 by making
7.9.2.
the change of variable 1 e 2x z 2 in the + =
integral.
R dx 7.9.3. Use the approximate formulas (7.9.8)
s = (7.9.13)
i Y — 7?
2 2 to find the arc length of the catenary curve
between points with x 0 and x 2 = —
 7.10 Curvature and the Osculating Circle 265»

(a = 1). Compare the result with exact value

of the arc length.
7.9.4. Find approximately the arc length
of the hyperbola xy — —
1 between the points
with x =
0.5 and x =
1. [Hint. In this case
we cannot obtain an exact value because the
integral in (7.9.2) cannot be expressed in terms
of elementary functions.]
7.9.5. Obtain approximate values of the
number n by computing the arc length of a
circle with a central angle of 45° (retain three,
four, and five terms in the series).

7.10 Curvature and the Osculating

Circle arc the tangent rotates only in one-
direction when we move from point m
Related to the arc length is the problem to point N, the angle cp shows the
of determining the curvature k and the
magnitude of the rotation of the tangent
radius of curvature R of a curve at a on arc MN
(or the magnitude of the
point. For a circle the radius of cur- “rotation” of the curve, since the direc-
vature R is simply the circle’s radius, tion of the curve is specified by the
and the reciprocal of the radius is the direction of the tangent). To determine
curvature, k =
HR. The smaller the the extent to which the curve is curved
radius R of a circle (i.e. the greater the we must also know the arc length s of
curvature k ), the tighter the circle MN: if s is large, then even a slow change
(Figure 7.10.1). When R is very large, in the direction of the tangent can lead
on the contrary, the curvature is hardly to a large value of cp.
noticeable, that is, the circle is Thus, the ratio
“almost” a straight line; for instance,
the fact that the earth’s radius is so fc av — (7.10.1)
large means that we do not notice the
specifies the average (or specific) cur-
curvature of the earth’s surface (to vature of the curve over arc MN. The
notice it, we would have to observe the
earth from outer space, and this is
curvature at point M of the curve is
defined then as the limit of the average*
not common practice).
Let us take an arbitrary curve T now.
curvature when the length of arc MN
tends to zero:
It is clear (Figure 7.10.2) that the curve
is curved more on the portion PQ where
,

it sharply turns to the left, than on the

portion MN
where its direction does
,

not change so rapidly. But what is the where As is the length of a (small) arc
exact meaning of the last statement?
The explanation is as follows.
MM X curve
of T, and A a is the incre-
ment of angle a corresponding to this
Consider the angle cp between the arc (a is the angle between the tangent
tangents of the curve at the endpoints
M and N
of the arc MN. If over this
to the curve at point M and the x axis),
or the angle of rotation of the tangent
to the curve (the angle between the
tangents t and t 1 to T at points M
and 1 M
Figure 7.10.3). But by virtue
;

of the definition of a derivative (see*

Section 2.1), we can rewrite the last
limit as

Figure 7.10.1 (7.10.2)

 , -

266 7 Investigation of Functions . Some Problems from Geometry

We note also that since a is a dimen-

sionless quantity (we employ, as usual,
radians for measuring angles) and s
has the dimensions of length [Z], the
curvature k of a curve has the dimen-
-1 -1
sions [Z] (say cm ) and the radius of
curvature (as any radius), the dimen-
sions of length [Z], say cm.
The notion of the radius of curvature
(and hence the notion of curvature) of
a curve may be introduced in a more
geometrical manner. To this end we
draw normals MQ
and M
X Q (perpendic-
ular lines) to the tangents to two close-
lying points M
and M
1 on T (by Q we
have denoted the point of intersection
of the normals; see Figure 7.10.3). The
Thus, curvature k can be defined as
the rate of rotation of the tangent to F
angle MQM 1 between the normals is
equal to the angle Aa between the
as the point moves along the curve
with a unit (linear) speed, that is, tangents to points and M M
x (in accord

under conditions where point travels M with a familiar theorem of geometry

the distance L ds in time dtMM = = on angles with mutually perpendicular
sides). From this we can find the dis-
ds.
Now we need only recall that tan a = tance MQ from the curve to the point
y' so that a = arctan y' and, hence,
of intersection of the normals.
We regard the small portion of the
da = d (arctan y') = \
curve as an arc of a circle. normal A
to the circle is clearly a radius, and the

moreover, ds — ^finally, point of intersection of normals is

2
1-1- (j/) ds (see Sec-
tion 7.9). Thus, clearly the center of the circle. If the
curve was a circle of radius i?, then ds
y"
k = da ~r 2
dx] would be equal to R
da or da/ds =
Ll+UO 2 dx]l[Vi+(y')
,
ds
1 /iZ; this quantity, da/ds (the curva-
ture of the circle), is constant for any
(7.10.3)
[ l+o/'m2 portion of an arc of the circle. Now we
and, hence, select the arc of a circle in such a man-
ner that it is as close as possible to the
= 4- _ [l+UO 2 3/2
i?
k
1 !
(7.10.3a) arc MM X (“hugs” the arc). The words
“as close as possible” are understood
curve T is specified parametri-
If
as a requirement that the first three
cally, x =
x(t) and y — y ( t ) (see Sec- terms in Taylor’s expansion of the
tion 1.8), then dy/dx y’lx' and = = equation y =y (x) of the curve,

~ (see formula There- — a) +

(4.4.9)).
y =y(a)+y’ (a) (x
fore, from (7.10.3) and (7.10.3a) it
follows that + \y" (a) (
x — a) 2

') 2 “1 3/2
(y

(x'Y J + ^y"(a)(x-a) + 3
... (7.10.4)
f
x y" — x"y'
[(*') 2 +Q /') 2 ] 3 / 2
’
(we assume that the point on the M
l _ [QeT + OO 2 3/ 2
]
curve corresponds to the value x a =
x y — x y'
, ,, ,, (7.10.3b)
k of the abscissa), coincide with the first
 x )

7.10 Curvature and the Osculating Circle 267

three terms in Taylor’s expansion of

the equation y (x) of the circle, —Y
y=-Y(a) + Y' (a) (x — a) +
+ jY" (a) ( x —a )
2

+ -g-y («)(*- o)*+ ... (7.10.4a)

7 29
(see formula In other words,
(6.1.18)). -

we require that not only the values

y ( a and Y
(a) coincide but so do the
first two derivatives of y (x) and x) Y (.

at point x a: =
y(a) = Y (a), y’ (a) = Y' (a),

y" (a) = Y" (a). (7.10.5) sign as we

pass through point M
(i.e.

as dx changessign), the osculating circle

But by virtue of (7.10.3), all this guar-
antees that at point the curvatures M 2 at point M
of the curve T intersects
the curve. 7 31 This makes it possible
-

of the curve T with equation y y (x) = to describe the osculating circle thus.
and of the circle 2 with equation y = Let us consider various circles cr that
Y {x) coincide. This means that the
= MQ
touch point M
of T and have at point
radius R
by formula
of circle 2 is expressed
(7.10.3a).
M the same sense of convexity (the di-
rection of bending), in other words, have
The circle
point
that passes through
M
2
curve T and is the clos-
of the
the same tangent at point M
as T and
have centers that lie on the side of
est to T is known as the osculating circle convexity of T (i.e. inside the curve)
of the curve. 7 30 The radius R of this
-

(Figure 7.10.4). Some of these circles

circle is the radius of curvature
of the lie completely inside T (i.e. on the side
curve, and the center Q of circle 2 of convexity of T), and their curvature
is said to be the center of curvature of is greater than that of T. Others lie
the curve (at point ). Thus, the M outside T, and their curvature is smaller
radius of curvature of a curve can be than that of T. But there is only one
defined as the radius of the osculating circle that goes from one side of T to
circle; the curvature of the curve is the the other (this happens at point Af),
reciprocal of the radius of the osculat- and this circle S 0 lies closest to T and
ing circle. is called the osculating circle 2.
We
note also that since by virtue of If conditions (7.10.5) are met for a
(7.10.4), (7.10.4a), and (7.10.5), the straight line (the tangent t to T at
difference (x) y (x) Y (1/6) X — = point Af), then the radius of curvature
[ Y" {a) — y’" (a)] dx + . . . changes of T at this point (which is calculated
via (7.10.3a)) becomes infinite and the
7 29
A circle in a plane is defined by three
-

numbers (parameters); these may be the coeffi-

curvature k vanishes. Such points M
cients a, 6, and r in the equation of the circle are sometimes called the rectifying
( —
a) 2 (y b) 2+ 2 —
r — 0. To find these
— points of curve T. It is clear that at
numbers (i.e. to define the circle) we must have
three conditions, which may be the conditions 7 31
-
Exceptions in this case are only points
in (7.10.5). where, in addition to conditions (7.10.5), we
7 30
An
osculating circle 2 of a curve T can
-
,
have Y'" (a)=y"' (a). Examples are the vertices
also be defined as follows. Consider the circle S of an ellipse (points where the ellipse intersects
that passes through three points Af 2 the axes of coordinates); at such points, as
M
,

and s of curve T that lie close to point M. symmetry considerations suggest, the osculat-
When all three points move toward circle S M ,
ing circle cannot interesct the ellipse (by anal-
tends to the osculating circle 2 of curve V ogy, all points of a curve T that possess this
at point M. property are called the vertices of T).
 , . x

268 7 Investigation of Functions. Some Problems from Geometry

so the second derivatives, y"

do =
Y" = that is, at such points the
0,
tangent t obeys the conditions (7.10.5);
in other words, at a point of inflection
M the tangent t acts as the osculating
circle E, or M
is a rectifying point
ofT.
For a straight line Z, which of course
can also be considered as a special case
of a curve, the angle a between the
tangent to l (which is simply Z) and the
x axis does not vary; this means that
the curvature k of a straight line is
identically zero (Z is simply not curved
Figure 7.10.5 in any way). For a circle S, the angle
a changes, but the rate k da/ds at =
a rectifying point the curve T has no which a changes along the circle re-
center of curvature Q (this center lies mains constant; in other words, we may
at infinity). say that circles (and straight lines,
From (7.10.3) it follows that the for that matter) are curves of constant
signs of k and R(up till now we spoke curvature. It can be shown that these
only of the absolute values of k and R are the only types of such curves.
and ignored their signs) coincide with Suppose that ( M
y, (x)) is a point ,

the sign of the second derivative y" of a curve T. The tangent t to the curve
that is, characterize the sense of con- at point M
has a slope y' (=tan a), and
vexity of the curve (see Section 2.7). If the normal n is perpendicular to Z,
the curvature is negative, that is, at that is, forms an angle (3 = a 90° +
point M we have a negative y", the with the x axis (see Figure 7.10.3); the
curve at this point lies below the tan- slope of the normal is tan P cot a = —
gent at point Af that is, the curve is
,
= —My' Therefore, the (directed) pro*
convex (ox convex upward] Figure 7.10.5a); jections MP
and PQ of the axes of
the center of curvature Q then lies coordinates of the line segment MQ
below point M. If the curvature is of length R are, respectively, p =
positive, that is, y" > 0, the curve R cos P = —
R sin a and q R sin $ =
lies above the tangent and is said to = R cos oc, which means that the coor-
be concave (convex downward Figure ;
dinates of the center of curvature Q aro
7.10.56); in this case the center of cur-
vature Q lies above point M Finally,
| = X + p = x - R sin a,
= +q= + Rc os a.
.

if the curvature at point M vanishes,

x\
y y
that is, y" =0, then, as we already If now we take into account formula
know, the curve at point M changes its (7.10.3a) for R and the obvious fact
sense of convexity; to one side of M that
the curve is convex upward and to the
a—
1 1
other it is convex downward (Figure cos
7.10.5c). At such a point the curve
y 1 +tan 2 a V i+a/') 2
’

tan a y'
passes from one side of the tangent sin a=
to the other, changes its sense of con- y + tan 1
2
a V'l+UO 2
’

vexity (the direction of bending), and we finallly get

the point is called a point of inflection .

At' points of inflection, not only the £ =* — V [l +(</') 2 ] 3 y'

first derivatives of the equation y =

y"
V 1 +(y') 2

y (x) of curve T and of the equation

— Y n =y+ V [l+0/') 2 3 l
l

y (x) of tangent t coincide, but V + (*/T

l
2
 , , ,

Curvature and the Osculating Circle 269

7.10

that is,

1+0/') 2
l = x _±m.y. =v y"

(7.10.7)

(This agrees with the condition that

r) > y for k >
0, or for y
"
0, or the >
condition that v\ y for k 0, or for < <
If the curve is specified parametric-
ally, x = x (t) and y = y (t), we can
write (see (7.10.3b)) Figure 7.10.6

l = x — 00 2 +0/T
y
x'y" —x"y'
2
(7.10.7a) where c = — 4/3 ]/ 3 and r]i = r\ —
(sT+G/')
n=-!/ +-x'y"—x"y' a 1/2.
Thus, the evolute of y x 2 is the =
(see Exercise 7.10.4).
semicubical parabola (7.10.8) whose cusp
The
locus of the centers of curvature coincides with point (0, 1/2) and whose
of a curve T constitutes a new axis of symmetry is the axis of ordinates
Q (£, r))
(Figure 7.10.6).
curve y known as the evolute of curve T.
The curve T with respect to its own Example 2. The cycloid x =
evolute y is called the involute.
1 32 a (t —
sin t), y a (1 cos t).= have — We
Example 1. The parabola y = x2 .
x' =
a (1 cos t), y' — a sin t x" = ,
=
Here y' — 2x and y" = 2 (a constant), a sin t, and y" =
a cos t (the primes
whereby the curvature k and the radius denote derivatives with respect to par-
of curvature R of the parabola at ameter t). Whence,
point (x, y) M (=Af (x, x 2
)) are
(x') 2 + (y = a r

)
2 2
(1 — cos t)
2
+a 2
sin 2\t

y, - 2
o V (1 +4x2)3
~ a 2 2 — 2 cos = 4a 2 sin 2 ^-
(
t)
V(i+ 4z2 )
3 ’
2

The center curvature y"x' — x"y' = a cos t-a (1 — cos t)

of Q (|, ri) of
the parabola with respect to point — (a sin 2 = a (cos — 1)
t)
2
t

M (x, x2 ) has coordinates

= -2a 2 sin 2

| =x • 2x = — 4x 3
,

Thus, in view of (7.10.7a) we have

T) =x 2 -|- Lt^!L - 3a;2 4- -i-
= a — sin + 2a sin
(t £
£ £)

that is, = a + sin(t £), (7.10.9)

rj = a (1 — cos — 2a (1 — £) cos t)

= — a + a cos £.

= °r “
= CT1 ‘
/2
’

To understand what curve is repres-

(7.10.8) ented by these equations, we introduce
a new parameter, t n (or t = + =
7 - 32
The word “involute” is the Latin for t1 — n); since sin t— — sin t and
“unwinding”, while the word “evolute” is the
Latin for “unwound,” that is the curve that cos t — — cos £ 1? we have
is being unwound; these names are related to
the properties of evolutes and involutes dis- ? = a (U — sin U) — an,
cussed at the end of this section (see the (7.10.10)
text in small print). T) — )!
a (1 — cos tx ) — o
2a.
 R ,

270 7 Investigation of Functions. Some Problems from Geometry

curvature Q (§, t|)) forms an angle p a 4- —

90° with the x axis, that is, coincides with
the normal to T. And since point Q belongs to
the normal n to curve T at point M, the respec -
tive tangent to y coincides with n, that is,
tangents to the evolute y of a curve T coincide
with the normals to T (Figure 7.10.8). 7 33 -

On the other hand, from (7.10.11) it fol-

lows that

d^ 2 -\-dr\ 2 = sin 2
a dR 2 -\- cos 2 a dR 2 — dR 2 ,

Thus, the evolute (7.10.9) (or (7.10.10)) i.e. do=dR, (7.10.12a)

of a cycloid is also a cycloid ,

= where a isthe length of the arc of evolute y
a (t x — sin tx) and = % a (1 — cos f x ), of curve T (both do and dR are assumed pos-
where = i + an and r\
t
— r) 2 a, + itive).
The last remark has the following meaning.
only shifted in relation to the initial MM
Let us consider an arc X 2 of curve T such
cycloid by 2a units downward (2a is that on it the derivative R' dR/ds retains =
the diameter of the generating cirle) its sign. We will agree to reckon the arc length

and na units to be left (jc a is the half- along T in such a manner that the derivative
of R is positive in other words, we assume that
width of the arch of the initial cycloid) ;

the arc length is reckoned along T in the di-

^’jCgUYb rection in whicn tne ra’dius ol curvature
grows Then the increment
.
2Q2 ^iQi M — =
discuss the properties of evolutes and
To 7? 2 —R± =
AT? of the radius of curvature
involutes in detail, we must first differentiate along arc 1 2 is MM
equal to the length Ao
in (7.10.6): of the evolute arc. And since the line segments
d\ — dx — R cos a da — sin a dR, M
MiQ 1 (=R 1 ) and 2 Q 2 (=R 2 ) are tangents to
evolute 7 of curve T (see Figure 7.10.8), the
d^ — dy—R sin a da + cos a dR. process of reconstructing the involute T from
its evolute 7 resembles the unwinding under
But since (see (7.9.2b))
tension of a nonexpandable string from a spool
dx
ds
— cos a ds=cosa —da
— da = R cos a da shaped like 7; the end of the string de- M
scribes curve T. That is the origin of the names
evolute and involute (see footnote 7.32).
dy — da —
= sin a ds= sin a-zact
ds
sin a da, By way of an example, let us construct the
(parametric) equation of the involute of a
we have circle of radius r and equation x
2
y
2
r2 + = ,

d'B, = — sin a dR, dr\ = cos a dR. or x —

r cos t and y r sin t =
where t is ,

the parameter (angle). Let us assume that the

(7.10.11) involute passes through point A (r, 0) and
that the unwinding of the string wound on
Dividing one by the other, we get
the round spool occurs counterclockwise (Fig-
-^1L= — cot a = tan (a + 90°). ure 7.10.9). The tangent to the circle MT
(7.10.12)
at point M
(r cos t r sin t) forms an angle
,

a = + t 90° with the x axis, and the line

From this it follows that a tangent to evolute 7 segment MT
ol length rt along the tangent
of a curve T (i.e. to the locus of the centers of rt is the length of arc AM
ol the circle) has
(

projections rt cos a = —
rt sin t and rt sin a =
rt cos t on the axes of coordinates. This read-
ily yields the following parametric expres-
sions for the coordinates X, of point T (X, Y
Y) (i.e. the equation of the involute of the
circle):

X — r cos t—rt sin t,

Y = r sin t + rt cos t. (7.10.13)

The theory of evolutes and involutes was

first developed by Christian Huygens in his

7 33
In this connection it is often said that
.

evolute 7 of curve T is the envelope of the

Figure 7.10.8 normals to T.
7.11 Selid Geometry Applications of Integral Calculus 271

Figure 7.10.11

(see Figure 7.10.7 and Eq. (7.10.10)). He

also demonstrated that if at the point of sus-
pension of the string of the pendulum certain
“lips” in the form of a cycloid are attached, so
that in the process of oscillations the string
winds about them, then the end of the string
(with the bob) of constant length describes a
cycloid (Figure 7.10.11). The first pendulum
clock constructed by Huygens had just such
“lips”; however, later he dropped this idea
because he found that the “lips” had practi-
cally no effect on the precision of the clock,
since in view of the smallness of angles a
Horologium Oscillatorium (1673) in connection in Eq. (7.10.14) we can always replace sin a
with the following problem. We wish to con- with a and the equation m ( d2 s/dt 2 ) mg a =
struct the equation of motion of a pendulum, implies that the period of oscillations of a
that is, a material particle of mass m (the bob) pendulum of constant length l is constant (see-
moving along a curve y: Section 10.3).
d2 s
m ~~mg sin a, (7.10.14)
~fa 2 Exercises
where the arc length of curve y and t is
s is 7.10.1. Calculate the curvature (in an ar-
time, so that d 2 s/dt 2 is the (tangenrtial, i.e. bitrary point of the curve) of (a) an ellipse
directed along the tangent to y) acceleration x = a cos t and y =
b sin t, (b) a hyperbola
of the particle, g is the acceleration of grav-
y = 1/x, and (c) a fourth-order parabola y =
ity, a is formed by the tangent l to curve y ax 4 .

and the horizontal, and mg sin a is the (ver- 7.10.2. Find the evolutes of (a) an ellipse
tical) projection of the force of gravity mg x = a cos t and y —
b sin f, and (b) a hyper-
on l (Figure 7.10.10; see Chapters 9 and 10). bola y =
1/x.
The period of oscillations of the pendulum is, 7.10.3. Prove that the points of maxima
strictly speaking, dependent on the swing or minima in the curvature of a curve corres-
(i.e. on the uppermost point on curve y at pond to the cuspidal points of the evolute of
which the pendulum starts oscillating). Huy- this point (see Figures 7.10.6 and 7.10.7).
gens posed the question of what shape should 7.10.4. Prove the validity of formulas
curve y have so that the pendulum bob swing- (7.10.7a).
ing along this curve would take exactly the
same time to complete swings of large and of
7.11 Solid Geometry Applications
small amplitude. By analyzing the differen-
tial equation (7.10.14) (which was not a sim-
of Integral Calculus
ple task because at that time there was no
differential and integral calculus) Huygens
In Section 3.6 we obtained the formula
found that y must be a cycloid But how do we
. b
make a pendulum move along a cycloid?
It was in this connection that Huygens devel-
V=^S(x)dx, (7.11.1)
oped the general theory of involutes and found a
that the evolute of a cycloid is also a cycloid
where S the area of a section of
(x) is
a solid by a plane x
constant per-—
pendicular to the x axis (we advise the
reader to repeat the derivation of this
formula). This formula was used to
obtain an expression for the volume of
a pyramid The volume of a cone was
.

obtained in exactly the same manner.

Place the origin of the system of coor-
Figure 7.10.10 dinates at the center of the circle of the
 ) s

272 7 Investigation of Functions. Some Problems from Geometry

Figure 7.11.1

Figure 7.11.2
cone base and send the x axis along
the altitude of the cone upward (Figure
7.11.1). Let S x be the area of the (.
have equal altitudes and if sections made
section of the cone by a plane per- by planes P and Q parallel to the bases
pendicular to the altitude and distant and at equal distances to them always
x from the base. This section is a circle have a given ratio the volumes of the
,

of radius r x From the similarity of

.
two solids have this given ratio to each
triangles we have rjr (H x)/H, = — other. Indeed, in calculating the vol-
where r is the radius of the base and H umes of such solids using formula
is the altitude of the cone. Whence, (7.11.1), where the x axis is perpendic-
r x = ( rlH ) (H —
x) and S (x) = nr x
2
= ular to planes P and Q, we see that the
nr 2 (H —
x) 2 IH
2
and, consequently, ratio of the volumes is the same as the
ratio of the integrands S (x).
H
Let a solid be generated by revolving
V= n H-x) z dx the curvilinear trapezoid depicted in
]
0 Figure 7.11.3 about the x axis. In this
nr 2 (H — x) 3 H __ Ttr 2H 3
nr 2 H case, the section is a circle of radius
H 2
3 o
~ 3 H 2 ""
3
*

y = / (x) and S (x) =

ny 2 .Using
(7.11.1), we find the well-known formula
To obtain the volume of a sphere of
for the volume of a solid of revolution :

radius R, we place the origin at the

center of the sphere (Figure 7.11.2). b

The section cut by a plane perpendicular V —n \

z
y dx. (7.11.2)
to the x axis and distant x from the a
origin is a circle of radius R x By the .

Let us find, say, the volume of a solid

Pythagorean theorem
and so S {x) = nR x2
Rx = YR— — x
2 2
generated by the revolution of the
= n {R 2 x2 ), upper half of the ellipse x 2 /a 2 y /6 =
2
+
2
whence
1 about the x axis (make a drawing).
R This solid is called an ellipsoid of
V= |
n(R 2 — x 2)dx revolution. Since for an ellipse y =
-R

Formula (7.11.1) results in CavalierV

theorem, 1 ** which says that if two solids
7. 34
This theorem played an important role
in formulating the concept of the integral.
Bona ventura Cavalieri (1598-1647), a pupil of
Galileo, gave this theorem (without proof)
in his Geometria hidivisibilibus (The Geometry
of Indivisibles) (1635).
 a z

Geometry Applications of Integral Calculus

273
7.11 Solid

The surface area of a sphere is readily

found by means of this formula. Indeed,
a sphere is generated by the revolution
of the upper semicircle about the x
axis. The equation of a circle is x
2
+
y
2 = j?
2
,
whence

y •= V R —x 2 2
,
Yr —* 2 2

(7.11.4)

Figure 7.11.4 Substituting into (7.11.3) yields

R
(b/a) Y —x 2 2
,
formula (7.11.2) yields
F = 2n\ Vi? 2 — 2
* - dx
a J Y
V = jt
j
~ (a? —x 2
)
dx
-R
= 2nRx\\-R — 4nR 2
R
.
-a

Moreover, from (7.11.3) and (7.11.4)

For a — b =R we
obtain, as expected we can easily find the surface of a
spherical layer {spherical zone) cut out
(4/3) nR
3
the volume of a sphere of
of a sphere by two parallel planes
radius R.
Now let us derive the formula for the
x =
a and x b
—
=
a (see Figure 7.11.2 >
if, say, b R, the layer becomes a
surface area of a solid of revolution
We consider a solid spherical segment or spherical cap). In-
(Figure 7.11.4).
bounded by sections passing through deed, we have
the points x and x + dx. Denote by dF b

the lateral surface area of this solid. F = 2n [

J
/i? 2 -x 2 -£M=
Yr —* 2 2
Regarding it as the frustrum of a cone, a
we get
dF ~ +
= 2nRx &
-- 2nR {b — a).
a [y (a:) + y (x dx)] ds, |

[a

where ds is the length of the small por- Thus, the sough t-f or surface area
tion of the curve by revolving which depends only on the altitude b a of —
we get the surface of the solid; ds = the layer (segment) and the radius R
yr+WW
sum + (x)
dx (see Section 7.9). The
y (x
2

+
dx) can be replaced
of the sphere, which is really a remark-
able and beautiful result.
y
with 2 y (x), disregarding the quantity
7 35 Formulas (7.11.2) and (7.11.3) can be
y {x) dx as compared with y (a:).
*

written also in a different way. Suppose the

Therefore object depicted in Figure 7.11.3 is a metal
plate of constant thickness. In this case the
dF ~2ny yx) Yi + (y' {x))
2
dx. mass M of this plate is

The entire surface area of the initial b

solid of revolution is
M=p ydx = \iS, (7.11.5)
j
b a

F=2n J
y(x)Y 1 + W Wf dx. (7.11.3)
2
where \i is the density of the material of the

a plate (density per unit surface area; see Sec-

b
7 - 35 in the expression dF of the
Note that —
tion 9.13), and S ydx is the surface area
sum y (x) y (x + +
dx) is multiplied by ds , so ^
that the quantity we ignore is of the order of a
dx ds c^. dx 2 . of the plate. The distance yc from the center

( 8-0946
 274 7 Investigation of Functions, Some Problems from Geometry

Figure 7.11.5

of gravity to the x axis is expressed by the

formula
b b

Vc
j
^ dx
~~w aj yidx ( 711 6)
-
S (x ) = 3X [yt (x)]
2 — jx [y 2 (a;)]
2

a = n {y\ —
(see (9.13.12)). Comparing (7.11.2) and (7.11.6),
and, therefore, the volume of the solid of rev-
we see that
olution is
V= 2n Sy c = 2 ny c S, (7.11.7) b

or that the volume V of a solid revolution is V=n ^(y\—yl)dx. (7.11.8)

equal to the area S of the figure whose rotation a
generates the solid multiplied by the circumfer-
ence 2ziyc of the circle described in the process
On the other hand, the distance yc from the x
axis to the center of gravity (centroid) C of
of rotation by the center of gravity of the figure.
(Here the center of gravity, or centroid, of a
a homogeneous plate shaped as in Fig- K
ure 7.11.5 is
flat figure is assumed to coincide with the
center of gravity of a homogeneous plate whose b
shape would be that of the figure.) This theo- b
i (vl —Vl)dx
rem is known as the (first) theorem of
Pappus 7 *
36 yc = ~2s 1

.
J
Formula (7.11.7) and the theorem of Pappus ° 2 (vi-y^ dx
can be modified that they incorporate
so j
the case of a solid generated by the rotation (7.11.9)
of a figure of arbitrary shape, and not only
a curvilinear trapezoid.
(see 9.13.11)). Hence here also V 2nycS. =
Take, for example, the solid generated by Formula 7.11.3) can be interpreted in a
the rotation about the x axis of the figure de- similar manner. Suppose that the surface of
a solid of revolution is generated by the rota-
picted in Figure 7.11.5. In this case the section
of the solid by a plane perpendicular to the x
tion of an arc AB
(see Figure 7.11.4) and imag-
ine that the arc is a homogeneous heavy string
axis is a circular ring (annulus) bounded by
the circles of radii y 1 (x) and y 2 (x). Hence with a constant density a (mass per unit
length; see the text between formulas (9.12.1)
and (9.12.2)). In this case the distance yc
7 36
Pappus of Alexandria (end of 3rd cen- from the center of gravity of the string C
tury A.D.) was the last of the great Greek to the x axis is given by the following for-
mathematicians; in his works he set forth mula:
(usually without any proof) many results ob-
tained by his predecessors and also by himself. si Si
The fact that the theorems of Pappus are often ds=
Vc
~~m~ °y ( x ) y (x) ds
called the Pappus-Guldin theorems or even j ~r f
simply Guidin' s theorems in the literature is
completely groundless. Both theorems were o
formulated by Pappus, while their proofs, put
forward by Paul Guldin (1577-1643), a Swiss — j
v(x) VI + (»' (z)) 2 dx; (7.11.10)
monk and an amateur mathematician, were a
highly doubtful and substantially weaker than
the proofs of the theorems given by B Cavalieri .
here a:(5 0 ) = a, ;r( 5l ) =& ?
ds is the element of
si b
and the famous Johann Kepler (1571-
1630) (the latter proofs were harshly, but un-
justly, criticized by Guldin).
length of arc AB, L= jds=
j
V 1 +(i>'(*))
2 *:
7.12 Curve Sketching 275

Exercises

7.11.1. Find the volume of a cone using

the fact that the cone is a solid generated by
rotating a right triangle about one of its legs.
7.11.2. Find the volume of a solid generat-
ed by rotating a figure bounded from above

Figure 7.11.7
by the curve y =
axis,
^
x, from below by the x
and on the right by the vertical x = 2
about the x axis.
is the length of the string, and oL is M— 7.11.3. Using the second theorem of Pap-
pus, find the center of gravity of a semicir-
the mass of string (see formulas (9.13.4) and
(9.13.5b)). cumference.
Comparing (7.11.3) and (7.11.10), we get
F = 2n Ly c = 2ny c L , (7.11.11)

which means that the surface area of a solid of 7.12 Curve Sketching
revolution is the product of the length L of the
arc whose rotation generates the surface by the
circumference 2nyc of the circle described by
The most primitive method of construct-
the centroid of the curve in the process of rotation. ing the graph of a function / (x) is to
This result (known as the second theorem of compute the value of / {x n ) for a large
Pappus) can be generalized so as to incorporate number of points x n The usual proce- .

the case of a curve of arbitrary shape (say the

dure is to choose points x n in the form
one that is the boundary of figure in Fig- K
ure 7.11.5) and not only arc (see Figure AB xn = x0 + na ,
with n = 0, ±1, ±2,
re 7.11.3), which is the boundary of a curvi- . . This, clearly, is quite an extrava-
..
linear trapezoid. gant method. In order to see the variation
Thetheoremsof Pappus (7.11.7) and (7.11.11)
of the function on an interval Ax we
often prove very useful. For instance, let us ,

find the volume and surface area of a torus have to choose a step a much less than
{anchor ring), which is generated by the rota- Ax or a <C Ax And if the step (subin-
,
.

tion about the x axis of a circle of radius r terval) is small, a very large number of
whose center is d distant from the axis of ro-
points are required to embrace the whole
tation (Figure 7.11.6). Here, obviously, the
center of gravity C of the circle (and of the range we are interested in. And yet we
circumference of the circle) coincides with the encounter the problem of constructing
(geometrical) center Q of the circle. Hence, the curves very often, since knowing the
volume of the torus in Figure 7.11.7 is
graph of a function provides extensive
V = 2nd X nr 2 = 2 n 2 r 2 d, (7.11.12) information about the properties of the
and the surface area is
function; for one, it immediately gives
— the number of real roots of the equation
F 2nd X 2nr = An 2 rd. (7.11.13)
—
/ {x) 0, gives the intervals within
The theorems of Pappus can be used in the which these roots lie, shows the maxima
reverse direction, so to say. The sphere in
and minima of the function.
Figure 7.11.2 can be obtained by rotating a
semicircle about the diameter that is part of The techniques considered in Sec-
the boundary of this semicircle. Then, if the tions 7.1 and 7.2 make it possible to
radius of the sphere is R, the volume will be construct graphs much faster and more
reliably and to gain a general picture
of the shape of the curve. This requires,
first of all, that we find the character-
Since we know that the volume V of a sphere
is
3
(4/3) Jii? ,we can find the distance yc istic points of the graph— maxima,
from the center of gravity of a semicircle to minima, discontinuities, salient points,
the diameter: points of inflection, etc.
Let us illustrate this fact by using
yc nR»/(n*R*) = ~ OAR. the example of the graph of a third-
degree polynomial
Similarly, using the second theorem of ,

Pappus, we can find the center of gravity of

a semicircumference (see Exercise 7.11.3). y = ax3 + bx 2 + cx + d. (7.12.1)
 276 7 Investigation of Functions. Some Problems from Geometry

tfi

5
n
\
2

! 1 I
1
i
V 1 1
i
1 1
>
~3 r2 -7 0\ 2 J 4
“/
X
1
-2
l J

-J

Figure 7.12.1 Figure 7.12.2

To be specific, we wish to construct the Let us find the points of inflection ,

graph of the function that is, points, where /" (x) 0 (see —
Section 7.10). From (7.12.3) it follows
y = 0.5x 3 + 0.75a;
2
- 3x + 2.5.
that y" 3x= —
1.5, that is, y” 0 =
(7.12.2)
(7.12.2) only at x =
0.5. Note that the graph
of a third-degree polynomial always
First we find the maxima and minima.
Equating to zero the derivative of has a point of inflection, and it is
we have unique (at this point the tangent to
,

the graph intersects the graph; see

(7.12.3)
y
'
= 1.5a;
2
- 1.5a; -3 = 0, (7.12.3) Figure 7.12.1). Indeed, if y is a third-
degree polynomial, then the equation
•whence we find the two roots of Eq. y" = 0 is a first-degree equation. It
: x1 = —1 and x 2 == 2.
always has a unique root, x 0 . At this
Let us investigate each of these values
separately. To do this, we find y". We
point the difference y y, where y is —
see that y" = 3x —
1.5, and y" (—1) =
the ordinate of the tangent to the curve,
— 4.5 <C 0 and y" (2) = 6 1.5 = — has the form A (x x0) 3 that is, — ,

= changes sign when x passes through the

4.5 >
0. This means that at x —1
value x 0 This means that at this point
the function has a maximum, .

the tangent is sure to intersect the

z/max— — 0*5 — 0.75 + 3 + 2.5 curve.
= 4.25, We return to the construction of the
graph and compute the ordinate y of the
while at x = 2 the function has a min- point of inflection to get y 0.875. =
imum y m:n = — 2.5.
,
,
Let us also determine the direction of
Now let us see how the polynomial the tangent to the curve at the point
behaves for very large x (in absolute of inflection. Using (7.12.3), we get
value). Note that for very large x the tan a = y'( 0.5) =
3.375. Using all the
term containing x3 will appreciably foregoing arguments, we get the shape
exceed the other terms in absolute value. of the curve (7.12.2) depicted in Fig-
Therefore, the sign of the polynomial ure 7.12.2.
(7.12.2) is determined by the sign of Of course, if we do not compute any
3
0.5a; large absolute values of x):
(for other values of the function, the result-
for x 0 we have y 0, that is, the » ing graph will give only a very rough
right-hand branch of the curve goes qualitative idea of the behavior of the
up, while for x<0
we have y <C 0, function, but even such a graph enables
that is, the left branch goes down. (It is us to count the number of roots (i.e.
clear that if a in (7.12.1) is negative, the number of points of intersection
the left branch goes up and the right of the graph with the x axis) and to
branch goes down.) draw certain conclusions about their
 7 .12 Curve Sketching 277

point (the polynomial has one real root;

see Figure 7.12.1).
Finally, the derivative of the poly-
nomial may have only one (double) root
x 0 Then the derivative will be of the
.

form
y' =A (x — x0 ) 2 , (7.12.4)

whence, integrating, we get

y = ±(x-x )^C. 0 (7.12.5)

Figure 7.12.3
From (7.12.5) we see that in this
values. In our example, Figure 7.12.2, case the polynomial differs from a per-
we see that there are three roots, that fect cube only in a constant summand.
one of the roots lies somewhere between It is clear that y has neither a maximum
0.5 and 2, that the second root must nor a minimum (see Example la in
definitely be positive (it even exceeds 2), Section 7.1). The graph of this function
while the third root is negative (it is intersects the x axis at one point, which
less than —
1). can be found by equating y to zero:
The graph may be improved by comp-
uting a few more values of the function A.(x - Xoy+c= o,
for certain values of x. For example, let
us compute three more values of the that is,

function in this example. For x 0 =

we have y =2.5. This permits us to (x~
/
x0 )
3 = 3C
j- ,
x =x 0 —y.VIST
-j-.
get a better picture of the variation
(7.12.6)
of the curve between maximum and
minimum. For x =
3 we have y = Finding the maximum and minimum
0.25. We
computed this value so as to of a third-degree polynomial (7.12.1)
get an idea of the rate of climb of the and, hence, the investigation of its
right branch of the curve. Similarly, graph can always be completed because
to get an idea of the rate of fall of the by equating the derivative to zero we
left-hand branch of the curve we take get a quadratic equation whose roots
x = —
2 and get y =
1.5. Using these are not hard to find. (In general, how-
values, we obtain the curve shown in ever, the situation is not all that
Figure 7.12.3. simple.) A polynomial (of any degree) is
With this graph we can draw more preferable over all other functions in
accurate conclusions concerning the that we can always find its value for
roots: one root lies between x 0.5 = any value of x and its graph has no
and x =
1, the second between x 2 = discontinuities or corners; when x in-
and x —
3 (closer to 3), and the third creases without bound in absolute value,
is less than x = —2 (its value is most the absolute value of y also increases
likely close to x = —2.5). without bound (and the higher the
It may happen that after equating degree of the polynomial, the greater
the derivative to zero we will not the rate of this growth). But for an
obtain any real roots. This will mean arbitrary function (even if it is algebra-
that the polynomial does not have a ic) the situation may be quite differ-
maximum or a minimum. Since all ent.
that has been said about the behavior In the general case, the construction
of the polynomial for very large abso- of a graph requires studying the behav-
lute values of x remains valid, the graph ior of the function at infinity, that is,
will intersect the x axis only at one the behavior of the function asx-> oo —
278 7 Investigation of Functions. Some Problems from Geometry

function. It is also easy to see that

near x = —1 the value of y is positive
to the left of l and negative to the right
of l (Figure 7.12.4).
Further details of the behavior of the
function are related to the fact that
the values of

2/
= l--7TT (
7 12 8)
- -

for large absolute values of x will get

as close to unity as desired (since the
fraction 21 (x +
1) becomes very small
in this case). Therefore, the straight
and x -> oo. If, say, the function comes line y 1 =
serves as the horizontal
closer and closer (in value) to a fixed asymptote of the graph of the function.
number C as x increases without bound, Finally, by virtue of (7.12.8),
the function has a horizontal asymptote
2 4
y = C, that is, as x grows, the graph
__
^
/

“ y ~ ““
comes closer and closer to the straight
(Z + 1)
2 ’
(«+l)»
*

line y = C. Similarly, if at x = a the

r
that y is positive for all values of x
is,
function undergoes a discontinuity from the domain of the function, while
and, say, grows without bound as x y" is negative for x —1 and positive >
approaches a, the straight line x = a for x <C —
1. Thus, for all values of x
is a vertical asymptote of the graph of (except x =—
1 for which the function ,

the function y = / (#), that is, the graph is not defined) our function increases ;

moves closer and closer to the straight for x >—

1 the graph is convex upward ,

line x = a as x approaches a. It is also while for x <Z 1 it is convex down- —

clear that the construction of a graph ward the graph has no points of in-
;

must begin by determining the domain flection (i.e. points where y" vanishes).
of the function (the range of the indep- It is also important that y 0 only =
endent variable); in addition, we must at x =
1, which means that the graph
determine whether or not the function intersects the x axis at point (1, 0),
is even or odd (it may be neither) (see while £ 0aty =1, which means =—
Section 1.7), since in the case, say, of that the graph intersects the y axis at
an even function we need only construct point (0, —1) (Figure 7.12.4).
the branch of the graph corresponding Here are two somewhat more diffi-
to positive values of x and then apply cult examples. If
symmetry considerations.
Let us illustrate what we have just » = -fcr PJ2-9)
said by several simple examples. Take
the function the investigation is simplified by the
fact that the function is even: the
P.12.7) graph is symmetric about the y axis.
Our function is defined for all values
It is clear that this function is defined of x except x 1 and x 1; at= = —
for allvalues of x except x =— 1, where these points the function becomes in-
the denominator vanishes and the func- finite. Thus, the straight lines x 1 =
tion undergoes a discontinuity. Since as and x = —
1 are the vertical asymptotes
x approaches —1 the absolute values of the graph of the function. On the
of y increase without bound, the straight other hand, the straight line y 1 =
line l whose equation is x = — 1 is the is the horizontal asymptote (see the ex-
vertical asymptote of the graph of the pression within the parentheses in
 2

279
7.12 Curve Sketching

(7.12.9)
). It is also clear that the equa-
tion y (x) =
0 has two roots: x =
± y 2, that is, x ~ 1.41 and x ~
—1.41.
(7.12.9)
If we employ the second form of
,
we readily see that

^ , _ 2x _ _ 2
—
2-2i-2j
—l
y ^2 — 1)2 ’ y (x 2 l)
2 (a: 2 )
3

— 6z 2
+2
(*
2 —l )
3 *

Thus, the derivative y' vanishes only

at x 0; =
since y" (0) 2> 0, the =
value x =
0 corresponds to a (local)
minimum in the function, y mln =
y (0) =
For negative x the function
2.
decreases and for positive x it increases; |
x |
<
1 and x |Y2 | > ~
1.4. Since
it has no points of inflection because the values of this function can be found
y" (x) is never zero; at x2 1, that is, < by extracting the square root of the values
for —1 x C C
1, the second derivative corresponding to the function depicted
is positive, or the function is convex in Figure 7.12.5, we can immediately
downward, while for x 1 the | | > construct the graph (Figure 7.12.6). A
function is convex upward. The graph fuller idea of the behavior of the curve
of the function is shown in Figure can be obtained by calculating several
7.12.5. values of y (x) corresponding to certain
Finally, we take the function points on the curve or the values of
y' (#), which give the slopes of the
z2 — tangents to the curve at various points
y=]A x — 2 1
=
(
x y (^)) (for instance, for the curves
,

1 in Figures 7.12.5 and 7.12.6 it is

(7.12.10)
advisable to find the slope y' (x) of the
tangent to point (1.41, 0)).
whose domain coincides with the do- Of course, the details with which a
main of positivity of function (7.12.9), specific curve is built depend on the
that is, the function is defined only for reasons that prompted us to investigate
the curve in the first place; in some
cases it is quite sufficient to obtain a
rough qualitative picture of this behav-
ior, while in others we may wish to
obtain a fuller idea of the behavior.
What we have discussed here in
connection with curve sketching per-
tains, of course, to “school” methods of
constructing curves that are specified
by equations. Quite different possibil-
ities have been opened up by the use
of electronic computers. It is possible
to program a computer in such a way
that it plots the curve of a function
on the monitor screen when you key
in the various values of the function
and those of the independent variable.
280 7 Investigation of Functions. Some Problems from Geometry

Exercises 12* + 15 =
— 0,x —
(b) 4a: + 15a: - 18* - 2 =
3 2

0, (c) 2a:
3 2
4a: + 3 = and (d) x — 0,
3

7.12.1. Find the maxima and minima of the

7.12.3.
x 2 -j- 2 0. =
Construct the graphs of the fol-
following functions and plot the graphs of these
functions: (a) y =x3 — 3a:
2
+= 2, (b) y = lowing functions:
x3 — 3a:
2
+
3x — 15, and (c) y x3 3a:
2— + (a) = -
(b) y
2 =x 3 — x), (c) 2 =
6a: -f- 3.
y
Jz_ f t (1 y
7.12.2. Determine the number of real roots
3 — 2 — a:
4 (l
+ a:), and (d) y* =x 2
(1
— x) 2
.

of the following equations: (a) 2a: 3a:

 )

Part

Higher Math Applied to Problems

of Physics and Engineering

Chapter 8 Radioactive Decay and Nuclear Fission

8.1 The Basic Characteristics dimensions of the probability of disin-

of Radioactive Decay tegration to are 1/s. 8 1 *

The initial condition consists in speci-

The basic law of radioactive decay fying the number of atoms at the initial
which was established in experiments,
,

time: N=N
0 at t t0 — .

states that the ratio of the number of Solving Eq. (8.1.1) by the method
atoms disintegrated in unit time to the given in Section 6.6 (see Eqs. (6.6.9)
total number of atoms is a constant and (6.6.10)) and using the initial con-
that depends only on the species of dition, we find that
atom. It is understood that the total
number of atoms
extremely large.
is
N (
t = AV'®* (8.1.2)

This ratio is called the probability (we advise the reader to perform all the
of disintegration. Denote by N(t) the computations). However, when the de-
quantity of atoms that have not disin- rivative is proportional to the desired
tegrated by time t. At time t dt + function, a simpler solution to the
there will be N(t +
dt) untransformed equation can be offered.
atoms. For this reason, during time dt In Chapters 4 and 6 we discovered
(from t to t + be N(t)
dt) there will — that the derivative of the exponential
N(t +
dt) ~ —dN disintegrations of
atoms. If we divide the ratio of the 8 1

number — dN of atoms that have disin-

Consequently, probability here is not to-
-

be understood in the sense of the assertion

tegrated during the time interval dt that, as in coin tossing, the probability is
to the total number of atoms N (this one half that the coin will fall heads. The
definition of the probability of disintegration
ratio is the fraction of disintegrated
as the ratio of the number of disintegrations
atoms) by dt we get the probability
, per unit time to the initial number of atoms
of disintegration co =
—dNIN dt from , holds true only for the case where the num-
which it follows that ber of disintegrations per unit time (say, per
second) constitutes a small fraction of the to-
tal number of atoms. The exact definition of

<
8- M )
probability of disintegration is given by the*
formula co =—
(1 /N)dN/dt, that is, the prob-
ability of disintegration is equal, by defini-
tion to the ratio of the number of disinte-
From this relation, recalling that the ,

grations during a small time interval to the-

dimensions of dNIdt are the same as total number of a atoms and to the magni-
those of the ratio N/t, we see that the tude of the time interval.
 —

282 8 Radioactive Decay and Nuclear Fission

function is proportional to the function was created, there were N 0 atoms.

utself: During the time interval from t to
t + dt the quantity of atoms that dis-
d(a—x )
---
7
— const
,
x x
ar. integrated was approximately
dx

In particular,
— dN = oA
c dt.

All the atoms of this group lived rough-

d (Cekx) _ dtgkx ly the same lifetime t. Among the
dx
atoms taken at the initial time t 0 =
if C and k
are constants. Recalling this there are groups of atoms that will have
property of the exponential function, different lifetimes: from the common-
let us suppose that the solution to to-all-atoms time of creation to the
Eq. (8.1.1) is of the form distinct-for-various-atoms time of dis-
= integration. To find the mean lifetime,
.N Ce ht (8.1.3)
,

we must multiply the lifetime of each

and let us try to choose C and k such
(8.1.3) group by the number of atoms in the
that both the equation and the initial group, add these quantities for all
condition are satisfied. Differentiating groups, and divide the result by the
we get total number of atoms in all groups.
Since we have to add a very large
-4?-
dt
= Cke ht = kN. number of very small terms, the sum
can be replaced by an integral, and
Substituting into Eq. (8.1.1) yields therefore (compare with Section 7.8)
.kN = —
coA, whence k — — co. Assum- oo oo
ing t =
0 in (8.1.3) and using the
initial condition, we get C And =N taNdtj aNdt. (8.1.5)
0 .
j
so N=N 0 e
~ cot
where the quantity
,
0 0

— co£ in the exponent is dimensionless, We substitute the expression for N

as it should be. from (8.1.2). The denominator on the
Radioactive atoms are characterized right-hand side of (8.1.5) is
by their half-life T, which is the time oo oo
during which the number of atoms N oN dt=
c
diminishes via disintegration by one ^ j
half the original amount. Let us deter- o o

mine the half-life T. From (8.1.2), - (Of

N (
T) 0 e~®
=N T On the other hand, . = (dWo e~ at dt = CO An
0

= N0 ,

A (T) 0 1 2,
=N by definition. There- j
o
fore i\V -0)T = NJ 2, or e -0)T = 1/2,
as was to be expected, since the integral
that is,

— (i>T = — In 2, T = 7
—CO (0
(8.1.5)
in the denominator yields the total
number
which, clearly,
disintegrated atoms,
of all
equal to the number is
(8.1.4) of atoms existing at the initial time.

Hence, the inversely pro-

half-life is
We
integrate the integral in the
numerator on the right-hand side of
portional to the probability of disin-
tegration.
by parts, setting t /, e~
at dt = =
Prior to disintegration, every atom
dg, and —(1/co) e~ g. As a result =
exists a certain period of time, which
we have
oo
is called the lifetime of the atom.

We
wish to find the mean lifetime t coA 0
j
te dt
of an atom of a given radioactive ele- o

ment. Suppose that at the initial time I"

i = 0, say when the radioactive element

=wN 0
(
—Lte-»*+ \
4-
 (

Measuring the Mean Lifetime of Radioactive Atoms 283

8.2

was the content of radium in the rocks 10 000

—M- From we now obtain
years ago, 10 6 years ago, 5 X 10 years ago
(5 X 10 9
is the age of the earth)?
9

(8.1.5)

_ l 8.2 Measuring the Mean Lifetime

7 " (
8 . 1 . 6)
(dN o
“
(0
’
of Radioactive Atoms
or that the mean lifetime of an atom is The mean lifetime t of various radio-
exactly the inverse of the disintegration active atoms is extremely diversified.
probability. Using this fact, we can To illustrate, let us take the several
write the basic equation (8.1.1) and known isotopes of uranium. One, with
its solution (8.1.2) as atomic weight 238 (U 238 ), has a mean
9
lifetime of 7 X 10 years. Another
235
(U ) has a mean lifetime of 10 9 years
(the fission of uranium-235 in nuclear
N = N e-W. 0 (
8 . 1 . 8)
power plants is the main source of
We must not forget that the time t atomic energy). The mean lifetime of
isthe independent variable; the number radium is 2400 years. 8 2 -

of atoms depends on t. On the other However, it would be wrong to think

that the mean lifetimes of all radio-
hand, f is a constant that describes
active atoms are measured in thou-
the given species of radioactive atom.
sands of years. Among radioactive sub-
From (8.1.8) it is evident that during
stances that occur in nature and were
time t the number of atoms diminishes
studied by Marie and Pierre Curie and
from N 0 to N 0e
_1
= Nje, by a factor
Ernest Rutherford we find polonium
of e, which is roughly equal to 2.72.
with a mean lifetime of about 200 days,
By formula (8.1.7), the initial rate
radium A with a mean lifetime of 4 mi-
of disintegration is such that if the
nutes, and radium C' with a mean life-
number of atoms decaying per unit ~
time of 2 X 10 4 second.
time did not fall off, all the atoms would
During recent decades, a huge
disintegrate in time t. Indeed, at t 0 = number (over 1000) of different radio-
there were 0 atoms N
and the rate of active substances with a vast range of
disintegration was (dNldt) t==0 = mean lifetimes have been discovered
— NJt . At that
complete disin- rate, in connection with the development
tegration requires a time equal to t. of nuclear physics and the use of atomic
From (8.1.4) it follows that co = energy.
(In 2)/7\and so t = 77 In 2 1.45 T. ~ If at time t there are N(t) untrans-
formed atoms, then there will be n(t) =
Computationally, the quantity t is
more convenient than the half-life T.
(x)N(t) = —
dNldt disintegrations (of
atoms) per unit time. The quantity
n ( t ) (=n 0 e~ (3)t , with n 0 = —
dN/dt ) t=to ;

Exercises compare with (8.1.2)) is the rate of

8.1.1. The mean lifetime of radium is 2400 disintegration of the atoms (see
years. Determine the half-life of radium. Section 8.1).
8.1.2. We
start with 200 grams of radium. Suppose we observe the decay of
How much radium will be left in 300 years? uranium for ten years. During this time
8.1.3. Ten grams of radium disintegrated
in 500 years. How much was there at the be- about 4 X 10 12 atoms in one gram of
ginning? uranium will have disintegrated. It
how much time will elapse
8.1.4. Determine would be extremely hard to detect that
for 1%, 10%, 90%, and 99% of an original the original amount of 2.5 X 10 21 atoms
supply of radium to disintegrate.
8.1.5. The amount of radium in the earth 8- 2
Reference books frequently give the
in various rocks (the ratio of the number of
radium atoms to the number of atoms of rock) half-life T « 0.69* instead of the mean life-
comes out to about one part in 10 12 What . time t\ see Section 8.1.
 t c

284 8 Radioactive Decay and Nuclear Fission

would be diminished by 4 X 10 12 atoms. “manufactured” simultaneously at time-

Measuring the number of desintegra- t = 0. They are all prepared in the-
tions often proves to be easier than same fashion and at the same time. We
measuring the number of atoms that know that radioactive atoms are un-
have not disintegrated. stable and are capable of disintegrat-
However, by experiments involving ing. We can suppose that the disintegr-
radioactive substances with relatively ation of the atoms requires a definite
short mean lifetimes (from a few mi- time. Imagine that after the atoms are
nutes to a few days) it has been possible ready, a certain time must elapse before
to verify formula n — (which they are mature enough to disinte-
follows from the fact that n is pro- grate. But then we should expect all
portional to N) with great accuracy the atoms to mature in the same period
and, thus, to corroborate formulas of time and, at the expiration of that
(8.1.1) and (8.1.2). To do this, let us time, to disintegrate simultaneously.
calculate the number of disintegrations Imagine, further, that we have models
over small periods of time. Dividing of guns with stretched springs and gears-
+1 +1
cne ^iiuiiVoei ~fi± ^ciisrAtegrAiioflo c He ^Oi^lxokKwolVATm^auiicek /wia irxJneasu
time interval, we get the disintegration their shells when the gears (or clocks)
rate at various instants of time. reach a particular position (time). Fir-
We construct the graph of the disin- ing will be regarded as disintegration
tegration rate as a function of time. of the model. If all models are the same
How can we be sure that this curve and manufactured at the same time,
is the graph of the exponential func- the shells should be fired after the lapse
tion? We
can compute the logarithms of an identical time interval.
of the resulting values of the rate of But this picture of model disintegra-
disintegration and, on this basis, con- tion (firing of the shells) has nothing
struct a graph of In n ( t ) as a function whatsoever in common with the actual
of time t. The result should be a straight behavior of radioactive atoms. Though
line, which can be checked roughly by created at the same time, they disinte-
eye. Numerous experiments do indeed grate at all imaginable times. Let us try
yield a straight line. to find out what percentage disinte-
Thus, In n (t) is a linear function of grates during a time less than the mean
time, or lifetime. From (8.1.2) we find that the
rate of disintegration (the number of
\n n (t) — a bt , (8.2.1)
atoms that decay in unit time) is
which means that n ( ) — e
a+bi = dNIdt = —
a)W 0 e -o) * (it is clear that
a bt
e e = ce
bt
.The quantity b turns dN is negative). During time dt there
out to be negative on the graph, b = will be
— co, where co is the probability of
disintegration. Thus, experiments con- dt = dN= — oN e~ at dt 0

firm the basic result of the preceding

section and enables us to determine co atomic disintegrations, while during
by computing the tangent of the angle the time from t = 0 to t = t the follow-
of inclination of the straight line
ing number of atoms will disintegrate:
(8.2.1) to the time axis.
Actually, this is a very remarkable 7

result. 8 -
3
Imagine N 0 radioactive atoms M— \
(nN 0 e~ (i)t
dt — —N 0 e~
at
\
o
=
o
8* 3
Niels Bohr, speaking on radioactive
transformations, said in this connection:
=A r
0 (1 — e~ (dt
)
atoms.
“The meaning of the discussions on the mean
lifetimes of atoms without any indication of they begin to decay; this means that the
a definite instant of time lies in the fact that probability of disintegration is the same for
the atoms, so to say, do not grow old until them at any time.”
>8.2 Measuring the Mean Lifetime of Radioactive Atoms 285

models and atoms and the mean life-

times of the models and atoms are the
same. And yet the curves are so strik-
ingly different! Experimentation with
radioactive nuclei irrefutably rejects
the type of curve obtained for the
models. The more accurate the experi-
ment, the more precisely is the law (8.1.2)
confirmed.
We examined thissystem with models
so that the reader would not accept as
ordinary and natural the relationship
Since co = 1 It, it follows that (8.1.2) for radioactive decay and would
have cause for surprise and curiosity:
M—N 0
(l — -j) — 0.6SN 0. “Indeed, why does radioactive disinteg-
ration proceed in this fashion?”
Thus, 63% of the atoms will disinte- What is the physical meaning of the
grate during a time less than t. Similarly, probability of disintegration? A long
we compute that during the time from time ago, at the beginning of the cent-
i to 2 1 23 % of the atoms disintegrate, ury, it was suggested that radioactive
and during the time exceeding 2 1 only decay requires some kind of external
action, say the entry of a particle from
14% of the atoms (the remainder).
In Figure 8.2.1 we see two curves: outside. Then we would imagine that
one for the number of disintegrations one atom disintegrated earlier since
it was hit by an incoming particle,
in unit time for radioactive atoms (curve
while some other atom remained un-
1) and for the gun models (curve 2).
touched. But this hypothesis did not
The gun-model curve has a certain
fit the facts which stated that radio-
width here. We can figure, perhaps,
active disintegration proceeds at the
that the models were not quite exact
same rate under all manner of condi-
and therefore did not fire off quite at
tions, irrespective of temperature, colli-
the same time. The more precise the
sions of atoms among themselves, the
models, the narrower curve 2 in Fig-
action of cosmic radiation. Also, energy
ure 8.2.1.
is strictly conserved in radioactive
The area under the curve represents
disintegration, which likewise rejects
the total number of disintegrated atoms
for one curve and the total number of
the idea of some kind of outside influ-
ence.
all models for the other curve. We can
take the number of models to be equal to A second possible hypothesis is that
the number of atoms. Then both curves at the initial time of creation of the
will have the same area. The abscissa radioactive atoms they were actually
of the center of gravity of both curves
not quite alike and for this reason de-
is also the same.
8 4
This means that
-
cayed at different times. This is in
we consider models in which the mean keeping with the clock-driven models
lifetime (prior to firing) is the same with clocks set for different times. This
as the mean lifetime of the radioactive hypothesis presumes that an exact
atoms. knowledge of the state of every atom
We
have thus done everything in our completely determines the whole sub-
power to make the curves similar: we sequent history of the atom and, in
particular, determines with exactitude
took as many models and with mecha-
nism such that the total number of when the given atom will disintegrate.
If atoms disintegrate at different times
8-4
As will be demonstrated in Section 9.12, after their creation, this means that
this follows from Eq. (8.1.5). the whole business was foreordained:
 286 8 Radioactive Decay and Nuclear Fission

when created, the different atoms of What is so remarkable in the fact

one and the same radioactive substance that 100 atoms with a given atomic
were not exactly the same and the di- weight and a given number of electrons
verse decay times were predetermined are the same? If these were nonradio-
in the creation stage. active atoms, there would indeed be
This view does not hold water either. no cause for surprise. But for radio-
For each specific mode of generation active atoms there certainly is cause
of atoms of a radioactive element we enough when we recall that out of the
should have a definite relationship 100 atoms 63 disintegrate in time t
between the rate of disintegration and and the other 37 disintegrate after t.
time. Experiment refutes this supposi- What is strange here is that the disinte-
tion.
gration time is different although the
One and the same species of radio- atoms are the same.
active atom can often be obtained in It is not fruitless to wonder in this
a variety of ways: say, atoms of Mo" fashion. In the phenomenon of radio-
(molybdenium with atomic weight 99) active decay we already perceive cer-
are produced in nuclear reactors in the tain peculiarities in the laws of motion
process of fission of uranium atoms. of atomic and nuclear particles that
These same atoms were earlier obtained differ from the laws of motion of the
under the action of the nuclei of heavy bodies we are accustomed to in classical
hydrogen (deuterium) on the atoms of mechanics and
ordinary life. These
ordinary, naturally occurring, non- peculiarities studied in quantum
are
radioactive molybdenium. Experiments mechanics. All this is of course outside
have shown that, irrespective of the the scope of our book. Our aim is modest
mode of production of the atoms, the enough. It is to show that the necessity
disintegration rate is given by formula for elaborating radically new concep-
(8.1.2) with a constant value of co, tions that differ drastically from those
which characterizes the given species of ordinary mechanics stems from the
of atom. Consequently, it is precisely very simple facts about radioactivity
the basic equation dN/dt coN that = — that can be comprehended by any
all experiments confirm, with the char-
school child. To realize that the old
acteristic co of atoms of a given species conceptions were insufficient, it was
quite independent of external condi- necessary to doubt, to wonder, to be
tions.
surprised.
This equation is pregnant with mean- In his autobiography, Albert Ein-
ing: all radioactive atoms of a single
species are identical. The probability
stein —the greatest physicist of the
twentieth century— notes the surprise
of disintegration does not depend on and wonder that he experienced when
how and when the atoms were obtained. he first saw a compass and perceived
One hundred freshly produced atoms the mysterious action of a magnetic
disintegrate in exactly the same fashion force that passes through paper, wood,
as in the case where 10 6 atoms are the earth and acts on the compass needle
generated, a time interval elapses such without any direct contact. He wrote
that 100 atoms are left, and we consider that this wonderment served as a tre-
the fate of these 100 remaining atoms. 8 5 -

mendous impetus for a further search.

8 *6
He wrote of curiosity, which he claims
Characteristic of the exponential func-
tion is that any portion of the curve is similar
“the modern methods of teaching have
to the whole curve. Indeed, let us begin reckon-
ing time anew from time t v Denote by t the Thus, the law of disintegration for the num-
time reckoned from this instant: t t t
±
= — ,
ber of particles 1N =N 0
e(Otl
remaining after
t = *!+ t. ThenN = N e~ = N e~^ tl+X ^=
0
(Ot
0
the previous decay is exactly the same as the
law of disintegration of
iV 0 e'°)il e"
cox
= N e‘ where = N^*
1
co r
'

,
1
freshly obtained
particles. This is precisely what is asserted
is the number of atoms at time tx . in the text.
 -

8.2 Measuring the Mean Lifetime of Radioactive Atoms 287

all but stifled.” Einstein himself evinced In this study the 98th element, call ~
an extraordinary capability for wonder- fornium with atomic weight 252 was-
,

ment and he was able to derive inspi- irradiated with neutrons. A neutron is
ration and impetus for the creation of captured by the nucleus of a Ca 252
theories out of the most mundane facts atom, which transforms into a Ca 253
'

of everyday life. For instance, under- atom. Californium-253 ejects an elec-

lying the brilliant general theory of tron and turns into Element No. 99,
relativity is the simple fact that caused which is called einsteinium (symbol Es).
Einstein to wonder why different bodies and has the same atomic weight 253.
fall with the same acceleration. About 10 9 atoms of einsteinium (this
-13
Quite naturally, to wonder is not is equivalent to 4 X 10 gram) wera
enough, and to merely pose a problem deposited on a gold plate and subjected
does not suffice. Einstein combined to alpha-particle bombardment (a par-
the ability to pose a problem and to ticles are nuclei ofhelium) in a cyclo-
solve it, which means mastering the tron. This generates Element No. 100,
requisite mathematical techniques. And fermium (symbol Fm), in accordance’
yet, among a galaxy of outstanding with the reaction
scientists, Einstein is the celebrated
physicist of the twentieth century be- 99 Es 253 + 2 He 4 = 100 Fm 256 + t p
l
,

cause of his capacity to wonder and pose and mendelevium via the reaction
a problem where others were not able
to see anything out of the ordinary. 99 Es 253 + 2 He 4 = 101 Md 256 + oft 1
.

Perhaps this analysis of radioactive

decay will help the reader to see what In this notation, each chemical symbol
depths of content are hidden behind has a subscript indicating the number
simple facts and formulas. in the periodic table (that is, the number
To conclude this section, we will of protons in the nucleus). The super-
illustrate the curves of radioactive script indicates the atomic weight
decay obtained experimentally in 1955 (rounded to a whole number, which is the
by Glenn Seaborg and his associates number of neutrons and protons in the
in the United States (Figure 8.2.2). nucleus). The helium nucleus, or the a
They were first to observe Element particle, is denoted by 2 He 4 the nu- ,

No. 101 of the periodic table, to which cleus of a hydrogen atom, or simply the
they gave the name mendelevium (sym- proton, by jp 1 and the neutron by
,

bol Md) in honor of the great Russian o^

1
In
. a nuclear reaction the sum of
chemist Dmitri Mendeleyev. the subscripts on the left-hand side of
the reaction equals that on the right-
hand side, and the same is true of the
superscripts, since in nuclear reactions
allwe have is an exchange of neutrons
and protons between the nuclei.
After the a particle bombardment,
the gold plate together with the newly
formed fermium and mendelevium was
dissolved in acid, and then the fermi-
um and mendelevium were extracted
chemically. As Seaborg writes, it was
the periodic table of Mendeleyev that
permitted foreseeing the chemical pro-
perties of an element that had never
before existed in nature and had never
Minutes after bombardment
been studied. After chemical separa-
Figure 8.2.2 tion, measurements were made of the
288 8 Radioactive Decay and Nuclear Fission

radioactive decay characteristics. Fer- After chemical separation, the nu-

mium-256 (with atomic weight of 256) cleus of mendelevium rapidly (in half
disintegrates radioactively with a an hour) captures an atomic electron
half-life of about 3.5 hours. It breaks and transforms into a nucleus of fer-
up (that is, fissions) spontaneously mium. And so the precipitate contai-
into two nuclear fragments of roughly ning mendelevium also yields (when the
the same mass (the details of the fission radioactivity is measured) a disinte-
process are given in Section 8.3). gration of atoms into two fragments
The upper curve in Figure 8.2.2 with a half-life of 3.5 hours. The decay
shows the number of nuclei of fermium curve for fermium obtained from men-
as a function of time in this experi- delevium lies in the lower left-hand
ment. Plotted on the horizontal axis corner of Figure 8.2.2 Six disintegra-
is the time in minutes. Plotted on the tions were experimentally observed.
vertical axis is the number of atoms avail- Special experiments demonstrated that
able at a given time. 8 6 The scale on these six atoms could not have appear-
this axis is not uniform: the distance ed as a fermium impurity in the pre-
from the horizontal axis is proportio- cepitate being measured but most de-
nal to the logarithm of the number of finitely had formed from the mendele-
atoms. In particular, the horizontal vium.
axis (y =
0) corresponds to one re- Seaborg and his associates observed
maining atom (In 1=0), while for a total of 17 atoms of mendelevium in
zero atoms we have —
oo on the ver- that series of experiments.
tical axis. The disintegration of each The foregoing example is not very
separate atom changes the number of good as an illustration of how exactly
atoms by 1; in between disintegrations the exponential law (8.1.2) holds true
the number of atoms is constant. In in radioactive decay (see also (8.2.1)).
the experimental set-up in which each Experiments demonstrating the valid-
separate disintegration is recorded, we ity of the exponential law were success-
have a polygonal (step-like) curve in- fully carried out with more common
stead of a smooth curve. On the step- radioactive substances. On the other
like curve, each disintegration is asso- hand, the example of mendelevium and
ciated with a vertical line connecting fermium shows what peaks of experi-
two steps. The upper straight line in mental technique modern physicists
Figure 8.2.2 corresponds to the decay have attained in synthesizing new ele-
law n =
n 0 £“ t/T where t,
=
77 In 2 ~ ments and recording the disintegration
5 hours, T ~
3.5 hours. It will be of each separate atom.
seen from Figure 8.2.2 that, in all, In Seaborg’s experiments, the
there were recorded 40 disintegrations counter recording mendelevium disinte-
of fermium. The more atoms there are, grations was hooked up to an amplifier
the closer is the polygonal line to a in the loudspeaker system of the insti-
straight line. When there are fewer than tute and every disintegration event was
five atoms left, then quite naturally heard by workers in the various labora-
the probabilistic nature of radioactive tories on different floors so they could
decay leads to appreciable deviations celebrate the birth (actually the recor-
from the exponential law, which holds ded death) of every atom of the new ele-
true for large number of atoms. ment created by man (true, before the
work was over, the local fire department
got interested in these goings-on and
8-6 It is not possible to count the number the disintegration news broadcast was
of atoms available at a given instant. What stopped).
is recorded experimentally are the disintegra-
In the Soviet Union the synthesis
tion events of the atoms. The number of atoms
N at time t is computed after the exper- and investigation of the heaviest ele-
iment, when all N
atoms have disintegrated. ments are being successfully conducted
 )

Disintegration (Radioactive Family) 289

8.3 Series

by Academician G. N. Flyorov and his unknown. Indeed, neither B nor C

associates at the Joint Institute for enter into Eq. (8.3.1). From it we there-
Nuclear Research in Dubna, a city fore find that A (t) A0 =
where
near Moscow. A 0 is the number of atoms of element A
at the initial time t 0. (We assume —
that B C =0 at t =0). —
8.3 Series Disintegration Substituting the expression for A ( t
(Radioactive Family) into (8.3.2), we get an equation involv-
In a number of cases, radioactive decay ing only one unknown function B ( t ):
produces atoms that again decay ra-
dioactively, so what we have is a chain + (8.3.4)
of disintegrations: an atom of element
A transforms into an atom of element B, How does one go about solving this
which in turn disintegrates into an equation? We can find a solution if we
atom of element G, and so on. Let us first consider the fate of the group of
consider the mathematical problem of atoms of B that have formed during the
determining the dependence on time same interval from t to t
of time, +
of quantities of elements A, B, C and At. We consider the number of
will
ways of solving the problem. We de- atoms of this group that are still “alive”,
note the quantities of substances (ele- A B (that is to say, atoms that have
ments) A, B, C that have not yet de- not disintegrated by time t ), as a func-
cayed by time t by the italic letters tion of time t. So as to avoid confusion
A, B, C; thus, A A (t), B =
B (*), = about the time t when we measure
and C = C (
t ).
the number of atoms and the time of
Let the probabilities of disintegra- formation of the group, let us denote
tion of A, B, G be equal to co, v, u ,
these times by different letters, t and
respectively. Then t, respectively. At time t, the rate of
formation of atoms of element B was
)A (8.3.1)
dt oA
c (t). During the small time inter-
A
termed the parent element ).
val At, a total of Ai? 0 = co^l (t) At
(here, is
atoms of element B were formed.
We write the equation for element B How does the number of atoms in the
(the daughter element ). In unit time a
group at hand depend on time t ? For
total of vB atoms of element B disinte-
t <Z t is it equal to zero: the atoms of
grate. On the other hand, during the
interest have not yet formed since the
time we have coA disintegrations of
group itself is still nonexistent, A5 =
element A, and since each disintegra-
tion of an atom of A given rise to one
0. Let £ t. >
Observe that a time

atom of B, there are coA, atoms of B

t —t has already passed since the
group began to form. The decay proba-
formed in unit time. Therefore bility of element B is v Therefore, af- .

ter the lapse of time t t from the —

4^- — — vj5 + cdA. (8.3.2) time of formation of the group the num-
ber of untransformed atoms will be
Similar reasoning yields
AB (t) = A B e - V ^-^
0
J£L=-uC+vB. (8.3.3) = coA (t) e~ v ^~ x) At.

Equations (8.3.1)-(8.3.3) form a sys- To find the total number of atoms of

tem of differential equations In the . element B
at time t we have to add the
,

given instance, we can solve these equa- number of atoms in all groups that formed
tions one by one, having to deal each prior to t. If we take At (and| hence
time only with one equation in one AB as well) very small, then the sum
19-0946
 t )

290 8 Radioactive Decay and Nuclear Fission

turns into the integral If we set A(x) = ^L 0 e

_COT
,
we get the
concrete solution
A B (t)
B <» = At
dx =
! (8.3.6)

The solution could also have been

e -v(t-x )dx. found without resorting to a considera-
tion of the separate groups of atoms.
Now that the solution has been found,
Observe that here the variable of it is already a simple matter to guess
integration denoted by x. The argu-
is the mathematical technique that will
ment t, upon which B depends, enters lead us to our goal. The solution (8.3.5)
into the integral twice: as the upper is of the form
limit and in the integrand. When in-
tegrating with respect to x the quantity
B(t) = e- vi
I(t ), (8.3.7)

t is to be regarded as a constant. We where / (£) stands for an integral that

can therefore write depends on t. We will seek the solu-
g-v(t- T) — g-vtgVX tion in the form of a product of e~ vt
by the unknown function I ( t ) and will
and take (oe~vt out from under the in- set up an equation for / ( t ):
tegral sign as a factor that is indepen- dB d dl
dent of t. We then get dt dt
(e~
vt
I) = —ver vt I + er vi
dt
t (8.3.8)
B (t) — we- vt ^A(x)e vx dx. (8.3.5) Substituting (8.3.8) and (8.3.7) into
o Eq. (8.3.4) yields
It is easy to verify, without evaluating ~ vt
the integral, that the solution (8.3.5)
e
^r= (S)A ^'
satisfies the original equation (8.3.4) or
for any function A(x). Indeed, let us
find the derivative dB ( t)/dt By the . = ®e vt
A(t). (8.3.9)
rule of differentiating a product we get
By hypothesis, at the initial time
t
t = 0 we have B = 0, and hence I =
=
~J~ — — (ove~ A (t) e vx With
vt dx 0 at t 0. this initial condi-
j
o tion, the solution to Eq. (8.3.9) has
t
the form

-\-(oe~ vt
i(W> o
e vx dx^ .

1 (
t = j
t

a) e
vx A (t) dx .

o
Since, by the property of the derivative
of an integral (see Section 3.3), Thus, the final result is

t
t

m B (t) = e~ vt
I (t) - e~ Kt
co^l (t) e” dx.
-jf
A (t) e drj =A ( t) evt , j
( j 0
0
(8.3.10)
it follows that In this formula it is essential, so as
t
to avoid confusion, to retain strict de-
signations and not to denote the vari-
—U)ve- Vt A (x)e vx dx
j able of integration x by the same letter
o
we use for the upper limit of integra-
+ (oA ( ) = — vB + coA. tion, t.
8.4 Investigating the Solution for a Radioactive Family (Series) 291

8.4 Investigating the Solution for a the signs so as to be dealing with posi-
Radioactive Family (Series) tive quantities in the parentheses and
in the denominator of the fraction. Then
In the preceding section we brought to
completion the solution of the problem B(t)=A0 -£T (e-«-e-«). (8.4.1)
in the case of two radioactive substances
(elements). Let us now investigate this Since v <C co, it follows that co/ (co —
solution for two particular cases: v) co/co = 1.
(1) a short-lived parent element A We consider the expression e~vi —
and long-lived daughter element B, e -(M f or two successive stages. First,
(2) a long-lived parent element A and
when t<^ tB = l/v, it will be true that
short-lived daughter element B.
vt<^ 1. Then e~vt c* 1. Since t can be a
Below, in addition to the decay prob-
abilities co and v we will make use of
quantity of the order of t A and co£,
,
consequently, of the order of unity, it
the mean 1/co _and
lifetimes tA = follows that e ~ must be calculated
t B = ilv. where t A <C
In the first case, exactly. From formula (8.4.1) we get
the nature of the solution can be
tB ,

readily grasped without calculations

B(t)~A 0 1— *-«*)• ( (8.4.2)

and formulas. The entire process breaks

In the second stage, when t^> t A =*
down into two stages. First, when t
1/co, it will true that co£^ 1,
be
t A (here, by hypothe-
of the order of
is
We can disregard e -co< in this stage,
sis, <C tB and
£a so also t<^t B in the since e~ oH
is small not only with respect
first stage), element A is transformed to unity but also in comparison with
~
into element B, while during this time e vt because v <C co. We get
,

there is hardly any disintegration of

element B. During this period the
B(t) =A 0 e-
vt
. (8.4.3)

amount of B is equal to the difference Thus, the exact formula does indeed
between the original amount A 0 and the yield the same results as those obtained
amount A remaining at time t: from simple qualitative reasoning.
B (£) = A 0 — A = A -A(t) 0 0
e-«t Now let us take up the second case,
that of the long-lived parent element A
= A 0 l — e-ot), £<
( t B .
and short-lived daughter element B:
By
the end of this period, practically
tA > tB , co<i;.
element A has been convert-
the whole of
ed into B, and the quantity of B be- We consider the period when a time t

comes equal to the original amount of considerably exceeding tB has passed

the parent element, A 0 The quantity .
since the onset of the process. In that
of element A becomes zero. Then fol- case, element B that was formed at the
lows a slow and protracted disintegra- beginning of the process has already ful-
tion of B: ly disintegrated by time t. Since B dis-

=A integrates rapidly and in a short time,

B(t) vt
0 e~ , *>*A .
at every given instant of time there is
We
will show how these qualitative only availabbe the quantity of ele-
ideas follow from the exact formula. ment B that has recently formed. What
For the case of two radiactive elements we have here is a steady state (also
A and B, we obtained in the preceding known as a stationary state): element B
section the formula is formed from A and straightway disin-
tegrates; element B does not accumulate
because it decays rapidly, but does
not disappear completely because A is
In our case £ A <C t B and co^>i;, and producing new quantities of B all the
so it is more convenient to interchange time. In a steady state system, there
19 *
 0

292 8 Radioactive Decay and Nuclear Fission

are just as many atoms of B disintegrat- Thus, the steady state is attained in
ing in unit time as there are atoms of a time roughly equal to the mean life-
B being formed from A, so that B time of element B (we recall that our
here changes but little. Mathematically, assumption that the rate at which sub-
this condition is written as v B oA, ~ stance B is formed is valid only approx-
whence From the condition £ A <C
imately).
t Bevident that the quantity of
it is
B(t)~^A{t) = 4?-A (t). (8.4.4) element A changes but little during
this time.
In the steady state the instantaneous
On the whole, the approximate exam-
quantity of B is proportional to the quan- ination in the case of a short-lived
tity of A and always represents some
daughter element yields the following:
small fraction of A. This fraction is
small because in the case at hand (the BW= i~ir) = 0
t=ti)A o t if *<^6,
t
second case) t B <^~t A and hence tn/tA^
(8.4.5)
1, for otherwise there would be no
steady state. B — B6t — AQe-®* = (titsAQe-®* if
(t)
1

How do we obtain the steady-state

equation from the exact differential
equation dBldt vB coA? Evi- = — +
t > tB •

dently, if we take it that dBldt is small We get the function B = B (t) in the
in comparison with each of the two form of two lines: first the straight line
terms on the right, replacing dBldt 1 (Figure 8.4.1), or a linear function,
by 0 we approximately get then an exponential curve. Figure 8.4.1
0= — vB + coA, or vB = oA. was constructed on the assumption that
tA — 10lB ;
instead of the exponential
Let us now examine the beginning curve we depicted the straight line 2
of the process. At t = 0 we have A =
.

A 0 and B = 0. This means that at It easy to verify that for t

is tB =
the two formulas in (8.4.5) yield almost
the beginning we do not have a steady
the same result.
state, since by steady-state formulas
Let us see what the exact solution
"we should first have
(8.4.1) to Eq. (8.3.4) gives us in the
case at hand, where u^> co and ^ B <C
t
A . denominator
In we neglect
the
(the subscript on B denotes a steady, or co compared with v. For ut^> 1 we al-
stationary, state). At t 0, element B = so neglect e~ vt in the parentheses. This
is forming at the rate dBldt coA 0 = , yields
while there is no disintegration of B
at all at the initial time since B 0. = B ~ A -^e- at 0 , (8.4.6)
It is possible to determine the time
t x during which, for the initial (con- which is precisely the steady-state so-
stant) rate of buildup of B, the quan- lution.
tity 5 st will be attained. Indeed, if
the rate of formation of B remains con-
stant, equal to (dBldt) t = 0 then ,

B= t (-%.-)
\ dt } t = .

Putting 5 st = (co/v)A 0 and (dBldt) t=0 =

<»A 0 ,
we get the desired time

Figure 8.4.1
 ~

8.4 Investigating the Solution for a Radioactive Family (Series) 293

We determine how close the solution tA is greater than all other times,
is to a steady state by how fast e~vt and
then all the results re-
t-B, tc ,
£ d ,
falls off. For very small values of t
ferring to the steady state are obtained
(when vt < 1, sothat cot is sure to be
just as easily as in the case of two ele-
small), we get the following by expand-
ments, A and B.
ing e~ xi and e ~ in a series and confin-
Sometimes the easiest way consists
ing ourselves to the first two terms:
in obtaining an exact solution valid
B ~ Aj (1 — —1+
co< vit) = for arbitrary v and co (in our case), from
which we then (for z;<C co or v co)
(8.4.7) obtain, via mathematical manipula-
which also coincides with the approxi- tions, some simple approximate for-
mate result. Actually, however, the mulas for the two extreme cases. But
exact formula yields a single smooth this is not yet all! If a simple approxi-
curve without discontinuities and sali- mate formula has been obtained in a
ent points (the dashed curve 3 in Fig- simple yet long-winded manner via
ure 8.4.1). The approach of this curve to the general solution, then alongside
the steady-state solution depends on this there should be another, simple,
how fast e~ vt diminishes. For instance, way of obtaining the approximate for-
for e~ zt to make a correction of the or- mula. One should always attempt to
der of 10%, we must have vt 2.3, find simple pathways because there

or t 2.3/v =
2.3£B Here, owing to -
will invariably appear problems in

the smallness of cot, we assume that

which the approach to an exact solu-
tion is insuperably complicated and
^
i Thus, true enough, the tran-
only a simple approximate approach
sition from the stage of initial buildup
to the stage where the solution is equal,
makes it possible to advance in the so-
lution.
with sufficient precision, to the steady-
In practical situations, exact for-
state solution takes place in a time
of the order of the time of disintegra-
mulas come up just as rarely as algebra-
ic, trigonometric, or other equations
tion, £B .
with solutions in whole numbers, al-
The exampleof a radioactive family
though most of the textbook problems
highly instructive in the sense
(series) is
lead to exact formulas, just as problem
that obtaining a general exact solu-
books for junior classes abound in equa-
tion does not in the least signify that
tions that can always be solved in
the work is at an end. The construc-
whole numbers.
tion of approximate theories for various
Observe that the conceptions of ra-
limiting cases is an absolutely neces-
dioactive families account for the strange
sary part of the work and the existence result of exercise 8.1.5 about the
of an exact formula does not at all take
amount of radium in the past: radium
the place of an approximate theory.
is a descendent (true, not direct, but
Approximate, yet clear-cut and picto-
via a number of intermediate substances)
rial conceptions serve as a check on an
of uranium-238. It is therefore not
exact formula.
correct to regard the present-day supply
Also, approximate theories give us
of radium as the result of the decay of
important new qualitative concepts
primordial radium. Actually, radium
such as that of the steady state. These
is in a steady state with uranium. From
can more easily be remembered and
the equation
they possess a broader range of appli-
cation than do the exact formulas.
For instance, in the case of a radioactive 5= —A V
family consisting of several genera-
tions, A B C D, the exact for- we find that the quantity of radium
mula is extremely unwieldy. But if B = 10 12
corresponds to the uranium
294 8 Radioactive Decay and Nuclear Fission

content of these nuclei lies within the range

from 35 to 57, the sum of the charges of
A = 4A- 5 = 3 x iOeB = 3 x 10-6 . two fragments always being equal to
the charge on the nucleus of uranium,
We have approximately found the pres- that is, 92 elementary charges. The
ent-day amount of uranium-238 in sum of the atomic weights of the two
rocks. The original abundance, 5 X
9
fragments is equal to 235 + 1 v, —
10 9 years ago, was twice as much, of where 235 is the atomic weight of ura-
the order of 6 XlO -6 These magnitudes
.
nium-235, 1 is the atomic weight of
are quite reasonable, unlike the the neutron that caused the fission,
results of the exercises in Section 8.1. and v is the sum of the weights of the
neutrons generated in the act of fission.
An enormous energy of 6 X 10 10 J/g
8.5 The Chain Reaction in the Fission
(per gram of fissioned uranium) is re-
of Uranium
leased in the fission process. Thanks to
In 1938, Otto Hahn and Fritz Strass- this great energy, right after the fission
mann in Germany and Irene and Frede- process the fragments rush apart at
ric Joliot-Curie in France demonstrat- speeds of about 10 cm/s; then they de-
ed that when a neutron enters a nu- celerate and their kinetic energy is
cleus of uranium, fission occurs in which transformed into heat.
the nucleus breaks up into two large The source of this energy is the elec-
fragments with the simultaneous emis- tric repulsion of two like-charged frag-
sion of two or three new neutrons. Ura- ments. Before the nucleus is separat-
nium with an atomic weight of 235 ed into two parts, the nuclear forces
(uranium-235 for short) is very active between the particles that make up the
in this respect. Naturally occurring nucleus balance the electric repulsive
uranium contains about 0.7% of ura- forces. But as soon as the nucleus has
nium-235 atoms and 99.3% of uranium- broken up into two separate fragments,
238 atoms. 8 - 7
The
fragments of
fission the repulsion of these two fragments is
uranium-235 are medium atomic-weight not countered in any way and so they
nuclei from 75 to 160. The charge fly apart at high speed. The fragments
are very quickly brought to rest in a
8 7
-
This was followed in 1939, in the labo- dense substance. Their time of flight
ratory of I.V. Kurchatov in Leningrad, by the is between 10“ 13 and 10 -12 s. In this
demonstration (carried out by the Soviet
scientists G.N. Flyorov and K.A. Petrzhak)
time they traverse distances from 10 -4
to 10~ cm. The kinetic energy of the
3
that uranium-238 is capable of undergoing
spontaneous fission without the entry of any fragments is converted into heat. The
neutron, althought the probability of this neutrons produced in fission have velo-
event is extremely low. The probability of
cities of about the same order as the
radioactive decay (with the emission of an a
particle) of uranium-238 corresponding to fragments (about 2 X 10 9 cm/s).
a half-life of 4.5 X
10 9 years is a) 5 = X Of crucial importance for the practi-
10 -18 s _1 while the probability of the sponta-
, cal utilization of the energy of nuclear
neous fission of uranium-238 is lower by a fac-
fission is the fact that a fission event
tor of 10 6 that is, it is equal to about 5 X
10~ 24 s -1 Thus, in one second in one kilogram
.
caused by one neutron gives rise to more
of uranium (which is about 2.5 X 10 25 atoms) than one neutron. It is quite clear that
there occur roughly 10 7 radioactive disintegra- if the neutrons do not leave the system,
tions and only 10 events of spontaneous fission.
their number will increase in geomet-
On the other hand, in the very heaviest ele-
ments, spontaneous fission becomes the most ric progression with time, that is, in
probable decay process (see the end of Sec- accordance with the law of the exponen-
tion 8.2, in particular Figure 8.2.2, where the tial function. The rate of energy release
decay curve of mendelevium is given). The
will build up by the same law, in
roblem of the chain reaction that we consider
elow does not involve spontaneous fission proportion to the number of neutrons.
at all. And even if at the onset of the process
 "

8.6 Multiplication of Neutrons in a Large System 295

there were few neutrons, their number

builds up so fast that the energy will
be released at a rate convenient for
practical use (for instance, as a source
of energy for a nuclear power plant), and
in just a short additional space of time
the energy release will build up to such
an extent that an atomic explosion will
take place. In reality, part of the neu-
trons leave the system, some are cap-
tured by other nuclei without causing fis-
sion. Wecan utilize this to control
the number of neutrons and, in a parti- How often will a neutron in flight
cular case, attain a steady-state system inside the metal hit a nucleus of ura-
in which the number of newly formed
nium?
neutrons in unit time is equal to the In a small time interval dt a neutron
number of used up 8 neutrons, so that traverses a distance of vdt. Let us pic-
the number of neutrons in the system ture a cylinder whose axis is the route
remains the same in the course of time, covered by the neutron; the radius of
and the energy can be released at a con- the cylinder is equal to the radius R
stant rate. That precisely is the regime of the uranium nucleus. The neutron
we need if atomic energy is to be used collides with the nuclei whose centers
for peaceful purposes. lie inside the cylinder. If this is so,
Our immediate task is to set up and the path of the neutron will pass at a
investigate the equation describing the distance less than R from the center of
number neutrons is a system as a
of the nucleus, and so the neutron will
function of time. This will be done in hit the nucleus and enter it. The vol-
the sections below. ume of the cylinder is equal to Jii? 2 vdt.

Figure 8.6.1 clarifies these ideas. The

path of a neutron is shown by a dashed
8.6 Multiplication of Neutrons line along the cylinder’s axis. In the
in a Large System case A the center of the nucleus lies
inside the cylinder, with the result
Let us derive an equation for the
first
that the neutron hits the nucleus, while
variation in the number of neutrons in the case B the center lies outside
with time in a very large system (say, the cylinder and the neutron does not
in a large chunk of uranium-235), hit the nucleus.
when loss of neutrons to the outside There are N
atoms in a unit volume
can be neglected. 8 The neutrons can
-

of metallic uranium and hence N

all be regarded as having the same nuclei (the dimensions of N -3
are cm ).
speed; we denote it by v. Therefore, in the volume nR vdt that
2

Fission of a nucleus occurs in roughly interests us there are 2

NnR
vdt nuclei.
half of all cases when a neutron enters There will be just as many events of a
a nucleus of uranium-235. In the other neutron hitting a nucleus during the
half, the neutron emerges leaving the small time interval dt Not every neu-
.

nucleus in the same state, with the num- tron hit makes a nucleus fission. Let a
ber of neutrons remaining unchanged. be the portion of cases a neutron hit-
The uranium nucleus is a sphere of ra- ting a nucleus causes fission (a ~
1/2
dius R of the order of 10 12 cm. in the case of uranium-235). Then the
number of fissions during time dt
is equal to NanR 2 vdt
8. 8
We
will consider the simplest case ©f
.

metallic uranium -235 without graphite mod- The quantity a nR 2 which has the
,

erator. dimensions of area since a and jc

 1

296 8 Radioactive Decay and Nuclear Fission

are dimensionless, is called the cross popular literature (see footnote 4.14).
section of fission and is denoted by Physicists and engineers speak rather
a f (the subscript f on the Greek letter of an exponential growth (in accord with
sigma stands for “fission”). the law of exponential increase) and
If there are n neutrons in the bulk rarely use (8.6.5) and (8.6.5a). The ex-
of metallic uranium, the number of ponential law (8.6.4) is characterized
fissions in time dt is equal to by the growth rate a .

Let us find the dimensions of a

nNot vdt. (8.6.1)
.

In (8.6.4) at is a dimensionless quantity

Each act of fission produces v neutrons, and, consequently, the dimensions of
but this involves the absorption of one -1
a are s The same result can be ob-
.

neutron, so the number of neutrons in tained if we recall that

every fission event increases by v 1. — —
Associated with the number of fissions
a =N (v 1) OfV,

(8.6.1) is the variation in the number of where N

measured in cm
is cr f in
-3
,

neutrons cm 2 and
,
v in cm/s.
dn = nN (v — 1) Of vdt. (8.6.2)
Let us find the approximate value of
the constant a. The density of uranium
From this equation we get is roughly equal to 18 g/cm 3 The num- .

ber of nuclei per cubic centimeter, A,

-~- — nN (v— )p,v.
can be calculated by recalling Avogad-
23
ro’s number, which is equal to 6 X 10
Set
atoms in one gram-atom of any
N (v — 1) G{V = a. (8.6.3) substance. Hence, 235 grams of ura-
Then nium-235 contain 6 X 10 23 atoms, or
6 X 10 23 nuclei. One cubic centimeter
of uranium-235 contains (18/235) X
6 X 10
23
^
4 X 10 22 nuclei, that
We already know that the solution to is, N~ 4 X 10 22 1/cm 3 Substituting .

this equation is the mean value v 2.5, v 2 X ~ ^

n (t) = n 0 e at , (8.6.4) 10 9 cm/s, and a, — (1/2) n (10~ 12 ) 2 ~
1.6 X 10“ 24 cm 2 into the expression
where n 0 is the number of neutrons in for a, we get
the system at t = 0.
To summarize, then, if the number a~ 4 x 10 x 1.5 x 1.6 X
22 10" 24

of neutrons in a system varies solely x2x 10 ~2x 10 s" 9 8 1

,

because of fission, then the number of

or
neutrons increases in geometric progres-
sion while time increases in arithme-
,
a" 1 ^ 5 X 10
-9
s.

tic progression .
To summarize,if the neutrons do not
Indeed, if we take a number of equal- leave the system, their number in-
ly spaced intervals of time, creases by a factor of e in approximately
4" ti 4“ 2Af, ti -f- 3 A £, 5 X 10" 9 second. At this rate of build-
£i, . . .,
-6
up, in one microsecond, or 10 s,
(8.6.5)
the number of neutrons has increased
then the corresponding number of neu-
~
by a factor of e 2x]08xi0 6 e 200 that = ,

trons is is, approximately, 10

0 43x20 °
10 86 -
= .

n i =^n 0 eat ^ fn u f
2
nu fn lf ..., (8.6.5a) One metric ton of uranium-235 con-
tains roughly 2.5 X 10 27 nuclei. If the
where / = e aAt .
neutrons do not leave the system, this
This way of describing the process, quantity of uranium will fission in less
growth in the number of neutrons in than one microsecond. This process is
geometric progression, is common in the an explosion of enormous force.
 8.7 Escape of Neutrons 297

Such a rate of buildup is not permis- from the mean density. Then the num-
sible if we want
to use the fission pro- ber of neutrons in this layer is
cess for generating electric power. It is
necessary that neutrons leave the sys- CSv dt — ^y-vdt.
tem and thus reduce the rate of neu-
tron buildup. Therefore the flux (the number of neu-
trons leaving in unit time) is

8.7 Escape of Neutrons nkicr 2 3^

(4/3) nr 3 ^ r
Picture a mass of uranium-235 in the
form a sphere of radius r. We have to Actually, the neutron density near
setup an equation for the variation of the surface is less than the mean densi-
the number n of neutrons inside the ty and, what is more (this was noted
sphere. Assume for the sake of simplic- above), the neutron velocities have
ity that the sphere is fixed to a thin different directions and not all the neu-
support so that it is surrounded by a trons leave the sphere. The neutron
complete void and a neutron that has flux is thus less than we obtained:
left the sphere will never enter it
again. (8.7.1)
How can we determine the neutron
flux (the number of neutrons leaving where & is a numerical factor less than
the sphere in unit time)? We make a unity. Later on, in Section 8.10, we
rough calculation. Consider a small time will compare our results with experi-
interval dt. During this time each ment and find that k is close to 0.3.
neutron covers a distance of vdt. Where If nuclear fission does not occur inside
are the neutrons that leave the sphere the sphere and no new neutrons are gen-
in time dt ? Evidently, they will have erated, then for the number of neu-
to be inside the sphere in a thin layer trons inside the sphere we get the equa-
adjacent to the surface of the sphere but tion dnldt = —
q or, using (8.7.1),
at a distance not exceeding vdt from
dn 3 kv
the surface, otherwise during time dt
they will not reach the surface, cross
nr ~ r~
n -

it, and leave it for good. But neither Setting

will all those neutrons that are inside
the layer of thickness vdt be able to -^- = 6, (8.7.2)
leave in time dt since not all neutrons
inside the layer have velocity directed we obtain
outward along the radius. In our very dn
rough calculation we will ignore this dt
— bn.
latter circumstance.
How is it possible to find the number The solution to this equation is famil-
iar:
of neutrons in the layer? There are a
total of n neutrons in the whole sphere. n = n 0 e~ ht (8.7.3)
The volume of the sphere is V =
3 The mean residence time of neutrons in-
(4/3) nr while the volume of the
,
side the sphereis, by (8.7.3),
thin layer that interests us near the
surface is approximately equal to Svdt 1 _ r
if vdt is small. Here, S =
4nr 2 is the t
i

b 3 kv
(8.7.4)

surface area of the sphere.

The mean density of neutrons (the If k = 0.3, then (8.7.4) yields t ~ r/v.
number per unit volume) is C = n/V. Therefore, the mean residence time is
Suppose that the density near the sur- roughly equal to the time during which
face in the thin layer does not differ a neutron moving at a speed of v trav-
 298 8 Radioactive Decay and Nuclear Fission

els a distance equal to the radius r positive negative. Indeed, from

or
of the sphere. (8.8.2) it isevident that when c is neg-
An exact consideration of the escape ative the number of neutrons n falls
of neutrons requires extraordinarily off with increasing t which means that ,

laborious computations. It is very im- n tends to zero as t -> oo. But if c is posi-
portant from the start of one’s studies tive, then n increases with t that is, ,

to get used to approximate calculations n grows without limit in the course of

of all quantities of interest. Exact cal- time. Only the effect of new physical
culations are frequently very involved factors not accounted for in Eq.
and require quite a different range of (8.8.1) can halt the growth of n.
knowledge, at times even the collective Thus the value c 0 is a critical =
-efforts of many workers and the use value, for it separates the distinct
of electronic computers, and so on. But types of solution with increasing and de-
does this mean that a student engaged creasing number of neutrons. Since
in self-instruction should give up the c =
a — 6, for a given a we can speak
desire to consider a problem? There al- of the critical value b CT a, since =
ways exist simple, even though rough, c = —
a 6 >> 0 for b <C b CT and c =
methods (similar to the one just consid- a — <
b 0 for b b cr The quan- > .

ered) for an approximate approach to tity a is determined by the properties

a problem. To stop short of an approxi- of the fissionable substance: according
mate solution because the exact compu- to (8.6.2), a = Nv cr f (v —
1). The quan-
tations are complicated is merely to tity b depends on the amount of
hide one’s lack of courage. Very often, fissionable substance taken:
just such hesitancy is distructive of
3 kv
the first steps of a scientist or inventor! £
r

The concept is therefore introduced of

8.8 Critical Mass
the critical value of the radius r cr for
Up to now we have
considered separate- which b = b cr = a. From (8.6.2) and
ly two the multiplication
processes: (8.7.2) it follows that 3 kv/r CT =
of neutrons without regard for their NvOf (v — 1), whence
^escape and the escape of neutrons with-
out regard for their multiplication. 3k
Let us now consider a system in
r° r— Not (v — 1)
’

which neutrons multiply and can es-

cape. As we know, in unit time, a times The mass of the sphere of radius r cr
n neutrons are formed and b times n is called the critical mass , cr It is m .

neutrons escape from the system. Since clear that

the variation of the number of neutrons
in unit time is dnldt, it follows that m cr = jnrl p, T (8.8.3)
dn
-— = an — bn, 7

dt with p the density of the fissionable

or substance. 8 9
For r >
r cr (this is the same as
m>m CT ), c 0 and we >
have a multi-

with c — a —
plication of neutrons. For r r cr <
b For a given initial
.

number of neutrons n 0 Eq. (8.8.1) has

(m <mcr), c <Z 0 and the original
, quantity of neutrons (exponentially)
the solution diminishes. Suppose we have a sphere
n = n0 e ct .
(8.8.2)
8 9
*
As before, we consider the mass of
This solution leads to quite differ- a fissionable material, say, uranium, in the
ent results depending on whether c is form of a sphere.
 6 ~ b

8.9 Subcritical and Supercritical Mass for a Constant Source of Neutrons 299

of radius r. Its surface area S and vol-

ume V are
S = 4nr 2 ,
F = -|- nr 3 ,

whence
S 4jir 2 _ 3
T ~~
(4/3) nr 3 ~7 ’

If r is small, this ratio is great. But if

r is great, the ratio is small. No wonder

that when
the radius is small, that is,
when the ratio of surface area to vol-
ume is great, the neutron escape in- Figure 8.8.2
creases and the conditions for neutron
multiplication deteriorate. It is surpri-
sing how sharply the number of neutrons The curves corresponding to ( =
varies with b: if b 6 cr in a short> 15 X 10 -9 s and t 30 X 10~ 9 s =
time the number of neutrons becomes
,

intersect with the vertical axis ( 0) =

practically zero, irrespective of wheth-
at n =
20 n 0 and n 400ft 0 re- — ,

er b =
1.016 cr or b 2 cr If & <C = .
spectively.
& cr the number of neutrons increases
As is seen in Figures 8.8.1 and 8.8.2,
,

without limit both for b 0.99& cr = the greater the time t the more diver- ,

and for b —
0.5 b CT although the rate,
gent are the n versus t curves in Figure
differs. This is precisely why one speaks
8.8.1, the steeper are the n versus b

of the critical value of 6, the criti-

curves in Figure 8.8.2, and the more
cal value of r, or the critical value of
sharply is the criticality of the value
mass. When above critical, the mass b = 2 X 10 8 s
-1
(in this example) man-
ifested.
is said to b q supercritical, when less
than critical, it is called a subcritical If we take t >
10 6 s, the n versus b
curve cannot be distinguished from the
mass.
In Figure 8.8.1 are given the curves vertical line b b CT =
2 X 10 8 s" 1 =
n =
0 for b >> b cr and n oo for — ;

n = ~
n 0 e( a b)t for a number of values
of b. Let us construct the curves of n
b <b CT .

as a function of b for a few definite val-

ues of time t. In the computations, a 8.9 Subcritical and Supercritical Mass
8 -1
is taken equal to 2 X 10 s Figure . for a Constant Source of Neutrons
8.8.2 shows the n versus b curves for
t =5 X 10" 9 s, t 15 X 10~ 9 s, and— In the preceding section we considered
t = 30 X 10" 9 s.
the problem of the variation with time
of the number of neutrons for a given
initial number n 0 of neutrons. We now
pose a somewhat different problem.
Suppose at time zero ( t 0) the number =
of neutrons is zero and a neutron source
is switched on at this instant of time,
emitting q 0 neutrons per unit time. This
problem leads to the equation
~- = cn + ? 0, (8.9.1)

with a
c = —
b We seek the solution
.

to thisequation with the initial condi-

Figure 8.8.1 tion n = 0 at t 0. =
 .

300 8 Radioactive Decay and Nuclear Fission

A method of solution was given in

Section 8.3 for a similar problem. We
give a brief review of the reasoning
there.
We seek the number of neutrons at
time t =?£= 0. The entire time interval
from 0 to t is partitioned into subinter-
vals At. We
consider one such subin-
terval from t to t At. During this +
time the source emitted q 0 Ax neutrons.
If the source operated only during one
subinterval of time At, then we would
be dealing with the problem of the pre-
ceding section with the initial number and the initial condition n — 0 at
of neutrons n 0 =
q 0 Ax the only differ- , t = 0.
ence being that these neutrons are emit- The same formula (8.9.2) yields a
ted at time t x and not at time t = 0. = solution for a positive or negative val-
Therefore, instead of the solution ue of c, though the shape of the n
n — n 0 e ct we would have the solution versus t curve is essentially different.
n n 0 e c( ~ x
= q 0 At£
‘ t c( *~ t)
'>
=(this so- For c >
0, that is, for a 6, the expo- >
lution refers to t t; for t >
t we have < nent ct is positive, so that e ct quickly
n =
0), since clearly it is precisely exceeds unity with increasing t. For
on the time that elapsed after the ini- large t and positive c we have
tial number of neutrons was fixed that
the number of neutrons depends, that n £:< — e
ct
.

c
is, in the given case, on the quantity

t — T. Fc: c <0
we have ct 0, and so < e ct

Actually, however, the neutron source becomes much less than unity with in-
is in constant operation during the creasing t, and the values of n approach
whole time from 0 to t, and so we have the number qjc (this number is—
to add the contributions of all neutrons positive since c 0). The n versus t <
emitted by the source in the various curves are shown in Figure 8.9.1.
sub intervals of time At, the sum of Note the curious case of c 0. =
these intervals covering the entire If c = 0, formula (8.9.2) cannot be
time interval from 0 to t. Such a sum, used directly. Expand ect in a series:
given small subintervals At, is an in-
tegral, and so e
ct
= i-\- ct\ + -f ....
t

c ^~ x) d%. Substituting this into (8.9.1) yields

n (t) = j
q0 e
o

It is easy to evaluate this integral:

— C 2Jr

—
<7o
F Jr ~2 ^'

n (t) =qe 0
ct

j
e~ cX
dx = q$e ci
e~ cx
*

This formula is valid for c = 0, too.

o
We then have (c = 0)
= q0 e ct
(
e~
ct
(e
ct - 1)
n = (8.9.3)
(t) q 0 t.
(8.9.2)
This result is also readily obtainable
It readily seen that
is this solution from (8.9.1). Indeed, Eq. (8.9.1) at
satisfies the equation c = 0 has the form dnldt = q 0 whence ,

{gCt ~ 1} ] = «° = cn +^ n ( t ) = q 0 t+ A, where A is the con-

stant of integration. For t = 0 it must
- eCt
-f = ir [-t
8.10 The Critical Mass 301

be true that n — 0, whence A — 0 and it is than unity. Let us find this

less
we arrive at (8.9.3). coefficient by comparing the formula
As was shown above, when c 0 < with experiment. Experiments show
the concentration of neutrons in the that the critical mass of uranium-235
course of time attains a constant value is about 50 kg. A uranium sphere weigh-
qjc or, what is the same thing, ing 50 kilograms has a radius of about
qj\ c The smaller the value of
|
. 8.5 cm, so, in the given case,
|
c |
(the closer we are to the critical
state), the greater this constant value. k ^ — /-s- /
0 3
~~
30
Thus, for a very weak source (small
q 0 ), a mass close to critical can yield Let us examine the physical signifi-
an arbitrarily large number of neutrons, cance of the formula for the critical
a large number of fissions, and a great radius. The neutron velocities cancelled
quantity of energy. Such in principle out in the expression of r cr which ,

is the mode of operation of nuclear re- means that the formula for r cr can be
actors. obtained without regarding the course
The maintenance of such a regime is of the process in time and without exam-
no easy task, since small variations in ining the rate of neutron multiplica-
b and c drastically alter the magnitude tion and the rate of neutron escape from
of qjc when c is close to zero, and opera- the system.
tion at c close to zero is necessary if we If we disregard the dimensionless
wish to obtain a big power output for factor 3 k (it is of the order of unity),
small q0 However, this engineering
.
the formula for the critical radius be-
problem can be solved by means of au- comes
tomatic control: when n gets out of
bounds, the control system changes a r cr-W a l — _
^~T ' (8 ' 10 - 1)
or b Besides, there are also natural fac-
.

tors that facilitate control: for instance, What is the quantity on the left? The
when n increases, the temperature volume height equal to
of a cylinder of
of the active material rises and then the radius and with area of the base
it turns out, c diminishes, so that to a equal to o* f is r cr cr f Recall that if a
.

certain extent the system is self-regu- neutron is in motion along the axis of
lating. such a cylinder, then it causes fission
of those nuclei of uranium-235 whose
centers lie inside the cylinder. Since
8.10 The Critical Mass
N is the number of nuclei in unit vol-
We now know how sensitive the prop- ume, we conclude that Nr Cr cr f is the
erties of a system are depending on mean number of nuclei in the volume of
whether we have a supercritical or sub- the cylinder.
critical mass. Let us examine in more We can now
give a different state-
detail the condition of criticality ment of the criticality condition.Ear-
lier, we learned that the mean path,
3k
rcr ~ Nat (v — 1)
‘ inside a fissionable material, of a neu-
tron born inside the material (via fis-
Substituting the numbers for uranium- sion) is of the order of the radius r.
235, Of 1.6 x 10 ~
-24
cm 2 v 2.5, ,
~ After a neutron has traversed a distance
N~ 22
4 X 10 1/cm 3
we get (in centi- ,
of about r, it leaves the fissionable
meters) material and is lost to the process.
The criticality condition means that,
r °r “k 4 X 1022 X 1 .6 X 10-24 X 1 .5
— 30 *’ on the average, prior to leaving the sys-
tem, a neutron should produce one
We do not know how to determine neutron over this distance. In fission,
the coefficient fc, all we know is that v —1 new neutrons are generated. Hence,
302 8 Radioactive Decay and Nuclear Fission

it is necessary that the neutron, but merely supplement a clear-cut grasp

prior to escape, produce approximately of the qualitative physical aspect of
l/(v — 1) fissions, that is, that there the matter. In particular, the reader
be roughly (l/(v — 1) nuclei in the should pay special attention to the
volume of the cylinder of volume ra f . principle expressed at the beginning
This is the condition that leads to for- of the section: if some quantity (z;)
mula (8.10.1). enters into the derivation of a formula
Quite naturally, these arguments are but is cancelled out in the final result,
not rigorous, but they are necessary this means that there is a derivation of
for an understanding of the physical this formula that dispenses altogether
essence of the matter and cannot be re- with that quantity. And one should al-
placed by any kind of calculations, even ways find that simpler derivation be-
the most precise ones performed on mod- cause a different derivation of a formula
ern electronic computers. Computer is tantamount to a fresh view of the
mx^ci/odx tfcfixwfo dm nntt qy/uvsvssb hranig .
 t

Chapter 9 Mechanics

9.1 Force, Work, and Power expressed not as a product but as an

integral:
The relations existing between the most
b
important quantities of mechanics ad-
mit of exact formulations only by A = j> dx.
means of integrals and derivatives. In a
Chapter 2 we examined the relationship
between the distance covered by, or
We assume as known the motion of a
body given by a known function x —
the position of, a body, z, and its veloc-
x ( ). The translation dx of the body
ity, y, and also between the velocity
during a small time interval dt is equal
v and the acceleration, a, precisely,
to the product of the (instantaneous)
v = dz/dt and a — dvldt We now go .
velocity v by the time dt:
on to examine the relationships between
quantities such as force, work, ener-
dx = v dt — dt.
gy, and power. dt
Let us consider the rectilinear motion
of a body along the x axis. Suppose a Therefore, the expression for work can
force F acting on the body is also direct- he written thus:
ed along the x axis. The work A 3 3
performed by this force is defined as the A = \F^dt=\Fvdt, (9.1.1)
product of the force F by the distance a a
traversed by the body, l = b a, —
where a is the initial position of the where the moments t = a and t = fi

body and b is the terminal position: correspond to beginning and end of the
body’s motion.
A = FI =F (b - a). The product Fv in this formula is
the work performed in unit time and is
Obviously, the situation is the same called the power Indeed, in the case of .

as in the case of the relationship be- constant velocity and force, the dis-
tween velocity and distance, that is, tance is equal to x vt, the work is =
the simple formula— work is equal to A =
Fx =
Fvt and the ratio of work
the product of force by distance is — ,

to the time elapsed (that is, the work

valid only for the case where the force performed in unit time, or power) is
is constant Now if the force varies
.
Alt =
Fv. Denoting power Alt by W,
during the process of the body’s trans- we can write
lation, the whole process has to be par-
P
titioned into separate small intervals
(subintervals) so that over every subin- A =
jf
W dt. (9.1.1a)
terval the force may be taken to be a
constant (for this to be true each subin- Recall that in the SI system of units
terval must correspond to a small in- the unit of velocity is measured in m/s
crement of time or distance). Then for and the unit of acceleration is measured
a small segment A x\ of the translation in m/s 2 The unit of force has a special
.

corresponding to the ith subinterval name, the newton (denoted N), which is
(from position of the body to posi- the force that imparts an acceleration
tion x i+1 ) the work is of 1 m/s 2 to a mass of 1 kg. Clearly, the
unit of energy or work is 1 N • m —
AAi = FiAxi = Fi (x i+1 — x\). 1 kg-m 2 /s 2 it is called the joule
;

(denoted J). Finally, the unit of power

This means that in the general case of is 1 N*m/s — 1 kg*m 2 /s 3 and is known
a variable force F == F (x), the work is as the watt (denoted W).
304 9 Mechanics

the spring pulls the body leftwards, or

F
I
F2 | >
F 1 |), the body, which ini-
|

tially was in the state of rest, will

move from left to right. Figure 9.1.2a
shows the initial position of the body
Figure 9.1.1. and Figure 9.1.26 the terminal posi-
tion: (6 — a) >
0 and F x <
0. The
A body can be acted upon by sever- work A x performed upon the body by
al forces simultaneously, say, F 1 and the tension force of the spring, or more
F2 . Then we can speak of the work per- briefly, thework of the spring, in this
formed by the first force, A lf and that translation is negative, while the work
performed by the second force, A 2 ,
which you have performed is positive,
during the time that the body was trans- A2 >0. The total work A = A 2 + A x
lated from the initial position a to the is also positive, but A <C A 2 since
terminal position 6. Regarding forces A1 C0. This means that only part
F x and F 2 as constant, we obtain (A) of the work performed by you (A 2 )

A x = (b — a) Fx ,
A2 = (b — a) F2 .
was received by the body, the other
A x ) having gone into stretch-
part ( | |

Note the signs of the quantities in ing the spring. Observe that in all
these expressions. A
force is taken to be cases the force of friction against a
positive when it acts in the direction of stationary surface is directed against
increasing x (Figure 9.1.1), while a the velocity of motion of the body, and
force acting in the opposite direction (to so the work of the friction force against
the left) is regarded as negative. If the a fixed surface is always negative ,

body is translated in the direction of irrespective of the direction of the mo-

the acting force, the work of that force tion of the body.
is positive. But if the body is translated The force F x with which a spring,
in the direction opposite that of the one end of which is fixed, acts on a body
force, so that F x and b a have differ- — differs in one very important way:
ent signs, the work A of the force is it depends exclusively on the position
negative. Now picture two forces acting of the body. Not all forces, by any
on a body (Figure 9.1.2a): the force means, have this property. For example,
F 1 of a stretched spring and the force F 2 the force of friction between a moving
of the tension of a rope which you (the body and a fixed surface always retards
reader) hold in your hand. F1 acts left- the motion of the body: it is directed
wards, F x C
0, while you are pulling leftwards if the body is in motion
rightwards, F 2 >
0. If you pull with rightwards, and it is directed rightwards
more strength (which means that the if the body is in motion leftwards. Thus,

absolute value of the force with which the direction of the force of friction de-
you pull rightwards is greater than the pends on the direction of motion of the
absolute value of the force with which body. Besides, the force of friction can
depend on the magnitude of the veloci-
ty of the body. Thus, the force of fric-
tion depends on the magnitude and
direction of the body’s velocity and
not only on the position of the body
(in fact, it may even be independent of
the position of the body).
The force F 2 with which you pull
the rope in the example of Figure 9.1.2
can vary in any fashion, at your plea-
(6 ) sure. The body can, say, move to the
Figure 9.1.2 right and then to the left. In so doing
 i

9.1 Force Work, and Power

;
305

(b) the expression for the force F t

—
F t
{x ,
t, u), which in general depends
on x ,
t, and v.
Knowing
the functions x ( t ) and
v and substituting them into the
(£)
expression for F ( x t, v), we arrive at t ,

an expression for F as a function of t

time and we can describe the work as

Figure 9.1.3 an integral with respect to time. Note
that other forces may also be acting on
the body, with the result that each sep-
arate force F cannot be expressed in
itwill twice pass through the same po-
terms of the acceleration and mass of
sition: the first time in the rightward
the body.
movement at the second time on the
Suppose we have a force
Example .

return route at time 1 2 .

