Naumov Ip A Course of Mathematics Book2 Construction of Comp
Uploaded by
Strahinja DonicNaumov Ip A Course of Mathematics Book2 Construction of Comp
Uploaded by
Strahinja DonicBook II. Construction of complex numbers and functions.
Basic mathematical analysis.
ISBN 978-952-94-0574-9
ISBN 978-952-94-0575-6
ISBN 978-952-94-0576-3
ISBN 978-952-94-0577-0
Copyright © Naumov Ivan, 2018. Illustrator, designer
Gudalova Tatiana
Finland 2018
Preface
This book continues the course of mathematics. We will start from the construction of vectors and
complex numbers. And then we will continue with a course of basic mathematical analysis, which
will be used as a tool for construction of power functions and radians. Both power functions and
radians are usually defined very briefly, without any detailed explanations. But some explanations
actually must be done. Consider the simple function f ( x) 2 x . What is the value of this function
3
at the point x 3 3 ? How can we find the number 2 ? It’s not a problem to understand how to
3
raise real numbers to integer or even to rational powers, but it’s not enough. If we want to consider
the function f 2 x as a function of a real variable x , we need to understand precisely what is a
value of this function at any irrational number. Put simply, we need to define the numbers like
7
3
3 8 8
2
2 ,2 , ( 5 ) , 4
5 . The notion of radians is also extremely important, I’ve never come across
any books that provide some rigorous explanation of it. There is always just a brief definition,
where radians are defined through the notion of a length of an arc of a unit circle, but there is no
explanation what is a length of such arc. How do we actually measure a length of an arc?
Does there exist an arc of a circle with a length 5 ? Usually people like to say we can straighten
an arc and then use a ruler. At first, it’s not mathematics, and secondly, there is no ruler that can
measure lengths like 5, 2 / e, 3
3. And as a result there is no normal definition of a length of an
arc, which leads to the fact that the definition of a radian argument is based on nothing, so it is not
a definition. That’s why I decided to provide a complete explanation of this notion.
As far as I’m concerned, it’s always very important to specify the most important notions, because
if we don’t have any “foundation” (like definitions and rules to follow), it’s easy to make a wrong
reasoning and to finish with a wrong result. Really, many things that seem obvious to us in reality
appear not only not obvious, but even wrong. I’ll give the simple example. Suppose two guys
compete in arm wrestling. Most of people will agree that one guy is winning because his palm
exerts a greater force on the palm of the other guy. So his palm pushes the other palm stronger
than the other palm pushes his palm, and that’s why he is winning. It sounds like an obvious
explanation, but it is completely wrong. According to the basic laws of nature (3-rd Newton’s law),
at any given moment, the forces, which palms exert on each other, are precisely equal in
magnitude. So at any given moment, palms push each other with equal forces. And something that
seemed “obvious” turned out to be wrong. And there are a lot of examples like this one. That’s why
it’s always very important not to rely solely on our intuition, but stick to the basic laws and
definitions, which are precisely defined.
It is initially planned to provide a complete course of mathematics, which consists of 8 books.
1-st, 2-nd and 3-rd book are dedicated to explain the foundations of mathematics, and the final
6,7,8 books will include more advanced subjects, like mechanics, differential equations and
equations of mathematical physics.
All The Best, Naumov Ivan.
[email protected]
Table of contents.
Vectors ……………………………………………………………………………….…….. 5
Determinants……………………………………………………………………………… 29
Complex Numbers………………………………………………………………………… 48
Analysis…………………………………………………………………………………….. 61
Literature..………………………………………………………………………………... 158
I express my sincere gratitude to: Galochkin Alexandr Ivanovich
Drutsa Alexey Valerievich
Kozko Artem Ivanovich
Because their lectures and seminars made me interested in math
and formed my approach to this subject.
I also would like to thank Alexander Bebris, as his educational web site and the project
“English Galaxy” provided an amazing course of English grammar, which is essential for writing.
Vectors
Vectors
In this chapter we will “build” vectors by using the basic principles and axioms of stereometry
as a foundation. It’s much more common to use more “intuitive” approach to vectors, where most of
the properties of vectors are taken for granted, or “proved” by a picture, which usually reflects only a
particular case. Such approach also uses the term “vectors have the same direction”, without
defining what is a same direction of parallel vectors. And as a result, everything depends on the eye
of the observer, which is not a good thing. We will stick to another approach, that doesn’t have such
flaws, but in return we will have to “pay” the price for it, e.g., in some cases we will have to pay more
attention to details and consider several possible variants. Let’s start.
Def1. Let AB is any segment in the space. Let’s choose a “direction” of AB .
For example, let A is a “start point” and B is an “end point”. The end point
B must be marked like an arrowhead [pict1]. Now we have the object which
must be denoted as AB . The small arrow above the letters shows us that A pict.1
is a start point and B is an end point.
AB (an arrow in the space) is called a “vector”.
The length of the segment AB is called a module of AB , and we write AB AB .
If points A, B coincide, so A B , then we say that AB is a zero vector, and we write AB 0 .
Next, let’s fix any line L in the space and turn it into a coordinate line.
If we move along L in a certain direction, we always meet
greater numbers, than we met before.
Such direction is called “a positive direction” of an axis.
Any axis in the space must be depicted as a segment with a
certain direction on it, which indicates a positive direction of
an axis. Let AB lies on some axis L (it means that A, B L ).
We say that AB has a positive direction if x A xB
(the coordinate of A is not greater than the coordinate of B)
[pict2], and we say that AB has a negative direction
if xB x A . Any zero vector 0 , which lies on L, has
pict.2
a positive and a negative direction at the same time.
Two vectors AB and CD are called collinear if they both lie on the same line L.
Vectors, that do not lie on the same line, are called not collinear.
Def2. Let AB and CD are collinear vectors, if they both have the same direction (both positive,
or both negative) then we say that these vectors are co-directed. If collinear vectors AB and CD
have different directions, then we say that these vectors are anti-directed.
6
Vectors
We will say that two lines are parallel if [A] or [B]:
[A] there exist some plane which contains both lines, and these lines do not have any common
points.
[B] lines coincide: these lines are in fact
the same line.
Exercise1. Lines AB and CD lie on the same
plane , and they intersect some line L
at points A and C . Vectors AE and CF
are collinear (and lie on L ) [pict3.1].
If vectors AE and
CF are co-directed, then
[A] BAE DCF AB || CD
AE and CF are anti-directed, then
If vectors
[B] BAE DCF 180 o AB || CD
Put simply. Lines are on one plane are parallel if
pict.3.1
and only if they form equal angles with co-directed
collinear vectors. And lines on the plane are parallel if and only if they form angles, which sum is
180 o , with anti-directed collinear vectors.
Def3. Two not collinear vectors AB and СD are equal if AB CD and AB || CD and AC || BD
(let’s call these requirements [S])[pict3]. Note: from [S] immediately follows that ABCD is a
parallelogram and also AC BD . Two collinear vectors AB and СD are equal if AB CD and
these vectors are co-directed.
Note. When we write BD || AC we mean that the line, which passes
throughB and D , is parallel to the line which passes through
A and С . We can also say “vectors are parallel” if vectors lie
on parallel lines, and we can write AB || CD .
From the definition of vector equality [S]: two not collinear vectors
pict.3
are equal if their modules are equal, they are parallel, and lines,
which connect their start and end points, are parallel.
Assertion1. Two not collinear vectors AB and СD are equal
The vector AC which connects their start points is equal
to the vector BD which connects their end points.
Proof. Let AB СD , by definition it means that AB CD
and AB || CD and AC || BD , from here follows that pict.4
ABCD is a parallelogram and AC BD .
We want to prove that AС BD , so we need to show [S],
7
Vectors
which is obviously true, really AC BD, AC || BD, AB || CD . Therefore AС BD. [pict4].
Let now AС BD , from here, in the exactly similar way as above, we can get AB СD.
Auxiliary1. Сollinear vectors AB and СD are equal xB x A xD xC .
Let collinear vectors are equal AB СD . [A] If both these vectors have a positive direction,
then x A xB and xC xD . The length of the segment AB is equal to xB x A and the length
of the segment CD is equal to xD xC . Vectors are equal, then AB СD xB xA xD xC .
[B] If both vectors have negative direction, then x A xB and xC xD . The length of AB is x A xB ,
and the length of CD is xC xD . As vectors are equal, then AB CD
x A xB xC xD xB x A xD xC .
Conversely. xB x A xD xC . [A] If x A xB , then there must be xC xD . Then both vectors
AB and СD have a positive direction (so these vectors are co-directed), and xB x A is a length
of AB and xD xC is a length of CD , then AB CD . Therefore AB СD .
[B] If x A xB then there must be xС xD . Then both AB and СD have a negative direction.
And x A xB is a length of AB and xC xD is a length CD . From xB x A xD xC follows that
xA xB xC xD AB CD . Then AB СD .
We can use auxiliary1 to prove the assertion2 which is very similar to the assertion1.
Assertion2. Two collinear vectors are equal vector that connects their start points is equal to
the vector that connects their end points
Proof. Let AB СD are equal collinear vectors, then xB x A xD xC [A].
We have to show that vectors AC and BD are equal, by auxiliary1 they are equal if and only if
xС x A xD xB - this equality immediately follows from [A].
vectors, which connects start and end points of AB and СD , are equal, it means that AС BD .
Then xС x A xD xB - from this equality follows the equality [A] and therefore AB СD .
We can generalize, from the assertions 1,2 follows the
Theorem1. Two vectors are equal vector that connects their start points is equal to the vector
that connects their end points
Theorem2. Vector equality "" is symmetrical: AB СD СD AB .
Proof: For not collinear equal vectors: AB СD , then AB CD and AB || CD and AC || BD and
also AC BD . If we want to show that СD AB (by [S]) there must be:
8
Vectors
СD AB, СD || AB, СA || DB - and we already have it. Let now AB and СD are collinear and
equal, then there is nothing to prove, symmetricity follows right from the definition, we can also
write AB СD xB x A xD xC xD xC xB x A СD AB . We showed that
AB СD СD AB . The proof of the converse assertion is exactly similar.
Theorem3. Vector equality "" is transitive: AB СD and СD EF , then AB EF .
Comment. Despite the fact that this theorem looks very simple, it appears the most “demanding”
theorem in the vector construction. It happens because we have precisely defined what are equal
vectors, that’s why we have very strict and clear requirements to vectors in order them to be equal.
Remember that if we don’t stick to the initial definitions, all the process loses it’s value.
Simple exercise. For any vectors AB СD BA DC .
Comment. If one of vectors AB, СD, EF is a zero vector 0 , then the proof is obvious
(each vector must be a zero vector in such case). From now on, there are no zero vectors among
AB, СD, EF . Also, if some two of vectors AB, СD, EF coincide, i.e., some two of these vectors are in
fact the same vector, then there is nothing to prove. From now on, there are no coinciding vectors
among AB, СD, EF .
Proof. Let there are no collinear vectors among
AB, СD, EF [pict5]. We have to show that AB EF ,
by definition, it means that we need to prove that:
AB EF and AB || EF and AE || BF .
From AB СD follows that AB СD, AB || CD .
From СD EF follows that CD EF , CD || EF .
pict.5
Therefore AB EF and AB || EF .
The last thing that we need to show: AE || BF . As AB || EF , both these segments lie on the same
plane , this plane also contains the segments AE, BF [T]. We have to show that these two
segments are not diagonals of the parallelogram ABEF (we can’t use [pict5] as an argument,
there must be a normal proof), we will show that the segments [T] lie in two parallel planes 1 and
2 and therefore they don’t have any common points (then [T] must be two opposite sides of the
parallelogram ABEF , then [T] are parallel and everything is proved). From AB СD follows that
ВD || AC and fromСD EF follows that DF || CE . Let’s draw the plane 1 through the lines
ВD and DF and the plane 2 through the lines AC and CE . These planes must be parallel
1 || 2 (because ВD || AC and DF || CE ). The plane 1 contains В, D, F , then ВF 1 ,
the plane 2 contains A, C , E , then AE 2 . Then [T] lie in two parallel planes.
If all three vectors AB, СD, EF are collinear, then the proof is obvious:
9
Vectors
AB СD xB x A xD xC || СD EF xD xC xF xE .
Then there must be: xB x A xF xE AB EF .
Let now some two of three vectors AB, СD, EF are
collinear. There are exactly 3 different variants
(if 2 of these 3 vectors are collinear).
Let AB and СD are collinear [pict6] they lie
on the same line L . So we have AB СD and also
СD EF . We want to show that AB EF .
As СD EF , then СD || EF L || EF and we can
draw the plane through the line L and the vector pict.6
EF , then contains all three vectors AB, СD, EF .
In the exactly similar way as in the previous reasoning
we deduce: AB EF and AB || EF . The last thing that we need to show: AE || BF (then
AB EF by definition). From СD EF follows that СE || DF and СE DF . From AB СD
follows that vector which connects their start points is equal to the vector which connects their end
points: AС BD (collinear vectors) then also (simple exercise from above) СA DB .
Let’s sum up: СE DF and СE || DF and СA DB . Lines CE and DF are parallel,
then (exercise1) they form equal angles with collinear co-directed vectors СA and DB , so
ACE BDF . Then we have: CE DF and AC BD (because СA DB ) and
ACE BDF , therefore ACE BDF . From here follows EAC FBD (as these angles
lie in front of equal sides in equal triangles). Also AС and BD have the same direction (because
they are equal), then (exercise1) there must be EA || FB and everything is proved.
The case when СD and EF are collinear is exactly similar to the case that is considered above.
The last case that we need to consider: AB and EF are collinear
(they lie on the same line) and СD doesn’t lie on that line
[pict7]. From AB СD follows that AC || BD, AC BD .
From СD EF follows that CE || DF, CE DF .
Let’s show that ACE BDF :
we have AC BD, CE DF [G] (we also need equal angles).
pict.7
10
Vectors
Comment: we can’t just say that ACE and BDF are equal because the sides of one angle are
parallel to the sides of the other angle. In such case angles mustn’t be equal. If sides of one angle are
parallel to the sides of the other angle, these angles either equal, or their sum is 180o (and they
mustn’t be equal in such case). So we need to provide a normal reasoning, based on our assumptions,
which we made at the very beginning of this chapter. Let’s consider two possible cases [A] and [B].
[A] Let AE and BF are co-directed. As AC || BD , then (exercise1) these lines form equal angles
with collinear co-directed vectors AE and BF , and we have: CAE DBF .
As AE and BF are co-directed, vectors EA and FB are also co-directed, and we have parallel
lines CE || DF , then (exercise1) CEA DFB .
ACE and BDF we have [G] and we have just deduced that the angles near
Let’s sum up: for
the base AE (in ACE ) are equal to the angles near the base BF (in BDF ), as the sum
o
of angles of any triangle is 180 , there must be С D ACE BDF ),
(or the same
therefore ACE BDF . As these triangles are equal, their bases are also equal: AE BF ,
we know that AE and BF are co-directed, therefore there must be AE BF (by definition),
then (theorem1) vectors which connect their start and end points are also equal, i.e., AB EF
it is exactly what we need.
[B] Let AE and BF are anti-directed. We have AC || BD then (exercise1) there must be:
CAE DBF 180 o CAE , DBF 180 o .
As vectorsAE and BF are anti-directed, vectors EA and FB must be also anti-directed,
and we have CE || DF , then (exercise1) CEA DFB 180 CEA , DFB 180 .
o o
Then in ACE and BDF the sum of all 4 angles which lie near the bases AE, BF is 360 o ,
it is impossible because the sum of all 6 angles of these triangles is 360 o . In this case there
must be ACE 0o , BDF 0o A E, B F , then AB and EF coincide, so they are
co-directed, which contradicts to [B]. Then the case [B] is impossible. The theorem is proved.
Theorem4. For any point A in the space and any vector СD there exist only one vector AB
(which start point is A ) that is equal to СD .
Proof. Existence. If СD is a zero vector, then С D ,
then we take the vector AA which is also a zero vector.
Let СD is not a zero vector. If A and СD do not lie
on the same line L, then we can draw the plane
through A, C , D [pict8]. And (in ) we can build pict.8
11
Vectors
a parallelogram ABCD : AB || CD, AC || BD . Then AB is (obviously) the vector we need. If A
and СD belong to some line L, then we turn L into the axis. Let’s choose the positive direction on
L in such way that СD has a positive direction. Then we can take the point B L such that
AB CD and AB has a positive direction. Then AB СD .
Uniqueness. Let’s assume that AB СD and AB1 СD . By transitivity (theorem3) we have
AB1 AB [V]. From AB1 AB follows (theorem1) that AA1 BB1 0 BB1 B B1 and
vectors AB1 , AB coincide.
Coordinates. Let’s fix some plane and an axis Ox on this plane. We will consider only vectors
AB which start point A belongs to Ox .
For any vector AB , the difference xB x A
(where xB is a coordinate of B and x A is a coordinate of A )
is called an x -coordinate of AB [pict9].
So, for any vector, the difference of it’s end point coordinate
and it’s start point coordinate is an x -coordinate of a vector. pict.9
Property1. Equal vectors have equal x -coordinates: if AB СD , then xB x A xD xС .
Proof. From AB СD follows that ABDC is a parallelogram, where BD || AC BD || Ox
(do not forget that we consider only vectors which start point belongs to
Ox ). Let h is a length
of the side BD . Then xD xB h and xC x A h [pict10], or vice versa xB xD h and
x A xC h . In any case we can subtract our equalities and get: xB x A xD xС .
pict.10
12
Vectors
Def4. The angle of inclination of any vector AB is the angle [0o ,360 o ) between AB and Ox ,
which is counted from the positive direction of Ox in the counterclockwise direction [pict11].
Let’s pay more attention to the def4, because it has
a great importance in math.
We need to have some standard according to which
we can always understand how to determine the angle
between a vector and an axis.
And our “standard” is an ordinary clock dial [pict11].
Let we have some vector AB and some axis Ox , where
A Ox . The dial-center must be placed at A such that pict.11
the direction from 9 to 3 is a positive direction of Ox ,
then the direction of rotation from 3 to 9 is
our direction, and we must calculate the angle
[0o ,360 o ) (between AB and Ox ), by moving from
Ox in this direction.
Property2. Equal vectors have equal angles of inclination.
Proof. Let AB CD , if A C then vectors
AB, CD coincide and there is nothing
to prove. Let A C , then ABDC is
a parallelogram and BD || AC BD || Ox .
Without loss of generality, let x A xC .
Let’s just consider separately all the
possible cases for the angle of inclination
of AB . So: (0o ,90o ) ,
(90o ,180 o ) (180 o ,270 o ) , pict.12
(270 ,360 ) [pict12]. In each case
o o
parallel lines AB and CD form equal acute angles with Ox . By using these equal acute angles in
each case, we can easily check that the angles of inclination of AB and CD are really equal.
And finally, the cases 0 , 90 , 180 , 270 are obvious.
o o o o
Def. Any vector OB , which start point is located at the origin O , is called a radius vector
(the origin O is a point with a zero coordinate on Ox ).
13
Vectors
Property3. For any radius vector OB, the next is true: xB OB cos [F]
(where is an angle of inclination of OB ).
Proof. If 0o , 90o , 180 o , 270 o ,
then the equality [F] is obviously true.
Let we have any other case.
We fix any OB and draw the unit circle
with the center at O .
OB may not intersect (if it’s length
is less than 1 ), in any case we draw
the ray OB and mark the point С
at which it intersects [pict13].
Let’s draw perpendiculars BBx and pict.13
СCx . In any case the right triangles
OB OBx
OBBx and OССx are similar, then [as OC 1] OBx OB OCx - from this
OC OCx
equality follows the result we need (formula) [F]. We just have to consider all the possible cases for
the inclination angle .
[1] (0o ,90o ) , then xB OBx and cos OCx and OBx OB OCx xB OB cos .
[2] (90o ,180 o ) , then xB OBx and cos OCx and
OBx OB OCx xB OB ( cos ) xB OB cos .
[3] (180 o ,270 o ) , then xB OBx and cos OCx and
OBx OB OCx xB OB ( cos ) xB OB cos .
[4] (270 o ,360 o ) , then xB OBx and cos OCx and OBx OB OCx xB OB cos .
Vector coordinates in the space
Let AB is any vector in the space.
Let’s draw the plane Ox
B
through the point B , which
is perpendicular toOx , it intersects the axis Ox at
the point Bx with coordinate bx [pict14]. Let’s draw
the plane Ox
A
through the point A , which is
perpendicular toOx , it intersects Ox at the point
Ax with coordinate a x . By definition, x -coordinate
of AB is the difference bx ax . In the exactly similar pict.14
14
Vectors
way we can define: y -coordinate of AB is the difference by a y . And z -coordinate of AB is bz az
And finally, coordinates of AB is the set of numbers (bx ax , by a y , bz az ) .
Theorem5. Equal vectors have equal coordinates.
Proof. [pict15] Let’s show that
pict.15
equal vectors have equal
x -coordinates.
Let’s fix some vector AB .
Let draw the planes Ox
B
and Ox . The picture is done
A
for the most general case, when
Ox
B
and Ox do not coincide,
A
so these are different planes.
We can build the vector Ax B1
such that Ax B1 AB (theorem4). Let’s show that B1 must belong to the plane Ox
B
.
[A] Let Ox
B
and Ox coincide. As Ax B1 and AB are equal, they must be parallel Ax B1 || AB .
A
Obviously A, B, Ax lie on the same plane Ox Ox . If
B A
B1 doesn’t belong to Ox
B
Ox
A
, then
Ax B1 , AB are skew lines and they are not parallel, which is not true. Then B1 belongs to Ox
B
.
[B] Let Ox
B
and Ox do not coincide. Then these are parallel planes (because they are both
A
perpendicular to Ox ). As Ax B1 || AB , we can draw the plane through the lines Ax B1 and AB .
The plane intersects Ox
B
and Ox and there are some lines of intersection Ox L1
A B
and Ox
A
L2 . There must be L1 || L2 (it is one of the axioms of stereometry).
As L2 passes through A, Ax , then L2 AAx . And L1 passes through B .
As AB, Ax B1 are equal, lines AAx and BB1 must be parallel.
Let’s sum up: L2 AAx is parallel to BB1 and L2 AAx is parallel to L1 Ox
B
which passes
through B . Then (by transitivity) BB1 must be parallel to L1 , both these lines pass through
the point B , then they must coincide, therefore B1 Ox
B
.
Next, AB, Ax B1 obviously have equal x -coordinates (because their start points lie on Ox
A
and
their end points lie on Ox
B
).
15
Vectors
Let’s take any other vector СD which is equal to AB . In the same way we build Сx D1 СD .
And С x D1 has the same x -coordinate as СD .
By transitivity of the vector equality (theorem3) we have Ax B1 Сx D1 , then Ax B1 and Сx D1 must
be parallel. Then we can draw the plane 1 through Ax B1 and Сx D1 , obviously 1 contains
the axis Ox . Then we have the equal
vectors Ax B1 , Сx D1 and the axis Ox
on the plane 1 , the start points
of these vectors lie on Ox [pict15.1] .
Obviously B1Bx Ox (because B1Bx
lies on the plane which is
perpendicular to
Ox ) and also
D1Dx Ox . Then x -coordinate of pict.15.1
Ax B1 when we consider it as a space vector is the same as x -coordinate of Ax B1 when we consider
it as a plane vector on 1 . Obviously Ax B1 and С x D1 are equal vectors on the plane 1 , then
their x -coordinates on 1 are equal: bx ax d x cx (property1), in the same time bx ax and
d x cx are x -coordinates of the space vectors Ax B1 and С x D1 . Therefore x -coordinates of the
space vectors AB and СD are equal. Similarly we can show that y -coordinates and z -coordinates
of AB and СD are equal. Then equal vectors have equal coordinates. Everything is proved.
By definition, when we want to find coordinates of some point A (in the space) we need to draw
the planes x , y , z through point A which are perpendicular to Ox, Oy, Oz . These planes
intersect axises at some points Ax , Ay , Az with coordinates ax , a y , az . The numbers ax , a y , az are
coordinates of A . Let’s notice that Ox x and therefore Ox is perpendicular to any line which
lies on x , in particular Ox AAx . Similarly Oy AAy and Oz AAz , then AAx , AAy , AAz are
perpendiculars from A on Ox, Oy, Oz , and coordinates of any point A in the space can be found
by drawing perpendiculars from A to coordinate axises.
Exercise. For any vector a with coordinates (ax , a y , az ) , the length of a is
a (a x ) 2 (a y ) 2 (a z ) 2 .
Solution: a connects it’s start and end points, these points have some coordinates
( xstart , ystart , zstart ) and ( xend , yend , zend ) . The distance between these points is
( xend xstart ) 2 ( yend ystart ) 2 ( zend zstart ) 2 and here we see that
16
Vectors
xend xstart , yend ystart , zend zstart are coordinates of a (by definition),
so these differences are ax , a y , az .
Def: a radius vector is any vector OA , which start point O is located at the origin.
The start point O of OA has coordinates (0,0,0) , then: coordinates of any radius vector are exactly
coordinates of it’s end point A.
Theorem6. If two vectors have equal coordinates, then these vectors are equal.
Proof. Let AB and CD are vectors with equal coordinates
(bx ax , by a y , bz az ) (d x cx , d y c y , d z cz ) (a, b, c) . Let’s build the radius vector
OB1 AB . We have already shown that equal vectors have equal coordinates, therefore OB1 has
coordinates (a, b, c) . Then B1 has coordinates (a, b, c) .
Let’s build now the radius vector OD1 CD . Equal vectors have equal coordinates, therefore OD1
has coordinates (a, b, c) . Then D1 has coordinates (a, b, c) . We see that B1 and D1 have the same
coordinates, therefore they must coincide: B1 D1 . Then OB1 OD1 and, by transitivity of
the vector equality (theorem3), we get AB CD .
Let’s sum up: from the theorems 5,6 follows
Theorem7. Two vectors are equal coordinates of these vectors are equal.
Another representation of vector coordinates.
Let AB is any vector in the space, it has coordinates
(bx ax , by a y , bz az ) . Let’s build the radius vector
OB1 AB [pict16]. OB1 has the same coordinates:
(bx ax , by a y , bz az ) . Then B1 has coordinates
(bx ax , by a y , bz az ) . Then, if we draw
the perpendicular B1Bx to Ox , the point Bx has
x -coordinate xB1 bx ax . Now we can consider OB1 as pict.16
a plane vector on the plane x which contains OB1 and Ox .
The start point O of OB1 is located at the origin. According to the property3 (from above), we have:
xB1 OB1 cos bx ax OB1 cos where [0o ,360 o ) is the angle between Ox and OB1 ,
which is counted in the counterclockwise direction from Ox .
17
Vectors
Then bx ax OB1 cos bx ax AB cos and similarly by a y AB cos and
bz az AB cos . Therefore, coordinates of AB are ( AB cos , AB cos , AB cos ) .
Def. We will say that AB forms the angle with the axis Ox if the radius vector OB1 AB
forms the angle with the axis Ox . The angle is called an inclination angle of AB with
respect to Ox . And similarly, we define the angles of inclination , with respect to Ox, Oy, Oz .
By definition, cos , cos , cos - “direction cosines” of AB .
Then: if we multiply a module of a vector by it’s directional cosines, we will get exactly coordinates
of a vector. Such representation of vector coordinates is very important for problems of mechanics.
Addition of vectors
Def. The sum of vectors AB and CD is the vector
AE AB CD , which we get from the next process: we build
BE CD [pict17], then we build AE , then AE is a sum of
AB and CD .
Property4 From this definition immediately follows that:
for any points A, B, C in the space we have: AB BC AC
pict.17
Assertion3. When we add vectors, their coordinates must be added. For any vectors AB and CD :
x -coordinate/ y -coordinate/ z -coordinate of AB CD is a sum of x -coordinates/ y -coordinates/
z -coordinates of vectors AB and CD .
Proof. AB has coordinates (bx ax , by a y , bz az ) and CD has coordinates (1 , 2 , 3 ) .
Let’s find the sum AB CD , we need to build BE CD [pict17]. As BE connects B and E,
it has coordinates (ex bx , ey by , ez bz ) . On the other hand BE has the same coordinates
as CD (because these are equal vectors), then (ex bx , ey by , ez bz ) (1 , 2 , 3 ) [J].
By definition, the sum of our vectors is the vector AE which connects A and E , so it has
coordinates (ex ax , ey a y , ez az ) . We obviously have:
(ex ax , ey a y , ez az ) ((ex bx ) (bx ax ), (ey by ) (by a y ), (ez bz ) (bz az )) ,
then, by [J], we can rewrite:
(ex ax , ey a y , ez az ) ((ex bx ) 1 , (ey by ) 2 , (ez bz ) 3 ) . Everything is proved.
18
Vectors
Sometimes when we speak about vectors we do not have any necessity to specify their start and end
points, especially when we speak about general properties of vectors, or consider some collections
of vectors, in such case we denote vectors just like a, b , с ..., it’s very convenient, because such
notation takes less place.
Assertion3 can be rewritten as: for any vectors a and b with coordinates (ax , a y , az ) and
(bx , by , bz ) , their sum a b has coordinates (ax bx , a y by , az bz ) .
Theorem8. Addition of vectors is commutative: a b b a (for any vectors a , b )
and associative (a b ) c a (b c ) (for any vectors a , b , c ).
Let’s prove commutativity. Let (ax , a y , az ) and (bx , by , bz ) are coordinates of a and b .
Then a b has coordinates (ax bx , a y by , az bz ) , and b a has coordinates,
(bx ax , by a y , bz az ) . We see that vectors a b and b a have equal coordinates.
then, according to the theorem7, a b and b a are equal.
Next, associativity. [1] a b has coordinates (ax bx , a y by , az bz ) , then (a b ) c has
coordinates ((ax bx ) cx , (a y by ) c y , (az bz ) cz ) .
[2] a has coordinates (ax , a y , az ) , and (b c ) has coordinates (bx cx , by c y , bz cz ) .
Then a (b c ) has coordinates (ax (bx cx ), a y (by c y ), az (bz cz )) .
We see that (a b ) c and a (b c ) have equal coordinates, then (theorem6) these vectors are
equal. pict.18
Consequence1. Sums of equal vectors are equal vectors:
a1 b1 , a2 b2 .... an bn , then a1 a2 ... an b1 b2 ... bn
(really, as vectors a1 a2 ... an and b1 b2 ... bn have equal
coordinates, then these vectors are equal).
Parallelogram rule. For any two not collinear vectors AB , СD
the sum of these vectors can be found in the next way [pict18]:
we build AH СD and we draw the parallelogram, which adjacent
sides are AB and AH , then the diagonal vector AE is a sum
of vectors AB and СD .
19
Vectors
Chain rule [pict19]. If we want to find the sum
of several vectors a1 , a2 ,... an [CH],
we can construct the “chain” from these vectors:
the end point of a1 must be taken as a start point
of b2 a2 , the end point of b2 must be taken as
a start point of b3 a3 …, the end point of
bn 1 an 1 must be taken as a start point of
pict.19
bn an . The vector , which connects the start
point of a1 with the end point of bn , is a sum of
all vectors [CH]: a1 a2 ... an .
pict.20
Subtraction of vectors . The difference of vectors AB and
СD is the vector HB AB CD [pict20], which we get from
the next process: we build AH СD , then we build HB ,
then HB is a difference of AB and СD .
Property5. HB AB СD HB СD AB .
Proof. We have HB AB СD , let’s consider AHB
(property3) for any points A, H , B in the space we have:
AH HB AB / commutativ ity / HB AH AB
[consequence 1 : HB HB, AH CD] HB СD AB .
Theorem9. When we subtract vectors, their coordinates must be subtracted, i.e., for any vectors
a , b with coordinates (ax , a y , az ) and (bx , by , bz ) , their difference a b has coordinates
(ax bx , a y by , az bz ) .
Proof. Let a b c and c has some coordinates (cx , c y , cz ) . From the property5 follows that
a b c a c b . We have proved that when we add vectors, their coordinates must be added
(assertion3): (ax , a y , az ) (cx bx , c y by , cz bz ) ax cx bx , a y c y by , az cz bz
cx ax bx , c y a y by , cz az bz , then (cx , c y , cz ) (ax bx , a y by , az bz ) .
Everything is proved.
20
Vectors
Theorem10. c a b c b a .
Proof. The part c a b c b a is proved (property5).
Conversely. Let c b a , we want to show that c a b . From c b a follows that
coordinates of a are sums of coordinates of c and b , so
(ax , a y , az ) (cx bx , c y by , cz bz ) ax cx bx , a y c y by , az cz bz
cx ax bx , c y a y by , cz az bz [M]. According to the theorem9, a b has coordinates
(ax bx , a y by , az bz ) , from [M] we see that c has the same coordinates as a b ,
Then (theorem7) vectors c and a b are equal: c a b .
Multiplication by numbers [pict21].
AB is any vector. We define
0 AB / by definition/ AA 0
For any positive number 0 we define:
AB is the vector which start point is A,
which is co-directed with AB , and which’s pict.21
length is AB .
For any negative number 0 we define: AB is the vector which start point is A, which is
anti-directed with AB , and which’s length is AB .
Def. a and b are non-zero vectors, a is called proportional to b if there exist some constant
R a b
such that
Exercise. Show that if a is proportional to b , then b is proportional to a .
From here follows: if one of non-zero vectors a , b is proportional to the other,
then we can just say “these vectors are proportional”.
Exercise. Non-zero vectors a , b are proportional these vectors lie on parallel lines
(in particular, on the same line).
Theorem11. If some vector is multiplied by , coordinates of that vector must be also
multiplied by .
Proof. Let’s take any vector AB , it has coordinates (Another representation of vector coordinates)
( AB cos , AB cos , AB cos ) . Let is positive, then AB forms exactly the same angles
21
Vectors
, , with coordinate axises, and it’s length (by definition) is AB , then AB has
coordinates (( AB) cos , ( AB) cos , ( AB) cos ) .
We see that coordinates of AB are exactly coordinates of AB , which are multiplied by .
When 0 the theorem11 is obviously true.
Let now 0 . So AB forms some angles , , with Ox, Oy, Oz , then AB || 0
(which direction is opposite to AB ) forms the angles (180 o ), (180 o ), (180 o ) with
Ox, Oy, Oz , and it’s length is AB , then it has coordinates
( AB cos(180 o ), AB cos(180 o ), AB cos(180 o ))
( AB cos , AB cos , AB cos ) [as ]
( AB cos , AB cos , AB cos ) and we see again that coordinates of AB
are exactly coordinates of AB , which are multiplied by .
Def. For any vector a we define a (1) a .
Exercise. For any vectors a , b we have: [A] a (a ) a a 0 and [B] a b a (b )
[C] a (b ) a b . [D] ( ) a ( a ) (for any , R ), [E] a a ( ) a
(for any , R ), [F] a b (a b ) (for any R ).
Hint. In each case it’s very easy to show that vectors on the left and on the right part of equality
have equal coordinates. And if vectors have equal coordinates, then these vectors are equal (theorem7).
Def. Let a1 , a2 ... an are some vectors and 1 , 2 ... n are some real numbers.
The vector 1a1 2 a2 ... n an is called a linear combination of a1 , a2 ... an with coefficients
1, 2 ... n , or just a linear combination of a1 , a2 ... an .
Property5. Obviously, x -coordinate of 1a1 2a2 ... n an is a linear combination of
x -coordinates of vectors a1 , a2 ... an with coefficients a1 , a2 ... an .
And y -coordinate of 1a1 2 a2 ... n an is a linear combination of y -coordinates of vectors
a1 , a2 ... an with coefficients 1 , 2 ... n . And the same is true for z -coordinate .
22
Vectors
Basis
Def. Let’s take the “unit vectors” ex , ey , ez with
coordinates (1,0,0), (0,1,0), (0,0,1) [pict22].
The start point of each vector is located at the origin O,
these vectors are called the standard basis vectors.
Any vector a with coordinates (ax , a y , az ) can be represented
as a linear combination of these vectors:
a ax ex a y ey az ez (property5) and this representation
is unique. Really, let a ex e y ez (where , , are pict.22
unknown variables). As vectors a, ex e y ez are equal, they must have equal coordinates.
a has coordinates (ax , a y , az ) and ex ey ez has coordinates
(1,0,0) (0,1,0) (0,0,1) ( ,0,0) (0, ,0) (0,0, ) ( , , ) , and there must be
(ax , a y , az ) ( , , ) , then ax , a y , az and the representation is unique.
Projection on an axis. Let’s fix an arbitrary axis L in the space.
(Notice: when we fix some axis in the space, the scale on it must be the same as a scale on
Ox, Oy, Oz ).
Let AB is any vector. Let’s draw two planes BL and LA through the points B and A, which are
perpendicular to L, they intersect the axis at some points with coordinates x BL and x AL ,
the difference xBL x AL is called a projection of AB on the axis L. It’s easy to notice now that
x -coordinate of any vector a is just a projection of a on the axis Ox , and the same is true for
y, z -coordinates of any vector. In fact, any axis in the space can be chosen from the very beginning
as a coordinate axis Ox (for example). And we can get the same results, as we got earlier for Ox ,
for any axis L in the space. Let OL is a point with a zero coordinate on L, let’s build OL B1 AB .
By definition, the angle between AB and L is the angle [0o ,360o ) between OL B1 and L ,
which is counted from L in the counterclockwise direction.
Then AB cos xB x A (the exactly similar result, page 17).
L L
Theorem12. For any vectors a1 , a2 ... an and any axis L , the projection of a sum-vector
a1 a2 ... an on L is equal to the sum of projections of a1 , a2 ... an on L .
Comment: this theorem is exactly similar to the assertion: x -coordinate of a sum of vectors is
a sum of x -coordinates of these vectors.
23
Vectors
Dot product. For any vectors a and b with coordinates (ax , a y , az ) and (bx , by , bz ) , the number
axbx a yby az bz is called “a dot product of vectors a and b ”, and we denote it
a b axbx a yby az bz .
Def: Let a and b are some non-zero vectors in the space. Let’s draw the radius vectors b1 b
and a1 a [pict23]. By definition, the angle between a and b is the angle [0 ,180 ]
o o
between a1 and b1 .
Assertion4. The dot product of any non-zero vectors a , b
can be calculated as: a b a b cos , where is an angle
between a and b .
Proof. Let a and b are vectors with coordinates (ax , a y , az )
and (bx , by , bz ) , we build b1 b and a1 a [pict23].
Let’s denote b1 OB and a1 OA . Then OA has
coordinates (ax , a y , az ) and OB has coordinates (bx , by , bz ) .
As the origin O has coordinates (0,0,0) , then A has
pict.23
coordinates (ax , a y , az ) and B has coordinates (bx , by , bz ) .
Let’s consider OAB , here OA a and OB b and AOB .
Let’s simplify the expression a b cos by using the cosine law for OAB :
a 2 b 2 AB 2
a b 2ab cos AB cos
2 2 2
, then
2ab
a 2 b 2 AB 2 a 2 b 2 AB 2
ab cos ab
2ab 2
a 2
x a 2y az2 b
2
2
x by2 bz2 (b a )
2
x x
2
(by a y ) 2 (bz az ) 2
2
2
2 2 2 2
ax a y az bx by bz (bx ax )2 (by a y )2 (bz az )2
2 2
2
axbx a yby azbz / by definition/ a b . Everything is proved.
The assertion4 allows us to find an angle between any non-zero vectors a and b , if we know
coordinates of these vectors.
For any (non-zero) a and b , we can fix an arbitrary point D in the space and build DA a and
DB b , the angle [0o ,180 o ] between DA and DB is obviously equal to the angle between
24
Vectors
a and b (because the triangle, which is built on the radius vectors a1 OA and b1 OB , is equal
to the triangle which is built on DA and DB ).
Exercise1. Find the angle between vectors a and b with coordinates (1,2,3) and (4,5,6) .
Solution. Let’s write the dot product of these vectors in two different ways.
32
1 4 2 5 3 6 12 22 32 42 52 62 сos сos
14 77
32 32
сos .
7 2 11 7 22
Now we have the cosine value of the angle [0o ,180 o ] . Any angle from [0o ,180 o ] is uniquely
32
defined by it’s cosine. So, our answer is: o o
is such angle from [0 ,180 ] that сos .
7 22
Comment: we haven’t defined the inverse trigonometric functions yet, but anyway, the answer
“ is arccosine of something” is not better than our answer.
Exercise2. Find the area of the triangle ABC which adjacent sides are the vectors AB and AС
with coordinates (1,2,3) and (4,5,6) .
