I

DOCUMENT. RESUME

ED 214 787

.SE 0 6478

AUTHOR
TITLE
INSTITUTION
SPONS AGENCY
PUB DATE
GRANT
NOTE ,.;

Bchoenfield, Alan H.; And Others

UMAP Modules-Units 203-211, 215-216, 231-232.
Education Development Center, Inc., Newton, Mass.
.National Science Foundation, Washington, D.C.

EDRS PRICE
DESCRIPTORS.'

MF01 Plus Postage. PC Not Available from DRS.

Biology; Calculus; .Chemistry; *College athematics;
'Economics; Engineering;.Geography; Harvesting; Higher,
Education; *Learning Modules; *Mathematical
Applications; Mathematical Models; Medicine; Physics;
*Problem Solving; Social Sciences; Supplementary
Reading Materials; Topology
*Integration (Mathematics); *Linear Algebra

80

SED76-19615; SED76-19615-A02
382p.; Contains occasional light type.

ABSTRACT
,

One module is presented in units 203, 204, and 205,

as a 9uide for students, and presents a general strategy for solving
integrals effectively. With this material is a solutions manual to,
exercises. This document set also includes a unit featuring
applications o'f calculus to geography: 206-Mercator's World Map and
thelCalculus. Unit 207-Management of A Buffalo Herd, fewtures,,a
;eslie-type model covering applications of linear algebra tb
harvesting. Two units include applications of linear-algebra to
'economicv 208-Economic Equilibrium-Simple tinear"Models, and
209 Genial Equilibrium-A Leontiet Economic Model. Unit 210-ViCou
Fluid FloW,And the Integral Calculus, contains applications of
'caloUius to ',engineering. Moduli 211-The Human- Cough, views calculus
applidatiOns to physics, biologidal,, and mediCal sciences. Social
science applications of calculUs.are viewed in 215-Zipf's Law and His
Efforts-to Use Infinite Series-in LiaggIstteS: Unit 216-Curves and.
their Parametrization, and '231 -The Alexander Horned Sphere, focus on
introductory topology. .Finally,-232- Kinetics of Single Reactant
Reactions', views calculus' applications to chemistry. (MP)

of

a' 4 *** #*******************************************************************

,.

*,

Reproductions supplied by EDRS are the, best that can be made

*
*
.
from the original document.
a
***************************************4r*******************************
4
.

ti

umap

Units203, 204, 205

MODLILESAND MONOGRAPHS IN tarDERORADUATE

MATHEMATICS AND ITS APPLICATIONS PROJECT
1

US. DEPARTMENT OF EDUCATION

NATIONAL INSTITUTE OF EDUCATION
leiD 'CATIONALReSOURCES INFORMATION

.
'

CEVTER (ERIC/

This document has been reproduce as

received from the person or organization
originating it
..2, Maw changes have been made to improve
reproduction quality
Points of view or opu4ons stated in this docV
ment do not necessarily represent official NIE

position or policy

Alan H. Schoenfeld
"PERMISSION TO REPRODUCE THIS

MATERIAL IN MICROFICHE ONLY

HAS BEEN GRANTED BY

P. 13A SE5
INTEGRATION:
r

TO THE EDUCATIONAL RESOURCES

INFORMATION CENTER (ERIC)."

Getting It All Together

A guide for students, presenting

general strategy for solving

integrals, effectively

do.

June 1977

eldc/umap/p5chapeittinewton,mbss.02160'
st

a.

.t

Intermodujar Description Sheet:

Unit 203, 204, 205

1lODULE1 AND MONOGRAPHS IN UNDERGRADUATE

GETTING IT ALL'TOGETHER

INTEGRATION:

Tftle:

MATHEMATICS AND ITS APPLICATIONS PROJECT (UMAP)

Author/Correspondent:

RevieWiStage/Date:

Alan H. Schoenfeld
Group in Science and Mathematics Education
c/d Physics Department
'University of California
'
Berkeley, CAL94720

IV, 6/30/77

The goal of UMAP Is to develop,. through a community

of users and developers, a system of instructional modules
in undergraduate mathematics which may be used to supplement existing courses and from which complete-courses may
eventually be built.

\
.

Calculus, Indefinite Integration, Review

Classification:

Second quarter calculus class at University of

Use Experience:
California, Berkeley. See paper by A. Schoenfeld,
"Presenting a Strategy for Indefinite Integration,"
Sesame, University of California, for details.

The Project is guided by a National Steering Committee of mathematicians, scientists, and educators. DMAP is
To of many projects of,Education Development Center, Inc.,
a publicly supported, nonprofit corporation engaged in
educational research and development in the U.S. and
abroad.
t
PROJECT STAFF

7 hours

length:

Suggested Support Material:

Ross L. Finney
William U. Walton
Solomon Garfunkel
Joyce Taggart
Thomas Good

None

Be able to'tolve problems in indefinite

integration knowing that a specific technique (i.e
substitution, paetiaI fractions, integration by parts,
trigonometric substitution) is apRlicable. Recall and use
the essential formulas given in the tableton the inside of
th-eback cover.

Prerequisite Skills:

I a.,

W. T. Martin
Steven J..firams

Llayron Clarkson

Skills:. Given an indefinite integral problem solvable.by

one of the above techniques, find a technique which is
appropriate and solve it. Specifically,

James 0. Forman/'

'

As

Robert H. tamarin
Alfred B. 01-1ptox.

None'

r--

rnell University
Unlversityt f Michigan Press
Michigan State University
Indiana University
"
Boston University
Mathematical Association
, of America
Noetheastgrn University

413oTexas

George Spriger

'

MIT (Chairman)
New York University
Southern University''
Rochester Institute of
technology

University of Muston

Ernest J. Henley
William F. Lucas
Walter E. Sears
Peter Signet!

'Simplify integrals-by algebraic substitution

Classify integration problems into the appropriate technique
Modify integration problems so that they can be'classified
And solved by one of these techniques!

Other Related Units:

(N:0,;1:

NATIONAL STEERING COMMITTEE

Output

Codirector
Codirector
Consortium Coordinator
Editorial Assistant
Editorial Assistant

;kiency J. Kopill

This material was.prepared with the support 9f

National Science Foundation Grant No'. SED 76- 19615. Rec-'
ommendations expressed are thoie of. the authors and do not
necessarily reflect the views of the NSF, nor of the
National Steering Committee.

r
4

-,, 5

0111.

ti

INTRODUCTION

J.

In order to solve problems in indefinite integration

gables of Integrals '

inside cover-r

4.

effectively, students need both a mastery of the special techniques

of integration and a general procedure for choosing and applying

Contents '

these techniques to problems.

11-

Introduction $I4

How to use these materials

time in this subject area is devoted to teaching and practicing the

,.Chapter 1 : SIMPLIFY'

Most textbook'space and classroom

special techniques of integration.

It is.generally issumed.tBat

with much practice and the help of insightful comments from their

Sec. 1; Easy Algebraic Manipulations.

teachers, students will develop a,"feel" for the material that

enables them to solve problems effectively.

1.

-Sec. 2: Obvious Substitutions

Exercises

Sec. 2:.Products

problems in integrati4en systematically and effectively: even after

Is

.Sec. 1: Rational Functions

16

. ... .

27

31

Scic-3:Trigonometric Functions
Sec. 4: Special Functions

lengthy classroom discuisions of" problem solutions.

Tolovercome this problem, this, booklet provides students

directly'with a general procedure for approaching"andsolving

'problems in integration.

37

Exercises

Chapter4-: MODIM

however, many students havhad difficulty learning to approach

14

Chapter 2 : CLASSIFY'

..

Based on observations of "experts"

working on integrals, the procedure has three steps: SIMPLIFY,,

44,
.

In my experience,'

CLASSiFYi and MODIFY.

Sec. 1: Problem SimilaritieA

Sec. 2: Special Manipulations

In step 1, kMPLIFY, we try to reduce a problem to one which

.

can be Solved by a formula or can be done easily. -'if this fails

.

to solve the problem we proceed to step 2, CLASSIFY.

fSeC. 3: Needs Analysis

Here we use the

., form of the integrand to decide which special technique (integration

,

Exercises
Appendix I : Pre-Test
Appendix II: Answers to Pre=Tcst.

62

by parts, by partial fractions; etc.) to use on the_problem.' If we

Al

are unable to CLASSIFY the integrand, we go to step 3, MODIFY. There

we tr%to:manipulate the integrand into a more familiar oi manageabld

: A2

form.

Tables of Integrals

A3

Detailed4Cdtline of the,General Procedure

A4

We alMays check for simple alternatives before begi9ing

complicated calculations, and start the process over with step 1,,

Intermodular Description Sheet.

-

SolutionsMapual

whenever. we have succeeded in transforming the integral to something

easier., The general procedure is outlined in the
table on page 4,

Front Cover

and summarized in full detail on the,last page of this booklet.

:Bound

.Separately

I
4

VP'

.3

/HOW TO' USE THESE MATERIALS

A

-Work the pre-test in Appendix I.

These materials are written

If

for people who have mastered the basic techniques of integration.

you miss more than one of the pre -test problems, of if you find them

INTEGRATION: e

difficult, you should review your textbook's sections on basic anti-

A GENERAL PROCEDURE

derivatives and substitutions before you start Chapter 1.

Before you

work on Chapter 2, you should be familiar with the techniques of

partial .fractions, integration by-parts, and trigonometric substi-

Proceed from one step to the next when the .techniques of

that step fail to solve the problem. Always look for easy
alternatives before beginning comptizated calculations.
If you succeed in transforming the problem to something
easltr, begin again at Step 1. .'

tutions.

1 This booklet is organized like the General Procedure, gin in

the chart on page 3.

The three chapters in the booklet and the

sections the are divided into correspond to the three steps in the
general procedure and their subdivisions.

You should work through

this booklet following.the procedure,closely, until

automatic.

it becomes

If it does; you will be ableto solve problems-in

integration like an expert.

/

Btep 2:.

Rational
.Functions

,,,"CLASSIFY!

!Products

Each section begins with a description of some technique of

integration, which is summarized in table fop.

Trigonometric
Functions

by sample problems, which serve }s review problems and epmples. -Iyal

Special
Functions
.

'should try to solve each sample problem yourself. Then compare your

(1.

answer with the solution _given.

MODIFY!
ti

will not be enough!

.

.;

Special
C Manipulations

which is as important as the answer.

'

Stitilarities

Just reading through the solutions

You should focus on the process of solution,

Step 3:

Problei.

The table is followed

heeds
Analysis

Each chapter.ends.with exercises-designed to reinforce-the

procedures you have just learned.
were a test.

Work the exercises as if they

Detailed solutions are in a separate solutions manual.

Rote: It's easy to "lose" terms in an,intergal if we're not careful..

I've chosen to write all the terms in an integral at each stage of
the process, Apct I suggest you do the tame.

time, but it helps prevent Iltly mistakes..

--

This-tikes some extra (\

Section 1

Chapter 1
.0

EASY ALGEBRAIC
MANIPULATIONS

SIMPLIFY!
40111111,1'

Sole algebraic manipulations are easy enough to use that

There is 'one general rule that you shodld keep in mind

it's worth considering them'autoitaticaily before going oh to

whenev r you are solving problems:

k1/4

lyALWAYS CHECK FOR EASY ALTERNATIVES

'anything else.

For example, We almost always break the integral

of. a sum into

sum of integrals and then integrate term by term.

Before doing this, however, we should lodk fOr other alternatives.

BEFORE BEGINNING ANY CCO4P.LICATED

Sometimes an algebraic or.trigodometric identity will siMplify

OR TIME- CONSUMING OPERATIONS,

the term facing us, before we try to integrate it.

Another operation

which. is more complicated but also worth considering is simplifying

I

As the sample peoblems below illustrate;

t is worth taking

a few moments to look for h Ruicksor easy solution to a problem

before jumping into a compiicated prce/re.

This is especially

rational functions by long division.

'We call a rational functiOn (thb quotient of two pOlynemials)
a "proper fraction"

if the degree of the numerator is less than

true in integration, where a timely observation can save tremendous

the degiee oP the denominitor.

amounts of work. The twb types of SIMPLIFYing opeiations we will

to manipulate than others.

Proper fractions are usually easier

discuss are summarized below.

of partial fractions to proper fractions.- Thus we.should consider

Also, we can only apply the technique

division as a preliminary simplification.

Step 1:

In sum,we hive:

SIMPLIFY

EasyAlgehraic
Manipulations

' ObviOus
Substitutions

EASY ALGEBRAIC MANIPULATIONS

(1) Break integrals into sums
(2) Exploit Identities

(3) Reduce rational functions to

Proper Fractidns by division

10

11

a
Solulions, Continued

SAMPLE PROBLEMS

Each.of the following sample problems canbe SIMPLIF4ED

Try to solve each problem

by an easy algebraic manipulation.

x + cos x)2 dx

The first thing we should notice is that the integral can't

be done directly, so some sort of manipulation is called for.
If we square the term (sin x + cos x), we obtain
2

before you read the solution, and then compare your method with

+ 2 sinx cos x + cos x) dx.

f(sin

.mine.

STOP!
While the integral can be broken into three terms and
each done separately, there are simplifications. Do you see them?
2

1 + sin x

2.

dx

'cos2x

Ssin

x + cos x)

Recalling the trig.identities (sin x + cos x = 1) and

(sin 2x = 2 sin z cos x), we can write the above as

dx

v(sin x + cos x) dx + /2 sin x cos x dx

1

= x -

dx + fsin 2x dx

cos 2x + C.

x3

NOTE: The terml2sin x.cos x dx can also be solved by the

substitution u=sin x. or u = cos.x.
These give two equivalent
solutions to the problem,

..,

it+ sin x +

and

x - cos2x + C.

SOLUTIONS

rl

+ sinx
-

cos x

This integrOd contains i,_sum, so we should consider breaking

the problem into a sum tf integrals. This gives us

3
2

jr

rSin X

--2- dx

.?(

cos x

fisec 2

x dx

jrsin.x
+
2

cos x

dx

The integrand inthis problem is an "improper fraction",

so we should perform a division. The division gives us a
quotient of (x) and a remainder of (-x), so we obtain'

cos x

.1

The first integral can now be done directly. In the second,

we notice that the denominator contains the term cos x,
Since the nunerator is sin x , which (except for A minus
sign) is the derivative of cos x, this suggests that we
make the subStitutions
.

du = -sin x dx.

Then the integrals become

isec2x dx

= tan x

.%. -sin x dx

tan

-A-20u

tan x - (-u
=

1)

1,nce,

=Rx

+ 1

dx -

1 fau
2

4.

= tan x

dx

fx

x_

idu

cos x

3
2

+ 1

dx =fx dx
x

dx.

In the second integrand, wenotice that the numerator is one

half, the derivative of the denominator. If we make the
substitUtions i= (x2+1), du = (2x dx), the above becomes

= cos x,

dx
+ 1

+C

tan x +

tan x + sec x + C.

1^

lnlui + C

In fx2+11 + C.

+C

uA

13

SeMoil

10

c
1

x2_9

Note: If the problem I just discussed were

substitution

u = x2 -9 'would not have helped.

substitution u = f(x)

dx, the.

In general, a

will only help if you cantfind the tern

'1

"OBVIO US" SAiliSTITUTIOWS-

If you try

du = f'(x)dx somewhere in the integral

substitution

and it looks like you're getting involved in a complicated procedure,'

The procedurei of chapter 1

stop to consider other alternatives.

Using substitutions is one of the...dosi powerful tools welave
for simplifying and solving integrals,

I always look for substituThere are two guidelines

tions before.I try more complex procednres.

are designed to help SIMPLIFY and solve an integral rapidly:

You

should explore all Simple alternatives before trying anything

complicated.

Ifineed be, you can always return to a complicated

substitution later.

I use in looking for substitutions:

(1) Does the integrand contain a function ofa,function?

OBVIOUS SUBSTITUTIONS

It it does, try a substitution with u ...as the_-"inside" function.

if

Consider the'integrals

(1) "Inside" functions

(2) "Nasty" terms and
denominators

Ix'tan-1(x2)
so

1 + x

and

dic

.cos 2x

.The

term

tan

-1

(x ) appears in

first integril, with" x

I mould try the substitution

inside function.
problem.

the

2'

as an

u = x2 in that

SAMPLE PROBLEMS

The denominator of the second integral iscOs2x= (cos x)2,

an'tifride function.

r Would try 'u = cos x.

Each of problems 1 through 3 can be.solved by a

substitution.

Try to solve each problem before you read the

'solution, and then compare your method with mine.

(2) Does theiintegrand contain a complicated or "nasty"

function, particularly in tht denominator of;a fraction?

trya substitution with

as the "nasty" function.

If so,

(tan-lx

t an-1 x

dx

Consider

x.1/ x2+2
'I x2+ 1

1 + x
-1

(fx :2 +2

x +,x) .-.2--)dx
x +1

and 'jr-Tr---dx:
)(tan

- 9

In the first problem I would try u = (tan -lx + x), and hope that
it helps. [It does; see sample problem 2.]
the denominator .isn't

substitution

1 Ir
2

2x dx

x2 - 9

u = x
s

particularly
- 9. Then

7iydu
17.-

In the second problem

"nasty", but it's worth trying the

du =,..2x dx, and the integral is

4.

-x

ex +
e

dx

- e

One of the following two integrals is much easier to solve

than the other. - Decide which it is, and solve it.

1

lnix2-91 + C.

(a)

4 5

x (1+x )

dx

(b) jr(l+x4)5 dX

11.

15

:ors

11

12

SOLUTIONS

4*"

7,

+ e

-1
e

- e

Is.

-x

dx
2

1
I'd like to work this problem using all the methods of,
this chapter, to illustrate how I would think about this
k;
problem if I didn't know where it came from.

"..1 + x

tan

'In this problem we have the term

If we try'
is an "inside" function.

-1
x, so

then du=

u = tan 1x,

1 + x

tan IX

As a first step, I look for algebraic simplifications.

The numerator is a sum, so I might consider breiking the integral up into.

dx.

2
.

..

Since du does appear in the integral, we can make the substitution. The integral becomes

.dx

-x

wietan -1 x (

------7
1 + x

, u.

dx) = dreu du

eX

+.1
e

+ C

.
.

dx

= etan-lx +
,

e-x

dx.

e-x

This doesn't seem,to help; so I look for substitutions. j

might be tempted to try the substitution u = ex at first,
since all the terms in the integral are expressed in terms
of'ex. But
du = exdx, and I don't see that in the integral.
For that reason_I on't explore the substitution further now.
If necessary, I can return to it.
Finally, I might try a substitution for the denominator,
-x
u'=qex - e ).

This gives

du = (ex + e-x)4x,

which does appear in the integral. From here on the problem is

easy.

an

-1

(x2+2

1--)dx

x + x)

We halle

[(e

x +1

and see if it helps.

dx

1 + x

1+x

1 + x

= In lul + C

1x

ln

e-

egt

ex
e

-x

+ C.

The kntegral then becomes

4(12+2 dx) =

jru du

X +1
=
2

(tan 1x

u
1

)ex

+ x

)dx,

1 + x

-- du

We obtain

du =E-L--71+

(tanla

1 r

In his expression the term Itan-ix + xj is rather

"nasty".
We might consider the substitution
-1
u= tan x + x,

and, we're in luck.

ex) dx]

iex

One of the folloing two integrals is muckeasier to solve

than the other.

(a)
+ 111,

Decide which it is, and solve it.

+ C

x (1+x

4)5

dx

(b)

1/(1+x4)5 dx

+ C.

17

EXERCISES

13

EXERCISES FOR C3AP2'ER 1

As always, I start working on a problem by looking for

algebraic simplitications.' In both parts (a) and (b) of
this problem, I can multiply (1+x4) by itself five tines,
and then integrate tern by term. That, seems \t'
complicated,
however, so I look for other alternatives.
In both parts of the problem

I see (1+x

term (1+x ) iis an "insiae" function.

u = 1 + x4,

then

Detailed solutions of these exercises are available in a

separate solutions manual, The order of the 'solutions is scrambled,
to keep you from accidentally seeing the answer to the next problem
you are working on._ The solution number of the exercise you are
woAXing on is underneath the exercise number.
For.ex'ample,-

so that the\

14

If I try

du = 4x

means that solution 5 presents a discussion of exercise 1.

{1+

dx.

sol. 5

Since the term (x. dx) appears in part (a), that

be easy to solve.
It becomes
.1x3(1+x4)5 dx

1
=11+x4j5(4x3dx)
4
=

17(1+x

4 6
)

ntegral will

In each of the .fallowing exercises, one problem can be done

easily. Use the techniques of easy algebraic manipulations and
obvious substitutions to determine which it is, and solve it.,

= -1.115 du

dx
+ sin x

(a)

+ C.

1.

sol.

x + 1

x
The sample problems you've worked through in this Chapter

Isol.

dx
+ 1

dx

61 (a) firr. )(11 +

2
(b)

x + 1

dx

CO

On tests I've seen students spend ten or fifteen

minutes trying to solve

sol. 4
j(Px3 + x2 + 1

may have seemed very easy, because you were on guard for simple
solutions.

+ x

dx

(b)

J2 + sin x

* * * WRUNG * * *

(a) viln(ex) dx

sol. 3

tcos x dx
(b)

vir(I + ./i-

X- 9
Je*
by partial fraetions or by using the substitution

0.1
it
4
(a) j tan x sec x dx

x = 3 sin 0 !

(b) d sec x tan x dx

{sol:8

dx

ex-

(b)

37-= 1 dx

The moral of this alter is:

ex

(a),

sol. 6
e

When you start working 'on a problem,.alatays

check fort an easy algebraic manipulation or
obvious substitution. , Only wheii you're sure
Aft'
the problem _cannot be SIMIFIED should you
try anything else.
,

4. Pa)

sol. 1

tan

fx2

+ 1 dx.

I ,8, I
sol. 7

(b)

si tan

-1

x dx

(a)

x2

dx
4x + 3

/ x-2
x.

-4x
+33

dx

15

Chapter 2

r Section 1

16

'RA T.IONAL FUNCTIONS

`CLASSI.FYI.
As we noted in the introduction,:experts generally follow a=
three-step procedure when solving integrals. The first step, which
we discussed in Chapter 1, consists in kooking for simplifications
or easy solutions to a problem. The second- step, if necestary,
consists df choosing and applying the technique most likely to
solve a problem.

This choice-of technique is tally bated. onthe FORM of the

integrand. Ask,'an expert why he chooses to solve fx sin x dx
using integration by parts, for example, and he'll say "because
it's a product of dissimilar functions." The solution to a problem
follows routinely once the right technique has.been chosen.

A rational function , is the qhot ient of two, polynomials .

Procedure for integrating rational functions is straightforward,

although it may sometimes be long and involved. A large part of
that procedure is purely algebraic, and consists of, "breaking up"
complicated rational functions into sups of simpler ones. We 0/11
begin by examining the simple or "basic" rational functions, and then
discuss tow -to break up thelmore complicated ones.

Part 1:

second-box in the General Procedure:

Step 2:
Rational
Functions

Definition: A Basic Rational Function is a "proper fraction" of the

.

Products

Trigonometric
Functions

Special
Tune-Lions.

yotir goal 'in working through this section. should be to

appropriate 'classify integrands by form and recall the techniques
4.
to them. If you systematically use\tze simplifications of enectbr 1
and tRe'classification scheme of this section, you should be able

/
lye most of the problems at the end of xourtext's chapter
on integration.
to so

20

r
r ,I
+b - (ax + b)n

r)s, + s

Or

',. ax2'+ bx + c

O.

z.

'Basic rational functions -of the first two types are easy to
integrate. If the denominator it (ax+b) or (ax+b)n, 'she
substitution u = (ax+b) will solve ft. problem. See sample
problems 1\ and 2.

CLASSIFY

RATIONAL-FUNCTIONS

form

In this chapter we wil I classify integrals in-four basic

categories, and' discuss the techniques most often effective in
dealing with them. Our classification is summarized by,the_.

The

'Things are more 'complicated kf the denominator is quadratic:

If the denominator factors easily, we use partial fractions to

break up the integiand. .For exarife,
(x-5) dx
,x2

- f bC-51cc
(x-l) (x-3)
+3
2 In

..f( 2x-1

,..

In

dx

r!"

We will discusst the technique of Partial fractions-in part 2

of-this section.

21
.11

.4E11E11
tI

r4.

18

17

o
Suppose the denominator'does not factor easily.

Then

complete the square andMake a sbbstitution for the

INTEGRATING BASIC RATIONAL FUNCTIONS

fprm in

14'

There are two possibilities.

the denominator.

(1) If ttfe denominatords (ax+b) or 'Saic+b)n, substitute

u r4,(ax41)). This reduces the problem to standard form.

(i) If the denOkinator is of thelorm (u +a"), we will obtain

(2) If the denominator is quadratic and factors easily,

use partial fractions to. finish tbsproblem.

something of the form

b+c
A

du.

u a

u a

(3) If_the denominator is quadratic and does not factor

easily, complete the square. If the denominator is then

'

bf,ia-T du

y--y qdb

u +a4"

044.

is

The fitst integral on the right will yield.a logarithm, and the

2
2
(u +a ), integrate directly to obtain a logarithm

and/or inverse tangent.

seconggives an arctangent.

ii:

(u -a

2 ),
'

(ii)

either use partial fractions or break up the

integral and use the formula onp.17.

If the denominator is of the form (u -a ), we obtain

-4 -,

Note:

f bu+c

Make sure you have checked for SZATLIPICANOW beforeeyouprocedure for rgtiomailunctirons.

-use the

du. .

;at

PS

ere aretwo ways to continde.from here. One is to factor the

denominatikand use partial fractions to break up the expression

SAMPLE PROBLEMS
bu+c

43

THe solutians.to these problems illustrate the techniques

described above. Try to solvethem before you read my solutions.
If they cause you a_great deal of difficulty, you should probably
practice on some simrlalr problems from your textbook.

fu+a)(u-a).
If tfie factors (u+) and (u-a),Jook. reasonable, this is probably

egood way to finish te problem.,,WR'do have another alteynativd:

however':

*.14 can write the integral as

---

bf4'

u -a

The first integral is

1.

2 j7--E

,f4

dx

(4x

5x +./

3)

u -a

logarithm, and the second can be solved

3.

easily using the formula given; below. See sample problems 3

through S.

j
x

3x

4.

dx

+ 4x +'13

x; 2

fx2

+ 4x + 13

44

t
1

27---72u - s

dx

2 du*

i;

u - a

u'+
1 a

x+S

oix

22

dx

+ 4x + 2

23
fP

dx

Solutions to Sample Problems

19

Solutions, Continued

f 31777

.the denominator is of ihe form (u2 + a2), where u = (x4'2)

and du=dx,
and a = 3. Making the substitutions u=x+2
.
.
. we obtain

dx

There is no algebraic simplification possible. Since the

-denominato.r is (5x+7), we make,the substitution
u = 5x+7;

(3x+7) Aix
.4,1_2

du = S dx.

The integral then becomes

4 f 5 dx

4
lniul
S

=
sti u

)37-477

+'C

20'

jr3u du

1[3(u-2)+71du

du

_.

x +4x+13

lnISx +71 + C.

tr+7

+ 3

= 2 -

lnIu2+321 + Itan-1(3) + C

In

J: ir51

tan-1(!ir21) + C.

dx

(4x + 3)

Again, I lee-no algebraic simplification.

denominator is (4x+31,-the substitutions

Since the

tf

du = 4 dx

u = 4x+6;

;+,_2
x

dX

+ 4x + 13

are called for. The integralthen becomes

dx
4

Jr' du

(4x+3)

( -1-

(u-s

+ C

-1

4(4x+3)

+C.

As always, I begin work on this problem by looking for easy

algebraic manipulations. The integral can be broken into a
sum of two integrals, tut, this does not look especially
promising. I see no useful identities and this is already
-a "proper fraction", so I look for obvious substitutions next.
The "nasty" term is the denominator, so I' should
consider the substitution

u = x

+ 4x + 13.

This would give

3x + 7

4.41

dx

du = (2x + 4) dx,

+ 4X f 13

which is double the numerator in this problem!

easy. The integral is

As a preliminary algebraic manipulation I would considers

breaking the integral into a-sum, but that doesn't look like
will help yet. Checking for obvious substitutions, I' would
consider substituting for the denominator, u = x2+4x;3.
This givqs du = (2x+4)dx, which does not appear,in the
numerator. I can't factor the denominator, so I should
complete the square. Since
x2

= '(x

4x +
132

s'" 4x + 4) + 9 = (x + 2)

(3)i,

'11p2x+4) dx
2

x2+4,043

lfdu
2

The rest is'

lnlul+ C

2',

In lx2+4x+131

C.

Notice: This problem could haie been done by completinilthe

The
square in the denominator, like we did in problem 3.
advantage of the SIMPLIFY step is that it saved us the trouble.

25

21

Solutions, Continued

x + S

Jr
2

22

dx

Part 2:

+ 4x + 2

DROWSING RATIONAL FUNCTIONS,

A preliminary check indicatet that none of the SIMPLIFYING
Since I cannot factor
the denominator easily, I complete the square to obtain
procedures will be .of: assistance here.
c4.

+ 4x + 2

(x

+ 4x + 4) - 2 = ( x+ 2 ) 2 - ( ir)
2

2
.

In part 1,of this section we learned tPintegrate the basic

rational functions. It is a fact that any rational function can be
decomposed into a sum of basic rational functions.
The techniqes we
use are summarized in the following table.

Thus the denominator is of the form u -a , where u = x+24.

and a = /2". Making the substitutions u = (x+2) and
.
du = dxs we obtain
DECOMPOSING RATIONAL FUNCTIONS'

jr(x+Sidx

ju du

jp(u-2)+S)du

2--

x t4x+2

- 2

jr3 du

(1) If the'function is On "improper fraction", divide to obtain

the sum of a polynom)al and a proper fraction.

- 2

- 2

The first integral is easy, and yields a logarithm. For the

second integral we can use the formula on page 17 to obtain
1

4,,

125

- 2

,.(3) Use the technique pf partial fractions to decompose the

proper fraction into a sum of simpler terms.

(2) Factor the denominator as far as you can, into a-product

of linear and quadratic terms.

u +11-

and the integral becomes

,oe
3

ubl:du

--7

We have already discusssed step (1) in theSIMPLIFY chapter.

If you are trying to integrate an improper rational function, your
first step should always be to divide, and then to look-for further
simplifications.

vir

du,

u411

2/T

- lnlu +1i1)+ C

(lnlu

In 1u2-21 +

Step (2), factoring the denominator, can sometimes be

difficult if.the denominator is complicated.
The following rules
from algebra often make this task easier.

2r2lin 1 ,12_21

u -a
24-1.

1
-2- In

Rule 1:

If a polynomial with whole,numbers for coefficients

has a root which is a whole number, that root
is a divisor of the constant term of the polynomial.

Rule 2:

For any polynomial P(x),

the term (x-a) is a
factor of P(x) if and only if P(a) = 0.

4.

u +If

In (x+2)

lx2+4k+21 +

242

-al +

C.

(102)

,To see how these rules work, let's factor the polynomial
P(x) =

01.

26

x3

+x +x

+6.

By Rule 1, any number which is a root of P(x) must-be a divisor

of the constant term 6. Thus the only candidates for a whole
number tbot of F(x) are
+1, -1, +2, -2, +3, -3, t6, and -6.

:27

41041.-

/a

/24
23
A

2
there will be
If the term (ax +bx+c) appears in the denominator,
a term of the form

Now,we use Rule 2 to see if any of these are roots of P(x).,Testing

'the candidates,one at a time, we obtain

Cx+'D

P(+1) =11,

12.+. 1 + 6 .1 9, so (+1) is NOT a root of P(x).

'

in the decomposition.

ax +bx+c

P( -1) = C-1)3 fr (-1)2 + (-1) + 6 =.5,

Youwill rarely, if ever, encounter terms like (ax +bx+c)

functions here.
denominator. We will not dedl with such
To use partial fractions, follow this procedure:

so'( -1) is NOT a root of P(x).

PC:2) = 23 + 22-+ 2 + 6 = 20, so (+2) is NOT a root of P(x).

in the'

?(-2) = (-2)31* (72)2 + (-2) * 6 =

x - (-2) = (x + 2) is a 'Actor

Using Rule 2, we now have that

....

..x

x + 2

jx3
3

..

Step 2: Multiply both sides of the equation by the denominator

of the fraction you are trying to break -up. Write
both sides of the equation as polynomials in x.

We can divide to find the other factor:

of P(x) = x3 + x2 +.x + 6.
s

Step 1: Decide what terms will appear in'the decomposition,using

the guidelines given above. Write an equation, with the
coefficients still to be determined.

0.. `thus (-2) IS a root of P(x).-

-8 + 4 - 2 + 6

are the'coefficients of x on both sides of the

Step 3: C
These enable you to solve for the terms A, B,
equa i
C, etc;_inthe-decompositiol.

- x
+ x2 + x + 6
2

x +2x

-X

4,

='x2 -2x
----.7"17-+ 6
3x + 6

SAMPLE PROBLEMS

2
Thus P(x) = x3 + x. + x + 6 = (x + 2) (x2 - x + 3).
term cannot be factored further, so we stop here.

Decompose these two functions into Sums of,basic functions,

e to try the
Make
using the techniques we have just discussed.
rr work with
y
solutions.
Then
compare
problems before you read ay

The quadratic

mine.