= — kx
F (x) acting on a body and
A
possible graph of the motion of the
let the motion of the body be given by
body (the dependence of the x coordi-
the equation x = B sin oo£, that is,
nate on the time t) is shown for this case
F x t) = kB sin ayt and v = dxldt =—
in Figure 9.1.3. We can, at our will,
(, ,

Bcocoscot; we do not require that

at time t x pull the body to the right,
the force F be equal to the mass of the
F 2 (£i) >
0, and at time t 2 let go of the
body multiplied by the acceleration:
rope so that F 2 (f 2 ) 0 or even push = it is assumed that other forces Q act
the body leftwards so that F 2 (£ 2 ) 0. < on the body, which together with F
But x (*x)
= x (t 2 ) = x x and so, speak-
ensure that the motion of the body obeys
ing generally, an arbitrary force F 2
the given law x = x (t). The work
cannot be regarded as a function of the
performed solely by force F can easily
x coordinate.
be found:
The foregoing examples of the force
of friction and the force applied by a
*2
person acting on his own free will serve
to demonstrate that the dependence A -= —B 2 k(jy
^
sin co£ cos co£ dt
of force solely on the position of a body, ti

F1 =
F x (x), which is characteristic of H
the force F x with which a spring acts = —B o—
k 2
f*

\
.
0
sin Zco£ dt
. B2 k
=— v— cos 0 _

Zcotl
1*2

on a body, is not a general property of 2 J 4 In

all forces, but is a particular property
11

associated with the elasticity of a

spring. ~ 4 (
cos 2coi 2 — cos 2 g)£ 1 ) (9.1.2)
To find the work A t
performed by a
given force F (where the subscript i
t (verify this).
shows that we are speaking of one of In this case, where the force depends
the several forces, F ly F 2 acting , • .
solely on the coordinate, it is much eas-
on the body) one must use one of the ier and convenient to take advantage
formulas of the expression of work as an integral
b 3 with respect to x:

A t
= j
F dx
t
or A t = \
F t
v dt.
b b
a oc

A = [ F (x) dx — fc x dx
We have to know two things: (a) what a
j
a
the motion of the body was, that is,
the dependence of the x coordinate of kp 2 , kb 2
~~~2 ~'
the body on time t or x = x ( t ), and ,
2
20-0946
 t 1

306 9 Mechanics

Substituting x = 5sinG)£, we can al- force will not be the same. This diffi-
so easily obtain an expression for work culty can be avoided when integrating
over a specified interval of time from with respect to time, since to every in-
^ to t 2 , stant of time t there corresponds one
definite value of the x coordinate, of
A _ kB 2
sin 2 (04 kB 2 sin 2 co 2

2 2
^g ^ 2^ the force F and of all other quantities.
,

It is easy to find the work by integrat-

It is easy to see that this expression co- ing with respect to time:
incides exactly with the preceding one
t2 *2
since
A = Fvdt= f cos at Bo* cos cot dt
cos 2 cd£ = cos 2 cd£ — sin 2 cot j
ti
j
1

= 1 — 2 sin 2
a)t , *2

whence — fBco j
cos 2 at dt.

cos 2 co^ 2 — cos 2 co^i

tl

= 1 — 2 sin 2 co ^2 — (1 — 2 sin 2 co^) Let us take advantage of the above tri-

gonometric formula cos 2cp = 2 cos 2 cp —
= 2 (sin
2
co^i — sin 2 co£ 2 )*
1 or, which is the same, cos 2 cp =
Substituting this identity into (9.1.2), 1/2 -f- (cos 2qp)/2. Then
we arrive at (9.1.3).
A
good deal of caution is required
when using the expression for work A = fB«,
as an integral with respect to the x tl

coordinate of a force F (x, v , t) de- — + -j fB (sin 2co

pending, generally, on x v, and t.
= fB(x) (t 2 *i) t2
,

Indeed, in principle, if the motion — sin 2co^). (9.1.4)

x —
x (t) is given, this equation can
be solved for t and we can determine Motion in accordance with the law
t (x). But one must bear in mind that x =B sin a)t represents oscillations of
t may not be a single-valued function of the body (see Chapter 10). As evident
x that is, one and the same position
, from (9.1.4), the work
increases without
x may correspond to two distinct in- bound with the passage of time, that
stants of time, which means that one is, as t 2 increases. This is due to reso-
and the same value of x is associated nance that occurs between the force and
with two distinct values of t (see Fig- the oscillations of the body (resonance
ure 9.1.3). Then the overall motion has will be examined in detail also in Chap-
to .be divided into separate periods du- ter 10).
ring which the velocity does not change Consider the work performed by the
sign and t is a single-valued function force during one half-period, choosing
of x. But for different periods, t is ex- for the initial time t x 0, x x 0, = =
pressed by unlike functions of x. and the terminal time t 2 ax/co, =
For example, let a body be moving sinco£ 2 =
sin n 0, x 2 =
0. Then, in =
via the law x = B sin co£, as in the pre- (9.1.4), sin 2 cd£ 2 sin 2 co^ =
0, and =
ceding example, but the force be given the work is
as a function of time, F / cos co t. =
The force is then not a single-valued A =i-/5(o (9.1.5)
function of position x. Indeed, let
t= 0, then x 0 and F — /. But if = The body has returned to its initial
we put t Jt/co, =
again x 0 but = state, while the work performed by the
F = —so that the body will be in force is not equal to zero but has a def-
the same position x 0 at different = . inite magnitude. is this result How
times (t =
0 and ^ =
Jt/co) though the to be understood from the viewpoint
 )

9.1 Force, Work, and Power 307

2C2
is expressed in terms of x by different
of the first formula A = F dx ? At formulas for increase of x from 0 to
j
XI B and for decrease of x from B to 0.
first glance, if we substitute xx = In this case, Fz (a:) = —F 1 (x). Sub-
x2 = 0, we get stituting the expressions for F 1 (x) and
o F 2 (x) into (9.1.6), we obtain
A = F dx~ 0.
j*

o A
=>W ‘-( t )
2
*
Actually, however, we have to consider 0

separately the process of buildup of x

from 0 to # max B and the process of
decline of x from x max B to 0. Dur-
=
=
-/
B
J V 1 - (x)V-
ing buildup, each value of x is asso-
In the second integral we can interchange
ciated with a definite value of the force
the limits of integration (this changes
F, which we denote by F ± :
the sign of the integral) to get
F i — f cos (tit — f Y 1 “ sin 2
b
2
A = 2f ]/" 1 — dx. (9.1.7)
\
(-i-)
o

During decline of x, the same positive

If we put z = x!B, then we have
values of x are associated with a dx =B dz and
(
negative value of force. 91 •
We denote i

this negative force by F 2 : A = 2Bf yT^dz.

j
r 2 o
F2 {x)=-iy \- i-f) .

But the integral

Thus the integral with coordinate x l

for the variable of integration breaks 7= /i=z*dz = ZL

up into two, J]
0

? H (the area of a quadrant of a circle of

A= \ F (x)dx+
it
\
F 2 (x) dx. (9.1.6) radius equal to unity), and so from
o ill b (9.1.7) we get

These two integrals cannot be combined A = 2Bf-%- = ~Bf,

by the formula

aba which coincides with formula (9.1.5)

b c c

cp dx -J- cp dx = cp dx, obtained by integrating with respect to

^ h time.
Thus, in the case of a force that is de-
since the integrands in the two inte-
pendent on time and can assume differ-
grals on the right-hand side of (9.1.6)
ent values for the same value of x,
are expressed by different formulas,
the work A is not a single-valued func-
although their meaning is the same
tion of x either. In the foregoing case
(force). This is due to the fact that F
of oscillatory motion, F = / cos cot,
is given as a function of t, while t
x = B sin (tit the quantity x again
,

9 * 1
The equation cos 2 g)£ + sin 2
G)£ = 1 is
and again passes through the same val-
true for all values of co t. From this it follows ues in the course of time, and the work
that cos (tit ±(1 =
sin 2 ^) 1 / 2 —
while the , performed by the force (for positive /)
sign depends on the value of (tit It is easy to
—
.
continues to increase all the time.
see that for ji/2<c (tit jc/ 2 one has to take < If the force is a function of velocity
the plus sign and for ji/2 (tit 3ji/ 2 the < <
minus sign, which was done above. (as is the case of friction), the situation

20 *
 ~

308 9 Mechanics

will be similar to the one discussed to the wind). Find the work done by the force
above: the body can return to its original of the wind in moving the boat b meters and
the power of the wind force. (It can be as-
position, but the work of the force will
sumed that the boat is in uniform motion, that
not be zero. In the case of friction the is,the speed v is constant). Determine the work
force is negative (see the exercises
,
and power as functions of v. Finally, find the
below). maximum
9.1.6. power for v 30 m/s, a = = 1,
S = 100 m2 and express the result in watts.
,
0

A body is moving according to the

Exercises law x — B cos (cof +
a) under several forces,
including a force that is time-dependent,
9.1.1. Find an expression in the form of an F =f cos cot Find the work performed by
.

integral for the work of friction, the force of the force during the time interval from t =
friction being proportional to the velocity of t l to t f
2
=
in particular, during one period
,

the body and in the opposite direction, F = of operation (from t 0 to t 2jt/o>). De- = =
— hv, with h positive. Demonstrate that the termine the average power of the force.
work is negative.
9.1.2. Suppose the force of friction is con- 9.2 Energy
stant in magnitude and opposite in direction
to the velocity, that is, F for v — —h
0 > We consider the case of a force that
and F =
-f h for u 0. Suppose <
the body depends solely on the position (coordi-
moves in accordance with the law x B sin cot. — nate) of the body, F = F (x). As we
Find the work of the force of friction. during_ 0 ^ 'iWv ~ £
= — y ^ OfO (Jtu , ’em’ ~fji!lS CcCllCtJ U1
the time interval from t 0 to t ji/co.
9.1.3. The force acting on a body is given
this kind
force is the force withof
by the formula F —
f0 sin oo 0 £, with / 0 con-
which a spring acts on a body, the
stant. Since the body is also acted upon by other end of the spring being fixed. 9 2 In
other forces, it moves according to the law X2
x — B sin (Dj*. Determine the work performed
by force F during the time interval from
that case the expression A = F dx
j
t = 0 to t =
T. Consider the case co 0 g^. = X\
9.1.4. A body is falling according to the may be applied without any complica-
law x =2
gt / 2 (the x axis is directed down- tions (compare with Section 9.1).
wards). Find the formula for the work result-
ing from the air-resistance force F aS =— In particular, in this case if the body
pv 2 /2, with a being a constant of proportion- first moves in one direction from x
x
ality dependent on the shape of the body to %ax =X
and then in the opposite
(see Sections 9.14 and 9.15), S the cross- direction returning to the initial posi-
sectional area of the body (in cm 2 ), p the
tion, we have x 2 = x l9 and the total
air density (about 1.3 X 10~ g/cm3 ), and v
3

the rate of fall (in cm/s). Also find the formula work done by the force is actually
for the work done by the force of gravity F = equal to zero:
mg, where m
is the mass of the body.
X2—XI
Perform the computations and compare
the results for a wooden ball of diameter 1 cm, A= F(x)'dx = 0.
a = 0.8, and for a steel bullet of length 3 cm, j
diameter 0.7 cm, a 0.2, for t =1 s, 10 s, = XI

100 s. Dividing the path length into sections

Remark The idea behind the calculation
.
only corroborates this conclusion:
is thatwe assume the force of air resistance x X2
to be small compared with the force of grav-
ity and not noticeably affecting the law of A= F dx + F dx
free fall. Computing the work done by the air J j

resistance and comparing it with the work per-

x1 X
X X
formed by gravity, we verify the correctness
of our starting assumption concerning the — F dx— F dx,
small role of the force of air resistance. In ^ j
Xi X*
Section 14 we give an exact solution of the
problem of free fall with air resistance. and A = 0 at xx = x2 .

9.1.5. A wind blowing with a speed vQ

2
acts on the sail of a boat with a force F equal
9
the’second end of the spring is allowed
-
If
to aSp (v0 — v)
2
/2 for v < vQ and — aSp (u0 .
— to move
at random, the force acting on the
for v >> v0 is the speed of the boat, S
v) 2 /2 ,
body will depend not only on the position of
the area of the sail, p the air density, and a the body but also on the position of the second
a dimensionless coefficient (a is approximate- end of the spring and this does not satisfy
ly unity for the sail put up perpendicular the stated condition.
 x x ]

9.2 Energy 309

In mechanics, potential energy is tion u (x) is determined from the for-

defined as the capacity to do work. mula (9.2.1) ifin it we put x 1 = x 0
A spring possesses a definite reserve of and x2 =
x:
potential energy depending on how com-
X
pressed or stretched it is. If one end is
u (x) — u — F x ) dx.
(, (9.2.2)
fixed in position, the potential energy of j
0

the spring depends on the position of X0

the body to which the free end of the

That is how the problem of determining
spring is attached. Thus, the potential
the potential energy from a given force
energy u =
u (x) is a function of the
is solved.
x coordinate. If in the initial position
= Wecan pose the converse problem,
x x x the potential energy is u (x-f),
namely, knowing the potential energy as
after the body has been displaced to
a function of x, or u = u x ), find the
position x =
x 2 when the spring has ,
force F x ). To solve this problem, take
(.

performed work A equal to

(.

the derivative of both sides of (9.2.2).

oo
The derivative of the integral (with
X2
respect to the upper limit) is equal to
A — F (x) dx
j*
, the integrand, so that
XI

the remaining potential energy is equal

^1=— F(«). (9.2.3)

to u —A . Thus,
X2 The minus sign here is essential. The
u (x 2) — u (xi) — A = u (.x — F (x) dx .
force is positive (in the direction of in-
j creasing x) if duldx is negative, that is,
XI
as x increases, the potential energy u
(9.2.1)
decreases. The force is negative (in the
Get a good feeling
of the sign affixed direction of decreasing x) if duldx is
to A
in this expression: if the spring positive, that is, as x increases, the
does (positive) work, the reserve capac- energy u increases, too. In the latter
ity of the spring to do work will dimi- case, obviously, as # decreases, the ener-
nish! The work performed by the spring gy u also decreases. This means that
is taken from the reserve of potential the force is always in the direction of
energy. For this reason the work done diminishing potential energy.
(that given up by the spring) is equal Let us examine in more detail the
to the difference between the initial example of the spring. Let the body be
and final energy of the spring: at the origin when the spring is not un-
der tension (Figure 9.2.1). When the
A = u (x-f) — u (r 2 ). body is pulled to the right, the force of
tension grows in direct proportion to
All formulas involve the difference
the displacement of the body and is
of potential energy in two positions of a
directed leftwards:
body. Therefore, if we replace u (x)
by u ) (. +
C, where C is an arbitrary
F = — kx, k> 0. (9.2.4)
constant, this will in no way affect the
physical results. Indeed,
Assume that u 0 = 0, at x = 0, that is,
[u (xff) +[C\ — [
u (.
2) + C regard the potential energy for the
= U (x x ) — u (x 2 ).

The value of u (x) at some given

point, call it^ 0 can be chosen quite ar-
onnr*
1 O 'O 0
-^
s

X
,

bitrarily. Denote it by u 0 Then at some .

i 77-/7
*

other point x the value of the func- Figure 9.2.1

 310 9 Mechanics

nontense spring as zero. This yields force of attraction acting on a body is

X X directed towards the earth’s center and
2
is Clr in magnitude, where the con-
u (x) =— F dx — k x dx —k .
^ ^ stant C is positive.
The constant C can readily he deter-
mined from the condition that the force
It is easy to see that this u (x) is asso- acting at the surface of the earth
ciated, via formula (9.2.3), with the (r r0 = 6400 km ~6.4 X 10 6 m) is =
force (9.2.4). known: F
Clr 0 that is,
(r 0 ) = mg ~ ,

Weconsider a second example, the C =

mgrl, where g is the acceleration
force of gravity. Send the z axis upward. of gravity at the earth’s surface equal

The force of gravity acts downward and approximately to 9.81 m/s 2 ). We finally
is equal to — mg
where g is the accele-
,
have
ration of gravity. It is independent of
(9.2.7)
the height z but a constant quantity is
,

merely a special case of a function whose

For zero we again take the potential
values are the same. The important
energy of the body at the earth’s sur-
thing is that the force of gravity does
face. Allowing for the fact that as the
not depend on time or velocity. We
distance r from the earth’s center in-
can therefore make use of the formulas
creases the work performed by force F
derived above. We
take as zero the po-
in the process is negative, we get
tential energy of the body at the earth’s
surface, where z 0. Then = r r

z
u= — j F dr = mgrl j
7-
u (z) = — F dz = ( — mg)'dz ro ro

j j
~7
= mgz.
o

(9.2.6)
-mgrl ( p - mgr\ (
-7 + 7-)
=- m g 7- (f — r 0) • (9.2.8)
The potential energy grows linearly
with increasing height of the body
At a small height z = r — r0 <C r 0 ,

above the earth’s surface. but slightly from

the ratio rfr differs
In the latter example we assumed
unity, whence
that the distance z is small compared
with the radius of the earth. Let us now u (
r) ~ mg (r —r = 0)
mgz ,
(9.2.9)
examine the attractive force on the as-
sumption that the distances can be which coincides with formula (9.2.6)

arbitrarily large. By Newton’s law of obtained earlier. But, as can be seen

gravitation, the attractive force is in- from (9.2.8), the potential energy does
versely proportional to the square of not increase without limit as r increases,
the distance between the bodies. We as would have been the case in ac-
know that for a body above the earth’s cordance with the approximate formu-
surface the force of gravitation towards la (9.2.6), but tends to a definite limit

the entire globe is equal to the force of

u (oo) = mgr 0 . (9.2.10)
attraction to a mass equal to the earth’s
mass and concentrated at the earth’s Thus, making allowance for the de-
center. 9 3 It is therefore convenient to crease of gravity with distance, we can
reckon distances from the center of the say that the energy of a body at an
earth and denote them by r. Then the infinite distance from the earth is the
same as, by the approximate formula,
9 3-
This does not hold true for a body at a distance of r 0 from the earth’s sur-
inside the earth, in which case we must allow
only for the portion of the earth’s mass be- face, or at a distance of 2r 0 from the
tween the earth’s center and the body. earth’s center.
9.2 Energy 311

In this problem we encountered a problem can either be replaced by infi-

physical situation involving infinite nity or not, depending on the aspect
distance. In this respect we must note considered. The possibility of such a
that in any physical problem we are substitution depends not only on the
always interested in finite quantities, quantity R itself (and its comparison
or in our case finite distances. For in- with other quantities of the same di-
stance, if we consider the motion of a mensions entering into the formulas, r 0
body and the energy of the body as de- in the given case) but also on the struc-
pendent on the earth’s gravity, then ture of the formula in which it occurs.
we can be interested in attaining the Returning to the question of poten-
moon, Mars, or other planets, or even tial energy of a body attracted to the
stars. All these objects involve distances earth, let us find the numerical value
that are very very great relative to of u (oo) per unit mass— it is equal to
that of the earth’s radius, but they are gr 0 ~ 9.81 X 6.4 X 10 6 6.28 X ~
finite! And, obviously, for a physicist 10 4 J/g, which is 30 times the heat
the expression r =
oo means only that of evaporation of water and 10 times
r> r 0. the chemical energy of explosives.
Suppose we consider the problem of In problems of celestial mechanics
launching a rocket to a great height, and in physics it is advisable to choose
to a considerable distance from the for zero the potential energy of a body
earth. We are interested in the energy located at an infinite distance from the
required and the time of flight. Here are mass attracting it. Then for the poten-
two cases. tial energy of a body at a distance r
(a) A space vehicle is to traverse a we have
distance of R = 10r 0 ,
where r0 is the r

u (r) = u (oo) — F (r) dr = —

earth’s radius.
(b) A space
,
vehicle is to traverse j
R = 100r 0 .
OO

The work needed to tear away from

(9.2.12)
the earth and go a distance R from the where C the constant in the expression
is
earth’s center is for the F
force, C/r 2 = —
and can ,

be determined from the formula C =

A = m gr\(-^ (9
2 11 )
-
mgrl if we know the acceleration of
r --I*)- -

gravity g at the earth’s surface and the

Recalling that r 0 ~
6.4 X 10 6 m, we radius of the earth, r 0 .

get A x mg X 5.76 X 10 6 in the first We can obtain a different expression

case and A 2 ~
mg X 6.34 x 10 6 in the for C. By Newton’s law of gravitation,
second. F = — GmMlr 2 where , m
is the mass of
A ten-fold change in distance caused a body attracted to the earth, is the M
a relatively, small changp in the. eaeryy
required. If we were to replace R by center of the earth, and G the gravita-
infinity, we would get A «, ~ mg X tional constant equal to approximately
6.4 X 10 6 (note that oo is not a num- 6.7 X 10- 11 N 2/kg 2 6.7 X
• m =
ber). A1 from A oo by 10%, and
differs m
10 -11 3 /kg-s 2 Therefore
. C GmM. =
A 2 differs by 1%. That is why, when Using this formula, we can easily de-
computing work, R may be replaced by termine C if we know G and M.
infinity. But a change in R has a strong Indeed, by measuring the attraction
effect on the time of flight, and for this of two heavy balls of known masses we
reason when considering the time of can find G Only after this, by measur-
.

flight we must never replace R by infini- ing the attraction of a body to the
ty. earth can we find the earth’s mass.
To summarise, then, one and the The form of the function expressing
same quantity R in one and the same Newton’s law of gravitation, that is,
 312 9 Mechanics

the fact that the force is inversely ty to do work but also as the work re-
proportional to the square of the dis- quired to bring a system to a given state.
tance, is proved hy comparing the A stretched spring can do a definite
attraction to the earth of a body at amount of work in returning to the un-
the earth’s surface with the attraction stretched state. That, clearly, was the
of a remote body, the moon, as well work that had to be done to stretch the
as by comparing the attraction to the spring in the first place. Similar as-
sun of planets that orbit the sun at sertions may be made in the case of a
different distances —
from Mercury to body raised to a definite height above
Pluto. the earth or for a system of two
The problem of potential energy of charges.
two electric charges e 1 and e 2 is quite sim-
ilar to the preceding one. The inter-
action force between the charges is 9.3 Equilibrium and Stability
equal to
We consider a body that can move with-
F=k- (9.2.13) out friction along a straight line, which
we take for the x axis. Let the
Here, the charges are expressed in elec-
if body be acted upon by a force directed
trostatic units (the unit of charge is along this axis and dependent on the x
(1/3) X 10~ C (coulomb)) and the force
9 coordinate. We can again picture the
in dynes, (1 dyne =
10~ 5 N) k is equal spring. Below we will examine other
to unity; in SI, where charge is mea- examples as well.
sured in coulombs and force in newtons, The equilibrium position of a body is
k = (1/9) X 10" 13 . There is no minus defined as that position for which the
sign in (9.2.13) that we see in the expres- force is zero and the body is at rest.
sion for the gravitational force. In- Denote the point of equilibrium by x 0 .

deed, if e x and e 2 are like charges (both Then / (x 0 ) =0. Expanding the func-
positive or both negative), the product tion F =F (x) in a Taylor series and
ex e 2 is positive. But like charges re- ignoring all powers of x —
x 0 except
pulse one another, that is, the force F the first, we see that two versions of
is positive. the function F (x) are possible in the
Again defining u (r) so that u (oo) = neighbourhood of point x 0 (provided
0, we obtain =
F (a: 0 ) 0), namely, F x )
(, k x (x —
x0) and ~—
F (x) k 2 (x — x 0 ); in
both formulas it is assumed that k x
u(r) = k-^~. (9.2.14) and k 2 are positive quantities. The first
case is shown in Figure 9.3.1a, the sec-
The potential energy of two like ond in Figure 9.3.16.
charges separated by a finite distance is These two cases are associated with
positive: they repulse and, moving from an entirely different character of equi-
r to oo can perform work equal to librium. In the case of Figure 9.3.1a,
if the body is somewhat to the right of
u (r) — u (oo) = u (r).

The potential energy of two unlike

charges is negative. Indeed, e x e 2 0 <
if, say, e x >
0 and e 2 0. This is clear <
physically: since unlike charges at-
tract each other, energy must be ex-
pended to pull them apart to infinity.
Note that thanks to the law of conser-
vation of energy, the potential energy
may be defined not only as the capaci-
 9.3 Equilibrium and Stability 313

point x = x0 ,
then it is acted upon by is converted into kinetic energy. But
a positive force, that is, a force which if the body is in the state of minimum
pulls it farther rightwards. Thus, the energy, energy from an outside source
equilibrium at point x x 0 in Figure = is required to move it to any other po-
9.3.1a is unstable. slight deviation A sition. This energy will go to increase
of the body (whether to the right or left the potential energy. A small expendi-
makes no difference) suffices for a force ture of energy will displace the body
to begin to act on the body, and this only a small distance. These properties
force will increase the deviation. On of a body in a position of minimum po-
the other hand, in the case of Figure tential energy fully accord with the
9.3.16 the force is negative (pulls concept of stable equilibrium.
leftwards) when the body deviates to Take the case of gravity near the
the right. Deviation of the body from earth’s surface. The potential energy is
the equilibrium position gives rise to a mgz where z is the height above the
,

force that tends to return the body to surface. The curves depicting the func-
the position of equilibrium. Here we tion u (x) can be visualized as curves
have to do with stable equilibrium. It indicating the altitude z of a body as
iseasy to see that for a body attached a function of the horizontal x coordi-
to a spring the second case is realized. nate. We
have to imagine a body in mo-
In accordance with the above expres- tion along a curve like a bead on a stiff
sions for force, we find the expressions wire. The u versus x curve corresponds
of potential energy via (9.2.2). In the to the shape of the wire if the plane of
case of unstable equilibrium, the drawing is vertical. Then it is clear
that the maximum of u (x) corresponds
u(x) ~u (x — ~ k 0) t
(x —x 0)
2
.
to (see Figure 9.3.2a) a point on the wire
from which the bead slides downward
In the case of stable equilibrium, at the slightest touch, moving to the
right or to the left away from the maxi-
u(x) ~u (x + Y k 0) 2 (x —x 0)
2
.
mum, and the minimum of u x ) (Fig-
(.

ure 9.3.26) corresponds to the lowest

The appropriate curves are shown in
point, at which the bead is in a stable
Figure 9.3.2a and 6.
position, and any other beads on the
Thus, in the case of unstable equi-
wire would strive to take up that posi-
librium, the potential energy has a
tion.
maximum while in the case of stable
,
Thus, the graph of u (x) gives a pic-
equilibrium it has a minimum. In
torial visualization of the direction of
both cases the force is zero at the point
forces and character of equilibrium.
of maximum or minimum, that is
Let us examine a few examples.
F = — {du!dx) x=X(i = 0. Example 1 Let a charged body be in
.

motion along a straight line (which we

is quite natural. If a body
This result
take for the x axis) on which two iden-
is the state of maximum potential
in
tical charges e x are fixed symmetrically
energy, energy is released during dis-
about the origin at a separation of 2a
placements in both directions. This ener-
(Figure 9.3.3).
gy can be used to overcome inertia and
It is quite clear that at the origin
the body is in a state of equilibrium.

Figure 9.3.2 Figure 9.3.3

 x .

314 9 Mechanics

Figure 9.3.5

that in the case e x e 0, the equilib- >

rium is stable. Here, if the body is dis-
placed rightward, the (greater) repul-
sion on the right will tend to move it
into the initial position.
Figure 9.3.4
Let us find (d2 uldx2 ) x== Q. Using (9.3.1),
we get
Indeed, the forces acting on the body
from the fixed charges are equal in mag-
~d*r
= 2e,e + (a — x)» ] 9 3 2)
' '

nitude and opposite in direction so [ (a+*) 3 ’ (

that they balance, which means their =

Putting x 0 in (9.3.2), we find that
resultant is zero.
The potential energy u (x) of a body d2 u I
_
with a charge e is dx 2 jx=0 ~ a3 ’

Hence,
=
ifthen d?u!dx 2 0
e x e is positive, >
at x 0, function u (x) has a
the
minimum at x 0, and the equilibri- =
where the distance to the left-hand
r' is um is stable. But if e x e is negative,
•charge, and r" the distance to the right- d^uldx 2 at x <0
0, the function =
hand charge, that is, r' =
x +
a and u (,has a maximum at x 0, and the =
r" =a —
x. Whence
)

equilibrium is unstable.
Example 2. We consider a situation
«(*) . (9.3.1) in which the charges are spaced in the
same way from the origin, but along
The appropriate curves are shown in a straight line perpendicular to the line
Figure 9.3.4. The upper curve corre- (the x axis) along which the charged
sponds to e x e > 0, which is the case of body is in motion (Figure 9.3.5). In this
like charges of body and fixed charges, the potential energy is
while the lower curve corresponds to e^e
exe <0, which means that the body u (a:) — 2
has a charge opposite to the fixed
Ya + x 2 2

charges. The graph of the function u (x) at a = 1

In the case e x e < 0, equilibrium at and |
exe |
=
in Fig- 1 is shown
the origin is unstable. Indeed, the body ure 9.3.6. Here equilibrium at the ori-
is attracted both by the left and the gin is unstable for ex e 0. If the charge >
right charge and at the origin the e of the body is opposite to the fixed
forces of attraction balance. But if the charges e 1 (e x e < 0), the equilibrium is
body is displaced the slightest bit in stable
any direction, say to the right, the at- This is easy to establish if we exam-
traction on the right will exert a stron- ine the force acting on a moving charge
ger effect and will continue to pull it (Figure 9.3.7). Let e x e be positive. Dis-
rightward. Similar reasoning shows place the body rightward from the
 $.3 Equilibrium and Stability 315

a In proved that
electrostatics it is
no point
this result is general: there is
of equilibrium in the space between
external fixed charges such that equi-
librium is stable relative to displace-
ments in any direction.
The general proof of this fact given
below may appear too hard for the
reader and he can skip it without any
loss of continuity in the book.
For proof in the general form, note
that the potential energy of a charge
e at point (x, y, z) as a function of the
position of equilibrium. The resultant
charge’s distance r from a fixed charge
force of repulsion is also directed to the
e x located at point (x x y x z x ) is
right, further increasing the deviation. , ,

The equilibrium unstable. If e x e

is
is negative, the resultant force is in
the direction of decreasing deviation eie

and the equilibrium is stable.

These results are also readily arrived
Consider the motion along the x axis
at by considering d u!dx at x = 0 (do
2 2
and find d 2 u!dx 2 for y and z constant
this yourself).
(this quantity is known as the second
Note that for e x e >
0, when stabili-
partial derivative of u with respect to
ty occurred in Example 1 (Figure 9.3.3),
x and is denoted by d 2 uldx 2 ). Then in
in Example 2 (Figure 9.3.5) we had
a similar manner we find d 2 uldy 2 and
unstable equilibrium. For e x e <0 (un-
d 2 u/dz 2 which refer to motion along
like charges) the situation was reversed: ,

the y and x axes, respectively. It turns

the equilibrium is unstable for the arra-
out 9 4 that for arbitrary x, y, z, xx
*

ngement of charges as given in Figure ,

9.3.5 and is stable for their arrangement yn and z 1 the sum of the second deriva-
tives along the three perpendicular axes is
shown in Figure 9.3.5.
zero' 9 *
Turning Figure 9.3.5 through 90°, we
note that actually it refers to the same d2u b 2u ,
d2u
(9.3.3)
initial distribution of charges in the dx 2 dy 2 '
dz 2

equilibrium position as in Figure Obviously, this property will be

9.3.3. We can say that Figures 9.3.3 preserved for the sum of any number of
aha rdier 'to "tne same’ rrittidi "dis- terms ot tlie Torm e k e/r k where e k ,

tribution of charges, but the directions is the fixed charge at point (x k yk , ,

of motion under consideration dif- z k ), and r k is the distance of charge e

fer. Then the equilibrium will al- from this point. Consequently, formu-
ways (for any signs of charges) be un- la (9.3.3) is valid for any distribution
stable in one direction or in the other. of fixed charges in space.

9 4
*
The reader should convince himself
of this. Note that the point (x ±J y ± z t ) at ,

which the fixed charge is positioned is special

in that the function and its derivatives become
infinite; we will ignore this point in our dis-
cussion. Various ways in which such points
can be considered are discussed in Chapter 16.
9 5
Functions that satisfy this condition
-

are known as harmonic functions they are ;

widely used in many fields of physics (see

Figure 9.3.7 Chapters 15 and 17).
 t )

316 9 Mechanic sr

In particular, this formula is valid is the derivative of the coordinate of

at the point at which charge e is in the body with respect to time. Thus;
equilibrium. But for equilibrium it is
necessary that the sum of the projec- ma = m ~= F, (9.4.1)
tions of the forces on each of the axes at
or
this point be equal to zero. For this we
must have (9.4.2)
du du du
~dx~
~U
p,
’
~dy
U ’
~dz
U
p|
’
We
begin with the case where the
force given as a function of time,
is
since if the projections of a force on three
perpendicular axes are zero, so is the
F —
F (t). This means that the deriva-
tive d 2 x!dt 2 is given as a function of
force (a vector quantity), that is, the
projection of the force on any direc-
time. Using Newton’s law (9.4.1), we
tion is zero. 9 6 can easily find the velocity at any
given instant of time. Besides the ap-
For equilibrium to be stable rela-
plied force we also have to specify the
tive to motion along all three perpen-
velocity at some time t 0 Then
dicular axes, it must be true that
.

t
d*u d 2u
dx 2
>o, dy 2
>0 ,
dZ 2
>0 .
v(t)=v(t0) + -^-
[
F(t)dt. (9.4.3)

(9.3.4) to

But this contradicts (9.3.3) since the Knowing the velocity as a function of
sum of three positive quantities cannot time, v =
v (t), and the initial posi-
be zero. tion x (t 0 ) of a body, we can find the po-
sition of the body at any given moment
in time:
Exercises
t

9.3.1. A charge e moves along a straight

x — x (t + v (t) dt (9.4.4)*
( t) 0)
line on which are fixed two positive charges e 1 J
,

and e 2 = Ae ± at a separation of 2a. Find the *0

point on the straight line at which equilibrium
where u (
t is given by (9.4.3).
of the charge e is possible and determine the
type of equilibrium. Consider two cases, e 3> 0 relationship between velocity
The
and e 0. < and distance is considered in detail
9.3.2. Solve exercise 9.3.1 when the sign together with examples in Chapter 2.
of e 2 is opposite to that of e x .
On the whole, formulas (9.4.3) and
(9.4.4) solve the problem of finding
x (
t
)
from Eq. (9.4.2), which is a
second-order ordinary differential equa-
9.4 Newton’s Second Law
tion involving the second derivative of
,

Newton's second law states that the the unknown function x ( t ). The answer
product of the mass by the acceleration is includes not only the given function
equal to the force applied 9 7 Accelera- ,
-
F ( t ) but also two constants defined from
tion a is the derivative of velocity v the initial conditions, the position and
with respect to time; in turn, velocity the velocity of the body at a given time
t0 .

9 -
we have a nonzero force F (a
6
If vector) If the law of motion of the body is-
acting in some direction, then there will be
given or has been experimentally estab-
a force acting along each axis equal to the
projection of the force F on the axis. lished, that is, we know the function
9 - 7
Newton’s first law the law of inertia
, ,
x ( t ), it is easy to find the force applied
states that any body on which no forces act to the body: to do this we must find
moves translationally and uniformly. This the second derivative of the function
means that the acceleration is equal to zero
for a force equal to zero. Thus, the first law is x ( ) and multiply it by m (formula
contained in the second as a particular case. (9.4.2)).
9.5 Impulse 317

Exercises
9.4.1. Find the law of motion of a body
acted upon by a constant force F if at time
t= 0 the body is at rest at the origin x 0. =
9.4.2. The same provided that (a) x 0 —
and v =
v0 at t 0, and (b) x =
x 0 and v = =
v0 at t = 0.
Figure 9.5.1
9.4.3. A body
of mass 20 kg begins to
move under a force of 1 N from the origin
without an initial velocity. What distance
will it cover in ten seconds?
is known as the impulse of the force
9.4.4. A
ball falls from a height of 100 me- during the time interval between t 0
ters (initial velocity zero). How long will it and t. Formula (9.4.3) may be written
take the ball to reach the ground? (Disregard as follows:
air resistance.)
9.4.5. Under the conditions of the preced- t

ing problem, the ball starts falling at a veloc- P(t)-P(t 0 = F dt = I (t0 t)).
ity v0 =
10 m/s. Examine two cases: (a) the
)
j
{ ,

initial velocity of the ball is directed downward to

and (b) the initial velocity is directed upward. (9.5.2)

Determine the time required to reach the
ground and the velocity the ball will have In other words, impulse equals change
at the time of impact. Verify that in cases
in momentum.
{a ) and (b) the velocity of impact is the same.
9.4.6. A
body is acted upon by a force pro- There are forces that act during very
portional to the time that elapses from the brief time intervals (an instance is the
beginning of motion (the constant of propor- blow of a hammer and the rebound af-
tionality is equal to k). Find the law of mo-
ter striking a body). Both prior and fol-
tion of the body if it is known that the body
begins moving from point x 0 at an initial = lowing the blow, the force is equal to
velocity v0 . zero. It is clear that in the absence of
9.4.7. A body is acted upon by a force pe- other forces (other than the brief blow)
riodically varying in time, F f cos o)£, = the body prior to the blow moves with
where / and co are constants.
(a) Find the law of motion of the body pro- a constant velocity and after the blow
vided that x 0 and v = 0 at t 0. Establish = — with another, also constant, velocity.
that this is oscillatory motion. Determine the Let F ( t ) differ from zero only dur-
period of oscillation, the maximum value of
ing the interval between t x and t 2
x ( t ), and the greatest value of the velocity.
(b) The same for a force F f sin cot — (Figure 9.5.1). We consider the in-
and x =
0 and u 0 at t 0. = = tegral
9.4.8. A body is in motion under a con-
stant force F. At time t t0 the body is at = t 2

oint x —
x 0 Find the velocity the body must
. / = F(f) dt. (9.5.3)
ave at t =
t 0 so that at t t± the body will = j
reach point x = xx .
1 1

It may be called the total impulse in

9.5 Impulse the sense that the integral is taken over
The problem of finding the law of mo- the entire interval of time during which
tion of a body for a given dependence of the force acts.
the force on the time was, in principle, The expression (9.5.1) involves an
solved in the preceding section. Here integral from t0 to t. If t 0 tx < and
we will examine the properties of the t > t2, then
solution and certain new concepts as- t

sociated with the solution. I (t 0 , t) — F dt = /.

The product P mv of mass by veloc- = j
to
ity is called the quantity of motion or ,

momentum while the quantity ,

Indeed, we write
t t n t 2 t

I (t 0 , t) = F (
t
)
dt (9.5.1) F dt--= F dt ^
F dt-{- F dt.
J j | ^
t0 to tQ 11 *2
 t )

318 9 Mechanics

Figure 9.5.2

Figure 9.5.3

The first and third integrals on the

right-hand side are equal to zero, since The shorter the time of action of a
F =
0 over the respective intervals, force, the shorter the time interval dur-
and the second (middle) integral is I. ing which the velocity of a body
Thus, from (9.5.2) and (9.5.3) we get changes from the initial value v 0 = 0
P(t) =
P(t 0 ) I if t0 + tx and < to the final value v = Hm
(Figure
t > t2 .
9.5.2 6). Thus, in the limit of an extre-
mely great force acting over an extre-
From formula (9.4.3) we see that the
velocity following the blow depends so- mely short time interval, the graph of
lely on the impulse of the force, that is, velocity takes the shape of a step (com-
on the integral of the force, and not on pare Figure 9.5.3a with curve 2 in Fig-
the particular type of the function ure 9.5.2 6). It is not essential here
F = F ( t ). For example, several diffe- which of the curves in Figure 9.5.2 a
rent curves of F (£) shown in Figure we compressed— the step is characte-
9.5.2a all yield the same impulse, rizedby only one quantity, v = I/m y

which is to say, they all change the ve- and this quantity is the same for all
locity of a body by the same amount. the curves.
It is not difficult to draw the appropri- If prior to the application of the

ate graph of velocity, v =

v ( ), for each force the body was at rest at point x 0J
of these curves representing F ( t ). Fig- after a brief application of a great force
ure 9.5.2 6 depicts these graphs un- the body begins to move with a con-
der the general assumption that the stant velocity equal to I/m. If the force
initial velocity is equal to zero. The acted at time t ± (we consider the in-
common element of all the curves in terval between t x and t 2 to be small and
Figure 9.5.2 6 is the finite value of ve- therefore do not distinguish between t 2
locity: all the curves go into a horizon- and £j) the position of the body as a
tal straight line on the right at a height function of time is given by the formu-
v = Hm. las

Each of the curves representing F (

t
x—x if t<Ct i,
0
in Figure 9.5.2a may be compressed (9.5.4)
along the axis of time and proportionate- x— -xQ —— —
-\
/
(t tj)
.

if
ly stretched along the axis of force.
The area under the curve of F, that is, The appropriate graph is shown in Fig-
F dt , the total impulse, does not ure 9.5.36. Note that x = x (t) satis-
I*
fies the equation
change in the process. That is precise-
ly how, say, curve 2 in Figure 9.5.2 a
was obtained from curve 1 .
 ) v .

9.5 Impulse 319

We recall that on the graph of x (t) The third term on the right-hand sidn
the first derivative dxldt is connected can be transformed by the formal rules
with the slope of the tangent to for handling iterated (sometimes called
the curve, while the second derivative multiple) integrals. However, since we
d 2x!dt 2 describes the rate of change of have not mentioned
such rules any-
the first derivative, that is, the second where, we get the transformed expression
derivative is associated with the cur- (in the form of a single integral) by
vature of the curve x x (t) (see Sec- = using the law (9.5.5) of the motion of a
tion 7.10). body under the action of a single im-
In Figure 9.5.36 the curve x x (t) = pulse of force.
has a salient point at t t ly x x0 = = . The action of force F(t) during thn
The salient point can be regarded as a time interval At from t to t At +
point at which the curvature is infinite, can be approximately replaced by tho
so that the existence of such a point (instant) action of the impulse A I =
corresponds to consideration of a very F (t) At. We already know the mo-
large force (infinite in the limit). How- tion of a body under the action of such
ever, both before and after the sali- an impulse— see formula (9.5.5) in
ent point the derivative dxldt is finite. which / is to be replaced by A I (t).
This means that a very large force act- It then only remains to combine thn
ed over a very short time interval, so contributions of all intervals A t to t

that the impulse is finite. 9 8 The impulse coordinate x ( t ) to get

can readily be found from the graph
(see Figure 9.5.36) by computing the x (0 —2x i (*> x)
velocity after the application of the
force and employing formula (9.5.2). = Z^-( —'t F t ) (
x) Ax
The law that we have found concern-
t
ing the motion of a body that up to
time t =
t was at rest and at that mo-
j
Fix) (t t) t dx (9.5.7)
ment received an impulse I will help to

us to refine the formulas (9.4.3) and

Here, as usual, we replaced the sum of a
(9.4.4). For this we need a special case
large number of terms corresponding
of the formula (9.5.4) where the body
to small intervals At by the integral
is at the origin x 0 0 at t t. We = = (the right-and left-hand sides of this
introduce the notation
formula can now be connected by a
rO if tC t, strict equality, ignoring the interme-
diate terms). Formula (9.5.7) does not
« <><• take into account the initial coordinate
(9.5.5) x (t 0 ) and the motion with the initi-
If we
substitute v ( t from (9.4.3)
al velocity, (t —
t 0 ) v (£ 0 ), since in
(9.5.7) we assumed that such motion
into (9.4.4) and use more accurate des-
is absent ( ( t 0 ) = 0 and x (t 0 ) = 0).
ignations (so that the upper limit and
A more general expression can be ob-
the variable of integration have differ-
tained if we add these terms to the right-
ent letters), we get an expression that
hand side of (9.5.7):
at first glance is rather unwieldy:

x(t)=x(t 0 ) + (t — t0 )v(t0 ) x{t)^=x(t 0) + (t — t0 )v(t 0 )

+~
t
t n
+ -i- dt l F (t 2 ) dt 2 . (9.5.6) j
F (t) (t — t) dx. (9.5.8)
j j U
to to

9 8
The advantage of formula (9.5.8)
. For greater details involving the vari-
ous mathematical constructions associated over (9.5.6) that in (9.5.8) we have
is
with this example see Chapter 16. to integrate only once. We did not state
 )

320 9 Mechanics

why we can merely add the terms This formula shows that the action
corresponding to individual impulses of one body on the other does not change
(involving the initial velocity and the the sum of the momenta of the bodies.
initial coordinate). This is examined
in more detail in Section 17.4. Here
9.6 Kinetic Energy
it suffices for us that we can directly
verify the values of x (t 0 ), (dx/dt) t==0 , Let us consider a body moving under
and d 2 x!dt 2 using formula (9.5.8). To the action of a force F (t) and known
this end we must differentiate x (t) between the work
find the relationship
in (9.5.8) in a manner similar to what done by the force and the velocity of
was done in the verification of formula the body.
<8.3.5). Multiplying both sides of the basic
will recall that by New-
The reader equation m (du/dt) — F(t) (Newton’s
ton’s third law, in the interaction of second law) by velocity y, we obtain
two bodies the force with which the sec-
ond body acts on the first, is equal mv-^=--F(t)v. (9.6.1)
in magnitude and opposite in direction
to the force with which the first body But according to the rule for calculat-
-acts on the second, F2 :
9 9
* ing the derivatives of composite func-
tions (see Section 4.3), the identity
F 2 (
t) = -F, (t).
dv d v2 \

As applied to the first body and force v I

IT IT \ ~2T j
F u formula (9.5.2) yields
t
is valid no matter what the function
v (t). Using this fact, we can rewrite
PAt)-Pi(to)= Fidt. (9.5.9)
(9.6.1) as
j
to

The same formula applied to the second m lTt {

J
t)= F W v '

body and force F 2 yields

t
or, since m is constant,

P 2 (t)-P 2 (f 0 )= F 2 dt. (9.5.10) =-((>-.

j
t°

-Since F2 = —F 1 by Newton’s third Introducing the notation

law, it follows that
(9.6.2)
t t

F^ dt == — F i
dt.
we finally obtain
j j
10 to

That is why (9.5.10) takes the form ^r = F(t) v. (9.6.3)

t
Recalling the expression (9.1.1) for
P i (t)~P i {t 0 )= —j F,dt. (9.5.11) work, we can write
to
it n
Comparing (9.5.9) and (9.5.11), we find
A= F(t)vdt=
that j j
to to

Pi (
t)
- Pi (f 0 ) = P2 (h) - P2 (t),
whence
whence - K
A = K (ti) (t 0 ). (9.6.4)
Pi ( + P2 ( = Pi (t 0 ) + P2 (h).
t t)
The quantity K is the kinetic energy
The subscripton F indicates the body
9 -
9
of the body. Formula (9.6.4) expresses
acted upon by force F; the subscript on P also
the conservation of energy : the change in
denotes the number of the body to which the
the kinetic energy of a body is equal to
momentum . refers.
9.7 Inertial and Noninertial Reference Frames 321

the work done by a force. Formula (9.6.3) form motion on physical phenomena
expresses the law that the rate of change holds true not only for mechanics but
in kinetic energy is equal to the power also for the propagation of light and
developed by a force. electric and magnetic phenomena. From
When the force is given by a definite this fact Einstein drew conclusions of
function of time, the impulse and, hence, tremendous importance in developing
the change in momentum caused by his theory of relativity (we do not ex-
the given force are dependent neither plain the theory of relativity in this
on the mass of the body nor on its ini- book). Newton, in formulating his
tial velocity, since the impulse and the laws of motion, assumed that there is
n absolute time which flows in the same
,

change in momentum are F dt. On way in the entire Universe, and ab-
^
to solute space in which all processes in
,

the contrary, the work done by a force the world take place. In the foregoing
and the change in the kinetic energy of we tacitly assumed all these concepts
a body under the action of this force to be valid. One can imagine, for in-
are essentially dependent, as may be stance, that somewhere in space there
seen from (9.6.2)-(9.6.4), not only on is an origin of coordinates 0, given, say,
the force itself but also on the mass of by the point at which three metal rods
the body and the initial velocity. In- placed at right angles to each other
deed, by acting with a given force over and having markings on them (or three
a specified time interval on a heavy body axes of coordinates, as is commonly
at rest at the start of motion we im- said) intersect. The same point has a
part only a small velocity that will re- watch attached to it. Then everything
sult in only a small displacement, and is simple: the coordinates of a body are
the work done by the force will likewise defined as the projections of the body’s
be small. Alight body will take up position on the metal rods (the axes
appreciable work and will acquire a of coordinates); the body is at rest if
large energy. If prior to the action of its coordinates remain constant. 910
the force the body was in motion in the In the general case, knowing the de-
opposite direction to the force, the force pendence of the x, y, and z coordinates
can reduce the energy of the body. on time t, one can find its velocity and
acceleration (more precisely, the three
9.7 Inertial and Noninertial Reference components dxldt, dyldt, and dzldt of
its velocity and the three components
Frames
d2 x!dt 2 d 2 y/dt 2 and d 2 z!dt 2 of its accel-
, ,

Picture a body participating in two mo- eration). Experience (in this case as-
tions at once, say, a man walking in the tronomical observations) has shown that
cabin of a ship in motion, or a ball drop- Newton’s laws are valid when the ori-
ped in the cabin. Suppose that one of gin of coordinates lies at the center of
these motions (that of the ship, in our gravity of the solar system, one axis
case) is uniform. The question then ari- points at the North Star, the second po-
ses whether by observing
it is possible, ints at an appropriate star in the equa-
the ball falling in the cabin or the mo- torial plane, and the third is perpen-
tion of some other body under the ac- dicular to the first two and points at an-
tion of an applied force, to establish
9 10
whether the ship is moving or not. To For the sake of simplicity we always
.

assume that the body is very small (that is,

put it differently, does the uniform mo- we are actually talking of a material particle).
tion of the ship influence the character The theory that studies the motion of finite
of motion of objects on the ship? No, bodies, that is, bodies that can deform and
it does not affect such motion in any rotate, is too complicated for this book. If
we speak of the motion of a finite body, we
way. Experiments have demonstrated always mean the position and motion of its
that the absence of any influence of uni- center of gravity.

21-0946
 b

322 9 Mechanics

other star (in other words, this star lies x (0lf t =

0) = a, — then in the new
on the axis that is perpendicular to the system of coordinates at t = 0 we will
first two axes). The sun’s mass is have x x = x +
a. If, in addition, the
2 X 10 30 kg, Jupiter’s mass is 1.9 X new origin is moving leftward with a
10 27 kg, and the distance between velocity (or speed) —6, then at time
Jupiter and the sun is 777.8 X 10 9 m; t =7^= 0 its position is x (O x
)
a bt. = — —
the mass of the rest of the planets is The position of the body reckoned from
much less than that of Jupiter. All this origin is
this positions the center of gravity of
the solar system at a distance of about
x\ = #(£) + a + bt , (9.7.2)

0.78 X 10 9 m
from the side of the sun which is just the right value for New-
facing Jupiter. The sun’s radius is ton’s second law to be valid (see above).
7 X 10 8 m, so that the center of gravi- The new system is moving, therefore,
ty of the solar system and, hence, the with a constant velocity with re- —
origin of coordinates used in astronom- spect to the old system, that is, it moves
ical calculations lies outside the sun’s by inertia in the same manner as a body
surface. on which no forces act. For this rea-
A system of coordinates can also be son it is said that Newton’s laws are val-
constructed say, a historical
using, id in inertial systems of coordinates ,
landmark. We
could place the origin that is, in systems that move by inertia
of coordinates at Trafalgar Square in with respect to one another without any
London, sending one axis “up” Nel- forces acting on them.
son’s Column, the second horizontally Even before Newton’s time the prin-
to the North Pole, and the third along ciple by which all inertial reference
the latitude line in the west-east direc- frames are equivalent was formulated
tion that passes through the chosen by Galileo, who wrote that in the cabin
origin. Newton’s laws operate in such of a ship sailing steadily in quiet water
a. at caaadJj^aa wUJl lass, accu- alL qatjui ucl tha mjmnpji
racy than in the above-noted astronom- as if the ship were at rest. Thus,
ical system; the reasons for this will among the great variety of systems of
he discussed later. coordinates, which can be associated
Even without conducting any spe- with a star, planet, or comet, there
cial experiments, Newton’s laws them- emerges a narrower class of inertial
selves imply that the system in which systems, or systems in which Newton’s
these laws operate is not unique. In- laws hold.
deed, Newton’s second law is In a system of coordinates rotating
with respect to an inertial system, on
r- ”* <
9 - 71 > the contrary, there appear additional
terms in the equations of motion (cen-
If to x we add a linear function of trifugal and Coriolis’s forces), which
time, the acceleration does not change, cannot be considered in a book of this
since if (t) x (t) a = bt we + + ,
scope. Rotation of the earth intro-
have duces certain correction terms into the
equations written in terms of coordi-
dx-i dx . . d 2x x _ d 2 x __ „
x% nates associated with the earth. Fortu-
dt 2 dt 2
'

di dt ’
1
nately,* ii we settne axes nyTar-cfn stars,
But if in the old system of coordinates we introduce, with a great accura-
the position of the body was character- cy, a practically nonrotating inertial
ized by the function x x (£), what = system of coordinates (which is simply
system should we take so that the po- one of the infinitude of inertial refer-
sition is characterized by x t ( t ) = ence frames). There was a time when
x (t) +a bt?+ If at time t 0 — some scientists concluded from this
we shift the origin to point Ox such that fact that the very property of inertia
 9.7 Inertial and Noninertial Reference Frames 323

depends on the presence of far-off stars; ing with a certain acceleration are
now, however, this view is considered similar to those that occur under the
arrJiAifv,. On a, shonJiL not. difi&uiSL + hfc fnrA5c ofx Thifb. mmab^ simian,
obvious property of inertia; rather, one line of reasoning led Einstein to his
should discuss the properties (precisely, basic assumption when he constructed
the positions) of the stars. The stars the modern theory of gravity, or the
closest to us are separated from the sun general theory of relativity .

by distances of about 10 16 to 10 17 m Let us examine the famous thought

and move (in relation to the sun) with experiment involving a very special
speeds ranging from 10 km/s to 50 km/s. noninertial system of coordinates, or
Although these speeds are not low, the noninertial reference frame. Consider
angular displacements of the stars are an observer in an elevator. As long as
infinitesimal; for this reason all sys- the elevator is at rest, the observer feels
tems of coordinates associated with only the force of gravity. For instance,
these stars can be considered inertial an observer standing still presses on
and in which the sun is at rest. the floor with a force Afg, where g
But can we do without inertial sys- is the acceleration of gravity, and M
tems of coordinates? One interesting is the observer’s mass.
case is a system in which the orientation Now suppose that at time t 0 the =
of the axes does not change, the axes do elevator is not at the ground floor and
not rotate with the passage of time, but is cut free of the cable. The elevator is
the origin moves along one of the axes therefore in free fall:
of an inertial system of coordinates
with constant acceleration.
dt*
~ 2
9

Thus, we take an inertial system of

coordinates (sometimes called a refer- where Z isthe altitude at which the ele-
ence frame if a clock is attached to the vator is at time t above the earth’s lev-
origin) and specify the law by which el. The observer inside the elevator
the origin of a new system of coordi- reckons his position z x from the eleva-
nates, 0 1 (x), moves in relation to the old tor’s floor. The position of the observer
origin (the motion is along the x axis): with respect to the earth’s surface, z
x (£)
= —/ (£). Then x 1 (
t) = x (t) + is related to z 1 in the following manner:
,

/ (£) and, hence,

d*x x 2
__ d x d2 f 1 d2f Zt =z Z=z z0 -j
.
.

_ px ,

dt 2
dt 2 '
dt 2 m '
dt 2
Let us write the equation for coordi-
Measuring x x (t) and applying Newton’s nate z x of a body (material particle) act-
second law to this quantity, the obser-
ed on by the force of gravity and other
ver will conclude that there is a force
forces F (elastic forces of springs and
m (d 2f/dt 2 ) acting on the body, in addi-
the like). The force of gravity is gm —
tion to the force F x If the observer
,
.
where m is the mass of the particle.
compares the behavior of various bodies In the system of coordinates associated
in the new system, he or she will
with the earth the equation of motion is
arrive at the conclusion that this addi-
tional force is proportional to a body’s
mass, since only in this case the addi-
m^ = F-gm.
d*z „

tional acceleration d 2f/dt 2 (which occurs

in the system) will not depend on the
Using this equation, we can find the
equation governing the motion of the
mass of a body.
particle in the system of coordinates
force we have just studied resem-
The
zx associated with the elevator:
bles the force of gravity. In other words,
the phenomena which take place in a
noninertial system of coordinates mov-
21 *
324 9 Mechanics

We see that there is no force of gravity of other stars in our Galaxy and other
in this system of coordinates. In a free- galaxies are negligibly small. Since
ly falling elevator the g-force is zero. the density of matter is evenly distrib-
This state isreferred to as zero g. For uted (on the average), both the poten-
all bodies on which no outer forces act tial and the force of gravity are expres-
We have cPzJdt 2 = 0, which yields sed by integrals that assume infinite
zx =
constant, or the state of rest in values (this is known as the gravitatio-
relation to the elevator. (The reader will nal paradox). But the absolute value of
recall that from the viewpoint of an the gravitational potential never ap-
observer on the earth, the elevator and pears in the theory, and the derivative of
the particle inside it fall with the same the potential, or the force due to the at-
Velocity and the same acceleration g.) traction of other bodies, is also com-
The state of zero g is terminated only pletely compensated for by the free fall
when the elevator hits the ground (even of the solar system regardless of whether
if the impact demolishes the elevator). this force is finite or infinite. This jus-
Up to this moment there is no way in tifies a study of the solar system irre-
which experiments conducted inside spective of the rest of the universe and
the elevator can determine the pres- removes the gravitational paradox.
ence of the earth’s gravity.
The work on an earlier version of this
hook was started before man went on a
9.8* The Galilean Transformations.
space mission. Most of you have seen TV
Energy in a Moving Reference Frame
programs beamed from space stations and
satellites. As you could see, the astro- Let us go back to inertial reference
nauts experience the state of zero g. frames and consider in greater detail the
But a spaceship is usually only some relationships between the description
200 to 300 km from the earth, where of one and the same phenomenon in
the force of gravity diminishes (accord- different inertial reference frames. Sup-
ing to Newton’s law of gravitation) pose that an object is in motion in the
only by several percent: coach of a train that is moving with a
constant speed v 0 in a direction fixed
Ag by the rails of the track, the direction
8
__ 9 Ai?
R
0
A
300
6400
— — 0.09
'v/
being that of the x axis. We
willassume
that the train is moving in the direc-
1 = -9%. tion in which x increases. Then the
coordinate x x of the object with respect
This leads us to the conclusion that to an observer who is at rest in relation
the total weightlessness (zero g) exist- to the track, say, an observer standing on
ing in the spaceship is due mainly not the platform of the railroad station
to the fact that the force of gravity gets which the train has left, will be related
weaker as one moves away from the to the coordinate x of the object in a
earth but to the fact that with the rock- reference frame fixed within the coach
ets of the spaceship switched off the (the origin of the coordinate system may
latter is in a state of free fall, just as be fixed at the back wall of the coach,
the falling elevator is. for instance) as follows:
We
can now approach the problem of
xx = x + v0 t + a,
inertiality of the reference frame asso-
ciated with the sun from a different where a depends on the choice of the
angle. We
do not insist that both the origin of coordinates x x if this origin
\

gravitational potential and the force coincides with the station which the
of gravity acting on the sun and other train has left, then a is the ^-coordi-
planets (and, in fact, all bodies in the nate of the back wall of the coach at
solar system) because of the attraction time t =
0 (cf. formula (9.7.2)). If, in
9.8* The Galilean Transformations. Energy in a Moving Reference Frame 325

addition, we assume that the initial tions with respect to the two observers
moment of the “train time” t (the ini- will be the same:
tial moment may be assumed to be the dv d i , dv du o
time of departure of the train from the a i--=—=-jr( v -^ v o) = ^r
dt dt
station) differs from the initial mo-
dv
ment of the “absolute time” t 1 (not con-
dt
nected with any train), say, we can as-
sume that t x =
0 coincides with 00:00 since the constant term v 0 constant =
GMT at the station from which the train in the expression for the speed cannot,
departed, 911 then we must write the of course, change the acceleration. There-
additional formula t x t 6, where = + fore, a force acting on the object is
b is the “absolute” time t x at t 0 (the — the same for the two observers, or
moment of departure). The transforma- F = ma x — ma.
tions Thedifference in speeds before and
after the force is also the same for the
x1 = x + vQ t + a, tx = + t b (9.8.1)
observer on the platform and the observ-
determine the relationship between the er in the coach. Indeed, let the speed
coordinates (x x tf) and (x, t) of one and
, of the body in relation to the observer
the same object
in two inertial refer- on the platform be v' prior to the force
ence frames, one of which (the train) and v" after the force, while for the ob-
moves uniformly with respect to the other server on the platform these speeds
(the station) with a velocity (or speed are v[ and v ", respectively. Then v[ =
in our one-dimensional case) u 0 and are , v +v0 and v'' v whence = +
called the Galilean transformations .

Let us now assume that the object — v[ = v

n
+ v0 — v' — v0
is moving within the coach (in the di-

rection of increasing xs) with a certain

speed v . Then The situation is more complicated with
x = x0 + vt. (9.8.2)
kinetic energy. Not only the kinetic energy
itself but even the differences of kinetic ener-

We then have gies are distinct for different observers. For

the observer standing on the platform,
xx = x + v '+ a = (x +
0t 0 vt)
m 2 m (v[) 2
K'i-K[ = ~~2 ( v'l )

H~ v0 t + a = (x + a) 0 2

+ (v Q + v) t. (9.8.3) m (^" + ^o)

2
m (v'+v 0 )
2

2
(If we
substitute t1 — b for f, we can m (v") 2 m (v f

rewrite the last formula as xx — 2 2

)
mv 0 v — mv 0v

l(*o + a) — (v + v 0 ) b] + (v + v0) t lt
= K" — K' -\~mv —
but for sake of simplicity
the o (v" v').

we will not distinguish between t x In this formula K\ and K{ are the final and
and t.) From (9.8.3) it follows that in initial kinetic energies calculated by the ob-
relation to the observer on the platform server on the platform, and K" and K' are,
respectively, the kinetic energies calculated by
the object is moving with a speed v1 = the observer in the coach.
v +
v0 This speed will, of course,
.
The work done by a force and the power are
be different for the observer traveling also different for different observers, since
in the coach. However, the accelera- although the force is the same, the distances
and the velocities are different for the observer
standing on the platform and for the observer
9 n
It goes without saying that on a long
-
in the coach. However, the law of equality
journey the moment 00:00 GMT does not of change in kinetic energy and work is valid
change but a new time zone may force us to for any observer, although each of these quan-
set our watch back or ahead, which makes it tities taken separately differs for different ob-
especially appropriate to distinguish between servers (see the exercises to this section for
“train time” t and “station time” t x . examples corroborating this fact).
 — ^ ,

326 9 Mechanics

Note the following remarkable formula

Valid for a body moving under the action of
only one given force F (t):
ti
mv\
F ( t) v ( t)
x ,
dt = mi?n

2 2
to

= -y- f'Ji — 'Joj (’-’I — ft)

= =J2±a. f ,„*. Figure 9.8.1

j
*0

We see that in this case the velocity (speed)

v may be taken out from under the integral
(f)
ful factsand quantities are those which do not
sign and replaced by the arithmetic mean of change under transformations (9.8.5). For
the initial and terminal velocities. instance, the linear equation
This conclusion holds true only for the x — vt -f- a (9.8.6)
case where v (t) is the velocity acquired by
an object under the action of only one force expresses the fact of uniform motion of a
F (t). acted upon by a number
If the body is point along the x axis (the very straight line
of forces, say, Fx F 2 and F 3l then the work , ,
along which we consider the motion), but the
performed by all these forces is equal to the specific value of v depends on the choice of
product of the mean velocity by the sum of the the concrete reference frame and has no phys-
impulses of all forces: ical meaning. (It is for this reason that New-
ton’s laws contain only the acceleration and
t L
not the velocity of a material particle.) But
a =ei±o± (Fi+F2+jFs) dt if we are dealing with two uniform motions,
J expressed via (9.8.6) and x =
vx t +a l7 re-
to
spectively, the difference v1 —
v has a concrete
physical meaning: it is the relative velocity
=-a±2*-ij of one motion with respect to the other motion
and does not depend on the choice of (inertial)
*0 to reference frame. There is another physical
tl quantity with a definite meaning: the inter-
J val t = —between two events, (x x
+- ^
l>0

j
F3 dt. (9.8.4) tx ) and (x 2
t2 tx

t 2 ); if the two events occur simul-

,
,

to taneously, tx t2= we can speak of the spatial

,
(9.8.4) distance (or space interval) d —
x2 —x x be-
However, the work done by each these of tween these events (the “distance” x between the
forces (say, F2 )
separately is not equal to the events is measured in units of time, say, sec-
ti
onds, while the second “distance,” d, which
corresponding summand — ^ j* F 2 dt in has a meaning only at t =
0, is measured in
entirely different units, say, meters).

,
since the force F2
to
acting separately
An arbitrary curve x =
x (t) in the xt-
would impart a velocity to the object that plane (Figure 9.8.1) fixes the law of motion of
differs from u ( t ) (see Exercise 9.8.6 below).
a (material) point along a straight line. The
Above we saw that two inertial reference slope k —
x' (=dx/dt) of the tangent to the

frames, (x t) and (x\ t '), specified on the x curve at point (or event) (x, t ), which it would
7

axis be more appropriate to denote by v, gives the

are related through the following for-
velocity of motion at this point. If we substi-
mulas:
tute for curve T (equation x —
x (£)) the tan-
x' = x-j- vt -j- a, t' = t -j- b, (9.8.5) gent l at the given point (this tangent consti-
tutes a good approximation of T in the neigh-
where v is the velocity of the reference frame
borhood of this point), we transform the ar-
(x, in relation to the reference frame (x\ t'),
t)
bitrary motion into the “approximating” uni-
so that the velocity of x' t') in relation to
form motion with a velocity equal to the in-
(
x , t) is
(
— v. The Galilean transformations
,

stantaneous velocity at the given moment in

(9.8.5) or (9.8.1) play in mechanics a role time. The second derivative x" ( 2
=d
x/dt 2 )
similar to that played in geometry by motions plays the role of the curvature of the curve and
(cf. Section 1.9). The transformation from one
also expresses the acceleration of motion. If
inertial reference frame to another inertial
reference frame has no effect on physical phe- d^x
- - = a= constant
nomena; hence, the only physically meaning- -
,
2
 )

9.9* The Path of a Projectile. The Safety Parabola 327

we are dealing with the law of motion 9.8.6. A body is under two forces: Fx — at
and F 2 = a (0 —t). The impulses of these
forces over the interval from 0 to 0 are the
same. At time t =0 the body has a velocity v0 .

Find the work done by each force during the

with b and c arbitrary constants, which law time interval from 0 to 0 and compare it
corresponds to uniformly accelerated (if a is with the product of the impulse by the av-
positive) or uniformly decelerated (if a is nega- erage velocity.
tive) motion and geometrically corresponds to 9.8.7. A man standing still on the ground
a parabola (9.8.7) in the zJ-plane. Replacing acts on a given mass m with a force during the
an arbitrary
9 curve x =
x (t) by the approxi- time interval t. As a result the mass, which
mating parabola IT (9.8.7) with the same values was originally at rest, acquires a velocity vx =
of derivatives x' (=v) and x" ( —a means re- Ft/m and a kinetic energy mv\/2 equal to the
placing an arbitrary motion by a uniformly work performed by the man.
accelerated (or decelerated) motion with the Consider the same experiment done in a
same (instantaneous) velocity and accelera- train traveling at a velocity v0 The mass m.

tion. Geometrically this replacement means that had a velocity v0 prior to the experiment and
for curve x = x (t) in the zf-plane we have
substituted the osculating parabola (cf. Sec-
v0 + vx after the experiment. Find the change
in the kinetic energy of the mass m. What
tion 7.10). Physically, the transition from an work was done by the man? Assuming that
arbitrary curve to its tangent (at a certain point, the man rests firmly against the wall of the
of course) means eliminating all forces, since railway car and v0 does not change, find the
a straight line in the zZ-plane represents uni- work of the force done by the train (locomo-
form motion, which occurs in the absence of tive) during the experiment.
all forces (motion by inertia), while the transi-
tion from a curve to a parabola (9.8.7) is equi-
9.8.8. A man of mass M
standing in
skates on ice (friction between skates and ice
valent to the assumption that all forces are is neglected) acts with a force F on a mass m
constant. 12 -
during time t. What kinetic energy will be im-
parted to the mass ml What kinetic energy
will the man acquire? What is the total work
Exercises
done by the force acting on mass m and on
9.8.1. Find the formula for the kinetic the man? Why is it greater than in Exer-
energy of a body moving under a constant cise 9.8.7?
force F (with zero velocity at the initial time) 9.8.9. The same experiment as in Exercise
as a function of time and also as a function 9.8.8, but the man has an initial velocity z>0
of the distance traveled. and moves together with mass m. The velocity
A body of mass m is, after the action of the force,

F
9.8.2.
— f cos G)£, and
is
v —in 0motion
att
under a force
= 0. Find the v0 + Ft/m and the velocity of the man is,
,

expression for the kinetic energy of the body, respectively, v0 —Ft/M. Find the change in
and determine the maximum of kinetic energy. kinetic energy of mass m and the man as a re-
9.8.3. A body is moving in accordance with sult of the action of the force. Find the work
the law x = x (t) = A cos (coi a), with A, + done by the force, which work is equal to the
a), and a constants. Determine the average change in the total kinetic energy, and com-
kinetic energy provided that t increases with- pare it with the result of the preceding exer-
out bound from t = 0. cise.
9.8.4. A
ball of mass falls from a heightm 9.8.10. Prove that under the Galilean trans-
H from a state of rest. Demonstrate that the formations (9.8.5) the following quantities
kinetic energy of the ball, A, is equal to retain their values: (a) the temporal interval t
mg (H —
h ), where h is the height of the ball between two events, (b) the spatial distance d
above the ground at a given instant of time. between two simultaneous events, and (c)
9.8.5. A
train with a mass of 500 metric the relative velocity ux —
v of one uniform
tons started out from a station, and in 3 min- motion relative to another.
utes developed a speed of 45 km/h traveling 9.8.11. Verify that a transformation to a
1.5 km. Determine (a) the work and the ave- new (inertial) reference frame via the Gali-
rage power of the locomotive on the assump- lean transformations (9.8.5) does not change
tion that friction on the rails is absent, and the acceleration a of (uniformly accelerated
(b) the same but having regard for friction or decelerated) motion in (9.8.7).
(the coefficient of friction k is equal to 0.004;
the force of friction is equal to the force of
attraction of the train to the earth, or the 9.9* The Path of a Projectile.
train’s weight, multiplied by k). The Safety Parabola
Let us consider the problem of the
9
See I. M. Yaglom, A Simple Non-
- 12

Euclidean Geometry and Its Physical Basis flightpath of a projectile (shell) fired
,

Springer, Berlin, 1969. from a gun with initial velocity v 0 .

328 9 Mechanics

We take the point of ejection of the mentary case in which the shell in
shell from the barrel of the gun for the upon solely by the force
flight is acted
origin and send the y axis vertically of gravity directed earthwards, it fol-
upward, while the x axis is assumed lows that F x 0 and F y — =
mg. And —
horizontal (the motion of the projec- so the equations in (9.9.1a) have the
tile occurs, therefore, in the £g-plane). form
For the sake of simplicity we disre-
gard air resistance because this would m -7r= ~ mg -
(
9 9 2>
- -

introduce considerable complications.

By Newton’s second law, What remarkable (and important)
is
here that our equations have sepa-
is

<
9 - 91 > rated: the first contains only one un-
known function v x and the second con- ,

We applied this law earlier only for tains the unknown function v y .

rectilinear motion. However, in the Let us find the initial conditions for
flight-path problem the direction of u the functions v x (£) and v y ( t ). At the
varies with time (the velocity is always time of emergence of the shell from the
directed along a tangent to the path barrel, t = 0,
of the shell). For this reason we are
forced to assume that the quantities
v x (0) = v 0 cos cp, v y (0) = v 0 sin cp,

v and F in Eq. (9.9.1) are vectors or , where v 0 =

v 0 ) is the initial (or
(
= and F = F x F y ).
| |

v (v x , v y ) (. , muzzle) velocity of the shell. The first

Any motion in the xy - plane may be equation in (9.9.2) yields dvjdt 0, =
regarded as the result of combining from which it follows that v x is con-
two motions: one occurring along the stant, and hence
x axis under force F x with velocity
v x and the other along the y axis un-
,
Vx (t) = v x (0) = v 0 cos cp. (9.9.3)
der force F y with velocity v y Applying
.

Newton’s second law to each of these The second equation in (9.9.2) yields
motions separately yields dvyldt g, = —
whence, integrating both
parts from 0 to t we get v y (t) v y (0)
,
— =
m-^ = F x) m^-=F v (9.9.1a) —gt, or
vy (t) = —gt + sin cp. (9.9.4)
(see Figure 9.9.1). In each of these
equations the force and the velocity
are directed along a single straight To determine the displacements x
line (the x axis in the first equation and and y along the coordinate axes, we
the y axis in the second). take advantage of the fact that v =
Denote by cp the angle which the bar- drldt where r = (x, y) is the radius
,

vector of the shell. This vector equa-

rel of the gun makes with the horizon-
tal; call cp the angle of departure. tion splits into two scalar equations:
Since we are considering the most ele- dx
Vx , y> (9.9.5)
dt

or, in view of (9.9.3) and (9.9.4),

— n0 cus cp, -^-= —£f + i>

0 sin<p.

(9.9.6)

The equations are again separated: the

firstdeals only with the unknown func-
tion x ( t ), and second deals with the
Figure 9.9.1 unknown function y ( t ).
 —
,

9.9 * The Path of a Projectile. The Safety Parabola 329 '

At
the initial instant of time, the termine the coordinate x 1 of the maxi-
shell at the origin, with the re-
was mum altitude of the shell, y max ,
we
suit that set up the equation dyldx = 0 to get
x = 0 and y = 0 at t = 0. (9.9.7)

Integrating Eqs. (9.9.6) from 0 to i

and using the initial conditions (9.9.7),
we find that

x =v 0t cos cp, y—v 0t sincp — -y-

< 7^2
.

(9.9.8) For this value of the height y is at

x,
its maximum physically clear
(it is
These formulas enable us to determine that this is precisely the maximum; this
the position of the shell at any can be verified easily by finding the
instant of time t. sign of d2 y!dx 2 ). Substituting this
Taking various values of t we can , value of x into (9.9.9) yields
find the position of the shell from for-
Vq sin 2 cp
mulas (9.9.8) at different times and
2g
plot a graph of the path of the shell.
Thus, Eqs. (9.9.8) yield a curve in the To determine the range of the shell,
.ry-plane. They determine a parametric it suffices to find the value x 2 for which
representation of the path of the shell, y = 0 (see Figure 9.9.2):
with t the parameter (see Section 1.8).
From Eqs. (9.9.8) we can easily ex- x2 tan cp — x\ g
2v% cos 2 cp
0.

clude t and obtain an equation of

the path in the ordinary form, as a func- Disregarding the solution x 2 = x0 = 0
tion of y in x. Indeed, the first equa- which does not interest us, we find
tion in (9.9.8) yields t xlv 0 cos cp; = that
then from the second we get v% sin 2cp
x2 (9.9.10)
g
y = xianw — x 2 -0-o - g b
— • (9.9.9)
The range of fire depends on the ini-

From this we see that y is a second-de-

tial velocity and the angle of departure.
gree polynomial in x, the graph of which
For what angle of departure (initial
is a parabola. Consequently, disregard-
velocity u 0 unchanged) is the range of
fire the greatest? Formula (9.9.10) shows
ing air resistance, we can say that
the path of a shell is in the shape of a
that this happens when sin 2cp 1 =
parabola. Figure 9.9.2 depicts the path
or cp —45°, and the range in this case

specified by (9.9.9) for the case where

is v^/g. Also, since sin 2cp sin 2 (90° =
cp) and in view of (9.9.10), for angles
v0 — 80 m/s and = 45°.
From evident that for
(9.9.9) it
cp

is
cp 0 and 90° —
cp 0 (say, at angles of de-
one and the same v 0 the shape of the ,
parture cp =
30° and cp 60°) the range =
is the same and shells hit the same tar-
trajectory depends on the angle of de-
get; artillerymen call fire with shells
parture cp. We will find the maximum
altitude of ascent of the shell and the
having angles of departure less than 45°
grazing and fire with angles of depar-
range of fire for given cp and u 0 To de- .
,

ture greater than 45° plunging.

9> Let us determine the time t x during
100 which the shell rises upward. All we
50 need to do is solve the equation dy(tf)!dt
0 = 0 because at time t £ 1? when =
y attains its maximum value, the shell
Figure 9.9.2 will cease to rise and will begin to fall.
 l

.330 9 Mechanics

The condition dyldt = 0 yields v 0 sin cp

— gt = 0, whence
P° s 9
ti = . (9.9.11)
-f
The total flight time t 2 can be found
from the fact that the flight ceases when
x = x 2 Combining (9.9.8) and (9.9.10),
.

we find that

sin 2(p
v 0 t 2 cos cp
j

this can be verified by directly calcu-

whence lating the slope of the tangent to both
2 u0 sin curves at point (9.9.14). Parabola (9.9.13)
t2
g
cp
(9.9.12) can be called a safety parabola —
if a target A (say, an aircraft) is out-
Comparing (9.9.12) with (9.9.11), we side the “dome”, there is no angle of de-
see that the total flight time t 2 is twice parture at which it can be hit. The tar-
the time of ascent t x for any angle of get can be hit only by using another
departure, or that the ascent time of a type of gun.
shell is equal to the descent time. Note, in conclusion, that the actual
Now let us turn from the problem of flight paths of shells are not exact pa-
firing at targets on the ground to rabolas. In reality the shell experi-
antiaircraft fire. If for a given value of ences air drag, which we ignored; the ac-
the shell’s initial velocity (this veloci- celeration of gravity g depends on the
ty depends on the type of gun) we change distance from the surface of the earth;
the angle of departure, we obtain a rotation of the earth also introduces cer-
whole family of trajectories (Figure tain distortions into the fall of the shell;
9.9.3), where the plunging-fire trajec- and, finally, Newton’s laws of motion
tories, that is, trajectories with of a material point cannot be applied
45° < <
cp 90°, touch a single curve without certain reservations to the mo-
*’* 1 /
110 x 1 ~bi~ a“prirjet4 ne Vcrfino a]>if ac-
9 9 13) tor not so important). This, however,
v=-$-*^3T (
- -

is
is
not a unique situation— we en-
(the grazing-fire trajectories lie inside counter it whenever we apply mathemat-
the “dome” (9.9.13) and do not touch ics to real-life problems; we always ig-

it). Indeed, solving Eqs. (9.9.9) and nore certain factors and are content
(9.9.13) simultaneously, we find that with a rough picture, which enables us
the curves (parabolas) have only one to use mathematical tools we are accus-
common point: tomed to. The actual range of fire, the
altitude of flight of a shell, the time of
* = -y-cotcp, !/=—|L(i_ C os 2 <p) flight, and so on depend on the mass of
the shell, the shell’s shape, and the air
(9.9.14) density. For instance, when the initial
velocity of a shell is low (see Figure
(this point lies in the upper half-plane 9.9.2 that refers to the case with u 0 =
only for cp ^
ly plunging-fire trajectories in
45°, in view of which on- 80 m/s), the role or air drag is in-
(9.9.9) deed insignificant, while for a great v 0
touch the curve specified by (9.9.13)). there is no way in which the drag can
The fact that the appropriate parabolas be ignored; for a 305-mm caliber gun
only touch (that is, do not intersect) fol- with a muzzle velocity of u0 =
lows from the sole fact that the common 800 m/s and cp = 55°, the air drag
point of (9.9.9) and (9.9.13) is unique; that acts on the flying shell diminishes
 9.10 The Motion of a Body in Outer Space 331

the flight range from 61 km (which of g with altitude; the value of g at

value is given by formula (9.9.10) the earth’s surface is denoted by g 0
to 22.2 km. (it is this value that was substituted
into formulas (9.10.1) and (9.10.2)).
Exercises
Indeed, by Newton’s law of gravitation,
a body placed at a distance R from the
9.9.1. A shell leaves the gun with a velo-
center of the earth is attracted to the
city of 80 m/s. Determine the range of fire
and the maximum height reached by the shell earth with a force F GmMlR 2 where =
M
,

if the angle of departure <p =

30°, 45°, and 60°. m is the body’s mass, and the earth’s
9.9.2. Determine the maximum altitude mass. In addition, by Newton’s second
at which a shell with initial velocity v0 = law, F =
mg where g ,
g (R) is the =
80 m/s can hit a target located 500 meters from
the gun.
acceleration of gravity at a distance R
from the earth’s center. Comparing the
two expressions for F we can write
9.10 The Motion of a Body ,

in Outer Space g =
GM/R 2 If R . r0 then g g0 = , = ,

whence g 0 = GMlr\ which , yields

Let us now turn to the problems of the G = gorf/M and, therefore,
motion of objects (artificial satellites g(R) = g0 rl/R 2 .

and spaceships) in outer space. For a

body launched from the earth to be- The condition (9.10.1) for the centrifu-
come a satellite of the earth, that is, for gal force being equal to the force of
it to orbit around the earth, the centri- gravity then assumes the form
fugal force on the body must be bal-
mu\!R = mg 0 rl/R 2 ,
anced by the gravitational force of the
earth. The respective velocity u x is which means that the satellite velocity
commonly known as the satellite (or u-l is given by the following formula:
913 To find
orbital) velocity . u l9 we ui = 2 i'2
(g0 r 0 /R) .

set up the equation

The greater the distance R the lower
m ~ = mg, (9.10.1) the velocity u ± necessary for a satellite
,

to remain in orbit. However, this does

where g is the acceleration of gravity, not at all mean that it is easier to
m the mass of the body in question, and launch a satellite to a higher orbit than
R the orbit’s radius (the left-hand side
to a lower orbit, since to launch a sat-
of Eq. (9.10.1) is the centrifugal force, ellite to a high orbit we must spend a
and the right-hand side is the force of lot of energy in overcoming the force
gravity of the earth). If the orbit is
of gravity on the trip from the earth’s
low, that is, the body flies close to the
surface to the orbit. Of practical impor-
earth’s surface, R
is close to r 0 the ,
tance is the motion of satellites on a
earth’s radius, whereby on the right-
stationary, or 24-hour, orbit, which
hand side of (9.10.1) we can put the lies approximately at an altitude of
force of gravity at the earth’s surface.
36 000 km above the earth’s equator; a
Formula (9.10.1) then yields satellite launched into such an orbit will
= Y gR ~ V gr ^ 8 km/s.
0 (9.10.2)
“hang” over a spot on the equator (see
Exercise 9.10.1).
For a satellite that orbiting the
is Let us now consider a more compli-
earth at a distance R from the earth’s cated problem. For a body to leave the
center that exceeds r 0 considerably, we bounds of earth’s gravity, its initial
must take into account the variation kinetic energy must be greater than the
difference between the potential energy
9,13 We are interested here in steady-
state motion with a velocity u x (and later with of the body at a far-off point and that
velocities u 2 and u 3 ) and ignore the transient at the earth’s surface. This difference
launching stage. was found in Section 9.2 (formula
 —
332 9 Mechanics

(9.2.8)). Here it is assumed that the By Newton’s law of gravitation, the

body acquires its speed rapidly over a force with which a body of mass is m
path that is short compared to the attracted to the sun is F kMpnir 2 = — ,

earth’s radius, so that the variation in where 1 M

is the sun’s mass (M 1 ^
potential energy over this path can be 2 X 10 30 kg), r the distance from
ignored. In other words, we must as- the body to the sun’s center, and k
sume that the thrust of the reactive force a constant. The potential energy of the
operating while the rockets launching body separated by a distance r from
the satellite are burning (see Section the sun’s center is
9.11) is very high and the force of grav-
ity can be ignored.
914 u (r) — ^^~m. (9.10.4)
The minimal velocity that a moving
body (as a rocket) must have to escape Here the zero of potential energy is
from the gravitational field of the earth taken as its value at infinity (cf. Section
or of a celestial body and move outward 9.2).
into space is called the escape velocity. The value of the potential energy of
Let us find this velocity for the earth a body which orbits the sun together
(we denote it by u 2 ). By (9.2.8), the with the earth can easily be expressed
initial energy required to reach a point in terms of the speed with which the
in space lying at a distance from the R earth orbits the sun. Indeed, on the
earth’s center, if prior to motion it was earth’s orbit the force with which the
at a distance r 0 (on the earth’s surface), earth is attracted to the sun is balanced
is given by the formula K — mg(rJR) x
Q
by the centrifugal force, that is,

(R —r 0 ). In our case R is much larg-

er than r 0 whence R ,
—r 0 cz. R, which
yields K ~ 0 mgr 0 . We assume this to
be equal to the kinetic energy of the where vx we already know, r x is the

rocket: radius of the earth’s orbit (r x 150 X ~

10 6 km =
1.5 X 10
11
m), and 2 M
mu\!2 = mgr 0 .
is the earth’s mass (this quantity is

This yields cancelled out from Eq. (9.10.5)).

Whence, kM 1 = v\rx , and (9.10.4) as-
u2 =. ]/ 2gr 0 ~ 11.2 km/s. (9.10.3) sumes the form
Combining (9.10.2) with (9.10.3), we u (rj) = — v\m .

find that
For a body at a distance r t from the sun
u.1 —y2u i
c^\Au u and u\ — 2u\. to leave the bounds of the sun’s grav-
itation, the sum of the body’s kinetic
For the sun, this velocity (denoted and potential energies at this distance
^ 3 ), when imparted to a body, takes the must be nonnegative. This leads to the
body outside the solar system. We will
following inequality:
find it using the fact that the speed with
which the earth orbits the sun is known: m + u (f\) = m -y- — mv\ > 0,
vx ~ 30 km/s.
(9.10.6)
14
9 -
has been found that it is more advan-
It
tageous to burn the rocket fuel fast (less fuel where v 2 is the velocity that the body
is required) than to extend the burning process
must have to leave the solar system,
over the entire time necessary to travel a dis-
and is the known orbital velocity of
tance of the order of the earth’s radius. Only in
the layer where the density of the atmosphere the earth.
is high, and so is the air drag, high velocities We have already encountered a sim-
are not advantageous, but since the thickness ilar situation when we considered the
of this layer is much smaller than the earth’s
radius (see Chapter 11), we will not take
motion of a body in the earth’s gravi-
this layer into account. tational field: the velocity necessary
9.10 The Motion of a Body in Outer Space 333

for a body
leave this field corre-
to this to happen (it is this velocity that
sponds energy that is twice
to a kinetic is called the escape velocity for the
the kinetic energy corresponding to the sun)? It can be found from the obvious
velocity that ensures that the body relationship
is an (artificial) satellite of the earth.
Inequality (9.10.6) yields the mini- m mgr 0 -(- m — -
. (9.10.7)
mal velocity v 2 that a body must have
to leave the solar system: Here the first term on the right-hand
side the energy necessary for over-
is
v2 — ]/ 2u 1 ~ 1.4 u 1 ^ 42 km/s. coming the attraction of the earth, and
the second term is the energy that the
Thus, to leave the solar system, a rocket (or spaceship) of mass must m
body on the earth’s orbit must have an have after it has overcome the earth’s
initial velocity (with respect to the gravity for its speed v 2 to be sufficient
sun) that is at least 42 km/s. If the ve- (together with the 30 km/s of the
locity is greater than 42 km/s, the body earth’s motion in orbit) for the rocket
will leave the solar system irrespective ;
to leave the solar system. Formula
of the direction of this velocity, that (9.10.7) then yields
irrespective of whether the body
is,

moves along the radius away from the

u\ = 2 gru + (v' z = ul +
2) (9.10.8)

sun (path 1 in Figure 9.10.1) or along whence

a tangent to the earth’s orbit (paths
u 3 =(u 22 + (v'
2
yy/2^ (
11 2 2 .
+ 12 2 ) 1/2
2 and 3) or even in the direction of the
sun (but at a certain angle so as not to ~ 16.4 km/s. (9.10.9)
land on the sun’s surface, path 4). Only
Note that for the rocket to get clos-
the shape of the trajectory depends
er to the sun or reach, say, Mars or
on the direction of the initial velocity
Venus, u 2 is insufficient. Indeed, hav-
(see Figure 9.10.1).
ing this velocity, the rocket will leave
It is clear that the most advanta-
the earth and move along the earth’s
geous trajectory for launching a rocket
orbit with the velocity of the earth, or
from the earth’s surface is 2: the earth
30 km/s.
moves with a speed of 30 km/s, so that
Although the potential energy does
only 12 km/s is required to obtain
diminish as the rocket moves toward
42 km/s if we launch the rocket in the
the sun, but the centrifugal force that
direction of the earth’s orbital motion.
acts on the rocket moving along the
This speed v'2 the rocket must have af-
orbit prevents the rocket from doing
ter it has left the earth’s gravitational
this. To get closer to the sun, the rocket
field, that is, after it has traveled a
must diminish its speed, which is as
distance large compared to the radius
difficult as increasing the speed. For
of the earth’s orbit.
instance, to hit the sun, the rocket must
What
initial velocity u 3 must the
be stopped, that is, it must move with
rocket have at the earth’s surface for
a velocity of 30 km/s with respect to
the earth (after it leaves the earth’s
gravitational field). For this the rocket
must have an initial velocity
uA ~ (30
2
+ 11.2 2 ) l/2 ~ 32 km/s
(9.10.10)

at the earth’s surface.

Wesee that it is more difficult to
hit the sun than to escape from its
Figure 9.10.1 gravitational bounds. More “profitable”
 —

334 9 Mechanics

variants of space travel can be obtai- ed into the thermal energy of the pro-
ned by using the gravitational pull ducts of combustion, which stream out
of other planets. However, we will not of the nozzle, the thermal energy turn-
discuss such questions in this book. ing into the kinetic energy of motion.
When a reaction (rocket) engine is
Exercise 9
fixed on a test bed, the combustion
9.10.1. Find the radius of an orbit in which products are exhausted at a definite
a satellite circles the earth every 24 hours. velocity u 0 The kinetic energy they
.

(If a satellite is launched into such an orbit

have per unit mass constitutes a definite
in the equatorial plane, it will “hang” con-
stantly over a single point on earth.) portion of the chemical energy of the
propellant:

9.11 Jet Propulsion and Tsiolkovsky’s

Formula
4-K = «Q, (
9 11
. . 1)

where a is a dimensionless number, the

In the case of motion in airless space, efficiency of the processes of combus-
the only method of flight control, tion and ejection of gases. 16
From now -

which is widely used and is highly effec- on we will consider the exhaust veloc-
tive, that is, a method that enables ity u 0 to be a given known quantity.
changing speed and direction, consists It is roughly 2 km/s for powder and
in ejecting a portion of the mass of the about 3 km/s for liquid propellant. It
flying object itself, which means apply- is easy to see that these quantities are
ing the reaction (jet) principle. associated with the values of a ~
The Russian scientist K. E. Tsiol- 0.5 (an efficiency of the order of
kovsky (1857-1935) was the first to 50%).
fully realize the significance of the jet Prior to combustion, the propellant
principle and to investigate the funda- was at rest. Suppose a mass dm of
mental regularities of reaction or jet , propellant is burnt and exits from the
propulsion. He was also the first to nozzle. In so doing, it acquires a mo-
suggest the possibility of conquering mentum equal to u 0 dm. Clearly, the
outer space by means of rockets. From impulse dl of the force with which the
him, via his pupils and followers rocket acts on this mass is equal to the
Soviet scientists and engineers stems — momentum acquired by the mass, 9 17
the scientific tradition that saw final or
embodiment in artificial earth satel- dl = F dt = u 0 dm.
lites, space probes, and space sta-
tions with astronauts on board. By Newton’s third law every action —
Let us derive the basic equation of has an equal and opposite reaction—
the rectilinear motion of a rocket. The the impulse of the force with which
propellant, whether gunpowder or a the mass dm of the combustion products
mixture of fuel (alcohol or gasoline) acts on the rocket vehicle is equal to
and oxidizer (oxygen or nitric acid), the same quantity with sign reversed.
possesses a definite supply Q of chemi- Suppose, for instance, that the exhaust
cal energy per unit mass (of the order velocity u 0 is in the direction of de-
of 5 X 10 6 J/kg for smokeless powder creasing x Then u 0 is negative, or
.

and 10 x 10 6 J/kg for a gasoline (or u0 = — u0 For the impulse of the

| |
.

petrol) and oxygen mixture 915 ). In force acting on the rocket we have
burning, this chemical energy is convert-
dIT =F dt = — u
r 0 dm = |
u0 |
dm.
9 15
*
The heating value
of gasoline is about (9.11.2)
50 X 10 6 J/kg, but burning 1 kg of gasoline
requires 3.4 kg of oxygen. In a rocket launched 9 16
If in (9.11.1) Q is expressed in J/kg,
*

into airless space, the oxygen has to be carried then u 0 will be expressed in m/s.
along and the energy must be referred to the 9 17
The designation dl is due to the
*

oum of the masses of fuel and oxidizer. fact that we consider a small mass dm.
 •

9.11 Jet Propulsion and Tsiolkovsky’s Formula 335*

The quantity The equation of motion, or Newton’^

second law, is
K I < 9 -“- 3>
T.du
M nr =F =\ u «\iu' •
i
dm,

is the impulse per unit mass, what is

called the unit impulse This quantity .
which can be rewritten as M du
is equal to the exhaust speed of gases |
u0 |
dm ,
or, using (9.11.4),
from a rocket at rest.
Let us check the dimensions in for- (M0 -m)^=\u0 \. (9.11.5)-

mula (9.11.3). The force F has the di-

mensions of kg* m/s 2 or N, and the The possibility of cancelling out dt
,

impulse I is the product of force by has the physical meaning that, in the*
absence of other forces acting on the
time, so its dimensions are N*s (or
rocket, the velocity of the rocket de-
kg* m/s). The dimensions of dlldm are
— pends only on the amount of exhausted
(kg*m/s)/kg mis which are the di- ,

mensions of velocity. For powder powder gases (for a fixed value of u 0 ).

gases, u0 =
2 X 10 3 m/s 2 km/s, = By the time a given amount m of pow-
der gases is exhausted through the noz-
while for liquid fuel u 0 3 km/s. = zle, the rocket has acquired a definite
The force acting on the rocket is, by
formula velocity u irrespective of the time dur-
,
(9.11.2),
ing which the given amount of powder
gases was released.
F r = u o\ It is easy to solve the differential
\
-JT •

equation (9.11.5). At t = 0, u = 0
It is proportional to the quantity of
and /Ti — O. We thus have
gases exhausted in unit time. m r

Now let us examine the derivation of

the formula for the velocity of the ro-
cket vehicle. If the rocket is itself in = — Kl In (M — m)\™* 0
motion with a velocity u then the ex-
haust velocity of the gases differs from
,

= — In (M — m + In M
|
iio | [ 0 x) 0]
u 0 and is equal to u u0 u u0 + = — | |

(recall that when the vehicle is at rest,

the exhaust velocity of gases is equal
to — u 0 ). It is obvious that such
| |
where m x is the total mass of powder ga-
quantities as the difference between the ses exhausted by a given time, and
velocity of powder prior to combus- M M —
— 0 mx is the mass of the rock-
tion and the velocity of the exhausted et at this moment in time. We have
powder gases and as the force with therefore arrived at Tsiolkovsky's for-
which the powder gases act on the rock- mula
et are independent of whether the
rocket vehicle is in motion or at rest u=|uo|ln-^L. (9.11.6)
(for the sake of definiteness we speak of
powder as the propellant). If we are interested in the terminal
Let us denote the initial mass of the velocity u ter at burnout, in formula
rocket together with the powder by (9.11.6) we must substitute M ter (ter-
M 0 and the mass of exhausted powder
minal mass of the rocket after all the
gases by m. The quantity is a func- m fuel has burnt out) for M, that is,
tion of time, or m=m
(t). The mass of M ter = 0 M —
m toU with m tot the
the rocket with powder at time t is total mass of the propellant. We get
equal to

M=M (t) = M -m0 (t). (9.11.4)

M ter= l«ol ln
l (9.11.7)
 r

336 9 Mechanics

This formula can easily be used to gases,u 0 To get an idea of the difficulty
.

.solve the inverse problem: what initial of the problem of launching a rocket,
mass 0 M
of the rocket vehicle must one should bear in mind that ter M
be taken so that to a given terminal includes the mass of the fuel tanks,
mass M
ter is imparted a definite vel- etc.
ocity W-ter • Let us find the efficiency of a rocket
M " as a whole. We
define this quantity as
1 r< o — ^ter
the ratio of the kinetic energy of the
Mter K( ’

rocket at burnout, ter ul er /2 to the M ,

whence chemical energy of the burnt fuel,

mQ = (M 0 —M Q. The efficiency
M =M 0 ter exp -jjfj- . (9.11.8) is
ter )

Mj eT uter
in our analysis we ig-
Everywhere n # (9.11.9)
2Q (M 0 -Mter)
1

nored the (transient) process of “slow”

combustion, when both the force of grav- Substituting the expression for u ter
ity and the reaction force act on the from (9.11.7) and expressing u\ from
body (rocket). Of course, in real life (9.11.1), we finally obtain
this situation is always present we — —a Mte r Mp \2
only hoped that by separating these n (in
Mq Mter Mter )
two forces we do not introduce a large
error into the results; the forces of grav- The efficiency proves to be the product
ity were studied in Section 9.10, of the “internal efficiency” a (which
while jet propulsion has been studied characterizes the completeness of burn-
in the present section. 918 will now We ing of the fuel and the conversion of
try to unite the results of these two sec- thermal energy into the kinetic energy
tions by using formulas (9.11.6)- of the gases) and a second factor that

(9.11.8) estimate the values of

to depends solely on the choice of the ra-
M /M 0 necessary for attaining the
ier
tio between the mass of propellant, m,
and the mass M We
satellite velocity u x and the escape vel- ter of the payload.

ocities u 2 and u 3 will again as- We denote ml Mter by 2 Then * M =M + 0 ter

sume that the propellant is gun powder,

.

m= M Xer (1 + 2 ), and
for which u0 2 km/s. Using for- = npa-gjffl-hn M ^+ m 2
=
| |

\
mula "(y'.lO), we get '
m \ Mter J

e
4 ~
54 for u x 8 km/s, 0 /M ier 2 = M =
e5
6 - ~
270 for u 2 11.2 km/s, and =
=-
-J-
[In (1 + z)] 2
. (9.11.10)

M /M 0 ter 3 = e 8 2 - ~ 3640 for u3 = At first glance it might appear that

16.4 km/s. In
the case of liquid pro-
due to the fraction 1 /z the efficiency is
pellant, u0 3 km/s, and similar = very great for small z’s. In reality, for
M
| |

calculations yield 0 lM terl

M ~
14.5,
M ~ small z’s we have In (1 z) z (see + ~
0 /M ter 2 42, and 0 /M ier 3 245. =
From the foregoing we see that the mag-
Section 4.9) and so r\ (aJz) z 2 ~
nitude of MjM
is strongly depen-
a z for Cl.
efficiency is pro-
z< The
te
portional to z and, hence, is small for
dent on the exhaust velocity of the
small z’s, with the result that the rocket
9 18
Note that such a “coarse” approach to
-
moves slowly and almost all the energy
a real process was undertaken in Section 9.10, iscarried away by the gases. For very
when we separated into two stages the laun- great z’s, the efficiency again falls be-
ching of an object into outer space: in the cause of diminished payload mass. 919
first stage we assumed the acceleration of grav-
ity g to be constant and equal to g 0 the ,
9 19
For large z’s, the quantity [In (1
.
z)]
2
+
acceleration of gravity at the earth’s surface, grows more slowly than z. Indeed, denoting
while in the second stage the object moved y =
In (1 z), we get z+ eV 1 and the = —
in a mode in which g diminished as the dis- function grows faster than any power y (see
tance to the earth’s center increased. Section 6.5).
 .

9.12 The Mass Center

, of Gravity , and Moment of Inertia of a Rod 337

9.12 The Mass, Center of Gravity,

and Moment of Inertia of a Rod
We consider a thin rod. The x axis will
lie along the rod. Denote by a the mass
per unit length of rod. Thus, on a por-
tion dx between x and x dx there +
will be a mass

dm — a dx . (9.12.1)

Since the terminal velocity of the rocket The quantity a (g/cm) is the product
is also dependent solely on z, we can of the volume density d (g/cm 3 ) of the
say that the efficiency of the rocket is rod’s material and the cross-sectional
determined by the requisite velocity. area S (cm 2 ) of the rod, or a Sd = .

At small velocities the efficiency of the The rod may have a cross-sectional area
rocket vehicle is low and so it is and a density that depend on x, so
disadvantageous to employ jet propul- that a is a function of x, or cr a (x). =
sion in automobiles and other cases of The quantity o' should be called the
relatively slow motion. At high veloc- linear density or the density per unit
ities, the energy efficiency of the rock- length but since the real density d
,

et again diminishes, but the use of (volume density) does not enter into
the rocket is nevertheless justified since the subsequent computations, we will
we do not possess any other means of call o, for short, the density consid- We
accelerating bodies to high velocities. er the thickness of the rod to be small
In conclusion let us find the value of and depict it merely as a straight line,
z =m/M ter which yields maximum a line segment of the x axis. The total
efficiency r\ and the magnitude of this mass of the rod is clearly
maximum. In view of (9.11.10), the b
problem reduces to determining the m -= o(x)dx (9.12.2)
maximum of the function F (z) z
-1
= j
a
,

[In (1 z)]
2
+
that is, to solving the
,

equation where a and b are the coordinates of the

ends of the rod.
rv /_\
F( Z )—
2? In (l + z)j] [In (l+z)l 2
0 Let the rod be fixed on the x axis,
(i+Z)z Z2
which is assumed to be a rigid rod so
or
thin and of so small a mass that it can
ln(l+*) = -ifr. (9.11.11) be considered a massless line; the x
axis is horizontal, while the y axis is
The simplest way of solving this equa- directed vertically upward. The force
tion is by graph. To do this, we deter- of gravity acts on the rod (as indicated
mine the point of intersection of the in Figure 9.12.1) tending to pull the
curves fx = In (1 + z) and / 2 (z) = rod downward.
2z/(l + z) (curves I\ and T 2 in Imagine that the x axis is a weight
Figure 9.11.1). This yields z~ 4, whence lever. Indicated schematically in the
F (z) 0.65 ~and rj max 0.65a /(it ^ drawing is a prism supporting the x
is easy to verify that z ~ 4 corresponds axis at the origin. The x axis can thus
to a maximum of F(z) and not to a
minimum). dm
a x x+dx b
Exercise * 0 7^77/77?^777\ ....
J
9.11.1. Prove that the value z ~ 4 cor-
responds to the maximum of the function
F (
z) = z
_1
[In (1 + z)]
2
. Figure 9.12.1

22-0946
 338 9 Mechanics

rotate about the axis perpendicular to

the plane of the drawing. Let us find
the load M
on the left at a distance R
along the x axis that is needed to
balance the rod on the right.
By the laws of a lever, the element of
mass dm distant x to the right of the
support 0 is balanced by element of
mass dM to the left if the masses are
inversely proportional to the distances ,

that is, if

,
or R dM = x dm. (9.12.3)

The element of mass dm is equal (as

we learned above) to a dx. To balance
the entire rod we need a mass that M
satisfies the equation anced by the rod (Figure 9.12.2)?
b We find that

RM = j xg (x) dx. (9.12.4)

a
RM = x c m— j
XGdx, (9.12.5)
a
This equation is the result of inte- whence
grating the left and right members of b b b
(9.12.3). To the right of the axis, differ-
ent elements of mass dm are located
xc = —j xo dx = xg dx a dx.
j j J
a a a
at different distances x from the sup-
port. This is why the quantity xg (x) (9.12.6)
appears under the integral sign. (In The quantity x c is the coordinate of
this way the integral determining the
the center of gravity or, as it is also
mass M
that balances the rod differs called, the center of mass of the rod, C.
from the integral determining the rod’s It is highly important that point C
mass m.) To the left of support 0, all is indeed a definite point of the rod: if
elements of mass dM
(which balance the we displace the rod as a whole along the
distinct elements dm of the rod) are
x axis, say, a distance l to the right
collected together at the same distance (Figure 9.12.3), x c will also increase by
R from the support. R is a constant and the same quantity Z, so that the point
so R dM = R dM = RM. C with coordinate x x c is always =
j j (for agiven rod) at a very definite dis-
The question now is: if we concentrate
tance from the endpoints of the rod.
the entire mass of the rod, m, at one
point C, then at what distance x c must
We will prove this.
Let us consider a rod displaced a dis-
this point be from the support (from
tance l to the right from the original
the origin, that is) in order to balance
the mass M
at distance R that is bal-
position (Figure 9.12.3&); the center of
gravity of the shifted rod will be de-
noted by C x Then .

/M 0 /77

Ak 37
R Xc xc =
a

dx
Figure 9.12.2 a (9.12.7)
9.12 The Mass Center
, of Gravity, and Moment of Inertia of a Rod 339

= -2 1
g
=zx c -\-l,
the field of gravity is equal to the po-
a dx a dx tential energy of its entire mass con-
^ ^
a a centrated at the center of gravity of the
which constitutes a result that is obvi- rod. We consider the position of the
ous from the start. rod as indicated in Figure 9.12.5. The
The most convenient thing is to choose potential energy du of an element of rod
the system of coordinates with ori- with mass dm is equal to gz dm where ,

gin at the center of gravity of the rod z is the altitude and g is the accelera-

(Figure 9.12.4). The quantities in this tion of gravity. The potential energy u
coordinate system will be denoted by of the whole rod is found by integrating.

a zero subscript. It is clear that For the variable of integration we choose

a length reckoned along the rod from
bo
its center of gravity; the density at
o0 (x) dx=-m. point x is denoted by {x). We express
^
a0 the altitude z in terms of x. As is evi-

The coordinate of the center of gravity dent from Figure 9.12.5, z {x) zc = +
x cos a, where z c is the height of the
x Co is zero in this system and so
center of gravity of the rod. We get
bo
bo bo
xo 0 (x)dx= 0. (9.12.8)
j
j
gz o 0 (x) dx =g j*
(z c -\-x cosa) o0 (x) dx
do
do &o

In other words, if we wish to balance bo bp

a rod supported at the center of gravi- = gz c a0 (x) dx + g cos a xo 0 (x) dx

ty by a load M
situated at a fixed dis- j
do
j
CLo

tance R from the support, we will find

that the mass required is zero (cf. = gx c nt 9
(9.12.4) and (9.12.8)): if the point of
support coincides with the center of
gravity of a rod, the rod is balanced since the second integral is equal to
without applying any additional load. zero by formula (9.12.8). Thus, the po-
We will now show that for any posi- tential energy depends only on the
tion of the rod its potential energy in mass of the rod and the height of the
rod’s center of gravity but does not de-
pend on the angle between the rod and
the horizontal: rotation of the rod about
its center of gravity requires no energy;
a rod fixed by a hinge at the center of
gravity is capable of freely rotating about
the support without release or expen-
diture of energy.
Now let us investigate the concept of
Figure 9.12.4 the moment of inertia of a rod. This con-
22*
 . ) -

840 9 Mechanics

Thus, E — 2
7co /2, or the kinetic energy
of rotation is expressed in terms of the
moment of inertia and the angular ve-
locity in exactly the same way that
the kinetic energy of simple translation
is expressed in terms of mass and linear
velocity, E mv 2 / 2. Note also that =
the moment of inertia (9.12.9) is al-
Figure 9.12.6
ways positive (as the mass m is).

cept comes up in a consideration of the If we

are dealing with a system of point
rotational motion of a rod. Let a rod masses ., m1? m 2
k lying at points .
± . m M
M M
,

be in rotation about an axis perpendi- 2, ., k in a plane, the moment of iner-

. .

tia of this system about a straight line l is

cular to the plane of the drawing and
passing through the origin. Then each 2m^ = m df-f-m d2+ 1 2 • • • (9.12.10)

point of the rod will describe a circle of the products of mass m* by the square of
centered at the origin and having a ra- the distance from point Mi to Z, or d\ (here
dius equal to the abscissa x of the given i =
1 2, ...,&). If we have a body (a plate
,

point in the initial (horizontal) posi- or a rod, as in our case) instead of a system of
point-like masses, we are dealing with a con-
tion of the rod (Figure 9.12.6). Denote
tinuous distribution of mass and the moment
by co the angular velocity of rotation of inertia is approximately equal to the mo-
expressed in radians per second. This ment of inertia of a system of point masses ob-
means that during time dt the rod ro- tained through partitioning the body into
The = many little parts and replacing each part by
tates through the angle dcp codt.
a point mass equal to the mass of the part.
arc length traversed by an arbitrarily It is clear that this definition, where it is as-
chosen point with abscissa x is dl = sumed that the number of these parts tends to
xdip =
x(x) dt\ hence the linear ve- infinity while the size of each part tends to
zero, leads to a formula for the moment of
locity of motion of a point on the rod
inertia in the form of an integral (cf. formu-
with abscissa x is v (x) = dlldt = (ox. la (9. 12.9)).° 20 A sum similar to (9.12.10)
Let us find the kinetic energy of rota- but with distances di between points Mi and
tion of the whole rod. An element of the straight line l substituted for the squares
of such distances, d\, is known as the static
mass dm distant x from the center of
rotation (i.e. from the origin; here we
moment of a system of masses about Z:

are speaking of a segment of the rod of ^PJ midi =m 1d 1 -\-m 2 d 2 + . .

(9.12.10a)
length dx with endpoints at x and Here, however, we must allow for the sign of
x +
dx) has kinetic energy di , depending on what side of Z the particu-
lar point Mi lies. It is clear, that the transi-
v2 , G) 2 # 2 tion from a system of point-like masses to a con-
o (x) dx.
IT 2 tinuous body leads in this case, too, to an in-
tegral replacing the sum; for instance, the sta-
Hence, the kinetic energy of the whole b

rod is tic moment j* xg (x) dx of a rod about point 0

b a
is present in formula (9.12.6) for the center of
x2o ( x ) dx. mass of a rod.
j
a
We nowtake up the evaluation of 7.
integral in this formula is called
The For a rod whose center of gravity lies
the moment of inertia of the rod about at the origin 0 (the center of rotation),
the axis passing through the origin and
is symbolized by /, or 9 - 20
In the case of aflat plate or a solid
we have to deal with a multiple integral (doub-
b
le or triple over the area of the plate or the
/= x2a(x)dx (9.12.9) volume of the solid. In this book we will not
j consider multiple integrals, however.
a
 9.12 The Mass Center , of Gravity , and Moment of Inertia of a Rod 341

the moment of inertia assumes the

value I 0 :

bo

I0 = ^
x 2 u 0 (x)dx. (9.12.11)

Let us determine the moment of iner-

tia of a rod for the case where the cen-
ter of gravity C 1 is distant l rightward
from the origin, so that x Cl —
l. In this case a = a0 + Z, b = We can picture a rod hinged at the
b0 + Z, cr (x) = (? 0 (
x — and ^ = center of gravity. Rotation of the axis
b
and the hinge need not be accompanied
x2a (x) dx. If we set z = x — l,
by rotation of the rod, that is, we can
a visualize a motion with successive
then x =z+ Z and dx = dz. When x stages as indicated in Figure 9.12.7. The
varies from a to 6, the quantity z
kinetic energy E' of such motion is
varies from a 0 to b0 Therefore,
.
mvcj 2 with u Ci the velocity of the
,

b0 bo center of gravity of the rod. But

I = (z -f- Z )
2 a0 (z) d:- = l
2 a0 (z) dz v Cx = co Z, so that E'
2 =
(co /2) ml
2
.

j j
The motion we considered earlier
&0 Go

bo bo (see Figure 9.12.6) differs from that of

Figure 9.12.7 in that in the former case
+ 21
j
a0 (z) z dz + j
a0 (z) z 2 dz.
the rod itself was in rotation with an
do &0
angular velocity co about its center of
(9.12.12) gravity. For this reason, the kinetic
b0 energy of rotation in Figure 9.12.6
Note that a0 (z) dz = m, while the proves to be equal to the sum of the
j energy of rotation of the type in Fig-
do
second integral on the right-hand side ure 9.12.7 and of the energy of rotation
by formula (9.12.8).
of (9.12.12) is zero about the center of gravity, which is
Finally, the third integral in (9.12.12) equal to 7 0 co 2 /2.
is/ 0 by (9.12.11). It is evident from the derivation of
Thus, formula (9.12.12) assumes the the formula that such a simple addition
form of energies in the combination of two
motions only results when we consider
r = ml 2 + /0 . (9.12.13)
the motion of the center of gravity;
only then do we find the integral
The quantity ml 2 clearly the moment is
(9.12.8) to be equal to zero.
of inertia of a point mass distant l m
from the axis of rotation (from the ori-
gin). Thus, the moment of inertia of a We can approach the concept of the center
of gravity of a rod from another angle. From
rod about an arbitrary axis perpendic- well known that the
elementary physics it is
ular to the rod is equal to the sum of resultant of two parallel (and pointing in the
the moment of inertia of the rod about same direction) forces / x and / 2 applied at
the parallel axis passing through the points M x and M 2 respectively,
,
is a force +
center of gravity plus the moment of f 2 applied at the point C that divides the seg-

inertia of a mass equal to the rod mass

ment MM 1 2 in the ratio M X C -i- CM — 2 f2
E (Figure 9.12.8a). If /x and / 2 are forces of
about an axis parallel to the first two gravity acting on masses m 1 and m 2 than point
,

axes and separated from this mass by C (such that X C M CM =2 / 2 H- h —m 2 -f-

mj) is called the center of gravity of the two
a distance equal to the distance of the
masses; if, in addition, we take the straight
center of gravity of the rod from the ar- line MMx 2 as the abscissa axis and assume
bitrary axis. the abscissas of points and M 2 to be x x
 .

342 9 Mechanics

with coordinates

2mx i i

XC = , yc= Y Z° = V m
m / m / m
’
n,
2j i i i j

i i i

(9.12.15)

(see Exercise 9.12.4b).

Now let us take a rod with a linear density
a = a (z). If we partition the rod by points
Xq = a, x ly x 2J .. M %n b into n small —
parts with masses m% —a (x{) A xi, where
A xi — Xi — xi_ x and i 2, ., 72, and = 1, . .

and x 2 then abscissa xq

,
of point C is (; mx xx + then consider n point-like masses mt corres-
m 2 x 2 )/(mi + m2 ). Indeed, then ponding to the 72 parts of the rod (each part is
assumed homogeneous since within each A xi
the density changes little) and concentrated
M C + CM
X 2 at the endpoints of each part, x xt, the re- —
sultant force of gravity for these point masses
m xx x — m 2x2 |— m x x x —|— m 2 x 2 will be applied to the center of gravity C x (xa),
Xl
m x + m2 mx — j— tt2
2 where xc x = (
x i) x i (%i) Ax*. Send-
772 2 (x x —x 2) 772 1 (X X —X 2)
m2 -
i

ing 72 to infinity and all the A xi to zero and

i

-
mi.
-j- 772 2
772 1 -f- 772 2 772 1
replacing the sums with integrals, we arrive
at formula (9.12.6) for the center of gravity
of the rod.
Similarly, if we have three masses m lf m 2 ,
and m 3 lying on a single straight line at points
M 1 (sq), M
2 (x 2 ), and s (x 3 ) (Figure 9.12.86), M Exercises
the resultant of the forces of gravity applied
to the first two masses is the force applied to a 9.12.1. Find the moment of inertia about
mass m 1 +m 2 lying at point C12 / mi x i+ m 2 x 2 \ _
the center of gravity of a rod of length l with
\ / a uniform distribution of mass.
The resultant of forces will be
all three 9.12.2. A rod is made up of two pieces:
applied to the center of gravity of the masses one piece of length l x has a constant density a lf
mx + m2 and m3 ,
that is, to the point C and the other one of length l 2 has a constant but
with the abscissa different density cr 2 Find the position of the.

center of gravity of the rod.

m ± xi-{-m 2x 2 9.12.3. Find the position of the center of
*C = [(rox + TOjj) m 3 x3 " j[(m l gravity and the magnitude of the moment of
m -\-m 2 1
inertia about the center of gravity of a rod
77i
i i a?m 2x 2 — m 3 x3 |— in the form of a thin triangle of length L.
+ TO2)+m 3]
ITl
j —j— 777-
2 —J— 772 3
Express these quantities in terms of the length
L and the mass m of the rod. [Hint. If the x
axis lies along a median and the origin is cho-
Quite similarly we can prove that the resul- sen at the respective vertex of the triangle,
tant of the forces of gravity of a system of then g (x) =
ax with a a constant.]
masses m x m 2 .,77i fe lying at points x (a:!),
. . M ,

9.12.4. Prove (a) formula (9.12.14) and

M 2 (x 2 ), .,
,

. .
,

k (x k ) is applied at the point M (b) formula (9.12.15).

( \
/ tn x x x -f- m 2x 2 -f- • + m k x h \ .

9.13* Centers of Gravity of a String

\ 772! + m 2 + •+»h / .
|
and of a Plate
^ /
(9.12.14) In Section 7.11 we encountered the no-
tions of the center of gravity of a flat
(see Exercise 9.12.4a).
Reasoning along the same lines, we can plate and of a string that was bent in an
establish that if k masses m ± m2 . . ., mk arbitrary manner. Suppose AB, with
lie in space at points y x z x ), 2 (x 2 M x (x x
, ,

M , A — A (%, b x ) and B B (a 2 b 2 ) is a —
M
, , ,
,

y 2 z 2 ),
, ., k (x h y h zk ), the resultant of
. .
, ,
string of arbitrary shape of linear den-
all the forces of gravity is a force applied to
... -f- m k at the center sity a ( x y) (Figure 9.13.1). In other
the mass m x m2 + + ,

words, we assume that a small sec-

of gravity of the system, a point C ( xc yc, %c) ,
 ( (

9.13* Centers of Gravity of a String and of a Plate 343

B
here stands for integration along the
j
A
curve AB. If x = x (t) and y = y (t)
are parametric equations of AB,
the
and t 1 and t 2 correspond to the endpoints
A and B, then along this string we have
a =
a (x(t), y(t)) o(t), the total mass =
of the string is

tion of the string of length ds = m= j

a (t) Y (x’) 2 + (y 1

)
2
dt,

V [x'Y + (y')
2
dt, with x = x (t) and ti

y— y the parametric equations of

(t)
and the coordinates of the center of
AB, has a mass dm a ( x y)ds, where = , gravity are
(x, y) is a point within the considered
section, say, its end. If we replace the
continuous distribution of mass, or xc = -^ x (
t )° (t) V x
'

)
2
+ 2
y ') dt >
\
our string, with a system of n point- u
like masses dm g (x y t ) ds (i t
= t , t
= t,

vc = 4
2 2
1, 2, . .each corresponding to
., ft),
" y (t)°(t)V( x ') +(y') dt-
5
the section ds of the string with e nd-
t ti
points ^4 (xi and Ai(xt y ), , t (9.13.3a)
where A 0 A A l9 A 2 =An B , ,
. . =
are the points that partition the string If the form of the string AB is given
into ft small parts, we conclude that explicitly, 9 - 21
y = y with
(x), a^. x^,
the center of gravity C1 of these point- b, then o = o (x, y (x)) = a (x), the

like masses has the following coordi- mass is

nates: b

X[G fait y i) dsi

m= j
a (x)V 1 + (y') z dx ,

X Ci = -sn *
a

Zj o fa t , y t ) dsi and
i
b

=-±-
2 yi& fait yi) dsi
xc = — xa (x) V + (y')
l 2 dx ,
y Cl (9.13.1) j
a
Zj G fail yfj dsi
i b

(cf. and (9.12.15)). Going over

(9.12.14) yc = 4" J
y (*) a (*) V 1 + {y') 2 dx -

to the limit n oo, ds t 0, we arrive

at the following expressions for the (9.13.3b)
coordinates x c and y c of the center of
gravity C of the string: In the particular case of a uniform
B string of constant density, o con- =
§
xa fa , y) ds
stant, we find that aS, where m=
XC — ~A '»
S —
b

ds is the length of the string,

j
J
G fa, y) ds
A whereby formulas (9.13.2), (9.13.3a),
B
yG fa, y) ds 9 21
Here we assume that each value of x
-
l
corresponds to a single point of the string; if
; (9.13.2)
this is not so, we can simply break up the
Gfa , y) ds string into several sections in each of which
$
A the above condition is satisfied.
 344 9 Mechanics

and (9.13.3b) can be rewritten thus: We denote by p =

p ( x y) the density ,

B B of the plateper unit surface area; in

xc = xds, y c = -j- ^ yds, (9.13.4) other words,we will assume that the
-J- j mass of a small rectangle with sides
A A
dx and dy with dx and dy very small
,

or Figure 9.13.2; the sides of the rect-

(see
t*
angle are taken parallel to the coor-
xc = -j-
j
x (t) ywv+W )
2 dt > dinate axes), is equal to p x 0 y 0 ) dxdy,
where dxdy the area of the rectangle, is
(, ,

u obviously, and p (x 0 y 0 ) is the density ,

at a point ( x 0 y 0 ) of the rectangle, say,

yc= -y \
y (t)V (x ')
2
+ (y') 2
dt,
at one of its vertices (since the rectangle
,

tl
is small, the density changes but little
(9.13.5a)
within its limits). Next we “cut” the plate
and into n vertical strips by the straight
b lines = x = a, x = x x = x
x 0 x, 2, . . .

xc = -J- xY 14- (y')

2
dx, . . = x n = 6, the width of each
., x
j
a strip being A^ = x — Xi^(i =1,2, t
. . .

b
. . here a and b are the smallest
., ft);

= -j- and greatest of the abscissas within

yc
j
y(x)V i + (y') 2 dx. plate D
(Figure 9.13.2). may also We
assume that the ith strip (or the £th rod)
(9.13.5b) has a linear density that depends on
Note that all formulas for the centers of y*
gravity contain the static moments of Qi {y) = P (Xi, y) A Xi, (9.13.6)
the string about axes Ox and Oy (see and a mass
p. 340).
f(x )
The situation is somewhat more com- t

plicated when we deal with a plate Am* = \ [p (x t , y) Ax dy

t]

D (Figure 9.13.2; the thickness of the

plate can be assumed so small that it
i{Xi)
can be ignored). Actually, the problem _
of finding the center of gravity of a = p(x t ,
y)dy^Ax t , (9.13.7)
[ (
plate is reduced to evaluating certain g(X-)

double integrals over the area of the — =

where y g (x ) and y f {x ) are t t
plate, which integrals are not discussed
the smallest and greatest values of the
(or even formulated) in this book. How-
ordinates y of the points of the ith rod;
ever, there is a certain method that
precisely, (x t g {x t )) and (x t f (**)) are, ,

makes it possible to overcome this diffi- points of intersection of the straight

culty.
line x =
x t with the boundary 9 22 of *

plate Z>. (Note that the integrand in the

term on the right-hand side of (9.13.7)
is a function of a single variable y,
since the value of X\ is fixed.) This mass
m
A t distributed “over” the rod can be
replaced with a point-like mass A t m
concentrated at the center of gravity

9.22 p or sa h e 0 f simplicity we assume

that plate D has the “oval” shape as depicted
in Figure 9.13.2, that is, is restricted by two
simple curves y g (x) and y / (x) taken = =
Figure 9.13.2 within the limits x a and x b. = =
 m .

9.13 * Centers of Gravity of a String and of a Plate 345

of the rod at point Ci(xa, ya ) with Ax t

possesses a mass
coordinates
/(**)

Anti = dy Ax t = [f (x — g (z*)] t)
Ax it
x Ci =Xi, ^
g(.x.)

f(x.) (9.13.9)
/(*P

yc i = yp( x n y) dy p (x u y)dy
i / j which coincides with the area of the
g(0C.) g(x
t
)
rod (the product of its height / {x ) t
—
g (Xj) by the width Ax ), and the t
(9.13.8)
center of gravity C lies, naturally, in t

the midpoint:
9 - 23
(cf.(9.13.6) and (9.12.6)); note that
the factor Ax t in the density (9.13.6)
xc ~x i » yc.=
/ (Xj) + g(Xj)
cancels out in the fraction in the expres- i 2
sion for the ordinate y Ci of the center (9.13.10)
of gravity C t
.

Thus, we have replaced the force of The sum of all masses (9.13.9) is
gravity acting on the plate with n
forces acting on the masses A m x
A (9.13.7)) at points Amn (cf.
,
m=2 =2 [f (z,-) —g (
x t )] A xi
2 ,
. . ., i i

C2 C n (see (9.13.8)). What re-

,
. .

mains to be done is to find the resultant Going over to the limit n -» oo

of these forcesapplied to the center of A x -> 0, we get
t

gravity of the system of rods considered

6
here and then (as usual) to pass from
sums to integrals via the limiting pro- M= j
[f{x)—g(x)\ dx,
cess n oo, Ax t 0 (for all i’s from a
1 to ri).

The derivation of the corresponding which is obviously the area S of plate

formulas poses no difficulties (see Exer- D Section 7.5). The coordinates of

(cf.

cise 9.13.2), but still it lies outside the the center of gravity C' of the strips
scope of this book, since it involves (rods) with the masses (9.13.9) concen-
evaluating iterated integrals, that is, trated at points (9.3.10) are
integrals that contain functions express-
ed in terms of other integrals. For this Xc =S x i If (
x t) —g (Xi)) A xjs,
reason we will not clutter our exposi- i

tion with such formulas. Instead, we con-

sider the case of a homogeneous plate yC - =2 [/ (Xi)-g (Xi ) ] A x /s(

with constant density p constant, = %

which of course can be set at unity:

p (x, y) 1. =
Then a rod of width (cf. (9.12.14) and (9.12.15)), or

xC ’ =2 x t [f (z,) — g (z ; )] A xJS,
i
9 23
Formula (9.12.6) refers to the case of
•

a horizontal rod, but already from the definition

of the center of gravity of a rod as a point C
Vc=\ 2 {[/ (^f)J — 2 2
(^i)] } Aorf/^.
such that when the rod is fixed at this point,
it is in equilibrium (see p. 339), follows the
equal status of the x and y axes (one is trans- The same
limiting process ra oo T
formed into the other if we rotate the rod
through 90° about point C whereby the second
Ax 0 leads to the following (exact)
) ,

formula in (9.13.8) follows directly from formulas for the coordinates x c and
(9.12.6). y c of the center of gravity C of the ho-
 g t )

346 9 Mechanics

mogeneous plate D with surface area It has been established in experi-

S: ments that if a body is moving in a
b liquid or a gas and the speed is low and
x M( x)— the body is small, the force of resistance
*c= 4" j ( x)] dx »
is proportional to the speed :
a
b F (
t
)
= —kv (t). (9.14.1)
Vc = ls~ {[/W]
j
2
— g{x)?}dx,
Here the coefficient of proportionality,
a
k is positive, and the minus sign shows
,

(9.13.11) that the force of resistance is in oppo-

sition to the velocity of the body. The
which we will now write out in full for
number k depends on the properties of
the special case of a curvilinear trape-
the medium and is proportional to the
zoid D =
ABCD which is formed by
,
viscosity of the medium. 9 24 In addition,
=
-

the x axis, the straight lines x a

= k is dependent on the shape and dimen-
and x 6, and the curve y / (x), = sions of the body. For example, if the
that is, for the case with g (x) 0: = body is a ball of radius R formula ,

b form of the
(9.14.1) assumes the
XC — y j xf (
X) dx, Stokes law: 9 2b
a
F =— (£), (9.14.2)
b
where r) is the viscosity of the medium.
*/c = ^s\[1{x)?dx, (9.13.12)
For air, rj = 1.8 X 10~ 4 g/cm*s, while
a
for water at 20 °C, rj — 0.01 g/cm-s.
where, as we already know, S — We consider the problem of decelera-
tion of a body. Suppose that some force
b
has imparted a velocity to a body
j
/ (x) dx. and at time t = t 0 has ceased to act.
The body continues to move and is act-
ed upon by the force of resistance
Exercises alone.

9.13.1. Find the center of gravity of a ho- 9 24

mogeneous plate that is (a) a triangle, (b) a
Viscosity rj can be defined as follows.
’

trapezoid, and (c) a half-disk. Let a liquid (or gas) be in motion along the x
9.13.2. How can we find the coordinates axis, and the velocities of the various parti-
cles be distinct and dependent on the y coor-
xc and yc of the center of gravity C of a plate
D = dinate. It is clear that a solid could not move
of density p p (x, y) limited by the curves
in that fashion and would be destroyed. In
y —g ( x ) and y =
/ (x) (with a x 6)? c c a liquid or a gas, in this case there arises,
between adjacent layers, a force of friction
9.14 of a Body in a
The Motion proportional to the difference in the velocities
of the adjacent layers, that is, to the deriva-
Medium that Resists this Motion with
tive dvldy. The proportionality constant in the
a Force Dependent Solely on expression for force / per square centimeter
the Velocity of horizontal surface area is called the viscosity
f =q ( dvldy (here the force / is expressed in
When motion, every body experi-
in N/cm 2 rather than in N, which implies that
ences a counteraction from the medium the dimensions of viscosity are g/cm*s).
9 25
Sir George Gabriel Stokes (1819-1903)
-

in which the motion is taking place.

was a British mathematician and physicist.
If the resistance is slight, say, in the Formula (9.14.2) is valid for vRp/r\ 5, <
motion in air with a low velocity, it where p is the density of the medium. The
can often be neglected. However, in reader can easily see that the quantity vRp/r\
is dimensionless It is called the Reynolds
some cases this approach is not satisfac- .

number (abbreviated: iV Re ) after the British

tory and the resistance has to be ta- physicist and engineer Osborn Reynolds
ken into account. (1842-1912).
 ) t 1 a

9.14 The Motion of a Body in a Medium that Resists this Motion 347

By Newton’s second law, the x axis downward toward the ground,

m ( du/dt ) = —kv. and place the origin at a height above
the ground (i.e. x on the ground). =H
Dividing both sides by m and setting Let the motion begin at t 0 =
klm = a, (a > 0), we get with a velocity u 0 Then x (0) . =0
and v (0) =
v 0 The body is acted upon
.

< 9 - 14 - 3> by two forces: the force of gravity (as-

sists the downward motion) and the
The solution to this equation, as we al- force of air resistance (inhibits the mo-
ready know (e.g. see Chapters 4 and 8) tion).
is By Newton’s second law,
v =v exp (—a —
m ~jj = mg — kv.
(it
0 (t t 0 )),
(9.14.7)
(9.14.4)

where v 0 ( =
v ( t 0 )) is the value of the Dividing all terms of this equation by
velocity at time t t0 Since a is pos- = . m, we get dvldt = g av (since —
itive, it follows that for t t0 the > klm = a), or
exponent in (9.14.4)
exp (— a ( t 0 )) 1,
is
—
and,
negative,
hence, < ir
=a (4 — ")
• <
9 -' 4 8 )
-

v (t) y0 <
that is, the velocity falls
,

We establish the dimensions of g/a.

off with the passage of time the medi- — Since a =
klm and k Flm, it = —
um retards the motion of the body, as
expected. follows that a has the dimensions of
Let us find an expression for the dis- s" 1 The dimensions of gl
. are then
tance traveled by the body. From cm- s/s 2 =
cm/s, which means that
{9.14.4) it follows that gl a has the dimensions of velocity. 926
Set g!a vv —
Equation (9.14.8)
dx
— vo exp — a(t — £0)) or ( ,
takes the form

dx — y exp — a (t — )) dt.
0 (9.14.5)
( 0
|jr =a(v t
— v). (9.14.9)

Suppose that at = t (initial time) t 0 Suppose that v 0 is less than v v Then

the body is at the origin: x (t 0 ) = 0. the right member of (9.14.9) is positive
Integrating (9.14.5) we find that at the start of the motion and, hence,
t the left member is positive too, dvldt >
x t
( )
=y 0 e~ a ^- t
^dt i
0; therefore, the velocity v ( t ) in-
j creases. And the closer v is to i? 1? the clos-
t°
er dvldt is to zero and, consequently,
that is,
the slower v grows. If at some time t x

x(t)=-%-[ 1 — *-«<*-*.>]. it is true that v (i x ) v x after this v = ,

(9.14.6) remains constant, since v is the =
solution to Eq. (9.14.9) with initial
Using formula we can
the entire distance covered by the body
(9.14.6), find
condition v (£ x ) vx =
Similarly, if at .

after time t 0 which is after the force

the start of motion v 0 v u velocity >
,
v approaches v x too, but in this case
ceases to act, up to the moment when ,

v decreases. For this reason, a cer-

the body stops. To this end, we note
tain time after the start of motion the
that for very large values of t, the quan-
body begins to fall at practically a con-
tity exp ( a (t £ 0 )) —
is extremely —
small (exp ( a (t —
1 0 ))<C 1) and can — 9.26 The calculation performed here of the
be neglected. Therefore, the body trav- dimensions of g/a is a check. The dimensions of
els a total distance of vja. g/a are evident from formula (9.14.8). Since
only quantities having the same dimensions
Let us now examine the fall of a body can be subtracted, it follows that g/a must
in air (allowing for air drag). We send have the dimensions of velocity, v.
 ) )

348 9 Mechanics

resistance is proportional to the square

of the velocity It has been experimen-.

tally established that in this case

F=-kSp-^-, (9.14.13)

where S is the cross-sectional area of

the body, and p is the density of the
Figure 9.14.1 medium. 9 27 We
see that the force of
resistance in this case is practically in-
stant velocity of iq g (a), irrespec- = dependent of the viscosity of the medi-
tive of the velocity it had at the start of um. The coefficient k in this formula is
fall. The graph of velocity for v 0 0 = a dimensionless number, and its magni-
is shown in Figure 9.14.1, where v = tude depends on the shape of the body
g/a is the horizontal asymptote to (for streamlined bodies, k can drop to
this graph. 0.03-0.05, while for bodies with poor
The foregoing examination shows that streamline characteristics, k can reach