Solution. We have already found the angle between these vectors. In order to find the area we can
1
use the formula S ab sin . So AB 12 22 32 14 and AC 42 52 62 77 ,
2
32
the angle between these sides is such that сos and [0 ,180 ] , then
o o
7 22
49 22 32 32
2
32 54
sin 1 cos 1
2
.
7 22 49 22 7 22
1 54 1 3
Then the area equals S 14 77 54 6.
2 7 22 2 2
Properties of a dot product.
[1] Non-zero vectors a and b are perpendicular if and only if their dot product is zero.
Proof. As we know a b a b cos . If vectors are perpendicular a b , then 90 , then
o
a b a b cos 90o a b 0 0 . Conversely, if the dot product is zero a b 0 ,
then a b cos 0 . Vectors a , b are non-zero vectors, it means that a 0, b 0 , then from
the equality a b cos 0 follows that cos 0 90 o .
By using [1] we can understand very quickly, are given vectors perpendicular or not. For example,
vectors with coordinates (1,2,3) and (3,2,1) are not perpendicular, because
1 (3) 2 (2) 3 (1) 10 . And vectors with coordinates (1,2,1) and (1,1,1) are
perpendicular, because 1 (1) 2 1 1 (1) 0 .
25
Vectors
[2] The dot product has a linear property:
[A] For any number and any vectors a , b we have: (a ) b a (b ) (a b )
[B] For any vectors a , b , c we have: (a b ) c a c b c and c (a b ) c a c b .
Proof. Both proofs are very simple and immediately follow from the definition of a dot product:
a b axbx a yby az bz . Let’s denote coordinates of our vectors:
a (ax , a y , az ), b (bx , by , bz ), c (cx , c y , cz ) then, for example, a b has coordinates
(ax bx , a y by , az bz ) and the dot product
(a b ) c (ax bx ) cx (a y by ) c y (az bz ) cz
ax cx a y c y az cz bx cx by c y bz cz a c b c and etc.
Change of approach. In mathematics equal vectors are usually understanded as a same vector.
So usually people point on equal vectors (which do not actually coincide) and say that these vectors
are a same vector. It is only a mathematical approach, and it obviously doesn’t work in mechanics.
Really, suppose that several forces exert on some object. Then each force is unique and the point of
application of each force is very important. If some two forces are equal as vectors, but they have
different points of application, these are different forces anyway, it’s not right to understand these
forces as a same force.
But in mathematics equal vectors are usually perceived as a same vector.
Most important trigonometric formulas
Remark1. For any angle and for any integer number k : cos(3600 k ) cos and
sin(3600 k ) sin .
Remark2. We have shown (Book I, page170) that for any angle [1] sin( ) sin( ) and
[2] cos( ) cos . Therefore, for any angles , we can take and then
sin( ) sin( ) and cos( ) cos( ) .
Assertion5. For any angles and : cos( ) cos cos sin sin [A]
Proof. Each angle and can be represented as: 360 o k 1 , here k is an integer number
and 1 [0o ,360o ) and 360o m 1 where m is an integer number and 1 [0o ,360o ) .
Let’s simplify both sides of [P] separately:
cos( ) cos(360o k 1 (360o m 1 )) cos(360o (k m) (1 1 ))
// remark 1 // cos(1 1 ) .
cos cos sin sin cos(360o k 1 ) cos(360o m 1 ) sin(360o k 1 ) sin(360o m 1 )
// remark 1// cos1 cos 1 sin 1 sin 1 . Therefore it is enough to prove that
cos(1 1 ) cos1 cos 1 sin 1 sin 1 for 1 [0o ,360o ) and 1 [0o ,360o ) .
26
Vectors
We can discard indexes here, so we have to prove
the equality [A] only for , [0o ,360o ) , let’s fix any
, [0o ,360o ) and let’s build the angles AOX
and AOY [pict24]. We have two unit vectors
OX and OY , the angle between these vectors is equal
to ( ) (if ) or ( ) (if ) in any case,
the dot product of these vectors
pict.24
OX OY OX OY cos( )
(because cos( ) // remark 2 // cos( ) ).
As OX OY 1 , then OX OY cos( ) [A1].
Let’s write the dot product in another way. Let e1 ,e2 are unit basis vectors, e1 has coordinates (1,0)
and e2 has coordinates (0,1) , then OX cos e1 sin e2 and OY cos e1 sin e2 , then
OX OY (cos e1 sin e2 ) (cos e1 sin e2 ) (cos cos ) e1 e1 (cos sin ) e1 e2
(sin cos ) e2 e1 (sin sin ) e2 e2 . As vectors e1 and e2 are perpendicular, then
e1 e2 0 e2 e1 and e1 e1 1 e2 e2 , therefore OX OY cos cos sin sin [A2].
From [A1] and [A2] follows [A].
From [A] immediately follows [B] cos( ) cos cos sin sin , really:
cos( ) cos( ( )) // from [ A] // cos cos( ) sin sin( )
cos cos sin sin and [B] is proved.
We have deduced two very important formulas [A] and [B]. There are also two basic formulas for
sin( ) and sin( ) . We have shown earlier that sin cos(90o ) (for any angle ).
Then
sin( ) cos 90o ( ) cos (90o ) ) cos(90o ) cos sin(90o ) sin
sin cos cos sin . And we got sin( ) sin cos cos sin [C].
The formula for sin( ) can be derived from [C]
sin( ) sin( ( )) sin cos( ) cos sin( ) sin cos cos sin .
Then [D] sin( ) sin cos cos sin .
Almost all the other trigonometric formulas can be very quickly derived from [A],[B],[C],[D].
For example:
sin cos cos sin
sin( ) sin cos cos sin cos cos cos cos tg tg
tg ( ) .
cos( ) cos cos sin sin cos cos sin sin 1 tg tg
cos cos cos cos
27
Vectors
There is also an important formula sin 2 sin( ) sin cos cos sin 2 sin cos .
And two important formulas for cos 2 cos( ) cos cos sin sin cos sin 2
2
(1 sin 2 ) sin 2 1 2 sin 2 [the first one] and
cos 2 cos2 sin 2 cos2 (1 cos2 ) 2 cos2 1 [the second one].
For some reason, most of students have difficulties with trigonometry, they try to remember by
3
heart formulas like cos sin , ctg tg , cos cos ,
2 2
despite the fact that there is no need to do that, everything can be quickly derived if we apply the
basic formulas for sin( ), cos( ) .
3
There is also a faster way to simplify formulas like cos . We just need to imagine as
2
3
a small positive angle. Then we imagine a unite circle and angles and , then we draw
2
perpendiculars to coordinate axises in order to obtain two equal right triangles, by using these right
triangles, we can quickly understand the connection between cosines and sines of given angles
3
(in our case, the connection between cosines and sines of and ).
2
There are also several very important formulas for a sum/difference of sines/cosines.
Let and are any angles. Let’s show how to derive the formulas for
sin sin , sin sin , cos cos , cos cos [T]. The main idea here is to represent
each angle and as a sum/difference of angles , and to use the formulas for
2
2
a sine/cosine of sum/difference. So , .
2 2 2 2
Then sin sin sin
sin sin cos
2 2 2 2 2 2
cos sin sin cos cos sin 2 sin cos .
2 2 2 2 2 2 2 2
And similarly for any other sum/difference in [T].
28
Determinants
Permutations
Let’s consider the set of several consecutive natural numbers {1,2,3} . The one-to-one mapping
1 2 3
f : 1 2, 2 3, 3 1 of this set onto itself can be denoted as - this symbol is called
2 3 1
1 2 3
a “permutation of numbers 1,2,3 ” . And conversely, the permutation [1] defines
1 3 2
one-to-one mapping f : 1 1, 2 3, 3 2 of {1,2,3} onto itself.
1 2 3 2 1 3
Let’s change positions of any two columns in the permutation [2]-this
2 3 1 3 2 1
writing is also a permutation, it defines exactly the same one-to-one mapping
f : 2 3, 1 2, 3 1, and we say that [1] and [2] are equal permutations, because they define
the same one-to-one mapping f .
Def1. {1,2,3... n} are consecutive natural numbers. Let j1 , j 2 ... j n are the numbers {1,2,3... n}
which are rewritten in some order (all the numbers j1 , j 2 ... j n are different, j k j m for k m
and each j k is some natural number from the set {1,2,3... n} ). And similarly v1 , v2 ... vn are
the numbers {1,2,3... n} which are rewritten in some order. The symbol
j1 j2 j3 ... jn
[S] is called a “permutation of numbers {1,2,3... n} ”,
v1 v2 v3 ... vn
or just a “permutation”.
And [S] is NOT a matrix. This symbol represents one-to-one mapping of {1,2,3... n} onto itself:
f : j1 v1, j2 v2 .... jn vn [3]. We say that two permutations are equal if they represent the
same one-to-one mapping f .
Obviously, if we change positions of any two columns in [S] (if we permute any two columns), a new
permutation will be equal to an initial one (because they both define the same one-to-one mapping).
Obviously, two permutations are equal if and only if one of these permutations can be obtained from
the other by several permutations of it’s columns.
In any permutation we can always permute several colums in order to turn the first row into
1 2 3 ... n
1,2,3 ... n , we will get the permutation - which is called the “standard form”
a1 a2 a3 ... an
of an (initial) permutation .
The main idea here is that any permutation can be rewritten in a more convenient way, when it’s
first row consists of consecutive natural numbers 1,2,3 ... n . And we achieve it just by permuting
several columns, so we will get exactly the same permutation as earlier, but now it is written in a
“better way”.
30
Determinants
Exercise1. Show that there are exactly n! 1 2 ... n different permutations
j1 j2 j3 ... jn
1 2 3
v v v ... vn
(where j1.. jn and v1..vn are both numbers 1,2.. n which are written in some order)
Hint. Use that any permutation can be written in the standard form.
Sign of a permutation. For any permutation, we say that the pair of elements (a, b) in the first
row is an inversion if a is on the left of b and a b . In the same way, the pair of elements
(c, d ) in the second row is an inversion if c is on the left of d and c d .
1 3 4 2
Let’s consider the permutation . In the first row: (3,2) is an inversion,
3 1 2 4
because 3 is on the left of 2 and 3 2 . Also (4,2) is an inversion. So, there are exactly two
inversions in the first row.
In the second row: here (3,1) and (3,2) are inversions. Then there are four inversions in .
Def2. If the total number of inversions in ( the number of inversions in the first row the
number of inversions in the second row ) is even, then we say that is an even permutation and
we write sgn 1. If the total number of inversions in
is odd, then we say that is an odd
permutation, and we write sgn 1. (The number sgn is called a “sign of a permutation”).
Theorem1. If some permutation is even/odd, then any permutation , which is equal to ,
is also even/odd.
Proof. Let’s permute some neighbor columns in , then the number of inversions k in the first
(upper) row will become k 1 or k 1, the number of inversions m in the second (lower) row will
become m 1 or m 1. So, there initially were m k inversions, and now we have m k 2 or
m k 2 or m k inversions. The parity of the total number of inversions in stays the same.
I.e., if the total number of inversions was even, then it is still even after a permutation of any
neighbor columns. And similarly if the total number of inversions was odd.
Let now we have two columns in and there are exactly h other columns between them.
Let’s take the right column, we will move it to the left towards the left column, by permuting
the right column with it’s neighbors from the left side. Soon our columns will become neighbors and
we will change their positions, then we will move the other column to the right, until it becomes
on the right place. Eventually two needed columns will be permuted, it is done by several
permutations of columns that stood near by. The parity of the total number of inversions does not
change when we permute any two neighbor columns, therefore the parity hasn’t changed.
Let and is even or odd. We can obtain from by several permutations of columns of .
31
Determinants
Each permutation of columns gives a new permutation with the same parity.
Then eventually has the same parity as .
The easiest way to determine the parity of any permutation (and therefore to determine it’s sign)
1 2 3 ... n
is to rewrite this permutation in the standard form . In such form there are
a1 a2 a3 ... an
no inversions at all in the first row, all the inversions may be only in the second row. So, it’s easier
to count the total number of inversions and to determine the sign of a permutation.
a11 .. a1n
Determinant. is n n matrix .. .. .. (from now on we consider only square
a
n1 .. ann
matrixes). For any square matrix A, the determinant of A is the number det , where
a11 .. a1n
сan be
[V1] det sgn a1k1 ... ankn
denoted as .. .. ..
all different permutations a
1 .. n n1 .. ann
k1 .. k n
So, the determinant of n n matrix A is a sum of n! products. Each product consists of matrix
elements which stay at different rows and columns, and there is a sign or in front of each
j j j ... jn
product. As any permutation 2 1 3 can be written in the standard form
2 1 3
v v v ... vn
1 2 3 ... n
, and different permutations have different standard forms,
1 2
a a a 3 ... a n
we can rewrite the definition of det in the next way:
[V2] det all different
sgn a j1k1 ... a jn kn
permutations
j2 j1 j3 ... jn
v2 v1 v3 ... vn
In practice we never calculate concrete determinants through these definition-formulas, it’s enough
to remember the formulas for determinants of “small” 2 2 and 3 3 matrixes (even for 2 2
matrixes is enough). Any determinant of a bigger matrix (like 4 4 or 5 5 ) can be simplified and
represented as a sum of determinants of smaller matrixes. There will be the examples soon.
Exercise2. Prove the next simple properties of det (by using [V1] or [V2], which one is more
convenient). [1] For any matrix : det det . [2] If has a row of zeroes, or a column of
32
Determinants
zeroes, then det 0 . [3] If we permute any two rows in A, or any two columns in A, then a new
matrix has the determinant det .
From [3] follows [3.1]: If A has two equal columns, or two equal rows, then det 0 .
[4] If we multiply by all the elements of some row/column of A, the determinant of a new matrix
is det .
[5] There is some determinant, where every element of some row is represented as a sum of two
elements, then such determinant can be represented as a sum of two other determinants:
a11 .. .. a1n a11 .. .. a1n a11 .. .. a1n
b1 c1 .. .. bn cn b .. .. bn c .. .. cn
1 1 .
.. .. .. .. .. .. .. .. .. .. .. ..
a n1 .. .. a nn an1 .. .. a nn a n1 .. .. a nn
The same representation is true for any column.
From [5] and [3.1] follows: we can take any row and add to it any linear combination of any other
rows, a determinant will not change. And the same is true for any column.
0
Def. is a block matrix if it looks like , here B and С are some square matrixes
0 С
(and therefore is a square matrix). Every zero 0 denotes a rectangle of zeroes.
Exercise3. Show that for any block matrix the next is true: det det det C .
Def. is a square matrix. Let’s fix any aij and replace it by 1 , in the same time all the other
elements in the row with number i and in the column with number j we replace by zeroes
(and all the other elements of A stay the same). The determinant of the new matrix is called
an algebraic complement of aij , and we will denote it like ij . (Note: ij is NOT a matrix, it is
a number, a determinant of a concrete matrix ).
For example,
0 a12 .. a1n 1 0 .. 0
1 0 .. 0 0 a22 .. a2 n
-an algebraic complement of a 21 and -an algebraic
.. .. .. .. .. .. .. ..
0 an 2 .. ann 0 an 2 .. ann
complement of a11 .
33
Determinants
Let’s derive a very important formula:
a11 a12 .. a1n a11 a12 .. a1n
a21 a22 .. a2 n (a 0... 0) (0 a22 0 ... 0) .. (0 ... 0 a2 n )
21
.. .. .. .. .. .. .. ..
an1 .. .. ann 0 an 2 .. ann
a11 a12 .. a1n a11 a12 .. a1n a11 a12 .. a1n
a 0 .. 0 0 a22 .. 0 0 0 .. a2 n
21 ...
.. .. .. .. .. .. .. .. .. .. .. ..
an1 an 2 .. ann an1 an 2 .. ann an1 an 2 .. ann
a11 a12 .. a1n a11 a12 .. a1n a11 a12 .. a1n
1 0 .. 0 0 1 .. 0 0 0 .. 1
a21 a22 ... a2 n
.. .. .. .. .. .. .. .. .. .. .. .. .
an1 an 2 .. ann an1 an 2 .. ann a n1 an 2 .. ann
Let’s consider the first determinant, we can annul in it almost all the first column. Really, let’s take
the first row and add to it the second row, multiplied by a11 (the first determinant will not change,
it follows from [5] and [3.1]). Then we take the third row and we add to it the second row, multiplied
by a31, then we take the fourth row and we add to it the second row, multiplied by a41 and etc.
0 a12 .. a1n
1 0 .. 0
Finally, the first determinant will look like: 21 - it is an algebraic complement
.. .. .. ..
0 an 2 .. ann
of a21 . And similarly, for any other determinant in our sum, then:
a11 a12 .. a1n 0 a12 .. a1n a11 0 .. a1n a11 a12 .. 0
a21 a22 .. a2 n 1 0 .. 0 0 1 .. 0 0 0 .. 1
a21 a22 ... a2 n
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
an1 .. .. ann 0 an 2 .. ann an1 0 .. ann a n1 an 2 .. 0
det a21 21 ... a2n 2n . The same formula is obviously true for any other row
det ai1 i1 ... ain in [D].
(with number i ), i.e.,
And similarly for any column (with number j ) det a1 j 1 j ... anj nj [E].
These formulas are not our final formulas, so we need to continue.
34
Determinants
Def. Let’s fix any element aij of and delete the row number i and the column number j .
We will get (n 1) (n 1) matrix ij , the determinant of that matrix must be denoted as
ij det ij
Assertion1. ij (1)i j ij for any i, j .
Proof. Let’s take some determinant ij , we take it’s column with number j (it consists of several
zeroes and 1 ) we will consecutively permute this column with it’s left neighbor columns.
Finally this column will become the first one, and the row with number i will look like (1,0,0....0) .
We will permute this row with it’s upper neighbor rows. Finally this row will become the first one
1 0 ... 0
0
and the determinant will look like: . This determinant is obtained from ij by j 1
... ij
0
consecutive permutations of columns and i 1 consecutive permutations of rows.
Each permutation of any two rows/columns changes the sign of a determinant, therefore
1 0 ... 0
0
ij (1) j 1 i 1 - there is a block matrix on the right side.
... ij
0
According to the exercise3, the right determinant is equal to det ij ij . Therefore
ij (1) j 1 i 1 ij ij (1) j i ij ij (1) j i ij .
Then from [D] we have: det ai1 (1)i1 i1 ... ain (1)in in [D1].
And similarly, from [E] we have: det a1 j (1)1 j 1 j ... anj (1) n j nj [E1].
Notice that any ij is a determinant of a matrix which is smaller than A. Formulas [D1] and [E1]
are our basic formulas for determinants-calculation.
a b
The simpliest determinant: ad bc .
c d
a b c
Let’s calculate d e f by using [D1] (we take the first row).
g l p
35
Determinants
Then
a b c
e f d f d e
d e f a (1)11 b (1)1 2 c (1)1 3 a(ep fl) b(dp fg) c(dl eg )
l p g p g l
g l p
aep bfg cdl ceg bdp afl . It’s very easy to remember this formula if we notice
the geometrical pattern (the positions of our factors in the initial 3 3 matrix).
a b c
So d e f aep bfg cdl ceg bdp afl .
g l p
Example1. Examples of determinants:
1 2 2 4
(a) 1 3 2 2 3 4 1 , (b) (2) 4 4 3 8 12 20 ,
2 3 3 4
1 0 2
1 3
(c ) (1) 3 3 (2) 3 6 3 , (d ) 2 2 3 the second column contains zero,
2 3
1 1 4
so let’s expand this determinant along the second column:
1 0 2
2 3 1 2 1 2
2 2 3 0 (1)1 2 2 (1) 2 2 (1) (1) 3 2
1 4 1 4 2 3
1 1 4
2 (1 4 (2) 1) 1 (1 3 (2) 2) 12 7 19.
1 2 3
(e) 4 5 6 let’s expand along the first row:
7 8 9
1 2 3
5 6 4 6 4 5
4 5 6 1(1)11 2(1)1 2 3 (1)1 3
8 9 7 9 7 8
7 8 9
1 5 9 6 8 2 4 9 6 7 3 4 8 5 7 3 12 9 0 .
We could also understand that the determinant (e) is zero almost without any calculations.
In any matrix we can take any row/column and add to it any linear combination of other
rows/columns, a determinant will not change. In particular, we can take any row/column and
1 2 3
subtract from it any other row/column. We have 4 5 6 , let’s take the second row and subtract
7 8 9
the first row from it:
36
Determinants
1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
4 5 6 4 1 5 2 63 3 3 3 3 3 3 3 3 3 2 3 3 3 20 0
7 8 9 7 8 9 7 8 9 7 1 8 2 9 3 6 6 6 3 3 3
1 1 3 4
2 2 2 4
Example2. ( f ) - it is a determinant 4 4 . It’s always very important to notice
3 3 4 6
4 9 8 10
if we can make some elementary manipulations with rows and columns to simplify the determinant
(like to take some row/column and add/subtract to/from it some other row/column), we want to make
so many zeroes as possible in some row/column, and after we will use [D1]/[E1].
1 1 3 4 1 1 (3 1) 4 1 1 4 4 1 1 4 (4 4) 1 1 4 0
2 2 2 4 2 2 (2 2) 4 2 2 4 4 2 2 4 (4 4) 2 2 4 0
3 3 4 7 3 3 (4 3) 7 3 3 7 7 3 3 7 (7 7) 3 3 7 0
4 9 8 10 4 9 (8 4) 10 4 9 12 10 4 9 12 (10 12) 4 9 12 2
now we can expand this determinant along the last column
1 1 4 0
1 1 4 1 1 4
2 2 4 0
2 (1) 4 4 2 2 4 2 2 2 4
3 3 7 0
3 3 7 3 3 7
4 9 12 2
1 1 4 1 1 4
2 2 2 1 2 2 (1) 4 2 4 2 4 0 12 and finally, we expand this determinant
3 3 7 3 3 7
along the second column
1 1 4
4 12 1 4
2 4 0 12 2 (1) (1)1 2 3 (1)3 2 21 4 7 12 3 312 16
3 3 7 3 7 4 12
2 8 12 8 .
Def: a matrix A is called diagonal if aij 0 for any i j . The only non-zero elements in such
matrix may stay on it’s main diagonal a11, a22... ann .
a11 .. 0
diag{a11, a22... ann} .. ... .. .
0 ... a
nn
The matrix diag{1,1... 1} is called a “unit matrix”.
If all the elements below the main diagonal are zeroes ( aij 0 for any i j ), then A is called
upper triangular.
37
Determinants
If all the elements under the main diagonal are zeroes ( aij 0 for any i j ), then A is called
lower triangular.
If a matrix is lower triangular or upper triangular, then we call it just “triangular”.
Exercise4. Show (by using [V1] or [D1] or [E1]) that the determinant of any triangular matrix is
a product of it’s diagonal elements. So, A is triangular det a11 a22 ... ann .
In particular det 1.
The main theorem. For any n n matrixes A and B: det( ) det det .
This theorem is very important, and we will provide a very simple proof for it. But at first we need
to introduce some simple theory, which will allow us to change our approach to determinants.
38
Determinants
Linear spaces
Def. F is a field and L is a commutative additive group. There exist the operation
(“multiplication”) "": F L L that for every pair , v || F , v L compares some element
v L . And "" has the next properties:
[A] 1 v v || v L and ( v ) ( ) v || , F , v L .
[B] “sort of distributivity” ( ) v v v and (v m) v m for any
, F , v , m L .
Then L is called a linear space under the field F. Elements of L are called “vectors”.
And elements of F are called “scalars”. We can just say “L is a space under F ” or “L is a space”.
The most important example is the linear space of 3 d vectors which was built earlier.
Notice, in mathematical model equal vectors must be considered as the same vector.
So when we say “a vector” we imply the whole class of equal vectors. 3 d vectors form
a commutative group L, we can multiply vectors by real numbers R, so we have a linear space
under R, it’s easy to see that [A] and [B] are true.
Let’s notice that the set of all m n matrixes which consist of real numbers, form a vector space
under R (in this case every matrix is a “vector” and any real number is a “scalar”).
a1
Two most important vector spaces (of matrixes) are: the linear space R of n 1 columns ...
n
a
n
and the linear space Rn of 1 n rows a1 ... an .
Def. L is a linear space under F . For any vectors v1, v2 ... vn L and any scalars 1, 2 ... n F
the vector 1 v1 2 v2 ... n vn is called a “linear combination of v1, v2 ....vn with coefficients
1, 2 ....n ” or just a “linear combination of v1, v2 ....vn ”.
Def. L is a linear space. If there exist some set of vectors {e1... en } L such that any vector v L
can be represented as a linear combination of these vectors v 1e1 ... nen and for any vector
v it’s representation is unique, then the set {e1... en } is called a basis of the space L.
Example. Prove that the “standard basis vectors” e1 (1,0... 0) e2 (0,1... 0) .... en (0,0...1) form
1 0 0
0 1 0
a basis of Rn and the “standard basis vectors” e1 , e2 .... en form a basis of R .
n
.. .. ..
0 0 1
Exercise. Provide some basis of the space of m n matrixes.
39
Determinants
From now on, L is always a linear space with some fixed basis {e1... en } L .
And L is a linear space under the field R of real numbers.
Def. L is a linear space. Every rule f that for any vector v L compares some real number
f (v ) R is called a form on L.
For any vector v , the number f (v ) is called “the value of f on v ”.
Two forms f , g , which are defined on L , are equal if f (v ) g (v ) for any v L . The form f is
called linear if [A] f (v ) f (v ) for any v L and any R
[B] f (v u ) f (v ) f (u ) for any v , u L .
Assertion1. f is a linear form. Then the value of f on any vector v is uniquely defined by
the values of this form on the basis vectors e1 , e2 ... en .
Proof. Really, for any vector v L there is a unique representation v v1e1 v2e2 ... vnen , then
f (v ) v1 f (e1 ) v2 f (e2 ) ... vn f (en ) , then the assertion1 is true.
Therefore: two linear forms f , g are equal they are equal on the basis vectors e1 , e2 ... en :
f (ek ) g (ek ) for any k .
Def. Every rule f that for any ordered set of vectors (v1 , v2 ... vn ) (every vector vk is taken from L )
compares some real number f (v1 , v2 ... vn ) R is called an “ n -form on L ” or just “a form on L ”.
A form f is called normalized if f (e1 , e2 ... en ) 1.
f is called polylinear if f is linear in all it’s arguments, i.e., for any number T [1, n] :
f (v1.. vT uT .. vn ) f (v1.. vT .. vn ) f (v1.. uT .. vn ) and f (v1.. vT .. vn ) f (v1.. vT .. vn ) .
f is called symmetric if it doesn’t change when we permute any two of it’s arguments.
For any T , J [1, n] : T J there must be f (v1 .. vT .. v J .. vn ) f (v1 ,.. v J .. vT .. vn ) .
And f is called skew-symmetric if it changes the sign when we permute any two of it’s arguments.
For any T , J [1, n] : T J there must be f (v1.. vT .. vJ .. vn ) f (v1,.. vJ .. vT .. vn ) .
Let A is n n matrix. Rows of A are some vectors from Rn .
The determinant det of any matrix A is a real number. Any matrix A can be considered
as exactly n vectors from Rn , which are the 1 st , 2 nd , .. n th rows of A.
Then det is an n -form on the space Rn , really
40
Determinants
components of v1
det f (v1.. vn ) .... ...... . For any vectors v1, v2 .. vn we have: f (v1.. vn ) det
components of vn
is a concrete real number. Then det is really an n-form on Rn .
According to the simpliest properties of determinant, the form f det (on Rn ) is normalized
(because det 1), polylinear and skew-symmetric.
From now on, (k1, k2 ... kn ) are the numbers 1,2... n which are rewritten in some order.
Let sgn(k1 , k 2 ... k n ) 1 if the total amount of inversions in the row (k1, k2 ... kn ) is even and
sgn(k1 , k 2 ... k n ) 1 if the total amount of inversions in the row (k1, k2 ... kn ) is odd.
1 2 3 ... n
Let’s notice that sgn sgn(k1, k2 ... kn ) , because there are no inversions at all
k1 k2 k3 ... kn
in the first row.
Theorem2. There exist exactly one n -form on Rn which is normative, polylinear and
skew-symmetric.
Proof. The existence is proved (the determinant is such form). Let’s prove the uniqueness.
Let f (v1.. vn ) is some normative, polylinear and skew-symmetric form on Rn . As f is polylinear,
it changes the sign when we permute any two of it’s arguments. It’s easy to understand that from
here follows: f (vk1 .. vk n ) sgn(k1.. kn ) f (v1.. vn ) [E]. Let’s notice, as f is normative,
for the basis vectors e1.. en the equality [E] looks like
f (ek1 .. ekn ) sgn(k1.. kn ) f (e1.. en ) f (ek1 .. ekn ) sgn(k1.. kn ) [E1].
Each vector v1.. vn can be uniquely represented as a linear combination of basis vectors:
v1 a11e1 a12e2 ... a1nen || v2 a21e1 a22e2 ... a2nen || .... || vn an1e1 an 2e2 ... annen ||
then f (v1.. vn ) f (a11e1 a12e2 ... a1nen ,..., an1e1 an 2e2 ... annen ) [linearity]
a1k a2 k
all different sets
1 2
... ankn f (ek1 .. ek n ) //[ E1] //
( k1 , k 2 ... k n )
all different sets
a1k1 a2 k 2 ... ankn sgn(k1 , k2 ...kn ) a1k1 ... ankn
all different permutati ons
sgn
( k1 , k 2 ... k n ) 1 .. n
k1 .. k n
we came to exactly the same formula that we had in the determinant-definition.
41
Determinants
So, the value of our form f (v1.. vn ) on any set of vectors v1.. vn Rn is exactly the same
vector v1
vector v2
as a determinant of the matrix which is formed from these vectors .
... ...
vector vn
The uniqueness is proved.
Theorem3. f and g are both polylinear and skew-symmetric forms on Rn .
If f (e1.. en ) g (e1.. en ) , then f g on Rn .
Proof. We have derived above the formula for any polylinear form:
f (v1 .. vn ) a
all sets
1k1 a2 k 2 ... ankn f (ek1 .. ek n ) [ skew symmetricity ]
( k1 , k 2 ...k n )
a1k
all sets
1
a2k2 ... ankn f ( e1.. en ) sgn(k1 , k2 ... kn ) , from here we see that the value
( k1 , k 2 ... k n )
of the form f on any set of vectors (v1.. vn ) is uniquely defined by the number f (e1.. en ) .
And exactly the same formula is true for g . So if f (e1.. en ) g (e1.. en ) , then f g on Rn .
Let’s prove now the main theorem: det( ) det det .
Let 1... n are the columns of B, then has the columns: 1 ... n (it follows from
the definition of matrix multiplication). So, we can write:
1... n and 1 ... n .
We want to prove that:
det( ) det det det 1 ... n det det1... n .
Let v1.. vn are any vector-columns from R . Let’s denote det v1 ... vn f (v1...vn ) - it is
n
obviously a polylinear and skew symmetric form and vectors v1...vn are it’s arguments.
And det detv1...vn g (v1...vn ) is also a polylinear and skew symmetric form with arguments
v1...vn . It’s easy to see that f (e1... en ) g (e1... en ) det then, according to the theorem3, f g
on any set of vectors v1... vn from R , in particular for 1... n we have:
n
f (1... n ) g (1...n ) det( ) det det .
42
Determinants
Inverse matrix
Def: a matrix A is called singular if det 0 and nonsingular if det 0.
Def. A is n n matrix. If there exist some n n matrix B such that ,
then B is called an inverse matrix, and we denote 1 . So 1 1 .
1
Comment1. If is singular, then an inverse matrix does not exist.
1 1
Proof. Let’s assume that exists, then
, then (the main theorem)
det 1 det det det 1 0 1 0 1 and we have a contradiction. Then 1 doesn’t
exist.
Def. Let is m n matrix. The transposed matrix is n m matrix C such that cij a jii, j
Put simply: rows of are the columns of and columns of are the rows of .
Examples.
1
2 3 2 1 1
|| 1 2 || 2 1 2 3 ||
1 4 3 4 2 3
1 2 1 2 3 1 4 7 2 0
1 3 5
2 3 5
3 4 || 4 5 6 2 5 8 || 3 1 .
5 6 2 4 6 7 8 9 3 6 9 0 1 6 5 6
When we go from to we say that we “transpose our matrix”.
Exercise. Prove the next basic properties: [1] ( ) for any matrix A,
[2] ( ) for any m n and n k matrixes A and B.
Theorem4. For any nonsingular matrix A there exist a unique inverse matrix 1 .
1
Let’s prove at first the existence of .
Lemma1. If we multiply any row of A by algebraic complements of any other row, we will get zero.
Comment. Let’s take for example the first row a11, a12... a1n and 21, 22... 2 n are algebraic
complements of the second row, then there must be a1121 a1222 ... a1n 2n 0 .
Let’s take the fourth row a41, a42... a4 n and 11, 12... 1n are algebraic complements of
the first row, then a4111 a4212 ... a4n 1n 0 and etc.
Let’s show for example that a1121 a1222 ... a1n 2n 0 .
Let’s take the determinant detA and replace it’s second row by it’s first row:
43
Determinants
a11 a12 .. .. a1n
a11 a12 .. .. a1n
a31 a32 .. .. a3n [T] then the sum a1121 a1222 ... a1n 2n is exactly the determinant [T].
.. ... .. .. ..
an1 an 2 .. .. ann
And [T] is a determinant of a matrix with two equal rows, such determinant is equal to zero, then
a1121 a1222 ... a1n 2n 0 . The same is true for columns: if we multiply any column of A
by algebraic complements of any other column, we will get zero.
1
Let’s build now.
[step1] We take the initial matrix A and instead of every element aij we write it’s algebraic
complement: aij ij (1)i j ij .
ij
[step2] Every element of the new matrix must be divided by det . I.e.: ij .
det
1
[step3] We transpose the matrix from the step2. The new matrix is exactly
a11 a12 .. .. a1n 11 12 .. .. 1n
a11 a22 .. .. a2 n 21 22 .. .. 2n
a31 a32 .. .. a3n a
ij
31 32 .. .. 3n
ij / det
ij ij
.. ... .. .. .. .. ... .. .. ..
an1 an 2 .. .. ann n1 n 2 .. .. nn
11 / det 12 / det .. .. 1n / det
21 / det 22 / det .. .. 2 n / det
31 / det 32 / det
ij / det
.. .. 3n / det Transpose
ij
.. ... .. .. ..
n1 / det n 2 / det .. .. nn / det
11 / det 21 / det .. .. n1 / det
12 / det 22 / det .. .. n 2 / det
Transpose
13 / det 23 / det .. .. n3 / det 1 .
.. ... .. .. ..
1n / det 2 n / det .. .. nn / det
44
Determinants
Now, by using the lemma1, we can easily check that
a11 a12 .. .. a1n 11 / det 21 / det .. .. n1 / det
a11 a22 .. .. a2 n 12 / det 22 / det .. .. n 2 / det
1 a31 a32 .. .. a3n 13 / det 23 / det .. .. n3 / det ,
.. ... .. .. .. .. ... .. .. ..
an1 an 2 .. .. ann 1n / det 2 n / det .. .. nn / det
and also that
11 / det 21 / det .. .. n1 / det a11 a12 .. .. a1n
12 / det 22 / det .. .. n 2 / det a11 a22 .. .. a2 n
1 13 / det 23 / det .. .. n3 / det a31 a32 .. .. a3n .
.. ... .. .. .. .. ... .. .. ..
1n / det 2 n / det .. .. nn / det an1 an 2 .. .. ann
The existence is proved.
Uniqueness. Let’s assume that there exist some other inverse matrix B such that
. We already have 1 1 . Let’s multiply both sides of
1
by from the left side, then
1 ( ) 1 1 ( ) 1 [associativ ity of matrix multiplication]
(1 ) 1 1 . So, for any nonsingular matrix A an inverse matrix 1 is
unique.
Properties of an inverse matrix. [A] det 1 1/ det . Really 1 1 , then
1 det(1 ) det() det det(1 ) det() 1.
[B] If or , then 1 (the proof is the same as in the uniqueness).
[C] A is a nonsingular matrix. If for some matrixes , we have , then
[D] For any nonsingular matrixes , : ( )1 1 1 .
Proof. [C] is obvious. Let’s prove [D]. By definition of an inverse matrix
( ) ( )1 / by def / . Let’s calculate ( ) 1 1 . As matrix multiplication is
associative, we can rearrange brackets in any way we want:
( ) 1 1 ( 1 ) 1 () 1 1 . Then: if we multiply
1 1
a nonsingular matrix AB by any of the matrixes ( ) , 1 from the right side,
we will get the unit matrix . Then, according to [C], there must be ( )1 1 1 .
45
Determinants
[E] For any nonsingular matrix : ( 1 ) ( )1 .
Let’s show that if we multiply a nonsingular matrix by any of these matrixes from the
right side, we will get the unit matrix (then, by [C] these matrixes are equal). So,
( )1 / by def / and
(1 ) // use the equality () // (1 ) () .
1 8
Examples. Find the inverse matrix for .
2 15
[step1] a11 1 11 (1)11 15 15 || a12 8 12 (1)1 2 2 2 || a21 1 21 (1)2 1 8 8 ||
1 8 15 2
a22 15 22 (1)2 2 1 1 Then .
2 15 8 1
1 8
[step2] The determinant of the initial matrix is det 1 15 8 2 1 , then
2 15
15 2 15 / 1 2 / 1 15 2
.
8 1 8 / 1 1/ 1 8 1
15 2 15 2 15 8
And finally, [step3] 1 . It’s quite important to check our
8 1 8 1 2 1
1 8 15 8 1 0 15 8 1 8
solution: . Notice, according to the property [B],
2 15 2 1 0 1 2 1 2 15
1 8 15 8 1 0
it’s enough to check only that or to check only that
2 15 2 1 0 1
15 8 1 8 1 0
. So there is no need to check both equalities (and similarly in any other
2 1 2 15 0 1
case), which is a very convenient thing, it is especially convenient for big-size matrixes, because
multiplication of such matrixes is quite a tedious task.
1 2 3
Example. Find the inverse matrix for 1 1 2 .
1 3 5
[step1]
1 2 1 2 1 1
a11 1 11 (1)11 1 || a12 2 12 (1)12 7 || a13 3 13 (1)13 4 ||
3 5 1 5 1 3
2 3 1 3 1 2
a21 1 21 (1) 21 1 || a22 1 22 (1) 22 2 || a23 2 23 (1) 23 1 ||
3 5 1 5 1 3
2 3 1 3 1 2
a31 1 31 (1)31 1 || a32 3 22 (1)32 5 || a33 5 33 (1)33 3 ||
1 2 1 2 1 1
1 2 3 1 7 4
Then: 1 1 2 1 2 1 .
1 3 5 1 5 3
46
Determinants
1 2 3
[step2] The determinant of the initial matrix is det 1 1 2 . Let’s calculate it (expand along
1 3 5
the first row):
1 2 1 2 1 1 1 2 1 2 1 1
1 (1)11 2 (1)12 3 (1)13 2 3 1 2 (7) 3 (4) 1 ,
3 5 1 5 1 3 3 5 1 5 1 3
1 7 4 1 / 1 7 / 1 4 / 1 1 7 4
then det 1 and 1 2 1 1 / 1 2 / 1 1 / 1 1 2 1 .
1 5 3 1/1 5 /1 3 /1 1 5 3
1 7 4 1 1 1
And finally [step3] 1 2 1 7 2 5 1 , now we can check that 1 .
1 5 3 4 1 3
For Reader’s practice :
sin x cos x cos x
[1] Solve the equation cos x sin x cos x 0 on the segment 45o x 45o . Answer: x 45o .
cos x cos x sin x
3/2 1/ 2 1 1 1 2005
[2] P and A and Q PAP T . Find PT Q 2005P . Answer: .
1/ 2
3 / 2 0 1 0 1
sin cos sin 2
[3] Show that for any : sin 120o
cos 120o
sin 2 240o 0 .
sin 120o
cos 120o
sin 2 240o
sin cos sin( )
[4] Show that for any , , : sin cos sin( ) 0 .
sin cos sin( )
12 22 32 42
22 32 42 52
[5] Calculate 2 Answer: 0 .
3 42 52 62
42 52 62 72
1 2 1
[6] Find the absolute value of the determinant 3 2 2 2 2 2 1 Answer: 16 2 .
32 2 22 2 1
a b c
[7] Calculate A b c a and AAT 1, abc 1 . Find the value of a3 b3 c3 Answer: 4 .
c a b
47
Complex Numbers
Construction of complex numbers
Def. С is the field: С consists of real numbers R and the element i such that i 2 1 .