Step (3) in the procedure calls for using the technique of

partial fractioni. -Since your textbook describes it in, detail,'
,
1111 just 'summarise it here. 2,'
......._
,
The technique of partial fraCtions is used to decompose a
.

f(x)=

'''

x3 + x2 - 6x + S
2

a g6)-

+ x - 6

- 1

Over fraction into a sum of basic rational functions. Make. sure

you have a proper fraction before you try.to use the technique:,
,

_facht4rm in the denominator of.the fraction you are

trying to, break up will give-one or lore terms when you use
.
partial fractions.
If (arftTaippears in the denominator, there will be a term
'

of the form

SOLUTIONS

x
1.

'in the decomposition.

,11--;--.....

,
.

- 6x + S

x2 + x - 6

If (ar+b)n appears in the denoidnator,'there will be terms

of the form ..
Al
. .An
212____ 2
in the decomposition.
'2' "'
'
(ax+b)n
(ax+b)
(ax+b)

+ x

f (x)=

.--

The first thing we should do is reduce the "improper fraction"

That division has a quotient of (x) and a remainder

1* division.
of (S) , so

29

25

Solution's to Sample Problems

26

Solutions, Continued

;IL

f(x) =

term in the denominator is 1, the only clndidates for roots

are x = +1 and x = -1. Since

+ x - 6

(x+1) is a factor of (x -1). We can divide

This tells us that
to find the other factor. This gives us

.(x+3)(x-2)
e'

is a root of x -1.

x = +1

- 1 = 0,

(1)

x3- 1

Since the terms in the denominator are both linear, the partial
fractions decomposition Will be of the form
Thus

g(x)

x + 3

(x+3) (x-2)

+ x + 1).

= (x - 1)(x

2
(x-1) (x +x+1),

x - 2

and our problem is to decompose

Multiplying both sides of this equation by (x+3)(x-2), we get
5

3
2

(Oix. + 5 =

(A + B) x

(x-1)(x +x+1)

or

A(x-2) + 1)(x+3),

into a sum of basic functions.. Using the criteria on pp.23-24,

we.see that the decomposition will be of the form

+ (-2A + 38).

(Remember that if a term does not appear, its coefficient is 0.)

3
2

(*)

(x-1)(A +x+1)

,Comparing coefficients, we obtain the equations

B-=

I+

-2A +,3B = 5

.IB = 1

Bx+C

x-1

x2+x+1.

JP

Multiplying through by (x-1)(x2+x+1), we obtain

A = -1

, so that

=
=

A (x2 +x +l) +,(Bx+C) (t-1)

.Ax2 + Ax + A + Bx2 - ix + Cx - C.

Thus

Thus
5

-1

1(0)x

x + 3

(x4.1)(x -2)

x - 2,

f(x) =

x - 2

g(x) =

4
x

- x
-

+ (A - B + C)x + (A - C).

= 0
A - B + C

B) x

This gives us the three equations

and

x + 3

+ (0)x + 3

A = 1
so that

- C

= 3

B = -1

C = =2

Plugging these three values back intOef'), we obtain

+ 3
3

(_,1)x +(-2)

x2 r x + 1

x-1

(x-I)(x -+x+1)

so that

This function is also an improper fraction, so we divide to

obtain
g(x) =
.

3
+

g(x)
x3 - 1

x + 2
1

x-1

+ x +

Our next step is to- factor the denominator. Since the constant

we

-27

28

Section 2
sks

This formula also has special application to the integration

.
of single terms that we can't integrate otherwise. Since if(x)dx

PR.ODIdCTS

can be written as Af(x)][1 dx], we can think of that inthrand

as a'product and try integration by parts with u=f(x) and dv=dx.

If the integrand is a product, an0 especiayy if the,inteo

,See sample problems 3 and 4.

is a productlrof dissimilar functions, you should consider Using.,

integration by parts to solve the problem. 'The formula is derived

SAMPLE PROBLEMS

',from- the. formula for the differential of a product,

+ v duo

d(uv) =.11

As usual, try these problems before you read my solutions.

Pay particular attention to the reasoning I use in making my

Integrating each term, we obtain

=fu

uv

choices of u

Rearranging this gives

=

uv

du.]

To apply this formula, we separate the integrand into two parts'. Ne

and the other

and integrate dv

dv.

to obtain

v.

,6 du, the Problem is solved.

We differentiate u to obtain du,'

(ln x)

dx

Sr

SOLUTIONS

is easier to solve

by differentiation.

comments are summarized

sin lx dx

The goat of this procedure, then,

As. the sample problems illustrate, this

is

3f

tan-ix dx

If we can then integrate the term

is to choose u and dv such that the term A du

than the original problem.

usually happens when u

2 ix'

.1. 1x cos x dx
dv

call one u

and dv in each problem.

dv +,fv.du.

fx cos x dx

These

There are two possible Choice's of u

Ow.

u = x

and

dv= cos x dx

-14 GRATING PRODUCTS

first case we obtain

in this problem;

To see which is more

dv= x dx

promising, we should determine

Consider integration by parts. The formula is

and dv

u = cos x

and

du

du = dx
,

in each.

In the

ind in the second

flu dv

uv - jry du

du = .. sin x
I

and your f,hoice of u

and

dv

should be governed by two things:

(1) You must be able to integrate the tezm yob call dv.

Clearly fv du

V =....2- X Q

is easier to solve in

the first case, so we make the substitutipns

u=x, dv=cos x dx

(2) You want

du
to be easier than the original integral.
This often happens when u i$ simplified b differentiation.

32

of Then

jr(x)(cos x dx) =

se
u

dv

(x) (sin x) u

Asin x) (10,

Z sin x +icos x

v*
+

du
C.

33

.01
Os

29

'Solutions; CohtinUel

;Note:" The functions -e ,` sin x, and_cos x- are affected - about the

same by either integration or differentiation. On the other .
hand, polynomials are usually ''complicated" by integration
and glade simpler by differentiation. This suggests the following guideline:
*t,

Solutions. Continued

(
3.

sin -1 x dx

This vitegrana cali.be considered as a product, if we

1x)(1 dx). 'Since, as in kroblem.2,
write the problem as, if.(sin
we obtain the greatestsimplification by differentiating an
inverse trigonometric function, we set

Let P(x) be any polynomial. All of the integrals

!Nal cos x dx
fiNd) sin x dx,
should be done by parts,' with u = P(x) and dv the remainder.'
/Pad ex dx,

sin-11

u=
-,

'41

1 dx

tanlx

or

(-0

dv

x sin

will have to integrate dv = tan lx dx, and that is no simple

matter. In the.second case, we Will have'
v=

and

Like priblen 3, thiscan be done by parts if we write it-as

A,

)-

ri

:dx

.----- 1+x
V.,

The second 'integral canfnow b. done by- the procedure for

ions: After us ng the procedure, we'obtain
a-1

1 Z
74;

+.6 In(1+x ) +

and

a"),"=i ax

7---dv

V.

ln
idu = -2
x

x dx)
-i (x)(-2 In,,,,__.
s"' du

x (In.., i) 2 - 211n x dx.

x di

v=x
.

(x)
j(ln 302(1 dx) = 7 (ln x) 2 -e-

f(tan-ix) (x2 dx)- = (tanlx) (- x

u =On x),,2

if (In x) )11 dx]. yilth

Here du is much simpler than u, because-we've replaced

an inverse tangent by a rational function! Nits -this choice .
,

11

we obtain-.

x3.

<

x +1/1,71t2 + C.

(ln x) 2 dx.

In the first case

we will have that .du = 2x dx, which is rather nice. But we

we,obtain.

du

tan-lx

Letts- examine 'which choice will help more.

,,du

1-x

dv = x2 dx.

m tan.4x--dx
dv4e=

v = x.

(sin-ii)(x) -1(x) ' dx

As ih problem 1, there are two reasonable choices 'for u and

:1 u = x

e.-

dx

Vi-:72

Thei

2.1x2 tan -Ix dx

2

du

so that
p

Asin-lx)(1.4

dv:

30

, f-r----4.

We havenq solved the problem, but we'fo simplified it:-we now

have to.integratfiAln.X dx) instead of f(ln x)2. cit. A second
-integration by tarts with U = ln x,' dV = 1 dx gives
is

x(ln x)

x) (x) -IGO
- 2/(1n. x)*(1 dx) = x(ln x)- -2[(ln
a...4(-dx)].
x
,.....y.

U,, di,'

'
2

U.

V te

dU
.

x)+2x + C.

Note: Like many problems in integiation' this can'be done in

x (or eW =x)
more than'one,way.. The s(ibstitution W
transforms,arinlx)
ts. adx to r 2eWdW, which is, done by parts (twice)
,
,

35,

Section 3

Twin Pairs of Trigonometric Functions

,

dx

-(cos

TRIGONOMET`RIC

x)= cos x

--(sec x)= sec x tan x

dx

x)=-sin x

--(tan x)= 4sec x

dx

.FUNCTIONS

sin x + cos x = I

dx

csc x)=-csc x cot x

dx

2
(c- ot x) = -csc x

1 + cot2x = csc2x

tan x + 1 = sec x

. If we are unable to exploit the "twin pairs", we turn to a

different approach.

The next thing we try to do is-to reduce the

+her\ are many,special techniques for integrating combinations

powers of the trigonometric functions appearing in the integrand.
of the trigonometric functions, and trying to keep track of all of
This is usually done with the help of the formulas
_them,,can be.difficult.

Instead we can keep some general guidelines

for approaching trigonometric integrals in mind.

The basic idea is

sin2k =

to exploiIthe'relationships among the trigonometric functions

theinselves, in order to.simplify the integrand.

-12.(1

-cos 2x)

cos x = --(1+cos 2x)

2

'

or by a reduction formula obtained by using integration by parts.

,....r.

The first kind of manipulation we look for is a simple

substitution of the kind u = sin x,

See sample problems 3 and 4.

u = cos x, etc.I'For this kind

of substitution to be successful, the ighrand should consist of

Finally, there is a last resort" technique based on the

u =Ian 2 . =Admittedly, this formula seems to come
substitution

an expression involving one trigonometric function, multiplied by

"out of the blue";

ttie.derivative of that function.

f
where

For example,

However:.if nothing else seems to work when you

are trying to integrate a rational function of

sin x

and

cos x,

the substitutions
d.1._Tc7:

f(sin x) =

is of the form flf(sin x)[d(sin x)],

u = tan(Z)

2
l +sin x

and

d(sin x) = cos

sin x =

2u
2

cos x
'

dx =

1+u

14.0

dx.

1-u

will transform the integrand to arational function of u.

In this problem we would make the substitution u = sin x.

.

be finished by the techniques of section 1.-" See problem S.

2 du

2
1+u

It can then
In sum,

if( an integral can be expressed as

.,/f(sec.,,x)(sec x tan x dx), we would set' u = sec .4,

INTEGRATING TRIGONOMETRIC FMNCTIONS

(1) Exploit "twin- pairs" to prepare for substitutions. 'Try

to obtain jrf(sin x)(cos x dx), etc.

Our first/object; then, is to manipulate an integral into the -

_form4f(sin x)(cosx,,dx), etc, To do this, we try toexploit the

Reduce powers of trig functions in the integrand, by

half-angle-formula or integration by OnFts.

(2)

"twin pairs",oftrigonometric functions:

sec- x and'tarvx,

and

csc x and cot x.

sin x and cos_x,

(3) As a last resort, the substitution u = tan(Z) transforms

rational functions of sin x and cos x to rational
functions of u.

The "twin pair" relation-

ships are summarized in the table on page 32:

We datcuss how-to

use then in sample. problems hand 2.

a

.36
s

41'

37

33

*0'

SAMPLE PROBLEMS

1.

JrcPs Sx dx

3.

.fsec 3 x tan x dx

2.

3
Lec

34

x tan x dx
A
on.

issin4x dx

Since this integrand involves sec x and

if we can express it as

tan x, we should see

(a) a function of sec x, multiplied by (sec x tan x), or

/5. f2

isec3x dx

Solutions, Continued

+ sin x

(b) a function of tan x, multiplied by (sec 2 x).

In this case we can achieve (b), since factoring out theLerm

(sec x tan x) leaves us with (sec 2 x tan 2 x), and the even

SOLUTIONS

power of tan x

can be expressed in terms of Secant.. We have

Aec3x tan3x dx

iosSx dx

=4r(sec2x)(tan2x)(seC x tan x dx)

=Asec2x)(sec2x - 1)(sec x tan x dx),

As a first approach to the problem, we shpUld try to exploit
sin x and cos x.

the '"twin pair" of

and the substitution

u = sec x

4u2(u221) du =/(u4 - u 2)du

to obtain either

an terms of powers of sin2x.,,,In this problem we can write

4
cos 5x = (cos x)(cos x), which gives us4
2

ticos x dx

(
x)(cos x dx) =
111114

ur[ 1

x) (cos xx)

3.

= -- sec x (b) a function of sin x, multiplied by (cos x).

2
Notice that we can achieve (b1) Since cos x can be expressed
2
in terms of sin x, then any even power of cos x can be expressed

u5 - y
1 u3
= 1yu
u +G

or

(a) a function of cos x, multiplied by (-sin

<

gives us

Thus we should ;try

sec 3x + C.

f4

sin x dx

Technique (1) doesn't help us in this

separate out (sin x dx), we're left with
which can't be expressed as a polynomial
Instead we turn to technique (2) and use

problem: if wv try to
the teim (sin's dx),
in its twin, cos x.
the double-angle formula:

,
-

2sin2x + sin4x ](cos x dx).

Ain4s
_

This is now in the form jrf(sin x)(cos x dx),

substitutions u = sin x, du = cos x dx

2u

] du' =

u -

yu 3

dx = fisin2x) 2 dx

5.45

+C

= sin x

-s

sin x +

os 2x + cos 2x) dx.

The first two terms in this expression can be.integrated easily,

and wb can again call on a double-angle formula to express
-

(1

A ,

Al-1-cos 2x)]2 dx
2

and the

give u;

+u

cos 2x =

+ cos 4x).

.This gives us

sin x + C.

4ix

44os

2x dx

1Lrjrdx

kfcos

4x dx

Note: This technique will work exactly in this manner for any

odd powers of cos x and sin.

sin 2x

38

sin 4x
32

sin 4x - 8sin 2x + 12x

32

)39

+ C

Solutions`; Continued

Sec .x dx-v'

U=

This, is' a ditficuli," problem. We?11' go through it slowly and

.1ndetail, so that the reasoning for it and prbblems like it

becomes apparent. We begin by noticing, that.tekehnique (1)

doesn't work for us here and that the double-angle formulas

.don't apply, so we decide to use integration by parts. There
are three'reasbnable choices:
-

.
.--..--.

..__, ..

(b) f(secix)(sec x dx) ,

(a) f(seex) (1:. dx) ,

(sec x tan x) - U + (islet x dx).

Solving this equation for U, we obtain

I III

2U = sec x tan x + f sec x dx

= sec x tan x + lnlsec x + tan x1 + C.
<Dividi3ng both sides of this equation by 2, and replacing U by
fiecx dx, we finally obtain

and

ilsec 3x dx

[ sec x tan x +

x + tan x1] +

lnjsec

C',

where C' =C/ 2.

(c) f (seq x)(sec 2x dx)

u

36

dv

Inchoite (a), settinrdv=dx would lead.to v=x, and the term

t (v du) would involve both x and a combination Of trig

Note: - This is a long and involved procedure. With 4nor

modifications, it will provide reduction formulas fol., powers
of all the triginkrietric functions. Because of its complexity,
however, you should only consider using it after checking othat
le-angle formulae don't
technique (1) and the

functions. Integrating that looks difficult, -so we go on to try

choice' (b), setting :1v = (sec x dx) leads
something else:
to -,v =- lnlsec i + tan x 1 , which is nasty. Thus we examine
choice (c)..Since
.
.
.
(du = sec x tan x dx!
,,.,
u
=
see':x
.
.
dv = sec2x dx ,- "'" --, `v = tan x
,

and this is the best of the three 'alternatives, we proceed :

(sec x)(sec x dx) = Asec"X)(tan

014)4

-f(tan x) (sec x tan x dx)

v

5. f2

In this problem, neither the "twin pairs" or reduction

formulas seem to help, so we make use- of the "last resort"
substitutio given in-technique (3). The substitutions
x
u =tan(); iin

du

= sec x tan--x -f(sec x) (tan2x dx).

2u

2 du.

COS X =

ItU

transform the integral to

7
1

2
Vecen use the Adentity (tan 2x =,s.ec x-1) to obtain+

-fsec3x

dx
+ Slit X

2 du
l+u

= sec lc tan x -1(sec x) (sec2x - l) dx, or

2+

1112

2u +2u+2

u +u+1 .

1+u

fS4;t3 x dx = sec x tan. x -fsec3Z-.'dx:+f'sec xdx.

$p This is a rational. funttion, and is done.bi-cerepleting the

, square in the -denominator:

For a moment it looki as if weove gone around in circles,

f.. du

because we now have the tern

u +u+1

-it =

jrSOC X dx

on both sides a equation (*).. Notice, *however, that

*veers With I negative sign onthe right-hand side of

du

(u+1/2)+(3/4)

1 tan-1
/377

Fe

We 4311. then consider (*) as-an algebra c 4eqUition,

-112+1/2
an .[/3741]

tan(y ) + (1/2)
*""

1-374b'

o-

41

37

Section 4

38

Once the triangle has been drawn and labeled, we can "read" whatever
substitution we need from it.

Follow this procedure.

To.obtain an expression for u , use the trigonometric /Unction

u and

that involves

SPECIAL FUNCTIONS

function of 0,

a.

differentiate to find du.

To obtain an expression for

function that involves the sides
In this section we will discuss,three kinds of substitutions

triangle.

which occur often enough that they are worth singling out
for special mention.

The first type of substitution deals

with terms of the form

2

(a

+u

(a

-u

and

(u

2-a 2
)

(a2412)1/2,_

The result will be a trigonometric

"'read," the functions from the trilgle.

procedure. See page 41 for the second /41d third kinds of substitutions
we discuss in this section.
0

SAMPLE PROBLEMS
Or

(u2-142)1/2..

The substitutions can be memorized, but I find it easier

to draw a triangle and derive them.
'

in the

integral, which you can solve in terms of 0. To express the answeg

substitutions for one of the terms

(a2+112)1/2,

n/2

We deal with functions like these by making a:trigonometric

( *)

lsomething)112 and

Sample problems 1 and 2 will illustrate how to late this

n/2

)-

lsomething)112, use the,trigonometrie

Make the substitutions.

in terms of x

ni2

u as a trigonometric

Once.you have

1.

dx

f) .

All of the substitutions

come from the Pythagorean theorem,

y2
x2
which says that , x2

dx

SOLUTIONS

in the triangle to the right.

9

If we place the sides

and u

on the triangle carefully, we

can make the third side of the triangle be any of the terms
in (*)
Seethe triangles below.

1.

fdx

(x +9)

Before'we try a trigonometric substitution, we should 'Check for

any SIMPLIFICATIONS. Unfortunately there are none, so we
draw a triangle.
In this case the term we wish to substitute
for is
2
1/2
(x +9)

tski

so we draw the triangle

14.

42

Solutions, Continued
.

39

Solutions, Continued

To Obtain the substitution. for x,we use the function that

sin 0 =

To substitutefor

, so

and

3 tan 0

'

In this case

involves the'sides (3) and (x).

tan 9 =

dx= cos 0 do.

and.

x = y sin 0

so
2

, we use,the trigonometric function

c'4 -9x2

dx me 3 sec20 do.

involving that term'and the constant: Here

1-x

that involves

+9)

2/2

2
1/2
= 3 sec 0.
(x +9)

or

cos 0 =

This gives us

and (3):

2"" , 1/2
sec 0 =

(x2+9)1/2, we use the function

To obtain the substitution for

rt:Tx2

44-9x2 = 2 cos 0.

so

At this point we're ready to substitute 4n the Integral. We

ear.

obtain

We are now ready to Substitute these into the problem. We get

1,

dx

dirdxa

(x +9)

3/2

1(x\+9)

1/2 3

3 sec

20

(- sin 0)2

dx

2 cos 0

(3 sec 0) .

cos 0 dO)

(-3-2

de

p
2

3 sec 0 d0

:/-27-;;Tc7,
1

9
.

f cos

fsin2
52.7.1(2)(1

+C.

sin

0 d0

We, mw return to the triangle to obtain the value of sin 0. This

j dO

1
...

cos 20) dO

f.

7.7../ cos 20 dO

sih 20

gives us the final answer

=

( 0 - sin 0 cos 0 ) + C.

(x +9)

To complete the problem, we need only read off the values of

the fiinctions of 0 from the triangle.

40-

x dx

,3x

r and

.4.--$7c2

since

sin 0:

and cos 0

Sin 0

respectively.
in 0 *

3x

To find 0, we can use the 'function

0 = sin

-1 3 x
2

In thisproblem the, term /4-9x2

Thus

is of the form 41-z7.u2,

and suggests

8 triangle'with hypotenuse 2 and let

dx

4-9x

27

[ sin -1(3x)

'2 '

" b

'1(1(4797)1
'2 J`

3x, like the one drawn to the fight.

We first determine, x

and

dx

function involving.(3x) and (2).

44

are

by using the trigonometrics

This gives us

'45

C.

41

MORE SPECIAL IviicrroNs

42

The second and third types of substitution we discuss in this

INTEGRATING SPECIAL FUNCTIONS

section are really special cases of A suggestionwe discussed in

Chapter 1, where we noted-that it is often worth considerin
(1)

If the integrand includes terms of tip form

substitutions for the "nasty" terms in integrands. Expressio s

involving

ex

and

2
2 n/2
(a -u )
,

occur often enougtito justify T fisting

1/11iTc7b

these substitutions.

2 )/1 /2

or

2
(a

n/2

2
+.11

(a) Draw a right triangle.

(b) Place a and u so that the third side of the
triangle is the term you want.
(c) "Read" the' substitutions from the triangle.

..__

We frequentlytehcounter integrals like

(u -a

'

3x

dx , 4(

+ 1

dx,

- e

a----

and

+ 1

(2) If the integrand is a rational function of SK

make the substitutions

dx,
'

and

dx =

which are rational functions of ex. At first glance it looks like

x

the substitution
t

u,= e

ddkr ex dx

(3) If the infegrand is rational function of x and

make the substitutions

will not be of assistance, because the

is missing.

You should make the substitution

in)71; = u,

If

= e

du = e

then

= u dx,

,dx

u-b), and

x- = 1.(

anyway!

dz.=

n-1

du.

so

dx = -Li- du.

SAMPLE PROBLEMS
If you are trying to integrate a rational function of ex, make the
substitutions

= u

and

dx =

3.

du.

The.etult will be a rational function of14.

flex

- e

-x

4.. f

dx

2x+1

dx,

*4,3,

-44 SOLUTIONS

A similar comment holds for integrals which include terms of

'the form
n

ax +b.

If we set

u = ;1i7n 7g,

then

Differentiating, N e obtain

and

= ax+b,
n

dx = -u

n-1

x =

n-b).
-

y174

1.

- e

-x

The inteKrand in this probl'em is a rational function `of

*Therefore we should make the substitutions

make the substitutions

= u,

x = '1(un-b),

and

dx =

un-.' dx.

The result of these substitutions will be a rational function of u.

See sample problems 3 and 4 Tor these sUbstitutions. The table
on page 42 summarizes this section.

46

dx

4
du.

If you are trying to solve an integral which is a rational function

of x and n axtb,

r.

= u,

dx = 2-du,

even though (exdx) doss-not appear in the numerator: Since

e

-x

.1

, the integral becomes

e.

Solutions% Continued
du

(Lryli. du)

(u -

, u - U

44

du

-14 (4)

-u2

- 1

Jr

A
EXERCISES

43

, tERCISES FOR CHAPTER 2

Using partial fractions or the formula on page 17, this is

The purpose

PART 1:

lu-11 -

du'_
uI-1 )

lniu+11]

of these

exercises is to give you pri.ctice in

the SIMPLIFY end CLASSIY steps of the General.Procedure.

DO NOT SOLVE 1W INTEGRALS AT THIS POINT.

1,1111u11+

I u+1

2.

1/1

lex-1

simplifications, classify them, and

2
ex+ 1

decide which technique

Then compare your reasoning with

you would use to solve'then.

mine, in the solutions manual.

sol. 6

'Examine them for

Remember:

means that solution #6 presents a discussion

of exercise 1.

ti

The integrand in this problem is a rational, function :of

and

so we should make the substitutions

32/7---c,

u/4

The inte

x =

= 2x+1 ;

-1)

dx =

du.

I1 f

xvx-+4

x tan, x dx

I s9.81

then becomes
-.

AO (3

du)

7+1dx)

)(u)(11-12

I so,. 1 12

11;ing the technique for rational functions, this becomes

3

+2

'7A3

ln I u-11

3u +

+.;:r

2+u+1
1n

/2. tan4x- dx

1,01. 3

u =

sx3

Vx -

cbc
1

dx
1

Isol.

3,./C2+6:c'dx

11

du

i ol.9.9

6x2

(if)

V2:7+1.

5.

I
O

14 1

I5

ex4-

v-3-

where

-7.

!soh

du.

u -1

dx

so1.861

1501. 13Ii 4/

48

9x
375"
+

dx

110,1
RO,. 2

dx

cos x,

The exercises are.contvmea on page

45...

sss; corr

45.

Chapter
-414441i5,

'MO.DIFY

es

Chapters 1 and 2 of this booklet contain the basic techn

ems.

necessary for solving most first-year calculus integration p

s a

Once we'can'SIMPLIFY or CLASSIFY an integrand, its solution

routine (although not necessarily easy) matter.

jar

We encounter the most difficuly with problems of unf

fOrm,ihose which.1;isit classification by the methods of M apter

2.

a * pART 2.. a

Kith such Problems our goal is to MODIFY the integrand manipu-

lating it until it is in a more convenient or recognizabl- form.

Once this has been done, we return -to the SIMPLIFY and/el
Solve each of the exercises from part 1.
in the solutions mqnual.

Detailed solutions are

This table lists the number of the solution

IFY-

stepi of the Geh4ral Procedure to'finish the problem.

"J

The three sections of this,chapter are:

to each exercise below the number of the exercise.

(1) Problem Sisilarities: looking,foz and explo ing

resemblances, between the problem we-zre-wo
ingi on
aid:probleti we know how to intfgrate

Exercise #
.Solution I

22

1,

ME

10 11 12 13 14.15 16

291:11:11331132

23 21 31 19 28

(2) Special Manipulations: techniques for exp essing

complicated integrands in more convenien form
(3) Needs,intlysis: looking to see what add tional terms

_might:help,Solvezproblem, and:Modify

the,

integrand -to include them.

LOgether,--:these form-the-third step of the G eral procedure:

MODIFY

'81:ep4:,.

Problem
Similarities
-

Special
Manipulitions
,

Needs
Analyiis'

51 f
4

.).
/

48

47

'Section

SAMPLE PROBLEMS
Use the suggestions given on page 47 to try to solve these
problems.

?Wen compare your solution with mine.

PROBLEM

SIMILARITIES

tf:tc

2.

dx

Jrx 6

1 3 - 9x + 8

Some integrals can be classified easily, but look so compli

dx

42

3.

cated that tAe standard procedures for solving them promise to be

messy..

Other integrals may notfit into the classification

/scheme
/scheme of Chapter 2, and)we may not knZw an appropriate way to
solve them.

One way to approach such problems. is to look for

similarities between them and problems we know how to do.

SOLUTIONS

If the

form of a difficulI problem resembles that of a "standard" problem,

.

there are two possibilitiets.

We might be able to reduce the

dx

difficult problem to that "standard" form.

/Thdwould

Or, the techniqi2I--!xe,___

use on the easier problem might help us solve the more

ifficult one.
approach.

The integrand in this problem is a rational function, so

Th# sample problems will illustrate this kind of

Summarized in table form, we have

7

we could solve the problem by the procedures of chapter 2.

The denominator is difficult to, factor, however, so we look

.4

for another approach.

'

The problem would be easy of the denominator were

PROBLEM SIMILARITIES

411

.41) Look for easy problems similar to the one you

ate working on.
(2) TiY to reduce the difficult problem to the form
of the easy similar prdblea.
(3) Try the techniques you would use on the similar
problem.

Yes, because of

instead of (1 + x4); can that be arranged?

the

term En the numerator.

U=X

du

(1 + x2)

Making the substitution

2x dx,

we get

fr3.7

1 f du
"22"

.1
tan u

1 + u

-1
1

= I tan

+ C.

--c

+ C

r.

x
x

50

Solutions, Continued

49

Solutions, Continued

- 9x

dx

dx

+ 8

(x+1) fx2+2x
a

This problem; like problem 1, can be solved directly by

the procedure for rational functions.' The denominator factors

As a preliminary simplification, we might consider a substitution for the "pasty" term in the denominator: u = x2+2x.

without diffrculty.to give us

"

Thisleads to

du = (2x+2)dx, and at first glance this

Sr"'

looks promiiing.

denominator, instead of the numerator, where we would like

Lir(x3-l) (Z -8),

itl

Mind rontini'w factoring Iliet_denominator,,use parties

Unfortunately,,the term (x+1) is in the

ip.we abandon this substitution temporarily, in the hope

we can find something easier.

lbat momisei to be

fractions, and then integrate term loy,tetm.

Looking foi similarities, we can ask;.;are there any

a very involved procedure, howeirer. 'We.should stop and look

"standard forms" that include square roots in the denominator?

for other alternatives.

Note that the integrand reseMbles 1 simple rational function

.4114*e.

with a quadratic denominator: instead of (x-1)(x-8),, we have

3

(x -1)(x
2

3'
-8). .Can IAIIT44fy the denominator?

term (x

-du

du

Yes, terms like

Yes, since theme

dx) appears in-the numerator. With the substitutions

3
2
u =-x ; du = 3x dx,

etc.

completing the square, in the hope that we get something

easier to handle.

We have

[x2+2x] = [(x+1)2 - 1], which

suggests the substitution u = (x +1). Then

we obtain
.

dx

4 du'

(Z.Ov-

/7171115;1

(7)

du

u- 1 + u - 8

'

21.ft u - 8

.11 1 1

fu-1
.

x-81

In

, x

Vote:

+ C.

-1

Substitutions like this might have occurred to-you after--

workingthrougfichapter 1.

If so, terrific!

Our guiding,

-printiple is: at every stage of a problem; ,look for easy

alternatives.

As, you gain experience, your

SIMPLIFYING techniques will grow.

.

(.4.1) ATTC

= ssec-1(x+1)

t2logue of

This suggests

177:777

du
=

11727

sec

-1u

+ C

Si

52

Section 2
-

A. RATIONALIZING DENOMINATORS

SPECIAL.

(A+B)(A-B) =

This technique is based on the relation

by Vii and

rf we replace A

MANIPULATIONS

(AT +

. r,
e

In this section we discussfour techniques designed,to express

by 47 ,

-B

we- obtain

)(ru" - .17 ) = u

v.

hose

1.fiC41UV

ie of. the JImm

(riti IF ), multiply both the numerator and denominator by its

"conjugate,

complicated integrands in more convenient formtfor integration.

They are
a

(15'.7. 1737).

The denMinator'orthe resulting fraction

,

--14rsimpty-(u-- v).

See sample pioblems 1 and 2.

SPECIAL MANIPULATIONS
B. SPECIAL .USE OF TRIGONOMETRIC IDENTITIES

A. Rati nalizing jenoilnatori of quotients

ti

0.Spe sal use of trigonometrictillentities

-t

C. "

The basic trigonometric,identities, like the terms discusied

For example,
n (A), can 174ritten as the difference of two squares.

moron denominators' substitutions

D. "Desperation" substitutions

(l+cos x)(1=cos'x) = 1-,cos2x = sin x;

(1sin x) (1-sin x) = 1-sir2x = cos2x;
2

(sec x +tan x)(sec x-tan x) = sec2x - tan x = 1;

-Theie.techniques oftensinvolv9 complex.manipulations.

It may

(csc x +cot x)(csc x-cot x) = csc x

cot x = 1.

not be clear that they ate Helping to solve 'a protlem until we have
done some complicated calCulations:

For that reason these tech-

The terms paired above, like

niques differ from the simplifications of Ckapter 1.

When we first

With practice you will

discover which.apprpaches to integ-rats you-can examine rapidly, and

which are time-consuming.

(t -cos x),

Only

if that fails, or if the standard techniques look'very complicated,.

do we looks for alternatives such as these.

and

called conjugates.

examine an .integral, we look for fast-and easy waysto solve it. If

that fairs, we try sb..claisffyl: and.use.standard techniques,

(1.cos x)

This knowledge should.gOverp tie order

If the integrand contains any of the terms (1 t cos x);

(1 sin x),

(sec x t tan x),

or

(csc x t cot .c),

eitherNid

the denominatorofa fraction or inside a square toot,

consider

mu:tiplying and dividing the integrand by its conjugate.

in which-fou apply them.

See sample problems 3 and.4.

5b

57"

53

54

D. "DESPERATION" SUBSTITUTIONS

C. "COMION DENOMINATOR" SUBSTITUTIONS

When an integrand involves a single term like

we sake the substitution u = x

l/n

or equivalently,

../1'

substitutions that are quick and easy to use, and we postponed

= x.

The result of this substitution is an integrand which has

(whole number) powers of u

Our guideline'in Chapter 1-was-that we should only consider

l/n

integer

instead of fivotimem/ powers of

x.

looking at any substitutions that looked complicated or unpromising.