a number of properties of v (t) can be 1.0-1. 5). Denoting /c£p/2 by x, we ob-
detected even without solving Eq. tain
(9.14.9). Now let us solve this equation. F = — xz; 2
(9.14.14)
pu t v — =
z. Then dz/dt —dv/dt, = (f).

and Eq. (9.14.9) can be rewritten as It is clear that x has the dimensions of
dzldt — —
az. At t 0 it must be true — g/cm.
that z iq = —
v 0 The desired solution .
Let us solve the problem of decelera-
~ at
is z (
t)= (v — t
u„) e Passing to the tion for the force of resistance (9.14.14).
function v (
t ), we get The appropriate equation is of the form
iq — v (f) = (iq — u e~ at
0) ,
or m (
dv/dt ) = — xv 2 .

v (
t) = iq -f-
(v 0 — iq) e (9.14.10)
Dividing both sides by m and setting
Considering this equation, it is easy xhn = p, with P 0, we get dvldt > =
to see that we can draw the same conclu- "Pz;
2
,
whence dv/v 2 = -P dt .

sions as were evident from our rough tegrating, we get = -P*

analysis of Eq. (9.14.9). If v 0 iq > where v 0 is the velocity of the body at
l
vu
\Vo I

then v t iq, since (u 0

(: >1 — v )~ e>0; time
at
= t t Therefore, v' 1 -f Vq 1 — =
but if v 0 iq, then (u 0 <
1 )e~
at
—v < 0 —P — (t
0.
t 0 ), whence
and, therefore, v (t) iq. < Second-
no matter what u 0 is, the quantity v = vo
(9.14.15)
ly,
at 1 + P*4)0 — ^o)
e~ is small for sufficiently large t,
and, practically speaking, v (
t = iq. From the formula p xhn we find
-1
=
Using (9.4.10), we find that that P has the dimensions of cm .

| Note that for t — t0 ^> l/p^ 0 the ve-

»

--^v^ivo-vje-at (9.14.11)

9 *
27 is valid for Reynolds
This formula
(th^ reader will recall that v == dx/dt),
numbers Rvp/r\ >* 100. The meaning of the
whence, knowing that x (0) 0, we = formula given in the text is that for the mo-
have tion of a large body, the energy expended on
overcoming the resistance of the medium is
X (t) = Vit+ — not spent on the friction of layers of the
liquid but on increasing the kinetic energy of
(9.14.12) the liquid, which is forced to move in order to
let the body pass through it. The reader is
the body has a large velocity and
If advised to derive the formula for the force
considerable dimensions, the force of (cf. Section 9.15).
 )

$.14 The Motion of a Body in a Medium that Resists this Motion 349

locity practically coincides with the (g/P)

1/2 = v x Then g ! P = yj, and Eq.
.

initial velocity: v ~ ~
Q/
p(t
1
— ,
t
.
.)
)
(9.14.18) assumes the form
0)
Let us find the formula for the dis- -§- = PK-y2). (9.14.19)
tance. Using (9.14.15), we get
An
exact solution to Eq. (9.14.19)
dx = dt is given in the answers to Exercise
1 + (t— to)
,

9.14.1. Let us consider the general prop-

whence erties the solution. Reasoning in a
of
manner similar to the way we did for
Eq. (9.14.9), we can show that in this
case a velocity of (g/p)V2 should =
to
set in. We
will show that after a long
(9.14.16) enough time lapse following the start
of fall, the formula
Assuming that the body begins to move
from the origin, x (t 0 ) = 0, we get, v — vx = C exp (— 2(k>1 £), (9.14.20)
from (9.14.16),
with C a constant, will hold true. Equa-
tion (9.14.19) can be rewritten thus:
j:(0 = yln[l + pi; 0 (*-*a)]. (9.14.17)

easy to see that this formula cor-

It is
±.^-^(v + v)(v -v).
l 1
(9.14.21)

responds to the velocity being an ex- For t large, v ~ v

x and
so in this equa-,

ponential function of the distance trav- tion we replace 4- v by 2v x But if .

x
eled: v =
v Q e~^ If we now wish to
. we replace v by vx in v x v, we get —
find the entire distance traveled by the dv/dt =
0, which yields v constant = =
body after the force that imparted the vx .Since we are interested precisely
velocity ceases to act, we will see that in the small difference between v and
this distance (formula (9.14.17)) in- u 1 (the law by which v approaches u 1 ) 1

creases without limit with time. (Note it is not permissible to ignore the differ-
that by formula (9.14.17) x oo ence v —
u x Thus, we replace (9.14.21)
.

as t-* oo.) with

Actually, this is not so. The point is
that when the velocity of the body is ~=2^ Vi ( Vi -v). (9.14.22)
small, the relation (9.14.14) no longer
holds true. We have to resort to (9.14.1) Put v1 —v= z. Then dz/dt —
and, for the distance, to formula —dv/dt, and Eq. (9.14.22) assumes
<9.14.6). the form
Let us consider the problem of a body
falling in air for the case of air drag
-~= — 2(Ji; 1 z. (9.14.23)
being proportional to the square of
The solution to this equation is
the velocity. At start, let the body fall
from the origin with an initial velocity z = C exp (— 2(3i;1 £), (9.14.24)
v 0 In analogous fashion to the case
.
which coincides with (9.14.20).
where the resistance is proportional to
The value of C in (9.14.24) cannot be
the velocity, we get the equation
determined from the initial condition
v (0) = v 0 (or z (0) = v 1 y 0 ) because —
^r=s-0» 2 -
or = P • Eq. (9.14.22) holds true only for suffi-
-df
ciently large fs (near t = 0 we cannot
(9.14.18) replace v +
v1 by 2v 1 ).
Also observe that the formula
It is easy to establish that (g/p) 1/2 has F ( t) =
—kv2 ( t is valid only when
the dimensions of velocity. Set v > 0. Indeed, if v 0, then it must <
 “

350 9 Mechanics

be true that F (t) = ku 2 (t), since the dium (g/cm 3 ), and k a dimensionless
force of resistance in opposition to is quantity (cf. (9.14.14)). If we do not
the velocity and, hence, is positive if restrict our discussion to slow motion,
the velocity is negative. Both cases the coefficient k is a function of two di-
(u >
0 and v <
0) are embraced by the mensionless quantities, namely, the
formula Reynolds number Re (see footnote A
F (*)
= —m (t) v ( t) |.
9.25) and the Mach 9 28 number Ma A ,
|

which is the ratio of the velocity u

of a body with respect to the surrounding
Exercises fluid to the speed of sound c in the me-
9.14.1. Find the expression for the velocity dium, or Am a = v!c. Hence
as a function of time from the equation dv!dt = = k(N Re ,N M
P (y? — v2 ) with the initial condition v (0)
= k a ). (9.15.2)
v0 . Using the formula for v, show that a veloc-
ity of vx =
(g/p) / sets in. Show that C in
1 2 In the particular case of a body moving
(9.14.24) (or (9.14.20)) is equal to 2v1 (v0 — with a velocity that is low compared
»i)/( v i +
v0 ).
9.14.2. In the problem of falling bodies
to the speed of sound, or ^Ma 1, we <
(the resistance is proportional to the velocity) have
have regard for the fact that the body is acted
upon by an expulsive force in accord with the k (A Re , Am a = 0) = k (A Re ). (9.15.3)
Archimedean law.
9.14.3. Applying the result of the preced- The behavior of this function depends
ing exercise to a ball and noting that for a on the shape of the body, on the orien-
sphere k =
6jii?r), where R is the ball’s radius, tation of the body with respect to the
and the viscosity of the medium, demon-
r] is
direction of the velocity, and, finally,
strate that for large values of t the velocity of
the ball, y, is equal to 2 R 2 g (p p')/9q, where — on what quantity is chosen as the char-
p is the density of the material of the ball, and acteristic size in the definition of the
p' is the density of the medium.
Reynolds number. Only in the simplest
case of a ball do all these three factors
9.15* The Motion of a Body in a Fluid play no role.

Let us now discuss in somewhat greater

For low Reynolds numbers, Re A < 1>
the asymptotic behavior of (9.15.3) is
detail the physical nature and the re-
constant
gularities that characterize the forces k (9.15.4)
that act on a body moving in a con-
N Re
tinuos medium, say, in air or in water. where we have left out the terms that
In view of the great importance of this are small compared to the leading term.
question to technology, especially to Substituting into (9.15.1) the value of
aviation, it has been studied in great k obtained through disregarding such
detail and constitutes the subject of a terms in (9.15.4) and allowing for the
special branch of science, fluid dynamics .
fact that Re A —
vRp/r) (by definition),
The particular cases discussed in Sec- we get
tion 9.14 correspond to extreme simpli-
constant X pv 2
Fr ~
r\
fication of the true laws of motion of 2 Sc
fluids.
^p
An assumption that is more exact
— constant x r] Rv, (9.15.5)

than the one formulated in Section where the constants in the middle and
9.14 is that the force of resistance of right members of (9.15.5) are, of course,
the medium on a given body can be different .

written in the following form: We have again arrived at the Stokes

law Eq. (9.14.2)), but what im-
F= _ kS
^ ,
(9.15.1)
(see is

where S is the cross-sectional area of 9. 28 Ernst Mach (1838-1916), an Austrian

the body (cm 2 ), p is density of the me- physicist and philosopher.
9.15 * The Motion of a Body in a Fluid 351

portant here is that in this approxima- ture of the motion of a liquid around a
tion the force of resistance is indepen- body and the origin of the forces acting
dent of the density of the fluid. The last on the body. Here it is more convenient
transformation in (9.15.5) is based on to transform to a system of coordinates
the fact that the cross-sectional area S in which the body is fixed, that is,
of the body is proportional to the square picture a fixed body in a flowing liquid.
of the linear dimensions of the body, Prior to contact with the body, that
2
i? (here we consider geometrically sim- is, far from the body, in the direction
ilar bodies), and it is for this reason opposite to that of the velocity of the
that only the first power of R remains liquid, the entire liquid moves with
in the expression for the force. One must the same velocity (both in direction
bear in mind, however, that for and magnitude). Near the body the flow
any finite value of 7VRe formula changes its behavior. Clearly, the com-
(9.15.4) is only approximate— it con- ponent of the velocity perpendicular
tains correction terms that are infinite- to the surface of body is zero— no liq-
simal compared to the leading term uid flows in, or out of, the solid, that
only for very low values of iVRe . is, no liquid crosses the solid boundary

In the other limiting case, iVRa of the body. But if the liquid has viscos-
oo, the situation proves to be more ity, it can be stated that the other

complicated. Here k can be assumed component, the one tangent to the sur-
constant, as was done earlier; although face of the body, is zero too — in this
the assumption that k is constant is ra- case the velocity of the liquid at the
ther crude, but in a book for beginners surface of the body is zero.
it is quite appropriate, we believe. In Two forces act at the surface of the
many textbooks on mechanics and exte- body: the force of pressure normal to
rior ballistics, the equations of motion the surface and the viscous force tan-
are integrable explicitly on the assump- gent to the surface and directed in the
tion that k is constant, and F con- = same direction as the velocity of the liq-
stant X v 2 But the aerodynamicist
.
uid at a small distance from the sur-
will say that k depends, even if weakly, face. The viscous force is equal to
on iV Re for large values of the Reynolds T] (du t Idn s
,)
where v t is the tangential
value. For instance, for a ball at component of the velocity, and d/dn
Nne =100, k ~
1.2, while at iV R e = stands tor the derivative along the
2 X 10 5 k,
0.4. ~
Then, in a rela- outward normal n to the surface of the
tively narrow region of Reynolds num- body; at the surface v t = 0, but the
bers from 2 X 10 5 to 5 X 10 5 the force ,
derivative dvjdn differs from zero.
of resistance diminishes by a factor of If it is known how the liquid moves at

approximately three, after which, up each point near the surface of a body,
to N-R e = 10 the calculation of the pressure and
8
it may be assumed that
,

k = 0.12. On the whole, we see that the viscous force poses no difficulty.
when iVRe changes by a factor of a mil- However, the calculation of the motion
lion (10 8 -f- 10 2 == 10 6 ), k changes by a is extremely difficult, and in general

factor of 10 (from 1.2 to 0.12), that is, form is inaccessible at pressent— the
over a broad range of values of A^ Re calculation has been carried out only
the following interpolation holds true: for a few simple cases.
/Q In slow motion, the viscous forces
kCCNnl In elementary (or “crude”)
-

play the principal role. The velocity

calculations we can always substitute
distribution in various fluids with
a constant for the extremely slowly va-
/6 various viscosities will be similar when
rying function iVR(J (provided that
the fluids flow around geometrically
the velocity varies within an interval similar bodies of different dimensions;
that is not very broad). for this reason the derivative du t Idn at a
Let us now discuss the physical pic- given point on the surface of the body
352 9 Mechanics

is proportional to the velocity v divid- tant of the forces of pressure is identi-

ed by the size R of the body, which cally zero In reality this situation ne-
.

means that the viscous force per unit ver occurs since even the smallest vis-
surface area the order of F t
is of = cosity modifies the flow considerably.
r\v/R. The force is F
total SFi = ,
A new phenomena emerges: although
where the surface area S for geometri- the oncoming flow is time independent,
cally similar bodies is proportional to near the surface of the body and be-
the square of the characteristic linear hind it the motion of the liquid changes
-dimension: S =
constant X R 2 Thus, . in a random manner (the picture
of the motion is like that of a flame of
F= constant X (9.15.6)
a bonfire). The phenomenon is known as
It can be demonstrated that, in the turbulence , or turbulent flow .

case of slow motion, allowing for the The rise in pressure at the leading
pressure distribution over the surface side of the body (on which the stream
changes only the constant. The Stokes hits the body) is not compensated for
law (9.14.2) for a ball of radius R, by the pressure behind the body. The
where the constant is equal to 6jt (here modification of the flow brought on by
the characteristic size of the body is the viscosity of the liquid proves to be
the ball’s radius), is a particular case significant and results in a nonzero
of the general formula (9.15.6). The force of resistance, or drag. As noted
product 6n in (9.14.2) was obtained earlier, the numerical value of this re-
through calculations and agrees well sistance depends very little on the
with experimental data. However, for Reynolds number, that is, depends but
an arbitrary nonsymmetric body the weakly on the viscosity.
calculation becomes very complicated In the system of coordinates where
even in the limiting case ARe -»* 0. the body is in motion and the liquid is
Even more complicated is the case initially, that is, in the absence of the
where a large body is in the stream of a body, at rest, the cylinder of liquid
liquid with large Reynolds numbers. through which the body travels during
The natural way to approach this prob- time t (the mass of this cylinder is
lem is to allow for the inertia of the M 1= p Svt) will part to let the body
liquid’s motion; a change in direction through and will acquire a velocity of
and magnitude of the velocity depends the order of v. Hence, the energy of this
then on the nonuniformity in the pres- cylinder is E=
2
xv !2 M =
tpSv*!2.
sure distribution. This energy has to be taken away from
When the stream (or jet) is decelerat- the body; in other words, the body per-
ed, there appears a pressure of the or- forms work. The work done by the body
der of pv 2 Since the viscous force is
. over time t is force times the distance:
proportional to the first power of the A = Put. Bearing in mind that
velocity, it can be ignored in rapid mo- Fvt = Spv3 l2, we arrive at the fol-
tion with large Reynolds numbers. It lowing formula:
would seem that the result is simply
Fi = pv 2 ,or F =
constant X p v2 S, F=^f~, (9.15.7)
with a k that is constant.
However, reality proved to be more which correctly yields the order of mag-
complicated than this. In the 18th nitude of this quantity. The liquid
-century the French mathematician, needs a certain time to “quiet down”
philosopher, and physicist Jean Le after the body has passed through it,
Rond d’ Alembert found that for a li- but its energy transforms into heat in-
quid whose viscosity is zero there is a stead of returning to the body.
solution to the equations of motion in The way in which a liquid flows around
which the pressure at different points a body depends essentially on the
on the surface is different but the resul- body’s shape. Certain shapes have been
9.15 * The Motion of a Body in a Fluid 353

found that have ft’s that are smaller grows. Even for streamlined bodies and
than the ft’s for a ball and disk by a high Reynolds numbers, ft is proportion-
factor of several tens (it is assumed al to an extremely low power of r),
that the ft of the disk is measured when precisely, it changes like r| 1/6 (Some
.

the disk is perpendicular to the flow). researchers prefer formulas of the type
The general idea is that the liquid, k =
(a +
fringe)" 1 .)
after it has let the body through, smooth- This is also the situation in the extreme
ly ranks” behind
“closes the body. case of a liquid flowing around a thin
If the contribution of the
this is so, plate lying along the flow. It would
difference in pressures diminishes and seem that in the limit of small viscosi-
becomes 10 to 15% of the entire force ties the liquid should slip along the
of resistance. Even at high Reynolds plate, since the contributes
pressure
numbers, the greatest part of the force nothing and the friction force is pro-

of resistance (85 to 90%) directly acting portional to the viscosity and is small.
on the body’s surface is provided by But in reality turbulence sets in (at
the viscous force. But this does not mean N^ e >
10 6 ) and the drag grows. 9 29
that the total resistance force and If the shape of the body or its position in

ft are proportional to the viscosity for in relation to the flow is nonsymmetric,

streamlined bodies for high values of for instance, the plate is at an angle to

^Re. the flow, there appears, in addition to

,
The viscosity,
it appears at hrstf the force parallel to the velocity of the
glance, enters the formula for ft (and F) liquid, a perpendicular component—
in the first power. However, the viscous the hydrodynamic lift (or aerodynamic
force is proportional to r\ (dv t ldn) and, lift if the fluid is air), which can exceed

all other things remaining the same (the the drag by a factor greater than 20.
dimension of the body, the velocity of The flight of airplanes and gliders is
the liquid at infinity, and the like), based on this fact.
the transient layer within which the 9 29
In calculating the Reynolds number we
-

velocity varies gets thinner as r\ de- substitute the length L of the plate for the
creases, that is, the derivative dv t ldn radius R of the ball: iV Re =
p vL/r\.

23-0946
 Chapter 10 Oscillations

10.1 Motion Under the Action Integrating, we find that

of an Elastic Force

Let us examine the case where the force

acting on a body depends solely on V0 Xq
the position of the body, F = F (x).
Above (see Section 9.2) we considered or H*-™±=^F(x)dx.
in detail the work done by such a force
X0
and found out that in this case the sys-
tem has a definite potential energy The physical meaning of this expression
u (x) with which the force is connected quite clear: the change in kinetic ener-
is

by the relation gy is equal to the work done by the force.

Using potential energy u u (x), =
we can write Eq. (10.1.2) as
d / mv 2 ^ _
du
Let us now turn to the problem on dx V 2 I dx ’

the motion of a body under the action

I =0
of such a force. The basic equation of or
ib( Tr+ u (
x)
)
motion is of the form
If the derivative of an expression is
m^r =F{x). (10.1.1) identically zero, the expression itself
is a constant, therefore

This equation cannot be solved di-

rectly since it involves the derivative
+u (x) = constant. (10.1.3)

with respect to time, while the force is

In this form, formula (10.1.3) expresses
given as a function of the x coordinate.
the law of conservation of energy ,
It is therefore natural to seek the quan-
namely, when a body is in motion due to
tities of interest as functions of the x
the action of a force depending exclu-
coordinates. For one thing, we will seek
sively on the position of the body (its
the velocity as a function of the coordi-
coordinate), the sum of the kinetic
nate, that is, v v (x). = will then We energy of the body, mv 2 /2 and its po- ,
represent the derivative with respect
tential energy u {x) remains constant.
to time, dvldt, as the derivative of a
This sum is called the total energy
composite function, since the x coor-
of the body.
dinate itself depends on the time:
The purpose of these lengthy trans-
formations here and in Section 9.14
dv [x (£)] dv [x (£)] dx (t)

dt dx dt
was to show that the law of conserva-
tion of energy (as applied to mechanics)
is a consequence of Newton’s second
But dxldt is nothing other than the
law. Also a corollary to Newton’s sec-
velocity v (x). Thus, dvldt = v ( dv/dx ),
ond law is the fact that the kinetic
whence
energy of a body is precisely mv 2 l 2
dv dv d mv 2 and not some other function of the ve-
m ~dt =mv -w=ta l

hr)
\
•
/ A r\
( 10 -
a
1Ja)a \

locity of the body. Of course, the value

of the law of energy conservation is not
Substituting this into the equation of limited by the simple examples consid-
motion (10.1.1), we get ered in this book. This law remains
valid within a much broader class of
phenomena, even such phenomena whose
£(=?) = f m - I* 0 1 - - 2* essence has yet to be clarified.
 10.1 Motion Under the Action of an Elastic Force 355

Howdo we continue the solution of (the straight line u = E), the picture is
theproblem of the motion of the body? vivid in the extreme. Intuitively, it is
Using the value of the velocity v 0 and clear that here the body will move be-
the x coordinate x 0 of the body at the tween two extreme points, or that the
initial time (t = t 0 ), we find the total motion is oscillatory. The velocity is
energy E of the body, which is a quan- proportional to the square root of the
tity that remains constant throughout difference E —u (x). For instance,,
the motion: E = mv\/2 + u (x 0 ). With when x == x A the velocity is propor-
,

the aid of formula (10.1.3) and know- tional to the square root of the length of
ing E, we can find the velocity of the
j the line segment AB (see Figure 10.1.1).
body as a function of x: To the right of point C and to the
left of point D is a region where E <
V(x)= / d~[E-u(x)] . (10.1.4) u (x), that is, a region that a body
with a given total energy E cannot
It remains to find the relationship reach for lack of energy (this is reflected
between x and t. From (10.1.4) we get in the fact that in (10.1.5) there appears
a root of a negative quantity). In the

•£=/ 1-'*— <*»' region where E >

u (x), the square
root yields two possible values for
whence v x ) in accord with two signs of the
(.

radical:

f 2 ’
v= ±y -^[e-u{x)\ .

At the initial time, both the quanti-

—7=—=
X
dx ty and the sign of u 0 are determined by
,

* = *o + , f
. (10.1.5)
the initial conditions. From then on r
y — —
)
X* [E u (x)]
the motion occurs in the direction spec-
ified by the sign of the initial velocity
Thus, the time t is expressed as a func-
tion of the x coordinate: t t (x). = v 0 Evidently, at v =^= 0 the sign of the
.

velocity cannot change suddenly. For

This function is given by an integral.
= instance, if at the initial time the body
We can find x x (t) by solving (10.1.5)
for x. Since v is expressed as a function
was located at the point x x 0 (ly-=
ing somewhere between points x D and
of x by means of a square root (Eq.
(10.1.4)), even a simple expression for
xE) and v 0> 0, the body will reach the

u =
u (x) frequently leads to extreme- extreme admissible point xc-
ly awkward expressions for t t x ). = At point xc, where the body’s velo-
(,

To get a general idea of the nature city vanishes, the formula

of the motion, it will be useful to draw v = Y
(2/m) [E —u (x)l is replaced by
the curve of u (x). Suppose the graph v = —]/ (2/m) [E —u (x)]. Since
of u =
u (x) has the form depicted in v = 0 at this point, the change in sign
Figure 10.1.1. If on the same graph we takes place without a jump (disconti-
plot the horizontal line at height E nuity) in the velocity. The situation is
similar at the point x =
x D Thus, in.

the case depicted in Figure 10.1.1, the

motion of the body will be in the form
of oscillations between the two ext-
reme positions, x c and x fJ .

Let us consider another example.

Suppose the potential energy of a body
is given by the function u = ax, with
a positive. Find the law of motion of
Figure 10.1.1 the body.
23 *
 356 10 Oscillations

We denote the values of the coordi-

nate x and the velocity v at the initial
instant of time, t 0 by x x 0 and v ,
= =
v 0 Then the total energy E is equal
.

to mv\l2 +
ax 0 Using (10.1.4) yields
.

v ( x ) =: + axo— ax)

with y = 2a/m. Knowing v (x), we

find the time
X Figure 10.1.2

f —f -U
C
\
dx

Y»l+ Y(*o — *)
= .
in Figure 10.1.2a), that is, the shorter the in-
x0 terval A =
Xi —
Xq over which the rise in
u (x) occurs, the greater is the force in absolute
Since under the integral sign we have value. The force is zero where the function
—y~ x u~1,zdu, where a = v\ y (x 0 —x), + u —u (x) is constant (to the left of x 0 and to
we can write the right of x x ).
Let a body start moving from x 0 (see Fig-
t =t 0 —y V ul +y (
xo x) Uo ure 10.1.2a) with a velocity i; 0 Suppose the
total energy of the body is equal to For what
.

E .

values of E
can the body reach point x ± ?
= — x) — vj. =
0, it follows that E =
t0
~Y Wv\ + y (x 0 Since u (x 0 )
On the other hand, E muf/2
is the body’s velocity at point x ±
—
mv%!2
u v where vx
and u x is
+
.

From this we find x:

the potential energy for x x v Therefore =
,

-f-
(t -t0)=— YY+n^) + »o-
-^=E- Ul . (10.1.7)

Transposing u 0 to the left, squaring, From this formula it is evident that if E

and canceling y, we obtain is lower than u v the body cannot reach x 1
because then vf becomes negative, which is
x = — |
- (t
— ^o )
2
~
h* impossible. For this reason, the body can reach
x =
x 1 only if E u± ^ .

(10.1.6) For this case we determine the work done

by the force F in displacing the body from x 0
Finding d 2 x/dt 2 we are convinced that to xy.

what we have is a case of uniformly de-

A_
mv\ mvl _ mvf
celerated motion This was to be ex- . 2 2 2
pected since F = —du!dx
that = —a, Using (10.1.7), we find that
is, the force is constant and negative, A =E— ux — E = —u x.
which means that the motion is uniform- The force F does not perform any work in fur-
ly decelerated. In this elementary ther motion of the body to the right of point
case where the force was actually inde- x l7 since F =
0 for x >> x x .

pendent of x there was of course no rea-

,

Exercises
son for employing such a compli-
cated computational procedure. 10.1.1. The potential energy is given by
the formula u =
kx 2 /2, where k is positive.
In the next example (in this connection see Construct a graph and show that the motion is
also Chapter 16) we will consider the potential oscillatory.
energy whose graph is in the form of a step 10.1.2. The potential energy u (x) is zero
(Figure 10.1.2a). Such a function u (x) is as- for x c0, equal to 2x for and 0<x<l,
sociated with the graph of force given in Fig- equal to 2 for x ^
1. At the initial time t0

ure 10.1.26 (to convince oneself that this is a body of mass 1 gram leaves the origin and
so, the reader should recall that F du/dx ): =— moves rightward with a velocity u0 (cm/s):
the force is extremely great and negative, that (a) v0 = 1, (b) v0 =
1.9, and (c) v0 2.1. =
is, in the direction of decreasing x. The steeo- In each case investigate whether the body can
er the curve representing u (x) is (the curve continue moving indefinitely to the right.
 (

10.2 The Case of a Force Proportional to Deviation 357

If it cannot, find the point at which it will tion 10.1. Instead we will guess the
stop.
10 . 1 . 3 u (x) x3
. = — +
Ax 2 At the initial .
type of solution and concentrate our
instant of time a body of mass 2 grams starts attention on investigating the proper-
out from a point x 0 at a velocity v0 (cm/s): ties of the solution.
(a) z0 =1, v0 = 1, (b) x0 = —2, v0 = 1, Thus, we assume that
and (c) x 0 —2, v0 — = —
1. In each case in-
vestigate the nature of the body’s motion (points x = a cos cot. (10.2.2)
at which the body stops, regions which the
body cannot reach). In the case of stopping
points, give at least a rough idea of their coor-
This form of the solution is chosen be-
dinates. cause the cosine is one of the simplest
10 1 4 The same requirements relative to
. . . periodic solutions, and the process we are
u (x) =
x 2 /{l +
x 2 ) as in Exercise 10.1.3, interested in (“oscillations”) is without
with m=2 grams: (a) x 0 0, v0 2 and = = doubt
(b) Xq =
1/2, v0 =
1/2. Express the time t
a periodic one. Substitut-
as a function of x in terms of an integral. ing (10.2.2) into the basic equation
(10.2.1) yields

10.2 The Case of a Force Proportional — ma co

2
cos co£ — — ka cot of ,
c (10.2.3)
to Deviation. Harmonic Oscillations
since
Consider a body acted upon by a force
F = — kx. As we know, to this force v — dx
dt
= — aco sin cot,

there corresponds
potential energy a
u =2
fcr /2. The origin is the position — tf(o
2
cosco£.
of stable equilibrium. The curve of po-
tential energy (parabola) has the shape
Formula (10.2.3) is valid for any t
shown in Figure 10.1.1. if raco
2 = k. Therefore, the function
The motion of a body under the ac-
(10.2.2) does indeed satisfy Eq. (10.2.1)
tion of such a force constitutes oscilla-
ii m<x>
2 = k, whence
tions to the left and to the right of the
position of equilibrium. can ima- We 10 2 4 > -

gine a ball rolling from one branch of < '

the parabola, building up speed and,

Thus
by inertia, rolling up the other branch,
rolling down again, and so forth. 10 1
By Newton’s second law, the equation
x—a cos [t]/ .
(10.2.5)

of these oscillations is
Observe that the square root in the ex-
r] 2 y
m ~lk = ~~ kx ' (10.2.1) pression for co does not lead to two solu-
tions since cos co t — cos ( — co£).

We will not solve it by the general and We sought the solution to Eq.
rather involved procedure given in Sec- (10.2.1) the form (10.2.2), but it
in
is quite clear that the sine is no worse
10 1
Of course, this picture does not con-
-

than the cosine. 10 2 Equation (10.2.1) -

stitute an exact description of the process we

then has another solution:
are interested in, since the velocity of the
body rolling from one branch
is
of the parabola
directed along a tangent to the parabola
x (t) = b sin cot. (10.2.2a)
and can be decomposed into two components,
the horizontal v x and the vertical v
z The .
10 - 2
a sin co t and
Both functions, x ± =
law of energy conservation includes, naturally, x2 = cos a)£, are characterized by the prop-
b
both quantities, u x and vz in view of which ,
erty that their second derivative is propor-
the force on the body (more precisely, the tional to the respective function (the propor-
horizotal component of this force) will not be tionality factors are negative: x'{ o2 x 1 = —
exactly equal to —
kx (think of the reasons and x 2 =
—co 2 x 2 ). The difference lies only
for this). But if the parabola is so gently slo- in the initial conditions (at t 0): while =
ping that v z is moderate, the “rolling from x i (0) =
0, for x 2 the initial rate of its change
the parabola” can be described as oscillatory is zero (x 2 (0) =
0) instead of the function
motion obeying formula (10.2.1). itself.
 ) . C

358 10 Oscillations

Indeed, d 2 (sin co t)/dt 2 2

co sin coL = — a = C cos cp ,
b = — sin cp, (10.2.8)
Substituting this function x (£) and where C and cp can be expressed in
its second derivative into Eq. (10.2.1)
terms of a and
and canceling sin cot, we get co =
Y k/m, which coincides with C = y a z+b 2 ,
tan q> = b/a,
10 3
Whence (10.2.8a)
(10.2.4). *

r
cp == arc tan ( — b/a z
.
.

sin (
t y - (10.2.5a) Let us find the period of oscillation,
T, that is, the time during which the
The reader can easily see that the body returns to the initial velocity
1 J

))• formulas (Iu.'2n0) "arid. (IU.z.ij

(see'
zos x = a cos co t b sin co£ (10.2.6) The function cos a (or sin a or <

lue of the solutions (10.2.2) and (10.2.2a) (a cp)) +

returns to its initial va
re- to Eq. (10.2.1) is also a solution.
tion. (To when the angle makes a complete
the (10.2.1) volution (changes by 2 ji). Thus, in
verify this, find the second derivative
erivative
ild of (10.2.6) and substitute it into Eq. expression a cos cot quantity co shoi
re- a and b can be chosen
,
en abso- vary through 2n in one period T. The
on lutely arbitrarily.) Thus, we have
ye the so- fore, T is found from the condit:
lution (10.2.6) to Eq. (10.2.1) contain-
ing two arbitrary constants, a and b.
T) = at
The solution (10.2.6) to Eq. (10.2.1) whence
can be written in a somewhat different
.9) way. Let us consider a solution to Eq. co T = 2jt, T -

(10.2.1) of the form (10.2.2) shifted by

rn- a phase angle cp: The quantity v == i/T gives the nu
nd x = C cos (cd£ + cp), (10.2.7)
ber of oscillations per unit time a
di- is called the frequency It has the .

where the numerical factor of the tri- =

d). mensions s _1 1/s (reciprocal secon
gonometric function is now denoted by
or
= — The unit of frequency— one cycle,
C Since here, too, x co C sin
the
(co£
.

+ cp) and, hence, x" = — o

2
C X
oscillation, per second goes by — i

special name hertz (abbreviated: H

c

in-
z )>
cos (co£ #, the +
function cp) = — co
2
in honor of the German physicist He
(10.2.7) solution to isEq. also a
nt, rich Hertz (1857-1894). It is evide
(10.2.1) provided condition (10.2.4) is —
In,
= from formula (10.2.9), that v co/!

1U-
met. 10 4 But since cos (a
*

P) + but more convenient in all forn

v,
cos a cos P sin a sin P, we can — las
it is
to deal with co and not with
rewrite (10.7.7) thus:
in 2 otherwise the coefficients 2 n and *

ity x = C cos cp cos ot

c — C sincp sin co t, will appear throughout. The quant
re- which means that (10.2.7) and (10.2.6) © =
2n/T is called the circular f,
10 6
represent the same solution only if quency and in (10.2.7) is cal’
led

The frequency co =
10 • 3
y k/m does not —
— the initial
,

phase angle
cp

yield a new solution since b sin t) = l0 .

0f
the simple formula ((
course,
)er-
b sin (y k/m t). Thus, the value co — arctan (—b/a) not sufficient for de is
ac- — V k/m yields the same solution (10.2.5a), mining cp completely— we must take into
ned but with a different coefficient b. count the signs of cos cp and sin cp determi
10 - 4
Clearly, a shift in the independent from (10.2.8).
10
on- variable of a trigonometric function does not • 6
To grasp
the origin of this term, c
ing affect the differential properties of this func- sider a line of length a revolt segment
ro- tion if x cos (t a) —
cos t, where + = counterclockwise. The similarity between
mt, t = +
t a with a constant, then tation and oscillation is quite appan
i
to dx/dt — (dx/dx) (dx/dt) (
dx/dx X 1 — = since a revolving hand of a clock returns
just dx/dx. (Going from (10.2.2) to (10.2.7) is its original position after each revolution ;

nal the same as a shift in time: t-> t 4- t 0 with as an oscillating body returns to its origi
=
,

rdi- cp/ co.) position after one period. Here the x coo
 ) a

10.2 The Case of a Force Proportional to Deviation 359

The constant a in (10.2.2) cannot be tions From (10.2.2) and (10.2.10) it

.

determined from Eq. (10.2.1) because follows that the amplitude is equal to
the equation is satisfied for arbitrary a ,
the original deviation of the body if at
which can be cancelled from both mem- the onset of oscillations the body was at
10 7
bers of (10.2.3). The same is true for rest. *

the constant b in (10.2.2a) and for Let us note in passing that, general-
both constants a and b in (10.2.6), ly speaking, if A (t) =L cos at (or
the constants C and cp in (10.2.7) are A(t) =
L sinco£), then L is the grea-
also arbitrary (check this statement!). test value of the quantity A (£) and is
If the motion of the body is described termed the amplitude of note A (£). We
by (10.2.2), the velocity is also that the oscillation frequency co
is independent of the amplitude a.
v= - —a sin Let x — xx (t) be a solution of
co (10.2.10)
Eq. (10.2.1), that is, let it be true that
m (d2xjdt2 ) = —
kx x We consider the .

From the relation cos 2 cot-\- sin 2 co£ =1 function x 2 (t) =

cx 1 ( t ), where c is a
it follows that for cos (of ±1 = constant. Substituting into Eq.
it will be true that sin oot 0. Con- = 2
(10.2.1) the quantities x 2 and d x 2 /dt
2

= —
,

sequently, the velocity v is equal to we get me {d2xjdt 2 ) kcx 1 ( t ), or,

zero at times of maximum deviation of cancelling out c,
the body in one direction or the other
d 2x 1
(x = a or x =
~a). Imagine that at 1

t< 0 the body is placed at the point

x = a and is held at rest in this point Thus, if x =
xx (£) satisfies Eq. (10.2.1),
with the aid of some external force (say so does x 2 (t) =
cx 1 ( t ). Similarly, if
a hook) until time t =
0, and then the x x ( t and x 2 (t) are two solutions to
hook is disengaged. At this instant the Eq. (10.2.1), x (£) x x (£) x 2 (£) is = +
body is at rest, and oscillations are also a solution. It is in this manner
initiated under the action of a force that we constructed the general solu-
F = — kx. In this case, the dependence tion (10.2.6) from the solutions (10.2.2)
of the coordinate x of the body upon and (10.2.2a). The velocity correspond-
time t is given by the formula x a = ing to solution (10.2.6) is
cos cot Since the absolute value of
.

dx
cos (x)t does not exceed unity, it follows v = = —a co sin co£ + &co cos cot.
that a is the largest value of x, or
the maximum deviation of the body from (10.2.11)
the position of equilibrium The con- .
Suppose we have taken x = a cos co£
stants b in formula (10.2.2a) and C as the solution to Eq. (10.2.1). Putting
t = 0, we get x 0 = a where x 0 = x (0)
in (10.2.7) have the same meaning.
,

The number a (or b or C, respectively) is the initial position of the body, hence
is called th e amplitude of the oscilla- x = x 0 cos cot. But then v = dxldt =
— x 0 co sin co£, so that at t = 0 we
nate of the end of the clock hand varies accord-
have v = 0. Therefore, using the solu-
ing to the law x =
a cos cot if the hand re- tion x = a cos cot, all we can solve is
volves with an angular velocity co. If T is the the problem with zero initial velocity .

period of one revolution, then v =

i/T is the Let us try the solution x = b sin co£.
number of revolutions per unit time and
= Here v = dxldt = 6co cos co£, and at
o) 2jtv is the angular velocity of rotation
expressed in radians per second. Since the
radian is a dimensionless quantity (see Appen- 10 • 7
We defined the amplitude as one half
dix 5), (o has the dimensions of 1/s. In view of the full span of one oscillation. Between
of the simple meaning of co in the case of the extreme left-hand point x and — —
circular motion, this quantity in problems the extreme right-hand point x = a, the body
involving oscillations is termed the circular travels a distance of 2a, which is twice the
frequency . amplitude.
 .

360 10 Oscillations

t = 0 we get u 0 fcco,=where v 0 = problem. Up to

now, in contrast to the
v (0) is the initial velocity, and general problem, we have only consid-
ered particular problems in one of
b = v 0 l(x), (10.2.12)
which x = x 0 and v = 0 at t = 0 and
whence x = in another v = u 0 at t = 0 but x = 0
(y 0 /co) sin co£. But at
t =
0 we have only x 0. Again, we = (at t = 0).
are unable to solve the general problem With
the help of (10.2.6) we can solve
with the aid of this solution we must — the general problem of the motion of a
body with an arbitrary position and an
assume that the oscillations started at
the point of equilibrium x — 0. arbitrary velocity at the initial time:
The relation (10.2.12) offers a prac-
x =
x 0 and u u 0 at t =
0. Using the =
tical and convenient device for mea- initial conditions, we find from (10.2.6)
suring impulse and velocity that is and (10.2.11) that a x 0 and b = =
widely used in mechanics and goes by vj(ti. Whence
the name ballistic pendulum. If a body
is suspended in the form of a pendulum
x =x 0 cos (tit -j — si n cot (10.2.13)
or held in a position of equilibrium
is
by springs and its frequency is known, The solution (10.2.6) is termed the
the initial velocity following a blow general solution to Eq. (10.2.1), while
may be determined from the amplitude (10.2.2) and (10.2.2a) are particular
of oscillations caused by the blow. solutions to Eq. (10.2.1).
The motion described by (10.2.6)
We will now show how (10.2.12) can be ap- or (10.2.7) is known as harmonic oscil-
proximately obtained via some general elemen- lations.
tary reasoning. The dimensions of amplitude
are cm, those of velocity cm/s, and those of
If the solution is written in the form
the oscillation period are s. Therefore, on di- (10.2.7), then it is clear that the ampli-
mensional grounds, the amplitude must be tude is equal to C. Hence, if the solu-
of the same order as the product of the initial
velocity into some portion of the period. Since
ti on is of the form (10.2.6), C =
motion caused by a blow lasts a quarter period ]/ a2 + b2 . Suppose x = x 0 and v =
up to maximum deviation and v v0 (be- < v0 at t = 0. Then a = x0 and
cause the motion is retarded), it follows that b = vjtti and so the amplitude is
b <u0 Tl4. If the motion has a constant decel-
eration, the average velocity would be equal
to half the initial value and, consequently,
b ~u0 T/8. Actually, however, as follows from
formulas (10.2.9) and (10.2.12),
C =V x\ + -W- (10.2.14)

v0 T _ v0T Exercise
b '
’
2 2jt 6.28
10.2.1. A body is in oscillation according
which is quite close to the result obtained via to the law d 2 x/dt 2 = — x. Find the function
the approximate approach. The important thing x =
x (t) and determine the period for the
here is that because the period is independent following cases: (a) x — 0 and v = 2 cm/s
of the amplitude, the latter is directly pro- at t = 0, (b) x = and v = 0 at = 0, and
1 t

portional to the initial velocity. (c) x =1 and v = 2 cm/s at = 0. For case t

(c) write the solution in the form of (10.2.6)

and as (10.2.7).
We have
verified that two different
functions, (10.2.2) and (10.2.2a), satis-
fy the equation m (<d 2 x!dt 2 ) = — kx. 10.3 Pendulums
Suppose we want to solve the problem
of the motion of a body having a given A model problem closely associated
initial position and a given initial ve- with the entire theory of harmonic oscil-
locity, that is, x = x 0 and u = u 0 lations is the pendulum problem.
at t = 0 (and the values of x 0 and v 0 Imagine a small (point-like) mass m
are arbitrary, each may not be equal at the end of a weightless cord of length
to zero). We
will call this the general Z, with the other end suspended from a
10.3 Pendulums 36$

(10.3.1)
is then replaced with the fol-
lowing approximate equation:
d 2 cp
(10.3.1a)*
dt 2

(10.2.1)
Clearly, Eq. (10.3.1a) is simply Eq.
which describes harmonic os-
,

cillations, with g/l playing the role*

of klm (if we divide both sides of Eq.
(10.2.1) by m). Therefore, the solution
to (10.3.1a) is given by (10.2.7), with,
Figure 10.3.1
to =
Ygil. The oscillation period of a
simple pendulum in this approxima-
fixed point (Figure 10.3.1). The system tion is given by the formula
is called a simple pendulum , and the
mass is called a bob. Suppose the bob r = -?5.
0)
= 2nl/T.
V
(10.3.2>
g
(and the cord) is deflected through an
angle cp from the vertical. One of the For instance, if we require that T 2 s,. =
forces acting on the bob is mg exerted
or that it takes one second for the pen-
by gravity, where g is the acceleration dulum to swing in one direction, then
of gravity. This force is directed down-
we get (putting g
2
9.8 m/s ) l = =
ward. The normal component of this 2
gt /4n 2 ~
0.993 m, or l 1 m. 10 * ~
force, that is, the one perpendicular to Here is a somewhat different approach
the path of the bob and directed out-
to the same simple pendulum prob-
ward from the fixed point, is balanced lem, which will be more convenient
by the force exerted by the cord, while when we turn to the problem of a physi-
the tangential component of the force cal pendulum. Suppose the simple pen-
of gravity, that is, the one tangent to
— dulum of Figure 10.3.1 at x 0 (the =
the bob’s path, is equal to mg sin cp; equilibrium position) has potential
it is the latter component that drives
energy u 0 its kinetic energy at this
,

the pendulum. (The minus sign in the point is zero. We deflect the pendulum
expression for the tangential component through an angle cp; its horizontal
shows that the component tends to deflection is then x (see Figure 10.3.1).
diminish the angle cp). Since the path In this position the pendulum potential
s that the bob travels from the equilib-
energy is u± = u0 + mgz ,
where
rium position 0 is hp, the bob’s velocity —
is l (dy/dt) and the acceleration is 2 - —Y l l
2
x 2
andi the
,
kinetic
2 energy is (m/2) dx/dt ) 2 (see Section
l (d\/dt ). Newton’s second law for (

9.6). The sum of potential and kinetic

this case is

In the days of the French Revolution,

10 * 8

ml = — mg sin cp, when the metric system was being adopted

in France, two measures of length were candi-
(10.3.1) dates, so to say, for the basic unit of length.
One was the length of 1/10 000 000 of the
d 2 ip
g distance from the equator to the North pole,
or-^f=- T sincp.
.

measured along the circle (called a meridian

circle) passing through both poles and Paris
(this length was called a meter and formed the*
The exact solution of Eq. (10.3.1) base of the new metric system), while the
is in no way simple (see Eq. (7.10.14) other was the length of a “one-second” simple
and the material that follows). But pendulum (the latter definition depends, obvi-
ously, on location because g varies in different
if cp is small, we can replace sincp
parts of the world). Both units are of the same
with the first term in the Taylor expan- order of magnitude and agree with the charac-
sion of sin cp (see Section 6.2). Equation teristic dimensions of the human body.
 362 10 Oscillations

energies will not change in the oscilla-

tion process, whence

u 0 + mg(l-~Y l 2 — x2)+ J (-^r) =u

j o,

Now we employ the fact that ,

x<^l (the deflection angle cp is samall),

x/l<^i 1. This enables representing
2
x in the form

Y p— x
to
A
2 1

X 2"| _ x2
i__

Y [TJ ( \ “ 1
,
~~ position of equilibrium through a small
J 21
angle cp, the difference between the
(here we retain only the first two terms current height z of the center of gravity
in the expansion of ]/" 1 — (
xll )
2 = of the rod and the lowest position of
11 — (xll)
2
]
1! 2
in powers of xll via this center, corresponding to the posi-
—
"Maclaurin’s formula; see Chapter 6), tion of equilibrium, is l l cos cp
nfter which the equation of pendulum (cf. the aforesaid). The potential ener-

^oscillations assumes the form gy u of the rod can be written as

u = mgz = mg (l — l cos cp)

= mgl (1 — cos cp ). (10.3.3)

Differentiating both sides of the above We expand cos cp in a series in powers
equation with respect to t twice, we get of and, since
cp cp is small, confine our-
dx d 2 x __ x dx selves to the first two terms: cos
‘
^
Z
~dt dt 2
“ “ 0/v^ T dt ’ cp ~1— cp
2
/2. Therefore approximate-
whence
ly we have u ~ mgl cp
2
/2. Clearly,
this relationship implies that the po-
d2 x _ g energy grows with angle
dt 2
~ ~"T (10.3.1b) tential cp,

that as the distance from the posi-

is,

This constitutes a different form of the tion of the center of gravity to the po-
equation of small oscillations of a sition of equilibrium (where cp 0, =
(simple) pendulum; the reason why it 2 =
0, and u 0) increases. =
is similar to Eq. (10.3.1a) is that for
The kinetic energy of rotation of the
small x and cp we have sin cp xll cp, = ~ rod about point A is

that is x L ~ )0 3 4)
' '
The ideas developed here can be ap- <

plied to the study of the oscillatory mo- where I is the moment of inertia of the
tion of a solid fixed at a definite “point rod (about the point of suspension).
<of suspension”. For the sake of simplic- But according to (9.12.13), I ml 2 = +
ity we will speak of a rod suspended at
a point A (Figure 10.3.2). Let the cen-
/0 ,
whence K -- ~~ (ml 2 + I 0 ) ,

ter of gravity be below the point of sus- where I 0 is the moment of inertia of
pension, the distance between the point the rod about the center of gravity.
of suspension and the center of gravity Suppose that the rod performs harmo-
being Z. Let us determine the period nic oscillations, that is, cp a cos cot. =
oi oscillation of the rod (the physical By the law of conservation of energy,
^
pendulum ). u +K= constant and ^max ^-max? —
If the pendulum is deflected from its which means that max is attained at K
 . ’

10.4 Oscillation Energy. Damped Oscillations 363

the position of equilibrium, where A

u = 0, while M m ax is attained at the
position where K= 0. Since dyldt = B
— aco sin coZ, we have
Figure 10.3.3

-^raax = i (ml + I 0 2
) «
2
W2 ,

ity of the rod, or when no oscillatory

^raax = mgl 7

2
»
motion takes place. 10 Z = 0 is
- 9
But
not a Eq. (10.3.6), that is,
root of
whence do)/dl does not vanish at this point.
Here we are dealing with a so-called
•tngl =\ m (
l
2
+- /0) « 2(° 2 . cuspidal minimum attained at the
boundary of the region over which Z
that is, varies (see Section 7.2).

<
,0 3 5 >
- ' Exercise

10.3.1. A pendulum is in the form of a

The oscillation period is
triangular lamina (tin or cardboard) (Fig-
ure 10.3.3). Determine the period of oscillation
T = 2k/co. (10.3.5a) if the pendulum is suspended (a) from the
acute end A and (b) from the middle of the
Formulas (10.3.5) and (10.3.5a) imply ,

base, or point B. In both cases, indicate how

that the greater the moment of inertia the point of suspension is to be displaced so as
J0 of the rod, the lower the frequency to obtain a minimum oscillation period.

of oscillations of the physical pendulum

and, hence, the greater the period T. 10.4 Oscillation Energy.
If the entire mass of the rod is con- Damped Oscillations
centrated at the center of gravity, then
/0 =0. In this case we obtain the well-
Let us write down the general solution
to Eq. (10.2.1) (the general harmonic
known formulas (10.3.2) for the oscil-
lation period (and frequency) of a sim-
oscillation) as x cos (coZ a). The — +
ple pendulum.
potential energy of the oscillating body
is equal, at every instant, to
If / 0 =/== 0, there is a definite position
of the point of suspension for which the
frequency is at a maximum Since the .
u (x (it
))
= 2
^=— 2
~ G0S
M+ °0
position of the suspension point is
characterized by Z, to find the position and the kinetic energy is
we desire let us solve the equation
dtoldl =
0. After simple algebraic ma- K (t) = -= -^- [
— Cm sin (coZ -f a)]
2
.

nipulations we arrive at an equation

mC 2 co 2
for Z:
2
sin 2 (coZ + a).
mg (ml 2
+ I 0) — mg l x 2 ml = 0,

(10.3.6) The oscillation frequency, as we al-

ready know, determined by the formu-
is
whence we have Z max ]// 0 /m. = la co
2 = klm. Substituting co 2 into the
For a rod of length L with uniform expression for the kinetic energy, we get
distribution of mass, 7 0 = mL 2 ! 12
(see Exercise 9.12.1) and therefore K (t)
— ^
' — sin 2
(co Z + a)
2
Jmax - L/V12 ~ 0.3 L.
Note that the minimum frequency, 10 • 9
The value co = 0,
of course, purely is,

o) — 0, is attained, obviously, at nominal; it means that

0 as l-> 0 (the
closer the suspension point is to the center
l = 0, that is, when the point of suspen- of gravity of the rod, the greater the peri-
sion coincides with the center of grav- od T).
 )

364 10 Oscillations -

so gy. This is not surprising since oscil-

=sLn 2(a)t+cC)
2?
<
lations with a frequency different from
=^
0 1
xxxxx /r
= cos 2(cot *(<)
0
( klm do not satisfy the basic equa-
tion of motion. Hence, for such oscil-
lations to occur, it is necessary that
the body be driven by some kind of
Figure 10.4.1 other, external, forces besides the force
F = —
kx (this force is associated with
Thus, the factor in front of the tri- the potential u (.x ) =
kx2 l2). Because
gonometric function in the expression of the work of external forces, the total
for potential energy and in the expres- energy mv2 ! 2 +
kx 2 ! 2 will no longer be
sion for kinetic energy is the same. The conserved.
functions themselves, cos 2 (co£ a) and + Let us now investigate the problem of
sin 2 (co£ +
a), are very much alike, damping of oscillations. Let a body be*
one being derivable from the other by acted upon by the force of friction in
a displacement along the time axis addition to the force F kx of a = —
amounting to At nl 2co (Figure = sping. Suppose the friction is so small
10.4.1). Each of the quantities u and over one period that the work of the
K oscillates between maximum and friction force is small compared with
zero: is at maximum, the oth-
when one the oscillation energy. We
can then as-
er at zero. Observe that the func-
is sume approximately that the oscilla-
tions cos 2 (co t -j- a) and sin 2 (co£ a) + tions occur as in the case of no fric-
describe oscillations about the mean tion: x (t) =
C cos (cd£ +
cp) (see Eq.

(average) value equal to half the maxi- (10.2.7)). The

oscillation energy is
mum. This is clearly evident from Fig- kC 2 /2. In the case of friction, the ener-
ure 10.4.1 and also from the familiar gy of the oscillations diminishes with
formulas the passage of time. Thus, friction will
cause the amplitude C in (10.2.7) to
cos 2 P =~ (1 + cos. 2(3)
— -~+ cos 2p, decrease slowly instead of remaining
constant. The law of decrease of C
sin 2 13 =-~ (1 — cos 2(3)
= -^ ^-cos 2(3. is defined by the condition that the de-
crease in total energy E is equal to the
(10.4.1)
work of the force of friction F x .

It is clear that the quantity

here
With respect to unit time, these quan-
tities are related thus:
taking on positive
(1/2) cos 2(3 oscillates,
and negative values alternately, and 1/2
2
te
= W&^ = kC dC
=F w
is the mean value of cos p. The sum at dt dt

of the potential and kinetic energies (10.4.3)

(that is, the total energy of the system).
where v is the velocity of the body, and

E= K+u= —y- [cos 2

(©f + a)
W 1 the power of the friction force.
is
In the oscillation process, both v and
F x vary periodically. The friction force
+ sin2((of + a)] = ig- , (10.4.2) F x is always dependent on velocity v and
is directed opposite to v so that the
,

is constant, as was to be expected. product Fpu always remains negative.

if we were to specify the
Note that In the case at hand of small friction,
motion with a frequency that did not which results in a slow damping of the
satisfy the formula co ]/ klm, the = oscillations, we can assume that the
sum of the potential and kinetic ener- change in amplitude C ( t is small over
gies would not be a constant and the several oscillation periods. The product
maximum kinetic energy would not be F ± v may be understood as the average
equal to the maximum potential ener- value of the product over one period.
 .

10.4 Oscillation Energy. Damped Oscillations 365

Formula (10.4.3) holds true only for

time intervals exceeding one oscilla-
tion period.
By way of an illustration, let us
examine the effect of a force of fric-
tion proportional to the first power of
the velocity of the body:

F = — hv F v= — hv 2
i , t . (10.4.4)

If v = dx/dt— — Cco sin (cot -f- cp), then

I’ v = — feC
Figure 10.4.2
1
sin (co£ + 2
co
2 2 cp). (10.4.5)

Note thataverage value of the forces, the elastic force kx and the —
force of friction h(dxldt). By New- —
2
•sin (co/ (p) +
over one period is 1/2
ton’s second law,
(see formula (10.4.1) and Section 7.8).
Using (10.4.3) and (10.4.4), we finally
( 10A8 )
get

We will seek the solution x(t) in the

s(¥) *T. same form as was obtained for a small
friction.
whence
dC ha 2 x(t) =C e-v 'cos(a)
0
t
i
t-{-a). (10.4.9)
~ r (10.4.8)
dt 2k This solution depends on four para-
meters: C 0 y, tOi, and a. ,
The values
that is (cf. (6.6.10)), of y and co t are determined from Eq.
while C 0 and a remain arbit-
€ = C 0 e~ yt
, (10.4.6)
rary,
,

Eq. (10.4.8) does not

that is,

where determine them, and require specifying

the initial conditions. Indeed, from
y = ha 2 /2k. (10.4.6a)
(10.4.9) we obtain
=
But since
can
co

be simplified:
2
klm formula
,
(10.4.6a)
—=— - yC^e-vt cos (co^ + a)
y = h!2m. (10.4.6b —C 0 (o i e~^ t sin (co^ -f a),

In formula (10.4.6), C 0 is determined

by the initial conditions. Multiplying
%r = y*C0e-y*co3 (©,* + <*)

both sides of (10.4.6) by cos (co£ + cp) + yC e~Y sin (co^ 4- a)

0 (o i
t

and employing (10.2.7), we get

+ C °i7^ _vt sin
,

oc +a )

x (£) = C 0 e~ yt
cos (cot + cp), (10.4.7) — Coale'Vt cos (co^ + a)
Substituting into (10.4.8) the expres-
where co = ]/ klm (see a characteristic 2 2
sions for x, dx/dt and d xldt and can-
graph of damped oscillations in Figure ,

yt
10.4.2). This is an approximate formula celing out [C 0 e~ > we get
obtained on the assumption that the my 2 cos (o)it + a) + my(£> 1 sin (co^+a)
friction is low and is proportional to
the first power of velocity. + my (iq sin (co^ + a)
In all cases where the friction force — mco cos 2
+ a) = — k cos^J+a)
((o ± t
is proportional to the first power of
+ hy cos (co^ + a)
velocity, the problem has an exact so-
lution. The body is acted on by two + fecoi sin (co^ +a),
 k k

366 10 Oscillations

or into Eq. (10.4.8), we arrive at quadratic

(
my 2 — mco
2
)
cos (co^ + a)
equation for y with two solutions:
y = =
y 2 The sum C x e~
y x and y .
yit
+
+ 2myco sin (co^ + a) 1
C 2 e~ yzt
of the two solutions corre~
= — + hy) cos (co^ +
(
a) spondingtoyj and y 2 will yield the gen-
+ ha> sin (co^ + a).
1 (10.4.10) eral solution to Eq. (10.4.8) and will
enable solving the problem for arbitrary
The last equation holds true for ar- initial data. This case of large h is con-
bitrary t if sidered in detail in Section 13.10 in

my 2 — mco 2 = — + hy ,
connection with electrical oscillations*.

2myco 1 = h (o^ (10.4.11)

Exercises
Canceling out of the second equa-
(o 1
10.4.1. Find the law of damping of oscil-
tion, we obtain y h/2m. Then from = lations for friction proportional to the second
the first we can find co x :
power of velocity (this friction is characteris-
tic of rapid motion of a body in a low-vis-
h2
CO?
2
= k
7—2 5— (10.4.12)
cosity fluid). Show that after the lapse of a large
1
m 4m .
time interval, the amplitude C (t) =
i/bt,
where b is a constant independent of C 0 which,

Consequently, is the amplitude at the initial time.

10.4.2. Find the law of damping of oscil-
lations for friction that is independent of ve-
x(t) = C<fi 2m
locity (this is characteristic of the friction be-
r tween hard dry surfaces). Determine the time
XC °S (y ~t~-^2- t + a ) (10.4.13) interval after which the oscillations cease.

When friction is low, that is, h is 10.5 Forced Oscillations

small in comparison with fc, we can ig-
and Resonance
nore fo2 /4m 2 under the radical sign as
compared with khn Then (10.4.13) . Consider a body acted upon by an elas-
becomes (10.4.7). Thus, in the approx- tic force F = —
kx. We
have estab-
imate consideration we correctly ob- lished that this force causes the body to
tained the law of decrease of the oscil- oscillate with a definite frequency
lation amplitude but did not notice the co = y
klm, which is known as the
small variation in due to frequency natural (or free) frequency. From now
friction. Note also that of the two laws on we will denote the natural frequency
that govern the decrease in amplitude, by co 0 so that co 0 = ]/ khn.
(10.4.6), (10.4.6a) and (10.4.6), Now let the body be acted upon by
(10.4.6b), only the second holds true
the elastic force and a periodic exter-
for arbitrary friction (not too large, that
nal force with frequency co, which gen-
is, such that h 2 4 km and the root < erally may differ from co 0 It then .

in (10.4.13) is meaningful), while the

turns out that the amplitude of the os-
first coincides with the second at small
cillations brought about by the exter-
k’s but in the general case is invalid
nal force is very strongly dependent on
because it contains co (which is not how close the frequency co of the exter-
generally a constant). nal force is to the natural frequency
If the friction is so high that h 2 > co 0 that is, to the frequency of natural
imri "tire rati icanu xn ( lb 1 id) n eco m es
f *
1 ,

, ' oscillations occurring in the absence of

negative and the formula becomes external forces. The phenomenon in
meaningless. This means that motion which the oscillation amplitude in-
involving appreciable friction is no creases as co approaches co 0 is called res-
longer oscillatory. In this case the so- onance and has a wide range of appli-
lution to Eq. (10.4.8) is to be sought in cations. It refers to any systems that
the form x Ce~~
yt
=
Substituting this . admit oscillations and vibrations. In
 a ,

"
10.5 Forced Oscillations and Resonance 367

mechanical systems (machine tools or For this equation to be true for all t’s,
motors) such vibrations can result in the separate terms involving coscof:
deformation and destruction of the and sin (tit in the right- and left-hand
equipment. On the other hand, resonance sides must be equal. Thus, we arrive at
can be useful, at times it is purposely the following system of equations:
used to produce, via a small force, vi-
brations of the operating tool with great — ao = — 20 —
2
<
)q 2yb(ti -f- ~~
amplitude.
In electric systems, resonance en- — =— b(ti
2
2ya(ti b(ti\ -f-
(10.5.3)

ables us, using several periodic forces

(cf. Eq. (10.4.11)). The second equa-
with different frequencies (say, a num-
tion yields
ber of radio transmitters), to achieve a
situation in which the oscillations in
our system depend solely on one of
6 = cog
2y(ti
— (0
2
a.

the periodic forces (the one whose fre-

quency is close to the natural frequency Substituting this into the first equation*
in (10.5.3) yields
of the system). This allows for tuning
a radio set to a definite station. 2
o-(ti 2
Let us set up the oscillation equation: a = l
m —
<ti

(10.5.4a)
(0)2 (o
2 2
) + (2v(o) 2 *

m — kx — + f cos (tit,
Then

(10.5.1) h = 1m (0)2— 0)2)2 +(2Y0))2

2yo)
• (10.5.4b)?

where the terms — kx and —h (dxldt)

correspond to the elastic force and the Going over to the form x =
friction force (see Eq. (10.4.8)), and C cos(co^+cp) for oscil lations and recalling
/ cos represents the external force.
(tit that C = Y 2 + b 2 (see (10.2.8a)), we
Divide both sides of (10.5.1) by m obtain the amplitude of oscillations
and set klm =
co^ in accordance with produced by the external force:
the fact that (in the absence of friction)
co 0 is the natural frequency of oscilla- 1
(10.5.5}
tions of the body. Denote the ratio V" (COg — 0)2)2 + (2 T o)) 2
*

him by 2y (see formula (10.4.6b)).

Then Eq. (10.5.1) assumes the form We see that C
the greater the closer is

— =- %x-2y — + —
d^x 2 dx .
f
co Sti>l.
,
(ti is to co 0 of C as a function
. The curve
of co for a given co 0 is shown in Figure*

(10.5.1a)
10.5.1 for two values of y (with flm 1 =
and co 0 —
1). The less the friction h r
It is natural to expect that under the that is, the less the quantity y h!2m y =
action of a force having a frequency co the sharper the rise in amplitude for
the body will oscillate with that fre- the frequency of the external force equal
quency. We
therefore seek the solution to the natural frequency.
to Eq. (10.5.1) or (10.5.1a) in the form

x = a cos at + b sin (tit. (10.5.2)

Substituting the expressions for x and

its derivatives into Eq. (10.5.1a), we
obtain
— 2 cos —
a(ti (titsin b(ti
2
(tit

=— cos (tit—b(ti\ sin (tit-\-2ya(ti sin

a(til (tit

— 2yb(ti cos + cos (tit — (tit.

*368 10 Oscillations

not hard to see that the sum of

It is different from ©§. Find the deviation of
the solution (10.4.9) to Eq. (10.4.8) w^ ax / (t>o from unity as a function of y.
and the general solution (10.5.2) to Hint. Test the radicand in (10.5.5) for a
Eq. (10.5.1),
minimum, denoting co
2 = z.

.x = a cos co£ + & sin oo£

yt 10.6 On Exact and Approximate
+ C 0 e~ cos (co^ + a), (10.5.6) Solutions of Physical Problems
where a and b are given by (10.5.4a)
In the preceding section we had the luck
and (10.5.4b), and co 1 and y by (10.4.12)
to find with
comparative ease the
and (10.4.6b), is also a solution to Eq.
exact solution of the problem involving
(10.5.1). Formula (10.5.6) provides a
oscillations of a body under the action
solution that satisfies any initial data
of a periodic external force in the case
by choosing C 0 and a. Indeed, suppose
of an elastic force that tends to move
that x = x 0 and v = v 0 at t — 0.
the body to the position of equilibrium,
Then, using (10.5.6), we find that — kx ,
and friction, —h (dx/dt). With
.x 0 = a + C 0 cos a, the exact solution at our disposal, we
v0 = b(j) — C 0 (y cos a + (Oj sin a). can easily find a number of important
limiting cases.
(10.5.7)
(1) The frequency co of the external
From this system of equations we can force is extremely low compared with
-determine C 0 and a (see Exercise co 0 ,
where co 2 klm. Neglecting co =
10.5.1). in (10.5.5) in comparison with co 0 we
Thus, (10.5.6) is the general solution get
of the problem involving oscillations
of a body under the action of an elastic C = f/m&l = flk. (10.6.1)
force and a periodic external force.
(2) The frequency co of the external
This general solution confirms the as- force is much higher than co 0 Then we .

sumption, made at the beginning of can neglect co 0 in (10.5.5) and assume

this section, that under the protracted that
action of an external force with fre-
C = -tm
1
quency co a body will oscillate with the (10.6.2)
+ 4y
,

same frequency co. This is true because y co 4

2 co 2

4
no matter what the initial conditions, But y 2 co 2 <C co (friction is not very
they only affect the values of C 0 and great, since otherwise there would be
•a, which is to say, only the last sum- no oscillations); neglecting the term
mand in the solution (10.5.6). However, 4y
2
co
2
,
we obtain
in the course of time, this term, which
Jhas frequency g) 1? becomes very close C -- —^ TttCO
2 .
(10.6.2a)
V 1

yt
to zero, due to the factor e~ which ,

tends to zero as t->- oo, and we can (3) The

friction force is small that ,

is, small Then in (10.5.5) we can

h is
neglect it for large t. The remaining
.

neglect the term containing y. As a

terms describe oscillations with fre-
quency co, which do not decay with the result we have
passage of time since they are main-
tained by the action of the external
force, whose amplitude is constant by as-
(The appearance of the absolute value
sumption. due to
of the difference in (10.6.3) is

Exercises the fact that in (10.5.5) we take the pos-

itive value of the root, since it is clear
10.5.1. Determine C 0 and a from the sys-
tem (10.5.7) that C and /, that is, the oscillation am-
10.5.2. Because of friction, the maximum plitude and the external force ampli-
amplitude C is obtained at somewhat tude, can be only positive.)
 - ,

10.6 On Exact and Approximate Solutions of Physical Problems 369

(4) The phenomenon of exact reso- cases that one can acquire some exper-
nance: the frequency co of the external ience in the proper choice of approxi-
force is exactly equal to the natural mations and be sure of the results, since
frequency co 0 Then (10.5.5) is replaced
. the correctness of the limiting cases
with that follow from the general formulas
can serve as a global check on the en-
c = ±m =
1 /
/ ____
(10.6.4) tire process of solving the general equa-
hio hio 0
tion, both the setting up of the equation
These limiting cases actually make that describes the process of interest
up over 90 % of the content of all the to us and the solving of this equation.
results obtained. When one obtains a Let us now return to the first case:
general result, it is always advisable the frequency co of the external force
to simplify it by considering various is low. This is clearly a slow motion.
limiting cases, as we have just done. Therefore, in the original equation
The simple formulas relating to the limi t- (10.5.1).
ing cases are more easily remembered
and more frequently used in practical
situations. Only once in a while does
m —
d^X
-
dt 2
5
— — kx — h —
dX
7

dt
- + / cos
7
. e ,

(tit,

one have to resort to general formulas. we can drop terms m (d 2 x/dt 2 ) and
Although limiting cases do not cover h (dxldt) involving acceleration and
everything but we know almost every- velocity, since if x C cos ((tit ^
(p), +
thing that is contained in the more com- then dxldt is proportional to co and
plicated exact formula. d2x/dt 2 to co 2 and for low frequencies
,

The question arises quite naturally these quantities are small. In this case
as to the possibility of obtaining limit- we get 0 kx ~—
/ cos (tit, +
ing formulas directly via simplifica- whence
tions in the equation itself rather than
in its solution. To solve an involved x ~ k 1
f cos (tit = C cos (tit

equation in exact fashion and then sim- with C = flk. (10.6.5)

plify the solution is just as senseless as
to use intricate machinery to package Therefore, for low frequencies of the
goods elegantly and then immediately external force, at each moment the ap-
tear open the package to get them since plied external force / cos (tit is balanced
each piece is to be used separately. There- by the elastic force kx. This result is
fore it is important to learn to extract clearly very general, for it refers to
from the general solution of the problem any motion with a low frequency. This
the particular (and limiting) cases with- limiting case is called a static case.
out solving the equation itself. For one, the elastic force G (x) may be
To obtain limiting (approximate) any function of the coordinate and not
expressions directly is particularly im- only a linear function of the coordi-
portant for the added reason that an nate like G = —
kx and the external ,

exact solution is very sensitive to the force F (£) may be any function of time.
slightest variations in the statement of The oscillation equation takes the form
the problem. A slight complication in
the problem and one finds it impossible
m J*L = G(x)-h%- + F(t). (10.6.6)
to get an exact solution. An approxi-
mate solution is rougher but more sta-
ble with respect to variations in the It is not always possible to obtain the
problem. exact solution of this equation, but the
Of particular importance to students approximate approach is preserved. In-
are cases where it is possible to obtain deed, neglecting in the case of slow mo-
and compare both solutions, exact and tion the terms in ( 10 6 6 ) that are pro- . .

approximate. It is precisely in such portional to velocity dxldt and accel-

24-0946
 c

370 10 Oscillations

we get G {x) + F
eration d?xldt 2 ~ (t) pare not their instantaneous values but
~ — F (x). From this we
,

0, or G (x) their amplitudes. The ratio of the exter-

find an approximate relationship be- nal force / cos (tit to the elastic force
tween x and or x ~ x t, Substitut- (t). (10.6.9) (the ratio of the amplitudes)
ing this into the exact equation
x (t) is /-f- (cOq/cd
2
The ratio is
)/ — 2
co / dJ.

(10.6.6), we can find the order of the the higher the greater co is. Similarly,
error introduced by neglecting the the ratio of the external force to that
terms h(dxldt) and 2
(d 2x/dt ). m of friction, / (hf/nuti) (m/h) co,~ =
Let us take a look at the second lim- grows without limit with co. For this
iting case, very high frequency co. reason, given high co, the external force
The time of action of the external force appreciably exceeds both the elas-
and, hence, the impulse during each tic force and the force of friction. This
half-cycle (while the force is acting in supports the possibility of an approxi-
one direction) are small because of the mate consideration of motion under the
short duration of the half-cycle in this action of an external force alone. 1010
case. Thus, for a given amplitude / The third limiting case, neglect of
of the external force, the higher the friction, presents no difficulties at all.
frequency co, the smaller the velocity Here Eq. (10.5.1) takes the form
that the body can acquire and the smal-
ler the displacement of the body. Neg-
m ~ -kx-\-j cos (tit
lecting the terms kx and h (dx/dt) in
Eq. (10.5.1), we arrive at an equation = — rmtilx + / cos (tit, (10.6.10)
of motion of a free body with no forces
acting except the external force:
since co^ and, = klm
hence, k =
f cos (tit. (10.6.7) rri(til. We seek
the solution to this
equation in the form x C cos (co£ = +
We seek the solution to this equation cp). Substituting into this equation
in the form x =
B cos (tit. Then the expressions for x and d^xldt*, we
d 2xldt 2 = —
B co 2 cos (tit, and whence get cp =
0, cp Jt, and (ti\ =co
2
X m \
—
— ~
|

Eq. (10.6.7) takes the form Bm co 2 x C cos (tit f cos cd£, whence
cos (tit ~
f cos (tit from which it fol-
~
,

lows that B -//mo) 2 and hence /

10 6 11 )
,

m |
cog — co 2 1
*
(
. .

x ~ — jjj—p cos (tit . (10.6.8)

Comparison of this formula with the
In the standard form x ~ C cos exact formula (10.5.5) enables estimat-
(cd£+ cp) the solution (10.6.8) can be ing the conditions under which this
written as follows (with C positive): approximation works.
Clearly, at co — co 0 approximation
x~ cos (ctit + n). (10.6.8a) (10.6.11) not valid and the fourth
is
limiting case (the frequency of the ex-
Here, the elastic force is ternal force is exactly equal to the nat-
ural frequency of oscillations, or re-
— kx ~ cos a>t = -3* / cos co t, sonance) must be treated separately.
Here we will not ignore friction, how-
(10.6.9) ever.) We seek the solution to (10.5.1)
and the force of friction is
10 - 10
It is essential that the friction force
—h dt
— —
m(ti
sin co t. (10.6.9a)7
v
considered above
closer the velocity
closer to zero the
to zero. In dry friction
is
be

(force of friction independent of velocity), an

When comparing forces that depend external force less than that of friction will
periodically on time, one should com- not cause oscillations at any frequency. >
 o ) .,

10.6 On Exact and Approximate Solutions of Physical Problems 371

in the standard form x — C cos(c 0 £ + energy considerations whose value con- ,

<p). Then sists in the fact that they also enable

solving approximately certain problems
m ~^T = — mCa\ cos (<B 0 f + <p), that have no exact solutions. The power
developed by an external force / cos co*
and, since (x)
2
0
=k/m , in the case of motion specified by

m— d 2x
== — kC cos (co 0 £ + <p)
= — kx. the expression x C cos (co£ cp) is = +
Substituting into the expres-
W = ex / COS (x)t -~—
at
(10.5.1)
sions for x and the time derivatives = — fCo: cos cot sin (cot + cp)
yields
Let us determine the average power
hC co 0 sin (cd 0 £ + <p) + / cos co 0 £ = 0. of the external force during a large (more

This equation will hold true for any precisely, infinite) interval of time:
t if C = f/h(x) 0 and cp = — jt/2. Hence,
W = —fC
the solution of our equation is
ex co cos co£ sin (oot + cp)

here the bar stands for averaging (see

7^ cos (10.6.12) Section 7.8). Note that

The oscillation amplitude at resonance cos cot sin (cot -f- cp)

is C = flh (D 0 and we clearly see that

,

there is no way in which we can ignore

= ~2 sin 2(0 * cos <P + cos 2
co£ sin cp,

the friction, at resonance. or h,

Referring to Figure 10.6.1, let us look and so
at the relationship between C and co
provided by the approximate formulas
cos co t sin (at + <p)
= ~ sin 2 at coscp
(10.6.3) and (10.6.4). Formula (10.6.3)
yields two branches of the curve C = + cos 2
at sin rp — ~ sin
|
(p,

C branches that go off to infinity

(co),
as co ->• co 0 while formula (10.6.4)
,
since sin 2 at = 0 and cos 2 at = 1/2. Con-
yields a finite value C C 0 ( //feco 0 ) = sequently,
at co =
co 0 If we construct the curves
w =
.
fC(D
of (10.6.3) and place the point A = TIT
ex 2 — Sln <P»

A (co 0 C 0 ) that corresponds to for-

,

mula (10.6.4), it is then easy to draw which can be written as

free-hand a smooth curve (dashed in
Figure 10.6.1) which, far from resonance, ^ex = -^cos (cp-f-y) . (10.6.13)
coinsides with the curves of (10.6.3)
and has a maximum C 0 at co co 0 = .
Now let us determine the average pow-
In the case of resonance, the ampli- er of the force of friction. Since
tude C can be determined by means of F fr — — hv it follows that
W ex - hv 2
.
(10.6.14)
But

72=
(ir)^ C 2co 2 sin 2 M+ cp) = -^.
see (10.4.1)). Therefore (10.6.14) yields

W ex
= —hCz^/2.
Since the work of the external force
goes to overcome friction, the average
24 *
 372 10 Oscillations

powers of the external force and the Let us determine Wf r The instantaneous .

force of friction must be equal: value is [r vF f r W =

hv 2 u n ~1 since = — ;

=
\

=
|

v2 v 2 we can write W\ T
\ |
—h v n+1
, |
.

w
|

u . =w „ e (10.6.15)
Substituting here the value of y, we find that
W = —hCn+1
fr co£ +1 |sin((D 0 *+(p)| n+1 .

that is,
(10.6.18)
fC(D C2G )
2
Using this, we get l0 u
cos {’P + -5-) -

W = ~hCn+1 0^+1 A, A—
tT ( | sin (<o 0 f + q>) |
n+1 .

or / cos + =hC(o , The condition (10.6.15) yields hCn+1 a)^ +1 A =

' -
fC C0 0
1

cos whence
whence (»+-f)i-
n
C = -jt| cos + T) (10.6.16) f { \
The maximum
v-
i
,
P cos •
( 2kA(,i” I'p+t)
amplitude attained at resonance is
The maximum possible amplitude
(at resonance) results, as may be seen
from (10.6.16), at cos (cp
that is, at <p
n/2) 1,
= —
n/2, and this deter-
+ = C =
C0 0
1

V 2 hA
(10.6.19)

A
particular case of formula (10.6.19) for
mines the initial phase angle cp. Here n = (friction proportional to the first power
1

<o — co 0 and C = flh(o 0 Hence, the .

of velocity) is the earlier-found formula C —
f /h(& 0 Note that at n =£ 1 the equation of
solution of the equation of motion in
.

motion becomes nonlinear and we cannot write

the case of resonance is
a general solution to this equation that
would be similar to the one found in Section
10.5— there is no way in which an exact solu-
tion can be found via elementary methods
this case.
We again have the formula (10.6.12).
Let us return to formula (10.6.13).
10.7 Combining Oscillations. Beats
From this formula we see that at reso-
nance W
ex has the largest value since Quite common
the situation when a is

at resonance cos (cp n/2) 1. For + = body performs oscillations si- several
this reason, in the case of resonance an multaneously. Visible light, for in-
external force develops maximum aver- stance, is the sum effect of many vibra-
age power and, consequently, performs tional processes related to the various
maximum work. electron transitions in the atoms of the
body emitting the light (see Chapter
12). Similarly, the sound vibrations
These arguments of an energy nature make
that we perceive when listening to a
it possible to determine the amplitude at res-
onance also in the case of a more complicated symphony orchestra are the sum of vib-
•dependence of the force of friction on velocity. rations of the air generated by each mu-
Let the force of friction be proportional to a
<fixed) power of the absolute value of velocity:
sical instrument in the orchestra and —
the conductor’s goal is to control the
"1
F fr =—hv M 71
,
(10.6.17) various instruments and hence all these
vibrations to produce an effect in
where h is positive, and therefore the first factor
ensures that the signs of v and Ff r are differ-
which all the vibrations merge into a
ent. At v >
0 this formula yields Ff r = beautiful melody. (The effect of purely
— hvn while at
, K0
we get Ff r h v n
that is, in all cases the force of friction is in
= | | , random mixing of many sound vibra-
tions is known to anybody who has lis-
opposition to velocity. Since we assume, as
tened to an orchestra tuning up before
before, that the motion is on the whole deter-
mined by the external periodic force F =
10 11 For reference we give values of A
} cos (0 tj that is, is of an oscillatory nature with
-

frequency co, or x C cos (g)£ —

/P), the av- + for a few n: n 0, A 2/jt 0.64; n 1, ^ =
erage power of the external force is given by A = 0.5; n 2, A ~ 4/3ji~ 0.42;
= and
formula (10.6.13) for this case, too. n = 3, A = 3/8 = 0.375.
 10.7 Combining Oscillations. Beats 373

a each musician tuning his

concert, + B and (ON) 2 = C2 + D 2 (cf. (10.4.2)
2

instrument without regard for the and (10.2.8a)), while the energy of the
other instruments.) There are also many combined oscillation (10.7.2) is pro-
cases (sometimes important) when two portional to (OP) 2 If the initial phases
.

ormore harmonic oscillations are super- in (10.7.1) are the same, the seg-
imposed; more than that, almost any ments OM
and ON point in the same
motion can be represented as a sum of direction and the length of the segment
many harmonic oscillations (see Sec- OP equal to the sum of the lengths
is
tion 10.9). of segments OM
and ON; the energy of
Westart with the addition of two oscillation (10.7.2) then proves to be
harmonic oscillations. The simplest much higher than the sum of the ener-
case here is when the oscillations are of gies of oscillations (10.7.1), since the
the same frequency, for instance, when energy is specified by the square of the
we are considering the sound produced amplitude of oscillation (OM, ON ,

by two violins playing in unison. The and OP ); if the oscillations in (10.7.1)

result is a harmonic oscillation of the are the same, the energy of their sum
same frequency. Indeed, if the two oscil- will be four times the energy of each
latory processes are described by the oscillation. If the initial phases in
formulas (10.7.1) are opposite (that is, differ by

=A
Jt), the segments and OM
are in ON
xx cos cot + B sin cot,
opposition; here the amplitude OP of
x2 =C cos cot A- D sin cot, (10.7.1) the sum oscillation is equal to the
difference of amplitudes OM and ON
then the sum x = x1 + x has the form
2 of the component oscillations. The oscil-
lations may even “cancel out” if their
x — {A C) cos co£ + (B + D) sin cot.
amplitudes (or energies) are the same.
(10.7.2) But if the energies differ only by a small
quantity and the phases of x x and x 2
The relationship between oscillation differ by jt, the sum oscillation will
(10.7.2) and oscillations (10.7.1) are have an extremely low energy (see
most conveniently described as follows: Exercise 10.7.1).
if to oscillations (10.7.1) we relate The case of addition of oscillations
points M and N with coordinates (^4, B) with
sum
different frequencies, when the
oscillation is not harmonic, is more
and (C, D) (or vectors OM and ON ), complicated. Let us consider, byway
then oscillation (10.7.2) has correspond- of an example, a mechanism known as
ing to it the vertex P of the parallelo- the crankgear which transforms trans-
,

motion into rotation, and vice

gram OMPN (or the vector OP = lational
r7
/ .2, snows'sucn a mec.
”'vui!s9. r?gfrnj- iu.
OM + ON; see Figure 10.7.1). The anism. crank OA is connect*
The
energy of oscillations (10.7.1) is then with the slide block B (a piston movii
fixed by the quantities (OM) 2 A 2 4- = backward and forward inside a clos*
cylinder) by means of the connectii
rod AB; the length of the crank will ]

denoted by R and the length of the coi

necting rod by L. Let us assume th,
the crank is in uniform rotation aboi
the axis O with a frequency co, so th;
the projection C of point A on the O
axis of the cylinder in which the slic
block moves performs a harmonic osci
lation about point <9. The slide bloc
Figure 10.7.1 B will also perform oscillations aboi
 —

374 10 Oscillations

Figure 10.7.2

its midpoint, but these oscillations

will not be harmonic. Figure 10.7.3
Indeed, if Z_ AOB = co £, then OC =
R cos cd£, AC — R sin cot, and. A peculiar situation arises when two
hence, oscillations with close frequencies co x
and co 2 are combined. If, for istance,
5C = l/L 2 ~i? 2 sin 2 «* Xl =
A cos co^ and x 2 A cos (o 2 t =
= Ll/l-(i?/L) 2 sin 2
(of. (note that the amplitudes of both oscil-
lations coincide), then
Thus, we have
x= x + x = A cos o + A cos
i 2
c
{
t o2t
c

x = OB — R cos (tit
= A (cos co^ cos co 2 £)

+ L ]/ 1 — (/?/L) 2 sin 2 (tit. (10.7.3)

= A X 2 cos ^1=^- cos t *•

But in the real mechanism R is usually (10.7.5)

much less than L, that the ratio
so
RlL is small (often of the order of 1/5, Here, obviously, the factor 2A X
so that {RlL) 2
(10.7.3) ~ 1/25); for a small z
cos — — — changes very slowly, and
we can replace Y For— 1 z with 1 — zt 2
t

0)1 0)2
(see Section 6.4). this reason the the factor cos t corresponds to
comparatively complicated formula
can be simplified: oscillations with a frequency
0)1
~F -2
(0
~
R2 \ co.Oscillations specified by (10.7.5) are
1 sin 2 con known as beats (see Figure 10.7.4a);
( 2 JT
often they are described (not quite
— 2
(1— COS2©0
= RD cos +
(tit
, V
+ L^l r a
~2fT
7?
2
"I

= j
+ i? cos (o^-f-^-cos 2co£.
(10.7.4)

The term L
first
2
i? /4L specifies —
the midpoint in the position of slide
block5 (the slide block oscillates about
this midpoint), participating in two
harmonic oscillations simultaneously,
that is, in the oscillation with frequency
co and amplitude R fixed by the motion

of point C and in the oscillation with

,

double frequency 2co and amplitude

R 2 IAL fixed by the change in altitude
CA of point A (Figure 10.7.3). Figure 10.7.4
10.8 The Vibrations of a String 375

correctly) as harmonic oscillations with have frequency of 1.5 x 10 7 Hz

a
(or 15 MHz), while the amplitude of
frequency co ----y 2 and a slowly
j these oscillations varies within the au-
varying amplitude A (t) = 2A x dio-frequency range, that is, the range
cos --LT
0)2
t (
= C cos Q£) . A similar of frequencies to which the human ear
is sensitive, approximately 15 to
situation arises when the phases of the 20 000 Hz.
oscillations x 1 and x 2 differ only in —
this case the phase of the sum oscilla-
Exercises
tion x also experiences slow oscillations
(see Exercise 10.7.2a). 10.7.1. What are the (a) energy and (b)
amplitude of the oscillations obtained through
What happens if the frequencies co combining the oscillations x1 and x 2 with the
and Q of the component oscillations same frequency: x 1 =
A x cos (cot cpO and +
differ considerably ? Suppose, x 1 = A X x2 —
A 2 cos (cof q> 2 )? +
cos cot and x 2 = B cos with Q^> co. 10.7.2. (a) What type of oscillations is the
Then the sum oscillation x x± x2 = + sum two harmonic oscillations, x 1
of =
a cos (oV+tPi) and x 2 a cos ((D 2 £—cp ), if
2 + (Oi
can be represented in the form of a har- -auk po-
'xx»
2 "duTe 'fib
monic oscillation with amplitude B tion for oscillations xL and x 2 with different
and frequency Q about the slowly amplitudes, a x and a 2 .
varying position of equilibrium, a =
a (t) =A cos cot (see Figure 10.7.46, 10.8 The Vibrations of a String
where A^> B).
We encounter the phenomenon of In Section 10.7 we studied the function
beats when we study ocean tides: the x =
f (t) that was the sum of two har-
monic oscillations (with the same or dis-
water level in oceans changes due to
the attraction from the moon and the tinct frequencies). Next we will inve-
stigate the extremely important ques-
sun; in view of the earth’s rotation about
axis, the level oscillates with a
tion of representing any function (contin-
its
period of 12 hours. But since the moon uous or even discontinuous) in the form
orbits the earth, the period of the level of a sum of harmonic oscillations.

oscillations caused by the moon’s at- Physicists and mathematicians arrived

traction differs somewhat from the pe- at this question in their studies of a
riod of the level oscillations caused by wide range of physical phenomena. One
the sun’s attraction: while the period of these deals with the well-known

of the “sun” tides is exactly 12 hours, problem of a vibrating string .

the period of the “moon” tides (which Let us consider in the xy- plane a
string fixed at points x 0 and x = l\ =
are stronger) is close to 12^- h. Since the string is understood to be a thin
these oscillations have different ampli- thread that can bend freely and along
tudes, the amplitude of the sum oscil- which there is a constant tensile force
lation is never zero, that is, the sum acting. Let us disturb the equilibrium
oscillation does not coincide exactly (a state in which the string lies along

with the beats depicted in Figure 10.7.4a; thex axis), that is, we shape it in a cer-
however, here also we can speak of a tain way specified by a function y —
slow variation of the oscillation ampli- / (x) (Figure 10.8.1). We then let the
tude of the water level, the greatest string go. The question is: how do the
height of the tide exceeding the lowest
height by a factor of almost three (see
Exercise 10.7.2b). Oscillations with a
varying amplitude are also used in ra-
diobroadcasting; for instance, the elec- ctx C a?
tromagnetic waves in the band SW
with a wavelength of about 20 meters Figure 10.8.1
 ) l x

376 10 Oscillations

tensile forces change the shape of the to the x axis (and hence cause the string
string? We
will restrict our discussion to vibrate in the transverse direction)
to small deformations, or strains, of the are, respectively, sin a and x R R
string, that is, we assume that y is sin (a + da), where a and a + da are the
small (the deflection of the string from angles between the tangents to the
the x axis is small) and dyldx is small string at points {
y) and {x M ,
N +
(the deformed string still closely re- +
dx y , +
dy) and the x axis (see Fig-
sembles the undistorted string, so that ure 10.8.1). The resultant of these
the angle a =
arctan {dyldx) between a components is
tangent to the string and the x axis is
small). We
write dyldx instead of
R sin (a + da) —R sin a ~ Rd ((sin a)
dyldx because the deflection y of the (cf.Section 4.1). But we already know
string changes in time too, or y — that tan a dyldx and sin a =
y {x, t) is a function of two variables, tan a a, so ~ that we can write
the abscissa x and the time t. will We sin a cz. tan a ~
dyldx (here we again
now derive the differential equation ignore quantities of the order of
that governs the behavior of y (r, t). a2 dyldx) 2 ) or higher). For this
(or
Let us take a small segment of the reason, d sin a d {dyldx) = ~
string, MN,
restricted by points 2 2
{d y/dx ) dx since for the time being
M {x, y) and N (x+ dx y + dy). The
,
,

t in our formulas is assumed to be fixed.

length of this segment is obviously Thus, the force that causes the seg-
2
ment MN
of the string to vibrate is
ds = V dx 2 + dy = dx 2
]/ 1 + (-|f-
close to 2 2
R
{d y/dx ) dx and the equation ,

of string vibrations can be written thus:

d 2y
R Py d2 y d2 y
a dx — dx or x>2

dt 2 dx 2
,
dt 2 dx 2 *

(cf. Sections 7.9 and 6.4), since we as- 10 8

. . 1)
(
sume the derivative dyldx to be small
and neglect terms of the order of {dyldx) 2 where c
2 = Rio (already a positive
and higher. Since we are considering quantity). The solution to this equation
string vibrations solely in the direction must satisfy two boundary conditions
perpendicular to the x axis, the velocity
y (0, t)
= 0 and y { t) = 0 (10.8.2a)
v of point M
is dyldt and the accelera-
for all values of t (the endpoints of the
,

tion a coincides with the second (partial)

string are fixed) and two initial condi-
derivative d 2 yldt 2 while the mass of
the segment MN
,

of the string is ads ~ tions

0)
adx, where a is the (constant) li-
y (x, 0) = / (x) and =-- 0
near density of the string (which is as-
sumed homogeneous). The equation of (10.8.2b)
motion of the segment MN
of the string
for all values of x (the last initial condi-
can be obtained if we equate the product
tion means that at t 0 the string is =
ma =
a dx {d 2 y/dt 2 ) to the force that motionless: we pull the string away
acts on this segment.
from the position of equilibrium, and
The only type of forces that act on the this is specified by the function / (x),
string is a tensile force; in other words,
and then let it go 1012 ).
here we are dealing only with free
vibrations of the string and disregard 10,12
Instead of (10.8.2b) we could require
all external forces that might act on it. that the velocity v (x, 0) dy/dt)t= 0 °f the =
On each end of the segment there MN
(

string for all xs at initial time t 0 be =

acts a tensile force R that stretches the a given function, that is, v (x, 0 ) g (x), where =
g (x) is a known function (this constitutes
segment along the respective tangent. the general problem for a fixed string; see
The components that are perpendicular Exercise 10.8.2).
 ) a .

10.8 The Vibrations of a String SIT

(10.2.1)
The vibrating equation was string Just as with Eq. (10.2.1),
.

first formulated and solved by Jean its particular solutions can easily be*
Le Rond d’Alembert in 1747 (see Exer- guessed: x e
hx
and 2 X =
e~
hx
X = -
.

cise 10.8.3). But of more importance These solutions yield X' ke hx y =

was the somewhat later solution of this x: = k 2 e hx = k2 X x ,
x; = -ke~ hx ,

equation carried out by Daniel Bernoulli and X; = (-k) e~ 2 hx = k2 X (cf. Sec-

2
in 1753, who reasoned as follows. tion 13.10). An arbitrary linear combi-
Let us seek the particular solution to nation of these solutions,
Eq. (10.8.1) that satisfies the boundary
conditions (10.8.2a) and the second ini-
X - aX + bX 2 = ae hx + be~hx
x y

tial condition (dy/dt) 0 at t 0 = = is a (general) solution to Eq.

also
in the form of a product y x t) (, ,
= (10.8.5) that satisfies any boundary
X {x) T(t) of a function X (x) that conditions: xQ (0) A, X' =X = =
depends only on x and a function X'(0) =
B. (Why?) But the require-
T (Z) that depends only on t. Substitut- ment that conditions (10.8.2a) be
ing this product into Eq. (10.8.1) satisfied leads to the following:
)
yields
X (0) = a + b = 0, X (l) = aehl + be~ = hl
0^

XT" = c*X"T or
,

that is, b = — and a (

e
hl — e~
hl
)
—0,,
(10.8.3) which are satisfied only if a=
where, obviously, X" d X/dx and = 2 2 b = 0. Thus, even the assumption
T" = d 2 T!dt 2
since both ,
and T X that the ratios in (10.8.4) are positive
does not lead to a solution to Eq..
are functions of a single variable, x
(10.8.1) that is not identically zero:
and t, respectively. But since c~ 2 T”/T
depends only on t and X"/X only on
X (x) =0, and this means that
x the fact that these two ratios are
,
y [x, t) = 0.
Finally, suppose the ratios we are
equal means that they are independent
considering here are negative that is^
of x and t that is, ,
,
X7X = c- 2 T"IT 2
0, or = -k <
=z — constant. X"=-k X,
^Y~= —k
'~r^f' (10.8.4) 2 2
. (10.8.6a>

Clearly, the ratios in (10.8.4) are

equal to
if
then X" == 0
zero ,
and
-±.I -=-k
T
2
,
T" = — c2k 2 T (10.8. 6b>

X = ax -f b, which is a linear func-

In contrast to the two previous cases,
tion in variable x. But since, in view
=X here we have meaningful solutions te
of (10.8.2a), (0) (
lX 0 (be- = 10 8 1
cause?/ =X
(x) T (t), where, of course, (
. . .

Clearly, both (10.8.6a) and (10.8. 6b)*

T (t) 0, since otherwise the particu-
are equations of the (10.2.1) type of
lar solution is of no interest at all),
= harmonic oscillations for the functions
the fact that (x) ax Xb means + X= X {x) and T T (t) (the equa- =
that b =
0 (since (0) 0) and X =
al =
0, or a 0 (since =(l) 0). X = tions differ from (10.2.1) only in nota-
tion). From (10.8.6a) it follows that
Thus, we have arrived at a zero solu-
tion to Eq. (10.8.1), which is of no in- X —A cos kx +B sin kx (10.8.7)
terest to anyone.
If the ratios in (10.8.4) are positive, (cf. Section 10.2). Since (0) 0 in X =
say, equal to k 2 ,
then view of the first boundary condition in
(10.8.2a), the coefficient A on the right-
^~ = k 2 ,
that is, X" =k 2
X. (10.8.5) hand side of (10.8.7) is zero: Bx X =
sin kt. Substituting this into the sec- X
This equation is in many ways similar ond boundary condition X (Z) = 0
to the equation of harmonic oscillations yields Bsin(kl) = 0, which for B 0
 — ) 0 x )

378 10 Oscillations

{the case A B = = 0 is, of course, of tions: each fixed point of the string (a
no interest to us) is possible only if fixed value x =
x 0 ) oscillates according
to the simple law
,kl = nn or k = ^j-n, (10.8.8)
,

y{x 0 ) = d n cos (a n Z) (10.8.12)

with n an integer. Thus, the final solu- with an arbitrary amplitude d n —
tion (10.8.7) to Eq. (10.8.6a) assumes d n (x 0 ) = b n sin (rmxjl) but with a
the form fixed frequency co n = (cn/l)n. This val-
ue co n of the frequency depends not
X = Bsin (-55-
*) ,
(10.8.9) only on the physical characteristics of
where B is arbitrary, and n is an arbit- the string (its length Z, the tensile force
rary integer. B, and the linear density a, through
Similarly, the second equation of har- the ratio B/a = c 2 ) but also on an arbit-
monic oscillations, (10.8.6b) (with k rary (but fixed for a particular solu-
^defined in (10.8.8)) admits of the solu- tion) natural number n : the various al-
tion lowed frequencies

T = C cos (ckt) -f D sin (

ckt
0) =^= —c jt
,
(02 = —2cxi
-= 2 co,

— C cos .

co 3 = 3co, ... (10.8.13)

(
10 8 10)
. .

Since form an arithmetic progression — all

frequencies are multiples of the mini-
m *-dT = cnn „ cnn
T 5r r
- Csin .

(—
( , \
mum frequency co.

tions (10.8.11) states that all the seg-

The law of oscilla-

+ -^Z>cos(-2p-f) ,
ments of the string vibrate in unison:
the frequencies co n do not depend on x ,

the second initial condition in (10.8.2b), and the various segments differ only
{dyldt) t===0 0, that is, =r'(0) 0, = in the (smoothly varying) amplitudes
which we assumed to be satisfied, yields dn =dn == b n sin (nnx/l).
(,

D =0. Thus, We
)

have therefore found the solu-

T = C cos (
cnnt / Z) tion (10.8.11) (more precisely, a family
«and, hence, of solutions (10.8.11)) to Eq. (10.8.1)
satisfying the boundary conditions
y — XT = b n sin (-^p- x
)
cos t
)
(10.8.2a) and the second initial condi-
tion in (10.8.2b). But how about the
(
10 8 11 )
. .

first initial condition? It is clear that,

where bn = BC is an arbitrary number, generally speaking, we have no control
and n a natural is number (which must over this condition: although we have
be selected to fix a particular solu-
the arbitrary constant b n which can be
tion). 1013
chosen at will, it is obvious that the
We
have thus arrived at a whole fam- initial shape of the string, y ( x 0) ,
=
ily (10.8.11) of solutions to Eq. a sinusoid rather
b n sin(mtx/l), is
(10.8.1), since the natural number n than an arbitrary curve (with a half-pe-
in (10.8.11) can assume any value. All
riod that is a multiple of Z; see Figure
these solutions are harmonic oscilla-
10.8.2, where n = 3).
10 13
The reader will recall that previous-
-

ly n was assumed to be an integer and not

only a natural number (i.e. a positive integer).
It is clear, however, that a change of sign of n
in (10.8.11) leads only to a change of sign of b n
.and yields no new solutions to Eq. (10.8.1),
while n =
0 corresponds to the trivial zero
solution y == 0 of the equation.
 )

10.8 The Vibrations of a String 379

So how should we proceed if the ini-

tial shape of the string, y {x, 0 ) =
/ (x), is not a sinusoid? The brilliant
idea of D. Bernoulli (which was de-
veloped in the 19th century by
J.B.J. Fourier into a powerful method Figure 10 . 8.3
for solving various problems in mathe-
matical physics; see Section 10.9) con-
sisted in reducing an arbitrary vibra- (
10 8 can be reduced, via a method
. . 1)

tion (a seemingly complicated function °f separation of variables x and t, to

of x and t) of a string to a system of sim- ordinary (i.e. not containing partial
pie harmonic oscillations (or “pendu- derivatives) and simple (i.e. represented
lums”) (10.8.11), namely, D. Bernoulli by elementary functions) equations
suggested employing the superposition (10.8.6a) and (10.8.6b). Of course, the
principle, that is, the fact that any lin- representation (10.8.14) of the solution
ear combination of solutions to Eq. Eq.
(10.8.1) is of value only if
[t reduces to a finite (and preferably
( 10 8 1 )
. is also a solution.
.

Let us set up the sum few) number of terms on the right-hand

side of (10.8.14) and yields a complete
i/=: b sin
A
(
{
— 1 1
cos (
K
CTCt

1
)
>
description of
y = y(x,t).
the sought function

2nx 2cnt Bernoulli put a lot of effort into the

-L&osin ( ) cos ( \ 4- '
* ‘V ”
study of the theory of vibrations
and V ~i ) i !
'

3 of (10.8.14) was well-acquainted with the proces

in- expansion of an arbitrary vibration
ple of the particular solutions (
10 8 11 )
.
to separate harmonics (i.e.
.
sin
to Eq. (10.8.1). Clearly, this sum also harmonic oscillations) : / (£)
satisfies Eq.
both boundary (10.8.1), V ck cos (co k t + <p ft ). Using his
conditions in ( 10 8 2 a) and the second . . k
hat initial condition in ( 10 8 2 b), since each . . perience as a base, he assumed t
ted term in this sum satisfies the equation any function / (x) can be represen
no- and the specified initial and boundary in the form of a sum (10.8.15) of han
14) conditions. nics, that is, that formula ( 10 8 . .

ua- Substituting t 0 into (10.8.14) and = yields the general solution to the eq
the allowing for the first initial condition in lion of string vibrations ( 10 8 1 ); . .

ion ( 10 8 2 b), we get

. . coefficients & 1? b 2 b 3 ... in expans
, ,

ion (10.8.14) are found from condit

ely y (x , 0) = / (x) — b t
sin (10.8.15). This statement was sever
> 0 ). criticized by L. Euler (see p. 0(
im- + & 2 sin
' 1 *
4 . &3 sin |
\ l J
+ . . .,
’
Euler argued that in the case of a s:
pie initial function / (x), say, the
>ne <

we (10.8.15) depicted in Figure 10.8.3, where

xis pull the string away from the x a
thus reducing the initial problem to de-
> a termining from (10.8.15) the coefficients at the string’s middle, x Z/2, tc =
let fixed (but small) height h and then
^i» b 2 b3 ... in the representation
, ,

by it go, the function / {x) is given

(10.8.14) of a solution to Eq. (10.8.1).
of different formulas in the two halves
If this were to be done, it would be a re-
the string:
markable achievement, since it would
mean that
“continuous” vibration a
y = ycan be construct-
(x, t) of a string
ed from a discrete set ( 10 8 11 ) of se- . .

parate harmonic oscillations (“pendu-

[ 6)
lums”) and that the general equation
 x x

380 10 Oscillations-

while formula (10.8.15) gives a single sion in a trigonometric series,

of f (x)
expression for / (x) on the entire range that representing f(x) in the form of a-
is,
from 0 to l. D’Alembert, who agreed sum of harmonics 1, cos x, sin x co& ,

with Euler, also thought it impossible 2x sin 2x cos 3x, sin 3x, etc., with ap-
, ,

to represent an arbitrary function by a propriate numerical coefficients. To*

single formula of the (10.8.15) type (the this end we write
example (10.8.16) of a case that seems
to contradict Bernoulli’s statement was / (x) = -y- + cos x + &i sin x-\-a cos 2x 2
first suggested by d’Alembert). But
Bernoulli did not yield a bit: his expe- +b 2 sin 2x +a 3 cos 3x + b sin 3x +3 . . . r

rience in mechanical phenomena told (10.9.1)

him that any function y (x, 0) could be
represented in the form (10.8.15). This
where the numbers aj 2, a 1? 6 1? a 2r
debate played an extremely important
b2 ,
have yet to be found (the term
. . .

in (10.9.1) that is independent of x

role in clarifying the very notion of a
is written in the form aj 2, instead of
function and in representing functions
a0 only for the sake of convenience).
by trigonometric series, that is, sums ,

First let us note the property of

of harmonics (or harmonic oscillations)
orthogonality of trigonometric functions:
given by formula (10.8.15) (see Section
10.9). 31

^
cos mx cos nx dx
Exercises -jt

jt

10.8.1. Suppose l jt, c =

1, and / (x) = = sin mx sin nx dx = 0
sin 3 x. Find the solution (10.8.14) to Eq. (10.8.1)
that satisfies the conditions (10.8.2a) and - jt
(10.8.2b).
10.8.2. How will D. Bernoulli’s solution
for all natural numbers m and n with
,

of Eq. (10.8.1) change if the initial conditions

m n\
of Eq. (10.8.1) have the form y (x, 0) / (x) = jt

and (
dy (x, t)/dt)
t=Q = g (x) (the initial shape
cos mx s\nnxdx = 0 (10.9.2)
of the string, y (x, 0) = / (a:), and the initial
velocity, v (x, 0) (dy = ( ,
t)/dt)
t= 0 = g ( ),
— jt

are specified, and then the string is released)?

10.8.3. (d’Alembert’s solution of Eq.
for all natural numbers m and n\ and
(10.8.1)). Prove that every function y (x, t) = jt jt

(p (x + ct) (x + —
ct), where cp and ip are
cos 2 mx dx = sin 2 mxdx— n
arbitrary functions of one variable, satisfies j
Eq. (10.8.1). How must cp and be chosen so -jt -jt
that the solution y (x, t) satisfies the boundary
conditions (10.8.2a) and the initial condi- for all natural numbers m. These for-
tions (10.8.2b)? What happens if the boundary mulas follow from the well-known fact
conditions are those specified in (10.8.2a) that
but the initial conditions are y (x 0) / (x) ,
=
and v (x, 0) =
g (x) (cf. Exercise 10.8.2)? cos mx cos nx
A

— ~y [cos (m — n) £-fcos (
m-\-n ) x] y
10.9 Harmonic Analysis. Fourier Series
sin mx sin nx
Suppose that we have an arbitrary func- A

tion fixed on a finite interval; it will

— ~y [cos (m — n) x — cos (m -\-ri) x],
be convenient to assume that this func- (10.9.3)
tion, y = f {x), is fixed on the inter-
val —
K ^jt (see, however, p. 383 and cos mx sin nx
Exercise 10.9.2). The problem consists
in finding the coefficients in the expan-
-=-|-[sin (m-f/?) x — sin (m—n) x)\
 ) —

J0.9 Harmonic Analysis Fourier Series .

381

particular cases of the first two formu-

las in (10.9.3) corresponding to n m= / (x) cos mx dx = a m zt, or
j
are the well-known relationships -jt

mx — — (1 -f-cos 2 mx),
jt
*cos 2
a m ==-^- f (x) cos mxdx,
(10.9.3a) j
sin 2
mx = y \
( 1 — cos 2 mx)
771=1, 2, 3, ... (10.9.6)
(since cos 0 = 1). To verify (10.9.2), Similarly, multiplying both sides of
we need only integrate from —ax to it (10.9.1) by sin mx and integrating
the right-hand sides of all formulas from — Jt to Jt, we get
(10.9.3) and (10.9.3a) employing the jt
rules discussed in Chapter 5.
Suppose that we already have the / (x) sin mxdx — 6 m Jt,
j
representation of y / (r) in the form = -71

{10.9.1). The function / (x) is assumed or

known; we wish to find the coefficients jt

a 0 lt b u a 2 b 2 etc. in the trigonometric

,
ffl ,

series. To find a m , we multiply both

,
bm =— 1 f
\ f (x) sin mxdx,
-jt
sides of (10.9.1) by cos mx and inte-
grate the products from n to n. On the — 7n=l, 2, 3, ... (10.9.6a)
Jt

Finally, simply integrating both sides

left we have j*
/ {x) cos mx dx ;
on the
— and employing
of (10.9.1) from jt to Jt
-Jt
the fact that
right all terms will be equal to zero
jt jt
by virtue of (10.9.2) except
n Jt cos nxdx= sin nxdx — 0
j j
a m cos 2 mx dx = am
j
cos 2 mx dx
j
n
= am n, (10.9.4)
and dx = 2n
j
since all the other terms in the right-
hand side of (10.9.1) after integration yields
transform into
/ (x) dx— ~dx==-Y-x2n,
j j
an cos nx cos mx dx -Jt
S
or

an cos nx cos mx dx = 0,
j a° = dx ' (10.9.7)
"n 1
(10.9.5)

b » sin
si nx cos mx dx which a particular case of formula
is
1
(10.9.6) with m = 0. 10 14 *

The formulas for the coefficients of

=K sin nx cos
jsi mx dx — 0 a trigonometric series simplify when
- Jt the function in question is even or odd.

^at n =0 the first formula in (10.9.5)

For an even function, / ( x) f (x), =
Jt
Jt

mxdx = -y X / (
x sin mx dx= \ f (x) sin mx dx
becomes -y- cos j
j
-Jt
Jt
10 • 14
precisely for this reason
It is we wrote
cos mx dx = Oj . The final result is aj 2 the right-hand side of
in (10.9.1)
^
instead of a 0 .
 — n

382 10 Oscillations

+ J / (x) sin mxdx

o
o

_ j / (
— x) sin — mx) d — x) ( (

-jt

Jt o

+ / (a:) sin mxdx= / (x) sin mx dx

j J
0

Jt

4- / (x) sin mx dx = 0,
^
o

since / {—x) f (x), sin (—mx) = =

—sin x, andj d ( x) dx. ——
Therefore, for an even function we have

/ (x) = +a t
cos x +a 2 cos 2x

+u 3 cos3a:+ .
.
(10.9.8)

where Thus, even functions can be expand-

JT 3T
ed in a trigonometric series (10.9.1)
containing only cosines, while odd func-
«o=4" j
/ =4 1 /(,t) dx '
tions can be expanded in a trigonomet-
-JT 0
ric series containing only sines.
JT
Example 1. Suppose that
am== ~k fl{x) cos mxdx
5
-Jt f ji +x for — jts^x^O,
0
^ ^ 1 jt —x for O^x^ji
== -i-
/ (x) cos mxdx
( j (Figure 10.9.1a; cf. Figure 10.8.3). This
- JT
function is even and therefore can be ,
JT
expanded in a trigonometric series con-
-f / (x) cos mx dx taining only cosines, with coefficients
j j
o
jt jt
jt

= j
/ (x) cos mx dx. (10.9.9) ®« —
~ j f(x)dx = ^~ j (n — x)dx
o o
o

In a similar manner, if we are deal-

/
1 Jtx
£2
---7T- \ I
jt
J2 fa Ji
2
\

ing with an odd function, / ( x) — = 2 I 0 Jt ^ 2 y

— / (x), then
JT

JT
am = (jt — x) cos mx dx
j
am =0 and bm =^ / (x) sin mx dx. o
j
jt
o

(10.9.10) = ( cos mx dx x cos mx dx

j j >
o o
(Prove this!). The result is
jt

/ (x) = sin x + b 2 sin 2x + — ~ [n ^

cos mx dx
& 3 sin 3x + ... (10.9.11) o
 kn ^

10.9 Harmonic Analysis Fourier Series . 38$

n (x), which is depicted by two heavy

g
solid lines:
(x) (4/k) sin x (the
if)! =
dotted curve), {%) (4/k) x —
(sin x +
(sin 3x)/3) (the dashed curve)

= f
k
j
cos mx dx —~ (x sin m)\
and i|) 3 (x)
+
—
(4/k) (sin x
(sin 5x)/5) (the thin solid
(sin3x)/3
curve).
+
o
Up till now we spoke only of func-
tions defined on the interval — k^
4-1
—m \ sin mx dx 1 = —
n J
(
\
—m sin x ^ k. But the case of a function
J
o defined on an arbitrary interval — Z^
1
x ^ introduces nothing new, since-
l
—
mx m 2 cos mx
sin “) obvious that we can always in-
is
m o /
it
troduce a new variable in such a way
0 for m even,
that the new interval will be reduced

{ nm 2
4
for m odd. to the interval from to k (cf. Sec- —
tion 1.7).
The final result is Let us introduce a new variable,
x1 = {nil) x, or x = (Z/k) xx . Then
/(*)=4+4 C0SX +"Jr C0S 3x / (x) can be rewritten thus: / {x ± l/n)
=
fi (
xi)> where the function / x of the new
•cos 5#+ • • • . (10.9.12) independent variable x x = {nil) x is
25ji
defined on the interval xY k. — ^
Figure 10.9.1a presents three func- The expansion (10.9.1), in which x~
tions that successively approximate and f{x) are replaced with x x and

/ (;z), which is depicted by a heavy so- f{x i), generates the following expan-
lid line: the function cp 0 (x) k/2 (the = sion of the initial function y = f {x):~
horizontal dotted line), (p x (x) k/2 = +
(4/k) cos x (the dashed curve), and y = x+«i cos [t x + bi sin (t x ) )
cp 2 {x) k/2 =
(4/k) cos x +
(4/9k) X +
cos 3x (the thin solid curve). +a 2 cos ~~ a: -|- b 2 sin —— a:
( j ( j
Example 2. Suppose that
1 for C x^n,
0
+a 3 cos (-y- x'j
+b 3 sin(-y- x
)
+ • • •»-

8 ( X) — [

—| — K^X < 0 (10.9.14)

| f 0r
with
(Figure 10.9.16). This is an odd func-
tion, and therefore it can be expanded = -f x) cos dx
into a trigonometric function contain-
a-h
j
rt( (~r) >

ing only sines, with coefficients ~j (10.9.15).

sin mx dx = ( — cos mx) '

31 bk = j
sin
(^T~) dx
ran '
0 -l
(see Exercise 10.9.1).
—-
mn
— for m odd, Note also that a function y f (#)' =
{ 0 for m even,
defined, say, in the interval 0 x l ^ ^
can always be continued onto negative'
so that values of x by setting / (— x) f (#),. =
which results in the following represen-
£(*)=-, x 4 /
(sin
.

*+— . sin 3a: sin 5a: . \

tation for f {x):

(10.9.13) f(x) = ’Y + a l
cos
(y a;|'+<i 2 cos ( —x )
Figure 10.9.16 presents three func-
tions that successively approximate +a 3
cos xj -f- . . . r (10.9. 16)>
 » )

384 10 Oscillations

where, obviously,

'
Uk m T j ^ cos (^T~) dx / '

,k = 0, 1, 2, ... . (10.9.17)

'(Why? See Exercise 10.9.1.) The func-

tion / {x) can also be continued onto
negative values of x by assuming that (b)

j (—x) = —
/ (x). This leads to a sine Figure 10.9.2
expansion of / (x):

=b sin ;r)+& 2 sin (-y- s) 10.9.2 we see the (“complete”) functions

f (x) {
(y / {x) and g x (.
of Examples 1 and 2.
Of course, all the representations of
fe sin -)-••• (10.9.18)
3
functions in the form of infinite sums of
trigonometric functions, that is, for-
where
mulas (10.9.1), (10.9.8), (10.9.11),
(10.9.14), (10.9.16), and (10.9.18), or,
= yj* / (*) sin to use a term employed many times, in
0 the form of trigonometric series are ,

- valuable only if the respective series

.ft 1, 2, 3, . . .. (10.9.19)
converge if one is to use the terminology
,

We encountered formula (10.9.18) when of Section 6.3. All this means that a
we studied string vibrations in Section sum of a finite number of terms (pre-
10.8 (see formula (10.8.15) for the func- ferably, a small number of terms) in,
tion y (x, 0 ) / (#)). = say, the series (10.9.1),
Note that the right-hand side of Eq.
(10.9.1) can be understood as a function sn = sn (
x) — yr + a i
cos xJrb 1
sin x
^defined not only on the interval from
— k to k but on the entire number axis, + a cos 2x + &
2 2 sin 2x-\- . . .
-f a n cos nx
.since all trigonometric functions on the + b n sin nx ,
(10.9.20)
right-hand side of (10.9.1), cos x, sin x,
provides a fairly good approximation
cos 2x, sin 2x, cos 3x sin 3x, etc., are ,
to the right-hand side of (10.9.1) and
periodic with the least common period
that, as n 00 the sum s n tends to
‘2k: if cp (x) is any one of the functions ,

considered, then cp (x 2kn) cp (x) + = / (x), so that the greater the n, the more
exact is the approximate equality
for every integral k. For this reason the
:sum of the series on the right-hand side / (x) ~ s n (x). (10.9.21)
of (10.9.1) (we will, as usual, write
In D. Bernoulli with
the debate of
this sum as / (#)) is a periodic function
L. Euler and J. d’Alembert it was
'with a period of 2k, or / (x 2 kn) + = Bernoulli who was right: it has been
j (x) for all integrals ft, since all
established that practically any func-
terms on the right-hand side of (10.9.1) tion (periodic or defined on a finite
retain their values if we replace x
interval) can be expanded in a trigono-
'with x +
2 kn. Similarly, the right-hand
metric series. 10 15 This assertion means
;side of the more general formula
(10.9.14) defines a periodic function 10 15
-
For it to be impossible to expand
with a period of 21. It is even some- a function / (x) in a trigonometric series,
times said that the formulas (10.9.1) / (x) must have on the interval on which it
is defined (or over one period if / (x) is perio-
and (10.9.14) de-fine the expansion of
dic) an infinite number of discontinuities (or
periodic functions with periods 2k and jumps) or an infinite number of maxima and
:2Z, respectively. For instance, in Figure minima.
 1

Harmonic Analysis. Fourier Series 385

10.9

say, each function / {x) that is and sin mx present in the integrands in the
that,
— ^famrdras iui ~u m "cruL uu m
f

defined on the interval from n to n .

Indeed, if m is high, the period of, say,

(or, which is the same, each periodic sin mx proves to be so small that the (smooth)
function / (x) with a period of 2jc) can function / (x) has no time to vary appreciably
over the period. This means that the fraction
be represented in the form (10.9.1),
that is, the approximate equality of / (x) sin mx dx (which yields bm) corre-
^
(10.9.21), with the right-hand side de-
sponding to one period of sin mx will be prac-
fined by (10.9.20), is valid. At points tically zero, since / (x) may be assumed con-
where the function continuous,
f (x) is stant and the values of the integrals corre-
(obviously, sponding to the two half-periods of sin mx
the sums s n are close to f (x)
will almost completely cancel out (since they
the greater the number n, the closer s n have opposite signs and their absolute values
is to / (x)), while at points where f(x) are practically the same).
has a discontinuity, or where it experi- A
more exact calculation shows that, if m
is high, the sum of all the fractions of the in-
ences a jump, so that its value on the
tegrals in the right-hand sides of (10.9.6) and
left, f(x —
0), differs from the value
(10.9.6a) corresponding to separate waves of
on the right, f (x +
0), the sum s n the cosine y —
cos mx and the sine y —
approaches, as n — >• oo, the arithmetic sin mx proves to be extremely small; this en-

mean (1/2) [/ (x — 0) + / {x + 0)] of sures the smallness of the expansion coefficients

in (10.9.1) and, therefore, the convergence of
10 16
these values. The rate
convergence of
the series in (10.9.1). If / (x) is not smooth,
of the sums s n {x) to the function / (x) that is, / (x) suddenly changes its direction at
depends strongly on the behavior of the certain points (see Figure 10.9.2a), the decrease
in the expansion coefficients proves to be not
function: it is the highest in the case of
so rapid, compared to the case where the func-
smooth functions; the presence of dis-
tion is smooth everywhere. The convergence is
continuities in the derivative of the even worse if the function experiences a jump
function (see Example 1) reduces the (see Figure 10.9.26), in the vicinity of which
the function / (x) in no way can be considered
rate of convergence, while if the func-
a constant. Indeed, we see that the coefficients
tion has discontinuities (as in Exam- brn corresponding to the discontinuous func-
ple 2), the rate is reduced still further, tion g (or) defined by (10.9.13) decrease only
which requires using a large number of like i/m as m increases, while in the case of
the continuous function of Example 1 the
terms in the trigonometric series to
coefficients a m in (10.9.12) decrease, roughly
represent the function with a given
speaking, like i/m as m 2,
oo; for a smooth
accuracy. function the rate of decrease of the coefficients
in the trigonometric series corresponding to
Let us explain, without laying claims to this function is even higher.
any mathematical rigor whatsoever, the rea-
sons why the “degree of smoothness” of a func-
tion / (x) affects the rate of convergence of the Let us take up once more the question
trigonomectric series representing the func-
of the famous debate between
tion. It is clear that to be able to discard in
the right-hand side of representation (10.9.1) d’Alembert and Euler, on the one hand,
all terms except a known (small) number we and D. Bernoulli, on the other, concer-
must ensure that the discarded terms rapidly ning the vibrations of a string whose two
decrease, so that even an infinite number of
ends are fixed. Particular solutions
such terms has little effect on the result. And
since the sine and cosine are bounded functions (10.8.11) of this problem were obtained
(neither can exceed unity in absolute value), in 1713-1715 by Brook Taylor (already
the coefficients defined by (10.9.6) and (10.9.6a) mentioned earlier in this book), who
must decrease rapidly as m->oo. But for smooth
functions the decrease of these coefficients
assumed that no other solutions were pos-
follows readily from the oscillatory (with al- sible (this assumption, however, lacked
ternating signs) nature of the functions cos mx foundation). D. Bernoulli’s achieve-
ment lay in the fact that he assumed
10,16 For thej function considered in Ex- that separate vibrations, (10.8.11), can
ample 2 we must assume that g (0—0) = — mix, or that a string can generate tones
and g (0 0) + =
1; the sums s n (x) at x
—0 various frequencies simultaneously.
of
converge, as n-+ oo, to the value
(1/2) [(-1) + 1] = 0. The fundamental tone corresponds to

25-0946
 n

386 10 Oscillations

the first frequency co, while the other fre- quite natural that although many scien-
quencies, 2co, 3co, etc. (see 10.8.13)), tists before Fourier (Euler, D. Ber-
correspond to overtones or higher har-
,
noulli, Gauss, and others) gave exam-
monics. 10 17 It was the richness of the ples of expanding functions into trigo-
physical content of the string vibration nometric series, all such series are
problem provided by the representation today called Fourier series.
(10.8.14) of the function y (x, t) and The procedure by which a function
the complete agreement of the results f(x) is expanded in a Fourier series (or
with the appropriate experimental data in a trigonometric series) is known as
that made Bernoulli so sure of his re- Fourier analysis For example, the ex- .

sults and enabled him to withstand the pansion (10.9.12) of the function / (x)
criticism of his friend Euler (whom he ofExample 1 shows that this function
deeply respected) and of d’Alembert. consists of a constant term Jt/2 and har-
Further developments of Bernoulli’s monics cos x cos 3x, cos 5x, etc. taken
,

method are connected primarily with with coefficients 4/jt, 4/9ji, 4/25 k, etc.,
the name of J.B.J. Fourier, who was the respectively, while the function g (x)
first to give a coherent general theory of Example 2 consists of harmonics
for expanding functions in trigonomet- (4/jt) sin x ,
(4/3jt)sin 3x, (4/5n) sin 5x>
ric series, a theory based on the simple etc. (see the expansion (10.9.13)).
formulas for the expansion coefficient
Fourier analysis of (periodic) functions
obtained above. He also provided many plays an extremely important role in mathe-
examples of expansions of concrete matics and applications (for one, in the study
functions. But even more important of various oscillatory processes); it is often per-
were his applications of such expansions formed automatically via so-called Fourier
analyzers Fourier analysis is sometimes called
to specific problems of mathematical
.

spectrum analysis, which of course stems from

physics, say, the problem of heat pro- the fact that ordinary light consists of waves
pagation. These applications were based with definite wavelengths (or frequencies or
on the same idea as the one put for- periods), and a definite wavelength corresponds
to a definite color in the spectrum. Generally,
ward by Bernoulli: the initial condi-
the expansion of an arbitrary (nonperiod-
tions in the problem, conditions spec- ic) function contains harmonics of the
ified by an arbitrary function / {x), a ((d) cos (at cp (to))+type corresponding to
were first assumed by Fourier to be the entire range of frequencies (0 so that we are ,

forced to replace the Fourier series with the

sinusoidal, that is, he dealt only with
Fourier integral over this range (the case of a
functions of the sin mx or cos mx type. function having a continuous spectrum). Here,
With such simple initial conditions however, we will not dwell on this more com-
solving the problem was relatively easy, plicated question. 10 18 -

while the general solution was then

found via the superposition principle
Exercises
and the possibility of expanding / (#) in
a trigonometric series. And it appears 10.9.1. Prove formulas (10.9.10), (10.9.15),
(10.9.17), and (10.9.19).
10.9.2. Expand a function y / (x) de- =
Already in the fifth century B.C. the
10 * 17 fined in an arbitrary interval a x <: b in a c —
trigonometric series. [Hint. Substitution x
Pythagoreans of ancient Greece studied the
a
relationships between the fundamental tone Xl 4- transforms / (x) into a func-
and the overtones of a vibrating string, which
relationships reduce to simple ratios of the tion /x (x x ) defined on the interval —ft <
lengths of the periods of the corresponding xx < n.]
sinusoidal waves, and the relation of these 10.9.3. Expand the following functions in
integral ratios to the euphony of the sounds trigonometric series: (a) y x2 = — C c
jt x jt;
— < <
,

produced by the string. The results obtained (b) y = Cl x for l x 0 and y =c2x

by the Pythagoreans played an extremely for 0 <= <

^ Z; and (c) y =
sin x for 0 < <
z zi

important role in the formation of their and y 0 for —

belief in the simplicity and cognoscibility of
10 18 The interested reader referred to
the world. They also considered mathematics
- is

the key to studying nature. the book [15].

Chapter 11 The Thermal Motion of Molecules.
The Distribution of Air Density in the Atmosphere

11.1 The Condition for Equilibrium Then instead of p we can simply speak of
in the Atmosphere the density p, since the altitudes h and
Let us consider the question of the law h +Ah are almost the same and p
of distribution of air density in the at- differs very little from the density
mosphere in altitude. It is common p (h) at altitude h. Therefore, assuming
knowledge that at high altitudes the that on the left of (11.1.2) we have —dp
air is less dense and the air pressure is and replacing Ah with dh, we obtain
lower than at sea level. The reason for
the dependence of pressure upon alti- &=-*>• (M-l-3)
tude is obvious. Let us mentally select
volume (altitude Ah base
a cylindrical ,
We have thus obtained a differential
area S
volume SAh). The air in this
,
equation for the dependence of pressure
volume (mean density p, mass pSAfe) is p =
p (h) on height. This equation also
attracted to the earth, that is, experi- involves the air density p.
ences the force of gravity directed down- We will assume that the temperature
ward and equal to mg = pSgAh. How- of the atmosphere is the same at all
ever, this volume does not fall and is altitudes. Actually, the air temperature
in a state of rest forthe reason that at depends on the heat flux from the sun
altitude h it is acted upon from below and the removal of heat due mainly to
by a pressure p (h) that exceeds the heat radiation by the air into outer
pressure from above at altitude h A h, + space or, to be more exact, by the
which pressure is equal to p (h + Ah) water vapor and carbon dioxide in the
(Figure 11.1.1). The pressure on the low- air. A small portion of the solar radia-
er base of the cylinder is Sp ( h ); it tion is absorbed by the upper rarefied
balances the sum of the pressure on the layers of the air, while the larger portion
upper base and the force of gravity: reaches the earth, heats the ground,
and then the ground heats the air.
Sp (h) = Sp{h+ Ah) + pSgAh. This explains the actually rather com-
(
11 1 1 )
. . plicated distribution of temperature in
the atmosphere: at ground level the
Formula (11.1.1) can be rewritten as temperature is known to fluctuate rough-
p (h) —p (h + Ah) = pgAh. (11.1.2) ly between —40 °C to +40 °C depend-
ing on the geographical location and time
We will assume that Ah is very small. of year; at an altitude of about 15 km
the temperature is minimal (about
— 80 °C) and is approximately the same
both in summer and winter round the
globe. At considerable altitudes the
air temperature increases, reaching
+60 °C to +75 °C at altitudes be-
tween 50 and 60 km.
Recent measurements by means of
artificialearth satellites show that at
altitudes of 300 to 1000 km the air den-
sity is low but still is greater than was
earlier thought. As we will see later on,
high air density indicates a very high
Figure 11.1.1 temperature of the air in these upper
25 *
 1

388 11 The Thermal Motion of Molecules

strata. What is more, a substantial with b =

a/m 0 Thus the gas pressure
.

portion of the molecules of oxygen and is directly proportional to the density .

nitrogen break up at these altitudes in- It is easy to find the constant of pro-
to atoms, ions, and electrons. portionality for air at room tempera-
If there were no influx of heat from ture. We
know that the air pressure at
without or any removal of heat, that sea level,is roughly equal to
p0 ,

is, if we were to consider a heat-insulated 10 5 N/m 2 (=

10 5 Pa). The air density p 0
column of air, then the temperature at pressure p Q is equal to 1.3 kg/m 3 ,

throughout the column would eventual- approximately. 11 Substituting p 0 and

ly even out. Below we will consider p 0 into (11.2.1), we get p 0
112
6p 0 = ,

precisely this kind of an idealized case whence

of total equilibrium, both thermal and
mechanical. Thermal equilibrium means
b ~ 10 5 /1.3 ~ 7.7 x 10 4 m /s2 2
.

that the temperature is everywhere Later on we will need not only the
the same and so there are no fluxes of numerical value of b for air at room
heat (if the temperature differed at dis- temperature but also the general expres-
tinct points in the air column, the heat sion of the constant b for any gas and
would move from the hotter points to the any temperature. To this end we take
colder ones, with a resultant flow of advantage of the ideal gas law (some-
heat). Mechanical equilibrium consists times called the Clapeyron ideal gas
in the resultant of all the forces acting law)
on any volume of air chosen in the at- pV = RT , (11.2.2)
mosphere being equal to zero. Here we
have to consider the force of gravity on where V is the volume occupied by one
the air in the volume and the pressure gram-molecule (or simply one mole)
on the entire surface bounding the given of gas, Tabsolute
is temper-
the
volume. ature from absolute zero,
(reckoned
For the pressure distribution that sat- —273.16 K) 11 3 and R is the molar gas
-

isfies Eq. (11.1.3), the atmosphere can constant (sometimes called the univer-
be in a s±ate_ oL rest.. sal gas constant ). We know that at 0 °G
Since we consider altitudes h small by (equal to approximately 273 K on
comparison with the earth’s radius, g the absolute scale) and at atmospheric
(the acceleration of gravity) can be pressure at sea level, p 0 = 10 5 Pa, one
regarded as constant. mole of gas occupies a volume equal to
22.4 liters, or 2.24 x 10“ 2 m 3 ( Avogad -
ro's law), whence 10 5 x 2.24 x 10~ 2 =
11.2 The Relationship Between Density 273 R ,
or
and Pressure R ~ 2
8.3 (N /m ) • m 3
/mol • K
Equation (11.1.3) contains two un- = 8.3 J/moLK.
known quantities: the pressure p and the 11 - 1
This quantity can easily be found
density of air, p. We must therefore experimentally by weighing. A hermetically
start by establishing a relation between sealed vessel of known volume is weighed
them. with and without air (a vacuum pump is used
to evacuate the vessel).
By Boyle's law the product of the 11 2
Observe that b has the dimensions of
•

pressure of a gas and the volume occupied the square of velocity. Actually this quantity
by this gas is constant for a given mass 0 m is closely connected with the velocity of
of the gas and for a given temperature: molecules and sound: the square of the speed
pV =a, where a is a constant. If the
of sound is equal to 1.46 (we will not derive
this relation).
gas density is p, then m = Vp. Hence,
0 11
3 Ordinarily,
•
temperatures on the abso-
V = m 0 l p, and since p = atV we can
, lute scale are measured in kelvins (symbol-
write ized K), after the English scientist Lord Kel-
vin: 20 °G 293 = K
(read: “20 degrees Celsius
p = b Pl (11.2.1) is equal to 293 kelvins”).
 .

11.3 Density Distribution 389

We denote the fractional molecular potential energy of one molecule in the

weight of the gas by M. For hydrogen field ofgravity at temperature T. The
we have M= 2, for helium M= 4, for mean kinetic energy of translational
nitrogen M=28, and for air the mean motion of one molecule is (3/2) kT.
value of M
is 29.4. By definition, V
contains M
grams, or x 10“ 3 kilo- M 11.3 Density Distribution
grams, of substance. This means that
the density p is connected with V by From formula (11.2.1) we find p p/b = .

the relation Putting this ratio into the differential

p - 10" 3 M/V, or V = 10- 3 M/p. equation (11.1.3) for the air pressure,
we get
Here density p is expressed in kg/m 3
and volume V in m 3 Substituting this .

expression for V into (9.2.2), we get

The solution to the equation is p =
P = 10»p^l. (11.2.3) Ce~( f /b >\ where C must be deter-
mined from the initial condition. Let
Comparing this with (11.2.1), we find
p = p 0 at h =
0. Then
that
p = Po e-(g/b)h' (11.3.1)
b = 10*^. (11.2.4)
Dividing both sides of (11.3.1) by 6,

Finally, let us express the pressure in we get

terms of the number of molecules n = p^-(*/*>\ (11.3.2)
p
contained in a unit volume of gas.
We know that one mole of any substance where p 0 is the air density at h 0 (at =
contains about 6 x 10 23 molecules. sea level). From formula (11.3.1) it is
This quantity is known as Avogadro's evident that at altitude b/g above H=
number and is denoted by A. Thus, the sea level the air pressure diminishes by
mass of one molecule (in kilograms) is a factor of e Using H, we can rewrite
.

(11.3.1) and (11.3.2) thus:

m = 4xl0 (11.2.5)
p = pf-'**, p =p 0 ^. V^o?6 }

If one mole of gas occupies volume V ,

then the number of molecules per unit Let us compute H using the formula
volume is n = A/V The gas density H = b/g:
p = nm. The ideal gas law (11.2.2)
yields H ~ (7.7 x 10 4 )/10
RT = 7.7 x 10 3
(m 2 /s 2 )/(m/s 2 ) = 7.7 km.
p — n -j— = nkT,
1 rp

where k is the Boltzmann constant :

(This value of H to a mean
temperature around 20 C)
8
k= 4-
A
~ a
- 1.38 x 10-23 J/K. the altitude is increased in arith-
If
metic progression, the pressure and
The quantity R
refers to a conven- density fall in geometric progression:
tionally chosen amount of substance, one p p 0 and p = p 0 at h — 0; if h
= = H,
mole, and so the dimensions of in- R then p = pje — 0.368 p 0
ana P ^
volve the mole. The quantity k refers 0.368p 0 if h = 2 H, then p =
pje ~ 0.135^ 0 P — ®’~~?P°’
;

to one molecule, and so k has the dimen-

2,
an ^
sions of J/K. The quantity kT has the if h = 3 H, then p = pje cz. 0.05 pQ

dimensions of energy (J). In Section 11.4 and p ^ 0.05p 0 ;

and so on.
it will be shown that in the atmosphere We obtain a formula relating H and
the quantity kT is equal to the mean kT. We know that H~b/g. Using
 h c

390 11 The Thermal Motion of Molecules

(11.2. 4) and (11.2.5), we arrive at (11.3.3) and substituting z for hlH,

that is dh — Hdz we get
RT RT kT ,

H= 10 3
Mg Amg ’
or H mg *
oo oo

hp (h) dh = hp0e~ h/ H dh
(11.3.4) j j
0 0
Knowing the dependence of density oo

on altitude, we can express the total = p H2 0

z
ze~ dz =p # 0
2

mass of air, m
a in a column with base ,
j
o
area of 1 2
Indeed, m .

oo
( since via the formula of integration
oo

tn a = pdh = p 0 e~
h l H dh.
j j by parts, ze~
z
dz = — ze 2
+
o o j
o
Make the change of variable z — hlH.
loo
Then dz =
(i/H) dh and
J
e
-z
dz= (— — e~ z
) =1; compare
oo
o

ma = p H j 0 e~
z
dz= — p He' 0
z
\™ =p 0 H. with formula (7.6.19) . The final result
j
0
is
Using the relation m a = p 0 we can H h=^-=H. (11.3.5)
compute H
once again (by way of a R Po
check). Since the atmospheric pressure at Thus, the altitude at which the den- H
sea level is approximately 10 5 Pa = sity and the pressure of the air diminish
10 5 N/m 2 the mass of air in a column
, e-fold is at the same time the mean
with base area of 1 m 2 presses against the altitude at which the air is located.
earth with a force of 10 5 N, that is, this A
similar result was obtained earlier
mass is approximately 10 5 /g 10 4 kg. ~ when we considered radioactive decay
Thus, ma =
10 4 kg/m 2 . Knowing that (Section 8.1): if the probability of de-
p0
~ 1.3 kg/m
3
we get , cay is co, with dnldt = on and n = —
^ 0 e _cd< ,
then during time t = 1/co
H = J2s. =, i°l ~ 7.7 x 103 m = 7.7 km, the amount of radioactive substance
p 0 1.3 decreases by a factor of e and the mean ,

lifetime of a radioactive atom is equal

in accord with the earlier computation.
to the same quantity:
t x 1/co. = =
Finally, let us find the mean altitude h Remember that the simple depen-
at which the air is located, that is, the dence of density and pressure on alti-
altitude of the center of gravity of a ver- tude, (11.3.3,) refers to the case of a con-
tical cylindrical column of air. So as stant temperature. Actually, the dis-
not to introduce extra quantities, we tribution of density and pressure departs
consider a column of air with base area somewhat from formula (11.3.3) and
of 1 2
m
although it is clear that the
, depends on the time of year and other
altitude of the center of gravity is factors.
independent of the base area of the cyl-
inder. At a height between h and Exercises
h +
dh is a mass of dm p dh. The = 11.3.1. Find the air pressure in a mine at
mean altitude is the following depths: 1 km, 3 km, and 10 km.
11.3.2. Find the dependence of air pres-
oo oo oo
sure on altitude for air temperatures equal to
h =
oo
j
h dm =

Let us find the integral in the numera-

\
hp (
)
dhj
j
0
p (h) dh.
—40 °C and +40 °G.
11.3.3. Suppose the air temperature varies
with altitude by the law dT/dh
where T0 is the air temperature at the earth’s
T 0j

surface and a is a constant coefficient. Find the

= —a

tor. Using the second formula in air pressure as a function of altitude.

 ~ "

11.4 The Molecular Kinetic Theory of Density Distribution 391

11.3.4. We know that under the conditions between two molecules to take place it
of exercise 11.3.3 the constant a is approxi-
-5 -1 is necessary that the trajectory of the
mately equal to 0.037 X 10 cm Using .

the result of exercise 11.3.3, determine the air center of one molecule hit a target of
pressure in a mine at depths of 1 km, 3 km, radius 4 x 10 °
-1
m
about the center of
and 10 km. The temperature at ground level is the other molecule. The area of such
taken to be 0 °C. Compare the results with
those of exercise 11.3.1.
a target is or nr 2 =
5 x 10
_l9 2
m .

This means that over a path length of

1 mgiven molecule will collide with
11.4 The Molecular Kinetic Theory all molecules whose centers lie in
of Density Distribution a cylinder with base area 5 x 10
19 2
m
and altitude 1 m. The volume of such
In the preceding sections we found the
a cylinder is equal to am
3
and the ,
distribution of air density in altitude
number of molecules in it is ncr, where n
under the action of gravity in a state
of equilibrium. We
regarded the air as
is the number of molecules in 1
3
m .

Thus, a molecule experiences no col-

a continuous medium with a given lisions over a path of 1 m. Therefore,
dependence of pressure on density.
the mean distance of free flight between
Now let us take that result and ap- collisions is
proach it from another angle, namely, the