The addition and multiplication on С are extensions of the addition and multiplication on R, i.e.,
for any a, b R C we have:
a b (addition by the rules of C ) a b (addition by the rules of R) ,
a b (multiplication by the rules of C ) a b (multiplication by the rules of R) .
С is called a field of complex numbers and elements of С are called complex numbers.
Let’s build C . We fix the any set R of real numbers. Let’s consider the set Caux of all pairs (a, b)
where a, b are any real numbers. We define addition and multiplication on Caux .
Def. (a, b) (c, d ) / by def / (a c, b d ) and (a, b) (c, d ) (ac bd , ad bc) .
How to remember it? Imagine that any pair (a, b) denotes the element a b i of some field,
where i 2 1 . Then (a, b) (c, d ) (a b i) (c d i) (a c) (b d ) i (a c, b d )
and (a, b) (c, d ) (a b i) (c d i) (a c b d ) (a d b c) i (ac bd , ad bc) .
Assertion1. The set (Сaux ,, ) is a field.
Proof. These properties can be checked straightly. We will give an another proof.
a b
Let’s consider the set X of 2 2 matrixes of real numbers A || a, b R .
b a
We can describe the set X as a set of all 2 2 matrixes which elements on the main diagonal
are equal and elements on the secondary diagonal are opposite.
Auxiliary1. The set X is a field under the matrix addition and multiplication.
Really A, B we have
a b с d a b с d a с (b d )
A , B A B
b a d с b a d с b d a c
a b с d ac bd ad bc ac bd (bc ad )
and A B .
b a d с bc ad bd ac bc ad ac bd
So X is closed under the matrix addition and multiplication. As addition of matrixes is associative
0 0
and commutative, it is also associative and commutative on X. The zero matrix
0 0
a b a b
obviously belongs to X. For any matrix A , an opposite matrix A
b a b a
also belongs to X. Then X is a commutative group under addition.
49
Complex Numbers
Multiplication of any 2 2 matrixes is associative and also distributive over addition,
in particular it is true on X. Then X is a ring.
Let’s show that all non-zero elements of X form a commutative group with respect to multiplication.
At first we need to show that \ is closed under multiplication. Let’s notice that for any
a b 0 0
A \ A we have det A a 2 b 2 0 . And similarly, for any other
b a 0 0
matrix B \ we have det B 0 , then for A B we have det A B det A det B 0 ,
p k
in any case A B looks like and p 2 k 2 0 , then one of the numbers p, k is not zero,
k p
therefore A B is not a zero matrix A B A B \ . Next, multiplication is associative
on \ , because multiplication of any matrixes is associative. Multiplication is also commutative
a b с d ac bd (bc ad )
on X, really, and
b a d с bc ad ac bd
с d a b ac bd (bc ad )
.
d с b a bc ad ac bd
Any matrix A \ is a nonsingular matrix det A 0 , then it has an inverse one A1 ,
1
we have to check that A \ . Let’s take any
a b 1 a b
A \ A A1 2 2 \ , then \ is a commutative group
b a a b b a
with respect to multiplication. And X is a field.
a b
Next, the mapping f : Сaux such that f (a, b) is one-to-one mapping.
b a
For any matrix from X the mapping f compares a pair of real numbers which stay in it’s first
column. For any matrixes A, B we have f ( A B) f ( A) f ( B) and
f ( A B) f ( A) f ( B) (here and are addition and multiplication of pairs on Сaux ). These two
properties off are obvious, just look above at the places above where we calculated a product and a
sum of matrixes from X. Therefore (C ,,) is a field.
Comment. The Reader can refer to the 1-st book “Construction of numbers, length and area”, there
was the theorem4 in the chapter “Groups, Rings, Fields”. We use this theorem now to conclude that
Сaux is a field. We will use it again very soon (the next page) to conclude that С is a field.
Let’s replace every pair (a,0) Сaux by the number a . We will get the set С, which consists of real
numbers R and some pairs of real numbers: С R ( pairs (a, b) of real numbers where b 0) .
The setС is called a set of complex numbers. Let’s turn it into a field.
We need to define addition and multiplication on С.
50
Complex Numbers
We define one-to-one mapping f : Сaux C such that: (a,0) Caux f (a,0) / by def / a
and for any other pair (a, b) Caux f (a, b) / by def / (a, b) .
Any elements z, m C can be uniquely represented as z f (a, b), m f (c, d ) , and we define:
z m f (a, b) f (c, d ) / by def / f (a, b) (c, d ) and
z m f (a, b) f (c, d ) / by def / f (a, b) (c, d ).
Then f : Сaux C is the mapping, and for any (a, b), (c, d ) Сaux we have
f (a, b) (c, d ) f (a, b) f (c, d ) and f (a, b) (c, d ) f (a, b) f (с, d ) .
From here immediately follows that С is a field, and we also have R C . (Theorem4, Book1, page40 )
Assertion2. The addition and multiplication on С are extensions of the addition and
multiplications on R C .
Proof. Let’s fix arbitrarya, b R C , in order to find their sum/product in С we need to find their
~
preimages in Сaux . Let’s denote for a while ,~ -the addition and multiplication by the rules of С.
So, a f (a,0) and b f (b,0) , then
a~ f (b,0) // by def // f (a,0) (b,0) f (a b,0) a b R and also
b (in C ) f (a,0) ~
a ~ b (in C ) f (a,0) ~ f (b,0) // by def // f (a,0) (b,0) f (a b,0) a b R .
Everything is proved. As the addition and multiplication on С are extensions of the addition and
multiplication on R C , it is appropriate to use the same symbols , to denote these operations.
Def. С contains the pair (0,1) , this pair is denoted by the symbol i (0,1) .
Assertion3. The element i has the property i 2 1 . Every element z C can be uniquely
represented as z a b i || a, b R .
Proof. Let’s take i (0,1) C , then
i i f (0,1) f (0,1) // by def // f (0,1) (0,1) f (1,0) 1.
Existence of representation. Let z C . Then there exist the unique pair (a, b) Сaux such that
f (a, b) z , for (a, b) Сaux we have (a, b) (a,0) (0, b) (a,0) (b,0) (0,1) (in Сaux ).
Then z f (a, b) f (a,0) (b,0) (0,1) f (a,0) f (b,0) (0,1)
f (a,0) f (b,0) f (0,1) a b (0,1) a b i .
Uniqueness of representation. Let z a b i and z a b i , then
a b i a b i a a (b b) i , if b b , then a a and the representation is unique.
aa
Let b b , then a a (b b) i i there is a quotient of two real numbers on the right
b b
side, therefore the number on the right side belongs to R, then i R . But we have the equality
i 2 1 , which is impossible in R . Really, R is an ordered field, and for any element r R there
51
Complex Numbers
must be r 2 0 , so we have a contradiction, this contradiction came from the assumption b b .
Then b b and the representation is unique.
The assertion3 has a great practical importance. As any complex number z can be uniquely written
as a b i , this form of complex numbers is used in practice. Because it allows us to perform
addition/subtraction/multiplication/division of complex numbers much faster. Really, the field of
complex numbers С includes real numbers and pairs of real numbers, and in order to add or
multiply these elements we should initially refer to the field Сaux , so it’s not a fast and a comfortable
way to operate in С. The notation a b i allows us to perform all arithmetical operations right in
the field С , without referring to Сaux . Now we can discard Сaux , it was an intermediate auxiliary
field, and we don’t need it anymore.
Example1. [A] (5 2i) (6 7i) 11 9i and
(5 2i) (6 7i) 5 6 5 7i 2i 6 2i 7i 30 47i 14i 2 30 47i 14 16 47i .
[B] (2 i)3 (2 i) (2 i) (2 i) (4 4i i 2 ) (2 i) (3 4i) (2 i) 6 3i 8i 4i 2 2 11i .
1 i (1 i) (1 i) 1 i i i 2 2i
[C] i.
1 i (1 i) (1 i) 1 i2 2
Two complex numbers a bi and c di are equal a c and b d .
(it immediately follows from the uniqueness of representation).
Def. For any complex number z a bi the real number a is called a real part of z and b is
called an imaginary part of z. And we can write a Re z and b Im z .
Assertion4. The field С is not an ordered field, i.e., there is no way to introduce any order relation
"" on С.
Auxiliary2. The ring of integer numbers can be ordered in the only one way, i.e., the order on
, where all natural numbers are positive is the only possible order on .
Proof. Let’s assume that is ordered in some other way, according to the simpliest properties of
ordered rings/fields: a || a 0 a 2 0 , let’s take 1 12 1 0 so the number 1
is positive, any natural number can be represented as a sum of ones, then n we have
n 1 1 .... 1 0 , then any natural number in must be positive. In any ordered ring,
the element a , which is opposite to some positive element a , must be negative, then every
number from is negative. The only number that we haven’t considered yet is 0 ,
but a zero element is not positive or negative in any ordered ring/field. We have deduced that only
natural numbers are positive in , therefore our order relation on is exactly the same
as a standard order relation on .
Let’s assume now that С is an ordered field, there exist some order relation "" on С.
From the exercise6 (chapter “Groups, rings, fields” – Book I) follows that any subring/subfield of С
is also an ordered ring/field. In particular, C is an ordered ring, but the order relation on
52
Complex Numbers
is unique. We have 1 0 (in ) 1 0 (in C ) (exercise6), then 1 0 (in C ) .
If С is an ordered field, then for any z C there must be z 2 0 , in particular for i C there must
be i 2 0 , but i 2 1 0 , and we have a contradiction. So С is not an ordered field.
We know that in any ordered ring/field an absolute value (or module) | a | of any element a is
defined, this definition is made based on some order relation "" . By using the notion of absolute
value we defined a limit of a sequence and exlored it’s properties . All this theory can’t be applied
toС, because С is not an ordered field, so the old definition of | a | can’t be introduced in С.
Anyway, it’s very important to introduce the similar notion in С.
Def. For any complex number z a bi , the module (or absolute value) of z is | z | a 2 b2 .
Exercise1. Show that [A] z C | z | 0 and | z | 0 z 0 ,
[B] z1, z2 C | z1 z2 | | z1 | | z2 | and | z1 z2 || z1 | | z2 | .
Def. For any complex number z a bi , the complex number z a bi is called a conjugate
number.
Exercise2. Show that [A] z C z z | z |2 [B] | z || z |
[C] z1, z2 C z1 z2 z1 z2 || z1 z2 z1 z2 || z1 / z2 z1 z2 .
Geometrical representation of complex numbers
Any complex number z a bi can be depicted as a radius vector
with coordinates (a, b) [pict1]. The axis Ox is called a real
axis and Oy is called an imaginary axis. Such representation
of complex numbers (as radius vectors) has great advantages.
Addition of complex numbers can be represented as addition
of radius vectors which depict these numbers [pict2].
A module | z | of any complex number z can be understanded
pict.1
as a length of a radius vector which depicts z.
For any complex number,
which is depicted as a radius
vector, the conjugate complex
number can be depicted as a
symmetrical radius vector
[pict3] (with respect to Ox ).
Other advantages of
geometrical representation
will be described later,
when we introduce
the next theory.
pict.2 53 pict.3
Complex Numbers
Assertion5. For any non-zero complex number z a bi there exist the unique angle
[0o ,360o ) such that z | z | (cos i sin ) . The angle is called an argument of z, and
we can denote arg z .
Auxiliary3. For any pair of real numbers , such that 2 2 1 there exist the angle
[0o ,360o ) such that cos , sin .
Proof. Existence. Let’s take some coordinate system
Oxy . We draw the unit circle and the lines x and
y [pict4], these lines intersect at the point B with
coordinates ( , ) , as
2 2 1 , then B lies on the unit
circle. The angle AOB , which is counted from OA
in the counterclockwise direction, is obviously the angle
we need.
Let’s fix now any non-zero complex number z С ,
a b pict.4
then z a bi a 2 b 2 i .
a b a 2 b2
2 2
a b
For numbers , we obviously have
a b
2 2
a b2 2
2 2
a b
1, then (auxiliary3) there exist [0o ,360o ) such that
a b a b
2 2 2 2
a b
cos , sin , then z a 2 b2 (cos i sin ) | z | (cos i sin ) -
a 2 b2 a 2 b2
the representation we need.
Let’s show that such angle is unique. Let’s assume that there exist some other angle
[0o ,360o ) , , for which we also have z | z | (cos i sin ) . Then
| z | (cos i sin ) | z | (cos i sin ) cos i sin cos i sin , as complex numbers
are equal, their real and imaginary parts are equal, then cos cos and sin sin .
As we have , then 0o , then cos( ) 1.
But cos( ) cos cos sin sin cos sin 2 1, we have a contradiction.
2
Then is unique.
Def. Quite often we use the letter to denote | z | , so | z | and, as we proved above, for any
non-zero complex number z there exist the unique representation z (cos i sin ) [T],
where [0o ,360o ) . [T] is called a trigonometric representation of z, or a trigonometric form of z.
54
Complex Numbers
Assertion6. When some non-zero complex number z a bi is depicted as a radius vector with
coordinates (a, b) , the angle from the trigonometric form [T] is the angle between and Ox ,
which is counted from Ox in the counterclockwise direction.
Proof. From [T] follows that z a bi (cos i sin ) a cos and b sin .
The radius vector has coordinates ( cos , sin )
[pict5]. Let’s draw the unit circle and build the angle
AOB , which is counted in the counterclockwise
direction from Ox . Then OB has coordinates
(cos , sin ) , then OB must have the same
coordinates ( cos , sin ) as . Then OB
coincides with . (really, as coordinates of these vectors
are equal, vectors are equal, and we have two equal
vectors with the common start point O , so these vectors
must coincide). And OB for sure forms the angle
pict.5
with Ox , then the same is true for .
Comment. The number 0 C can be also written in a trigonometric form. Let’s find this form.
Suppose we have a representation 0 (cos i sin ) . Obviously, 0 0 0 i , then
(cos i sin ) 0 0 i . When complex numbers are equal, their real and imaginary parts
are equal, then cos 0, sin 0 . Then
( cos ) 2 0, ( sin ) 2 0 2 cos2 2 sin 2 0 2 (cos2 sin 2 ) 0
2 1 0 0 . So, in the representation 0 (cos i sin ) there must be 0 .
From here immediately follows that any angle is appropriate. So, let’s make an agreement
that o o
is any angle from [0 ,360 ) .
Assertion7. z1 1 (cos1 i sin 1 ) and z2 2 (cos2 i sin 2 ) are two non-zero complex
numbers. Then z1 z2 12 cos(1 2 ) i sin(1 2 ) and
z1 1
(cos(1 2 ) i sin(1 2 )) .
z2 2
And the geometrical meaning is: when we multiply two complex numbers, we must multiply their
modules and add their arguments. When we divide two complex numbers, we must divide their
modules and subtract one argument from another.
Proof. z1 z2 12 (cos1 i sin 1 ) (cos2 i sin 2 )
1 2 cos1 cos2 sin 1 sin 2 icos1 sin 2 sin 1 cos2
1 2 cos(1 2 ) i sin(1 2 ). By the way, from here immediately follows that for any
complex number z (cos i sin ) and for any natural number n we have
z n n (cosn i sin n ) [De Moivre's formula].
55
Complex Numbers
z1 1 (cos1 i sin 1 ) (cos1 i sin 1 ) (cos2 i sin 2 )
Next, 1
z2 2 (cos2 i sin 2 ) 2 (cos2 i sin 2 ) (cos2 i sin 2 )
1 cos1 cos2 sin 1 sin 2 isin 1 cos2 cos1 sin 2 1
(cos(1 2 ) i sin(1 2 )) .
2 cos2 2 sin 2 2 2
Example2. Calculate (cos 72 o i sin 72 o ) 40 . Solution: Let’s denote
z cos 72 o i sin 72 o 1 (cos 72 o i sin 72 o ) , then
(cos 72 o i sin 72 o ) 40 z 40 140 (cos(40 72 o ) i sin( 40 72 o )) cos(8 360 o ) i sin( 8 360 o ) 1 .
Example3. Calculate 1 3i
2018
. Solution: Let’s write this number in a trigonometric form
and use [De Moivre's formula] (cos i sin )n n (cosn i sin n ) .
We have the number 1 3i
2018
.
Let’s represent 1 3i (cos i sin ) , there must be 12 ( 3) 2 1 3 2 and
1 3 1 3
cos , sin cos , sin 60o , then 1 3i 2(cos60o i sin 60o ) .
2 2 2 2
So (1 3i)
2018
2(cos60o i sin 60o ) 2018
22018 cos(2018 60o ) i sin( 2018 60o ) .
22018 cos(336 360 o 120 o ) i sin( 336 360 o 120 o ) 22018 cos(120 ) i cos(120 )
o o
1 3
22018 i 22017 1 3i .
2 2
100
1
Example4. Calculate .
i 1
1 i 1 i 1 1 1 1
1 i 2
1 1
Solution: i (cos 45o i sin 45o ) .
i 1 (1 i)(1 i) 2 2 2 2 2 2
100 100
1 1
Then cos(100 45 ) i sin(100 45 )
o o
i 1 2
1 1 1
50 (cos(12 360o 180o ) i sin(12 360o 180o )) 50 (cos(180o ) i sin(180o )) 50 .
2 2 2
Example5. z1 , z2 .... zn are complex numbers such that | z1 || z2 | ... | zn | 1 .
n n 1
And z zk
. Show that z is a real number and 0 z n2 .
k 1 k 1 zk
1
Solution. For any zk we have | z k | 1 z k z k 1 zk .
zk
56
Complex Numbers
n
n 1
n n
Therefore z k
z z k z k ( z1 z 2 ... z n ) ( z1 z 2 ... z n )
k 1 k 1 z k
k 1 k 1
( z1 z2 ... zn ) ( z1 z2 ... zn ) z1 z2 ... zn - it is a non-negative real number, then
z z1 z2 ... zn 0 . Also z z1 z2 ... zn z1 z2 ... zn n . And we have 0 z n .
Example6. z1, z2 .... zn are complex numbers such that | z1 || z2 | ... | zn | 1 . Show that
n n
1
zk z .
k 1 k 1 k
n n
1 1
Solution. For any zk we have | zk | 1 zk zk 1 zk
zk
. Therefore zk z ,
k 1 k 1 k
z z 1 1 1 n
as z, h C , then
h h
, then
zk zk zk
zk
k 1
1n n 1 1 1 1 1 1 1 | z || z |
z ... ...
k 1 zk k 1 k z1 z2 zn z1 z2 zn z C
1 1 1 n
1
... z , everything is proved.
z1 z2 zn k 1 k
Assertion8. There are no zero divisors in C if z1 z2 0 in C, then z1 0 or z2 0 .
Comment. We have shown earlier that there are no zero divisors in any ordered ring/field.
That’s why we didn’t need to check the similar assertion for , Q, R .
Proof. Let’s assume that our assertion is not true, it means that there exist some non-zero numbers
z1, z2 such that z1 z2 0 . Any non-zero complex number can be uniquely represented in
a trigonometric form. So z1 1 (cos1 i sin 1 ) and z2 2 (cos2 i sin 2 ) , then
z1 z2 12 cos(1 2 ) i sin(1 2 ) 0 . From here it’s very easy to get the equality
12 0 . Then 1 0 or 2 0 , it means that z1 0 or z2 0 , and we have a contradiction.
So, there are no zero divisors in C.
Def. Let z C is any complex number and n . If there exist some C such that ( )n z
then is called an n th root of z , and we denote it n z .
Assertion9. For 0 C and for any n we have n
0 0 . And for any non-zero complex number
z C and any n there exist exactly n different complex numbers 0 , 1... n 1 C such that
(0 )n z, (1 )n z .... (n 1 )n z (the symbol n
z has n different values 0 , 1... n 1 ).
57
Complex Numbers
Proof. Let 0 C . Let’s fix any n , let ( ) 0 ... 0 . From the assertion8 follows
n
that one of the factors ... is zero, then 0 .
Let z C is any non-zero complex number. Then it can be uniquely represented as
z (cos i sin ) . Let there exist some complex number C such that n z .
If 0 , then z 0 , which is not true, then 0 and can be uniquely represented
as l (cos i sin ) . From the equality n z and [De Moivre's formula] follows that
l n (cos n i sin n ) (cos i sin ) -these complex numbers are equal, therefore their real and
imaginary parts are equal: l n cos n cos and l n sin n sin [M] from here follows
l n
cos n
2 2
cos and l n sin n
2
sin l 2n cos2 n 2 cos2 and
2
l 2n sin 2 n 2 sin 2 , let’s add these equalities
l 2n (cos2 n sin 2 n ) 2 (cos2 sin 2 ) l 2n 2 , as , l are positive real numbers,
from l
2n
2 follows that l n , then l n . Let’s return to [M], so (n )n cos n cos and
(n )n sin n sin , then cos n cos and sin n sin [V].
Here is a fixed angle, and is the angle we need to find. From [V] we see that cosines and sines
of n and are equal, then these angles define the same point B on the unit circle. We have
the restrictions: [0o ,360o ) and [0o ,360o ) . Then n must belong to one of the next sets
[0o , 360o ), [360o , 360o 2), [360o 2, 360o 3) ... [360o (n 1), 360o n) , and belongs to [0o ,360o ) .
As n and define the same point on the unit circle, only the next variants are possible:
n , n 360o , n 360o 2 , n 360o 3 ...... n 360o (n 1) it is
360o 360o 2 360o 3 360o (n 1)
equivalent to , , , ...... .
n n n n n
So, we have found l n and we have also found all the possible values of , they define exactly
n different complex numbers:
360o k 360o k
k n cos i sin || k 0,1, 2 ... ( n 1) .
n n
Now we need to check that all these numbers k really satisfy the equation n z , it’s very easy
to do by using the assertion7. Everything is proved.
During the proof we have deduced a very important formula, the explicit value of n-th root.
58
Complex Numbers
If z (cos i sin ) , then
n n cos 360 k i sin 360 k || k 0,1,2 ... ( n 1) .
o o
z
n n
And the geometrical meaning is: for any non-zero
complex number z all the values of n z can be depicted pict.6
as radius vectors, which end points are vertexes of some
regular polygon with n sides [pict6], the center of such
polygon is always located at the origin.
Example7. Calculate 4 256 .
Solution. We will use the formula
360o k 360o k
n
z n | z | cos i sin || k 0,1,2... n 1.
n n
In our case z 256 and n 4 , so n
4 | 256 | 4 .
Here arg z 0 , then we can write:
360o k 360o k
n
256 4 cos i sin || k 0,1,2,3
4 4
360o 0 360o 0
For k 0 4 cos i sin 4,
4 4
360o 1 360o 1
for k 1 4 cos i sin 4i ,
4 4
360o 2 360o 2
for k 2 2 cos i sin 4 ,
4 4
360o 3 360o 3
for k 3 2 cos i sin 4i .
4 4
Then 4
256 has four different values: 4, 4i, 4, 4i .
For Reader’s practice:
6
2k 2k
[1] Calculate sin
k 1 7
i cos
7
Answer: i .
334 365
1 i 3 1 i 3
[2] Calculate 4 5
3
Answer: 3i .
2 2 2 2
5 5
3 i 3 i
[3] Show that the number is purely real.
2 2 2 2
[4] Solve the equation z 2 | z | 0 (Hint, use the representation z x iy || x, y R ).
Answer: z 0, z i, z i .
59
Complex Numbers
[5] [Advanced] How many solutions does the equation z 3 z z have? Answer: four solutions.
5
1 3 27 10 3
[6] [Advanced] Solve the equation z 6 z z . Answer: 0, 5 2 , i , i.
2 2 2 2
Def. For any complex number z0 and any 0 the set of complex numbers
( z0 ) {z || | z z0 | } is called an -neighborhood of z 0 .
Any sequence {z n } of complex numbers uniquely defines two sequnces of real numbers {xn } and
{ yn } , where zn xn iyn and conversely. The point z0 is called a limit of the sequence {zn }
if any neighborhood of z 0 contains all the terms of {zn } , starting from some number.
Exercise3. Show that {zn } z0 (where zn xn iyn and z0 a bi ) {xn } a and { yn } b .
In other words, a sequence {zn } goes to z0 if and only if the sequence of real parts of {zn } goes to
the real part of z0 and the sequence of imaginary parts of {zn } goes to the imaginary part of z0 .
A sequence of complex numbers {z n } is called fundamental, if for any (small) 0 there exist k
such that m, n k zm zn .
Exercise4. Show that: {zn } is fundamental (where zn xn iyn ) both {xn } and { yn } are
fundamental.
Assertion10. The field of complex numbers С is a complete field: any fundamental sequence
{zn } С converges to some limit z0 С .
Proof. Let’s fix an arbitrary fundamental sequence {zn } , we have zn xn iyn || n , then
(exercise4) {xn } and { yn } are both fundamental, as R is a complete field, both these sequences
converge: {xn } a, { yn } b , then (exercise3) {zn } converges to a bi .
60
Analysis
Limits
Earlier we have defined 3 types of points: internal points, boundary points, external points.
If some set R is fixed, then any point a R is an internal point of X,
or a boundary point of X, or an external point of X.
It’s very important to define the next 2 types of points: limit points and isolated points.
Def. Let R is a set. And a R is some real number (any real number can be called a point).
If there exist some sequence {xn } || xn a n which goes to a, then a is called a limit point
~ and there is no any sequence {x } || x a~ n which goes to a~ , then a~ is
of . If a n n
called an isolated point of X [pict1].
A limit point of X mustn’t belong to X.
For example, for any half interval (a, b] ,
a is a limit point and it does not belong to (a, b] .
In the same time for any segment [a, b] ,
a is a limit point, and now it belongs to [a, b] .
Any isolated point of X belongs to X
pict.1
by definition.
Exercise1. For any set R , any point a is a limit point of X or an isolated point of X,
and no other variants.
Exercise2. Any internal point of X is a limit point of X. A boundary point of X may be a limit
point of X and may not be a limit point of X. An external point of X is never a limit point of X.
In order to define the first 3 types of points (internal, boundary, external) we used the notion
of a neighborhood (a) . For the new 2 types of points it’s much more convenient to use the notion
of a deleted neighborhood D (a) , in fact it is a neighborhood (a) where the central point a is
deleted.
Def. For any a R and any 0 , the set of numbers D (a) (a) \ {a} {x || 0 | x a | }
[pict2] is called an -deleted neighborhood of a , or just a deleted neighborhood of a .
The set D (a) can be depicted on the line as an interval without it’s middle point.
Exercise3. Show that: [pict2]
[A] a is an isolated point of X there exist D (a) which
does not contain any points of X.
[B] a is a limit point of X any deleted neighborhood
D (a) contains at least one point of X. pict.2
62
Analysis
Def [A]. X is a set of real numbers, f (x) is defined on X. And a is a limit point of X
( a may belong to X and may not belong to X, therefore f may be defined at a and may be not
defined at a). If for any sequence {xn } such that {xn } a || xn a n , the sequence
{ f ( xn )} always goes to some fixed number , then is called a limit of f at the point a,
and we write lim x a|| x f ( x) .
Once again. If for any {xn } || xn a (n) || {xn } a we have { f ( xn )} , then is a limit
of f at a, and we write lim x a|| x f ( x) .
Notice: from this definition immediately follows that f may have only one limit at any point a,
because the sequence { f ( xn )} (as any other sequence) can’t have two different limits A B.
There is another equivalent definition.
Def [B]. X is a set of real numbers and f (x) is defined on X. Let a is a limit point of X, and for
any (small) positive 0 there exist some positive 0 , such that for any x : 0 x a
we have f (x) . Then is called a limit of f at a, and we write lim x a|| x f ( x) .
Once again. If for any (small) positive 0 there exist some positive 0 such that for any
x D (a) f ( x) , then lim x a|| x f ( x) .
Notice: the requirement “a is a limit point of X ” is important, it guarantees that for any 0
the set D (a) is not empty and it contains some points x , for which the condition
f (x) must be checked.
We will show that this definition is equivalent to initial
one, but let’s explain at first it’s meaning.
The meaning. Let lim x a|| x f ( x) , according to
the Def [B]. [pict3]
Let’s take a very small positive , for example
0,000001, then there exist such neighborhood D (a) ,
pict.3
that for any point x D (a) the value f (x) differs
from so slightly, that the “distance” between f (x) and is less than 0.000001 .
X everywhere near the point a , every value f (x) is extremely close to .
So, in the set
We can also say: when x approaches to a , the value f (x) approaches “infinitely close” to .
Really, let x approaches toa, then at some moment x will appear inside any fixed
neighborhood D (a) , and when it happens, right away the value f (x) is so close to , that their
difference is less than , where can be fixed from the very beginning as a very small positive
number.
63
Analysis
An important comment. In the limit definition (Def [A] or Def [B]) we do not allow x to reach
the value a , we allow x to approach “infinitely close” to a , and we observe the value f (x) , does it
approach infinitely close to some fixed number, or not. But once and for all: we forbid x to coincide
with a , we allow x to approach infinitely close to a (and such approach is possible, because a is
a limit point of X , and there exist x so close to a as we want), but x can’t coincide with a .
There are a lot of reasons for such restriction, without it we couldn’t define a derivative of a function
(it is one of the most important notions in mathematics). Let’s explain briefly.
f ( x) f (a)
A derivative is defined as a limit of the function at the point a , and the function
xa
f ( x) f (a)
is not even defined at a . This function will be always defined on some deleted
xa
0
neighborhood of a . But if x reaches the value a , there appears an expression , and such
0
f ( x) f (a)
expression is not defined. So is not even defined at a , but it’s limit at a is one of
xa
the central notions in mathematics. That’s why the limit definition is made in such way that x
never reaches the value a .
An important comment2. When we calculate lim x a|| x f ( x) the only values we need are
the values of f on . The function f may be defined at a , or may be not defined at a , it does not
affect the existence of the limit lim x a|| x f ( x) and it’s value. And similarly, outside of the set
(on R \ ) the function f may be not defined at all, or f may be defined at some points of R \
and reach any values at these points, it does not affect the existence of lim x a|| x f ( x) and it’s
value.
Def. X is a set of real numbers and f is defined on X. Let’s fix any coordinate system Oxy .
The set of points with coordinates {(x, f ( x)) || x } is called a graph of f on X, or just
a graph of f [pict4].
pict.4
64
Analysis
Geometrical meaning of a limit. f has a limit
at a point a if for any small positive there exist
some positive , such that the graph of f on the set
(a) lies completely inside the strip which is
bounded by the lines y , y [pict5].
Let’s show that definitions Def [A] and Def [B] are
equivalent. Then in any concrete case we will be
able to choose which definition to use.
Both definitions are really important, but the initial pict.5
one Def [A] is very convenient for proving different
properties of limits.
Proof. Let Def [A] is true. We want to show that for any positive we can find some positive
such that x D (a) f ( x) . Let’s assume that it is not true, it means that there
exist some concrete positive , for which we can’t find any appropriate . Therefore, for any 0
there is always at least one “bad point” x in the set D (a) , for which we have f (x ) .
Let’s fix any decreasing sequence of positive numbers 1 2 3 4 .... 0 which goes to zero
{ n } 0 , then we have the sequence of nested neighborhoods
D1 (a) D 2 (a) D 3 (a) D 4 (a) ..... As we noticed above, in any neighborhood D n (a) there
exist some “bad point” xn D n (a) such that f ( xn ) [T]. Then we have the sequence
{xn } and {xn } a || xn a n . According to the Def [A], for the sequence { f ( xn )} there
must be { f ( xn )} , but we have f ( xn ) || n , it means that -neighborhood of does
not contain any terms of the sequence { f ( xn )} , therefore is not a limit of the sequence { f ( xn )} .
We have a contradiction. It proves that from Def [A] follows Def [B].
Conversely Let Def [B] is true. We want to show that for any sequence
{xn } || {xn } a, xn a (n) : there must be { f ( xn )} .
Let’s fix any sequence {xn } || {xn } a, xn a (n) and any positive 0,
according to the Def [B], there exist 0 such that x D (a) f ( x) .
As {xn } goes to a and xn a (n) , the set D (a) must contain all the terms of {xn } ,
starting from some number k . And for any xn D (a) there must be
f ( xn ) f ( xn ) () . It means that neighborhood of contains all the terms of
{ f ( xn )} , starting from the number k , then { f ( xn )} . Everything is proved.
65
Analysis
Sign conservation. Let lim x a|| x f ( x) . If 0, then there exist the deleted
neighborhood D (a) such that x D (a) the value f ( x) 0 . If 0 , then there exist the
deleted neighborhood D (a) such that x D (a) the value f ( x) 0 .
Put simply, if a limit of f at A is positive/negative, then all the values of f near A are also
positive/negative.
Proof. Let 0 , let’s fix for example / 2 0 , there exist D (a) such that x D (a)
we have f ( x) / 2 / 2 f ( x) / 2 / 2 f ( x) 3 / 2 , so the value f (x)
is “squeezed” between two positive numbers and therefore f (x) is positive (for any x D (a) ).
Similarly, when A is a negative real number we can take / 2 0.
Theorem1. Both limits lim x a|| x f ( x) and lim x a|| x g ( x) exist, then
[1] lim x a|| x ( f ( x) g ( x)) [2] lim x a|| x ( f ( x) g ( x))
f ( x)
[3] lim x a|| x (if g ( x) 0 on and 0 ) [4] lim x a|| x f ( x) for any real
g ( x)
number R [5] lim x a|| x f ( x)k k for any k .
An important comment. Notice that we require the existence of both limits lim x a| x f ( x) and
lim x a| x g ( x) , only after that we can guarantee that all the other limits [1]-[5] exist, but NOT
conversely. We can’t use any of the formulas [1]-[5] if we don’t know for sure that both limits
lim x a| x f ( x) and lim x a| x g ( x) do exist (there was a similar comment for sequence-limits).
Proof. The proof is very simple if we use the Def [A]. We have proved earlier quite similar theorems
for sequences Basic properties of convergent sequences (Book1): if {xn } a and { yn } b , then
x a
[1.0] {xn yn } a b , [2.0] {xn yn } a b , [3.0] n ( if yn 0 n and b 0 ).
yn b
Let’s fix an arbitrary sequence {xn } || xn a (n) || {xn } a from the condition
lim x a| x f ( x) follows that { f ( xn )} and from the condition lim x a| x f ( x)
follows that {g ( xn )} . Then from [1.0], [2.0], [3.0] we have { f ( xn ) g ( xn )} and
f ( xn )
{ f ( xn ) g ( xn )} and . From here follows that [1] [2] [3] are true.
g ( x n
)
Next, [4] immediately follows from [2] if we take a constant function g (x) for any x .
And [5] can be easily derived from [2] (show how).
66
Analysis
Squeeze theorem for functions. Functions f , g , h are defined on X and:
lim x a|| x f ( x) and lim x a|| x h( x) and there exist some deleted neighborhood D (a)
such that x DR (a) : f ( x) h( x) g ( x) , then also lim x a|| x h( x) .
Proof. Let’s fix an arbitrary sequence {xn } : {xn } a . From lim x a| x f ( x) and
lim x a| x g ( x) follows that { f ( xn )} and {g ( xn )} , then f ( xn ) h( xn ) g ( xn )
and from the squeeze theorem for sequences we have {h( xn )} . Everything is proved.
Exercise1 (Set replacement). lim x a|| x f ( x) . Then for any set , for which a is also
a limit point, there must be lim x a|| x f ( x) .
We have already provided the definition of a continuous function (it was done for the practical
purposes in the book1, we needed it to understand for which aR the value k
a exists).
Def [A1]. f is defined on X and a . If a is a limit point of X and for any sequence
{xn } || xn a n || {xn } a we have { f ( xn )} f (a) , then we say “ f is continuous at a ”.
If a is an isolated point of X, then f is continuous at a (by definition).
Notice, as any point a is a limit point of X or an isolated point of X, then for any point a
it’s possible to check, is f continuous at a or not.
When a is a limit point of X, continuity at a means that
[A] The limit lim xa|| x f ( x) exists [B] This limit is equal to f (a) .
So, continuity is just a particular case of a limit-existence. But now f must be defined at a.
By using the limit definition Def [B] we can write an equivalent definition.
Def [B1]. f is defined on X and a . If a is a limit point of X and
0 0 || x : 0 | x a | f ( x) f (a) , then “ f is continuous at a ”.
If a is an isolated point of X, then f is continuous at a (by definition).
This definition can be simplified, we can change the requirement 0 | x a | to
| x a | x (a) , really, by doing it we just include the case x a , but then
f ( x) f (a) f (a) f (a) 0 . So, when a is a limit point of X, we have the tantamount
requirement 0 0 || x (a) f ( x) f (a) [R]. Notice that the requirement
[R] is obviously true for any isolated point a , really as a is an isolated point, there exist some
(a) such that (a) has no common points with except a.
67
Analysis
Then for any positive 0 , we will always take 0 and then the only x (a) is x a,
for which the condition f ( x) f (a) is obviously true.
So, we can seriously abbreviate our definition Def [B1].
The main definition of a continuous function [B]. f is defined on X and a .
If 0 0 || x (a) f ( x) f (a) , then “ f is continuous at a ”.
(Show that this definition is (logically) equivalent to the Def [B1], we have just shown that it follows
from Def [B1], and it’s time to prove conversely). We can also abbreviate Def [A1] by throwing away
the requirement xn a (n) and the phrase about an isolated point a.
(Show that this definition is equivalent to Def [A1]).
The main definition of a continuous function [A]. f is defined on X and a .
If {xn } || {xn } a { f ( xn )} f (a) , then “ f is continuous at a ”.
By using the main definition of a continuous function [B] it’s easy to prove the next theorem.
Theorem2. f , g are defined on X and a . Both functions f , g are continuous at a , then:
[1] f ( x) g ( x) is continuous at a [2] f ( x) g ( x) is continuous at a
f ( x)
[3] is continuous at a (if g ( x) 0 x )
g ( x)
[4] for any constant R the function f (x) is continuous at a ,
[5] for any natural number k the function f ( x) f k ( x) is continuous at a .
k
Def. f is defined on X. We say that f is continuous on X if f is continuous at every point
a.
Obviously, if f continuous on X, then f continuous on any subset .
Consequence from theorem2. f , g and are continuous on X, then
f g || f g || f / g (if g 0 on ) || f ( x) ( R) || f k ( x) (k ) are continuous on X.
Composite function. Some function f is defined on X.
let’s consider the set of all values { f ( x) || x } f ( )
[pict6]. Let is any set which contains the set of all the
values f () , and a function g ( y ) is defined on . Then
for any number x the number g f (x) is defined, and
we have the function g f (x) of argument x which is
defined on X, this function is called a composite function. pict.6
68
Analysis
Theorem3 [Continuity of a composite function]. f (x) is defined on X and g ( y ) is defined on
f () . If f (x) is continuous at a and g ( y ) is continuous at f (a) , then g f (x) is
continuous at a .
Proof. Let’s use the main definition of a continuous function [B] for g f (x) . We fix an arbitrary
sequence {xn } || {xn } a , if we show that {g ( f ( xn ))} g ( f (a)) , then g f (x) is continuous
at a. As f (x) is continuous at a, then from {xn } || {xn } a follows that { f ( xn )} f (a) .
The sequence { f ( xn )} is a sequence of which goes to f (a) , and g ( y ) is continuous at f (a) ,
then there must be g f ( xn ) g f (a) . Everything is proved.
Consequence1. f (x) is defined on , and g ( y) is defined on f () .
If f (x) is continuous on , and g ( y ) is continuous on , then g f (x) is continuous on X.
Improvement of the Theorem3 [Limit of a composite function].
lim x a|| x f ( x) and g ( y) is defined on , where f () and .
If g ( y ) is continuous at , then lim x a|| x g f ( x) g ( ) .
Exercise4. Show that the improvement of the theorem3 is true.
We have provided above the most general and the most important limit-definition:
lim x a|| x f ( x) .
Ordinary limit. Let now X is some deleted neighborhood of a, so DR (a) .
The Reader can easily check that the equivalent limit definitions Def [A] and Def [B] turn into:
Def [ordinary limit A]. If for any sequence {xn } DR (a) || {xn } a the sequence { f ( xn )} ,
then we write lim x a f ( x) .
Def [ordinary limit B]. If for any (small) 0 there exist : 0 R such that for any
x D (a) we have f (x) , then we write lim x a f ( x) .
Left and right limits. Here are two similar definitions. Let f is defined on some
“left neighborhood” (a R, a) of a (for the left limit). Or f is defined on some “right neighborhood”
(a, a R) of a (for the right limit).
Def [left limit A]. If for any sequence {xn } (a R, a) || {xn } a the sequence { f ( xn )} ,
then we write lim x a f ( x) and we say that A is a left limit of f (x) at the point a.