1

If neither the SIMPLIFY nor CLASSIFY steps help us solve a problem,

we should now consider more complicated substitutions'in the hope

Some integrands involve more than one fractional power 9f

At this Atagre have little to lose.

tnat they will prove helpful.

x, like
For example, to solve
1/3
fx

1/2

-+ 4

+ X

dx

2/3

A1 +rx- )dx,
A.

u = 1 +47 or eVen

*we might try

=f-17-Tr

. See p;.Obl

To solve an integral like this, we would like to find a

substitution

u = x
.

of

To solve

1/N

such that all of the fractional powers

--

x. are replaced by integer 'powers of

u.- We choose N as follows.

1(

X +.1

)4x

Let N be the smallest common denominator of all the fractional

pavers of x which appear'in the integrand.

u =

IN,

so that

x = uN

and--

Make' the substitution

we might try

u =

dx = ; uN-1 du.

x + 1

or even

U=

Ix + 11
-x

MEMBER: Our goal'is to manipulate, the integrand until it takes,

The integrand which results from thissubstitution will be a rational
a familiar or convenient form.

function of u.

As:soon as we succeed, we return'

to the SIMPLIFY and CLASSIFY techniques of chapters 1 and 2.

In the problem above, the smallest common denominator of

%
2
and
-is
b. Thus we should sake the substitutions
-23
u = x

-4W

1/6
;

x = u

SAMPLE PROBLEMS
Try each problem before you read the ablution. Then compare your
reasoning with mine.

dx = 6uSdu.

-The integral then becomes

dx

rti

t 47:7

(2

i+27-14 )(6u4 du),

which can be solved 1:1y the.procedure for rational functions.

See sample problem S.

2. f
3.

58

1,

.dx
I -

of 1

dx
+ cos x

4.

5
6.

rx-

6;

Solutions to Sample Problems

55

54

Solutions, Continued

3.

SOLUTIONS

Wrdx

if;717

'dx

+ cos x

477

Since the integrand is a rational function of cos x, we could

use the substitution

denominatorliy the conjugate term ((i+1

is fairly easy,.we can try that first and see what happens,

- 47-71).

We get

This gives. us

(1-cos x) dx

dx

- 17-73 )

dx

j(16;71

f(i771 + trx7-1) (1371 -

jr(l+cos

x.

vfl+cos

(1371,

(x+1) - (x -1)

jrcos x dx

fdx

) .dx

2 J(x.1.1)1/2dx

(x+1)

3/2

(x-1)

1/2

dx

csc x

COL X +1 +

i--- ,.

COL X +

(x-03/2

(where u=sin x)

dx_-

sin2x

sin x

sin x

f (1-cot x )

x)(1 -cos x)

- 47-1 ) dx

2--

x+1--

to transform it to'if rational

Since working with conjugates in -this case

function of u.

To solve this problem, we multiply both numerator and

u = tan

'to

* C

sin x-

cot x + qsc x +C.

2.

.1 4. /di

x dx'

cos x' cix:.

In this problem the "nasty" term is inside the - square root.

a

Here too we multip1N2umerator and,Onominafor by the

conjugate term, (1 + V1=7).

-reason we can try the technique, in the hope that the result

This gives us.

is simpler to work with: If it isn't

41k

ix

1 + d1=7).dx

thing else.,

.jrx (1 + )7; ) dx

(1-47iii)(1.4)

jr(1

2
=

(1-x)

3/2

1/2

(1-x)

+ C.

)dx

fsin2x
l+cos xlk

dx

"
laCTn x

w e would look for some-

x)(1+cos x)(l+cos x)

dx

4.7cT7

dx

x" fi

sin x4x

417.7g7

.r
This may look as complicated:asthe integral 'lie started-with,
but is much easier and can be done by the techniques of
Chapter 1.

'

- (1-x)

f/F7:77).
p(1 + J1=7) dx

/*.

If we multiply (1-cos x) by -rts conj ugate (l+cns x), we obtain 4

_
,
2
.d'
sin x, and the square root of that i S just sin x.FOr that

We-have the term

cos x

in the denominator,

.s

3'
Solutions, Continued

58

57

and (almost) its deriviitive in the numerator.

Section 3

Making the

substitution .0 = co,1 x, the integral becomes..

J/10,11 -1/2

Jr. -du-

571

-2(1+u)

1/2

+ C

NEEDS ANALYSIS

+ C.

,
dx
J1/i

- x

1/2

The technique of needs analysis has been implicit in much of

,

This problem involves 'fractional exponents. The least

--

1
1
common-dencminator-ofand-----i-se-the

2
.

our work so far, and we, now state it formally as an integration

,

substitution

technique.

It consists'of asking what might enable us to solve a

problem, and then either adding it (and compensating for it) or

u = x1/6,

so

X L

'U 6

and

dx = 6 uS du.

changing something in the problem to it., Needs analysis explains

The integral becomes

the, reasoning behind our exploiting 'twin pairs" of trigoAometric

...

6u5 du

u du

1 - u

: u

-6f ( u2 + u + 1 +
-

_6fu 3 du

(2u3 +' 3u2 +

t;

(2x1/2 + 3x1/3

u - 1

u-i /

;functions, for example.

involving sin x, we search for a way to introduce the term

(cos

If an integrapd is a complicated expression

rf

tx).- Conversely, if (cos x dx) appeared in the integrand,

we might seek to express the rest of the integrand ih terms of'

sip x.

+ In lu-1) ) + C =

6x1(6

For an integrand involving e , we might Seek to Introduce

(exdx). (Thls\11' done automatically by the substitutions omex;

141/6_ 11 .,
C.

du=e. dx;

dx

This problem can be done by sequential substitutions

dx = -du.

An alternit&strategy is given in sample

problem 14

If'the iniegrand involves x , we can-look for a way

to introduce

[nxn-I dx'.

As usual, we summarize in table form.

u = 4.1

v = 1+u;

w = F. As an example of a "desperation" substitution,-hoWever, we might try

u =4CT7:7,7T .

Then

dx = 4u(e7 -1) du.

fr7Tir

dx

u2 = 1 +/II;

x.= (u -1)

2
;

and

NEEDS ANALYSIS.

(1) Look'for a term, or a form of the integral, that would.

enable you to solve it.

Then

f(u) (4u(u.,_1) f du

(2) Try to modify the integral to produce, the term or form

you/need.
= f(41:414 - 4u2) du =

(3) Try to introduce the term you'need, and compensate for

it.

'4

3u

+ C

4
=

(1 +,,tx )

5/2

(1 +
.

)5/2

+ C.

-.Sample

2.

'SAMPLE PROBLE'MS

60

:;olutiohs, Continue!

59

Probtems'aiii SoluilOns

ft

rsec `x dx

01,'

J /5 - sec2x

iry' to solve -each: of these ;roblems using a needa anaIysis.

Then compare your eclution with mine:

Since this integral involves a fUnction of sec x, our first

dx

.1e x

- e

dx

reaction is: we need the term (sec x tan x dx.).

x,ax +b!

10,

We can multipli

-x

numerator
and denominator-by-tan x to obtain
n9

it

sec-x dx

jv

((sec x)(see x'tan x dx)

dx

415-secx

v (tan

(sin x + 6) (cos.x)

a - sec x

Instead, we can ask: We have

but this looks verynasty.

the terii.seeh dx) in thi numerator. Cdn the rest of the

--soturteig-

integral be -expressed-in-terms -of-ran -x2--1741s-,---s-ince-7---

Using this in the denominator, we obtain

sec2x-.. tan2x + 1.
2
sec x dx

dx

eX

du

e-x
-

= tan x].

ji:r477:

tan2x

We solved this'problem before on page 42, where the

procedure for special functions called for the substitutions

= sin

-1

-1 tan x
sin
(-2) +.

(7)

and. dx = 3-du.

ex =u

Needs analysis provides another route to a solution...Since

the integiand is a rational function of ex,-Iwoad like to

make the substitution u = e

if

if x(axn+b)

'This would work most easily

One way to handle this problem might/be a "desperation"

egent in the integrand, I can

tear,

obtain it, if T multiply numera

r and denominator of the

a substitution
numerator.
.

sand now the substitution' u = ex gives

. du

-1

k(u

the

term causing diffuculty,

This gives

integrand by ex.

u = (axn+b). 'Another way /is to focus on the

substitution,

- 2 .

( ln fu-1

u+1 /

- In u+li )

,.

n-1

n1
-

./

in the denominator.

we would need

' Jr

dx

If du

nxn-I dx

(xn)(axn4)

(u)(au+b)

We can now solve the problem by partial fractions,

obtaining
=

1!/15H

+ C.

"1.1((l/b)

64

To make

nxn-2 in the

integral becomes

nx

(nxn-1) (x)(axn+b)

where u = x

We :an get it, if we muliply.numerator and

denominator by
nx-

like u = xn.

1.4/hild
au+

au+b

du

Solutions, Continued

(1n1u1 - In (au+b1)+ C

62

EXERCISES

61

iota:7;E I + C

EXERCISES POR CHAPTER 3

4i4

in l

n1

+ C.

axn+b

PART 1:

Examine each of these integrals and DECIDE how Lou

.

would solve it.

4..

Then compare your chosen approach

with mine, which.is giveri in the solutions manual..

dx

Lir(sil\x + 6) (cos x)

Since this integral involves sin x\ and

cos x

I (5

we. need

dx

Is

ol. 5

either

Isot.

x -3x-+2

dx

11177

(a) (sin x dx) in the nuaerato% with all the rest

expressed in terms of cos x, or

@) (cos x ax) iin the numerator, with all the rest

expressed in terms of sin x.
410

1 2I.,J
isol.

If we try (a), we obtain

sin Adz

i(sin2x + 6sin x)(cos x)

sol. 3

easily in terms of cos x. So we try (b):

dx .

Isol.78

dx
steac7xx+ 2

,/

2/3
xx+ 1

Isol.4 9I

dx

j(x+4)17+73ic

'9

dx

I8
isol. .1
//1":67

dx

That doesn't help, because We can't express the 'denominator

soh

dx

21 .fx.,2(1.x.,3)

cos x dx

cos x dx

(sin x + 6)(cos x)

3.

(sin x + 6)(cos2x)

(sin x + 6)(1 -sin2x).

1

Here the numerator il (coi x dx) anil the denominator is a

function-of sin,x. Now the substitution

.

._

gives us

PART 2:

du
(u +6)

u = sin x

du
06u+6111+u)(1-10

(1-u2)

Solve each of the isereisestgiven above.

solutions are in the solutioqs manual.

Detailed
The solution

numbers are given below.

1/10
1+u

(
1

-1
35

-1

u+6

lnlu+64+

10

1/141
1-u /

InIl+u1 -

,1

20

13

l9

14

17

18

1112

10

lni l-ut + C

In 'sin x + 61 + .11T 1n11+sin xl -

66

Exercise t

dx

lsol. gi /sec

sol. 4

Solution #

15

11111-4in xl + C.

Whew!
Y.

67

16

A2
.

Appendix I.

Pre-Test

a
,

Appendii

Answers to Pre-Test

...-

You should be able to do arc or ttOst--ppblems without difficulty.

If you have a lot of trouble; practice these-types of problems
before you-try to work through the booklet. Answers are on -the_
opposite page.
k

"

gX2+3x1/2+1)

(2)

..4

(9)

x2

6x(x241)2

x2 + 6

(4)

,,

''

dx,

'

`-

cm -:,-,.,*.

.
,_-.

iicsc 22x dx

(42)

(5).

i-

(tan'- 4x). ,"

_ --

(12)

-1

cot 2x + C

(6)

,,'

.
^
(14) i sec
x tan, x 'clx,

2 cos 2x
sin 2x

(7)

C
4

.1

'

(13). r sin 5x +,C

"cos Sx ch:

(13)
..,

6 .sec 26r-

,.

''
(6), ---(log(sin '2x))
.

(7)

.1. (x2+x)8.+ C
.

c-

,:c1

(1)

14x ei,, x +1

,(4)

(...

c
2,

(11) ./(2x+1)(x40t)

..-

).
.

-JAN +6i +, L

(10)

2 1/Z---1 x

f 2x ch

2
7x

'

(3)

d +2 7

clx

+C

(9)

(10)

+ C.

---b 3

clx

x + C;

"i xs

NOT 1(x 2 +1) 3

(2)

d-'.
(x24 l)3

1 xs f

r(x 2 *1)2 ch

(8)

(8)

:.,

lox . :.1 x'"

(1)

"4

(14)

iec x + C

1 + 16x '
.

Pk:34

1.

a
ei.41

.cb

13-

A3

This table contains the formulas whiCh are

USEFUL for integration. For short-term use
(on tests, for exarple) memorizing them will
save you time and trouble. For long-term or
ional use, you can look them up or
o
hem when you need them.
de

'

This table contains the formulas which are

ESSENTIAL for integration. You should know
them so well that you never have to refer
to the table when solving probloms.

lir11114
-.

ESSENTIAL FORMULAS'

Trigonometry
2

USEFUL FORMULAS

.
Trigonometry

cos x---sin x

(a) sin x + cos x = 1

(c) cos 2x=

(b) sin 2x = 2 sin.x cos

1 - 2

iintx

(4) /sin u du

un+1
=

n1

du
u'

In u

-cos u

Air du

e
.

(S) loos u du = sin u

(6)

(2)1

(g)

1 + cot x = csc x

cos2x =

cos 2x)

T4egrat ion

_ .

Ludu

(e)

- cos 2x)

(f) sin2x =

(d) tan2x + 1 = sec2x

2 cos2x - I

Integration

(1)

Pec2thu

- sin

-1

ar L72;

(--u )

+ C

(13) Stan u du = -In 'cos ul

(14) /cot u.du =

ln 'sin ul

(15)1sec u du =

lnfsecu +tan 14.1-

--J

= tan u + C

(11) f

(7) jrric2u du = -cot u + C.

du
a2 911 2

Ale

(3)

feu du = eu + C

(8) fsec u tan u du = sec u

+C

jr .du

(12)

(9) Ascii cot u du = -csc u

au2_at

sec

-1u
(3)

(16) `csc u dti = -In lcsc u

cot u:+C

A-Detailed Outline of the General Procedure

.

SIMPLIFY!'
.

ghTIAMMLCMffiniagtA

asvIcus suBsTITurtoNs

(1) Break integrals into Sums.

(1) Substitute for the "Ins de

Terme" in complex expressions.

(2) 'Exploit Identities.

(2) Try to substitute for "Nasty"

Terms or Denominatore (brief
tries'only).

(3) Reduce rational functions to

Proper Fractions by division.
.

..

'CLASSIFY!

411-

,INTEGRATING TitIGONIETRIC FUNCTIONS,

INTEGRATING RATIONAL FUNCTIONS

-,

..........,....

(1) Reduce to 'proper fractions" by division.

(2) Factor the denominatA r.

(1) Exploit'twin pairs to prepare for substitutions.

Try to obtain integrals of the form
4
.

'
i .

(3) Decompose by partial fractions into a sum of

"basic" rational functions.

(2) Use half-angle formulas ors

by. parts
to reduce powers of trigonometric' functions
100fi the integrand.

If the denominator is (ax+b) or (ax+b) n ,

use the substitution u = (ax+b).

(4)
.

(3) As a last resort, the substitution u = tan(2)

.transforms rational functions of sin x and cos x
to rational' functions of u. (see p.32)
e

If a quadratic denominator does not factok

easily, complete the scf9are. For the terms

(5)

(a2+u2)integratedirectly to obtain a

i:
+

.
.

lggariths.and/or,arctangent.

'

IaEGRATING SPECIAL FUNCTIONS

if(sin x)(covx dx); etc.

ii; _(u -a ), break intoa sum and use the formulaon p.17, or use Partial fractions.

(1)

If As integrand includes tDrme.of the form

'.

(a2_u2)n/2-, (u2../12)n/2, or (a2+u2)n/2,'

IkTEERATING PRIOCTS

(a) Draw a right triangle

(b) Place a and u so that the third side of
the triangle is the term 'you want.
(c) "Read" the substitutions from the triangle.
,,,

Consider integration by.parte. The formula is

, -,,
.

fu du = uv

'

,and your choice of. u

by two things:

- ,(i) du

and dv

(2) If the integrand ie a rational function of e

make the substitutions

should be governed

%: ,

(1) You must be able to inteirate the term- dv.

= u

and

dx =

L
u

du.

(3) 'If the 4ntegrand is a rational function

(2) You want dry du to be easier than the original

integral. This often happens when u is

of

and t/Fi.4,. make the substitutions

;;7"; 21

, simplified by differen ;iation.

x = --(un-b), and dx = Ru n-1 du.

u;

MODIFY!
.

PRCGLEM SIMILARITIES
(1) LoOk for easy problems similar
to the one you are working on.

(1) Rationalizing denominators

of quotients. I
.

(2)'- ,Cry to reduce the difficult

(2) Special uses of

trigonometric identities.

'

(1)'Look for a term, or a form

of the integral, that would
enable you to solve it.

problem to the form of the easy

similar pioblem.

(3) "Coimon denominator"

substitutions.

#3) Try the techniques you would

use on'the similar problem.

(4) "Desperation" substitutions

NEEDS ANALYSIS

SPECIAL: MAMIPUEATIONe

..,.-A-.

+!,

72

(2) Try to modify the integral

to produce theprm or form

you need.
'

(3) Try to introduce the term

you need; compensate for it.

Return to:
EDC/UMAP
55 Chapel St.
Newton, MA 02160

STUDENT FORM 1

Request for Eelp

Student:4 If you have trouble with a specific part of, this

unit, please fill
out this form and take it to your instructor for assistance. The inEbmation
you give will help the author to.revise the unit.
Your Name.

Unit No.

Page
Model Exam

'Section

() Upper,

OR

()Middle

Problem No.

OR

Paragraph

Text
Problem No.

() Lower
Description of Difficulty:

(Pledse be specific)`

2-

Instructor:0 Please indicate your resolution of the difficulty in this box.

COrrected errors in materials.

List corrections here:

Gave student better explanation, example, or procedure than in unit:Give brief .outline of your addition here:
.

t.

(:2)Assisted student in acquiring general learning and problem-solving

skills (not using examples from this unit.),

Instructor's Signature
Please use reverse if necessary.

,Return to:

STUDENT FORM 2

EDC/UMAP
55 Chapel St.
Newton, MA 02160

Unit Questionnaire
Name

Unit No.

Institution

Date

'Course No.

Check the choice for each question that comes closest to yourpersonal opinion.
1'.

.How useful was the amount of detail in the unit?'

Not enough detail to understand the unit

Unit would have been clearer with more detail

Appropriate amount Metall

Unit was occasionally too detailed, but this was not distracting
. Too much detain; I was often distracted
.2.

How helpful were the problem answers?

Sample sointions were too brief; I could not do the intermediate steps
Sufficient information was given to solve the problems
Sample solutions were too detailed; I didn't need them

3.

Except for fulfilling the prerequisites, how much did you use other sources (for
example, instructor, friends, or other books) in order to understand the unit?
A Lot

4.
o ._

A Little

Not at all

How long was this unit in comparisdn to the amount of time you generally spend
a lesson (lecture and homework assignment) in a typical math, or science course?

Much
Longer
5.

Somewhat

Somewhat
Longer

About
the Same

Somewhat
Shorter

on

MuchShorter

Were any of the following parts of the unit confusing or distracting? (Check
as many as apply.)
Prerequisites
Statement ofakills and concepts (objectives)
Paragraph headings
Examples
Special Assistance Supplement (if present)
Other, please explain

ti

6.' Were any Of the following parts of the unit particularly helpful? (Check as many
as apply.)
Prerequisites
Statement of skills and concepts (objectives)
Eicamples

Problems
Paragraph headings
Fable of Contents
Special Assistance Supplement (if present)
Other, please explain

Please describe anything in the unit that you did not particularly like.
1,"

Please describe anything that you found particularly helpful. (Please use the back of
this sheet if you need more space.)

74

This rablp contains the formulas which are

USEFUL for integration. For short-term'use

This table contains. the fpxmulas which are

ESSENTIAL for integration. You should knout
them, se well that/you never have to refer
to the table where solving problems.

(on tests, for 'example) memorizing ,them will

save you time:.and trouble.
For long-term or
occasional use, yolk can look them up or
derive them when you need them.

.
.

A5

4).

USEFUL FORMULAS

SSENTIAI; 'FORMULAS

Trigonometry

Trigonometry
2

cos

(a) sin x + cos x = 1

x .-

sin2x

(c) cos 2x =1 '2 cos2x ,(b) 'sin 2x = 2 sin x cos x

f 1--

(d) tan2x

sec2 x

(f) sin2x = 1/2(1

cos 2x)

1 + cot x = csc2 x

(g) cos2x = 1/211

cos 2x)

1
2

(e)

2 sin2x

Integration

Integration

0.

(4) fsin u du = -cos u + C

n+1
(1) fundu = "

+ C

(10).

(5) fcos u du = sin u + C

du

Jae -u2

.0

(6)fsec2u ciu.= tan u + C

071
`

"

fdu
u

(14)jcot u du = lnIsin ul- + C

1)

111,11.+

(13)ftan u,du = -1nIcos ul + C

- sin-!(11) + C

*du

(7) .jcsc2u du = -cot u + C

tan-1(,, 1.1)

4---6-

.Aa+u2

(3) feu du = eu + C
o

du

(12)

u42-a2

u du =.1n I sec u+ tan t4+ C

(16)fcsc u du = -ln

(8) isec a tan u du = sec u'+ C

(9) fasc ucot u du ,= -ctc u + C

(15) f.sec

= 1 sec-L(11)

I csc u + cot ul *4- C

'75

I.

76

Alan H. Schoenfeld

INTEGRATION:

Getting It All Together

Solutions Manual

.10
June 1977

edc/umap / 55 chapel stinewton, mass. 0216Q

7'7

Solutions, Chapter 1

(Exercise 4)

There are no

Part (a) can be solved easily.

easy

3,(Exercise 5) Part (a) can be solved easily.

aigMbraic manipulations in either part of the

problem, so we look for. substitutions.

the "qasty" term is

In both (a) and (b),

In this problem, a simple algebraic manipulation is all we

need.
_

In part (a),

If we try

tan -lx.

ifIn (ex) dx
u = tan lx,

ln(eX) = x, so

then

du =

jrx dx

x2 + C

dx,

x +1

and this term does appear in (a) . Using this substitution in

N

part (a), we obtain.

21 dx)
x +1

u((tan -lx)(

du

4, (Exercise 6) Part (a) can be solved easily.

Since both parts of this exercise involve a 5th power of a

1

rather nasty term, algebraic simplifications are out of the

(tan 1x)2 + C.

question. Voth because it is "nasty" and an

u= (1 + o5F ),

theh

du

nside function,

the term .(1 + f ) commands our attention.

we try
dx

211
The term
(EXercise 2j

appears in the denominator in part (a) of the

Part (b) can be solved easily.

exercise, so part (a) looks promising. We obtain

We begin by looking fdrhalgebikic simplifications.

Both
fdx

integrals (a) and (b) can be broken into sums, but that doesn't
look terribly promising at this point..

There are no identities

) (1 +. ,fie )5

(1

(1 +

)5.

(1 +MS 21i

ITC

.which apply to either problem. But we notice that part (b).

is an "improper fraction", so we should divide'to reduce it to

2,4

du

-4

+ C

a proper fraction. _Thd quotient is (x2) and the remainder is

(1).

p3

Now (b) is easy to finish:

2
+ X + 1
d
x + 1

-1

+ C

2 (1 + 137 )
.3(

fiX2

)dx
ar

3x3 + In lx+11 .+ C.

24 1977
Orr, ""' P.41

d1).

Solutions, Chapter 1

We might try exploring with trig identities in the hope of

simplifying either part of this exercise. If we do, a short
amount of exploration convinces us that this approach is
unpromising.

In both parts of this problem the "nasty" term

Part (b) can be solved easily.

7 (Exercise 8)

Part (b) can be solved easily.

5,(Exercise 1)

Chapter 1

Sdlutions

As in Sample Problem 6, the moral here is: look before you leap!
We can factor the denominator into (x-1)(x-3), vhith means that
both parts of Exercise 8 can be solved by the technique of
Partial Fractions.

In both la) and (b), hover,' -the "nasty"

2

term is the denominator, (x -4x+3).

is the denomintor, 0(2+sin x).. If we try

If we try

u = (2+sinx),

then

u = x -4x+3,

du = cos x dx.

then

du = (2x -4)dx,

which is twice the numerator of part (b). Part (b) can then

Since (du) is the numerator in part (b), (b) is the easier

be

solved almost immediately:

Ap

problem to solve. We get

1f12142lx

Jr (x -2)dx

dx
Lfirsi3lic

In luf +

in. 12+sin x I + C

"li

x -4x+3

x -4x+3

+ C

.=

lnix -4x+31 + C

8Exercise 3) Part (b) can be solved easily.

,

6, (Exercise 7)

We might try to exploit the relationship

Part (b) can be solved easily.

tan x + 1

sec x

in either part of this exercise, but' manipulations with this

In both parts of this example we have that the integrand

is a rational function of ex.

While we might be tempted to

.qot.

jump into the substitution

u = ex, let's follow the procedure.

may get complicated. We should hold d'(' using this until we

have checked for anything easier.

If we examine

(b), we see that this almost finiPhes the problem.

We obtain

feSx

e+

dx

u=

=1( e x

4x
=

dx

( e4x

e-x )

80

0.

x.

Unfortunately, we don't have sec x in part (a), or any easy

'way of getting it. So.we go on to part (b). There the "inside"
sec x,

suggesting

dx
du = sec x tan

u = sec x;

dx.

At first this doOtn't look helpful either, untifIke realize

- e

-x

+ C.

that we can "borrow" a C,sec x) from,(sec4x),= (sec300(sec x).

Then

du = sec

tan x;

function is

Sx

tan x

as the "inside" function, which suggests

The first of our algebraic manipulations calls for breaking

an integral of a sum into a sum of integrals.

In part (a) we have

I
4
isec x tan x.dx
= f u

du

Asec3x][sec x tan x dri'

sec x + C

'

,Solutions, Chapter 2

PART I.

Solutions, Chapter'2,'PART I

dx

I(Exercise

.:.

tan-lxdx

(Exercise .134'

1
1,

The methods of Chhpter 1 apply here. The "nasty term

The SINKIFting techniques of Chapter 1 don't seem to.

Our clue to approachingt1

help herp.

we studied in section 4,- r11..ax+b, and stalest

2

u `;=-- 3x+1

.(u -l) ;,dx

and if we set

= tan

-1

dx.

x, hen du-

From this

x2+1

substittitions
2

x.=

x;

point or the problem is easy.

ecial- functions

in the niimerator. This is one

,/33-c.i-

tan

lem is the term

u du .
7 dx

,
.5

(fiercise

".

1c
2+osrkx'
.5

9. dx

10)
,--

.(Exercise,l)

There are no apparent simplificatibps -for this problem.

Our clue for approaching, it is the

The methOds' of 'Chapter 1 dori't seem, to apply. iThe integran

li
I
e
.
is' a combination of trig functions, so
, we check for the
..,

There doesn't

appropriate. technique there.

see)k

x /372,

to he,any way

to' exploit "twin pairs", and there sr:nopowers to redice,

. .
so we are left with the "last fesort'l substitutions bas. on

term /;c:7: , which is one of the

'special f

*Le studied in section'.

'4. It suggests trig substitutions,

based on 'the triangle tb the right.

I.

.
(Exercise

/.(x3.x2)

liV

x2+x-2

.
5

cos2x) dx

(Exercise 15)

,
.
'The integrand in this exercise is a rational function, ,so

.tle should follow,the procedure for rational function' s.

,*

We could use he techniques of the Trigonometric.

..-- Functions sectioh, btit we should' check

first.

for

easy alternatives

If we remembpr the trigonometric identit)5

cos 2x = cbs

2x

.5

.`

the-problem can be' done easily by the methods of Chapter 1.

There are no apparent simplifications.: Here the integrand

is a product of di,ssimilar.functioni, so integrat,ion by parts

is a 'likely
..

IV

4 dx

(Exerci

3)

.se

..*ex.l.

.,

..

..

There are no

.
'apparent

simplifications,
7"

and

x
the' term e .

in th% denominator indi,car that we should conside

xi,
X
du
Substitittions - u = e
dp = e dx; dx 8
.

\.

82
,

ix tan-ix dx'

(Exercise 6)

s'in2x ,

I.

'

technique.

(ar:

u = x;. dv = tan

(b):

U,..! tan

-10

The two choices we have are..

x dx.

'

dv =5 x dx.

Choice la) doesn't look promisfpg; because we w

to integrate
u 2. tan

-1

clv*= tan lx Aix.'

x to;obtain. chi

d have

In choice .(b), we dif etentiate

dx
1---, which' is much simpler. So

we'use integration b,' parts:

41

with u = tan lx1

'

chi' = x dx.

83

Solutions. Chapter ")

Solutions, Chapter. 2, PART I

PART I

f cos(ln x). dx

914

1x4dx

.(Exercise 16)

u = In m.

dr2

tan

4x

12x +4

While the denominator suggests a trig substitution, we

should be careful and check for simplifications first. Since.
the integrand contains an "inside" function, (x2+4), we can
try u = x2 +4 and see what happens: ,du = 2x dx, and we're
in luck-. The.problem can be done by the means of Chapter I.

-.1

cj.

(Exercise 8),

ti

14(Exercise

' -4

15(Exercise 14)

S dx,
With u = (x+3), this
i(x+3)2 - 9, .1

is

9 dii

1777'

and a trig substitutir

.*.

84iiiii,

'.

'

5-

Asc2x cot3x dx

This problem can be done directly by the means of Chapter

2x
J. If we sett' u = cot x, then du = c sc dx, and the
3
-,/
ihtegrai becomes
- u d,.
. ...,

: `is suggsted; With tte help of the- diagram given above.

4.

we try u = x4 -1, then du=4x 3 dx. Since our numerator is

(Sx3 dx), we can finish the problem easily with the. techniques
of Chapter 1.

.
.
but this doesn'r'seem.te help muth. What. can we integrate?
Termt of the form ji du ;

squares tha denominator to obtain

x -1

The integrand is a rational function, bqt we shouldn't

rush into the techniques f Chapter 2 until we've checkerfol%
simplifications. The "nasty" term is the denominator,andi.f

....

se
.12i7
so, wb should.,consider completing the

dx

'

S dx

.,/(7)(7f6)

"/Sx3
.

As a prelimindry simplification-, we Might factor the

term in the denominator to obtain
''''':

9 x dx

1a(ixercise 5)

dx

There are no apparent simplifications. Since the proplem

involves trigonometric
-sunctiorfs, we should first try to
.
exploit the relationship between tan x and its "twin", sec,x.
If that fails, we fight look for a reduction :formula. ,

.1

Sirce the integrand here is a rational unction, we should

follow the procedure given in section 1 for integratingo
rational functions.

Since this integrand contains an "inside" function,

(In x), our first approach should be to try the substitution

(Exercise 2)

..'

x3-1
.

1.64Exeilci se" 14.)

i*

A*

..,/csCII, cot3x dx6'

tills problem can't be done iramediateliy by the moans of

Chapter 1. We have to eip loit "twin pairs ",'as in;gample
Problem 10.

'Solutioni, Chapter 2, PART II

(Exercise- 4) -'

Solutions, Chapter 2, PART II

dx

(Exercise 15)

,./i7---01

j(sin2x

- cos2x) dx

.
(ee solution [3] for our reasoning.

See so,lution jlt=fOr-our reasoning. With the substitutions

u =13x471- ; ;2 = 3x+1 ; x___= .1-(u 2 --1) ; dx = 23- u du,
--0the integral becomes
'

Asin2'x

cos x) dx

We have
1

licos 2x dx

r.
$

10

sin Zx + C.

0,

- je -(u -I) 2 .d

f 6 (3(u2-1)]2[3 u du]

j* 4 dx

(Exercise 3)

ex-1
11(u4-2u2+1)

= 4-

chi

- [-5. u

44
9

u 3 '--+ u]

-3

See solution [4] ,for our.reasoning. With the substitutions

u = ex; du = e x

(3x+1) 5/2 - 2 (3x+1) 3/2 + (3x+1) 1/2 + C

this is

4/(1-1-1.7 -

2 t cos x

See solution [2] for our reasoning. With the substitutions

2

2u
u = tp(1) ; . sin x = ---2; cos x =, 1-u
--2- ;
1+u

1+u

u2 +3

Now if we remember formula 11fr\om the table of useful

integrals, this is

V,

.,

u. +C
14

lnlex

ex

tan-ix dx'

See solution [5] for our rasoning. With the substitutions

a
dx
u= tan- ; ;
11

x +1,

rJ

) +C

---fr--,,,
4

u1

.the integral becomes

".

ir- tan -1( ian(1))+

1n1-1 + c,

(Exercise 13)

ltu

. (18 [ '1. tn -1 (

=4

2 du

du

Ill

dx =

4 (lnfu-11, - 1111110+ C

1-u 2 1

1+U

...

ithe integral becomes

f91+dj)

--, Using partial fractions,

41:

f 9 dx

..

'

Vetomes

th'e

Exercise 10)

du

dx =

'.

C.

-lx/ dx
x ti

du

-1

- [tan-lxj2
1

+ C.

RECEVP.-7D

lifr
.

siAr.; .04 1977

Pjel./..A

87

C:111, c.