viewpoint of molecular theory. We will
consider the separate molecules and their
l = —~
no
7.5 x 10“8 m.
motion. The idea that matter consists
This quantity is known as the mean
of individual atoms was first expressed
free path .
in ancient Greece. However, the motion
Because of collisions, a molecule
of molecules and its connection with
moves in a polygonal line, but the volume
heat was first examined by the great
of a cylinder formed from polygonal
Russian scholar M.V. Lomonosov, who
lines differs but little from that of a right
is thus the founder of the molecular
cylinder and so our computations re-
kinetic theory.
main valid.
The gaseous state differs from the Actually, one has also to consider
liquid and solid states in that in a gas
the motion of those molecules that are
the molecules may be regarded as inde-
hit in collisions. It can be proved that
pendent and noninteracting. The motion
this circumstance changes but slightly
of molecules in a gas is that of free
the mean free path of a molecule, reduc-
flight by inertia. From time to time
ing it by a factor of only 1.5, so that
the molecules collide. Under ordinary
in what follows we will assume that
conditions, such collisions occur with
extreme frequency and the path lengths
i ~
5 x 10~ m.
8

Molecules have velocities v of the

which molecules traverse between order of 300 to 500 m/s. Hence, the
collisions are extremely small.
mean free path time, that is, the mean
At atmospheric pressure and a tem-
time between collisions, is of the order
perature of 0 °C, 22.4 liters of gas com-
of
prise one mole of a substance, or 6 x
10 23
molecules, while 1 3
of gas con- m
tains n ~23
6 x 10 /2.24 x 10
-2

2.7 x 10
25
molecules. At first glance, the quantities l ~
For our crude purposes we will regard 5 10" 8 m and t
x ~
10 _1 ° s are
molecules as spheres of radius about extremely small. But they have to be
-10
2 x 10 m. 11 4 Then for a collision
*

compared with the size of a molecule,

-10
11 4
whose radius is r 2 x 10 m, and
•
In reality, diatomic molecules, say of
with the duration of the collision itself,
oxygen or nitrogen, are more like pairs
merged spheres, something reminiscent
of
of which is less than r/u 10 13 s. If ~
peanuts (two nuts to a shell). we do that, it will be apparent that the
 n

392 11 The Thermal Motion of Molecules

molecules of a gas collide very rarely: where B is a constant defined by the

at atmospheric pressure, the molecules value of density at sea level (h 0), =
of air spend 99.9% of the time in free
n(0)=Be- u (°V kT .
flight and only 0.1 % of the time in a
state of collision. A
remarkable fact is that the density
Molecular collisions in a gas do not of molecules at a certain altitude is
affect the pressure of the gas and do not only dependent on the potential energy
influence the law of distribution of den- of the molecules at the given site: the
sity of the gas in the atmosphere. Con- mass m of a molecule, the acceleration g
firmation of this fact lies in Boyle’s law of gravity, and the altitude h entered
and the ideal gas law. In Section 11.2 into formula (11.4.2) in exactly the
these laws are written as p nkT. = same combination {mgh) as they entered
The gas pressure depends on the num- into the expression for the potential
ber of molecules per unit volume, but energy u .

the radius r of the molecules and their Let us find the mean value of the
cross section a do not enter into the potential energy of a molecule, or
formula. This means that the quantities
r and a cannot enter into the formula u = mgh — mgh — mgH
for the density distribution in altitude.
(see formula (11.3.5)). Using formu-
Let us rewrite the formula for density
distribution (11.3.2) by expressing b in
la (11.3.4), we get
terms of molecular quantities. Since
b =
10 3 RTIM =
AkT/Am kT/m , = u = mgH = mg = kT.
we can write
Thus, the mean potential energy of one
p =pe 0
~ b
= p e ~ mgfi/kT
0 (
11 4 1 )
. .
molecule is kT.
(compare with (11.3.3) and (11.3.4)).
We have established that the distrib-
ution of air molecules in the atmosphere
Divide both sides of (11.4.1) by m ,
depends on the temperature and on
where m denotes the mass of one mole-
the potential energy of the molecules.
cule. Note that p/m n is the number =
of molecules in unit volume at altitude But for a given mean potential energy u
h while p Jm — n 0 is the number of
,
equal to kT
different molecules have
molecules in unit volume at sea level. different potential energies. In other
Formula (11.4.1) assumes the form words, to a given value of u there cor-
responds a definite distribution of mole-
n = n e~ mgh /hT
0
'
. (11.4.2)
cules in potential energy. Part of the
The quantity mgh is the potential molecules, those below altitude H,
energy of a molecule of mass m located have a potential energy less than kT.
at altitude h if for zero we take the Let us find the ratio of the number of
potential energy of a molecule at sea lev- such molecules to the total number of
el, since the potential energy of a mol- molecules. This ratio is
ecule at sea level, u (0), can be chosen H oo

arbitrarily (see Section 9.2). Then ndhj ndh

u (h) = u (0) mgh whence mgh + ,
= j j

u (h) —
u (0). Formula (11.4.2) can
o
H
o
CO
be rewritten thus:
=n 0 e~ mgh l hT dh j 0 e~ mgh l hT dh.
n e-[*W-u(0)]/*r ^ j
(h) n (0) o o

This the law of distribution in altitude

is Let us evaluate these integrals:
of the number of molecules. can We H
write kT H
it like this:
e -mgh/kT dh — e
~ mgh i hT

n(h) = Be- u (V/ kT ,

rr> £r
0
 -

11.5 The Brownian Movement and Kinetic-Energy Distribution of Molecules 39$

10 3 kg/m 3 ) displaced by the particle is

= T5-(‘ --j£ (1--«-*). half the mass of the particle. Taking this
oo fact into account, we can say that the
f g-mgh/kT — kT Q-mSh/hT |°° — !^L #
force of gravity “acts” only on half of
J mg lO mg the particle’s mass, or approximately
0
-17
6.5 x 10 kg. At room temperature,
Therefore T = 17 °C = 290 K, such particles arc
distributed in height (by formula
n dhj n dk = (1 e~ l
)j
(11.4.2)) in accordance with the law
J J
0 0 6.5 X 10~ 17 X 9.8 ,\
= 1 — e~ l
0.63. 290 X 1.38 X 10~ 23 /
’

To summarize, then, 63% of all the or n =n 0 exp (—1.6 x 10 5 h). Thus,

molecules have a potential energy less the number of particles per unit volume-
than the mean value, while 37% have decreases by a factor of e when the
a potential energy exceeding the mean altitude is increased by (1.6 x 10 5 ) -1
value. It is now easy to calculate that m~ 0.62 x 10~ 5 m.
14% of all molecules have a poten- By observing the distribution, in
tial energy exceeding 2 kT, of all 5% altitude, of particles of known size and
molecules exceeding 3 kT, and so on. density, it is possible to obtain the
Generally speaking, the portion of mo- Boltzmann constant k. On the other
lecules whose potential energy is great- hand, the ideal gas law yields the mag-
er than a given value u equals e~ ! hT
u
.
nitude of i? =kA, after which we can
This conclusion is very general and find Avogadro’s number. This work was
refers to the kinetic energy of molecules, carried out by Einstein and Perrin in
too, which we will study in the section 1903-1907 and served as a crucial exper
below. imental corroboration of the entire
atomic-molecular theory and played a
tremendous part in the development of
11.5 The Brownian Movement physics.
and Kinetic-Energy Distribution There is a constant transformation
of Molecules of energy taking place when molecules
move under the force of gravity: if a mo-
Over a hundred years ago, an English lecule is moving downward at a given
botanist Robert Brown observed the time, potential energy is being convert-
random movement of microscopic par- ed into kinetic energy, while if a mole-
ticles suspended in water or other liq- cule is in motion upward, kinetic
uid. Einstein advanced the idea that this energy is being converted into potential
movement of particles is due to their energy. When a gas is in a state of
thermal agitation. From this the con- equilibrium, that is, the pressure of
clusion was drawn that, for one thing, the gas is balanced by gravity, tho
the particles would not all lie on the gas molecules are actually moving at
bottom of a vessel but would be distrib- random with high speeds. However, if
uted in height by the same law as the we picture to ourselves a horizontal plane
distribution of molecules. in the gas, the number of molecules
If a suspended particle has the shape passing through it in unit time upward
of a sphere of diameter d ~
=
5 x 10 7 m, is equal to the number of molecules
then its volume is nd 3 / 6 6.5 x passing through the plane in the down-
10 2 ° 3
m an(j
? with a density p = ward direction, so that, on the average,
2 x 10 kg/m 3 the mass of a particle
3
the gas is at rest. In the equilibrium
is close to 1.3 x 10" 16 kg. We must also state, the transition of kinetic energy
take into account the Archimedean prin- into potential energy and the transition
ciple. The mass of the water (density of potential energy into kinetic energy
394 11 The Thermal Motion of Molecules

/ where n 0 is the total number of mole-

cules, and m is the mass of one molecule.
Observe that v% +
vy + v\ = v 2 where ,

v is the speed of a molecule. There-

fore, (11.5.1) has, in the exponent, the
quantity (j mv2 /2)/kT which
, is the ra-
(a) (b) (c) tio of kinetic energy to potential ener-
Figure 11.5.1 gy. The mean kinetic energy calculated
on the basis of (11.5.1) turned out to
equal (3/2)kT. For the number of
balance, since the number of molecules molecules n whose kinetic energy ex-
moving up equals the number moving
ceeds the given value .Ewe have a rather
down. unwieldy relationship. True, this com-
It is to be noted that in random mo-
plicated relationship can approximate-
tion the individual (identical) molecules
ly be described by the simple formula
have different velocities, or different
kinetic energies. This is true because n~n e~ E kT
0
/
. (11.5.2)
if two balls having identical speeds col-
This law yields an incorrect value for
lide at an angle, the velocities of the
the mean kinetic energy of the mole-
balls after the collision may differ.
cules:
Figure 11.5.1 illustrates a collision
oo oo
after which one of the balls (on the left)
is brought to a halt, while the other Ekl n=~ ndE= e~ E lhT dE = kT
j j
one, moving upward, shoots off with 0 0
double energy (Figure 11.5.1a depicts
instead of (3/2) kT. It gives perceptible
the balls prior to collision, Fig-
departures from the true value if E
ure 11.5.16 in collision, and Fig-
is of the order of kT. However, when
ure 11.5.1c after collision). Note the
E^> kT the divergence between the
,
positions of the balls at the instant of
exact and the approximate law is not
collision: if the second ball were, at
essential.
collision, located below the first one,
It will be noted that for the same
then it would stop and give up all its
temperature, molecules with different
energy to the first ball.
masses have the same mean kinetic
Since in molecular motion there is
energies and have the same distribution
a mutual conversion of kinetic and
as to the magnitude of kinetic energy,
potential energy, it is natural to suppose
since the mean velocity of a molecule
that the distribution of the molecules
as to kinetic energy is similar to that
is proportional to 1/]/ m, where m is
with respect to potential energy. the mass of a molecule.
We give without proof the results of Considering the collisions of mole-
computations carried out at the end of cules against the walls of the containing
vessel, we can find the gas pressure to be
the 19th century by Maxwell and Boltz-
mann. The number of molecules having
velocity components along the x axis P ~
hrtcwvmrc/ x ~dirfi 'o x V
&u x uVung^ue y a'x’is ,

between v y and u y +
dv y and along the z ,
Putting E k]n = (3/2) kT, we get the
axis between vz and vz dvz is equal to + ideal gas law

= p — n 0 kT.
dn
(2nkT/m) 3 l 2
Mutualcollisions of molecules give
m vx + vl + v
exp
( l)
dv x dv y dv z risenot only to an exchange of kinetic
2 kT
,
J energy between molecules but also
(11.5.1) to a conversion of the kinetic energy
 “

11.6 Rates of Chemical Reactions 395

of motion of the molecules into the a reaction to occur. This is reminiscent

energy of rotation of a molecule and in- of the Greek hero Achilles who was
to the energy of vibrations of the atoms vulnerable only in the heel.
of the molecule, which is to say, into A proper explanation was finally
the internal energy of the molecule. given at the end of the 19th century by
The converse is also possible: in a col- the Swedish scientist Svant e Arrhenius.
lision, part of the internal energy of It is this: reactions are initiated only
molecules is transformed into kinetic by collisions of molecules whose energy
energy. It is therefore natural that the exceeds a definite value, the so-called
distribution of molecules as to their activation energy Ea .

internal energy W
also obeys the law of For instance, when molecules of hy-
proportionality to the quantity e~ w/kT . drogen and iodine collide, they form two
The fact that the number of particles molecules of hydrogen iodine HI, the
with a given energy is an exponential energy of the colliding molecules must
~
function of the energy is a universal exceed 3 x 10 19 J. Compare this with
law of nature. the fact that at 0 °C the magnitude
of kT is 1.38 x 10 -23 x 273 ~ 3.8 x
10“ 21 J. This means that at room
11.6 Rates of Chemical Reactions
temperature (or at 0 °C, which is of the
Of what use is the law of distribution of same order of magnitude) only a minute
molecules as to kinetic energy? Such fraction of the molecules possess the
important characteristics of a gas as the needed energy, a = ~
e v ,
where v =
pressure it exerts on the walls of the 3 x 10
-19
/3.8 x 10
-21 ~ 80, whence
containing vessel, its heat capacity, a ~1 / 10
35
.

and the total reserve of energy in the

J
We get the reaction time by multi-
'vokiriire xfi ?ne gas are iidhned 'ny mean plying trie time between collisions \n
quantities, which is to say, they are is of the order of 10
10
s) by the mean
defined by the bulk of the molecules number of the collisions among which
whose energy is close to the mean value. there will be one collision involving
For example, why do we have to know the required energy. This mean number
that a minute portion (of the order of of collisions is of the order of 1/a ~
0.00001%) of the molecules have kinet- 10 35 . We
obtain the reaction time
ic energy exceeding 17 kT ? These sepa- at 0 °C of the order of 10 25 s 3 x~
rate molecules with very large energies 10 17 years. This result accords with
have practically no perceptible effect the fact that at 0 °C the reaction 2 H +
on the pressure and the general supply 1 2 =2HI is practically unobser-
of energy of the gas. vable.
However, the picture changes drasti- From the reasoning given above it
cally if we consider chemical reactions. follows that, depending on the tem-
It turns out that precisely these rare perature, the reaction time is expressed
molecules with high energy completely by the formula
determine the course of chemical reac-
tions. The mystery of chemical reac- t = xe E*' hT ,
tions stems from the fact that molecules
entering into a reaction collide every where t is the time between two colli-
10~ 10 second, whereas a reaction fre-
sions, and E a is the activation energy.
quently requires several minutes (some- This formula gives a true picture of the
times hours). Which means that only dependence of the rate of chemical
an extremely small portion of all colli- reactions on the temperature. A charac-
sions result in a chemical reaction. teristic feature of this formula is the
The idea was advanced that molecules extremely sharp decrease in reaction
have a certain very small “sensitive time and increase in reaction rate for
spot” that must be touched in order for slight variations in temperature.
 396 11 The Thermal Motion of Molecules

However, it frequently happens that cule, this comes out to

116
Q = 18 x
chemical reactions are much more in- 2260/6 x 10 23 - 7 x 10-20 But J.
volved because they may proceed via di- at T — 0 °C = 273 K we have kT
verse intermediate stages. By way of an 273 x 1.38 x 10~ 23 J ~ 3.8 x
illustration, let us examine the reaction 10~ 21 J, QlkT ~ 20. Only
whence
H + Cl 2 - 2HC1. those molecules can tear away from the
2
surface and evaporate whose energy
This reaction proceeds not via collisions exceeds the evaporation heat Q. The
of molecules of hydrogen and a molecule fraction of such molecules is equal to
of chlorine but by the scheme e -Q/ hT Therefore, the rate of evapora-
.

tion is also proportional to e ~^ hT For

H +
2 Cl 2 - H + 2 Cl + Cl,
.

computational convenience it is com-

Cl + H = HC1 + H,
2
mon practice to multiply the numera-
H + Cl = 2 HC1 + Cl. tor and the denominator of the expres-
As a result the actually observed reac- sion QlkT by Avogadro’s number A:
tion rate involves complicated rela- QA QA
Q __
tionships. However, for each separate
kT ~ kAT RT *

reaction, say for

Cl + H = HC1 + H (in the reac- The quantity QA is the evaporation
2
heat of 6 x 10 23 molecules, which is to
tion H + Cl = 2HC1),
2 2 say the evaporation heat of one mole of
the Arrhenius law holds true, and the substance. The quantity kA R is the =
reaction rate is proportional to e~ E * lkT ,
molar gas constant: R 8.3 J/mol-K ~
Ea having (see Section 11.2). The evaporation heat
with the activation energy
different values for each reaction. of one mole of water is equal to Q m ~
The Soviet scientist Academician 18 x 2260 40 000 ~
J/mol. Thus,
N.N. Semyonov made a thorough in- the rate of the evaporation of water is
vestigation of complex (chain) chemical proportional to e -40000/8 3r ^-sooo/r. *
~
reactions and elucidated the laws govern- Let us consider the saturated vapor
ing the course of such reactions and above a water surface. If the vapor is
the general causes that lead to such com- saturated, the number of molecules of
plicated reaction schemes. water evaporating per unit time is equal
to the number of molecules in the
vapor and adhering to the surface of
11.7 Evaporation. The Emission the water (condensing) in unit time.
Current of a Cathode The rate of evaporation is Ce~Q™l RT 7
The idea of Svante Arrhenius concern- where C is a constant proportional to
ing the role of a small number of the area of the water surface. The rate
molecules whose energy greatly exceeds condensation is proportional to the
the mean value of energy is helpful pressure of water vapor and to the sur-
in analyzing not only chemical reactions ^ ace area Hence, in the case of saturat- *

but also a series of other phenomena vapor, when the rates of evaporation
including the evaporation of a liquid. an(^ condensation are equal,
Evaporation requires the expenditure _
* bi‘ a^bonsraeradie Amount oi ^nerg$: °x/p ce ^ = ,

For example, the evaporation of 1 gram

of w ater at 100°C requires the consump-
T where D and C are quantities propo
>» tion of about 2260 J. 11 5 Per mole- tional to the surface area and on
-

g slightly dependent on the temperatu

11 5
The evaporation heat is but slightly
•

dependent on temperature; for water Q =

of 2260 J/g at 100 °Cand 2500 J/gat0°C. We 11 6
Water has a molecular weight
•

23
henceforth disregard this dependence. 18 and Avogadro’s number is equal to 6 X 10
11.7 Evaporation. The Emission Current of a Cathode 397

and totally independent of the pressure, (this electron-ejection energy can be

-whence determined by other methods).
An electron tube offers a marvelous
Q /Br
p = Fe~ m ,
method for measuring the distribution
of electrons leaving the surface of a cath-
ode in accordance with their speeds at
where the constant F does not depend
on the surface area of the water. Thus
a given temperature. When the cathode
a relationship is established between
isheated, we will apply a small negative
potential cp to the anode. With this
the pressure of saturated vapor and
potential, the anode will repulse the
evaporation heat.
electrons ejected by the cathode. For
Let us consider yet another process
this reason, a large portion of the elec-
similar to evaporation, that of the
trons will not reach the anode and will
emission of electrons from a heated sur-
fall back onto the cathode. However,
face. This process occurs on the cathode
there will be some electrons that will
of an electron tube. A cold cathode in
reach the anode over the repulsive force.
vacuum does not emit electrons. 11 7 *

For this to occur, the kinetic energy of

But at high temperatures the cathode
the electron ejected from the cathode
does emit electrons. Then if the anode
must exceed the difference in potential
(also called plate) has a sufficiently
energy of the anode and cathode, that
high positive potential, it will attract
is the quantity ecp, where the Greek
the electrons and each electron torn out
letter e (epsilon) stands for the electron
of the surface of the cathode will fall
charge (unfortunately, the letter e
onto the anode. The electric current
stands here for the base of the natural
flowing in a circuit through an electron
logarithms). The fraction of such elec-
tube is equal to the product of the
trons is equal to e~ s ^l hT Thus, for
number of released by the
electrons
.

a negative potenial cp across the anode,

cathode in unit time into the magni-
tude of the charge of a single electron.
the current is equal to j =
/ 0 e~
s<P/ feT
,

where j 0 is the current for a positive

Experiments show that in these con-
potential. In this experiment, it is
ditions the following relationship exists
necessary that the distance between the
between current j and temperature T:
cathode and anode be small so that the
number of electrons between them
j — ge~Q/ hT .
should not be great and the mutual
repulsion of electrons should not affect
The value of Q differs for different the result of the experiment.
cathodes. For instance, for a cathode The Soviet scientist Academician
made of pure tungsten, Q/k 55 000 °C = A.F. Ioffe proposed using this phenom-
while for barium oxide, Qlk = enon for direct conversion of thermal
30 000 °C and, hence, such a cathode energy into electric energy. If electrons
can operate at lower temperatures. go from the cathode to a negatively
7
Using the dependence of / on T we can ,
charged anode, such a system is a source
determine Q/k Here the quantity Q
.
, of voltage: the current in an external
which enters into the latter formula, circuit between the negatively charged
coincides with the energy necessary for anode and the positive cathode is in
tearing an electron out of the cathode
a direction such that it performs work.
This method of obtaining electric cur-
11 7 rent is remarkable in that there are
•
Here we do not consider the case of
*

a very strong electric field (10 6 V/cm and more) no moving parts and the circuit is fun-
capable of tearnig electrons even out of a cold
damentally simple. In this respect it
cathode. Neither do we discuss the ejection
of electrons from a cathode by the action resembles the generation of electric
of light or bombardment of a cathode by power by means of thermoelectric cells,
electrons, ions, or other particles. which was also proposed by Academi-
398 11 The Thermal Motion of Molecules

cian Ioffe. At present semiconductors, the heated cathode, which emits elec-
which contain free electrons already at trons according to the above-described
room temperature, have successfully law, is still widely used in cathode-
replaced electron tubes in radio and TV ray tubes.
sets and in electronic computers. But
 )

Chapter 12 Absorption and Emission of Light.

Lasers

12.1 Absorption of Light: of the cross sections of all particles in

Statement of the Problem and this cylinder be equal to S.
a Rough Estimate What is the physical meaning of the
condition thus posed? If it were possi-
Let us consider the absorption of light ble to arrange the particles so that the
in air containing black particles of areas covered by various particles do
soot. Suppose a unit volume contains N not overlap, then using the particles
particles. The area of a section of one in the cylinder of height L and base
particle by a plane perpendicular to the area S it would be possible to cover the
ray of light is denoted by cr. For short, whole area of the cylinder and achieve
we call a the cross section For example, . a complete absorption of all the light.
for a particle in the shape of a sphere of For x <
L, total absorption of the
radius r, cr is the area of a cross section light is clearly impossible: no matter
passing through the center of the sphere, how the particles of soot are placed,
or cr = nr 2 .
12 1
the total area of their cross sections
We
will assume that the light inci- does not suffice to cover the whole base
dent on the surface of the particle of of the cylinder. In other words, there is
soot is completely absorbed. The prob- not enough soot to ensure total absorp-
lem consists in determining the portion tion of light.
of absorbed light and the portion of It is clear that at x L or even for =
transmitted light as a function of the x >L there will not really be com-
quantities N, o', and the path length x plete absorption. For a random arrange-
that a light ray traverses through air ment of soot particles and for arbitrary x ,

containing the soot. there will remain certain directions

Webegin with the roughest kind of along which there will not be a single
estimate of the distance over which an particle in the path of the light, which
appreciable portion of light is absorbed. will then pass through.
We denote this distance by L. Just In the volume SL of the cylinder
what the pregnant expression “appreci- there are NSL
particles, the sum of the
able portion of light” means will be ex- cross sections of which is aNSL and ,

amined later on in the sections that fol- so we require that aNSL £, or =

low. Let us not be upset by the clumsy
statement of the problem, since such
L = 1/oN. (12.1.1)

a situation is quite common in physical Let us verify the dimensions in

problems where time and again refine- (12.1.1): a is the area, so its dimensions
ment of a problem is postponed and is are 2
m
and N
is the number of particles
,

carried out in the process of solution. per unit volume and has the dimensions
Consider a cylinder with base area S of m“ 3 . Consequently [L] = m" 2
x
and height L We
require that the sum
. m +3 = m, as required.
The energy transmitted through an
area in 1 second is called radiant flux.
12 - 1
An exact definition of the cross section
o for a particle of intricate shape is this: Let I be the radiant flux through an area
a is the mean ( average area of the shadow of 1 square meter. This is called the ener-
cast by a particle on a surface perpendicular
to a ray of light. The mean area is determined
gy flux or irradiance and its dimen-
, ,

from measurements of the shadow area at

sions are J -m^s’ 1 or 2
Below we ,
W/m .

different orientations of the soot particles in consider the energy flux of light I (x)
relation to the light ray, and the plane on as a function of the thickness x of a
which the shadow from the particle falls is layer. Clearly,
assumed to be perpendicular to the ray in
all measurements. / (x) = I 0f (*), (12.1.2)
400 12 Absorption and Emission of Light Lasers .

where / 0 is the energy flux of the in- sorbed by the soot particles, / would be
cident light, and / (x) is the desired constant.
function that characterizes the attenu- The layer under consideration con-
ation of the light. tains N
dx particles covering an area of
What can we say about the proper- oN dx of the total area of 1 2 of the base m
ties of the function / (#)? If x = 0, of the layer (we assume that the sha-
there is no attenuation of light, I (0) = dows from these particle do not overlap,
/ 0 and so / (0)
,
1. If x = 0, then > which is quite reasonable if dx is small).
the light is attenuated, I {x) <C / 0 ,
Hence, the layer absorbs a fraction
and therefore / (x) <
1. Clearly, / (x) oN dx of the energy incident on the lay-
•decreases with increasing x and ap- er: dQ =
INodx When light passes .

proaches zero, that is, / {x) is a decreasing through the layer dx the radiant flux is ,

function.-.. Thu i
7
eriv a live „ is -.nega- aitennaied^b^ an„aniouu.t.ftcja iaLt.o v t.bn^ 1

tive, dfldx < 0. quantity of absorbed energy dQ Prior .

We
have already said that there will to entry into the layer, the energy
not be complete absorption either for flux was I (x ) after emergency from the
,

x = L or for x >
L, and so we do not layer, it became / (x + dx), and so
•expect / (x) to vanish when x = L.
I (x) — I (x + dx) = IaN dx. (12.2.1)
However, we may assume that the value
x = L is a characteristic length. This But I (x dx) + — I (x) ~ dl ,
whence
means that when light is transmitted (12.1.1) yields
over a path x L, the fraction of
absorbed light is extremely small when iL ^ -I No. (12.2.2)
compared with the fraction of trans-
mitted light. Over a path x ~ L a per- The solution of this differential equa-
ceptible portion of the light is absorbed, tion, as we know, is

and over a path x^> L most of the ,

I=I e-oNx
0 (12.2.3)
light is absorbed and only a very small
portion is transmitted. (compare with formulas (6.6.10) and
As may be seen from formula (12.1.2), (6.6.9) and see Chapter 8). Here I 0 =
the function / {x) is dimensionless. We I (0) is the value of the energy
can assume that if a dimensionless vari- flux at x = 0.
able x!L is introduced, then the func- If the layer thickness is increased in
tion / (xlL) will always be the same for arithmetic progression, x x a, x 2 = =
any kind of particles, for any N and cr. 2a, xs 3 a, =
etc., the energy flux
These suppositions will be corroborated decreases in geometric progression. In-
and made precise in the sections that deed, if we introduce the notation a =
follow. e -oNa (where, of course, 0 <C
a < 1), we find using (12.2.3), that
=
I (x± ) =
I 0 a, I (x 2 ) / 0 a 2 I (* 8 ) = ,

3
12.2 The Absorption Equation /0 a ,
etc.

and Its Solution

12.3 The Relationship Between Exact
We conduct calculations for a col-
all
and Approximate Absorption
umn of air in the form of a cylinder with Calculations
base area 1 2
m
(in the preceding section,
when we considered a cylinder with base It will be highly instructive to com-
area S 2
m
the quantity S canceled
,
pare the exact solution (Section 12.2)
out anyway in (12.1.1)). consider We and the rough estimate (Section 12.1).
a thin layer of air between x and x dx + Such a comparison will help us to make
in the cylinder. A beam of light consists use of rough estimates in complicated
of parallel rays and is characterized by problems where an exact solution is
the energy flux /. If no light were ab- hard to find. Also such a comparison
 , )

7 2.3 The Relationship^ Between Exact and Approximate Absorption Calculations 401

helps one to understand the range of (12.3.2), the tangent line intersects the x
applicability of a rough solution (see axis at x =
L. Therefore, if absorption
Section 10.6). were to occur at the same rate, that
In the rough solution we found the is, so that the same amount of light
distance over which appreciable absorp- is absorbed on every unit of length
tion takes place, L 1/iVor. With the = (precisely, the amount absorbed on the
aid of the characteristic length L, the initial path), all the light would be
exact solution (12.2.3) can be absorbed over the distance x =
L.
expressed as follows: To summarize, then, the quantity L,
which was obtained via rough consider-
/ = / 0 *-*/l (12.3.1)
ations, is indeed of extreme impor-
tance in the exact solution as well.
Thus the supposition that the quanti-
Note once more the importance in
ty L, found in a crude chain of reason-
the practical work of a physicist or
ing, enters into the exact solution is
engineer of knowing how to find and
fully corroborated.
apply approximate, rough, solutions,
The exact solution is indeed of the
and one should take every opportunity
form
to develop skill in finding and under-
I = Iof (xl L ) standing such solutions. This is far
more important and fruitful than mali-
where (cf. (12.3.1)) f (
x!L — e~ x l L.
cious snickering over the drawbacks
We consider the distance x = L.
of rough solutions. We will be pleased
Approximate reasoning gave complete that a very rough solution yields 100%
absorption of light over this distance.
absorption where the exact solution
Actually, from the exact solution
yields 63 % the error is only by a factor
(12.3.1), putting x L, we find that = ,

of 1.5. The rough solution, for x = L,

I =
I^e~ x ~
0.37 7 0 which means that ,
yields 0% light transmission instead
37% of the light is transmitted through of the exact value of 37%, that is, the
a cylinder of length L and, hence,
approximate approach diminishes the
the absorption is 63%. For small x!L
transmission of light by a factor of oo,
we express e~ x ! L by means of the but that is not so bad either because
approximate formula e~ u 1 u (see ~ — from the very start it was evident that
Section 4.8), confining ourselves to
we could not expect good accuracy from
two terms of a series expansion of e~ u rough solution and the 0 % transmis-
a
(cf. Section 6.1) to get
sion of light must be understood only as
a hint that absorption is considerable.
e
- x /L ~ — xlL.
1 (12.3.2)
If it has been established that a prob-
Geometrically, this tantamount to is
lem does not have an exact solution in
the form of an explicit formula, one
replacing the curve y = e~ x > L by the
= should not be deterred in the least.
tangent to the curve at point x 0
(Figure 12.3.1). As can be seen from Seek only for a very rough solution of
the problem. But when using it, be
sure to remember that the solution is
a rough one, an approximate one, and
by no means an exact one.
Let us again dwell for a moment on
the question of dimensions. W
e have
verified the dimensions of L =
1 INo
and have established that this is length.
It is often possible to find an approxi-
mate expression for the quantity that in-
terests us when all we know are its
Figure 12.3.1 dimensions and the dimensions of the
26-0946
 : ~

402 12 Absorption and Emission of Light. Lasers'

initial quantities given in the state- cross-sectional area

12 2
ment of the problem. In the case at
-

9qS 2
hand, however, this is not possible. q(x) = 2

Indeed, a quantity having the dimen-

Si +s 2 5 •

sions of length can be constructed by In calculating the variation of q over

proceeding solely from the concentra- the length it is essential to assume
tion N (m -3 ), namely, Z x =N
_1 / 3
. (The that the particles occurring within the
quantity l x is the mean distance between layer between x and x dx are distrib +
the particles.) A quantity having the uted in this layer randomly that is, ,

dimensions of length can also be con- do not prefer illuminated or shadowed

structed out of the cross section a(m 2 ), areas.
namely, l 2 = cx.
1/2
(The quantity l 2 The reader
advised to investigate
is

characterizes the mean particle size.) the situationwhich the particles

in
It is then obvious that the quantity form an ordered array in the air like
l = ZfZ
1-0*
for any value of the expo- atoms in a crystal and how this in-
a
nent a, also has the dimensions of fluences q (x).
length. For instance, the L that interests
us is obtained at a = 3. 12.4 The Effective Cross Section
Thus, in the problem at hand, dimen-
sional analysis does not give a definite In the problem of attenuation of light
answer. To find L, that is, the quantity passing through dusty air, the quanti-
with dimensions of length entering into ty o* has a simple geometric meaning
the exact solution, it is precisely the of the area of the shadow cast by a sin-
rough solution to the problem that one, gle dust particle. The law of attenuation
it turns out, has to find. But even when of light (12.2.2) is the same for light of
dimensional reasoning yields a unique different wavelengths (that is, differ
answer, it is also desirable to get a rough ent colors), since a is independent of
solution to the problem so as to obtain the wavelength.
a clearer picture of the phenomenon. Contrary to the above situation, in
Finally, we note that the calculations the absorption of light by separate mo-
carried out in Section 12.2 were based lecules and atoms there is observed
on the assumption that the soot parti- a strong dependence of the law of
cles are distributed randomly for a given attenuation of light upon the wavelength
mean concentration N. We assume that of the light. For example, in clean
at the entrance x= 0) the energy flux q0 is
(,
air at atmospheric pressure, visible
uniform over the entire cross section. But light is hardly at all attenuated (atten-
after the light has traveled a certain dis- uation is less than 1% per kilometer
tance x through the air containing the of path length; accordingly, the atten-
particles, there appear, strictly speak- uation is by a factor of e over a distance
ing, certain parts of the cross section of about 100 km). Ultraviolet rays
(of total area S x ) covered by shadows of wavelength 1.8 x 10 -7 m = 1800 A
from the particles (and there q = 0), (A stands for Angstrom, 1 A = 10 10 m)
while other parts remain unshadowed are attenuated by a factor of e over
(of total area S 2 ) where q = q 0 (here a distance of L = 0.001 m= 0.1 cm.
’S ± S2 =
S, since the total cross- Still shorter ultraviolet rays of wave-
sectional area S remains constant). The length 1.1 x 10" 5 cm == 1100 are A
quantity we are calculating here is the attenuated e times over a path length
energy flux averaged over the entire of L =0.01 cm.
Consequently, the absorption of
light by air is not like the absorption of
12 2
This constitutes the [method of dimen-
•

light by a black dust particle, which

sions (dimensional analysis). The approach
has wide application in physics, mechanics, absorbs light of any wavelength to
and engineering. the same degree.
 ~ " ~

12.5 Attenuation of a Charged-Particle Flux of Alpha and Beta Rays m

ftc
1
°h^U D£>4t K IJ y ° «o mrj-
1
A f
W9 ^ n nt/ic , naUxVj ,
+
hnk i
° ^ infe br y
gle atom
in unit time, q, is proportional rarefied gas an atom excited by light,
to the energy flux I of the light at the as a rule, re-emits the light in an arbit-
point where the atom is located: rary direction, while in a dense gas the
excited atom transmits the energy to
q = al. (12.4.1) other atoms in collisions.
Thus, in a rarefied gas we have light
Here cf is the constant of proportionali- scattering while in a dense gas we have
ty. Let us determine the dimensions of a. true absorption. However, the “fate” of
The dimensions of q are J/s. The dimen- the energy taken away from the atom is
sions of I are J/m 2 -s. Hence, the di- unimportant in calculations of the de-
mensions of o' are m 2 The quantity a . crease in intensity of a directed light ray.
is called the effective cross section. For

a black dust particle, the constant of

proportionality coincides with the geo- 12.5 Attenuation of a Charged-Particle
metric area of the shadow. For mole- Flux of Alpha and Beta Rays
cules and atoms, a is strongly dependent
on the wavelength of the light. The exponential law of diminution of
In rough fashion, we can picture the particle flux as a function of distance,
cause of this dependence as follows.
/ = /^-*/l (12.5.1)
The amount of energy absorbed by an
atom when acted upon by light proves is based on a very general assumption
to be particularly great when the that the attenuation of the flux over
frequency of the light oscillations coin- a small distance dx is proportional to
cides with the frequency of motion of the intensity of the flux:
the electrons in the atom. This is
resonance
tensively
: the
and
electron
absorbs
oscillates
particularly
a
in-
-S-" —r 1 ’ (12.5.2)

large amount of light energy. Such where the constant of proportionality

a resonance is attained, for instance, 1 /jL is dependent solely on the type of
in the absorption by sodium atoms (in particle. Since (12.5.1) is the solution
the vapor state) o of yellow light of to the differential equation (12.5.2), it
wavelength 5890 A 5.89 x 10
-7
m.= is obvious that formulas (12.5.1) and
The very same yellow light is emitted (12.5.2) are equivalent.
by sodium atoms at higher tempera- Experiments have shown that in
tures, when
electron oscillations are certain cases the exponential law (12.5.1)
caused by energetic collisions of atoms is quite exact, but sometimes devia-
among themselves. tions from the law are observed. Let us
At resonance, a reaches 10~ 14 2
m ,
consider carefully the reasons that can
that is, is of the order of magnitude of give rise to deviations from formula
the square of the wavelength. Atoms
(12.5.1) or (which is the same thing)
and molecules are of size 10~ 10 to from (12.5.2).
10 9
m, which corresponds to a cross It easy to answer the question
is
section of the order of 10 20
to 10
18
m 2
.
about the meaning of deviations from for-
Thus, the maximum effective cross mula (12.5.2). Formula (12.5.2) pre-
sections are many times greater than sumes that when x and I vary, the light
the true cross-sectional areas of atoms (or other radiation) under consideration
and molecules. On
the other hand, for the does not vary qualitatively, otherwise
light, whose frequency does not corre- the number L would change. We re-
spond to the natural frequency of the write Eq. (12.5.2) as
atom, the effective cross section is
small, much less than the cross-sec- i_dl__ 1_
tional area of the atom. I dx L
26 *
 404 12 Absorption and Emission of Light. Lasers

From this we see that the quantity

7 -1(dlldx) is constant. If it turns out
that at different points in space this
quantity is different, this means that at
such points not only is the intensity of
radiation different but also its physical
characteristics (say a different color
Figure 12.5.1
light, that is, having a different mean
wavelength).
When considering problems of pro- todo so. Actually, different a-particles
tection against radioactive radiation (having the same initial energy) expe-
and questions of the passage of a-, P-, rience different numbers of collisions—
and y-rays and neutrons through vari- they are not necessarily equal exactly
ous substances, there is a different reason to 100 000. However, since 100 000 is
for departure from the simple law a big number, over a given path length
<12.5.2). the departures of the number of colli-
As applied to the process of light sions of separate a-particles from the
absorption, the law (12.5.2) signifies mean (100 000) are but slight (of the
the following: if the light encounters order of 300, which is about 0.3% of
a dust particle, some passes by the par- the total number of collisions). 12 3 For
-

ticle without any change while the this reason, a-particles of the same
other portion of light is completely energy always lose all their energy
absorbed by the dust particle. The over roughly the same distance. This
situation is more complicated in the path length depends on the initial
case of radioactive radiations: an a-par- energy of the a-particle. If a flux of
ticle is a nucleus of the helium atom a-particles of the same energy is di-
flying out of the radioactive parent rected along the x axis, the relationship
nucleus at high speed (of the order of between the intensity of the flux and
€.07 c, where c is the speed of light, the path length x is shown in Figu-
that is, at a speed of about 2 x 10 7 m/s). re 12.5.1. The curve is quite different
In passing through an atom, the a- from the graph of the exponential func-
particle gives up a small part of its tion. Over a considerable portion of
energy to the electrons. After roughly the path length, the intensity of the
50 000 collisions with atoms, the a- flux of particles does not change: the
particle will have lost half of its energy. same number of a-particles fly through
It will not cease to exist, but its energy an area of 1 cm 2 in the same intervals
and speed will have changed. After of time. Then the intensity falls off
100 000 collisions the a-particle comes sharply. This drastic fall was prepared
to a halt, ceases to collide with atoms over the section where the intensity
and to knock out electrons. The stopped remained constant, because over this
particle removes two electrons from portion the enegy of the a-particles
the surrounding atoms and becomes diminished with increasing path length.
a neutral helium atom at rest (with the The sharp drop in the flux intensity
exception of thermal motion). When
Rutherford found that incidence of an 12 3
•
Here the law of large numbers operates.
a-particle into a vessel with gas is This law deals with a multiple repetition of
accompanied by helium buildup in the random events of the same type, the events
being such that their outcome cannot be pre-
vessel, this fact signified the end of dicted. This law states that in a large number
a very important stage in radioactive of trials the fraction of cases with a definite
studies. outcome differs very little from the expected
(mean) number of such outcomes (this fraction
The number of collisions necessary
necessarily tends to the mean number as the
to stop an a-particle is such that number N of repetitions of the event tends to
it takes only several centimeters of air infinity).
12.6 * Absorption and Emission of Light by a Hot Gas 405

there is a considerable probability (of

the order of 37%) of the passage of
unaltered y-rays and neutrons. This is
why they require thick shielding. A flux
of y-rays and neutrons does not become
zero for a definite thickness of the
protective layer, as was the case for a-
and p-particles. As experiments and
Figure 12.5.2 complicated calculations show, for a
given large thickness of the protective
occurs where the energy of the a-parti- layer, a flux of y-rays and neutrons falls
cles becomes extremely small. off in rough accord with the exponential
The picture is similar in the case of law, with the exponent often differing
fast electrons ((3-particles emitted when from the one corresponding to the initial
a neutron is converted into a proton in section of the curve.
the nucleus of an atom). Here the situ-
ation is complicated by the fact that
12.6* Absorption and Emission
in radioactive decay electrons with
of Light by a Hot Gas
different velocities are emitted; what
is more, the electrons give up part of Let us imagine a hot gas through which
their energy to the atom near which a beam of light is passing. We wish to
they fly and also experience a consid- study in greater detail the effect that
erable lateral deviation. light propagation has on the atoms of the
The curve for the flux intensity of gas, say, sodium atoms. More precisely,
p-particles as a function of path length we wish to know what happens to
is of the shape shown in Figure 12.5.2. the atoms when they absorb or emit
We see that it differs greatly from the light and when they collide with other
one in Figure 12.5.1. Even for small atoms and molecules, bearing in mind
x's some of the electrons fall out of the that the temperature of the gas is high.
beam. These are mostly electrons which First we must note that the assump-
had low initial velocities. Therefore, tion that an atom is like a small ball on
near x =
0 the behavior of the curve is a spring, or an oscillator (a device
similar to that of the exponential func- with a definite natural frequency), in
tion. Further on, however, the curve many cases is not sufficiently accurate.
reaches the x axis, the intensity I be- An atom is a system consisting of
comes zero for a very definite value of x a nucleus and electrons that obeys the
corresponding to the maximum energy laws of quantum mechanics These laws
.

of the electrons generated in the given cannot be discussed here;, suffice it to

type of radioactive decay. say that a quantum mechanical system
The most important practical prob- of this type may be only in certain
lems (they are also the most difficult states with definite (or, as the mathe-
ones) are those connected with protec- matician would say discrete) ener-
tion against y-rays (gamma rays) emit- gy values. The lowest state, which is
ted by radioactive substances and against characterized by the lowest energy pos-
neutrons produced in the fission of sible for the system, is called the
nuclei in nuclear reactors and nuclear ground state ( we will sometimes call it
explosions. The situation here is con- the unexcited, or zeroth, state and
fused and complicated in the extreme label the quantities referring to this
by the fact that y-rays and neutrons state with a subscript “0”: the energy of
give up energy in large portions and the atom in this state is E 0 the num-
,

are strongly deflected from their orig- ber of atoms in the ground state per
inal directions in the process. Even in unit volume is n 0 and so on). In the
,

a thick layer of air (100 to 200 meters) same way we will speak of the first state
406 12 Absorption and Emission of Light. Lasers

(after the ground state) as the first ex- depends on the difference in the energies
cited state and label the respective quan- of the two states of the atom; this
tities with a subscript “1”: the ener- frequency (which can be measured using
gy E 1 the concentration of atoms n ly
,
a spectrometer) characterizes the transi-
etc. The second excited state has ener- tion from one state of an atom to an-
gy E 2 ,
a concentration of atoms n 2 and ,
other. The quantity v, more precisely,
so on. We will always assume that v li0 is not the frequency of electron
,

E 0 E 1 ^E 2 .... vibrations inside the atom in a defi-

At a low temperature all the atoms in nite state (zeroth or first excited) it —
a system have the lowest possible ener- depends on both states. Spectroscopists
gy, that is, are in the ground state: the (scientists dealing with the theory and
energy of each atom is E 0 and the ,
interpretation of interactions between
concentration n 0 of atoms is equal to matter and radiation) noticed this pat-
the concentration n of the entire col- tern long ago, since it is evident from
lection of atoms (in all possible tables of spectral lines in the spectra of
12 4
states). -
various atoms. Indeed, it was found that
An atom in its ground state can
absorb light, going over in the process
V 4,l + V l,0 — V 4,3 + V 3, 0 ==•••-
to another state, say, the first excited.
What does quantum mechanics say (Why? Verify this.) The validity of
about this? In the language of quantum such equalities can be verified without
mechanics, light consists of separate
knowing anything about atoms.
particles, called sometimes light quanta The famous Danish physicist Niels
and sometimes photons and the energy Bohr (1885-1962) was the first to under-
,

stand the true meaning of, and reason

of each such particle (or quantum of
light or photon) is hv where h = for, such a pattern. He formulated the
,

-34 following “quantum postulates”:

6.63 X 10 J-s is Planck's con-
there is a set of states (0, 1,
(1)
stant, and v is the f requency of the light
2, .), called stationary states, in
. .
(number of oscillations per second).
A which the atom does not emit light;
peculiar feature of quantum theory
is that while speaking of particles of (2) emission of light occurs during
the transition of the atom between two
light we do not give up such notions as
stationary states, and the frequency of
the frequency of the light and wave-
the radiation emitted equals the differ-
length, which strictly speaking belong
ence in the energy of the states divided
only to the concept of light as an electro-
magnetic wave.
by Planck’s constant.
Similar statements can be formulated
But no deeper into the
let us go
for light absorption by atoms.
philosophy of quantum mechanics.
Let us now return to the main ques-
Energy conservation states that when
an atom absorbs light (a single photon)
tion of this section how a hot gas ab- —
sorbs and emits light. The reader will
and goes over to its first excited state,
recall that at a low temperature all the
then
atoms in a gas are in their ground state
E0 + hv = Ev (12.6.1) and that light of resonance frequency
(we will again write v instead of v lt0 )
The same holds when an excited atom
is absorbed, so that
emits a photon, going over from the
first excited state to the ground state.
This means that the “resonance” fre- = - on 0
I, (12.6.2)
quency v of the absorbed or emitted light
where I is the light flux and a is the
12 - 4
Strictly speaking, n = n0 + n1 + so-called cross section (see formu-
ji
2 + . . but at a low temperature
la (12.2.2)). What will change if the
U0 n2 ,
etc -> an(l n l + U2 + • * • 72
0>
so that n ~ n0 .
gas is hot?
 —

12.7 * Radiation in Thermodynamic Equilibrium 407

In a hot gas, as a result of the ther- reasoning behind all this is as follows:
mal motion of the atoms and collisions in a layer dx cm thick and a cross-
of the atoms with atoms and molecules sectional area of one square centimeter
of other gases (if we are dealing with there are n x dx excited atoms, with the
a mixture of gases), a fraction of the rest of the reasoning done on dimension-
atoms will transfer to states with al grounds, namely, w has the dimen-
higher energies. In other words, we sions of 1/s s
-1= a is dimensionless,
,

will have excited atoms in addition to and n x has the dimensions of -3

so m ,

those in the ground state, that is, we that if x =

wahv, the dimensions of
can no longer ignore %, n 2 ..Accord- , . . xi2 1 are J/m 3 *s, which is the same as
ing to the general law of thermal mo- the dimensions of dlldx We see that .

tion (see Section 11.4) at a given tem- on dimensional grounds formula (12.6.5)
perature r, we have is correct. And yet it is not correct!
The question marks on both sides of
n0 n i -f- n 2 -r- . . . = e~ E ^ kT -f- e~ El / kT
Eq. (12.6.5) are to remind the reader
-f- e -E 2 /kT+. 9m9m (12.6.3) that this formula is incorrect. We will
see how this formula must be modi-
If the total number of atoms per
fied when we consider the thermody-
unit volume, tz, is fixed, then, combin-
namic equilibrium of light fluxes.
ing (12.6.3) and the fact that n =
n2 + ., we get+ . .

12.7* Radiation in Thermodynamic

-Eo/kT
n ° ==n
e Equilibrium
- E ^ hT
e + e~ E ^ hT + ... Suppose that the vessel containing
=
iJre -{Ei-Eo)/kT^_
~ ’ (12.6.4) the gas has
ideally reflecting walls.
The temperature of the gas is fixed and
is maintained constant.
n€ -(Ei~E 0
)/kT
ni = It is natural to assume that if ini-
l + e-iEi-Eo)/^ _j_ _ ’

tially there was no light in the vessel,

We see that for a fixed n the absorption it will appear as a result of emission
drops as the temperature rises, since by excited atoms, while if initially the
n0 <n. But in addition to this and — intensity of the light was too high, the
this is the most important aspect surplus of light will be absorbed by the
there appear excited atoms in the gas, atoms in the ground state.
atoms that are capable of emitting light. It can be expected that, regardless of
The frequency of the light they emit is the concrete initial state, a certain
the same as that of the light they intensity of light will set in in the ves-
absorbed earlier. sel as a result of absorption and emis-
It may be assumed that in the pre- sion of light by the gas with a fixed tem-
sence of excited atoms the equation perature. More refined reasoning shows
for the intensity of the light is that the intensity of light of a given
frequency depends only on the tem-
? — — onj -f- xn lt ? (12.6.5) perature of the gas. Indeed, imagine two
vessels filled with two different gases
Here the coefficient x must have the but having the same temperature. Next
meaning of the product of the probabi- connect these vessels with a light guide
lity w of an excited atom going over in containing an optical filter that trans-
one second to the ground (zeroth) state mits light only of a definite frequency
and emitting a photon with the probabil- in both directions. If the light intensi-
ity a of the emitted photon joining ties were different, then, in spite of
the light flux. Since I is expressed in having the same temperature, one ves-
units of J/m 2 *s, we must multiply wa sel would give off more energy than it
by the energy hv of one photon. The receives, while the other would receive
408 12 Absorption and Emission of Light. Lasers

more energy than it gives off, so that the we employ (12.7.2):

firstvessel would cool off while the
other would heat up, which is clearly Po + Pi + P + 2 . . . — a (1 + e- h v/* T
impossible. e -2hv/kT _j_
_f_ _ ^ )
Thus, we conclude
that there is 1-e - hv/kl
a certain equilibrium intensity of light, (12.7.4)
an intensity at which there is equilib-
and, hence,
rium between emission and absorption.
This equilibrium intensity at a given a—1 — e~ hv i kT
(12.7.4a)
frequency depends only on the (12.7.4)
note that
inside the parentheses in
temperature of the gas and not on the
the sum of the terms
we have
properties of the gas. It is natural then
of a geometric progression).
to try to find the equilibrium intensity
of light starting directly from the gen-
Now we can find the average number
of photons cp and the average energy
eral principles of thermal motion.
(more precisely, the energy density)
Let us consider radiation (or light) of
e in a given type of rays:
a certain frequency and direction in a
vessel. The energy of this radiation can- Cp — 1 X Pi + 2 X p2 + . . .,
not assume arbitrary values: according (12.7.5)
to quantum theory (in our case, the
theory olt photons), 'the energy may he
a = qhv
c (12.7.5a),

zero (not a single photon of a given (since cp is the average number of pho-
wavelength) or hv (one photon) or 2 hv tons per unit volume, the dimensions
(two photons), and so on. of this quantity are m~ 3 while the di- ,

From general principles, the ratios mensions of 8 are J/m 3 ). Substituting

of the probabilities p 0 p x p 2 of , , ,
• • .
(12.7.3) and (12.7.4a) into (12.7.5) and
the corresponding states are then the result into (12.7.5a), we get

Po -T- Pi +p + =1 -f- e~ hv ' kT -hv/kT a

2 • • .

y = e

_ -hv/ki
= hv/kT_ (12.7.6)
e -2hv/hT^_
’
-L. , (12.7.1) 1 e e !
.

in accordance with the fact that the hv

(12.7.6a)
v/ftn
energies are 0, hv 2hv, and the , . . .

probabilities are proportional to e~E i kT

(here we have left out the simple cal-
(since e° 1, =
the sequence on the
culations and have given the final re-
right-hand side of (12.7.1) starts with
sult 12 5 ). A given
-
frequency v corre-
unity). But the sum of all these prob-
sponds to a certain period 1/v of the
abilities must be equal to unity because
oscillations and a certain wavelength
we have exhausted all the possibili- =
X c/v of the wave.
ties —
the given light ray may be char-
We can now find the number of waves
acterized by one, two, three, etc., or
in a volume V with frequencies ranging
by the absence of photons. This means
that we can normalize, so to say, the
from v to v dv. One must bear in +
mind that electromagnetic waves have
distribution (12.7.1) by adding to these
a certain polarization: the electric field
ratios of probabilities the following:
in a wave is at right angles to the direc-
Po + Pi + Pi + • • •
= 1* (12.7.2)
tion of propagation of the wave (in
which that the sum of the
shows three-dimensional space, however, there
probabilities is unity. From (12.7.1) it are numerous mutually perpendicu-
clearly follows that lar directions that are at right angles

Pn =z ae~ nhv / h2\ n = 0,1,2, ...(12.7.3) 12 - 5

To derive (12.7.6) from (12.7.3),
(12.7.4a), and (12.7.5), it is sufficient to note
(i.e. po = a, px = ae~hv i kT ,
etc.), where that cp— e
~ hv
/ hT cp = Pl +p + 2 pz + . . .=
factor a has yet to be found. To find a, 1-Po (=!-«)• (Why?)
 Q ,

12.7* Radiation in Thermodynamic Equilibrium 40 9*

to a given direction). We must also unit volume, Q. To this end we take the-
allow for the conditions of wave reflec- energy of each oscillation with a defi-
tion from the walls of the vessel, but nite frequency, multiply it by the num-
it has been found that the number of ber of oscillations of the given type r
independent types of oscillations with and add the products (that is, integrate
frequencies within a given range de- them):
pends, to the first approximation, only on oo

the volume V and is proportional to hv 8jiv 2 dx

this volume. Thus, we can speak of .( e
hx/kT ^ c3
0
a certain number dK
oscillations of
within a frequency range dv per unit (12.7.9)-
volume; this number proves to be has the dimensions of energy density,
dK^ —~ c
,
(12.7.7)
('

J/m 3 ).It is convenient to introduce

the notation hv/kT x and isolate the* —
where c is the speed of light. Since the dimensionless integral thus:
frequency v (and, hence, dv) has the
dimensions of s" 1 and c the dimensions 4

of m/s, the dimensions of dK are m

-3
.
^=-7?-( Jx) J*r=r- (12.7.9a)

A full derivation of (12.7.7) is too

tedious to include in a book for begin- The
integral on the right-hand side-
ners, but it is easy to see that (12.7.7) of (12.7.9a) cannot be expressed in
is reasonable and can easily be under- terms of elementary functions. Ap-
stood. Note that vie MX, whence = proximately, the integral
v = c/X and |
dv |
= (c/X
2
)
dX. Then,
according to (12.7.7), the number of
(12.7.10).
oscillations with wavelengths ranging
0
from X to X +
dX is
can be evaluated if we note that
dK = ^-dX. (12.7.7a)
1 -X 1

This means that the number of oscilla-

ex — l
c
1 — e-x
tions with wavelengths ranging, say, = e~ x (1 + e~x -f e~ 2x -f- . . .)
from X 0 to 2X 0 is
= e~x -f e~2x + . . .

2X {

dX 8n and, hence,
K ^o> 2X 0
Q
8lt
f
T3-=-r(i^-
I 1 1
)
i
Xo
— x 3 e~ x dx -f- x 3 e~ 2X dx -\- ....
In 1 J j j
* (12.7.8)
3 XI (12. 7. 11)*

If we abstract ourselves from the numer- But (see (7.6.19))

ical factor, which differs but little, in oo
order of magnitude, from unity, we can
say that the oscillationsnumber of
x 3 e~ x dx — 3! = 6, x3 e 2X
dx
j j
with wavelengths close to X 0 in a unit o o

volume is approximately equal to the oo

number of oscillations that fit inside = -|r y 3<rV dy = jQ'X 6, etc.,

a cube with an edge X 0 On the average, .
|
o
each oscillation occupies a volume of XI ,
so that
which constitutes a result that not on-
ly can easily be remembered but is also
dx a a t
quite natural. 1 ex
f

(
1+ i6\

Now we can easily write an expres-

sion for the total energy of radiation per (12.7.11a)*
 TK

410 12 Absorption and Emission of Light. Lasers

The series on the right-hand side of emit as much radiation as it absorbs: we

(12.7.11a) is convergent; if we keep conclude, therefore, that a black body
only five terms, we get 6.482 for the with a temperature T emits per unit
-8
approximate value of integral (12.7.10). area per unit time 5.67 X 10 T 4 joules
The theory of functions of a complex of energy ( in kelvins). This quantity
variable (see Section 17.2) yields the characterizes what is called blackbody
exact value of (12.7.10), which is n 4 /15 radiation.
or approximately 6.494; thus, the error The order in which we have discussed
introduced by limiting the sum in all these questions is opposite to the
(12.7.11a) to five terms is smaller than order in which these concepts developed
0 2 %.
. historically. First the law (12.7.12) was
Thus, the exact expression for the discovered in experiments, that is, it
equilibrium radiation density (in J/m 3 ) was established that the total energy
at a given temperature T (in kelvins) is radiated by a black body is proportional
to the fourth power of the temperature of

^ = t?t£-- 7 5x -
10-16 74 -
(
12 7 12)
- -
the body (1888; the Stefan-Boltzmann
law or the fourth - power law or Stefan's
law of radiation ), with the propor-
Suppose that the vessel in which such
tionality coefficient equal to 5.67 X
a radiation density has set in has a hole 10~ 8 J/m 2 *s* 4 (the Stefan-Boltzmann
in one of its walls with an area of S
constant).
square centimeters through which ener-
In 1899 formula (12.7.6a), known as
gy flows out. The energy flux will be equal the Planck radiation formula was dis-
to RS, where R is measured in units of
,

covered. Planck's constant h =

6.63 X
J/m 2 -s, with 10 -34 J-s for the first time appeared
in this formula.
R = -^-Q~ 5.67 x 10~ 8 Ti . (12.7.13) Planck assumed that energy associat-
ed with electromagnetic radiation is
Electromagnetic energy “flows,” or emitted and absorbed in discrete amounts
propagates, with the speed of light c, proportional to the frequency of the
but at each given moment of time only radiation, with h being the coefficient
a half of the photons are moving toward of proportionality, but that quantum
the hole, with the other half moving theory has nothing to do with the
in the opposite direction. In addition propagation of the radiation. Later,
To this, even the photons that do move in 1905, Einstein established that light
toward the hole are flying at different consists of certain portions with the
angles to the wall with the hole, which energy of each portion equaling hv and
-diminishes the flux by a factor of two. momentum equaling hv/c. Finally, in
Exact allowance for these factors gives 1924, the Indian physicist Satyendra
the coefficient 1/4 in (12.7.13). Nath Bose (1894-1974) gave a derivation
Now let us close the hole in our box of the Planck radiation formula based
(vessel with mirror walls) with a black on the calculation of the probabilities
material heated to the same temperature p o, p u p 2
,
.., with which we started
.

as that of the box and the radiation our discussion.

inside it. When we speak of the material
as black ,
this means that any photon
with any frequency falling on the 12.8* Emission Probability and the
material is absorbed. Our “stopper” Conditions for Thermodynamic
.should heat up, since it absorbs radi- Equilibrium
ation. However, since the stopper has
the same temperature as the radiation, Let us return to the formula for the
it obviously cannot heat up to a higher intensity of a beam of light passing
^temperature. Hence, the stopper will through a hot gas.
 '

12.8 * Emission Probability and the Conditions for Thermodynamic Equilibrium 411

From Eq. (12.6.5) (the reader will energy density e has the dimensions of
recall that this equation is incorrect), J/m 3 and c has the dimensions of m/s,
the product cs has the dimensions of
= — onj+yait,? J/m 2 *s, which are the dimensions of I.)
Thus, the formula (12.8.1a), which
and the condition dlldx = 0 we obtain was obtained via naive assumptions,
the equilibrium intensity 7 eq :
expresses a dependence of I eq on v and
T that does not differ significantly from
= (
12 8 . 1 )
.
(12.8.2). For the similarity to be com-
plete, we must
require that the ratio
The ratio njn 0 of the concentration of the off by an excited
energy given
of excited atoms, n x ,
to the concentra- atom to a single wave in the form of
tion of atoms in the ground state, n0 , radiation (this energy is characterized
is equal to exp [ — (E —
1
E 0 )/kT] (see by coefficient x) to the cross section of
(12.6.3) or (12.6.4)). According to the absorption of light by the atom in
quantum condition (12.6.1), E t — E 0 = the ground state, cr, obeys the following
,hv, which means that n x /n 0 — e~ hv/hT , expression:
so that
~:=chv. (12.8.3)
? Ie q = e ~hv/hT ? (12.8.1a)
From the point of view of the theory
Formula (12.8.1a) for I eq is similar of resonance (and from the point of
to the correct formula (12.7.6a) obtained view of quantum mechanics, as well),
by Planck, but does not coincide with it. this formula looks quite natural. If the
We will start with the similar as- absorption cross section is small (e.g.,
pects. The equilibrium radiation den- because the charge on the vibrating
sity J eq does not depend on the con- ball is small), so is the probability of
centration of the gas whose atoms emission (which also depends on the
absorb (in the ground state) and emit charge on the ball).
(in an excited state) radiation; it de- Now we wish to go from the probabil-
pends only on the frequency v of the ity of emission of a single wave to the
light and the temperature T of the gas. total probability of energy emission by
In the limit hvIkT 1, the dependence an excited atom. If we divide the
of / eq on these two quantities is the energy emitted per unit time into
same as in the Planck radiation for- a single wave by the energy hv of one
mula. Indeed, if hvIkT 1, then photon, we obtain the probability of
•
e hv/kT and i n the Planck radiation emission into a single wave. The next
formula (12.7.6a), step is to add all these probabilities,
^ hv that is, to integrate over the frequency
8 “ e
hv/k l __r

1
’ range: 12 6 -

x 8jxv 2 dv _ C a8itv 2 dv
we can ignore the one in the denomi-
hv c3 c2
nator in comparison to the large term J
,
e hv/kT : f or we have (12.8.4)

hv ~
hve hv ,hT (the dimensions of w are 1/s s
_1
). =
y
“ e
hv/hT
'

,
Since w is the emission probability
(precisely, the probability of emission
from which it follows that
per unit time, or emission rate), we

Ieq ~ chve~ hy? l kT

, > 1, (12.8.2) 12 6
It is
•
unimportant that this range
includes frequencies which the given excited
atom will not emit, since for such frequencies
;since I is the energy flux, and c is the =
x 0 and the factor x (v) in the integrand
.speed of light. (Note that since the will cut off such frequencies.
 ) v

412 12 Absorption and Emission of Light. Lasers-

have the following expression for the oscillator depends on frequency via the'
concentration of excited atoms in the following formula:
absence of external radiation: 1
A oc
— (12.8.9)
^f=-wn i
. (12.8.5)
(V 0)
2
+7 2 ‘

Ithas been found that in the case of an

The solution to this equation is obvi- atom at rest the absorption of light,
ous: obeys a similar formula: 12 7 *

nx (
t = (t 0 ) exp [
—w (t — £„)] X* y
2
(12.8.9a>
(
12 8 6 )
. .
2ji (v —v 0)
2
-f- y
2 *

Accordingly, t = w
_1
is the mean life-
We see that the maximum cross sec-
tion of absorption of light is of the
time of the excited atom (cf. Sec-
order of the square of the light’s wave-
tion 8.2).
length. For visible light this is many
The reader will note that above
times the size of the atom: the yellow
only the first excited state was con-
line of sodium has a wavelength of
sidered, a state that can decay in only
one manner, that is, go over to the X =5400 X 10 -10 m, while the radius,
of an electron orbit is of the order of
ground (or unexcited) state. As a rule,
10 -10 m, which means that the resonance
for given ground state and a fixed
a
cross section of absorption of a photon
excited state (with a given spatial
is greater than the area limited by the
orientation of the atom), the absorption
electron orbit by a factor of several
cross section cr depends only on the
millions. Here a photon manifests itself
direction of propagation and the polar-
not as a point-like particle flying along
ization of the incident wave. The
a straight line and either hitting the
exact formula in this case assumes the
target or missing it. According to the
form
laws of quantum mechanics, the move-
W= ( ( ai + o )^dQ,
2 (12.8.7) ment of a photon is determined by the
wave equations of the electromagnetic
where and cr refer to the two polar- field, and in this lies the possibility
2
for an atom to capture a photon over
izations of the wave, and dQ is the
a distance of the order of the light’s*
solid-angle element. When o x o2 = = wavelength.
cr, that is, the cross sections are inde-
pendent of direction, we have cq + Up till now we emphasized the similar-
ity between the approximate formula
a2 = 2cr and dQ = 4jt, hence the fac- (12.8.1a) for the equilibrium radiation,
^
tor 8jx in the simplified formula density and the exact Planck radiation
(12.8.4). formula (12.8.2) (note that hv kT ).
Note the dimensions of the various But in what respects does the exact
quantities: a has the dimensions of 2
m formula differ from the approximate;
and c/v =
X the dimensions of wave- in other words, how should we modify
length, m, so that the quantity g/X 2 the expression for dlldx so that the*
in the integral equilibrium radiation density cor-
responds to the Planck radiation
w= (
(12.8.8) formula? The answer to this question
was given by Einstein in 1917. As
often happens in physics, the answer
is dimensionless. What is important is
that a depends strongly on the fre-
proved to be easier than formulating
quency — light a resonanceabsorption is
the question correctly.
process. The reader will recall the
results 12 7
-
Substituting (12.8.9a) into the integral
obtained in the theory of oscillations: on the right-hand side of (12.8.8), we get
the absorption of energy by a damped w —
 q

12.8* Emission Probability and the Conditions for Thermodynamic Equilibrium 413

Einstein’s answer was as follows. We in the parentheses in (12.8.10) corre-

xeplace Eq. (12.6.5), spond to the two types of emission,
and onj. Accordingly, the equation
dl
~on Ir + Kn
0
,

i, for the variation of the number of

excited atoms n 1 is
with the following:
^=-<m / + x» 0 1 (H-q>), (12.8.10) dni
~dt
— wn {
On±I
hv
Gn 0 I
hv
(12.8.14)

where according to what was established

earlier x = chvo. If cp is the average In the absence of radiation we have
number of photons in a given ray (see only one term, —
wn l7 as we wrote
Section 12.7), then earlier. However, in the presence of
light, an atom can not only “jump”
l = chvy, ^ = chv^. ( 12 8 11 )
. .
from the ground state to an excited (or
upper) state (this fact is represented
Substituting this into Eq. (12.8.10) by the term on 0 I/hv on the right-hand
yields side of (12.8.14)) but can emit stim-
= chvo[ — n^+rii (l + ulated radiation by going over from
<p)].
ax ax an excited state to the (lower) ground
(
12 8 12 )
. . state (the term —
onjlhv in (12.8.14)).
whether we achieved
First let us see
We also note that the cross section cr
our goal. Indeed, the condition I = in the terms describing absorption
on and stimulated emission
/ eq that is, dlldx
,
0, leads to the = (< 0 I/hv )

(—on-^I/hv) is the same. If the intensity

following formula for the equilibrium
I of the beam is high and we can neg-
number of photons:
lect the term —wn x describing spon-
c
Peq __ ni _ e _ hv /kT taneous emission in comparison to the
1 + <Peq no terms responsible for absorption and
that is, TIq CPeq + Wi (1 + *Peq)
~ *-*r stimulated emission, the equation for I
simplifies:
-hv/kT a

= (12.8.13)
''

i _ e -hv/kT'~" e
hv/hT__ i •

^.= _a(«o-»,)/. (12.8.15)

We have thus arrived at an expression
for the probability of absorption and Obviously, the condition for this is
emission of light which leads to the the “strong” inequality I >> 7 e where ,

Planck formula. Now let us stop and / eq is the equilibrium radiation in-
evaluate the price we paid for this tensity at a given temperature of the
result. gas.
In formula (12.8.10) we added to Let us substitute into (12.8.15) the
the term x^ another term, x^cp, equal equilibrium concentration of excited
to onj. Thus we (following Einstein) atoms, —
n 0 e~ hv ^ kT The result is a
.

maintain that there exists a process of formula for the effective absorption
photon emission induced by the pre- cross section:
sence of the light beam and proportional
to the number of photons, or radiation -g-= -o(l-e-W hT)n0 I. (12.8.16)
intensity, in the already existing beam.
This process is known as stimulated When the gas is heated, the effective
emission (the word “stimulated” refers cross section, that is, the coefficient of
to the fact that the emission is caused /, decreases not only due to the decrease
by the presence of a light beam in the in n 0 but also due to the countereffect
hot gas). produced by the stimulated emission,
Stimulated emission appears alongside which yields an additional factor inside
spontaneous emission; the two terms the parentheses on the right-hand side
414 12 Absorption and Emission of Light. Lasers

of (12.8.16). If hv <C kT, then damp oscillations we need a force

that is in resonance with the oscilla-
e-hv/kT^ i_*2L and tions, that is, has the same frequency.
The difference between a driving force
1 _ e -hv!kT^ *£.< 1 ,
and a damping force lies in the different
ratios of the phase of the force to the
and the decrease in absorption is con- phase of the oscillator.
siderable. If there are only two levels, However, there is no classical picture
0 and 1, so that the total number of that can completely explain the phe-
atoms is n = n 0 + n x then n 0 drops ,
nomenon of stimulated emission. An
by a factor not greater than two as exact and complete theory of stimulated
the temperature T is raised from 0 emission is possible only if we employ
to oo.
12 - 8 quantum mechanics and the quantum
For example, in radio astronomy a theory of electromagnetic radiation
marked effect is produced by the trans- (quantum field theory). The theory of
ition of hydrogen atoms from a low- stimulated emission enters as an inte-
lying excited state to the ground state gral part into the quantum theory of
accompanied by emission of radiowaves radiation, whose bases were developed
with X = 21 cm (the hydrogen line). by P. Dirac in 1927.
The energy of the respective photons This fact stresses even more the re-
is hv = hclX £2 6.5 x 10
-34
X 3 X
. markable intuition of Einstein who in
8
10 /0.21 10~ 24 ~
J. If the temperature 1917 was able to clarify the main fea-
of the hydrogen cloud is 100 K, then tures of stimulated emission proceeding
kT ~
1.37 X 10- 23 X 100 10- 21 J, ~ only from purely thermodynamic con-
and the value of the factor inside the siderations, approximately in the same
parentheses in (12.8.16) is (1 — manner as we have done. Note that
e -hv/kT )
^ hvIkT 10
-3
(cf. Section the very notion of light as a stream
4 8 ).
. of particles, photons, was first intro-
What are the qualitative features duced into science by Einstein in 1905,
of the radiation produced in stimulated and it was for this discovery that he
emission? The excited atom in this was awarded the Nobel Prize.
process emits radiation that amplifies
the forward wave (the wave that stim-
ulates the atom, so to say), since 12.9* Lasers
otherwise there would be no mechanism
for compensation of the energy absorbed Stimulated emission opens up a remark-
by the atoms in the ground (unexcited) able possibility for generating electro-
state. In this respect, stimulated emis- magnetic oscillations (light, for one
sion differs from spontaneous emission, thing).
in which the excited atom emits radi- Suppose that we have a medium in
ation in all directions and only partially which the concentration n x of atoms
in the direction of the forward wave. in an excited state is higher than n 0)
Using the language of classical me- the concentration of atoms in the
chanics, we could say that radiation ground state. Under ordinary con-
absorption is similar to resonance build- ditions of thermal equilibrium, with
up of oscillations (cf. Section 10.5), nJriQ =e~ hy i kT such a situation is
,

while stimulated emission is analogous impossible, since here rijn^ <

1. But
to damping of an oscillator, since to there are several ways in which the
reverse can be achieved.
12 8
Of course, it would be more correct to
•
One method of making nx greater
say that the temperature is raised from 0 to than n 0 will be discussed later. Now,
T >
(E 1 —
E 0 )/k, since from the standpoint
without going into details, we will
of physics the passage to the limit T-> oo is
meaningless— the atoms would simply decay. simply assume that n x tz >
0 so that
,
 =

12.9* Lasers 415

the exact equation (12.8.10) can be

written in the following form:

-g- = 6/ + xn 1 , (12.9.1)

with b g (n-L = —
n 0 ) positive. The re-
versal of sign of the coefficient of I
in the expression for dl!dx changes
drastically the behavior of the solution.
The changes in the properties of the
solution can be compared to those that
occur in the solution for the concen-
tration of neutrons in radioactive decay
(we will write /') and consider only'
when we go over from the subcritical
its absolute value.
mass of the fissionable material (ura-
The equation for 7' has the form
nium-235 or plutonium) to the super-
critical mass (Section 8.9). l\P-=~bx. (12.9.4).
Indeed, if b in the equation for dlldx
is positive, the light flux will increase We integrate this equation “from
exponentially as it passes through the right to left,” with the “initial condi-
medium. Neglecting the second term tion” being the quantity I'L I 0 $e bL
| |
=
on the right-hand side of (12.8.1), we corresponding to the value x L (see =
find that (12.9.3)), and find the flux value /'
I=I0ebx >I0 , (12.9.2)
corresponding to x = 0:
o
where 7 0 is the intensity of the light
at the origin of coordinates, that is,
|/;|=|/£| exp(-& jdx)
at x =
0 (at the entrance to the medi-
L
um). A definite volume of the gas (the = \rL \e bL =I$e 2b t. (12.9.5)'
substance) with b 0, >
say, a pipe
After reflection from the left mirror
of length L with the gas in it, operates
(reflection coefficient 8) we have a flux
like an amplifier with a flux 7 0 at the
:

(see Figure 12.9.1) equal to

entrance, we have a flux I L I 0 e bL = >
7 0 at the exit which is more powerful I'o pSe 2bL / 0 . (12.9.6)
than the incident flux.
Now suppose that at both ends of Now we must allow for the fact that
the pipe we have mirrors with reflection the total flux from left to right is
coefficients P and 8. If at the left end
/tot = /o + /J»
(x =
0) we have an influx of radiation
7 0 then at the right end of the pipe
where 7 0 is the intensity of the incident
,
light, and 7" is the intensity of the
we have, as we know, I L I 0e bL = .

twice reflected light. On the right-hand"

This flux will be reflected by the
side of the expression (12.9.6) for 7"
right mirror (reflection coefficient P),
and the reflected flux we must put 7 tot instead of 7 0 (above
we did not distinguish between 7 0 ,

|/lI=P/l = /uP^ l (12.9.3) (the incident intensity) and 7 tot because

double reflection was not considered a
enters the pipe from the right (this flux
possibility). We then have
must be considered negative since the
energy is transferred by this flux in the /tot — /<)+/<>“ /o + pSe 26L7 tot ,

direction opposite to the positive di-

rection of the x axis). So as not to
whence
have to think about the right sign, we 7 t0t (12.9.7).
put a prime on the respective flux l-p6e 2bL *
 *416 12 Absorption and Emission of Light. Lasers

From the standpoint of physics we booklets, and books. Here, in a textbook

can say that at $&e 2bL = 1 the system on mathematics, we will only briefly
consisting of a pipe with the active point out two methods for obtaining
gas and two mirrors (with reflection inverse population, or an active medi-
^coefficients P and 8) is in a “critical” um— this is the name that physicists
state: a stationary state is possible for use describe the situation with
to
any value of / If $be2bL is less than
0. nx >n 0 that is, b
, >
0, which is re-
unity, the system is “subcritical” (cf. quired for operation of a laser.
Section 8.9). If we ignore spontaneous One method (historically the first),
emission, there is only one solution, which was employed by the Nobel Prize
Io =0, while if we allow for spontane- winners N. G. Basov, A. M. Prokhorov
ous emission (that is, the term un x in (both USSR), and Ch. Townes (USA),
Eq. (12.9.1)), radiation with an inten- consists in sending molecules that are
sity proportional to x^L sets in in the in states 1 and 0 into a capacitor
system. Finally, if we allow for stim- placed in a vacuum, in which they
ulated emission, the amplification move without colliding with each other
{or gain, as it is called) is even higher: under an electric field. The electrical
it is proportional to the ratio 1/(1 — properties, namely, the polarizabilities,
$8e 2bL ). of molecules in states 0 and 1 are differ-
From the standpoint of technology ent. Using this fact, the inventors of
-such a situation is of little interest. the laser were able to set up conditions
-A remarkable situation emerges when in which the molecules in the ground
2bL is greater than unity. By analogy state were deflected away from a vessel,
p 8e
with a chain nuclear reaction, here we while the molecules in state 1 gathered
*can speak of the system being in the in the vessel. Thus, in the vessel, n x was
‘“supercritical” state. In this case, even greater than n 0l which is just the situ-
for a very small initial intensity and ation required for amplification of light.
low spontaneous emission there will The other method, which is more con-
be an exponential growth in the radi- venient and does not require a high
.ation intensity with time, approximate- vacuum but which historically was de-
ly proportional to e** 1 where y
,
~ veloped later than the first, consists
{c!2L) (P8e 26L —l)
-1
The radiation in-
. in employing a system with several
tensity will grow, but so will the levels: 0, 1, 2, 3. The system “works,”
number of excited atoms used up; n x that is, emits radiation, using the
:and b will decrease as long as the transition onto one of the intermediate
system remains noncritical; the system levels, say, the transition 3 2 with
will become critical only at a high the frequency v 3>2 (E 3= —
E 2 )/h. For
value of /. this to happen, we must excite state 3,
Thus, in a system with b positive, that is, create a high concentration n 3 .

radiation of high intensity can be Here, however, n 3 remains lower than

achieved. The radiation is strictly mono- n 0 since it is impossible to create an
,

chromatic (it is represented by a single inverse population, that is, to make n 3

oscillation wave), with the result that greater than n 0 by irradiating the gas
,

it can be focused into a point. A device with light of frequency v 3>0 because of
that converts input power into a very stimulated emission of light. (Why?)
narrow, intense beam of coherent visible But if molecules go over very rapidly
or infrared light is called a laser (for from level 2 to low-lying levels, 2 1
Zight amplification by stimulated emis- or 2 ->• 0, there is a high probability
sion of radiation) or maser (for micro- that n 2 <
n 0 and n 2 n3 <that is,
,

wave amplification by stimulated emis- the system becomes supercritical with

sion of radiation). respect to the emission of light of
The applications of a laser beam have frequency v 3>2 .

been described in numerous articles, There are also other methods for
12.9* Lasers 417

creating an active medium. Acom- ultrahigh temperatures and pressures,

mon feature of all lasers is the trans- which could be used in controllable
formation of input power into a thermonuclear fusion. Today lasers are
highly energetic, plane, monochromat- being widely used in surgery, in pro-
ic electromagnetic wave. If such energy cessing of materials, and in chemical
transfer is achieved, the energy can reactions, but all this, as we have al-
easily be focused into an extremely ready said, belongs to books on other
small volume. With high energy con- topics.
centration it becomes possible to achieve

27-0946
 Chapter 13 Electric Circuits and Oscillatory
Phenomena in Them

13.1 Basic Concepts and Units which lost part of its electrons (only
of Measurement in rare cases is a positive charge the
result of a body acquiring positive
In this chapter we consider phenomena charges). A
negatively charged body is
that occur in electric circuits The prin-
. one which has acquired a surplus of
cipal elements of an electric circuit are electrons. The direction of current is
resistance capacitance, inductance,
,
opposite to that in which the electrons
and sources of current (voltage). Our ex- move in a conductor.
position is not designed to take the place The electric potential
of a given point
of a standard physics textbook but is the potential energy that a positive
rather to supplement, develop, and charge of 1 C possesses when placed at
refine some of the knowledge contained the given point. The electric potential
in a school textbook. We will therefore of the ground (earth) is taken to be
confine ourselves to a brief review of the zero Hence, the point of a circuit
.

definitions of resistance, capacitance, connected to the ground by a metal con-

and so forth and to their units, on the ductor (we say it is grounded or earthed)
assumption that the reader is sufficient- has potential zero. The unit of potential is
ly acquainted with the basic notions the volt (abbreviated: V). The poten-
from the school physics course. tial of a point is equal to 1 volt (1 V)
The quantity of electricity is deter- if a charge of 1 coulomb placed at this

mined as the net charge, or the difference point has a potential energy of 1 joule.
between the positive charge and the (The reader will recall that the joule,
negative charge. We denote it by q . the unit of energy and work, is defined
The unit for the quantity of electricity as the work performed by a force of
is the coulomb (abbreviated: C). The 1 N over a distance of 1 m, or 1 J =
elementary charge, the charge of the 1 N-m = 2
1 kg*m /s .) The potential
2

proton, is e v — 1.6 x 10~ 19 G, and the energy u of a charge q placed at a point

electron charge is e e = —
1.6 x 10
-19
C. where the potential is equal to cp is
Electric current is defined as the
quantity of electricity flowing in unit
»(J) = ?(C)-<p(V). (13.1.1)

time through a cross section of a con- We have to imagine here that q is small,
ductor. We will denote current by /. because if a large charge (say 1 C) is
The unit of current is the ampere (abbre- placed at the given point, the potential
viated: A); this is a current in which cp will change. For this reason, it is
1 coulomb of electricity passes through better to say that the potential is the
a cross section of a conductor in one coefficient of q in (13.1.1), or the factor
second. Thus, 1 A = 1 C/s, but it is that connects the potential energy of a
more convenient to say that 1 C = charge and the quantity of electricity.
1 A-s, since in the SI system of units The work A performed by a field in
the ampere is a base unit and the cou- transferring a charge from a point where
lomb is a derived unit (see Appendix 5). the potential is equal to cp 2 to a point
For the (positive) direction of current where the potential is cp 2 is
we take the direction in which positive
A= u,x — u^ —
charges would "have to move in order
qhV
to produce a given current. Actually, Just as in mechanics only the difference
in metallic conductors positive charges of potential energies enters into all
are stationary, and the current flows physical results, so in electricity the
due to the motion of electrons. As a formulas always involve a difference of
rule, a positively charged body is one potentials but never the potential ener-
13.1 Basic Concepts and Units of Measurement 419

Figure 13.1.2
Figure 13.1.1

gies themselves. There will be no change

ever. 13 A change in the charge on
* 1

plate A of the capacitor is due to the

in the potential difference if to all po-
fact that a portion of the charge left
tentials at all points we add an identical
the plate and moved to some other site,
summand. It is therefore possible to
choose the potential of any point in
say, point D
in Figure 13.1.2.

a circuit or a piece of equipment in

If a current j is flowing from to A D
(from left to right), then in time dt
arbitrary fashion, say, set it equal to
a quantity of electricity j dt will flow
zero. However, after this has been
through the cross section of the conduc-
done, the potentials of all other points
tor, whence
become quite definite quantities. It is
precisely for this reason that we can
take the ground potential as zero.
dqA — 1 dt, or — i- = /.

Let us consider a capacitor (Figure

Let us now find out what a current
13.1.1) consisting of two parallel plates,
flowing in a conductor depends on. By
A and B. One of the plates (say A) can
be connected to some source of voltage.
Ohm’s law the current is proportional
to the potential difference across the
The quantity of electricity on this plate
terminals of the conductor, the current
is directly proportional to the potential
flowing from higher to lower potential.
difference across the plates of the ca-
Thus,
pacitor,

Qa = Ccp c , / = &(cpD — <pj = (q> D — q> A ). (13.1.2)

where cp c is the difference of potentials

defined as the potential of plate A The positive quantity k is called the
conductance. The reciprocal, 1 /&, is
minus the potential of plate B
Since .

called the resistance of a conductor and

in Figure 13.1.1 plate B
is grounded, it
is denoted by R. The unit of resistance
follows that cp c in this case is equal to
is the ohm (abbreviated: Q), which
the potential plate A.
of
is the resistance of a conductor through
The coefficient of proportionality C is
called the capacitance of the capacitor.
which a current of 1 ampere is flowing
The unit of capacitance is the farad when a potential difference of 1 volt
is impressed across the terminals, so
(abbreviated: F). It is the capacitance
of a capacitor in which the potential
that 1 Q 1 V/A. =
difference across the plates is 1 volt
We denote the quantity (p D cp A —
"
for a charge of 1 coulomb (10 6 F is
by (p B This is the potential across the
.

resistance R. The value of cp# is defined

a microfarad (jiF), 10~ 9 F a nanofarad
-12 (just as like that of cp c ) as the left-hand
(nF), and 10 F a picofarad (pF)).
potential minus the right-hand potential
An equal quantity of electricity (but
(the current flows from left to right).
opposite in sign) accumulates (builds
up) on plate B:
Ohm’s law (13.1.2) can then be written
as
qb
— — qa — — Ctpc*
/
= -
7f,
or <p r = RJ. (13.1.3)
The electric charge of a closed system
is a conserved quantity, that is, electric
13 1
The net electric charge of a system
-

charges of the same sign never appear remains unchanged when two particles of equal
or disappear in any process whatso- and opposite charge appear or disappear.
27 *
 a a

420 13 Electric Circuits and Oscillatory Phenomena in Them

o
/
1 o W o—
/ ^ 2
1
2 J R
Figure 13.1.4
Figure 13.1.3

As a source of voltage in a circuit we allow for a linear dependence of E on /:

can take a voltaic cell. There is a de- E = a - bf , (13.1.4)

finite potential difference across the
terminals of the cell. We can assume, where the (positive) coefficients a and b
roughly, that the potential difference are characteristics of the cell. call We
a cell whose emf is independent of the
is independent of the current flowing
current /, that is, b == 0 in (13.1.4) an
through the cell. In particular, in a cell
ideal cell.
the current can flow from a low poten-
tial to a high potential. Through a re-
Let us consider a series connection
of an ideal cell with an emf equal to a
sistance, the current always flows from
u high potential to a low potential, and a resistance b (see Figure 13.1.4).
like water in a tube connecting
Then
two
vessels flows from high level to low We = cpi — 92 = —a,
level.
A <Pb
= <P 2 — q>3 = bj;
voltaic cell is like a pump that
can take in water in a low-level vessel whence
and pump it up to a high-level vessel, —
that is to say, make the water move <Pi — 9.3 = (<Pi +
cp 2 ) — (<P2 cp 3
)

uphill. To operate the pump we need = — + - bj = — — ( bj)

some kind of external source of energy. = —E.
The same applies to the cell. When
the current flows from low to high We have again
arrived at formula
potential, chemical reactions take place (13.1.4) the emf E whereby b is
for ,

in the cell. The energy of these chemical often called the internal resistance of
reactions in the cell is transformed into the cell. The real cell, with which we
electric energy. usually deal, which has an emf that
The potential difference which the satisfies formula (13.1.4), yields the
cell yields is termed the electromotive same dependence of E on j as the series
force which we abbreviate to emf.
,
connection of an ideal cell and a re-
The potential difference across the sistance b. The name for a remains the
cell taken as the left-hand potential same, the emf of a real cell, bearing in
painus the right-hand potential (Fig- mind that E a when j 0, and = —
ure 13.1.3) is equal to minus the electro- the emf drop for j =/= 0 is characterized
13,2
motive force of the cell: by b .

In the sequel, when considering elec-

<Pi — qp2
= —E. tric circuits involving current sources,
for example a cell, and various resist-
In reality, the emf is slightly dependent
on the current flowing through the cell.
ances, we can imagine that we are
dealing with an ideal cell with constant
When the current is flowing (from left
to right in Figure 13.1.3) in the direc-
emf independent of the current, while
the internal resistance b may be com-
tion from low potential to high potential
bined with the external resistance R
(which is the normal operating con- .

Thus, a real cell with internal resistance

ditions of the cell when it is generating
•electric energy), the emf E diminishes
b connected in series with a resistance R
with increasing current flow. Approxi- 13 2
Since the current is zero for an open
•

mately, we can take it that the emf is circuit, the emf may be defined as the poten-
constant, but more exactly we must tial difference across a disconnected cell.
13.1 Basic Concepts and Units of Measurement 421

is equivalent to an ideal cell connected

in series with a resistance 1 6 R = R + .

The current flowing in the circuit is 1 L 2

then given by the formula
Figure 13.1.6
. E
R+b *

with the magnetic field that is produced

worth once again paying special
It is in the space surrounding a conductor
attention to the difference between re- that carries a current. The magnetic
sistance and source of voltage. If in field is particularly high if the conduc-
a circuit there is a potential difference tor has the form of a coil with a large
across a resistance, such that cp 2 <Pi > number of turns. The field is further
(Figure 13.1.4), then, by our definition, increased if the coil is wound on an
<Pn —
<Pi cp 2 —
0, that <
is, cp# is iron core.
negative. Hence, by formula (13.1.3), The magnetic field, in turn, gives
the current is negative, too, which rise to electric phenomena (if the mag-
means that the current flows from right netic field varies). As we know, given
to left, from point 2 to point 1. Now a varying magnetic field, each turn
suppose that there is a potential differ- (even every portion of a turn) of the
ence of the same sign across the ter- coil becomes a source of voltage, some-
minals of the voltage source, and the thing like a voltaic cell. In a coil in
dependence of E onj is given by formu- which the turns are wound so that the
la (13.1.4) (see Figure 13.1.3). Here, current traverses the core of the coil
let cp 3 >— cp x but cp 3 — cpi < a . Then in the same direction throughout the
bj = cpi cp 3 + a = —a (cp 3— cp x ) > 0, length of the coil, all these voltage
ys, 'bvmwb art vuinwiladi vn 'sulyus ou
flows from left to right despite the the overall voltage builds up (the volt-
fact that the potential cp x on the left ages are additive in a series connection).
is less than the potential cp 3 on the On the whole, a coil is equivalent
right. Thus, the voltage source is ca- to a voltage source with a potential
pable of overcoming the potential difference proportional to the rate of
difference and yielding a positive cur- change of the magnetic field. But the
rent (from left to right) for a negative magnetic field in a coil is proportional
potential difference ((p x 93 0 ), pro-— < to the current flowing in the coil.
13 3

vided that this negative potential dif- For this reason, the rate of change of
ference does not exceed in absolute the magnetic field is proportional to
value the emf of the source. Yet for the rate of change of current flow, that
a negative potential difference, the is, to the derivative dj/dt Referring to .

resistance always yields a negative Figure 13.1.6, we find that in the coil
current. In any circuit containing only
resistance, that is, circuits without 2=^’. o3 - 1 - 5)
emf, the current can only be identically
zero, while if there are cells, that is, and the positive direction of the current
emf’s (Figure 13.1.5), a nonzero solu- is taken to be from point 1 to point 2

tion (current) is possible. inside the coil, while the quantity

Now let us consider inductance The . is the potential difference across the
phenomenon of inductance is connected 13 3
•
We
will not discuss the case of two
coils wound on one core, which is a transform-
er connecting two electric circuits that carry
different currents. Neither do we consider
cases of a more complicated dependence of the
magnetic field on the current when an iron
core is inserted in the coil and the current is
Figure 13.1.5 so great that the iron is saturated.
422 IB Electric Circuits and Oscillatory Phenomena in Them

coil (it is defined as the potential cp x

on the left minus the potential cp 2 on

the right). When considering in detail
the direction of the magnetic field and
the emf induced by its variation, it is
found that the coefficient L (the in- 1

ductance) is always positive.

From formula (13.1.5) it follows that Figure 13.1.7

if dj/dt is negative, then cp x cp 2 is —

negative, too, that is, q 2 )
<p x >
Thus, . similarity will be discussed in more
if the current is positive (flows from detail in Section 13.8.
1 to 2) and decreases in magnitude, the From the standpoint of subsequent
coil plays the role of a cell sustaining computations, capacitance, resistance,
a positive current in the circuit, despite emf, and inductance have one thing
the fact that cp^ is negative. But if the in common, they all require two ter-
current is positive and increasing, djldt minals for connection in a circuit (un-
is positive, and so cp L is positive, too. like, say, a transformer, which requires
In this case, the coil plays the part of four leads, or a transistor, which has
an additional resistance, since the po- three leads: collector, base, emitter).
tential difference across the coil is Devices with circuit connections involv-
positive for a positive current (cf. ing two leads are called two-terminal
(13.1.3)). networks, while if there are four leads,
A from a voltage
coil differs essentially they are two-terminal pair networks
source and from a resistance in that (or four-pole networks). Each circuit
the quantity 9 ^ depends not on the element— capacitance, resistance, emf,
current intensity / but on the rate of and inductance— is characterized at any
change of the current, djldt . given time by a specific current passing
The coefficient L in Eq. (13.1.5) bears through it and a definite potential
the name “coil inductance” (also self- difference at input and output.
inductance ). 13 4 The unit of inductance
-
We can imagine a closed box (labelled
is the henry (abbreviated: H). If the “Box” in Figure 13.1.7) with two leads,
inductance of a coil is equal to 1 henry, A and B, sticking out of it. The interior
this means that when the current is of the box may contain anything:
changing at a rate of 1 ampere per R, E, L, and C. Connect an ammeter A
second, a potential difference of 1 volt and a voltmeter V. With the circuit
is induced in the coil. We obtain the connections as shown in Figure 13.1.7
dimensions of inductance from for- (the “+” and “ ” signs correspond to
—
mula (13.1.5): the labels at the terminals of the ammeter
and voltmeter), the ammeter records
1H = 1 V-s/A.
the current / passing in the direction
From the foregoing it is clear that from point A to point B the voltmeter ,

i n j ji J*h j p a* .indicates the potential

. difference
circuit just like an inert mass (flywheel) cpBox = 9a — 9 b* The relationship
affects velocity: inductance impedes between 9 e 0 x and / depends on what
any change in the current, and a mass is inside the box:
(by Newton’s second law) tends to
in the case of a resistance R,
impede any change in velocity. This
<Pbox = Rj, (13.1.6)
13 * 4
Often instead of saying “a coil with
inductance L" we simply say “inductance V\
in the case of an emf 13 - 5
E0
just as we say “capacitance C” instead of
cpBox = —E 0, (13.1.7)
“a capacitor with capacitance C\ or emf E
13 • 5 emf
instead of mentioning a voltaic cell or a volt- The internal resistance of the
age source. source is disregarded.
 •

13.2 Discharge of a Capacitor Through a Resistor 423

9 al> an d cp bl =
cp ae- Whence, adding
all the equations in (13.1.10) termise,
we obtain
?C + (PR + (Pl + CPje; = Cp Ac — (p BjE .

If the circuit in Figure 13.1.8 is

closed, then cp Ac = In this case,
cp BjE .

in the case of an inductance, L, consequently,

<Pc + cpR + cp L + cpjp = 0. (13.1.11)
<p Bo * =£§, This general equation, together with
(13.1.8)
the expressions (13.1.6) to (13.1.9),
in the case of a capacitance 13 * 6 fully describes all the processes that
C,
occur in the circuit. Below we will use
t
this equation to examine a variety of
^Box = ( < P BO x)« +4 (13. 1 .9) circuits beginning with the very sim-
<0 plest which consists of only two ele-
or ments.

^BOX = J_f1 • (13.1.9a)

~dT '
c 13.2 Discharge of a Capacitor Through
a Resistor
There are of course cases in which
more complicated relationships are in- Let us examine the process in a circuit
volved. For example, a rectifier (a va- with capacitance C and resistance R
cuum-tube diode or a semiconductor (Figure 13.2.1). We denote by cp the
diode) does not fit any of the formulas potential of point A (plate B of the
(13.1.6)-(13.1.9). However, in a large capacitor will be grounded). To begin
number of important problems we can with, let cp cp 0 =
The corresponding .

confine ourselves to considering the quantity of electricity on plate A is

circuit elements for which these for- Qa Ccp 0 .

mulas are valid to a high degree of Can we speak of a current flowing

accuracy. These are the circuits that through a capacitor? A capacitor con-
we will investigate (with the exception sists of two plates separated by an
of Section 17, where we give special insulator (say, air) so that in reality
consideration to the properties of a cir- an electron cannot go through the
cuit with a device that exhibits a com- capacitor, say, from A to B However, .

plicated relationship between current if a positive charge is impressed on

and potential difference). plate A then plate B will have a nega-
,

Let us study the circuit shown in tive charge, and a positive charge will
Figure 13.1.8. We
will first write down flow out of plate B along the wire (the
the voltage drops on the separate ele- current also goes from left to right).
ments of the circuit: Two ammeters, A x and A 2 one , of which
<Pc — <P ac — <Pb c , cp R = cp — cp
measures current in the wire connected
Afi Bf? , to plate A the other in the wire con-
,
(13.1.10) nected to plate 5, will have identical
<Pl = <Pal — <Pbl> Ve = Vae ~ 9 be
< readings. What precisely is it that flows
And observe that cp Bc = (Par, b r = cp

qA = =
6
Here
*
C7cpB 0X and dqA /dt /,
whence dy Bo Jdt = j/C. If at the initial time
i — i0 we have =
cp
Box (cp
Box ) 0 ,
theE
t

'(PBox — ((PBox)o + C 1
j dt.
j
to Figure 13.2.1
 424 IS Electric Circuits and Oscillatory Phenomena in Them

(positive charges or electrons) through We solved a differential equation like

different portions of the electric circuit this in connection with the problem
does not interest us, just as we are of radioactive decay (see Chapter 8)
not interested in whether the same and in many other sections of this book.
electrons pass through A 2 that have If cp =
cp 0 when t 0, then =
passed through A 2 or not. For this
reason we will henceforth only speak of
cp (t) =y e~ 0
t/ RC
, (13.2.4)

the current passing through the capacitor whence

and will have in mind the current
flowing in the conductors connected to
the plates of the capacitor. We can It can be seen from formula (13.2.3)
speak of current flowing in an electric that the quantity RC has the dimen-
circuit through a capacitance in the
sions of time. Let us verify this:
same way that we speak of current
flowing through a resistance or an in- [R] =Q = V/A - V-s/C, [C] - C/V,
ductance. The difference lies in the whence
different type of relationship between
V-s _C
current and potential difference as ex- IRC] s.
pressed by the formulas (13.1.9) and G V
(13.1.9a). During time t = RC the charge q on
When we close the switch P (see the capacitor and the current / di-
Figure 13.2.1), a current minish by a factor of e.
The discharging process in a capac-
>'
= TT<P* itor can easily be observed experi-
mentally. Buy a capacitor with capac-
will flow through resistance R. By itance C
20 (liF =
20 x 10~ 6 F and =
Eq. (13.1.11), (p + q)R = 0, whence a resistor with resistance R 20 MQ = =
(p# = — cp
and so 20 x 10 6 Q. For an RC circuit of this
type we get RC 400 s, which is =
7
= — -jf <P* (13.2.1) a very convenient time for observation-
al purposes.
Since current flowing from left to right The quantity is known as the RC
is taken to be positive, it follows from
time constant circuit consisting of a
(13.2.1) that for cp 0 the current is > of a capacitance and a resistance (recall
negative, it flows from right to left, and that in the case of radioactive decay
the capacitor becomes discharged. 13 7 -

the mean lifetime was an analogous

Recalling that / dq/dt (current = quantity).
flowing through a capacitance) and We will consider the problem of
q =
Cep, we find that charging a capacitor through a resistor.
The circuit diagram is shown in Fig-
i~ c -if- (13.2.2) ure 13.2.2. If switch P is closed, then,
by (13.1.11), cp E +
cp R + cp = 0, where
Comparing (13.2.1) and (13.2.2), we cp the potential of the nongrounded
is
obtain plate of the capacitor. Since (p s E0 = —
dep and cp H =
Rj, it follows that E0 — +
dt
(13.2.3)
£° J £
13 7
•
Observe that in having the
all circuits f| AM/ 1|

form of a rectangle (see Figure 13.1.8 and

1

R
subsequent figures), we speak of the direction
of current in the upper side of the rectangle; X/>*
O
the current flow in the bottom side that closes
the circuit will clearly be in the opposite
direction. Figure 13.2.2
 2

13.3 Oscillations in a Capacitance Circuit with Spark Gap 425*

Rj +
cp 0. =
The current flowing
= = cr jL
through the capacitance is /
C {dyldt), whereby
dqldt
"I I VA
R
o
'o —C 1|-
2

— J?o + i?C-^- + = 0, <p or

TT---ST<'P- £')- (13-2.5)

Figure 13.2.4

To howcp varies with time, it

find out
will be convenient to make the change From formulas (13.2.4) and (13. If .

of variable, z = cp E 0 with dz = dcp. — ,

it isevident that charging and discharg-
Equation (13.2.5) can then be re- ing the capacitor is the faster the
written thus: smaller the resistance R and the smaller
the capacitance C .
dz __ z
~~
lt'~' RC-
Exercises
ItS solution is
13.2.1. Referring to Figure 13.2.1, C =
z=z 0 e~ t / RC , (13.2.6) 10- 6 F and R 10 7 Q, R = 10 8 Q, R =10 9 Q. =
For each of these cases, determine the time’
where z 0 is the value of 2 at the initial lapse during which the current flowing
time. through the capacitor at the initial time falls^
Let us find the solution for the case off by 10%; decreases by a factor of 2.
13.2.2. Consider the process of equalizing
where at the initial time the capacitor the potential across a resistance R in series
is not charged: cp = 0 at t = 0. Then with two capacitors, C 1 and C 2 one of which
= —E = ,

Zq From (13.2.6) we get z = at time t 0 is charged to a potential differ-

=
.
0
— or
ence cpci (0) a, while the other is not charged
at all, that is, (pc 2 (0) =
0 (Figure 13.2.4).
= z-{- E = — E e~O RC E
<p 0 0 q\
13.2.3. Determine the variation of the time
constant of the circuit depicted in Figure 13.2.1
= E (i -e-tlRC
0 ). (13.2.7) if all linear dimensions of the circuit diagram,
are increased n-fold (for the case of a plane-
The graph of cp as a function of t is parallel capacitor). (The condition of the prob-
given in Figure 13.2.3. The curve cor- lem is to be understood in this way: the di-
mensions of the capacitor and the resistance are
responds to the formula (13.2.7), while increased but the materials of which they are*
the dashed horizontal line represents the made are not changed.) [Hint. The formula,
value cp =
E 0 which the solution ap- for the capacitance of a plane-parallel capac-

proaches with the passage of time. The itor is known from physics: C =
& S /4 n d,
where S is the area of a plate of the capacitor,
quantity z has the geometric meaning d the distance between the plates, and e a
of vertical distance from the curve to constant characterizing the material between
the dashed line. This distance dimin- the plates (the dielectric constant). The re-
sistance of a wire resistor is found from the for-
ishes exponentially with the passage
of time.
mula R =
pZ/cr, where l is the length of the-
wire, a the cross-sectional area of the wire,
During a time equal to RC the ,
and p a constant characterizing the material
charge of the capacitor reaches 63% of the wire.]
of its final value, during time 2 RC it
reaches 86%, and during time 3 RC
13.3 Oscillations in a Capacitance
it reaches 95% of its final value.
Circuit with Spark Gap

A typical circuit diagram involving-

a capacitance is shown in Figure 13.3.1.
The circuit includes a voltage source-
with emf E and resistance (the role- R
of R may be played by the internal
Figure 13.2.3 resistance of the voltage source) Under-
 e

426 IS Electric Circuits and Oscillatory Phenomena in Them

through the spark gap. Therefore, for

cpwe get the equation
dtp _
~
cp

dt rC ’

with cp = cpi at t — tA ,
whence we
obtain
-(t-< ^ )/rC
T
Figure 13.3.1
<p = cp 1 e . (13.3.2)

At time t t B (point 5), cp = cp 2 = ,

neath is a spark gap. For a potential
the spark gap again becomes an insu-
difference less than a certain value cp x ,
lator, the charging process is initiated
the spark gap is an insulator. At cp cp x = (section BC ), and so on.
a spark jumps the gap and the air be-
tween the wires heats up and becomes
Let us determine the time t B tA —
during which the capacitor discharges.
•a good conductor. denote the totalWe To do this, we take advantage of the
resistance of the leads and the incan- = =
fact that cp cp 2 at t tB Putting .

descent air by r. The quantity r is cp =

cp 2 and t =
t B in (13.3.2), we get
small and remains small as long as a
current flows maintaining a high tem-
perature of the air. For a definite small
whence
value of current / 2 the air cools and
the spark gap again becomes an insu- tB — tA = rC In (q>]/cp 2 ).

lator. This current value is associated

Over segment BC (charging) the rela-
with the potential difference (p 2 = hr -

tion (13.3.1) displaced in time by the

Here cp x > <p 2 : a higher voltage is
amount t holds true (in Figure 13.3.2,
.needed to initiate a spark than to keep
t is depicted by the line segment A Q B).
it burning.
For this reason
Figure 13.3.2 shows the dependence
of cp on t for such a circuit. The capacitor cp =E (1 — e-(f-t)/RC) #

is charged over the section OA, with

no current flowing through the spark
Putting t = tB ,
we get
~ T)/RC
gap. In this case the solution (13.2.6) cp 2 = £’(l— _(<B
).
is valid:
Similarly, setting t = tc ,
we find that

<p = £(l -e-V™). (13.3.1) e- (<c-Tl/RC

— i?(l — ).
<Pj

The potential difference at point A Combining the last two formulas, we get
at time t =
t A reaches the value cp l7
-«c-*BVM
E— =e
=I2s
the spark gap begins to conduct cur- | <Pi
.
t
’
or
rent, and the capacitor discharges.
t c — B = RC In
£—
Since in this case R r, the current > t
E—
q>2

q>r
from the voltage source can be ignored
as compared to the current passing The complete period (the charge-
discharge cycle) is

T= -tA = (t c -t B + (t B -tA
tc ) )

= RC In Eg—
- f rC In
q>2

q>2
.
(Pi

Ordinarily, the resistance R in the

circuit of the voltage source is many
times greater than that of the spark
Figure 13.3.2 gap, and for this reason the charging
13.3 Oscillations in a Capacitance Circuit with Spark Gap 427

tentials even out in the course of time.

Indeed, in Figure 13.2.1 the value
cp =
0 sets in, and in Figure 13.2.2,
cp =E
0 is the steady-state value (see
o
formulas (13.2.4) and (13.2.7)). 13 8 The
situation is quite different in the case
0
of a spark-gap circuit. Here we have
o undamped oscillations of cp (true, they
are very different from those that we

,
U
n
(

m u
n
studied earlier). These oscillations are
connected with certain specific prop-
erties of the spark gap, in particular
with the fact that until a definite
potential is reached (the breakdown
potential cp x ), no current flows through
Figure 13.3.3 the spark gap.
Many books have been written about
the properties of discharge through air
time is much longer than the time of in a spark gap. All we have given here
discharge. On the other hand, the is a smattering of information only —
discharge current is many times greater enough to understand the operation of
than the charging current, greater than the circuit shown in Figure 13.3.1.
the maximum current obtainable from This information does not even suffice
the voltage source (with an internal to answer the simple question: What
resistance of R 1 the voltage source does will happen if we connect the spark gap
not produce a current exceeding E/R x ). to a voltage source without a capacitor?
Thib 'vrrciiih °Juwn ; ru F’fgim? 'trL.Ux LirUwh ,
;
L nm ''tvrreih firowb, + hfb T v\)i-c-
transforms a long-term small current age across the spark gap will be E 0 .

generated by the voltage source into Since E 0 is greater than cp l7 breakdown

a strong current, which, however, is should occur. But if this occurred, the
of short duration (it is customary to resistance of the spark gap would
speak of “short pulses” of current). become small, equal to r. Then a
This circuit operates like a system in potential difference equal to E 0 r/(r R) +
which a tiny flow of water gradually would appear across the spark gap and
fills a vessel (Figure 13.3.3). The vessel the current would be j =
E 0 /(r R)- +
is fixed in such a manner that when If R is great, the current / is small,
a sufficient amount of water has accu- less than ; 2 and the potential difference
,

mulated, the vessel turns over and across the spark gap is small, less than
the water pours out. The vessel then cp 2 .But then the air will not heat up
rights itself and the process begins and the resistance of the spark gap
°cSv?^v .
T i\x firmed* wi 1L "xivl u +
bx^ "mpAL ,

on a horizontal axis 00' below the which means the potential difference
midpoint. A weight is attached at the will be great and equal to E We have .

bottom of the vessel so that the center a contradiction.

of gravity of the empty vessel lies below Actually, under these conditions we
the axis. But when the vessel fills up have an electric discharge of a different
with water, the center of gravity of type, the glow discharge (small current
the full vessel lies above this axis and without heating of the air), instead
the vessel tips over. of the spark with incandescent air.
Let us return to the circuit diagrams
in Figures 13.2.1 and 13.2.2. In these 13 8
-
Below we will see that the potential
circuits, which consist of capacitances, tends to a steady-state value as oo in
resistances, and emf sources, the po- a circuit with an inductance, too.
 428 IS Electric Circuits and Oscillatory Phenomena in Them

13.4 The Energy of a Capacitor quantity of electricity that has flowed.

When the capacitor is being charged,
A charged capacitor has a definite the process stops at cp E 0 In this = .

supply of energy, which can be given process, the voltage source will have
up very quickly if the capacitor is performed work
discharged through a small resistance.
Let us find the supply of energy of E»q« — E CE —
0 0 CE\.
a capacitor of capacitance C, one plate What supply ofenergy will the ca-
of which is grounded and the other pacitor possess? This can readily be
has a potential cp 0 Then the quantity .
computed from formula (13.4.3):
of electricity q 0 C cp 0 = .

It would appear at first glance that W = ^CE\.

the energy is equal to the product
g 0 cpo (= Ccp^). In reality, this expres- Where has half the work performed
sion not exact, though it is correct
is by the source gone? We will show that
as to order of magnitude: it differs it went to heat up the resistance R.
from the true value by a factor of 2. Recall that if a quantity dq of elec-
Let us consider the charging of the tricity flows through a resistance, the
capacitor. When the capacitor’s po- energy released will be
tential is cp and the charge is q the
dA 1= R dq
,
cp ,
(13.4.4)
addition of a small quantity of electri-
city dq increases the energy by where cp R is the potential difference
dW — cp dq (13.4.1)
across the resistance. Using the fact
=
.

that dq =
j dt and j cp r /R, we can

The essential thing is that during transform (13.4.4) to the familiar form
charging the potential cp changes, since
cp = qlC . Substituting this value of cp dA = -^-dt = j 2 R dt.
into (13.4.1), we get
The quantity pR = cp r/R is the amount
dW — ~q dq. (13.4.2) of energy released on the resistance in
unit time, which is to say, it is the
Integrating (13.4.2) from q = 0 (un- power output the resistance. of It is
charged capacitor) to q — q0 ,
we get measured, as expected, in watts:
qo
2 [ fR] = A -Q = A (V/A) = A-V
2 2
-

W{q») = ±)qdq = ±-^ = A-(W/A) = W.

0
The time dependence of j in the case
= -f = 4 C(P»- <
13 4 3 )
- -
of the charging of a capacitor through
a resistance was found in Section 13.2:
Thus an exact evaluation yields the
coefficient 1/2. / (t)
= -jf e~ tlRC
Now us examine the charging of
let
a capacitor from a voltage source Therefore dA =
{E\!R) e~ 2 t/ RC dt.
through a resistance (see Figure 13.2.2). The energy released during time T is
The voltage source has a constant emf, T
E 0 Therefore, when a quantity of A e~ 2t ' RC dt,
.

electricity dq flows, the voltage source

(T) =^- j
0
does work E 0 dq (this work is performed
at the expense of the chemical energy whence
of the voltage source, that is, the
A(T)= e~ 2t/RC
chemical energy diminishes). Hence, the
total work done by the voltage source
is equal to E 0 q 0 where q 0 is the total
= ~-CE\{\ — e~ 2T RC l
). (13.4.5)
,
 .

13.4 The Energy of a Capacitor 429

Weknow that as the time interval T

increases without limit, the potential
<p approaches the value 0 As can be E .

seen from (13.4.5), A then approaches

CE\!2. Therefore, the total energy re-
leased on the resistance is

A= ±CEl (13.4.6)

Thus, calculations confirm the fact

that in the charging of a capacitor
half the energy is lost on the resistance.
The efficiency of charging is only 50 % and so on. The gain is clearly seen in

Note that if we directly connect the the graph: lay off the charge q of the
voltage source to the capacitor, nothing capacitor on the axis of abscissas and
will change, the efficiency remains at
the potential cp on the axis of ordinates.
50%, the role of resistance being taken The two are connected by the relation
by the internal resistance of the voltage cp =qlC, which represents a straight
line (Figure 13.4.1). The energy of
source, which will then heat up. From
formula (13.4.6) it is evident that the a capacitor is equal to the area of the

energy lost uselessly on the resistance triangle OAB. The work done by the
in charging the capacitor is independent
voltage source is equal to the area of
of the magnitude of R and hence does
the rectangle OABD. The energy lost
not depend on how fast the charging on the resistance is equal to the area
of the triangle ODB.
takes place.
If the capacitor is charged in stages,
Since R
did not appear in (13.4.6),
the sum of the works performed by all
this formulamay be obtained without
introducing R into the intermediate voltage sources is equal to the shaded
transformations. Indeed, for the circuit area in Figure 13.4.2. We
leave it to
of Figure 13.2.2, <p B+ +
cp fl cp c 0, = the reader to find the efficiency for
whence = — E — c = E — qlC.
cp H cp <p 0
the case where the charging process is
Therefore dA = (E — q/C) dq. Inte- divided into n stages:
0
grating this expression from q = 0 to
q = Qo = E 0 C yields E — -2n
^ i
-
i
F
^2 — —^
2(
y

A = ±CE%. E*
3cp
,
En =(f>.

The last derivation holds true also for

the case where the resistance R varies In the foregoing, one plate of the
with time, while the previous deriva- capacitor was grounded, that is, its
tion^ hpJiLtrau^anl^ fan. 9* const ant. , sinco, potential was a»i A =
0. Here., the energx
only in this case could the formulas of of the capacitor depends on the po-
Section 13.2 be applied.
In order to reduce losses in charging
a capacitor, we should have done as
follows: first take a voltage source
with small emf E x and charge the
capacitor to the potential E x then ,

disconnect the first voltage source and

connect a second source with greater
emf E z Having charged the capacitor
.

to potential E 2 disconnect the second

,

source and connect a third with emf E 3i

 )

430 IS Electric Circuits and Oscillatory Phenomena in Them

tential of the other plate, (p 2 , that is, is q0 — C 0 cp 0 and the energy of the
capacitor is

TJ/ ffp
~ ,2
“ 2C 0 *

neither plate is grounded, the

If
energy of the capacitor depends on the Let us disconnect the capacitor from
potential difference across the plates, the voltage source and move the lever.
(p c ,
that is, W= —Cyc- Indeed, we The charge will then remain constant
(the potential varies in inverse pro-
know that the charge q on each plate portion to the capacitance), and the
of the capacitor depends on the po-
energy will change:
tential difference, the charges on the
plates being equal in magnitude but
opposite in sign:

Qa — Ccp c q b — — Ccp c = — £a,

,
The electric energy of a charged
dq A = — dq B
capacitor is similar to the elastic energy
.

of a spring: W
x increases if an exter-
(,

When computing the variation of nal force applied to the lever does work.
energy in the charging process, one Then the external force overcomes the
has to take into account the variation forces with which the plates of the
of charge on both plates. Let the po- capacitor act on the lever. Contrariwise,
tential of plate A be <p 1? the potential if W (x) decreases, the lever is displaced
of plate B be cp 2 and cp x cp 2 cp c ,
— = . and does work in opposition to the
Then external applied forces. We may con-
clude that the force with which the
dw = q> x dq A + <p 2 dq B = <p x dq A
plates act on the lever is
— cp 2 dq A = (q)! — <p 2 ) dq A
dW
- cp c dq A .
p= dx dx
q%
2 C (x)

But since (p c = q A /C ,
we can write ~ ql dC(x) __ cp
2 (x)
dC(x)
2 [C (x)] 2 dx 2 dx *

dW = ±qA dqA . (13.4.7) (13.4.8)

Integrating (13.4.7) from 0 to q A yields The force is directed toward increas-

ing capacitance. For instance, if the
^=#=4 ^- capacitor consists, say, of two identical
parallel plates, the capacitance is in
Knowing the expression for the ener- inverse proportion to the distance be-
gy charged capacitor as a function
of a tween the plates. This means the ca-
of the capacitance, we can find the pacitance increases when the plates are
mechanical forces acting between the brought closer together. Quite true r
plates of the capacitor. Imagine the because when a capacitor is charged,
plates be connected mechanically
to the charges on the plates are of opposite
with some kind of lever and suppose sign and so the plates attract with
that the capacitance C depends on the more force if they are close together.
position of the lever. If the position of Formula (13.4.8) enables us to find
the lever is characterized by the value the force in more complicated situ-
of the x coordinate, the capacitance is ations, as, for example, in the case of a
a function of a;, or C C {x). At a — variable capacitor, in which one plate
definite position x 0 of the lever the can move in and out between two fixed
capacitance is C (x 0 ) = C0 . If in this plates.
position the capacitor is charged to the It is important to note that we took
potential cp 0 the charge on the plates
,
the derivative dW/dx for a given con-
 . — .

13.5 Inductance Circuit 43H

stant charge q However, it is not per- Fa J

missible, when seeking the force by the
formula F = —
dW/dx, to take the
i| O-,
9
derivative of W=
C {x) cp 2 /2 assuming
cp constant and having regard solely
for the fact that C depends on x We
would then obtain an incorrect sign for Figure 13.5.2

the force. Indeed, if the capacitor is

disconnected from the voltage source, Wesee that the current in the circuit
(p is not constant: cp qlC and C = = of Figure 13.5.1 falls oft exponentially.
C (x). If the capacitor is connected The current diminishes e-fold during*
to a voltage source, then cp remains a time interval T = L/R .

constant as the capacitance varies. But Let us verify the dimensions of L/R.
then the charge q varies, which means L
is measured in henrys, that is,
that a current is flowing through the V*s/A and R in ohms, or V/A. There-
voltage source, that is, the voltage fore,
source is doing work equal to cp dq [L/R] = (V*s/A)/(V/A) = s,
(as C increases). Hence, when applying
the law of conservation of energy for so that L/R indeed has the dimensions;
constant cp and variable capacitance, of time. We
will call L/R the decaij
one has to take into consideration not time. In the circuit diagram shown in*
only the variation in the energy of the Figure 13.5.1, in which there is
capacitor and the work of the force voltage source, the current tends to
but also the work done by the voltage zero with the passage of time. How te
source. set up an initial current of j0 will be-
discussed later on.
For the present, let us consider a cir-
13.5 Inductance Circuit
cuit consisting of a source having an.
Let us consider a circuit consisting of emf 0 £
of a resistance /?, and of an
,

resistance R and inductance L (Fig- inductance L (Figure 13.5.2). From the

ure 13.5.1). By formula (13.1.11), condition
cp* + <Pl = 0. (13.5.1) <Pe + 9b + <Pl = 0,

Since cp# = Rj and =L dj/dt ),

using (13.5.1) we find
cp L
that
(
recalling that cp# = —E 0 ,
we find that

-E + Rj + L% = 0.
0 (13.5.4>

Thus, the current in the circuit of Fig- We rewrite this equation as

ure 13.5.1 satisfies the equation
di = JL (Jk—A
dj_
__ __ .
dt L \ R J
1
'

dt~~ L h (13.5.2)

This equation is similar to Eq. (13.2.5)'

The solution to this equation is
and is solved by the very same proce-
j(t)=j 0 e-W L )t. (13.5.3) dure. We get

wL
L / (T) = -^+Ae-WU t
, (13.5.5).
v onrm
R where A is determined by the initial
condition. Suppose the switch is closed
at t =
0. Then / (0) 0 because there =
was no current flowing in the circuit
Figure 13.5.1 when the switch was open. Given this
432 13 Electric Circuits and Oscillatory Phenomena in Them

say that the voltage is absorbed entirely

condition, we that A
find =
—EJR, by the inductance. As time passes, the
and (13.5.5) assumes the form
current approaches the constant value

= -|> (l_ e -(R/Df). EJR dj/dt tends to zero, and the volt-
,

j (13.5.6)
age is absorbed by the resistance.
It is interesting to compare the solu-
In the course time the current ap-
of
tions that formula (13.5.6) yields for
proaches the value
the same E
0 but different values of R
and L. Let R
x be small, 2 great, L y R
small, and L 2 great. For different com-
binations of R
and L we get four curves
nrs'^aruV fs~ Axaepeiia^n tire Aicnlci- of current as functions of time (see
ance L and is simply obtained from Figure 13.5.5). The final current / (oo)
Ohm’s law in a circuit with an emf E 0 depends on R alone; it is the same for
and a resistance R. However, this the pairs, R
L x and R±, L 2 j (oo) is
x , \

current sets in not at once but gradu- also the same for the pairs R 2 L x and R 2 , ,

ally (Figure 13.5.3), and the time re- L 2 The initial rate of buildup of cur-
.

quired for / (oo) to set in depends on L :

rent depends only on the inductance L
in time L/R the current becomes and does not depend on the resistance:
0.63; (oo), in time 2 (L/R) the current the curves specified by the pairs R u
becomes 0.86/ (oo), in time 3 (L/R) L x and R 2 L x and R L 2 and R 2 L 2
, x , ,

the current becomes 0.95/ (oo), and so all merge at the point O (0, 0).
forth. Thus, the ratio L/R can also be On dimensional grounds, it is clear
called the buildup time for the current. that the steady-state current is pro-
According to the basic equation portional to the initial rate of current
(13.5.4), the sum of the difference of buildup and the time of buildup. For
potential across the resistance, Rj, and our definition of buildup time, the
across the inductance, L (dj/dt), is equal formula is correct without any addition-
to the emf E 0 It is interesting to follow
.
al coefficients. Indeed, the initial rate
each term separately. They are shown of current buildup, (dj/dt ) t=0 is equal ,

in Figure 13.5.4. At the initial time, to EJL, and the buildup time, T,
/
= 0, R/ = 0, and E0 = L (dj/dt). We is equal to L/R, whence the established
current is

/ (°°) = T (dj/dt) t= = jjr o —£ = -g .

Now comes the question we posed

at the beginning of our discussion of
setting up the initial current / 0 in the
circuit in Figure 13.5.1. can take We
the circuit diagrams of Figure 13.5.6.
Figure 13.5.4 We start by closing switch A and leav-
13.6 Breaking an Inductance Circuit 433

is very small. And at point A there will

appear a very large (in absolute value)
“
negative potential:

A
(p = L^-~—dt T
.

Figure 13.5.6 The potential difference across the re-

ing switch B open. Then a current will

sistance R (it is equal to Rj) and the
emf the voltage source change but
of
flow and soon attain the value E 0 /(R +
slightly when the switch is opened.
RJ in accordance with formula (13.5.7).
For this reason, the great potential
We choose E 0 in such a manner that
difference that appears across the in-
E 0 /(R + Ri) = 7o* We for the wait
ductance L when the switch is opened
steady state to set in when the current
falls entirely on the switch; by this
is equal to / 0 with A closed and B open.
we mean that the potential difference
Then in this state we close B and
across the open plates of the switch
open A. The result is the circuit shown
becomes very great, of the order of
in Figure 13.5.1. At the initial instant
of time (when closing B), a current ; 0
LjJ t, and can exceed many times over
the emf of the voltage source, E 0 If .

flows. The potential at point 1 (see

the potential difference is great, the
Figure 13.5.5), prior to closing the
air gap between the open contacts will
switch, is cp x =
0, since in the steady
break down and a spark will jump
state, when j 0 is constant, the voltage
across.
drop on L is zero. Prior to closing the
The problem of current variation in
switch, the potential at point 2 is
a circuit when a switch is opened proves
equal to cp 2 =
Rj 0 When switch B
-

to be very complicated; this is due to

is closed, point 2 becomes grounded
the involved nature of the laws of elec-
and so the potential at point 2 is
tric discharge in air between plates.
<p 2 = 0. There is then a
corresponding
readjustment of potentials at all other
Indeed, prior to breakdown, when cp <
cp b there was no current; but when
points of the circuit. In particular, the ,

—Rj. = breakdown occurs, the resistance of the

potential at point 1 is now (p x
spark falls drastically, a big current
flows at a potential difference consider-
13.6 Breaking an Inductance Circuit ably less than cp b . Here we only note
the basic fact: large potential differences
Above we considered the process of
arise in inductance circuits in circuit
a steady-state current setting in in the
breaking. When such a circuit is closed,
circuit shown in Figure 13.5.2, which
the potential difference does not exceed
consisted of a voltage source, a re-
sistance R, an inductance L, and a
j
E o (the emf of the source) anywhere.
"Tne 'two circuits “fen own ‘m figure
switch. Figure 13.5.3 shows the curve
13.6.1 give a quantitative idea of the
of current buildup when the switch is
—
0. In time, the cur-
phenomenon of sudden brief increases
closed at time t
in potential difference. These circuits
rent reaches the value j 0 E 0 /R What = .

differ from that of Figure 13.5.2 in

will happen now if we suddenly open
that current can also flow in inductance
the switch 5? If the current ceases to
flow in a very short time t, the deriva-
L when switch B is open, so that circuit
breaking occurs without any spark.
tive of the current with respect to time
However, if the resistance R is much
can be represented as follows:
greater than the resistance r, a large
dj
^ j(t + T) — j(t) _/o
’
potential difference appears across the
dt T T
inductance at break.
which is to say, that the derivative will As an example, let us consider the
be very great in absolute value if t circuit shown in Figure 13.6.1a. We
28-0946
 434 IS Electric Circuits and Oscillatory Phenomena in Them

Now let us examine breaking the

circuit after a time lapse of t^>
after the circuit is closed, that is to
say, after a constant current / «, Ejr — :

has been set up in the circuit. When

the circuit is broken j r jE 0 and = =
Jr +Jl =
0, whence j R jL This = — .

means that the entire current passing

through L must pass through R in
the reverse direction. As before, of
course, (pL^^n* Therefore, cp r Rjr = =
— RJli or cp L = —
Rj l But since cp L . =
L ( djjdt ), it follows that L (dj L /dt) =
— R}l-
assume R
r. If the switch is closed, We have thus arrived at Eq. (13.5.2),
at an arbitrary time the current in the which is quite natural since the right-
left portion of the circuit (r, E) is hand part of the diagram in Figure
equal to the sum of the currents in the 13.6.1a (after the circuit is broken)
parallel connection RL : does not differ from the circuit diagram
in Figure 13.5.1.
It — Je = Jr + 7l- The current diminishes by a factor