[right limit A]. If for any sequence {xn } (a, a R) || {xn } a the sequence { f ( xn )} ,
then we write lim x a f ( x) and we say that A is a right limit of f (x) at the point a.
69
Analysis
[left limit B]. If for any (small) 0 there exist : 0 R such that for any x (a , a)
we have f (x) , then lim x a f ( x) .
[right limit B]. If for any (small) 0 there exist : 0 R such that for any x (a, a )
we have f (x) , then lim x a f ( x) .
Exercise5. Both left and right limits at a exist and both limits are equal to A The ordinary
limit at a exists and it is equal to A. Shortly: lim x a f ( x) lim x a f ( x) lim x a f ( x)
Ordinary continuity. f is defined on some neighborhood R (a) .
Def [A]. If for any sequence {xn } R (a) || {xn } a the sequence { f ( xn )} f (a) ,
then f is continuous at a.
Def [B]. If for any (small) 0 there exist : 0 R such that for any x (a)
we have f ( x) f (a) , then f is continuous at a.
The left\right continuity requires from f to be defined on (a R, a] or on [a, a R) .
And in these cases we say “ f is left-continuous at a “ or “ f is right-continuous at a”.
From the exercise5 follows: f is left and right continuous at a f is continuous at a.
Def. Let {xn } is some sequence and n1 n2 n3 n4 .... is any increasing sequence
of natural numbers, then the sequence xn1 , xn2 , xn3 , xn4 .... {xnk } is called a subsequence of {xn } .
Exercise6. {xn } has a limit a, then any it’s subsequence {xnk } {xn } has the same limit a.
The converse assertion is obviously not true, if some subsequence {xnk } a , the sequence {xn }
mustn’t converge to a.
Lemma1. Any bounded sequence has a convergent subsequence.
Proof. Let {xn } is a bounded sequence. Then xn C for any n C xn C . Let’s divide the
segment [C , C ] into two equal parts: [C ,0] and [0, C ] . At least one of these segments
(for example [0, C ] ) contains infinitely many terms of {xk } , there must be the term with the least
number n1 , we fix it xn1 [0, C ] it is the first term of our subsequence.
Let’s divide now the segment [0, C ] into two equal parts: [0, C / 2] and [C / 2, C ] . At least one of these
parts contains infinitely many terms {xk } (if both parts contain infinitely many terms, then we
choose the right one), let it be [C / 2, C ] . There must be the term with the least number n2 n1 ,
we fix it xn2 [C / 2, C ] and we continue the process. We will get the sequence of nested segments,
each next segment is a half of a previous one, therefore there exist exactly one point
h, which is
common to all these segments. Obviously h [C, C ] . And we also have the subsequence {xnk } ,
let’s show that {xnk } h .
70
Analysis
Let’s fix an arbitrary positive 0 and consider the neighborhood (h) , as h is a common
point of all segments, therefore (h) contains all the segments, starting from some number k
(really, if (h) does not fully contain some segment such that h , then the length of the
segment is not less than / 2 0 , it can’t be true for all segments, because their lengths go to
zero ). Then (h) contains the segments with numbers k 1 , k 2 , k 3 .... Every segment with
number m contains the term xnm of our subsequence, then (h) contains all terms of {xnk } with
numbers k 1 , k 2 , k 3 .... And is an arbitrary small positive number, then {xnk } h .
Def[U]. f is defined on X, f is called uniformly continuous on X if for any 0 there exist
0 such that x1, x2 : x1 x2 f ( x1 ) f ( x2 ) .
Assertion1. Uniform continuity on X is “stronger” than the ordinary continuity.
If f is uniformly continuous on X, then f is continuous on , but not conversely.
Proof. Let f is uniformly continuous on X, let’s fix an arbitrary a .
And let’s take x2 a in the Def[U]. Then for any 0 there exist 0 such that
x1 : x1 a f ( x1 ) f (a) it means exactly that f is continuous at a .
Then f is continuous at every point a and f is continuous on X.
Comment. The converse assertion is not true. If f is continuous on X, then f mustn’t be uniformly
continuous on X.
Lemma2. A sequence {xn } [a, b] converges to some limit c , then c [a, b] .
Proof. If c [a, b] then there exist some neighborhood (c) which doesn’t have any common
points with[a, b] , this neighborhood contains all the terms of {xn } , starting from some number,
which is impossible, because {xn } [a, b] . This contradiction proves that c [a, b] .
Cantor’s theorem. f is continuous on [a, b] , then f is uniformly continuous on [a, b] .
Proof. Let’s assume that there is no uniform continuity, then there exist some “bad” 0 for which
it’s impossible to find any appropriate 0 , it means that for any 0 there exist some points
x, z such that x z , but anyway f ( x) f ( z ) . Let’s fix an arbitrary positive sequence
{ n } 0 , for every n there exist xn , zn [a, b] such that xn zn n and f ( xn ) f ( zn ) [Z].
All the elements of both sequences {xn } and {zn } belong to [a, b] , then both these sequences are
bounded, then (lemma1) each of these sequences has a convergent subsequence, let’s take just
one subsequence {xnk } c , this subsequence defines an increasing sequence of natural numbers
n1 n2 n3 n4 .... and this sequence defines the subsequences {znk } and { nk } . As { n } 0 ,
then the subsequence { nk } 0 . We also have xn zn n , from here follows that
xnk znk nk nk xnk znk nk xnk nk znk xnk nk , both outer sequences
71
Analysis
{xnk nk } and {xnk nk } converge to c (because {xnk } c and { n } 0 ),
then, according to the squeeze theorem for sequences, {znk } c .
So we have {xnk } c and {znk } c , both these sequences belong to [ a, b] ,
then (lemma2) с [a, b] .
As f is continuous at the point c, there must be { f ( xnk )} f (c) and { f ( znk )} f (c) .
As sequences { f ( xnk )} and { f ( znk )} go to the same limit, their difference { f ( xnk ) f ( znk )}
is an infinitely small sequence.
But from [Z] we have f ( xn ) f ( zn ) , it means that the absolute value of any term of
k k
the sequence { f ( xnk ) f ( znk )} is not less than , then neighborhood of 0 does not contain any
terms of this sequence. Then { f ( xnk ) f ( znk )} is not an infinitely small, and we have
a contradiction. This contradiction proves our theorem.
Def. A function f (x) is defined on X, f (x) is called bounded on X if there exist some С 0
such that f ( x) C || x .
1-st Weierstrass theorem. f (x) is continuous on [a, b] , then f is bounded on [a, b] .
Proof. Let’s assume that it is not true, then for any natural number n there exist at least one point
xn [a, b] such that f ( xn ) n . For n 1 there exist x1 [a, b] such that f ( x1 ) 1.
For n 2 there exist x2 [a, b] such that f ( x2 ) 2 and etc. Then we have the sequence {xn }
(notice that there we do not require from the terms of this sequence to be different numbers, so there
may be for example that x1 x2 , or x2 x3 and etc). All terms of {xn } belong to [a, b] and therefore
{xn } is bounded. According to the lemma1,{xn } has a convergent subsequence {xnk } h .
As {xnk } belongs to [a, b] then (lemma2) h [a, b] . As f is continuous at the point h , from
{xnk } h follows that { f ( xnk )} f (h) . So the sequence { f ( xnk )} converges, and in the same
time we have f ( xn1 ) n1, f ( xn2 ) n2 , f ( xn3 ) n3 ..., where n1 , n2 , n3 ... is an increasing
sequence of natural numbers. From here immediately follows that { f ( xnk )} is an unbounded
sequence. Then { f ( xnk )} does not converge to any limit (really, any sequence that converges to
some number f (h) must be bounded, we have proved it earlier). And we have a contradiction.
Everything is proved.
Let some function f is bounded on [a, b] . Then the set of all values { f ( x) || x [a, b]} is a
bounded set, therefore it has the least upper bound and a greatest lower bound m: so
m f (x) for any x [a, b] . But there is no guarantee that f reaches the numbers m,
~
(i.e., there may no be any x [a, b] such that f ( x ) m and there may no be any x [a, b] such
0 0 0
that that f ( ~
x0 ) ).
72
Analysis
Example [pict7]. Let’s consider the function f on [1,8] :
f (1) f (8) 2 and f ( x) x x (1,8) . Here m 1 and 8 ,
but f doesn’t reach any of these values on [1,8] .
There is only a guarantee that f (x) approaches arbitrary close to m
(such that f ( x) m ), for some values x [a, b] . And the closer f (x)
goes to m, the less the value f (x) we have.
And similarly, f (x) may not reach the value on [a, b] , but f (x)
pict.7
may approach arbitrary close to (such that f (x) ). And the
closer f (x) goes to , the greater the value f (x) we have.
So, for any function f , even if f is bounded on [a, b] , the notions “ the maximum value of f on
[a, b] ” and “ the minimum value of f on [a, b] ” are not correct. There may no be any
maximum/minimum value. But these notions are correct if f is continuous on [a, b] and the
2-nd Weierstrass theorem proves it.
2-nd Weierstrass theorem. f (x) is continuous on [a, b] , then f reaches on [a, b] it’s
maximum and it’s minimum m .
Proof. From the 1-st theorem follows that f is bounded on [a, b] . Then the set of all values
{ f ( x) || x [a, b]} is a bounded set, and it must have the greatest lower bound m and the least
upper bound .
[A] Let’s show that f reaches on [a, b] . We assume that there is no x0 [a, b] such that
1
f ( x0 ) and we consider the function g ( x) on [a, b] . The numerator 1 can be
f ( x)
considered as a constant function, such function is obviously continuous on [a, b] , the denominator
f (x) is also a continuous on [a, b] function (because f is continuous on [a, b] ).
1
So g ( x) is a quotient of two continuous functions on [a, b] . The function f (x) is
f ( x)
positive everywhere on [a, b] , and as f (x) may approach arbitrary close to , the difference
1
f (x) may be an arbitrary small positive number, then may be an arbitrary big
f ( x)
positive number. So g (x) is not bounded on [a, b] , and in the same time g (x) is continuous on
[a, b] , it contradicts to the 1-st Weierstrass theorem. This contradiction proves that f actually
reaches it’s maximum at some point of [a, b] . [B] In order to show that f reaches m on [a, b]
1
we can consider the function ( x) .
f ( x) m
73
Analysis
Def. Let f is bounded on [a, b] , the number m , where
sup[ a,b] f ( x) and m inf [ a,b] f ( x) is called an oscillation
of f on [a, b] . And we denote it: [ a,b] ( f ) m . [pict8].
Exercise1. f is bounded on [a, b] , and for any x1, x2 [a, b] we
have f ( x1 ) f ( x2 ) , where is some fixed positive number.
Then sup[ a,b] f ( x) inf [ a,b] f ( x) [ a,b] ( f ) . pict.8
Theorem4. Let is some figure on the plane, the boundary consists of several graphs of
continuous functions like y f ( x) || x [a, b] or x g ( y) || y [c, d ] [pict9].
Then is measurable.
Proof. It’s enough to show that is a zero area figure. pict.9
We will show that for any positive 0 the graph
y f ( x) || x [a, b] can be covered by several rectangles,
without common internal points which total area is not
greater than . All these rectangles together form
an external measurable figure which area is not greater
than . Then the graph of y f ( x) || x [a, b] is measurable
and it’s area is zero. Then the boundary of the initial
figure is a union of several zero area figures (several
graphs), then the boundary is also a zero area figure,
then is measurable. Let’s start.
We have y f ( x) || x [a, b] is a continuous function, then (Cantor’s theorem) f is uniformly
continuous on [a, b] . Let’s fix an arbitrary small positive 0 . For the positive number 0
ba
there exist 0 such that x1 , x2 [a, b] : x1 x2 f ( x1 ) f ( x2 ) .
ba
Let’s divide the initial segment [a, b] into several consecutive segments 1, 2 , 3.... m which go
one after another, and the length of each segment k is less than . Then for any points x1 , x2 k
we “automatically” have x1 x2 and therefore for any x1, x2 k we have
pict.10
f ( x1 ) f ( x2 ) . The function f is continuous on
ba
every segment k , then f is bounded on every k and the
numbers sup k f ( x) and inf k f ( x) are defined (supremum
and infinum of f on k ).
From x1 , x2 k f ( x1 ) f ( x2 ) (exercise1)
ba
74
Analysis
~
follows that sup f ( x) inf f ( x) [V ] . The graph of f on the segment k [c, d ] lies
k k
ba
inside the rectangle which is formed by the lines x c, x d , y inf k f ( x), y sup k f ( x)
[pict10]. The area of such rectangle is equal to
~
(d c) sup k f ( x) inf k f ( x) k sup k f ( x) inf k f ( x) [V ] k
ba
(here k is the length of k ).
So, all the graph y f ( x) || x [a, b] is covered by rectangles, every rectangle is built under some
segment k , the total sum of areas: (Total sum of areas ) 1 2 .... k
ba ba ba
1 2 ... k (b a) . Everything is proved.
ba (b a)
Def. f is defined on X, f is called monotonically increasing if for any
x1, x2 : x1 x2 f ( x1 ) f ( x2 ) . And f is called “strictly increasing” if for any
x1 x2 f ( x1 ) f ( x2 ) . And f is called monotonically decreasing if for any
x1 x2 f ( x1 ) f ( x2 ) and strictly decreasing if for any x1 x2 f ( x1 ) f ( x2 ) .
If f is monotonically increasing/decreasing on then we can say “ f is monotonic on ”.
If f is strictly increasing/decreasing on X, then we can say “ f is strictly monotonic on X”.
Theorem5: f is monotonically increasing on [a, b] . Then at any point h (a, b) there exist both
left and right limits: lim x h f ( x) and lim x h f ( x) . Moreover, the left limit at any point
h (a, b) is exactly the supremum of all values { f ( x) || x [a, h)}. And the right limit at any
point h (a, b) is exactly the infinum of all values { f ( x) || x (h, b]}.
In particular, at the end points a, b there exist the right limit lim x a f ( x) and the left limit
lim x b f ( x) .
Proof. We fix an arbitrary h (a, b) , let’s consider the set of
values { f ( x) || x [a, h)}, this set is bounded above by f (b) .
Therefore this set has the least upper bound [pict11].
Let’s show that lim xh f ( x) . According to the limit-
definition, if we show that for any sequence
{xn } [a, h) || {xn } h there must be { f ( xn )} , then
pict.11
lim x h f ( x) .
Let’s fix an arbitrary sequence {xn } [a, h) || {xn } h .
75
Analysis
Let’s fix an arbitrary positive , as is the least upper bound of the set { f ( x) || x [a, h)}, the
half interval ( , ] must contain at least one value f ( x0 ) || x0 [a, h) (here we use the least
upper bound criterion). Let’s consider the interval ( x0 , h) , as {xn } h || {xn } [a, h) , the interval
( x0 , h) contains all the terms of {xn } , starting from some number k . Let’s write it:
k : n k x0 xn h x0 xn || xn [a, h) . As f is monotonically increasing, there must
be f ( x0 ) f ( xn ) || xn [a, h) , remember that f ( x0 ) belongs to ( , ] , then f ( xn ) also
belongs to ( , ] . Really, we already have f ( x0 ) f ( xn ) and also f ( xn ) , because is
the least upper bound of { f ( x) || x [a, h)} and xn [a, h) . Then f ( xn ) belongs to ( , ] ,
and it is true for any n k .
So, for any n k all the terms of the sequence { f ( xn )} belong to () ( , ] [, ) ,
and is an arbitrary small positive number, then { f ( xn )} . We have proved that
lim x h f ( x) .
Similarly we can show that there exist the right limit lim x h 0 f ( x) m . We consider the set
{ f ( x) || x (h, b]} . This set is bounded below by f (a) , then it has the greatest lower bound m .
And we can show that lim x h f ( x) m .
Exercise2. Formulate the theorem, which is similar to the theorem5, for a monotonically
decreasing function f on [a, b] .
Let’s remind that if some function f (x) is one-to-one f : , then we can speak about
an inverse function f 1 : .
Graph of an inverse function. In practice it’s very convenient to use the next idea. Suppose we
have a graph of one-to-one function f on X and we need to understand quickly how the graph of
the inverse function f 1 looks. All we need to do is to reflect all our picture over the line y x ,
such symmetry changes the places of Ox, Oy [pict12], the new graph (after the symmetry) is the
graph of f 1 .
pict.12
76
Analysis
Inverse function theorem. A function y f (x) is strictly increasing and continuous on [a, b] .
Then: [A] f is one-to-one mapping f : [a, b] [ f (a), f (b)] and
[B] An inverse function x f 1 ( y ) , which is defined on [ f (a), f (b)] , is also strictly increasing
and continuous on [ f (a), f (b)] .
Proof. Let’s fix any number [ f (a), f (b)] , according to the (Book1,page145, consequence1
intermediate values of a continuous function), there exist c [a, b] such that f (c) . It means
that f : [a, b] [ f (a), f (b)] covers the segment [ f (a), f (b)] . And f obviously doesn’t “glue
together” numbers from [a, b] . Really, for any c d from [a, b] we have f (c) f (d )
(If c d f (c) f (d ) , If d c f (d ) f (c) , because f is strictly increasing).
Then f is one-to-one mapping [a, b] [ f (a), f (b)] . So [A] is proved.
As f is one-to-one, then there exist the inverse one-to-one mapping f 1 : [ f (a), f (b)] [a, b]
which we denote as x f 1 ( y ) . Let’s show that this function is also strictly increasing.
Let y1, y2 [ f (a), f (b)] || y1 y2 , let’s show that f 1 ( y1 ) f 1 ( y2 ) [Z].
1
Let’s notice that f ( y1 ), f 1 ( y2 ) are some numbers from [a, b] . If f 1 ( y1 ) f 1 ( y2 ) , then
f f 1 ( y1 ) f f 1 ( y2 ) y1 y2 , and we have a contradiction. If f 1 ( y1 ) f 1 ( y2 ) , then
f f 1
( y ) f f
1
1
( y ) y
2 1 y2 , and we have a contradiction again. The last variant is
f 1 ( y1 ) f 1 ( y2 ) (and in this case we don’t have any contradiction). So x f 1 ( y ) is a strictly
increasing function.
Let’s show that f 1 ( y ) is continuous on [ f (a), f (b)] . We need to show that at any concrete point
y0 [ f (a), f (b)] the limit lim y y0 f 1 ( y) exists and this limit is equal to f ( y0 ) .
(when y0 f (a) or y0 f (b) we speak about right/left limit).
Let’s consider the most general case, we fix any point y0 ( f (a), f (b)) inside the segment
[ f (a), f (b)] . As f 1 is strictly increasing (and therefore monotonically increasing), according to
1
the theorem4, there exist both left and right limits lim y y0 f ( y) and lim y y0 f 1 ( y) .
1
And the most important: lim y y0 f ( y) sup{ f 1 ( y) || y [ f (a), y0 )} [L1] and
lim y y0 f 1 ( y) inf{ f 1 ( y) || y ( y0 , f (b)]} [L2].
As f
1
is strictly increasing on [ f (a), f (b)] , then any element of the set { f 1 ( y) || y [ f (a), y0 )}
1
is strictly less than f ( y0 ) , and f 1 ( y0 ) is strictly less than any element of the set
{ f 1 ( y) || y ( y0 , f (b)]}. From here immediately follows that
sup{ f 1 ( y) || y [ f (a), y0 )} f 1 ( y0 ) inf{ f 1 ( y) || y ( y0 , f (b)]} [J].
77
Analysis
If sup{ f 1 ( y) || y [ f (a), y0 )} f 1 ( y0 ) , then there exist some number x1 between these
1
numbers sup{ f ( y) || y [ f (a), y0 )} x1 f 1 ( y0 ) inf{ f 1 ( y) || y ( y0 , f (b)]} [J1].
Then from [J1] we see that x1 is not a value of the function f 1 at any point y [ f (a), f (b)] ,
so f 1 doesn’t reach the value x1 at any point of [ f (a), f (b)] , in the same time x1 is a point of
the segment [a, b] , then f 1 is not one-to-one [ f (a), f (b)] [a, b] and we have a contradiction.
Then in [J] there must be an equality sign "" , i.e.,
sup{ f 1 ( y) || y [ f (a), y0 )} f 1 ( y0 ) inf{ f 1 ( y) || y ( y0 , f (b)]}. If we assume that the second
sign "" is actually a sign "" we will get a similar contradiction. So, the second sign "" must be
1 1 1
also "" . Now we have sup{ f ( y) || y [ f (a), y0 )} f ( y0 ) inf{ f ( y) || y ( y0 , f (b)]} [J2].
Let’s sum up. From [L1] and [L2] and [J2] we have:
lim y y0 f 1 ( y) f 1 ( y0 ) lim y y0 f 1 ( y) [V]. As both left and right limits at y0 exist, and
these limits are equal to f 1 ( y0 ) , then the ordinary limit at y0 exists, and it is equal to f 1 ( y0 ) .
So lim y y0 f
1
( y) f 1 ( y0 ) , it means exactly that f 1 ( y) is continuous at y0 .
1
So f ( y ) is continuous at any point y0 ( f (a), f (b)) . Similarly f 1 is right-continuous at f (a)
(the proof is similar and even more simple), and f 1 is left-continuous at f (b) .
Then f 1 is continuous on [ f (a), f (b)] .
Exercise3. Formulate the theorem, which is similar to the Inverse function theorem, for a strictly
decreasing function f on [a, b] .
Exercise4. f is continuous on [a, b] . Show that f is strictly monotonic on [a, b] if it reaches
different values at different points, i.e., for any x1 x2 from [a, b] we have f ( x1 ) f ( x2 ) .
78
Analysis
Limits and equivalent functions
Def. We say that {xn } goes to infinity if for any (big) С 0 there exist some number k ,
starting from which n k we have xn C . And we write: {xn } or lim n xn .
In particular, we say that {xn } goes to plus infinity if for any С 0 k : n k xn C
and we write {xn } or lim n xn . And we say that {xn } goes to minus infinity if
for any С 0 k : n k xn C and we write {xn } or lim n xn .
Notice! When we say “ {xn } converges” or “ {xn } has a limit” we imply that there exist some concrete
real number a such that lim xn a . The symbols / / are not limits!
These symbols are just auxiliary symbols, and we use them in order to describe the behavior of our
sequence {xn } . And when we say that some sequence converges (has a limit) we always imply that
it is a “finite limit”, which is some concrete real number a.
It’s easy to notice that and are more detail cases of . So when we have {xn }
we can also write {xn } , and when we have {xn } we can also write {xn } .
Def. When some sequence goes to (in particular to or ) we say that we have
an “infinitely large sequence”.
There exist the similar definition for functions:
Def. f is defined on and a is a limit point of . If for any sequence
{xn } || xn a n || {xn } a we have { f ( xn )} / / , then we say that f (x) goes to
infinity/plus infinity/minus infinity when x goes to a , and we write lim x a|| x f ( x) / / .
And again, when we say “ f has a limit at a”, we always imply that f has some “finite limit” ,
which is a concrete real number. The symbols / / are not limits, these are just auxiliary
symbols which we use to describe the behavior of our function f when x approaches to a
(by staying in ).
And the equivalent definition: lim x a|| x f ( x) / / if for any (big) number C 0 there
exist some 0 such that x : x D (a) we have f ( x) C // f ( x) C // f ( x) C .
And there are standard variations: if, for example, f (x) is defined on some (a R, a) and f (x)
goes to plus infinity when x goes to a from the left, then we write lim x a f ( x) .
If f (x) is defined in some (a R, a R) and f (x) goes to minus infinity when x goes to a
(from any side), then we write lim x a f ( x) .
79
Analysis
Let’s consider the function f ( x) 1/ x , this function is defined on (,0) (0,) and when x
goes to zero from the right, the value 1 / x goes to plus infinity, then we can write lim x 0 (1/ x) .
When x goes to zero from the left, the value 1/ x goes to minus infinity, then lim x 0 (1/ x) .
And finally, let f (x) is defined on some set ( R,) . If there exist some number A such that:
For any (small) 0 C R such that x (C,) f ( x) , then we say
there exist
“ f (x) goes to when x goes to “ and we write lim x f ( x) .
Let f (x) is defined on (, R) (here R 0 ). If there exist A such that
0 С R : x (,C ) f ( x) , then we say “ f (x) goes to A when x goes to
“ and we write lim x f ( x) .
Let f (x) is defined on (, R) ( R,) . If there exist A such that
0 С R : x (,C ) (C,) f ( x) , then we say “ f (x) goes to A when x
goes to “ and we write lim x f ( x) .
Exercise1. By using the previous information, define next the limits:
lim x / / f ( x) / / .
For practical purposes it’s very important to define infinitely small functions.
Def. (x) is defined on and a is a limit point of . If lim xa||x ( x) 0 , then we say that
(x) is “infinitely small when x a ” ( “infinitely small when x goes to a ”).
Exercise2. Let lim xa||x f ( x) , then the function ( x) f ( x) x is infinitely small
when x a . So, for any function f (x) such that lim xa||x f ( x) the next representation on
is possible: f ( x) ( x) , where (x) is defined on and (x) is infinitely small when x a .
And conversely. If f ( x) ( x) , where (x) is infinitely small when x a , then
lim xa||x f ( x) . It immediately follows from simpliest properties of limits, really:
lim x a|| x f ( x) lim x a|| x ( ( x)) lim x a|| x lim x a|| x ( x) 0 .
Exercise3. If (x) and (x) are infinitely small when x a , then their sum and product
( x) ( x), ( x) ( x) are also infinitely small when x a .
Exercise4. ( x) || x is infinitely small when x a . And g (x) is bounded on , then
( x) g ( x) is infinitely small when x a .
From the exercise4 follows that: if (x) is infinitely when x a , then for any constant R
the function (x) is also infinitely small when x a .
80
Analysis
Local comparison of functions
Let f ( x), g ( x) are defined on X and g ( x) 0 || x , and a is a limit point of X.
f ( x)
Def. Suppose that the limit lim x a|| x k exists (it means that k is some concrete real
g ( x)
number). If k 0 , then we say that f and g have the same order when xa
(“when x goes to a ”) and we can use the symbol f O(g ) .
In particular, when k 1 we say that f and g are equivalent when x a , and we can write
f g . And finally, if k 0 , we say that f is negligible with respect to g when x a,
and we can use the symbol f o(g ) .
This simple definition has a great importance in math and it has a very simple meaning.
f ( x) f ( x)
Let f is negligible with respect to g , so f o(g ) lim x a|| x 0 , then is infinitely
g ( x) g ( x)
f ( x)
small when x a , so the values of near the point a approach arbitrary close to zero,
g ( x)
it means that near the point a , at any concrete point x , the value f (x) is so much less
(in absolute value) than the value g (x) , that their ratio is almost zero. And even more, for any
small 0 there exist some deleted neighborhood of a such that for any point x from this
neighborhood, the absolute value of f ( x) / g ( x) is less than .
f ( x)
Let f is negligible with respect to g when x a . So we have lim x a|| x 0 , it means that
g ( x)
f ( x)
the ratio ( x) is infinitely small when x a , from here f ( x) ( x) g ( x) - and we can use
g ( x)
this representation instead of f (x) in our reasonings, so instead of f (x) we will write ( x) g ( x) ,
where (x) is infinitely small when x a . In many cases this approach is very convenient.
And conversely. If we have the representation f ( x) ( x) g ( x) || ( g ( x) 0 x ) on ,
where (x) is infinitely small when x a , then
f ( x) ( x) g ( x)
lim x a|| x lim x a|| x lim x a|| x ( x) 0 and f is negligible with respect
g ( x) g ( x)
to g when x a .
Exercise5. Explain the meaning of the other symbols "O" and "" in the same manner as we
explained above the meaning of "o" .
81
Analysis
The next theorem for equivalent functions may be very helpful in many cases:
Theorem1. f ( x), f1 ( x) and g ( x), g1 ( x) are defined on X and a is a limit point of X and
g ( x) 0, g1 ( x) 0 x . And f ( x) f1 ( x) and g ( x) g1 ( x) when x a .
f1 ( x ) f ( x)
If the limit lim x a|| x exists, then lim x a|| x also exists and these limits are equal.
g1 ( x) g ( x)
f ( x)
Consequence. When we calculate some limit lim x a|| x we can replace functions f , g by
g ( x)
any equivalent functions f1, g1 (so, f f1 and g g1 when x a ) and calculate the new limit, if it
exists, then the initial limit also exists and their values are equal.
f ( x)
Proof. We have f ( x) f1 ( x) and g ( x) g1 ( x) when x a . Let’s designate h( x) and
f1 ( x)
g ( x)
v( x) , we know that lim x a|| x h( x) 1, lim x a|| x v( x) 1 and obviously v( x) 0 on .
g1 ( x)
h( x ) h( x) lim x a|| x h( x) 1
The ratio is defined everywhere on and lim x a|| x 1 [T].
v( x) v( x) lim x a|| x v( x) 1
f1 ( x ) f ( x)
Let the limit lim x a|| x exists, so lim x a|| x 1 . By using [T] and the simpliest
g1 ( x) g1 ( x)
properties of limits we can write:
f ( x) h( x) f1 ( x) h( x ) f ( x)
lim x a|| x lim x a|| x lim x a|| x lim x a|| x 1 1 ,
g ( x) v( x) g1 ( x) v( x) g1 ( x)
so the initial limit exists and it is equal to A.
Next, there are similar definitions of symbols "O, , o" in the cases when x or x ,
and they also have a great practical value.
Let f ( x), g ( x) are defined on ( R,) and g ( x) 0 || x ( R,) .
f ( x)
Def. Suppose that the limit lim x k exists.
g ( x)
If k 0 , then we say that f and g have the same order when x goes to and we can use the
symbol f O(g ) . In particular, when k 1, we say that f and g are equivalent when x goes to
, and we can write f g . And finally, if k 0 , we say that f is negligible with respect g
when x goes to , and we can use the symbol f o(g ) .
f ( x)
Exercise6. Formulate the theorem, which is similar to the theorem1, for the limit lim x .
g ( x)
Exercise7. Give the definitions of symbols "O, , o" , when x .
82
Analysis
Radians
Assertion1. For any k the function f ( x) k x is continuous on [0,) .
Proof. Let’s fix an arbitrary a [0,) and take any sequence {xn } [0,) || xn a || {xn } a ,
we need to show that k xn k a , then (by definition) k
x is continuous at a.
Auxiliary1. If some positive sequence { y n } is infinitely small, then {k yn } is also infinitely small.
Proof. Let’s fix an arbitrary positive , and take the positive number k , as { yn } is infinitely small,
there exist the number m , starting from which yn 0 k yn k k yn k yn 0 .
So for any positive there exist the number m , starting from which k yn 0 , then {k yn } is
infinitely small.
Auxiliary2. Auxiliary inequality: k
a k b k a b a 0, b 0 .
Proof. We can assume that a b , if not, we can just permute a, b in both (left and right sides).
Then we can discard the absolute value | ... | signs. And we need to prove that k
a k b k a b,
this inequality is equivalent to k
a k a b k b and this one can be checked by raising both sides
to the k th power.
Let’s prove now the assertion1. {xn } [0,) || xn a || {xn } a , then a xn is an infinitely small
positive sequence, then (auxiliary1) the sequence k a xn is infinitely small. Then (auxiliary2)
0 k
a k xn k a xn from the squeeze theorem for sequences immediately follows that
k
a k xn is infinitely small. Obviously, for any sequence { y n } the next is true:
{ yn } is infinitely small { y n } is infinitely small. As k
a k xn is infinitely small k a k xn
is infinitely small, then x
k
n
k
a , everything is proved.
In particular, the square root function f ( x) x is continuous on [0,) .
Assertion2. sin is strictly increasing on (90o ,90o ) .
Proof. Let’s fix arbitrary 1 2 from the interval (90 ,90 ) , let’s show that sin 1 sin 2 .
o o
This inequality is equivalent to sin 2 sin 1 0 . Let’s use the formula for difference of sines:
2 1 2 1
sin 2 sin 1 2 cos sin [T]. All the factors on the right side of [T] are positive.
2 2
83
Analysis
2 1 2 1
Really, the angle (90o ,90o ) , then cos 0.
2 2
2 1
And as 2 1 2 1 0o 2 1 (0o ,90o ) , then sin 0 , so, the right part of [T]
2
is positive. Then sin 2 sin 1 0 and everything is proved.
Assertion3. sin is continuous at the point ~ 0o.
Proof. We want to show that lim
0 o
sin sin 0o 0 .
Let’s fix an arbitrary positive (0,1) . We can draw the
line y , it intersects the unit circle at two points [pict1],
one of these points defines the angle AOB (0 ,90 ) .
o o
Let’s designate o AOB and we also consider angle
AOB o , which is symmetrical to AOB with
respect to Ox . Let’s take any angle
( o , o ) o o , as sin is strictly
pict.1
increasing on (90o ,90o ) (lemma2), then
sin( o ) sin sin o sin o sin sin o and, according to our construction, sin o .
So we have sin .
Let’s sum up: for any positive (0,1) there exist the positive o such that ( , ) we
o o
have sin . We can rewrite: (0,1) 0 (0 ) | sin 0 | - it means
o
exactly that lim 0o sin 0 .
Assertion4. sin is continuous everywhere (o ,o ) .
Proof. Let’s fix an arbitrary angle ~ (o ,o ) , the value sin ~ is a constant. If we show that
~ is infinitely small when
the function sin sin ~ , then lim ~ sin sin ~ and sin
~ ~ ~
is continuous at . Let’s consider sin sin , we have: sin sin 2 cos
~ ~ sin [M].
2 2
~ ~ ~
Let’s show that sin is infinitely small when . We consider sin as a composite
2 2
~
function ( ( )) , here ( ) and ( ) sin . The function ( ) is continuous at ~
2
and (~) 0o and sin is continuous at 0o (assertion3), then the composite function ( ( ))
84
Analysis
~ ~ ~
is continuous at ~ , and we have lim ~ ( ( )) ( (~)) lim ~ sin sin 0
2 2
~
and sin is infinitely small when ~.
2
The expression [M] can be considered as a product of a bounded function:
~ ~
2 cos , 2 cos 2 (o ,o ) and an infinitely small (when ~ ) function
2 2
~
sin ~ function. Everything is proved.
. Then [M] is infinitely small when
2
Exericise1. Show that cos is continuous everywhere (o ,o ) .
Exericise2. Show that each function tg , ctg is continuous on it’s domain.
The unit circle is a measurable figure. Really, it’s boundary consists of two graphs of functions
~
f ( x) 1 x 2 || x [1,1] and f ( x) 1 x 2 || x [1,1] .
Let’s show that both these functions are continuous on [1,1] , then (theorem4) is measurable.
The function x is continuous on [1,1] (we showed in the 1 st book that
2
x k || k is continuous
everywhere). Then ( x) 1 x 2 is also continuous on [1,1] , the values of (x) on [1,1] belong to
the segment [0,1] , and the function ( y) y is continuous on [0,1] (because ( y) y is
continuous on [0,) ). Then the composite function ( ( x)) 1 x 2 is continuous on [1,1] .
~
And therefore f ( x) 1 x
2
is also continuous on [1,1] .
Def. is a unit circle [pict]. Let’s fix an arbitrary angle
[0o ,360o ] and let’s draw any central angle of ,
it defines an arc of a circle, which we will call an -arc.
There also appears a figure, which is called an -sector.
The unit circle is measurable, then it’s boundary is a zero
area figure (Book1, “Area construction”, [3-rd criterion of
measurability]). Any -arc is some part of a zero area figure , [pict]
then (assertion8, “Area construction”) any -arc is also a zero
area figure. Let’s consider now any -sector. It’s boundary consists of two equal segments (which
are both zero area figures) and an -arc, then the boundary of any -sector is a zero area figure,
then any -sector is measurable, and it’s area (the number ) S ( ) is defined. Let’s describe the
process which gives us the area of any -sector.
85
Analysis
pict.2
The process. Let’s fix any -sector.
For any natural number, which is a power of two 2 , we divide the angle into 2 equal angles .
n n
We build the internal figure int
n [pict2] by connecting several points on the circle, and
the external figure ext
n , by building a tangent line at the middle of every small arc.
The area of any triangle from int
1
is equal to S 1 1 sin n
sin / 2
n
n
. There are 2 such
n
2 2 2
n 1
triangles, so the area S (int
n ) (2 / 2) sin( / 2 ) 2
n n
sin( / 2n ) .
Next, any triangle in n
ext
has a height 1 , it divides the triangle into two equal right triangles, with
angles , so their cathetuses, which are opposite to these angles, are equal to tg .
2 2n 2 n 1
Then the area of each triangle is tg tg 1 2 tg n 1 , and there are 2 such triangles,
n
2n 1 2n 1 2
n 1
n ) 2 tg ( / 2
S (ext n
so ).
n 1 n 1
We have two sequences {S (int
n )} {2 n )} {2 tg ( / 2
sin( / 2n )} and {S (ext n
)}
let’s call it an internal and an external sequence.
86
Analysis
Assertion5. Both internal and external sequences go to the same limit.
n 1
Proof. Let’s notice that the internal sequence {S (int
n )} {2 sin( / 2n )} is strictly increasing.
Really, for any n we have int
n n 1 S (n ) S (n 1 ) [pict3]. Let’s explain it.
int int int
pict.3
Let we have a figure int
n , now the angle is divided into 2 equal central angles.
n
For each of these angles the angle bisector can be drawn. Then will be divided into 2 2n 2n 1
equal central angles. If we connect now several points on the circle, we will get the figure n 1 and
int
obviously n n 1 S (n ) S (n 1 ) . Notice that the area of n1 differs from the area of
int int int int int
int
n by the area of several triangles, which are built on the chords of the figure n , that’s why the
int
area of n1 is strictly greater than the area of
int
int
n . But in fact, for us it’s even enough the
estimation S (n ) S (n 1 ) , and this inequality immediately follows from n n1 .
int int int int
An internal sequence is also bounded above, really as any internal figure int
n belongs to the
n ) S ( ).
-sector, then it’s area is not greater than the area of -sector, so S (int
According to the theorem about a limit of a monotonic sequence, the internal sequence
n 1
{S (int
n )} {2 sin( / 2n )} converges to some limit A. Let’s consider now an external sequence
n 1
n )} {2 tg ( / 2
{S (ext n
)}.
Let’s take the term of this sequence with number n 1, so
n 1 n 1 sin( / 2 )
n
(2 n 1 sin( / 2 n )) S ( int
n )
S (n 1 ) 2 tg ( / 2 ) 2
ext n
. So we have
cos( / 2n ) cos( / 2 n ) cos( / 2 n )
S (int
n )
{S (int
n )}
{S (ext
n 1 )} n
. The sequence {S (n )} goes to A, and the sequence
int
cos( / 2 ) {cos( / 2 )}
n
{cos( / 2n )} goes to 1 .
(Comment: is a fixed angle, then the sequence { / 2n } 0o and the function cos is continuous
at 0 , then {cos( / 2 )} cos(0 ) 1).
o n o
87
Analysis
Then, according to the basic properties of limits,
S (int
n )
n
{S (ext n 1 )} {S ( n )} . So, both internal and external
ext
cos( / 2 ) 1
sequences go to the same limit , these sequences are the sequences of areas of internal and
external figures (for the initial -sector).
In the “Area construction” (Book 1) we have proved the assertion5 [2-nd criterion of measurability]
from which follows that the area of -sector is also equal to S ( ).
Let’s sum up: for any angle [0o ,360o ] , an -sector is a measurable figure and it’s area S ( )
n 1
can be determined as a limit of the sequence {2 sin( / 2n )} {S (int
n )} , or as a limit of
n 1
the sequence {2 tg ( / 2 )} {S (ext
n
n )}.
Let’s consider now any internal figure int
n , it defines n equal
chords on the circle [pict4]. These chords are equal because they
are bases of equal isosceles triangles.
In order to calculate such chord we can take any isosceles triangle
with angle and draw it’s height, it divides the triangle into two
2n
equal right triangles with angles , so the base is
2 2n 2n 1
sin n 1 sin n 1 2 sin n 1 .
2 2 2
n pict.4
And there are 2 such chords in total, then the sum of all chords of
the internal figure n is n 2n 2 sin n 1 2n 1 sin n 1 [V].
int
2 2
And the sequence of internal figures { n } defines the sequence { n } of sum of chords.
int
Assertion6. The sequence { n } of sum of the chords converges, and it’s limit is 2 times greater
than the area S ( ) .
Proof. We have { n } {2n 1 sin( / 2n 1 )} and {S (int
n )} {2
n 1
sin( / 2n )}, we know that
n )} S ( ) . Let’s take the term of the sequence {S (n )} with number n 1.