So u ions Cha ter

PART

r2

See solution [8] for our reasoning. Using integration by

parts with
dv = x dx
u = tan -lx
4

.See solution [6] for our reasoning. Based on the triangle

to the right, we obtain khe
substitutions x = 2 tan 0;
dx = 2 sec 20 ; and
27; = 2 sec 0. the integral
becomes

du =

(tan ..1x)(x dx)

Usirig.fora

L.

7f

cr.

./ ics c 0 dO.

2x2 van lx:-

T .1n

lf 2

dx

=.

x2

tan 1x -

el

C.

2 5.(Exercrs-v4)

=-

4i)

1:2
x

'

dx

tan

tan -lx -

Ylco; u du

"tf2 tarrix

2 6.0-xePcise 2)

J(x

inix+21 _+

(a-_.Z
+2

21441.) dx-

lnlx-1] + C.

= sin

= sin(ln x) + C.

dx

See solution [10] for our reasoning. Exploiting the relationships between tan x and its "twin," sec x, we'obtain

2 itan4Z dx

2J(tan2x) (sic2x - 1) -dx

2,/ (tan,2Z)(4A2x)(ix - 2

tan2x dx

2,/(tan2x)(sec2x dxr- 2f sec2x dx

2
c

C.

21 ((tan2x) (sec2x dx) - .2fisec2x -1) dx

;cmosc

Al(

z.,2

See solution [9] for our reasoning. With the substitutions

du = -1 dx , the integral becomes
= In x ;

u);( (

(x+2),7x

71

jcos(ln x) dx

x-2.
FollOwilig the procedurefor rational
functiOns, we 'obtain
en
x

du

X2+

Jcos(ln x)[(1) dx]

P2221-1c2
cl3c-

dx

2-

+ C.

Going back to the triangle to "translate" csc 0 and cot 0

in terns of xt'we obtain

(x+529) dx

(1'6) , we obtain

12)

x +1

-=

2 a(Exercise
..

(tan -1 x)(-21 x 2 ) -

1111t4 V sin 0

, we obtain

tax-7. G

d0

-7 in icsc Q + cot

71 x2

Z.

9.

-711/cos '0] (10'

2 [sin Ii/cos 0]

dx

x +1

7 /sec 0 al.

f) sec20 de
(2 tan 0) (2 'sec .0)

,6, tail-4X dx

44111

117:i

12

.0

_.(ExerCise 6)

:4

Chapter 2, PART II

Solutions,

11

141

r 7 dx

Exercise 1)

e-

tan 3

2 tan x 4%2 x

_+

f dx
.

89

fx dx

for os,r

reasoning. After completing the square

the denotinator and making the substitution

u =

1 f

d;c

3,J

x-1

lf(u-3/2)du

1 f

= jx dx

x+3,

if du

u2 +31 4

.
1

-32

+3
1

lnlx-11

(i 2 31
lniu 4 --1

From the triangle

1

= T x

to the -right, we obtain the

u2+3/4

+ 3/4

u du

the above integral becomes

S du.

14

Solutions, Chapter_2..PART II

fi;,...kx

See solution [111

13

5 dx

f(exercise 8)

in

PART

Chaster 2

Solutions

+ -E ln Ix-1( --6 lnix2+x+11+

--

2 v3/4
r-

tan

2x+1
(,)
+ C.

-1

..,i. tan

,/3

#16

substitutions

f
----z.

iks

u = 3 sec 0; du = 3

sec 0 tan 0

1../7-37= 3 tan.4

do;

tiis transforms the integral to

/3

see.0 tin-0-d0
3 tan 0

5 In Isec 0 + tan

=5

In

f 9x

Exercise 5)

See solution [13] foi our reasoning. With the substitution

01+ C

'

36x

I x.+3

f3

111.

1+

u =

x2+4, we have

if 9x dx,
I

C.

1 x4dX

3jExercise 16).
,

2x dx

7-7

I/2

7-7

x-1

.,

9f

riT

9)1747

j5x43

3a(Exercise 7)

dx

x -1

dx

px:
(Ix":5f 4x3dx
x4
-1

SY4): f(13)X + (1/3)

du

1`ro

- -4-

rdx
3 J x- 1

integral, to obtain

+ C.

(x-1) dx
3

2
(xs.)
+
2
s.

We now make the substitution

in lui + C

dx

x +x+1
rx dx

With the sUbstitution

-1); the integral is

u :1 (

(x-1)(x2 +x+1))

-1/2

C.

See solution [14] for our reasoning.

fx'.

viri(x

+ C

9 f du

./ x -I .

Following the procedure forrational functions,we obtain

4'
X. ax

dx

),P7.4

10

5.fieC 0

..

-1( u+ C

(xs-2

in the third

91"

10

Solutions, Chapter 2, PART II

31 o(Exercise 14)

Solutions, Chapter 3, PART I

15

(Exercise 8)

fcsc2x cot3x dx.

It)

dx

77+ 7-Te

See solution [15] for'our reasoning.

2

With the substitutions

''

Like Sample Problem (9), this4problem can Ve approlched a

u = cot x;- do = - csc x dx, the above becomes

number of ways. With such a nasty expression, we might be

3

'

- ,d('( cot x) (-

csc x dx)

-fu 3 -du

32

-1

tempted to make a "desperation" substitution with

u

4
cot 4x + C

=174Tx

2. (Exercise 9)

(Exercise 11)

orcsc 3x cqt

3x

dt

(1 + x

1/3

) dx

This is not a "common denominator" substitution problem.. If

-41
drcsc3x cot3x dx

1/2

we multiply the two terms in the integrand, the problem can

21(csc2m)(cot2x)(-c.s6 ettx dx)

-be handled basily by the methods of Chapter 1.

.

At this poinethb integrand has been expressed in terms of

csc x and.its d!rivative.

-.AcsC3x)(csC2x - 1)(- csc x cot x dx).

With the substitutions

u = csc x,'

du = - csc x cot x dx, we obtain

:341.P
(Exercise

3)

ci x2/3 dx
x+1

This is a "common denominator" substitution problem, where

2/3
the terms in the integrand-are---x------and

2
(u-2 -) ( ti-1,) (du}

denominator is

--=

-1

f'S''

-- csc .x +

u = x

3*

.---The-common

'3 , so we should make the substitution

= 1/(.1z4+u2)du"

-1

1/3
.

+ C

(Exercise 5)
-

s.-

dx

o17 o 1477
.-

In this problem we should rationalize the denominator, and

then see whatever else is called for.

5. (Exercise 1)

x dx
x4 -3x +2

The form of$this problem is similar to national functions

92'

with quadratic denominators.

denominator bo setting
\;\

We could obtain a quadratic

u = X2

46.

Solutions, Chapter 3, PART I

dx

exercise 10)

Lec

tan2x + 1 = sec2x,

%/7
or

See solution [1],for our rbasoning. With the substitution

sec x - tan x = 1.

u.="7+77T, we have

We can multiply the denominator by its conjugate, sec x - tan x.

j4u(u2 -1 )du

4
a

4.)r(u 2-1) d u_. =

xS dx

(Exercise 6)

= 1 +Or;

u2

x =

u2 -1 =,15F

(u2-1

dx = 4u(u2-1)du. Thenthe integral becOMes

and

7.

18

dx

LExercise 8)

x + tan x

The easiest way to handle this problem is to recall the

identity:

PART II

Solutions, Chapter 3,

17

...-

4
3/2
= - [1 + ./71r]
3

There might be any of a number of approaches to thif problem.

5

- 4 u + C.

1/2

- 4 [1

+ g.

The key.bbserVation to'make is that the numerator, x dx, :can

1

be written as - (x )(3x dx). This makes the substitution

= x

look promising as a beginning; we can go anfrom there.

Op-

2(Exercise

1/2

dx

(:),(Exercise 7)

fx1/2(-+ x1/3)

9)

5/6

)dx

a x3/2

/(x l /2

11/6

+ C.

(x+4)/;7:87
As-in Sample Problem 3, we need a way to get started on this
problem.

Perhaps completing the square in the denominator

13

will give us a lead.

fx2/3 dx

.(Exercise 3)

x+1

,See so1 gon [3] for our reasoning. With the substitution

9,(Exercise 4)

dx

u = x

There doesn't seen to be,any easy way to approach this

1/r we

have

= x

and (3u ,du)v= dx. The integr 1

becomes

1(x1/3)3

problem. It might be worth trying a desperationpubstitution,

dx

_11.612)(3u2 du),,3 u4 du

x+1
u

1.1W

u +1

40

3u -

Su

) du

3u

3u

(u+12(u -u+1))

x dx
sec x + 2

Since the integrand contains an expression involving sec x

in 'the denominator, we

du
2

ft=

2)

If we now follow the procedure for rational functions, we

obtain

/0,(Exercise

u.+1

=.1(3u

(1771

an ask': -what do we need to integrate-

u+1
2
u -u+1

)du =hu

u+1
I 2
+(3/4)

u+1

du.

such an expression? The derivative of sec.x, [sec x tan x dx].

For the third integral, we set

w = (u - -2). This gives -um

We ctn obtain this by Multiplying both numerator and denominator by sec x.

Au
V

94

du .

fu' _p21/1/ilt.
u+1

+(3/4)

ti

(continued...)

95.

10

Sblutions, Chapter 3, PART II.

f3u du .+

du

w dw

13+

jr

19

Solutions, Chapter 3, PART 11

'

dx
sec x

(Exercise 10} ;

dw

tan x

w-+(3/4)

w +(3/4)

See 'solution [6] for our reasoning.

3

4.

lnlw2.2 I

1111u-11

20

tan t
65-

TTr

Multiplying numerator

anu denominator of the above integral by (sec x

we obtain

tan xi,

1u2-u+11-

1111U+1.1

-tan

(sec x - tan x] dx
)(1(sec x - tan xj[sec x

/(sec x - tan xj dx
tan x]

.sec

7*''''

1 VS

2/3

3
-2- x

lnix

4,(Exercise 5)

'

1.2/3
1/3+
'nix
-x
11

+1 -

4 .5

tan-1
t

1/3
.(z-1

jils et x - tan x] dx

2x

- tan

2x

jisec x - tan xj dx

InIsec x + tan x 1 + In 'cos x 1 -+ C

x dx

4-7x-

NOTE: As'usual, there is more than one way to approach this

To rationallze.the denomimator in this problem, we multiply

both numerator and*denominatop by

- A=7

[ .5-47i

This yields

problem. if we don't notice that we can multiply by the

conjugate of the denominator, .or .if we feel uncomfortable with
-sec x and tan x, -we can express the integrand in terms of

077 - 4;7 ], x dx

j[

x dx

WITT

+x

",

sin x and cos x.

j'cos x dx

x
I

JL

x dx

./1 -7 ]

2x

This, gives us

(I-x)

(l +x)

sin x
cos x

cos
cos x

.117-7]

sin x + 1

In (sin x + 11 +,,C

We can sho4reatily that these -are the same answer. In this

( (1+x) 3/2 +(1-x)
.

case 'the second alternative gives us a faste1 solution than

3/2 )
+

ut
the use of conjugates. That can happen; the important thing

C.

to have is an organized, logical procedure for approaching

.4
.(Exercise '1) . .
.

integrals.

x dx
x4-3x 2+2

...,-

if

See solution [5] for our reasoning.. With the substitution

u =
1

.F.

, du = 2x dx, this integral becomes

2x ar ,

(x2) 2.3(2)42

,2

du

,1 jr

u -3u+2

,JD

du

1 jr

it

(u-2) (u-1)

see solution [7] for our reasoning.

=x3,

u-1

u-2

\ du

(lnju-21- 111111-1,1)-0 C

The

2
1

:14+

C,

dx)

..1731 44

.)/('

du =,3x2 dx, thi; integral is

00)(3;2

With the substitutions

3
4.

xs dx

.( Exercise 4)

o.

ubstitutkons

liu du
3 Vfl+u
.

v = (1+u), dv = du,

reduce this to

Solutions, Chapter 3, PART II

r(v-1) dv

21

.
These substitutidns transform the integral to
2

2 v3/2

2 v1/2 + C

2
3

z4. 1-(1+u)

f (sin 0)(c'os 9 dO)

- 2 (1+x 3 ) 1/2

2fisin 0)2(

`cos .9 J

cos40

3/2 - 2(1+u) 1/ 2 + C
.
=

2
-3 (1+x 3 ) 3/2

22

jtv1/2_ v112] dv

Solutions Chapter 3, PART II

2.ftan20 sec 9 dO

- 1) (sec 9) d0

2.fsec39 d9 - 2fsec 9 d9.

+C

2f(sec20

1
; dO
cos 91

Finally, we can-see our way to e end of the troblem. The

first integral can be done by parts, the 'second by formula 15.
We obtain

dx

18.(Exercise 7)

(x+4)471;
2

See Solution [8] for our reasoning.

Codpleting the'square in

(1-[sec 0 tan 9 + In IseC 0 + tan 01)- 2(lnisec 0 + tan

the denominator, we obtain

r.

=.(sec 0

( x+4),/(x+452-42

fu a;2_42

19(ExScise

4))

, and the substitution u = x+4 yields

.)+ CC.

fgr

x
x )e
I

Solving fOrx, we obtain

'4'

2%

1-u

Jo (2u du

(1-u2)2)

. ,

_.

du

derived

IT-71

Cl

were)

2-7

.
x dx
sec x + 2
See solution [10j for our reasoning. Multiplying numerator and
denominator by (sec x) and making the substitution- u= sec x,

A we obtain

sec x tan x dx

J (sec .x) (sec x + 2)

du

fu(u+2)

Th

12

u+2

du

(In dui - In 1u+21)+ C

In u-72-1 + C
I

(1-u2)2

....

The team in the denominator suggests

the pbstitutions 1-1--T--12= cos 0;

u = sin

1 +u

pan

a(Exercise 2)

(1-u2)

ur

1-u

xtl-

2u du %
, so dx 1i-3----*

171_,

1-n-e

I sec -1 (-)
u + C = -1 sec -1(x+4
4
4
4

This is the hardest problem in these'aaterials; don't get

'dismayed if you h a lot of trouble! With the desperation '
i, x
substitution
, we obtain u 2 = --- , so u2 (x+1)*= x.
integral becomes

id

i-u2)

(W-712

tan 0)- lnisec 0 t tan 01 + C

dx

44+

V 1 int sec-x
2

isec x + 2 + C.

dd F cos 9 dO, which.we

from the 'triangle to, the iiiht.

I-u

99

umap

MERCATOR'S WORLD MAP'AND THE CALCULUS

Philip H. Tuchinsky
Department of Mathematical Sciences
Ohio Wesleyan University
Delaware, Ohio 43015

UNIT 206
MODM7SANDMONOGRAPTISINUNDERORADTIATE
VATIdMATICS AND ITS APPLICATIONS PROJECT

6/26/78

TABLE OF CONTENTS
I.

HtRCATOR'S ACHIEVEMENT
1.1

MEReATORS WORLD MAP AND THE CALCULUS

I:2
1.3
1.4
1.5
1.6

by Philip M. Tuchinsky
2.

p Strategy for Navigation with Map and Compass

Rhumb-Lines
The Need for a Map on Which Rhumb Lines are Straight
HercatOr's Successful Map
Modern Navigators Use Mircator Char4s
The Integral fsec 41 4 is Involved

5 V(

.'-'

8
10

CALCULUS AND THE MERCATOR HAe

The FraMewOrk of the Mercator Map
Horizontal Distances at Latitude $/Get Stretched
Mercator's Insight: Vertical,Distances Must Also be
2:3
Stretched
2.4 The Vertical and Horizontal Stretchi;g at Latitude
A
\ 4: Must Be Equal
2.5_,Summary: How We Get Straight Rhumb Lines
....
2.6 'Now to Place the Parallels of Latitude

10

2.1

/,2

imitagrolitup

tieliallpail
tIll IIM IMO. /IP
1

.tal

IMUMMIC
EMINISSUMFABENWAR;

3.

ONEORP44

Os MEN mosamem

3.2

3.3

Mil MIMI EMI IS

Illegarlin
wrAmihr

4.
.

MArcStfor's Map: Cartography in His Times

Edward Wright's DisCovery
Later Mathematical History

SEVERAL CALCULATIONS OF fsec.lx dx

5.F

.5.2
,6.

15

16

22

TErlisual Integration
A Partial Fractions Integration
Gregory's Form of the 'Integral

22

dx

27

A SERIES FOR ilec

APPLICATIONS OF CALCULUS TO GEOGRAPHY

14
15'

0
4.2
4.3

5.

13

16

II

MORE HISTORY
3.1

wm-'

21
24

Derivation.of allis' Series

Numerical Approximation of the Integral.

./71.

28
29

WHAT HAS CALCULUS CONTRIBUTED TO THE MERCATOR PROJECTION?

29

edc /umap /55chapel stinewton, mass. 02160 \

REFERENCES

8.

ANSWERS TO EXERCISES'

9.

7.

33

35

SPECIAL ASSISTANCE.SUPPLEMENT

43

101
4

Intermodular Documentation
Title:
ow Author:

Sheet: UMAP Unit 206

MERCATOR'S WORLD MAP AND THE CALCULUS

From Section 4:
Three calculations of fsec x dx.
Exp=lain why X-2-= Ixl, not x.
lntegrate'using both radians and degrees,
Confidently use the easier trigonometric identities.

Pl lip M. Tuchinsky
De
rtment pf Mathematical Sciences
Phi Wesleyan University
Delaware, Ohio 43015

Review Stage /Date:

Classification:

IV

From Section 5:

6/26/78

Approximate fsec x dx, ftan x dx and

Integrate a series term by term.

APPL CALC/GEOGRAPHY

Suggested Support Material: None is essential. -Ailencartdm

wall map and globe are helpful.
A spherical titanicboard (a globe painted dull black on which chalk Tines
9
can be drawn) is an outstanding classroom aid =ficrr -thhs
module.
Physics, geology, and/or geography departments
often have such a globe and will lend it.

References:

See Section 7 of text.

Prerequisite Skills:

'Discuss the advantages and shortcomings of the Mercator,piojection.

Discuss the historical need for and development of the Mercator
map and fsec x dx as interdependent problems.

For the basic application in Se.ctions 1, 2, 3:'

Definition of the trigonometriC relationships in a trhangbe.
The identity sin2x + cos2x = 1.
Recognition of integral sums and the Riemann.integralsIttery
approach.
Definition of the natural logarithm .function, tbasir ,knowtedgeoff
latitude and longitude.

For SeCtion"4 add:

Integration of x-1 to In Ix'.
Partial frActions integration.
Change of variables in integration.
Derivatives of the trigonometric functions.
Trigonometric relations like.sin-(- -x) = cos
Double angle formulas from trigonometry.

x nurerically.

arc.

From Boxes:
Briefly discuss the'achieVements of Gerber:dos Mercator, Janes
Gregory, and John Wallis.

Suggested Uses:

Tlie unit can be done all at Qnce or in several

pieces Section 1-3 plus 4.1 and exercises lb, 2, 5, 6 are appropriate as soon as fsec * dx is discussed. Section 4.2, 4.3
and exercises 1, 3, 7, 8 call for more knowledge of integration.
Section 5 and exercises 9-13 require knowledge of the series
portion of calculus and might be done much later than Sections 1-4.
The unit is also appropriate for independent reading by honors
'students and for seminar presentation-by advanced undergraduates.

Other Related Units:

MAP. editor for this module:

William U. Walton

The:author is indebted to V. Frederick Rickey of the

Mathematics Department at Bowling Green State University in
Otia for much of the material in this paper. His first
introduction to thisapplication was through ProfessorMckey's presentation "An Applicaton of Geography to
isec0(10 and its History" at the May, 1975
meeting the Ohio Section of the Mathematical
Assoelation !
of-America, and he has generously helped the author with
source material.

lnir:(eb) = In (a) + In.(b).

For Section 5 add:

.Convergence and sum of geometric series.
' attegration of fcos x sin4 x dx,,.
-For the exercises add: so
Chajn rule.

DiffiCult jirrgonometric identity work (you might,ive-bbnt).

1

From Sections ,1-3:

lutput
Describe an application which
a) causedifsec x dx to be calculated before calmultis,was
b) leads :to an. approximation sumlfor fsec x dx.
Sketch the bass frame work of a Mercator map.

Show erhumb line path between any two cities onallimrd7r7mp

Describd the mathematical principles that make 'the 1Merzatarrmap
useful to sailors.

102

0'1978 EDC/Project UMAP

All Rights Reserved

103

,
MERCATOR'S WORLD MAP AND THE CALCULUS

MODULES AND MONOGRAPHS IN UNDERGRADUATE

rHEMAfICS AND ITSYAPPLICATIONS PROJECT (UMAP)

Philip Tuchinsky
Department of Mathematical Science
Ohio Wesleyan University
Delaware, Ohio 43015

The goal of UMAP is to,develop, through a community of users

and developers, a system-of instructional modules in undergraduate
mathefnaticsand its applications which may be used to suppldMent
existing courses and from which complete courses may eventually be

6/26,L78

built.

The Project is guided by a National Steering Committee of

mathematicians, s ientists, and educators: UMAP is funded by a
grant from the'N'a Tonal Science Foundation to Education Development
Center, Inc.", a publicly supported, nonprofit corporation engaged
inLeducational research in the U.S. and abroad.

Ross L. Finney
Solomon Garfunkel

A Strategy for Navigation with Map and Compass

Imagine y'oiNtsel'f piloting a ship .at sea,-hotv can you

Suppose you have brought

the most basic of navigational aids.:
a magnetic compass
and good maps. The simplest way to use your compass

Director
Associate Director/Consortium

Coordinator,,

Felicia DeMay
Barbara Kelczewski.
Paula M. Santillo
Zachary Zevitas

1.1

MERCATOR'S ACHIEVEMENT

reliably get to your destination?

PROJECT STAFF

1.

Associate.Director for Administration

Coordinator for Materials ProduCtion
Administrative.Assistant
Staff Assistant
.

NATIONAL STEERING COMMITTEE

would be to hold its needle still by keeping youipship

moving in a constant compass direction. Thus, if you ,
travel steadily northeast, your compass needle ,(which
1
points north) will make a steady145 angle with your.
direction of motion/and the needle will shay still.
'

W.T. Martin
Steven .r.,Brams

\klayron Clarkson
nest J. Henley
William Hogan
Donald A. Larson
William F. Lucas
R. Duncan' Luce

George Miller
Frederick Mosteller
Walter E. Sears
George Springer
Arnold A. S't.vassenburg

Alfred B. Willcox

(Chairman)

New York University

Texas Southen4 University
University of Houston
Harvard University '
SUNY ht Buffalo,
Cornell University
Harvard University
Nassau Community College.,
Harvard University
University'ofeMichigan Press
.

t Indiana -Univel-sity

SUNY at Stony Brook

Mathematical Association of America
4

This material was prepared with the,support of National

Science Foundation Grant No. 5E076-19615 A02. Recommendations'
expressed are those of the authors and do not necesarity reflect
the views of the NSF, nor of the National, Steering Committee.

Firs 1-5 show such a northeaS"tward journey (an

airflight from the Galapagos Islands in the Pacific Ocean's
to Franz Josef /Land in the Artic) as it would appear on
five types of map. The airplane's course makes 'a 45 P
angle with all the meridians (the north-south lines, great
circles through the north'-and south poles) on each mop.
Which map would be the easiest one on which to lay out
the course? Figure 1 may give the best overall ,view'of
.the earth as a sphere, but the Mercator projection in I
Figur 5 is the best for navigation because
your ship's

In truth, the compas's points to the north magnetic pole,

the North (geographic) pole.
Diserepancies of this kind are
/discussed later in the Special Assistance Supplement. [S-/]

not

105

II
_

11 111011E111111111 1

.211

iu1.w :A1
_

iii1111Pgiilrilll.lillaaUU

-1143111171111Califil 11 INIPM
Tr 1,I Ai 11/2111114111111111111111
El 1E11/6111111111i11 1
IVA1111112iMmild111111111
i I 111121J11 NI 11111=1111111,11
.121111111LHINIPAINIIII II
li
/ s 1,1410111110
.
Negag11 i
islIms I.
1

Ii.

1_1_

11

nos
1111

IN 1111111

wilillimllgom
1011111111 IN I
11 II NOUN 1

,..

II

1111111M1

Figure 1.
A flight with a constant bearing of 45 E of N from the
Galapagos Islands in thevPacifi5rto Franz Josef Land in'the Arctic
Ocean.

.1

Figure 3.

Plot of flight on "plane chart" such as was in use for

charts of small areas in Mercator's rime. Angles are not true and
a straight line would not give a path of constant bearing.

1111111111111/1111

i.'

WM

1.

Illimiligin
ummunugn elmillimmul
.-iffilommorim
INECWSIFINNII

Bin
MON_
BM=

,,

ppm E

--,

Almll ,_.,
1,_

Oen

hull

WM
BM

2.-.--i'
--; -1--i- -

',

MIN"
111

HO

Ld-h

1_,.

-!
,

BMW
MIR

,44111011111

i.

vi

AM
i i
1

4Pigure 2.

The flight of Figure

plotted orf one...form of conformal
map.' (Theangles to the meridians are constant but because the "
meridians converge the patty is curved and would be difficult to Piot
and Measure.)
1

iH

Figure 4.
F ight.On a cylindrical projection, a Map often confused
with. Mercato 's. Again, the path of constant bearing is not a
straight lin
See' Exercise 5.
3

106

107

Ir4 M

III

CIAMMIMITANIMICVSMAM

de4+

1.3
1

14,1

The Ned for a, Map On Which Rhumb Lines Are Straight

-ik,

If you

ElbliiiniMIXNANIO
In
iiiiinintrirdiiiMil
wAtta

II MIMINIW' .21 WIMICIIIIIIIIIIIII

IIMOMMOINOMMEMBIgaMMOMMENIEMOM

NOMMOMIONIMMOMMIAMMUROMOMMIEINOMM
simmismagiommemmummmumwmaymminumm

draw that line with a ruler and read the compass direCtion

MINIMEMMEMPAIMUMMTAMMEMEMMIMMMOMMON

FM
imam 11131MONIIIIIMINE
... MN IIMMUMMUMMEMEMMOMMIN
IIIII MOVIIMOOMMIMUMMIUMWMA

ish to follow a thumb line course, you must

know what co stant compass direction to use from your

ftatting poi t S to your destination D. If you had a map
on which the rhumb line path between any points S and D was
simply the straight line between those points, you could
\,

by measuring,(wth a protractor) the angle at which

_1'

meridians are e cwt.

.

Before.lkiercatoes
time
sailors attempted
.
...

to use pla,iiichSrts (charts in which the lines of latitude

.

were equally spaced) forthisiurPose.

Figures 6a and h

11111111111111111111111111111111111

and 7 ,show the e

or that arises when a straight line nn

a plain chart is' ssumed to be a rhumb line course.

1111111111111MMOMIIIIIMIN
0

Figure 5. Frght. at constant bearing on a Mercator projection.

Straight line path is easily constructed, measured, and followed.

course appears there as a straight line, not a curve.

In the sixte
It's

easy to construct the course with a protractor (compass

succeeded in creati g such a map--his famous world map

Pablished in 1569. This map was recognized as a gigantic
achievment, the first signifidant improvement in map design

Rhumb Lines

Sailors have used thelcompass and followed lines of

in 1400 years.

constant compass direction since at least the thirteenth

century.
line.

th.century, Gerhardus Mercator recognized

that such a map, '41 which all rhumb lines would appear as
)
straight lines, wolii& be very useful to sailors.
He

rose) and straight edge because it is-a straight line course.

1.2

d.4 ,Mercator's S ccessfuj Map

They called such paths on a map or chart. rhumb

Cartographers and mathematicians found that the

sailors' rhumb fines became IFStal-like curves on the globe

and named them "loxodromic curves" Or loxodroma (from the

A Standard reference on cartography calls

the Mercator projection a "radical departure and improvement
over methods existing before his (Mercator's) time. In
contemporary judgment he was styled as 'In cosmographia
longe primus', which, translated, means: In cosmography
by far the first."
(Deetz and Adams, p. 104)

Latin loxos--slant and drome--running) because they cut all

The Mercator world map has become such a fixture in

the meridians they cross at the same slant angle (See

Figure 1). You should trace a loxodrinne On a globe to see,

our culture that it is familiar to every school child.

%remember this map as a very unsatisfactory early view of

that it spirals.' As a rhumb line moves north and the

)

''

our planet, because my teacher convinced me more of its

shortcomings than of its value.
TheshortcOmings are serious:.
distances are hard to Measure on the map because: northern

meridians get closer together,' the line must turn steadily

toward the pole to. cut all the meridians at thesail.ce angle.
It spirals toward the pole without ever reaching'it.

regions appear grossly exaggerated in' area (compare, Greenland

to Africa on a,globe and on a Mercator map--or in Figure 1

Historians disagree as to the origins of the magnetic

compass. You will find an interesting account of the
compa5p*
and its history under !'compass" in the Encyclopedia Brittantca.

and Figure 5); the polar regions cannot be shown at, all but
must be inset as separate maps; distances ar'e hard to

te

109

EV3111113 04
1.1
11111 f

Mal
II 1- Ill

Ira

NM

MIMI

*4

1
firsi

HiI1111 P.
IR
s Mk
lrd4 I IN
or il Briii
LI

The navigational advice obtained from charts 6a and

'6b leads to 'dHferent results.
The solid line shows' the course
from 6b, a rhumb'line that does join the Panama Canal and Land's
End.
If weobeyed.the plane chart in 6 and followed a constant
500 compass bearing we would be far off course, as the dashed
path shows.

GERHARDUS MERCATOR is the Latinized name of Gerhard Kramer,

born in Flanders id1512.
He was,the expert engraver of mapsections for a globe made by Gemma Frisius in 1536, a craftsman of mathematical and astronomical instruments, and a land
surveyor. His major achievements as a cartographer include
a globe in 1,541, a large (132 x 159 cm) map of Etdope (1554)
which made his reputation and was reprinted with corrections
in 1572, a map of,the British Isles in 1564 and the great world
map of 1569. His major-work was done at Duisberg, Germany
where he was cosmographer to the Duke of Cleves. Mercator spent
his final years creating a collection of maps of west and south
Europe, of high accuracy for the period.
It was published in
1595, a year after his death, as Atlas - or Cosmographic Meditations on the Structure of the World. Thus the word "atlas"
was first applied to a collection of maps. He.should not be
confused with Nicholas Mercator, 16147,1687, mathematician and
astronomer, nor were they related.
(Source: 'Dictionary of
Scientific Biography Vol. IX, Am. Council of Learned Ncie-

k6

Figu're 7.

Figure 6a. A straight line course joining the Panama Canal to

Land's End, England deawiron a plane chart.
It advisesus to use
Aa compass bearing of 5Q0 as shown.

MIME

ties, 197, p. 309.)

Figure 6br

The comparable straight linetcourse on a Mercator map.

The correct compass direction to follow is 569 east of compass

north.
6

111

measure becatise the scale change's as we look"along vertical

S.
lines
t
.

In the schoolroom where studentAare learning about the

relative .sizes and locations of ccuntries the map is, at its
.

,worst.

A s a navigational.aid, the map has been unsurpassed,,

for.400 years because loxodimmes appear as straight lines

and angles measured on the map are the same as those measured
on the globe.
1.5

Modern Navigators Use Mercator Charts

The shortest,path between two points on a sphere is the

Modern air and sea navigators
naturally prefer, to follow that shortest route. To do so,'
great-circle route (S-,2].

they begin by plotting ihe'course with a straightedge on a

gnomonic map (Figure 8) on which all great circle routes
appear as straight lines.

However, the compass direction

changes continually along the great ci.icle route (which,

except. in special 'cases' is not a rhumb line), and pilots

still expecf to be told to follow a fixed compass direction.

It is thus convenient and usual to replace the
great circle route with a sequanp? of rhumb lines.
t/

Because angles cannot be mbastired readily on a gnomonic

or great circle chart, the navigator selects convenient
intersections with the meridians along his great circit
)

Figure 8.
A great circle route appears as a straight line on this
gnomonic projection.
(The path appears curved becApse of an opti.cal
illusion; sight along it to verify that At is a straight line.)

course and plots these points on a large scale Mercator map.

Straight lines drawn between these points on the Mercator
projection give rhumb lines which are easy to follow as a
course and which use ully approximate the great ci le path.
Figure 9 shows the re4ulting course on the Mercator `ap.
The,,
extra distance involved in following the rhumb line p ces

found himself ih need of the integrar

instead of the great circle itself is minor in comparis

n to
the improved ease and certainty of navigation.

We'll briefly cover the mathematical history of this integral'

as well.
For-all practical purposes this integral was
evaluated long before the invention of the calculus although

1.6

fisec
-

The Integral (secs do IF

In this paper we'll explore the construction of the

Mercator map in some detail. We will see why, a century
before Newton and Leibnitz created.the calctlus, Mercator

no proof appeared until 1668, when the calculus was newborn

but known.
9

112

D113

4.

Two of out a,s-Umptiot-S-ahould-be made explicit':.

1

Our map is in "one-to-one" scale:

we will dup-

licate distances along the equator-mile for mile

(although other distances will be distorted).

That does not yield a pocket siie map but scaling

it down to printable size is an easy matter.

Thus

we'll soon talk of "stretching" earth distances, to

put them on the map!

We take the earth as a sphere.

Cartographers
can include the planet's equatorial bulge, but we

.b

will not attempt to do that here:

2.2

Horizontal Distances at Latitude

Get Stretched

So far we have placed a family of parallel meridians

on the map at right angles to the horizontal equator.
'50

70

60

50

40

30

20

troubles begin when we try to place the parallels of

10

Our

A series of rhumb line paths (straight Line segments On

this Mercator map) approximating the great circle route of Figure 8.