It is then always true that cp L = cp^. of e during time x 2 LIR. Here, =
Let the switch be closed at time x2 T 1? since R r. At the time of

t =
0. At this instant all the current break, the current has the value j Loo =
will go through the resistance R, so that Ejr After the break has taken place,
.

7ro 7jro =
EJ(R =
r), by Ohm’s law. + but prior to the current falling off
Then perceptibly, that is, for break time

— Eq R R r
t <t 2, we get
TrO “E <PrO
0 »

+ ’

hence <Pr = <Pl ^ — Rj L = oc —E+.

d r
ik = Vro_
dt t= o L L 7? +r Thus, at break we can obtain a po-
After a sufficient time lapse after mak- tential difference that is many times
ing the circuit, a constant current will greater than the emf of the voltage
flow. In the steady state, the entire source. This fact is extensively employed
current will go through the inductance. in engineering, in particular in the ig-
Indeed, if the current / does not vary nition systems of internal combustion
with time, then djldt 0, therefore = engines. Observe that this large poten-
<Pl = 0, and so cp# = 0, whence jR = 0. tial difference occurs over an extremely
In the steady state, cp r OO E o and small time interval.
jT oo = /loo — easy Ejr, whence it is The foregoing is a rough considera-
to obtain the order of the time tion of the problem without the use of
during which the steady-state current derivatives and other tools of higher
sets in: mathematics. An exact consideration
dJ l of the problem of closing a switch in
7 LOO — 7lo :

dt t=0
'4
X* or
the circuit of Figure 13.6.1a yields the
Eo R following. Proceeding from the rela-
L R+r tions
whence
L(R + r) L <Pe + <p r + <Pi = 0, j = jR + jL ,

rR ~ R *
Wr — <p Li
 *

The Energy 435

13.7 of Inductance

of heat Q =
Rj 2 is released on the re-
sistance in unit time.
What is the source of the electric
energy that is converted into heat in
the resistance? The energy is given up
by the inductance, which has a certain
supply of energy.
Let us find this supply of energy by
Figure 13.6.2 considering the elementary circuit dia-
gram shown in Figure 13.5.1 and let us
calculate the entire thermal energy re-
we get the differential equation
leased onR. Suppose at the initial time,
dj L
i
rR .
= E0 R t =0, this circuit has a current / 0 The .

jL
dt (R + r)L (R + r)L *
current will then decay in time in ac-
'\nndrcmv5b ’w’Jhi
+
hfb Vaw
At the initial time, t = 0, the current
flowing through the inductance is zero: j ( t) =j 0 e- Rt f L .

fL =
0 at t 0. Therefore, = The quantity of energy released on the re-
sistanceR in unit time, that is, the rate
= ~r~ {
4 eXP
(F^)T*]} of energy release (dimensions: W/s),
is the instantaneous power output h.
= (13.6.1) Using the time dependence of / given
above, we find that
The current in the circuit is
h= Rf =•--
Rfie- Rt ' L .
(13. 7.1)

J = Jl + J Knowing h, it is easy to find the total

amount of heat released from time t 0 =
+ (
13 6 - 2 )
-
to infinity ( t =
oo), that is, to complete
decay of the current. To do this, it
In Figure 13.6.2 the approximate solu- suffices to integrate (13.7.1) from t 0 =
tion is shown by the broken line and the to t —
oo. This yields
exact solution (13.6.2) by the smooth oo
curve.
We advise the reader to examine the Q= (
Rjle- 2Rt ' L dt = R)\ f
e~ 2Rt / L dt
0 0
process of variation of current and po-
tential difference at make and break RilL __ L)%
(13.7.2)
in the circuit of Figure 13.6.16. It is
2 R 2 *

useful to solve the problem twice: once This heat is equal to the supply of
by setting up the differential equation energy of the inductance through which
and seeking its solution in the form of the current j 0 flows. This supply of
an exponential function, and the second energy does not depend on the magni-
time in the approximate fashion, as we tude of the resistance R. An inductance
did for the circuit depicted in Fig- L with current ; 0 has a definite supply
ure 13.6.1a. of energy which, ultimately, is trans-
formed completely into heat, irrespec-
tive of the magnitude of the resistance
13.7 The Energy of Inductance
R ,
which only affects the rate of trans-
We have seen that in a circuit con- formation of energy into heat but not
sisting only of an inductance L and a amount of energy.
the total
resistance R, current continues to flow Formula (13.7.2) can also be obtained
after the voltage source has been dis- by considering the process of current
connected. The current gradually falls buildup in an inductance. Indeed, the
off in time. In the process, a quantity power of the current (that is, work per
28 *
 / " ~

436 13 Electric Circuits and Oscillatory Phenomena in Them

unit time) This work is

is equal to cp /.

done by external sources and of voltage

goes to increase the inductance energy
W:

(13.7.3) Figure 13.7.1

Using the fact that cp = L (dj/dt), we conduction electrons, or roughly 10 22

find,from (13.7.3), that electrons per gram, or 10 25 electrons per
kilogram. Copper has a density of about
dW dj djn
8 g/cm = 8 x 10 kg/m 3 and so one
r .
1
r 3 3
1 *
(13.7.4) ,
dt dt '
2 dt
cubic meter of copper wire contains
roughly n = 8 x 10 28 conduction elec-
We will assume that / = W=0
0 and
at t = 0, and j = ; 0 and W = W at 0
trons. Imagine a piece of copper wire
of length v dt and a cross-sectional area
t = £0 . Then, integrating (13.7.4) from
t
— 0 to t = £0 , we obtain S O (Figure
to the left of cross section
13.7.1). the electron velocity is v
If ,

then through S there will flow Snv dt

w^=4- electrons in time d£. 13 9 In time dt the -

electrons at cross section A will move

To be specific, imagine the circuit in to cross section 0, which means that
Figure 13.5.2 (cp == cp^) and carry out
during this time all the electrons in the
the detailed computations for buildup
volume between O and A will pass
of the energy of inductance. In the
through O. That is, they will pass
steady-state mode, when the current has
through a cylinder of altitude v dt and
reached a constant value 0 the potential ,
base S.
<p A is zero and the energy of inductance
Denote by e the electron charge in
does not vary, but the source of emf
needed to maintain the constant current
coulombs, e = —
1.6 X 10~ 19 C. The
quantity of electricity which the Snv dt
j o must continue to generate energy, electrons transfer in time dt is equal to
which is released as heat on resistance R. the current in amperes multiplied by the
The energy of inductance is pro- W time dt. Therefore, Snvedt= ] dt, whence
portional to the square of the current, = Snve, or y = j/Sne.
j Substituting
which is to say, it is proportional to the = 1 A, S = 10 -7 m
7
2
8 x 10 28
-3
m
— 1.6
, ,
square of the rate of motion of the elec-
e= x 10 -19 C, we obtain
trons. Therefore, externally, re- W
sembles kinetic energy. But is the W v = ~ 8x 1
0" 4
kinetic energy of the electrons? 10’ 7 x8xl0 28 xl.6xl0- 19 *

Let us consider the orders of magnitude

of W
and the electron energy. Using The dimensions are that of velocity:
a copper wire of length 100 and dia- m [v] = A/m -m 2 _3
*C = A*m/A*s = m/s.
meter 0.35 mm
(cross-sectional area
equal to approximately 10 7 2
), we m Now we have
to calculate the kinetic
can wind a coil having an inductance of energy the (conduction) electrons.
of
0.02 henry. A current of 1 A flowing in The electron mass is 9 X 10
31
kg. m
this coil will release W = 0.02 x The total number of electrons moving
l
2
x 0.5 10 J. —
will now find
-2
We in the wire ( S
"
10 7 2
and l = m =
the kinetic energy of the electrons.
"
10 2 m) is 10 2 x 10 7 x 8 x 10 28 ^
We will assume that for each atom of
copper there is one electron carrying
current (the conduction electron). The \y e have in view the mean velocity of
1. 3.9

theirmotion in the direction of current flow

atomic weight of copper is about 63, so and not the velocity of random thermal mo-
63 grams of copper contain 6 x 10 23 tion.
13.7 The Energy of Inductance 437

10 24 . The kinetic energy of one electron The essential practical difference be-
is tween a capacitance and an inductance is

~ — X 9 x 10
that a capacitor disconnected from a
-31 10" 8
mv 2 x 64 X voltage source can retain its supply of
—
energy for a very long time hours or
~3x 10" J, 37
even days. The discharge time of a ca-
pacitor is equal to RC, where C is the
while the total kinetic energy of the
capacitance and R the leakage resis-
electrons moving in the wire is
tance. Using good insulators, we can
obtain enormous values of i?, that is,
r=rc-^-~io 24 x3x 10- 37
long discharge times. An inductance in
=3x 10- 13 J. the form of a coil and short-circuited
(minimal resistance) retains its energy
Thus, the kinetic energy of the elec- (if current is flowing) for only a fraction

trons constitutes a minute fraction of of a second.

the inductance, although it depends on The decay time of the current in an
the current via the same law (it is pro- inductance is of the order of L/R, but
portional to / 2 ) as the inductance ener- even with the best conductors (copper,
gy. Physically, the inductance energy silver) it is impossible to make L/R
is the energy of the magnetic field which greater than a few seconds for an ordi-
appears in the coil when current flows nary laboratory-type coil. It will be
through it. noted that if we increase the number of
Let us point out some similarities and turns in the coil for a given volume by
differences between capacitance and in- using thinner wire, L will increase, but
ductance. Both can serve as reservoirs of so will i?, their ratio, however, to within
energy, both can be used to accumulate order of magnitude, does not change.
electric energy from a weak primary Therefore, under laboratory conditions,
source and then release it quickly at inductance is conveniently used for
the required place and time. increasing voltage but not for long-
A capacitor can be charged with a term storage of energy.
small current j 1 during a long time t x ;
Circuits involving capacitances and
then rapidly discharging it through a inductances can be used to accumulate
small resistance during a short time t 2i energy from, say, a flashlight battery
we can obtain a large current / 2 ~ which with an internal resistance of
fitjtz,and the potential difference across several ohms yields a few volts so that
the capacitor does not exceed the emf the maximum power output is of the
of the primary source. In other words, order of one to two watts. Using circuits
a capacitor enables us to increase the of this kind, we are able to obtain pow-
current but not the voltage. ers up to hundreds of kilowatts. But
We
can send a large current through a power output of this kind lasts for a
an inductance under a small voltage time interval of the order of 10~ 6 s.
(small emf) E 0 of the primary source. It has been noted that the electric
The only requirement here is that the energy in an inductance is quickly con-
resistances of the inductance and the verted into heat due to resistance. This
primary source be small. Then it takes assertion holds true for coils of the
a comparatively long time t 3 for a large ordinary laboratory kind and at ordi-
current to build up in the inductance. nary (normal) temperatures. In two
When the inductance is shunted by a extreme cases, however, this does not
high resistance, we can obtain a large hold true.
potential difference cp for a short time (1) At very low temperatures of the
i4, with cp ~E 0 t 3 /t 4 . An inductance order of —
260 °C down to absolute zero
enables us to increase the voltage but (— 273 °C), many metals (for instance,
not the current. lead, mercury, but not copper) pass in-
438 IS Electric Circuits and Oscillatory Phenomena in Them

to what is known as the su perconducting between the resistances of the coils:

state. Their specific resistance (resistiv-
ity) becomes exactly equal to zero. R\ = pl 1 /S 1 , R2 = pills?,!
The Dutch scientist Heike Kamer-
lingh Onnes (1853-1926), who discov- where p is the resistivity (specific re-
ered this phenomenon in 1911, observed sistance) of the coil material, l is the
a constant current in a ring circuit of length of the wire, and S the cross-
superconducting material that lasted sectional area of the wire. Geometrical-
many days without any decrease in in- ly, it is clear that l 2 == nl x and S 2 =
tensity. The presence of current in such n 2 S 1 and, hence, that R 2 — RJn. The
a ring circuit is detected via the magnet- resistance is inversely proportional to
ic field of the current. n that is, to the dimensions.
,

The practical application of super- It can be proved that the inductance

conductors is limited not only by the of the large coil is exactly n times the
difficulty of generating low tempera- inductance of the small coil, L 2 =
tures. A strong magnetic field converts a nL x which means that increasing the
,

superconductor to the normal state linear dimensions of the coil n times

(with finite resistance). This is why large increases the inductance n times, too.
currents cannot be transmitted through The decay time of the current, t, is of
a superconductor. 1310 the order of L!R\ consequently, % 1 =
(2) The relationship between induct- L 1 /R 1 and t 2 L 2 IR 2= =
n 2 L 1 /R 1 =
ance and resistance and the conditions n2 tx .

of current decay vary drastically when Thus, the decay time of the current
all the dimensions of the coil are increased , is proportional to the square of the di-
particularly when passing to astro- mensions. If the earth were to consist
nomical phenomena (on the astronomi- of copper, the decay time of a current
cal scale). 1311 flowing in it would be of the order of
Picture two geometrically similar 10 15 to 10 16 s, or approximately 10 8
coils, one of which is n times the other in years.
size, the number of turns in both being The conductivity of ionized gases is
the same. In the large coil, the diame- of the same order of magnitude as that
ter of the coil is n times greater, but of copper. For this reason, the decay
so is the height of the coil and the diam- time of a current in astronomical phe-
eter the wire. Suppose the coils
of nomena is enormous. This means that
are made of the same material. Quan- resistance and Ohm’s law do not play
tities relating to the small coil will be any role whatsoever in these phenome-
labeled with the subscript 1, those refer- na. Recall Figure 13.5.5: the current
ring to the large coil will have the sub- on the initial portion of any curve de-
script 2. Let us calculate the relation pends on L alone but not on i?, and in
astronomy we are always on the ’’ini-
tial portion.”
i3 10
.
n 1901 an a iioy was discovered of the
j Terrestrial magnetism
is due to the
rare element niobium and tin in which a cur- magnetic the current flowing
field of
rent up to 100 000 A/cm 2 and a magnetic
field up to 25 T (that is, 250 000 gauss) in the viscous molten mass of the cen-
were not able to destroy superconductivity. tral core of the earth. The slow motions
At present superconducting magnets are wide- of this molten mass in the magnetic
ly used in science and technology, for instance, field sustain the currents, just as the
in particle accelerators. There is reason to
believe that in the near future superconducti-
motion of the armature in a dynamo in
vity will be employed in electric transmis- a magnetic field sustains the current
sion lines. in the armature winding and in the
13 11
Compare with Exercise 13.2.3; for
-

winding of the electromagnet. The

a circuit consisting of a capacitance and a resis-
tance the discharge time does not change
same is true of the magnetic field of the
when all dimensions are altered. sun. The theory successfully predicts
 AA

13.8 The Oscillatory Circuit 43

the periodic changes in the direction of describe oscillations with a circular fre-
the sun’s field (the half-period is 11 quency 1 /]/" LC The oscillation period
. is
years).
T= ~ = 2nYLC. (13.8.4)
13.8 The Oscillatory Circuit
Let us check the dimensions in
Let us consider a circuit consisting of (13.8.4):
a capacitance C and an inductance L
(Figure 13.8.1). Let point B of the cir-
1C] = F = C/V = A-s/V,
cuit be grounded. By formula (13.1.11), [L] = H = V/(A/s) = V-s/A,
<Pc + cp L = 0, where cp L =L (dj/dt)
and cp c = q!C. The voltage drop on the so that [LC] = s and T cc LC does
2
Y
capacitance, cp will be denoted simply indeed have the dimensions of time, s.
c ,

(p. Then Let us examine in detail the solution

to Eq. (13.8.2). The solutions cp =

+ lJ- = 0. (13.8.1)
A sin co t and <p = B cos (at are actual-
(P
ly indistinguishable since the sine
Note dqldt and so djldt
that / = = curve is obtained from the cosine curve

d^qldt 2 since d^qldt 2

. But C i\!dt 2 ), = (<
by a shift along the t axis; the two can
it follows that, using formula (13.8.1),
be combined into a single expression:
we find that = B cos
cp ((at + a) (13.8.5)

<p + LC-^ = 0, or (cf. Section 10.2). The amplitude B of

the oscillations and the initial phase a
(13.8.2)
dhp (in Section 10.2 we denoted this quan-
dt 2 tity by cp) may be arbitrary. For a
We considered a similar equation in given cp (t) we find the dependence of
Chapter 10 in the study of mechanical current on time:
vibrations. It was established that the
functions A sin (dt and =
cp = j = C 4jj^= — CB(a sin (cot + a). (13.8.6)
(p
B cos (dt are solutions to Eq. (13.8.2)
Let us find the energy of capacitance
for arbitrary A and B and a suitably
and the energy of inductance:
chosen co. Let us verify this, say, for
=
A sin (dt and in passing we will
cp

define co. Substituting cp and d\!dt 2

Wc = = £f- cos* (at + a),
into (13.8.2), we get
55i.i Q . (w , + a
— ALC(o 2
sin (dt = — sin co£,
).

Substituting the expression (13.8.3) for

or, canceling out — sin co t, we get co yields
LC co 2 = 1, whence
WL = sin 2 (cof-f-a).

"W- (13 8 ' '

3)
The total energy is independent of
Consequently, for a solution to time, as was to be expected. Indeed,
Eq. (13.8.2) we have functions that
P= W c W L = [cos 2 (cot-fa)

+ sin 2 (cot-fa)] =-^i.

To summarize, the motion of charges
in an LC circuit is similar to the mo-
Figure 13.8.1 tion of a mass attached to a spring.
440 13 Electric Circuits and Oscillatory Phenomena in Them

The energy of a charged capacitor may

be likened to the elastic energy of a
spring, which is at a maximum when
the mass is in the extreme position of
maximum separation from the equilib-
rium position. The energy of an induc-
tance may be likened to the kinetic
energy of a moving mass. When the
Figure 13.8.2
charge on a capacitance is equal to ze-
ro, the current reaches its maximum
whence (as well as from (13.8.6)) we
(in absolute value), at this instant the
also obtain
capacitance energy is zero and the in-
ductance energy is equal to the total jm*x = Vn+(C/L)<?l (13.8.9)
energy (cos
2
(co£ a) 0 and + =
sin 2 (cd£ a) +
1). =
This is exactly (note that to = Y 1/LC ,
in view of
what happens in the oscillations of a (13.8.4)).
mass on a spring: when the mass passes The circuit diagram for generating
through the position of equilibrium, oscillations is shown in Figure 13.8.2.
the potential energy is zero and the Here we have a source of voltage with
kinetic energy is equal to the total ener- emf E 0 If we close A and leave B
.

gy of the oscillations. open, after a lapse of time x BC

Let us use the term general problem after closing the circuit the capacitance
for the problem of finding the potential will be charged to potential 0 Open E .

in a circuit provided that at the ini- A and at time t =

0 close B. Then oscil-
tial time t = 0 we have j = j 0 and lations in the LC circuit will set in
cp = Neither the particular solution
cp 0 . with cp =
(p 0 =E
0 and at t 7=00. =
<p =A sin out nor the particular solu- Note that with these oscillations, the
tion cp — B cos (x)t enables us to solve potential difference across the electrodes
the general problem. To solve the gen- of the open switch A will vary pe-
eral problem we will need the general riodically from 0 to 2 E 0 .

solution (13.8.5), with the two (inde- There is another way of setting up
terminate) parameters B and a at our oscillations in the circuit of Fig-
disposal. ure 13.8.2. First close both switches A
Setting t = 0 in (13.8.5) and (13.8.6) and B. The current flowing in the cir-
yields cuit will be 7 0 =E
0 /B. At t 0 open =
A. Then oscillations in the LC circuit
cp (0) — B cos a = (p 0 ,
will set in with cp 0 =
0 and j 0 = EJB
/ (0)
= — CB 0 sin a - 70, at t =
0. For these oscillations, the

whence we find the solution with maximum amplitude of the potential

given and will reach
cp 0 j0 :

B = Vvl + ti/(C<») 2 ,
y ma x = ioVuc=Eo^-Vuc.
tana^ — ; 0 /C(ocp 0 . (13.8.7)
It will be recalled that in a circuit
We already know that for such oscilla- without capacitance, when we break
tions the energy is conserved and, hence, the circuit containing inductance L,
"hie q/dieitcfcli liffrerenue irevtbro'peh on
is equal to the initial energy C^\12
2
+ the switch is the larger the greater the
Lj J 2. The expression for the energy
may also be written as C5 2 /2, which resistance of the air gap between the
enables writing the first formula in electrodes of the switch. When such a
circuit (with no capacitance) is broken,
(13.8.7) as
there is always a discharge in the
B = ^=V^l + (LIC)jl, (13.8.8) air gap between the contacts of the
 2

13.9 Damped Oscillations 441

Figure 13.8.3 Figure 13.9.1

switch. In the case of a capacitance,

the maximum potential difference be- When a resistance is present in the

tween the electrodes of switch A does not circuit, electric energy is converted in-
exceed a definite value, E 0 to thermal energy. The power output of
cp max If + .

the resistance, h, is given by the fol-

this value is less than what is required
to initiate the discharge in the air gap
lowing formula:
of the switch, there will be no discharge. h= Rj 2 = Rjm sin 2 a>t
We say that the capacitance C ex-
tinguishes the discharge when an induct- = (l-cos2©f)- (13.9.4)
ance circuit is broken. Note that the
The power output in the case of ele-
quantity
than unity.
R~ X L/C may be greater
Then by opening switch B
Y ctric oscillations does not remain con-
a quarter-period after opening A, we
stant: over each period (cycle), h twice
obtain a potential on capacitance C reaches a maximum and twice becomes
zero (the sign of course does not change,
that is higher than the potential on the
voltage source, E 0 .
that is, the power output is always
positive). Let us find the average value
Exercises of h over one period. From (13.9.4) we
find that
13.8.1. Consider the variation of potential
with time in the circuit shown in Figure 13.8.3.
Determine the greatest value of cp and the time
h= ~ 2- (1-cos at).
required to attain it. Assume that switch A
is closed at time t 0. = Recalling that the average value of the
13.8.2. In the preceding problem, find the cosine over one period is zero, we ob-
energy of capacitance and the energy released
tain
by the voltage source when cp is at a maximum.

13.9 Damped Oscillations

Heat release on resistance R can occur
Let us consider a circuit with a resist-
ance R in series with an inductance L
j
only as a result of reduction in the elec-
tric energy P. Therefore
(Figure 13.9.1). We assume R to be
small. If R is not taken into account at
all, we get the circuit diagram of Fig-
§-=-’> (13.9.5)

ure 13.8.1. If cp = cp 0 and 7 = 0 at

We assumed that R was small, and
t — 0, then, by (13.8.7), we have so h is small. The oscillation energy
cp = cp o cos cd£, / = / m sin (co£ + ji) = falls off slowly and an appreciable
change in energy becomes noticeable only
—7m sin (x)t , (13.9.1)
after several cycles. Considering time
with intervals that are large compared with

— /max “ ®" — (13.9.2)

the oscillation period T, we replace h
7m r •

\ LiL by h in the right-hand side of (13.9.5):

The total energy P is then Cip®/2 or,

(13.9.6)
using (13.9.2),
Since the energy P varies slowly, from
P = -^- l
, (13.9.3)
(13.9.3) we see that j m too is a slowly
 h

442 13 Electric Circuits and Oscillatory Phenomena in Them

varying quantity. -Expressing jm from lations (see Section 10.4, and compare
{13.9.3), we get the solution (13.9.12) of Eq. (13.9.11)
with formula (10.4.7)), so that here we
jm = yWL. (13.9.7) could simply refer to the results ob-
Using this, from (13.9.6) we get dPldt = tained in Section 10.4. However, we will
— CRiL) P (here ~ was replaced with = ).
not complicate matters by referring the
The solution to this equation is reader to Chapter 10 to compare the
different notations, and will briefly
P = P 0 e-WDt, repeat the required derivation.
where P is the
0 value of P at t 0. = Weput the expression (13.9.12) for cp
Therefore, according to (13.9.7), J m — and the expressions for the first and sec-
yWJLe-Rtlu. Then ond derivatives of cp that follow from
(13.9.12) into Eq. (13.9.11) and cancel
= Y ZPJL e~ Rtf2L sin the common factor Ae~ Kt out of all
j (co t + n). (13.9.8)
terms to get
Recalling that = cp cp 0 cos cot and cp 0 =
j m /C co, we get
LCX 2 cos (tit + 2LCk(ti sin (tit

— LC co
2
cos (tit = — cos (tit

<p=-^cosat
+ RC'k cos (tit + RC(ti sin (tit.

= ~^\ ~
/t e Rt/2L cos ©*• (13.9.9) For this equation to be valid for arbit-
rary t, it is necessary that the coefficients
Formulas (13.9.9) show
(13.9.8) and of cos (tit and sin (tit be equal sepa-
that if a small resistance is present, the rately on the right and on the left:
electric oscillations damp out via an
exponential law.
LCV - LC(ti
2
= RCX - 1, (13.9.13)

The solution given above was obtained 2LCl(ti = RC(ti. (13.9.14)

by means of an approximate calcu-
lation. Note that in this approximate The condition (13.9.14) yields the val-
ue of A, and by substituting this value
solution the relation j = C (dcp/dt) is
not satisfied, although it holds the more
into (13.9.13) we get co:

exactly the smaller R is. Let us now

try to solve the problem exactly. For
*= -&• I
13 9 ' 15 *
-

the circuit shown in Figure 13.9.1 we

= The constant Ain (13.9.12) has yet
have + cp+ cpjR 0, whence
to be determined. To do this we must
—
V + RJ + L-1 = 0, (13.9.10) specify the initial condition (say, (p
— Finally, knowing
cpo at t 0). cp (£),

and j = C {dq/dt). Substituting the we can easily find j = C (d(p/dt). We

expression for j and d]ldt into (13.9.10), then have
we find that ]
= — CAe~te s i n co^ A cos co£).

(13.9.16)
LC T$- = -i’- RC i <
13 - 9J1 >

Comparing the exact solution with

We
will seek the solution to this the approximate, we note the follow-
equation in the form obtained in the ing: (1) in the approximate considera-
approximate consideration, or tion of the problem we correctly de-
termined the number A, which describes
<p = Ae~u cos at ,
(13.9.12)
the rate of decay of the oscillations
where A, co, and A are constants that (however, the approximate solution
we have to determine. The procedure does not yield the dependence of fre-
that follows agrees completely with the quency co on resistance R), and (2) the
one encountered in the theory of oscil- formula for current is somewhat differ-
 ) —

13.10 * The Case of a Large Resistance 443

C —Wv
R
only for B that are not too large. In-
deed, from (13.9.15) it is evident that
ft
onrm
L
if B is greater than 2 YL/C, the formula
we have for co is meaningless since the
radicand is negative. In that case
Eq. (13.9.11) has a different kind of
Figure 13.9.2 solution. We will seek the solution in
the form cp =
Ae~ 0* (and accordingly,
ent from the one that was obtained in j
= —AC$e-$ ). Substituting
l
into
approximate fashion. (13.9.11) the expressions for cp and its
In exactly the same manner we can derivatives and canceling Ae~ 3* out
show that Eq. (13.9.11) has yet another
of all terms, we get LCp
2
—1 = ;

+
solution, This is a quadratic equation in
p. Solving it, we find that
<p = Be~ u sin co£, (13.9.17)

where co and % are the same as in *= “8 10J »

(13.9.15). The corresponding current is
-k±V§?-Tc- '

/ = CBe~u (co cos cot — X sin co£). The radicand in (13.10.1) differs in sign
(13.9.18) from the radicand in (13.9.15). Hence,
in those cases where it is impossible to
The sum cp = e~ Xt (A cos co t + find co we can find p, and vice versa.
B sin co£) of the solutions (13.9.12) and Formula (13.10.1) yields two distinct
{13.9.17) is also a solution to Eq. values of P, and so we can set up two
{13.9.11). It is only with the aid of solutions to Eq. (13.9.11), cp = Ae~ Pd
this sum that we can solve the general and cp = Be~^K Their sum is also a
problem: to find the solution to Eq. solution:
{13.9.11) with the initial conditions
<p = cp 0 and j j 0 at t =
0. Indeed, = cp = Ae~^ t
Be~^K (13.10.2)
for the coefficients A and B we then
Accordingly
have the equations cp 0 A and j 0 = =
— CAh C5cd, whence A
+ cp 0 and = /= -AC^e-M-BC^e-M. (
13 10 3 )
. .

B = (CXap 0 + j 0 )/C(d .

If cp — cp 0 and= j 0 at = 0, then, j t
Exercises assuming that = 0 in (13.10.2) and t

13.9.1. Find j ( t for the circuit of Fig- (13.10.3), we get A + B = and cp 0

ure 13.9.1 if C 1, L = 1, and R =0.1, 0.5, = —AC p! — BC P = 7o- We can find A 2

1. Assume that cp 1 and / = 0 at t 0. = = and B from this system of equations.
=
,

13.9.2. The same question if cp 0 and

Let us consider in more detail the
—
0 at t 0. =
j
13.9.3. Using the approximate method, expression for p. Let B 2]/ L/C. >
find the rate of decay X of oscillations in the Then
circuit shown in Figure 13.9.2 on the assump-
tion that R is very great. 4L
1
= JL 1/'
LC 2 L V '

R2C
13.10* The Case of a Large Resistance
can be expanded by the binomial theo-
The. casr, r/msidereil. hem a largp re- oi_ mm Sncdinm confine ojiv - WA
sistance is mainly of mathematical selves to two terms:
interest and is not connected with the
sequel. It may therefore be skipped in a 4L

first reading. i?
2
C — 2L \ R2C )

The solution of Eq. (13.9.11) ob- R. 1_

tained in the preceding section is valid 2L RC'
 i

444 13 Electric Circuits and Oscillatory Phenomena in Them

Therefore 13.10.2. Find cp (t)for L = 1, C= 1,

and R = 2, 4, provided that cp^ = 1 and /n = 1
o R R \ R l 1 at = 0.
~ ~ t
,

P1 2L 2L RC~ L RC L
since R is great, similarly, 13.11 Alternating Current

n ~ _RL R_ _ J_ We will now examine circuits in which

P2 “2 2L ^ _J_
RC “ RC *
the voltage source has an emf that
varies periodically with time with a de-
These values of and p 2 are famil-
finite given frequency co. These pro-
iar from Sections 13.1 to 13.5. Indeed,
blems are very important in radio-cir-
P x corresponds to current decay by the
cuit work. The frequency of alternating
law £-(#/£)*, which means that this is
current exerts quite a different effect
an RL circuit (see Section 13.5). The
on the passage of current through an in-
second root p 2 corresponds to current
ductance and a capacitance. The higher
decay by the law which means ?
the frequency, the faster the current
that this is an RC circuit (see Sec-
varies and the “harder” it is for the
tion 13.2).
current to pass through an inductance
Of mathematical interest is the particular and the greater the potential difference
case where the radicand in (13.10.1) is exactly that a current of a given intensity can
zero: set up. Contrariwise, the potential differ-
2
i? 1 ence of the plates of a capacitor is the
4 L2 ~ LC ’
smaller the greater the frequency. When
so that both roots, P x and P 2 ,
coincide. We the frequency is increased, the period
obtain only one solution to Eq. (13.9.11). diminishes and, consequently, the time
But in order to solve the problem with initial
— interval decreases during which the
condition q> q> 0 and j /0 at t= 0 we = current flowing in one direction can
need two solutions.
How can we find the second solution? Sup- charge up the capacitor. Therefore, as
pose that P x =4= p 2 but the difference fh p2 — the frequency increases, the charge on
is small. Then we have two solutions: e~® lt the capacitor decreases and so does the
and e~~^ zt . Their difference is also a solution. potential difference on the plates of
We write this solution as
the capacitor.
We have already pointed out (see Sec-
tion 13.8) that the movement of charges
Since p2 — iPi is small, it follows that
in an LC circuit may be likened to
)t
1 -|- (p 2 — (in the Taylor series the oscillations of a body suspended
only two terms are retained), whence from a spring: whereas in the case of
-e-V 2t ~ e-^t (p — pi). 2
vibrating body the distance of the body
from the origin and also its velocity
The last expression suggests that if p2 = fh
= vary periodically with time, in a cir-
P, the second solution must be taken in the
cuit the potential and current vary pe-
form cp =
Bte~^ 1 Substituting this cp into .

riodically.
Eq. (13.9.11) and noting that p R/2L, =
we see that the equation is indeed satisfied. The frequency with which a body vi-
Thus, when P x P2 P, we must take cp = = brates under the action of the elastic
in the form force of the spring (in the absence of
y = Ae~$ + Bte~$ any other forces) is called the natural
t t
.

This
frequency. Similarly, the frequency of
(p (and the corresponding/) permits solv-
ing the problem with arbitrary initial cp 0 oscillation of potential in an LC circuit
and /0 . is termed the natural frequency of the
circuit.
Exercises Developing this analogy further, we
can assume that if the circuit is con-
13.10.1. Find cp (t) for L= 1, C = 1,
and R = 2, 6, 10. Assume that cp 0 = 1 and nected to an alternating current network
Jo = 0 at t = 0. (which means the potential impressed
 p

13.11 Alternating Gurrent 445

on the circuit will vary periodically), oscillograph (a so-called loop oscillo-

we will have what is known as reso- graph with two loops or a cathode-ray
nance. What resonance means is that oscillograph with two beams). The pos-
the amplitude of oscillation is at a max- itive direction of current is shown by
imum when the frequency co of the an arrow. The voltmeter V measures
current is equal to the natural fre- cp = cp A —
closing one or an-
cp B . By
quency cd 0 of the circuit. The amplitude other of the switches we can investigate
increases sharply when the difference the current flowing through the re-
co —co 0 approaches zero. Resonance sistance, inductance, or capacitance.
actually does take place and we will Suppose the current varies with time
consider it in Section 13.13. in accordance with the law
For every two-terminal network (see
Figure 13.1.7) connected to an alter- 7 = 7o cos (<*>* + a). (13.11.1)
nating-current circuit, there is a de-
If this current flows through resistance
finite relationship between the poten-
i?, by Ohm’s law we have
tial difference and the current. Let us
find this relationship first for the simpl- cp^ = Rj = Rj 0 cos (co£ + a). (13.11.2)
est case of separated elements R, L
and C and then, in Sections 13.13 and For the sake of generality, let us write
13.14 for more complicated circuit. this as
We will consider alternating current <Pr (0 = <Pi cos (<*>* + a i)>
of a definite frequency co; as before, co
is connected with the period through where x = Rj 0 and a L = a.
c

the relation co = 2 n/T. For example, Let the current (13.11.1) flow through
in the USSR the standard current is inductance L. Then
50 cycles per second (or 50 Hz), that is,
T = 1/50 s and (o =
2n x 50 ^ <P l = L~- =— Lco/oSin (co£ + a).
314 s- 1 .
Set
We refer to the diagram in
circuit
Figure 13.11.1, which contains an am- <Pl = <P 2 cos ((x)t + a 2 ). (13.11.3)
meter A that indicates current / and a
voltmeter V measuring the voltage Then cp 2 = Lco/ 0 and a2 = a + n/2.
(difference of potentials). Suppose that
Indeed, we know
that cos (P + n/2) =
the ammeter and voltmeter are so iner-
— sin p for arbitrary P and so

tialess (high-speed) that they permit cos (co£ + a + n/2) = —sin (co£ + cc).
measuring the instantaneous value of
current at each instant and, hence, their Thus, in the case of alternating cur-
readings vary with a period equal to rent the relationship between the am-
that of the current. This experiment is plitude of current, / 0 and the amplitude ,

usually accomplished with the aid of an of voltage, cp 2 in the inductance is the

,

same as in a resistance equal to R 2 =

L co. If L is expressed in henrys (H)
and co in reciprocal seconds (s _1 ),
then R 2 will be expressed in ohms (Q).
Inductance differs from resistance in
that the curve of the voltage is displaced
a quarter-period from the current
curve (Figure 13.11.2). This is quite
evident from the formula

—sin (co£ + a) = cos (co£ + a + n/2)

Figure 13.11.1
= cos [co (t + n/2co) + a].
 ) »

446 IB Electric Circuits and Oscillatory Phenomena in Them

Figure 13.11.2

frequency in reciprocal seconds, we ob-

Let the function cos (co£ + a), to
tain R 3 in ohms.
which the current
is proportional,
In a capacitance, the voltage curve
reach a definite value at time tp.
is shifted forward with respect to the
cos (co^ + a) = a. current curve by a quarter-period (Fig-
ure 13.11.3). Thus, the curve of vol-
The function cos (co£ + a + Jt/2), to tage cp c in a capacitance is shifted in
which the voltage on the inductance is the opposite direction to the curve of
proportional, reaches the same value
voltage cp L in an inductance.
at a different time £2 , so that
For a given current, cp L and cp c are
cos (cd £ 2 + a + n/ 2 ) = a of opposite sign. If the curves of cp L
and cp c are brought to coincidence, we
= cos ((ot x + a).
will see that the current flowing through
Therefore, co£ 2 + Jt/2 = co^, whence the capacitance and the current flowing
£2 = h— 774, which
k/2co = — tx
through the inductance have opposite
means the voltage leads the current signs. Indeed, all formulas expressing cp
by one quarter of a period. as a function of t can readily be trans-
Quite naturally, we can add any in- formed into formulas expressing 7 as
tegral number of periods to £ x and write a function of £:
t2 = h — Z74 + T = £1 + (3/4) T or
cp = cos ot + a),
cp 0 (c
£2 — + (7/4) 77 The formula indi-
cates the smallest (in absolute value) cp
H = Rj cos (cd£ + 0 oc),

time shift that carries the current curve cp L = Lcd/ cos (a)£ + a + Jt/2)
0
into the voltage curve.
Let us consider the case of capaci-
= — Leo/ sin (co£ + a), o

tance. Here, 7 = C (dcp c /d£) and so

<Pc = 7^7 7° cos (<i>t + a — -j -

<p c =— 1 f
\ j dt == — 1 f
\ 70 cos (oof + a) dt
= 7o siD +a )

= sin (co£-|-a) (13.11.4) and

(the constant of integration is equal to

7 =
'

70 cos(u)f + a), jR = cos (at -f- a)

c but for alternating current it is

<p ,
7l =-£-cos(a>i + a-^-)
always true that cp c = 0). Writing cp c

as cp c = cp 3 cos (co£ + o^), we get cp

3 = = -^- sin (<at + a),
(1/Cco) and a 3 = a
7 0 jt/2. —
Thus, in an alternating-current cir-
cuit the relationship between the am-
7C = Cco cos [at + a +
cp 0
j

plitude of current and the amplitude of = — cpoCco sin ot a). (c -f-

voltage is the same on a capacitance as
on a resistance equal to R 3 = l/Cco. The opposite phase shift and the oppo-
Expressing capacitance in farads (F) and site signs in the formulas referring to
 ) -

13.12 Average Quantities. Power and Phase Shift 447

inductance and capacitance are of cru- 13.12 Average Quantities,

cial importance when considering L and Power and Phase Shift
C connected in one circuit. T .. A < i .

In the preceding section the current

In alternating-current experiments,
and voltage were regarded as functions
one frequently makes use of a single-
beam cathode-ray oscillograph. A voltage ime H° wever in many cases it '
’

proportional to the current is impressed

su fices know the average (constant}
on one pair of deflection plates (deflec-
lues of these quantities (cf. Sec- ™
tl0
tion along thex axis) and a voltage pro- f an elementary
portional to cp is applied to the other
As case, let us consid-
a Seating device with resistance R.
pair of deflection plates (deflection along
e know that in a direct-current (dc)
the y axis). The beam moves along a .

line whose equation has the form x =

circuit the power output (that is, the*

=
6cp, where the coefficients a
Quantity of energy released per unit of
aj and y
and b depend on the sensitivity of the
time ) 1S h 9/ Rj == cp IR. In an = =
oscillograph. Since 7 and cp are period-
alternating-current (ac) circuit the pow-
er ou tput varies, that is, the (instan-
ic functions of time, the beam sweeps
taneous) power output at time t is=
out the same curve on the screen all the
time. At 50 cycles per second, the hu- =
man eye cannot detect any motion of
h(t) cp (t) j
(
t Rj 2 (t) =
the ray and sees a solid luminous curve. cp
2
(t)/R. =
If the potential difference from the
resistance cp#, is impressed on the ver-
, Therefore (see (13.11.1)),
tical deflection plates of the oscillo-
graph, me fay aescrines' a siraigm uni. n (t)~= rt]£ cos (co£ 4- a). (13112.1)
21

True enough, for

Over one period, h ( t ) becomes zero-
x = aj = aj 0 cos (co£ + a), twice and reaches a maximum (equal
V
— = bRj 0 cos (at + a )-
2
to Rj 0 twice. When considering electric
)

heaters, however, we are usually in-

Eliminating t, we find that y = (
bRla ) x.
terested in the amount of heat generat-
If to these plates the poten- we apply ed during a time interval t that is
tial difference from the capacitance cp c , ,
many times greater than the period T
of alternating current (in the USSR
the result is an ellipse :

this period is usually 0.02 s). It is there-

= aj fore sufficient to know the average
x 0+ a)
cos (cot ,
value of the power output over a large
y = b-^j sin + a), 0 (coi time interval t. In view of (13.12.1),

whence h = Rjl cos 2

(cd£ + a) = Rj\ cos 2
(co£ + a)

= cos 2
(cof + cx)
(13.12.2).

-f- sin 2
(cd£ + a) = 1 .

since, as we repeatedly noted in this

Also, an ellipse results from L (induct-

cp
book, cos 2 Sec-
(cd£ + a) = 1/2 (cf.
tion 7.8 or 10.4). Equation (13.12.2) i&
ance). If we connect cp H cp +
L or
approximate being the more exact the
<Pn + <Pc> Ike axes of symmetry of the
greater t is.
,

ellipseno longer coincide with the x

and y axes. Thus, the shape of an oscil- The average value of an alternating
logram tells us what the circuit is made current , 7 ,
is ordinarily defined as the*
up of (C, L, or i?), what the “innards of intensity of a direct current that gener-
the box” consist of (see Figure 13.1.7). ates an equivalent power output on a re
 .

448 IS Electric Circuits and Oscillatory Phenomena in Them

rent and voltage vary along curves that

13 12
distance i?: are shifted with respect to one another,
although the frequency is the same. Let
Rp = Rjl/ 2, (13.12.3)
us consider the power output in the gen-
whence eral case of an arbitrary phase shift
of ; with respect to cp (that is, an arbi-
J — 0.71 /o (13.12.4) trary shift of one curve with respect to
another). Let
In the same way, the average value of
determined from the
/ = 7o cos (tot + P),
the voltage, cp, is
condition cp = cp
0 cos (tot +p+ a).

Then
h (
t)
= / 0 cp 0 cos (cot + P) cos (tot
whence
+ P + oc).

<p=- V\p 2 = —Lqpo ^ 0.71_(p 0. (13.12.5) But

cos (at +P+ a) = cos (tot + P)
Instruments that measure alternating
current (ammeters and voltmeters) are X cos a — sin (tot + P) sin a,
calibrated so that they give the average
so that
value of current or voltage, / or cp.

From formulas (13.12.4) and (13.12.5) cos + P) cos (tot + p + a)

(tot

it follows that the maximum values of = cos cos + P)

oc
2
(co£
current and voltage attained in an ac
circuit exceed the average values by a
— sin a cos + P) sin + P) (tot (tot

Y ~ == cos a cos (tot + P) 2

factor of 2 1.41. For example, in a
circuit with an average voltage of 220
volts the maximum instantaneous vol-
— sin a sin (2tot + 2P)
tage (peak voltage) reaches ±310 volts.
From the relations (13.12.2), (13.12.4), And since

(13.12.5) and the formula cp 0 = Rj 0 it

cos 2 (co£ ± P) = sin (2co t + 2P) = 0,
follows that cp Rj and h cp/,=so j = ,

that Ohm’s law and the relationship we have

between power, current, and voltage on
a resistance hold true for mean values. h =/ 0 cp 0 cos ax = /cp cos a.
When we considered alternating cur-
rent flowing through a capacitance and Thus, the average value of the power
an inductance, we saw that the cur- output in the general case where there
13 12
is a phase shift a is proportional to
*
Clearly, if j is given by (13.11.1), the
cos a. In the particular case of a re-
average value j defined by the formulas of =
Section 7.8 is zero. For this reason it is ad- sistance, a 0, cos a 1, and we re- =
visable to define the average value (al so cal led turn to formula (13.12.2).
the mean value) as we do here: j 2 1 /2
(in = In the case of a capacitance, a =
O’ )
mathematics this quantity is known as the — 3t/2 and cos a 0, while in the case =
root -me an- square of /). The square root is re- of an inductance, a +jt/ 2 and =
quired so that j should have the dimensions of cos a 0. =
Hence, in both cases the
current, while the operation of squaring makes
it possible to avoid the mutual compensation of
average power output is equal to zero.
positive and negative current values. Finally, This result is quite understandable from
the sign =
here means “equal by definition,” the standpoint of physics. In a capaci-
that is, we introduce a new definition of /, tance and an inductance, electric ener-
2
while j is defined in the usual manner. gy is not transformed into heat, it is
13.13 An Alternating-Current Oscillatory Circuit Series Resonance
, 449

merely stored up. In an ac circuit, a cp = cp x cp 4 — cp# + Wl “b <Pc- Re-

capacitance during one half of a period calling the formulas (13.11.2) to
takes electric energy from the circuit (13.11.4), we get
and stores it, only to release it back
into the circuit during the other half.
cp — jR/ 0 + a) — Laj sin {at
cos {at 0 -f- a)
The same goes for inductance in an ac
+ sin {at + a) = Rj cos {at + a) 0
circuit.
An
ordinary transformer without any
load is actually a pure inductance (if we + 7° (~ — La} sin(cot + a). (13.13.2)
ignore the slight losses in the wires). A
current flows through the transformer We see from this formula that the
=
How- potential difference on the inductance
with an amplitude j (pi La.
ever, as already stated, no power is and that on the capacitance have differ-
taken from the circuit because a jt/2 = ent signs, and so the coefficient of
and cos a =
0. It is an interesting fact
sin {at +
a) is the difference of two
that electric meters are designed to mea- terms. We write cp as
sure just the quantity /cp cos a. For cp = b cos {at + a + P), (13.13.3)
this reason, a nonloaded transformer
will hardly add anything to your ele- where b is the amplitude of the poten-
ctric bill, it will only increase the total tial difference, which is to say, the
current flowing in the wires. maximum value of potential difference
If a large number of inductances (the peak voltage). To find fe, we rewrite
(transformers, unloaded electric motors, (13.13.3) as follows:
and the like) are connected in parallel, cp = b cos P cos (cd£ + a)
the total current can become large,
and then the losses in the electric wir-
— b sin P sin {at + a). (13.13.3a)
ing will be rather noticeable. This Comparing expression
this with
effect is an important factor relative to we find that
(13.13.2),
the electric networks of a whole city.
The laws that govern the flow of cur- b cos P — i?/ 0,

rent through circuits with inductances

and capacitances show how such losses b sin p = 70
(Leo — . (13.13.4)
can be minimized.
Squaring both equations in (13.13.4),
adding, and taking the square root of
13.13. An Alternating-Current the sum yields
Oscillatory Circuit. Series Resonance
= j 0 yr R*+(L(o —^) 2
Let us now
consider an ac circuit com- b . (13.13.5)
prising resistance, inductance, and ca-
pacitance in series (Figure 13.13.1). It This formula shows that for a given
is obvious that in this system the cur- amplitude of the current, / 0 the peak ,

rent flowing through R, L, and C is voltage b is at a minimum when

the same. We write the current in the
form Lco = ^. (13.13.6)

7 = ;o cos {at + a). (13.13.1) Writing (13.13.5) as

The potential difference in the series is
+ ,

-L- C
p0~0-o —Wv— L
<^nnrnrv>-|(-o--j we see that for a given peak voltage, 6,
the peak current, / 0 is at a maximum if ,

condition (13.13.6) is met. This con-

Figure 13.13.1 dition may be written thus: co =
29-0946
 . ,

450 IB Electric Circuits and Oscillatory Phenomena in Them

1 iVLC, which is nothing other than rent for a given peak voltage as com-
the natural frequency of an LC circuit. pared with that at any nonresonance fre-
Therefore, (13.13.6) is the condition of quency 0 =£ co 0 .

resonance the condition of coincidence

,
It is of interest to investigate in de-
of the natural frequency of the circuit tail how the peak voltage and the peak
and^the xJrjvjacr frgrrjiPqcK^nf, tbr alie- current varv iq the .case,.at denar^ur-
nating current. Observe that at reso- from exact resonance, that is, when
nance the circuit voltage is 0 =^= 1 ty LC. To do this, let us take
cp = Rj o cos (cd£ + a). (13.13.7)
advantage^ of formula (13.13.3). We
find that 6 /]/ 2 whence b
cp = , = ]^ 2 cp.
Using (13.13.1), we find that at reso- Substituting this into (13.13.5), we get
nance
cp = Rj. (13.13.8) +
Let us now pass to average values.
We determine the average values of But / 0 /]/2 = /, and so
current and potential difference in ac- = 1 yi? +(L©-l/C©) 2 2
<p .

cord with formulas (13.12.4) and

(13.12.5). From cp L =
Lcoj 0 sin ( 0 i a) + Finding 7 from this formula and put-
we obtain (see footnote 13.12) ting it into (13.13.9) and (13.13.10),
we obtain
q> L = L(i>j — <p. (13.13.9)
Leo -
L —
Similarly, from cp c = (1/Cro) / 0 x Vh 2
+(L(d—i/C(d )
2
q> ’

sin (at -j- a) we get 1 -

(PC C(0 l^i? 2
+ (Leo — 1 /Cat) 2
(P ‘

= i-ds- < l3 - 13 - 10 >

If we denote by 0 O the natural frequen-
Formulas (13.13.7) to (13.13.10) are cy of the circuit, then 0 2 1/LC. =
valid only in the case of resonance, that These formulas can now be written
is,when 0 O = 1 /]f LC. Putting the val- thus:
ue of 0 into (13.13.9) and (13.13.10), co 2 (p
we get
]/ + (<»§-® 2 ) 2
<Pjl
= <Pc=-^- .y/"L-
1
c~<P-
ogq>
9c (13.13.11)
For this reason, when we have reso-
nance, the voltage on the inductance
and the capacitance is the greater the
/ i? 2 C0 2

L2
~ -CO 2 ) 2

smaller the resistance R and cp and

In this form it is quite evident how the
, L
ratio or cp c /cp depends on the close-
cp
L /cp
(p c may exceed the ac source voltage cp
ness of the natural frequency of the cir-
many times over.
In a series circuit the resistances are
cuit, 0 O to the driving frequency 0
, .

additive. But the ’’resistances” of the The ratio cp L /cp as a function of 0 near
capacitance and the inductance are of co = 0 O is shown in Figure 13.13.2. The
opposite sign and are different functions graph is constructed for the case of
of the frequency. At resonance frequen- i?/Lco 0 =0.05. It is illustrative of the
cy they are equal in absolute value typical resonance curve
and hence cancel each other. If RIL($ 1, the dependence < of
Thus, at resonance, (0 0 O ) an = cp L /cp oron 0 is mainly deter-
cp c /cp

RLC series circuit has a minimum “re- mined by the second term of the radicand
sistance” and carries a maximum cur- (
0* — 2 2
co ) . When 0 = 0O ,
this term
 ,

13.14 Induction and Capacitance in Parallel... Parallel B/uuujjwrji 4&L

N?

Figure 13.14.1

extremely small (in other words, we

neglect it). Then cp c and cp L coincide
and are equal to the voltage cp in the
circuit (that is, of the ac source), while
vanishes, and under the assumption the current / is made up of the cur-
that jR/Lco <
1 the denominator has a
rent j c flowing through C and the cur-
minimum and the ratio q) L /(p or cp c /(p rent j L flowing through L Let cp L . =
a maximum. The ratio constitutes 70% cp c = cp =
cp 0 cos (co* a). Using the +
of the maximum value when (co
2 — formulas of Section 13.11, we find that
2 2 = 2
jR co
2
/L 2
that is, at co
2
— 2 =
— — Ccocp sin (co£ + a
co ) , co

zfuRco/L, whence jc 0 )

Jl —
R
co 0 — = co
co 0
co

+ ca
-j-
,

2L
*
sin (at 4- a).

Therefore
The variation in frequency for which
the square of the ratio cp L /cp or cp c /cp / = 7c +/'n
fallsto one half of the square of the
maximum value is called the half- =<Po (t^ — C<0 ) sin (
co *+ a )>
width of the resonance curve. If the ra-
tio is 70% (0.7) of the maximum, its
whence, assuming that co
0 = 1/]/ LC ,

square comes to 0.7 0.5 of the ma- 2 ~

ximum value of the square. Therefore,
the half-width of the resonance curve,
co —
co 0 constitutes RI2L, which means
,
l/Z-G) — C(0 '
C ((Dy — (0 2 )
•

that the half-width (also called the

half -width at half -maximum) is equal In this case too we see a typical re-
to the quantity that characterizes the sonance relationship: for a given cur-
rate of decay of oscillations in that cir- rent /, the voltage cp is the greater the
cuit (see Section 13.9). closer co is to co 0 . It is easy to see that
Consequently, the smaller the resist- when co is close to co 0 j L and j c are ,

ance R, the smaller the half-width

j
much greater than the current j in the
of the resonance curve and the steeper circuit, that is, an ac circuit containing
the curve near co co 0 From formu- = .
L and C experiences strong oscillations.
las (13.13.11) we see that the smaller A small external current suffices to
the resistance R the greater the maxi- sustain much stronger currents in the
mum of the ratio cp L /cp or cp c /cp. That is circuit.
why the phenomenon of resonance is It will be recalled that in a parallel
particularly evident if R is small. circuit, conductances (which are the re-
ciprocal of resistances) are additive:
13.14 Inductance and Capacitance
in Parallel. Parallel Resonance

Consider the circuit in Figure 13.14.1, The “conductances” (that is, the ra-
which differs from that in Figure 13.13.1 tios of current to potential difference)
in that L and C are in parallel. We of a capacitance and an inductance have
take the resistance of the circuit to be opposite signs and depend differently
29 *
452 13 Electric Circuits and Oscillatory Phenomena in Them

R L tial equation, and the unknown quan-

tity appears in all terms in the first pow-
er. the superposition prin-
Therefore,
ciple is valid, that is, the sum of two
or several particular solutions of the
equation is also a solution.
Figure 13.14.2
In Section 13.13 we obtained a for-
mula for the half-width of a resonance
on the frequency. At resonance (co = —
cancel each other and the total
curve, co co 0 =
RI2L, which shows
<x>
0 ) they that the slower the decay of oscilla-
“conductance” is at a minimum, which
tions the smaller the half-width. This
is to say, the current is the smallest
does not occur only with respect to elec-
for a given potential difference and,
hence, the potential difference cp AB is
tric oscillations. We
can consider any
system capable of oscillation. Let an
at a maximum for a given current in
external force give rise to oscillations in
the external circuit.
such a system and then cease to operate.
In a simplified circuit without resist-
The system is now on its own. The
ance, the amplitude of oscillations
oscillations begin to decay. If the am-
grows without limit as co approaches co 0 . ~
plitude of oscillations decays like e '?*,
In reality, the resistance in the circuit
makes the amplitude finite when co
— then y characterizes the rate of decay
_1
(it has the dimensions s ). During time
<o 0 .
connected in parallel with L
t = 1/y the amplitude diminishes by a
If R is
factor of e, or by 63%.
and C, all calculations become very Let us now consider the resonant step-
similar to those of the preceding section.
up of such a system by a periodic exter-
But this case is rarely encountered nal force. The amplitude of oscillations
in practical situations. Ordinarily, the
at a given time is the sum of the ampli-
inductance has a perceptible “resistan-
tudes acquired during the time of step-
ce” and therefore the typical circuit
up. In the presence of damping, an
diagram is that shown in Figure 13.14.2. amplitude acquired a long time before
In this case the calculations are some-
will have time to decay and will not
what longer than in the preceding sec- play any part or make any contribution
tion and we will not carry them through
to the amplitude of oscillation at the
in detail. The result for co close to co 0
given time.
and for small ratios i?/Lco is
The decay time is clearly 1/y. In this
time the amplitude of free oscillations
j l ^ ic = ^
T'T '

/ (i?0)/L) +
2
(0)o
—G )
2 2
)
• will have diminished e times, or by
63%. Hence, even when the stepup
It thus turns out that current amplifica- force is constantly operating from t =
tion at resonance in a parallel ac circuit — oo, the oscillation amplitude will still
obeys the same law as voltage ampli- be determined solely by the time inter-
fication in a series circuit discussed in val from t —
1/y to t, where t is the
Section 13.13. time of observation. The action of the
have already
force at earlier times will
decayed.
13.15 General Properties of Resonance
For the difference between two perio-
in a Linear System
dic forces with somewhat distinct peri-
Note in our reasoning that when we ods, F o sin co 0 £ and F 0 sin co£, to mani-
added oscillations caused by a force at fest itself conspicuously, a time T of
different moments of time, we assumed observation is needed during which their
all along that the system without that phases will have separated by approx-
force was linear The behavior of the
. imately n units: coT co
0T
=
rc, so ±
system is described by a linear differen- that co
|
—
co 0 |
—
n/T. Consequently,
 13.16 * Displacement Current and the Electromagnetic Theory of Light 453

if the oscillations of a system “remem- varies when

the charge on the plates,
ber” only the action of the force during varies, is to say, when a current
which
time T =
1/y, then in such a system a flows in the right- and left-hand conduc-
difference n/T = ny in the frequencies tors. We can now do one of two things:
of the exciting force hardly affects the (1) either beg the reader’s pardon and
amplitude. From this we see that the explain that whatever we have spoken
half-width of the resonance curve is of current flowing through a capacitor
proportional to the decay factor y. (capacitance) this was not so actually —
On the other hand, since the system there was no current, the only current
“remembers” and accumulates the action flow being in the conductors on the
of a force during time 1/y, the oscil- right and left;
lation amplitude at resonance (hence (2) or regard the varying electric field
also the height of the resonance peak) is in the space between the plates on a
inversely proportional to y. Calculations par with ordinary current (the motion
confirm this reasoning. of charged particles). Maxwell, who
We would like to remind the reader suggested this view, was able to draw
once more that the system we are con- conclusions of tremendous significance.
sidering is, in the absence of an exter- It had long since been known that
nal force, linear. The behavior of such electric current (the motion of charged
a system described by a linear differ-
is particles) gives rise to a magnetic field .
ential equation, thus making the su- But if a varying electric field is similar
perposition principle formulated above to an electric current, an electric field
quite applicable. For nonlinear systems, varying in a vacuum should also set up
the superposition principle fails, and a magnetic field. This hypothesis of
the theory of such systems proves much Maxwell led to a remarkable symmetry
more complex. Also note that a linear between electric and magnetic fields.
system does not generate oscillations Faraday experimentally discovered in-
even if a constant emf source (an ideal duction, that is, the fact that any vari-
source) is present in the circuit. ation of a magnetic field gives rise to
an electric field. Maxwell, in strictly
13.16* Displacement Current and the theoretical fashion, hypothesized the
Electromagnetic Theory of Light existence of a similar phenomenon in
which any variation of an electric field
Up to now we have almost everywhere gives rise to a magnetic field. Only then
considered current flow through a capac- did the theory of electric and magnetic
itor without any reservations. Indeed, fields acquire its modern form.
if we connect a capacitor in an ac cir- The mathematical theory of Maxwell
cuit in series with an ammeter, the is written in the form of differential
ammeter will indicate a definite cur- equations that are too complicated for
rent j
= C cocp. On
the other hand, no thisbook and so we do not give them. 13 13
current flows through a capacitor be-
13 13
cause the plates of the capacitor are sep- •
The
characteristics of electric and
arated by an insulator (air or a mate- magnetic in space are functions of
fields

rial or even a vacuum), and so the in-

several variables —
they depend on the three
coordinates of a point in space and on time.
dividual current carriers (electrons) in Accordingly, the derivatives that are present
the left-hand conductor and plate will in the Maxwell equations are the partial
derivatives of functions of several variables,
never get over to the right-hand plate
and the equations themselves belong to the
and conductor. Consequently, no charged class of partial differential equations (in the
particles are in motion in the space Conclusion we touch on such equations).
between the plates, that is, there is no Moreover, these characteristics are vector
electric current in the sense that we have quantities. Finally, note that complex num-
bers and the theory of analytic functions of
spoken of current up to now. All there a complex variable (see Chapter 17) fit ele-
i^ in this space is an electric field that gantly into the theory of electromagnetism.
454 13 Electric Circuits and Oscillatory Phenomena in Them

The solutions to these equations describe an inductance, where cp depends on

the propagation of electric and magnetic djldt, neither is it like a capacitance,
fields in empty space. Both fields must where depends on However,
cp j dt.
be present at all times: a variation in j
the electric field gives rise to a magnet- the “box” differs from an ordinary resist-
ic field, and any change in the magnet- ance in that the function j (cp) differs
ic field generates an electric field. from Ohm’s law / =
cp !R. The “box”
At the time when Maxwell worked, has a more involved function / (cp).
Faraday’s experiments had already been This function is called the characteristic
completed, that is, the relationship curve of the “box”.
between a varying magnetic field and The only general assertion that can
an emf induced by it was known. Also be made with respect to j (cp) is that cp
known was the magnetic field of a cur- and j cannot have different signs if
rent. Finally, the relationship between batteries or some sources of energy are
the charge on a capacitor and the ele- not hidden in the box. If cp and j are
ctric field between the plates was like- of the same sign, energy is absorbed in-
wise known. These findings sufficed for side the box during current flow and
writing down the equations for fields the box takes up electric energy from
in empty space. the circuit to which it is connected. In
Maxwell found the rate of propaga- the box this electric energy is convert-
tion of the fields in vacuo. This velocity ed into heat and is dissipated. Since
proved to be equal to that of light! From cp is constant and negative when j 0 <
this it was natural to conclude that and positive when / >
0, it follows
light is nothing other than electromag- that cp =0 at / =
0. In all other re-

netic oscillations. Furthermore, the theo- spects the functional relationship be-
ry predicted the possibility of the exist- tween / and cp can be of any kind. For
ence of electromagnetic oscillations example, for current rectifiers we have
of anywavelength including X rays boxes whose characteristic curve is
(whose wavelength is thousands of shown in Figure 13.17.1: the current
times shorter than that of visible light) flows easily in one direction for a small
and radio waves with very large wave- potential difference and hardly at all
lengths. It was thus that the investiga- in the other direction. As can be seen
tions of Faraday and Maxwell began from the graph, the current is small
the work that culminated in the discov- even for a large negative potential differ-
ery of radio waves by Hertz, the in- ence. Such are the properties possessed
vention of radio as a means of communi- by diodes made of two semiconductors.
cation by the Russian A. S. Popov and In 1958 the Japanese devised a “box”
the Italian Marchese G. Marconi, the made up of specially chosen semiconduct-
discovery of X rays by Wilhelm K. ors, the tunnel diode (in reality, this
Rontgen, and the creation of a complete “box” is in the form of a minute cylin-

theory of the electromagnetic field der just a few millimeters in diameter

and the interaction of this field with and altitude), which has an unusual
curve of j (cp) with a minimum
matter. (see

13.17* Nonlinear Resistance and

the Tunnel Diode

Let us consider a two-terminal network

(a “box”) that is similar to a resistance
in the sense that current flowing through
the “box” depends solely on the instan-
taneous value of the potential differ-
ence. In this respect, the “box” is not like Figure 13.17.1
 — i 1 ——
13.17* Nonlinear Resistance and the Tunnel Diode 455

ure 13.17.2). The current through R is

equal to the current through the box,
and so R Rj cp = =
Rj (cp). This equa-
Figure 13.17.2 in which typical values tion is conveniently solved by graph.
of cp and j are indicated ). 1314 This curve We write it down as
does not contradict the principle ex- cp =E— Rj (cp) (13.17.1)
pressed above: the sign of cp is the same as
that of 7 everywhere, which means the and we construct in the cp/-plane the
“box” only absorbs energy. We will not straight line cp =E— Rj This line
.

go into the physical reasons for such a may be called the load curve of the
strange curve, but we will examine the battery-resistance system. The solution
consequences for a circuit involving a to the problem is given by the intersec-
tunnel diode. For the sake of brevity tion of the straight line (13.17.1)
we will continue to call it a box. (that is, the straight line j (E cp)/ — —
We start with the simplest type of R) with the j (cp) curve, which is the
circuit consisting of three parts: a bat- characteristic curve of the box. In Fig-
tery with emf E
a resistance
,
(the or- R ure 13.17.4 we have a graphic solution
dinary kind that obeys Ohm’s law), of the problem involving one battery
and the box (Figure 13.17.3). We in- and three different resistances: R±
clude the internal resistance of the bat- (small), 2 R
(medium), and R s (large).
tery in R. From the graph we can see that for a
The equation defining the current and sufficiently large E we can choose an
distribution of potential in the circuit R such that it will not be too small or
is of the form — +
E Rj (cp) 9 0, + = too large and there will be three points
where cp is the potential difference across of intersection, A, B and C, and thus ,

the box and / (cp) is the function defined three solutions to Eq. (13.17.1).
by the properties of the box (see Fig- For three solutions to exist, the j (cp)
curve must have a descending portion.
E R Box

Hi w — It is clear that the line on which dj/dy

is positive everywhere can only once
intersect the load curve, no matter
what E and R 0. >
Figure 13.17.3 Now us consider a somewhat more
let
complicated circuit diagram involving
13,14 By the end of the 1970s both physi- capacitance in parallel with the box
cists and designers had greatly advanced the (Figure 13.17.5). For this circuit we
miniaturization
devices.
resistor)
of all parts of
The modern diode (or capacitor or
now comes in the form of a tiny spot
a dwindling fraction of millimeter in diameter
and consists of multiple layers with the speci-
electronic

—
£
1
— —— — /?
*—
'
Box

ii
*

fied properties, hundredths or even thousandths £

of a millimeter thick. Figure 13.17.5
456 13 Electric Circuits and Oscillatory Phenomena in Them

find that the current through the box, £ R

/ (cp) ,
and the current flowing through
the capacitance, C (dyldt), together equal
Hf AN\i — —Q— <j>
Box
9—i

the current flowing through the re-

kjuuLHH
L C
sistance,

dxp E — cp Figure 13.17.7

i ( 9) +c dt R ’

We now consider a system (Fig-

whence ure 13.17.7) consisting of an inductance
and a capacitance connected in parallel
with the box. We again denote by cp the
potential difference across the box,

The
intersection points of the char- 7 =
7 (cp) the characteristic curve of the
box, and ]\ the current flowing through
acteristic curve 7 (cp) of the box with
the load curve ( E —
cp )/R correspond to
L and C :

the solutions cp =
constant, dyldt 0 = .

Let us examine the sign of dyldt near

these points. A
glance at Figure 13.17.4
shows us that dyldt is positive if cp < <P — 9c = <Pi,
= £-^-. (13.17.2)
cp A or 9 b 9 < <
9 c an d negative if
9a 9< < 9b or 9 9c- > We
consider the process in the circuit
The arrows in Figure 13.17.6 indicate when the current and the potential in
the direction of variation of (p with the box are close to the intermediate
time. We
can see that the intermediate point B of intersection. Let
solution at point B is unstable all we
9 = 9b + A — 9b + g,
:

need to do is depart slightly to the left 9c

or to the right, and dy/dt acquires a
sign such that the deviation of <p from
i —1 (<Pb) + / -4~ m B =/(q> B +
“9 lv=(p
I )
fc/.

cp B increases. On the other hand, the

points A and C in Figure 13.17.4 charac- where k stands for the derivative dj/dy
terize two stable solutions, which cor- at cp = cp jg. We
used the first two terms
respond to stable states of the system. of the Taylor series to represent cur-
The existence of two stable states rent; the intermediate potential of the
permits using these boxes in mathemat- capacitance is cp B and cp# and 7 (cp B ) ,

ical machines as memory cells. By satisfy the condition 7 (cp B ) = (E —

making a lot of such circuits and trans- cpb )/B. Substituting these expressions
ferring (by an external means) some in- into Eq. (13.17.2) and simplifying,
to the A state and others into the C we get
state, we can record (“remember”) any de-
sired number or other information. kf + h = —i-, cJW==h,
Using such systems, we record informa-
tion in coded form as AACACACCC
..., f —g (13.17.3)
where each letter A or C indicates a
state of the appropriate system (the first
From these equations it follows that
(13.17.4)
in A the second in A the third in C, = 1
, ,

/ ~ r]u (13.17.4)
the fourth in A, etc.). l/R + k ]i
-1 =
where r i?
_1
+ k = +
(dj/dq>),p=«p .
B Equations (13.17.3) and
_JfA also yield

r
= df__ j£_ — _ Hi L Ji i
Figure 13.17.6 dt 2 dt dt dt C ’
 '

13.17* Nonlinear Resistance and the Tunnel Diode 457

and finally > R _1

since R~ x is precisely the
,

slope of the load curve / (E =

cp)//? —
intersecting the characteristic curve at.
-S
L+
xt- + -5T>'. = 0 ’ <
1317 5
-
>
the point B. Consequently, the total,
resistance of the oscillatory circuit,
which is the ordinary equation describ- r =(R' 1 fc)
_1
+ is negative.
,

ing an oscillatory circuit with capaci- The equation for an oscillatory cir-
tance C, inductance L, and resistance cuit involving L, C, and r, with r posi-
r. The resistance r reflects the fact that tive, yielded damped oscillations. For
the capacitance and inductance are in r <0 this equation will yield stepup>
parallel to two circuits, the circuit in- oscillations, which grow with time.
volving the battery and resistance R Thus, a tunnel diode is capable of gen-
and the circuit with box and resistance erating oscillations in a circuit.
k~ x = dy/dj. Since the two circuits are The capability of generating oscilla-
connected in parallel, the conductances tions is a consequence of the instability
(reciprocals of resistances) are additive, of the solution at point B. The oscilla-
whence follows the expression for r. tion energy is taken from the battery.
The currents and potentials in the The oscillation amplitude increases with
system break up into two terms: the time by an exponential law only so long
constant term (<p B / (cp B )) and the os-
,
as it may be considered small and we
cillatory term (j 1 ( t ), / ( t ), g ( t )). Here, can employ the Taylor series expansion
for the oscillatory term the role of re- of the characteristic curve / (cp) about,
sistance of the box is played by the de- the point B. Roughly speaking, the
rivative d<p/dj taken along the charac- maximum amplitude is limited by the
teristic curve. If the box consisted of points A and C in Figure 13.17.4.
an ordinary (ohmic) resistance, <p i?/, = Already by 1961 oscillators with,
the derivative would be equal to the efficiencies up to 25% and power out-
resistance, dy/dj = R. puts of 0.5 MW
at 7500 MHz (at a 4-cm
What does the unusual characteristic wavelength) had been tested. For us,
curve / (cp) of the box (tunnel diode) of tunnel-diode circuits are interesting
Figure 13.17.4 lead to? At point R the from the standpoint of a mathematicaL
derivative djldg) is negative, which means consideration of a nonlinear problem,
that with respect to oscillations the questions of power output, and the re-
box has a negative resistance! What is presentation of currents in a system as a .

more, from Figure 13.17.4 it is evident superposition of a constant solution and*

that at point R we have k | |
=
djldcp |>
|
oscillations.
 3 —

Part

Some Additional Topics

Chapter 14 Complex Numbers

14.1 Basic Properties of Complex dd is equal to (ad -f- bc)lbd and the pro- ,

/Numbers duct of ( a) by (—b) is -\-ab. u l

That numbers can be multiplied with-
out limit leads to the operation of rais-
The use of complex numbers can shed ing a number to a (integral positive) pow-
additional light on some topics discus- er: a2 a-a, a 3 = a-a-a, and gener- =
sed in this book. Originally, the number ally an =
a -a -a ... a. But the in-
(natural) characterized the number of
objects in any (finite) their set, for n times
verse operation, that is, taking roots of
instance the number of children in a
family, of boats on a river, or of the
ordinary numbers, can not always be
performed: there is no (positive or nega-
fingers of the hand. Numbers can be ad-
tive) number x whose square would be
ded: if we unite two groups of people
consisting of a and b persons, we will
equal to, say, —
7, that is, there is no

get a + b people altogether. Numbers number x =Y

—7. In what follows we
can also be multiplied: a bunches of b will call ordinary numbers, that is, pos-
flowers each will yield ab flowers. But itive, negative, and zero, real num-
sometimes natural numbers cannot be bers 14 2 Attempts at taking (square)
.

subtracted or divided. Manipulating root of any real number lead also to the
with just natural numbers we cannot notion of complex numbers.
subtract 5 from 8 or divide 3 by 2. 14 - 1
These conventions only make it possi-
The quest for carrying out operations ble to apply all well-known operations per-
«of subtraction and division led to formed on “old” numbers to “new” numbers
the appearance of negative and fraction- (when extending the notion of number we
have just to refresh our knowledge and not to
al (rational) numbers it was accepted
:
start from the very beginning; so, according
to designate the difference 5 —
8 as — to our rules a -f- b all b/i =
(a X 1 + = +
and the dividend 3 2 as 3/2. New 1 X b)l 1; X 1 = a
(a +
b). All
b)f 1 = +
numbers had quite a different meaning the general rules of operation are also valid

since neither group of people can contain

(for instance, such as a b b a or + = +
(a +
b) c ac =
be). +
The same reasoning
— 3 or 3/2 persons. (Not without rea- also holds in considering the rules of operation
son, Samuel Marshak, an outstanding on complex numbers we are dealing with in
Soviet poet and translator, wrote about this chapter.
14 2
To make it possible to take square root
•

a schoolboy who on solving the problem of any positive number, we have to include
^obtained 2 2 / 3 diggers.) Operations per- in the real numbers not only all (positive and
formed on such numbers are also defined nonpositive) fractions (rational numbers),
in a new way; indeed it is senseless, but also irrational numbers, such as 2 V =
say, to unite —3 boats and 3/2 boats in 1.41 .. (In this connection see,

I. Niven,
. . for instance,
Numbers Rational and Irration-
fleet. We have to assume, for instance,
:

al, ,
New York, 1961, which is intended for
that the sum of the fractions alb and a wide circle of readers.)
 1 a a

14.1 Basic Properties of Complex Numbers 459 '

We
introduce a new “number” ]/" — Whenfinding the quotient k a/6, =
defined by the condition that its square we can use the fact that at any (com-
equals the number 1 Clearly, none of — .
plex) number 6 6 X 4- *6 2 the pro- =
real numbers either positive or nega- — duct of 6 by the number 6* bx i6 2 = —
tive —
possesses such a square, and there- (the number 6* is often denoted 6" and
fore we call it the imaginary unit in ; called the conjugate to b) is always real:
mathematics it is usually denoted i.
Expressions of the form c a ib or = + bb* = (b + ib t 2) (b x - ib 2 )

z x = +
iy, where a and 6 (or x and y)
are real, are named complex numbers.
= k- (ib
= K + b\.
2
)* = b\- m
= 14 1 1)
Numbers of the form ib ( 0 ib) or + ( . .

iy are sometimes called pure imaginary

(or simply imaginary). (If 6* = 6, the number 6 is real; the
Four arithmetical operations on com- square root of the product 1^66* =
plex numbers a = a x + ia 2 and b =
b 1 + ib 2 directly follow from the de-
Y b\ 4- b\ denoted 6 and called is | |

the absolute value or modulus of the , ,

finition of the imaginary unit. They are complex number 6.) Therefore
introduced as the natural generalization
of ordinary operations on real numbers: 7
a_
__ ai + *q 2 __ (gi + ta 2 ) (frj — *fr 2 )

if a 4- b /, a b =
g, and ab h, — = = 6 &i 4"^2 (^1 4 *
^2) (^1 — ^2)
then __
~
(^l^l 4 g 2 ^ 2 ) 4~ i (#2^1 — ^1^2)
/ = /i + if = a + id + (61 + ib
2 ( i 2) 2)
-
gl&l 4 #2^1
— (a + &i) + {a + b
1 i 2 2 ); “
fl 2^2 |
• gl&2
'

m* m*
S —g + 1 = (% + ^2) — (&l + ^2)
Our rules permit representing any
= («1 — 61) + (a ~ 62); i 2
rational function of the complex
h = h + ih 2 = {a + ia (b + ib
x x 2) x 2) variable z x =
iy (for instance, / + =
= a b + ia b +
1 1 + ia i6 x 2 2 2
2
z/( 1 4- z )) as the sum / u 4~ iv, =
where u and v are two real-valued func-
= (a 6 — a & + (a 6 + a b
1 i 2 2) i
x 2 2 x)
tions of two real variables x and y
{in calculating h we take into consider- (see Exercise 14 . 1 1 ).
.

ation that i
2 — 1); therefore The problem of taking the square root
of a complex number can be solved in
fi — a + bv =a +6
f2 2 2;
— 4-
i
a similar way. Indeed, if r rx
£1 —a —b i 1, £2 = ^2 — ^2; ir 2 =y = y 4“ id 2, then (r t 4~
x
= a^ — a 26 h = a b + a
2 , 2 x 2 2bv ir 2 )
2
= — (1
r\ 4- i 2 r x r 2
r\) a ax = = +
ia 2 , which results in the following sys-
Similarly, if k = alb = k + ik then determining the
x 2, tem of equations for
a = 6&, that is, unknowns r x and r2 :

+ 162) (&i + 2) = (6^ — 6 2 /c 2 ) r\ — r\ = av 2rxr 2 = a2 .

4" i {b±k 2 4~ b 2 k x ) = a 1 4* ^2* This system is readily solvable. Since

r\ + (— rl) =% and r\ (— r\ )
=
For determining and /c 2 we have now
— (a 2
4 ), r\ and
/ r\ are roots of the —
two linear equations, quadratic equation X2 ax X — —
*1*1 —
6 2 & 2 = a x and b 2 k x 4~ b x k 2 = a 2 , (a^/ 4 ) =
0 whence ,

which can be solved in all cases where —

6^0 (that is, when b x and 6 2 are ri = |// 1 \a\+ai
2 ’
„
r2 “ 1V
/* |a|

2
’

not zero simultaneously):

#l4 l 4~# 2&2 #2&1 &1&2 where, as usual, a "J/^a
2
4~ =
7. 7
2_ (Note that both integrands are positive.)
| |

1
^ 4-^2 ’
b*+bl •
 x —
.

460 14 Complex Numbers

where c is real, that is,

1
c* =c , and b - (b 1 — ib 2 ).

Indeed, if z =x+ iy, then the left-hand side of

(14.1.2), as is readily seen, equals b x x -\-b 2 y + c.

The equation (x 2 + y
2
) + b±x + b2y + = c 0'

of the perimeter of the plane of complex

variables can be rewritten in the form similar
to (14.1.2):

zz* + bz + 6*z* —f- c = 0, (14.1.2a)-

with c* =
c and b = (1/2) (6 X ib 2 ), since —
zz* = x 2 2
y and bz b*z* = b x x
+ b 2 y. + +
Equations (14.1.2) and (14.1.2a) can be-
combined as follows
Figure 14.1.1
azz* + bz + 6*z* +c= 0, (14.1.2b)

where a and are real (a* a, c* c = = c). Equa-

Finally we have tion (14.1.2b) describes a circle if a =£ 0,
and a straight line if a 0.
—
Y a = Y at + ia 2 The position of point zona complex
a| + ai +i |/ |g| plane can be also characterized by its
1
-- 1
)
g -
polar coordinates the distance p of ,

point z to the origin O (which is the

Thus, introducing the new, imagin- absolute value z of number z) and | |

ary, number solvable

£ in order to make the angle cp between the ray Oz and the
the problem of taking the square root of positive ray of the real axis (angle cp
— 1, we unexpectedly were able to is argument (amplitude or
called the
take square roots of all numbers phase) number 2 sometimes it is
of ;

“old”, real, and “new”, complex. denoted Arg z). Here (see Figure 14.1.1)
Before going further we will consider # = pcoscp, y = p sincp,
the geometric representation of complex 2 2 —
numbers. The complex number z = iy
P -=V x -\-y |z|,
x
is usually considered as a point of the
.

tan =T y
’
c °s cp = ,
sm cp =
.
y
.

so-called complex plane (or the plane of

the complex variables) with (Cartesian)
Thus
coordinates x and y (Figure 14.1.1). z = p (cos cp + £ sin cp) (14.1.3)
Here the .r-axis is called the real axis (which is the trigonometric form of com-
(its points correspond to the real num-
plex number).
bers x + £0 = x) and the y-axis, the
The form (14.1.3) is especially con-
imaginary axis. To each complex num- venient for multiplication and division*
ber there corresponds a single point
of complex numbers
of the plane and vice versa; to conjugate
complex numbers there correspond zi — Pi ( cos Ti + £ si n Ti) an(3
points which are symmetrical about the z2 = P 2 (cos cp 2 + £ sin cp 2 ).

real axis. The absolute value ]/ zz* = Indeed, it is clear that

|
z=Y |
2
+ y
2
of the complex num- z f z 2 == p A (cos + sin cp 4
p 2 (cos cp 2
ber z = x
cpj i
)
-\- equal to the distance
iy is
Oz of point z to the origin (corresponding -r £ sin cp )
2 = pjp 2 [(cos <p 4 + £ sin cp
4)

to the number 0). X (cos cp

2 + i sin cp 2 )]

= p p (cos cos — sin sin cp d cp cp 4 cp 2 )

The
straight line b±x + + —0
b2y c of
4

+ (cos sin + sin cos

2 [ 2

complex
the plane of the z variable =x+ iy £ cp 4 cp
2 cp
A
cp 2 ) ]

can be given by the “complex equation” = P P cos + + sin + T

1 2 [ (cp 4 cp
2) £ (q> 4 2) b
bz + b*z* + = c 0, (14.1.2) (14.1.4)
 , r ,

14.1 Basic Properties of Complex Numbers 461

or in words: when multiplying complex

numbers values are mul-
their absolute
tiplied and
arguments are added.
the
This leads to the following. If zjz 2 =
z 3 — P 3 ( cos 93 & sin
93 )> then z 1 = +
Z 2 z 3l that is, pi = p 2 p 3 9i 92 9s. ,
= +
whence p 3 = p x /p 2 93 = 9i 92 ,
— -

Thus,

iL
Z2
= -£l C ° s
[ (cp, _ cp 2 ) + i Sin (cp, — cp 2 ) ]
P2 Figure 14.1.2
(14.1.4a)

when dividing complex numbers their (see 14.1.2). Note that for-
Exercise
absolute values are divided and the argu-
mula (14.1.6b) yields three values of
ments are subtracted. the cube root of number z: since in-
Note that the rules stead of a the argument z can be writ-

IZ 1 Z 2 = Z ll Z2 »
(14J - 5)
ten as a 2k or a +
4 k, we have +
I l l I

1ST
completely coincide with those for real
zt = Y (cos + sin (3 i (3),

numbers; while the rules for the argu- z 2 = V r cos (P + -y-)

ments of complex numbers: [

Arg(z,z 2 ) = Argz,+ Argz 2 ,

+ isin (p + -y-)] ,

(14.1.5a)
Arg = Arg z, — Arg z2 ,
z3 = 3
/r [cos (p +
j

are new to us. In form, Eqs. + isin (p + -y-)] ,

{14.1.5a) resemble the rules of operation

on logarithms, but it is not yet quite
= with (3 = a/3 (Figure 14.1.2).
clear what relates the angle Arg z c
p
to logarithms (do complex numbers
Thus, the use of complex numbers
have really logarithms? We will answer makes it possible to find three roots
this question later on).
z lt z2, and z3 of the cubic equation
Multiplying, according to Eq.
z
3
c —
0, where c = r (cos a — +
sin a) is an arbitrary complex num-
{14.1.4), the number z p (cos cp = + i

ber. Moreover, we can show that each

i sin cp) by itself n times we obtain the
algebraic equation of nth degree
Moivre formula
zn — n cos n(p _[_ 1 s[n (14.1.6)
P = a z n a^
(z) 0 -f-
71-1
-f- a 2 zn
~2
+ . . .

p (

This formula is also valid for negative

. .
. + a n _iZ + a n = 0
and fractional n. For example, noting where a 0 =7^ 0, a l5 a 2 ., a n are ar- ,
. .

that 1 = 1 (cos 0 + i sin 0) we get bitrary complex numbers, has n (and

only n ), generally speaking, complex
Z" 1 = y = y [cos( — cp) + isin( — cp)] roots z ±1 z 2 ., z n which are not nec-
,
. .

essarily distinct (some time ago this

= — (cos — cp i sin cp) (14.1.6a) important proposition was called the
fundamental theorem of algebra). Here
and the polynomial P (z) is the product
of n linear factors:
z i/z =Y r (cos a -f i sin a)
P (z) = a0 (z — z,) (z —z 2 ). . .

= 3
/r (cos -y + i sin
y )
(14.1.6b)
(z — zn
. . . ).
 * aa

462 14 Complex Numbers

For example, the quadratic equation then

Q {z) = z 2
+ pz + q = 0 (14.1.7) «0 (
Z o) 4 + «l* (z*) 3 + «2 (z*) 2 + «3 i
Z o)-T at = 0*.
But a$ a 0 ...,= , af =a 4 and na-
has two roots turally 0* 0, therefore =
Q z *) = «o 2 o) + «i z o) + a z *)
4 s 2

=-f±|/ 4—9- 2
( ( ( (
Zl 2 (14.1.7a)
,

+ «3 ( o) +
r =
with that is, z% = Xq — iy is also a root of the same 0

<? (z) = (z — Zl) (Z —Z 2 ).

equation (distinct from z 0 for y Q =£ 0).
From the foregoing it follows that each
real polynomial (i.e., such that all its coeffi-
The roots (14.1.7a) of Eq. (14.1.7) will cients are real) can be expanded into (real) li-
be real for real p, q and distinct for near and quadratic factors (e.g., see the expan-
2
p /4 q — >
0 (recall that the real num- sion (14.1.8)). Indeed, the polynomial P (z)
a 0 zn
=
ber # # =
iO is a particular case + %zn-1 + a n _iZ +
a n has n (real . . . + +
or complex) roots z1? z 2 ., z n which nul- . .

of a complex number); the roots will P (%) == P

,

= =
,

P —
be the same (equal to —p/2 for p 2 /4 — lifies it:
Here P (z)
(z 2 )
can be expanded into a product
. . . (z n ) 0.

q =
0; and complex-conjugate (equal
P = a0 — zj)(z — —
±iY q — p
(z) (z z 2 ). . • (z z n ).
to —p/2 2
/4, the radicand
being now positive) for p 2 / 4 — q 0. < But if z 0 0 0 = x + iy where
y x =£ 0 is a com- ,

(When we say that for real p, g and

plex root of the polynomial P (z) and z% =
x0 —
iVo is a root complex-conjugate to z 0
2
p /4 <
q the quadratic equation (14.1.7) then
,

/ms /2o solution, we are not mistaken

and just want to say that manipulating
(* —z — z%)0) (z

only with real numbers we have to

= (z— x — iy — x + iy 0 0 ) (z 0 0)

consider in this case that Eq. (14.1.7)

= l(z— x — iy — x + iy
0) 0 ] [{z 0) 0]

has no roots.) Equation = (*— x — (iy = — x + 0)

2
0)
2
(z 0)
2
yl
= z — 2x z + (xl + yl).
2

Z 5 __ 32 4 + 7z s
_ + 12z __ 4
0

= (z - l)
3
(z
2
+ 4) = 0 (14.1.8)
Thus, to each pair of complex-conjugate roots
Zq z% of the polynomial P (z) there corresponds
,

the quadratic factor z 2 2x 0 z (x\ yl) in

— + + =
has five roots, of which only three are its expansion, and to a real root z j xx
distinct: + iO =
x x (if only such a root exists) there
=z = corresponds a linear factor z zx z xx — = —
Zi 2 z3 = 1, z4 = 2i, and of the expansion of P (z). This reasoning
zh = —2 i.
proves the required assertion. For example, the
inexperienced reader may consider the ex-
pansion
all the coefficients of the equation (of
If
any
an
order) a 0 zn
—
a^71 -1 0?i-i z
0 are real, then to each “essentially com-
+ + -
+• . + x* -f- a 4 = x* + 2a 2x 2
+ a — 2a x 4 2 2

plex” (i.e., not real) root z x iy of the — + = (x + a2 —(V2ax)

2
)
2 z

equation there corresponds a second root

z* x = iy, —
which is complex-conjugate to = (x + )+/ 2 ax] (x2 + a2 — 1^2 ax]
[
2 2
[ )

the first root. This follows from the rules of

operation on complex numbers, according to
= {x y^2 ax-\-a (x2 — 1^2 ax-\-a
2 -\- 2
)
2
)

which (how could we

to be accidental and artificial
(Zl + Z 2)* = Z * + h Z (
Z1 — Z 2 )* = Z* — Z*, know that 2 a 2x 2 should have been added to
(*1*2)*= *1*2
and 2a 2 x 2 subtracted from the expression x 4 +
a 4 ?). But the knowlegeable reader will under-
»

stand that this expansion follows with neces-

sity from the general fact formulated above of
expandability of real polynomials and from
(check By
virtue of (14.1.9), if, for in-
it). the formula
stance, z 0 =
x 0 -\- iyo is the root of the equation
—
Q (x) a0 x 4
= a4 x 3 a2x2 +
a zx a4 0 + + + = x = yr — = 4 <2
y~ 1

with real coefficients, that is,

for the roots of the polynomial x 4 + a 4 (see
Q ( 0)
=a o
z
o +a + izo + H = 0, Exercise 14.1.4).
14.2 Raising a ftumber to an Imaginary Power and the Number e 46&

The fact that in the region of com- 14.1.3. Prove that n values of the nth
root of an arbitrary complex number c (n
plex numbers the square root can be roots of an nth order equation zn c 0) — =
taken of any (positive or negative, real are represented geometrically by n points of
or complex) number and that each the complex plane, which are the vertices of
quadratic equation has here two roots a regular polygon of n sides. ____

undoubtedly rejoices us since it proves 14.1.4. Write all the values of y^— 1.
Using the above formulas factorize the poly-
that the set of complex numbers is, in
a certain sense, “constructed more sim-
nomial z 4 +
a 4 (use real factors).

ply” than the set of real numbers.

14.2 Raising a Number to an
However, at least we could have expect- Imaginary Power and the Number e
ed this result; indeed, the imaginary
unit i was especially introduced to We pose the problem of determining an
make it possible for us to take the square imaginary power. How can we find the
root of —
1 (or to solve the quadratic
number 10 i ? Can we multiply 10 into
equation x2 1 + =
0). Of course, here itself i times, that is, ]/ — 1 times? At
we have obtained more than we had first glance this question seems sense-
reason to expect (that is, the existence less. However, the question of the
of not only one quadratic root ]/" 1, — result of multiplication of 10 into itself
previously nonexistent, but of all pos- — 3 times or 1/2 time seems at first
sible quadratic roots of all complex also senseless, since literal application
numbers; the solvability of not only the of the rule of raising a number to an
especially chosen equation x
2
1 0, + = integral positive power (an aa a) = . . .

but of all quadratic equations), but

n times
this just indicated how successful the
is impossible here; yet the num-
introduction of ]/ 1 i was. How- — = bers 10- 3 (= 0.001) and 10 1 / a (=V 10 ~
ever, the fact that application of com- undoubtedly exist.
3.16)
plex numbers z x =
iy permits taking+ Let us recall the logical sequence of
any roots of all (both “old” and “new”) determination of negative and fraction-
numbers and solving algebraic equa- al powers of numbers, the example being
tions of the third, fourth, fifth, and inspiring. First, we only define raising
any other orders so that to make all a number to an integral positive power
equations solvable we have no need n:
to introduce any new numbers and
can manage with the complex numbers an = aaa ... a.

we used to solve quadratic equations, n times

this fact is certainly a surprise. All
this suggests that “there is something”
From this definition — for integral pos-
itive powers as before— two rules fol-
in complex numbers, that they mean
low:
more than just the taking of roots of neg-
ative numbers, that complex numbers n+m times n times m times

are not at all accidental and can be +m = aa =a

a" ... a ... aa ... a
useful in many areas of mathematics
and mathematized natural sciences. = a amn
, (14.2.1)
These hopes have materialized. n times n times n times

Exercises
m times
14.1.1. Let z x =
iy and / (z) + u = + (14.2.1a)
iv. Find u and v as functions of x and y if the
function / (z) has the form: (a) / z3 3z +1, = — (cf. definitions of addition and multi-
(b) / = z/( 1 + z 2 ), (c) / = plication of natural numbers on p. 458).
14.1.2. Prove formulas (14.1.6a) and
Generalization consists in that we re-
(14.1.6b) and the validity of de Moivre’s for-
mula (14.1.6) for any (real) n. quire the rules (14.2.1) and (14.2.1a)
 ,

''464
14 Complex Numbers

be also fulfilled for negative and frac-

II
tional powers. N N

Then the rule (14.2.1) yields

a*a
~3
— a 4+ (" = a = a.
3) 1
!
k/n
and therefore 0 9 / X
ta
-3 —JL = JL. Figure 14.2.1
a4 a? ’

and similarly for any natural m in absolute value, in particular also

for pure imaginary numbers ri whose
absolute value ri r is small. In | \
=
order to determine the number e hi ,

In a similar way, by virtue of (14.2.1a), with k being any (not small) real num-
JL. 3 ber, we use the fact that, by virtue of
• (a 1 / 3 ) 3 =a 3
=a = a, i
(14.2.3),

and therefore e
hi/n^,
1 + Aj for (14.2.3a)
-# 1 / 3 — / a and generally a^ q = / a.
q

where n follows from the notation

as
Now for any fraction n — plq we have n >> 1,
,

understood to be a very large

is

, uv/q = ( a i/*>p^ («/ a)* = q/a?. (14.2.2)

number (we will consider it to be natur-
al) so that kin <C 1.
Since for any real (irrational) number The point 2 = 1+ kiln (= p (cos cp +
we can
find an arbitrarily close rational i sin complex plane has Car-
cp)) of the
fraction, the problem of raising a num- tesian coordinates (1, kin) and polar
ber to any real power is completely coordinates (p, cp) (the absolute value
•solved. However, the techniques we and argument of the complex number
used to find negative and fractional z) where
,

powers are themselves insufficient to

•determine the imaginary power of a ik
(a and k being real numbers); for this
purpose we have to use the data we
(14.2.4)
have gained in studying the derivative
= tancp = —
k/n . k
of the exponential function. sin cp ,

Let us use the fact that

*er ~ 1 + r for |
r |
<C 1 (14.2.3) (Figure 14.2.1). For n >> 1, that is when

by the definition of the number e (cf.

kin <
1, the (exact) equalities (14.2.4)
can be replaced by the (approximate)
.Section 4.8, formula (4.8.2)), the ap-
relations
proximate equality (14.2.3) being the
.more exact the smaller r 14 3 Let us -
.
p~l, cp~+,
| |
(14.2.5)
agree now that Eq. (14.2.3) is also
valid for complex r’s which are small
the approximate nature of (14.2.5) in-
14 • 3
The
difference a er (1 r) is = — + dicates that terms of the (kln)HYi and
«of the second order of smallness with respect higher orders of smallness are neglected
tor: for small |r| the ratio a/r 2 remains “finite”, (see the text below printed in small
that is, not too small and not too large.
type). Since, by virtue of de Moivre’s
(Instead of referring to Eq. (4.8.2), we could
have, which is the same, proceeded from the formula (14.1.6),
definition (4.8.3a) of an exponential function,
= +~
n
lim (1 k/ri)
n-> oo
+
assuming that the ,
e
hl
~ 1 IP ( cos <P + i sincp)]
n
^
-atter formula is also valid for complex (pure
.maginary, in particular) powers k.) = p n (cos ncp + i sin racp)
 l

465
14.2 Raising a Number to an Imaginary Power and the Number e

Further we can naturally assume that

we have =
n
if z x + iy, then
(l + Ai) ~l"[cos (»-£-) z
e z= e
x+iy = exe = e x (cos y +
ly
i sin ,y)
(14.2.10)
+ i sin n )
= cos k + i sin k
( ]
(14.2.3b) (cf. Exercise (14.2.1)).
Formula (14.2.10) can, of course, be
and consequently immediately taken as the definition
e
hi
= cos k + i sin k. (14.2.6) of the exponential function e z of the
complex argument z = x + iy (com-
This relation can be assumed to be the pare with the above reasoning for the
definition of the exponential
function
follows from Euler formula (14.2.6)). This definition
an imaginary argument; it
of
(14.2.3a). Formula is justified by that if z = x + iO = x
the supposition z
then e becomes the exponential func-
(14.2.6) is known as the Euler formula.
tion e* of the real variable x, and also
The Euler formula (14.2.6) helps find
by the fact that the function (14.2.10)
the number 10* retains the basic property (14.2.1) of
_ ^ln 10)i = £lnl0.i^e 2 -3i
exponential function:
= cos 2.3 + i sin 2.3 ~ — 0.67 + 0. 77z. if z = z + z2 1 ,
where z1 — x1 -\-
tyi and
z2 = x + iy 2 2, then
Another method for deriving the Euler
formula is connected with the expan-
sion of functions into series. As we e= exp
z
+ x2 + (y + y )] [(as, ) i t 2

know (see Chapter 6) = X4 + X2 [cos (y + y + sin (y + y2)]

e< >
t 2) i t
2 a3
e
x ~ +
1
a: . ,

4!
= (e xl eX2 [(cos y + sin y,) (cos y2
) t
i

(14.2.7)
+ sin y2 = [ex (cos y + sin y,)
i
)]
i
t
i ]

X [eX2 (cos p2 + sin p = e e i 2 )]

zl Z2 .

this being convergent for all

series
x's, which gives us courage to consider
By virtue of formulas (14.2.4) and (6.4.6),
the number x in the series to be complex
as well. Suppose, by definition, *2 1/2 k*
(ix) 3 P
/,
1 + 7^")
. \
~ X+ ~2~^ + ---
1
,

lx V
*** = ! + -§- + (

2!
^,

”~3!
k2
= 1- 2n 2
,

,
Qz) 4
"+ '

4! 5!

z2 where the dots indicate terms of order

2! ^ |

4!
higher than the term k 2 /2n 2 in accord-
ance with the (more general) formula (6.4.3),
;

x b
X3 '

we have
'3T
+ ”5f (14.2.8)
k2
But since (see Chapter 6)
k
l + 2n2

k2 k2
cosz = l— (14.2.9) =1+ 7?. . =1
2n2 2n 2

sin x =x —— h-ff
• • -
(14.2.9a) whence, owing to the smallness of the second
term k 2 /2n for n 1, it follows that the ab- <
solute value of the number (1 ki/n) n (~e hi ) +
(the series (14.2.9) and (14.2.9a) are can be assumed to (approximately) equal 1.
also convergent at all x’s), we again Similarly, we can also verify that it is not a
obtain the Euler formula big mistake if we assume that the argument
of the number (1 kiln) equals kin (see Exer- +
e *
2
= cosx + isinx. cise (14.2.2)).

30-0946
 . — .

466 14 Complex Numbers

Exercises whose modulus is 1 and changes pe-

14.2.1. Prove that the relation e x+ iy ~ riodically with n The value of ax .

changes in the same manner, where

(l for n ^1 (that is, e x+i y = a =
cos cp i sin + cp = e
i(
P is an arbi-
x+iy ' n trary complex number with unit
{ \
lim 1 -I I means the same as modulus: in this case
n-*oo
(
' n J
)
'

formula (14.2.10).
ax = {e l(v)
= e ® x = cos (cp^) + £jsin (cpx)
x l

14.2.2. Prove that in the formula 1 + Now, if a = p (cos + sin = pe i(

kiln =
p (cos cp i sin <p) for 1 we have + cp i cp) t>

<p =
kin 4- 1 ., where the points indicate . .
is a complex number whose modulus p
terms of the
smallness.
1 In
2
and higher orders of is not unity (p 0 and p ¥= 1), then >
14.2.3. Prove that for any complex (and a x
= (pe = p e^x
i(P) x x

not only for pure imaginary) number A z small

= = px [cos (qxr) + sin i (cpa:)] (14.3.2)
in absolute value, we have e* z 1 + Az +
.
., the points indicating the
. terms whose can be expanded into a product of two
2
absolute values are of order Az and higher.
| |
cofactors: one of them (which deter-
mines the absolute value of the number
ax ) will monotonically increase or mo-
14.3 Trigonometric Functions and notonically decrease with increasing x y
the Logarithm while the other equal to e i(px (it is re-
The Euler formula reveals a sponsible for the argument of the num-
(14.2.6)
deep inner relation of periodic functions ber ax ) will change periodically. If
cos x and sin x to the exponential pCl, then the real part and the coef-
function. At first glance the exponential ficient of the imaginary part of a x 9

function ax has nothing in common with that is, the expressions p* cos ((px) and
periodicity: for a 1 the value of a
x
> p* sin (cp.r), will simulate damped oscil-
lations (see Section 10.4 and, in partic-
monotonically increases with increas-
ing x while for a ,
1 it always de- < ular, Figure 10.4.2). Later (see Sec-
tion 17.1) we will see that the relation
creases; there is no periodicity at all.
But if a = —
1, then the powers of the
between the function ax and damped
number a assume the following values: oscillations is by no means accidental.
The Euler formula (14.2.6) enables
-1. +1, -1, +1, -1, +1, . .
constructing many other important
(14.3.1) functions of a real, imaginary, or com-
that is, they change periodically 14 - 4 plex variable. Let us begin with re-
.

If we write writing this formula and expressing

trigonometric functions via an expo-
— = cos n +
1 i sin n = e in , nential function, rather than an expo-
nential function via trigonometric ones.
then we can consider the power n in
From
the notation ( l) n to be any real
number: e'w = cos + cp i sin cp, (14.3.3)

(
— l)
n
= (
e i7l
)
n = e inn = cos { nn ) it also follows that
+ i sin (nn ) e~ i(P
= cos — (
cp) 4- i sin ( — cp)

Here (generally speaking, complex)

= cos cp— i sin cp. (14.3.3a)
quantity ( — l) n assumes various values Adding (14.3.3) and (14.3.3a) term-by-
We have just limited ourselves to
14 • 4 term and then subtracting (14.3.3a)
natural (integral positive) values of the power. from (14.3.3) we readily obtain
If we let n also assume integral negative
values and vanish, we will infinitely continue COSCp^rr
e
i<P
+— £
,-i<P
,
.

Sinq>=:
g*P_e-i<P
777 .

to the left the periodic sequence (14.3.1) of

numbers ( — l)
n .
(14.3.4)
 i i

14.3 Trigonometric Functions and the Logarithm 467

The corollaries (14.3.4) of (14.3.3) are such an angle since the cosine of any
also often called Euler formulas . angle does not exceed unity. Here it
Note that for real cp the terms on the is tacitly implied that the angle is a
right-hand sides of (14.3.4) are certain- “real” angle, that which can be drawn
ly real (they are the cosine and the or measured with an instrument. But
sine of an angle): indeed, substituting now that we have introduced imaginary
the expansion (14.2.8) and a similar angles we can obtain a definite numer-
one for e i(P and e~ i(P in these expressions ical solution: the equation
we again obtain formulas (14.2.9) and
(14.2.9a).There is another approach cos q> = 1.5 (14.3.7)

to readily convince ourselves that the has the (approximate) solution cp i. ~

expressions obtained are real. Let us Obviously, the value cp is also ~—
replace i by in the expression z— = a solution of Eq. (14.3.7); moreover,
P +
iQ for a complex number z (P and there are infinitely many solutions
Q being real); we obtain a complex- (14.3.7)differ by multiples
which of 2n. Thus,
conjugate number z* P iQ. If z = — before we had no solutions of Eq.
did not change here, that is, z* z, = but now we can consider them
then Q =
0 and z is real. But we can to be infinitely many:
see that (for real (p) the right-hand
sides of (14.3.4) do not change on sub-
cp ==: ziz i “I
- 2kn ,
k = 0, ztl, +2, ....
stituting i —
for £, and consequently (14.3.8)
they are real.
Further, using expressions (14.3.4)
The situation hereis quite similar

for cos cp and sin cp we can also deter-

to that we encountered in the
case of algebraic equations. When we
mine trigonometric functions of the
used only real numbers, the quadratic
imaginary argument via exponents. We
replace in (14.3.4) cp = ix]), assuming yp
equation x 2 —n=
with n 0, had
0, >
real, and so cp is pure imaginary. We get two solutions x — ±]/"ra, while
(roots),
the equation x 2 n = 0 had no solu-
+
= tion. But when we used complex num-
cos (ii|)) sin (i^)
2 bers, the second equation also had two
solutions, x = zh i^ n. Similarly, the
(14.3.5) equation cos cp = n, for n 1, had ^
infinitely many solutions (for example,
Thus we are back again to the exponen- to the value n =
1/2 there correspond
functions of the real argument yp
tial
and more accurately to certain com-
angles 9 zizJi/3 = +
2 fcrc, k being any

binations of exponents we are going to

integral number), while for n 1 the >
equation had no solution; but when cp
discuss in the next section of this
is complex, the equation has infinitely
chapter.
many solutions.
In discussing the functions cos (tip)
The problem of solving equations can
and sin (h|)) we can solve still another
be formulated as that of determining
problem. For example, assume yp
= 1.
an inverse function. Let us consider a
Simple calculations yield —
(“direct”) function t x2 the deter- ,

e 1 -f- e~ x _ 2.72+0.36 mination of its inverse reduces to

cos i = 2 solving the equation x 2 t where t is = ,

(14.3.6) given and x must be found. Here for

real variables the original function
2 7 2 -°- 36
(and sin i ~ i
-
_ =l,l8i) . We exists for all values of the argument x ,

while the inverse function only exists

pose the question:
whose cosine equals 1.5? Before we
What is the angle for the argument t 0; in the complex
plane the inverse function is also de-
^
could have answered that there is no fined for all t. Similarly, the function

30 *
 u .

468 14 Complex Numbers

cp =
arccos s the inverse of s
,
cos cp, = ber w = u iv: p = \
u? \
=
also exists for real variables only in the Yu + 2,
v 2
cp = Arg w = arctan (v/u),
interval —1 ^ s ^ 1
Section 4.11); (cf. we write (14.3.10)
,

thus
while for complex numbers the function
is defined for all s . In ( u + iv) = In Y + v + 2 2 i arctan ( vlu )
The situation proved to be the same (14.3.11)
with the logarithmic function z = In w,
which is defined as the inverse of the We have finally arrived at a new
exponential function w e We know = z
.
notion of the logarithm of a number,
that real numbers do not always have a notion wider than we learned at
logarithms, for example, negative num- school. Formula (14.3.10) (or (14.3.11))
bers have no logarithms. Logarithms enables us to find, for example, the
of complex numbers, as we will see logarithm of a negative number. Indeed,
later, are always defined. Note, finally, we can write, say,
that all the three inverse functions, —5= 5x (—1)= 5 (cos ji + i sin it)
~x =Y t, cp =
arccos s, and z In w, =
are not single-valued, the nature of this
being different; the reasons for this whence it follows that
^difference will be discussed later.
In (—5) = In 5 + in
We
rewrite formula (14.2.10) thus
~ 1.6 + 3.14t. (14.3.12)
<e
x+iy
=p (cos y+ i sin y ) = pe ly , (14.3.9)

where p = e
x
Since
that x = In Since (cf. Section 4.9) log 10 x =
the equation w
,

e* can be written —
is, p.
In z/ln 10 — In x/2.3, we have
as z =
In w, from (14.3.9) it follows log 10 (~5) - (1/2.3) In — 5) (

that if w =
p (cos cp i sin
9), then + ~ 0.7 + 1.37 i.

z = In w = In p + jcp
the modulus p of the complex
= In |
w |+ i Arg w. (14.3.10)
If
number z is unity, that is, if z =
This relation can, of course, be con- cos cp + i sin cp, then In z — cpj =
i Arg z. This relation of logarithms of
sidered as the definition (unknown to
us before) of the number In w: we “de- complex numbers and their arguments
rived” (14.3.10) from (14.3.9) on the is responsible for that the properties
assertion that e w means the same z
= of the arguments of complex numbers
as z —
In w, which is only valid for are close to those of logarithms: for
real z and w and not for complex (or any two (complex) numbers u and v
aven for negative real) w for which we have
In w had no meaning before. The ex- = log u + log v
log ( uv ) y
pediency of the definition (14.3.10) is
connected with that for cp Arg z 0, = = log =logu — logy,
that is, when z is a real positive number ,

formula (14.3.10) leads to the usual Arg (uv) = Arg u + Arg v,

notion of logarithm (when introducing
logarithms of complex numbers we have Arg = Argi^— Argy, -

just to refresh our knowledge and not

to start from the very beginning). that is, when multiplying complex
Besides, by (14.3.9) here we always numbers their arguments (and loga-
^for any w) have e ln
w w. = rithms) are added, and when dividing
Expressing the numbers p (that is, they are subtracted. (Of course, here
|
w |) and cp (that is, Arg w) via the the symbol log u can also mean the
real part and t the coefficient of the logarithm to any base, since all loga-
imaginary part of the complex num- rithms are in proportion.)
14.4 * Trigonometric Functions of a Purely Imaginary Independent Variable 469

Note that the angle cp, which is the 14.4* Trigonometric Functions of
argument of a complex number and a Purely Imaginary Independent
enters into the formulas, can be re- Variable. Hyperbolic Functions
placed by the angle cp 2fcrt, where + Let us return to formulas (14.3.5) for
k is any integer. Therefore, the loga-
trigonometric functions of the imag-
rithm of any number is not single-
inary argument:
valued, but infinitely many-valued For .

= —+
example, the complete solution of e* e
-*
COS (l\j))
(14.3.12) is
(14.4.1)
In ( — 5) = In 5 + i (n + 2fcrt), sin (i(p) — i
e yb__ e -ty
^

k — 0, zb 1 , ±2, . . ..

The combinations of exponents e^ and

e~^ contained in the formulas have
The same refers to the logarithm of any
special names and notation. The former
positive number. For example,
is called the hyperbolic cosine and the
In 5 = 1.6 + i x 2&jt, latter the hyperbolic sine of the argu-
k — 0, +1 ±2, ,
. . ., ment i|), and denoted by

or
cosht|) = — ,
sinh yp
— *

In 1 = i X 2/br, k — 0, -4-1. ~4~2. ...,

(14.4.2)
since
(The letter h reminds us of hyperbola.)
1= cos 0 + sin 0 = cos 2n i
Thus,
+ sin 2n =
i = e° = e 2ni . . .
cos (hf)) = cosh if),

^4 — Jit
sin (n|>) = sinhi yp. (14.4.3)
Later we
will return to the question as Some properties of hyperbolic func-
to why
the logarithm is not single- tions copy the properties of (usual
valued, why every number has not one, or circular) trigonometric functions, re-
but infinitely many logarithms. producing them sometimes in a some-
Note, finally, that logarithms enable what distorted form. For example,
us to raise to any, (complex), power of f
rr\3Wi *1 1,

any (complex) number. For example, the function cosh even: cosh ( \|))
ip is — =
let us find the number i
%
(what will e ~^
(i
+ e^)/2 = cosh yp, while the func-
we obtain on multiplying the number i
into itself i times?). Obviously,
tion sinh yp is odd: sinh ( =. —^
i = cos ~ + i sin =e
and, consequently,
ji
— (erti/2)i __ e (Ui/2) i _ g-rt/2
~2.7 _1 - 57 = frr
157
~0.21
*
2.7

(the number is even found to be real).

Exercises

14.3.1. Let u
p (cos cp =
i sin <p) and +
z = x + iy. the values of u z
Write all .

14.3.2. (the jocular problem). What is

larger: e e or i*? e i or i e ?
 r

470 14 Complex Numbers

(, e
~ = — sinh On the other
— e't>)/2 ip- metric functions), whence it follows, in

hand it is obvious that cosh (0) = 1 particular, that

and sinh (0) = 0 (compare the curves
of the hyperbolic sine and cosine repre- sinh ij; ^ i|), cosh ^ ^1 j?2_
2 ’
(14.4.6)
sented in Figure 14.4.1). Figure 14.4.1
also presents the curves of the hyper-
tanh of) ~^
bolic tangent: for small yp (we have here neglected
—e ^ terms of the third and higher orders of
sinh
tanh ip — cosh ip gii’-f.g-il’
smallness with respect to i|>). Further,
it is obvious that

clearly, the function tanh ip is odd (sinh o|))'= g

and tanh (0) =
(
(tanh (— ip) i
= —tanh ip)
2 )

0.
14 - 5
Note also that to the well-known e^+e-y
sin 2 cp = 1 there cor-
— cosh oj),
formula cos2 <p + (14.4.7)
responds the following formula of “hy-
perbolic trigonometry”: (cosh i|))'
= ( 2 j

cosh 2 ip — sinh 2 ip=l. (14.4.4)

= sinh t|),

Indeed,
whence, with account taken of (14.4.4),
cosh 2 ip — sinh 2 ip
= ( g
) we have
- 21
*—e~* \2 _ e 2^ +2 + e l)

( e - (tanh tp)' = (S|)

~l 2 ) 4

g 2H>_
(sinh yy cosh ip sinh — ip (cosh ip)'
2+g -2M> _ 4
cosh 2 ip

= cosh 2 ip— sinh 2 ip _

— 1
The expansions into series
cosh 2 ip cosh 2 ip

'“ 1 + T+f + ! + £- + 2
(compare with the formulas for deriva-
tives of trigonometric functions in Sec-
! 4!

2
if tion 4.10).
e-^— 1- + 2 ! 3! + 4!
+ .

* ‘ *

Of no resemblance
less surprise is the
of the weil :fnmmlnfo Known '(‘"Sihftal”';
and formulas (14.4.2) yield
of addition for trigonometric functions:

cosh ij) = 1+ ^2
i

' ^ ,

‘ ‘ ‘ ’
= sin a cos P + cos a sin p,
sin (a-)- P)
2i 4! 6!

(14.4.5) cos (a + P) = cos a cos p

— sin a sin p,
<^h3
u . ta n a + tan. [3

sinh ip = ip -)- -j-

+ -JT+ tan (a + P) — _ an a tan
. .

p \ t

and similar formulas for hyperbolic

(compare with the expansion (14.2.9) functions
and (14.2.9a) into a series of trigono-
sinh (u + v) = sinh u cosh v
14 • 5
We know
that the (ordinary) cosine -f cosh u sinh v,

and sine (of a real variable) are bounded:

|
cos <p 1 sin <p | c
1 while the tangent
, | | c ,
cosh (u + v) = cosh u cosh v
can assume arbitrarily large values. But the -(- sinh u sinh v,
hyperbolic cosine and sine are not bounded
. sinh (n+O
since as is readily seen cosh
sinh \|?-»- oo as
oo as \p-»- oo,
-> oo; while tanh \|? is
tanh (u + v)— cosh + v)
bounded: tanh \p 1 (and obviously < sinh u cosh v -f- cosh u sinh v
=
| |

lim tanh (14.4.8)

1).
\J?-*-oo cosh u cosh n+sinh u sinh v
 u A — p

Imaginary Independent Variable 471

14.4 * Trigonometric Functions of a Purely

sinh u! cosh a -f- sinh p/cosh u

~
1 (sinh u sinh v)/(cosh a cosh u)

— tanh u tanh + u

1 + tanh u tanh v

(the last of these formulas was ob-

tained by dividing the numerator
and denominator of the fraction
sinh ( u + a)/cosh ( u + v) by cosh u
cosh u; as to the first and the second
formula see Exercise 14.4.1). From
Eq. (14.4.8) it follows that
Figure 14.4.2
sinh 2u = 2 sinh u cosh u,
cosh 2 = cosh u + sinh 2 u 2 (14.4.9)
,
sinh (—'ll)) = — sinh and tanh
o|), = ( ty)
— the equalities cosh (0) = 1, sinh (0)
tanh 2 u —
2 tanh u tanh'i|));

+ tanh u 1
2 = tanh (0) = 0; and some others. 14 - 6

(we have obtained these formulas by Wecan now look from a new angle at
setting u =
v in (14.4.8); all formulas the relation (14.4.3) of circular and
(14.4.4)-(14.4.9) resemble very much hyperbolic functions. Previously, the
those you already know, don’t they? sine and the cosine of angle q) were de-
fined as the segments in a circle of
The names of hyperbolic functions cosh i|)

are connected with the following.

radius 1 (see Figure 14.4.2a) or as the
and sinh \1)

Ordinary (circular) cosine and sine can be de- ratios of the sides of a right triangle
fined as the abscissa and the ordinate of a with acute angle cp. But these defini-
point M
of a unit circle x 2 y
2
1 (Fig- + = tions tell us nothing of, say, what sin i
ure 14.4.2a); the argument here is the angle
— or cos i is. What is the strange angle
L.AOM (p, which can be considered as a
doubled area of sector S aom of the circle (eur of magnitude i? How can we turn through
circle has the radius 1). Similarly, the hyper- such an angle? How can we construct
bolic cosine and sine can be described as being a right triangle with such an angle?
the abscissa and the ordinate of a variable
— — The last questions, of course, remain
point M
of a unit hyperbola x 2 y
2 1
without any answer, but the values
(Figure 14.4.26), where the argument is the
hyperbolic angle which is the doubled area
'll), of sin i and cos i can quite possibly be
of sector AOM
of the hyperbola (see Exer- found with the help of Euler’s formulas.
cise 14.4.5). Figure 14.4.2 also presents geo-
metrically the circular and hyperbolic tangents:
Note that the series (14.2.9) for the
cp 2 cp 4

sin cp MP AT _ cosine, cos cp = a

1 —^ ,

~r
^
•••>
tan cp
cos cp OP O contains only even powers of angle cp;
(see Figure 14.4.2a) therefore, on substituting cp i (or, =
sinh _ MP AT generally, cp — tyi, where is a real
tanh \[) = \()
— AT number) all the terms of the series
cosh ip OP OA
remain real: the cosine of a pure imag-
(see Figure 14.4.26)
inary angle is a real number. The
(note that in both cases OA 1); the seg- = series (14.2.9a) for the sine, sin cp =
ment A T is called the tangent line, and the
segments MP
and OP the sine and cosine (p — + . . ., on substituting cp = i\

lines. Many properties of hyperbolic functions

follow from the foregoing: formula (14.4.4)
(the relations cos 2 cp sin 2 cp 2
1 and cosh aj>+ = 14 6
Note that formulas (14.4.8) for addi-
•

— sinh 2 \J) =
1 are no other than the equations tion and the definitions (14.4.2) for hyperbolic
x2 + y
2
1 —
and x 2 y
2
1 of the circle — — functions can also be derived using the geometric
and the hyperbola); the even nature of the representation of hyperbolic functions pre-
function cosh and the odd nature of the func- sented in Figure 14.4.26 (see, for example,
tions sinh \1) and tanh-ip (from Figure 14.4.26 V. G. Shervatov, Hyperbolic Functions Heath,
it is readily seen that cosh ( cosh \p, — = 'll)) Boston, 1963).
,
 .

472 14 Complex Numbers

yields a pure imaginary expression and m

whence the coefficient i appears in for-
mula (14.4.3) relating sin (n|?) and the
real function sinh i|).

The analogy between trigonometric

and hyperbolic functions can also be
seen when we discuss an important
topic on differential equations. We
know that the differential equation

^-=-kx (14.4.10)
for k >0 admits of a general solution
x = A cos co£ + Z? sin co£, of the base of the hill, that is, the com-
co = Yk (14.4.11) ponent of the force of gravity, which
is tangent to the surface of the hill and
(seeSection 10.2). On the other hand,
the equation
tends to move the ball, is zero. However,
once the ball moves even slightly, say,
(14.4.10a) to the right, a force F begins to act
on it in the direction which increases
where k is also greater than zero, as we the motion, the force growing in pro-
have already seen (see Section 13.10), portion to the increasing displacement x
(we can consider the force to be pro-
has particular solutions x = e^hi and
x = e-Vhi so that its general solution
portional to the displacement: F kx —
,

can be written in the form

for k >
0). By Newton’s second law (see
Section 9.4) the equation of motion of
x^CeV^t + De-V^K the ball is

where the constants C and D are arbi- m-^- = F=kx. (14.4.12)

trary. If we write this solution as
The solution to this equation (cf. Eq.
elTZt+e-V** (13.4.10a)) has the form
x=(C + D)} 2
ft

+• (C-D)
eV’bt g-V'kt x = ae^ m be ^ \ (14.4.13)
2
or, which is the same,
and denote C +D = A, C —D = B,
x = A cosh -p B sinh eot, where
we, by virtue of definitions (14.4.2),
obtain a general solution of (14.4.10a),
which is similar to the general solution ®=]/ (14.4.14)

(14.4.11) of (14.4.10): The existence solution that

of a
x—A cosh ot+B sinh
c cot, where co = .
grows exponentially 14 7-
z/
x =
(14.4.11a) indicates that the position of the ball
on the top of the hill is unstable . We
The relation trigonometric and
of
can expect the initial conditions to be
exponential (or hyperbolic) functions
sheds new light on the initiation, in
generally such that in (14.4.13) a 0 ^
similar conditions, of periodic motions 14 7 Or the existence of the solutions
Yx = •

(oscillations) and aperiodic motions. cosh co£, Y 2 = sinh w t (growing exponential-

ly as well). It is readily understood that
Let us consider a ball lying on a cir- as grows the functions u = cosh and u — \|? \|?

cular top of a hill (Figure 14.4.3a). sinh grow by the exponential law: \|?

On the very top the force of gravity is _ cosh

= lim sinh 1
— —
nr— = ^
'll? _ . \|?

1 lim 1
com jnbteiy naianceki ny ’me reaction aj)-*oo
‘
 k , 3
-

14.4* Trigonometric Functions of a Purely Imaginary Independent Variable 473 ;

and b 7^
0 (it is quite unlikely that the
= (14.4.12a) is bounded for all t which ,

conditions ensure
initial the strict is closely connected with the steady
observation of the equality a 0 ). = state of the ball in the well.
Consequently, in a certain period of time
the first term on the right-hand side of Imaginary numbers (such as the square root
(14.4.13) will necessarily become rather of a negative number) were first introduced
(although in a rather formal manner) by the
large and the ball will go down.
prominent Italian Renaissance mathemati-
Now we consider another (quite op- cian Girolomo Cardano (1501-1576). It was
posite in a certain sense) situation. Let by no means accidental that complex numbers
the ball lie at the bottom of a well appeared when algebra was making such
strides or that Cardano was the man who intro-
which has oval walls (Figure 14.4.3b). duced them. (In ancient times, geometry was
When the ball is displaced a distance x the center of attention; in the Renaissance,
from the equilibrium position, a force F it was algebra that paved the way for the rise
appears which is directed toward the of mathematical analysis in the 17th century.)
Cardano won himself a place of honor in the
bottom of the well, the “returning” history of mathematics chiefly for his pioneer-
force. (Such direction of the force in- ing publication of the formula for solving am
dicates that the equilibrium is stable arbitrary 14 8 cubic equation 14 9 if
- -

in the case of the well as distinct from =

the equilibrium of the ball on the hill.)
xs + px -f- q 0, (14.4.15).

Suppose that the force F is proportional then

to the displacement x in magnitude, so A
that F = —
kx where k is positive
, *“]/ ~t+I// "

(t) +(t)
s

and the minus sign in the formula in-

+V zi z7W^W-
dicates that F tends to return the ball
to its initial position. Then the equa-
tion of motion of the ball will be
(14.4 .16)4

-kx. (14.4.12a) But seemingly simple Cardano formu-

the
la (14.4.16) is tricky.For instance, if (14.4. 15)*
Formally, this equation looks like the has the form x s 3x —
0, that is, if p = — —
and q =
0, the equation has the following sim-
previous one (Eq. (14.4.12)). Therefore, = =
ple roots: x 1 0, i 2 =|/' 3 1.73, x 3 ~
we can use the solution of Eq. (14.4.12) V" 3 ~ —
by substituting —
for k, since this
— 1.73, then (14.4.1 6) yields the re-

substitution transforms Eq. (14.4.12) markable result x — / y — 1 V/ -/=T=

into Eq. (14.4.12a). Here we obtain the f
7
i + V—
—
that is, we suddenly have square'
roots of 1.
solution to the new problem The procedure by which square and cube
roots can be extracted from complex numbers-
was developed by one of the Cardano’s follow-
x= ae^
1

+ be~^ = aemt ers, Raphael Bombelli (c. 1530-1572), who

was able to explain how formula (14.4.16)
guarantees in all cases a correct way of find-
-\-be ~ i(0t (o=
, j/ ,
(14.4.13a) ing the roots (three roots, though not alb
necessarily distinct) of Eq. (14.4.15) (see Exer-
cise 14.4.4).
or, with the change of variables, The most influential French mathemati-

14 8
Cardano did not discover it, however;
-

y — A cos co£ + 5 sin co£ he borrowed the formula— (14.4.16)— from,

Niccolo Tartaglia (whose real name was Nic-
(14.4.14a) colo Fontana) (c. 1500-1557) (e.g. seeM. Kline,
Mathematical Thought from Ancient to Modern
(cf. Eq. (14.4.11); expressions (14.4.13a) Times Oxford University Press, New York,.
,

and (14.4.14a) coincide if we set A = 1972, p. 263).

14 9
On reducing the arbitrary cubic equa-
{a +b)/ 2, B = {a —
b) i/2). Here the
tion ax z
-

bx 2 cx +d +
0 to (14.4.15),. + =
(real) solution (14.4.14a) of Eq. see formulas (1.5.6) to (1.5.6a).
 ,

474 14 Complex Numbers

rcian of the 16th century, Francois Viete (1540- found results in the field of “complex algebra
1603), showed that when ( q/2 )
2
(p/3) +
3
is and analysis.” For one, Gauss created a non-
less than zero, that is, when the cube roots in trivial “arithmetic of whole complex numbers”,
Cardano’s formula (14.4.16) are extracted from numbers like a +
ib, where a and b are whole
complex numbers, it is possible to reduce the numbers, or integers. In this unusual arith-
right-hand side of (14.4.16) to a simple com- metic, 3 is a prime but 5 is not since 5 =
bination of trigonometric functions. This (1 + 2 i) (1 —2 complex
i). The theory of
paved the way to using trigonometric functions numbers can be said to have become a full-
in the theory of raising complex numbers to fledged branch of mathematics only beginning
.a power (even negative and fractional powers), with Gauss. Before Gauss, a geometric inter-
that is, to the general formula (14.1.6). But pretation of complex numbers (see Figure
in complete form, formula (14.1.6) was first 14.1.1) was presented by Caspar Wessel
^expressed by the French mathematician Ab- (1745-1818), a self-taught Norwegian-born sur-
raham de Moivre (1667-1754). L. Euler, how- veyor, and the Swiss Jean-Robert Argand
ever, gave it its modern form. (De Moivre fled (1768-1822), who was also a self-educated math-
from his native France to England to escape ematician and a bookkeeper. However, the
the persecution of the Huguenots, and he made publications of these two practically unknown
.a big contribution to the progress of mathe- scientists failed to attract any attention at
matics in England.) the time, and Gauss arrived at his ideas inde-
The famous Leonhard Euler made wide pendently.
use of complex numbers. He was the one to
introduce the notation i for V—
1, and he also
found most of the results of this chapter. Euler Exercises
was convinced of the validity of the funda-
14.4.1. Prove formulas (14.4.8) for adding
mental theorem of algebra, which states that
hyperbolic functions.
>every polynomial equation of degree n (with
2 tanh (t/2)
real or complex-valued coefficients) has exact- 14.4.2. Prove that sinh t- '
ly n roots (in general, complex-valued), and
2
1 — tanh (t/2) ’

iie made many attemps to prove it, but this

cosh t — 1 + tanh 2 (t/ 2)
and tanh t =
requires a full understanding of the algebraic i - tanh 2 (t/2)
.and the geometric meaning of complex num-
2 tanh (t/2)
bers (see Figure 14.1.1), an understanding that
the mathematicians of those days did not 1 + tanh 2 (t/2) *

yet have. Jean d' Alembert, the second out-

14.4.3. What curves are given by the fol-
standing mathematician of the 18th century,
lowing parametric equations:
came closer to proving the fundamental theo-
rem of algebra, but he failed to carry it through -* 2
,
a)
v

x = tt l
y =-
21
to the
We
end.
first proof (proofs, to be more
owe the
(
+ 1 1
2 .
l-M 2
*exact, there were several) of the funda-
for l-M 2 y=-
21
mental theorem of algebra to Carl Friedrich
(b) x~-
l —t 2 1 — t
2

Gauss (1777-1855). He also provided the mod-

ern idea of the plane of the complex variable What geometric meaning does parameter t
z = +
x iy, thus freeing complex numbers have in these formulas?
from the last traces of mystery, even of mys- 14.4.4. How does formula (14.4.16) lead
r
ticism, 14 10 and produced many other pro-
-
to the values 0 and ±\ 3 of the roots of the
cubic equation x 3
Sx 0 considered in — =
14 - 10
Yet even the much simpler negative the text?
numbers were firstintroduced only by Car- 14.4.5. Prove that cosh = OP and of
dano and Viete,and even in Viete’ s time math- sinh = MP in the notation of Figure 14.4.2a,
ematicians viewed them with great sus- where is twice the area of the hatched hyper-
picion. bolic sector.
 2

Chapter 15 Functions the Physicist Needs

15.1 Analytic Functions of a Real (color of dress as function of name of

Variable girl) cannot be expressed by a formula,
so one cannot find the derivative or
The idea of a function and of a func-
integral of such a function, with the
tional relationship have undergone much
result that all the tools developed
change through the history of mathe- over the centuries by mathematicians
matics. Different approaches to the
prove to be useless, since these tools
notion of a function in various periods
cannot be applied to such functions.
sometimes led to stormy discussions 151 ,
Functions obtained as a result of
and at present are reflected in the differ- experiments in the form of lists of
ent ways in which a function can be
experimental data always prove to be
defined.
given for a certain set of values of the
The broadest possible definition of independent variable within a certain
a function states that any relationship
accuracy 15 2 Gan the derivative or inte-
.
-

between the element of two sets can be gral of such a function be found? By
called a function. If Mary is wearing
the very meaning of a derivative (or
a white dress, Ann red, and Dorothy
integral) such calculations involve find-
blue, we can say that the color of the
ing the values of the function in a small
dress is a function of the name of the
neighborhood of a fixed value of the
girl (or that the dress is a function of
function and the independent variable.
a certain person). Here the domain of
But it is impossible to obtain such
definition of the function (or “input,”
values from a table or experiments
as mathematicians today say) consists
directly, especially if one takes into
of three names (or three girls) Mary, — account the errors of measurements.
Ann, and Dorothy— while the range of One must assume that there exists a
values of this function (or “output”)
smooth functional relationship between
consists of three colors— white, red,
the quantities that are measured, since
and blue (or three dresses). Experi-
only then can we use the tools of higher
mentally observed dependences of one mathematics; the experimental data
type of physical quantities on another
or tables are needed to construct such
type, for instance, electrical resistance
a relationship.
of a wire on the wire’s temperature,
Of tremendous importance is the fact
are also functions. Finally, one can
that a fundamental theory always leads
write many formulas of the type y = to quite natural mathematical formulas.
ax2 bx+ c and y+ e~x / 2 =
and all ,
This is also true when we are forced to
these are functions, too.
reverse certain relationships (say, solve
But is it advisable to gather all these algebraic equations) and arrive at solu-
different ideas under one name, func-
tions that are discontinuous, that is,
tion? In some respects it is, since this
a gap between the roots of an equation.
stresses the enormous generality of the
A physical theory is always constructed
concept of function. This generality, in such a way that it preserves the
however, is attained at a certain price. possibility of using that precious tech-
A function given purely verbally nique for studying phenomena, differ-
ential and integral calculus.
15 - 1
Best known of all is the controversy Thus, functions given by formulas
on this subject (mentioned in Section 10.8)
involving three outstanding mathematicians possess enormous advantages. It is
of the 18th century —
Leonhard Euler, Jean clear, firstly, that specifying a formula
d’Alembert, and Daniel Bernoulli. (Actually
15 2
they never gave up their original opinions.) •
There are inevitable errors when
The real question was: can a graph with a the graph of a function is the only source
break in it or a salient point be considered of information about the function (say, plotted
to represent a single function? by a self-recorder).
 f

476 15 Functions the Physicist Needs

enables calculating a function to any correspond numbers of the form

to
accuracy and for any value of the inde- m!n = where m and n x are
ml2n x ,

pendent variable. In contrast to the relatively prime positive integers and

case where the function is given by m is odd; there is an infinitude of such
a table or lists of experimental data, numbers on any given finite segment
repeated calculations by a formula of the x axis. The techniques developed
always yield the same result the accu- — in this book are not suitable, of course,
racy can always be'preassigned. Another for studying such functions.
convenient feature of a formula is that the other hand, we have, say, the
On
calculating the values of the function “ideal” function y 1 x, whose = —
for different values of the independent graph is a smooth (even straight) line
variable is reduced to a sequence of and which at each point has a tangent
similar, single-type operations; this
,
line (which coincides with the graph
feature is especially important when of the function itself), that is, the
computers are employed. A formula function has a derivative everywhere.
also enables calculating, to any ac- The inverse of this function can also
curacy, the difference of two values of be easily found (it coincides with the
the function at different (but closely initial function); the same is true of
lying) values of the independent vari- the function / ( (x)) (which is simply
able, which makes possible the numer- F {x) = x); it is easy to find the inte-
ical calculation of the derivative. Ad- gral of this function:
ding the values of a function at adjacent
points, we can calculate the integral
of this function, and thus both the der-
f(x)dx=(x-^-)l
ivative and the integral can be calcul- j
a
ated to any degree of accuracy Today, in .

the computer age, such direct compu-

tation of derivatives and integrals is
quite simple.
-<<-»> ( 4i - 1
)
;

However, even among functions for and so on.

which one can state an exact method Intermediate cases are also possible,
for calculating / (x) from the value of x that is, functions that are not as “bad”
there are monstrosities. Take the follow- (or, as mathematicians say, pathologi-
ing function: / {x) =
1 if x is an irre- cal) as the “fence” we described above
ducible fraction with an even denom- and yet not as “good” as the function
inator and / {x) =
0 for all the other we have just described. So how is one
values of x: lh Z to classify the “good” and “bad” func-
tions? What functions are considered
1 if x — mfn ,
where m and n “good”? One approach that clarifies this
are relatively matter considerably is the transition
= prime and n is to complex numbers and (complex-
f (x)
•

even, valued) functions of a complex vari-

0 otherwise. able— but we will try to answer
first
these questions while remaining in the
real domain for the time being.
Just try and construct such a function,
One of the main results of Chapter 6
and you will see that this is impos- func-
is the possibility of representing
sible. You would have to build a dense
tions in the form of series:
fence of vertical segments of unit length
each erected at points of the x axis that /'( 0 ) r (Q)
/(*) = /( 0 )
1 ! 2!
15 3
-
Even such an exotic function (but
mathematically cerrect) can be expressed /"'( 0 )
still
by a formula, albeit a rather complicated one.
,

3!
X3 + (
15 . 1 . 1 )
 2

15.1 Analytic Functions of a Real Variable 477

{Maclaur ill’s series) or It is clear that an integral rational

function (a polynomial)
f (x) = f (a) + (x — a) / (x) = a0 + ctiX + cc 2 x2 + . • . + d n xn

+ J^(*- a 2 + )
T - (^-«) 3 + •

(15.1.2)
• •

is analytic: here / (0)

/"( 0) -
2 !a 2 /"'(0)
a0 =
=
,
/'( 0)
3!a 3
(15.1.5)

=a
. .
1?
.
, ,

{Taylor’s series, which generalizes Mac- . .

n
., /< n\a n
> (0) =
and /< ,
m)
(0)
—0
laurin’s series). These series make it for m >
n (thus, representation (15.1.1)
possible, if we know the value of a of such a function coincides with
function at one point and the derivative (15.1.5)). We can say that analytic
at the same point, to reconstruct the functions, such as e x 1 -f x x 2 ! 2! = + +
function (with as high an accuracy as £ 3 /3! +
... or cos x 1 x2 I2\ = — +
desired) in the vicinity of the point z 4 /4! —
... or In x 1 z x2 !2 = + + +
and continue it to another point. If,
3
x /3 +(see Chapter 6), are the
. . .

say, formula (15.1.1) holds true, then, most natural generalizations of poly-
differentiating it term by term, we nomials: a general analytic function
find the series expansion for the deriva- (15.1.1) or (15.1.2) is, so to say, a poly-
tive nomial of infinite degree.

r (*)=/'
f" (0) r (Q) x 2 -f- Of course, a function that is analytic at
(0)
1! 2! a point x —
a is “infinitely smooth” at this
point, that is, has derivatives of all orders
(15.1.3)
at this point, since the derivatives are present
in representation (15.1.2), the existence of
and, required, for higher-order deriv-
if
which is equivalent to the function being ana-
atives. Similarly, integrating (15.1.1), lytic. However, there is no rule by which the
we can write existence of all the derivatives of a function
leads to the analyticity of the function; in
$f(x)dx = C + f(0)x+l$La* other words, analyticity is a stronger require-
ment than the condition that the function have
an infinitude of derivatives. Indeed, the fact
r (Q) x3 4 r (Q)
x*+ (15.1.4) that a function has all the derivatives implies
3! 4!
only that we can write the series on the right-
hand side of (15.1.2), while for the function
It is to functions of this type that the / (x) to be analytic it must be represented by
techniques developed in this book can the series, which means that the series must
converge (the sum exists) for all x close to a.
be applied most successfully. Functions
Take, for example, the function y -1
c /*
2
=
that can be represented in the form (Figure 15.1.1). At x —
0 the expression e
-1 /* 2

{15.1.1) or (15.1.2) are known as ana- has no meaning, but since for small (in abso-
lytic functions. lbA lute value) x the number 1/a; 2 is extremely large
and, hence, e~ x i x is very small, it is natural
15 * 4
to assume that y (0) =
0. The reader can easily
Of course, formulas (15.1.1) and see than all the derivatives also exist in this
<{15.1.2) can be applied usually only in a lim- case: y( n ) (x) =
d n y!dx n with n , 1, 2, 3,
—
., . .

ited range of variation of x (that is, in a lim- which can easily be found by the formulas of
ited vicinity of point x 0 or x =
a), a situa- = Chapter 4 (say, dy/dx e —
-1 2
/* [d ( 2
1 /x )/dx]=^ —
tion discussed in detail in Section 6.3. In 3
(2/a; ) e
-1 /* 2
), but are all zero at x 0 (see —
accordance with this, only the concept of Exercise 15.1.1), which means that the right-
a function analytic at a point (i.e. such that hand side of Maclaurin’s series (15.1.1) re-
it can be expanded in a power series (15.1.1)
or (15.1.2) in the vicinity of the point under
consideration) has a rigorous meaning, a func-
tion that is anal ytic on an interval is analytic
at each point of that interval. Here we will
not dwell on the possibility of term-by-term
differentiation and integration of formu-
las (15.1.1) and (15.1.2); this possibility
usually materializes for the majority of
functions that a scientist or engineer needs.
478 15 Functions the Physicist Needs

presenting our function is identically zero, 15.2 The Derivative of a Function