{S (int int
So S (int
n 1 ) 2 sin( / 2
n n 1
) and we have n 2n 1 sin( / 2n 1 ) , then n 2 S (int
n 1 ) , and we
n 1 )} and the sequence {S (n 1 )} goes to S ( ) , then { n } goes to 2 S ( ) .
{n } 2 {S (int int
have
So, for any [0o ,360o ] the numbers L( ), S ( ) are concrete real numbers.
Def. For any -sector, the limit of the sequence { n } is called a length of -arc, and we denote
it L( ) . We got the formula L( ) 2 S ( ) (remember
88 that we operate on the unit circle ).
Analysis
Assertion7 [Additivity]. For any angles , [0o ,180o ] their sum [0o ,360o ] defines
an -sector. And S ( ) S ( ) S ( ) and L( ) L( ) L( ) .
Proof. Let’s fix any angles , [0o ,180o ] . We could use the formulas
that we obtained above to prove this assertion, but there is a more
simple way [pict5]. Let’s build a central angle and
a central angle right next to it. So we have an -sector and
a -sector which form together an -sector. Both , -sectors do
not have any common internal points, then, by additivity of area:
S ( ) S ( ) S ( ) [T]. We showed above that for any angle
[0o ,360 o ] we have L( ) 2S ( ) , in particular it is true for the pict.5
angles , , ( ) . Let’s multiply by 2 both sides of [T],
we will get L( ) L( ) L( ) .
Consequence1. Let we have some -sector (where [0o ,360o ] ).
Let’s divide it into 1, 2 ... n -sectors. Then the area of -sector is a sum of areas of
1, 2 ... n -sectors, and the length of -arc is a sum of lengths of 1, 2 ... n -arcs.
Def: Let’s take any 180o -sector, it defines a 180o -arc [pict6]. The length of that arc L(180o )
must be denoted by the letter , so L(180o ) .
A length of any arc is a limit of a concrete sequence, in this case we
n 1 180o
have lim n 2 sin n 1 L(180 ) [V].
o
2
From here follows that the number can be calculated as a limit
n 1 180o
of the sequence 2 sin n 1 .
2
In order to get a very good approximation of we can calculate
180o
n 1
some concrete term 2 sin n 1 for some big value of n . pict.6
2
Practical calculations show that 3,1415 .
Consequence2. The full circle can be represented as two 180o -arcs, then (consequence1)
the length of the circle is 2 . The area of a half-circle is two times less than it’s length, so it is / 2
(because L( ) 2S ( ) ). The area of the full circle is . By using simple ideas (as above):
o
the length of 30 arc is / 6 , the length of 45o arc is / 4 , the length of 60o arc is / 3 .
89
Analysis
Consequence3. For any [0o ,360o ] we have L( ) L( ) and S ( ) S ( ) .
It immediately follows from the assertion7. Let’s fix any , then the angle ( ) is a positive
angle. Let’s build an -sector and a ( ) -sector right next to it, together they form a -sector
and, by additivity, (assertion7 ) we have S ( ) S ( ) S ( ) S ( ) S ( ) . Now we can
multiply by 2 both sides, then we get L( ) L( ) .
Auxiliary3. For any angle (0o ,180o ) the length of -arc
~~
is not greater than the length of the segment AB 2tg ( / 2)
~ ~
[pict7]. Where A, B are the points of intersection of
a tangent line (which passes through the central point of -arc)
and rays OA, OB , where AOB .
Proof. Let’s fix any central angle (0o ,180o ) .
Let OX is an angle bisector and l is a tangent line at the pict.7
Any internal figure n is symmetrical with
int
point X.
respect to OX (because in order to build such figure,
the angle must be divided into 2 n equal parts, so
there will be a symmetry with respect to OX ).
Let’s consider all chords of the internal figure int
n . The
figure n defines the points
int
A, A1, A2 , A3 .... Ak , B on
the circle (in fact the number k 2 1 but we don’t
n
need it). Let’s build the rays [pict8]
OA, OA1, OA2 , OA3 .... OAk , OB , these rays intersect
the tangent line l (which passes through X ) at
~ ~ ~ ~ ~ ~
some points A, A1, A2 , A3 .... Ak , B .
We will show that each chord is strictly less than
the appropriate segment on l .
~~ ~~ ~ ~
So AA1 AA1 , A1 A2 A1 A2 , A2 A3 A2 A3 ...[G].
pict.8
Let’s consider for example the chord A1 A2 , this chord is
a base of the isosceles triangle OA1 A2 , then OA2 A1 is an acute angle. Let’s build the line
~ ~
A2T || A1 A2 , the angle OA2T is equal to OA2 A1 , so it is also an acute angle. Let’s notice that
~ ~ ~
OA2 A1 is an obtuse angle, because it is an exterior angle of the right triangle OXA2
~ ~ ~ ~ ~
(where OXA2 90 ). So OA2T is acute and OA2 A1 is obtuse, it means that A2T lies inside
o
90
Analysis
~ ~ ~ ~
the triangle OA2 A1 . As OA2T is an isosceles triangle, the angle OTA2 is acute, then the
~ ~ ~~ ~ ~
adjacent angle A1TA2 is obtuse, then A1 A2 is the greatest side in A1TA2 (because it lies in front
~~ ~ ~ ~
of an obtuse angle), in particular A1 A2 A2T and obviously A2T A1 A2 (because A2T and A1 A2
~
are bases of similar isosceles triangles with the similarity coefficient OA2 / OA2 1 ), then
~~
A1 A2 A1 A2 .
~~ ~~ ~ ~
So we have [G] AA1 AA1 , A1 A2 A1 A2 , A2 A3 A2 A3 .... The sum of all chords
AA1 , A1 A2 , A2 A3 .... is n (the limit of such sums is the length of -arc). And the sum of all
~~ ~~
segments AA1, A1 A2 , A2 A3 .... is A B . From [G] we have n AB , so every term of the sequence
~~ ~~
{ n } is less than A B , then the limit of this sequence is not greater than A B , it means that L( )
~~ ~~ ~~
is not greater than A B , so L( ) AB . And finally, the segment A B [pict7] consists of 2 equal
~~
cathetuses of right triangles, with the common hypotenuse 1 and angles / 2 , then AB 2tg ( / 2) .
Everything is proved.
Until this moment, when we were talking about an -sector,
we meant virtually any -sector of a unit circle. Let’s agree that
from now on any central angle must be built in the standard
way, as we do it in trigonometry (Book I , “Angles”). So, any angle
[0o ,360o ] must be drawn as an angle BOA , where B is
moved along the circle in the counterclockwise direction from A
[pict]. And from now on, any -sector is a concrete figure, which
[pict]
we get after we build the central angle [0o ,360o ] according
to the standard rules. For any -sector, the numbers L( ), S ( ) are uniquely defined.
And we can say now that L( ), S ( ) are concrete functions on the segment [0 ,360 ] .
o o
From the consequence3 we know that for any [0o ,360o ] we have S ( ) S ( ) and
L( ) L( ) . So, the functions L( ), S ( ) are both strictly increasing on [0o ,360 o ] .
Assertion8. The function L( ) is continuous on [0o ,180 o ) .
Proof. We will use the estimation from the auxiliary3 here. Let’s show that L( ) is continuous at
any point ~ [0o ,180 o ) . We fix an arbitrary ~ [0o ,180 o ) . If ~ , then
L( ) L(~) L( ~) (the equality L( ) L(~) L( ~) || ~ immediately follows from
~~ ~~
the additivity of L). Obviously ( ~) (0o ,180 o ) , then (auxiliary3) L( ~) AB , where A B
91
Analysis
is a segment of a tangent line which passes through the middle
point of ( ~) -arc [pict9].
~~ ( ~ ) ~ ( ~)
And A B 2tg . So we have L( ) 2 tg ,
2 2
~
we can rewrite it L( ) L(~) 2 tg ( )
2
~
( ) pict.9
0 L( ) L(~) 2 tg [1] for ~ .
2
( ) ~
If we assume ~ , we will get the similar inequality 0 L(~) L( ) 2 tg [2].
2
~) 2 tg ~
We can unite the estimations [1] and [2] and write 0 L( ) L( [E].
2
Here we will use the squeeze theorem for functions. Obviously lim ~ 0 0 and if we show that
~
also lim ~ 2 tg 0 , then there must be lim ~ L( ) L(~) 0 and it means that
2
L( ) L(~) is infinitely small when ~ , then L( ) L(~) is infinitely small when ~ ,
then lim L( ) L(~) and L( ) is continuous at ~ . Everything is proved.
~ ~
Let’s show that 2 tg [F] is infinitely small when ~ . Let’s consider at first tg
2 2
~
as a composite function ( ( )) , where ( ) and ( ) tg .
2
Obviously ( ) is continuous at ~ and ( ) tg is continuous at ( ) 0o , then the
composite function ( ( )) is continuous at ~ , it means that
~ ~ ~ ~ ~
lim ~ ( ( )) ( ( )) lim ~ tg tg 0 , then tg is infinitely
2 2 2
small when , then [F] is infinitely small when
~ ~.
Consequence4. L( ) is continuous on [0o ,360o ] .
We already know that L( ) is continuous on [0 ,180 ) . Let’s consider L( ) on [180 ,270 ] .
o o o o
~ [180 ,270 ] there exist the unique [0o ,90o ] such that ~ 180 and according
For any
o o o
~) L( ) L(180 o ) L(~) L( ) . So the function L( ) on [180 ,270 ]
to additivity: L(
o o
differs from L( ) on [0 ,90 ] only by a constant
o o
. Then L( ) is continuous on [180 ,270 ] .
o o
In the exactly similar way we can show that L( ) is continuous on [270 ,360 ] .
o o
After we sum up our results and conclude that L( ) is continuous on [0 ,360 ] .
o o
Similarly, as L( ) is strictly increasing on [0 ,180 ) , then it is strictly increasing on [0 ,360 ] .
o o o o
Then L( ) is strictly increasing and continuous on [0 ,360 ] .
o o
92
Analysis
According to the inverse function theorem, L : [0o ,360o ] [0,2 ] is one-to-one mapping,
and there exist the inverse mapping ( L) : [0,2 ] [0o ,360o ] which is also continuous and strictly
increasing.
~
The function L( ) takes the angle ~ [0o ,360 o ] and compares to it the length L L(~) of ~ -arc.
~
The function (L) makes the inverse action, it takes the number L [0,2 ] and compares
~
~ ( L~) [0o ,360 o ] to it such that the length of ~ -arc is L
the angle .
Let’s extend our function L on [0 , ) . For any angle
o o
360 o there exist the unique
representation 360o k ~ || k , ~ [0o ,360o ) and we define: L( ) 2 k L(~) .
Geometrical meaning [1]. By definition, any angle 360 o we represent like 360o k ~
and we must understand such angle (by definition) as a combination of k full rotations around the
~ 360o . According to our
circle, starting from A in the counterclockwise direction and some angle
definition, for any 360o the length of -arc is the total length of k full circles (i.e., 2 k ) plus
~ 360o .
the length of ~ -arc, where
Now L( ) is defined on [0o ,) , this set can be divided into segments
[360o k , 360o (k 1)] || k , k 0 . On any concrete segment [360o k , 360o (k 1)]
the function L( ) differs from the function L( ) on [0o , 360o ] only by a constant 2k .
Then L( ) is also strictly increasing and continuous on every segment [360o k , 360o (k 1)].
Every segment [0o , 360o n] can be represented as a union of segments
[0o , 360o ], [360o 1, 360o 2] .... [360o (n 1), 360o n] , on every of which L( ) is strictly
increasing and continuous, then L( ) is strictly increasing and continuous on [0o , 360o n] .
Theorem1. L( ) is strictly increasing and continuous on [0o ,o ) and L( ) is one-to-one
mapping [0 , ) [0,) . The inverse mapping ( L) : [0,) [0o ,o ) is also strictly
o o
increasing and continuous.
Proof. This theorem is very simple, but anyway, let’s give the detail proof of it.
[Increase of L( ) ]. Let’s fix any 1 2 [0o ,o ) there exist the segment [0 , 360 n] that
o o
contains both these points 1, 2 . As L( ) is strictly increasing on [0o , 360o n] , then
L(1 ) L( 2 ) , then L( ) is strictly increasing on [0o ,o ) .
93
Analysis
[Continuity of L( ) ]. We just need to show that L( ) is continuous at any point ~ [0o ,o ) ,
~ [0o ,o ) , by definition,
so let’s fix now any L( ) is continuous at ~ if for any sequence
{ n } ~ || { n } [0o ,o ) we have {L( n )} L(~).
~ } ~ || {~ } [0 , ) , as this sequence converges, it must be
Let’s fix an arbitrary sequence {
o o
n n
bounded, then there exist the segment [0o , 360o n] which contains all the terms of the sequence
{~n } and ~ . As L( ) is continuous on [0o , 360o n] , then it is continuous at ~ [0o , 360 o n] ,
and for the sequence {~n } ~ || {~n } [0o , 360o n] we have {L(~n )} L(~) .
Then L( ) is continuous on [0 , ) .
o o
Let’s show that L( ) is one-to-one mapping [0o ,o ) [0,).
[Step1] we need to show that for any number b [0,) there exist some b [0o ,o ) such that
L(b ) b .
Let’s consider the numbers 2 , b . By the Archimedes axiom there exist n~ such that b 2 n~ .
Let’s consider the segment [0o , 360o n~] , as we noticed above, L( ) is strictly increasing and
continuous on [0o , 360o n~] . According to the inverse function theorem, L( ) makes one-to-one
mapping [0o , 360o n~] [0, 2 n~] , as b (0, 2 n~) , then there exist some b [0o , 360o n~]
such that L( b ) b .
[Step2] Let’s show that L( ) doesn’t “glue together” elements of [0 , ) . Really as L is strictly
o o
increasing on [0o ,o ) , then for any 1 2 [0o ,o ) we have L(1 ) L( 2 ) .
From the [Step1] and [Step2] follows that L( ) is one-to-one mapping [0o ,o ) [0,) .
Then the inverse mapping ( L) : [0,) [0o ,o ) is defined.
Let’s show that ( L) : [0,) [0o ,o ) is also strictly increasing and continuous.
As L( ) is strictly increasing and continuous on every segment [0o , 360o n~] , then the inverse
function (L) must be strictly increasing and continuous on every image [0, 2 n ~] of such
~ the function
segment. So, for any n (L) is strictly increasing and continuous on [0, 2 n
~] ,
it’s very easy to deduce from here that (L) is strictly increasing and continuous on [0,)
94
Analysis
(the proof consists of exactly the same steps as [Increase of L( ) ], [Continuity of L( ) ],
but now these steps must be done for (L) ). Everything is proved.
Let’s finally extend our function L on (o ,o ) .
Now L( ) is defined for any [0o ,o ) , let’s define L( ) for any (o ,0o ] .
For any negative angle (o ,0o ] . We have , and we define: L( ) // by def // L(| |) .
So, for any negative angle , the length of -arc is defined like the number which is opposite to
the length of | | -arc.
Geometrical meaning [2]. Let’s build any negative angle [360o ,0o ] on the unit circle.
Then the value L( ) is a minus length of the arc which is opposite to the negative angle .
For any 360o there exist the unique representation 360 k ~ || k , ~ (360 ,0 ] ,
o o o
and this angle (by definition) must be understanded as k full rotations around the circle in
the clockwise direction and some negative angle ~ . And according to our definition, L( ) is
a minus total length of k full circles, minus the length of the arc which is opposite to the negative
angle ~.
Now L( ) is defined for any angle (o ,o ) .
As L( ) is strictly increasing and continuous on [0 , ) , then L( ) is strictly decreasing and
o o
continuous on ( ,0 ] . As L( ) on [0 , ) is one-to-one mapping [0 , ) [0,) , then
o o o o o o
L( ) on (o ,0o ] is one-to-one mapping (o ,0o ] (,0] .
As L( ) is continuous on (o ,0o ] and on [0o ,o ) , then L( ) is continuous everywhere
(o ,o ) .
Let’s also extend the function (L) . Now it is defined on [0,) . We extend it in the exactly similar
way as above, for any L (,0] we have L | L | , and we define ( L) // by def // (| L |) .
As (L) is strictly increasing and continuous on [0,) , then (L) is strictly decreasing and
continuous on (,0] . As (L) on [0,) is one-to-one mapping [0,) [0o ,o ) , then (L)
on (,0] is one-to-one mapping (,0] (o ,0o ] .
As (L) is continuous on (,0] and on [0,) , then (L) is continuous everywhere (,)
Exercise3. Show that ( L) || L (,) is an inverse function for L( ) || (o ,o ) .
(use that (L) on [0,) is already an inverse function for L( ) on [0o , o ) and use also
the simple rules, according to which we extended these functions).
95
Analysis
Let’s sum up, now L( ) is defined on (o ,o ) and the inverse function (L) is defined on
(,) . Let’s notice that when L goes from to 0 , the angle value (L) goes from to
o
0 o , when L goes from 0 to , the angle value (L) goes from 0 o to o . So in general,
when L goes from to , the angle value (L) goes from o to o .
Let’s consider trigonometric functions sin , cos || ( , ) [R1]. Each of these functions
o o
must be turned into a composite function sin ( L) , cos ( L) || L (,) [R2].
When L goes from to , the angle value (L) goes from o to o .
And the new representation does not “restrict” the angle value (L) , and now the variable L
(which is a real number) is our main variable.
Def: For any angle (o ,o ), the number L( ) (which can be understanded as a length of
-arc ,Geometrical meaning [1],[2]) is called a radian measure of .
And we can say “ L( ) is in radians”. For example, “ is 180o in radians” and “ 2 is 360o in
radians” and “ / 3 is 60 in radians” and “ / 2 is 90 in radians”.
o o
In general, any number L (,) can be called “ L radians”. And now our functions
sin ( L) , cos ( L) || L (,) are the functions of a radian argument L .
Assertion9. sin ( L) , cos ( L) are both continuous on (,) .
Proof. The function (L) is continuous on (,) , the values of (L) on (,) belong to
(o , o ) . And the function sin is continuous on (o , o ) , then (continuity of a composite
function) the function sin ( L) is continuous on (,) . And similarly for cos ( L) .
Exercise4. What are the domains of composite functions tg ( L), ctg ( L) ?
Explain why each of these functions is continuous on it’s domain.
96
Analysis
sin ( L)
Theorem2. lim L 0 1
L
sin ( L)
Proof. Let’s consider the right limit lim L 0 .
L
We fix any positive number L (0, / 2) , it defines
the unique angle ( L) (0o ,90o ) such that the length
of (L) -arc is L [pict10]. We have shown earlier
( L) pict.10
the estimation L 2tg [ES] (auxiliary3).
2
We need to find the lower estimate of L . Let’s denote ( L) AOB , we draw the perpendicular
BH , obviously BH AB and AB AX XB , where X is a central point of (L) -arc, let’s notice
that the sum AX XB is the first term 1 of the sequence { n } , the limit of this sequence is
the length L of (L) -arc. The sequence { n } is always strictly increasing (we can understand it
by simple geometrical reasoning, or just by using the connection between { n } and a strictly
increasing sequence S (n ) ). Then we have: sin ( L) BH AB AX XB 1 L .
int
( L)
Let’s take [ES], then we have sin ( L) L 2tg , let’s divide all the sides by sin ( L) 0 ,
2
( L) ( L) ( L)
2tg 2 sin cos
L 2 2 2 1
then 1 and we have
sin ( L) sin ( L) ( L) ( L) ( L )
2
2 sin cos cos
2 2 2
( L) sin ( L)
2
L 1
1 from here follows that cos 1 [Z].
sin ( L) ( L)
2
2 L
cos
2
And we will use the squeeze theorem for functions here. Obviously lim L 0 1 1.
( L)
2
Let’s show that lim L 0 cos 1 , if it’s true, then from [Z] there must be
2
sin ( L) ( L)
2
lim L 0 1 . We will show that the ordinary limit lim L 0 cos 1 (then the right
L 2
( L) ( L)
2
limit lim L 0 cos exists and equals 1 ). Let’s consider cos as a composite function
2 2
( L) ( L)
( ( L)) , where ( L) and ( ) cos . The function ( L) is continuous at
2 2
~
the point L 0 (because (L) is continuous everywhere) and the function ( ) cos
97
Analysis
~
is continuous at (0) 0o , then the composite function ( ( L)) is continuous at L 0 , and
( L) (0)
we have lim L 0 ( ( L)) ( (0)) lim L 0 cos cos cos 0o 1.
2 2
( L) ( L)
2
We showed that lim L 0 cos 1, then, obviously lim L0 cos 1 1.
2
2 2
sin ( L) sin ( L)
Therefore lim L 0 1 . The function ( L) is an even function which is defined
L L
on L (,0) (0,) , and it’s graph has a symmetry with respect to Oy .
It follows from ( L) ( L) for any L 0 . For any such function: if one of the limits
lim L0 ( L), lim L0 ( L) exist, then the other limit also exists and these limits are equal.
sin ( L) sin ( L)
So there must be lim L 0 1 . The function has both left and right limits at
L L
~ sin ( L)
the point L 0 , and these limits are equal to 1 , then the ordinary limit lim L 0 exists,
L
and it is also equal to 1 . Everything is proved.
In mathematics all the functions sin, cos,tg, ctg are usually considered as functions of a radian
argument L . And instead of writings sin ( L), cos ( L), tg ( L), ctg ( L) people write just
sin x, cos x, tgx, ctgx , where x is assumed to be a radian variable L . So we assume x L and
we just neglect the letter . Let’s notice that all the basic trigonometric formulas are still true.
Any formula sin 2 2 sin cos , which is true for any angle , can be rewritten for (L) ,
so sin 2 ( L) 2 sin ( L) cos ( L) and we replace L by x and neglect the letter , so we have
sin 2 x 2 sin x cos x . And the same can be done with any other trigonometric formula.
sin x
The theorem2 can be written as lim x 0 1 , it is a very important limit in mathematics.
x
We will use it to obtain the derivatives of all trigonometric functions. From now on we consider all
trigonometric functions as functions of a radian argument x (,) , such approach is extremely
important, only now we can define inverse trigonometric functions, later, by using radians, we will
be able to provide Taylor serieses for trigonometric functions and etc.
Inverse trigonometric functions
Let’s consider sin x || x , , cos x || x [0, ], tgx || x , , ctgx || x (0, ) ,
2 2 2 2
each of these functions is continuous and strictly increasing/decreasing on it’s segment/interval.
Then (according to the inverse function theorem) each of these functions is one-to-one mapping:
98
Analysis
sin x : , [1,1] || cos x : [0, ] [1,1] ||
2 2 and for each of these functions the inverse
tgx : , (,) || ctgx : (0, ) (,)
2 2
function is defined, which is also continuous and strictly increasing/decreasing on it’s domain.
arcsin y : [1,1] , || arccos y : [1,1] [0, ] ||
2 2
The inverse functions are denoted like:
arctgy : (,) , || arcctg y : (,) (0, )
2 2
.
These functions have very simple meanings. For example, arcsin y is the angle x
2 , 2
(in radians) such that sin x y . And arctg y is the angle x , (in radians) such that
2 2
tgx y and etc.
The graphs of the inverse trigonometric functions can be easily obtained from the graphs of the
initial trigonometric functions, we just need to reflect each of these graphs over the line y x
(red line) [pict1], [pict2].
pict1
99
Analysis
pict2
Exercise1. If we pay attention to the pictures, it’s easy to notice that for any x 0, the line
2
y x lies above the graph y sin x , but below the graph y tgx , it means that for any x 0,
2
there must be: sin x x tgx . Show that this equality is really true (it’s easy to do by using
the estimations that we got during the constructions of radians).
For Reader’s practice:
[1] Show that for any y [1,1] we have arcsin y arccos y .
2
[2] Prove that sin( 2 arcsin x) 2 x 1 x 2 .
[3] Given that arccos y arccosz arccosh , prove that y 2 z 2 h2 2 yzh 1.
5 2
[4] Solve the equation (arctgx) (arccosx)
2 2
. Answer: x 1.
8
12 4 63
[5] Prove that arcsin arccos arctg .
13 5 16
100
Analysis
Power function
In any multiplicative group G , an integer power a || m of any element
m
a G is defined.
In particular, for any real number a R the real number a m R is defined ( m is any integer
2 1/ 5
number). We want to define the meaning of symbols like ( 2 ) , 3 5 , 18 and etc., i.e., we want
to throw away the restriction for the power m “to be integer”.
We have already defined what is k
a || a R, k . And we showed that for any a 0 and for
any k the real number k a is defined, now we define: a1/ k k a .
If we want to define a || a R, m R , we have to make the restriction for a to be positive
m
(at least non negative). Really, even now we see that for any negative number a 0 the number
a1/ 4 4 a is not defined, by definition 4
a is such real number b that b 4 a , but in any case
b 4 0 , and we have a 0 , then b 4 a , and the element a1 / 4 is not defined.
So, from now on, a is always a positive real number a 0.
According to the standard properties of integer power, in any commutative multiplicative group we
have: [A] a a a
m n m n
, [B] (a b) m a m b m , [C] a mn (a m ) n (a n ) m , [D] (a m ) 1 (a 1 ) m .
In particular it is true for any a, b R, m, n (in the last case [D] there must be a 0 ).
Auxiliary1. If for some positive real numbers a, b we have a b for some natural d , then a b.
d d
Comment. The field of real numbers R is an ordered field, and it’s very easy to get Auxiliary1 if we
use the Property4 [C] of ordered rings/fields, Book1, page43.
Def1. Let a R, a 0 is any real number and q Q is any rational number. There exist the
m
representation q || m , k . We define a q a m / k // by def // (a1 / k ) m || m .
k
Assertion1. The def1 is correct. The main point here is that any rational number can be
represented as a quotient of an integer and a natural in many ways, let
~
q
m m ~ , k , k~ we need to show that a m / k is equal to a m~ / k~ .
~ || m, m
k k
~ ~
The main idea: a m / k and a m / k are positive numbers, we will show that they are equal if they are
~
raised to the natural power k k , then, according to auxiliary1, these numbers are equal.
~ ~ ~
(a m / k ) k k , the number a m / k // by def // (a1/ k ) m , then (a m / k ) k k ((a1 / k ) m ) k k ,
Let’s calculate
1/ k ~
here the element a is some real number and m, k , k are integers, so we can rewrite
~ ~ ~ ~ ~ ~
((a1 / k ) k ) m k a m k [1] and in the exactly similar way we can get (a m / k ) k k a m k [2].
101
Analysis
m m~ ~ ~
It’s easy to see that the expressions [1] and [2] are equal, really ~ mk mk , then
k k
~ ~ ~ ~
a m k a m k and [1] and [2] are equal. Then (the main idea) we have a m / k a m / k .
Assertion2. For any positive real numbers a, b 0 and for any q, q1 , q2 Q we have:
[A] (a b) q a q b q , [B] a q1 q2 a q1 a q2 , [C] a q1 q2 (a q1 ) q2 (a q2 ) q1 .
m
Let’s prove [A]. We fix any rational number q Q and it’s representation q || m , k .
k
Then (a b) (a b)
q m/ k
/ by def / ((a b)1/ k ) m and a q b q a m / k b m / k (a1/ k ) m (b1/ k ) m .
We already have the property [A] for any integer power m, so (a1/ k ) m (b1/ k ) m (a1/ k b1/ k ) m .
And we need to show that numbers (a1/ k b1/ k ) m and ((a b)1/ k ) m are equal. It’s obviously enough
to prove that a1 / k b1 / k and (a b)1/ k are equal. Let’s raise these numbers to the natural power k ,
if we get equal results, then (auxiliary1) these numbers are equal.
So (a1/ k b1/ k ) k (a1/ k ) k (b1/ k ) k a b and ((a b)1/ k ) k a b .
q
a aq
Consequence from [A]. For any positive a, b 0 and any q Q we have q .
b b
q
a
Proof. This equality is equivalent to b a the left part, according to [A], can be simplified
q q
b
q q
a a
as b b a .
q q
b b
For [B],[C] we fix an arbitrary pair of rational numbers q1 , q2 Q and their representations
m1 m
q1 , q2 2 || m1, m2 , k1, k2 .
k1 k2
Let’s prove [B]. a q1 q2 a m1k 2 k1m2 / k1k 2 (a1 / k1k 2 ) m1k 2 k1m2 and
a q1 a q2 a m1 / k1 a m2 / k2 (a1 / k1 ) m1 (a1 / k2 ) m2 . We need to show that (a1 / k1k 2 ) m1k 2 k1m2 [3] is equal
to (a1 / k1 ) m1 (a1 / k 2 ) m2 [4]. We have here some positive real numbers which are raised to some
integer powers. Let’s show that both these numbers are equal if they are raised to the
natural power k1 k2 . So (a )
1 / k1k 2 m1k 2 k1m2 k1 k 2
(a1 / k1k 2 ) k1 k 2
m1k 2 k1m2
a m1k 2 k1m2 and
(a ) (a1 / k 2 )m2
1 / k1 m1
k1 k 2
(a1 / k1 )k1 k 2 m1 (a1 / k 2 )k 2 k1 m2 (a1 / k1 )k1 k 2 m1
(a1 / k 2 )k 2
k1 m2
a k 2 m1 a k 2 m1 a m1k 2 k1m2 . We see that numbers [3] and [4] are equal if they are raised to
the natural power k1k 2 , then these numbers are equal.
102
Analysis
m1 m2 m1 m2
Let’s prove [C]. We have q1 q2 , then
k1 k2 k1 k2
a q1 q2 a m1 m2 / k1 k 2 / by def / (a1 / k1k 2 ) m1 m2 ((a1 / k1k 2 ) m1 ) m2 [5] and
(a q1 ) q2 (a m1 / k1 )m2 / k 2 (((a1 / k1 )m1 )1 / k 2 )m2 [6]. If we show that ((a1 / k1 ) m1 )1 / k 2 [7] and
(a1 / k1k 2 ) m1 [8] are equal, then expressions [5] and [6] are equal, and it is exactly what we need.
Both [7] and [8] are positive real numbers, if they become equal after they raised to the natural
power k 2 , then these numbers are equal. So [7] (((a1 / k1 ) m1 )1 / k 2 ) k 2 (a1 / k1 ) m1 a m1 / k1 and
[8] ((a1 / k1k 2 ) m1 ) k 2 (a1 / k1k 2 ) m1k 2 a m1k 2 / k1k 2 . And obviously a
m1 / k1
and a m1k 2 / k1k 2 are equal,
m1 m k m m~
because the fractions and 1 2 are equal (in the assertion1 we have proved that if ~,
k1 k1 k2 k k
~ ~
then a m / k a m / k ).
Theorem1. If a positive sequence of real numbers {xn } goes to a positive a, then for any rational
number q Q we nave {( xn ) q } a q .
m
Proof. Let’s fix any representation q || m , k .
k
[Step1] {xn } a , then {( xn )1/ k } a1/ k . In “radians” (assertion1) we showed that for any k
the function f ( x) k x continuous on [0,) . The sequence {xn } and a both belong to [0,) ,
as {xn } a , then { f ( xn )} f (a) {( xn )1/ k } a1/ k .
m
[Step2] Let’s take the representation q . If m is a natural number, then according to
k
the simpliest properties of sequence limits, we have
{( xn )1/ k } a1/ k {(( xn )1/ k ) m } (a1/ k ) m {( xn ) m / k } a m / k {( xn ) q } a q .
m
q . If m 0 , then q 0 and {( xn ) } {( xn ) } 1,1,1.... and
q 0
Let’s return to representation
k
a q a 0 1, then {( xn ) q } a q .
m
Let’s return to representation q . If m is a negative number, then m | m | , where | m | .
k
m
} a
m /k m /k
The number is a positive rational number and we already know that {( xn ) - here
k
we have a positive sequence that goes to some positive number, then:
1 1 m /k m /k
{( xn ) } a {( xn ) m / k } a m / k {( xn ) q } a q .
( xn )
m /k
a
m /k
103
Analysis
Assertion3. For any a R || a 0 we have lim n a1 / n 1.
Proof. Let a 1. Let’s denote n || n n a 1 0 n a 1 n . Then, according to the
a 1 a 1
Bernoulli inequality, (Book1,page 118) we have n , then finally 0 n which is
n n
a 1
equivalent to 0 n a 1 , let’s use here the squeeze theorem for sequences, we will get:
n
lim n (n a 1) 0 lim n n a 1 .
1
If a 1, then lim n n a lim n n 1 1 . And the last case: 0 a 1 a || b 1, we already
b
1 lim n n 1 1
know that lim n b 1. From here lim n a lim n
n n n 1.
b lim n n b 1
Consequence1. For any a R || a 0 we have lim n a 1/ n 1 .
1/ n 1 lim n 1 1
Proof. lim n a lim n 1/ n
1/ n
1.
a lim n a 1
Assertion4. For any a 1 the function a q is strictly increasing on Q , i.e.,
q1 q2 Q a q1 a q2 .
Proof. [part1] a 1 then for any pair of consecutive integer numbers n, (n 1) we have a n a n 1 .
Really: a n is a positive number and a n 1 [as 1 a] a n a a n1 .
Consequence2. For any integer numbers n1 n2 we have a n1 a n2 .
m1 m m m
[part2] Let q1 , q2 2 || m1, m2 , k1, k2 and 1 2 . Both a m1 / k1 and a m2 / k2 are
k1 k2 k1 k2
positive numbers, if they are both raised to the natural power k1 k 2 and we see that one of these
numbers is greater than the other, then it was true from the very beginning (auxiliary1).
As we showed above: (a m1 / k1 ) k1 k 2 a m1 k 2 [9] and (a m1 / k1 ) k1 k 2 a m1 k 2 [10], we know that
m1 m2
m1k2 m2k1 (these are integer numbers), then from the consequence2 we have
k1 k2
a m1 k 2 a m1 k 2 . So [9] is less than [10]. Then a m1 / k1 a m2 / k2 .
Assertion5. For any a 1 we have lim q 0||qQ a q 1.
Proof. Let’s fix an arbitrary rational sequence {qn } 0 , if we show that {a
qn
} 1 , then
lim q 0||qQ a q 1. We fix an arbitrary 0 . According to the assertion3 and consequence1 from
104
Analysis
it: lim n a 1 / n 1 and lim n a1 / n 1 . For 0 there exist the natural k such that n k
we have a 1/ n 1 and a1/ n 1 , let’s rewrite: 1 a 1/ n 1 [E1] and
[E2] 1 a1/ n 1 . As a 1, then for any n we have 1 a1 / n , from here follows that
1 1
1/ n 1 a 1/ n , then a 1 / n 1 a1 / n [E3]. From [E1] and [E2] and [E3] we have:
1 a
1 a 1/ n 1 a1/ n 1 || n k . We need to fix here any natural number n such that
1 a 1/ n 1 a1/ n 1 [E4].
Let’s take now the sequence {qn } . We take 1/ n , as {qn } 0 , then, starting from some number
m , there must be qn 0 1 / n 1 / n qn 1 / n - from this inequality and assertion4 follows
a 1 / n a qn a1 / n , then from [E4] we have 1 a n 1 a n 1 and it is true
q q
that
for any n m .
We had started from an arbitrary positive 0 , and we deduced that there exist some m , such
that n m we have a qn 1 , it means that {a qn } 1. Everything is proved.
We are ready to define the power function in the case a 1.
Def2. a 1 is a constant and x R is any real number. Let {qn } x is any sequence of rational
numbers that goes to x , then the sequence {a qn } converges and it’s limit must be taken as
a value a / by def / lim n a n .
x q
Assertion6. The def2 is correct. At first, any real number x can be represented as a limit of some
rational sequence (Book1, page 94, [Example L]). Secondly, if some rational sequence {qn } x ,
then {a qn } converges, and for any sequence {qn } that goes to x , the sequence {a qn } always goes to
the same limit. Let’s prove it. We fix any x and any sequence {qn } x .
qn
[1-st part] We will show that {a } is fundamental.
The sequence {qn } converges, therefore it is bounded, then there exist some rational
0 : qn n . Let’s fix an arbitrary 0 , according to the assertion5, lim q 0||qQ a q 1,
then for the positive number / a there exist 0 such that q :| q 0 | a q 1 / a [H].
And the final step: as {qn } converges, then {qn } is fundamental, then for 0 there exist k
starting from which m, n k we have qm qn . Let’s consider the absolute value a qm a qn
105
Analysis
for any m, n k . We have: a qm a qn a qn a qm qn 1 [S], here qm qn is a rational number,
and it’s absolute value is less than , then from [H] we have a qm qn 1 / a .
And a qn a qn [assertion 4 qn ] a . Then from [S] we have a qm a qn .
Let’s sum up: we had started from an arbitrary positive 0 and we deduced that there exist k
such that m, n k we have a qm a qn . It means that {a qn } is fundamental.
R is a complete field, then {a qn } converges in R to some number, which must be denoted as a x .
[2-nd part] Let’s show that for any other rational sequence {vn } x the sequence {a
vn
} goes to
the same limit ax (and therefore ax is uniquely defined). Let’s fix any other sequence {vn } x ,
and we have {qn } x . Let’s consider the sequence q1, v1, q2 , v2 , q3 , v3... , this sequence also goes to x.
As we have shown in the [1-st part] the sequence a q1 , av1 , a q2 , av2 , a q3 , av3 .... converges to some limit.
We already know that it’s subsequence a q1 , a q2 , a q3 , a q4 .... goes to a x . When a sequence converges
to some limit, any it’s subsequence converges to the same limit. Then a q1 , av1 , a q2 , av2 , a q3 , av3 ....
may converge only to ax (if we assume that it converges to some other number, we will immediately
get a contradiction). Then it’s subsequence a 1 , a 2 , a
v v v3
.... also goes to a x .
So, for any a 1 and x R , the number a x is defined. Therefore, for any concrete a 1 we have
the function f ( x) a . Let’s explore the properties of this function.
x
[1-st property]. For any a 1 the function a x is positive everywhere x (,) .
And for any positive x we have a x 1.
Proof. Let’s fix an arbitrary real number x and
pict.1
any rational sequence {qn } x .
Let’s fix some neighborhood (x) , it contains
all the terms of {qn } starting from some number
k . Let’s fix any rational number v “on the left”
of (x) [pict1], then we have v qn || n k .
Then (assertion4) we have a v a qn , then obviously a v lim n a qn a x .
And for rational number v we obviously have
0 av (because a is positive), then there must be
0 a x . So a x is positive everywhere (,).
Next, let’s fix any positive x and any rational pict.2
2
106
Analysis
sequence {qn } x . Let’s fix for example the neighborhood ( x) || x / 2 of x [pict2].
(x) contains all the terms of the sequence {qn } starting form some number k .
Then k : n k we have qn (x) . Let’s fix any rational v between 0 and x / 2 [pict2],
then 0 v qn || n k.From here, according to the assertion4, we have
a 0 a v a qn 1 a v a qn . From a v a qn || n k follows that a v lim n a qn a x .
Then 1 a a x 1 a x . Everything is proved.
v
[2-nd property]. For any a, b 1 and for any real x, x1 , x2 R we have:
[A] (a b) x a x b x , [B] a x1 a x2 a x1 x2 and [C] (a x1 ) x2 a x1 x2 .
Proof. Let’s fix any real x and any rational sequence {qn } x . Then a x lim n a qn and
b x lim n b qn . The element (a b) x (by definition) is a limit of the sequence (a b) qn .
We know that [A] is true for any rational powers, so we can write (a b) qn (a) qn (b) qn
and according to the standard properties of limits we have
(a b) x lim n (a b) qn lim n (a) qn (b) qn lim n (a) qn lim n (b) qn a x b x .
a x a, b 1
x
a
Consequence from [A]. x || .
b b x R
For the cases [B] and [C] we fix any rational sequences {qn } x1 and {vn } x2 .
Let’s prove [B]: a x1 lim n a qn and a x2 lim n a vn , as {qn } and {vn } go to x1 and x2 , then
the rational sequence {qn vn } goes to ( x1 x2 ) and a x1 x2 lim n a vn qn . We know that [B] is
true for any rational powers, and according to the standard properties of limits we have:
lim n a vn qn lim n (a vn a qn ) lim n a vn lim n a qn a x1 x2 a x1 a x2 .
Let’s prove [C]: a x 1 is a limit of the sequence {a qn } , so a x1 lim n a qn , then (a x1 ) x2 is a limit
q v q v q v q v4
of the sequence (lim n a n ) 1 , (lim n a n ) 2 , (lim n a n ) 3 , (lim n a n ) .... [S1].
Next, the rational sequence {qn vn } goes to x1 x2 , then a x1 x2 is a limit of the sequence
a q1 v1 , a q2 v2 , a q3 v3 , a q4 v4 ....... As [C] is true for any rational powers, then the last sequence can
be rewritten as (a q1 )v1 , (a q2 )v2 , (a q3 )v3 , (a q4 )v4 ... [S2].