.Figure 9

latitude on the map.

In Figure 10, distances along the

parallels of latitude between specified meridians are seen

to shrink to zero as we move toward the poles, but those

2.

CALCULUS AND THE MERCATOR MAP

distances will

meridians are p
2.1

The zquator.is.

a rhumb linein the east-west direction and will have to be

a straight line on the map; let's place it horizontally
across the middle.

The meridians-of'lengitude are the,

and the stretching will have to increase as we move toward

the poles. The vertical placement of these stretched hori-zontal lines will have to be skillfully done to keep the
rhumb lines straight.
North Pole

'northxsouth rhumb lines and must also appegr oh the map as

straight lines.
''evenly.

illel lines.

Thus horizontal distances on'

the map will have to be longer than the true earth dist'ances,

The Framework of the Mercator Map

Let's begin to create Mercator's map.

ve Co be equal on the map because, there,

Let's place them vertically, and space, them

This gives. us accurate right angles between the

north-south and'east-west meridians and the equally divided

equator on the map.

The'other east west rhumb lines

include-the arctic and antarctic circles,:the tropics of

Cancer and Capricorn and all the other pazallels of latitude.

As we will see; our main proglem isto plaCe them as

horizontal lines with such a spacing that rhumb lines will
turn
out to be straight lines oir the mat.
10'

Figure'10.

Correspqnding points on meridians and map: EF on the

globe 'stretches to E'F' on the map.
'4:

115

.,,ft

To. see how to placethe parallels of intitulle wee ratusstt

study. the horizon al stretching. A wedge !al the eatatet
its
associated m p are shown in Figure 11. Eegimant 3AB its
f
a part of the e itor sRanning 6 '(radians);
IIff
R (aa approrimat ly 4000 miles) is the earth's raxiilalt-tieen
Re is the actual length of AB and of the co-r4;tspolidin z WEE'
_

on the map.

The horizontal stretching an now be understood.

eas=t-west length PQ' = (R cos 0) 'at latitude. 0 must be

An

tre.tr_itek into
R e on the map. (Why does P'Q'' =
R61) We must stretch PQ` by the factor
to

convert it into P'Q' for the map, because

Since

zns

)11"V

ptir
PQ

R'

(R cos 0 ) 0

PQ
F91-]".
1

cos 0

sec 0
a

lid get- as the length of P'Q,' on the map_

-22. 3:-

P'Q' = (PQ) sec 0 .

Mercator's Insight: Vertical Distances Must Also
He Stretched 0,c

Mercator's g-&a,t insight was that each piece o f vertical

must be stretched,, when put on the map,
lxy tire same factor sec 0. As we shall see,,he thus succeeded
in- pres'ervjng'angles from the earth onto the map. That is,,
iE any two lines, meet at an angle on the earth, their images
copied_ carto the map will meet at that same angle. This
he true for all "angles ever w T re on globe and map. (A
maw
pie_serves angles is ca Ledo ' conformal; -the study of,
-7e,riciioss at latitude

Figure 11.

. -map.

1 wedge of the earth and its rorrespondi-ng4firr.toff ttite

Now consider

at 0 radians north 1n1j,/mbie.

E:17)ii5Saa

part of a circle centered at 1', where T is xlizezz1 1,7111Taltith

of the earth's center C, inside the earth.. Siatme Q;1T rata

BC are parallel,, the angle 0 of 'latitude ..t.ppealas iin ttrAixaglhe

QTC where stlown.

Since QC is an earth-radius.3.4ee

PQ is a ,part of a circle of radius QT

*.

PTQ has the same central angle e as does se-L-Mar

the actual length of PQ is

!rzae taxa,zsactimit

t; thus

1/4

116'

PQ = (R cos 0)a

ttx appear as straight lines? On .the earth, recall, 'a

rifamiir*Line cuts all the meridians at a constant angle. If

the mar is=tonformal, the rbumb tine on .the, map will cut
x1111 the vertical parallel meridian lines at that fixed angle
aa.rcb will, taus be a Straight line, for straight lines are
eeccact.Wthe curves that cut a family of parallel lines all
at the same angle in plar& geometry. Thus to have rhumb

aneplutas straight lines, the.whole secret is to space

mitt tite horizontal lines correctly, placing them at such

dia.tances from the equator line' on the map that angles will
lie preserved. (Of coarse, stretching the meridian lines and

spacing the garallels of latitude around: the equator are two

PQ = (QT) 0
OT

jus=ti which 'globe-to-map functions yield conformal maps is

air impartant.part of advanced mathematics and cartography.)
Why does making the map conformal cause the rhumb lines

IL

mane

Eon
-0,#

Same
ame

task.)

2.4

The Vertical and Horizonta

Must Be Equal

Stretching at Latitude

- 2.5

How Me Get Straight Rhumb Lines

Summary:

A concise summary of our logic aw'reads as follows:

Let's explore the stretching-fbrther. _Figure 12 shows

To get rhumb lines_to appear as straight lines on

1.

a'rhumb Ha-cutting a meridian at angle a.

It cuts thee

the Map, weLmeedlo preserve angles from the earth

1Lhrorizontal.parallel of latitude at
.

onto the map.

Hoxizontardistances,at latitude M are stretched-

0 = I -

the complementary angle

a.'

2.

by a factor sec M as tht5 are shifted

Suppose we move a small distance AZ along this rhumb line

to map.

'away from the crossing 15dint.

This movement is the combinedeffect of simultaneously moyirlAz cos Ounils horizontally

(eastward) and AZ gin B units vertically (northward).3
will this moveme:Alook ,like on our Mercator map?

To-preserve angles, we must also stretch the

3.

vertical lengths along the meridians by the. same

What

factor sec M at latitude M.

If an

initioint was,at latitude M, the Az cos B.horizontal

,portion of the movement is ostretChed by a factor sec M.

globe

2.6

If

How To Place The Parallels of Latitude

As we move

the angles a and B are going to 6e preserved oh the map,

continually.

the vertical component of the motion must also be streCched

north,

along a meridian, the latitude changes

What will it mean "to stretch the vertical

lengths along.the meridian by the sane lector sec M at latitude

by the same factor sec M, becoming CAz: sin B) sec M.

Then Az.
on the earth is mapped as Az sec M qn the map. (See Fig. 13)-o

4,"?

Integral calculus

rovides a method.

Let's, tTy to

calculate D (M0), the distance on the map along the meridian

40

from the equator to the parallel at latitude Mo.

AE sin$

at latitude 00 on the Mercator map.).

enla

ed

(If we knew

the number D (4O) we would know hoto locate the parallel

AE.cps

First, we cut the

interval- from 0 to 00 into many small pieces:

let Am

represent a bit of angle located near 4, where 0 < m < M

This small bit of latitudinal Angle subtends a bit of

.

meridian RAO on the globe (Figure 14), a length of

On Earth

Figure 12.. The local scene on a globe at la'?

meridian locatedroughly St latitude O.

As.this

N, north

pol e

Wilt need

parallel at

(AZ sin 8) sec 0

latitude 00

latitude 00

latitude 0

latiludr

enlarged

C, center

of earth
Figure 13.

Dvt'ol

RAO

118

A'

On 'Earth

may help here.

Figure 14.

14

sec0

after stretching

Sime lodall scene on the map.

3RecallEng that cos2g + sin2B

+ AO

latitude 4

4:11

(Ak. cos B) sec 0

Bt

it-eqUa tor

On Map

Setting up the ittervii1 or D($0).

15

119
qo

bit of meridian is shifted from globe lo map, it is 'stretched

by the factor sec
Thus D(0

c 0 on the map.
V

is approximately the sum of such bits of

60 as m moves from 0 to 0
)

MI

length Rsec

.c.5Y

(1)'

D(00) =.I Rsec

.'
.

,60

11111111

IIIi

If we let ) ail the A0 lengths tend to zero and use more and

TreAof them, we get better and better approximations of

D(40); in the
we get

t oo
D(00) = io R sec 0d0

(2)
e....

R Jo

pko

-; 9 '-.--

1.:

it-

._

,,

sec 0 d 0.

.s.

.0.

To ,place all"the parallels of latitude o the Mercator

map, we will need D(00) for all values 0; 00 1 1 . Thus we
need 4sec 0d 0 to construct th Mercator map!
3.

MORE HISTORY

1101141

go,....0

3.1

Mercator's Map:

Cartography In His Time

Figure 15.

....

Sketch of Mercator's map of 1569.

Mercator did not )(flop that he needed the calculus to

on ;he earth's surface.

\make his map.

An almost-correct value of 68.5

miles per degree was known to Eratosthenes (200 B.C.) but not
'accepted by Ptolemy. thus distances were stretched across
too many degrees of latitIde and longitude.
Ptolemy 'took

He did know that he must place the equally\spaced-on-earth parallels of latitude further and further
apart., His map contained" minor errors in the placing of the
parallels of latitude; it also contained misplaced mountain
rangest rivers and continents, as the sketch (Figure 15) of
the original 131 x 24 centimeter map shows very clearly.
Mercitor's sources were the written itineraries of travelers

the east-west length of the Mediterranean%Sea as 62.

Mercator's value of 52 was a substantial improvement but

correct value of less' 'than 42 was not known until after

1700 A.D.
One resuieekof this error is worth mention:

and the older maps of his day, both notoriously inaccurate.

Where modern mapmakers spend their energy on the accurate

geggraphers of the generation before Columbus had stretched

accumulation of data, Mercatormainftask was to reconcile,

the Europe-Asia land mass.much too far around the globe;

the inevitably contradictory reports that reached him.

Columbus had reason to bglieve that a journey of reasonabtle('

One severe example will show the inaccuracies of

mapping at that time. Mercator's-map constituted the first
usefill, dramatic improvement in mapping the known worldt
since the time of Ptolemy (the great astronomer and

length tO the west would'bring him to the orient.

geographe 0 1400 years earlier.

Maps_of
modern quality, did Aot appear until nineteenth century
exfiorers began to carry sophisticated instruments 'into the
field.

Mercator, facing' these complex problems,, spaced his

An, important error

parallels of latitude as best he could.on the map of 1)69.

His exact method is not known. And sailors, properly

on those early maps resulted from taking 10 as 56.5 miles

4The cartogrqphic history in this section istaken
from Crone (1966).
The mathematical history is drawn from the notes of Professor V.
F. Rickey
and from Cajbri 0915)
16

""

distrustful of mapmakers, did not a.lopt the Mercator map at

17

120

12

once.

By 1600er so, Mercator maps of portions of the earth

began to appear.

3.3

,Geographers really needed an integration formula for

corrections ihmthe placement of the parallel; of latitude due

'to the work df Edward Wright.
Steadily.

Later Math0Latical History

These were of larger scale and incorporated

the integral so that lengthY.summations could be avoided.

Acceptanceby sailors grew

The first atlas of Mercator projections wds the

The following fifty years saw a search for Such a fqrmula

through.hon-calculus techniques.

Ammo del Mare of Sir Robert Dudley',. published in 1646.

By

In 1614 Napier published

tables of sines and logailthilis of sines, although these,

that time Mercator maps were_the,navigator's standard.

were not quite logarithms as we know them today.

3.2
,

In 1620

Edward Wright's Discovery

Edmund Gunter published a table of log(tan e).

The matherlidtical history that arose from Mercator's

Henry Bond discovered by chance and published'in Norwood's

achievement is astonishing.

Epitome of Navigation, as Edmund Halley tells us half a century

As mentioned earlier, we do

not know whether Mercator really understood' where to place

later,

"that the Meridian Line was Analogous to a Scale of

the parallels'of latitude to straighten the rhumb lines.

By the time,Edward Wrightorpublished

By 1645,

Certaine Errors in Navigation

Logarithmick Tangents of half the complements of the

.Latitudes."

in 1599, the secret was'out:

[From Halley (1698, p.202) as quoted in

Cajori (1915,11.314)j

"the parts of the meridian at every poynt of latitude

must needs increase with the same roportion wherewith

Bond's discovery is that

the Secantes or hypotenusae of the arke, intercepted

betweene those pointes of latitudeandthe.aequinoc-

(3)

sec 0d0 =

In

tan[.

Ci -

Of.],

a,correct formula not usually seen by calculus students.

tiall (equator) do increase... by perpetuall addition
Bond did not prove the formula, but, led a number of prominent

of the Secantes answerablA the latitudes of each

mathematicians, including John Collins, M. Mercator,

point...we may make astable which shall.shew the sections and points of latitude in the mer dians of the

William Oughtred and John Wallis, to attempt the proof.

Bond's conjecture came from comparison of tables and graphs.

nautical planisphdere: by which section , the parallels

are to be drawne." [Prom Wright (1599, pp. 17-18). as
quoted in Cajofi (191t, pp.:31'2-313)0

During the 1660's, Newton and Leibnitz produced a

systematic calculus and by 1668, via a nastily complicated
geometric argument, James Gregory proved the truth of (3).5

Wright recognized that a sum of secants was needed;* by his

"perpetuall addition" we assume he meant the,eontinuous
summation of integration. He could not have known.of
integration as an anti - differentiation' process, but the

an accurate Mercator projection and they consistently

intuitive notion of a limit of integral sums was afloat in

the intellectual seas of-that time. lorprovide the navigati
tional
corrections, Wright published "table of summation-

approximations of the integral (2) for 0 between 0 and 45

at intervals of 1 minute of latitude.
.

18

122

During theNnext decade or two, simple calculations of ftec0d0

were found. Throughout this period, mathematicians were quite
conscious
at they were providing the theory necessary for
regarded the task as an important and worthy one.
,

!Thus

jsec m do

is one case where an integral was first

treated by-summation long before the birth of the calculus

-

Cannot resist including'a quote, ascribed to Edmund

from Cajori's article.
Halley, in reyiewing-the research on
says about Gregory's'proof that it involved "a tong train of
and Complication of Proportions, whereby the evidence of the
is in a great me/inure lost, and the Reader wearied before he

Halley quoted
our problem,
Consequences
Demonstration
attain it."
19

123

.r

lt

'and was eventually made part of.the-calculus mainstream.

degree at the equator,,

This integral is one of the most esoteric that calculus

approximation

stude* are asked to handle because the' usual integration

Without knowing it, he was using the

/30

methas given seem unmbtivated., and "magicar:" - But the.

sec.i

sec 4) id.

i=1,

integral's importance for applications makes its Study

worthwhile, and we will next examine several techniques for
calculating it.
(If Sections 4 and or S are omitted, it is

Try this approximation yourself to find the distance in earth radii

still appropriate tp9read Section 6.)

use steps of 5Q.and then of 1 (Remember 1 = 4-1 radians).

(radians) from the equator to 30 north latitude on the map.

1

o0

your results to the exact value given by equation 3.

First

Cohpare

Do you

Exercises

think Mercator's probable method was sufficiently accurate. for a

Differentiate to confirh that

1.

a small scale world map?

4

a).

sec x dx = - In

I
0

tan

x))

Exercise 12.

.1

5.

4).
0

When you have completed Section 5, compare

your sesult to the value given,by the series approximation of

sec x dx = In (sec x-+ tan x) for 0 < x < m/2

I was taught, erroneously, that a

Mercator map is obtained when) a paper

41.

cylinder is wrapped around the earth,

K
x
c).

tangent it.the equator as in:the

tan (2(1+ x))

-sec x dx = in

sketch, and points on earth ere,

d'
2.

Starting with blank graph paper, make part of a Mercatoromap,

projected onto the cylinder as though

as followt:

by a point-size light at the earth's

put in the equator, the meridians at 0, 30, 60,

What spacing of the

330 latitude; and parallels of latitude at 10, 20 , 30, 40,

center.

50, 600' 70 ane80 north and south.

longitude-circles does this

(Arrange the scale so'that

these parallels do fit.

projection involve (1nstead of the

Now, using a globe or non-Mercator world

map as a source of data, sketch in Greenland andoAfrica.

Do your
D(4)) = RI

results look about SlieFigure 5?

Do-Africa and Greenland have

roughly equal areas, as the map seems to say?

(look up the actual

placemelit of the circle at latitude

.

on the MeiPcator.map)?

The formulas in Exercise 1 are for x measured in radians.

Convert

any one'of.themrso that it gives

eir

sec 4) d

area in the almanic or atlas.),

3.

'

4)

"

.....)

Are the

Ongltude lines more spread out on

CYLINDRICAL PRPJECTION

this map or on the Mercator map?

Point A is mapped to A'

Figure 4 shovA a map. made with

sec y: dy

this cylindrical projection.

for y measured in degrees.
'

4.

(Hints for Exercises 3 and 5 may be found on page 32.)

It is probable that Mercator Constructed'his map grid by using a

table of secants-at one degree intervals.

Adding up the secants

from one degree to 30 degreesWould give him the approximate

spacing of the 30 line of latitude in terms of the size of one

20

MM.

Ide

6.

and no motivation other than "look, it works" seems possible.'

An'swer each question,z_supporting your answer with speCific

We have used the well-knowh result

evidence from the unit:

ii)^ Do we know enough about integration when we have

b)

ln

ill

Did Newton and Leibnitz create 'the calculus in response

ly1+ c.

learned to calculate integrals by antidifferentiation ?1

4.2

A Partial Fractions Integration

to'scientific questions as part of the.intellectual

A lit- tle obvious trigonoietry permits 'us to calculate

growth 6T:their age or did they create it out of thin

c)

air because.of its internal logic and beauty?

the integral by partial fractions.

Can fsec (x)-dx be Calculated between limits x = a and

been marked for further comment:

x a b without the use-of a "closed integration formula?"

dx

isec x dx
d)

What is the advantage of sailing a rhumb line course' as

oppoied to another course?

Are there disadvantages in

cos x dx

cos x

On -an accurate Mercator map of the world, how or where

f)

at CI

The multiplication by 1 = co,s x/ cos x

ahead.

sly

is done so that

can be done, a modest example of planning

the next 'step (1.;)

On an accurate Mercator world map, does an inch of map

cos x.dx
J (1-sin x)(1+sin x

is- the north pole located?

0.

J cos2

jcos'x dx

sailing the rhumb line and, if so, what are they?

)

Some equal signs have

Here are the partial fractions

distance along_ the parallel of latitude at 400 north

(1-sin x) (-1+s

x)

distance along the parallel at 30 north?

g)..

Thus

Is fhe Mercator map an easy one to use to measure the

distance between Chicago and New Orleans?

4.

4.1

cos x

fsec x dx =

SEVERAL CALCULATIONS OF fsec x dx

The Usual Integration

1/2

1/2

represent the same. eacth-distance as and inch of map

4177373'i x

,r

cos x

L- ln (1-sin x)

ln

l+sin x

ln

il+sin x

dx

ln. (l+sin x)]+c

+c

1-4in x

A typical, sneaky calculation of this integral is

A_
fsec x

dx = 5 sec

x`.

sac x + tan x
dx
sec x + tan x

0.1

h(sec x + tan x)

sec x + tan x

*This sect ion may be !Olmi ited.

1+tin x

+s.rinx)+C

(l+sin x)24.,

1-sin2x

dx

(l+sin

in 'isec X + tan xl +c

(4)

-six

ln

See Suggested Uses on inside

of title page.

22

=In

cos2x
l+sin x

Nilcos x

127,

x)2

+ c

+ c

23

Il+sin x I
cos ,x

ln

Here it is:

jsec x dx =

in Isecx.+ tanxf + C.

dx

dx
,

i 'cos x

Again, the decision at Eg leads to improvements at the

following 0 steps.

dx

The step marked

(5)

sinr7/2-x
'
2

rests on the' fact tliat

1747 " IYI

dx

although you might think more immediately'of

cos y
(6)

ilFri

cost y

'

cos(212-x)

where y ='(n/2-x)/2

' ST?

seey dx
Both (5) and (6) are correct when Y
when y is negative;
/(73)70

1-51

0 but (S) is-still true

tan y

Now change variables to y, using dy = -4 dx, a74 get

S
j

while (6) is not:

sec2y dy
tan y

ln 'tan yf +c

IFTTP # -S.
Thus, the absolute value bars arise very
integration formula.

turally ih the

(7)

In

The algebra at
The calculatiQn involves no t ganometry more
sophisticated than sin2x + cost
1 and a/cos x sec x , but
requires a little algebr
zation.
It was apparently
first done by Isaac Ba

t 1670 and may be the

earliest use of partial

4.3

Gre o

's Form

oninsintegiation..

interval xe(04), sec x > 0 and the integral y 0.

But (n/2-x)/2
is ,between 0 and n/4, its tangent is between 0 and 1, and the,

In would be negative.

The minus sign straightens that out.

Another form of the integral is given in Problem 7.

And it should be possible tp, convert In (sec 0+ tan 0) into

e Inte ral

It is not difficult to derive (3), Gregory's'form of

the.integral.
The needed 'trigonometry this time is that

cosx=sin (I

is sneakier than one might like, 1, adMit.

The minus sign here is not unreasonable.
For our basic

- 1n4tan((n/2-x)/2)) via trigonometry, should it not?

are asked to do so in Problem 8.

You

x)

and the doutle angle formula

sin

-'x)

2 iin(2.G- - x)) cos

24

25

Efenercise_

7;- (ae ideas close to those we used to `derive Gregory's formula,

JAMES GREGORIE or Gregory..1638-1673, =eat:odium:Chu:cm

of infpoctance in mathematics and optics than ite*gas,gineenoree-

Alit for in his own day. A great technological achlevenrat oft

that time was the design of efficient low-xlistorticntedizscomes:

isecxdx = -In tan 4ce-.)

r:1

Gregory contributed eiperimental designs that influenced:atmton' s reflector telescopes; the Cassegrain xlesign
ife72%aas

irr section- 4- to derive this formula:

4

the ultimate successful result of this effort. :4

fsec x

Gregory put much effort into finding the lengths, aereeas

and volumes associated to the conic sections. These'-oesdtts
were needed for engineering work such as design cffoapti:mil
instruments. Difficult integrations were involved, mini *gene
done by geometric methods using classical areek.kicnovikeigeoff
the conics, physical principles, etc. ',;The calr.ulatixmcoff
MercatorProjectivn is xmeaancarodte.
Jsec x dx for use
His later mathematics centered on calculatism xsf -asalasaff.optimi-

8E

+x) + c.

Th

a). Show.via trigonometry that, if 0< x< /r/2,'

tan (1-(X+

WI N

.= sec x + tan x .

This may be done in many ways. - My own inethod 'started from

the for:1;11as for tan (A/2) and gn (me) .

nomials. He used approximation methods that -were -redilscovened

by Newton, Simpson, Taylor and Cotes some .years later-, and

credited to them. He was also a pioneer in the -use xff inffirithe

series; see Exercise 13.
.

to.

Now show cot (-12-

- tan f(-I; -vx). (Hint:

draw

graphs of the tangent and cotangent curves and give a

geometric sort of proof.) Thenexplain why the two
integratiohs lisited_in problem 7 are guivalent...

His work was not influential hecausvgragory amactiinzgaat

isolated Universities, was too much out )5f zommuniamtficln*.:Ith
his proper peers. His generation saw its -work absuched ass
small portions of the deeper, richer,,systematir caitctIlpsddeveloped by Newton and Leibniz. (Source: Diationra2 )2-7..-Zdi-

entific Biography, Vol. V% pp 524-530 and X..31. lloyet.,

1 If tan

(Mitt for- Exercise 7 may be found on page 32.)

HiiartoTy

of Mathematics)

5! A SPITES, FOR fsec x dx

S

JOHN WALLAS (1$16-1703) was the greatest Inglbsii-rmathe-

Rern11 that Mercator's need was to calculate

matician of the generation that preceded 'Newton. lie.wasmn

important leader of the transition from Zreek geometric retebindes
to modern algebraic methods. Two books publ ishexi ob also Hz)
1655, one on analytic geometry, the other ?An.
ices-hail
methods, were bothnfluencial. He deriAd-man ^Einviapertscil

results of the'calculus, Including

a

n +l

by algebra -based methods that were a great-iinTiffixacti:onoff

the geometric derivations used earlier by tavalizri wridcathess.Some of his "prdofs" were incomplete or erreneouswndgeere

critized during his life even Though the resu

he thus helped mathematics.make enormous

ageoecooreett.;
rhaaihqg

the vigorous and most efficient perivatidns to 'found *y

others later.
sa
c 1 ergyman, and - , . lain a:DK/Lim
Charles
.
Source: C. B.
r, A History Weraencrbiga,
John Wiley, New York, 1968, Chapt- XVIII.

sec 43 df)

fbmirmanr values df 4), even every 1/60 of adegree.' While

Ornevacr"s proof

al',:logarithiick Tangent" formula was

Gomm= fox-tbois integral was valuable, it helped the task

.-----afrcmmputmtion only to the extent th',4t tables of log (tan14-1r--77
wexmaarailahre
In 1685, John Wallis published a series
liaxmmoE the integral; offpfing a wholly new and. fairly

anammitartdomputational method.

section- may be- omitted.

See Suggested Uses'on inside of

27

130
131
-1

Derivation of Wallis' Series

5'71

5.2

This series is very easy to derive.

From section 4.2,

in thepartial fractions derivation, we have

Numerical Approximation of the Integral

We can use (8) to approximate
ix Sec x dx

josecxdx

..

xo

-fg
0

cos x dx

1 -sine

by getting y = sin x from a table and taking partial sums

of (8) until the desired level of convergence is obtained.

icos x [1 + sin2x + sin4x

0

+ sin'x + ...]dx.

All we have done here is to use the geome.tric series formula

You are asked to do so on the computer in Exercise 12.

I The series is not an exceptionally fast-converging one.
For x = W6 some partial sums are:
Highest power of

= 1 t a + a2 + a3 + a" + ... if lal

sin x- included

< 1

Partial Sum

.49999999
.54166666
.54791666
.54903273
.54924975.
.54929414
.54930352
.54930556
.54930601

with a

..sin2x, (which does satisfy 'al< 1 for the xc(O,n/2)

that concern us).

Thb next step is to convert this integral-of-an-infinite-

11

13
15

into an infinite-sum-of-integrals, which we calculate

rm by term6.
Continuing:
x

The correct value, fOr, comparison, is

cAtN

cos x

sec x dx =

1 dx +

cos x

17

fn/6

dx +

in

cos x

.s in

x dx +..,

JO
= sin x + sin3x + sinsx

sin2x

+ ---r

+ ...

'tan x dx

y + Xs + XI + X

... with y = sin x.

4

0.54930614.

Many integrations that lead to obnoxious formulas can

be convected 'into series calculations leading to convexgent, computable answers. Wallis published a series for

L.-

( x

cos x sinkx dx +

(8)

-sec x dx

This series is convergent for any xc[0,7/2) as

Exercise

11 asks you to show.

in 1685, too, and we include this one as Exercise gas one

See alsoExercise 10.

example.

6.

The reader shoUld be warn9d that, in general, it is not true

that the integral of an infinite sum is equal to the term-by-term
sumof the intertals. However, as you will see proven in more
advanced courses, the Calculation here is legal because the series
involved Is convergent for all values in a closed interval [0,x]
where 0< x< it. and the functions.inolved, including the sum, are all
continuous.
8

WHAT HAS CALCULUS CONTRIBUTED TO THE

MERCATOR PROJECTION?'

The map that Mercater published in 1569 was revolutionary because it simplified the task of navigation
at

,sea -- sailors could plot rhumb line cour,rs by the simple

use of straight lines.
By about 1600 corrected versions
of Mercator world map were accurate enough to satisfy the

practical reqdirements of navigation at sea and the map

1.32

133

29

tf

soon came into wide use.

-,0

But all of this was accompli.shed

before the invention of the calculus; what has calculus

S4

1,

tan x dx = Tts-

As, ore and more detaile'dlieicator charts of smaller

) dx

S6

S6

where s = sin (x).

over the centuries, a more and more accurate-spacing of

For what x does this converge and why?

10.

What goes wrong when we try to carry out Exercise 9 for

precise mathematical Calculation of fseco dO was known,

cot x dx?

this spacing could be immediately accomplished to any

JO

The only limitations in the produc-

tion of Mercitoi maps were those imposed by the printing

II.

Give a proof that Wallis' series

v3

I-

process, size and quality of paper, and so on.

1-sin2x

+T+T4""7.+

smaller parts of the earth's surface have been made

and
__/

degree of accuracy.

The answer will, be

x

Once the

= f sin x cos x

really added to the achievement& of Mercator?

the parallels bf latitude has been necess ,pry.

lir

NoAmath-

y7

with y = sin x for some xc [007/2), is convergent

ematical barriers stood in the way of the cartographer,

because methods had been provided to create the Mercator

v5
4- 1--

ar

a)

,by a comparison test

projection both in theory and to any level of accuracy

b)

by another test

in practice.

c)

For what y (and then what x) does this series converge?

12.

The influence of Mercator on the course of mathematical development was important.

A computing project: use Wallis' series

3(

T 7T+

sec x dx = s

Along with many other

... where s = sin (x)

technologi5al-problems of that age, the problem of reto calculate on the computer successive partial sum approxima-

fining the Mercator projection to",a high Ovel of accuracy

inspired the mathematicians and guided their efforts 'in

developing the Calculus. They did not qjt working on

Your instructor wiJ1 tell you what

interval [0,x] to use.

Continue until you have_the integral

approximated within .5x10 -6.

cartography- inspired problems once the Mercator pt'oblem

Sin1101600 the Mercator

projection' Ls`been further refined (to take intoLaccount

the machine?
13,

a)

Derive Gregory's series:/

use ofmOre mathematics and many other projections have

arctan x =

been developed on a sound mathematical basis.

Exercises
9.

How will you decide when you have

calculited enough terms t'o the series and are ready to get off

had been solved, of course.

the equatorial bulge of the earth, for example) through

tions,of the integral.

)0

Hint:

dx

= x -

x3

x2

x7

Replace 1/(1+x2) by a geometric series before

integrating term by term.

Use this start to get a series-for, also given by Wallis (n

b)

1685, for

'tan x dx
J

c)

x dx
cos x
sin x cos x dx
cos2%

or What x does this alternating series'Conver

Use the computer and this series to get a tablet:pf arctan

x forx-vaiues that your instructor assigns.

30

What is an estimate of the error if yotstop after soNnie

terms?).

134

(How-can you

easily decide when to get off the machIgE, for xc(0,1)?

31

135

d).

What does the series tell you about a relationship between

arctan rand arctan (-x)?

7.

Caikri, Florian, 1915:

Hints For Exercises

180

3.

Change vaeiables using

5.

In Figure 13t angle e is the latitude of points A, A', and D2(e)

Carslal, H.S., 1924:

"The Story of Mercator's Map. A

Chapter in the History of Mathematics." The Mathe-

gives the spacing of the parallel of latitude just as D(e) did

for the Mercator map. 'Try to find D2(e) and then show

matical Gazette XII (Jan. 1924) pp 1-7.

D24) 10(0.
.7.

dx

Start with isecxdx =

cos x

Crone, G.R., 1966:

Maps and Their'Makers.

Books, New York, 1966:

dx

sin(x+i)

CaprIcorn

Deetz, Charles H.0 and Oscar S. Adams, 1944:

Elements of
MaR,Projection With Application to Map and Chart Construction, 5th Edition, revised, 1944. Special.Pub. licitidn No. 68 of The Coast and Geodetic Survey, U.S.
Department of Commerce.

9: Reread the beginning of Section 5.1.

10. The lower limit of integration, 0, causes the.integral to be

improper:

"Oman Integration Ante - dating

he Integral Calculus" Bibliotheca Mathematica, 3rd

Series, 14 (1915) pp. 312-319.

y x. The result wilrbetan integral

in x, where x irmeilsured in radians.

C

4F8RENCES

'

Is it finite/

Halley, Edmund, 1698:

"An Easie Demonstration of the

Analogy of the Logarithmick Tangents to the Meridian
Line or suit of the Secants: with various methods for
Computing the same to the utmost Exactness", PHilosophical Transactions XIX (1695-1697), London, 1698.

Hobbs, Richard R., 1976:

Marine Navigation I; Piloting.

Fundamentals of Naval Science Series, Naval Institute
Press, Annapolis, Md., 1974.
Reprinted 1976 with
corrections.

Parsons, E.J.S. and W.'F. Morris, 1939:, "Edward Wright

and his Work",hmagio Mundi 3 (1939) pp 61-71.

Taylor, E.G.R. 1971:

The Haven-Finding Art.

.Hollis. and

Carter bondon 1971.

A

hq

/
Sadler,'D.H. and E.G.R. Taylor, 1953( "The Doctrine of
Nauticall'Triangles. Compendious.
Part I
Thomas
Hariot's Manuscript (by Taylor)., Part II :Calculating the-Meridional Parts (by Sadler)." J. of the
institute of Navigation, London 6 (19,53) pp 131-147.

137

Wright, Edward, 1599:

,Certaine Errors in Navigation,

8. ANSWERS TO 'EXERCISES

Arising. either of the ordinatie erroneous making or

rising of the sea Chart, Compass, Cross Staffe and

tibles.of declination of the Sunne, and fixed Stirres

a)

crx

[1 ro

(2 (2-x)11

tan

detected.and corrected, London, printed by Valentine

Sims, 1599.
Note;

1-x))

sec2

Historical,material in this paper has been drawn

;-

(-1)

tan

almost totally from"(Cajori, 1915) and (Crone, 1966)

only.

Other references are given to allowLthe reader

2

quick access to the literature for'further research.