while the function proper does not vanish, of
of a Complex Variable
course, at x 0. Fortunately, functions like
e~ x ! x% are very rare and are almost never en- Let us take a (complex-valued) function
countered in physical and technical applica-
tions. 16 - 5
w of a complex variable z, that is,
w =f where w = u + iu and z =
(z),

The requirement that a function be x + iy, with the numbers x and y

analytic is really very strong. For and the functions =u and_ u. (<z. 7 u,\

instance, an analytic function (within v = v (x, y) being real-valued.

the range of applicability of formula If a function / (z) can be expressed

(15.1.1) or (15.1 .2)) may be reconstructed by a formula, then its derivative

from its values on an infinitely df/dz = lim [/ (z A z) +
/ (z)]/Az can —
Az-*-0
small segment (arc) of the graph of the
be found by the general rules of Chap-
function, since knowing f [a) and the
ter 4 for (real-valued) functions of a real
value of f (x) for values of x that are
variable. Here are some examples: if
arbitrarily close to a is quite sufficient
for calculating all the derivatives of
w z
2 =
then dwldz , 2z; if w
hz
e y = =
the function at point a. Therefore,
then dwldz =
ke kz (here k may be com-
analytic functions stand rather far from
plex-valued); if w In z, then dwldz = =
1/z; and so on. There are numerous
the definition of a function as an arbit-
examples of this type; for instance, we
rary correspondence between the values
can form various combinations, such
of x and the values of / (x). However,
almost all functions given by simple
as sums (say, z
2
+ e
:

), products (say,
zV), and functions of functions (say,
formulas are analytic; functions for
which the analyticity condition is vio-
'
1
e 1 ), with d(z 2 + z
e )/dz = 2z e\ +
lated at separate points (like the func-
d(z 2 e )/dz
z
2 ze -f and = z
zV ,
z2
de ldz =
2 ze z \ It is quite natural, therefore,
tion e~y x2 considered above) can be con-
that all the rules referring to functions
sidered as pathological.
of a real variable remain valid when
Of course, in physical problems we
may also encounter functions that are we go over to functions of a complex
variable, since these rules are based
not analytic at some points (say, func-
on the general laws of algebra, say, the
tions that have discontinuities or salient
points), and Chapter 16 is devoted to
distributive property {a b) c ac + = +
even more remarkable functions (it is
be ,
which shows how
remove paren- to
theses in calculations, and similar prop-
questionable whether the name “func-
tion” can even be applied). In the major-
erties. These laws, or properties, do
not depend on the nature of the quan-
ity of cases, however, the functions
tities z, Az, w, A w. The rules for finding
that a scientist or engineer needs are
“good,” or analytic; this statement is
the derivative of / (z) z 2 or of the =
product of two functions, w^w^, are
especially evident when we go over to
based on the following relationships:
functions of a complex variable (see
below). (z + Az) 2 = z 2 -f 2zAz+ (Az) 2
,

A(z 2 ) = (z + Az) 2 — z = 2zAz +2

(Az) 2 ,
Exercise
and, respectively,
15.1.1. Suppose / (x) = 0
e
- 1 /* 2
for x
and / (0)
a function at x
=
0. What is the derivative of such
=
0? The second derivative
(w ± + Aw = ^1^2
+ A wj (w 2 2)

f" (0)? The rath derivative /( n ) (0)? + w kw + w Nw + A^ A^

1 2 2 1 x 2?

15 • 5
The reasons for such strange behavior A {w w = (w + A w (w + A^
x 2) x x) 2 2)
of this function are clarified if we go into the
complex domain, that is, if we continue the
— w w — A^ + w Aw + Az^A^v
1 2 t^ x 2 2 1
function in such a manner that x can take on
any complex values (in this connection see The
raising to an imaginary or com-
Section 15.2). plex-valued power was defined by us
 r

15.2 The Derivative of a Function of a Complex Variable 479*

Suppose dy = 0, that is, dz — dx. Then*

dw ~ + dx) — w (z) — [u (x + dx y)
ir (z ,

+ iv(x + dx, y)] — [u(x, y) + iv(x y)) ,

^i£ dx + -t dx i -

Therefore, the derivative in this case*

is equal to

dw du
W (z)
dev
dz dx dx

Figure 15.2.1 H" i

-fa
— 2x-\- i2y. (15.2. 3>

Now assume that dx = 0 and dz = idy.

in such a way that the (approximate) Then w + dw = w [x + + dy) il
equality e AZ 1 ~ +
Az, where we have and, hence,
(y

discarded terms whose order of small-

ness is greater than that of A z, remains dw dy+
valid (see Section 14.2), that is, we
=% i
4r d y>
employed a formula that lies at the dw dw _ . dw i du , dv

base of calculations of derivatives of the dz i dy dy dy ‘

dy

exponential function of a real variable, —— i ( — 2 y) T“ 2x — 2x i2y

e
x
.We can therefore be sure that all -f- .

the rules for finding derivatives are (15.2.4)

applicable to functions of a complex
In transformations (15.2.3) and
variable: after we have defined that
AZ ~ (15.2.4) we took into account formulas
e 1 +
A z, all subsequent reason-
(15.2.1) and (15.2.2) only in the last
ing is on a stable foundation.
stages, while all the other reasoning
Note that the variation of the inde- retains its value for all functions, that
pendent variable z of the function / (z)
is, if the derivative of the function
can occur in different ways: we can = =
vary only the real part of z, that is,
w w (z) u (x, y) iv (x, y) does, +
not depend on the choice of the incre-
dxz =
d x x and d x y =
0, or only the
ment dz of the independent variable,,
imaginary part of z, that is, d 2z id 2 y = we have
and d 2 x =
0, or the two parts simultane-
ously (Figure 15.2.1). The function w dw du . . dv . du dv
l 1
will change differently depending on dz dx '

dx dy dV

the way we change z, but if dz is small, (15.2.5)

the ratio dwldz, or the derivative, will Since both u and v are real-valued func-
remain the same in all cases. tions, the above relation yields two
Let us verify this using the function formulas, the famous Cauchy -Riemaim
w =
z
2
as an example. We^ave equations :

du dv [du
w — (x + iy) = 2
x2 2 ixy
dx
dv
dy ’ dx
~~
dy *
(15.2.6)
= u + iv, (15.2.1)
These equations are satisfied automatic-
ally if w is given by a formula. Here
y
2
and v = 2 xy\ here
are some examples:
(1) Ifw~ = z2 (x-Hy) 2 thenu = x 2— ,

1Hl
a*
dx
= 2x ^L=—2u
dy v = 2zy,% = %. = 2z, and -£ =
(15.2.2)
dv dv
vv 0 -—•—- = 2 y (compare with (15.2.2)).
dx
 x x

480 15 Functions the Physicist Needs

(2) = (2 + 3i)[(x —
If w=(2 + 3i) z2 2 x and y). The function w = / (z) also
y
2
) -j-
2
i2xy] 6.n/] + — [
2(x 2 —y — )
consists the real part u and the
of
2— 2= imaginary part v whereby to each
y -\-ixy], then
,
i [3 (x )
value of this function there corresponds
= --£-=-6 x-4y.
4x-6y and
(3) If w = e = e cos y
x x
ie sin that
^
z
z/,
a point in another plane (with coor-
dinates u and u). To each z, that is,
to each pair (x y), or to each point in
is, u = e cosy and v = e siny then
x x ,

i
the ;rz/-plane, we assign a pair of values
du dv ~ , da dv
of u and v that is, a specific value of w
dx dy * dy dx
,

—e x
sin y.
(which can also be understood as a
point in the w-plane, and it is best to
Note that the Cauchy-Riemann equa-
distinguish between the two planes so
tions not only connect u {x, y) and
as to avoid confusion, the complex z
v but also impose certain con-
(x, y)
plane and the complex w plane). If
ditions on each of these (real-valued)
functions. Indeed, in view of (15
we assume that u ( x y) and v (x y) , ,

are smooth functions, we can even

d / du \ d dv . d*u obtain the (partial) derivatives duldx,
l \ i.e.
dx \ dx ) dx \dy) ,
dx 2 du/dx duldy, dv/dy, dwldx, and dwldy.
,

d2v
JLI dv \ _ JL( du 1

However, even then we cannot be sure,

dxdy ’ dy \ dx ) dy \ generally speaking, that the derivative

i.e.
dx dy
d 2v d2u
dy 2 *
—
dw
dz
= —r—++— —
du
dx
idv
;

idy
i i cl
has a definite value, i

that is, a value independent of the

d2u choice of dx and dy. Thus, the general
From this we get = and ’
,

X correspondence under which to each

o 2v d*v
.

similarly,
.
i i
in other point z x = +
iy there corresponds (or

words,
is assigned) another point w u (x, y) = +
iv y) cannot be used to determine
(.
,

d2u d2u d2 v d 2v dw/dz.

.

^ .

^
dx 2 '
dy 2 ’
dx 2 ‘
dy 2 we write a formula that expresses
If
z
(15.2.7) w as a function of z, such as z
2
e or , ,

3 z
z e \ the real and imaginary parts of w,
Any
function cp cp (x, y) of two = that is, the functions u (x and
y)
treal-valued) variables x and y that
,

v (
x ,
interconnected
y), prove to be
satisfies = 0 is known as via the Cauchy-Riemann equations
harmonic (see footnote 9.5). Thus, if (15.2.6) and each obeys the harmonicity
w = / (z)
y) y) =a u ( , + iv (x, is conditions (15.2.7). Thus, a formula that
differentiable function of the complex connects z and w restricts considerably
variable z x =
iy, the real and imag- + the choice of the functions u (x, y)
inary parts of w
(i.e. functions u ( x y) ,
and v (x, y), but in return it enables
and v («r, respectively) are harmonic finding the derivative dwldz', moreover,
y),
functions. the fact that w and z are related through
At
this point it is advisable to return an analytic function results in other
to the question of how functions are important and neat corollaries. In this
defined or classified, that is, the question case one usually says w {z) is an analytic
posed at the beginning of the chapter. function of complex variable z. All simple
z
If we are speaking of functions of a
functions, such as z 2 and e are ana- ,

complex variable, the most general lytic.

approach would be as follows. The

complex variable z consists of two One must understand that analytic func-
tions of a complex variable emerge when we
parts, the real numbers x and y or,
specify the dependence between w and z by
which is the same, it corresponds to a formula; this does not mean, however, that
a point in the plane (with coordinates all functions (that is to say, not all maps of
 =

15.2 The Derivative of a Function of a Complex Variable 481

x2 + y
2
Obviously, u
. x2 2
y and v 0, = + =
or du/dx 2x, =
du/dy 2 y, and dv/dx — =
dv/dy —
0, which implies that the Cauchy-
Riemann equations are not satisfied in this
case. In general, there is not a single function
of a complex variable involving 2 * that is ana-
lytic— not one of such functions can be written
in the form of an algebraic (or analytical)
formula. 15 6 The same is true of functions of
-

variable 2 that involve 2 or (the more so) Arg 2 1 |

.

The properties of analytic functions of a

complex variable sometimes shed light on the
properties of ordinary (i.e. real-valued) func-
tions of a real variable, properties that often
seem mysterious if one does not resort to com-
plex variables. As an example, we can take the
function y (1 =
x 2 )- 1 discussed + in Sec-
tion 6.3 (the graph of this function is shown in
(b> Figure 6.3.3). The function can be expanded
in a Maclaurin’s series (formula (6.3.8)),
Figure 15.2.2 which unexpectedly converges only in a limit-
ed range, x <1, although this function
| \

the complex z plane into the complex w possesses no singularities in the real domain.
plane) are necessarily analytic. For instance, it The reason for such behavior is that the (ana-
lvjicl .function w 2
( 1 4-, z )-
1 =
of the complex
maps each point in thecomplex 2 plane into variable 2 becomes infinite at points 2 =M, =
a point w (2) in the complex w plane that is which lie on the circle |
2 |
= 1, and it is these
symmetric about the origin (Figure 15.2.2a),
to z singularities that make
impossible to con- it

is analytic; here u —x and v y, and, = =— tinue Maclaurin’s series for this function out-
of course, side the circle 2 | <
1. Of similar origin is
=
|

the singularity at point x 0 of the function

d2 u d2u d 2v . d2v
y = 2
e-1/* discussed in Section 15.1 (the graph
r
(

~
dx 2 dy 2 dx 2 dy 2 of this function is shown in Figure 15.1.1).

—=—=—=
'

d 2v
For the function e~l/z 2 of the complex variable
d2 u d2 u
f
(since
. .

in this
, .

case 2 the point 2=0

is, as mathematicians usu-
ally say, an essential singular point at such ;

d 2v a point the function cannot have any value

dy 2
~0 \

)
’
at all. Indeed, when we send 2 to zero along

du_ the real axis, that is, 2 x, with the (real) =

du .
_ dv dv
number x being small in absolute value, we
dx dy '
dy dx find that z~ 2 — 2 = —
x~ is a very large (in
absolute value) negative number and e i/z
- 2
=
On the other hand, if we take the function e -l/x
2
i s extremely small. But if we send 2
w = f (z) = z*, which maps point z of the to zero along the imaginary axis, that is,
complex 2 plane into point w 2 symmetric ( ) 2 =iy, where y is very small in absolute value,
to zabout the real axis (Figure 15.2.26), we see then 2
~2 — (iy)~ 2= — y~ 2 is a very large =
that this is not an analytic function. Indeed, positive number and, hence, e~l/z
2
el/ 2/ 2 00 =
here u x and v =
y, and although, of = — as y -* 0. If we send 2 to zero in another
course, direction, the function e~i/z can
2
tend to
d2u d2u d2v d2v any arbitrary value. This means that our func-
,

__ ~
^ tion can in no way be considered analytic at
,

dx 2 dy 2 dx 2 dy 2
point 2=0
(even no value can be assigned to
we have it at 2 =
0 ), and it is also clear that in the
neighborhood of this point the function can-
du dv du dv
1 = = —1, . i
but -t—
.

~° not be expanded in a power series (however, in

dx dy ’ dy dx the neighborhood of any other point the func-
tion can be expanded in a power series).
(for a function to be nonanalytic, at least one
of the Cauchy-Riemann equations must not
be valid). Similarly, the function w 2 2 = = Note that, say, the function w = / (z)
<
|

=
|

zz* is nonanalytic, too. (But is there an alge- such that w 0 for x 0 and w = 1
braic formula that can express the dependence
of 2 2 on 2 without involving 2 *? Of course 15 6
is not true of such ar-
This, of course,
=
-
| |

not!) The same is true of the function w tificially functions as w

constructed 2 **, =
/ (z)
— 2 = V zz*. Examples of such func- or w (2 =
*)* 2 =
where the asterisk has a
tions abound. Take the function w =
|

2
|

= | |
2
purely formal function.
,

31-0946
 x

482 IS Functions the Physicist Needs

for x >
0 is not analytic, and the reason not necessarily mean that the function
for this is not so much the jump in w has derivatives of higher orders, for
at x =
0 as the fact that the quantities analytic functions of a complex vari-
x and y, that is, the real and imaginary able this is not so (other functions in
parts of z, are not of equal status, so the complex domain are of no interest
to say, in relation to function w. For to us). The fact is that a function w =
similar reasons the function w z* = = f (z) that has a first derivative w' =
x —
iy is not analytic either, since in f (z) automatically has a second deriv-
constructing z* we approach x and y ative/" (z) —
dw' (z)/dz and, in general,
from different “angles,” with the real all higher-order derivatives (see Exer-
part of z remaining the same and the cise 15.2.2). In addition, in the neigh-
imaginary part of z changing its sign. borhood of a point z z 0 where the =
On the other hand, and this is impor- function w =
/ (z) has a derivative we
tant, if a function incorporates z and have
not x or y separately or z* or z |, it is
+ --fp- z — z
|

analytic even if it cannot be expressed f ( z) =/ (

z o) ( «)

in the form of a finite combination of

elementary functions. For instance, the
function given by a series say ,

+ r^L(z-z,y+ .... (15.2.10)

“’<’> = 2n w- <
15 2 8 '
- -

This implies that, knowing the function

in an arbitrary small neighborhood of
is analytic, and so is the function given
point z 0 (this is sufficient for finding all
by a definite integral say ,
the derivatives of the function at point
z
z =
z 0 that is, all the expansion coef-
=
,

w(z) e~v 2 dp (15.2.9) ficients in (15.2.10)), we can continue

J
0 the function w =
f (z) outside the neigh-
borhood by defining it via series
(on the meaning of (15.2.9), where z
(15.2.10). Moreover, finding the value
and p are complex numbers, is touched
on below), and the function that is a so-
of the function w f (z) =
at point

lution of a differential equation, say

z — zx in this manner (and in neigh-
boring points), we can write an expan-
=. z 3 —w 3 where w=0 at z = 0, sion at point z =
z 1 similar to (15.2.10)
,
dz and, with the help of this expansion,
find the values of w at new points. This
although in all three cases w cannot be
process can be continued indefinitely.
expressed in terms of z via elementary
functions. But the example of the
The very possibility of expanding ana-
lytic functions of a complex variable
Cauchy-Riemann equations (or the con-
in a Taylor’s series (15.2.10) justifies
ditions of harmonicity for functions
the use of the term “analytic” for
u ( y) and u (x, y)) shows that we
functions of both real and complex
,

can make certain statements concerning

variables. It also suggests that the idea
an analytic function even if we only
of a “reasonable,” or analytic, function
know that it is analytic and know noth-
of a complex variable stands rather far
ing more about its concrete form.
from the general idea of an “arbitrary”
More than that, the general theory
function, the definition of which re-
of functions of a complex variable
quires specifying all its values, that is,
states that every analytic function can
specifying the entire range of the inde-
be expressed in the form of a series or
pendent variable.
integral While for an arbitrary (real-
.

valued) function of a real variable the The properties of analytic functions of a

presence of the first derivative does complex variable are widely used in theoretical
15.2 The Derivative of a Function of a Complex Variable 483

physics. As yet there is no complete theory of feature characteristic of Euler was that he
elementary particles. But if we assume that paid special attention to analytic functions of
such a theory can be constructed, that there a real variable, which he called “continuous” 35 7
-
.

exist formulas (unknown to us) expressing Euler was also inclined to believe that these
the main relationships of this theory, we can were the only functions worthy of the atten-
arrive at meaningful assumptions about the tion of mathematicians. Euler’s work abounds
theory. It is natural to suppose that the for- in results that today belong to the theory of
mulas expressing the functional relationships analytic functions of a complex variable, in-
are analytic. Then, even if we do not know cluding many examples of the expansion of
their exact form, we can make some state- such functions into series and of the application
ments about the properties of elementary par- of these series. However, Euler did not ad-
ticles. For instance, the mass of a particle and vance very far in studying the properties of
that of the corresponding antiparticle must analytic functions, both real-valued and com-
coincide. As for the properties of unstable par- plex-valued. For instance, he thought that
ticles, we can say that the lifetime of such a knowing the values of a real-valued analytic
particle and that of the corresponding anti- function y =
/ (x) on an arbitrarily small
particle coincide, too, although the decay pro- interval over which the independent variable x
ducts may be quite different. There is no way varies permits reconstructing the function on
of explaining how scientists arrive at these the entire real x axis, which is, of course, not
far-reaching conclusions. However, it must the case.
be stressed that the assumption about the D’Alembert made a fundamental step for-
analyticity of the formulas that have still to ward in the theory of functions when he clearly
be discovered plays the main role here. stipulated that both the independent variable
The first attempts to apply the operations and the value of the function can be both real-
of differentiation and integration to complex valued and complex-valued. On this point
quantities were made by Gottfried Leibniz Euler’s investigations followed d’Alembert’s.
and Johann Bernoulli. For instance, Bernoulli The analytic function (that is, one fixed by a
a X
found the simple formula \ 02 f 2 -^-arctan
- = formula) w =
f (z), where z x = +
iy, was
J b + x b b written by d’Alembert as u (x, y) +
iv (x, y).
For instance, by differentiating z k with k
in 1712 by using manipulations connected ,

with differentiation and integration of com- the constant and z the variable (both complex-
plex-valued expressions, manipulations which, valued), he found the correct formula (in the
although absolutely correct, had not yet been form of u +
iv) for raising a complex number
sufficiently substantiated. In the same work to a complex power. He was also the first to
Bernoulli obtained, in the process of his transfor- arrive at equations that are now known as the
n a ~7* N n Cauchy-Riemann equations, (15.2.6), in his
mations, the beautiful formula /- ft — Essay on a New Theory of Resistance of Flu-
V tan a+i J )

ids (1752), where he considered the motion

tan na i
a nj e( j to t jien s tiH un- of a body through a homogeneous, weightless,
tan na +
i ideal fluid. (The relationship between this
known de Moivre formula (14.1.6). However, problem of mechanics and the theory of ana-
neither Leibniz nor Bernoulli had fully under- lytic functions of a complex variable will be dis-
stood the situation, as is testified by the de- cussed below in Section 17.3.) In 1761 d’Alem-
bate between the two on the meaning of the bert derived Eqs. (15.2.7) from the Cauchy-
logarithms of negative numbers, a debate con- Riemann equations (15.2.6). Joseph Louis
nected with the subject of complex-valued func- Lagrange did much to advance all this work
tions and their analysis. Leibniz regarded these on the mechanics of liquids and on the theory
logarithms as imaginary, while Bernoulli saw of analytic functions of a complex variable.
them as real quantities. The publication in Continuing the work of d’Alembert in hyd-
1745 of the correspondence between Leibniz romechanics, Euler, in 1755, also obtained
and Bernoulli immediately attracted the at- (independently of d’Alembert) Eqs. (15.2.6),
tention of the most prominent mathematicians to which he attached great importance and
of the period, Leonhard Euler and Jean d’Alem- from which he obtained a number of corolla-
bert, with Euler taking the side of Leibniz and ries (including results of a geometric nature
d’Alembert supporting Bernoulli. Euler’s ar- connected with what are now called conformal
ticle entitled “De la contro verse entre Mrs.
'
transformations 1 0 8 ) In 1776/1777 Euler (who
•
.

[Messers] Leibnitz et Bernoulli sur les loga-

rithmes negatifs et imaginaires” (On the Dis-
pute Between Leibniz and Bernoulli Concer- 15
7
At present this term has quite a diffe-
•

ning Logarithms of Negative and Imaginary rentmeaning (notice the way in which Rie-
Numbers), published in 1749, contained the mann uses it in his doctoral dissertation, from
modern theory of logarithms of complex num- which we quote later on).
15
bers set forth in Section 14.3, and Euler em- 8
A
conformal transformation (of a plane
•

phasized the fact that the logarithmic func- or space)is one that preserves angles be-
tion is many-valued, something which Leibniz tween curves and that, in the neighborhood of
and Bernoulli had not even suspected. Another a point, possesses the properties of a simila-

31 *
 484 15 Functions the Physicist Needs

was then about 70) launched on a pioneering today call “the geometric theory of analytic
study (and also application) of the functions.” (The dissertation produced a whole
integrals of functions of a complex vari- course of lectures that Riemann delivered in
able (see Section 17.2). However, it was not Gottingen in the winter of 1855/1856 and the
until 1811 that Carl Friedrich Gauss in a letter ,
summer of 1856, in which he presented an amaz-
to well-known German mathematician
the ing wealth of material and many profound
and astronomer Friedrich Wilhelm Bessel ideas. 15 -
9
)

(1784-1846), gave the first clear-cut definition An

altogether different line in the theory
of the integral of an analytic function of a of functions of a complex variable was begun
complex variable and an idea of the basic pro- by Karl Theodor Wilhelm Weierstrass (1815-
perties of such integrals. Incidentally, this 1897) of Berlin, a man who was indisputably
letter was published only in 1880, when all one of the most distinguished mathematicians
the results set forth in it were already known of the 19th century. Unlike Riemann’s rea-
from investigations by Augustin Cauchy, who soning, the Weierstrassian theory of functions
had worked on this theme from 1813 onwards. could quite aptly be called “algebraic.” Weier-
In some textbooks on the theory of func- strass, to whom geometric concepts were
tions of a complex variable Eqs. (15.2.6) are completely alien and whose thinking followed
called “d’Alembert-Euler equations.” It seems a strictly formal direction, harshly criticized
to us, however, that the traditional name, Riemann (for whom, at the same time, he had
“Cauchy-Riemann equations,” is more justi- profound respect and whose investigations he
fied. The fact is that neither d’Alembert nor held in the highest esteem) for drawing on
Euler created the theory of functions of a geometric pictorialness and for a somewhat
complex variable. For instance, their works casual attitude toward logic; the modern (very
contain no full definitions of the notions of strict) criteria of “mathematical rigor” owe
the derivative and the integral of an analytic their origin largely to Weierstrass. In the theo-
function; neither did they establish the rela- ry of functions Weierstrass’s main tools were
tionship between Eqs. (15.2.6) and the very power series and operations performed on such
fact of the existence of the derivative of a func- series; he greatly advanced the “formal al-
tion of a complex variable, gebraic” theory of analytic functions of a com-
plex variable. For one thing, following an alto-
dw w(z-{-Az) —w (z) gether different approach, he obtained a large
"=
2™ Al ’
number of results that Riemann had first stat-
ed but had not backed up with flawless proofs.
a derivative that is independent of the way The rivalry between Riemann and Weier-
in which A z tends to zero. It was Cauchy who, strass can be compared with the relations be-
in a whole series of lengthy papers (the most tween Leibniz and Newton. Here Weierstrass
important of these was a paper on “complex followed Leibniz, and Riemann followed New-
integration” submitted for publication in 1825), ton.
gave a consistent theory of (differentiable, that Riemann was a most profound thinker
is, analytic) functions of a complex variable. in the realm of physics and to a certain extent
Riemann’s brilliant doctoral dissertation of anticipated the general idea of Einstein’s theo-
1851 entitled “Grundlagen fur eine allgemeine ry of gravity and created a consistent mathe-
Theorie der Functionen einer veranderlichen matical formalism without which the general
complexen Gr'jsse” (Fundamentals of the Ge- theory of relativity would have been impos-
neral Theory of the Functions of a Complex sible. The mutual respect in which these two
Variable) played what was perhaps a greater scholars held each other, although their think-
role in advancing the theory of analytic func- ing was completely different (they did not
tions of a complex variable. This was the even understand each other very well!) shows
firstconsistent and full exposition of the new that differences in scientific ideology do not
theory and it laid the foundation for what we necessarily lead to personal enmity. The exam-
ple of how Riemann and Weierstrass, proceed-
rity transformation. It is easy to see that the ing “from different angles,” built the theory
transformation, or map, 2 w ( 2 ), of the of analytic functions of a complex variable,
plane of the complex variable 2 on itself spe- serves as another reminder of the fruitful-
cified by the analytic function w f ( 2 ) with = ness of different approaches in the process of
derivative /' ( 2 ) is a conformal transformation: scientific investigation.
from the fact that w w0 —
f' ( 2 0 ) (2 ~
2 0 ), — We conclude this chapter with an extensive
where w 0 =
f ( 2 0 ), in the neighborhood of a quotation from the above-mentioned disser-
point z =
z 0 and from the rules of operation
on complex numbers (see Chapter 14) it follows 15 9 This course of lectures, which played
that near z 0 the transformation consists of a
a great role in the history of mathematics,
translation of the complex 2 plane to the
was attended by only three persons. (By the
point w0 (by the vector z0 w 0 ), a p-fold dila- way, on p. 210 we mentioned a most impor-
tion, where p = /' (20 ) |, and a rotation tant lecture course by Johann Bernoulli which
=
|

through the angle cp, with cp Arg/' (z 0 ). was attended by only one student.)
 — , ,

15.2 The Derivative of a Function of a Complex Variable 485

tation by Riemann. 15 10 It embraces the main

*
mathematical operations, it is found that the
points of everything we have said above and value of the derivative dw/dz is independent
expressively demonstrates how it all began: of the value of the differential dz. lb n Thus, -

it becomes clear that not every dependence of

Let 2 denote a variable that can change the complex variable w on the complex vari-
without limit and assumes all possible able 2 can be expressed in the above manner.
real values. If to each of its values there This characteristic property of all func-
corresponds only one value of another tions that can be defined in terms of elementary
variable, w, then w will be called a func- mathematical operations will be put at the
tion of 2 . If, furthermore, while variable 2 basis of our further investigation..., namely
changes continuously and runs through all we will proceed from the following definition...:
the values contained between two con- A complex variable w is said to be a func-
stant limiting points, the magnitude of w tion of another complex variable 2 if both
also changes continuously, then within the vary in such a manner that the value of the
aforesaid this function is called
interval derivative dw/dz does not depend on the value
continuous. clear that this defini-
It is of dz.
tion in no way establishes a connection ...If we represent the ratio (du + dvi)/(dx~r
between separate values of the function: dyi) in the form
if the behavior of the function is indicated
only in a certain definite interval, beyond
this interval it can he continued quite
arbitrarily.
dx-\-dyi
The relationship between variable w and
variable2 can be fixed by a mathematical for-
it becomes clear that the ratio has one and
mula in such a way that the value of w cor-
only one value for all values of dx and dy
responding to each value of 2 is obtained
provided that
through certain mathematical operations con-
ducted on the numerical value of 2 .... du dv , dv du
Suppose that the range of variable 2 is Z '
and
~dx ~~dy ~fa ~~dy~'
not restricted to real values but spr eads to
complex numbers x yi, with i + 1, as=V
well. Let x +
yi and (x dx) (y + dy) i+ + Thus, these conditions are necessary and
be two infinitely close values of 2 to which w
sufficient for —
u -r vi to be a function of
there correspond two values of variable w z x= +
yi. This leads us to the following
namely u vi +
and ( u du) (v + +
dv) i. + conditions imposed on the separate terms of
In this case, if the dependence of variable w the function:
on variable 2 is assumed to be quite arbitrary,
d2u ° 2u _a d2v d 2v _
the ratio (du +
dvi)/(dx +
dyi), generally speak- 1

~ dx 2 dy 2
~

ing, changes with dx and dy. If we assume dx 2 ‘

dy 2 ’ 1

that dx +
dyi =
we clearly see that
which serve as a basis for studying the prop-
erties of each term of such functions when
it is considered separately from the other term.

Exercises

15.2.1. of the Cauchy-

Check the validity
Riemann equations for the functions of a com-
plex variable that were encountered in Exer-
cise 14.1.1 and for the function w In 2 = .

15.2.2. Suppose that w / (2 ) u ( x y) = = ,

+ iv (x, y) is an analytic function of the com-

=
plex variable 2 and f (z) dw/dz u x (x, y) =
+ iv 1 (x, y) is its derivative. Prove that
the Cauchy-Riemann equations for / ( 2 ) auto-
But each time the function w of variable 2 matically imply the validity of the Cauchy-
is defined by a formula containing simple Riemann equations for f'(z).

ib 11
15 - 10
G. F. B. Riemann, Gesammelte mathe- This statement is obviously valid
.

matische Werke 2nd ed.. Dover

,
(reprint), in all caseswhere dw/dx can be obtained from
New York, 1953, pp. 3-43. the expression of w in terms of 2 via the ordi-
nary rules of differentiation.,.. (Riemann’s
footnote.)
 Chapter 16 Dirac’s Remarkable Delta Function

16.1 Various Ways of Defining closely related to the applications of

a Function the exponential function and of tri-
gonometric functions to many problems
The functions we have been studying of physics, say, to problems involving
up to now have ordinarily been defined oscillations, than are the definitions
by formulas. This means that a pro-
cedure was always indicated for com-
puting the values of the function for [“=(‘+t)T
any given value of the independent n
variable. We could call this an algo-
e
x — lim [i
'nil ) ) ,
n-*-oo \

rithmic representation (algorithm mean-

ing a procedure for computing some- and of the sine and cosine as certain
thing). To illustrate, take the function segments in a circle or as ratios of sides
y = / (x) 2x =
3x 2 It +
actually .
of a right triangle.
.amounts to this: “take x, multiply by 2, It is curious to note that the defini-
x
square x, multiply by 3, and then add tions of e sin#, and cos# via differ-
,

the two numbers to get the value of y ential equations prove to be convenient
for the given value of x.” Trigonometric for electronic computers. If in a cal-
functions were defined differently, by culation the investigator has to sub-
x
means of geometric concepts, by mea- stitute into a formula the values of e
suring arcs and line segments in a circle. for different values of x, he copies
However, here too we can speak of an them out of a table. When operating
algorithm for calculating their values, a computer, it is more convenient and
which amounts to turning to trigono- faster to have the machine compute e x
metric tables or to a pocket calculator step by step via the approximate for-
*
with the appropriate keys. mula /x/ Ax) in accor-
e (l _|-

Up to now we have studied the prop- dance with the equation dyldx y for =
erties functions specified in this
of the function ex (or via more exact
fashion, the laws of increase and de- formulas based on the same equation)
crease of functions, the laws for finding than it is to refer to a table of values.
the maxima and minima of functions, The same goes for the functions sin x
etc. The study of these properties and cos x. It is easier to compute them
leads to new ways of defining functions. every time simultaneously: 161
For instance, the function / == ex may sin (x + Ax) ~ sin x — Ax cos x,
be defined as a function whose deriva-
tive is equal to the function itself:
cos (x + Ax) ~ cos x — Ax sin x.

dfldx =
/, with the supplementary con- Thus, one general approach to the
dition that / (0) 1. The sine, the = concept of a function lies in specifying
function cp =
sin x, and the cosine, or a procedure for computing it and in
—cos x may be defined as functions
, subsequently investigating it. There
that satisfy one and the same equation is also another approach. We can seek
a function with definite general prop-
d 2 cp
dx 2
~ <P>
dx 2
~$ erties, the aim being later to attempt
(on the basis of these properties) to
under different initial conditions
find the formula that describes the
<p (0) = 0, dff
dx
function at hand. Such is the usual
o
procedure when handling experimental
d^>
*( 0) =1 ,
dx
=0 .
16,1 Where do you think these formulas
o
come from? Check to see whether they accord
These definitions are in many respects with the equation for the second derivative
more to the point, so to say, and more of the sine and cosine.
16.2 Dirac and His Function 487

data and finding empirical formulas by

trial. Our object here is to construct
in this way a remarkable function that
is useful and important both in mathe-
matics and in its applications.

16.2 Dirac and His Function

Paul Adrien Maurice Dirac (1902-1983),

the celebrated English theoretical phys-
icist, came to fame in 1929. He had
elaborated a theory capable of de-
scribing the motions of electrons in
electric and magnetic fields with arbit-
rary velocities almost up to the velocity
of light. This was the quantum theory
which also accounts for the fact that
the electrons in an atom move only in Paul Dirac
specific orbits with definite energy val-
ues. Dirac knew that an electron po-
foundation and leads to fresh pre-
ssesses a definite rotational moment, or
dictions.
is similar to a spinning top, and he
Dirac was not only one of the best
took this into account in building his
theoretical physicists in the world, he
theory. When the theory was con-
was a marvelous mathematician. In his
structed, it turned out that a conclusion
classical Principles of Quantum Me-
could be drawn that Dirac had not
chanics Dirac introduced and made wide
foreseen, namely, the existence of par-
use of a new function, which he denoted
ticles with mass the same as the electron
by 8 (x). It is known as Dirac's delta
mass but with opposite (positive) function or, simply, the delta function.
charge. For two years it was thought
The delta function can be defined
that Dirac’s theory was good for de- —
as follows: 8 (x) 0 for any x 0,
scribing electron motion but that the
conclusion concerning particles with
that is, for x <
0 and x 0, and>
8 (0) =oo. In addition, the delta func-
positive electron charge was erroneous
tion must satisfy the following con-
and that as soon as he got rid of it the dition:
theory would be a very good one.
But in 1932 a positively charged + oo
particle with the electron mass, called b{x)dx = \. (16.2.1)
the positron (also called the antiparticle j
— 00
of the electron), was discovered. The
big drawback of Dirac’s theory became Figure 16.2.1 gives a pictorial view of
its triumph, its principal contribution: a function similar to the delta function.
Dirac’s discovery was the first instance The narrower we make the strip between
of a new particle being discovered “at the left and right branches, the higher
the tip of a pencil.” This is an instruc- must the strip (the integral, that is)
tive example from the standpoint of be to retain the given value of 1. As
the relationships of theory and experi- the strip becomes ever narrower, we
ment. Theory rests on the findings of approach the condition 8 (x) 0 for=
experiment, but a consistent, logical x =7^= 0, and the function approaches
and mathematical, development of a the delta function. In one of the follow-
theory takes the investigator beyond ing sections these arguments will be
the confines of the material used as its utilized in the construction of formulas
 ,

488 16 Dirac's Remarkable Delta Function

(tall and narrow “steps”) which completely

define the function. 16 2 .

Dirac and other physicists have for many

years used the delta function freely and pro-
ductively, without any strict definitions exist-
ing for such strange objects, just like mathe-
maticians for centuries used irrational num-
bers until, at the end of the 19th century, the
first theories of the real variable appeared.
Mathematical justification for such remark-
able entities as the delta function (“genera-
lized functions”, or distributions, as they are
Figure 16.2.1 usually called) was given by the Soviet ma-
thematicians S. L. Sobolev and I. M. Gel’fand,
the French mathematician Laurent Schwartz,
that yield the delta function. Here we and others; this topic is discussed in many
continue the study of its general prop- books. 16 3 -

erties.
Possibly, the following important for-
It must be stressed, of course, that
mula gives a better description of the
the “equations” defining the delta func- Dirac delta function than (16.2.2):
tion,
+ °o
0 if f(x)6(x)dx = f( 0). (16.2.3)
j
8 — oo
oo if £= 0,
(16.2.2)
+ 0O Indeed, since 8 (x) 0 for x 0, we = ^
\
8 (x) dx = 1 conclude that the value of the integral
on the left-hand side of (16.2.3) cannot
depend on the values of / {x) no matter
must not be taken too literally, since what x =^= 0. The only essential value
from the standpoint of classical mathe- of / (x) is the one where 8 {x) =7^ 0 that ,

matics (that is, the mathematics we is, for x =

0. This means that in the
are accustomed to) these conditions narrow region where 8 (x) =t^= 0 (in the
are meaningless and contradictory. limit the width of this region tends to
Obviously, the above-described process zero; see Figure 16.2.1) 8 (x) is multi-
of stretching (and contraction) of the plied by / (0). Hence, formula (16.2.3)
strips clarifies the meaning of the delta follows from the conditions (16.2.2).
function, the “intuitive” definition
(16.2.2) of the delta function can only 16 2-
For those interested, here are the “de-
stimulate our imagination in relation cimal” approximations of the delta function:
to this process. y = 0 for x 1, y >
0.5 for x =
1; <
y = 0 for
|

x 3> 0.1. y
|

5 for x —
0.1;
| |

<
y = 0 for
|

x p> 0.01, y
\
|

\
50 for x — | |

| |
<
It is clear that by defining the delta func- 0.01; ... (it is more convenient to “smooth
tion we introduce a completely new object, out” these steps, so that all functions are made
so unlike any other object discussed earlier continuous); below we will give other (and
(there is no way in which we can say that our still more convenient) definitions of the delta

new function is analytic in the sense of Chap- function.

16 3
ter 15). However, the description of this new
-
Possibly, the simplest of these are two
object is fairly simple and must not put us small books by J. Mikusinski and R. Sikorski,
The Elementary Theory of Distributions, pub-
off. For example, the number Y
constantly used in mathematics, is not a “real”
2, which is
lished in English in Warsaw in 1957 and 1961.
For instance, in the first book the authors in-
number from the standpoint of arithmetic:
troduce generalized functions as limits of se-
it is not given by its exact value but by the
quences consisting of ordinary functions, say,
sequence 1, 1.4, 1.41, 1.414, 1.4142, con- . . .

the functions (p^ (x), where cp n (x) 0 for =

sisting of approximate values of ]/" 2 (which, \x\> l/i0 n ,y n (x) 10 n /2 for x = 1/10”; | |
<
incidentally, characterizes Y2 sufficiently in
order to freely use it). Similarly, 8 (x) is given
see footnote 16.2. (It goes without saying that
the concept of the limit of a function requires
not by its exact values but by approximations a rigorous definition.)
 , — —

16.3 Discontinuous Functions and Their Derivatives 489*

We can also argue in reverse. We can smaller than the area under 8 (x).~

say that 8 (x) is a function such that Therefore,

no matter what form the auxiliary
function / (x) has, we always have for- / (x) b8 (x) dx= bf (0), (16.2.4).
j
mula (16.2.3). This condition alone
brings us to all the conclusions con- / (x) 6 (cx) dx = -jij-/ (0). (16.2.5>
j
cerning the form of 8 (x) that were
earlier employed in the definition of A comparison of (16.2.4) and (16.2.5))
8 (, x ). From formula (16.2.3) it follows enables us to say that
that 8 = 0 at x =#= 0,
(x) that
j
8 (x) dx= 1
S(cx)==-y!r5(x).
and that 8 (0) — oo. 16 * 4

Note that, of course, to grasp the Formula (16.2.4) is quite obvious, but.
meaning of formula (16.2.3) requires, we can also hope that the reader who
like the definition (16.2.2) of the delta has gone through the trials and trib-
function, a certain effort of thought, ulations of the preceding chapters oF
since the integrand on the left-hand this book will be able to grasp formula
side of (16.2.3) contains a thing that (16.2.5) as well. Formally, it is readily
can only loosely be called a function. obtained by the change of variable*
However, a “rough” treatment of the
delta function makes it possible to u—\c\x, 1
dx = -^-r du.
\c\
justify both the doubtful definition
(16.2.2) and formula (16.2.3), that is Here we also make use of the fact that,
to say, both formulas are sufficient for the function 8 (x) defined by formula
applying this new concept and must (16.2.2) is an even function of its argu-
always be taken into account. ment: 8 ( x) 8 (x). =
Let us carry through a few more
obvious consequences of the definition Exercises
of 8 (x). By the general rule of a change
16.2.1. Show that for a function q> (x)<
of variables (discussed in detail in having a unique zero x 0l so that q) (;r 0 ) 0* =
Section 1.7), the function 8 (x a) — and cp (x) =£ 0 at x =£ x 0 we have the formu- ,

is displaced a units to the right in

la 8 =- 1
8 (x —x What
relation to 8 (x), that is, 8 (x a) oo — = (q> (*))
|
q>
y
(x 0 ) |
0 ). is.

at x =
a and is zero otherwise. Accord- the function 8 (\|? (x)) if the function
vanishes at several values of x ?
(x)i

ingly,
16.2.2. Evaluate the integral
+ oo
+oo (a:) 8 (sin x) dx.

^
/ (x) 8 (x — a) dx = f (a). (16.2.3a)
j
— oo
— oo

16.3 Discontinuous Functions

and Their Derivatives
Now easy to see, if we consider
it is
a curve of the form depicted in Fig- Let us consider the integral of 8 (x) as'
ure 16.2.1, that 88 (x) is b times higher a function of the upper limit, that is r
than 8 (x) and that 8 ( cx ) is c times | |
X
narrower than 8 (x), so that the area 0 (x) = b(z)dz. (16.3.1)*
under the curve 8 (cx) is c times | J
j
— oo
It is easy to see that the graph of this*
function has the form of a step (Fig-
16 4
Here the integral sign without any
•

ure 16.3.1). Indeed, as long as x <C 0 r

limits of integration will always be under-
stood as being over the entire range of the the domain of integration in (16.3.1)’
variable of integration, from oo to —
oo. + is wholly located where 8 (x) 0V =
 ,

490 16 Dirac's Remarkable Delta Function

J x
“/ 0 x
1 Z
/ -r\ '
Figure 16.3.1
Figure 16.3.2

which means that 0 (x) = 0 when £<0. If we apply it to the expression (16.3.1),
But if x >
0, the domain of integra- we obtain
tion involves the neighborhood of the
origin, where 8 (0) oo. On the other = -^• = 8 (x). (16.3.2)
hand, since 8 (x) 0 when x =
0, the > Thus, we do not need to make an excep-
value of the integral does not change
tion for the derivative of a discontinuous
when the upper limit changes from 0.1
(or even 0.000001) to 1 or to 10 or to oo.
function. We
merely say that at the
Hence, for any x > 0 we have point of discontinuity the derivative
is equal to a “singular” function, the

X +oo delta function.

= =1 Wehave learned to handle the deriv-

*0 (x) = j
8 (z) dz
j
8 ( z ) dz ative of an elementary discontinuous
— oo — oo
function and can now very simply find
derivatives in more complicated situ-
is shown in Figure
ns 16.3.1. ations. Here are some examples. Let
Thus, with the aid of the delta func-
tion we have constructed the simplest y = x for x < 1 and
•
discontinuous function 0 {x) such that y = x — 2 for x > 1. (16.3.3)
when x 0 <
then 0 (^) 0, and =
We
-0 {x) =1 in the domain x 0.
16 5
These > -
refer to the graph of this function
in Figure 16.3.2. The jump occurs at
•simple considerations enable us to ap-
proach the problem of the derivative x =
1, and the magnitude of the jump

•of a function having discontinuities

is y (10) +
y (1 —0) —
—2. (Here =
in a more consistent fashion, without we use the notation y (1 0) to denote +
the limiting value of y as x approaches
apparent exceptions and extensive re-
1 from the right that is, from the
servations. ,
.

we did not know about the delta

If
direction of x >
1, and y (1 0) de- —
function, we would have to say that notes the same on the left; see Fig-
derivatives cannot be found at points ure 16.3.2.) From this we formally get
where a function is discontinuous. But
we have just constructed a discontinu- -g- = l-25(z — 1). (16.3.4)
ous function, 0 (x). The general rule
of the relationship between an integral This notation is better than the dreary
and the derivative (see Section 3.3) is: statement that dyldx 1 everywhere =
X except at point x —
1, where the func-
if F (x) = y (z) dz, then y ( x) = .
tion has a discontinuity and does not
j have a derivative.
X0
The delta function
is a typical brain-
child twentieth century. The
of the
16 5
At point x
•
=
0 the function 0 (x) nineteenth century had a passion for
undergoes a jump, and there is no need to
-define the value of 0 (0) any further; suffice
investing all its arguments true, not —
it to say that at this point the value of 0 (x)
quite true, and simply false— in the
-changes from 0 to 1. form of “impossibilities.” It is impos-
 ) , ) :

J6.3 Discontinuous Functions and Their Derivatives 491

sible invent a perpetual motion

to
machine, transformations of chemical
•elements are impossible, it is impos-
sible to change the total mass of a sub-
stance, it is impossible to prove the
existence of atoms, it is impossible
to determine the composition of the
stars, it is impossible to find the deriv-
ative of a discontinuous function. The
twentieth century has found numerous
constructive solutions to what appeared
to be impossible in the nineteenth cent-
ury, say, to the problem of element trans-
formations or to the reduction of the to-
tal mass of a substance by transforming
mass into energy. To take an example,
the delta function resolves the problem
of the derivative at a point of discon-
tinuity (at any rate for a discontinuity
in the form of a finite jump). Indeed, full justification, write y' = —2x for
the notation of (16.3.4) contains in one x < 1 and y' = 2x + 46 (x — 1) for
line the fact of discontinuity of the (Note how the signs (less <
function y (x) (since dyldx involves the than) and ^
(less than or equal to), >
delta function), the nature of the dis- (greater than), and (greater than or ^
continuity (a jump), the site of the equal to) are placed in the formulas.
discontinuity {x 1), =and the mag- Be careful not to write the delta func-
nitude of the jump (the coefficient —2 tion twice. 16 6 ) We can also write
-

of the delta function).

Integrating (16.3.4) with the con- y' = V(x) + 46 [x — 1),
dition x 0, y = =
0, we can restore — 2x if x< 1
the graph of y (x) in its entirety. True, where cp x —
x>
(

in the case of the function (16.3.3) we

2x if 1.

.had it easy in the sense that a simple

case was especially chosen where the To verify the notation, integrate the
derivatives on the left and on the expression of the derivative. You will
right are expressed by a single formula. obtain the original discontinuous func-
This of course is not obligatory. Why tion.
.should the derivative be continuous if Sometimes mathematics use is
in
the function itself suffers a discontinu- made of the so-calledsignum function
a T J
" sgu"x ~ or csgir’jl! ^
L
1 1 Ts a§nnda Ynus
* J
‘
ity?
Let us consider a more complicated sgn x = —1 for x 0 and sgn x = +1
<
example: y = —
x 2 for x 1, and y =
<
x 2 for x
.
>
1 (Figure 16.3.3). For x
—
1 < 16 6
Thus, the derivative of a discontinu-
we have y' = > •

2x while for x
, 1 ous function / (x) at the point of discontinuity
we have y' = -\-2x. The discontinuity x = x0 has a definite “value” 6(x x 0 ). —
is associated with y' = 46 (x —
1). We As for the discontinuous function / ( x itself,
can now, at our pleasure, adjoin the there is no sense in asking for its value at the

point x = 1 to the left-hand region

actual point of discontinuity. At any rate,
this question is meaningless in nearly all
and then write y' = —2x + 46 (x 1) — applied problems.
for x 1 and y' — 2x for x
^ >
1. Or,
16 7•
Sgn x is often read as “signum x” or
“the sign of x” since the Latin for sign is
another version (quite similar to the
signum. Sometimes it is assumed that
first), we can adjoin x = 1 to the right- =
sgn 0 0, but we will never use this physic-
Ihand region and then, with the same ally meaningless notation.
492 16 Dirac's Remarkable Delta Function:

Figure 16.3.5
Figure 16.3.4

not simply say that the second deriva-

for x > 0.can also write sgn x =
We tive is equal to zero everywhere in this
xl\ x
|, where x is the modulus
| |

case? Of course, the crux of the matter

(absolute value) of x The curve of the .
=
lies in the salient point at x
signum function, y = sgn x, is shown
0,
where the two straight lines meet. But
in Figure 16.3.4a. It is easy to see that
how are we to be sure that it is at the
sgn x = — 1 + 20 (x) salient point that the value of the
X second derivative is 28 (x)? To assure
= —1 + 2 8(x)dx. (16.3.5)
ourselves of this, let us round off the
j salient point: take the function
— oo

Now let us examine the function yi = Y x2,Jr a 2

- (16.3.8)
x itself. Its graph is depicted in
| |
The smaller the value of a, the closer
Figure 16.3.46. We find the derivatives =
this function is to the broken line y
to be d x \ldx — —1 for x and < x (Figure 16.3.5a). Clearly, at a = 0
\
1

d\ x \ldx = +1 for x 1 or, briefly, > | |

formula (16.3.8) yields y 1 = x The |.

with the aid of the new function, |

function specified by Eq. (16.3.8) is^

smooth and differentiable everywhere.
-^- = sgn.r. (16.3.6)
Employing the rules of Chapter 4, we
can easily find that
From formula (16.3.5) it then follows
that dy i __ x d2y1 _ a2
dx 2 {x 2 -\- a 2 ) 3 ! 2
dx Yx + 2 a2
’

4^ = 25(z). (16.3.7)
(16.3.9)
This formula is so important that one Figure 16.3.56 depicts the graph of
would do well to get a good feeling d 2 y 1 /dx 2 (both Figure 16.3.5a and Fig-
of it. We know that the second deriva- ure 16.3.56 are constructed for a =
tive of a function is connected with the 0.5). It is easy to see that this curve
curvature of the curve on a graph (see becomes higher and narrower as a
Section 7.9); the curvature of a straight 2
decreases; the integral ^ffiyjdx dx is-
line is zero since the second derivative )

of a linear function is zero. It would equal to 2 for all values of a (see Exer-
appear then that if the graph of x | |
cise 16.3.3); whence, in the limit when
consists of two straight lines, on each a = 0 we obtain the expression
of which d 2 x \/dx 2 0,
1
then why = d2 x 1
\ldx
2 = 28 (x) given above.
 16A Representing the Delta Function by Formulas 493

^Exercises

16.3.1. Draw the graphs of the following

(discontinuous) functions and write down the
(first) derivatives of the same functions: (a)
y — x if i 1 and y < x 1 if x^> 1; — —
(b) y =
el/x/(l e 1 /*). +
16.3.2. Find the curvature of the curve
Vi
— */i ( x )
representing function (16.3.8). How
does the curvature change when we send a
to zero?
oo

16.3.3. Prove that \ y'{ (x) dx = 2 for the

Figure 16.4.1
J
— OO
function in (16.3.8).
graphs are brought to the same height
(see below), it is hard to distinguish
16.4 Representing the Delta Function them at a glance. Incidentally, the
ly Formulas last (exponential) function is much
At closer than the others to the axis of
the end of the previous section we
inadvertently obtained an expression abscissas for large values of x far away
from the maximum. 16 9 Now recall -

y = (16.4.1) what needs to be done to increase the

2 (x 2 +a 2 3/ 2
)
’

height of the bell rc-fold and decrease

which approaches 8 (x) in the limit as a the width m-fold: take ftcp (mx). If the
tends to zero. 16 8 Let us examine in -
area under the bell is to be preserved
detail the problem of an analytic expres- choose n =
m. To summarize, then,
sion for the delta function. the function rccp ( nx ) 8 {x) as n oo

Let us take a function cp (x) of which or, to put it otherwise, 8 {x) =

we only demand that it vanish for lim racp (nx). It is also easy to formally
n-+oo
•x =± oo (that is, we require that verify that
•cp (x) 0 as x ±oo) and that the
integral I = ^cp (x) dx be nonzero. It ^
racp (nx) dx — cp (z) dz
|
is always possible to make this integral =
<equal to unity by multiplying cp by an
= j
cp (x) dx 1

appropriate constant. Suppose that via the substitution z nx. =

this has already been done, so that Thus, to the three variants of cp (x)
cp (x) dx = 1. It is clear that cp (x) correspond the following three repre-
j
sentations of the delta function:
lias a maximum somewhere between
— oo and + °°. The simplest examples 6
are functions that are everywhere posi- L 2 (l +* 2* 2 3 / 2
)
’

tive and even, that is, symmetric about

the y axis (see Section 1.7). Here are 6W= ‘™ <.«+»*> (16 4 2>
' '

some concrete instances:

6 (x) — lim — 1

~r~—
r e~ n2x2 .
r

<Pl
= 2(l+;r 2 ) 3 / 2 ’ ^ = H 1+z 2 ’
n-+ oo V n

Let us verify that the procedure that

" x2
<Ps(*) =-r^ e * was proposed earlier,
V ji

The graph of each of these functions 6 ( x) = 2 (*• + **)*/* ’

is bell-shaped (Figure 16.4.1). If these

16 9
The
intersection of all three curves cp lt
•

16Note the “2” in the denominator

• 8 of cp
2, at just about the same points in
and cp 3

(16.4.1); without it we would have 28 (x). Figure 16.4.1 is of course purely accidental..
 494 16 Dirac’s Remarkable Delta Function

fits this definition. To do this, we action,

that is to say, some kind of
write finite function t|) (x), with the aid of
the delta function is possible and desir-
a2 _ a2
2(x 2 +a 2 3/ 2
) 2a 3 (x 2 /a 2 +i) 3 / 2 able precisely when the detailed form
of the action (which is to say, the true
— 1
dependene of it on x) is inessential, the
2a(x 2 /a 2 + l ) 3/ 2
important thing being only the inte-
and set n =
1 la to get the first repre- gral.
sentation of the delta function in The examples given above do not
accord with (16.4.2). exhaust by any means the diverse
We conclude that there is no single cp (x) from which we can “manufacture’ 7

simple formula that can yield 6 (x). 8 (x). We can give up the symmetry of
Clearly, the fact that 8 (0) oo is not = cp as we pass to racp (nx) and increase
(x)\
enough. To define 8 (x), it still remains n, the distance of the maximum from
to be demonstrated that this is precisely x = 0 diminishes, that is, even an
the infinity that is needed. However, asymmetric function approaches 8 (x).
8 (x) can be obtained as the result of Here is an example: the function
passage to the limit (n —>- oo) from well- n- 1 / 2 ?-^- 1 ) 2 passes into the function
behaved (well-defined) functions of x Jt -V
2 ^-^- 1) 2
whose maximum lies at
that involve the auxiliary quantity n x = \ln.
as a parameter. We must stress partic- We
can give up the notation of cp (x)
ularly here that 8 (x) may be obtained by means of a simple unified formula
by such a limit process from different that ensures the smoothness of (p (x).
functions cp. As long as n is finite, the For instance, we can take the function
m
functions p ( nx ) differ from one another cp (x) to be discontinuous, say,
and, in particular,
1/2 for — Cx<Z 1,
1
|

/ mp (nx) dx^=f (0). 0 for x<Z — 1 and x>l.

= j
/ (a:) (16.4.3)
1

(16.4.5)
And only in the limit, as n -> oo, do
all the distinct functions rccp (nx) tend
The limit process consists in our taking
to a single limit 8 (x) and the corre- cp n (x) = nl2 for — 1 Cnx < 1,
sponding integrals 1610 (16.4.3) tend to i.e., — Hn < x < 1/rc,

/ (0):
<Pn ( x) = 0 for x <— 1 In and
lim 7-/(0).
n->oo
(16.4.4) x > 1 In (16.4.6)

The arbitrariness that is evident in the and sending n to infinity. (Sketch the
choice of the original function cp (x) graph of cp (x) according to (16.4.5) and
from which we obtain 8 (x) is in full also cp n (x) according to (16.4.6) for
accord with the essence of the matter. n = 3 and n 10.) —
In Section 17.4, we will consider exam-
Finally, we can reject the condition that
ples of the application of 8 (x) to phys- cp (x) be positive. A curious and important
ics. The description of some kind of example is

+0 )
16 - 10
Strictly speaking, if / (x) is discontin-
uous at certain values of x and tends to
infinity at these points (or / (x) -> oo as x -» oo - 0)
or as x-+ —
oo), not all cp (x) can be used to (16.4.7)
obtain (16.4.4). Furthermore, the function
/ (x) must not be discontinuous or at least its The graph of R (x) for a given co is shown in
discontinuities must not fall on the point Figure 16.4.2. The value of R (0) is equal to
x— 0 where 8 (x) =
oo since otherwise we will
,
(o/ji (the indeterminate form involved in the
have those meaningless questions about the vanishing, simultaneously, of the numerator
walue of a function at a point of discontinuity. and the denominator at x 0 is evaluated=
16.4 Representing the Delta Function by Formulas 495 >

^ R(x)
1 \

/
/
1
TA7V A
/ Figure 16.4.3
\
/
\ l

Figure 16.4.2 (see Exercise 16.4.1). The graph of P (x) for'

q =10 is shown in Figure 16.4.3. As q oo,

in elementary fashion). The function R (#)

the function P (x) behaves near x 0 just =
like R (x) does as co
passes through zero and changes sign at x — oo, but differs from
R (x) in that its high maxima repeat period-
±ji/a), ±2
ji/(d, dz3ic/o), .... The oscillations —
0, ±2ji, ±411, .... In other
of R(x) damp out as they recede from x 0 = ically at x
words, P (x) represents, as q oo, a sum
due to the denominator. The curve does not
go beyond the curves y 1/ji x shown =± | |
of delta functions:

dashed in Figure 16.4.2. It can be verified that

+ oo P (x) = 6 (x) + 6 (x — 2n + 6
) (x + 2n)
R (x) dx — 1 for all values of co. + 6 (x — 4:rc) + 6 (x + +
4jt) . . . .

— oo
It turns out that we can regard R {x) as a The functions R and P and their connection
delta function if we send oo to infinity. This with the delta function are not mathematical
is likely since as oo is increased, the altitude
oddities. Recall how R and P were construct-
ed: R is an integral of the cosine and P is
(o/n of the principal maximum on the R axis
grows, while the width of the half-wave jx/co — a sum of cosines. If from cosines it is possible

C C x jt/oo decreases. But how are we to deal

by addition (integration is a kind of addition)
with the fact that the amplitude of oscilla- to construct the delta function, then 6 (x a) —
tions does not decrease when co increases; as can be constructed from cos co (x a) — =
before, R attains ±1 /ji x and the dashed | |
cos ooz cos cDa —
sin (ox sin ooa, that is,

lines do not move toward each other? Let us from cosines and sines with constant coef-
ficients. But then any function
/ (x) can be
consider the integral f (x) R(x) dx. The great- represented as a sum of cosines and sines.
^
Any function can be replaced by a series of
er the value of oo, the more frequent the oscil-
steps / (x{) A x t , and each step is actually
lations and the more exactly the positive and —
6 (x f (x t ) Ax t Thus, with the aid of
.
negative half-waves compensate each other,
yet the contribution of the first half-wave and
R and P, that is, essentially via delta func-
tions, the possibility is proved of expanding
the ones closest to it are all the time the same.
functions in a Fourier series (if the function is
It is for this reason (we do not give the proof)
periodic; cf. Section 10.9) and into a Fourier
that lim \f (x) R (x) dx = f (0), but this means integral (if the function is nonperiodic).
Reread Sections 10.8 and 10.9 after fin-
that lim R (x) has the properties of the delta ishing with this section. Ordinarily, mathe-
(D-*-oo
matics textbooks do not mention the delta
function.
function. Many mathematicians prefer to
Of interest is a similar function:
keep such “physical heresy” away from the
k=q student as long as possible. The realization
that actually the delta function is being used
p{x)= coskx )-
"iir (t+2 fe=l
(is.4.8)
in the proofs will help you to grasp the mean-
ing of these proofs.

Using the formulas of elementary trigono-

metry, we can obtain the expression Exercise

x
16.4.1. Prove that the right-hand sides-
of (16.4.8) and of (16.4.8a) coincide.
 Chapter 17 Applying Functions of a Complex Variable
and the Delta Function

17.1 Complex Numbers and Mechanical tude f) equation (17.1.1) takes the
Oscillations form
The third part of the book is just
tended to excite the reader’s curiosity in
in-
+ ^ = / cos CD* (17.1.2)

the functions that seem to be unlike the

ordinary functions, such that the func-
(see (10.5.1)); here k, h, /, and co m ,

are constants. 171

tions of the complex variable and the
•delta function, and to show the pos-
We have considered all the quanti-
sibility, in principle, of extending the
ties in (17.1.1) and (17.1.2) to be real-
valued. Let us allow for the time being
techniques of higher mathematics to
(since in the final solution we have to
these unusual objects as well. This last
eliminate this assumption again) com-
chapter of the book outlines some pos-
plex-valued solutions x to our equation;
sible ways of applying functions de-
scribed in Chapters 14, 15, and 16 to
and let all the remaining quantities,
some problems of physics and engi- which are the coefficients in (17.1.2),
to be as before real-valued. It is more
neering.
convenient now to replace the expres-
We
will show how calculations can
sion / cos (x)t for the external force with
be simplified by using complex varia-
bles using an example of forced oscilla-
tions of a heavy body (see Chapter 10, fe
i(*t
( := / cos cd£ + if sin o>£) (17.1.3)
-especially Section 10.5). We
start with
the equation of such oscillations. Sup- adding to the right-hand side of (17.1.2)
pose a body of mass m
(the material the imaginary term if sin cot.
point of mass m
since we neglect the We find the solution to the equation
dimensions of the body) is acted on by
an elastic force —
kx which is propor-
m intr + h ir+ k:c = feiat (17.1.2a)
tional to deviation x (to make it more
clear imagine a spring trying to return
the body to equilibrium), a friction in the form
force —
h (dx/dt) which is proportional
to velocity dxldt (and directed in oppo- x—ae i(dt
. (17.1.4)
sition to the velocity), and an exter-
nal force F =
F {t). Here equation Such a form of the solution is suggested
(9.4.2) for the body’s motion (Newton’s by the fact that all the derivatives of
second law, see Section 9.4) assumes an exponential function are as we know
the form proportional to that function; therefore
after substituting this expression for

* mJSr=- k *~ h t+ F >
x in the left- and right-hand sides of
(17.1.2a) both its sides will contain
or the same factor e i(dt which we can sim-
ply cancel. Indeed, since by (17.1.4) we
mJlF + hJW + k * = F -

11 • 1
Since m has the dimensions of kg
Weconfine ourselves mainly to the case and x ,
dxldt, and d2 x/dt 2 the dimensions of m,
of a periodic external force F —
f cos at
m/s, and m/s 2 respectively (cos co t
is of
course dimensionless), it is clear that the
(compare this with what is said below
coefficients k, h and / in equation (17.1.2)
,
at the end of this section). For a per- have the dimensions of kg/s 2 kg/s, and kg-m/s 2 ,

iodic force (with frequency co and ampli- (i.e. N, since / is the force).
 \ ’

497
27.2 Complex' Numbers and Mechanical Oscillations

have dx/dt = ia(»e iat and d2 x/dt 2 = Let us now return to real variables.
-awV®', equation (17.1.2a) yields Since the initial equation (17.1.1) is
linear, the principle of superposition
— am a2 e iat -f iha(oe iat + ake iat = fe ia>t ,
is valid for the mechanical system which
or the equation describes, that is, the
motion caused by a composite force
a ( — mw 2 •+ A + zTito) =f ,
Fi +F 2 can be obtained by adding the
that is, motions produced separately by forces
F x and F 2 (we have frequently used
ffl = — mtifi-\-k-\- ihm t . (17.1.5) this principle before). In our case the
force (17.1.3) is composed of the real
Thus we have found that (complex) component / cos art and imaginary
amplitude a of oscillations x = ae
i(£>t
component if sin art. But all the coef-
(= a (cos co£ +
i sin art)) is proportion- ficients in except for the free
(17.1.1),
al to the applied force /, which is
quite term F (the term which does not con-
natural for a linear system (for the tain the unknown function x), are
solution of the linear equation (17.1.2a)). real; therefore, to a real force F there
When the friction force is not large must only real value x
correspond
(i.e. when h is not large), the
maximum and to a (pure) imaginary force F an
of the amplitude (the amplitude’s mod- imaginary x In other words, our linear
.

ulus, to be more exact) is attained at system generates a real response to a

a frequency co such that 77U0
2
k, that — real external force, and an imaginary
is, at a frequency of the external force response to an imaginary force. There-
close to the natural frequency co 0 = fore we can be sure that to the exter-
Y him of the system Section 10.2); (cf. nal force F — f cos art there corre-
thus, the phenomenon of resonance sponds the solution
occurs here (Section 10.5).
x = fr cos (co t + cp) (17.1.7)
Expression (17.1.5) for the complex-
valued amplitude a can be rewritten as of equation (17.1.2), where r and cp are
(17.1.5a) given by (17.1.5b), and to the force
a — fre i(p
,
F = if sin art there corresponds the so-

where lution x =
ifr sin (co£ cp) (i.e. to the+
force F f =
sin at of equation (17.1.1)
rel<t = — there corresponds the solution x =
fr sin (co£ cp), + which, by the way,
that is, supplies us with no new information,
_ 1 since this solution can be obtained

Y — ma +
{
)
2 /c )
2
+ /j
2 co 2 ’ from (17.1.7) by substituting t n/2co +
1 for t).

<
P
= arctan ( mX-k ) •
The same result can be obtained some-
what differently. By Euler’s formu-
Then, las (14.3.4) the real force F f cos co£ =
is the sum of (complex-conjugate) forces
x ~ ae i(0t = fre^+v (17.1.6)
F1 = (1/2) fe^ and F 2 = (1/2) fe~^;
where and cp are given by formula
r here due the fact that equation
to
(17.1.5b). Thus, the concise complex (17.1.1) is linear (by the superposition
notation (17.1.4) of the real solution principle) the motion brought about by
to equation (17.1.2) gives both the the sum F x +
F 2 of the forces is the
amplitude \x fr and the frequen- \
= sum of the motions produced by each
cy co and phase cp of the solution, that of them separately. But we have al-
is, the phase shift of the oscillation ready seen that the force F 1 generates the
with respect to the phase of the force motion x = (1/2) fre
l ^ t+®\ where r and
assumed to be zero (see (17.1.5b)). cp are given by (17.1.5b). Similarly,
32-0946
 / )

498 17 Applying Functions of a Complex Variable and the Delta Function

we can find the solution corresponding riod of time is made up of many similar periods
to the force F2 and then the sum of the
(cycles)).
2. The quantity a (see (17.1.5)-(17.1.5b))
two solutions (see Exercise 17.1.1). can be considered as a function of the complex-
valued frequency co. Here a becomes infinity
Note finally some properties of the solu- for co =
co r where (o r is the resonance frequen-
,

tion obtained which are even retained in more cy and equals % +

jpi with the imaginary part
general cases where the oscillatory system necessarily positive, i.e. pi >> 0. That a is
at hand consists of a number of masses, springs infinite, i.e. to be more exact that the ratio
(“returning forces”), and friction surfaces (a means that f/a
a/f is infinite, 0, i.e. that at =
number of friction forces). the frequency cor free oscillations may set up
1. The imaginary part of the “complex am- without the action of any external force F.
plitude” negative and equals
(17.1.5a)
— ((/co/z) t)/r
2
(Why?) This corresponds to a
is
But in this case the solution x (t) — ae*®**
.

satisfies the equation

negative angle (p (or to the angle cp within the
limits n < <
is
<p 2 ji). >

The negative nature of the imaginary part a

connected with the work performed by the
m l&r + h
W + kx=0 ’ (17 - 1 - 2b >

force F on the average during many oscilla- which obtained from (17.1.1) by substitut-
tions or exactly in one cycle. If an (external) ing F —is 0, whence it follows
force is given by the equation F f cos (Dt — — mco +
and the solution has the form (17.1.7), then
2
ihco
r + =
k 0 (17.1.9)

the work of the force is (cf. the denominator in (17.1.5) for a). Clearly,
such free oscillations (when friction is pre-
sent) can only be damped. Therefore in the solu-
tion e
l(S>rt
= q uan _
—
tity — pi must be negative, i.e. pi must be posi-
= ^
/ cos (tot) [
co/r sin (co^-f-cp)] dt tive.
In our case the quadratic equation (17.1.9)
yields
=— 2 rco
sin ((Dt + (p) cos ((Dt) dt
J
= ^-/
2
r0) [sin (2(o^ + + sin
cp) cp] dt
J

= 2
/ rco
£ — J sin cp dt
If here k/m h 2 /4m 2 — =
x 2 >> 0, then we have
—^ + two solutions co rl x = +
ih/2m and co r2 =
sin (2coi cp) dt
J
. (17.1.8)
(1
— +x ih/2m with the same (positive) imagi-
nary part ih/2m. (These solutions are almost
Clearly, the second integral on the right- complex-conjugate, more exactly co ri = o)*
2
—
hand side of (17.1.8) extended over the time t, and o) r2 = —
w*j.) If k/m —
h 2 /4m 2 <C 0, i.e.
which corresponds to one cycle of oscillations 2
h /^m 2 —
kim = xf >* 0, then the solutions
(t = t = 2ji/(d), vanishes, so that the work Xj) i and (o r2 —
— have the form co ri = (h/2m +
is equal to 2
(1/2) / co (sin 9 ) t. But this (h/2m —
x x i which are two pure imaginary
)
work must be positive since during the consid- numbers having positive coefficients of i ,
ered time the system returns to the initial
position corresponding to the same values of
since x x h/2m.<
both the kinetic and the potential energy;
therefore the work of the force F will not be The same techniques based on using
expended on the displacement of the system complex variable x can also be applied
and will be spent to equalize the friction force, in the more general case of an arbitrary,
which is always negative: the friction force
is opposite to velocity dx/dt and hence to the and not only sinusoidal, force F. In
displacement dx. the case of an arbitrary periodic force
Note that during the time differing from F we can just expand the force in a
the full cycle of oscillatiors, for instance, dur-
Fourier series (see Section 10.9) and
ing the short time which is much less than T
7

the work of force F can be both positive and then use the superposition principle
negative, since during this short period the considering separately the effect pro-
energy of the body (the sum of its potential duced by each sinusoidal force (17.1.3);
and kinetic energies) can decrease. We can be here the complex form of a Fourier se-
sure of the sign of work only if we consider
a very long period of time of functioning of the ries, that is, representation of an arbi-
system (or the time of one cycle— a large pe- trary periodic function F ( t with pe-
 ) ,

17.1 Complex Numbers and Mechanical Oscillations 499

riod T in the form of a linear set of where co = ]/ k/m (cf. Section 10.2) and
functions e ni0}t where co 2 n/T (fun-
,
= A = y km.
damental frequency; cf. (10.2.5)) and
n = . . ., —2, —1, 0, 1, 2, is an inte- For the derivation of formula (17.1.11)
ger, that is, in the form of the sum of or (17.1.11a) see Section 17.4; in particular,
functions e in(dt taken with some coeffi- compare (17.1.11a) with (17.4.5a). Without
considering the reasoning used in obtaining
cients f n In the case of a nonperiodic
.

formula (17.1.11a) it is also easy to verify

force F it must be expanded not in a directly that it gives the correct solution to
Fourier series but in the so-called Fou- equation (17.1.1) for the case h 0. Indeed, =
rier integral extended over all possible the function on the right-hand side of (17.1.11a)
values of frequency co (and not over depends on t in two ways: first, t is the upper
limit of the integral and, second, the integrand
frequencies co n =
tzcd that are multiples of
depends on t (in the expression sinfco (t —
the fundamental frequency co as in the t)]). Hence the derivative dx/dt must involve
case of the Fourier series); here the two terms: the first term is the result of differ-
entiation of the integral with respect to the
most convenient form of the Fourier
upper limit, and the second term is obtained
integral is also its complex form. in differentiating the integrand with respect
Without dwelling on the details of the to t. 17 2 But the derivative of the integral with
-

corresponding constructions (see, how- respect to the upper limit is equal, by the New-
ever, Exercise 17.1.2) we note that an- ton-Leibniz theorem, to the value of the inte-
grand at the upper limit (see Section 3.3);
other approach is possible to the problem thus we have
of oscillations generated by an arbi-
trary force F =
F {t): by decomposing dx
t

force F into separate delta-like compo- sinO'F (£) + ^4

f
)
co cos [co (t — t)] F (r) dx
nents which correspond to a narrow — oo
strip of the graph of function F that ,
t

the replacement of this function = Aa>

is,
cos [o> (t — x)] F (x) dx (17.1.12)
with the sum of functions F x (£), where j
— OO
F (t) for T^^T + dT, and consequently
0 for all different t

(17.1.10)
17 * 2 It
follows from the fact that if y = F (x) =
and then using the principle of super-
position. (The solution corresponding to / (
X ,
x) dx, then
j
the delta-like force F T ( t is termed the
Green function of the respective differ- Ay = F {x + Ax)~ F(x)
ential equation; in this connection, x-f Ax
see Section 17.4.) Here we arrive at the
solution of equation (17.1.1) having,
= f (x-\- Ax, r) dr —j / (x, r) dr
J
a a
in a more simple case when friction
forces are absent (when h 0), the = x+Ax x

following form =£ j
/ {x-\-Ax, r)dr—^f (x-\-Ax, r) axj
t

x(t) =A e^-^F (r) dr, (17.1.11)

X x
j
— oo + £ j
/ (x-\-Ax, r) dr — j
/ (x, r) dr 1 .

where A
and co are defined by the coef- a a

ficients m
and k of equation (17.1.1), or Thus, the ratio Ay! Ax decomposes into the
in the “real” form sum of two fractions, the limit of the first one
X
t

x(t) = A sin [co (t — t)] F (t) dr, is the derivative of the integral f / (x
J
+
j a
— oo
Ax, r) dr with respect to the upper limit and
(17.1.11a) equals the function / (x Ax, r) at r x + —
32 *
 — ,

500 17 Applying Functions of a Complex Variable and the Delta Function

i
supplied to the circuit (emf). Here as
d 2x
dt 2
— Ada cos 0 -F (£) + ^ I- a) sin [co(£ t)] well methods similar to those outlined
above (consideration of complex solu-
tions of real differential equations) are
F (t) d%
t
used rather widely, e.g. the concept of
“imaginary current” is familiar to all
= AoaF (t) — Am 2
J
sin [co (t — t)] F (t) dx
those acquainted with electrical engi-
— oo
neering.
= A(oF(t) — <* 2 x(t). (17.1.13) We omit the discussion of all these
questions.
Therefore, if ca = k/m, i.e.
2 co=|^ k/m, and
Exercises
A(o= 1/m, i.e. A = l/com = 1 / km, then
d 2x
~dt 2
—m F(t ),
17 . 1 1 , Find the solution (17.1.7) to equa-
.

tion (17.1.2) as the sum of two solutions to

equation (17.1.1) corresponding to the values
which is equation (17.1.1) at h — 0.
Fx =
(1/2) fe i(Dt and F 2 (1/2) fe~i®t of force —
F .

Finally we note that the same meth- 17 1 2 Prove that the (real) function
. . .
/ (x)
ods of “complex analysis” can be ap- defined in the interval from n to n can be —
plied to more complicated problems, say, represented as a sum (the complex form of the
Fourier series):
to those related to the establishment of
oscillations, where x 0 at t t 0 and = < =2
oo

(=... + c_ 2 e- 2i *
x =/= 0 only at t t0 or to the case of > ,
f (*>
— OO
c h ihx e

several oscillating bodies when in the

system there act a number of “return- + +c +c 0 ±e
ix
+ c e 2ix
2 -|- . .
.),

ing forces” (“springs”) and there are JT

many
friction are
friction surfaces (many forces of
present which are propor-
with c fe = — i

j
r

- JT
f(t)e-ik*dt (here k = . .
.

tional to derivatives dxldt). In the — 2, — 1, 0, 1, 2, ...) and c_^==cj; if the

theory of electric circuits (see Chap-
ter 13) we also frequently encounter values of are complex numbers (while x
/ (x)
then the same formulas are still
is still real),
oscillatory processes— in equations re-
valid, with the exception that c_
lated to equation (17.1.1) the role of k is not nec-
essarily equal to c%.
coefficients m, h, and k is played here
by inductance L, resistance R and ,

the quantity 1/C, the reciprocal of ca- 17.2 Integrals in the Complex Plane
pacitance C, and the role of the “exter-
nal force” F is played by the voltage
We consider the definite integral
z

(which transforms to / ( x x) as ,
kx 0); w (
z) = / (z) dz, (17.2.1)
the second fraction is equal to j
a

where the lower limit is a complex

-^7 j [/(s + Az, T)— /(«, T)] dx number a, while the upper limit z is
a variable. The integral here is understood
in the same way as when we dealt with
f f(x + Ax, T )—/(*, t) real variables (see Chapter 3), that is,
“ J Aa:
d
the right-hand side of (17.2.1) is approx-
a
imately equal to the sum
and under ordinary conditions its limit is

/ {z)o Az0 -\- / (

z l) &Z 1 + ••• T“ / (
z n- 1) & z n- 1

dx = 2 / (Zi) A z it
i=0
 1 z

17.2 Integrals in the Complex Plane 501

Figure 17.2.1 Figure 17.2.2

where a zx z 2
z0 = z n -i, z n
, , , •
= dz — Or the integral dz of
/ (z) / (z)
z are thepoints which divide the path C ^
(connecting a and z) of integration into an analytic function over a closed con-
rather small portions, and A z t
= tour is always zero if inside the contour

*i+ — z i9
0, 1, i .,
= . . n — 1. the function has no singularities. Con-
Let us find the integration path C. versely, if the integral is equal to zero
Figure 17.2.1 shows three such paths as over any closed contour, it is indepen-
an example. We can prove, however, dent of the path of integration.
that if within the region bounded by two In order to prove the above state-
paths C and C x the function has no sin- ment, let us first consider a small square
gularities that is, it has sense every-
,
with vertices 1 (x 0 y 0 ), 2 (x 0 A, y 0 ), , +
where and nowhere equals infinity, 3 (^0 A, y 0+ A), 4 (x 0 y 0 -\- A) (Fig- + ,

then integral (17.2.1) taken over these ure 17.2.2). We denote by §f (z) dz
paths has one and the same value. Of
the integtal over a closed contour join-
course, when we mention the derivative
/' {z) of the function f {z) of a complex
ing consecutively points 1,2,3, 4,
variable we mean that the function is
and again 7; we evaluate this integral.
considered analytic. Let f {z) = f (x + iy) = u (x, y) +
The property of integrals (17.2.1) iv (x, y); we set A so small that we can
given above admits of another formula- assume that
tion equivalent to the original one. Let us 2 3

denote by
qp /
(z) dz the integral extend-
§ / (z) dz = j
/ (z) dz + j
/ (2) dz
1 2
ed over a closed contour, say, the con-
4 1
tour which obtains if we go from point a
to point z along path C 1 and then re- + /(z) dz+ f(z) dz~f(Zi) A z,
j j
turn from point z to point a along path 3 4
— C 1 (path C 4 joins a and z, but we go + / (z 2) Az 2 + / (z 3 ) Az 3 + / (z 4 ) Az4 ,
along the same path in the opposite di-
rection; see Figure 17.2.1). Since (cf. where Zj is understood as point and
7

Section 3.2) f (z) dz dz ,

we A Zj= z j+1 — zp, here = 1,2, 3, 4 7
j
-Ci Ci and z 4+1 = z is again point 1 Since x .

have A = z — z = A = —A z = z —
Zj_ 2 x 3 3
z and A z = z — z = jA = —A
4 2 = 3 2 4
/ (z) dz = / (z) dz + f (z) dz z4 —z l9 we have
j j
C -Cl
— / (z) dz ~ {[ u (x 0 , y 0) + iv (x 0 , y 0) I
= j
f (z) dz f / (z) dz,
C Cl — [u (x 0 + A, y0 + A)
and if f(z)dz=^f (2) dz, then + iv (x 0 + A, y0 + A)]} A
j - a, 4- v (*«
Ci c 4- {(u (^0 “I y<>) i
502 17 Applying Functions of a Complex Variable and the Delta Function

+ A, — l> (*0» y + A)
J/
0 )1 0
higher than A, and the integral by a quan-
tity of order of smallness higher than the
+ iv X y + A)]} iA
( 0, area A 2 of the small square (of A 3 or A 4 or
= — {[w + A, y + A) (ar
0 0
still lower order). And even this implies that
the integral is equal to zero.
— u (x + [v (x0 + A,
0, i/
0 )] Indeed, let us subdivide the initial square
y ~0)v (*o» Vo + A)]} A
with side A into n 2 squares with side A/n.
The integral over the contour of the original
— {y {x + A, y + A) 0 0 square is, as can be easily understood, equal
— v {x y )+ [u (x„, y + A)
0, 0 0
to the sum of n 2 similar integrals over the
contours of all the small squares (we go along
— u (x + A, y )]} iA
0 0 all the contours in the same direction, say,
= -A- A — 5- iA, counterclockwise): in fact, when summing up
all the integrals we go along each internal

where ^4 and B denote the expressions line segment two times in two different direc-
tions and the sum of these summands will
in the braces. But cancel. But while the value of the total inte-
gral is of the order of A 3 the value of each sim-
u + A y A) — u x y
(*o , 0
-\-
( 0, 0) ple square is of the order of ( A/n ) 3 and the
,

= \u(x + A, + A) — u (x
o j/ 0 o, i/
0 + A)] sum of these integrals is of the order of n 2 (A/n) 3
= A 3 In. Since this estimate is valid for any n,
+ {u{x y + A) + u(x y )]
o, 0 0 , 0
our integral must necessarily vanish.
Similarly, that is, by dividing the region
^ du{xo, y 0 + A) du(x 0 yp) 1 .
inside a closed contour into small squares we
— rL
, ,

3a: ^ % J can prove that the integral taken along the

contour of this region equals zero, thereby
and proving that integral (17.2.1) is independent
y(a; 0 — v(x y + A)
+ A, y 0) 0 , 0
of the path of integration.

==[y(x«-|-A, — y(*o + A, Po + A)] I/,,)

Example. Consider the integral
+ [y(a: + A, p + A)— y(x
0 + A)] 0 0, z/
0 1 +i
^ dvjxo + A,
+ ^ dxy + 1J
A (a: /= z*dz. (17.2.2)
" fL go) , 0 , 0 ) .
j
% ’
0

where we, as before, use the approxi- Suppose path A goes from point z 0 0 =
mate formula: / (x + A) — / (x) - ^ to point z = 1 + i first along the real

/' {x) A. Thus, axis (from z 0 = 0 to z x =

1) and then
along a straight line which is parallel
dujxo, j/p + A) dv (t 0 -|- A, y 0 )
to the imaginary axis (from z x 1 to =
dx dy
it z = 1 + 0* Here we have
Vo) i
dv(x 0 , j/p + A)
+t
dy dx ])
i,

r du dv I=^z 2 dz=z (x 2 —y 2 +2ixy)\ u = 0 dx

~ \l~te f
dy J [w + w]}* = ° A
j
0

1^

by the Cauchy-Riemann relations. (In

+ j
(a:
2
— y 2 +2ixy)\ x=1 idy
the last expressions the arguments of 0
the functions du/dx, dvldy, duldy, duldx
can be equated to x 0 y 0 that is, we , ,

can consider that the partial deriva-

tives are calculated at the vertices of The numerical result is the same if we
the square shown in Figure 17.2.2; here choose another path along the B first
the error is of the order of A.) Similarly, imaginary axis from point z 0 0 to =
we can show that B is also equal to point z 2 =
i and then along the straight

zero. line which is parallel to the real axis

(from point z 2 i to point z 1 —
i. = +
Of course, we have really only established
that the integral is almost zero, that A and
Verify this). Finally, we can take in-
B differ from zero by a quantity of order tegrals along diagonal C of the rectangle
 b . —

17.2 Integrals in the Complex Plane 503

(the square here) having vertices at small (in absolute value) A z (irrespec-
points z 0 ,
z l9 z, z2 : put z = (1 i) t + 9
tive of the increment A z, that is, of
with 0 ^ ^ t 1; then dz = (1 + 0 dt the ratio of its real part A^ to the coef-
and ficient Ai; of its imaginary part) we
l + i 1^
obtain
/ = z 2
dz = (1 + i) t ]
2
(1 + i) dt
j j
[
A F(z)
o o
Az ~f (z). (17.2.6)
i

= (1 + 0* t
a
«ft = -|-(l+0* From (17.2.6) it follows that the value
j
0 of integral I (a, z) can only differ from
the function F (z) by a constant. The
value of the constant can be found from
the condition / (a, z) 0 at z a = = ,

Integral (17.2.2) can be evaluated whence it follows that I (a, z) F (z) =

without resorting to all these simple F (a) in accordance with (17.2.4). But
arithmetic calculations. To this end, since (see Section 15.2) for complex
as in the case of integration of real- numbers we also have 3
(z )' = 3z 2 ,

valued functions, it is sufficient to find

the antiderivative F (z) of function z 2 .
z 2
dz = z
3
/3 +C from which the val-
j
Let F (z) be the function (of course, ue of integral (17.2.2) obtained above
analytic) of the complex variable such follows immediately.
that We have presented all the reasoning
so briefly because there is no difference
^ ==/(*)• (17.2.3) here between (smooth) functions of the
real variable and (analytic) functions
Then, by the Newton-Leibniz theorem, of the complex variable— all the re-
Z sults of Chapter 3 are applied to the
f(z)dz=F (z) + C, case of functions of a complex variable.
J\ Now let us consider specific proper-
a
ties of integrals of functions of complex
b
variables which have no analogs in the
I (a, b)= jf(z)dz = F(b)-F(o), real region. Consider the integral

(17.2.4)
dz (= In z). (17.2.7)
where a ,
b, C F ,
(a), F (
)
are complex ^
numbers. Equations (17.2.4) can be
proved as in the case of functions of We take a circle radius p
of fixed
a real variable: we find the change in with center at point 0 as the path
z =
=
z
of integtation, that is, we set z pe iq) = ,

the integral I (a, z)

j
/ (z) dz at a where p =
constant and 0 cp^ 2n; ^
a then dz =
i pe ? dtp and dz!z= idcp (equal-
i(

small change in the upper limit of ity d (e i(P) =

ie i(v dcp follows from the
integration: fact that by the definition of the quan-

/(a, z + Az) — /(a, z)

tity e i(P for small Acp we have e iA Q ^
z+Az
1 +i Acp). Therefore, integral (17.2.7)
taken along our closed contour is
= j
f (z) dz = f (z) Az+ . . (17.2.5) found to be equal to
2

(we have omitted here the summands 2n

of A z 2 and even smaller orders
| |
^)-^-dz=j idq> = 2ni 9 (17.2.8)
of magnitude), whence in the limit at o
504 17 Applying Functions of a Complex Variable and the Delta Function

that is, its value is far from being zero. 8

The value (17.2.8)integral is
of the
independent of radius p of the circle;
moreover, it is also independent of
the shape of the contour enveloping
point 0; we will obtain the same result,
2ju, when integrating along any other
closed contour containing 0 inside, say,
along the boundary of a square with
Figure 17.2.3
vertices ±1
i ±
(see Exercise 17.2.1).
Why is the main result concerning
the equality to zero of integral
We can go along circle C not once but
n times and only then go along the seg-
/ (z)
dz found to be invalid? The ment A of the real axis; here the value
^
reason is that in the given case the in- of integral (17.2.9) will be
tegrand / (z) 1 /z =
has a singularity
inside the contour of integration:
its value is infinite at point z 0. = 1 nC A
No matter how small the subregions
we take inside the contour, the value = + n j "§ = ln 2 + 2nni
1 "T"
>

/ (z) =
oo is inevitably inside one of A C
the subregions; it will make its con-
tribution —
murder (and the infinity in- where nC denotes the contour obtained
side a contour) will out! in an rc-fold rotation along circle C in
Let us examine in detail our inte- the same direction (counterclockwise).
gral (17.2.7). This integral taken along On the other hand we can also go along
the contour surrounding point 0 is the same circle C in the opposite direc-
equal to 2 ju. But when integrating the tion, clockwise; if we go along this
function Hz along any other path, we path m
times we obtain the value of
cannot also ignore the fact that the integral (17.2.9) equal to ln 2 2mm. —
integrand has a singular point. Here the value ln 2 0.69 of integral ^
Let (17.2.9) can be obtained over a multi-
2 tude of integration paths: all paths
(17.2.9)
which do not go around point 0, like
path B shown in Figure 17.2.3; the
1
2

The value of this integral taken along result I = dzlz = ln 2 — 4ji i can al-
j
the shortest path A going from point l
z =1 to z —
2 along the real axis so be obtained over many (other) paths
(Figure 17.2.3) is equal to In 2 0.69. ~ and so on.
But we can also go from point z 1 — Thus, the final solution is
to point z =
2 by another path. For
=
example, let us go from point z
along circle C of radius 1 with center
1
/=
2

j
—= In 2 + i2nk,
at point 0 and only then go along the 1

segment A of the real axis from point 1

to point 2. Here we obtain a different where k = 0, ±1, d=2, .... We may
value of integral (17.2.9): say that this is the most general deter-
mination of the natural logarithm of 2,
while the number 0.69 is only one of
many values of ln 2, this is its prime
1 C A — = =
value. Indeed e 1 e ln2+i2nh 2e i2nh
= 2m + In 2 ~ 0.69 + 6.28 i. 2.
 i i

17.2 Integrals in the Complex Plane 505

Thus, we have kept our word given the equality <jp

dz/z
n = 0, where the
in Section 14.3 and explained in a new
way the reason for the logarithm w — integer n greater than unity.
is

In z being many-valued. The many-val-

To sum up, the value of the integral

ued nature is connected, on the one along a closed contour depends not
just on whether the function is infinite
hand, with the fact that the function
inside the contour, but on the presence
e ^ is periodic,
that is, that the func-
1

2 k k. On the = or absence in the expression for the

tion equals 1 for (p
function / (z) (in its expansion in
other hand, the logarithm can be de-
series extended over both positive and
fined as the integral dz/z\ in such an negative powers of z a, where a is —
approach, as well, the fact that the the singular point of the function in
logarithm is many-valued is connected which / (z) =
do) of the term which is
with the behavior of the function Hz proportional to l/(z a) and on the —
at z —
0 and that In integrating we coefficient of this term.
17 3
In view of -

can avoid the point z 0 at which the= the fact that the coefficient of (z a)' 1 —
function f (z) =
Hz is infinite (such a in the expansion of / (z) in powers of
point is termed the pole of the func- (z —a) is so important, it bears a spe-
tion of the complex variable), cial name, the residue of / (z) at point
Note that it is the function Hz that, z — a Clearly, at any point where the
.

upon integration along the contour which function is analytic, that is, where it
does not surround point z 0, yields = assumes a finite value and can be expand-
a nonzero result, 2n (while similar ed in a Taylor’s series (in the neigh-
integration of function dz, where c is a borhood of such a point), the residue
fixed, generally speaking, complex num- is zero. If we take the function / (z) =
ber, gives the result c-2ni). Clearly, 1/z
2
which becomes infinite at z
,
0r =
the integral along a closed curve of it has a zero residue at the same point.

function / (z) =
a, where a is a con- In contrast, the function
stant number, or of function / (z) =
az n where n
, >
0 (for example, of Z*+l (z+j)(2— j)
function 2 z 2 or 4
£z ), is— zero— in
fact, these functions have no singular = _1
2i z
1

—
1

2i
1

z-\~i
(17.2.10)
points (poles). But quite unexpectedly
has singular points: z = i and
two
integration of functions that are negative
(and greater than unity in absolute
z =— The residue of cp (z) at z = i
i.

is equal to H2i = —i/2, and the resi-

value) powers of z, say 1/z 2 or 1/z 3
along a closed contour surrounding
,
due at point z =— i is equal to
the pole z 0, —
in which the func-
— H2i = i/2. Indeed, if, say, we tako
-1
tion is infinite, also yields zero: for
the expansion of (—1/2 i) (z i) it + ,

has no terms with negative powers of

such functions / (z) we again have
(z —
i), so that the total coefficient of

(jp/
(z) dz = 0. Indeed, substitute into (z —
i)
_1
here is equal to l/2i.
There exists a remarkable applica-
integral
^
dzlz
2
the values z pe P, = Z(

tion of integrals of functions of a com-

dz = ipe ® dcp, where p
1
=
constant and plex variable to the evaluation of inte-
0 ^ ^ cp 2 k; we get grals of a real variable. Here is a typi-
2jx
cal and yet important example:

0 j T^r dx - (17.2.11)

17 • 3
We do not consider more complicated
cases, where the integrand itself is many- val-
In the same simple way we can prove ued (like, say, the function / (z) = In z).
 ,

506 17 Applying Functions of a Complex Variable and the Delta Function

and along y we have

l^icDzJ _ |^iO)(ac+iy)| __ |g-CDy| |^icox|

= |
e-®v\ X 1 < 1

(since in the upper half-plane y is great-

er than zero and thus e~ 1), while
m<
I 1
—
111
|
1_|_ 2 2 Z2 I R2 t

we conclude that along y the absolute

value of the function / (z) e 2co 72 X =
(1
2
z ) is +
extremely small; therefore,

the integral 174 / {z) dz along y is

^
v
small, too. This implies that for R
Figure 17.2.4 large, the integral
A ft AdZ
(17 ' 2 14)
This is the integral of a product of a •BthT*-
r
'

rapidly varying function cos ax (where

it will be assumed that w is much high- where T is a closed contour consisting
er than unity) by a slowly varying of the line segment x of —R ^ ^ R
function y x (1
2
a: )*
1
=
The curve + - the x axis, with R
very large, and the
representing the integrand is depicted semicircle y in the upper half-plane
in Figure 17.2.4a, with the dashed (Figure 17.2.5), is very close to /+
curves being the graphs of the functions along the real axis. But the integral of
yx = (1 + 2
a: )
-1
and y 2 = —(1 + x*)~ l . an analytic function over a closed con-
According to Euler’s formulas (14.3.4) tour depends, as we know, only on the
singular points of / (z) lying inside this
cos (ox = e i0)K + 4r e~ iax . (17.2.12) contour. But does the function / (z) =
£icoz/(i z
2
)
= e i(* z cp (z) have any sin-
gular points inside our contour?
Substituting this into the right-hand
It is clear that inside T the function
side of (17.2.11), we I into two
split
where e idz never becomes infinite and has no
integrals, 1+ and /_,
17 - 4
We see that it can be assumed that
(17.2.13) | / (z) |
< R~ 2 along y; this together with
— oo the inequality
|
j
/ (z) dz | c [ \ f (z) \ \
dz |

The second integral /_ can be written (see Exercise 17.2.2), which follows from the

in a similar manner, only e i(OZ must be very definition of an integral, and the fact
that the length of the semicircle y increases
replaced with e~ i(0z The (dummy) .
like the first power of R, imply that the inte-
variable of integration in (17.2.11) is

denoted here not by x but by z which ,

gral / (z) dz decreases, as R grows, in any
|
suggests that it assumes not only real v
case no slower than R-1 .

values but also complex values (this

idea central to our discussion).
is r
Suppose thatz runs along the semi-
circle y: z R in the upper half-
| |
=
plane (z x =
iy, y + > 0), with R
being very large. Since
icoz i(0z

U(Z)\—
1 * _ 1 l* l

2 l + *2 2 |l + z*| Figure 7.2.5

 . ° )

17.3 Analytic Functions of a Complex Variable and Liquid Flow 507

singular points. The function cp (z) = of the “amplitude function” y1 =

1/(1 + z
2
as noted earlier, has inside
), (1 + x2 )" 1 .

T the singular point z i, and the res-= Integrals similar to (17.2.11) play
idue of cp ( z ) at this point is i/2. — an important role in the theory of vi-
This implies that the function f (z)
= brations and in theoretical physics in
(1/2) e
i(0z
cp (z) at its singular point z =i general (compare with the integral in
has a residue —
(i/4) e i(0i (i/4) e
~
By =— 3
. (17.1.11a)). A slowly varying pulse
the general properties of analytic func- (whose shape is defined by the function
2 -1
tions, the integral (17.2.14) over the of time / ( t (1 £ ) ) =
does not ex-+
cite a high-frequency oscillating system
huge contour T is equal to
^
e i(* z (1 + because at high frequencies (co 1) the
cr
oo
z-)' 1 dz, a small circle with
where cr is
its center at point z i; this integral= integral f ^ *s sma ll* Only the
— oo
is equal to
theory of functions of a complex vari-
able makes it possible to establish that
44- e-®2ni this integral falls off exponentially as
CO —17.2.1.
oo.

(17.2.15)
Exercises

The same value has the integral Prove by direct calculation that
(17.2.14), which, hence, is independent ^dz/z — K
17.2.2. 2ji£, where the integration contour
of R. For this reason integral (17.2.13) is K
-0
also equal to (1/2) Jte .
coincides with the boundary of the square
In a similar manner we find that whose vertices lie at the points ±1 ±i.
/_ =
term in
(1/2)
the
05

integral
w
which is the second
,

I. Thus, the final

Prove the inequality j/w dz
result is
17.2.3.
= f (z) \dz\, where A is an arbitrary arc
I ue~® (17.2.16) i
| |

It is extremely difficult to arrive at in the complex plane.

this formula without using functions of oo

a complex variable. Numerical inte- Prove that e

_<2/T2
cos cot dt =
gration is also very involved since for j
co >1 the expression on the right-hand
rjfne 0)2x2
(This result is physically im-
side of (17.2.16) or (17.2.15) is very
.
2 2
portant: the so-called Gaussian profile e~x /t
.small. Separate positive and negative
of a slow pulse (see Figure 17.2.46) has little
half-waves of the periodic function influence- (does not excite) high-frequency os-
cos ($x will not be especially small: cillating systems, corresponding to high fre-
their amplitudes, if we are speaking of quencies co.)
-waves corresponding to moderate val-
ues of x, are of the order of unity, 17.3 Analytic Functions of a Complex
which implies that the contribution of Variable and Liquid Flow
a half-wave (of the type we are consid-
ering here) to the integral I is of the In this section we conclude the large
order of zhl/co. (Why?) On the other topic of complex numbers and func-
hand, for co ^>1, the total value of I tions of a complex variable. One can
given by (17.2.15) is much smaller only marvel (we already mentioned
than 1/co. Hence, positive and negative this before) at how the introduction of
half-waves cancel each other with a an “imaginary unit” i ]/ 1 not on- = —
high degree of accuracy, which is rela- ly resolved all the difficulties that ap-
ted to the analyticity, or smoothness, pear in solving the various quadratic
508 17 Applying Functions of a Complex Variable and the Delta Function

equations but also introduced harmony integral that cannot be evaluated in

into the entire theory of algebraic terms of elementary functions (of the
equations (or simply algebra) as well z
22
as the theory of functions and mathemat- type l e~ /2 ^z) or by the solution of
ical analysis. All functions can natu- a
rally be generalized to functions of a differential equation that cannot be
complex variable: not only polynomials solved in the traditional sense (i.e. we
and algebraic functions (only here do cannot express the dependence of the
the questions concerning the number of unknown function on the independent
zeros of polynomials and the power func- variable by means of a formula) or by an
tions with fractional exponents gain infinite series. What is important is only
completeness), but also the exponential that the value w =
/ (z) of the function
function, the logarithms, and trigono- be dependent on z x =
iy and not on +
metric functions. We replace the general x or y separately; in other words, the
definition of a function as correspondence function must not break up the natural
between two sets of variables (even bond between x and y in their combi-
in the real domain we are not able to nation x +
iy.
preserve this definition of a function in Functions of a complex variable
all generality) by the definition of an prove to be a powerful tool for solving
analytic function. By restricting our- mathematical, physical, and engineer-
selves in this way, we gain colossal in- ing problems. In such problems, natu-
formation about the functions: the en- rally, the data is given by real quanti-
tire apparatus of higher mathematics ties. The answer must also be expressed
(i.e. the concepts of a derivative, an in real values, since “the square root of
integral, a differential equation, a pow- minus one piece of chalk” cannot be
er series) can be applied to analytic allowed into everyday life. Neverthe-
functions of a complex variable; how- less, at some intermediate stage in a
ever, many statements that are true for problem, somewhere between the for-
analytic functions prove to be invalid mulation of the problem and the ans-
for functions in general. Of great impor- wer, it often proves to be useful to in-
tance to analytic functions are the troduce complex numbers or quantities
points (the values of the variables) where and taking (but only temporarily) a
the functions become infinite. It has complex number of pieces of chalk, so
been established that the behavior of to say (compare with Section 17.1).
analytic functions at such points de- For instance, in Section 17.2 we saw
oo
termines a broad spectrum of properties
C COS (DCC
of these functions; for instance, the that to evaluate \
dx (which is
1 x2
presence of such points enables evalu- — oo
ating integrals of functions over various a purely real integral) it is advisable to
i(* x
contours without actually calculating write cos co£ as (1/2) e i(OX (1/2) e~ +
them. and thus introduce what seem to be
It has been found that by specifying quite unnecessary imaginary units.
a function of a complex variable on a Even if we do not know the precise for-
small segment (as small as desired) mulas that express a law of nature, of-
we can, at least in principle, determine ten only the assumption that there exists
the function in the entire complex an analytic function of a complex vari-
plane, over the entire range of the in- able that expresses the law (even a
dependent variable. It turns out that function that cannot be expressed by a
analyticity is determined not by the formula) is fruitful and informative. The
fact that there exists a simple formula very fact that such a function exists
for expressing the function in terms of leads to certain relationships generated
the independent variable z: an analytic by the analyticity of the function (a
function may be expressed, say, by an function unknown to us). Such an ap-
 x #

17.3 Analytic Functions of a Complex Variable and Liquid Flow .509

proach is widely used at the forefront of

modern theoretical physics, say, in
what is known as dispersion relations
(unfortunately, nothing more can be
said about these relations here). The
greatest miracle is that the imaginary
unit and complex numbers have proved
to be necessary for quantum theory, the
physical theory of the microworld. The
place of complex numbers in quantum
theory is so important that today in
teaching analytic geometry to students In many cases it is admissible to as-
of physics and chemistry it is often sume that there exist functions cp (x, y)
thought necessary almost at the start and (x , y) such that 17 - 7

to speak of a complex plane and a space

where points are specified by complex- V *~ ~ dcp(;r, y) „
V V~ _ d<p{z, y)
dx ’
dy
valued coordinates, although of course
there is no way in which such a space (17.3.1)
can be visualized. 17 5 -
d^(x, y) „ _ di>(x, y)
Apparently, the first serious applica-
U x— dy
9
U y ~ dx
tions of the theory of analytic functions (17.3.1a)
of a complex variable (the very crea-
tion of this theory was due probably to cp is known as the velocity
The function
these applications; see p. 483) were in potential and the motion of liquids
,
fluid mechanics which is the theory of
,
that obeys the above assumption is
the movement of gases and liquids. called potential flow. It can be said that
Suppose we have a plane-parallel flow the flow of a liquid in the vicinity of
of an (incompressible) liquid of con-
stant density. The velocity vector v
point M=M
{x, y) is potential if and
only if it is reduced to the transfer of a
of any particle of the liquid is at all mass of liquid in the direction of v and
moments of time directed parallel to in the deformation of a small volume of
the (horizontal) xy- plane and the entire
the liquid (which is admissible in view
column of the liquid corresponding to of the fluidity of the liquid) but does
given x and y coordinates and different not include rotation of the liquid about
altitudes z moves as a whole (in “one
piece”). In this case we can ignore the
point M
(such rotation is often observed
as whirlpools in rivers or even in a bath-
third dimension and consider the mo-
tub). In accord with this, potential
tion as occurring in the ary-plane. Sup-
flow is sometimes called irrotational (or
pose that v (a vector) of the flow at
vortex-free) and the points (generally,
point M M
= (x, y) has components isolated) where this condition breaks
v x and v y (Figure 17.3.1); it is clear
that v x = v x (z, y) and Vy = v y (x, y) 17
Since the functions cp and ^ are defin-
- 7

depend on point M, that is, are func- ed only through the values of their deriv-
tions of this point, or functions of atives in (17.3.1) and (17.3.1a), they are
the coordinates x and y. 17 6 -
similar to potentials, that is, are defined only
to within arbitrary summands C 1 and C 2 ,

which enables choosing the zero values of cp

17 5*
E.g. see I.M. Yaglom, Complex Num- and at our will. Only the difference of values
bers in Geometry Academic Press, New York,
,
of cp (or ip) at two different points but not the
1968. values themselves has any physical meaning.
17
For the sake of simplicity we have
• 6
(Note also that since v and v y have the dimen-
restricted our discussion to the case of steady- sions of velocity, m/s, and the coordinates
state flow, when the velocity does not change x and y have the dimensions of length, m, the
in time and the vx ( y) and vy (x, y) are
,
functions cp and ^ defined by (17.3.1) and
independent of time. (17.3.1a) have the dimensions of
2
/s.) m
510 17 Applying Functions of a Complex Variable and the Delta Function

down are known as vortexes. The differ- (In Figure 17.3.1 the equipotential
ence curves are depicted by dashed curves.)
fay
Thus, the mesh of curves cp — constant
dx
dv x
dy
(17.3.2) and i]' =
constant provides a complete
picture of flow of the liquid as a whole.
which is zero if the flow is potential Now let us assume that in the neigh-
(why?) and, hence, is nonzero only at borhood of a given point x y) in a M (, ,

vortexes, is known as the strength of the liquid there are neither vortexes nor
particular vortex, or simply vorticity. sources and sinks; in other words, both
The function (x, y) is known functions, cp and 'ip, exist. Obviously,
as the stream function . It can be proved from (17.3.1) and (17.3.1a) it follows
that if at a point M
(x, y) there exists that
the amount of
a stream function op,
liquid flowing into a small contour sur-
_ x
_ d<p _ ^ 9
dx dy
rounding point M
will be equal to the
_ Jhp_^ d^
(17.3.3)
amount of liquid flowing out of the UV ~ dy dx
'

contour every second (the liquid is as-

sumed incompressible). Thus, the ex- Equations (17.3.3) are close in form to
istence of a stream function at a given the Cauchy-Riemann equations (15.2.6),
point M=M (x, y) guarantees that which connect the real and imaginary
point M neither
is a source nor a sink parts of an analytic function w =
of liquid. The sum w (z) = u (x, y) -f iv (x, y) of a com-
dv y plex variable z = x + iy. Thus, if we
du x .
(17.3.2a) assume the xy - plane to be the complex
dx dy
z plane, then we naturally arrive at
(which identically zero if at point
is the analytic function
M=M {x, y) there exists a stream
w =w (z) = cp + hp (17.3.4)
function ty) characterizes the strength
of the source or sink, while the sign of of complex variable z, a function that
this sum characterizes the nature of is closely linked to liquid flow.
the “singular point” M : if (17.3.2a) is
The function the (17.3.4) is called
positive, we have a source at point M complex potential of liquid flow It .

(i.e. more liquid flows out of the con- completely characterizes the flow of a
tour surrounding M
than in), while if liquid, so that the study of various
(17.3.2a) is negative, we have a sink. plane-parallel flows is reduced to the
(The sign of the difference (17.3.2) cha- study of analytic functions w {z). The
racterizes the direction in which the li- derivative of the complex potential,
quid rotates at the vortex.)
The reader can easily see (Exercise dw d cp dip
= —v«
v
17.3.1) that the curves ^ constant are = dz ( dx dx
(17.3.5)
simply the streamlines that is, the ,

curves along which the particles of the (see (15.2.3)), is also an analytic func-
liquid flow: along each streamline (the tion of complex variable z. This func-
tangent to each such line at each point tion is called the complex velocity of
M has the direction of the velocity vec-
flow . This name suggests a relationship
tor v) the magnitude of ip remains the
between the complex number v and the
same. Equipotential curves which are ,
velocity v; obviously,
specified by the condition cp (x, y)
=
constant, on the other hand, are per- I
v I
2
= (— v xf + {VyY = vl + vl
pendicular at each point to the velocity = \
V I
2
,

vector (see Exercise 17.3.2): the liquid

moves across these curves in the direc- that is, the absolute value of v is equal
tion in which the potential decreases. to the length of v; on the other hand,
17.3 Analytic Functions of a Complex Variable and Liquid Flow 511

Arg v characterizes the direction of complex potential of the electric field*

vector v (see Exercise 17.3.3). Flows The derivative
specified by the complex potentials
w =cp +
ity and iw £cp can be

called conjugate because the stream-

=— + E=—
'
dz
(
\
= — dx
dV
dx
,

lines of one flow coincide with the equi- =— + iEy )

(17.3.8)
potential curves of the other, and vice
versa. of the complex potential can be called
Quite close to the above scheme of liq- the complex electric field strength since ,

uid flow is the use of analytic functions this complex quantity is very close in
in the classical theory of the electromag- meaning to the electric field vector E;
17 8
netic field. Let us consider a flat
-

in particular, E |. The ele- E =

electric field. We
will assume that with-
ctric fields with
| |

W=O+
V¥ and
|

in the plane region we are considering iW =

there are no free electric charges or vari-
—W + i<& aresometimes called
conjugate ,
the streamlines of one coin-
able magnetic fields (the presence of cide with the equipotential curves of
such fields is equivalent to the presence the other, and vice versa.
of charges, in accord with the law of We will not dwell any further on the
induction). Suppose that (x, y) ® ® various ideas of the mathematical theo-
is the potential of the electric field at a
ries of liquid flow and electric field,
point M=M
(x, y). Then the electric
which have to do with functions of two
field strength at this point is fixed by variables x and y and, respectively,
vector E =
(E x E y ) with coordinates ,
with differential equations involving the
d cp partial derivatives of these functions,
Ex = dx ’
dy *
(17.3.6)
which are simply known as partial
differential equations (see Section 10.8).
In addition to potential ® we can intro- We note only that the theory of analy-
duce the streamline of the field T* ,
= tic functions of a complex variable
W (x, y), such that 17 9 -

plays an extremely important role in

T these purely applied problems.
hFy -
d dW
(17.3.6a)
dy ’ dx
Exercises
The curves Y=
constant are the lines
=
17.3.1. Prove that (a) the curve con-
of force of our electric field, while the stant is a steamline of liquid flow; in other
curves <J> =
constant are the equipo - words, that the tangent to this curve at each
tential curves By virtue of (17.3.6) . point M
{x, y) of this point coincides in di-

and (17.3.6a), the function rection with the vector v ( x y), and (b) if ,

AB is an arbitrary arc (in the plane of liquid

W=®+W (17.3.7) flow)

B = B
endpoints A
with A y± )
x 2 y 2 ), then liquid flow across the
and —
( ,

arc per unit time is equal to the difference

is an analytic function of the complex —
(#!, yf) (the increment of the
^ (^ 2 2/ 2 )
variable z x =
iy; it is called the +
?

streamline along arc AB).

17.3.2. Prove that the equipotential curves
17
Precisely, here we can speak only of
• 8
cp =constant and the streamlines con- =
stant are perpendicular at each point in the
electrostatics and magnetostatics since two ,
liquid flow (precisely, the tangents at point
coordinates (x and y) prove to be insufficient
for the general problems involving electromag-
M (
x y) to the streamline and equipotential
,

curve passing through this point are perpen-

netic phenomena— three spatial and one time
dicular to each other).
variable are required.
17 9
Since
•
and have the dimensions W 17.3.3. Suppose V = V (x, y) is a com-
of voltage, V, and the coordinates x and y plex number such that OV = v (where OV
have the dimensions of length, m, it is clear is a vector with the initial point at 0 (0, 0)
that the dimensions of E x and E y are those and the terminal point at P, and v is the ve-
of V/m (we can also say that V/m are the locity vector at point (
x y)). Prove that M ,

dimensions, of E). points v and V in the complex plane, where v

 z a 2

512 17 Applying Functions of a Complex Variable and the Delta Function

is the complex velocity of flow, are symmetric mass of the rod without the loads is
with respect to the imaginary axis (and, hence, b
Arg v = ji — Arg V).
a (x) dx, while with the loads it is
17.3.4. Describe (plane-parallel) flow of
an (incompressible) liquid for the following a
•complex potentials W= +
cp ity: (a) w nz, = b
where a is a real or purely imaginary number,
(b) w =
az 2 (a is a real or purely imaginary M= a ( x) dx-\~ 2m i->

number), (c) w =
a! ( is a real or purely
j
a
imaginary number), and (d) w =
In z and
i

w— i In z. where the summation is extended over

all the loads attached to the rod. The
position of the center of gravity is
17.4 Application of the Delta Function
b

Jf Sections 17.3 logically be-

17.1 to
x = ~w ( j
xo W dx+ 2 x m i i)
a i
longed to the material of Chapters 14
and 15, since in these sections we dealt
The moment of inertia with respect to
with complex numbers and analytic
the origin is
functions of a complex variable, the
b
present section is a continuation of
Chapter 16, which must be looked 1= x2o (x) dx-y 2 x\m
j
through before going on. We note also a
that the first example, which starts this
But with the aid
of the delta function
section, closely related to the content
is
it possible to include the separate
is
of Section 9.12, while the second exam-
masses in a generalized density func-
ple, which deals with Newton’s second
tion, which we denote byr} (x). It is de-
law (17.4.1), continues the discussion
fined by the formula
in Section 9.5 about the impulse and
the motion of a particle moving under a T] (3?) — O' (x) +S (
x —X i) •

short impulse, say, a blow. Therefore, i

Vdrore ^tuhynxg "hrese xwanrjJreb ’ib

advisable to go over the appropriate Indeed, if we consider the general dis-

sections of Chapter 9. tribution of mass along the rod, we can
First of all, we will show you how say that at the points where the loads
the delta function permits abridging are attached the density exhibits in-
and making the writing of the condi- finite jumps. With the aid of the new
tions in many problems more convenient. function r] (x) all the quantities can be
Let us consider a rod with a variable written in uniform fashion and more
cross section to which a number of sep- succinctly:
arate point loads of masses 1? 2 m m ,
b b
are attached (Figure 17.4.1). Let
etc.
the mass per unit length of the rod be
M= v\(x) dx, X = -J—
j
xt\ (x) dx,
a a
expressed by the function a (x). The
b

I = xh] (x) dx.

j
a

The concept of the delta function

permits combining continuously dis-
tributed masses and point masses in a
single general expression.
Another example of the use of the
nij m
delta function refers to the motion of
Figure 17.4.1 a mass point (or material particle). The
 ) ) t

17 A Application of the Delta Function 513

basic equation, it will be recalled, ir(t)\

expresses Newton’s second law:

m
= (17.4.1)
0 T t

(see Section 9.4). Recall the arguments

in Section 9.5 to the effect that the
action of the impulse is independent of
the law of variation of the force, pro-
vided that the force is sufficiently
brief. These considerations are similar
to what was said in Section 16.4 to the Figure 17.4.2
effect that the delta function can be
constructed out of a variety of functions
function of time t (Figure 17.4.2a):
<p (x) and concerning the conditions
when it is possible to replace a finite v = 0 for t <C x and v llm for t =
x. >
function cp (x) with the generalized,
The next step consists in determining
singular function 8 (x).
the path. From the fact that v dxldt =
If the concrete form of the function of
we get, by integrating both sides of
Eq. (17.4.3),
the force is not essential in the problem
about a blow, this means that F (t) f 0 for t <! x,
may be replaced by the delta function,
F (t /8 (t —
t), where x is the in- 1 — (* _T )
f° r t >% -
(17.4.4)

stant of the blow, and J — \^F (t) dt

(Why?) The graph of the path is shown
is the impulse. We
will carry out the in Figure 17.4.26.
integration of the equation of the mo- Characteristic of the curve x ( ) of
tion under a unit delta force formally the path is the salient point at t = x.
and according to all the rules. Let the Here again we are convinced that the
particle, prior to the blow, be at rest second derivative of a function having
at the origin: x 0 and v = dxldt 0 = = a salient point contains the delta func-
at t = —
oo. We
will also assume that tion: the function x ( t ) has a salient
the (unimportant) factor J in the ex- point. According to the equation of mo-
pression for the generalized force is of tion, the force is proportional to d^xldt
2
,

the order of unity. Then Eq. (17.4.1) and x (£) with a salient point was ob-
takes the form tained precisely for a force that was pro-
portional to 8 (t —
x), so that in the

i7 - 42 > case of a salient point, d^xldt 2 contains

<
a delta function, which is what we set
out to prove.
Integrating (17.4.2) and bearing in Now let us take the next step. The
mind that d 2xldt 2 =
dvldt, with v ( t problem of the motion of a body under
the velocity of the particle, we get a given force is linear This means that .

t if there are two solutions, x x (£) and

v (t) = 4r S(t-r)dt x 2 (t ), corresponding to two distinct
j
— OO forces, F^t) and F 2 {t), the sum of the
solutions, x ( t ) x ± (t) =
x 2 ( t ), is a+
= 4-0 (t-x), (17.4.3) solution that corresponds to the action
of the sum of the forces, F {t) =
where the function 0 {x) is expressed Fi{t) + F 2 {t). This
property (the su-
by formula (16.3.1) (see the graph of perposition principle) is a consequence
this function in Figure 16.3.1). Thus, of the simple fact that the second deriv-
the velocity is expressed by a step-like ative of a sum of functions is the sum
33-0946
 t t t

514 17 Applying Functions of a Complex Variable and the Delta Function

of the second derivatives of the func- F (t) At8 (t t). —

already know the We
tions: solution of the equation of motion for
d*x __ d 2 (x x +x 2) d 2x x . d 2x 2 the delta function. denote it by We
dt 2 dt 2 dT 2 572“ -
xi (£» t). The solution as the function
of time t depends on the instant t of
Taking into account that d xjdt = 2 2
application of the force. Recall that
Fx t)/m and d 2 x 2 /dt 2 = F 2 (£)/m, we
(;

get
f 0 for tCr,
dH __ £i_(£) . F 2 (t) _ F(t) x,(t, t) (17.4.4a)
dt 2 t?i '
m m ’
v m
which is what we set out to prove—
One of the strips into which the force
that the sum of the solutions, x x x2 + ,

describes the motion produced by the has been decomposed, from t to t + At,,
sum of the forces. is 8 (t —
the t)
coefficient with
Only one reservation is in order: so- ^(t) At. Thanks to the linearity of the
lution of the equation of motion depends equation, the solution for a force in
not only on the law of force but also on the form of such a strip is obtainable
the initial conditions, that is, on the by multiplying x x into that coefficient:
initial position and velocity of the giv- F (t) At x x ( t t). This is the solution
,

en mass. If we choose these conditions for the action of a single strip.

thus: x x == 0 and dxjdt 0 at t = — Nowlet us take advantage of line-

— 00 and ^2 0 and dxjdt 0 at = = arity and write out the solution for the
= —
,

t 00, then the sum of the solu- function F (t), which we consider as
tions, x , will also satisfy the same con- the sum of the strips. It is clear here
dition: x 0 and dx/dt =0 at t = = that the summation should actually be
— 00. replaced by integration. The result is
Let us now combine the reasoning 00

concerning linearity and the familiar x(t)= (t, t) F (t) dx. (17.4.5)
solution for the delta function so as to J
- 00
obtain a general solution for a force
that depends on time in an arbitrary At firstglance this formula is rather
fashion. We partition the graph of the strange: the ^-coordinate at time t is
force F
into strips of width At (Fig-
(it
)
expressed by an integral from 00 to —
ure 17.4.3). What does a separate strip +00 with respect to t, that is, the force
located between % and t At of time + enters into this expression at all in-
stants of time. Yet it is clear that the
t represent? Let us change the designa-

tions, leaving t for “current” time law of force subsequent to time t does
varying from —
00 to +00, whereas t not affect the preceding motion. How-
will refer to the given chosen strip. The ever, there is no error in the expression
height of the strip is F (t), the width for x ( t ). The properties of the func-
is At, and the area (i.e., the impulse) tion x L ( t) ensure reasonable proper-
,

is F (t) At. Since the strip is located ties of the solution. Indeed, x x ( t ), is ,

at t —
t, it is obvious that it can be zero for t <Z t. Hence, when integrat-
replaced by the delta function with a ing with respect to t we actually do not
coefficient equal to the impulse: need to take t >
t, since the integrand

is identically zero due to the factor

x x (t t) being equal to zero. Recalling
,

the expression (17.4.4a) for x x ( t ), we* ,

obtain
t
l
*(«)=-- (t-x)F(x)dr. (17.4.5a)
j
Figure 17.4.3
17 A Application of the Delta Function 545

This method of obtaining a solution

has very great general significance. To
summarize then: if for a linear system
we know a solution referring to the
action of the delta function, the solu-
tion referring to the action of an arbit-
rary function (F(t) in our example) is
obtained by simple summation (inte-
gration, to be more precise).
The ideas of linearity and addition
(the real term is superposition) of solu-
tions apply not only to such simple prob-
lems as the motion of a point; they
hold true in vast areas of mathematics, the introduction of the delta function
physics, and the natural sciences. It some- that clearly and succinctly explained
times happens that a system is very the essence of the Green function.
complicated and it is impossible to Examples of this nature abound in
solve the equations even for the most mathematics, for we know numerous re-
simple action of the delta function. A sults pertaining to tangent lines, areas,
solution corresponding to the delta and volumes that were obtained before
function can occasionally be obtained the invention of derivatives and inte-
experimentally. In other cases such a grals. The advance of science lies not
solution can be obtained from physical only in the attainment of new heights
reasoning (see Exercise 17.4.1). Then and new results but also in popularizing
linearity comes into play and we obtain and simplifying earlier derivations. The
the answer for an arbitrary acting func- aim of this book is precisely that: to
tion. The solution that corresponds to simplify the understanding of our clas-
the delta function (%(£, t) in our exam- sical heritage, the fundamentals of high-
ple above) is so important that it has er mathematics.
a special name, the Green function of
the problem. Curiously enough, the
English mathematician George Green Exercises
(1793-1841), for which the function was
17.4.1. Consider a string held taut by
named, lived in the 19th century and a force kand with ends fixed at points x 0 =
quite naturally knew nothing about the and x —l. Regarding the deviation as small,

delta function. 1710 But it was only determine by the parallelogram-of-forces law
the shape of the string under a unit load at
point x = x x (Figure 17.4.4a). Obtain the
17 10
•
The author of excellent papers in formula for deviation of the string under a
mathematics and mathematical physics, Green force distributed along the length of the string
came from a family so poor that he did not via an arbitrary law / (x) (Figure 17.4.46),
receive any education at all; it was not until 17.4.2. Find the motion of a pendulum un-
comparatively late in life that he became der a force expressed by the delta function,
acquainted, through his own efforts, with that is, solve the equation m cPx/dt 2 ) =
mathematics and physics. At the end of his — +
kx — 6 (t t) provided that x
(
= 0 and
life, though, chiefly thanks to support from =0
dx/dt — — at t oo. Using this solution,
the influential William Thomson (Baron find the motion of the pendulum under a
Kelvin), Green was offered the post of pro- force dependent on time via an arbitrary law.
fessor of mathematics at Cambridge Univer-
sity.

33 *
 t

Conclusion. What Next?

Higher mathematics, or, to be more to follow, let us briefly state a common

exact, differential and integral calcu- feature of all the problems we have dealt
lus, makes it possible to solve a large with up to now. These were problems
class of problems that do not lend them- involving the motion of a single par-
selves to solution by the methods of ticle in mechanics or the flow of current
elementary mathematics (arithmetic, in a circuit. We dealt with functions of
algebra, and geometry). Of tremendous one variable (time). The number of
importance is the very formulation of functions was one (current as a function
such new concepts as instantaneous of time) or two (the position of a body,
velocity, acceleration, impulse. These x =x (t), and the velocity of the body,
notions (and numerous others in diverse v = v (; )).
fields) are formulated exactly only in Quite naturally there follows the
the language of derivatives and inte- purely quantitative generalization: prob-
grals. lems involving the motion of two bod-
The knowledge you have gained in ies, three bodies, etc., will ultimately
reading this book constitutes but a lead to problems of the motion of a gas
small part of the whole science of mathe- or a liquid. But one gram of hydrogen
matics, to be exact, a small part of consists of 3 X 10 23 molecules, hence
those divisions of mathematics that find 3 X 10 23 separate bodies. It must be
application in the natural sciences and clear at this point that the old method
technology. In the Preface we said is now useless and new methods are
that a cultivated person should have a needed. Not only is it impossible to
general picture of mathematics, irre- solve 3 X 10 23 equations, there is nei-
spective of his or her field of interest. ther paper nor time enough to write
The three parts of our book provide them down. Even writing them down is
enough general information for that out of the question because we can never
purpose. But if your field is connected know either the exact number of mole-
with engineering, chemistry, or (espe- cules or their initial positions, which
cially) physics, you will need more. Here would then yield the initial conditions
we wish to outline in brief the fields imposed on the equations.
of physics and the associated divisions To our aid come new sciences that
oi x lLmAvsi J ikeo
ly study in the future. integral calculus. These are hydrodynam-
Note that up till now the exposition ics and gasdynamics, which did not
has been that of a textbook, and if you come into being until the 19th centu-
have put your mind to the matter at ry. Both differ substantially from the
hand, you will have mastered the mate- mechanics of a single point in the for-
rial in all its details. What now follows mulation of problems and in the meth-
is a very short outline of the difficult ods used to solve them (compare Sec-
problems that lie ahead. The style is no tions 9.14 and 9.15 and Section 17.3).
longer that of a textbook. We do not Here we are not interested in the actual
hope to explain the content of complex number of molecules of a gas contained
mathematical theories on so few pages, in a volume. Instead, such notions are
of course. We wish merely to give the introduced as the distribution of the
reader a general impression of the prob- density of the gas in space, p = p (x, y, z)
lems of these theories to arouse inte- (the mass of the gas per unit volume), GA
rest in them. Some of the books listed
G#1
in the Selected Readings will help you Strictly speaking, the introduction
of the notion of density at a point in space,
to obtain a broader view of mathema-
p (x, y, z ), requires an operation similar
tics and of mathematical physics. to differentiation. This quantity is defined
For a better understanding of what is as the limit of the mean density of the gas
 Conclusion. What Next? 517

the pressure of the gas, p = p (x, y, z), But there are other physical theories,
and the velocity of the gas at different primarily the theory of electromagne-
points in space. What is more, all these tism , where such an approach is im-
quantities depend on time, t so it is ,
possible.
natural to write p p (x, y, z, t). = Consider two point charges at rest.
Thus, from problems of several func- The force acting between them depends
tions of one variable we pass to func- on their position (the distance r between
tions of several independent variables. them). This would seem to be a prob-
Accordingly, in setting up the equa- lem involving six functions (the coor-
tions that describe the motion of a gas dinates y u and z 1 of one charge and
and other characteristics, we encounter the coordinates x 2 yz, and z 2 of the
,

partial derivatives with respect to time other) of one variable (time). To a first
and the spatial coordinates, for ins- approximation, the motion of the charges
tance, dpldt, dpldx, dpldy, and dp/dz ,
changes but
little —
one merely has
where, say, to take into account the magnetic in-
teraction that appears between the
dp _ Hty1 Pl£i y±AHi z
i £hl£i£i h z
i £) charges, an interaction that depends on
** Ay™ the velocities of the particles.
An
extremely important division of An extremely important factor— one
mathematical physics is the investi- demanding a fundamentally new ap-
gation of partial differential equations ;
—
proach is the existence of a lag in the
we gave a very rough outline of them interaction due to the propagation of
in Section 10.8. The equations describe the interaction with a finite speed (the
the motion of liquids, gases, and sol- speed of light). The action of one charge
ids, the propagation of heat in various on the other at time t depends on the
media, the diffusion of atoms and mole- position (and velocity) of the first charge
cules. They are so important that they at some earlier time t — t, which
are often called equations of mathe- lags behind t by a finite amount T =
matical physics . r/c, where c is the speed of light. What
Another point that must be mentioned takes place in this interval of time?
is that velocity is a vector quantity, In a vacuum the space between the
v = v (x, y, z) or even v = v (x, y z, t), ,
charges contains an electric field and a
which means that at every point in a magnetic field. At each point in space we
gas (or liquid) the magnitude and di- specify two vectors, E and H, called
rection of the velocity are specified. In the electric field strength and the mag-
other words, we can say that three com- netic field strength , respectively. The
ponents of the vector v, namely, v x v y , ,
magnitude and direction of each vector
and vz are given and that these compo-
,
depend on the presence and motion of
nents depend on the point in space and charges. Small (“test”) charges placed at
on time. This leads to more complica- certain points in space provide the pos-
tions (or simplifications) in problems of sibility of measuring the fields E and
hydrodynamics and gasdynamics; how- H at the given points. But there is
ever,we will not go into them here. still more to the theory of electromag-
In all these theories
it is possible, at netism. In the vacuum, where there are
least principle, to continue regard-
in no charges, the electric and magnetic
ing the system as consisting of separate fields act on each other: a change (over
particles and as being described by time) in one field is related, via Max-
many functions of one variable (time). well’s equations, to the spatial deriv-
atives of the other field. The time de-
contained in a small volume (the mass-to- rivative of the magnetic field generates
volume the volume becomes ever
ratio) as
a circular electric field, while the time
smaller. Changes in the density due to one
or two molecules leaving or entering the derivative of the electric field plays the
volume do not interest us here. same role as electric current and gener-
518 Conclusion. What Next?

ates a magnetic field. The result is a hydrodynamics and electromagnetism,

neat pattern: charges generate fields and also the similarity of the equations
at the points where they exist, and the of the two theories, remains; however,
interaction of the fields carries the in- the physical meaning of the electro-
formation concerning the position and magnetic theory has proved to be differ-
motion of the charges throughout the ent and does not reduce to mechanics.
entire space. After prolonged and intense debates
Decisive in the development of this scientists completely abandoned the
theory is a consideration of the two idea of any ether.
vectors E and H, or the six quantities When speaking of a mathematical
E x ,
y
E z , xE y ,
andH H
, z ,
H
as func- , theory, one must not only speak about
tions of four quantities: the three coor- the statement of the problem and the
dinates x, y, and z and the time t. But initial equations but also about the na-
specifying these functions at a certain ture of the results.
time amounts to the fixing of an infini- We can name two types of solutions
tude of their values at all points of for partial differential equations. One is