We need to show that sequences [S1] and [S2] have the same limit. Then we need to show that
their difference is an infinitely small sequence.
(lim n a qn ) vk
Let’s denote k (a ) (lim n a ) , then k (a ) 1 .
qk vk qn vk qk vk
(a qk ) vk
a, b 0
q
a aq
According to the consequence from [A] (for rational powers) we have q || .
b b q Q
107
Analysis
Here we have lim n a qn 0 (follows from the [1-st property]) and also a qk 0 . Then we can
lim
vk
a qn [J]. When k the sequence lim n a is
qn
qk vk
rewrite: k (a ) 1 n
a qk a qk
lim a qn vk
positive and goes to 1 , then (theorem1) the sequence n goes to 1vk 1, then
qk
a
lim n a n
q vk
the sequence 1
qk vk
qk goes to zero 0 . And the sequence {(a ) } goes to some real
a
x1 x 2
number (it actually goes to a , as we can notice above). Then the sequence [J] goes to zero 0 ,
and the sequence { k } (which is [J]) is infinitely small. Everything is proved.
[3-rd property]. For any a 1 the function a x is strictly increasing and continuous everywhere
x (,) .
Proof. Let’s fix an arbitrary pair of real numbers x1 x2 . Then ( x2 x1 ) 0 is a positive real
x 2 x1
number. According to the [1-st property] we have a 1, let’s multiply both part by a x1 0
(it is positive according to the [1-st property]). Then
a x2 x1 a x1 1 a x1 [2 nd property] a ( x2 x1 ) x1 a x1 a x2 a x1 , then a x is strictly
increasing on (,).
Let’s fix now an arbitrary point x0 (,) . If we show that for any sequence {xn } x0 we have
{a xn } a x0 , then a x is continuous at x0 (by definition). Let’s fix an arbitrary sequence {xn } x0 .
For every term x n , we can find the rational number q n such that xn qn 1 / n [L], from here
immediately follows that {qn } x0 . Then a x0 lim n a qn (by definition of a x0 ).
We want to show that the sequence {a xn } also goes to a x0 . We already know that {a q n }
goes to a x0 , then it’s enough to show that the difference of these sequences {a xn } {a qn } is an
infinitely small sequence (then they go to the same limit). Let’s consider {a xn a qn } . And let’s
estimate a xn a qn [P]. From [L] we have qn 1 / n xn qn 1 / n , as a x is strictly increasing,
qn 1 / n
then a a xn a qn 1 / n , let’s subtract a q n from all sides, in order to get the estimation of [P]:
a qn 1 / n a qn a xn a qn a qn 1 / n a qn [P1] and now we will just use the squeeze theorem for
sequences. Let’s show that {a n
q 1/ n
a qn } is infinitely small, we represent a qn 1/ n a qn a qn (a1/ n 1) .
When n the sequence {a1 / n } goes to 1 (assertion3), then {a1 / n 1} goes to 0 .
And the sequence {a q n } goes to some real number (which is a x0 ), then the sequence {a qn (a1 / n 1)}
goes to zero. So, the sequence {a q n 1 / n a qn } is infinitely small. Similarly, in [P1] the sequence
108
Analysis
{a qn 1 / n a qn } is infinitely small. Then from [P1], the sequence {a xn a q n } is infinitely small.
Everything is proved.
Let’s extend the definition of a power function.
Def. For a 1 we define 1x 1 || x R . For any a (0,1) we take the representation
1 1
a || b 1 (which is unique) and we define a x x || x R .
b b
This definition is correct, in the [1-st property] we showed that b || b 1 is positive for any
x
x (,) , then a x || a 1 is really defined on (,) .
From this definition immediately follows that for any concrete a (0,1) the function a x is strictly
decreasing, continuous and strictly positive everywhere, and the [2-nd property] is true for this
function. For a 1 the power function is a constant function (and in this case it is also continuous,
strictly positive everywhere and the [2-nd property] is true). Now the power function a x 1 is
defined for every concrete a 0 [pict3]. And we know it’s properties.
pict.3
Logarithmic function
Let’s fix any a 1. We know that a x is strictly increasing and continuous on (,) .
Exercise1. Show that lim x a x and lim x a x 0 .
Exercise2. Show that a x is one-to-one mapping (,) (0,) .
Exercise3. By using the inverse function theorem show that there exist the inverse function
f 1 ( x) : (0,) (,) which is also strictly increasing and continuous on (0,) .
Def. The inverse function from the exercise3 is called a logarithmic function and we denote it
log a x f 1 ( x) .
So, log a x is defined on (0,) and it is strictly increasing and continuous on (0,) .
The simple equality: a
loga x
x || x 0 (which is exactly f ( f 1 ( x)) x ) shows us a very simple
and important meaning of a logarithmic function: log a x is the power, to which a must be raised in
order to become equal to x.
109
Analysis
When we see the writing log a x , we usually say that a is a base of the logarithm log a x .
And log a x can be called a “logarithm of x to the base a ”.
It’s easy to prove that for any a 1, b 1 and for any x1, x2 (0,) the next properties are true:
[1] log a a x x [2] log a ( x1 x2 ) log a ( x1 ) log a ( x2 ) [3] log a xb b log a x || b R
log b x
[4] log a b x (1/ b) log a x || b R, b 0 [5] log a b log b a 1 [6] log a x .
log b a
We can also consider the power function a x where a (0,1) and we can perform exactly the same
reasoning as above (the only difference here is that a x is strictly decreasing, then log a x must be
also strictly decreasing), and we can get exactly the same properties [1]-[5] as above.
Def. The logarithm log e x , where the base is an exponent e , is called a “natural logarithm” and
we denote it ln x log e x .
Upper and lower limit
We have already defined the notion “subsequence”, and we have proved that any bounded sequence
has a convergent subsequence [T]. We have also noticed the next obvious thing: if a sequence
converges to some limit, then any it’s subsequence converges to the same limit.
Def. Let we have some sequence {xn } and it has a subsequence {xnk } that goes to a .
Then a is called a partial limit of the sequence {xn }.
From now on we consider only bounded sequences {xn } . For any bounded sequence, the set ,
which consists of all partial limits of this sequence, is not empty. (We have proved [T] that
contains at least one number). And is a bounded set. Really, {xn } is bounded, therefore all it’s
terms belong to some segment [a, b] and any partial limit of this sequence must also belong to [a, b].
Then [a, b] and therefore the numbers inf and sup are defined.
Assertion1. The numbers inf and sup are also partial limits of {xn } , i.e., there exist some
subsequence {xnk } inf and some subsequence {x pk } sup .
Proof. Let’s fix any positive sequence { k } 0 . As inf is an infinum of , the segment
[inf , inf 1 / 2) contains some partial limit p1 , as p1 is a limit of some subsequence, the set
( p1 1 / 2, p1 1 / 2) contains all the terms of that subsequence, starting from some number,
let’s just fix any term which belongs to ( p1 1 / 2, p1 1 / 2) , it becomes the first term xn1 of our
subsequence. Obviously xn1 inf 1 . The same thing must be done for the number 2 , and we
will find xn2 such that xn inf 2 (It’s important to notice here that we must choose n2 n1 )
2
110
Analysis
and etc. As a result we have the subsequence {xn } : xn inf k and { k } 0 ,
k k
then {xnk } inf . Similarly we can find the subsequence {x pk } sup .
Def. The numbers inf and sup are called a lower and an upper limit of the sequence {xn } and
we denote them as lim n xn inf and lim n xn inf . From the assertion1 follows that we
can say “ lim n xn is a minimal among all partial limits of {xn } ” and “ lim n xn is a maximal
among all partial limits of {xn }”.
Assertion2. The upper limit lim n xn has the next property: for any positive 0 there exist
k , starting from which n k xn lim n xn . And the lower limit lim n xn inf has
the similar property: for any positive 0 there exist k , starting from which
n k xn lim n xn .
Proof. Let’s assume that it’s not true, it means that there exist some “bad” 0 for which there is
no any appropriate k , i.e., for any concrete k there exist at least one n k such that
xn lim n xn . So, for k 1 there exist n1 k 1 such that xn1 lim n xn . For k n1
there exist n2 k n1 such that xn2 lim n xn . For k n2 there exist n3 k n2 such that
xn3 lim n xn and etc. Then we have the subsequence {xnk } , and every it’s term is not less
than lim n xn . As the initial sequence {xn } is bounded, then {xnk } is also bounded, then it has
a convergent subsequence {x pk } and {x pk } is also a subsequence of {xn } , and every it’s term is not
less than lim n xn , then it’s limit is also not less than lim n xn , then this limit is greater
than lim n xn , and we have a contradiction (because lim n xn is a maximal among all partial
limits). This contradiction proves our assertion for an upper limit.
For lim n xn the proof is similar.
Assertion3. The sequence {xn } converges to some limit a lim n xn lim n xn a .
Proof. Let {xn } converges, so lim n xn a then any it’s subsequence goes to the same limit a,
in particular, the sequences {xnk } and {x pk } (from the assertion1) both go to a,
then inf sup a lim n xn lim n xn a .
Conversely We have lim n xn lim n xn a . Let’s fix an arbitrary positive 0 , according to
the assertion2 there exist k such that n k we have xn lim n xn xn a and there
exist m such that n m we have xn lim n xn a xn . Let’s fix the maximal
k max(m, k ) , then starting from k , n k we have a xn a xn (a) .
111
Analysis
So, for any positive 0 , there exist the number k , starting from which every term of
the sequence {xn } belongs to (a) , it means that lim n xn a .
Let’s consider now the case of an unbounded sequence {xn } .
Assertion4. If non-negative (in particular positive) sequence {xn } || xn 0 n is unbounded, then
it has some subsequence {xnk } which goes to . If non-positive (in particular negative) sequence
{xn} || xn 0 n is unbounded, then it has some subsequence {xnk } which goes to .
Proof. Let’s fix any non-negative sequence {xn } || xn 0 n . As it is unbounded, for any positive
0 there exist some number k such that xk . Then for 1 there exist n1 such that
xn1 1 , for 2 there exist n2 n1 such that xn2 2 . (if there is no such n2 , then all the terms
of {xn } starting from the number n1 are not greater than 2 and there are also n1 first terms
x1, x2 .... xn1 which can be any real numbers. Such sequence is bounded, and we have
a contradiction). Then for 3 we can find n3 n2 such that xn3 3 and etc. As a result we have
the subsequence {xnk } such that xnk k || k , then {xnk } .
Let now we have a non-positive unbounded sequence {xn } , then { xn } is a non-negative unbounded
sequence, and, as we showed above, it has some subsequence { xnk } , then {xnk } and
{xn k } is a subsequence of the initial sequence {xn } .
Consequence (from the assertion4). For any non-negative sequence {xn } only one of the next two
cases is true:
[1] {xn } is bounded and therefore (according to the assertion1) the upper limit lim n xn is
a concrete real number.
[2] {xn } is unbounded and therefore it has some subsequence {xnk } which goes to .
In this case, by definition, we write lim n xn .
This consequence has a great importance for serieses and power serieses, both will be considered in
the next 3-rd book.
We have proved enough for practical purposes.
112
Analysis
Derivative
f ( x) f (a)
Def. f (x) is defined on some neighborhood R (a) of a , then the function is defined
xa
f ( x) f (a )
on the deleted neighborhood DR (a) , if the limit lim x a exists, then it is called
xa
a derivative of f (x) at the point a , and we denote f x (a) . And f is called differentiable at a
in this case.
Comment. In many books almost immediately after a derivative is defined, the “geometrical
meaning” of it is described like a slope of a tangent line. But there is never a definition, what is a
tangent line actually is. There is the chapter in this book about curves, and there will be a normal
definition of a tangent line, and then we will formulate normally the geometrical meaning of a
derivative. Now a derivative is a concrete limit of a concrete function.
Assertion1. f is defined on R (a) and f is differentiable at a the next representation is
possible on R (a) : f ( x) f (a) ( x a) ( x) ( x a) || x R (a) , where f x (a) is
a constant and (x) is infinitely small when x a and continuous at x a
(i.e., lim x a ( x) 0 (a) ).
Proof. Let f is differentiable at a , it means that the next limit exists
f ( x) f (a) f ( x) f ( a)
lim x a f x (a) . The function f x (a) // by def // ~( x) [T] is defined
xa xa
on DR (a) and the limit of the left part of [T] when x a is equal to zero, then the same is true for
~( x) 0 . From [T] follows that x D (a) the next
the right part of [T], then lim x a R
~( x) ( x a) [E]. And now we have the
representation is true f ( x) f (a) f x (a) ( x a)
representation [E] which is almost what we need, but this equality is true on DR (a) , and we can’t
~( x) isn’t defined at the point x a (really, look at
even substitute the value x a in [E], because
the definition [T] of this function). But remember that we are proving now that the needed
representation [E] is possible, then in our power to find any appropriate function (x) which
satisfies to [E]. Let’s do it. We define the new function (x) on R (a) : x DR (a) ( x) ~( x)
and (a) 0 . Then the representation [E1] f ( x) f (a) f x (a) ( x a) ( x) ( x a) is true
for any x R (a) and lim x a ( x) 0 (a) . Everything is proved.
Conversely Let f ( x) f (a) ( x a) ( x) ( x a) || x R (a) , then the same
representation is true on DR (a) , so f ( x) f (a) ( x a) ( x) ( x a) || x DR (a)
f ( x) f (a)
let’s divide all the sides by ( x a) , then ( x) the right part has a limit A
xa
when x a , then the left part also has a limit when x a , and we have
f ( x) f (a )
lim x a , then f is differentiable at a and f x (a) , everything is proved.
xa
113
Analysis
Consequence1. f is differentiable at a f is continuous at a.
Proof. Let’s use the representation from above:
f ( x) f (a) f x (a) ( x a) ( x) ( x a) || x R (a) [V] every summand on the right side has
a limit when x a . So lim x a f (a) f (a), lim x a f x (a) ( x a) 0 (because f x (a) const ),
lim x a ( x) ( x a) 0 , then the right part of [V] has a limit f (a) when x a , then the left part
of [V] has the same limit f (a) when x a , i.e., lim x a f ( x) f (a) and it means that f is
continuous at a.
Assertion2. If f , g are differentiable at a. Then
[A] f ( x) g ( x) is differentiable at a , and it’s derivative at a is a sum of derivatives f x (a) g x (a)
[B] f ( x) g ( x) is differentiable at a , and it’s derivative at a is f x (a) g (a) g x (a) f (a) .
[C] If g (a) 0 , then f ( x) / g ( x) is differentiable at a , and it’s derivative at a is
f x (a) g (a) g x (a) f (a)
.
[ g (a)]2
f ( x) f (a)
Proof. f , g are differentiable at a , so both limits lim x a f x (a) and
xa
g ( x) g (a )
lim x a g x (a) do exist. f is defined on some R (a) and g is defined on some P (a) ,
xa
let’s fix the smaller of these neighborhoods R (a), P (a) (let it be R (a) ) both functions f , g are
defined on R (a) , then f g and f g are defined on R (a) , and therefore we can raise a
question about their derivatives at a.
If the next limit lim x a
f ( x) g ( x) f (a) g (a) exists, then, by definition, it is a derivative of
xa
f ( x) g ( x) at a. So,
f ( x) g ( x) f (a) g (a) f ( x) f (a) g ( x) g (a) the right part
xa xa xa
has a limit f ' (a) g ' (a) when x a , then the left part has the same limit f ' (a) g ' (a) when
x a.
f ( x) g ( x) f ( a ) g (a )
Next, if the next limit lim x a exists, then, by definition, it is a derivative
xa
of f ( x) g ( x) at a. Let’s rewrite: [1]
f ( x) g ( x) f ( a ) g ( a ) f ( x) g ( x) f ( x) g (a ) f ( x) g ( a ) f ( a ) g ( a )
xa xa
f ( x) ( g ( x) g (a)) g (a) ( f ( x) f (a)) ( g ( x) g (a)) ( f ( x) f (a))
f ( x) g (a) .
xa xa xa
The right part has a limit f x (a) g (a) g x (a) f (a) when x a . Then the left part [1] has
the same limit when x a and [B] is proved.
114
Analysis
In the case [C] as g (a) 0 and g is continuous at a (because g is differentiable at a ) we can fix
the neighborhood (a) of a , on which g (x) has the same sign as a value g (a) , then
g ( x) 0 x (a) . As f (x) is defined on R (a) and g (x) is defined on (a) and
g ( x) 0 x (a) , we take the smaller of these neighborhoods (let it be (a) ).
Then f ( x) / g ( x) is defined on (a) and we can raise a question about it’s derivative at a.
f ( x) / g ( x) f ( a ) / g ( a )
If the limit lim x a exists, then, by definition, it is a derivative of
xa
f ( x) / g ( x) at a . Let’s rewrite:
f ( x) / g ( x) f (a ) / g ( a ) f ( x) g ( a ) g ( x) f ( a ) 1
[2]
xa xa g ( x) g ( a )
f ( x) g (a ) f (a ) g (a ) f ( a ) g (a ) g ( x) f (a ) 1
xa g ( x) g ( a )
f ( x) f ( a ) g ( x) g ( a ) 1
g (a) f (a) this expression obviously has a limit
xa x a g ( x) g (a )
g (a) f x (a) f (a) g x (a) 21 when x a , then [2] has the same limit when x a and [C]
g (a)
is proved.
In addition: [D] f is differentiable at a and R is a constant, then f (x) is differentiable
at a and it’s derivative at a is f x (a) .
Def. For any functions f , g and any constants , , the function f g is called a linear
combination of functions f , g .
Consequence2. If f , g are differentiable at a , then any linear combination of these functions
f g is differentiable at a and it’s derivative at a is f x (a) g x (a) . In particular, for
1, 1 the difference f is differentiable at a , and it’s derivative at a is f x (a) g x (a) .
Def. Next symbols can be used in order to denote a derivative of f at a point a:
f f ( x)
f x (a), f ' (a), (a), |x a .
x x
By using these designations, the previous result (assertion2+consequence2) can be written as:
( f ( x) g ( x)) ( f ( x) g ( x))
|x a f x (a) g x (a) and | x a f (a) g x (a) f x (a) g (a)
x x
( f ( x) / g ( x)) f (a) g (a) g x (a) f (a)
|x a x .
x g 2 (a)
115
Analysis
Derivative of a composite function: x x(t ) is differentiable at t 0 and f f (x) is
differentiable at x0 x(t0 ) . Then the composite function f ( x(t )) is differentiable at t 0 and it’s
derivative is ft ( x(t0 )) f x ( x0 ) xt (t0 ) .
Proof. x(t ) is differentiable at t 0 , then on some (t0 ) the next representation is true
x(t ) x(t0 ) (t t0 ) (t ) (t t0 ) || t (t0 ) [3]. And f (x) is differentiable at x0 ,
then on some ( x0 ) the next representation is true
f ( x) f ( x0 ) ( x x0 ) ( x) ( x x0 ) || x ( x0 ) [4].
The function x(t ) is continuous at t 0 (because it is
differentiable at t 0 ), then we can squeeze
(if necessary) the neighborhood (t0 ) [pict1],
in order to be: t (t0 ) the value x(t ) belongs to
( x(t0 )) ( x0 ) . Suppose we have already
squeezed the neighborhood (t0 ) .
So, for any t from (t0 ) the representation [3] is
true, and the value x(t ) belongs to ( x0 ) , then for
pict.1
x(t ) the representation [4] is true, then t (t0 )
we have f ( x(t )) f ( x0 ) ( x(t ) x0 ) ( x(t )) ( x(t ) x0 ) we can substitute here x0 x(t0 ) ,
so f ( x(t )) f ( x(t0 )) ( x(t ) x(t0 )) ( x(t )) ( x(t ) x(t0 )) || t (t0 ) [5] and finally,
we substitute the representation [3] of difference x(t ) x(t0 ) in [5]. Then t (t0 ) we have
f ( x(t )) f ( x(t0 )) (t t0 ) (t ) (t t0 ) ( x(t )) (t t0 ) (t ) (t t0 ) it may look
quite voluminous, but let’s just regroup the summands on the right side:
f ( x(t )) f ( x(t0 )) (t t0 ) (t ) (t t0 ) ( x(t )) (t t0 ) (t ) (t t0 )
f ( x(t )) f ( x(t0 )) (t t0 ) (t ) ( x(t )) (t ) (t t0 ) [6].
Here is a concrete constant (number), and the second expression in the square brackets .....
is an infinitely small function when t t0 and this expression is also a continuous at t 0 function.
Really, let’s consider: (t ) ( x(t )) (t ) [R]. We obviously have here
lim t t0 (t ) lim t t0 (t ) 0 . Next, acc. to the theorem3 [Continuity of a composite function]:
x(t ) is continuous at t 0 and x(t0 ) x0 , and (x) is continuous at x0 , then the composite function
( x(t )) is continuous at t 0 , it means that lim t t0 ( x(t )) ( x(t0 )) ( x0 ) 0 and finally,
lim t t0 (t ) . From here follows that the limit of [R] when t t0 is zero, then [R] is
an infinitely small function when t t0 . All the separate functions from [R] are continuous at t 0 ,
then [R] is continuous at t 0 . Then we can write (t ) ( x(t )) (t ) (t ) where (t )
is infinitely small when t t0 and continuous at t 0 . Then we can rewrite [6] as:
116
Analysis
f ( x(t )) f ( x(t0 )) (t t0 ) (t ) (t t0 ) - it is the representation of difference
f ( x(t )) f ( x(t0 )) of a composite function f ( x(t )) on the neighborhood (t0 ) , and we see that
this representation is exactly the same as we had in the assertion1.
Then f ( x(t )) is differentiable at t 0 and it’s derivative at t 0 is . From [3],[4] we see that
f x ( x0 ) and xt (t0 ) . So ft ( x(t0 )) f x ( x0 ) xt (t0 ) .
f ( x) f (a)
Def. f (x) is defined on some half-interval (a R, a] , then is defined on (a R, a) .
xa
f ( x) f ( a )
If there exist the next left limit at a, lim x a , then is called a left derivative
xa
of f at a, and we denote it f ' (a) . In this case we say that f is left differentiable at a.
f ( x) f (a)
Def. f (x) is defined on some half-interval [a, a R) , then is defined on (a, a R) .
xa
f ( x) f (a )
If there exist the next right limit at a, lim x a 0 , then is called a right
xa
derivative of f at a, and we denote it f ' (a) . In this case we say that f is right differentiable
at a.
For the left/right derivative it’s easy to prove the assertions that are exactly similar to
the assertion1,assertion2 and consequence1, consequence2 (formulate them). Even the theorem
about a derivative of a composite function can be extended: the function x(t ) is left/right
differentiable (or just differentiable) at t 0 and f (x) is left/right differentiable (or just differentiable)
at x0 x(t0 ) . And f (x) is defined on every value x(t ) , then f ( x(t )) differentiable at t 0 in exactly
the same way (left/right/ordinary) as x(t ) is differentiable at t 0 . And the formula
ft ( x(t0 )) f x ( x0 ) xt (t0 ) is true (there may be right/left/ordinary derivatives in this formula).
From the basic properties of function limits we have:
f is differentiable at a f is left differentiable at a and f is right differentiable at a, and left
and right derivatives at a are equal f ' (a) f ' (a) f ' (a) .
Def. f (x) is defined on some neighborhood R (a) . We say that f (x) has a local maximum at a
if f (a) f ( x) x DR (a) , in particular a strict local maximum if f (a) f ( x) x DR (a) .
And we say that f (x) has a local minimum at a if f (a) f ( x) x DR (a) , in particular
a strict local minimum if f (a) f ( x) x DR (a) . If f has a local minimum or a local maximum
at a, then we say that f has a local extremum at a.
117
Analysis
Lemma1. f has a local extremum at a and f is differentiable at a, then f x (a) 0 .
f ( x) f (a )
Proof. f (x) is defined on R (a) . The next limit exists lim x a . Let 0 , then
xa
f ( x) f (a)
we can fix some deleted neighbourhood D (a) , such that x D (a) the value [J]
xa
is positive. Let x a in D (a) , then x a 0 and in order [J] to be positive there must be
f ( x) f (a) . And similarly, when x a in D (a) , there must be f ( x) f (a) (so on the right
of a we have f ( x) f (a) , and on the left of a we have f ( x) f (a) ). Then there can’t be any
local extremum at a. And we have a contradiction. Then the assumption 0 was false.
The similar contradiction will appear if we assume 0 . Then 0 .
Def. f is defined on [a, b] . We say that f is differentiable on [a, b] if f is differentiable in
the ordinary way at any point x0 (a, b) and f is right differentiable at a and left differentiable
at b .
Obviously, if f is differentiable on [a, b] , then f is continuous on [a, b] .
Roll’s theorem. f is differentiable on [a, b] and f (a) f (b) . Then there exist x0 (a, b) such
that f x ( x0 ) 0 .
Proof. f is differentiable on [a, b] f is continuous on [a, b] . From the Weierstrass theorem
follows that it reaches it’s maximal and minimal values at some points с, d [a, b] .
c is a point of local extremum of f (x) , then (lemma1) f x (c) 0 and c is
If c (a, b) , then
the point we need c x0 . And if d (a, b) we get exactly the same result. If c (a, b) and d (a, b) ,
then c, d are the end points of the segment [a, b] . As f (a) f (b) , then f (c) f (d ) and
the maximal and minimal values of f on [a, b] are equal, then f is a constant function on [a, b] ,
so f const on [a, b] , for such function it’s derivative is equal to zero at any point x0 (a, b) ,
so any point x0 (a, b) is appropriate.
Cauchy formula. f and g are differentiable on [a, b] and g x ( x0 ) 0 x0 (a, b) .
f (b) f (a) f x ( x0 )
Then there exist x0 (a, b) such that: .
g (b) g (a) g x ( x0 )
Proof. Let’s define the auxiliary function ( x) [ f (b) f (a)] g ( x) [ g (b) g (a)] f ( x) - it is
a linear combination of differentiable on [a, b] functions, therefore (x) is differentiable on [a, b] .
As (x) satisfies to the Roll’s theorem, there exist x0 (a, b) such that x ( x0 ) 0 .
Both [ f (b) f (a)] and [ g (b) g (a)] are constants, then the derivative x ( x0 ) is
x ( x0 ) [ f (b) f (a)] g x ( x0 ) [ g (b) g (a)] f x ( x0 ) , and we have
[ f (b) f (a)] g x ( x0 ) [ g (b) g (a)] f x ( x0 ) 0 , now we can divide both sides by g x ( x0 ) 0 ,
and we will get the Cauchy formula.
118
Analysis
Consequence1 [the mean value theorem]. f is differentiable on [a, b] .
Then there exist x0 (a, b) such that f (b) f (a) f x ( x0 ) (b a) .
Proof. Let’s take the pair of functions f ( x), g ( x) x on [a, b] , the Cauchy formula can be used here,
obviously g x ( x0 ) 1 for any x0 (a, b) , then we have:
f (b) f (a) f x ( x0 )
f (b) f (a) f x ( x0 ) (b a) .
ba 1
Comment. It’s very important to notice that both the Cauchy formula and the mean value theorem
can be applied even in the case when b a . So we shouldn’t care and check all the time that b is
greater than a . If f is differentiable on some segment with ends a, b , then the mean value
theorem f (b) f (a) f x ( x0 ) (b a) is true. Here x0 is some point in the interval with ends a, b .
And the same for the Cauchy formula (explain why).
Consequence2 from [the mean value theorem]. f is differentiable on [a, b] . If f x ( x0 ) 0 for any
x0 (a, b) , then f ( x) const on [a, b] . If f x ( x0 ) 0 for any x0 (a, b) , then f is strictly
increasing on [a, b] . If f x ( x0 ) 0 for any x0 (a, b) , then f is strictly decreasing on [a, b] .
Proof. Let’s fix any x1 x2 || x1, x2 [a, b] . As f is differentiable on [ x1, x2 ] then (the mean value
theorem) there exist x0 ( x1, x2 ) such that f ( x2 ) f ( x1 ) f x ( x0 ) ( x2 x1 ) [V], where x0 is some
point from the interval ( x1 , x2 ) . If f x ( x0 ) 0 || x0 (a, b) , then f ( x2 ) f ( x1 ) and it is true for
any x1, x2 [a, b] , then f is a constant function on [a, b] . Similarly, if f x ( x0 ) 0 || x0 (a, b) ,
then the right part of [V] is positive, then f ( x2 ) f ( x1 ) and f is strictly increasing on [a, b] .
And similarly for the other case.
Derivative of an inverse function. y f (x) is defined on [a, b] and f x ( x) 0 on [a, b] .
Then the inverse function ( y) || y [ f (a), f (b)] is differentiable at any point f ( x0 ) and
y ( f ( x0 )) 1/ f x ( x0 ) .
Proof. As f x ( x) 0 on [a, b] , then (consequence2) f (x) is strictly increasing on [a, b] also f is
continuous on [a, b] (because f is differentiable on [a, b] ), then f is one-to-one mapping
[a, b] [ f (a), f (b)] , and the inverse function : [ f (a), f (b)] [a, b] is defined (we showed it in
the inverse function theorem), and ( y ) is also strictly increasing on [ f (a), f (b)] and continuous
on [ f (a), f (b)] . Let’s fix an arbitrary point y0 [ f (a), f (b)] , there exist the unique x0 [a, b]
such that y0 f ( x0 ) .
( y) ( y0 )
We are interested in the next limit lim y y0 || y[ f ( a ), f (b )] [L].
y y0
119
Analysis
Let’s fix an arbitrary sequence { yn } [ f (a), f (b)] || yn y0 n || { yn } y0 . If we show that
( yn ) ( y0 ) 1 1
, then (by definition) the limit [L] is equal to , and it is exactly
y n y0 f (
x 0x ) f (
x 0x )
what we need.
( yn ) ( y0 )
We consider the sequence , every y n can be uniquely represented as yn f ( xn ) ,
y n y0
( yn ) ( y0 ) ( f ( xn )) ( f ( x0 )) xn x0
then [L1]. Let’s notice that
yn y0 f ( xn ) f ( x0 ) f ( xn ) f ( x0 )
xn x0 n (because if some xk x0 , then f ( xk ) f ( x0 ) yk y0 , which contradicts to the choice
1
of the sequence { y n } ). Then [L1] can be rewritten as
f ( xn ) f ( x0 )
[L2]. What is a limit of the
xn x0
sequence [L2]? We have { yn } y0 and is continuous at y 0 , then
{ ( yn )} ( y0 ) {xn } x0 .
f ( x) f ( x0 )
As f (x) is differentiable at x0 , then the limit lim n f x ( x0 ) 0 exists, and for
x x0
f ( xn ) f ( x0 ) 1
the sequence {xn } x0 we have f x ( x0 ) , then the sequence [L2] goes to ,
xn x0 f (
x 0x )
everything is proved.
Comment. The same theorem is true when f x ( x) 0 on [a, b] (the proof is similar).
f ( x) f (a)
Let f (x) is defined on R (a) , then is defined on DR (a) , and if this function has
xa
a limit at a , then it is a derivative f x (a) . In practice it’s usually more convenient to calculate the
f ( x) f (a) f (a x) f (a)
other limit, let’s denote x x a , then x a x and we have .
xa x
We understand intuitively that when x goes to a , the difference x goes to zero, and we can
f (a x) f (a)
calculate the limit lim x 0 in order to find f x (a) . Let’s formulate it as:
x
f (a x) f (a)
Exercise1. If f (x) is defined on R (a) , then (x) (a function of a variable x )
x
f (a x) f (a) f ( x) f (a )
is defined on DR (0) . If the limit lim x 0 exists, then the limit lim x a
x xa
also exists and these limits are equal.
120
Analysis
sin x
Exercise2. Prove that for any a R we have |x a cos a .
x
sin( a x) sin a
Solution. Let’s fix any a R . We are interested in the next limit lim x 0
x
let’s use the formula for difference of sines:
a x a a x a cos a x sin x
2 cos sin
sin( a x) sin a 2 2 2 2
[U].
x x x
2
x
sin
sin x 2 1 and
We have proved earlier that lim x 0 1, from here follows that lim x 0
x x
2
x
obviously lim x 0 cos a cos a . Then the limit [U] exists and it is equal to cos a .
2
cos x
Exercise3. Prove that for any a R we have |x a sin a .
x
sin x cos x
Consequence. As we have tg ( x) , ctg ( x) , then by using the formula
cos x sin x
( f ( x) / g ( x)) f (a) g (a) g x (a) f (a)
|x a x it’s easy to derive that at any point a , where tgx is
x g 2 (a)
tgx 1
defined, we have: |x a and at any point a , where ctgx is defined, we have
x sin 2 a
ctgx 1
|x a 2 .
x cos a
Exercise4. arcsin y is defined on [1,1] , prove that for any y0 (1,1) we have
arcsin y 1
| y y0 .
y 1 ( y0 )2
Solution: Let’s consider the function sin x on the interval , , for any x0 , we have
2 2 2 2
sin x
|x x0 cos x0 0 , then, according to [derivative of an inverse function], the inverse function
x
x arcsin y is differentiable on (1,1) and for any y0 (1,1) || y0 sin x0 we have
arcsin y 1 1 1 1
| y y0 [as cos x0 0] .
y sin x cos x 1 sin x0
2
1 ( y0 ) 2
| x x0 0
x
121
Analysis
arccos y 1
Exercise5. Show that for any y0 (1,1) we have | y y0 .
y 1 ( y0 ) 2
arctgy 1 arcctgy 1
And also, for any y0 R we have | y y0 and | y y0 .
y 1 ( y0 ) 2
y 1 ( y0 )2
At this moment we can’t find derivatives of logarithmic and power functions.
At first we need to derive several auxiliary limits.
Theorem1. lim x 0 1 x 1/ x e .
n
1
Proof. We already have: lim n 1 e (it is a definition of e)
n
Let’s show at first that the right limit lim x 0 1 x
1/ x
is equal to e , and then that the left limit
lim x 0 1 x is also equal to e , then the ordinary limit lim x 0 1 x
1/ x 1/ x
exists and equal to e .
Let’s fix an arbitrary sequence {xn } (0,) || xn 0 n || {xn } 0 if we show that
{1 xn } e , then there must be lim x 0 1 x e . Shortly speaking, {xn } is a positive
1 / xn 1/ x
sequence that goes to zero.
As {xn } 0 , then, starting from some number k , there must be: m k : 0 xm 1 , then we can
discard first k terms x1, x2 .... xk from {xn } which do not necessary belong to (0,1) . So, we can
assume that from the very beginning we have the sequence {xn } 0 such that 0 xn 1 || n ,
then we have 1/ xn 1 || n . For every real number 1 / xn there exist the pair of natural numbers
mn 1, mn such that mn 1/ xn mn 1 [S]. We obviously have {1/ xn } , then from [S] the
sequence of natural numbers {mn 1} also goes to .
1
mn 1
1
n 1
mn 1
Then 1 is a subsequence of 1 and therefore 1 e [V1].
mn 1 n mn 1
1
mn
As {mn 1} , then also {mn } and therefore 1 e [V2].
m n
It’s easy to get the next estimation for 1 xn
mn
, by using [S] and the basic properties of power
mn mn 1
1 1
function: 1 1 xn 1
1 / xn
. From [V1] and [V2] follows that the outer
mn 1 mn
sequences here both go to e , then from the squeeze theorem for sequences follows that 1 xn
1 / xn
also goes to e .
122
Analysis
Let’s show now that lim x 0 1 x e . We fix again an arbitrary sequence
1/ x
{xn } (,0) || xn 0 n || {xn } 0 we need to show that {1 xn
1 / xn
} goes to e .
Similarly, as we assumed earlier, we can assume now that 1 xn 0 || n .
1
1 / | xn | 1 / | xn | n
1| xn | 1| x |
1 | xn | | xn |
1 xn
1 / xn
1 | xn |
1 / | xn |
1
1 1
| xn |
[Y]
1 | xn |1 / | xn |
1 | xn
| 1 | xn
| 1 | xn
|
| xn |
Let’s denote now yn - it is a positive sequence, as {| xn |} goes to zero, then { y n } also goes
1 | xn |
to zero, and we can rewrite [Y] as
1 y
n
1 / y n y n 1
1 yn
( y n 1) / y n
1 yn
11 / y n
(1 yn ) 1 yn
1/ yn
[Y1]. We obviously have
(1 yn ) n1 and, as we proved above: {1 yn } n e (because { y n } is a positive
1/ yn
sequence that goes to zero), then [Y1] goes to e when n . Everything is proved.
ln(1 x) ax 1
Theorem2: [A] lim x 0 1 and [B] For any a 0 we have lim x 0 ln a .
x x
ln(1 x)
Proof. Let’s prove [A]. The function is defined on (1,0) (0,) , we fix an arbitrary
x
sequence {xn } (1,0) (0,) || xn 0 n || {xn } 0 from the theorem1 follows that
{(1 xn )1/ xn } e . The function ln( x) is continuous at e (because it is continuous on (0,) ),
ln(1 xn )
then there must be {ln((1 xn ) )} ln e 1 and [A] is proved.
1 / xn
x n
ax 1
Let’s prove [B]. We fix an arbitrary positive a 0 , the function
is defined on (,0) (0,) ,
x
let’s take an arbitrary sequence {xn } 0 from this set. If a 1, then [B] is obviously true.
a x n 1
Let a 1, we want to show that the sequence goes to 1 (it is exactly what we need).
n x
In order to do it we need to represent this sequence in another form. As a 1, the function log a x is
defined, and this function is one-to-one (0,) (,) . Then for any x n there exist
the unique positive real number, which we denote as 1 yn , such that log a (1 yn ) xn [R].
As xn log a (1 yn ) a
xn
1 yn yn a xn 1 [J]. As {xn } 0 , there must be {a xn } 1
х
(because а is continuous at x 0 ). And from [J] we see that { yn } 0 || yn 1 n .
123
Analysis
a xn 1 a loga (1 yn ) 1 (1 yn ) 1 yn
Then [Z]. Let’s use here the formula
xn log a (1 yn ) log a (1 yn ) log a (1 yn )
log b x ln(1 yn )
log a x , (we change the base of the logarithm) log a (1 yn ) , then
log b a ln a
yn y ln a yn
n [Z]. from [A] follows that 1 . And therefore, the limit of
log a (1 yn ) ln(1 yn ) ln(1 y n
)
the sequence [Z] is ln a . Everything is proved.
Now we are ready to find the derivatives of logarithmic and power functions:
Theorem3. Let a 0 || a 1 is a fixed positive number, then
log a x 1
[A] log a x is differentiable at any point x0 (0,) and | x x0 .
x x0 ln a
[B] Let a 0 || a 1, then the power function a x is differentiable at any point x0 (,) and
a x
|x x0 a x0 ln a .
x
Proof. [A] Let’s fix an arbitrary x0 (0,) . We need to calculate the limit:
x
log a 1
log a ( x0 x) log a x0 log a ( x0 x) / x0 x0
lim x 0 lim x 0 lim x 0
x x x
x x x
log a 1 log a 1 log a x ln 1
lim x 0 x0 1 1
lim x 0 x0
log b x lim x 0
1 x0
x x0 x0 x x0 x
log b a ln a
x0 x0 x0
x
ln 1
lim x 0 0
1 x 1
[Theorem 2 [ A]] .
x0 ln a x x0 ln a
x0
[B] Let’s fix an arbitrary x0 (,) . We need to calculate the next limit:
a x0 x a x0 a x 1 a x 1
lim x 0 lim x 0 a x0 a x0 lim x0 [Theorem 2 [ B]] a x0 ln a .
x x x
124
Analysis
Theorem4.