1/ cos y

X)

where y =

sin y/cos y

(iT

In this brief paper, complex historical ideas have

-naturally been compressed more than they deserve.

Any

.1

sirT)4,

2 cosy siny

= csc(2y),

inaccurate impressions that may be ponveyed as a result

are the sole responsibility of the author of this paper.

csc

b)

dx

x) = sec x.

( 1n. (sec x + tan x ))

1
.

sec x + tan x

(sec x tan x + sec x

(secx + tan x) sec x

sec x + tan x

te
'

c)

g. sec x .

dx

Cl

n tah

+ x))]

tan.

IA +

sec

xl)

ll

for z =

x)

just as in (a) above.

2 sin z cos z

si7-riz
1

csc 2 + xJ

sec x

2.. The area of Veenland is approximately-840,000 square miles,

and the area 'Of Africa is 11,706,727 square miles.

34

3 5

139

phi

fsecydy - fsec

( 180

ef

180

dx
(Sample answers)

6.

The construction of the Mercator map leads us to discover

fsecx dx as a limit of integral sums,: derivatives do
not
enter theoproblem. The integral was adequately approximated
decades before an antidifferentihtion formula was disCovered.

,.178,2"fsecill xjdx
180
w

In

seC

1.12.x

+ tan [180
x)1
V

b)

Ideas that are now part of the calculus and problems ialljng
for the calculus were "in the air" long before Leibnitz and
Newton. For example:
Cavalieri's integration of, xn;
Mercator's need for fsecx dxr Gregory's geometric calculation
of integrals.

c)

Yes, as a finite integral sum E(secx;)Ax or by use of

finitely many terms of Wallis' series.

dx

d)

A rhumb line is easy to sail because the pilot simply keeps

an eye. on his compass.
He wants to keep the compass needle
reasonably still. One disadvantage.se'the rhumb line-opath
is its greater length in comparison to the great circle path.
Extra distance costs time, fuel, and money,

e)

The north pole needs to b

in Isec y + tarry ( + c.

Change of variableS:
yAir

dy =

4.

180

10

1300

With 5 steps,

seNit d$ = 0.5564789339.

0
F

t
With 1 steps,-

ino
l'"
secf df = 0.5506730838.
)

0
o

2,,o

p(i)

away from the equator.

`Equation (3) gives j'" sec. df = - In tan 0 --'f))+=i0

. Rf:"'
se 0

4=0

. 0

located a distance

d0 =
is off the map!

It

If an inch of map diStance at 40o N represents A earthmiles and an inch at 50N represents B miles, then

= 0:5493061443.
5.

From Figure 15,

s

g)

D2($) = R tan f

must be compared to
D1f)

sec 40

-ie"C-5-g

because of the "stretching" of earth distances as they

are placed on the Mercator map' Thus A # B.

tan $

so,

The scale changes continually along north-south lines

of a'Mercator map. Thus a mostly north-to-south distance
like that from Chicago to New Orleans is quite hard to
calculate from the map.

ec $ d$ .

7.

dx

dx

sec x dx

cos x

The easiest way to show Do(f) > D(f) is to notice that the
deriyatives are easy to ampari:

sin [x +

2
Sit

dx
d.; D201 = R sect

j 2 sin y cos y

R sec f =t

D(f)

)2

all fe (04).

36

140

dx

Since 171(0) = D(0) = 0, we can conclude that 02(f) > D($) fqr

sin y
cos y

cos

}n tany +

tan y

141

37

a)

-sin A'

'tan
.

A
.2 sin2' I;

'"

A
A
- cos y- 2 sin 1- cos 1-

This

b)

Easier Ablution due to Ronald Shubert, Chairman, Department

-of-MatheastIci, Elizabethtown College; Elizabethtown; Pa.:

tan (4-+ y), y =

A proof will drag you into

straight-forward use of formulas like (**) above in 8a.

sinA

9.

jtan xdx
6x + ...)dx.
Jsinx cosx (1 + sin2x + sin4 x + sin
Put in y = sin 2x :

and thus,

tang (it 4.

1
=---

. csc

when the graphs are drawn.

1 - cos A

cscA- cot A,,

[it +
tan 2

see in the form

easy 't

1 - Cot (x + 12

1(1

+ y + y2 + y3 + ...)dy

(x + 2

c+y+

+,,.)

ow sec x + tan x .
.

My own s

utlon is longer: Recall that

).

A + tan
lea) tan (A + B) It 1 Stan
A tan B
and

sin 6x +

x
(c + sin 'x + sinie
2

..
10.

The integral is infinit and we get this absurd divergent series:

Ix cot x dx

tan

(Al
2

1 = cosA
1 + cos A ,

dx

cos x

in x

si

then

tan

tan

I co
sin

cos 'x sinx dx%68,

1 - tan2

i+

!.

1 + COS X.

-,177---SX

1 + COS X

+j1.1 +cos
X
COS X

it

or

COS X

-1 + 1.+ Ca X
1

+2

;s 2x

th y cos 2 x

v.

-1110SX
1+ cos X

1- COS X

1.+ cosx

'1 - -cos X
+ cos X

,
+ y + y2 + y. + ...)dy

2
1

sir

+1

u3

(c + y

2
+ 2 /(1 -cosx) (1 + cos x )
1-+ cos x
1 + cos x'

The left is +

e2

and the "right slifelooks negativel,

111K

2 cos x
1 + cos x
2 + 2 cos x
2 cos x

SieX + tan x

11.

a)

A term by term comparison with

r+Y 3 +r5 +yI

4. .

(a convergent geometric series with y sinx<) shows'

the convergence .of Wall is' series.

39

142

143
5

(
b)

The ratio test

tells us to talr.ulate, laro3urnef6is
,
..

.
.

9,

RIGHEtT POWtR
OF SIN(X) INCLUDED
IN PARTIAL SUM

v2n+1

)n o10 17----it
G

., 2 n+3 /on+3) .

'

R go lint

y2//+! /(2n+1)
.

The rat io- tartne in (b) tells -us That 'Mile

hemmer bas
divergent series. For y 1 -1 -we get a =memoirs aziteernating series. Thus we have convergence 'fcrr -41 yy 11
exactly. Now, what x gives -1 K _strix 42 .A11 mail >cc,

'

..
...

re, amid

1.28128n0

21

1.31169002
131604511

fRfiliTPAL SSW

.X1127135-10e

3
3

3335873f602

xx:.--- 60.

31

41

012159882904

aLUNt3352

mt=--- 71j0

co226184693
012268812221

J. =2,02758941

004449999995

3
5

Q15511 i66666
e055410666
m554903273

-7

'6 .54930614

1.31675696

09

1.31

79

1.,31695747

0.96592522
1-99537299
2:02425779
2.02717058
2.02753181
2.02758099
2.02758805
2.02758893

51
101

201
251
301
.

495'

1-31

71

151

1_31
-,.,

C0a647,51814

to'

1 gm .26484224

51
61

a:oda-B-043

30

0.88137329

OF S1N(X) INCLUDED
INePARTIAL SUM

R.

0.88137296

33

0.36682540

159

0.88137n6

31

.08737743

0.:83137078

29

11

x = 5
The correct integral

"

0.83136758

27.

HIGHEST POWER

fil33699789.

X".

0,:8888113241391278

0.88134566
0.88136067

23
25

is

0.88107761

21

except ....?

Computer results for x = '5, 15 , 300, 4313,

80 are below.

0A.138070933

17
19

< 1 (1 .e. , -1 < y -< 1 ) .and dd 'remiss tqkben

0.87785062 0.87965544

13
15

1 .--- .88137358

y2 > 1 (i.1,., y > 1 or y < -1). Tor .y==-1.weettaa

42,c,

0..86931325
0.87294015

7
11

20 ..=.115

also be used.

converges when y

00P...M1-87:71:55791:67811

,4

Since R < 1, the series IS convergent. Zither teejszs way

c)

3
5

335

m553914975
11

.13
15
17

00S493504
01$4990352
0DS4990556
01554990601

4y.
a

145

9.

HIGHEST POWER
OF SIN(X) INCLUDED
IN,PART1AL SUM

PARTIAL SUM

.101

151

201
251
301

351
401
451

80

I. 2.43624604

(S -l]

The system of latitude and longitude is trased

on the geographic north and south poles, but com-)

0.98480775
2.21986109
2.39058930
2.42057205
,2.43044352
2.43400011
2.43535054
2.43588126
2.43609499
2.43618263
2.43621907
,2.43623439
2.43624088
2.43624365
2.43624482
2.43624532
2.43624552
2.43624554

51

501

551
601
651
701
751
801
811

lz

pass needles do not usually ,point' to the geogra ic'

north pole.
Instead they point to the north ma e-'
tic pole at -75 N. latitude, 101 W. longitude i

far northern Canada, north of the Dakotas. Naviga-,\

tors are used to correcting compasS readings for
fi

that discrepancy.

Thus we will speak conveniently

of compass north ag the north pole, 90 N. when
that is not true.
Compasses, in fact, must also be corrected for
deviations due to the magnetic iron.in a ships hu1,,0A
or cargo holds and even for iron ores in nearbnd
masses.
The hncyclopedia Brittanica articleundeT
"compass" discusses this in more detail.

For x nearing 90, sin x near 1, we need quite a few terms in

a partial *sum to get good accuracy!

a)

SPECIAL ASSISTANCE SUPPLEMENT

arc tan x.=

0

dx

+ X2

x
X2

(i

X4

X6

X8

'") dx

A great circle on a sphere is a circle of the

largest possible circumference, like the Equator. All
the meridians (north-.south lines of constant longitude) are
great circles. The circles of constant

. X

3
b)

7"

use the ratio test:

x2n + 3/ (2n -+
R = lim
x2n 4- 1/ 12n +

latitude are (except for the equator, latitude 00)

not great- circles because these east-west directed
circles get smaller in size as we progress from. the

= X2
1

Thus the series 'converges when-x2 < 1, diverges when

x2 > 1. For x2 7 I (x = .4.- 1) we get a convergent alternating series.
d)

equator toward either pole'.

Thus the series converges for - 1-< x7< 1.

Plug in -x for x and thaw arc tan (-x) = - arc tan jx):
t

146

Notice that there is a full set of meridian'

ghat circles reaching from the North Pole to alil the
other points on the spherg, ant that all of them also

pass Through the South Pole, opposite the Nort Pole

on the Sphere. Thus there are infinitely man
great
circles routes \between the two poles.
If w/ need to
42.

43

147

6
connect the North Pole *wiih any other point P ex.

cept the South Pole, there is a unique great.circle

passing fhreugh N and P.

The shorter arc between N

and P. along that unique great circle is the shortest-

path on the globe connecting N and P, the gret circle

route between them.
Similarly, there is a full set of great circles
through any point Q on the globe.

Between Q and its

opposite p 'nt R there are infinitely many great
circle route
Connecting Q and an other point R
(not opposite to Q) on the sphere,
ere is a unique
great circle and along that circle lies the great
.

circle rout, again the shortest between Q and R.

ti

The Project would like to thank Marjorie A. Fitting of

San Jose State University,tBarbara Juister of Elgin Community
College, Roland Smith of Russell Sage College, and L.M. Larsen
of Kearney State College for their reviews, and all others who
assisted in the production of this unit.
This unit wat field-tested and/or student reviewed in
prelimary form by Alan Shuchat, Wellesley College, Wellesley,
Massachusetts;.Peter Nicholls, Northern Illinois University,
DeKalb, Illinois; Richard G. Mokgomery, Southern,Oregon
State College, Ashland, Oregon; Kurt Kreith, University of
California at Davis; Robert L. Baker, Jr., U.S. Naval Academy,
Annapolis, Maryland; James Bradley, Rqberts Wesleyan College,
Rochester, New York; and Jonathan Choate, The Groton School,
Groton, Massachusetts, and has been revised on the basis of
data received from these sites.
so

44

148

MANAGEMENT OF, A BUFFALO HERD

by

Philip M. Tuchi sky

Department of Mathem icall Sciences
"Ohio Wesleyan University
Delaware, Ohio. 43015

S.

TABLE OF CONTENTS
r-o

,..

INTRODUCTION , -7-r

1.

1.1

What Harvest Should You Take?

1.2 What Herd Shobid You.Start With?
1.3 Wildlife Management.
1.4 _The Task Abead

2
2

._

. -THE MODEL
...2if'llerd Components and Their Survival Rates
1.2 Basic Model Equations

,2.3
.

Ytt,YF

humbeWrifOlii e-Syear itigs;-fema le y

female cal
''.0,',:illirribarlot,!,iiiiies.;haeVeited.;."il
64,116ei: a -

adult

.Matrix Notatiori:
..'
The Before-Harvest Jiodel

2.4
rt ngs

2.:5

Survival Rates Would Be Larger on a Ranch

Exercises and Computer ?rejects

-2.6

s'

The After-Harvest Modqin Vector and

nuM4e'r- al. adilltisie;

year"

6
.6

..!:

telltales haivested ''next year"

:. ASSUMPTIONS, STRENGTHS, AND WEAKNESSES OF THE MODEL

.

3.4 14 Modes
l' Basic
3.2

St rengtps

Limitattoris of the Model

.. ..

.
, ... ....
.
,

..

1 0 10

APP,LICATioNii ,ESTABLiSHiNG A HERD

A Her:1 and- Rarvest That Continue Year After Year

4:2Cafcelatien2of' 01-0-1, : . . . : . L . . . . . . .
11.3' :A IteadY flarvest'FlUs.'eontrol led Growth

, 4.1

- of, the'
:' e
. 4:41Exeralles '. .,'.
:-.34.5

..,

-. .

Mathematical Ins ights

.... . t.

. .

. *.. .

11

..

11
13

16
16

t,

17'_:-,_

'4:6 'An'.Effietent.IY,Small Here.4.7 . Exercises . .

'18

22

M1

5.

APPLICATIONS:. CALCULATING THE HARVEST

22

% 5.1 -Steady AnnUal Harvests ancl:Hard

(L < LIE*YPEM4EL),

T..

5.-2-1-Constant-Harvests-From-a-Growirig-Nerd
'...

. 6,
7.

5.3

Exercisei

. .. .

. -.

'''.

25'

27
.,,

RetERENCES'-:

27

ACKNOWLEDGEMENTS
,

'ANSWERS' TO

22

:EXERCISES'

10_

hismodille is suitable for a first lineal algebra-course ,

'Intarmoduiar DOScription Sheet: .UMAP Unit 207'

or a post - ,linear- algebra course in mathematical modeling.

It is suitable for presentation by advanced students in a

Title:- MANAGEMENT OF A BUFFALO HERD

seminar.

Author:

Philip M. Tuchinsky
Department of Mathematical Sciences
Ohio Wesleyan University ,
Delaware, Obio 43015

Review Stage/Date:
Classification:

III

The matrix is not diagonal:sable: Tor pursue the calculations

in Section 4.5 further, the natural path is to seek the
eigenvalues of matrix M. But the more general

12/28/77

APPL LIN ALGAARVESTING/LESLIE-TYPE MODEL

'Suggested Support Material:

References(

Kay exercises call for computer use.

See Section 6 of text.

0
0
0
0
0

0
0
0
0

0
0
d

0
0
0
0

0
0
c

0
0

0
0
0
c

0
0

has

Jordan
form .

inverse by row (Gaussian) elimination (optional): overdetermined lineal equations;' elementary matrix algebra.
No eigentheory is used. No background in biology/ecology/
ranching is eeded.

det(M

Discuss major strengths and weaknesses of a linear model,lp

OA
f
a nonlinear reality.
king varied
Simulate several policies of hary estuse of a computer simulation.'
,

- AI)

A2 (a - A) ( -A3

**

aX 2 + bcel- 0-

The eigenspace of zero is unfortunately one-dimensional; thus

the 4ordan form above: The.square, cube, and higher power's of
Jordan form are diagonal.

1977 EDC/Project UMAP

All Rights Reserved.

"-

Predicteereaching Time:

2-3 class periods,, including discOssion

..of computer project results. This assures.that class time
is mostly related to the Math and studehts read the biolosical
.
content for themselves.

; -

Sections 4 and 5 are independent of-each other;

either can be done first. Sections 4.6 and 5.2 are hai-der
than the other three a0Olication examples in 4.1, 4.3, 5.1.
Section 4.3 -rnii

Suggested Uses:,

. .
kwide range Of basic linear algebra skflis can be tled'
'together by wprkinglhroughthis module.' Computer experience r
with a linear transformatIon is a key benefit of this
'module -- if at all possible;
:exercises .3 - 8, Section 2.6

inevitably arise'.

0 -0

9._

where a,b,c,d,e are all E(0,1), and the eigenvalues are zero
/twice), parameter a, real x>a, and a complex conjugate pair
u, u.
The characteristic equation that yieldS.these roots is:

Output Skills:6
1.
Calculate with linear difference equations on computer. /-"`
2.
identify overdetermined linear equations and decide when
,. they have a solution.
3.
Set tip and solve matrix equations from word problems.
4.
Sum finite geOMettic seria for matrix case.
5.
Despribe an application, hat uses linear equations to model
birth; aging and deathin a population.
Specifically, detail
use4mf a matrixAo transform that population through time.
6.Differentiate between matrix level and entry level calculatIonsand give example's where both are helpful.
7.
Explain context where polynomial functiOnslof. a matrix
8.

'

Prerequisite_ Skills:
1..
Matrix - multiplication; matrix-inverses; calculation of the

2.

'

I 'recommend use of some of

'15i

11

.62

ti

MODULES AND MONOGRAPHS IN UNDERGRADUATE

1:

INTRODUCTION
.

MATHEMATICS Aft ITS APPLICATIONS PROJECT (UMAP)

1.1
The goal of UMAP is to develop, through a community of users
and developers, ,a system of ihstructional modules in undergraduate
mathematics and its applications which may be used to supplemen't
existing courses and from which complete courses may eventually

What Harvest,Should You Take?

Imagi -ne yourself.as the.operator of a buffalo ranch.

YOu have a certain herd on,hand-, and each year you "har-

be built.

vest'Ya number of mature buffalo for their meat.

The Project is guided by a National Steering Committee of

mathematicians, scientists and educators. UMAP is funded.by a
grant from the.National Science Foundation to Education Development
Center, Inc., a publicly supported; nonprofit corporation-engaged
inveduqational research in the U.S.-and'abroad.

permit the remaining herd, for the next year, to replenish

itself through its own natural breeding.
The herd has
a certain known structure:
it is made up of Known- pro
portions of adult vs immature animals, of
vs.
./
males. Here are some questions you might ask while

PROJECT STAFF

.1

Ross L. Finney
Solomon Gardunkel

simultaneously trying to gain a goa frrveSt and maintain

.,the'herd for good future harvests:

Director
Associate.Director/Consortium
Coordinator
Associate Director for Administration
Coordinator for Materials Production
Project Secretary
Secretary/Financial Assistant

Felicia Weitzel
Barbara Kelczewski
Dianne Lally
. Paula M^,Santillo'
JacksAlexander
. Edwina Michener

You

What harvest policy will lead to'a herd next

year that has thesame size and structure as

this year's herd?

;Editorial Consultant

-1 What annual harvest will permit 'the herd to

Editorial Consultant

grow steadily.s0 thSt in ten ;rears it will

NATIONAL STEERING COMMITTEE

have doubled in size while keeping the same

,

W.T. Martin
Steven
Brams
,Llayron.Clarkson
James-D. Forman
Ernest J. Henley
Donald A. Larson
William F. Lucas
Frederick Hosteller
Walter E. Sears
George Sprtnger
Arnold A. Strassenburg
Alfred B. Willcox

MIT (Chairman)
New York Universiglt
Texas Southern University
Rochester Institute of Technology
University of Houston
SUNY at Buffalo
Cornett UniiterSityv
Harvard University
University of Michigan Press
Indiana University
SUNY at Stony Brook

proportional structure?
-- Do substantially different future heids result
if more, the time number, or less females are

'

harvested than males?

1.2

What Herd Should You Start With?

NeA,'imagine yourself as planning to enter the

'A.

buffalo-ranching industry.

You set goals (based on your

costs, capital, desired income, etc.) for-a-desired

Mathematical Association of America

'harvest.

The Projett would like to thank Edward L. Keiter for his

review and all others who assisted type production
of this unit.

This material was prepared with the support of Nationail

Science Foundation Grant No. SED76-19615.
RelEomendations
expressed are those of the authored do not.necessailli reflect
the views of the NSF nor of theNational Steering Committee.

.153.

That is, you select, as a basic parameter of

your business, a number of mature animals that you'

;(intend to harvest each year.

You might ask:

V
I

Although buffalo management is not a Major industry, this paper is

developed in terms of it because a widely available Computer program named BUFLO is based on the same model. The methods discussed
here are-the subject of research in human population dynamics; cattle, sheep, and other ranching industries; forest, fishing, and
wildlife management.
See the references.

154
A

Each 100 adult Cows will bear approximately 48 male

--' Mat initial'iize and structure of herd will

1.3

provide the desired harvest?

calves and 42 female calves each year in, late sprOV.

How sholad the quota be distributed among male

This 90% reproduction rate is almost unrelated:to the

and female animals to achieve a herd Hof smallest

number of adult males in the herd because male buffet()

size that will continue to yield the quota?

are polygaMdus.

Wildlife Management

preserVe4

'

Because of deaths at birth and sla

of the calves survive to become next year's year4ngs

ranch because. the buffalo herd lives among its natural'

and about 75% of the yearlidgs become adults.-,Once they

You have a limited amount

of land, and its vegetation must support the herd.

natural enemies as the wolf and coyote, only about 60%

Conditions here are quite unlike those on a

predators, such as theftolf.

Buffalo naturally suffer different death rates at

different ages.3

Finally, imagine yourself as the manager of a game

What

reach maturity, buffalo are quite safe frdm their,eneiei

until they Oeaken from illness, injury, or old age: 93%

quotas of male and female buffalo should you license

We will

of the adults survive from each year to the-next.

hunters to kill each year to maintain the herd at an

take these numbers to be the same for males apd females

appropriate size?

and the same year after year.

1.4

The Task'Ahead

4.

In this paper we will consider a mathematical model

based on linear algebra -- of a buffalo, herd.

2.2

Basic Model Equations

It is easy to organize this data into a mathe

It will be

model.

possible to answer the questions above using the model,

but the model is a much simplified version,of the situation in nature.

AM = number of adult males '-

We will consider'the underlying assump-

"

AF-= number of adultfemale;

tions of the model and their limitations to some extent.

ical

Let
a

YM,YF = humbers of male yearlings, female yearlings "

M,CF = numbers of male calves,-female

While we.will look at the model mostly as a management

tool, we will also be in a position (in the exercises in

or more specifically, let these _be. the numbers of

SectInn,2.6).-to study historical issues concerning the

.destruction of the vast U.S. buffAlo populaticuO.that

buffilo.at the' end of,iTlis year" jut after-the hlryest. :

Let 'AM" AF',.,., CF' be the comparable counts' for next

thrived on the Great Plains in the early 1800's.

year's herdt also.at the completion Af (next.year's)

harvest. - Let
2.

a9

THE MODEL

QM' = number of adult males harvested "next year"

2.1

Herd Components
We consider

QF' = numbas of adult females harvested "next year" t

Their Survival Rates

(It

is our pone); to harvest only adult buffalo.)

s of buffalo within the

six c

Then, the breeOng.process, followed by harvest, is

contained in these equations:

herdi'ecaves are in their first year of-life, Yearlings

in their second, and all older buffalo are adults. Each
age group is broken down in male and female categories.

This will be true in the wild. On a .ranch, survival to adulthood

J.-mid be more likely. Oata in-this paragraph appliesto wildlife.
See Sections ,2.5 and 3.1 The references offer similar data due
to Fuller.

2-

The mode] hat

taken' from computer program BUM. See the
references for fu/11 acknowledgement.

15

1. 5 e

AK'

.95 AM* .75 YM

AF'

.95 AF + .75 1F

OK'

QK'

(2)

YF' 00,60 CP

QF'

1111 al .60 CM

0
0
0

.
,

CM' 0..48 AF

CF' las

.42 AF

as an example of.

Let's read these equations In

harvest in jth year (14;t four entries are aliwaws

cause off

natural deaths, .95 AM represents tthe sera timans

year among this year's adult males and .73 '144 tips

zero).
rdeeott

With this notation .etablsbed, it is

tblee Tram--

her of this year's yearlings who survive 's thsozoree aakialtt

males. Thus .95 AM + .75 YM represents the =mil off aathaltt
males just before harvest next year. (-We 4ma4iree dal= tfiee

AF'

CM'

.[CF'''.

0
a

0
o

0.

0)

0)

.75
0

o
.79

_48
-.42

YF"

0
0
.6

0
.95

yiv

perhgps-on-e'da'the .41.....er-anarzwestt
total., AN', is correctly given in the
frilsieequatibamciff
(2).
The second eqUation treats the stbil-t

.
a

.6

m,
0

QT."

larly.

adult cows after harvest this year, the iteiblwwil'LL ggrom

by 48:. male calves and 42 female calves totiticelliarmnreext

.95

AM'

harvest takes place at one specific moment eea h yrear

The third and fourth .equations .Pay tthat WM off

this year's calves survive to become -next 7yeear' v)eetaclings. The last two equations say that, 14.1.4
MEatithfumed-r&d

time to re-

t.trite (1). as

ft

(3)

wit

M i's- the

60

6x

6 matrix just abore....

The

year-to-

year process is given more generally as

(44,

0.j

Q.

j = 0, r

3,,

year.

This: is the after harvest model because it learolres &era

2.3. The After-Harvest Model in Vector -an4 44411:71iiiCc

Notation

cnnunts

We may call,M\the transformation

Now label "this year" as "year

d so on. Define the vectors:

"neateneezrir... ams

0, N. herd structure after 'harvest .linmiteij .rtb ifeear

(Jo- 0, 1, 2, 3777)

..1-1x-edirmstr- 0
.

mudt, Cwi.r of our

riroxiel-,. fur- it transforms its input,,,

beta structure

just after harvest, ints the berd,

structtrze tkat birth

a.ggiat& and deatl w 11 .produce just

before
barwest nm the
N
-

1534low,ing. year.

In our earlier notation, the first pi' 'bhp

sional vectors are:

2:44 The Before-Harvest Model

The model just discussed is useful Rf we bare

a
herand. want to examine what next Feats

AK
AF

YK

taken just after the harkt.est is otarpletecl_

barireSt will

ggii.0m.a.

YF
CM

CF.

We must gather the harvest quotas 1=0

41:15CitrIXES, trail%
Put

as' a: new herd.

But Suppose we are trying to

stdec= this year's harvest so that next
year's bead
mar he studied. Then we want before-harwestAterdi

21M

cna art. to whiche can apply the

.

harvest.. They ieseawe

as notat:imr of .their own:

157
()E1 15

411

Hj = herd before harvest.in the jth year

f.,

Is the herd likely t

grow r shrink if you continue

j0, 1, 2, ...
this policy into th

Thus .6.

and the last paragraph of Section 2.3

2,

....
involving different

Clearly, Hj is diminished by Qj at harvest and the new

herd H. - Q. undergoes the breeding transformation. Thus
(5)

telly to show that,,over several years

Use equation (4) rep

2.

Thebefore-harveet-modelrelatesH.to Hsi.

Justify

your answer.

says that Hj.4.1 = MGj,j = 0, 1,

future, orlian't yOU ktell?

arvests, an initial herd G0 will 'transform

into

Hi+1 =

- Q.),

j = 0, 1, 2, 3,

t.
-fi'Mt0 -Qv

...

-62 .

2.5

M2Z0 - Q2 -

Survival Rates Would.Be Larger on a Ranch

M3b'0 - Q3 -

One more comment.

The birth and death rates were

given for a herd living in the wild, subject to its

Now, provide a b'ological meaning for each term in the equations.

natural predators. (The effects of man as a predator are

reflected in QM and QF, not the given percentages.) Most

For example, M

of our effort, however, will be with questions that

relate to ranching, where herds are fenced and natural
predators almost absent. We would expect much larger
fractions of ,each category to.survive the year. However,

correct this to account for the harvests.

MO,2 - mr11, etc.

' there are no accepted numbers to use in M, and, rather

than arbitrarily pick ,some, we will use the same matrix
114.for both wilderness and ranch applicatiofi's.

is the herd that resulti after two years if

no harvests are taken.

The other terms in the second equation

.

What does the MQ

1

term mean?

The remaining' problems if) this sec9,on call for the vse of

a computer.
3,

The

Write a computer program that will calculate next year's

results will be qualitatively the same for 'higher sur-

herd size from this,year's, using the after-harvbtt model.

vival rates (as the author has checked in some detail).

It should receive as inputs: (1) the initial herd structure

(2) the constant harvest; (3) the number of years the

2.6

Exercises and Computer Projects

herd is lo be studied.
1.

411 the.week before harvest last year your'ra

r.

herd with this struoture:

h had a buffalo

The program should loop tp calculate

the herd size year by yejr for the number of years requested.
It should print out the successive years and the herd structure
4

AM = 200

YM = 300

CM =,,520

AP= 1000

YF = 300

CF = 500

that would result for that year, using our model.

4.

The U.S. Buffalo Herd in 1830.

The authors of theBUFLO

Your harvesting policy each year Its_ to take 100 adult males

compdter,program (from which_cmeMOdel is taken; see the

and 200 adult females.

references) state that the total buffalo herd in the Unite?

Calculate the structure of the herd

a.

after last year's, harvest

States in 1830 consisted of 60 million animals distributed

b..

before this year's harvest

as follows:

c.

after this year's harvest

d.

before next year's harvest

e.

after next years harvest

15J

beware of this trap as %pm work your program: if you compute the
components,,of next year's herd in their usual order., a new value of
AF will be computed.before the old value can be used-to calculate
The old value of AF must be saved before it is
CM and CF.
replaced with the new one.

160

30% male adults

P.
.

27% female.adults

9% male yearlings
14% male calves

Repeat Exercise 7, but this time take constant annual

8.

8% female yearlings

harvests (in the catastrophic year and thelOthers) of

12% female calves

million or 4 million animals, splitting the harvest among

(These figures should be taken as good historic guesses;

males and females in the ways listed in Exercile 5.

Comment

estimates of the total herd size vary widely above and beloti

on the combined effects of catastrophe and harvest.

Which

60 million.)

harvests worsen the effects of the #atastrophe?

Let this data give your initial herd.

Take a

Which

overcome it?

constant 14rvest of 4 million animals annually for a period of

ten years. Distribute that harvest in various ways among

3.

males and females, trying to find a,harvest that leaves the

herd approximately unchanged after a years.

.

the harvest among

That is, split

le and female adults in a specific way

3.1
and trace the herd
for Exercise 3.
the same way.
loop.

The examples in Sections 4 and S will show that -.c

Then try other splittings of the harvest in

Several computer"runs can bebse

or you can

can really calculate with this model; it is a workable

ion into the

management tool.

A convenient way to get the number 60 mi

It does reflect the hasit processes

Of birth, aging, and death among buffalo. The equations

Start with you computer. program from 'Exercise 3 and the initial .-

herd given in ExerCise 4.

Take a 4 million animal harvest

in Sections 2, 4 and 5 all have teasonable biological

Orleconomic interpretations
The actual numbers used as birth7:hnd survival rates

annually for twenty years, using thgse strategies:

are reasonably, close to correct figures.

a.

hatvost 100% adult males

b,

harvest 75% adult males, 25% females.

c.

The Model's Basic Strengths

r ten years using your computer program

machine is 60.E6 in Fortrenkor Basic.

5.

ASSUMPTIONS, STRENGTHS, AND WEAKNESSES

OF THE MODEL

One,piece of
evidence fo vhis is that, among adult buffalo? a life

span of approximately 25 years was the rule

harvest 50% adult males, 50% females

d.

harvest 25% adult males, 75% females

e.

harvest 100% aduft females.

The results are strikingly different.

'

at the

time when great wild herds'roamed the plains.

Our model

predicts an A erage life span of 21.5 years (where we

count buffs o that!die between their 2nd and 3rd birthdays
as age PI, etc.)

Discuss the biological

reasons.

In Example 3, Section 4.6, we will show that no more

than about 14% of a herd may be harvested annually without
.

6.

7.

Repeat Exercise 5, takingra much larger harvest (say 12 million

animals) annually. Compare to other results you have.

evdntually depleting the herd'.- This vatue would vary in

Let's examine the effects of a natural catastrophe (flood,,

nature but the model is qualitatively.cvrect enough

range fire, etc.) onfa herd.

to convince me that a steady hdrvest of (say) 20% of

Take the initial herd from

Exercise 4 again; Pad set the constant annual harvest to

the herd would destroy the herd in time.

zero.

6 provide strong evidence of his.

Drastically.reduce the birth and survival

rates in
the matrix M and transform the herd forward 'for one ygar, to
simulate a catastrophe: Now put our usual nOmbers back in M

and trace the herd forward for nine more years, still-taking no

harvest-Othat are the long-term effects of the catastrophe?.

See E.D. Branch, The Hunting of the Buffalo, University of Nebraska

,Press, 1962, p. 11.
Branch's figure of 25 is presumably drawn
from journals of the 1800's and may well be high.

161

v,

Exercises S and

162

.)

3.2

Limitations' of the Model

Yet another weakness is thg/t no economics its inin;

The actual 'quotas harvested would.

cluded infihe model.