space. The theory is more difficult than characteristic of phenomena that devel-
the theory of the motion of one or sev- op in a limited volume: these are natu-
eral particles, but the results are ral oscillations with definite frequencies.
richer. A body of a given shape has a certain
Mathematically, the theory of the set of frequencies.
electromagnetic field is a theory of par- Recall the pendulum and its definite
tial differential equations and in this frequency of oscillation. If an external
respect is similar to the theory of elas- (periodic) force F acts on the pendulum,
ticity, acoustics, and gasdynamics. The we get the characteristic phenomenon
only difference is that in the latter case of resonance when the frequency of F
the equations are arrived at via an is close to the (natural) frequency of the
idealization: in speaking of the density pendulum. All these physical ideas fit

of a gas we abstract ourselves from the the theory of ordinary differential equa-
separate molecules and only in this tions: m
d^xldt 2 ) (i kx F(t).= —
In +
approximate sense can a gas be consid- the theory of partial differential equa-
ered as a continuous medium character- tions it appears that a body has many
ized by a continuous function p(a:, y, z t) , frequencies and behaves like a set, or
(the densitythe gas). An electric
of collection, of many pendulums with
(or, more precisely, an electromagnet- distinct frequencies, whence there are
ic) field is actually a continuous func- many resonances. c 2 -
You can verify
tion of the spatial coordinates and this at once if you have a piano at home.
time. Depress one of the keys slowly and
Hydrodynamics, the foundations of soundlessly, so as to release the string
which were laid in the 18th century by without striking it with the hammer.
the works of D. Bernoulli, J. d’Alem- Now strike the other key sharply and
bert, and L. Euler, prepared the mathe- listen to the response of the free string.
matical tools for the electromagnetic The other type of solution of partial
theory worked out in the second half of differential equations has to do with
the 19th century by James Clerk Max - matter or fields that fill all space (pro-
well (1831-1879). No wonder, then, pagation of waves). These are the waves
that at first attempts were made to of radio and light (in the electromagnet-
transfer the ideas of mechanics to the ic theory) and sound waves in elastic
electromagnetic theory. A
special sub- media. Waves have the remarkable
stance, called the ether , was hypothe- property of being able to carry infor-
sized as being responsible for electric
and magnetic phenomena. We know c * 2
If you wish to understand this aspect
that the mathematical analogy between in detail, go back to Section 10.8.
 —

Conclusion What Next?

. 519

mation: pressure or an electric field at of attributes (resonance, for instance)

one point (near the receiver) as a func- that are characteristic of mechanical
tion of time proves to be similar to vibrations.
the curve of that same source quantity The power of mathematics is connect-
(near the transmitter) as a function of ed largely with the repetition of one
time. and the same mathematical scheme
It is possible to construct the solu- applied to a large number of phenomena
tions of equations describing the direct- of different kinds. Mathematics might
ed beam of a searchlight or laser. The be described as a rather limited set of
beam of a searchlight and the jet of “keys,” each of which unexpectedly
water from a hose are strikingly simi- opens up numerous doors that appear
lar. Knowledge of the properties of so- to be different (compare, for instance,
lutions of different kinds of problems Sections 2.1, 2.2, and 2.5). This cir-
and analogies between phenomena de- cumstance is connected with the fact
scribed by similar types of equations that nature’s arguments are simple
have always been extremely important there are no special intricacies and —
in the development of physics. there are not so many simple mathemat-
The method of mathematical model- ical constructions. The fantastic gene-
ing consists in finding a mathematical rality of mathematical methods has
scheme so close to the phenomenon un- always amazed scientists. c 4 *

der consideration that an examination For a long time atomic spectra were
of the scheme can replace a study of a mystery to physicists. It was not so
the phenomenon. The scheme might be much the specific laws and numerical
either an ordinary differential equation values of the frequencies but the very
or a system of such equations. Some- fact that one and the same atom emits
times it is a partial differential equation or absorbs (via resonance) the oscilla-
with initial and boundary conditions of tions of several distinct but quite defi-
one kind or another (see Section 10.8). nite frequencies. The similarity to the
In other (and more sophisticated) ma- oscillations of elastic bodies suggested
thematical constructions we have to the formulation of the equations of
deal with the notion of probability on ,
quantum mechanics . Likewise, the si-
which we will dwell below. It is essen- milarity between a stream of particles
tial that the model should present a and solutions of equations particular
sufficiently accurate description of the to waves found its application in quan-
given real-life phenomenon. c 3 If, in so -
tum mechanics, where it gave rise to
doing, we come to an earlier-studied particle-wave dualism. For instance,
mathematical scheme, we can immedi- light can be regarded as a wave (the
ately transfer to the new case all the viewpoint of Christian Huygens) and
findings pertaining to it. For example, at the same time as a stream of parti-
from the fact that the mechanical vi- cles (or corpuscles) of light, or photons
brations (see Chapter 10) and electro- (Newton’s concept; see Section 12.6 and
magnetic oscillations (see Chapter 13) subsequent sections). In pre-20th-cen-
are described by differential equations tury physics these two approaches
of the same type it follows that the seemed to contradict each other, and a
theory of electric circuits has a number
G * 4
See, for instance, the following arti-
cles by
several distinguished physicists:
c * 3
The words
“sufficiently accurate” em- E.P. Wigner, “The unreasonable effectiveness
phasize the (that is, making
idealization of mathematics in the natural sciences” in
a rough approximation of the phenomenon Symmetries and Reflections, Indiana Univer-
under study by disregarding details that are sity Press, Bloomington, 1967 (reprinted by
secondary and do not interest us at the mo- Ox Bow Press, Woodbridge, Conn., 1979),
ment) that we invariably encounter when pp. 222-237, and G.N. Yang, “Einsteims
passing from physical reality to the mathe- impact on theoretical physics” in Physics
matical scheme describing it. Today 33, No. 6, pp. 42-49, 1980.
,
 520 Conclusion . What Next?

choice had to be made.Modern quan- (In this case the surfaces cp constant =
tum mechanics shows how and in what are half -planes and the surfaces 0 =
sense both points of view are correct constant are cones (why?).) Coordinates
and supplement each other. can also be introduced by countless
Wehave already presented several other methods.
examples pertaining to the theory of To summarize, then, we can introduce
equations of different types. curvilinear coordinates £, rj, and £,
An example of a different kind is and with a lot of effort, agonizingly,
offered by complex numbers and func- learn to compute point-to-point distances
tions of a complex variable (see Chap- and other quantities with the aid of
ters 14, 15, and 17). Complex numbers, these new coordinates.
which originally arose from the solu- At first glance this is a dull and to-
tion of algebraic equations, were for a tally useless effort. One must possess a
long time regarded with much mis- peculiar bent of mind to see beauty in
trust.However, progress in hydrodyna- overcoming difficulties and in devel-
mics and gasmechanics turned out to be oping a clumsy theory with arbitrary
closely connected with these unusual coordinates of the most general na-
“numbers.” The theory of functions of a ture. c - 5

complex variable was evolved by Augu- Elaboration of the theory of curvili-

stin Cauchy and Georg Riemann and near coordinates in Euclidean space
further developed by Karl W
eierstr ass might appear to be purely an achieve-
(1815-1897) just when it was needed to ment method. However, the general-
of
make further scientific and technical ized coordinates£, rj, and £ are equally
advances possible. The rise of aeronau- convenient (or equally inconvenient) in
tics at the beginning of the 20th cen- describing ordinary space (in which
tury was based in large measure on cal- Euclidean geometry holds true) and
culations of fluid flow by the methods “curved” space. The rectangular coor-
of “complex analysis.” One of the pio- dinates x y, and z are convenient for
,

neers in this field was Nikolai Zhukov- ordinary space but are absolutely un-
sky (1847-1921), the distinguished Rus- suitable for a description of curved space.
sian investigator in the field of me- These coordinates do not give even
chanics. a hint of the possibility of the existence
Geometry offers another marvelous of any other kinds of space.
example of the development of mathe- The transition to curvilinear coordi-
matics. nates, which seemed to be such a need-
Ordinary experience teaches us that less complication, actually prepared us
in space it is convenient to introduce for a vast range of spaces whose very
three coordinates: x, y, and z. Any fur- existence was totally unknown to us.
ther complications would seem to be Then it turned out that the force of
superfluous, “a trick of the devil.” Yet, universal gravitation is linked up with
coordinates £, rj, and £ can be intro-
duced in a different way so that the G 5
-
Such was the mind of the great Gauss,
condition £ =
constant corresponds to who worked out a “geometry in curvilinear
some curved surface (whereas x = coordinates.” Displaying a profound under-
constant for arbitrary y and z is the standing of the applied significance of mathe-
equation of a plane perpendicular to the matics (for example, Gauss arrived at the
idea of curvilinear coordinates from his
x axis). For instance, it is possible to
work in geodesy), he also possessed an insati-
introduce into space the three coordi- able curiosity concerning the secrets of the
nates p, cp, and #, where p = OM
is the world of mathematics, which prompted him
distance between the variable point M to devote hours and days to arduous cal-
culations in the hope of some new results (say,
and the origin <9, while cp and d are the
the conversion of common fractions to deci-
geographical coordinates (longitude and mal fractions with hundreds of figures in the
latitude) on the sphere p =
constant. period of the fraction).
 —

Conclusion. What Next? 52 1

the fact that space is somewhat curved. the idea of multidimensional spaces
True, this “somewhat” had to do with entered into theoretical physics. For
the conditions here on the earth and in instance, phase space for the simplest —
the solar system. In certain phenomena case of a moving point, the phase space
of a larger scale (catastrophic explosions is six-dimensional with coordinates^:,,

of stars, evolutionary processes in the y 2, x

,
y' and z' (x =[ dxldt y' —
, , ,

Universe), space may turn out to be dyldt z! —

dzldt ), which are the spatial
,

highly curved, and here one must inevi- coordinates and the projections of the
tably go over from conventional Eucli- velocity vector of the point. 0 7 Today,
dean geometry to non-Euclidean. Me- too, in theoretical physics we often have
taphorically speaking, the creation of to deal with multidimensional worlds
non-Euclidean geometry, a purely ma- that are simply impossible to imagine;,
thematical achievement, c 6 was a flash
-
finally, we have the ultimate monstros-
of lightning, and it was followed by a ity —
an infinite-dimensional world each)
—
clap of thunder the creation of the point of which has an infinitude of coor-
general theory of relativity, or the geo- dinates (expressed as numbers— and
metric theory of gravitation. Today we are lucky if the numbers are^real
physics is being invaded by completely and not complex). Instances of such
new kinds of geometry, which at first infinite-dimensional spaces are found in
glance seem quite removed from any quantum mechanics. It was most for-
reality. tunate for physicists that the famous;
Back in the 18th century, d’Alem- German mathematician David Hilbert
bert drew attention, in his article (1862-1943) evolved a theory of infinite-
“Dimensionality” (in the famous En- dimensional spaces just before the ap-
cyclopedic written together with Denis pearance of quantum mechanics.
Diderot (1713-1784)), to thefactthat our In recent years topology one of the- ,

world is essentially four-dimensional newest and most abstract sections of

besides the three spatial coordinates x, geometry, which took its start in the-
y, and z of a point, it is also necessary, works of Henri Poincare, has found unex-
in order to distinguish events taking pected applications. Topology deals,
place in the world, to know the time, t. exclusively with the “roughest” prop-
Moreover, the connection between the erties ofgeometric figures. For instance,
coordinates x, y, z, and t of the four- in topology a sphere does not differ
dimensional physical world later proved from a cube, but a doughnut (torus) does
to be more complicated than New- because it has a hole. Topology ana-
ton and d’Alembert had originally imag- lyzes these “rough” geometric properties
ined; in particular, in this “world” it is with great thoroughness and depth. All
impossible to distinguish unambiguous- of a sudden this not-so-trivial analy-
ly between the spatial coordinates x, y, sis aroused the interest of physicists. c 8 -

and z and the temporal coordinate t

G * 7
The representation of a physical pro-
(this is the essence of Einstein's special
cess phase space of a (mechanical)
in the
theory of relativity). Later (beginning,
system is known as the “ phase portrait ” of the
essentially, with the works of Lagrange, process. For instance, the phase portrait
a junior contemporary of d’Alembert) of the uniform motion of a point along the z
axis, z at = r
+
6, is the straight line z' a =
in the phase plane (z, z ) of the moving point
0,6
Again, how amazing that the non- (here z' =
dzldt is the velocity of the point).
Euclidean spaces that were so necessary for The phase portrait of harmonic oscillations,,
progress in physics came into being at just that is, oscillations characterized by a con-
the right time in the works of mathematicians stant total energy E kx 2 / 2 2 =
(x') l 2, is. + m
(Nikolai Lobachevsky (1792-1856) of Kazan, described by a point in the phase plane (x, x')
Russia, John Bolyai (1802-1860) of Hungary, moving along an ellipse E constant. =
0,8
and Carl Gauss; then Georg Riemann and, We
would like to draw the reader’s-
later, Felix Klein (1849-1925) of Germany attention to the excellent, though by m>
and Henri Poincare (1854-1912) of France). means easy, textbook by B.A. Dubrovin*.
322 Conclusion . What Next?

.And no one can say what other esoteric We often meet phenomena of this
-approaches to space will be concocted kind in the theory of combustion and
by physicists of the future. G 9 *
explosion, where the conditions imposed
At the intersection of (multidimen- on the chemical reactions are deter-
sional) mathematical analysis and to- mined via algebraic equations. The ex-
pology there lies a new theory, the ternal conditions enter the problem as
theory of catastrophes. (Catastrophe parameters, but the solution of the
-theory is the invention primarily of the equations may be a discontinuous func-
Trench mathematician Rene Thom tion of the parameters. Discontinuities
{b. 1923).) It immediately gained (per- are therefore specific features of the so-
haps excessively wide) recognition in lutions. Catastrophe theory shows that
^connection with the many different these features can be classified by di-
possibilities of its use in the natural, viding them into several “rough” (topo-
humanitarian, and social sciences. G 10 *
logical) types corresponding to the char-
The theory poses the question of the acteristic phenomena described by a
^conditions under which smooth and ge- given equation.

^
nerally well-behaved analytic functions

^explosions, that is,

c iiwilra*

a discontinuity in
*
In the study of nature, a diversified
i
desired: •
+

mathematical difficulties, the mastering

'xf *

the final result. Simple examples of of the mathematical apparatus, physi-

this kind were known long before the cal intuition, boldness of conception,
mame “catastrophe theory” appeared. experimentation, and mathematical mo-
Let a quantity a be given as a smooth deling. All these combined make it
:function of another quantity, b or ,
possible to advance science.
a =
/ (b), and suppose that a can be Modern physics would be absolutely
fixed and then b can be observed or inconceivable without the concept of
measured. The smoothness of the func- probability , which is quite different
tion a f(b)= by no means guaran- from anything discussed in this book.
tee tbit ejmvtfhnf&jb ALL Lb a, hue oA mnjiTjfviRfL
tion b = For instance, if a
cp (a). = —
above Newton’s second law mx F =
f ( b) has a amaximum
a max at = and its particular case, the law of iner-
b =
b m then for a,
a max there are < tia
=
(see Section 9.4), Boyle’s law
two values of b near b ni whereas for ,
pV constant and its generalization,
a >
a max there is not a single value of van der Waals’s law (see Section 7.3)—
b. Hence, b experiences a discontin- were of a dynamical nature, that is,
uity as a varies smoothly. they enabled us to make a precise pre-
A.T. Fomenko, and S. P. Novikov, Modern diction of the phenomenon: when the
*Geometry Methods and Applications, 3 parts,
:
volume occupied by a gas is halved, its
.Berlin, Springer, 1984, which is intended pressure doubles (Boyle’s law); if the
largely for theoretical physicists. motion of a body takes place in the
G 9 An
'

interesting but not simple booklet absence of any forces acting on it, its
which deals with the overall question of the
relationships between physics and mathematics
velocity in future will be exactly the
iromthe present-day viewpoint is Yu.I.Manin’s same as it was at the start; and so on.
.Mathematics and Physics Birkhauser,
,
The laws of the falling of, say, a ran-
Boston, 1981. The two papers mentioned domly tossed coin are altogether differ-
in footnote G.4 deal with the same question.
G,i0 ent. Here all one can say in advance is
See, for example, the article by
E.G. Zeeman “Catastrophe theory” in Scientific
that when a coin is tossed into the air
American 234, No. 4, 1976, pp. 65-83, and
,
many times, it will fall heads up in
"V.I. Arnold’s book Catastrophe Theory Spring- ,
approximately half of the cases, that is,
er, Berlin, 1984. A thorough treatment of the the probability of its falling heads up is
subject is presented in Catastrophe Theory and
equal to 0.5 (this result is guaranteed by
Its Applications, Pitman, London, 1978, by
T. Poston and I.Stuwart, which also con- the condition that the coin is true, that
tains many expressive examples. is, both sides are absolutely equal). The
Conclusion. What Next? 523

physical laws by which one cannot illustrates the concepts of thermodynamic

c 13
irreversibility and fluctuations.
*

predict the exact outcome of an exper-

iment but only the probability of one The case of two vessels is an idealization,
of course. Typical problems deal with the
outcome or another, that is, the fre- movement of particles in space, where we seek
quency with which the outcome is re- the law that governs changes in the particle
peated when an experiment is carried distribution depending on the coordinates and
time. When it is a question of large particles,
out many times under the same condi-
visible under a microscope, we speak about
tions, are called statistical laws , and Brownian motion. It is characteristic that
the branch of mathematics that deals the physicists who studied Brownian motion
with them is known as the theory of (A. Einstein, M. Smoluchowski, A. Fokker,
and M. Planck) turned to probabilistic pro-
probability GAi .
cesses long before the problem attracted the at-
tention of mathematicians (A. Kolmogorov,
The example of a coin being tossed many
A. Khinchin, W. Feller, Paul P. Levy, and
times is the simplest, almost a trivial, instance
others). For this reason the main equation
of ^probabilistic process characterized by the
of (continuous) processes allied to the “toy”
accidental nature of the outcome of its sepa-
process of transferring particles we have ex-
rate stages (in this case, separate tossings).
amined is called the Einstein- Smoluchowski
Here is a still more typical example of the
equation or the Fokker-Planck equation or
same kind, used by physicists as a simple mo- the Kolmogorov equation.
del of the process of the diffusion of gases (the
0,12
model of Paul and Tatiana Ehrenfest ).
Let us take a vessel divided into two parts, It is typical that present-day physics
A and B, by a porous membrane. It contains N pays careful attention to the statistical
particles (gas molecules, say) distributed in
laws of nature and, hence, makes exten-
some way between A and B at each moment
:

of time, one of the particles is chosen at ran- sive use of the mathematical tools of
dom and transferred from its part of the vessel the theory of probability. For instance,
to the other. We
can pose the question of how an aspect of the transition from the clas-
many times, on the average, a transfer must sical mechanics of Galileo and Newton
he made so that the initial distribution, 100 par-
ticles in A and 0 particles in B becomes 50 to quantum mechanics is the replace-
,

particles in A and 50 particles in B. It turns ment of the “dynamical” world of the

out that about 140 transfers are needed. The rationalists in the 17th to 19th centu-
reverse process is also possible (at least in ries by the random world of modern
principle): with an initial distribution of 50
particles in A and 50 in B, random transfers physics of the microcosm, in which one
•can lead to 100 particles in A and 0 in B , but cannot in principle speak of the exact
this requires 2 100 ~
10 30 transfers. This example trajectory of a particle (say, an elec-
tron) but only of the probability of
Cl1 The most elementary concepts of the finding the particle at one or another
theory of probability are explained for begin- point in space at a certain time.
ners in B.V. Gnedenko and A.Ya. Khinchin,
Finally, another relatively young
An Elementary Introduction to the Theory of
Probability W.H. Freeman, San Francisco, branch of mathematics that has acquired
,

1961, in K.L. Chung, Elementary Probability primary importance for physics is

Theory with Stochastic Processes 3rd ed.,, the theory of groups This theory stud-
.

Springer, New York, 1979, and in F. Mosteller, ies the “degree” of symmetry of physi-
R.E.K. Rourke, and G.B. Thomas, Jr.,
Probability with Statistical Applications, 2nd cal or other objects. It is clear, for exam-
«d., Addison-Wesley, Reading, Mass., 1970. ple, that a square (Figure C.la) is more
Two other good books are: W. Feller, An symmetrical than an isosceles trapezoid
Introduction to Probability Theory and Its (Figure C.16) and that the latter is more
Applications 2 vols., Wiley, New York,
,

1968, 1971, and J. Neyman, First Course in

symmetrical than a trapezium (Fig-
Probability and Statistics Holt, Rinehart,
,
ure C.lc). The exact meaning of these
and Winston, New York, 1951. statements is as follows. It is clear that
c 12 Paul (or in full, in Russian, Pavel Si-
*

gizmundovich) Ehrenfest (1880-1933) and Tati- C. 13 For further details and proofs of the
ana Alexeyevna Ehrenfest-Affanasjewa (1876- results we have just stated see, for instance,
1964) were prominent 20th-century physicists J. G. Kemeny and J. L. Snell, Finite Markov
who worked in Russia (and, later, in the Chains 3rd printing, Springer, New York,
,

USSR) and the Netherlands. 1983, Section 7.3.

 ?

524 Conclusion . What Next

accordance with their inherent “sym-

metry groups” and had no idea whatever
of the possible physical applications of
the concepts he had evolved. What is-
more, at the beginning of the 20th cen-
tury the prominent English astrophysi-
cist James Hopwood Jeans (1877-1946)
energetically protested against the in-
clusion of elements of group theory in
the course of mathematics for students
of physics because he thought physicists
would never have any use for it. Today,,
when the theory of groups has acquired
such great significance in physics,.
Jeans’s statement is often quoted as an
example of a prediction that failed.
When scientists speak of the role
that group theory plays in physics today,
what they have in mind mainly is its
Figure C.l application to the fundamental ques-
tions of the structure of matter, first and
foremost to the theory of elementary
the symmetry group of the square, a
particles. But the first practical appli-
group consisting of all the rotations
cations of the ideas of symmetry were in
that transform the square into itself, is
crystallography. After all, crystals are
broader than the symmetry group of
highly symmetric bodies. It was found
an isosceles trapezoid or that of a tra-
that crystals could be classified on the
pezium. The symmetry group of a square
basis of their symmetry, and that such
contains symmetries with respect to
a classification sheds substantial light
all the diagonals and the midlines and
on all their properties. A big event at
also with respect to rotations about the
the end of the 19th century was the
square’s center through 90°, 180°, and
270°, while the symmetry group of an
enumeration of all possible systems of
isosceles trapezoid is limited only to
crystal symmetry —there proved to bo
230. These are the Fedorov groups , so
symmetry with respect to the midline
MN (see Figure C.l b); the symmetry
named in honor of Evgraf Fedorov
(1853-1919) who, along with A. M. Schon-
group of a trapezium (Figure C.lc) is
flies, a German, and V. Barlow, an En-
even more limited, consisting merely
glishman, discovered the crystallogra-
of a single rotation through 360°, which
phic groups.
leaves all the points in place and does
Physicists also found use for the
not contain a single true movement!
The theory of groups was created by Shubnikov groups named after the So-
v iet cry st al lographer Aleksei Shubn ikov
a brilliant young French revolutionary
(1887-1970), who made a study of the
named Evariste Galois (1811-1832), who
was killed in a duel when he was only
symmetry of colored (for instance, black
and white) ornaments. His theory also
20. c 14 (The duel had evidently been
takes into account the colors of separate
provoked by the police.) Galois devel-
sections. The point is that other physi-
oped the ideas of this theory as applied to
cal properties of objects (for example,
algebraic equations in the early 1830s.
electric charge) can play the role of
He suggested classifying equations in
color (black and white).
0,14
Read Leopold Infeld’s book Whom the
An enumeration of the Fedorov (and
Gods Love : The Story of Evariste Galois ,
Shubnikov) groups required a thorough
McGraw-Hill, New York, 1948. analysis of the very concept of “a sys-
 0 a

«Conclusion . What Next? 525

tem of symmetry,” and an understand-

S'-
ing of the meaning of the operation of 9 A=<j
2
A
composition of rotations, which substi- 1

tutes one rotation of a crystal for an

1

equivalent of several rotations. 0 i X

1

It is clear that each movement of a

crystal or any other body that trans-
—

>
forms it into itself constitutes a cer- (a)
tain transformation in space; we can y\ \
pA=yr 4A
•call a composition, that is, a sequence
AfrtA
0^ /
_ /
of two transformations a and P (first (3 '''.r
and then a), a product: y ap. For = o\X
4
y^ x
example, the composition of two reflec- /
/ ^
"o _
tions a with respect to the straight line
/
Ar yr 3A
Ox or of four rotations n through 90° /\2 =X A Z

constitutes an identical transformation (b)

c, which returns each point to its place:
<y
2
= 8 and Jt 4 =
e (Figure C.2). Thus
Figure C.2

we arrive at the “arithmetic” (or “al-

gebra”) of symmetries and the possibil- whereas the product na (first the reflec-
ity of utilizing the precise tools of tion, then the rotation) is equivalent
mathematics in a “symmetry calculus.” to an entirely different operation—

This calculus of transformations (or reflection with respect to the straight

symmetries) is the subject of the theory line x —y =
0 (Figure C.3&).
of groups. A peculiar situation arises All these considerations played a sub-
here in that aP may not be equal to stantial role in the study of the sym-
metry of crystals. Recall what was said
Pa, or that operations in symmetry cal-
culus may be noncommutative. c lb For -
in Section 14.1 about the origin of com-

instance, the product an of a rotation plex numbers. Counting began with the
through 90° and a reflection with re-
natural numbers 1, 2, 3, .. Later . .

spect to the x axis (first the rotation, there was a need for negative numbers,
then the reflection) is equivalent to a — — 2, — 3,
1, and so on, and also zero;
single reflection with respect to the at the same time, fractions came into
straight line x y + =
0 (Figure C.3a),
y t

v ArJTA <

C* 15
The first system of noncommutative
numbers, called quaternions was discovered ,

by the famous Irish algebraist, astronomer,

and physicist William Rowan Hamilton (1805-
X
1865). His quaternions differ from the com-
mon complex numbers x
they contain three imaginary units,
and
iy +
2 in

so that the general quaternion is w =

u -f- xi —{— yi -f- zk, where w, x y and z
that
7
=

,
,
!

V
A2 =(j/TA
0

Ya)
4 >
x
n

are real numbers. The square of each imagi- y

nary unit is —1 but they are noncommuta-
tive, so ij —
k but ji —
,

, k. —
Hamilton’s
calculus of quaternions was the first version of
vector calculus ( he called a quaternion without
imaginary units, w Q =
u, a scalar and used
the term vector for a purely imaginary quater-
—
a
n. 1
x
nion, xi yj + +
zk ). The renowned
Treatise on Electricity and Magnetism (1873)
A
di frames ™iei K ihaxweii was written ’in tne
J
1

language of quaternions, to which Maxwell '(b)

was accustomed, and not in the language of
vectors so familiar to us. Figure C.3
 x ?

526 Conclusion. What Next

being and also irrational numbers, such of symmetry, regarded broadly, entered
as ]/ 2. A
further generalization of the theoretical physics. And as physics
number concept led to complex num- brought to light more and ever newer
bers; only after the introduction of these particles, the idea of symmetry in the
unusual numbers was the theory of properties of particles and the applica-
equations complete (in particular, the tion of the theory of groups became in-
elementary theory of quadratic equa- creasingly significant. The theory grad-
tions). In Chapters 14, 15, and 17 we ually became more complex as rota-
showed how fruitful the generalization tions were examined in an imaginary
of the number concept was, especially multidimensional space and account was
when combined with the derivative and taken of particle charge and spin. Today
the integral. the approach to physical objects from
However, mathematicians could not the standpoint of their inherent sym-
bring themselves to challenge the prin- metry is perhaps the prime method that
ciple ofcommutativity in numbers— physicists use as they try to analyze the
for allnewly introduced numbers it great variety of elementary particles
was always a + b = b cl and ab =
+ brought to light in the numerous exper-
ba . What
Galois accomplished should, iments and theoretical works of scien-
therefore, be regarded as a most funda- tists in the second half of the 20th cen-
mental advance: he was the first to con- tury.
sider groups whose elements are non- Modern physics makes extensive use
commutative; in general, in the “arith- not only of the theory of discrete ( cry-
metic of Galois” ab =^= ba. In everything stallographic ) groups , which, in a
else, the theory of crystallographic way, copies elementary arithmetic (non-
groups is close to ordinary arithmetic: it commutative arithmetic!) but also the
has a unit element such that the product theory of continuous groups , whose ori-
of this element with any member of gins lie in algebra and mathematical
the group, in either order, is that same analysis. Along with the question of the
member (in the “calculus of symmetries” group of operations that transform the
the role of the unit element is played by square in Figure C.la (or a crystal)
the “identical transformation” e), the into itself, one can also pose the ques-
operation of division is introduced as tion of the symmetry group for the entire*
the inverse of multiplication (a -f- b = plane or for three-and multidimension-
ab -1 =c if a =
cb), and soon. Howev- al spaces. For example, the symmetry
er, the fact that multiplication may be group of a Euclidean plane consists of
noncommutative later played a most all its motions
important role in physics. x' = x cos + y cos + a,
cp cp
Although the theory of groups was y' = — sin + y cos + b,
cp cp (C.l)
put to active use in the theory of crys-
tals in the 19th century, the problems where angle cp and the numbers a and
that arose seemed very remote from b are arbitrary (see formulas (1.9.6)).
the profound problems of the main “buil- The fact that the angle cp of rotation
ding blocks of the Universe” elemen- — and the quantities a and b characteriz-
tary particles. In 1932, however, the ing the translation of a figure in the
famous German physicist Werner Hei- plane may be arbitrary permits us to*
senberg established that replacing a speak of an infinitely small rotation
(positively charged) proton in a nucleus (through an infinitely small angle Acp or
by an (electrically neutral) neutron is dcp, with dcp 0) or an infinitely small
similar to a rotation in mathematics. translation. This in turn makes it pos-
This discovery was a brilliant page in sible to introduce the concept of a
the history of physics. Actually, it was “group trajectory” (defined, say, by the
from this moment that groups and non- fact that a and b are constant and cp va-
commutative operations and the theory ries at our choice) and also the concepts
Conclusion. What Next? 527'

of the differential, derivative, and inte- Cauchy died in 1857, and in the 1860s*
the French Academy of Sciences decided to
gral in the group space. Here we can
publish his complete works. The prominent
also define the main functions, such mathematician Camille Jordan (1838-1922)
x
as, for instance, the exponential e (giv- was put in charge of the project, and in con-
en, say, by formula (6.2.2), with x nection with this he made a close study of all
of Cauchy’s personal papers. Among them he*
the “group variable”). The noncommu-
discovered the letter from Galois, which,
tativity of the quantities that we meet 30 years after it had been written, made a tre-
in this theory creates absolutely new mendous impression on him. Jordan painstak-
problems, which have no analogies ei- ingly studied all the published and unpub-
lished writings of Galois. He gradually arrived
ther in the theory of functions of a real
at the conviction that he must write a book
variable (or several real variables) or about his young compatriot’s ideas. Jordan’s*
in the theory of complex variables. monograph, entitled Traite des substitutions et
Continuous groups, like the group (C.l) des equations algebriques (A Treatise on Sub-
stitutions and Algebraic Equations), which
of motions of a plane, were first stud-
was published in 1870, was the first textbook-
ied in detail by the Norwegian mathe- on the theory of groups to appear in the
matician Sophus Lie (1842-1899) and mathematical literature.
are now called Lie groups . During the In the period when Jordan was working-
on his treatise there were among his students-
past two decades, Lie group theory has
two capable young friends, the mathematicians
acquired enormous significance for the Felix Klein of Germany and Sophus Lie of
classification of elementary particles in Norway. They too took a keen interest in the
modern physics and for an understand- ideas of Galois and the theory of groups. Jor-
ing of particle interactions. dan was the first to note the existence of twa
different types of groups: discrete (crystallo-
graphic) and continuous. The two pupils of his.
The outstanding importance of the theory
divided these two branches of mathematics
of groups and, in particular, Lie groups, for
between themselves, so to say. Klein’s main
modern physics prompts us to dwell briefly achievements were in the sphere of discrete-
on the history of the origin of these depart-
roups: a special type of discrete groups which.
ments of mathematics. C 16 We
have already
ge distinguished, now known as Klein groups ,
spoken of the pioneering work done by the
has been arousing great interest lately. Along*
brilliant young Frenchman Evariste Galois
with that, in the dissertation which he sub-
in the theory of equations (his predecessors
mitted on the occasion of his admission to the-
were the famous Lagrange, the Italian Paolo
faculty of the University of Erlangen and
Ruffini (1765-1822), and the distinguished which later came to be called the Erlangen*
Norwegian mathematician Niels Henrik Abel Programm he suggested a system of symmetry
(1802-1829)).
C17 Galois’s work was not ap-
(the symmetry group) of any geometric mani-
preciated during his lifetime; what is more, it fold as a principle to be used in classifying the-
even failed to gain notice. Galois wrote letters separate branches of mathematics. According
to the famous French mathematicians Cauchy to Klein, the Lobachevsky space differs from
and Simeon-Denis Poisson (1781-1840). Pois- the conventional Euclidean space not only
son was unable to grasp the importance of because its parallel lines have different prop-
Galois’s ideas, which were far ahead of his erties (a secondary criterion which does not
time, while Cauchy evidently did not even read explain much) but also because it has a totally
the letter sent to him by an unknown youth. different symmetry group. Later this prin-
ciple was applied in physics, and today we are-
inclined to believe that, for instance, Ein-
c * 16
For more details see I. M. Yaglom, stein’s special theory of relativity differs*
Felix Klein and Sophus Lie Znanie, Moscow,
, from Newton’s classical mechanics in that
1977 (an enlarged English version of this book the former has a different symmetry-group«
will be published by Birkhauser-Verlag, structure in four-dimensional space-time; more-
Boston, New York). over, in “Einstein’s world,” this group is
c 17 The
*
difference betweenthe work done arranged so that the difference between the*
by Galois and that of his predecessors con- spatial and temporal coordinates (see Sec-
sisted not only, and not so much, in that tion 9.8 and the book mentioned in footnote*
Galois was the first to introduce the term“group” 9.12) is somewhat diminished. Klein’s prin-
and to define this concept rigorously, as in that ciple of symmetry is of great importance to the
Lagrange, Ruffini, and Abel worked exclusive- whole of modern physics.
ly with commutative (or Abel) groups, where Sophus Lie was that rare type of scholar
ab = ba for any two elements a and &, whereas who devoted his entire life and all of his pro-
Galois introduced general-type (noncommuta- digious research and writings (six very large*
tive) groups. books and a multitude of articles which were*
 9 —

528 Conclusion . What Next?

later collected in his multivolume Collected proved to be combinations of approxi-

Works) to the theory of continuous groups mately 100 types of atoms having ap-
(Lie groups). He worked out a wide-ranging
theory of continuous groups. He transferred
proximately 1000 different nuclei.
to differential equations the results that However, all these many nuclei were re-
Galois had obtained for algebraic equations: the duced to only two types of elementary
main difference here was that the symmetry particles, that is, protons and neu-
group of an algebraic equation (the Galois
trons. But by about the middle of the
group) is finite, whereas the symmetry group
of a differential equation (the Lie group) is 20th century an opposite trend appeared
continuous. Lie’s theory of the groups of differ- in theoretical physics —one toward an
ential equations (in particular, the classi- ever more complicated picture of the
fication of differential equations via their sym-
Universe. A very large number of ele-
metry group) has attracted much attention of
mathematicians and physicists in recent de- mentary particles, more than 100, have
cades. been discovered in cosmic rays and
Lie did not live to the 20th century and, also with the aid of accelerators. All
-unfortunately, could not see the rapid flow-
these particles are just as “elementary”
ering of his theory; for one thing, he did not
know that the theory of continuous groups as the proton or neutron. Here the nest-
would be applied in physics. The group-theory of-dolls model came into operation
period in the history of theoretical physics again: many particles are now describ-
began with in-depth investigations by Nobel
ed as consisting of simpler particles,
Prize winners Max Born and his pupil W. Hei-
senberg, both of Germany, and P. A. M. Dirac quarks, which have not yet been ob-
of Britain.Herman Weyl, G 18 a pupil of
* served. At present, however, we are not
D. Hilbert CA (already mentioned on many only, and not so much, preoccupied
-occasions) both of them the most distinguished
,
with the question of whether “still
mathematicians of the 20th century, and more elementary particles” exist. The
the physicist E. Wigner, were also among the
establishment of symmetry— that is, a
trailblazers of that period.
resemblance between various particles
leads to extremely important conclu-
The theory of symmetry and the theo- sions about their properties. The inter-
ry of groups play a very important
action of particles with one another,
role in the most fundamental and dif-
the possibility of the transformation of
ficult part of physics, the theory of
certain particles into other particles,
-elementary particles.
and other properties are very closely
For a long time this theory developed
connected with symmetry. The secrets
in one direction: physicists “split” com-
of quarks, electrons, and neutrino can
plicated objects and discovered the sim-
perhaps be revealed by a deeper study
pler parts that comprised them. It thus
of their symmetry. And today the the-
appeared that molecules consist of
ory of symmetry plays a cardinal role
atoms, atoms consist of nuclei and elec-
in all of the investigations that are
trons, and nuclei consist of protons
being conducted. And so physicists na-
and neutrons. Today we would add
turally take a great interest in all
that protons and neutrons consist of
aspects of the theory of symmetry and,
quarks. The evolution of physics was
first and foremost, the theory of groups
reminiscent of the way a child takes
(discrete and continuous).
apart a nest of Russian painted wooden
Not long ago, in a serious article, we
dolls and discovers a smaller doll in-
came across the following: “Since the
side each one it opens up. In each new
theory of groups is now taught in
layer, physicists saw a simpler picture:
high school, persons who recently
all the countless types of molecules
finished school can skip the next sec-
0,18
tion. However, older physicists should
See M. H. A. Newman, “Herman give it their attention in order to be
Weyl...,” in Biographical Memoirs of Fellows
of the Royal Society 1957.
able to discuss it with their children
,

u,ly
See the informative book by G. Reid, and do deatn tneir "Snftiertts?- ^Vv e diave
‘

Hilbert ,
Springer, Berlin, 1970. not yet decided to include elements of
 —

Conclusion . What Next? 529

the theory of groups in our mathemat- them, man has at his disposal a more
ics textbook for students who are be- powerful tool than ordinary “common
ginners in physics, but we may yet have sense,” which is based on traditional
to do so. logic.
An absolutely new situation has Using his hands, man makes simple
arisen in mathematics and mathematical tools with the aid of which he makes
physics with the appearance of com- machine tools; with the aid of these
puters. They range from pocket calcu- he constructs more complicated devi-
lators to computer complexes, each unit ces, and using these more complicated
of which handles a separate stage of the devices he does things which he could
operations. The functioning of these not do with his hands alone. Mathemat-
complexes can be compared in a way ics is very much like that. It devel-
to the higher nervous activity of man, ops more and more complicated theo-
in which the hemisphere of the
left ries, introduces fresh notions and en-
brain (which controls speech) and the ables us to comprehend and master the
right hemisphere (which handles visual most unusual phenomena of nature.
impressions) perform different functions. At the end of the course of mathema-
Computers have significantly changed tical physics we can again start a new
the approach to problems in higher chapter entitled “What Next?”, but do
mathematics in that they often make it not lose heart, for not far off is the
simpler to apply an approximate brute- boundary line that marks the end of study
force calculation than to use complicat- and the beginning of creativity and
ed mathematical models. Furthermore, the development of new theories.
computers have given rise to absolutely We have tried to picture the reader
new branches of mathematics. who in a few moments will close this
One occasionally hears the remark book with a sign of relief. Most likely,
that “mathematics is a mill that grinds you are finishing school or are in your
up only what is put into it.” In this first year at college. May mathematics
way poor results are explained by the always remain for you an exact and
fact that the original premises were beautiful language, a means of express-
faulty. In reality, the mill quite often ing ideas, a way of thought. May math-
turns out much more than is put in ematics be more than merely another
and the results are sometimes totally subject thatmust be “passed” at an
unexpected! examination and then left behind with-
Fundamentally, mathematics can be out a trace. Love mathematics
regarded as a variety of refined logic. and mathematics will love you and help
The remarkable thing is that having you in your work.
set up the rules of the logic and learned

34-0946
 , : ,

Selected Readings o

Texts on Higher Math for Beginners 4, Boundary Value Problems Integr ,

Equations and Partial Differential Equ

,

1. M. K. Potapov, V. V. Aleksandrov, and tions vol. 5, Integration and Function

;

P. Pasichenko, Algebra and Analysis

I. Analysis.
4.
of Elementary Functions, Mir Publishers, 19. “Mathematics in the Modern World
Moscow, 1987. Scientific American Sept. 1964. ,

2. L. S. Pontrjagin, Learning Higher Mathe- 20. P. Lax, S. Burstein, and A. Lax, Calcul
matics, Springer, Berlin, 1984. with Applications and Computing vol. ,

3. S. M. Nikolsky, Elements of Mathematical Springer, New York, 1976; 2nd printii

Analysis Mir Publishers, Moscow, 1983.
,
1984.
».S«.T^3Mr0 ,4- FJvsi.Cs>uxseJv.. FjjIc vlos £ tb„ , ?i_R v .
anrl_ Robbins.., What.
ed., Addison-Wesley, Reading, Mass., Mathematics! Oxford
University Pre
,

1978. London, 1941; reissued 1978: Chapters

5. I. Niven, Calculus: An Introductory Ap- to 7.
proach, Van Nostrand, London, 1961.
6. L. D. Hoffmann, Applied Calculus: A Year
Course McGraw-Hill, New York, 1983.
,
Texts on Physics for Beginners
7. M. Vygodsky, Mathematical Handbook
Higher Mathematics, Mir Publishers, Mos-
cow, 1971; reprinted in 1975, 1978, and 22. G. S. Landsberg, Textbook in Elementc
1984. Mir Publishers, Mosco
Physics, 3 vols.,
8. R. Courant and H. Robbins, What is 1988.
Mathematics'* Oxford University Press, 23. B. Marion, Physics and the Physic
J.

London, 1941; reissued 1978, Chapter 8. Universe, Wiley, New York, 1971.
24. L. D. Landau and A. I. Kitaigorodsk
Physics for Everyone 4 vols., Mir Pu
Advanced Texts (Higher Mathematics) lishers,Moscow, 1980, 1980, 1981, 198
reprinted 1983, 1984, 1986 and 1987.
9. A. D. Myskis, Introductory Mathematics for 25. J. Orear, Fundamental Physics, 2nd e<

Engineers Lectures in Higher Mathema-

:
Wiley, New York, 1967.
Mir Publishers, Moscow, 1972; 2nd
tics, 26. F. J. Bueche, Introduction to Phys
1975, reprinted 1979.
ed. for Scientists and Engineers, 2nd e(
10. I. P. Natanson, Theory of Functions of a McGraw-Hill, New York, 1975.
Real Variable, 2 vols., Ungar, New York, 27. F. A. Kaempffer, The Elements of Physic
1955, 1959. A New Approach Braisdall, Walthai ,

11. V. I. Smirnov, A
Course of Higher Ma- Mass., 1968.
thematics, vol. 1: Elementary Calculus. 28. Physics the PS SC text , D. C. Heat
,

Addison-Wesley, Reading, Mass., 1964. Boston, 1960.

12. L. Bers, Calculus, Holt, Rinehart, and 29. F. W. Sears, M. W. Zemansky, and H. !

Winston, York, 1969. New Young, College Physics 5th ed., Addiso ,

13. G. M. FikhtengoPts, Fundamentals of Wesley, Reading, Mass., 1980.

Mathematical Analysis, 2 vols., Perga- 30. L. N. Cooper, An Introduction to t

mon Oxford, 1965.

Press, Meaning and Structure of Physics Harp ,

14. J. Marsden and A. Weinstein, Calculus ,

and Row, New York, 1968.
3 vols., 2nd ed., Springer, New York, 31. E. M. Rogers, Physics for the InquirU
1985. Mind : The Methods Nature, and Phil ,

sophy of Physical Science , Princeton Ur

versity Press, Princeton, N. J., 1960.
Additional Reading (Higher Mathematics)

15. Ya. B. Zeldovich and A. D. Myskis, Ele- Advanced Texts (Physics)

ments of Applied Mathematics Mir Pub- ,

lishers, Moscow, 1976.

16. R. Differential and Integral
Courant, 32. D. Halliday and R. Resnick, Physic
Calculus, 2 vols., Wiley, New York, 3rd ed., 2 parts, Wiley, New York, 197<
1937, 1936. 33. R. M. Eisberg and L. S. Lerner, Physic
17. A. D. Myskis, Advanced Mathematics for Foundations and Applications, McGra\
Engineers'. Special Courses Mir Pub- ,
New York. 1981.
Hill,
lishers, Moscow, 1975; reprinted 1979. 34. P. Feynman, R. B. Leighton, ar
R.
18. V. I. Smirnov, A Course of Higher Ma- M. L. Sands, Feynman Lectures on Phy
thematics Addison-Wesley, Reading, Mass.
,
ics 3 vols., Addison-Wesley, Readinj
,

1964: vol. 2, Advanced Calculus vol. 3, ; Mass., 1963, 1964, 1965.

Part 1, Linear Algebra ; vol. 3, Part 2, 35. R. P. Feynman, The Character of Physicc
Complex Variables, Special Functions’, vol. Law Cox and Wyman, London, 1965.
,
Selected Readings 531

36. Berkeley Physics Course McGraw-Hill,

,
Readings in the History of Mathematics
New York: vol. 1,Mechanics 2nd ed.,
,

1973; vol. 2, Electricity and Magnetism,

1965; vol. 3, Waves 1968; vol. 4, Quantum

, 38. D. J. Struik, 4 Concise History of Mathe-
Physics 1971; vol. 5, Statistical Physics ,
,
matics, 3rd. ed., Dover, New York, 1976.
1967. 39. E. T. Bell, The Development of Mathe-
37. G. J. Kopylov, Elementary Kinematics of matics, 2nd ed., McGraw Hill, New York,
Elementary Particles Mir Publishers, Mos-
,
1945.
cow, 1983.

34 *
 2

Appendices

Appendix 1. Derivatives a
x
dx= + C.
5.
^

1. y
* = c, 9
-SL
dx
= 0. 6.
^
e
kx
dx = ~ e
kx 4-
C.

2 . y
y = x, ’
-^-
dx
= 1. 7. f
xV* dx = i-K xn e — 4
K
fex
f
dx.
J j

6 - y ~ x !&- ax ~ q
5
r & _ 1 i
111
* x *
J l + eft* k 1 + eft*

** ^- e ’
dx
e •

9. sin ax dx
^

5 . y = ax ,
— a x In a ~ 2 3 aK log 10
. a. — -j 2
ekx
(& s i n — a cos ax) + C.
c
jra

6. i/
* — In a;,1 4^- =— .
e
kx
cos axdx
dx x 10.
j

7 - y = log a a:, —— (k cos ax-\-a sin ax) + C\

dy
dx
_
xlnfl
1
J
0.434
logio a
1

x 11.
j
sin kxdx= — ^-cos kx-\-C.

8. i/ = sinx, -^|- = cosx. 12. cos/cxdx = -“Sin &x-f C.

j

= cosx = — sin x.
Jiro'- --T cotfe + C
9. y , 13 = '-
-

10. y = tanx, -g— -5^7. 14 = ^tankx + C.

18.

11. y = cotx, %-=-- - 4 r. sin 2 x 15. \ sin 2 /cx dx = '

-sin 2/cx Jr C.

1 19.
12. y== arc sin x,
dx 1— a;
2 16. cos 2 A:xdx = -|- + -^-sin2/^:x + C ,

j
xn
13. = arccosx, dy 1
xn sin kx dx— — cos kx
V 1
z/
dx <M
1
20.
dy _ 1 ‘~ 1
14. t/ = arc cot x, dx 1+* 2 * + -y j
xr cos kx dx.

15. z/ = aiccotx, dy _ 1
x n cos Arx dx = sin kx
dx i+*2 •
j

—^ j
:rTl
”
1 s*n dx *

Appendix 2. Integrals k—l)x

lxdx =
sin (
of Some Functions \ sin fcx sin
2(k—l)
- Sl
C
(fc+r +
if 1*1 (
if 1*1
1. dx — x-t-C. 2
j = |Z|, see No. 15).

2. jV<£r = -|^- + C (agt— 1).

j
cos kx cos lx dx = ^(fe — ij
j-f-=ln|*| +C.
3.
+ 2(fe+r +C if
Si
|A:| ^ UI (if •*'
dx 1 n
f
ax-\- b a
i
In |ax i

+ &l +
\ u\ \

C\
= |Z|, see No. 16).
 x x + 1 &

Appendices 533

21. (
J
sinkxQ,oslxdx= — C
°o^T!\
2 (k + l)
*
37.
j ax2 + bx+c y 4ac — b2

- C
2(f_lf +C if
v a retail'
2ax -\~b
l/4ac-&
— 2
i

f-
r*
f?
•
t
if
/
4ac— 2
>0,
22.
j
tanfcrd£= — ^-ln |cos kx \ + C.
J a* 2 + frr+c Yb 2 — 4ac
23. cot kx dx — In |sin kx \ + C.
2aa:+6— — 4ac |/~6 2
j v ln
2ax + b + Y b — 4 ac
+^ 2

24.
j
Y ax -f- b dx = (ax -f b)^ 2 +C .
if 4ac — 6 < 0, 2

dx 2
nr f dx 2 ]/* a#+ b .
^
i*

" ^
25. J ax 2 -\-bx-\- c 2 ax-\-b
j j/' a£-|-&

dx
if 4 ac — b = 0, 2
i.e., c — b /4a 2
,

’ )'«
= arc sin }-C.
J OO f <*£

J (a;z 2 -t- frr + c)

n

27.
j
£ ]/" a# +& — —6
2a# —|— 6
n_1
— 2b)(ax + b) s/ 2
(n 1) (4ac 2
) (
ax 2 + fcx -f- c)
_ 2(3ax ^
15a 2
,

,
(2n — 3)2a i* dx
'
(n— l)(4ac — b ) (ax + bx + 2 2 c)
n~ l
28. l/a 2 — ,r
2
d.r
)

j for n^2 and 4ac—b ^=0. 2

= 4" x a2 — + a2arc smx]

a;2 C-
= 1
[ 39 -
h
J a^ 2 + 6a:4-c 2a
In \ax 2 -\-bx-\-c\

29 j£2E2. & dx
.
j -11 ax 2 -\-bx-\-c
(see No. 37).

= - a ln Ji±£EE*. + C . dx _1_ i x2
| |
4°. j-; ~~ n
x (ax 2 -\- bx-\-c) 2c \ax 2 -)rbx-\- c\

30. xY z -{-mdx = -^- (x 2 +m) 3/2 -\-C.

j ..,11^. a <* No. 37)’
2c J
i
+ bx+c
ax^ (
v

31 *
l7fc =ln|a:+ ^^+™ l+c -

41 f
J
d ±
x m (ax 2 -\-bx-\-c) n

32. ^ (m — 1 )
ex 171-1 (ax 2
1

+ bx + c)
n_1

m — 3) a dx
= /^+^-gln| *+Vf+* + C \
.
— 1) c
(2^
(m
-f-

J
l

xm
~2
(ax 2 bx -j- c) n

(n~\~m — 2) b
33.
j
Y + m^ = y [a:]/ x
a:
2 2
-\-m
(m — 1) c
f

J a;
m_1 (ax 2
dx
-f- bx + c) n

+ m ln (for m > 1).

|o; + ]/a: + m|] + C. 2

34.
j
1.
j
y ax +
= V 2 — a2 — a arc cos + C.
n{
a(7z
^+ b
+ } v^+b+c. l)
r 1 '

^ _ ^—« 1 /q I* dx _ 72(a;r-|-fr) 1

+ a + 0.
C /~r
35 1^5T"-S- in " ,

- -a 2a "fe|
x
a;
2 2
1 "t '
J 'Y ax -\-b ( n |/ ax -\-b

36 -
lm— — dx n"
x 2 -j-a 2 =—
=
ia ar c tan
arctaII
1

T" +\-C.
C 1
'
44.
j
ln a: dx = x ln x — £ + C\
 a . i -

534 Appendices

45. (In x)
n
dx (— l<;r<;l
2. if m is not a positive
j integer).

= x (In x) n — n (In x)
n
^dx. — + —
j sinx = a: -gj- -gj 7 f+---
46. I
J
arc sin — dx
a
(any x).

X2 X4 X6
*= x arc sin + ]/”
a2 —# + C 2
O
3. cos*=l- A

lj --f 4
.

j—6j-+ ,

...

(any a;).

47. arc cos ~^dx 17

J 4.
6. tan;r = ;r-|--g-;r + 3
315
= x arc cos —
a
— V 2— r
C. 62
2835
48. arc tan dx
j x x2 a; 3 x4
r
5. e
y
— 1 + *jr
* .

+
.

"
2f +
,

3T-r 4f
.

+
.

• • •

*= # arc tan In (a 2 + # 2) + C. (any x).

49.
j
arc cot dx In (l + x) = x— -y- + -y- — "T+ • • •

— —- + — !<*<!).
a; arc cot In (a 2 + £ + C.
2
)
(

„ x3 1-3
7. arc smi=i+
• ,

+
,

2 ;^ 5
~ a:
.
5

Appendix 3. Series Expansions

+ ^4
3 5
+ ^7 • •• (
— l<z<l).
+ gr^l + mg + ^y)
6 7
2
1. (l s 7*3 7*7

arc tan x = x —+ 7*5

- -

m m (m
8. =-+...
( 1) 2 ) ja
-|
3 !
^

( — < *< i ).

Appendix 4. Numerical Tables

TABLE A4.1

X ** X e
* X e
*

0 1.000 1.000 1.5 4.482 0.223 3.8 44.701 0.0224

0.1 1 .105 0.905 1.6 4.953 0.202 4.0 54.598 0.0183
0.2 1.221 0.819 1.7 5.474 0.183 4.5 90.017 0.0111
0.3 1 .350 0.741 1.8 6.050 0.165 5.0 148.41 0.00674
0.4 0.492 0.670 1.9 6.686 0.150 5.5 244.69 0.00409
0.5 1 .649 0.607 2.0 7.389 0.135 6.0 403.43 0.00248
0.6 1.822 0.549 2.2 9.025 0.1108 6.5 665.14 0.00150
0.7 2.014 0.497 2.4 11.023 0.0907 7.0 1096.6 0.000912
0.8 2.226 0.449 2.6 13.464 0.0743 7.5 1 808.0 0.000553
0.9 2.460 0.407 2.8 16.445 0.0608 8.0 2 981.0 0.000335
1.0 2.718 0.368 3.0 20.086 0.0498 8.5 4 914.8 0.000203
1 .1 3.004 0.333 3.2 24.533 0.0408 9.0 8103.1 0.000123
1.2 3.320 0.301 3.4 29.964 0.0334 9.5 13 360 0.000075
1.3 3.669 0.273 3.6 36.598 0.0273 10.0 22 026 0.000045
1.4 4.055 0.247
Appendices 535
 536 Appendices

Appendix 5. The International System, or SI

Quantity SI unit

Name Dimensions Name of unit Unit symbol

Base units

length m meter m
mass kg kilogram kg
time s second s
electric current A ampere A
temperature K lei v in K
luminous intensity cd candela cd
amount of substance mol mole mol
Additional units

plane angle — radian rad

solid angle — steradian sr, sterad

Derived units

velocity m/s meter per second m/s

acceleration m/s 2 meter per second squared m/s 2
angular velocity 1/s radian per second rad/s
angular acceleration 1/s 2 radian per second squared rad/s 2
force kg-m/s 2 newton N
pressure kg/m -s 2 pascal Pa
work, torque, energy,
quantity of heat kg-m 2 /s 2 joule J
power, heat flux kg-m 2 /s 3 watt W
heat capacity (specific) m 2 /s 2 -K joule per kilogram kelvin J/kg-K
momentum kg-m/s kilogram-meter per second kg-m/s
impulse kg-m/s newton-second N-s
frequency 1/s hertz Hz
quantity of electricity A-s coulomb C
electromotive force, po-
tential difference kg-m 2 /A-s3 volt V
electric field strength kg-m/A -s 3 volt per meter V/m
electric resistance kg-m 2 /A 2 -s 3 ohm Q
electric capacitance A 2 -s 4 /kg-m 2 farad F
inductance kg-m 2 /A 2 -s 2 henry, weber per ampere H, Wb/A
magnetic flux kg -m 2 / A-s 2 weber Wb
magnetic field strength A/m ampere per meter A/m
magnetic flux density, tesla, weber per square
magnetic induction kg/A-s 2 meter T, Wb/m 2
magnetomotive force A ampere A

Appendix 6. Greek Alphabet

Aa Alpha It Iota Pp Rho
BP Beta Kx Kappa So Sigma
r? Gamma A% Lambda Tt Tau
A 5 Delta M p, Mu Tv Upsilon
Ee Epsilon Nv Nu ® cp Phi
Z£ Zeta Xi xx Chi
Hr, Eta Oo Omicron w^ Psi
00 Theta IT n Pi Q co Omega
 k s

Hints, Answers, and Solutions

1.6.1.
Part 1 (a) y~~^ 2; (b) y = ± Yx + 2;
Chapter 1
*={'£+! -1;
(c) 1.6.3. (d) y = ±V lAJ+l-l.
1.2.4. r = /2, a = 45°; r = 2]/T, a = (a) y= (b) y =

—45°; = a _== —135°; ==_4/2, a =

r 3 1^2, r
— —6
135°._1.2.5.
a/]/ 2); (alY 2,
2±2V 2;
0); (0,
2y
r
2;
—a/Y 2); (— a// 2,
lY 2. 1.2.6^ (0,

0)
/—
da: b
a: (c

a
2
/4a) 6 ,
(C)
x
" =
(make a drawing). 1.2.7. (a, 0); (a/2, a]/"3/2); — cz-j-a
( — a/2, a]/* 3/2); (-a, 0); (—a/2, -a/ 3/2); 1.7.8. Here we can easily prove in a purely
(a/2, — a]/" 3/2) (make a drawing). 1.2.8.__(a) formal manner that x and y appearing in the
Two ±aj/ conditions for the substitution of the variable
cases: (—a/2, 0), (a/2, 0), (0, 3/2).
preserve the sign of aB (see the solution to
(b) Four
1.3.3.
cases: (0, 0), (a, 0), ( a/2^ zta]/^ 3/2) Exercise 1.7.9). from which it immediately
and (0, 0) (—a, 0) (—a/2, ± a/ 3/2) (make a follows that the curves (1.7.10a), (1.7.10b).
drawing). 1.2.9. A2 (x x , —y x ),A* (— x u y x ), and (1.7.10c) cannot approach one another. A
A 4 (— a?!, —y (make a drawing). more interesting (“geometrical”) reasoning
= — x\
x)
(b) y 2x 1. 1.3.4. If
(a) y = — is also possible here. It is easy to see that neither
the straight line intersects the coordinate the translation of the coordinate origins, nor
axes at the points
1.4.3. and then a and b takenM N , the change in scales along the x and y axes
with the appropriate signs are the lengths of (which is equivalent to uniform stretching
the line segments and ON. OM of the curve along the axes) affect the presence
1.4.2. Use the fact that x 2 2x 2 — + = or absence of (local) maxima or minima of the
(x — l) 2
+ 1; 2x 2 + 4x = 2 (x + l) - 2. 2
curve y =
f (x) and also the presence or ab-
The stretching from the origin (homo- sence of salient points at which the tangent is
thety) with a coefficient Y transforms point horizontal. From this and from the fact that
M (x, y)_to point M' (
x ', */'), wither' = Yk x,
the curve (1.7.10c) has two maxima (the mo-
no tonically increasing curves (1.7.10a) and
y' =Yk y-kxy Whence,
k. (b) The x'y' = = (1.7.10b) have no maxima) and the curve
stretching along the x axis with the coefficient k
(1.7.10a) has the salient point (the tangent at
transforms point (x, y) to point x (x xMy x ), M ,
which is horizontal)— the origin— the requi-
with x x — kx, y 1 — y( whence x = —x x, y = red assertion follows immediately. 1.7.9. At

y±). 1.4.4. We must prove that for k, x lt x 2 >> 0

1.5.1.
\ k
B = c — —=
0 the relation (1.7.9) re-
1.8.1. 3 ct
the following inequality holds true duces to the form (1.7.10a); if B 0, this

J_ / _L Jl ) > !l or £i+^ > relation reduces to the form (1.7.10b) for

1.8.3.
2
2
\ x1 x2 / (^i + ^2)/2 ’
2x x x 2 aB = ac — — >> 0. and to the form (1.7.10c)
3
Xi +X 2
*
for ac ——< 0.

4
(c) The graph of the function y = 3
a: x 2 can be obtained by the “addition” of The curve is represented in Fig-
(a)
the graphs we have known of the functions ure 1.8.2. The circle x 2
H.2. y
2 1. + =
yx a: —
4
and y 2 =
x 2 in shape it resembles these An ellipse (Figure H.3). 1.8.4. The line
graphs. 1.5.3. See Figure H. 1 a-d correspond-
;

segment of the straight line y x between the =

ing to the limiting cases n -> oo; for large n '
points (—1, —1) and (1, 1). 1.8.5. x =
the graphs are close to those given in Fig-
ure H.l.
r arc cos
~V2n — / i/
2. 1.8.6. (a) The

Figure H.1
 k -

538 Hints Answers and Solutions

, ,

nates x ' , y' related to the old coordinates x

,

y by formulas (1.9.6) the equation of the same

straight line will have the form y' k'x' s', = +
where

k cos q'-f-sin <%' s-j-fca' — 6'

cos a '
— sin a'
’ s
cos a' — A: sin a'’

here a' = —a =
/_x'Ox is the angle between
the O'x' and Ox axes, and a' and 6 ' are the
coordinates of the new origin O' in the old
system xOy .

Figure H.2
Chapter 2

2.1.1. + + 3 tAt + (A
(a) t-
av = 3* 2 1 *)
2

y inst = 3f 2 = jnst + 1. 2.1.2. yav t> .

;

2.2.1. av = 0.3965 + 4.162 X

c 10" T — 3

15.072 X 10- T + (2.081 X 10" — 15.072 7 2 3

X 10“ T) AT — 5.024 X 10“ (AT)

7
= 7 2
c lrjst
0.3965 + 4.162 X 10~ T — 15.072 X 10
;
3 7 7* 2

c (0) = 0.3965; (100) = 0.7977; (500) = c c

;

2.1007. 2.2.2. cmst = 4186.68 + 16746.72 X

10- T + 3768 X 10- T
5 6 2
.

y — y' — 12x —
f
2.4.1. (b) (a) 4a: 3 ; 2

6a; + 2; y' = Sx + 4; (d) y' = —2/a;

(c) 3
(e)
y' = a — (1 y' = 2 ax — 26/a; 2.4.4.
2
1/a; ); (f) 3
.
;

(1.2) = 1.44; = = Aa; = 0.2,

2 2
y if (x) a; a; 1,
+ y' Ax = + 2 X =0.2 = 1.4;
,

Figure H.3 then y (x) (x) l2

2
the error 3%. Similarly, (l.l)
is 1.21 and
(l.l) ~ 2
+ 2 X 0.1 = 1.2 (the error 1%)
l2 is
and so on.
equations of the epicycloid x — (R+ r) cos t— 2.5.2. The tangent is horizontal at the
r cos iH-T g, */ = (i? + r) sin i — r sin points x — 0 and x = 2. 2.5.3. See Figure H.5;
r
(b) The equations of the hypocycloid
r
a: = the tangent is horizontal at x ±1/1^3 =
2.5.4. See Figure H.6. 2.5.5. See Figure H.7.
~~ r
(i? — r) cos t + r cos — - t, y = (R — r) X 2.5.6. (a) y (3/4) a; 1/4, =
(1/3, 0), (0,
-
d
r —1/4); (b) y 3a; 2, (2/3, 0), = (0, —2). —
sin / — r sin
r
1. Figure H.4 a-d shows 2.5.7. The tangent to the parabola (a) at the
r point (x 0 y 0 ) intersects the coordinate axes
respectively: the epicycloid for R r (car- =
,

at the points (x 0 /2, 0) and (0, */ 0 ); the tan- —

dioid); the hypocycloid for r (1/3) R (the — gent to the cubical parabola (b) intersects the
Steiner curve); the hypocycloid for r (1/4) R = coordinate axes at the points ((2/3) a; 0 0), ,

(astroid); the hypocycloid for r (1/2) R (the = (0, -2 y 0 ).

diameter of a circle of radius R). 2.6.1. x = 0; the minimum at a
= — 1, the;>ma-
(a) 0,
1.9.1. (a) y x; (b) x = —
0; (c) y x = = + the maximum at a < 0; (b) x
2; (d) y =
(—4/3) x. 1.9.2. For the first triple ximum; x = 1, the minimum; (c) x = — b/a j/"
of points: AA 2 A t A s 90°; SaajAzAs = 25; = (b/a >» 0), the maximum at a >* 0, the mi-
= 2/ 5. 1.9.3. (a) 2/ 2. 1.9.4. (a) nimum at a < 0, x = b/a (b/a >» 0), the y
3y~ 2/2.1.9.7. If y fcc =
s is the equation + maximum at a < 0. the minimum at a > 0.
of the straight line, then in the new coordi- There is no maximum or minimum at b/a c 0;

(b) (c) (d)

Figure H.4
 + =

Hints, Answers, and Solutions 539

For an exact evaluation of the integral

(the limiting sum corresponding to the value
m — oo) we write

Figure H.5

_ 2m 2 + 3m+i 1.11.11
6m 2 3
'

2 m 6 m2 '

whence it follows immediately that ^t 2 dt =

l

-|-(~0.33). 3.2.3. Let q = ™'r2; we denote

o
Figure H.6
30 = 1, Xi = q, x2 =q 2
,
x m = q m = 2. Fur-

ther 2 arfA 2 (
x l) s (
x l~ «Z-i) = S (^ X Z
)
3

l l l

m
il ~ 1
(qi—qi~ 1
) = (g—i) 2 q =(q— l)(g3 +g 7
+
i=i
^ 4771+3 aS
U + 4m" 1 = (?-!)•
9 ••• ? )
g
4 — 1

g
3
[(g
m 4-l]
)
3
g (2
4 — 1)
sending
g
3
+g +?+l 2
g 3_|_ g2 + g+ l
;

Figure H.7 2
15
’

x 3 dx — =
5
(d) x = —1
the maximum; x 1 the mini- —
mum; a 3> 0 x
,

0 the minimum; a— ,

=
y=
(e) 0,
=
, ,

.x =
0, the minimum; a 0, x 0, the max- < 3 3.75.

imum, x — —Y — a/2, the minimum, x =

Y —a/2, the minimum. 2.6.2. — 4°. t 3.4.1. (a) 1/3; (b) (1.1*— 1) c* 0.11033;
2.6.3. (a) ~ 4.08°, (b) ~ 3.92°.
t t

2.7.1. 2; 6a:; 12a: 2 ; 2a. 2.7.2. The acceleration (c) 1/2; (d) 2 (/3— 1) c* 1.464. 3.4.2. (a) S=
equals 2a, it is constant. 2.7.3. (a), (c) The b
curve is convex at x 0, concave at x < 0; > >

y=y(x) = — x the equation

(b) convex at x <c —
p/3, concave at x ]> p/3. — j
y (x) dx, with
o
b

Chapter 3
j
->

—^ sda: =
0
3.2.2. Below are given the values of the
u
sums when the interval is partitioned into m
parts (m = 10, 20, 50, oo): xdx JL
b
— 2
3.4.3. (a) 5=
m 10 20 50 OO
At 0.1 0.05 0.02 0 - 2 =
2.18 2.26 2.30 2.33 y (b) 5
y 3.4.4. 5

r
s4xm 2.49 2.41 2.37 2.33
#2 3.4.5. 0.7837 for n = 5;
r2

As seen, even at m =
50 both sums differ but 0.7850 for n 10. 3.4.6. = See Figure H.8.

little from the limiting value at m oo. = 3.4.7. See Figure H.9a-c.
 a - = =

540 Hints, Answers, and Solutions*

4 -4 -3- W 5-=
5 _ 1 &
~ a sin
1
°
d 2y
•
16 jij i ’
dx 2 2 3 *

4.5.1. (a) y' = 5x4 — 12x + 3x + 14x-2; 3 2

(b) = 2(x +x+l) (3x +

j/' y' 4(x - 3 2 l); (c) 2

x + l) (2x — (d) y' = 10 (3x — 1)» 6x-

3
1);
2

(e) y' = 1— y=x

y' — X (f)
2 /*>,
-jr-

v —
;
0 a:
2 1
Figure H.8
-e 4=- . 4.5.3. (a) If a: changes by 1%, then
3
y a:

Ay = n X 0.01 therefore when a: changes by z/;

k%, we have Ay = 72 X O.Oly/c. In the given

case x changes by 10% an Ay = nX0Ay.

(b)
Y (c)
Since — 1/2, Ay = O.ly = 0.05?/. There-
72

fore y (11) - y (10) + 0.05X 5 = 5.25; y( 9) ~

y (10) — 0.05 X5 = 4.75. Let us now obtain the
exact solution. We denote the proportionality
factor by k; then y = kY x Since at x = 10, •

7=5, we have 5 = Zcj/^10, whence k ~ 1.58

does not change. 3.6.3. (2/3) R3 X (the calculations are carried out to two decimal
3.6.2. It
i places) .^Therefore y == 1.58 j/^o:; y (11) =
tan a. 3.6.4. yY — y 1
2
dy = 1/3; note that 1.58]/TI 5.24; y (9) ~ ~ 4.74. (b) The ap-
proximate values are y (11) ~ 4.50; y (9)
~
5.50. The exact values are y (11) ~ 4.54;;
(9) ~ 5.56. (c) The approximate values are
the area of the cross section of the “hoof”
y
(11) ~ 6.00; y (9) ~ 4.00. The exact values
(Figure 3.6.8) with the plane, which is per-
y
pendicular to the plane parallel to AB, ABP ,
are y (11) ~ 6.05; y (9) ~ 4.05.
and distant by y from is equal to AB =
,
4.6.1. y
f
—x 2 (x 2 — 1) (7a: 2 — 3). 4.6.2. y'
2y V -V 1
2
-

(2a: —[— 1) xs (8a: +7) xs

3a: 2 Y x2 ~\~ x
2 Y x2 ~t x (
2 Y x2 + x
\
'
Chapter 4

4.2.1. (a) y' = x 2 )'

(. a:
2
+ = 2X
a:
2
(a:
2 )' 4.b.3. y = 5a: 4 y
2a: X x2 = 4a:
3
;
(b) y' = 5aa: 4 + 4&a: + + 3
3ca: 2
—
x3 2a:) 1 / 5

x2 — • 2xx b (x s — 2a:)
1
/
5 (

xs — 2a:
X
i

4.3.1.
dz _ dz dy
2ya = 2a (ao: -f- b);
(3a: 2 — 2) x5 \ x2 —
r
~
1 (simplify).

2 + (*
4.6.4. y'

+ ~k) y -

dx dy dx
z' = ( ax 2 -f- 2abx + b 2 )' 2 2x
-f- 2 ab. 3x 2
4.6.5. y' = 2x]/
r
yx-{-x -|-x 2 X.

4.3.2. (a) z'=? — (b) z' = 2 YX — 2 s

(ax +b )
2
1/3 j/g+1
*
(simplify). 4.6.6. y' =5
2a (f/x-f-
(c) z'
1
(d) y' = 2 Y y x+x v
(aa: + 6)
3 (X + 1)2
2a: (a: 2 + + 5)4a:
; (e) </'
x2
(r 2
— —2 2a:

-4- W
1
_\\44 /
x /
(_U_' 3 y^x 2
1

3 Yx
i

* |
x + [V x +
4.3.3. y"
fa
= ±1- 2 • 2/'gg' + 2/ (g')» - /gg" 1 \ 5
4 - 6 ' ,' !'
= x2 + l
4.6.8. y' =
ft )
•
(x 2 -i) 2

4.4.1. (a) y'

1
(b) y' = 1 — a:
2
4.6.9. y'
x2 + +5 2a:

|A( 2 x + 1)2
’
* (x 2 —x + l) 2 *
(*+i) 2 *

~ i2~~ 3 (3x —
(c)
1 1
4.6.10. y' = Y^T~2
1)

(x+1) YZ — 2 1 4 zyz 2x 2 Y * +2~ 3

 1 =

541
H ints ,
Answers, and Solutions

4.6.11. y'= -
(x 2 — l) 3 / 2
*
4.6.12. y' = — J-—
u
— v
,
whence (uu)
r
= u'v-{-uv\ (b) Differ-
entiate term-by-term the equation ln(w/y)=
2x 3 + = -
4a: 3 Y_ x
= =
ln^ — lnz;.
• 4.6.13.
3 (£+1)4/3 4.9.3. (a) y' ^
iY x*+x y'x
6a: y'^+l 4.6.15. y' = -i-- (2a:)' = -i- or y = In 2a: = In 2 + In a:, the-
4.6.14. y' . ,

6 j/"x 2 Y + x2 j/" a:
=— 1 2a:
refore y'=(ln 2)'+(lnx)' (c) y
— tf!
(1+^ 2 ) 3 ^T -— • 4.6.16. y'=*/'(2x+3) 2 +
;

= (e) y
'
=
j - 3/ ,
4
-.-.— . 4.6.17. i,' = 3a: 2 Yx - 1 + TTT ;
isfe rr'
3y"2a: +3
h) »'=
X3 — ^55=1. 1 2a: 2
. 4.6.18. y' =
(
f)
»'=-*r=r; or)
^'=6^ipij5 <

2 V"" a:
— 3y (a:
2 — l)
2
In a: + 1; (i) y' = 3a: 2 In (x + 1) + -q—j- ;

_ 2x2 4.6.19. y' = ]/|±| X since In y = x In x, differentiating we find

3 (a; — 1) 3 |/^ 2x —3 K 1 (j)

4.6.20. y' = Ax —
3 10a: 2 — 22a: — 11 — \

y
y’ = In a:+l, whence y' = y (In x-\- 1), ‘
or

(a;+l) 2 — 2) 2 >/>+l) 2 "

3
= -i-(/5) V xa "
(a:

- x5 2a: 3 -f- 2x 2 — y' = x*(lnx+l); (k) y'

1
X
4.6.21. y' = -f- 1

+ Y — (x 3 1)
2 x<t i
/ x In t
_ _|_
j£x 1_\ gee ^ jjj nt tQ
+ 2/3
1 x 1 \ -I- 2a:
\ Y — *x2 1 /
^ (j))_

( ,.^ +r)
4.6.22.
+ l) (a:
2 '

4.10.1. y' = 2 cos (2x + 3). 4.10.2. y'

4.6.23. = /i ^T /a:+/5 + /a -1 •X
j/'
2
3
— sin — 4.10.3. y' = —
(a: — 1) sin(a: —
1). (2a: 2

+ l). 4.10.4. y' = 2 sin cos x. 4.10.5. y' —

A+/5+^V —
a: a;

3 — (1+2 /x) a:]/ —! 4.6.24. y'= a;

2
3 cos cos x — 2 cos x sin x sin
3a: 4.10.6. y' —
2 3a:.

C0 ? X
6 V a: / (x+y a:)
2
(sin 2a:) x ( In sin 2a: -f-
^
-
\ (cf. Exer-
6/7
_j_r i—\ x *+ —
i( Yx)
(i
l 2Yx
Y x3 3 > else 4.9. 3j). 4.10.7. y' = tan x + — g2 .

,1/7
y' =
2a: + -yt)
x 4.6.25.
(
V a:

1— 6^ — 4a:|/~a 2
1 1
* '
3(x + l)f/x (x+1) 2 '
2 sin 2 (x/2)
4.7.1. y' ~ 2.3 log 2 X 2*. 4.7.2. y' ~ = — y' =
4.11.1. y' 4.11.2.
2.3 log 5 X 5* +1 . 4.7.3. y'~ — 2.31og2 (y)*. y i

-7—, —
1
5 4 . 11.0
4.11.3 y'
y = 7 1—= =.2 . 4.11.4. y'=
4.7.4. y' ~ 2.3 X lO ^ 1 y' ~ 1 +x 2 .
4x
o V- . 4.7.5.
2x —
2 y a: 3
4, *
, . ,
1

9x 2 +6x+2
* ' ' V
x4 — 2x +x + 3 2
l
’

2a: X 2.3 log 2 X 2**. 4.7.6. y' ~ (l — ^-) X arc tany*

4.11.6. y'=
2 3 log 2 X 2* +1 /*. 2Yx(x+1)
4.8.1. (a) y' = — «-*; (b) y' = 5e»— 3e *; 3
4.13.1. —1; 1. 4.13.2. — 1.
<c) y' = 2xe xi
(d) \
y' = 1
e*; (e) y' =
z y x
3 (a:
2 — 1) e x3-3;x:+1 . 4.8.2. More than by a Chapter 5
factor of 150 000.
log15- 5.2.1. 1/6. 5.2.2. 1/2. 5.2.3. 2.
15= ~ 1.6825.
4.9.1. log
&5
log 5
4.9.2. (a)
5.2.4. — (
= arc tan — arc tan 0 = 1 0
j
.

Differentiating both sides of the equation

(the logarithms are natural) we find --- = 5.2.5. e
e
5.2.6. 1.
 a . \ .
.

542 Hints Answers, and Solutions

,

5.3.1. z8 — z + * + C. 2
5.3.2. ~ — x*
/ = sin 3a:, dg = e 2x dx; then integrating by
5
A O parts yields e 2x sin 3xdx = -^- e 2x sin 3a; —
3_
4
^4 +^— 7*3 -7*2

5.3.3. -La*-- 1.*»+ j

i_
x 2 -\~C. 5.3.4. 2 -j- 2x 3 In \x\-\-C,
—^ e 2x cos 3xdx. In the last integral we put
2 -y a;
-J-
= cos 3a:, dg = e 2x dx and again integrate by
/

5.3.5. 2x + In |
a; — 1 | + C. 5.3.6. — x+ parts to obtain e2 *sin 3xdx—^ e 2x sin 3a: —
bj
^l«\
—~
+~ —
C x+d\

3) + ^(a; — 2)
3r
=
+ C.
tliei1
5.3.7. H ^_ o:,
~~2
(
~2
e2X C0S
j

— J
e2X s * n ^xdx
^
• C° n_
x 2 _l2> »
sidering the last expression as the equation
i.e.A-\-B)x — {2A-\-2B) = x and + 5=1, ^4
(
with the unknown e 2x sin 3 xdx we
therefore A — — 2, B = 3.
find
3 A + 2B = 0; j
Answer. — 2 In x — 2 + 3 In x — 3 C. -f- e 2x sin 3a;da: = —S 2 S1D 3* — 3 cos 3a;)
+
5.3.8. — In \x — l|+ln \x — 2\+C.
| | | |

C.
5.3.9. If I 13
x-\-i _ s-f-l A B C0S 2 X_+ 2 3 * a 2^)
—— (
x — 3a;+3 — 1) (x — 2) +c-
,

2 — (a; —2 a; 1 ‘

a;
’
5.4.10.
0
then A (x — 2) + B (x — 1) = £ + whence A —
— 2, 5 = 3 (in the previous equality first
1,
5.5.1. — 1
sin (3a: — 5)+5 /
/
put 3a: — 5=
assume that a:=l and then x=2 ). Answer.
— dx — dt
\ 5.5.2. — — cos + + 5.
21n |s— l|+31n \x—2\ C. 5.3.10. If + 3 )
9 (2a: 1)

x -\-2 B 5a:

+ ~V ~Y(3x — 2f + C.
-4
-
+ ,
* .

then 5.5.3. 5.5.4. Since at

*(a: 2 —
l) + *
a:
2 j— 1 a:
2 l
*

>1 +l) + 5a:+^ = ^ +

(a:
2
4 = 5 = 1, 2
2, i.e. 2,
Y x = z we have x — z 2 and dx~2zdz then
C = — 2. Answer 2 In — ,

+ arc tan In .
|
a;
|
a: (a:
2 -)-
X dx _ r z 2 »2z C?z
_ r z2 dz
1) + 5.
f
z 2! +2 ” 2
" 1+Z
x+y'5
(In \x\ —
J J J
5.4.1. cosa: + a:sina:+5. 5.4.2. a:
fH-1+i dz=2
l) + 5. (Hint. In formula (5.4.3) put / = lna:,
*J 1+z + T-p7 ] dz = j
[(z- 1)
dg=dx). 5.4.3. -y £ sin 2a;— f
-y a:
2
—y ) X — 2z + 21n(l+z) + C = x— 2 Y"x+2ln(l +
z2

cos 2a: + 5. 5.4.4. ( —x — 3 — — 6) +

3a: 2 6a;
V*)+C. 5.5.5. l/'i ^5 + C. 5.5.6.
sin x 2
4

5. 5.4.5. (2a:+l) cos x-\-(x 2 ~\-x — sin a;-f-C. 1)

5.4.6. Let (2a: 2
Ci) cos 3a; ( 2
2
2
+ 1) cos dx = (a x
+ a x + b x + sin
3a:

c2) 3a;.
x
2
-f-

Differen-
^aj-f- C, or
1

y cos2 x -|— 1
cos4 x + C

tiating both sides of the equation we get 1

5.5.7. c. 5.5.8.
(2a: 2 -j- 1) cos 3a; = (2aia; + b{) cos 3a: — (3 a^a: 2 -)- 3 sin 3 x sin x
3b\X —|— 3cj) sin 3a: —J— (2q, 2 x —J— b 2 ) sin 3a; —{— (3# 2 a;
2 ~j~ — In |
cos x | +C (put cos x — t). 5.5.9.

+ =
36 2 a;
( — 3«ia: —
3 c 2 ) cos
2
3a:,

— 3ci 2
3b x x
i.e.

-f-
(2a: 2 -f- 1)

2 a: +b 2)
cos
sin 3a:
3a:

+
arcsin —a + C. 5.5.11. In |
a;+ l/^ 2 + a;
2
|
+67

(3« 2 a;
2
-f- 3& 2 a: + 3c + 2«ia; + 6
2 X) cos 3a;. Thus we (In addition to the method indicated in
have 2a; 2 —|— 1 = 3cl 2x 2 —\~3b 2x-\- 3c 2 —j- 2cliX-\- 6^, p. 168, we can make here an artificial sub-

— 3b ix — 3 + 2a x stitution x a tan t (i.e. t =

arctan (a;/a)), =
0= — 3«ia: 2 2 -f- &2 , whence
dx =adt /cos 2 and a 2 -}-x 2 = a 2 /cos 2 t or the
t
3a = 2 3b = 3c -}-&i=
2, 2 -f- 2ct± 0, 2 1, — 3#i = 0, substitution Ya 2Jr x2 — z whence, squar- —
— 3&i + 2a = 0, — 3cx + = 2 b2 0. From this
ing both sides of this last equation we get
system of equations we find: ai = 0, & = 0, 2 a 2 = z 2 —2zx (the term with x 2 cancels out in
both sides of the equation) and thus x =
ci = 0, a2 =~, b1 = -~ ,
c2 = -~ Answer. .

Y cl
2 Jrx 2 = z — x=z — 2z
2z ’

~~ x cos 3a;
+ x2 -| sin 3a; +C
^ Jy j z2 +a 2
dx = 1 z2z
T-Jg).. ^z ^
5.4.7. a: a resin a; -j- l/* 1
— x2 + 5. 2z 2 z2

—— 1 z2 +a 2
whence
dx
5.4.8. a;arctana; ln(a; 2 -)-l ) + 5. 5.4.9. Set 2z 2
dz ,

V + a*
 —

Hints, Answers and Solutions , 543

(z 2 +a 2
) .
z2 +a 2
— 4p) 0-6065 0.7788 0.8825 0.9394
y 1
5[ 2z 2 ‘
2z (

= atanf, Ay 1.0422 0.5052 0.2506 0.1251 Aar

+a 2
5.5.12. Set z

da:
adt
so that
then a;
2

da;
COS 2
1
£

X
— Ay
1.042 1.010 1.002 1.006 1

cos 2 t
,
(a;
2
+a 2 2
)

cos 2 tdt== -^3 j(l+cos2£) dt= -~g- ^+

‘

j 6.1.6. Substitute into the numerator on

1
-7T-
2
sin 2/
\

)
]
+ <7. Since £ = arctan —
a
and
the right-hand side of (6.1.27) expression
(6.1.13a) for / (a Aa:) and / (a Ax) + —
sin 2 t—
2 tan t
— 2 x/a 2 ax (in (6.1.13a) set x a Aa: and x a — + = —
Al _ Aa; respectively).
1-j-tan 2 £ l+(*/a) 2 x2 +a 2
Apply Taylor’s formula (6.1.18) to
6.2.1.
dx 1
we finally have the function (6.1.1) (or Maclaurin’s formula
(a;
2
+a 2 2
) 2a 3
(6.1.19) to the function (6.1.1a)). 6.2.2. sia; =
arctan
2a 2
1

a;
2
+a 2 C. c +X H- — ...» where C =0 if

we assume that si 0 = 0.
Chapter 6 * +1
6.3.1. (a) y= =l + 2x+2x*+2x* +
6.1.1- y = +
(aa:§ + cx + d) + 6a;§ 1 X
(x — x
0

— x+ + + + + b) X
(3aa;§ 2bx 0
= ln(l+z) = z—A +
c) 0) (3aa: 0
(a:
0)
2
a (x — 3
All the subsequent
a;
0) .
2x*+...; (b) j,

terms equal zero; the sum of the above four

terms equal the polynomial. 6.1.2. Since
= = 1, y”{0) = 2, 6.3.2. y = lna:=(a: — 1)
0.
y( 0) 0, y'( 0) ym (0) = . .

n, . . y = X + x +?!+...= 2 (x— l) 3 (a: — l)

4
In Exercise 6.3.1
21 3 4
X
x
^1 +x+ —' + . .
.
. 6.1.3. y = e [1 + the series can be used for computations for
J
x 1; <
in Exercise 6.3.2, for 0 x 2, < <
- (* “ - |

i
|

(* 1)
+^ l)
2
+ (x l) 3
+ . . .].
6.3.3. / (x) g(x) = f (0) g (0) + [/' (0) g (0) +
~ m
*+A
6.1.4. It is easy to verify that
m ~ D r2
(1 + r)
g’ (0) / (0)] [/" (0) g (0) +2 /' (0) g’ (0)
+
1 + mr + 171
( -
(cf. Section 6.4), e mr ~
2 *•+...
2 2
^(0)/ (0)1 .

1 + mr + . Thus, for small r the quan- 6.4.2. These are the formulas (6.4.3) for
mr m by the term + —
tity
(1/2) mr
2
e from (1 differs
For example, if m
.
r)
50, r 0.02 = a—i and m —n
1
in which we retained two
(see the example in p. 143), then (1/2) mr 2 = first (a) and three first (b) terms respectively.
01, i.e. the error is less than 0.5%. (When
mr 2 is small, m can be large; here mr3 mr4 . . .
6.4.3. yT. 2 ~ ~ 1.062 (the table
1.067 and
1.063); f TI~ 1.033 and ~ 1.032 (the
, ,

will of course be small.) 6.1.5. (a) By formu- gives /

la (6.1.24), with y (0) = 1 and A y — y (Ax) — table gives 1.032). 6.4.5. For x = 0 even
1, we have
(Y x)' does not exist (becomes infinite).
Aa; 1 1/2 1/4 1/8 <1 t0 6.5.1. (a) 1; (b) -1/2; (c) 1/3; (d) oo;

y (Ax) 2.7183 1.6487 1.2840 1.1331 1

(e) 1; (f) 1/2. 6.5.4. (a) Substitute t =—
Ay
1.718 1.297 1.136 1.065 1
Ax (A)
and consider the ratio / g (-£-) ;

use the rule (6.5.2). (b) Prove that

By formula (6.1.25)1, setting A y=y |

(t)- - »
where c is a number
Aa: \
we get g (x)—g(x 0 ) g (c)
~2
J intermediate between x and x 0 (why?); fur-
ther assuming that x and x 0 are close to a
Aa; 1 1/2 1/4 1/8 4J t0 (and are on the same side of a), first let

y(4r) 1-6487 1.2840 1.1331 1.06451 x a with x 0 constant (

V
here —
g(x)—g(x 0)
/i£o)
 b — 6 — x —

544 Hints Answers, and Solutions

,

JS-) where by condition x-+a\ and then —— — sin 6, where a and 6 are
g(x) / Y + — b2
the angles formed
(c i r)2
let x Q -+-a. by the lines of the body’s
motion in the I and II media and the perpen-
6.6.1. (a) &= c ;
(b) y*—kx*=a; dicular to the line of the interface, we have
sin a vx
~~
(0 <•> sin p v2 *

(This the Snellius law that is, the point

is ,

= Ae sinx = 0. 6.6.4. Expand must move in the same way as the light ray
tan . 6.6.3. y
~Y does intersecting the boundary between two
x2/2 media with different velocities in the media.)
,e 0 (x) in a series in powers of x (to this
To prove that we have obtained the minimum
J2T
-end we must expand
ers of t and then integrate the obtained equa-
e~ t 2/2
in series in pow- of T it is sufficient to find - —
dx 2
;
it is easy to

tion termwise) ; prove that for an appropriate x 0

the product e X2 / 2 (D (x) can be represented by
ascertain that —
d T 2
- >» 0 for
dx 2
all x.

the series (6.6.23) (where a 0 0). = 7.2.1. y ra i n = 3. 7.2.2. At the moment

of time t the steamer S is represented by the
6.7.1. z = 2 st — (2st—2 0 )exp [— to)] ; line segment l whose endpoints are projected

on the line, which corresponds to the bank

the water level grows, tending asymptotically where the fisherman F sits (we take it for the x
to the value z =
z s t. 6.7.2. Find the steady-
axis) at the points with the abscissas xx =
state value z zs t = =
constant (=q 0 /k) which v (t — f 0 ), x» = The distance
v (t — t
0) — l.
satisfies the differential equation; then de-
note z = +%
and form the differential
from F to the point of the steamer S is M
z st
•equation for the function z x z x (t). — A (x) = Y + 2 h2 where x x <: x we
— min A (x). For xt 0 t t20 -\-l/v
, \

must find D C c
this minimum equals h; it is attained at the
Chapter 7
point x = 0, which corresponds to the con-
dition —
dx
— 0; for t <C t0 and t >t 0 +l/u there

7.1.1. x= *+ b -Vf + »-ab 6

7 i 2 Let are boundary minima Y (t — f0)
2
+^ 2 an ^
x be the side of the rectangle which belongs
Y\v (t — t0) — l]
2
+h 2
respectively. (This
problem can be solved geometrically without
to the base (equal to a) of the triangle, then
resorting to differential calculus.) 7.2.3. (a)
the area of the rectangle is S (x) — xh 1
— I/max 0 at x = 0; (b) J/ ax 1 at x — m = = 0.

x h
^
7.4.1. Prove that y" < 0.

— \
= hx — x 2 Solving the equation Tn
X\
i
rn
x 1 [n x 1 -j-x 2
n>
.

J 7.4.2. (a)
( 2
2
> 2
at 1,

=0 we =y
S' (x) find £ ;
then the adjacent
xx
,
x2 ,
...
(b)
/
^ ^~2
+— Y ,m <—iifi. for
h
side of the rectangle equals
y and the area
0 <m< 1, x 1 =£ x 2 ;
(c) 1 -r- + -
fe
ah
S(x) = 7.1.3. S = 2B 2 (the desired rec- /k

tangle is a square). 7.1.4. The radius of the

A-)]‘ > (^+fL) at k> 0 and x^x 2

3 f— = 1
—x 1
>
base = y y- ,
the altitude h 2 r. (and x u ^ 2 >0); (d)
y %i log x x -j- 2 log x 2

av x -\-bv 2 1
7.1.5. /=
v\ + v\
7.1.7. The time of
y +^ (^i 2) log [(^ 1 +^ 2 )/2] (in all these cases

motion T = — Va?+x*+-\- / 2 +(c-*)2 ,

we only give an inequality, which is a par-
ticular case of the inequality (7.3.1)).
7.5.1. n/2. 7.5.2. 1/6. 7.5.3. 2ji+ 4/3 and
where the magnitude of the projection
c is
37
of AB on the line l of interface, a and b are
the distance of points A and B to l. The
6ji 4/3. 7.5.4. (a) a In 2; (b) -y a (here a

is the amount of paint needed for a unit

condition —0 yields surface area). 7.5.5. 10 ji.
dx vx Va + x 2 2

—x c . x
= sm 7.7.1. F(a) + F(b)= + =
or since a, j -f- J-f-
7;
2
Y + +x 2 {c )
2
Ya +
,

2 *
 y .

Hints Answers, and Solutions

, 545

— ——
ab 1 1
-
5
1

. 7.8.8. If JT is the period of the

& Jl v

1 a 1
function y, then we have sin[(o(£+r)
Ojr
+
- _
2 +a]=sin (cof+a), whence (dT=2n, T— — . The
4
= x — ~2~ i r
= -^~. m
(0
7.8.1. y
2 \ x 2 dx pjgj*0
period of the function y 2
is —= 2 a)
; con-

y(i) = <y ~ i 1.33 < y ( 0) +y ( 2) _ 2

sequently, we must find the mean value of
the function y = sin 2
(cot + a) in the interval

7.8.2. (a )y =-^y( 0 )+~y(i)+ j y(2) = from t =0 to = n t -

We get
b Jt/O)

+ q)
y X 1+-^- X 4 = -|- ;
(b)
j
a
= + ra 3 + 1
sin 2 (of
Jl/O

1 |6 1 ,1
~
2
px 2 + qxj\
\
-jr(6 3 -fl 3 )ryp(& 2 -fl 2 ) +
,

3i J [2 2
cos2 ( w * + a)J df — y
9(6 -a) = (6 - a)
[ + (
6* + ab + a*) + 0

l-p( 6 + a) + According to Simpson’s

.
7.9.1. (a) s=
^
V^l + 4.x 2 dx; (b) s =
b

rule (7.8.3) we have y dx =(b — a) y= b 2x2

^
\dx .

(6-«z) [+(«)+ + (i) + + (6)].

0
7.9.2. On
Vl+ei
changing the
0
variable we get

Substituting y~-) » the values y (a), —

y ^
and y (6) and comparing the result with the
s==
1 W~[dz. But
j
(l +++* =
above expression we see that they are identical. K2
2R
„ n 0
7.8.3.
-
F ~ 2R — R
1 f

J
A
r
2 dr ~R
1 /

(
A
r)L
\ |2R H- J
|
+r =z + 2
I (-^T-^r) =
1
R + £)•
z-jr-Tr In 7 r
2+1 4
1

( Therefore, 5 = (
z +
The 2—1 \VT+72
R ('S' + 1) ‘""‘Tir- 4 In
\
/l+^-l/"2 +
value of the force of gravity at this section
z + l ) 1/1
is half as great as that at the earth’s sur- J_ 1+ C2 1
-yin —
142 1

face, F~= 0.5/%. 7.8.4. (a)F (R) F 0 F(2R) = J

= 2
11

1/1++2+1 / + 2 1
.7.9.3.

1 F(R) + F(2R) Partitioning the arc the catenary curve

4 °*
2
= 0.625 F 0 > 0.5F 0
from x =0 £=0.9 and from £—0.9
of
to to £—2 we
(the error is of the order of 10 %);

(b) F --t, 2 *
R )=^ F °' there - find sx 1.043, s2 -
e2 + e~ 2 e 0.9 + e
-0.9

( ) (4
2 2
,
fore
1
F„ ^ T -F
, 4 „ 1
* -109 F
F p° ^
~ 2
T 0 0
6 X 4 ° 216 _1_
2 -x dx . In the last integral we
0.9
0.505F 0 (the error is 1%). 7.8.5
72 + 1
*

set e x = — —— x
_ 2 e dx
m—n
———
can t, then f
- __
7.8.6.
In 772
—— In 72
. For t?2 = + v with v<72.
72 ^ ’

f 2dt f /
’

1
J **—<?-*
l
1 \
(
J (**)2_l
(e

J-^ZT- Finally we
we have In m = In (72 + v)=ln + ln^ 1+— = 72
j 7+rj^*
j obtain s2 ~ 2.624 and = Sl +s ~ 3.667. From 5 2

In 72 + .... 7.8.7. (a) Both mean

the exact formula we get s— 3.627. The error
272 2
72 is approximately l<yQ . 7.9 4 . . 5 ~ 1.146.
values are equal to
1

y ;
(b)
1
—
1
and 7.9.5.
considered
The length of the circle’s arc being

35 — 0946
 a h

546 Hints, Answers, and Solutions

years. 8.1.5. Suppose that at the initial time

t —
0 the number of radium atoms in 10 12
atoms of rock is 0 ; at the time t
10 000 N =
_ 10 0000
it equals 1. Therefore we have 1 =N 0e
2 400
T

whence N0 ~ ~ 65.
4
Similarly, we find
e
that 10 6 years 0 e417 ago N ~
10 181 Clearly, ~ .

the result is preposterious: 10 12 atoms of rock

contain 10 181 radium atoms! No less absurd
is the result of calculations of the amount
of radium 5 X 10 9 years ago. All this proves
that the initial assumption is incorrect on that
35 the present amount of radium can be regarded
r
18 432 as the result of disintegration of the original
supply of radium (when the earth originated)
Accordingly, we obtain the following esti-
(cf. Section 8.4).
mates of the number Jt: (1) Jt ^ ,/• = -(i+
V2
1 Chapter 9
.±) : 3.117; (2) jt~- 1
12
12 160/ V* t
t

4 1_ “

iIo+Jb) -3 - 133; < 3> 31 - 1+ ^

12
9.1.1. A(t)= — v 2 dt; A is negative
V2 (
to
j

35 t
3.138.
160 896 r 18 432
'

ab
since
J
v 2 dt >0 for t > t0 . 9.1.2. The motion
7.10.1. (a) k =- to
)/' ( 2 sin 2 t-\-b 2 cos 2 t)
z
of the body is periodic with the period
2*3 2n
(b) * = -
T ;
we have to find the work for a half
V (
i +* 4 )
3 CD
period. During the first quarter of the period
12a* 2 — h;
(c) = the velocity is positive, and therefore F=
y'(l-f 16a 2 *^ during the second quarter of the period the
=
velocity is negative and F h. During these
7.11.2. F=2it. = -R.
7.11.3. xc time intervals the force is positive, and the
JX
work equals the product of the force by
7.12.1. (a) 2/ max = 2 at * = 0; m in = —2 */
the distance covered by the body; for the
at * = 2; (b) there are no maxima or minima;
first and second quarters of the period we
the curve intersects the * axis at the point
have A 1 = hB and A 2 — h ( — — B) = — hB; the
a:=l-|-|/14^34 and the y axis at point work during the half period isA= +A =
y = —
15; (c) there are no maxima or minima; T
0

the curve intersects the * axis between the

points * 0 and * = —1 and the y axis at
= — 2 hB. 9.1.3. A = Bf 0 (a 1
^
sin co 0 t cos co^ dt.
the point y = 3. 7.12.2. (a) There are three 0
roots; (b) three roots; (c) two roots; (d) one This integral can be easily evaluated if we
root.
recall that sin co 0 * cos 0^ = — [sin (co 0 +coi)£+
Part 2 T

Chapter 8 sin (co 0 — coi) t]. Therefore A= —y- - 5

j
X
8.1.1. T 1600 years. 8.1.2. 176.5 g. ~ 0

8.1.3. 53.3 g. 8.1.4. know that (t) We N — [sin (co 0 + (Di) t + sin (co 0 — «x) t] dt =
N where N 0 is the quantity of substance B /pCOj r 1
- ,
1 costcoo+cox) T
at the initial time t 0. We are interested in = 2 LcOoH-COx
1 "
C0 0 — (Ox co 0 (Ox
the time t { by which there will be (100 — cos(co 0 ——©x)r
1) %=
99% untransformed substance: N (t± )=
«o — Wi
7-
1
J
.
,
In the case where 10
=
cox co ft

—N
100
0.
Therefore,
100
= N^e-tR whence we cannot use the last formula. In this

t
x = f In For radium t = 2400 years and
case, however, sin co 0 £ cos cox t = — sin 2co 0 t,

T
therefore = 2400 In -!~ = 24 (years). Sim- —
t
±
whenc e4=^y^ j
sin 2co 0 * dt =—jr~ (1
ilarly for the three remaining cases we find 0
t 2 ~ 250 years; ts ~ 5500 years; 11 000 ~ cos (o 0 T). 9.1.4. The work of the air resis-
 a

Hints, Answers, and Solutions 547

tance is Ay(t) = aS( g

t
4 The work of v =—
F
m
t. Therefore —•= —mF
dx
£; dx ——
F
tdt %
^
.
dt m
the force of gravity is A 2 (t)= —j- t
2. For

a ball: Ay (1)= -0.00965, Ay (10)= —96.5,

=
whence
]
dx= i \
tdt, since x = Oat = 0. £

Ay (100) =-
965 A 2 (1) 0.177, A t (10)
X 10 s =
,

=
177. For a bullet: 4 t (l) = — = v0 +
1.77, A
2 (100)
-1.18X10-®, Ay (10) —11.8, Ay (100) = =
Finally x -

2m
t
2.
9.4.2. (a) x
^ t t
2
;

— 118 10 3 ; A 2 {i)
X 0.435, 4* (10) 43.5, = = (b) x = Xq-\~ u 0 t -j- t
2
9.4.3. 2.5 m.
=
.

A (100) 4350 (the dimensions of A are in

2

joules — J). 9.1.5. A= ^p-v^ b

_ We 9.4.4. The equation of motion is m—^ = mg,dv

find the power W using the formula W = Fv, whence we = gt 2 2x

^ v
find x
2
’
f
= V
W= - - . Let us find the velocity 4.5 s for £ = 100 m. 9.4.5. (a) £ = 3.6 s;

which the power the

(b) £ = 5.6
Further, for the case s. (a) v =
v (for a given y 0 ) at is
gt -f~ vq. Let v\ be the velocity at the moment
when
dW
— = 0; we get =y of impact; then v\ gti-\-v 0 with = , £i the
greatest, i.e. -
z;i 0
landing time. From the equation v = gt+v 0
=v gt 2
and v2 0 /3. Clearly, we are not interested we find x =v 0 t- Let the ball
in the value v vQ — at which the power is
zero. Of interest here the value v v 0 /3 is = fall from the height H; then x — H at £=£l
(the reader can complete analysis make a therefore 2H— 2v 0 t i -\-gt\, whence t\ —
based on taking into account the sign of
= — v + Vv$ + 2gH
d 2 W/dv 2 ). For y 0
30 m/s and i> =
10 m/s we 0
and v\ = gh + v =
have max W ~
25.75 X 10 5 W. 9.1.6. The work
0

of the force for one period is A Bnf sin a] = Y + 2gH Inv%

the remaining
. the case (b) 0 v = gt—v and
is the same; the final velocity
the power is W = —^— sin a. obtained is the same as in (a).
k
9.3.1. We
take as the x axis the straight 9.4.6. + v 0 t. 9.4.7. (a) x =
line on which the charges are located; let
the charge e x coincide with the origin and 2n
the charge e 2 be at the point x 2a. The equi- =
f
COS G)£, T = (0
/
772 (O
2 ’

librium of the charge at x is only possible /

(b) x— — f— - £
/
sin w£.
*=
-

ifF=-^
dx
0. But
x* {2a — x) 2 ’
if 771(0 777(0

9.4.8. Let the desiredvelocity be u0 Then

0 < x < 2a, if the charge e lies between ex
the law of body’s motion is x =
.

-v 0 {t-
eie 4eye
r„ = i<0,
and „
e 2
.

,
x2 (2a — x) 2
if and
£o)+ tj2m — — (t £0)
2
. Since x —x ± at £ = t1 ,
we
ete
F= — x) if x > 2a. In the first
Z2 1

(2a 2 have xx = v o (ti ^o) +

case condition F
the 0 yields £ 1 2a/3, — = _^ 7?

x2 =
~2a (the second root x 2 is neglected whence vn — -
(*i“ *o)*
since we must have x >• 0); in the cases £i £0 2m
where and £<0the equation F 0 x>2 = F2
has no solution at all. Consequently, there
9.8.1. K = —-t
2m
2 =
F(x-x 0 ). 9.8.2. =
is one position of equilibrium x x 2a/3. Then = 2 2
d2 u
/
sin 2 at; Kmax = f
9.8.3. K=
we calculate at the point x x1 2a/3, = = 2m(o 2 2t77(0 2
mA 2 2
" VA 2 2
+ a) =
(n
/72 1/ sin 2 (g)£
777^4 (0
9.8.5. (a) A ~
we find that if e>0, then -- > 0, i.e. the 2
.

dx 20 4 X 10 7 J, W ~ 2.2 x 10 5
W; (b) A ~ 7 X 10 7
J,
equilibrium is stable, but if e<0, the equi- W ~ 38 X 10 9.8.6. We find the work 4 W.
librium is unstable. 9.3.2. There is one posi- of each separate force. First we determine
tion of equilibrium x x outside the charges. the velocity of the body. From the equation
If the coordinate system is chosen in the
same way as that in the solution to Exer- 777 = at-\- a (0 — = a0 £) we find v =v 0
-}-

then^!= — 2a. > 0, the equi-

cise 9.3.1,
librium
rium is
is stable,
unstable.
but if
If ee 1
ee x < 0, the equilib- —aQ
777
t. The work of force F i is 4i1 =
e

9.4.1. The equation of motion

=0
is m— —
= 0,
F;
j
at
(
vo + ~ t) dt = av 0 d 2
2
.

h ^T
a2 0 4
. Simi-
using the fact that v at £ we find

35 *
 — m - 2 °

548 Hints Answers, and Solutions

,

av 0 Q 2 the mass m acts on the man with the force

larly, the work of force F2 is A2 = —F
(Newton’s third law of motion). We find
a2 04
6m
We now obtain the product of the v i= ,
v2 = — The work performed
impulse by the average velocity / i
= by the force F on the mass
e e m and the man is A~Kx~\-K2 , where
— a0 2 a0 2
K = mu } FH 2

\ at dt
-
2= j « (0 '
- t) dt — -

—^— , i
2m
is the change in the kine-

tic energy of the mass, R 2 = -~ __

FH 2
• , .
2 2M.
v° + J t dt is
,
the change in the kinetic energy of the
j ( £r )
man. Whence A=
v=— 0 = v +2 °
Q2 ' and
change
. 9 8 9
. . . The
2 in the kinetic energy of mass m is
— “a0 / a \ -
=
^
,

therefore y o~r ^ 2^ A Km = mu 0 u 1
/ii>
(
’
+ where vx = t. The
2
,

m
02 We can see that + change in the kinetic energy of the man
y (7
( o+"^T
•
i is
) Mv\
j 2 )^=A + A 2 though hv=£A u I 2 v =£ A2 A Km = +
.

MvqU 2 where ,
.
y2 = p —
M *

(cf. p.
1
At the beginning of the
326). 9.8.7.
experiment the mass m had a velocity v 0
(it was moving together with the train) and
A
.
= —
F 2 t 2 (MA-m)
2

9 - 8 - 10 - w If t\=h+b
d = + &. = — t[=t -
ia then T'
—x -\-vt -{-a and x — x vtti=i.
t'
m Vq f“ TJs
It x
2
akinetic energy K — After the ac-
(b) l 1 -\-a, 1 2 2
-\-
2
then d'=x — x\ — x — x = d.
x .
with == t\ t 2l
2 2 x
tion [of the man the velocity of the mass 9.9.1. For = 30° we have z = 565 m, cp

+ v u K — — 0+ —
L 2
m V V ^max = 81.5 m; for = 45° we have x =
2
rr l)
with v 1 =
• . i
„ ( q>
650 m, y ax = 163 m; for
,

became v0 2
2 ,

ra = 60°, we have q)

Ft *2 = 565 m, y m ax — 244 m. 9.9.2. The equation

The change in the kinetic energy
m of the trajectory is y = x tan w—x - g -— 2 _
2 cp
A = »£«±a£-!El. This is
2 z/gcos
At a given x = 500 m we seek 9 for which
the work performed on the mass by the train 2/max > that is, we solve the equation
and the man together. In order to find the
work performed on the mass by the man, -=q.
~ it yiems
" tan q>-
^9 1
gx 1
COS 2 CD
note that the velocity of the mass with
respect to the man moving in the same — —
Vq —

6
-g
5x 3
2 X2
. Using this fact and the equation
train was zero before the experiment and
became after the experiment. Therefore of the trajectory we find = -^._|^.
y max
the work performed by the man is A 1 = ^~ — Setting v 0 80 m/s, x = = 500 m we determine
y max 135 m. —
n
u
— muf —t*t* .
Ti
It is
•

easy now . a ,
to find the 9.10.1. r= 30 000 km.
2 2/7i

work:. Ao A of_ the locomotive Ao = AK— A-,= 9.11.1. EincL t,he_ secnmL df\rj vatiya. of,
dz 2
mv 0 v 1 = v0 Ft.^The last result can also be the function y (compare it with what = F(z)
is said in pp. 337 and 216, note that
obtained in a different way. Indeed, the loco- F (z) -> 0 as z -> 0 and z -> 00 ).
t
9.12.1. Setting the origin of coordinates
motive’s work is A2 = vF dt, since the ve- at the center of gravity of the rod we get
j
o
a2 u2
locity v =v 0 is constant, we have A 2 — Iq— x 2 odx=o x 2 dx = o-j2> since
t ^ ^
-if 2 - if
i/
0 F dt =u 0 Ft. 9.8.8. Before the experi- m— ol the
,
^
last "result can be written as
j j 2
o
of mass m and that of the
I0 =m 1
. 9.12.2. Set the origin at the
ment the velocity
man are zero. After the experiment the mass joint of the pieces having different densities
m acquired the velocity v x and the man v 2 . so that for the first piece it is Oj at £<0
We find these velocities from the equations and for the second piece it equals o 2 at
° 212
m ^ -=F and M -
dt dt
V2
.
•— F ,
since if the x>0. Then xc = ^~
2 h +
'

lli
. 9.12.3. On
man acts on the mass m with the force F, choosing the coordinate system as indicated
 - —

Hints Answers, and Solutions

,
549

{Vq—Vi)
in the hint we obtain x c = -^L] the moment U — Vx~' 2vi^1+^0 _ 2&vi t.
comparing this
;

of inertia
L
with respect to the origin is
with (9.14.24)
Vi-\-v Q
we find C = - -- --fe-? —— .

aL 4 aL 2
a (x) dx — —^—
f
I — x2 . Since m= — ,
9.14.2. In the case where the resistance is
j
proportional to the velocity the equation of
o
—dv— = g A
2m mL 2 Jc

a=-jY~
L2 \
therefore 1= —q In view of
motion has the form
dt m
v
m
,

where A is the buoyancy. In this case the

(9.12.13) /o = /- mL \, with £,-4 L;
velocity v-l = —a ji
sets in. By the
a am
whence 0 *
Archimedean law A = Vp'g, where V is the
18 volume of the body and p' = Pi is the density
9.13.1. (a) Compare with the hint to Exer- of the liquid. Since m = Fp (p = pb is the densi-
cise 9.12.3. Answer. The center of gravity be-
longs to the median of the triangle and is ty of the body), we have A = and vi =
2/3 of the distance from the vertex along the
median (it coincides with the median point), (l . For p'>p the body floats up
(b) Let a and b be the lengths of the bases of j
the trapezoid, with b, the center of grav- a> (fxCO), while for p' <p it sinVs (v 1 > 0).

ity lies on the straight line connecting the

middle points of the bases at the distance
Chapter 10
g
-Lfo - ..., ~b from the lower base, with h the alti-
a+b (a) The body stops at x 0 = 1/4.
10.1.2.
tude of the trapezoid, (c) The center
of gravity (b) The stopping point is x 0 ~ 0.9. (c) There
lies on the symmetry axis of the half-disk at are no points of stop. 10.1.3. The position of

the distance — R from the diameter which

maximum of u (x) is u max ~ 9.5 at x = 8/3.
3ji
Noting that the left-hand branch of the graph
bounds the half-disk, with R the radius of risesand the right-hand branch descends we
the half-disk. can construct a rough graph of u (x), which is,
however, sufficient for solving the problem,
9.14.1. We write the equation thus -— = (a) From the fact that the sum of the kinetic
dv and potential energies is constant it follows
1 that two stopping points (at which the kinetic
whence accounting the
q—
— vSr for ini-
* /
0 2 ,
energy is zero) are the values of the least roots
"

v\
tial
13 (

condition
)

y(0) = i?
0 we find £ = of the equation x 3 4~ 4a: 2 6. This equa- — =
tion can be solved either graphically or by a
f " '

"o
—— — v2
t"
• (
\
In order to evaluate the
numerical method. We get x 1 1.09, x 2
1.57. The body oscillates between the points x t
~ — ~
and x 2 (b) One stopping point is x
. —2.04. ~
integral, it is sufficient to write the integ- However, this point is to the left of the point
rand in the form —— —r-— —- ^ — from which the body starts at the initial mo-
v\ — v* Vi —v 1

v i-f-y ment. Since the initial velocity is directed to

and find a and b by the method of undeter- the right, the body will move to the right
mined coefficients; see exercises to Section 5.3.) without reaching the point of stop, (c) One
vi +v stopping point is x 2.04. In this case = —
Thus we have In
—v In A -f-2 $vit, the body will move to the left of the stopp-
Vi
ing point and then travel to the right.
where A=
u +^o
i _ Vi +V -.AeW vlt 10.1.4. There are no points of stop. The body
Va—V^ ,

will 'travel ‘to 'fne right, (fij "T’nere are two

p- For very large t we stopp ing points: x1 = 4~ 9/11 , x2 =

1<
+ i — |/^9/ll The motion
body is the oscil-
. of the
have 2 Vlt
e $ >1 and therefore v~v 2 . We lation between the points x Y and x 2 In the .

write the solution to the equation in the

form u =v i
—A—— -j-e'
_2
e
-2$vit
. or u —v i
__ case (a) t -,# +
) V 1-f
4 -f 3z 2
X2
dx. In the case

-
2fivit
20 4- 20a: 2
2Vl
A + e~ W* For rather large t we (b) t = :

*°+j 1 9 — 11a: 2 dx at x <x 1

~ 2 $ vlt 0.5
can neglect in the denominator since e
it is small compared to A; therefore in this
-2$vit = / 204- 20 a 2 :

<x<
case we have v — Vi ~ — 2y x
a

or
and t t 1
J
f
V 9 — 11 a :
2 dx at x 1
~A
 — , +

550 Hints, Answers and Solutions ,

< x 2 where t 1 is the time when body reached

A=— —
,

4
x ± (Note that the integrals remain finite
.
obtain
1 r
\ (1 —x 2
)
dx = -^- whence
though at the stop point the integrand is »

infinitely large.) 1

dC hC 2 (P* X 4 4/kd 3
10 2 . . 1 . (a) x = 2 sin t, x (b);r = cosZ, (c) =
dt 3nk
Denote
3nk
= b;
cos t + 2 sin t. This solution can be written
in the form x = C cos + a), (t where C = l/"5, then — bC 2 and consequently =
a = arctan (
— 2) ~ — 1.11, i.e. x — dt dt
cos (t — 1.11). In all these cases T = 2n.
1
The solution to this equation is
bC 2
10.3.1. (a) Let L be the length of the pen-
_1
whence C
dulum. We know that co
i
/~ m gl
=y and in —— ,
b CCq ’
1 + ’

here C 0 is the value of the amplitude at

= —mL
2
our case 7
l = -^L,I
2 -
— r _ ,
(see Exercise the initial time t= 0 which can be deter-
mined from the initial conditions. (Note that
the same law we have obtained for the velo-
9.12.3). Therefore G)= j/~ city of the rectilinear motion in the case
of resistance proportional to the squared
=—= 3L /IT
V/ ^f- 5A3 V T*
2n , , velocity (see (9.14.15)).) 10.4.2. In this case
T 2n The
the work for a quarter of the period is
value of l corresponding to the maximum fre- — fC, therefore the average power is
quency
-fC-^r — — fC
is determined from the formula
. We get the equation
l
~V/J±
max _i ml
- Since in our case 70
kC
dC
^ ~
2/(70) , dC
— kn and
mL 2 L dt n dt
max =
,
we have Z therefore 2/o)
18
V 18 C = C0
,
therefore -
t. The oscillation s
A kn
-
g 1^18 will cease at time when C — 0, that is,
=V
tx

*°max Knowing 0) max we find

2L
when Zi = - (we suppose that > T).
—
t±
\2n L 2/o)
T min . (b) Here][Z = ,
/ = 70 + 10.5.1. From the first equation of the sys-
EL^max
mL 2 /-
~ ~ tern we get C 0 cos a. Using = —
this
— mL
q) a:
2
mL 2 2 0
ml 2
18
,

9 — 6
o)=
y —— .
expression we readily obtain from the second
—
The minimum period coincides here with equation
^ C U0 sinq>=&
Xq a
—
COi (0i 0)i
that obtained in the case (a), but the point Squaring these two equations and adding the
=( b —
of suspension is different (however, Z max is
as before). result yields C% — y
——
\ 0)] (Oi COi /
10.4.1. Let the oscillations be given (x 0 — taking square roots of both sides
a) 2 ;

by the
= — C(Dsin((Dt +
formula x=Ccos(a>t + a), then of the equation we find C 0 and then find
v a). We use the rela- co v0 x0 a —
—dC F = — hv v
& y
tion kC r
at
=Fiu where , x I I. x
tan cp —
cox coi o)i

— — — hC s o) 3 x0 a
.
—
F ± v= 2
hv \v \ |
sin + a) 3 (o)Z
|

10.7.2. Beats (cf. equation (10.7.5)).

and therefore the law of the change of the
10.8.1. Use the fact that sin 3 ;r =
amplitude C can be written as kC —= sm x sir x -
sin x(i— cos 2x)
— _ sin x
dt
2 “ 2
— hC^afiA, where A = |
sin 3 (coZ-f-a) |
.
sin x cos 2x sin x sin 3x — sin x
Note that sin(o)Z + a) retains its sign when t

varies from t, =
0)
to t2
0)
, with 3
-£-sin x — 1
sin 3x (representation of f (x) in
t2
the form (10.8.15)). Answer. y(x, Z) =
sin 3 (<£>t-{-a,) dt
j 3 1
~r cos r-cosSZsinSa:. 10.8.2. We are
sin x
t2 — — ~2 ti Therefore A=
t_L
_
4
t
4
n-a
.
t2 11
forced to abandon the condition D 0 so =
co
that instead of (10.8.14) we get y {x, t) =
-n Jf
sin 3 (o)Z + &) dt. Changing in the 2 bk
knx
*
kcnt
d b sin
l
knx
X
l
, .

kcnt
10.8.3. Conditions (10.8.2b) yield
+ =x
.

last integral the variable cos (coZ a) we

 ^ 0 ~ g
) — —

Hints Answers, and Solutions

,
551

<p (x) (x) = f (x); q>' (x) — o|)' (x) = 0, whence Chapter 13

y{x ,
f)=-
2
- [/ (a:-(-ci)+/ (a: — ci)] ,
where we 13.2.1. The current decreases according
^nnlsidi' the_ interval t,o_ the Jaw. f== .Atlb£Jim£^ ljL7which„
<0, l)assuming that the function is odd and 9
periodic with period 21. interests us, we have/--jQ- ;0 ;
therefore
C0 X
10.9.3. (a) y-=-^ 4 (cos x
^ + -

^r/o = /V
_<l/RC
. whence e
~ tl,RC
.

—C —
—pcos 3a: ,

r
\

oo
(b) y = (
[
Ci
1
2 \
j
n Taking logarithms yields
9
ln-^Q- =
2 (Ci — C2 sr\ cos [(2k — 1) nx/l] whence
,
£x = RC In -Jq- ~ O.IOS/?^.
9
In veiw of
)
•2 2k —
,

1
the formula we have
last ^ 1 s for
i? =
10 7 Q; *i~10.5 s for R 10 8 Q; ^ ~105 s for =
sin ( nrcx/l
=
R 10 9 Q. The time t 2 when the current ,

(C!+C 2 )2 (-1)" decreases by a factor of 2, is determined in

t2 / RC
n=l a similar way: 0.5j o j o e~ whence * 2 = 1
—
0.6937? £7. Here * 2 ~6.93 s for i? 10 7 Q; =
£2 ^69.3 s for R — 10 8 Q; 2 ~ 693 s for *
Chapter 11 7? = 10 9
Q. 13.2.2. We use formula (13.1.11)
= and find (P Ci + Tr + (P^ = 0, 2
whence q) R =
1.48p 0 p=1.13p 0
P=3.67p 0
where p 0
11.3.1. p
the air pressure at the earth’s
is
; ; ,
— (Tc + Pc 1
(
2
)- The curr ent in the circuit is

surface. 11.3.2. The dependence of the pres- d(p

sure on altitude of the form p -gh/b the same everywhere; therefore / =C i

Cl
—
is 0e
--p

where 6 = 3
RT
— For the temperature „ —dtd(fc
2 <Pr_ — (
Pci + Pc (
2
1 .
C2 R
~~R R
* We obtain the
— 40 °C we have T = 273 — 40 = 233 K, d(P
C, *Pc 2
6
8.3 X 103 X 233
on / — 6.6 X 10 4 cm 2 /s 2 in
equations
dt RCi ('Pc. + Vc.)* dt
29.4 ;

this case 77 = —b = 6.6 km. For the tempera- RC 2 (TCi + T^)’ adding them yields

ture 40 °C we have T=313 K, 6 = 8. 8X 1 + where we

77 = 8.8 km.
set
10 4 cm 2 /s 2 , 11.3.3. From the dt
'
RC dt
dT
equation ~^p~ = ~ aT o it follows that T= r C \C2
(C is the total capacitance of
— aT oii~C^2
/-t ,

0 h-{-C, where the constant C is deter-

mined from the condition T = T 0 at 6 = 0, two capacitors connected in series, C i and
whence C = T 0 therefore T = T 0 (l — ah). ]
C2 ).
Since q> Cj Tc 2 = a at + * = * ast

The basic equation for determining the den- tlRC

dp
equation yields ^ Cx Jr^>
C2 — ae~ . Clearly,
sity is
~[h~—
— £P- We use Clapeyron’s equa- d<? d ^c
RT
Ct n
z
2
=0 , or
tion p = 10 3 p -jy- and substitute the ex-
dt dt
-
C 2 cpC2 = 0. Therefore ^iTci
' =
RTAi — och)—
)

m i
p = 10
pression for
£
T to find 3p where is a constant. Using the initial
.

conditions cp
Cx = a, q) c =0 at ^ = 0, we find
RTp ~ t/RC
Set 10 3
M =
*o, then p = p6 0 (1 — och), A = Cia. Thus, q?
c +q> c =ae ,
C x q) c —
r<

whence rp=-,
P
The C2 ^>c 2 = C 1 a. Whence (
Pc t = a ~^~Zfc~ X
b 0 (l
— ah) .. differential equa-

dp ci
tion for p assumes the form-^- "— — s p 'l+4±e- -t/Rc\
a (-1 +
dh 6 0 (1 — ah) * Ci *
~ci+c2 r Vc

or
jlp
= b — gdh we
e
-t/RCy i3 2 3 We label all the quantities
# # #

— a6) o (1
’ integrating get
belonging to the circuit before all linear
dimensions are increased with the index 1
In p = "ft a
In (1 — ah) + C ,
whence p ~ and after the increase with the index 2. Then
A(i — ah) §/b ° a
, with A = e c p p 0 at . Since = zS 2 &n 2 S 1 _
R 2C2 C2 -
h — 0, A = p0 \ therefore p p 0 (1 =
ah) g / bo(X — .
-RiCi ,
4ji d 2 Andxn
11.3.4. p =p 0 (1
— 0.037x 10 5
6)
3 46
p
-

;
nCi, R2 = p
nl l _ Ri
Therefore
1.13p 0 ; p = 1.44p 0 ; p=-2.97p 0 . n 2 Oi
 ) 2 ) — x

552 Hints Answers , and Solutions

,

—D - n c 13.10.1. cp (t) = e~t + te~ l = cp (*)

r2 = =R C1 = T1 — 0Me~*- +
\

1 1 . The time constant — 0.03e -5 - 83f +1.03c“ 0 17 *; -

q>(*)= Qt

l.Ole- 0 1 *. 13.10.2. y(t)=e-t

-
+ 2le~t\ = cp (*)
has not varied.
13.8.1. Let the potential difference be cp.
— 0.37e-3- 7 3* 1.37e-°- 27 L +
For the circuit shown in Figure 13.8.3 we

have q>E+q>r + <Pc = 0, or Z + = £„ q> Part 3

d(p d2i Chapter 14

E= —E As = C
p
(since (p 0 ). j ,
LC ~jp-+ ,

14, 1.1. (a) u = x — 3 xy — 3x-f-l, u—3

s 2 2 y-

<p =E 0, LC
i.e.
— -E (CP o) or LC X us (b) u=
x -\-x-\-xy 3 2
3 y;
— 22
(l-J-a:
2
y )
-\-Ax 2 y 2
’

d2 z
_ with 2 = —E cp 0. The solution to y- -y*
u =
+ z — y + Ax y
(c)
2 2 2 2 2

the last equation has the form z = ( 1

~2
)

Z? cos (G)*-(-a), with (0=1 l\ LC', therefore

r
( 2 )
xn y
2
+ ( ^ )
* n-4 i/ 4 + • • »
v = ( \ )
x
cp = Z? cos (a)*4-a)+#o- But cp 0, y 0 at = =
* = 0; using this fact we get =
E Q a 0. B~ — xn~ y y — g xn ~ y s 3-
f- xn
~h
y
b — ..., where
Finally we have = £' (l — cosco*). The value
,
^ j ^ ^ j
cp
0
Tmax = 2Z£ corresponds to the equation — k\ (n — ‘

0 ----- is the number of combina-

? -

cos(D*=— 1 i.e. = n/(d = T/2; it is attained

l )
,
t \ k / k)\
at half period of oscillations. 13.8.2. The tions of n elements taken k at a time. 14.1.2.
energy of capacitance W— /2 = 46'£’§/2= C'cp 2 Formulas and (14.1.6b) can be proved
(14.1.6a)
2 CEl\ the energy released by the voltage proceeding from the equations z~ 1 z = i and
source P = qE = CyE = 2CEI. 0 0
(zi/3)3 =
Z the general formula (14.1.6) follows
;

from the fact that it is valid for any rational

13.9.1. f(t)= e~’Kt sin o)*, with 72 =
p/^ = (4/^) p, where the cases of p being
positive or negative must be considered sepa-
X — —-
ZL ,
o 2 = —LC ;
X2 . For the given three
rately.
r (cos a -f- *
14.1.3.
sin a) then
Use the
y
rc
fact
= p (cos p +
that
*
if c

sin p),
—
cases we have = — 1.0025e-0 * 05i sin
=
j(t)= — 1.031*- sin 0.97*; 0 - 25 *
j (*)
j (*)
=
t;
with p = y
/r
r, and P =P 0 H — »
with p 0
— lA5e-°- sin 0.87*. 5t
13.9.2. =
— — sin G)*j
j (*)
,
& = 0, 1, 2, ..., n — 1. 14.1.4. In accord-
/o |cos co* e~^ t \ formulas for X
ance with the results of Exercise 14.1.3 we
and
cases
(o see in Exercise 13.9.1.
we have / (*)=e-°- 05 — cos + 0.05 sin *
(
For the given
t t);
have
3ox/4,
y — ==—cos P+
5ax/4 (
1
3jx/4),
i sin p
7ji/4 (=
with
— p
Jt/4);
= Ji/4,
note
j (t) = e~°- 2bt ( — cos 0.97* + 0.26 sin 0.97*); also that in all cases cos p and sin p equal
j (t = e-°- bt (
— cos 0.86* + 0.58 sin 0.86*). ±1^2/2.
13.9.3. If H
very great, then the currentis
14.3.1. If lnii=w, then u = e w and u z —
e ~
flowing through the resistance is small, i.e. e wz \ use this fact. 14.3.2. Clearly e
the current flows mainly through the induct-
(2.7)
2 7
*
> > 1 i
l
while numbers ei = cosl +
ance. Therefore the greater the R the more
\ (

isinl, where the angles are measured in ra-

close this circuit is to that shown in Figure ln l =
( = e
e e eJt/2 ?

13.8.1, where cp cp cos ((D*-j-a).

0 If R is = dians) and e( )
)
cannot be i

great, we can assume that cp in the circuit compared (we can compare in magnitude only
of Figure 13.9.2 has the same form, while
real numbers and not arbitrary complex
dP numbers).
cp 0 varies slowly with time. Here -jj'— — A; 14.4.3. (a) The circle; (b) the hyperbola.

but h = Rj 2
lJ where j1 is the current flowing We have for (a) and (b) * = tan-^- and t =
through the resistance i?, /1 = -2-
R ;
therefore
tanh — \b
respectively if at the same time the
1
cpg cos 2 (co* + a) and h= Thus, curves are also given by the “standard”
R R ZR
.

parametric equations; (a) z cos(p, y sincp; = =

(b) x cosh — sinh\j?. ?/ =
- Noting that P- we find
S' 2
,

= d(p0 -
~ Whence
dq> o _ Chapter 15
dt dt 2 R dt 15.1.1. All derivatives have the form
2

2 RC
P — e” 1 /* ,
where P {z) is the polynomial
IRC
cp
0 . Therefore (p 0 = Ae ;
X=
2 RC'
^ j
=0 they are equal to zero
(in variable z ); at z
 ~ ) x

Hints, Answers, and Solutions 553

due to a high rate of decrease of the function streamline of the liquid whose direction at
y = e~
i ^ x2
as x-*-0 (cf. Section 6.5). For each point is y y(x, y)d ^ is equal to zero, =
example, /' (x) 2/x3 e
i / x2
with = (b) Let T be a small arc with the endpoints

t=
3

than
— 1/x
the
2
but e l
,
decreases more rapidly as t
function 2 t 3 / 2 (— 2/x 3
,

— oo
P = P(x y ), Q = Q (x + dx,
,

R(x -j-dx y); consider the

y dy) and R
flow along the bro-
+ =
—
,
in-
15.2.1. —5 oo.
| | | |)
ken line PQR assuming that the vector of the
creases as t
flow rate is v ( x y) and prove that (accurate
If w— lnz, then u = In \
r x 2 -\~y 2
to infinitesimals of the first order) this flow
,

and i; = arctan~ ,
whence -4~ — *4“ =r- equals dty ty (x-{-dx, y-\-dy) —
\|? (x, y). —
x dx dy 17.3.2 Use the fact that the tangent vectors
to the lines cp constant and oj) =
constant =
= =x
x 2 ~ry 2 dy dx 2 -\
-y 2
. 15.2.2. Use
at point M (x, y) are v<p = (
~ j
the fact that ,
v± — .
dx dx and V * =
IT) ie V * W> ~vx)
(“IF ’
>
‘ =
16.2.1. and v^-, =(v x v y ), and then take advantage
,

of the condition of two vectors being perpen-

Chapter 16
dicular to each other. 17.3.4. (a) When a is
Near the point x = x 0 we expand real, the streamlines are straight lines parallel

the function to the x axis; equipotential lines are parallel

cp (x) in a series retaining only
two first terms cp (x) ~ cp (£ 0 )-f- (p' (x 0 ) (x — x 0 )~ to the y axis; when a is pure imaginary, the
cp' (x 0 ) (x —x 0 ). Thus denoting cp' (£ 0 )=c,
flow is conjugate to that described, (b) When
x — x = y, 0 we get 6 (cp (x)) =6 (cy) = a is real, the streamlines are hyperbolas
=
xy constant; equipotential lines are hyper-
6(y) ^ In a more bolas x 2 y
2 — =
constant; the flow rate at point
T7T M is proportional to the distance OM.
general case the function 6 (op (z)) is the sum
(c) When a is real, the streamlines are the
of similar expressions valid for the zero’s
circles tangent to the Ox axis at point 0\
of \f>. 16.2.2. The function (p(^) sina: van- = equipotential
17.4.1. lines are circles tangent to Oy
ishes at x 0 kn, k 0 ±1, ±2, .., = = ,
.
axis at point 0. (d)For the stream- w—\uz
<P' (* 0 )
I
cosxq cos/cji
I
=1; therefore
I |
= | |
= lines are straight lines emanating from
-f-OO -f-oo
point O equipotential lines are circles with
6 (sin x)
16.3.1.
= 2— &(x — kn) and o|3 (x) X center at 0.
;

j Equating to zero (equilibrium) the

h= oo — oo
4-00
sum of the projections of forces onto the
direction of the y axis we get (for the case
6 (sin x) dx— ^ ^(fcn).
of small deviations, y <C l) 1 —k ——
Xi
(a) y' (x) = — 6 (x — 1 (b) Since
k ——
— — z= 0. Whence
1).
y ( 0) =
+16.4.1. 1 y ( — 0) = 0, we have Ay = 1 at l xx
x = 0. For
1

x =?= 0 we have y' (x) =

e i/x (l-x i)
"" and final] y y'(x)=S(x) — yi <*i)=i —x X1
l 2
( l -|- g l / X )2
>
/ ( Xx l ± kl

e l/x
^2(l_|_gl/X)2-
I (t)l(b-A)
Multiply the right-hand side of X (l —x
(16.4.8) by siny and use the fact that kl
±)
, 0 ^ x ^ xi.
y(x, * x ) =

/ —xl \ I t ,
k \

cos 17.3.1.
kx sin ~- =
y- j^sin a; — sin | \ l— X\ 1 1 \ xx l — 1 )
j
x — x)
1 (l

kl
xi ^x^ l.

The function y (x, x t ) is called Green's junc-

tion in the problem on the string. For an
Chapter 17
arbitrarily distributed force j(x) the deviation
of the string is given by the formula y(x) =
the streamline, then by the
If T is
l
definition of the function the tangent vector
/ (#i) y (
x ,
x t ) dx l . Note that Green’s func-
t to T (i.e. to the line J
^ j o
if> =
constant) coincides with the vector v of tion helped us find the solution to the func-
the flow rate which means that along the tion y (x) when we did not even know the
 )

354 Hints, Answers, and Solutions

position remaining zero. The solution to the

•equation of its behavior ^this equation has equation of oscillations with such initial
conditions at t x has the form =
the form »(0) = 0, y(l) = o) .

0 for — oo< t <t,

17.4.2. The general solution to the equation
x t) =
sino)(£ — t)
t, 1
witho ut th e force is x (^—Cx sinco^+6’ 2 coso)^ (
v '
for t<£<4-oo.
=y
1

k/m, Cj and C 2 are arbitrary constants- { 7TCG)

Since the delta function differs from zero only

C0.--^° x
at t = x, the solution to the equation with In other words, in formula (*)
v 7
C Ax = -

the delta-like force and the state of rest at mo) ’

t = —
oo has the form _
C2—
sin cot
,
mio
0 The solution to the problem with an
for — oo < £ <T, arbitrary force / t is
x (t) = cos at
(*)
(

Ci sin + oo
{
for t < < + oo;
t
X(t)= f (T)x(t, T )dX
j
the delta-like force imparts a unit impulse — 00
to the body, therefore after the action of

—
t
the delta force the body at rest acquires an
=— the
= /(T) Sin[0)(i — T)]dT.
initial velocity z;
0 ,
initial |
 Name Index

Abel, N.H. 527 Einstein, A. U t, 16, 286, 287, Kemery, J.G. 523
Aleksandrov, V.V. 530 321, 323, 393, 410, 412, Kepler, J. 121, 274
Archimedes 120 414, 523 Khinchin, A.Ya. 523
Argand, J.R. 474 Eisberg, R.M. 530 Kitaigorodsky, A. I. 530
Arnold, V.I. 522 Euclid, 121 Klein, F. 521, 527
Arrhenius, S. 17, 395, 396 Euler, L. 210, 211, 212, 379, Kline, M. 473
384, 385, 386, 474, 475, Kolmogorov, A. 523
483, 518 Kopylov, G.I. 531
Barlow, V. 524 Krylov, A.N. 16
Barrow, I. 124 Kurchatov, I.V. 294
Basov, N.G. 416 Faraday, M. 453, 454
Bell, E.T. 531
Fedorov, E.S. 524
Feller, W. 523 Lagrange, J.L. 123, 124, 210, 212,
Bernoulli, D. 209, 211, 378, 380,
Fermat, P. 121, 122, 130 521, 527
384, 385, 386, 475, 518
Bernoulli, Jacob 209, 210, 231 Feynman, R.P. 530 Landau, L.D. 530
Fikhtengol’ts, G.M. 530 Landsberg, G.S. 530
Bernoulli, Johann 209, 210, 211,
Flyorov, G.N. 289, 294 Lang, S. 530
483, 484
Bers, K. 530 Fokker, A. 523 Lax, A. 530
Bessel, F.W. 484
Fomenko, A.T. 522 Lax P. 530
Bohr, N. 284, 406 Fourier, J.B. 212, 379, 386 Leibniz, G. 10, 113, 122, 123,
Boltzmann, L. 394 130, 209, 210, 211, 261,
Bolyai, J. 521 Galilei, G. 113, 121, 322 483, 484
Bombelli, R. 474 Galois, E. 524, 527 Leighton, R.B. 530
Born, M. 528 Gauss, K.F. 213, 380, 474, 484, Lerner, L.S. 530
Bose, S.N. 410 521 L<§vy, P.P. 523
Briggs, H. 146 Gel’fand, I.M. 488 L’Hospital, G.F.A. 210, 211
Bronsthein, I.N. 170 Gibbs, W. 112 Lie, S. 527, 528
Brown, R. 393 Gnedenko, B.V. 523 Lobachevsky, N. 521
Btirgi, J. 146 Goethe, J.W. von 112 Lomonosov, M.Y. 391
Burstein, S. 530 Gradshteyn, I.S. 170 Lorentz, H.A. 17
Bueche, F.J. 530 Green, G. 515 Luzin, N.N. 13
Grenville, V. 13
Guldin, P. 274 Mach, E. 350
Cardano, G. 473,474 Maclaurin, C. 209
Cassini, G.D. 230 Mandelbrot, B. 15
Cauchy, A.L. 125, 212, 484, 520, Hahn, 0.294 Manin, Yu. I. 522
527 Hall, A.R. 125 Marconi, M.G. 454
Cavalieri, B. 272, 274 Halliday, D. 530 Marion, J.B. 530
Chung, K.L. 523 Hamilton, W.R. 525 Markushevich, A. I. 252
Cooper, L.N. 530 Heisenberg, W. 526, 528 Marsden, J. 530
Courant, R. 530 Hermite, Ch. 15 Maxwell, J.C. 394, 453, 454,
Curie, M. 283 Hertz, H. 358, 454 518, 525
Curie, P. 283 Helbert, D. 213, 521, 528 Mendeleyev, G. 287
Hoffmann, L.D. 530 Mikusinski, J. 488
Hooke, R. 115 Moise, E.E. 13
Huygens, Ch. 122, 123, 210, 270 Moivre, A. de 474
d’Alembert, 211,
J. 212, 352,
379, 380,
384, 385, 386, Monge, G. 212
474, 475,
483, 518. 521 Infeld, L. 524 Mosteller, F. 523
Descartes, R. 121 Ioffe, A.F. 397, 398 Myskis, A.D. 530
Diderot, D. 211, 521 Jeans, J.H. 524
Dirac, P.A.M. 414, 487, 528 Jensen, J.L.W.V. 238 Napier, J. 146
Downes, Jr., F.L. 13 Joliot-Curie, F. 294 Natanson, I.P. 530
Dubrovin, B.A. 521 Joliot-Curie, I. 294 Nernst, W. 17
Dwight, H.W. 170 Jordan, C. 527 Newman, M.H.A. 528
Newton, I. 10, 16, 113, 122, 123,
Ehrenfest, P.S. 523 Kaempffer, F.A. 530 130, 209, 321, 484
JShrenfest-Afanasjewa, T.A. 523 Kamerlingh-Onnes, H. 438 Neyman, J. 523
556 Name Index

Nikolsky, S.M. 530 Rogers, E.M. 530 Tartaglia, N. 193, 473

Niven, I. 458, 530 Rolle, M. 259 Taylor, B. 209, 385
Novikov, S.P. 522 Rontgen, W.K. 454 Thom, R. 522
Rourke, R.E.K. 523 Thomas, G.B 523 .

Orear, J. 530 Rufftni, 527

P. Thomson, W. (Lord Kelvin) 15,.
Rutherford, E. 283 388, 515
Pappus of Alexandria 274 Ryzhik, I.M. 170 Torricelli, E. 121, 204
Pascal, B. 126 Townes, Ch. 416
Pasichenko, P.I. 530 Sands, M.L. 530 Tsiolkovsky, K.E. 334
Perelman, Ya.I. 141 Schonflies, A.M. 524
Perrin, J.B. 393 Schwartz, L. 488
Petrzhak, K.A. 294 Seaborg, G. 287, 288 Viete, F. 474
Planck, M. 410, 523 Sears, F.W. 530 Vygodsky, M. 530
Poincare, H. 521 Semendyayew, K.A. 170
Poisson, S.-D. 527 Semyonov, N.N. 396 Weierstrass, K.T.W. 484, 520-
Polya, G. 17, 209 Shervatov, V.G. 472 Weinstein, A. 530
Pontrjagin, L.S. 530 Shubnikov, A.V. 524 Wessel, C. 474
Pope, A. 124 Sikorski, R. 488 Weyl, H. 528
Popov, A.S. 454 Simpson, T. 255 Wigner, E. 519, 528
Poston, T. 522 Smirnov, V.I. 530
Potapov, M.K. 530 Smoluchowski, M. 523
Prokhorov, A.M. 416 Snell, J.L. 523 Yaglom, A.M. 58
Sobolev, S.L. 488 Yaglom, I.M. 58, 327, 509, 527
Reid, C. 528 Steiner, J. 61 Yang, C.N. 519
Resnick, R. 530 Stieltjes, T.J. 15 Young, H.D. 530
Reynolds, O. 346 Stirling, J. 249
Riemann, G.F.B. 212, 213, 483, Stokes, G.G. 346 Zeeman, E.C. 522
484, 485, 520, 521 Strassmann, F. 294 Zeldovich, Ya.B. 13, 530
Robbins, H. 530 Struik, D.J. 531 Zemansky, M.W. 530
Robinson, A. 126 Stuwart, I. 522 Zhukovsky, N.E.
Subject Index

Abel groups 527 circle, involute of 270 stretching away from x axis
abscissa 26 length of circumference of 53
absolute maxima (minima) of 261, 264 translation of 50
function 226 osculating 267 vertices of 267
absolute space 321 circular frequency of oscillations curvilinear coordinates 520
absolute time 321 358 curvilinear trapezoid 99
absolute value of complex number Clapeyron ideal gas law .388 cusp 223 ' 1

459 coefficient of linear expansion 71 cuspidal maxima (minima) of

acceleration 69, 89, 113 coefficient of volume expansion 72 a function 224
activation energy 395 common logarithms 146 cycloid 61
algebra, fundamental theorem of commutative groups 527 length of 261f
461 complex electric held strength 511
algebraic equation of nth degree complex number(s) 459
461 absolute value of 459 damped oscillations 441, 466
algebraic function 139 amplitude of 460 daughter element 289
alternating current 444 argument of 460 decay time 431
amplitude of complex number geometric representation of decrease of function 85
460 definite integral 96
amplitude of oscillations 359 460 deformation 115
analytic functions 477 modulus of 459 delta function 487
one complex variable 480
of phase of 460 density, linear 337
analytic geometry 121 trigonometric form of 460 distribution-of-gas 516
angle 27f, 61f complex plane 460 dependence, empirical 26
angle of departure 328 complex potential 510 derivative 7 3
antiderivative 106, 198 of electric held 511 of composite function 134
approximations of functions 178 complex velocity of flow 510 of implicit function 158f
Archimedes spiral 120 composite function 134 of inverse function 136
arc length 260 derivative of 134 on the left 224
area bounded by curve 95 concave curve 235 logarithmic 147
area of triangle 62 concave downward curve 91, 235 of order n ill
argument of complex number 460 concave function 235 partial 154
argument of function 23 conductance 419 on the right 224
arithmetic mean 95, 236 cone, volume of 27 If second 89, 176
arithmetic progression 141 conformal transformations 483 total 161
astroid 61 conjugate flows 511 deuterium 286
asymptote(s), horizontal 278 conjugate numbers 459 Dido’s problem 218
of a hyperbola 36 conjugation of curves 225 dielectric constant 425
vertical 278 constant, Euler’s 246 differential 127
atmospheric pressure 117 continuous spectrum 386 differential equation(s) 116, 198
average (or specific) curvature convective rate of temperature general solution to 204
265 variation 160 initial condition for 199
average velocity 68, 149 convergent series 188 order of 198
Avogadro’s law 388 convex curve 234 partial 453
Avogadro’s number 389 convex function 234 partial solution to 204
axis, polar 28 convexity 214 second 131
axis of abscissas 26 of curve 41 total 157
axis of ordinates 26 point of 214 with variables separable 199
convex upward curve 91, 234 differentiation of composite fun-
ballistic pendulum 360 coordinates, Cartesian 26 ction 134
base-10 logarithms 146 polar 28 differentiation, rules of 132-142
beats 374 corner point 222 diodes 454
binary logarithm 58 cosine, hyperbolic 168, 261 Dirac’s delta function 487
bit 58 cosine line 420 direction of current 418
Boltzmann constant 389 crankgear 373 directions, field of 201
boundary conditions for diffe- critical mass 298ff direct proportionality 34
rential equations 376 critical point 88 disintegration, prob bilit f 281,
boundary value of independent critical temperature 234 289
variable 194 cross section 399 distance 27ff, 67
Boyle’s law 388 effective 403 distance from point to straight li-
breakdown potential 427 cubic equation 44 ne 63
cubic function 57 distribution, steady-state 160
curvature of a curve 265 divergent series 189
Briggs logarithms 146 curvature at a point 265 domain of integration 98
Brownian movement 393 curve(s), average (orspecific) cur- dummy variable 100
bulk expansion coefficient 72 vature of 265
catenary 261
calculus, differential and integ- center of curvature of 267
ral 101, 120 concave 235 e (the number) 142
capacitance 117, 419 concave downward 91 effective altitude of figure 243
Cardano formula 473 of constant curvature 268 effective cross section 403
cardioid 61 convex downward 235 effective length of figure 243
Cartesian coordinates 26 convexity of 41 effective width of curve 243
catastrophe theory 522 convex upward 91, 234 einstenium 287
catenary curve 261 effective width of 243 Einstein-Smoluchowski equation
length of 261 equipotential 510 523
cathode-ray oscillograph 445 integral 201 electric circuit(s) 418
Cauchy-Riemann equations 479 parametric representation of electric current 418
Cavalieri’s theorem 272 60 electric field strength 517
center of curvature 267 radius of curvature of 267 electricity, quantity of 117
center of gravity 274, 341 rectifying points of 267 electric potential 418
center of symmetry 43 shrinking along x axis 54 electromagnetic field theory 511
centroid 238, 274 shrinking to x axis 54 electromagnetic theory of light
chain reaction 294ff smooth 215 453
chain rule 134 Steiner 61 electromotive force 420
characteristic curve 454 stretching along x axis 53 elementary function 169
558 Subject Index

ellipse 53 function (s) 23 group(s), Abel (or commutative)

ellipsoid of revolution, volume algebraic 139 527
of 273 analytic 477 continuous 526
emission of electrons 397 function (s) discrete 526
empirical dependence 26 approximations of 178 Fedorov 524
empirical formula 26 argument of 23 Galois 528
energy, activation 395 of a complex variable 212, Klein 527
kinetic 115, 320 213 Lie 527
law of conservation of 354 composite 134 Shubnikov 524
total 354 concave 235 symmetry 524
energy density 408 convex 234 growth of function 85
energy flux (or irradiance) 399 cubic 57 Guldin theorems 274
energy of inductance 435 cuspidal maxima (minima)
energy of radiation 409 of 224 half-life of radioactive atoms 282
epicycloid 61 decrease of 85 half-width at half-maximum 451
equation(s), algebraic of nth dependent on a parameter half-width of resonance curve 451
degree 461 173 harmonic analysis 380
Cauchy-Riemann 479 elementary 169 harmonic functions 315, 480
cubic 44 even 56 harmonic mean 236
differential 116, 198 exponential 144 harmonic oscillations 360
Einstein-Smoluchowski 523 firstapproximation for 177 hertz 358
Fokker-Planck 523 Green 499 higher mathematics, fundamen-
Kolmogorov 523 growth of 85 tal theorem of 101
of mathematical physics 212, harmonic 315, 480 homothetic transformation 42
213, 517 hyperbolic 469ff Hooke’s law 115
quadratic 39 implicit 158 horizontal asymptote 278
of straight line 31 increase of 85 hyperbola 35
equation of state, van der Waals infinitely valued 152 of degree n 45
233 inverse 47, 135, 467 equilateral 35
equilateral hyperbola 35 inverse trigonometric 151
equilibrium, stable 313 linear 30
of order n +
1 45
hyperbolic angle 471
thermodynamic 407 linear approximation for 177 hyperbolic cosine 168, 261, 469
unstable 313 linear-fractional 51, 57 hyperbolic functions 469ff
equilibrium intensity of light locally monotonic 50 hyperbolic sine 168, 469
408 longarithmic 145, 468 hyperbolic tangent 470
equipotential curve 510 maximum of 38, 83, 226 hypocycloid 61
escape velocity 332 mean value of 253
essential singular point 481 minimum of 38, 83, 226
ether 518 monotonic 49 ideal-gas isotherms 233
Euler broken line 201 multiple-valued 152, 159 ideal gas law 233, 388
Euler formula(s) 465, 466 nonsmooth maxima (minima) imaginary axis 460
Euler’s constant 246 of 224 imaginary power 463
evaporation of liquids 396 odd 56 imaginary unit 459
even function 56 order of increase of 196 implicit functions 158
evolute 269 parametric representation of derivative of 158f
of a cycloid 270 136 improper integral 174
of a parabola 269 period of 55 impulse of force 317
exponential curve, arc length of periodic 55 unit 335
264 pole of 505 increase of function 85
exponential function 144 power 46, 138 indefinite integral 104, 198
extrapolation 26 primitive 106 independent variable 23
quadratic 37 index 29
rate of change of 177 inductance 421f
rate of growth of 86 energy of 435
Fedorov group 524 second derivative of 89 induction 453
Fermat’s principle 125, 221 signum 491 inequality, Jensen’s 238
fermium 287 sine integral 187 inertial systems of coordinates
held, electric 517 stream 510 322
magnetic 453 trigonometric 148, 466 infinitely valued functions 152
held of directions 201 two-valued 50 infinite series 181
first approximation for functions of two variables 154 infinite slope 33
177 functional 218 information, theory of 58
first theorem of Pappus 274 functional analysis 213 initial conditions for differential
flow, complex velocity of 510 functional relationship 23 equations 199, 376
irrotational (or votex-free) fundamental theorem of algebra initial condition for radioactive
509 461 decay 281
plane-parallel 509 fundamental theorem of higher initial ordinate of straight line
potential 509 mathematics 101 31
rate of 203 fundamental tone 385 initial phase angle 359
fluent 124 initial velocity 69
fluid dynamics 350 instantaneous velocity 69, 149
fluid mechanics 509 Galilean transformations 325 integral(s), definite 96
fluxion 124 Galois group 528 Fourier 386, 495
Fokker-Planck equation 523 gaseous state of matter 234 improper 174
forced oscillations 366, 497f gas pressure 388 indefinite 104, 198
force, impulse of 317 general solution to differential tabulated 168
work performed by 114 equations 204 integral curve 201
force of gravity 310 general theory of relativity 323 integrand 98
formula, empirical 26 geometric mean 236 integrating functions dependent
Fourier analysis 386 geometric progression 141, 188 on a parameter 173
Fourier analyzer 386 glow discharge 427 integration, changing the varia-
Fourier integral 386, 495 graphs, transformation of 50 bles in 168
Fourier series 386, 495, 498 graph of function 30 domain of 98
fourth-power law 410 gravitational paradox 325 limits of 97
fractional (rational) numbers 458 gravity, force of 310 by parts 167
frequency 55, 358 grazing fire 329 by substitution 168
natural 366, 444 Green function 499, 515 variable of 98
Subject Index 55 £

internal resistance 420 Mach number 350 oscillation(s) 357

interpolation 26 Maclaurin’s series 182 damped 466
inverse function(s) 47, 135, 467 magnetic field 453, 517 forced 366, 497f
derivative of 136 maser 416 period of 358
inverse proportionality 35 mathematical analysis 101 oscillatory circuit 439
inverse trigonometric functions mathematical modeling 519 oscillograph, cathode-ray 445
151 maximum of function 38, 83, loop 445
involute 269 214, 215 single-beam cathode-ray 447
of a circle 270 absolute 226 overtones 385
irrational numbers 458 cuspidal 224
irrotational (or vortex-free) flow at endpoint 222
509 local 44, 87 Pappus- Guldin theorems 274
isotherms 233 nonsmooth 224 Pappus theorem, first 274
ideal gas 233 relative 44 second 275
van der Waals 233 maximum of function parabola 37
smooth 215 arc length of 263
Jensen’s inequality 238 mean, arithmetic 95, 236 of order n 42
general 240 geometric 236 osculating 131
kinetic energy 115, 320 harmonic 236 semicubical 45
Klein groups 527 mean free path 391 squaring of 120
Kolmogorov equation 523 mean value of function 253 parallel lines 65
mendelevium 287 parameter 60
Lagrange’s theorem 258 method, rectangular 96 parametric representation of
laser 416 trapezoid 95 curve 60
law(s), Avogadro’s 388 minimum of function 83, 214, parametric representation offunc-
Boyle’s 388 215 tion 136
Clapeyron ideal gas 388 absolute 226 parent element 289
of conservation of energy cuspidal 224 partial derivative(s) 154
354 at endpoint 222 of second order 156
fourth-power 410 local 87 partial differential equations 458
Hooke’s 115 nonsmooth 224 partial solutions to differential
ideal gas 388 smooth 215 equations 204
of large numbers 404 modulus, of base a with respect pendulum, ballistic 360
of light refraction 221 to base b 56, 145 physical 362
of motion 113 of complex number 459 simple 361
Newton’s first 316 Young’s 115 period of function 55
Newton’s second 316 Moivre formula 461 periodic function 55
Newton’s third 320 molar gas constant 233, 388 period of oscillations 358
Ohm s 24, 36
moment of inertia 339 perpendicularity 62
of radioactive decay 281 momentum 317 pi (the number) 264
statistical 523 monotonic function 49 phase of complex number 460
Stefan-Boltzmann 410 locally 50 phase space 521
Torricelli’s 204 motion, law of 113 photons 406
law of radiation, Stefan’s 410 uniform 67f physical pendulum 362
leakage resistance 437 uniformly accelerated 69, 113 Planck’s constant 406, 410
Leibniz and discovery of calculus multiple-valued function 49, 152, Planck’s radiation formula 410
125 159 plane-parallel flow 509
leminscate of Bernoulli 231 plunging fire 329
L’Hospital’s rule 196 point of convexity 214
Napierian logarithms 146
Lie group 527 point of inflection 91, 268
natural frequency 366
lifetime of radioactive atom 282 polar axis 28
natural frequency of circuit 444
light, electromagnetic heory of polar coordinates 28, 460
natural logarithm 146
453 pole of function 505
natural numbers, sum of powers
equilibrium intensity of 408 pole of polar system of coordi-
of 245
spontaneous emission of 413 nates 28
negative numbers 458
stimulated emission of 413 positrons 487
neutron 526 potential difference 419
light absorption 406f neutron flux 297
light emission 406f potential energy 309, 354
Newton and discovery of calcu- potential flow 509
light quanta 406 lus 124
limit 72 Newton and Leibniz, controversy polonium 283
limits of integration 97 over priority in discovery
power 303
linear approximation for func- of calculus 125f
power function 46, 138
tions 177 Newton-Leibniz theorem 101 power output 428, 435
linear density 337 pressure, atmospheric 117
Newton’s first law 316 primitive function 106
linear expansion coefficient 71 Newton’s second law 316
linear-fractional function 51, 57 principal value of Arc sin x 152’
Newton’s third law 320 principle, Fermat’s 125, 221
linear function 30 nonsmooth maxima (minima) of
linear relationship 30 probabilistic process 523
a function 224
lines of force 511 probability 519, 522, 523
nth derivative 177
liquid state of matter 234 of disintegration 281, 289'
number(s), complex 459
locally monotonic function 50 fractional (rational) 458
problem of vibrating string 375
local maximum of function 44, 87 irrational 458
progression, arithmetic 141
local minimum of function 87 negative 458 geometric 141, 188
local rate of temperature vari- proportionality factor 34
pure imaginary 459 protons 526
ation 160 real 458
logarithm(s) 145f pure imaginary numbers 459
number of disintegrations 284
base- 10, 146
binary 58
Briggs 146 odd function 56
quadratic equation 39
common 146 Ohm’s law 24, 36 quadratic function 37
modulus of base a with order of differential equation 198
quantity of electricity 117,
respect to base b 56 order of increase of function 196 418
Napierian 146 order of smallness 80
quantum mechanics 405, 519'
quaternions 525
natural 129 ordinate 26
logarithmic derivative 147 origin of coordinates 26
logarithmic function 145, 468 osculating circle 267 radian 148
loop oscillograph 445 osculating parabola 131 radiant flux 399
 f

560 Subject Index

radioactive atom(s), half-life of specific elongation 71 thermal expansion 7 1 f

282 specific heat capacity 70 thermodynamic equilibrium 407
lifetime of 282 sphere, volume of 272 time, absolute 321
rate of disintegration of 245 spherical layer (or zone) 273 time constant of a current 424
radioactive decay, initial con- surface area of 273 time lag in interaction 517
ditions for 281 spherical segment (or cap) 273 topology 521
law of 281 surface area of 273 Torricelli’s law 204
radioactive family 289 spontaneous emission of light 413 torus 275
radium 283 spring 309 surface area of 275
radius of curvature 265, 267 squaring the parabola 103 volume of 275
random event 405 stable equilibrium 313 total derivative 161
rate of change of function 177 static 'moment 341 total differential 157
rate of disintegration 283 statistical laws 523 total energy 354
rate of energy release 435 stationary state 291 total rate of temperature varia-
rate of growth of function 86 steady state 291 tion 160
rates of temperature variation 1 60 steady-state distribution 160 trajectory 59
reaction (or jet) propulsion 334 steady-state mode 206 transformation, liomothetic 42
real axis 460 Stefan-Boltzmann law 410 transformation of groups 50
real numbers 458 Stefan’s law of radiation 410 translation of curve 50
rectangular method 96 Steiner curve 61 trapezoid, curvilinear 99
rectifier 423 stimulated emission of light 413 trapezoid method 95
rectifying points of a curve 267 Stirling’s formula 249 triangle, area of 62
relationship, functional 23 Stokes law 346 trigonometric functions 148, 466
linear 30 Straight line, equation of 31 of imaginary argument 469
relative maximum 44 initial ordinate of 31 inverse 151
remainder 180 stream function 510 trigonometric series 380
residue 505 streamline 510 Tsiolkovsky’s formula 335
resistance 419 stretching of wire 115 turbulence 352
leakage 437 string 375 two-terminal network 422
resonance 306, 366, 445, 450, subcritical mass 299 two-terminal pair network 422
452, 497, 518 sum of powers of natural numbers two-valued function 50
Reynolds number 346, 350 245
Rolle’s theorem 259 superconducting state 438
root-mean-square 236 superconductivity 438
uniformly accelerated motion 69,
rule, L’Hospital’s 196 supercritical mass 299
113
superposition principle 379, 453,
safety parabola 330
uniform motion 67f
497, 513 unit impulse 335
salient point 222 surface 157
universal gas constant 388
satellite (or orbital) velocity 331 surface area of solid of revolution
unlimited figure 242
secant line 80 273
unstable equilibrium 313
second derivative of function surface area of spherical layer
uranium 283
89, 176 273
second differential 131 surface area of spherical segment
-second-order partial derivatives 273 van der Waals equation of state
156 surface area of torus 275 233
second theorem of Pappus 275 symmetry, center of 43 van der Waals isotherms 233
-semicubical parabola 45 about axis 43 variable, dummy 100
series, convergent 188 symmetry axis 37, 43 independent 23
divergent 189 symmetry group 524 of integration 98
Fourier 386, 495, 498 vector 328
infinite 181 vector calculus 525
Maclaurin’s 182 velocity 67
Taylor’s 182, 214 tabulated integral 168 velocity, average 68, 149
series expansion 264 tangent line 80, 471 escape 332
Shubnikov group 524 tangent plane 157 initial 69
signum function 491 Taylor’s series 182, 214 instantaneous 69, 149
simple pendulum 361 theorem(s), Cavalieri’s 272 satellite (or orbital) 331
Simpson’s rule 255 Guldin 274 velocity potential 509
sine, hyperbolic 168 Lagrange’s 258 vertical asymptote 278
sine-integral function 187 Newton-Leibniz 101 vertices of a curve 267
sine line 471 Pappus-Guldin 274 vibrations of a string 375
single-beam cathode-ray oscillo- Rolle’s 259 viscosity 346
graph 447 theorem of algebra, fundamental volume 118
sink 510 461 of cone 27 1
slope, infinite 33 theorem on geometric and arith- of ellipsoid of revolution 273
slope of line 32 metic means 239, 241 of pyramid U8f
slope of tangent 81 theorem of higher mathematics, of solid of revolution 272
smooth curve 215 fundamental 101 of sphere 272
smooth maxima of function 215 theorem of Pappus, first 274 of torus 275
smooth minima of function 215 second 275 volume expansion coefficient 72
solid of revolution, surface area theory of continuous groups 526 vortex 510
of 273 theory of discrete groups 526 vorticity 510
volume of 272 theory of electromagnetic field
source 510 511
source of voltage 420 theory of electromagnetism 517
wire, stretching of 115
space, absolute 321 theory of groups 523
work performed by force 114
spark gap 426 theory of information 58
special theory of relativity 521 theory of probability 523 Young’s modulus 115