[A] The power function x n || n , x (,) is differentiable at any point x0 (,) and
x n n 1
| x x0 n x0 . For any negative integer number n the function x n is differentiable at any
x
x n n 1
point x0 (,0) (0,) and we have the same formula | x x0 n x 0 .
x
[B] The generalized power function x a || x 0, a R (which is by definition x a e a ln x ) is
x n a 1
differentiable at any point x0 (0,) and | x x0 a x 0 .
x
Proof. [A] Let’s fix any n and any x0 (,) . We are interested in the next expression:
n
( x0 x) x0
n n
binomial k 0 Cnk x0k x n k x0n
x theorem x
n2 n 1
Cn x0 x Cn x0 x Cn2 x0 x n 2 ... Cnn 2 x0 x 2 Cnn 1x0 x1 Cnn x0 x0 x0
0 0 1 1 n 1
n 2 n n
x
n2 n 1
Cn0 x0 x n Cn1 x0 x n 1 Cn2 x0 x n 2 ... Cnn 2 x0 x 2 Cnn 1x0 x1
0 1 2
x
n2 n 1
Cn0 x0 x n 1 Cn1 x0 x n 2 Cn2 x0 x n 3 ... Cnn 2 x0 x1 Cnn 1x0 - every summand here is
0 1 2
a function of x and every summand (except the last one) has a limit 0 when x 0 , and the last
n 1 n 1 n 1 n 1
summand has a limit Cn x0 when x 0 . Then all the sum has a limit Cn x0 when
( x0 x)n x0
n
n 1 n 1
x 0 . Then lim x 0 Cnn 1x0 nx0 .
x
1
Let now n is a negative integer number, then x n . Let’s fix an arbitrary x0 (,0) (0,)
x| n |
f (a) g (a) g x (a) f (a)
and now we will use the formula ( f / g ) x (a) x .
g 2 (a)
We can use it, because functions 1 and x , | n | are differentiable at any point
| n|
x0 (,0) (0,) .
| n| 1
0 x0 | n | x0 1 |n|
| n|
1
So |n| ( x0 ) | n| 1
| n | x (|n|1) | n | x |n|1 n x n 1 .
x x
2| n|
x0 x0
[B] Let’s consider x a ea ln x || x 0, a R . We fix any point x0 (0,) . The function e a ln x is
a composite function ( g ( x)) , where ( y) e y , g ( x) a ln x . And g (x) is differentiable at any
125
Analysis
point x0 (0,) (we showed above that any logarithmic function is differentiable) and ( y)
x
is differentiable at any point g ( x0 ) (because the power function e is differentiable everywhere).
( g ( x)) ( y ) g ( x)
Then | x x0 | y g ( x0 ) |x x0 , so we have
x y x
e a ln x
x
| x x0
e y
y
| y aln x0
(a ln x)
x
1
| x x0 e aln x0 a e ln x0
x0
a
a
1
x0
1 a 1
x0 a a x0 .
a
x0
Everything is proved.
Exercise6. Find the derivative of f ( x) ( x a)n || n at a point x0 R .
Solution. This function can be considered as a composite function ( g ( x)) , where ( y) y n and
g ( x) x a . As g (x) is differentiable at any point x0 and ( y ) is differentiable at any point
y ( x0 ) , then ( g ( x)) is differentiable at any point x0 and
( g ( x)) ( y ) g ( x)
| x x0 | y g ( x0 ) | x x0 .
x y x
(( x a)n ) y n ( x a)
Then | x x0 | y ( x0 a ) |x x0 n ( x0 a)n 1 1 n ( x0 a)n 1 .
x y x
Derivative examples
We will not consider the simpliest derivative-examples in this book, there is a huge amount of books
and internet resources where it is in abundance. Let’s take more advanced examples.
For Reader’s practice:
.....
sin x sin x
at x0 / 4 .
sin x
[1] Find the derivative of f ( x) sin x
Solution. We have here f ( x) sin x ln f ( x) f ( x) ln(sin x) let’s differentiate both sides
f ( x)
with respect to x (when we do so, we need to take into account the rules of differentiation of
a composite function).
1 1
(ln f ( x))' ( f ( x) ln(sin x))' f ' ( x) f ' ( x) ln(sin x) f ( x) cos x
f ( x) sin x
1 f 2 ( x) ctgx
f ' ( x) ln(sin x) f ( x) ctgx f ' ( x) .
f ( x) 1 f ( x) ln(sin x)
We have obtained the explicit expression for f ' ( x) , now we can substitute x0 / 4 , then
( f ( / 4))2 ctg ( / 4) ( f ( / 4))2
f ' ( / 4) f ' ( / 4) , we see that here appears
1 f ( / 4) ln(sin( / 4)) 1 f ( / 4) ln(1 / 2 )
..... .....
sin / 4 sin / 4 1/ 2 1/ 2
sin / 4
the value f ( / 4) , let’s find it: f ( / 4) sin / 4 1/ 2 a.
1/ 2
126
Analysis
We have here
a1/ 2
a 1 a11/ 2
ln(1) ln( a11/ 2
) 0 (1 1/ 2 ) ln a ln a 0 a e0 1,
12 1
then we have f ( / 4) 1 , then f ' ( / 4) .
1 1 ln(1 / 2 ) 1 ln(1 / 2 )
.....
ln x ln x
at x0 e .
ln x
[2] Find the derivative of f ( x) ln x
Solution. Obviously f ( x) ln x ln f ( x) f ( x) ln(ln x) and we differentiate both sides with
f ( x)
respect to x , so
1 1 1 1 f ( x)
f ' ( x) f ' ( x) ln(ln x) f ( x) f ' ( x) ln(ln x)
f ( x) ln x x f ( x) x ln x
f 2 ( x) f 2 (e)
f ' ( x) . Let’s substitute now x e , f ' (e)
x ln x 1 f ( x) ln(ln x) x ln e 1 f (e) ln(ln e)
1 1
we obviously have f (e) 1 , then f ' (e) .
e 1 1 1 ln 1 e
[3] Find all the functions f (x) , that are defined on R , such that f ( x) f ( y) ( x y)2 x, y R .
Answer: there must be f ( x) const on R . Hint. Show that f ' ( x0 ) 0 for any x0 R .
[4] f ( x), g ( x), h( x) are differentiable on [a, b] . Show that there exist c (a, b) such that
f ( a ) g ( a ) h( a )
f (b) g (b) h(b) 0 .
f ' (c ) g ' (c ) h ' (c )
f ( a ) g ( a ) h( a )
Solution. Let’s consider the auxiliary function ( x) f (b) g (b) h(b) , this function is obviously
f ( x ) g ( x ) h( x )
differentiable on [a, b] and obviously (a) (b) 0 , then there exist c (a, b) such that ' (c) 0 .
And it’s easy to understand (as f (a), f (b), g (a).... and etc. are constants) that for any x [a, b]
f ( a ) g ( a ) h( a )
we have ' ( x) f (b) g (b) h(b) , we already know that ' (c) 0 - it is exactly what we need.
f ' ( x ) g ' ( x ) h' ( x )
1
[5] Let f ( x) x , find f (100) and f ' (100) .
1
2x
1
2x
2 x ....
Answer: f (100) 10001 and f ' (100) 100 / 10001 .
Hint. In such cases we always need to create some simple auxiliary equality for f (x) (as above).
127
Analysis
[6] f (x) is differentiable everywhere (on R ) and it’s derivative f ' ( x) is continuous everywhere
(such function f is called a “smooth function”). We have f (0) 6 and f ' (4) 5 .
f (4) f ( x 2 )
Calculate lim x 2 . Answer: 20 .
2x
Solution. According to the mean value theorem, we can write f (4) f ( x ) f ' (c( x )) (4 x )
2 2 2
2
here c(x) is some point inside the interval with ends 4, x , we denote it in such way because
for every value x there is a concrete value c c(x) . Then
f (4) f ( x 2 ) f ' (c( x)) (4 x 2 )
lim x 2 lim x 2 lim x 2 f ' (c( x)) (2 x) [L].
2 x 2 x
Obviously lim x 2 (2 x) 4 . But what is the value of lim x2 f ' (c( x)) ? The simple explanation is:
when x 2 x 4 , it forces c( x) 4 and then f ' (c( x)) f ' (4) 5 (initial condition),
2
so lim x 2 f ' (c( x)) 4 and the limit [L] is 20 .
Notice. We got the equality lim x 2 f ' (c( x)) 4 , the explanation from above is quite convenient,
but not rigorous. Let’s give a normal explanation. We consider f ' (c( x)) as a composite function,
it is constructed from the functions f ' ( y) || y c( x) . The value c(x) is a number inside the interval
with ends 4, x . It’s easy to understand that lim x 2 c( x) 4 (just because c(x) is always between
2
4 and x 2 ) and z f ' ( y) is continuous at y 4 (because it is continuous everywhere).
Then (Improvement of Theorem3 [Limit of a composite function] ) lim x 2 f ' (c( x)) f ' (4) 5 .
f ( a 2h 2 ) f ( a 2h 2 )
[7] f (x) is differentiable at x a and f ' (a) 1/ 4 . Calculate lim h 0 .
h2
Hint. Use the mean value theorem. Answer: 1 .
[8] f (x) is such function that 2 x 2 f ( x ) 2 x f ( x ) find f ' (8) .
Solution: 2 x 2 f ( x ) 2 x f ( x ) 2 f ( x ) (2 x f ( x ) 1) 2 f ( x ) 2 x 2 x f ( x ) 1 2 x
2 x f ( x ) 2 x 1 log 2 2 x f ( x ) log 2 (2 x 1) x f ( x) log 2 (2 x 1) f ( x) x log 2 (2 x 1)
we got the explicit formula for f (x) , and now we can use the standard rules to find it’s derivative:
1 2x 1
f ' ( x) 1 (log 2 (2 x 1))'(2 x 1)', so f ' ( x) 1 ( 2 x
ln 2) 1 ,
(2 x 1) ln 2 2x 1 2x 1
1 1
then f ' (8) .
2 1 255
8
2
[9] f (x) is such function that f ( x) x x x ..... , find f ' (2018 ) .Answer: f ' (2018 )
8073 1
[10] Functions g and f are both differentiable on [a, b) , here b R , or b .
And g (a) f (a) and g x ( x) f x ( x) everywhere on (a, b) . Show that g ( x) f ( x) everywhere
on (a, b) . Hint. Consider the auxiliary function ( x) f ( x) g ( x) .
128
Analysis
Taylor’s theorem
Higher Order Derivatives. f (x) is defined on some neighborhood R (a) of a. Suppose that
f is differentiable at any point x0 R (a) . Then for any x0 R (a) the number f x ( x0 ) is defined
and therefore f x (x) is an independent function, which is defined on R (a) . And we can raise
a question about it’s derivative at a. If f x (x) is differentiable at a, i.e., the limit
f x ( x) f x (a)
lim x a exists, then this limit is called the second derivative of f at a, and we denote
xa
f ( x) f x ( a )
it f xx (a) lim x a x . Similarly, for any point x0 R (a) we can raise a question about
xa
f ( x) f x ( x0 )
a derivative of f x (x) at x0 , i.e., the question about the existence of the limit lim x x0 x ,
x x0
if this limit exists, we denote it f xx ( x0 ) and it is called the second derivative of f at x0 .
If f x (x) is differentiable at any point x0 R (a) , then f xx (x) is an independent function on R (a)
and we can raise a question about it’s derivative at any point x0 R (a) , if such derivative exists,
we denote it f xxx ( x0 ) and it is called the third derivative of f at x0 and etc.
Def. An n-th derivative of f at a can be denoted as f n (x) .
Assertion1. Let f (x) is differentiable on R (a) and f x (a) 0 . And the second derivative f xx (a)
exists.
[A] If f xx (a) 0 , then f has a strict local minimum at a.
[B] If f xx (a) 0 , then f has a strict local maximum at a.
f x ( x) f x ( a ) f ( x)
Proof. By definition f xx (a) lim x a lim x a x . If [A] f xx (a) 0 , then
xa xa
f ( x) f ( x)
lim x a x 0 , then there exist the deleted neighborhood D (a) R (a) , on which x 0.
xa xa
f ( x)
So for any x D (a) (a , a) (a, a ) we have x 0 . If x (a , a) , then x a 0
xa
and there must be f x ( x) 0 . So, for any x (a , a) we have f x ( x) 0 , then f (x) is strictly
decreasing on (a , a] . If x (a, a ) , then x a 0 and there must be f x ( x) 0 .
So, for any x (a, a ) we have f x ( x) 0 , then f (x) is strictly increasing on [a, a ) .
So f is strictly decreasing on (a , a] and f is strictly increasing on [a, a ) .
Then a is a point of a strict local minimum of f . And there is a similar proof for [B].
Def. f (x) is defined on OR (a) and it is n times differentiable at a.
f 1 (a) f 2 (a) f n (a)
The function n ( x) f (a) ( x a) ( x a) ... ( x a)n is called
1 2
1! 2! n!
a Taylor polynomial of n th degree for f (x) with center at a.
129
Analysis
Auxiliary1. f (a) n (a) and f x (a) (n ) x (a), f xx (a) (n ) xx (a) ..... f n (a) (n )n (a) .
(it’s very easy to check these equalities).
Auxiliary2. (x) and g (x) are defined on R (a) and both these functions are (n 1) times
differentiable on OR (a) and also:
[A] (a) g (a) 0, x (a) g x (a) 0 .... n (a) g n (a) 0 ,
[B] g ( x) 0 for any x DR (a) and for any k we have g k ( x) 0 || x DR (a) .
Then: for any x DR (a) there exist the point c from the interval with ends x and a such that:
( x) n 1 (c)
.
g ( x) g n 1 (c)
Proof. We fix any x DR (a) (a R, a) (a, a R) , let x (a, a R) .
Let’s apply the Cauchy formula for (x) and g (x) on [a, x] , then we have
( x) (a ) x (c1 )
|| a c1 x , according to [A], we have (a) g (a) 0 , then we get
g ( x) g (a ) g x (c1 )
( x) x (c1 ) ( x) x (c1 ) x (a)
|| a c1 x . According to [A], we can rewrite it || a c1 x .
g ( x) g x (c1 ) g ( x) g x (c1 ) g x (a)
Now we can apply the Cauchy formula for x (x) and g x (x) on [a, c1 ] , we will get
( x) x (c1 ) x (a) xx (c2 ) ( x) xx (c2 )
|| a c2 c1 x , so now we have || a c2 c1 x .
g ( x) g x (c1 ) g x (a) g xx (c2 ) g ( x) g xx (c2 )
( x) xx (c2 ) xx (a)
According to [A], we can rewrite it: || a c2 c1 x and now we apply
g ( x) g xx (c2 ) xx (a)
the Cauchy formula for xx (x) and g xx (x) on [a, c2 ] .
( x) xxx (c3 )
Then we have || a c3 c2 c1 x and etc. Eventually we will get
g ( x) g xxx (c3 )
( x) n 1 (c)
|| a c x . And there is a similar proof for any fixed x (a R, a) .
g ( x) g n 1 (c)
Taylor's theorem. f is defined on OR (a) and (n 1) times differentiable on OR (a) .
f n 1 (c)
Then for any x DR (a) the next formula is true: f ( x) n ( x) ( x a) n 1 [V] where c
(n 1)!
is some point from the interval with ends x and a , and n (x) is a Taylor polynomial (def above).
Proof. Let’s fix an arbitrary x DR (a) . From auxiliary1 follows that for ( x) f ( x) n ( x) we
have (a) 0, x (a) 0, xx (a) 0.... n (a) 0 . Let’s take the function g ( x) ( x a)n 1 ,
then the pair ( x), g ( x) satisfies to the auxiliary2. Then there exist c from the interval with ends
130
Analysis
a, x such that:
f ( x) n ( x) f n 1 (c) n (c) as n ( x) 0
n 1 n 1
( x) n 1 (c) f ( x) n ( x) f n 1 (c)
n 1
g ( x ) g (c ) ( x a)n 1 g n 1 (c) and g n 1
( x ) ( n 1)! ( x a ) n 1
( n 1)!
f n 1 (c)
f ( x) n ( x) ( x a) n 1 - it is exactly what we need.
(n 1)!
Notice that any value x DR (a) defines one point (the number) c (which lies between x and a),
for which [V] is true. Then c c(x) is a function which is defined on DR (a) and obviously
lim x a c( x) a (just because c(x) is always between x and a).
f n 1 (c)
Def. The function ( x) ( x a)n 1 is called a remainder.
(n 1)!
Assertion1. If f n1 ( x) is continuous at a, then the remainder (x) can be represented as
( x) ~( x) ( x a)n , where (x) is a function on D (a) , which is infinitely small when x a .
R
Proof. By definition, the remainder is:
f n 1 (c( x)) n 1 f n 1 (c( x))
( x) ( x a) ( x a) ( x a) n . The function
(n 1)! (n 1)!
f n 1 (c( x))
~( x) ( x a) is defined for every x DR (a) . We need to show that this function goes
(n 1)!
to zero when x a , i.e., to show that lim ~( x) 0 . Let’s fix an arbitrary sequence
x a
{xn } R (a) || xn a n || {xn } a , every x n defines one point cn cn ( xn ) from the interval with
ends xn , a , then we have the sequence {cn } {cn ( xn )} R (a) || cn a n || {cn } a .
As f
n1
( x) is continuous at a, there must be { f n 1 (cn )} f n 1 (a) or the same
n 1 n 1 ~ f n 1 (c( xn ))
{ f (cn ( xn ))} f (a) . Then the sequence { ( xn )} ( xn a) goes to zero 0 ,
(n 1)!
then lim x a ( x) 0 . So ( x) is infinitely small when x a , everything is proved.
~ ~
Consequence1. By definition of the symbol o , we can write (x) as ( x) o(( x a)n ) , which is
~ ( x) ( x a ) n .
equivalent to ( x)
Let’s sum up:
(n 1) times differentiable on OR (a) , then for any
[1-st result] If f (x) is defined on OR (a) and
x DR (a) there exist some c from the interval with the ends x, a such that:
f 1 (a) f 2 (a) f n (a) n f n 1 (c)
f ( x) f ( a) ( x a)
1
( x a) ...
2
( x a) ( x a) n 1 [T].
1! 2! n! (n 1)!
131
Analysis
[2-nd result] If in addition f n1
( x) is continuous at a, then the next representation is true on DR (a) :
f 1 (a) f 2 (a) f n (a)
f ( x) f (a) ( x a)1 ( x a) 2 ... ( x a) n ~( x) ( x a)n [M], where
1! 2! n!
lim x a~( x) 0 . Let’s define now the new function (x) on O (a) : x D (a) ( x) ~( x) R R
and (a) // by def // 0 , then for such function we have lim x a ( x) 0 (0) , i.e., (x) is
infinitely small when x a and also continuous at a. And for any x OR (a) the next
representation is true:
f 1 (a) f 2 (a) f n (a)
f ( x) f (a) ( x a)1 ( x a) 2 ... ( x a) n ( x) ( x a) n [V].
1! 2! n!
Really, if x DR (a) then the representation [V] is exactly the representation [M]. And if x a ,
then both sides of [V] are equal to f (a) , so [V] is true x R (a) . And from now on we will
always use the representation [V] (instead of [M]). In general, the Taylor’s theorem is a very
powerful tool for many purposes, we will show soon the applications.
It’s easy to get the Taylor’s representations [V] of elementary functions (i.e., trigonometric, power,
logarithmic functions), and it’s convenient to remember most of them.
Comment. In each case here the last term o( x ) is ( x) x k , where (x) is infinitely small when
k
x 0 and continuous at x 0 , i.e., lim x 0 ( x) 0 (0) .
x 2 x3 x 4 x 2 x3 x 4
e 1 x .... o( x ) x R and ln(1 x) x o( x 4 ) x (1,1] ,
x 4
2! 3! 4! 2 3 4
x3 x3
sin x x o( x ) x R
4
and arcsin x x o( x 4 ) x [1,1] ,
3! 6
x3 x3
tgx x o( x ) x ( / 2, / 2) and arctgx x o( x 4 ) x [1,1] ,
4
3 3
x 2
( 1) 2
cos x 1 o( x3 ) x R and (1 x) 1 x x o( x 2 ) || R, x (1,1) .
2! 2!
These representations allow us to determine values of really “difficult” limits. Let’s give some theory
and examples.
, where lim x a f ( x) 1 and
g ( x)
In practice we often deal with limits like lim x a f ( x)
lim x a g ( x) , in such case we say that “we have a limit 1 ” as lim x a f ( x) 1 , then f (x) is
obviously positive on some deleted neighborhood D (a) , and because of that x D (a) we can
ln f ( x )
ln f ( x ) g ( x ) g ( x )ln f ( x ) ln f ( x)
rewrite f ( x)
g ( x)
as e e e 1/ g ( x )
, where . Here we have the fraction
1 / g ( x)
ln f ( x)
numerator and denominator both go to zero when x a . Such limit lim x a can be
1 / g ( x)
132
Analysis
ln f ( x)
calculated if we use Taylor’s representations for it’s functions. And after that, if lim x a ,
1 / g ( x)
is equal to e (it follows from continuity of e x at every point
g ( x)
then the initial limit lim x a f ( x)
R ).
Def. The expression e v ( x ) , where v(x) is some function, can be rewritten like ev ( x ) exp{v( x)} .
It is convenient when v(x) is some voluminous expression.
Example1. Calculate the limit lim x 0 1 sin x arctg x 2 2
1 / sin x 2
. Answer: e 2 .
ln(1 sin 2 x arctg 2 x)
Solution. Here we have a limit 1 , let’s rewrite it like exp 2 [Ex].
sin x
x3 x3
Let’s use sin x x ( x) x , arctgx x ( x) x 4 . Here and everywhere later, every
4
3! 3
function (the last term) like (x) or (x) is such that:
lim x 0 ( x) 0 (0), lim x 0 ( x) 0 (0) . Then
2 2
x3 x3
1 sin x arctg x 1 x ( x) x 4 x ( x) x 4
2 2
3! 3
x 4
x 4
1 x 2 2 v( x) x 4 x 2 2 u ( x) x 4 , where v(x) and u (x) are infinitely small
3! 3
when x 0 and both continuous at x 0 . Then 1 sin x arctg x 1 2 x x t ( x) x ,
2 2 2 4 4
where t (x) has the same property as v(x) or u (x) or (x) or (x) , each new auxiliary function is
a function: infinitely small when x 0 and continuous at x 0 .
So, the numerator in [Ex]: ln(1 sin x arctg x) ln(1 [2 x x t ( x) x ]) . Let’s use the
2 2 2 4 4
formula ln(1 x) x h( x) x , then:
ln(1 (2 x 2 x 4 t ( x) x 4 )) (2 x 2 x 4 t ( x) x 4 ) h(2 x 2 x 4 t ( x) x 4 ) (2 x 2 x 4 t ( x) x 4 ) .
The function h(2 x x t ( x) x ) is a composite function, here 2 x x t ( x) x is infinitely
2 4 4 2 4 4
small when x 0 and continuous at 0 , and the same is true for h , then the composite function
h(2 x 2 x 4 t ( x) x 4 ) ( x) is infinitely small when x 0 and continuous at 0 ,
so we can rewrite: ln(1 (2 x x t ( x) x )) (2 x x t ( x) x ) ( x) (2 x x t ( x) x )
2 4 4 2 4 4 2 4 4
( 2 x 2 x 4 t ( x) x 4 ) ( x) ( 2 x 2 x 4 t ( x) x 4 ) 2 x 2 ( x) x 2 .
And the denominator in [Ex]: sin x , let’s use the formula sin x x s( x) x , then
2
sin x 2 x 2 s( x 2 ) x 4 . Similarly, the composite function s( x 2 ) can be rewritten as q(x) .
133
Analysis
And sin x x q( x) x . From [Ex] we see that we are interested in the next limit:
2 2 4
ln(1 sin 2 x arctg 2 x) 2 x 2 ( x) x 2 2 ( x) 2
lim x 0 lim lim 2,
x 2 q ( x) x 4 1 q ( x) x 2 1
x 0 x 0
sin x 2
then the initial limit is equal to e 2 .
4 35 x3 1 (cos x)arctgx
Example2. lim x 0 . Answer: 1/ 10 .
x8 sin x3
0
Solution. Here both numerator and denominator go to zero, so we have a limit . Let’s use Taylor’s
0
arctgx
representations. The most puzzling function here is (cos x) , it is obviously a generalized power
e arctgxln(cosx ) [S].
arctgx
function, so we need to rewrite it as (cos x)
arctgx
e ln(cosx )
x2 x2 x2
Here cos x 1 ( x) x3 , then ln(cos x) ln 1 ( x) x3 ln 1 ( x) x3
2! 2! 2!
let’s use the formula ln(1 x) x ( x) x , then
x2 3
x2 3 x2 3 x2
ln 1 ( x) x ( x) x ( x) x ( x) x 3 .
2! 2! 2! 2!
x2
The composite function ( x) x 3 is infinitely small when x 0 and continuous at x 0 ,
2!
then it can be rewritten as (x) and the last expression can be rewritten as
x2 x 2
ln 1 ( x) x ( x) x 2 . Next: arctgx x ( x) x 2 , then the power in [S] is
3
2! 2!
x2 2 x3
arctgx ln(cos x) ( x ( x) x )
( x) x v( x) x 3 .
2
2! 2!
x3
v ( x ) x 3
arctgxln(cosx )
Then [S] is e e 2! . Let’s use the formula e 1 x o( x) e 1 x h( x) x .
x x
x3
v ( x ) x 3 x3 3 x3 3 x3 3 x3
So e 2!
1 v( x) x h v( x) x v( x) x 1 t ( x) x 3 .
2! 2! 2! 2!
There is also the expression 3 x 1 (in numerator). Let’s use the formula
5 3
x3
(1 x) 1 x r ( x) x , here 1/ 5 , then 5
x3 1 (1 x3 )1 / 5 1 r ( x3 ) x3
5
x3
1 ( x) x 3 .
5
134
Analysis
Then the numerator of the initial limit is
x3 3 x3 3 3x3 x3
4 3 x 1 (cos x)
5 3 arctgx
4 3 1 r ( x ) x 1 t ( x) x
3
( x) x 3
5 2! 5 2
x3
( x) x3 . Let’s consider the denominator x8 sin x3 as
10
sin x x ( x) x sin x3 x3 ( x3 ) x3 , we can rewrite it as sin x3 x3 p( x) x3 .
So, the needed limit can be rewritten:
x3 1 1
( x) x 3 ( x)
4 35 x3 1 (cos x)arctgx 10 1
lim x 0 lim x 0 lim x 0 5 10 10
x8 sin x3 x ( x p ( x) x )
8 3 3
x (1 p( x)) 1 10
Taylor’s representations also allow us to find approximate values of functions.
Example3. Let’s calculate cos10o with the accuracy 0,002 .
Solution. We need to convert degrees in radians cos10 cos
o
. Let’s use the formula [T] for cos x ,
18
so there exist the point c in the interval with ends 0, x such that:
x 2 сosxxx (c) 3 x 2 sin(c) 3
cos x 1 x cos x 1 x
2! 3! 2! 3!
(we are free to choose how many terms we want to take in [T]).
x 2 sin(c) 3 x 2 sin(c) 3 x3
From here we have: cos x 1 x cos x 1 x ,
2! 3! 2! 3! 6
3,5 x3 (0,2)3 x2
if x 0,2 , then 0,002 . Then for x we have cos x 1 0,002 .
18 18 6 6 18 2!
x2
So when x , the expression 1 approximates cos x with 0,002 accuracy.
18 2!
( / 18) 2 2
Then cos 1 cos 1 .
18 2! 18 648
Taylor’s formulas [T] and [M] has a great importance in mathematics. When we use these formulas,
especially for limit-calculation, we should be careful and do not lose any terms. In some cases we can
use more simple tool in order to calculate some “difficult” limits, it is called a L'Hospital rule.
Notice that the L'Hospital rule is a great tool, but it doesn’t work in all cases. If we apply this rule to
the previous examples1,2 we will not get any result. Taylor’s representation is a much stronger tool,
but the use of it requires from us to perform much more work.
135
Analysis
f ( x)
L'Hospital rule. We want to calculate some limit lim x a , where f (x) and g (x) are defined
g ( x)
and differentiable on some deleted neighborhood DR (a) and g ( x) 0, g x ( x) 0 on DR (a) .
f ( x)
And we have the situation: lim x a f ( x) 0 and lim x a g ( x) 0 . So the limit lim x a is
g ( x)
0 f ' ( x)
a limit . In such case we can consider the limit lim x a , if this limit exists, then the initial
0 g ' ( x)
f ( x)
limit lim x a also exists and these limits are equal.
g ( x)
In practice this tool allows us to solve problems very quickly.
sin mx 0
Example. Find lim x 0 . Here we see a limit , let’s calculate
sin nx 0
(sin mx)' cos mx m m sin mx m
lim x 0 lim x 0 , then the initial limit lim x 0 is .
(sin nx)' cos nx n n sin nx n
~ ~ ( x) :
Let’s prove the L'Hospital rule. We define two auxiliary functions f ( x) and g
~ ~
f ( x) f ( x) x DR (a) and f (a) // by def // 0 and similarly:
g~( x) g ( x) x D (a) and g~(a) // by def // 0 .
R
~
f ( x) f ( x)
The functions and ~ are both defined and equal on DR (a) , then the existence of
g ( x) g ( x)
~
f ( x) f ( x)
lim x a ~ is tantamount to the existence of lim x a . If one of these limit’s exist,
g ( x) g ( x)
f ' ( x)
then the other also exists and these limits are equal. Let’s assume that lim x a .
g ' ( x)
~
f ( xn )
Let’s fix an arbitrary sequence {xn } DR (a) || {xn } a if we show that ~ ,
g ( xn
)
~
f ( x)
then lim x a ~ , and it is exactly what we need.
g ( x)
~ ~( x) are differentiable on D (a) . For every x we have:
Both functions f ( x) and g R n
~ ~ ~ ~
f ( xn ) f ( xn ) f (a) Cauchy f ' (cn )
. As {xn } a , there must be {сn } a (because every
g~( xn ) g~( xn ) g~(a) formula g~' (сn )
f ' ( x) f ' ( сn )
сn is between xn and a ). As we have lim x a , then , which is equivalent
g ' ( x) g ' ( сn )
136
Analysis
~
f ( xn )
to ~ , everything is proved.
g ( x n
)
f ' ( x)
Comment. L'Hospital rule can be applied repeatedly, i.e., we can apply it for lim x a
g ' ( x)
f ' ' ( x)
if necessary (and if possible) and then for lim x a if necessary and etc.
g ' ' ( x)
For Reader’s practice:
[1] Find the limit lim n n cos sin . Answer: .
4n 4n 4
ax bx a
[2] Find the limit lim x 0 . Answer: ln .
x b
100 k
x 100
[3] Find the limit lim x 1 k 1 . Answer: 5050 .
x 1
1 2
[4] Find the limit lim x 0 2 ctg 2 x Answer: .
x 3
x
1/ x
[5] Find the limit lim x tg Answer: 1 .
2x 1
1 1
[6] Find the limit lim x 0 Answer: 1 .
sin x arctgx tgx arcsin x
1 x sin x ln(cos x) x 7
[7] Find the limit lim x 0 Answer: .
3
1 x3 1 8
1 2 x 3 cos x 4
[8] Find the limit lim x 0 Answer: 3 .
tgx x
1 2tgx e x x 2
[8] Find the limit lim x 0 Answer: 2
arcsin x sin x
xetgx sin 2 x x 3
[9] Find the limit lim x 0
Answer: .
x x 3
tgx 4
x x2 ... xn
Exercise*: {xn } a , show that [1] the sequence 1 also goes to a .
n
[2] If {xn } is a positive sequence, then {n x1 x2 ... xn } goes to a .
137
Analysis
Integral
From now on, f (x) is any function that is defined on [a, b] and bounded on [a, b] .
Let’s divide the segment [a, b] into several consecutive segments 1 , 2 .... n , we choose
an arbitrary point k on every segment k , the sum f (1 ) 1 f (2 ) 2 .... f (n ) n is
called an integral sum of f with respect to the partition 1, 2 ... n { k } . This sum depends on
the points k k || k 1.. n . The maximal length max{ k } (i.e., the length of the longest segment)
is called a norm of the partition { k } .
Def. The number A is called an integral f on [a, b] if for any (small) positive 0 we can find
the positive 0 , such that for any partition { k } of [a, b] such that max{ k } , for any
points k k || k 1.. n , the integral sum f (1 ) 1 f (2 ) 2 .... f (n ) n is
“ close to ” , i.e., .
b
In this case we denote f ( x)dx and we say that f is integrable on [a, b] , and is an integral
a
of f on [a, b] .
b
Exercise1. Show that if such number f ( x)dx exists, than it is unique (there can’t be two
a
different numbers which are both integrals of f on [a, b] ).
As f is bounded on [a, b] , it is bounded on every segment k , so the set of all values { f ( x) || x k }
has a supremum k and an infinum mk . The sum m1 1 m2 2 .... mn n is called
low
a lower (integral) sum of f with respect to { k } and the sum 1 1 2 2 .... n n
up
is called an upper (integral) sum of f with respect to { k } .
Notice. For any concrete partition { k } , a lower and an upper integral sums and
low up
are
concrete real numbers, these numbers are uniquely defined by the partition { k } . And an ordinary
integral sum f (1 ) 1 f (2 ) 2 .... f (n ) n is not uniquely defined by { k } , this sum
is defined by the choice of points k k || k 1.. n .
For any partition { k } we obviously have up .
low
Exersice2. Let’s fix any partition { k } .
We have the set {} of all integral sums with respect to { k } . Show that the upper sum
up
is
a supremum of {} , and the lower sum
low
is an infinum of {} .
138
Analysis
Theorem1 [Integrability criterion]. f is integrable on [a, b] for any 0 there exist 0
such that for any partition { k } , which norm is less than , we have up low .
Proof. Let f is integrable on [a, b] . We fix an arbitrary positive 0 , for the positive number
/ 4 0 there exist some 0 such that for any partition { k } , which norm is less than ,
/ 4.
for any integral sum f (1 ) 1 f (2 ) 2 .... f (n ) n we have
Let’s fix any such partition. Any integral sum lies inside the interval ( / 4, / 4) .
Then the set of all integral sums {} lies inside ( / 4, / 4) . As
up
is a supremum of the
set {} ( / 4, / 4) , then there must be
up
[ / 4, / 4] , as low is an infinum
of the set {} ( / 4, / 4) , then there must be [ / 4, / 4] . So low and up
low
are some numbers from the segment [ / 4, / 4] , then the difference up low is not
greater than the length of this segment, so up low / 2 .
We had started from an arbitrary positive 0 , and we found 0 such that for any partition,
which norm is less than , we have up low .
Conversely. Let’s fix an arbitrary positive 0 , there exist 0 such that for any partition
{ k } , which norm is less than , we have up low .
~
Auxiliary1. Let’s notice an important fact. Any lower integral sum low is not greater than any
~ low
upper integral sum , i.e., up even if these sums correspond to different partitions
up
~
{ k },{ k } of [a, b] . Really, let’s fix an arbitrary partition { k } , suppose that we added only one
~ ~
new point to { k } , so it divides some segment j in two segments j , j 1 .
~
Let’s consider f on j . We obviously have j supx j f ( x) j supx~ f ( x) , because
j
~ ~ ~
j j . And similarly j supx j f ( x) j 1 supx~ f ( x) (because j j 1 ).
j 1
From here follows that
~ j
~ ~ ~
~ ~
~ ~
j j 1 j 1 j j j j 1 j ( j j 1 ) j j .
~ up
From here follows that the upper sum of f with respect to the new partition (where one new
point is added ) is not greater than the initial upper sum up of f with respect to the old partition.
Conclusion. When we add some new points to our partition { k } (i.e., we make a “refinement of it”)
an upper integral sum
up
may only become less. And it’s very easy to understand (similarly, by
adding one new point to the initial partition) that a lower integral sum
low
may only become
greater (after any refinement).
139
Analysis
~ low
Let now we have two different upper and lower integral sums , up which are built with respect
~ ~
to some partitions { k },{ k } . Let’s unite these partitions { k },{ k } (we just add all the points of
~
one partition to other partition). The new partition {k } is a refinement of { k } and it is also
a refinement of { k } , let ,
up low
are the upper and lower integral sums with respect to {k } .
~low ~
Then we have low and up up and obviously low up , then low up .
And the auxiliary1 is proved.
Let’s finish the proof of the converse assertion (theorem1). We consider the set of all possible lower
sums {
low
} (every sum is built with respect to some partition of [a, b] ) and the set of all possible
upper sums { } . Any element of { } is not greater than any element of {up} , then { low} is
up low
bounded above and there exist sup{ } , and {up} is bounded below and there exist inf{up} .
low
From the initial condition, for any small 0 there exist some concrete elements low {low}
and
up
{up} such that up low . From here follows that sup{low} inf{up} .
Let’s show that A is an integral of f on [a, b] . We fix an arbitrary positive 0 , there exist 0
such that for any partition { k } , which norm is less than , we have up low . From the
definition of follows that up , and for any integral sum , which is built with respect
low
to { k } , we also have up . Then both numbers , belong to the segment [low, up ] ,
low
b
the length of this segment is less than , then .Then, by definition, f ( x)dx .
a
Geometrical meaning. For any positive function
f on [a, b] , if f is integrable on [a, b] , then
b
f ( x)dx is an area of the figure which
a
“lies below” the graph of f [pict1].
Let’s fix an arbitrary partition of [a, b] .
Any upper sum is an area of a figure ,
up ext
which consists of rectangles and contains . pict.1
And any lower sum low
is an area of a figure
int
,
which consists of rectangles and belongs to . So we have ext and low S (int )
int
and
up
S (ext ). According to the theorem1[Integrability criterion], for any 0 we can fix some
measurable figures
int
ext such that S (ext ) S (int ) .
140
Analysis
From here (assertion4 [1-st criterion of measurability] “Area construction”, Book 1) follows that is
measurable, and the number S () is defind. Let’s fix now any positive sequence { n } 0 ,
for any number n let’s fix the internal figure int (n) and the external figure ext (n) such that
S (ext (n)) S (int (n)) n [T]. Then we have the sequence of internal figures
{int (n)} || int (n) and the sequence of external figures {ext (n)} || ext (n) .
As (n) (n) , then S (int (n)) S () S (ext (n)) || n [E].
int ext
b
In the integrability criterion (theorem1) we showed that f ( x)dx is a supremum of the set
a
{ low} of all lower integral sums and an infinum of the set {up} of all upper integral sums.
So, for any possible lower and upper sums we have up . Notice that every area
low
S (int (n)) is some lower integral sum (by construction), and every area S (ext (n)) is some upper
integral sum . Then there must be S ( (n)) S ( (n)) || n [E1]. From [E] and [E1]
up int ext
follows that both numbers and S () belong to the segment [ S ( (n)), S ( (n))] || n ,
int ext
from [T] follows that the length of that segment can be less than any positive number, then
b
S () S () f ( x)dx .
a
Theorem2. If f is continuous on [a, b] , then f is integrable on [a, b] .
Proof. As f is continuous on [a, b] , then it is bounded on [a, b] and we can raise a question about
it’s integrability. According to the Cantor’s theorem, f is uniformly continuous on [a, b] .
Let’s fix an arbitrary positive 0 . We take the positive number /(b a) 0 , there exist 0
such that for any x1, x2 [a, b] , if | x1 x2 | , then | f ( x1 ) f ( x2 ) | /(2 (b a)) .
Let’s fix any partition { n } of [a, b] which norm is less than . Let’s consider f on any segment
k , for any x1, x2 k we obviously have | f ( x1 ) f ( x2 ) | /(2 (b a)) , from here immediately
follows that k mk /(2 (b a)) where k sup k f ( x) and mk inf k f ( x) .
Let’s estimate the difference between upper and lower integral sums:
up low (1 m1 ) 1 ( 2 m2 ) 2 .... ( n mn ) n 1 ... n
2 (b a) 2 (b a)
1 ... n b a . So we have up low .
2 (b a) 2 (b a) 2
Let’s sum up, for an arbitrary positive 0 there exist 0 such that for any partition { n } ,
which norm is less than , we have low , then according to the theorem1[Integrability
up
criterion], f is integrable on [a, b] .
141
Analysis
Theorem3. f is monotonic on [a, b] , then f is integrable on [a, b] .
Proof. Let f is monotonically increasing on [a, b] (i.e., x1 x2 [a, b] f ( x1 ) f ( x2 ) ).
Then for any x [a, b] we have f (a) f ( x) f (b) , then f is bounded on [a, b] and we can raise
a question about it’s integrability on [a, b] . If f (a) f (b) , then f const on [a, b] and such
function is obviously integrable on [a, b] . Let f (a) f (b) , then we fix an arbitrary positive 0 ,
and we take the number /( f (b) f (a)) . Let’s fix any partition { n } of [a, b] which norm is
less than .
Any value k sup k f ( x) is obviously the value of f at the right end of k .
Any value mk inf k f ( x) is obviously the value of f at the left end of k .
So let’s denote 1 [a, x1 ], 2 [ x1 , x2 ], 3 [ x2 , x3 ]..... n [ xn , b], then
up low ( f ( x1 ) f (a)) 1 ( f ( x2 ) f ( x1 )) 2 .... ( f (b) f ( xn )) n
( f ( x1 ) f (a)) ( f ( x2 ) f ( x1 )) .... ( f (b) f ( xn 1 ))
f (b) f (a) f (b) f (a) f (b) f (a)
[ f ( x1 ) f (a)] [ f ( x2 ) f ( x1 )] [ f ( x3 ) f ( x2 )] .... [ f (b) f ( xn 1 )]
f (b) f (a)
( f (b) f (a)) . And according to the theorem1[Integrability criterion], f is
f (b) f (a)
integrable on [ a, b ] .