Whether one should blindly accept advice from the

model is anotherLtatter. The_ model is built on a number
of assumptions that do not correspond to nature.
The

surely be related to the price.of meat andk,the cost

The managft:Of a game,

preserve might not be troubled by such questions (if

most important of these is that the birth and survival

Lhis grazing lands are sufficient for the herd so that

rites used in M are assumed to be constant year after

year; this would not be true in nature.

of feeding the herdcon any ranch.

no feed is to be purchased).,

We can regard

there:is no single obvious'

the survival rates in M as averages for 'normal" years

wayto extend the model so tha;\lconomics is_effectively

that provide generally,favorable weather and feeding

conditions. Aknormalconditions like severe storms,
range fir
drought, floods, And disease might temporarily

included.

_cause muc

Our model

we have AsvmeA that all two-year-olds are full adults.

although they

And the.numberof calves born has been made a simple

fraction of tnenumber of'adult females. This is //

might not be rar,ein-4e.wild or on a'ranch.

..

hanisis of the model are not ideil.

wer birth and survival rates.

does not provide for such catastrophes,

The breeding m

'HI fact, _buffalo begin to reproduce at agesttwo onithree;

roughly true in a Volygamously mating herd if reasonable

The constant birth and survival rates ao not permit

the study of overcrowding or overpopulation... Instead,

numbers of adult bulls are in the herd.

the model assumes that unlimited land, food and water

a value AM = 0 would not interrulit)the mating process,

as it would in nature.
In fact, the actual' herd would .

.are available for.the herd,

In the wild, an overpopulated

be in danger of extinction if any of the six categories

herd would eat poorly anti its birth and survival rates

would decrease.

greW too small.

it would be more subject to disease."

model:

,-,Qn a welt-run riliFh`we' would not expect overpopulation.

We Will see in Sections 4 and 5 Oat the herd size

This can not be included.in a linear

In using the modal, one could.declare the herd

.

"extinct" if any category were to grow too small..

'

can be related to the harvest in ways that make overJpopulation manageable:

In.6ur model,

-Finally, we hive lumped'all adult buffalo into

In any case, the model is one of

two categories and declared them all equal in their

:UnliMited exponential growth for the herd, tempered by-

ability to survive and breed, ignoring the obvious

variations with

wthe harvesting process.

C

Another Weakness of the model is that harvesting-

Despite all these defects:, and othe

..is done only once'a year, rather than stead ly or several

:, times yearly. Ftealit; was different: Plat s Indian

simplificationef the herd.

Let'..s.put it tot use.

'4 4!

'tribes held lengthy summer and winter buffalo hunts.

White men slaughtered the buffalo continually in the. 11400's.
.

thet I've un-

doubtedly missed, the model as pregented offers a useful

4.

APPLICATIONS:

ESTABLISHING A HERD

On a ranch today, the:herd v4Oul4 be thinned as, meat

prices and the

availability of rangeland and water

A }lei:And Harvest That Continue Year After Year

4.1

dictate.

Example 1.

What -size and structure of herd do,

must we have (or put together) this 'year so that next

We have considered an obViot4 way to simulate a-catastrophe with
the-model in Exercises 7 'and 8, Section 2.6.

year we may take a pre ichosen harvest 1.1 and then have a
herd

10

163'
!-cW,".

such that e1
,

iF

164

11

A businessman planning to create a ranch mig t

to the'Aprobleril posed in Example 1 as

D-

He choos.eschis annual "product"

ask this 'question.

(9)

bti and wants to know what "capital'investment" d

,

he

.tb year'(
e

matrix level:,ele can write the silution in (9) ,without

_.

'110 with
t;

= d ) and yield product -0 l' Since we end

= G

the process of harvesting 14'1 and

Thainiaining a herd ofthi-Aaiiie-ilze and 4tructure'can

actually looking at any of the specific numerical entries

of M; we use M as a single itc4, not a collection of

continue year after year: Ke'call the heid.and hariest

36 mumbers,

vectors ateadktetate.1

of.M to establishthat (M-I)

0,

'f or the

because the year-long study-period for the herd progresses

from initial herd through the breeding process to the

4.2

Calculation of (M-I)
We would need (M-I)

-1

exists and to actually

-1

-1

to proceed 'further with (9),

so we have the opportunity to carry through an:unpleasant

.pre-set harvest at the end of the period.

Thus we know

Nowe'irex, we do have to use the entries

calculhte the solution in.(9): that work is at entry

level, not matrix level.

A'S'the chosen notation indiCateA-, it is natural,

to use the after- harvest -count

i(

Go = (M -I) -1 Q1

Notice that we've completely solved the problem at

...

'....

10.

should make so'that it -will "maintaTV itself from year

and want to solve' for do in

matrix pivbting Gaussian-elimination caac4lation by hand,

in detail.

G1 =GO
,i6)

[compare

(3i1

The reader who would benefit from such an example

is invited to follow along, eleCtronic calculator in

We can replace di with do in-the second equation of (6),

hand, verifying each step.

The reader who prefers to

see how the answer is used in the rest of this section ip
welcome to do so: skip to the paragraph containing equa-

getting,
4

t 4. Mt
0

(i
1

tion (10) at the end of this section.

and reArranp to read (I is the 6 x 6 identity matrix)

(7)

(M -I)t =
0

Recall thai, to find the inverse 7 , we list M-I

and adj in to it a six.-by-six identity' matrix to create

1 .

This is &set of six linear equation

d0;

-I I and

are known.
-.05

0 =

0
-:05
b

0,

.48

a 6 x 12 matrix:

for the unknowns

In fact we are asted t6.solve

0

.75

.75

-1

-1

0
0

0
0

.6
0

.75

-.05.

.75

-1

-1

,48.

0_

.42

0
0

0
.6

-1

-1

to

0
1

/0

.6

.6

-0 ,

,0

.0

O.

-1

0.

*1

0'

0.

Now we reduce the left side to a six-by-six

identity matrix using only elementary row operaMone:'
we may (1) multiply a row (all 12 columns) by a non-zero

There is a unique solutition-because M-I> iA nonWe will calculate (14-I) .1 in5ectioir1,2,

beloW.
In terms of it we can write our
,

singular.

-.05

'

1(
12.

'

7There are other', less efficient methods.

:

13

scalar, or (2) add a sealir multiple of one row to

1,

another row, or (3) interchange any two rows., To work

now: multiply the top two_rows by -20 each (to convert

the -:05's into l's for the 6 x 6 I).

check 4.
these
rows

. -15

0
0

-1

7.2

2.78

8.4

6.3

-20

-1

.6

You

-1

.6

matrix. The result is:

.48

.42

Al

-15

'

'*1

,o

-1

0,
0\

-20

0)
0

-15

0'

-20

-1

'0

.6

'o

0,

11/40-

-1

.6

0,

7.2 -1

6.3 ** 0

9.6
8.4

TOW 3 to kill the

new .row 3

to tow I:
check

0 4

'-20

"64

The first three

-0

=1

8.4

chehincy..,Multiply

-4.32

4.6

2.78

8.4

Firtally we multiply the 6th row by

sixth column to reach

31.080
7.1944
2.0720-

-20

4
4
4
4
4

0
0
I

1.8130

3.4533

23.310

-15

5.3958

-1

14.5540

-.6
0

35972
.2.5900

clear the

-1

13.986
3.2374

.93237
.21583

1.5539

0
0
3.0216
b
The matrix that has appeared.on.the right is (M-I)-1.
The first, third -and fifth ckinans are exact.,and,the
Others are correctly rounded tofive significant digits,
2.2662

is:

-20.

.
31.08

-15

13.99

23.31

abouiliO do in coludn 4 does

A.072

1.554

r.6

.0.

1.813

3.453

2.590

3.022

2:266

not.

to rishileairinu friendly'zeros
TOW 4 by *l to get a new row 4. Add
,

row4 to rows 2, 5,

-.

-5.396

zeroed.

-1

6.3

-9

4 is

-7.2

,277-- and

--1,

4.,

7.194

rbst of _column
4,
.

,appropriate moktiples of this new

,A.so"that. the

Pleaie

coluAs now Atch'a-6 x 6'1.

left

+,0

the inverse

-1

.disturb,-three.coiumifs: -14 gain this because

we-w*k,fros

-1

Keeping four significant digits, our final result for

notice that what. we-are

0'

6.3

-4.32 -.6

which is more than we can make'good use of below.

-5.76 -1

-.6

.9.6

-2.592

0 ,

.0

Os

7.2 -l

-1

-15

.6

0
t

-20

'0

:0

0
0

-9

-is

Cl

0'

-.6, .9

-9

-15, -64.8

-15

-20

-38.88 -20 -86.4

0
0

1OF

'Multiply row 3 by -1 and usetplat new

(in the 1,3 slot) Joy adding 15 of the

should be able to decide how we get to the next

0.

-0.

0
0

.0

-1

9.6

-15

-15

-20

0
0

'

-.6

.0

4.32

-20

-1

1.

check 4 0"
these. 4.

L%'

-9

-.6

1.

.1

rows

-15

-20

first

times row 2 from row '5, and .42 of row 2 from row 6.
The.se two elementary row operations give us

-these

0
gotten.* 0

The first column on the left is fine. Make the second.

coltIran fit the goal of a 6 -x 6 I by subtracting .48

13%.

new
row 4

Get

-9

0.

- .3597

0.
.-1
0

3.237'
.9324

.2158
1.554

.3597

and

Reach .

168

14

tiI

15

4.3 "A Steady Harvest Plus Controlled Growth of

the Herd

choose the herd -O0 so that, after growth and a harvest

next year, we will have a herd that is 12% larger.

Example 2.

Our ranch-planning businessman now wants

a herd that yields harvest. 4 next year (and every yearthereafter) while it grows by 40% during the first two
years. The larger herd is to have exactly the same,

is to have the same structure as

b.

harvest model.

).

Solve your equations from (a) at matrix level for "a-

harvest swill be taken. The resulting herd ti will be

allowed to grow another year, when a large( harvest 1.1i will

be taken.

Calculate t

After a year's growth and next year's

to

0'

as known and use the after-

harvest, initial herd

it

(i.e.`be 1.12

10. A buffalo herd G will be 'allowed to grow until next year, when

proportional structure as the original one.

Again we.regard

e0

so that this proces's leads to a final

resulting herd G2 such that G2 =

2

(which we will calculate)-will

become

0'

11. From this year's herd to a harvest (swill be taken

Jop

next year.

After another year's growth, a harvest 1.2 Q will be taken.

-a; = nto

The final resulting . herd G2 is to be 25% larger than GO

The next year's growth and eventual harvest yields

=14 1

-ft

(i.e. -4; = 1.25 to).

(same Q each year),

.a.

b.

= M(Me0- 4) - Q

Write equations for this situation comparable to (11a,b,c).

Solve for G. .
0

2-=

=M6O-MQ -Q

12. A herd GO

grows for five yeats with no harvest being taken.

In the flfth year,

and we want,40t growth (plus the harvests) after two years:

rvest ti is subtracted.

hslis

herd G5 is exactly ddu le e0'

(lic)

G2 = (1.4)c.

13: Find -a'

if, after 5 years during. which the same known harvest

0
Q is taken at the end of each year, the herd is
to double:

Frail (11 b,c) we conclude

M

2= .'

J.

- MQ

Z.

= 2 GO

0'

= 1.4 to

14. Find t if theherd is to'double in six years (46 = 2 GO).

and we rearrange this to

(12)

(M

Assume that the same, known "harvest Q is taken IP

after the second,

- 1.41) 60 = (M

fourth, and sixth years of growth.

known 6x6
a .known
matrix
unknown vector

4.5

a.

written at matrix level:

to =

1.4I)-1 (1+ 1) Q

4.4 Exercises
9.

a.

Write equations comparable to (6) or (lla,b,c) for'this

situation: Were given an annual harvest

Q.

Mathematical Insights

The example of Section 4:1 and 4.3 and the exercises

,of 4.4 should have provited you with experience that makes
these comments believable:

In (12) we have a set of 6 linear equations that have a

unique solution.
(We won't prove that M2 - 1.41 has an
inverse, but it's, true.)
Oueproblem has this solution,

(13)

The'resu-ltping

Find G0.

When calculating with matrices, we find that algebra

arises that is much like the algebra we learned long
Is
ago for numbers.
Most ofwhit we can do with numbers
is also correct for matrices. (Key exception:matrix
multiplication is not commutative.)
We can even sum
geometric series -- see Section 5.2 below.
It pays

We want to
17.

16

-169

(FOrr example, if weed up selecting a harvest a

We
= .75, q =
p
malli.es, and_ 25% females

to think of a matrix as the analog of a -si-7141t =vitraw.

b.

We may naturally need to calculite high _paws (41.i4ee

M10) of matrices.
very welcome.

Thus-, in (9), using

Ott

iihcatt

"elgenvalues and eigenVectors" you FM

technique.
c.

M2 m

ew scalars p q1 amxfl 'F. the

Expressions like M2 - 1.41, M +

t9

I ,

-+

148 -+ kM-7

and
.-+

They are polynomilat-s

-

3x -+

tD

We can think of

5 is a -pt ilyz

in x, i.e., they are sums of integer rata

Tf-

The herd -6-0 is now a multiple of the total harvest 1T_

Ai

In the same sense that 4x-

0)1'

0
0

in the (square) matrix M, enter our -work in a ruttartal

way and are worth study.

= (m

(see Section 5.2), called -palyncimild.1,5

M + I

tm

chaaw p- and q.

An easier, way to ibt) this Wainki htee

There is one: when you

T5%

COT,

0
0

equivalently, linear combinations of I -- 141.;14:M:1)P

qtc.

d.

All our calculations in the example Mere at nustfruoc,

thee "herd structure per animal harvesteE" ar thole minti-lberd1

meedeoLto produce one harveSted animal. because %diem mml-

..xs,

level and at that level we got a lol done. awm

xrr P3)

further progress with expressions like

tinned. by T, it

requires that we go to entry level (eouardual awTd1)..

Matrix algebra is a powerful too] ,

becomes.-the'

total herd Ga.

,---

The her.4gi'ae (the total number tuff animals im the

but -by .xled.kinT

with the matrix as a whole we are out xff.-tauthm..tit

herd),flai- a herd GO will' be

the individual entries, and their intcrimatixrnirmay

be critical:
4 -.6

FY)

(ro,r,t01.111-em-n1-1

(06) (1.1.1.1,1,1). to

An Efficiently Small 1-ierd

tD
tD

For any specified harvest Aucm*s$.:QM

Example 3.

a)

and QF", we have found an appropriate -s-teady-iiStxttehharld

(which will yield those quotas) in ExamOle 11.

addSs up; the entries in to.

perhaps our real goal is 'simply 'to 'harvest 1'

with T = QM + QF.

Naturally, we wish to jdo tttriz %wait

the smallest pnssible herd -(which would 7re:ciutre tilve littaa5St

land, feed, fencing, handling by employees, -papenwrkk,

etc.)

becausmultiplying by this vector

adt

4.

C111111lb

Since thi

we. simplify by studying

HS = "herd size per

= "herd size"/T

(1,.t,r,t,T,004-1311

We set _up the algebra in this .way: 'QM A..t.01bbe ssorre

fraction of T, say QM = pT where 0 cp <1.

QF = cT with 0.< q < 1.

Since T = QM -+

-23_:hnibadli);,

CIF, A? -*.:Aq .=- 11.

-UR

171

172

'

arrima311 Hanv.esttedl"

Is there some way to split up T imo XM aanki qt*F

so that the herd is smallest?

is a maniple aff

-20 p + 31.08q > 0

Again: our goal is to select p and q to make HS as

small as possible.

44.

P < 32008 q = 1.554 q

To this point, we have dealt at matrix level,

aside from setting up

- q < 1.554 q'

with scalars p, q, and T.

<4.

From here we must work at entry level, calculating the

individual equations.

Weylug in (M-0-1 from (10),

HS

and q =1 - p, the herd size may be taken close to minimal.

p + 31.08q
7.194q
`2.072q
1.813q
3.453q
3.022q

(1,1,1,1,1,1)

In Table I, various values of p and q are used.

The resulting values of HS and the resulting herds are

-20 p + 48.63q.

shown.

The percentage breakdown of the herd into its

six components is given (or equivalently, an actual breakdown for a herd of 100 animals is given). Retail that

Here we have rounded to two \ecimal places.

The goal was to select p and q such that

0 < p <

easy:

p < .60846.

ConclUsions: by takirig p < .60846 bust close to that value,

Section 4:2, and calculate

(17)

q > 1/2.554 = .39154

01;< q < 1, p +q = 1; and HS is minimal.

That's
as p increases, q must decrease and HS grows

steadily smaller; thus, p,= 1, q = 0 is the "right

ansier," and. the correct herd size per animal harvested.
is HS = -20! Clearly'nonsense!

NS is the size of the mini-herd that yields one animal

for harvest; thus\l/liS is the fraction of the initial
herd G0 (investment) that is harvested after a year.
Example: in the first column, each 6.90 animals breed
to become 7.90 animals and yield'a 176.90 or 14.5%
"output." These figures are given as "% harvest."
Table 1.
Structures of Seven Herds
of Various Efficiencies

We have ignored two biological restraints that will

correct this nonsense. -First, the herd size per animal
-

harvested,put be positive: HS>0.

condition oR p,q:

This imposes another

p

HS,- -20p + 48.63 q > 0

4. p< 48.63

Since p+ q = 1 we have 1
.2914 and p < .1086:

HS

% harvest

2.4313 4:

AM
-

q < 2.4315 q<->q "> 1/3.4315 =

itus our nonsense value p =

is ruled out.
'

4. S'condly, all'six components. of the mini-herd that

produces one animal for harvest [see (15)) must be

positive. Once we-plug in p sand q, these components are
'given by the column vector shO'wn in (L7).
(Trace the
calculations until yoti.see this) Adl six will )be positive if we insist that

./7

20

t.

173
*e.

AF
YM
YF
CM
CF

CD

C)

C)

CD

C)

C)

CD

.608
.392

.606
.394

.605
.395

.600
.400

.580
.420

.550
.450

.500
.500

6.90
14.5%
. 34%

40.9
11.8
10.3
19.6
17.2

7.04
14.2%

7.11
.14,1%

7.45
13.4%

8.82
11.3%,

10.88
.9.2%

14.3

1.8%

2.5%
39.9

518%
38.6

16.5%
34.2

27.4 %

38.7%

29.7

11.5

11.1

9.9'

10.1
19.3

10.1

8.6

16.9

164

9.7
18.5
16.2

25.1
7.2
6.3

40.3
11.6

19.2

16.4
14.4

8.6
7.5
14.3'
12.5

7.0%

12.1
10.5-

'

Here p = fraction Of adults that are males; q = 1

-'p = fraction
of adults that are females.
Herds of smaller size HS (animals
per animal harvested) result as p is taken tloser to .60846,
'which it cannot equal or exceed.
_

The structure and size of a herd that will yield a

harvest of T animals'annually,vdries
considerably as me

17j

21

go?

herd, i.e.

apportion the harvest differently among adult male and

female animals.

(The Process can then go on for

'many years, yier4ing steady -state harvests and herds.)

In a polygamous herd, there., is no need

This queStion arises before we harvest, of course;

to have anywhere near one bull per cow to achieve the

birth rates for calves we have assumed. In this regard,

thus,we use the haunt-before-harvest model. Then wemust

it is common in cattle ranching to run 1 bull with20-30

solve
9

cows. The first-three herds in the table above have cowto-bull ratios of 120 (= 40.9/.34), 22,-and 16; the
ether herds have much lower ratios. Thus herd O2
for

0'

1 = 1
Exercises

HO

M(H0 -.4'0)

to be practical and is fairly close to minimal size.

"-4.1

H1

when g

is known.

and g = g 0 = g

(compare (5)]

Simplify the notation to

and use algebra to reach

15.

a.

b.

16.

.HQ =

(ti - I)

H.

In Example 3, show that a 25:1 ratio of adult cows to

'(18)

bulls arises when p = .60624 is used.

(Here I is the 6 x 6 4dentitymatrix.)40 The,"obvious"

next step is to multiply through by !4-1 and-get the
Unfortunately, M-1 does
= M-1(M -*I)g.
"right' answer"

What value of p leads to a 30:1 ratio

Check our, work in Example 3 as follows: take a herd of one

million animals structured like.herd(pin Table 1.

are 18,000 adult males, etc.)

not exist!

(Thus thbre

So far we have woted at matrix level, i.e., we have

Use the after-harvest model as

used matrix algebra to calculate with the maqices as a

programmed in Exditise 3, and take a 14.2t harvest, using the

whole, not their individual entries,

values of p and q given in the table for herd 2 to calculate

the constant annual harvest.
initial herd for 20 years.

On the computer, trace this

dividual equations that make up the matrix level full .

It should remain roughly constant

system.

in size and structure.

Let's examine (18) in detail.

S.

To make:more Pro-

gress we must go down to entry level and look at the.in-

APPLICATIONS: CALCULATING THE HARVEST

We appear to have six

linear equations for the six unknowns in0..

-side is known.)

Now.

(The right

Howevei, four of the, sientries in

We will now ask what harvest should be taken from a

herd already in our possession, if it is toe preserved
in size for the future. We also will discuss harvests

were set as zerolrom the beginning. (We harvest only

adult buffalo.) Thus, in (18) we have six equations in
The equations are overdetermined.
two unknowns, QM and OF,

that provi4 for controlled groWih of the herd., This is

in contrast to Section 4, where we "designed" herds to

Usually., two conditions (equati-ons) suffice to determine

provide specified harvests.

two -unknowns.

Entirely different difficul-

ties will appear.L.

S.1

for QM and QF, will we have any solutions 4 all.

cf

When are we lucky?

Steady Annual Harvests and Herd

Example 4. Given "this year's" herd n, what harvest,

Qb should be taken from it so that next year's herd H 1
till have the same size and structure as this year's

.22

1Th

Only if we are lucky, by having the ext

four conditions here add no contradictory requirements

The six equations say inaetarlf

8
,

Up to now we hav,e used AM, AF, etc., as components of the hi,rd,

after harvest, QM and QF as the number of buffalo just, harvested.
In Section 5 these variables are components of the herd Wore "
'harvest and quotas of buffalo about to be harvested.
3.

t'

fn.

23

+ :75 1

f95 Q14

"a'

.95 QF = -.05-AF + .75 YF

{Alit- .95 (AM;QM) + .75 tym}

=
(AF-QF) + .75 YF

5.2 'Constant Harvests From aGrowing Herd

Example 5.

Cm}

(19b)

0.- -YF + .6
14! QF

qt.:t2' AF
'8

{.42

{YF =
CM

cF)

{CF
CF

We want'to'Setect a harvest 4 so bat,

taking the same harvest every year, the herd will double
in ten years while retaining the same proportional, strut;"

CM1

.6 CFI'.
:48. (AF-QF) )

.cure.

= .42*(AF-QF)

Now, the' valuei. of AM, AF/YM, YF,' CM,'

assumed

to be known, so we could *Solve for our` unknowns,

QF, using equations (19a) alone.

That is, if do is our initial herd before harvest

tlfis year, then at the end of ten years we want.to have

2d a* the herd structure.,
0

We use the before-harvest-count because, again, that

is when the,question of selecting a quota arises'. Let

Then equation's (19'b

lead to a'contradiction unlOSs'the values of AM, AF,'

YM, YF, CM, CF, QM and QF'already.known happen to satis-

Hj be the herd before harvest ih,the,jth Yea, .j ..0;1,'A

fy (19b;c): 'Any herd fore which these font equations

2,

Then

..., 10.

(191),C) are ITot satisfted cdnnot.'duplicate it*elf from

thii'year to'next no mattvr4lodtlzarvest'is taken:

(11ecAl that -we are requiting tif'= IL*, pith the strict:

C M ("To 2,

-7

'2

M (go

-0

This makes sense if we read eqdaticins

Consider (19b): to have lc,. docihis-year'i

yearlings (which, if they survive, are adults in Hi) must

ma2
!

M2-Qs

MQ

=t1 H

be-exactly replaCed id'H1 by the suiVivors of this year's

calves. Equations (19b) say that YM and CM;.YF and CF
in our lierd H = 110 = di:Aust be in the natural balance
of six yearlings per ten calves for each sex so that,
thesurvival rate of .6 will cause this year's calves to
exactly. replace theyearIingi3opulation asithe year

211

=H

...*

/*

10

M1GU-

0.

10

=, M

(I + m + m

- mit

,Equations (19c) simply *ay that theSe b4tths, forming,,,

the calf populations of al, MIlst exactly'replace CM and
CF in 1r

Thus, the fclir extra copditioni in lie overdetermined

system (19) simply require-that the/herd have a natural ag
' balance so:that, 'considering the survival rates, it will

riplenigh itself despite the harvest.

24

.-

Pig t

T-herefoi-e,

write it as\the set., of linear equations

-Now interpret (19c): This,years caliles must also be

..'i,1-pre,-eriselyreplacedbyrwribotncalvesiffti=kis to `be
.
true. After the,hafvest, there will be AF - QF adult
females and*Oey 41ill give-birth to .48(AF,--QF) new'calf
ialesand .42(AF - QF) new calf females by next. year.

In this equation, we- know do and want 4..

passes.

= M211

logically.

- MCI
1

- MO

.11

m athematical meaning pf equality for vectors.)-

. (20)

(I +
,

m + m2 +

,,

... +
,

M9)M

knowrr 6x6,
6x6 'matrix
'matrix

All

=,,

1.1

(il. - 2I)Iro
all.known

,'' :

unknown

'

of this has\been at matrix level.

in that spirit.

We push ahead.
9

Have you noticed t$t I t M + M2 +

+ M .loolc
geometric, series? When numbers are involved, we
know how to add up such expressiohs:
+

+a +

1+ a

cc

n -1

= T---a

if

a ti 1.

Can we do something similar here, when M and I are square

matrices?

Of course, we can approximately double the,herd, and

(22) will help us see-how.

'

Indeed we can

M + M2+

Put S =

Weehave_examined whether we

can precisely double it..

Thus S is a 6-x 6 matrix, and MS makes sense: MS =

10
M + m +
M
Subtraction leads to the familiar
Massive cancellation:

Exercises

5.3

Is the initial hprd given in Exercise 1 a "natural" one

a.

whichif a proper harvest QM and QF were taken, could

M10.

(I - M)S = S'- MS = I

17.

exactly reproduce itself next year?

In fact, (I

M)-1 does exist for our 6 x 6 matrix M.

Explain your answer.

b.

tlepeat a. for the initial hccl Of Exercise 4.

a.

Show that M

We.Calculated (M -WI in Section 4.2; of course

18.

,(21)

If we replace the OA entries in M with arbitrary numbers

b.

striking.' It might tempt is t

a,b;c,d,e:f,g,h, we get

elieve the infinite

a
0
0
0

0
c
0
0

geometr Y series'formula;
Al

= (I - M)

-1

analogous to. " T-7R

Indeed, thisfornwla'is_yalid for certain families of

LO

matr&ces- Wand infinite series.of'.maerices is a fascinat-

Stow that (M

ing subject mits own xight: We will not explore in

that direction now butope.thing is clear: a sensible
.1 def. r

on

011,

0
0
0
0

0
0

0
0

0.

e
0
0
0

'f

0
0

doo not exist, eitheer.

Thus the over-

about determinants will hive an advantage in this problem.

19.

Reviie Example 5 ,p that the herd.will grow by,50% in

a,

po-LI -'141)-mt- 041-- 201.0.

rough by I - M, and by

Mons analogous to

410--1
-m
)

(1

M)

Check our geometric series result in (21) by carefully

20.

(M10

21)46,

multiplying outrhe left side of-(I41)(1+M+M2+...4M9) = I-M10,

to gtt theright side. (Why doelthWconfirm equation (21)?)

Ilkt is as far as we can go at matrix level in this

exa,mple becaused4-1

does nit exist.

identirOC;11 the algebraic properties of matrix multiplication

The right side of

is known (although unp'l'easant to 5p6Urate).

ancrOditiontha you use,- such as the associati-velaWof

The

-miiitiplicatio7 n; left dIstribut,iye law, etc.

.systempie overdetermiced. Sdlne.ht9wanbe doubled in

ten years in-the way we suggested, but most'cannot

"41

4.79.)

Repeat a, with 50% growth over eight years.

b.

whichcloes exist (proof omitted):'

(1.-:100)7i

=X1.5 H and carry through

That Is, set H
0
JO
the algebt% of Example 5 for this hew cases .Reach equaten years.

(20) foil. We have made:progress: using (21) in (20)

we obtain:
,

V (22)

We,:.were interested in solving the linear equations

(22)

0
0
0
0
0

specific birth and survival rates. (The reader who knows,

o r first task:

We can multiply

determineenaturefofExamples 4 and 5 does not depend on

of ':convergence" for such series would be

.

In how many ways can you

The analogy to the numerical g ometric series formula'is

i+m+m-1

does not exist.

dd this?

S =

-1

26.
"'
C...

It

180

27

. .
a

6 a
o

REFERENCES-

I first met this model when Karl Zinn of the Center

ture.

.
.

kit models for populations with age struc-

The original paper's are!

Leslie, P.H., "The uses of matrices in certain

population mathema0.cs,""Bibmetrika '33 (1945),
pp. 183-212.

Maynard, Massachusetes 01745,

The program and-its .documentatidn are part of project EXTEND'and the Huntington
Two Computer Project.
Program.BUFLO Interactively

'

Lelie, P.H., "Some further,,notes on %be use of

matrices in population mathematics,"
Biometrika 35 (1948), pp... 213-245.

permits one to follow a buffalo herd throUgh many years

while applying a va.14,e.ty of management policies. 4

Much research-by Lelpfie and'others has followed, with the

goal of overcoming the limitations of Leslie'soriginal

Whiletquations'(2) acre ttkeil directly .from

_I am solely "responsible for the mathetatics that follows
in this paper.

CF = 2.0

important L

Polytechnic Institute of Brooklyn and distributed nationallY by the Program Library, Digital Equipment Corporation,

CM = 2.0

YF = 1.2

Our mo el is a simplified variant of the more

of Michigan introduced me to a computer program named

DUFLO; written, by L. Bran nd R. L. Siegel of the

qt

YM = L2

for Research on Learning and Teaching at the University

AM = 16.8 mi.11;on.
AF = 16.8

models.
*.

These 1.pertations areluch the same as the

Ar

ones we have dWussed,fdt outsimpleemodel: 'use of

constant coefficients from year to year and lineirity of

An alternative discussion of exactly the same model

with different survival rates based on an actual 'modern

,,.

the, model.

In addition, the Leslie approach has-been

applied to much more than buffalo herds. The interested

buffalo herd may be found in:

reader might start with:

.

Watt, Kenneth E. F., Ecology and Resource',

Management, McGraw Hill, 1968, p. 358 fft
This is an excellent betook for all readers
in applications of undergraduate-level math
to biology.

Pielou, E.C., An Introduction to Mathematical

4 Ec010

Wiley- Interscience ,-- New York,

Chapter III covers the Leslie model.

Pielou is a leading mathematical biologist; ...
her ooks are among the basic advanced
work in the field.
l
Usher, M.B., itA matrix approach to the management
.

The buffalo model discussed there is drawn from:,

Fuller, W: A., "Biology and Management of .the Bison

bf Wood Iuffalo National Park," Canadian
De artment of Northern Affairs Natural Resources
Wildlife'Management Bulletin, Series 1,
No. 16, 1962.
0
.

9f renewable resourceswith specialreference

to selection forests," Journal of Applied
Ecology 3 (1966); pp. 355-3q7.

As I read Watt, the survival coefficients matrix

used.
by Fuller and Watt is
..

.9

0'
.

40,

.75.

'0

.75

o
0

0
0

.4,

0
0

,36

0,

.1r 1 0

O
0
0

Usher:19.:B.,"A matrix model for forest managment,'"

Biometrics 25 (1969), pp. 309-315.

.9

0
0
0

-0
0

Fowler, Charles-W. and Smith, Tim, "A matrix.

method for determining stable densities and age
'distributions and its application to African
elephant populations.". University of Washington
Quantitative SCience Paper N. 31, Seattle,.
January 1972. (Write Fowler or SMith at
U. Washington, Seattle, 9819S for,more
information.)

.4

0
0

Usher, M%B.,"A matrix approach to the management

of renewable resources', with special reference
to selection forests -- two extensions,"
Journal of Applied Ecology 6'(19.69), pp. 347-8.

or'

and their "guqssiimated" 1,830 herd (4.40 million

buffalo
isstructured as:

29

181
P.

A Kell- written disCussion of the Leslie mod 1-with-b.

sheep ranching) is'

Anton, Howard, and Chris Rorres, Applications

of Linear Algebra, John Wiley & Sons, 1977-,
Chapters 9 and 10.

= Kt = (320, 985:"312, 300, 384 :336)

0

c.

ii ; = Hl --'4= (220, 785, 312-, 300, 384, 336)

d.

it

= mit1 '= (443, 971, 230, 202, 377, 330)

.

(Decimal results have been rounded.)

e.

6
ACKNOWLET)GEMENTS

7.

HI

foi--

application to harvesting of herds (including d

F.

G2 ="it.
H2 2

-4H. (343, 771, 23011E2, 377, 330)

The herd is shrinking slowly in

adult femalet.

I want to thank a number of people who have con-

key category of

This will continue for ;while, causing

the whoit herd to shrink slowly;

....

tributed as.this paper has evolved into this third,

edition.

Thanks to:

Ovel- two years G

2.

if left unharvested, would become M G

0'

daq.

Tarl Zinn of the Center for Researchon Learning

Q1, but the defy

and Teaching., Universitr. of Michigan;

of the tVo year period, MITI.