Exercise3. By using the initial definition of an integral prove the next properties:
[1] f and g are integrable on [a, b] . Then f ( x) g ( x) is integrable on [a, b] and
b b b
( f ( x) g ( x))dx f ( x)dx g ( x)dx .
a a a
[2] f is integrable on [a, b] and R . Then f (x) is integrable on [a, b] and
b b
f ( x)dx f ( x)dx .
a a
From [1] and [2] follows that: f and g are integrable on [a, b]. Then any linear combination of
these functions f ( x) g ( x) is integrable on [a, b] and
b b b b b b
f ( x) g ( x)dx f ( x)dx g ( x)dx , in particular f ( x) g ( x)dx f ( x)dx g ( x)dx
a a a a a a
142
Analysis
[3] f is integrable on [a, b] . Then for any c (a, b) : f is integrable on [a, c] and on [c, b] and
c b b
f ( x)dx f ( x)dx f ( x)dx .
a c a
b b
[4] f and g are integrable on [a, b] and g ( x) f ( x) on [a, b] . Then f ( x)dx g ( x)dx .
a a
b b
[5] If f is integrable on [a, b] . Then | f | is integrable on [a, b] and f ( x)dx f ( x) dx .
a a
Comment. The converse assertion is not true, if | f | is integrable on [a, b] , then f mustn’t be
integrable on [a, b] . Consider the “Dirichlet function”: f ( x) 1 when x is rational and f ( x) 0
when x is irrational.
Mean Value theorem. f is integrable on [a, b] . Then there exist the number
b
: inf [ a,b] f m M sup[ a,b] f such that f ( x)dx (b a) .
a
Proof. For any x [a, b] we have
b b b b
m f ( x) M mdx f ( x)dx Mdx m (b a) f ( x)dx M (b a) .
a a a a
b
Then really f ( x)dx (b a) (for some : m M ).
a
Consequence1. If f is continuous on [a, b] , then it reaches all it’s intermediate values on [a, b] ,
then there exist с [a, b] such that f (c) , and the mean value theorem looks like
b
f ( x)dx f (c) (b a) .
a
a
Let’s extend the definition of an integral: [1] We define: f ( x)dx 0 , [2] f is integrable on [a, b]
a
a
b
(here a b , as in all the previous cases). We define f ( x)dx // by def // f ( x)dx .
b a
b a
Now: if f is integrable on some segment with ends a, b , then both integrals f ( x)dx and f ( x)dx
a b
b
are defined. And now the symbol f ( x)dx makes sense in all the cases: a b, a b, b a.
a
143
Analysis
It’s very easy to check that the properties [1],[2] are still true for the new “extended” integral.
And [3] becomes true for any points a, b, c (if only all the integrals from [3] exist).
Also, the Mean value theorem and the consequence1 from it are still true.
Let’s sum up: the mean value theorem, and all the “linear properties” of integral are true in any case
a b, a b, b a.
But remember, that all the properties concerning the estimation of integral values (like [4] and [5])
are true only when a b (it’s very easy to understand why). So, when we estimate some integral
value, we need to make sure at first that a b .
Def. f (x) is defined on [a, b] . If there exist the function F (x) ,which is differentiable on [a, b] ,
such that Fx ( x) f ( x) x [a, b] (at the end points a, b we imply the left and right derivatives
of F), then F (x) is called an antiderivative of f (x) on [ a, b ] .
The symbol F ( x) f ( x)dx means that F (x) is an antiderivative of f (x) on some segment.
Assertion1. If F (x) is an antiderivative of f (x) on [a, b] , then for any constant C R the
function F ( x) C is also an antiderivative of f (x) on [a, b] . And if F (x) and G(x) are
antiderivatives of f (x) on [a, b] , then there exist some constant C such that
F ( x) G( x) C || x [a, b] .
The first part is obvious. In the second part we need to consider the auxiliary function
( x) F ( x) G( x) , we have x ( x) Fx ( x) Gx ( x) 0 on [a, b] , then x ( x) const on [a, b] .
Fundamental theorem of calculus.
x
f (x) is continuous on [a, b] , then F ( x) f (t )dt is an antiderivative of f (x) on [a, b] .
a
x
Comment. For any concrete number x [a, b] the integral f (t )dt exists (because f is continuous
a
x x
on [a, x] ). So, for any x [a, b] the value f (t )dt is a concrete real number, then F ( x) f (t )dt
a a
is really a function of variable x on [a, b] . We need to prove that at any point x0 [a, b] we have
Fx ( x0 ) f ( x0 ) .
x
Proof. We fix an arbitrary point x0 [a, b] and we consider the function F ( x) f (t )dt ,
a
F ( x) F ( x0 )
the derivative of this function at x0 is the next limit lim x x0 (if this limit exists).
x x0
144
Analysis
x x0
F ( x) F ( x0 )
f (t )dt f (t )dt
Let’s consider a a
[V]. It’s easy to notice that
x x0 x x0
x0 x x
f (t )dt f (t )dt f (t )dt (it doesn’t matter x0 x or x x0 ). Then [V] can be rewritten like
a x0 a
x
f (t )dt
x0
[V1]. Let’s apply now the mean value theorem for f on the segment with the ends x0 , x ,
x x0
x
as f is continuous, there exist some point c c(x) such that f (t )dt f (c( x)) ( x x0 ) .
x0
Notice, we have denoted c c(x) as a function of x , because every x x0 from [a, b] defines some
point (number) c c(x) which lies in the interval with ends x0 and x . Then [V1] can be rewritten
f (c( x)) ( x x0 ) F ( x) F ( x0 )
as f (c( x)) . And therefore lim x x0 lim x x0 f (c( x)) [V2].
x x0 x x0
Obviously lim x x0 c( x) x0 (because с(x) is always between x and x0 ), as f (x)
is continuous at x0 , the limit of the composite function lim x x0 f (c( x)) exists and equals f ( x0 )
(Improvement of Theorem3 [Limit of a composite function]).
F ( x) F ( x0 )
And from [V2] we have lim x x0 f ( x0 ) . Everything is proved.
x x0
Consecuence2. From the assertion1 follows that any antiderivative of a continuous function f (x)
x
on [a, b] must look like F ( x) f (t )dt С , where С const .
a
Consecuence3 [Newton-Leibniz formula]. f (x) is continuous on [a, b] and F (x) is any
b
antiderivative of f (x) on [a, b] , then f (t )dt F (b) F (a) .
a
x
Proof. Any antiderivative looks like F ( x) f (t )dt С , where C const, then
a
b a b b
F (b) F (a) f (t )dt С f (t )dt С f (t )dt С 0 С f (t )dt .
a a a a
145
Analysis
From here follows that the [Newton-Leibniz formula] can be applied not only when a b , but also
when a b or a b . So we can apply this formula for any kind of integral.
[Newton-Leibniz formula]: f (x) is continuous on some segment with ends a, b and F (x) is any
b
antiderivative of f (x) on that segment, then f (t )dt F (b) F (a) .
a
Consecuence4. f is differentiable on a segment with ends a, b , then f (x) is an antiderivative
b
of f x (x) on that segment and f x (t )dt f (b) f (a) .
a
Def. A function is called smooth on [a, b] if it’s derivative (as an independent function) is
continuous on [a, b] .
Theorem4 [Change of a variable]. f (x) is continuous on [a, b] . And x (t ) is smooth on
[ , ] , all the values of (t ) on [ , ] belong to [a, b] and ( ) a, ( ) b .
b
Then f ( x)dx f ( (t )) t (t )dt .
a
b
Proof. From the initial condition follows that both integrals f ( x)dx and f ( (t )) t (t )dt do exist
a
(because these are integrals of continuous functions). Let’s fix any antiderivative F (x) of f (x) on
[a, b] . And let’s consider the function F ( (t )) , it is a composite function, and it is defined on [ , ] .
For any point t0 [ , ] , the function x (t ) is differentiable at t 0 , and F (x) is differentiable at
x0 x(t0 ) (because F is differentiable everywhere on [a, b] , as Fx ( x) f ( x) by definition),
then the composite function F ( (t )) is differentiable at t 0 , and there must be
Ft ( (t0 )) Fx ( (t0 )) t (t0 ) [as Fx ( x) f ( x)] f ( (t0 )) t (t0 ) .
So we got the formula Ft ( (t0 )) f ( (t0 )) t (t0 ) || t0 [ , ] , we can rewrite it as
Ft ( (t )) f ( (t )) t (t ) on [ , ] . The last equality means that F ( (t )) is an antiderivative of
f ( (t )) t (t ) on [ , ] . Then the [Newton-Leibniz formula]
f ( (t )) t (t )dt F ( ( )) F ( ( )) F (b) F (a) [N1]. In the same time for the integral
b
f ( x)dx , according to the same [Newton-Leibniz formula], we also have
a
b
f ( x)dx F (b) F (a) [N2]. From [N1] and [N2] follows the equality we need.
a
146
Analysis
Def. For any function (x) on [a, b] , the difference (b) (a) can be denoted as
( x) |ba (b) (a) .
Theorem5 [Integration by parts]. f (x) and g (x) are smooth functions on [a, b] .
b b
Then the next formula is true: f ( x) g x ( x)dx f ( x) g ( x) |ba f x ( x) g ( x)dx .
a a
f ( x)dx F ( x) |a .
b
And the [Newton-Leibniz formula] can be rewritten as
a
Proof. Let’s consider the function f ( x) g ( x) , this function is differentiable on [a, b] and (according
to the standard formula) ( f ( x) g ( x))x f x ( x) g ( x) f ( x) g x ( x) the right part of the last
equality is a continuous function on [a, b] , then the same is true for the left part. Then the next
b b
integrals are defined and equal ( f ( x) g ( x))x dx ( f x ( x) g ( x) f ( x) g x ( x))dx
a a
b
f ( x) g ( x) |ba ( f x ( x) g ( x) f ( x) g x ( x))dx
a
b b b b
f ( x) g ( x) |ba f x ( x) g ( x)dx f ( x) g x ( x)dx f ( x) g x ( x)dx f ( x) g ( x) |ba f x ( x) g ( x)dx
a a a a
We have built the theory of integral and proved the most important theorems.
/2
/2
We will not consider the simpliest examples like sin xdx cos x |0 1 , because such examples
0
are in abundance in many books and internet resources. There will be several good exercises for
Reader’s practice.
For Reader’s practice:
20
cos2 x 1
[1] Show that 0 dx 7 . Solution. For any x [12,20] we have
12 1 x
8
10
20 20
cos2 x 1 1 cos2 x 1 1 8 1
0
1 x8 1 128 128
then 0 1 x8 dx 128 dx
128
( 20 12)
128
10 7
.
12 12
1 2 n
[2] f (x) is defined and continuous on [0,1] . Find the next limit lim n n f f ... f .
n n n
1
ln( f ( x )) dx
0
Answer: e .
147
Analysis
1 1 1
[3] Calculate lim n ... . Answer: ln 2 .
n 1 n 2 2n
sin x 2
[4] Calculate the integral dx . Answer: .
0 1 cos 2
x 4
[5] Calculate the integral sin x dx . Answer: 0 .
3
/2
1 2 3
[6] Calculate the integral 2 sin x
dx . Answer:
9
.
0
e
1
[7] Calculate the integral x(1 ln 2 x) dx . Answer: 4
.
1
1
[8] Calculate the integral arcsin x dx . Answer: .
0
4
/2
1
[9] Calculate the integral 1 sin x cos x
dx . Answer: ln 2 .
0
2 2
e1 / x
[10] Calculate the integral 3 dx . Answer: (e e1 / 4 ) / 2 .
1 x
1
3
[11] Calculate the integral 4 x 2 dx . Answer:
3
2
.
0
148
Analysis
Curves
Def1. x(t ), y(t ), z (t ) are some functions on [a, b] , then R(t ) ( x(t ), y(t ), z (t )) || t [a, b] is called
a vector function on [ a, b] .
For any t [a, b] the value R(t ) is a vector and let’s agree
to draw it as a radius vector [pict1]. The set Ω of points in
the space with coordinates {( x(t ), y(t ), z (t )) || t [a, b]} is called
a hodograph of R(t ) (it is exactly the set of end points of radius
vectors R(t ) || t [a, b] ). The variable t is called a parameter.
For any t [a, b] the point in the space with coordinates
( x(t ), y(t ), z (t )) is denoted like P(t ) . pict.1
Def2. R(t ) ( x(t ), y(t ), z (t )) || t [a, b] is a vector function.
And [A] x(t ), y(t ), z (t ) are smooth functions on [ a, b]
(i.e., these functions are differentiable on
[a, b] and their
derivatives are continuous functions on [a, b] ), and
[B] t1 t2 [a, b] P(t1 ) P(t2 ) [pict2] (i.e., different
values of parameter define different points in the space), and
[C] for any parameter t [a, b] the vector
v (t ) ( xt (t ), yt (t ), zt (t )) is a non-zero vector, i.e., pict.2
v (t ) 0 t [a, b] . Then the hodograph of R(t ) is called
a smooth curve, or just a curve.
A vector v (t ) ( xt (t ), yt (t ), zt (t )) is usually drawn as a vector,
which start point is P(t ) [pict3] (we will show later that such
vector is a “tangent vector” to a curve at a point P(t ) ).
P(a) is called a start point of a curve and P(b) is called
an end point of a curve.
pict.3
From the def2 follows that for any point P on Ω there exist
the unique parameter t [a, b] such that P P(t ) .
In particular, for any different points P D on , the parameters t , d [a, b] ,
where P P(t ), D D(d ) , are different t d.
Let’s make any partition of [a, b] , so a t1 t2 t3 t4 ..... tn b.
Any such partition defines some points A P(t1 ), P(t2 ), P(t3 ),.... P(tn ) B on the curve Ω.
149
Analysis
Let’s consider the polygonal chain
P(t1 ) P(t2 ), P(t2 ) P(t3 ), P(t3 ) P(t4 ),.... P(tn 1 ) P(tn ) [pict4], it’s
length (by definition) is the sum of lengths of all segments:
L P(t1 ) P(t2 ) P(t2 ) P(t3 ) P(t3 ) P(t4 ) ... P(tn 1 ) P(tn ) .
Def3. is some curve. If there exist some real number L()
such that for any (small) positive 0 there exist the positive
0 such that for any partition pict.4
a t1 t2 t3 t4 ..... tn b of [a, b] , which norm is less
than (i.e., max(tk 1 tk ) ), we have L() L ,
then L() is called a length of a curve .
In the def3 L is a length of a polygonal chain P(t1 ), P(t2 ), P(t3 ),.... P(tn ) .
Theorem1. If such number L() exists, then L() is a supremum of {L} , where {L} is a set of
lengths of all possible polygonal chains which correspond to all possible partitions:
a t1 t2 t3 t4 ..... tn b of [a, b] .
Consequence. The length L() is greater than the length L of any polygonal chain (which
corresponds to some partition of [a, b] ), in the same time, the length L of a chain can approach
arbitrary close L() .
Proof. [Step1] Let’s show that the length L of any polygonal chain is not greater than L() .
~ ~
Let’s assume that it is not true, then there exist some chain L such that L() L , such chain
corresponds to some concrete partitiona t1 t2 t3 t4 ..... tn b of [a, b] .
~ ~
Let ~ is a distance between numbers L() and L , i.e., ~ L L() 0 .
Let’s notice that if we add any one new point 1 to our partition
a t1 t2 t3 t4 ..... tn b , we will get a new partition, and
the length L of a polygonal chain, that corresponds to the new
~
partition, is not less than the length of the chain L that
corresponds to the initial partition. Really, let’s add one new point
1 between t1,t2 [pict5], then we have a new partition
a t1 1 t2 t3 t4 ..... tn b , obviously
P(t1 ) P(t2 ) P(t1 ) P(1 ) P(1 ) P(t2 ) (triangle inequality). pict.5
~
Then L L .
150
Analysis
Let’s take now the positive number ~ / 2 0 , there exist such that for any partition of [a, b] ,
which norm is less than , the length of the chain L which corresponds to such partition, satisfies
to L() L ~ / 2 .
Let’s refine the initial partition a t1 t2 t3 t4 ..... tn b by adding some new points m
. The chain L, which
between t k , t k 1 , in order to get the partition which norm is less than
~
corresponds to such partition, has the property L() L ~ / 2 and L L .
~ ~
Let’s sum up: we have [1] L() L and ~ L L() and in the same time
~
[2] L() L ~ / 2 and L L .
~
It’s easy to mark the numbers L(), L, L on some coordinate line (by taking into account [1],[2])
and to see that we have a contradiction.
~
Then the initial assumption L() L was false and the [step1] is done.
[Step2] Let’s show that for any positive 0 the half interval ( L() , L()] contains at least
one element from the set {L} (where {L} is a set of lengths of all possible polygonal chains).
0 , then there exist 0 such that for any partition, which norm is less than ,
Let’s fix any
~
we have L() L . So let’s fix any such partition, for the chain L , which corresponds to it,
~ ~ ~
we have L() L and L L() from here follows that L ( L() , L()] .
From the [Step1] and [Step2] follows that L() is a supremum of {L} .
Consequence1. From the theorem1 follows that if a curve Ω has a length L() , then the number
L() is uniquely defined. Really, if L() exists, then it is a supremum of some concrete set {L} ,
and any set may have only one supremum.
b
Theorem2. For any curve Ω the number L() exists and L() xt (t )2 yt (t )2 zt (t )2 dt
a
b
xt (t ) yt (t ) zt (t ) dt
. 2 2 2
Proof. Let’s notice that the integral do exist, because x(t ), y(t ), z (t )
a
are smooth on [a, b] (def2 [A]), then xt (t )2 yt (t )2 zt (t )2 is continuous on [a, b] and even
a strictly positive on [a, b] (def2 [C]), then the composite function xt (t )2 yt (t )2 zt (t )2
is continuous on [a, b] and therefore it is integrable on [a, b] .
Let’s fix an any partition a t1 t2 ... tn b, this partition defines the polygonal chain with
151
Analysis
length L P(t1 ) P(t2 ) P(t2 ) P(t3 ) P(t3 ) P(t4 ) ... P(tn 1 ) P(tn )
n 1
L ( x(t k 1 ) x(t k )) 2 ( y (t k 1 ) y(t k )) 2 ( z (t k 1 ) z (t k ))2 . Let’s notice that we have here
k 1
the differences of x(t ), y(t ), z (t ) at the end points of segments [tk , tk 1 ] , then we can apply the mean
value theorem. Let’s consider x(t ) on [tk , tk 1 ] , there exist some point hk (t k , t k 1 ) such that
x(tk 1 ) x(tk ) xt (hk )(tk 1 tk ) and similarly for the other functions y(t ), z (t ) .
n 1
L xt (hk )2 (tk 1 tk )2 yt ( pk )2 (tk 1 tk )2 zt (uk )2 (tk 1 tk )2
k 1
n 1
[ xt (hk )]2 [ yt ( pk )]2 [ zt (mk )]2 (tk 1 tk ) [S1] -this sum looks very similar to
k 1
the integral sum [S2] of xt (t )2 yt (t )2 zt (t )2 on [ a, b ] .
n1
[S2] [ xt ( k )] [ yt ( k )] [ zt ( k )] (tk 1 tk ) (here k [tk , tk 1 ] ). But anyway, the sum
2 2 2
k 1
[S1] is different, because points hk , pk , mk may be different, and they are determined by the mean
value theorem, these are some points which belong to the same segment [t k , t k 1 ] . And in the sum
[S2] we have only one arbitrary point k on [t k , t k 1 ] , so [S1] and [S2] are really different sums.
Let’s fix an arbitrary positive 0.
[Step1] We will use inequality [E] a b a b a 0, b 0 .
As xt (t ), yt (t ), zt (t ) are continuous on [a, b] , then [ xt (t )]2 , [ yt (t )]2 , [ zt (t )]2 are also continuous on
[a, b] , then these functions are uniformly continuous on [a, b] . Then the for the positive number
2
there exist 0 such that for any t1 , t2 [a, b] , as soon as t1 t2 , then, right
12(b a) 2
away:
2 2 2
[ xt (t1 )] [ xt (t2 )]
2 2
, [ yt (t1 )] [ yt (t2 )]
2 2
, [ zt (t1 )] [ zt (t2 )]
2 2
.
12(b a)2 12(b a)2 12(b a)2
Let’s fix an arbitrary partition of [a, b] which norm is less than , we need to estimate how close is
the sum [S1] to the sum [S2].
| L |
n 1
k 1
[ xt (hk )]2 [ yt ( pk )]2 [ zt (mk )]2 [ xt ( k )]2 [ yt ( k )]2 [ zt ( k )]2 (tk 1 tk ) [inequality [ E ]]
152
Analysis
n 1
k 1
[ xt (hk )]2 [ xt ( k )]2 [ yt ( pk )]2 [ yt ( k )]2 [ zt (mk )]2 [ zt (mk )]2 (tk 1 tk )
n 1
[ xt (hk )]2 [ xt ( k )]2 [ yt ( pk )]2 [ yt ( k )]2 [ zt (mk )]2 [ zt (mk )]2 (tk 1 tk )
k 1
n 1
2 2 2 (t t )
n 1
(tk 1 tk ) (b a) / 2
k 1 12(b a)
2
12(b a) 2 12(b a) 2 k 1 k
2(b a ) 2(b a )
k 1
b
[Step2] From the main definition of integral xt (t ) yt (t ) zt (t ) dt
2 2 2
follows that:
a
~
for / 2 0 there exist some positive such that for any partition of [a, b] , which norm is less
~
than , we have / 2 (here is an integral sum [S2]). And from the [step1], for any such
partition there also must be | L | / 2 .
Let’s sum up: from L / 2 and / 2 follows that | L | .
~
We had started from an arbitrary positive , and we showed that there exist some 0 such that
~
for any partition, which norm is less than | L | .
, we have
Then, according to the def3, is a length of the curve . Everything is proved.
Consequence2. Any point С on divides in two curves AC and CB [pict6] .
From the simpliest properties of integral and theorem2
we have: L() L( AC ) L(CB ) .
Let’s take any с d [a, b] . The length of the curve with
end points P(c) and P(d ) is equal to
d
xt (t ) yt (t ) zt (t ) dt [F]. The function
2 2 2
(t ) xt (t )2 yt (t )2 zt (t )2 is continuous on [ a, b]
and therefore it is continuous on [c, d ] , then it reaches it’s pict.6
maximum [ c , d ] and it’s minimum m[ c , d ] at some points of the segment [c, d ] . From the def2 [C]
follows that (t ) is strictly positive on [c, d ] . As there exist some points u, z [c, d ] such that
(u) [c, d ] and ( z ) m[c, d ] , then [c, d ] 0 and m[ c , d ] 0 . So we have
0 m[c, d ] (t ) [c, d ] || t [c, d ] . And obviously 0 m[ a,b] m[c, d ] and [c,d ] [ a,b] .
153
Analysis
Then 0 m[ a,b] (t ) [ a,b] || t [c, d ] . Let’s designate m[ a ,b ] m and [ a ,b ] . Then we have
0 m (t ) || t [c, d ] . Let’s estimate the integral [F] (the length of the curve between P(c)
d
and P(d ) ): 0 m (d c) xt (t )2 yt (t )2 zt (t )2 dt (d c) [V].
c
From [V] follows that: if the values of parameters с d
are very close, then the length of the curve between P(c)
and P(d ) is a very small number. And conversely,
if the length of the curve between P(c) and P(d ) is a
small number, then с must be close to d . We got [V]
from the assumption с d [a, b] .
If d c [a, b] we can get the similar formula.
And we can unite these results in one: for any
с, d [a, b] || с d , the length of the curve between
pict.7
d
xt (t ) yt (t ) zt (t ) dt
2 2 2
P(c) and P(d ) is and
c
d
the next estimation is true 0 m d c xt (t ) yt (t ) zt (t ) dt d c [U] [pict7]
2 2 2
(it’s very easy to check that this estimation is true in any case: с d , d c , the case c d is not
allowed, we do not need it. And our reasoning was done for different values of parameter с, d ).
And obviously, for any (a, b] the length of the curve between P(a), P( ) is
L( ) xt (t )2 yt (t )2 zt (t )2 dt . As we have a strictly positive function inside the integral,
a
then L( ) is a strictly increasing function on [a, b] . (it means that when moves from a to b ,
the length of the curve between P( ) and P(a) grows, which makes sense).
Exercise1. Show that L( ) is continuous on [a, b] (use the estimation [U]).
Conclusion. The length function L( ) is strictly increasing and continuous on [ a, b ] .
From now on, the points on a curve, which correspond to parameters c, d , we denote as C, D
(i.e., C P(c), D P(d ) ).
154
Analysis
Def4. is a curve and D is a fixed point on .
The line L, which passes through the point D ,
is called a tangent line to at the point D if:
for any (small ) positive 0 there exist 0 such
that for any point C on the curve, where
0 L(DC ) , the angle (C ) ( L, DC ) , i.e.,
the angle between L and DC (in radians)
is less than [pict8].
Exerise2. From the def4 follows that if there exist a pict.8
tangent line L at some point D , then it is unique (there
can’t be two different tangent lines at the same point D ).
Def5. L is a line, any non-zero vector v , which lies on L, is called a direction vector of P(d ) .
Theorem3. For any point P(d ) on Ω, the tangent line L at P(d ) exists.
And v (d ) ( xt (d ), yt (d ), zt (d )) is a direction vector of L.
Proof. As there can be only one tangent line at any point, it’s enough to show that the line L, which
passes through D and which direction vector is v (d ) ( xt (d ), yt (d ), zt (d )) , is exactly the line we
need (we will show that this line L satisfies to the def4 and therefore L is a tangent line).
Let’s fix an arbitrary positive 0 and any point D on the curve. The line DC goes through the
points D,CΩ, so we can call it “the secant DC”. And DC is a direction vector of the secant DC.
Obviously R(d ) DC R(c) DC R(c) R(d ) . We have
R(c) ( x(c), y(c), z (c)) and R(d ) ( x(d ), y(d ), z (d )) , then
DC ( x(c) x(d ), y(c) y(d ), z (c) z (d )) as C and D are
different points on the curve, then c and d must be different
1
numbers, therefore the number is defined (it can be
cd
a positive or a negative number, it doesn’t matter).
Let’s consider the vector which is proportional to DC ,
DC x (c ) x ( d ) y (c ) y ( d ) z (c ) z ( d )
the vector , ,
cd cd cd cd
it is still a direction vector of the secant DC .
The angle (C ) is the angle between vectors [pict9] pict.9
v (d ) ( xt (d ), yt (d ), zt (d )) and
155
Analysis
DC x (c ) x ( d ) y (c ) y ( d ) z (c ) z ( d )
, , .
cd cd cd cd
DC
Notice that v (d ) is a constant vector, all it’s components are fixed numbers. And is not a
cd
constant vector, it’s position is completely defined by the parameter c (and d is a fixed parameter,
because D is fixed).
DC
Let’s write the dot product of vectors v (d ) and in two different ways:
cd
x (c ) x ( d ) y (c ) y ( d ) z (c ) z ( d )
сos (C ) xt (d )
DC
v (d ) yt (d ) zt ( d )
cd cd cd cd
x (c ) x ( d ) y (c ) y ( d ) z (c ) z ( d )
xt (d ) yt (d ) zt ( d )
сos (C ) cd cd cd [T]
2 2 2
x (c ) x ( d ) y (c ) y ( d ) z (c ) z ( d )
[ xt (d )]2 [ yt (d )]2 [ zt (d )]2
c d c d c d
The expression on the right side of [T] is a function of a variable с that is defined for any
c [a, d ) (d , b] . It’s easy to understand (from the simpliest properties of limits) that
the right part of [T] has a limit 1 when с d , this expression is a function of a variable c,
let’s denote it (c) . As (c) is always a cosine of some angle, we always have 1 (c) 1.
Let’s sum up: we have [T] сos (C ) (c) || 1 (c) 1 || c [a, d ) (d , b] , then we can take the
arccos() of both sides of the first equality of [T],
arccos(сos( (C ))) arccos( (c)) (C ) arccos( (c)) . Then (C ) is a composite function
arccos( (c)) . As lim с d (c) 1 and arccos y is left continuous at y 1, then
(theorem about a limit of a composite function) the limit lim с d arccos( (c)) exists and equal to
arccos1 0 , then lim с d (C ) 0 .
Let’s fix an arbitrary positive 0 , there exist the positive such that as soon as 0 | c d |
we have (C ) 0 (C ) (because (C ) is always a positive angle).
We have the estimation [U] 0 m d c L(CD ) d c . Let’s take the positive number
m , then for any point C on the curve, for which L(CD ) m , then we have
m d c m d c and therefore (C ) .
156
Analysis
So, for any positive 0 , we can always find the positive m such that if L(CD ) m ,
then (C ) .
We showed that the line L, which direction vector is v (d ) ( xt (d ), yt (d ), zt (d )) , satisfies to the
def4, then L is a tangent line.
A graph of any smooth function f (x) on some
segment [a, b] can be considered as
a 2-dimentional curve , i.e., [pict10] the graph
of f (x) on [a, b] is the same as a hodograph of
the vector function R( x) ( x, f ( x)) || x [a, b] .
Notice that the graph of f (x) is really a curve,
because it satisfies to the def2.
Let’s fix any point D on the graph of f (x) ,
it has some coordinates (d , f (d )) . pict.10
According to the theorem3, there exist a tangent line L at this point, and it’s direction vector is
v (d ) (1, f x (d )) . In general, let L is a line on the plane which direction vector is (a, b) || a 0 .
Then for the angle between L and the positive direction of Ox we have tg b / a (the number
tg is called a slope of a line L). In our case, the direction vector is v (d ) (1, f x (d )) , then we have
the slope tg f x (d ) . Then the derivative of a smooth function f ( x) || x [a, b] at any point
d [a, b] is a slope of a tangent line (to the graph of f ) at the point (d , f (d )) .
Exercise3. Does there exist any curve which passes through every point of some unit cube?
Hint. Show that any curve has zero volume, i.e., for any 0 it can be covered by several
rectangular parallelepipeds, which total volume is less than .
[1] Find the length of the curve:
x (t 2 2) sin t 2t cost , y (t 2 2) cost 2t sin t || t [0, ]. Answer: 3 / 3
[2] Find the length of the curve:
x cos3 t , y sin 3 t , z cos 2t || t [0,2 ] . Answer: 10
[3] Find the length of the curve: y ln sin x || x [ / 3, / 2] . Answer: ln 3 / 2
[4] Find the length of the curve: x cost t sin t , y sin t t cost || x [0, / 3] . Answer: 2 / 18
157
Literature
Literature
1. Course of lectures of mathematical analysis.Part 1. V.I. Kolyada, A.A. Korenovskii.
Odessa, 2009.
2. Differential geometry, E.G. Poznyak. MGU, 1990.
3. Course of differential and integral calculus,Volume 1. G.M. Fihtengolz. FizMatLit, 2001.
4. Linear algebra, E.S. Kochetkov, A.V. Osokin. Moscow, 2012.
5. Linear Algebra. A.I. Kozko. MGU, 2018
6. Lectures on algebra, D.K. Fadeev, Moscow, 1984.
7. Mathematical Analysis. A.I.Kozko. MGU 2017
8. Course of mathematical analysis. A.M. Ter-Krikorov, M.I. Shabunin. FizMatLit, 2001.
9. Numbers and polynomials. I.V Proskuryakov. Moscow, 1965.
10. Mathematical analysis, Part I. M.V. Falaleev. Irkutsk, 2013.
11. Collection of problems in mathematical analysis. Parts 1,2. L.D.Kudriavcev, A.D.Kutasov,
V.I.Chehlov, M.I.Shabunin. Fizmatlit, 2003.
12. Mathematics for Jee Main & Advanced. Book2, Book4. Wiley, Plancess. Mumbai 2015.
13. Encyclopedia of elementary mathematics, Volume 4, P. Aleksandrov, A. Markushevich,
A. Hinchin. Leningrad, 1953.
14. Multiple and curvilinear integrals, G.M. Fihtengolz.
15. Encyclopedia of elementary mathematics, Volume 5. P. Aleksandrov, A. Markushevich,
A. Hinchin. Leningrad, 1966.
16. The video “Complex Numbers as Matrices”. Youtube, channel Mathoma.
17. Technique of calculating limits, G. S. Shukova. RHTU, 2005.
18. Theory of functions of a complex variable in problems and exercises, L.N. Posicelskaya,
Moscow, 2007.
158
You might also like
- Smith Hubert The Illustrated Guide To Aerodynamics100% (6)Smith Hubert The Illustrated Guide To Aerodynamics244 pages
- Blennow M Solutions Manual For Mathematical Methods For PhysNo ratings yetBlennow M Solutions Manual For Mathematical Methods For Phys348 pages
- Le Gall JF Measure Theory Probability and Stochastic Process100% (1)Le Gall JF Measure Theory Probability and Stochastic Process409 pages
- Evstigneev M Introduction To Semiconductor Physics and DevicNo ratings yetEvstigneev M Introduction To Semiconductor Physics and Devic325 pages
- Aliabadi Aa Turbulence A Fundamental Approach For Scientists100% (1)Aliabadi Aa Turbulence A Fundamental Approach For Scientists296 pages
- Deshmukh PC Quantum Mechanics Formalism Methodologies and AP100% (3)Deshmukh PC Quantum Mechanics Formalism Methodologies and AP659 pages
- Holmes MH Introduction To Differential Equations100% (1)Holmes MH Introduction To Differential Equations248 pages
- Sobot R Engineering Mathematics by Example91% (11)Sobot R Engineering Mathematics by Example474 pages
- Kaufmann Emil Architecture in The Age of Reason Baroque and100% (2)Kaufmann Emil Architecture in The Age of Reason Baroque and380 pages
- Tornetta GN Mathematical Quantum Physics A Foundational Intr100% (3)Tornetta GN Mathematical Quantum Physics A Foundational Intr187 pages
- Davis Basil S The Basic Physics of Quantum Theory100% (3)Davis Basil S The Basic Physics of Quantum Theory220 pages
- VECTORS and ANALYTIC GEOMETRY in SPACE PDF100% (1)VECTORS and ANALYTIC GEOMETRY in SPACE PDF120 pages
- Basic Electronic Circuits - Problems and SolutionsNo ratings yetBasic Electronic Circuits - Problems and Solutions304 pages
- Givental A. - Linear Algebra (2016, Sumizdat)No ratings yetGivental A. - Linear Algebra (2016, Sumizdat)218 pages
- Vectors in The Plane Zambak 1st Edition Ertuðrul Tarhan PDF Download100% (1)Vectors in The Plane Zambak 1st Edition Ertuðrul Tarhan PDF Download44 pages
- Mathematical Tripos: IA Vectors & Matrices100% (1)Mathematical Tripos: IA Vectors & Matrices148 pages
- SQA Higher Mathematics Unit 3: SCHOLAR Study Guide100% (1)SQA Higher Mathematics Unit 3: SCHOLAR Study Guide170 pages
- A Primer of Quaternions - Arthur S. HathawayNo ratings yetA Primer of Quaternions - Arthur S. Hathaway70 pages
- (Made Simple Books) William R. Gondin, Bernard Sohmer - Advanced Algebra and Calculus Made Simple (1959, Doubleday) - Libgen - LiNo ratings yet(Made Simple Books) William R. Gondin, Bernard Sohmer - Advanced Algebra and Calculus Made Simple (1959, Doubleday) - Libgen - Li228 pages
- Introductory Mathematics. Applications and MethodsNo ratings yetIntroductory Mathematics. Applications and Methods232 pages
- #Introductory Mathematics - Applications and Methods - Gordon S. Marshall - Springer (1998) .1100% (1)#Introductory Mathematics - Applications and Methods - Gordon S. Marshall - Springer (1998) .1233 pages
- A. S. Solodovnikov, G. A.Toropova - Linear Algebra With Elements of Analytic Geometry - Mir - 1990100% (1)A. S. Solodovnikov, G. A.Toropova - Linear Algebra With Elements of Analytic Geometry - Mir - 1990300 pages
- A Comprehensive Guide To Vector AnalysisNo ratings yetA Comprehensive Guide To Vector Analysis364 pages
- (Sheldon Axler) Linear Algebra Done Right Ch01 EspaciosVectoriales100% (1)(Sheldon Axler) Linear Algebra Done Right Ch01 EspaciosVectoriales20 pages
- Math 302 Lecture Notes Kenneth Kuttler - Download The Ebook Now For An Unlimited Reading ExperienceNo ratings yetMath 302 Lecture Notes Kenneth Kuttler - Download The Ebook Now For An Unlimited Reading Experience62 pages
- Trolling Euclid - An Irreverent Guide To Nine of Mathematics Most Important Problems (2016)No ratings yetTrolling Euclid - An Irreverent Guide To Nine of Mathematics Most Important Problems (2016)205 pages
- Introduction To Vector and Tensor Analysis: Jesper Ferkinghoff-Borg September 6, 2007No ratings yetIntroduction To Vector and Tensor Analysis: Jesper Ferkinghoff-Borg September 6, 200761 pages
- Introduction To Vector and Tensor Analysis: Jesper Ferkinghoff-Borg September 6, 2007No ratings yetIntroduction To Vector and Tensor Analysis: Jesper Ferkinghoff-Borg September 6, 200761 pages
- Section 3A Vector Representation-Euclidean SpaceNo ratings yetSection 3A Vector Representation-Euclidean Space11 pages
- Zucker Paul Styles in Painting A Comparative Study100% (5)Zucker Paul Styles in Painting A Comparative Study356 pages
- Koranga Bs Padaliya SK Nautiyal VK Special Functions and The100% (1)Koranga Bs Padaliya SK Nautiyal VK Special Functions and The124 pages
- Naumov Ip A Course of Mathematics Book1 Construction of Numb100% (2)Naumov Ip A Course of Mathematics Book1 Construction of Numb171 pages
- Porter Richard N Quantum Field Theory An Introduction For CH100% (1)Porter Richard N Quantum Field Theory An Introduction For CH282 pages
- Shah NH Panchal J Elementary Vector Calculus and Its ApplicaNo ratings yetShah NH Panchal J Elementary Vector Calculus and Its Applica228 pages
- Ross SM Pekoz Ea A Second Course in Probability100% (1)Ross SM Pekoz Ea A Second Course in Probability191 pages
- Dehli M Doering e Schedwill H Fundamentals of Technical TherNo ratings yetDehli M Doering e Schedwill H Fundamentals of Technical Ther622 pages
- Fabrizio M A Course in Quantum Manybody Theory100% (3)Fabrizio M A Course in Quantum Manybody Theory350 pages
- Bruma P Introduction To Relativity Indepth and Accessible Vo100% (3)Bruma P Introduction To Relativity Indepth and Accessible Vo293 pages
- Lucke Charles e Father John J Handbook of Thermodynamic Tabl100% (3)Lucke Charles e Father John J Handbook of Thermodynamic Tabl261 pages
- Bellantoni L Modern Physics The Scenic Route100% (3)Bellantoni L Modern Physics The Scenic Route205 pages
- Final Assignment & Solution: MEC 231 Basic Mechanics INo ratings yetFinal Assignment & Solution: MEC 231 Basic Mechanics I39 pages
- Practice Exercises: Graphing Functions Using MATLAB100% (1)Practice Exercises: Graphing Functions Using MATLAB2 pages
- Arts Lesson Plan Template: Developed by CPS Department of Arts EducationNo ratings yetArts Lesson Plan Template: Developed by CPS Department of Arts Education9 pages
- Course Material On CNC Turning and CNC MillingNo ratings yetCourse Material On CNC Turning and CNC Milling108 pages
- Newspaper Advertising: A Guide To Newspaper Display Ad SizesNo ratings yetNewspaper Advertising: A Guide To Newspaper Display Ad Sizes15 pages
- To Find Chord of A Windmill Rotor Blade As A Function of Radius RNo ratings yetTo Find Chord of A Windmill Rotor Blade As A Function of Radius R6 pages
- Fia 2025 Formula 1 Technical Regulations - Issue 01 - 2024-12-11 1No ratings yetFia 2025 Formula 1 Technical Regulations - Issue 01 - 2024-12-11 1178 pages
- Silo - Tips - Icem Surf Quick Reference November Dassault Systemes All Rights ReservedNo ratings yetSilo - Tips - Icem Surf Quick Reference November Dassault Systemes All Rights Reserved37 pages