'or

ntroducing me to the model and encouraging_

The

0'

harvest Q, is subtracted, of course.

We,also subtract, not

40 of the harvested sub-herd Q1 at the end

The linearity of the model assures

that these sub-herds can all be superimposed.

A

I.

me to develop its mathematical content.

3.

A FORTRAN, program is-listed in Table 2, pages 34 and 35.

4.

This may have been a frustrating problem -- it has no solution.

'Tom Hern and Fred Rickey of Bowling Green State

University, and David Staley of Ohio Wesleyan

The herd is inherently unstable because, in 1830, it was

Univergity for class-testing earlier editions

growing exponentially '(or would have beers, had not white man

in their linear algeba courses.

Bill'Cannon of the U.S. Department Of
culture Laboratory, Delaware, Ohio, for putting

interfered).

A harvest of 1.4 million males, 2.6 minim!, fe-

"

males will convert the initial herd Of 60 million into a herd..

of 59.984 million in ten years, but the herd stfuctureis

-ne in contact with much literature in this

drastically changed.

fieleand critically reading the fiist edition.-

The new herd.has many,fewer calves than

the-original, and theherd is in fact headed for extinction.

Edwgrd Kelly, California StatgiUniversiiy

Other harvests, of 4 million lead to herds that grow rapidly

at HayWard, for a careful review of th9. second

or decline.rapidly, but this herd is inherently unstable.

edition.
--

Sol Garfunkel foi administrative and editorial

work at ,Proj ect

a.

re dispflyed in Tables 3 and 4 (pp.36-42).

their potential progeny.

The effect of harvesting a lot of

Also, all of the herds that

involve 20% harvest (Exercise-6) meet a fast extinction.

We are given

=.(200, 1000,300, 300, 520.1500) and Q. = (100, 200,

,r

.1

I'll write vectors horizontally to save space.

0;0, 0, 0).

mow

The poini'is that, by slaughtering females we.also slaughter

females' is to destroy the herd.

.0*

OP

int.

Computer printouts

5,6.
j

ANSWERS TO EXERCrSES

8.

1.

And that's the whOl

7.

4111*

dne examplelis shdWn in Table.5 (Page 43)i With commentary.

You should try others.

Zo = Ho - (I"

9 :'

= (100, 800, 300, 300,320: 500)

0

a.

G1 = MGO
= 1.12

'30.

eOr'

-184

'
a.

b.

(m - 1.12/)-1:1/

1.44

Mkt

nu:at-ions

10. Equations

G1 I. M GO

- G2 = MG) - 1.1t

sG2 sG0

,,

lead to solution
s

to 44(M

.,

...

- /)"4 4

(M

-4-

e-4- - m

Equations ti = Mt0 -

,2 = MG' G

...?

tEt,

mt5

is

G.

condense to

lead to solution

i ....

'Go = 142 - 1.2.5I)1 (M -+ 1./1.)11

6.2.

_.s.

2Sc 7 G.& = 14

'

ea

,...,..

(1+ # Pit * MA-%

i

The mistrials- i s

ta'7 = (het - ZI)4(14.4 -14 fiX1.4-Z)cl..

(harvest is-il=33)

= P1G0

.4-

t'

1.2Q

Equations

6-

Gk. = 1.25 G0

12.

c.- t

t = Me,.

1.)-1 )32.

s
11.

11-2

..

t2

.606E4

155

171;

166

Ai. computer printou.t appears as rahlle & (hEri3P WO- 111z-sat's

ace right on- target..

=

4..

65

mt4

GS =

2t0

at:

13.. EquationsG1t

le

ms2

G3

Na; again-, equatians (.1'31:)! are not smtilsifiledi tby a3 Rena

1%. as.

Change the equations to

It19-'
=-tPli- - Met +Mi+

-1

ip-,

TM

... 1.5 ir
CD

s-

* mhti
..
MX s....

'

The& gz) is replaced by the cateadettamtEcerfi syistenn

/t

.04

= 313)1.

with- 12%. fettstIe calves and itt Fest:mile waacrilErsg3s..

= (M5 7::

= 520

are natt satisfied by

have solution
0

to: giVe lust one reasons:mans' rmerty mutations OW)

17;.

(final 'harvest)

latZ w

MG.3

tr.

n.crITP a

Air

NC t&

K(!tic - 11:51)A9-. 9

4- mi 4- pit

condense to
4

**, Tus;
h

At
0

T-511i;

(0

Itacktcx this rep/at:went frnr (g22))::

015- an-10 4;45 44.41i24*'

32

1
J

1.

186_

$3

IP

T TABLE OF PERCENTS
C PRINWRITE
104
FORMAT( -IHOrIBAr'PERCENTAGE DISTRIBUTION OF HERD')
104
WRITEIS.1021
LL.LONG*1

TABLE 2

(5,1

A listing of my FORTRAN program, used to create all the

It does more than Problem 6
printouts that follow, is given below.
asks, because it gives the results,irOpercentages and in actual
The program wasrun on an IBM1110 computer
millions of buffalo.
but should easily adapt to any standard FORTRAN,.

00 30 1(.1.LL

NyEAR. K-1

WRITE15.101
105
30

106

NYEAREISAMK,J10.101

211.12.5x,f5.1,6X,61F4.1,6X11
CONTINUE
Q11110(21
WRITE15,106)
FORNATITHO.,CONSTANT ANNUAL HARVEST IS 'F6.2' MALES,
FORMATI

' F6.20 FENA4

IES (To MIL1 LIONS)01

GO
ENO

#
e

THIS PROGRAM ACCOMPANIES THE 'APPLICATION PAPER

CC

C
C
-c

'

C
C
C
C

C
C
C

C
C
C
C
C

The data cards that produce the printout of Table 3, page 36,

'MANAGEMENT OF A BUFFALO HERD'

ANO ODES THE CALCULATIONS REQUESTED IN PROBLEM 3 OF THAT PAPER.
IT RECEIVES PAIRS OF OATA CAROS AS INPUT. THE FIRST CARD pima()
SIx CATEGORIES. IN
LIST THE SIZE OF THE INITIAL HERD IN ENE USLAL
6 F 10.1. FORMAT.'IN MILLIONS THIS TELLS THE PROGRAM THE INITIAL
DATA CARD LISTS THE CONSTANT
NUMBER OF ANrAF0NOF,CN,CF. THE ,SECOND
HAirvE5T oF MALES, INfN FINALES, 'IN 2 F r0.6,FORMAT (GIVE THESE IN
NILLTONS, ,TOO) AND THE NUMBER OF NEARS THAT THE HERD LS TO BE
TRACED, IN 12 FORMAT IN MAIMS 2'1,22.

are these, given as samples.

precedethe fake pair.

'PLACE PAIRS OF DATA CARDS BEHIND ONE 'ANOTHER. PROGRAM TERMINATES

WHEN A FAKE OATA-CARO-PAIR IS FOUND WITH A NEGATIVE ENTRY IN THE AM SPOT.
THUS MANy'HERDS_ MAY BE STUDIED WITH ONE COMPLTER RUN.

100

102

FORMAT

TOTAL

lx,0yEAR

NvEAR0
TOTAL

4.

O.

20

4.8

8.4

7.2

(initial herd)

ca

(harvest of males, females; years traced) A

(fake data-card-pair to terminate program)

Zg

blank card

Li

IH1.24x.TILLION.S OF BUFFALO')
FORMAT.(
-'
NRITE 15,102)
AM
IF

5.4

wRITE(5,,1011
101

16.2

-100.,

l
DIMENSION re(6$03161,SAvE(50.71,
REA0(2.100) H.0111.0121,LONG.
6F10.4/2F10.4,121
FORMAT(
CALL EXIT
IFIHIII.LT.D)

18.

OUTPUT IS GIVEN IN MILLIONS Of ANIMALS AND ALSO IN

AREAKOONN OF THE HERO. YEAR -or YEAR:
1

Many pairs of data cards can

AF

CE,I

.Cm

t
-,

'

'

MIS LIr6

All of (his
code simply sets
up the initial
herd properly
Note

TOTAL.TOTAL.HILI
NYEAR.TOTAL,H*
wRITE(S,1031
211.12,4A,F7.3.613A,F7.311
103 FORMAT)
CONVERT TO PERCENTS AND S AVE 'FOR LATER PRINTING.
SAvEII.11eQ.
600r
DO ID 1.2.7
LLL -I
5

'

$AvEt1.0HaL)/ToTAL100

10

8"

SAvE(1;IVX.SAVETTTI)Y,ttstetilL-1-

MAIN LOOP
,00 2S R.I.LON9

NYEARK

TENPOH121

TEMPO is used to avoid a key

Note
fail to save .
if
trap tn. the program
will not tave*
the old value of M(2).
, It to use in the correct calculation

.754(3) - 0(1)
.9500 f 1
H(11
.75.'4141 - 0(21
.gs.H(2)
H(2)
.614(51
H(31
i
11(41.#0.6eH161
.411.T6mPO
NISI
.42TENP04.. 0
H(61
TOTAL .0
00 IS 181,6
TOT AkeTOTAL *NIL I "
WRIT (5,1031 NYEAR.TWALOI
,

15

%AR.'

of m(S) and n(6)

..

sAve(oc,i),..p

00 20 L.2.7

fg

Ikliki.k1 H(LL1/TO TAA.100

) ..sAvEIKK.1).sAvfm,i,
_ .

ct4T(.4 t 1

...continued Rest *se

18,

lk

188

34

35
4%.

TABLE 3 (Continued)

TABLE 3.

Twentyuyear.printouts for the five cases

called for in Exercise 5 follow.

Case b)

MILLIONS OF BUFFALO

Case a)

YEAR
0
1

.2

.63.191

AM
18.000
17.150
16.072
14.768
14.131

67.453
72.276
78.165 14.021
85.242
14.348
31.417
93.521. 15.183
34.704
.103.11
16.567
38.344
114.141 18.524
42.365,
.126.736 21.094
46-.806
141:039 .24.322
1.713
157.210 28.257
.134
-175426 3X.854 /63.124'
195.884 38.477
69.742
218.803 '44,894'
77.053
244.425 52:284
85.431
4273.018 60.734
94.056
304.879 70.341
103.916
40.338 81,,21'^1,)47810
3
.759 93.468
126.846

4
5

6
7

8
9

10
11

12
13

,14
15
16

17
18
19

. 20

YEAR
0
1

TOTAL
100.0
100.0
100.0
100.0
100.0

4
5

'106.0
100.0
100.0'
100.0
100.0
100.0
99.9
100.0

6
7
'

TOTAL
50.999
60.079

8
9
10
11

12
13
14

.100.0
100.0
15' 100.0.
16
100.0
17
99.9
18
100.0
19
100.0
20
100.0
.

YEAR

MILLIONS OF BUFFALO
AF
YM
YF
16.200
5.400
4.800
18.990
5.040
4.320
21.280
4.665
4.082
23.278.
5.469
4.785
25.703
6.128
5.362
28.440
6704
5.866

CM

8.400
7.775
9.115
10.214
11.173
12.337

7.402
8.190
9.048

6.477
13.61
7.166
15.081R
7.917
16.658
8.745
18.405
9.9.95
'11.043
9.662
20:335
12.201
10.676
22.467
13.480 11:795
24.822
14%893 1`3.01
27.424
016.454
14.397
30.299
18.179
15.907' '33.476
20.085
17.575
36.985
22,191
19.417
40.863
24.51
21.453
45.146
27.088 23.702
49.879
29.927% 26.186
55.109
,

PERCENTAGE DISTRIBUTION OF HERD

AM.
AF
YM
YF
30.0
'27.0
9.0
8.0
28.5
31.6
8.3
7.1
25.4
(3316
7.3
6.421.8
1343.
8.1
7.0
19.5
35.5
8.4
'7.4
17.9
36.3
8.5
"7.5
16.8
36.8
8.6
7.5
16.2
37.1
8.7
7,6. ,/
46.0
37.1
8.7.
7.6
16-2
37.1'
4.7
7.6
16.6
36.9
8.7
7.6
17.2
36.6
8.6
7.5
17.9
. 36.3
8.5
7.5
18.7
35.9
8.4
19.6
35.6
,
8.4
7.3
20.5
35.28.3'
7.2
21.5
34.8
'8.2
7.1
22.2
34.11.
8.1
23.0
34.0'
8,07.0
23.8
33.7
7.9
6.9
,24.6
33.4
7:8
6.8 5
'

TOTAL
AM
AF. o
YM
59.999 18.000.. 16.200
5.400
60.079
18.150
5.040
'17.990
62.291' 48.022
19.330 , 4.665
65.158
17.620
20.425
5.181
68.251
17.625
.21.804
5.567
71.941 '17.919
23.367
5.882
'76.300
18.435
25.059
6.279
81.324 19.223
26.927
6_729'
87.075 20.309
28.998
7.217
93.631
21.707
31.284
7.755
101.063 23.438
33.809
8.351
109.452
25.529
9.009
36.599
28.010
118.890
39.682
9.737
129.480
30.913
43.088
10.40
141.332
46.851
11.1428'
34.273
154.574
38.130
51.008
12.409
169.341
42.531
55.602 13.493
185.788 47.524
60.677 14.690
204.084 .53.166
66.283
16.013
224.417 '59.518
72.478
17.474
246.995 66.648
79.322 9.089

CF
7.200
6.803
7.975

2
3

8'.937
5

9.776
10.795
11.944
13.195
14.576
16.104
17.793
19.658
21,719
23.996
26.512
29.291
32.362
35.755

6
7
8
9
10
11

12
13

15
16

17
18
19

39.,503

20

43.644
48.220

YEAR
C4
14.0
12.9
14.4
15.1

15.4
15.7
16.0
16.1
16.1
16.1
16.
15.9

15.7
15.6
.15.4

,99.9

6'
7

8
.

9
lb

14.1.

11

114.0

13.2
13.0.
12.9
12.812.6

12
".

13

'

100.0
99.9
100.0
100.0
100.0
.100.0
100.0
100.0

14 --.--100.0
15

16
17

18
19

20

100.0
100.0
100.0
100.0
100.0
100.0

' 8.400

'11..320

.7.775

44.082

8.635
9.278
9 804

CM

4.533
. 4.871

'5.147

7:883
8.520
9.223
10.000
10.858
11.806
12.854
14.011
15.290
16.703

34.-789

7.,307

'

..

8.1

7.1

24.1

32.8

8 ,2

23.6
23.3

33.1

8.2 .
8-2

7.2
7.2
7.2
7.2
7.2
7.2%

25.1

2535
2620
26.5
26.9'

14.200
15.371

14.7'

33.1

8.0
8.0
7.9
7.9
7.8

7.0
7.0

7.7
7.'

12.1

14.5

7.1..

32.9
32.8
32.6
32.4
22.2
'52,1

11.3

14.7 \

14.6
-14.5
14.4
14.3,1.

6.9
6.9
6.8
6.8
6.7

14.27'
14.1

14.0

CF
12.0

'

7.1

8.2 .

16.666
18.097
19.677
21,423
23.352
25.484
27,839
30.440

'

8.1
8.1

33.4

13.M

.33.3
'33.2

3333!114-

23.8
24.2
24.6

7.200
6.803
7.555
8.118
8.578
9.157
9.814
10.525
11.309
12.179

14.7
14.8
14.8
14.8
14.8
14.7

33.3

23.1

23.1
23.3,
-23.5

CF

10.466
11.216
12.028
12.925
13.919
15.016
16.228
17.567
19.047
20.682
22.488
24.484
26.689
29.124
t 31.816

5 494
5.888
6.315
6.785

PERCENTAGE'DISTRIBUTION .OF HERO

AM
AF
YM
YF
CM
30.0
27.0
9.0
.8.0
14.0
30.2
29.9
8.3
7.1
/12.9'
28.9
31.0
7.4
e.5
13.8
27.0
7.9
6.9
31.3
14.2
25.8
8.1
31.9
7.1
14.3

100.0 ".24.9 '1432.4

L4.1
14.1

5.1
14.'

13.2
13.5
13.8
14.0

13.3.

14.5 ,

140.0
100.0
100.0

. 12.6

15.2
14.9
14.8

CF
12.0
11.3

13.9
13:8
13.6
73.5

TOTAL
100.0

YF
4.800

12.4
12.5
12.7
12.8
12,9
.12.9
13.0
13.0
12.9
12'.9

12.8
12.8
12.7
12:6
12.5
12.4
12.4
12.3

CONSTANT ANNUAL HARVEST IS 3.00 MALES, 1.00 FEMALES (MILLIONS)

CONSTANT ANNUAL HARVEST IS 4.00 MALES, 0.00 FEMALES (mamas)

36

189

t-

37

10

TABLE 3 (Continued)

TABLE 3 (Continued)

Case c)
YEAR

TOTAL
55.999
60.079
614391
62.863
64.226
65.717
67.359

69:126'

71.038
73.120
75.389
77.864

0
1

2'
3

a,

10

Case d)*--.

MIL'LPONS OF BUFFALO
AM
AF
YM
YF'
'18.000 16.200
5.400
4.800
19.150
16.990
5.040
4.320
4.665
19.912, 17.380
4,082
20.473 17.573
4.893
4.281
21.119 17.905
5.005
4.379
21.817 1845
5.061 , 4.428
22.522 .18.701
5.156" 4.512
23.263 19.151
5.269
4.610
24.052 19.651
5.386
4.712
24.889 20.203
5.515
4.826
25.781
20.812
5.659
4.952
26.737 .21.486,
5.818
5.091
27.764 22.230
5.994
5.244
22..1371. 23.052
6.18$_ 5.414
30.069 23.960
6.402
5.602
31.367 24.964
6.639
5.609
32.778 26:072
6.900
6.038
.34,314 27.297
7.189
6.290
7.508
35.991 28.651
6.570
'.861
6.879
37.823''30.146
39.82.8, 31.798
8.251
7.220

80.571

12
13

83.533

14

86.781

Is

90.3.44

16
17

94.257
,98.559
103.290
180.496
114.;30

18
19

20

CM

CF

YEAR

8.400
7.775
8.155
8.342
8.435
8.594
8.781
8.976
9.192
9.432
9.697
9.99

7.200
6.803

0
I

6o.o79

7135

60.491,

11

60.568
60.20;
59.493
58.417
56.928
55.002
52.610
49.715
h6.277

12

42.251

13

37.587
32.229
26.115

0.931

i,

10.670
11.065
11.501

1E982
12.514
13.102
13.752
14.470

7.299
7.380
7.520
7.684
7.854
.8.043
8.253
8.485
8.741
9.024
9.336
9.682
10.063
16.484
10.950
11.465
12.033
12.661

TOTAL

W999

4
5

6
7

8
9
1.0

14
15

16

19.174
11.329
2.495
-7.423
-18.533

17

18
19

20

AM
18.000
20.150
21.922
23.325
24.613
25.715
-Y6.609
27.304

27.795
28.071
28.125
27.944
27,517
26.830
25.865
24.603
.23.025

21.105
18.816
16.129
13.008

MILLIONS OF BUFFALO
AF
YM
YF
16.200
15:989
15.430
14.720
14.006
13.222
12.343
11.374
10.304
9.122.

7.815
6.372
4.778
1.069
-i.080
-3.456
-6.081
-8.981
-12.185
-15.725

5.40o
5.040
4.665
4.605
4.443
4.239
4.033

4.808
3.555
3.275
2.907
2.627
2.250
1.835
1.376
0.868
0.308
-0.311
-0.995
-1.751
-2.586
_

YEAR

PERCENTAGE DISTRIBUTION OF HERD

AM
AF
YM
YF
CM
30.0
2.0
9.0
8.0
14.0
31.8
28.2
8.3
7.1
12.9
'32.5
28.3
7.5
6.6
13.2
32.5
27.9
7.7
6.8
13.2
32.8
27.87.7
6.8
13.1`
33.1
-27.4
7.7
13.0
6.7
33.4
-27.7
7.6
6.6
13.0
33.6
27.7
7.6
6.6
12.9
33.8 A, 27.6
7.5
6.6
12.9

TOTAL

100.'0

2
3

4
5

' 64.
7

8
9

100.0'
100.0
92.9
99.9
100.0
100.0
99.9
100.0

1,0

tom

11

99.9
10W.0

.1
1.3

14
'15

16
17
18
19 ,

20

100.0'
100.0'
.100.0
100.0
100.0
100.0
1Q0.0

14.0
,

34..3

.34.4
34:5
34.6
34.7

-34.7
34.8
34,8
34.8

348

27,6
27.Z
2i.5
27.5
27.5.
27.6
27.6
27.6
27.6
27.7
27.7
27.8

7.5
7.5
7.4
7.
.4

7.3
7.3
7.3
7.2
7.2
7.2
7.2

6.6
6.5
6.5
6.5
6.4
6.4
6.4
6.4

12.9
12.8
12.,8

12.8
.124.
1247

12.712.7
12.6

6.3.

6.3'
k.3
6..3

'

12.
12'16

12.6

CF
12.0
11.3
11.6
11.6
11.4
11.4
11.4
11.3
11.3
11.2
'11.2
11.2
1122
11.1'
11.1

YEAR

TOTAL
100.0

100.0
100.0
100.0

2
3
4

1,00.0.

1014.0

'6

9
10

100.0
loorp
100.0
100.0
100.0

11

100.0...

12

100.0
100.0
100.0

7
8

),

13
14

4.800
4.320
4.082
4.029
3.888
3.709
3.529
3.332
3.110
2.866
2.596
.2.298
1.969
1.605
1.204
0.760
0.269
-0.272
-0.871
-1.532
- 2.263,

CM
8.400
7.775
7.675
7.406
7.065
15,723

6.346
5.925
5:459
4.946
4.378
3.751

3.058
2.293
1.447
0.513
-0.518
-1.659
-2.919
-4.311
-5.849

CF

7.200
6.803
6.715
6.480
6.182
5.882
5.553
5.184
4.777
4.327
3.831

3.282
2.676
2.006
1.266
0.449
-0.453
-7.451
-2.554
-3.772
-5.117

PERCENTAGE DISTRIBUTION OF HERD

AM
AF
YM
YF
30.0
27.0
9.o
33.5
26.6
8.3
7.1
36.2
25.5
6.7
7.7
38.5
24.3
7.6
.6.6
40.8
23.2
6.4
7.3
41.2
22.2
6.27.1
45.5
21:1
6.9
6.o
47.9
19.9
6.6
5.8
50.5
18.7
6.4
5.6
53.3
6.2
,17.3
5.4
56.5
15.7
.5.9
5.2
60.3
13.1
5.6
4.9
65.1
11.3
5.3
4.6 ,
8.o
71.3
4.8
4.2
80.2
4.2
3.3
3.7

CM
14.0

CF
12.0
11.3

12.6
12.2
11.7

11.1
10.6,

11'.3

10.8
10.4
9.9
9.4
8.8
8.1
7.2
6.1
4.4

10.2
9.8
9.5
9.1

8.6
8.2
7.71,
7.0
6.3
5.3
3.9

11..1

`ii.I
11.0
-,-/11.6
11.0

.CONSTANT ANNUAL HARVEST*IS 1.00 MALES, 3.00 FEMALES (MIL IONS)

The date for years 15120 is nonsensical, and means that
e
herd is extinct: after the 14th year, the required hary t of
adult fema,les is not available.

CONSTANT ANNUAL HARVEST IS 2.00 %ALES, 2.00. FEMALES (MILLIONS)

/
.

191

o,

3g.

39

14
TABLE 3 (Continued)

Mao

FABLE 4

Twenty perOgnt harvests lead to early extinction in all fit,

cases requested ierjxercise 6:

Case e)

MILLIONS OF BUFFALO
YEAR

TOTAL
59.999
60.079
59.591
58.273
56.175
53.269
49.475
44.730
38.965
32.099

-I
2
3
4

5
6
7

8
9

AM
18.000
21.150
23.872
26.178
28.107
29.613
30.696
31.344
31.538
31.254

YM
5.400
5.040
4.665
4.317
3.882
3.418

AF
16.200
14.989
13.480
11.868
10.107

8.150
5.985
3.597
0.957
-1.958

2.911
2.347
1.723
1.035

YF
4.800

CM
8.400
7.775
7.195
6.470
5.696

4.320'

4.082
3.777
1
3.397
2.990
2.547
2.053
1.508
0.906

Case

CF,

7.200
6.803
6.295

4.851
3.912
2.873
1.726

...."

YEAR
0

5.661

4.98'4

4.2,45

3.423
2.514
1.510
0.402

0.459

YEAR
0
1

TOTAL
59.999
52 079
47.591
44.633

AM
18.000
9.149
d,472

-8.051

TOTAL

PERCENTAGE DISTRIBUTION OF HERD

AM
AF
-YM
YF

100.0
100.0
100.0

30.0
17.5
0.9

YEAR

TOTAL
100.0

0
I

2
3
4

5
6
7

PERCENTAGE DISTRIBUTION OF HERD

AM
AF
YM
YF

30.0
35.2
40.0
44.9
50.0
55.5
62.0
70.080.9

100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0

27.0
24.9 ,
22.6
20.3

9.0
8.3
7.8
7.4
6.9.
6.4
5.8
5.2

17.9
15.3
12.0
8.0

2.4 - 4.1

8.0
7.1

6.8'
6.4

6.0 -

..

27.0
36.4
44.7

9.o
9.6
9.8

8.o
8.2
8.5

s's;

CF

8.400
7.775
9.115
10.214

.803
7.975
8.937

CM

CF
12.0
13.0
16.7

14.0
14.9
19.1

v
CONSTANT ANNUAL HARVEST IS 12.00 MALE'S, 0.00 FEMALES (MILLIONS)

../

CM
14.0
12.9
12.0

CF
12.0
11.3
10.5
9.7

X11.1
10.1

5.6

9.1

5.1

7.9
6.4
4.4

4.5
3.8

MILLIONS QF BUFFALO
AF
YM
YF
4.800
16.200
5.400
18.990
5.040
4.3P,
4.665 t$4.082
21.280
23.278
5.469
4.785

Case b)

8.8
7.9
6.9
5.6
3.8 ,,"
..1

CONSTANT 'ANNUAL HARVEST IS 0.00 MALES, 4.00 FEMALES (MILLIONS)

Exting&jon occurs as a result of the hanoest fo)lowing the,
eighth ear.
.

Case c)

MILLIONS OF BUFFALO

YEAR

TOTAL.

59.999
52.079
42.191
30.863
18.446
4.627

2
3
4

as.

0
1

2'
3
4

.40

AF
16.200
12.989

9.580
6.163
. 2.310
-1.994

..YM

YF

5.400
4.800
6.040
4.320
4.665
.665
4.082
3.741" 3.273
2.759
.2.414
1.775
1.553

TOTAL

PERCENTAGE DISTRIBUTION OF HERD

AM
AF
YM
YF

100.00
100.00
100410
100,b0
100.00

30.0
29.0
28.8
29.3.
29.3

YEAR

AM
18.000
15.149
12.172
.9:063
5.415
1.214

27.0
24.9

22.7

9.0
9.6
11.0

19.9
12.5

12:1
.14.9

8.0
8.2
9.6
10.6
13.0

CM
A.8.400

CF

7.200
6.803
5.455
4.023
2.588
0.970

7.775
6.235
4.598
2.958
1.108

CM
14.0
14.9
14.7
14.8
16.0

CF
12.0
13.0
12.9
13.0
14.0

CONSTANT ANNUAL HARVEST' IS 6.00 MALES, 6.00 FEMALES. (MILLIONS)

41

193
194

TABLE

-Table.4 (coritinued)

.Case d)

mole, the initial herd was

or Exercise 7. As one
)Data
transformed for one year n this ca .strophic way:
.60

MILLIONS OF BUFFALO

YEAR
0
1

2
3

YEAR
0
1

TOTAL
59.999
52.079
39.491
23.978

TOTAL
100.00
100,00
100.00

AM
18:000
18.150
18.022
17.620

AF
16.200
9.989
3.730
-2.394

YM
5.400
5.040,
4.665

2.877

YF

4.800
4.320
4.082
2.517

PERCENTAGE DISTRIBUTIONAF HERD

YF
AF
YM'
AM
30.0
34.8
45.6

27.0
19.1

9.4

8.0

9.0
9.6
11.8

8.2
10.3

CM
8.400
7.775
4.795
1.790

CM
14.0
14.9

CF

7.200
6.803
4.195

CF

OF BUFFALO

59.999
52.079
36.791

AF
16.200
6.989
'-2.119

YM

YF

5.400
5.040

4.800

11.665

4.082

4.320'.

CM
8.400
7.775
3.355

OERCENTAGE DISTRIBUTION OF HERD

4'

YEAR
0
1

TOTAL
100.00
100.00

AM
30.0
40.6

AF
27.0
13.4'

YM

9.0
9.6

YF
8.0
8.2

CF
.9

7.200

10

6.803
?.935

11

12

13
14

CM
14.0
14.9

15
16

CF
12.0
13.0

17
,18

216.697 '72.803
239.423 80.445

19

CONSTANT AN,IJAL HARVEST IS 0.00 MALES; 12.00 FEMALES (MILLIONS)

20
YEAR
'

^ 100.0

'3

99.9
100.0
100.0
100.0

--400.0

100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1Q0.0
100.0
100.0
100.0

10
11

12

13
14

15

16

r.

42

TOTAL

2
4

17
18
19'

4%

.15
0
0

0
.15
0

0,

0'

.25
.20

20

70.438

3.417
3.666
4.116
4.557

2.990
3.208
3.602
3.987
4.394

5.022
5.549- 4.855

5.366
6:132
6.775
5.928
7.485
6.549
7.236
8.270
7.995
9.137
8.833
10.095/
11.153
9.759
12.323 10.782
11.913
13.614
13.161
15:042
14.541
16.619

PERCENTAGE DISTRIBUTION OF HERD

YF
YM
AF
AM
8.0
9.0
27.0
30.0"
3.1
3.6
34.0
37.8

100.0
99.9

195

15.823
17.438
19.267
21.295
23.526
25.991
X8.716
31.727
35.053
38.728
42.788
47.273
52.229
57.7Q5
63.754

17.962
59.212 15.814
65.418 21.911
72.286 24.233
79.885 26.788
88.275 29.611
97.543 32.130
1p7.783
36,175
119.095 , 39.980
131.5930 .44.184
145.400 48.828
160.655 53.958
177.507 59.625
196.126 65.886
53,.526

Case e)

AM'

16.20 14.294

48.371

18.000
21.150
23.872

0
.40

e/

and then transformed further for 19 more ye'rs using the usual
The results:
matrix M.
MILLIONS OF BUFFALO
CF
/YF ,\
CM
AF
YM
AM
YEAR TOTAL
7.200
4.800 \\8.400
16.200
0 '59.999
18.000
5.400
3.240
1.080
4:050
1.260
34.229, 12.960
11.639
4.888
1.943,
2:429
39.973 13.257
11.867
5.586
2
4.984
3
5.'96
44.113
14.416:`11.731
2.933
3.352

CONSTANT ANNUAL HARVEST IS 3.00 MALES, 9.00 FEMALES IMILLIONS').

MAR TOM,

.40

1 .566

0
.60'

0-

12.0
13.0
10.6

12.1

>

33.1

32.6
33.5
33.5
33.4
33.4

29.6
28.8
29.5
29,5
.29.4
'29.4
29.4
29.4

_33.5
33.5
33.5

29.4

ZLI

Ir

31.5
I 33.5
33.5
33.5
.33.5

33.5
3.3.5

33.5

21.4
29.4.
29.4

'

'.29.4
4

29.4
29.4
29.4.
29.4

6.0
7.5

4.8
6.6

.7.0

6.1

6.8
6.9
6.9
6.9.
6.9
6.9
6.9
6.9
6.9
6.9
6.9
6.9
6.9
6.9
6.9
,7 6.9

5.9
6.0
6.0
6.0
6.0

6.0
6.0
6.(1

6.0
6.0
Et.0

6.0
6.0 .
6.0
6.0
6.0

6.11

5.347
6.003
7.595 6.645
7.324
8.370
9.28 \8.092Vr
6:943
10.221
9.880
11.292
12.475, 10.916
13.784. 12.061
15.229 13.325
14.722
16.825
18.589 16.265

6.861

20'.538

22.691

25.070 '2f.936
23,698 24.216
30.602 26.777
CM

CF

14.0
11.8
13.9
12.9
12.6
12.8
12.8
12.7
12.7
12.7
12.7
12.7
12.7
12.7
12.7
12.7
12.7

12.0
9.4
12.2

12,7.

12.7
12.7
12.7

ANNUAL HARVEST IS 0.00 MALES, 0.00 FEMALES,(MILLIONS)

196

17.971
19.855
-t.

11 41.,

11.0
11.2
11.2
11.1
11.1
11.1

11.1
11.1
11.1
11.1
11.1
11.1

41.1
11.1
11.1
11:1

11.1

4'3

TABLE 6
Data for Exercise 16. The herd does indeed 'emain very stable.
There is some roundoff error: the harvests taken were .086 and .056
annually (males, females, in millions), rather than the .086052 and
-055948 that the table's data for herd 2 indicates.

YEAR TOl'AL'
0
2
3

4
5

6
7

18

9
10
11

12

13
14

15
16

17
19

0.997-

20

0.996

YEAR
0

TOTAL,

2
,3

4
5

how
1.000
1.000
0.999
0.999
0.999
0.999
0.999
,0,998
0.998
0.998
.0.997

0.999
1.000
1.000
1.00o
1.000
1.000
1.000

6
7
8
9

100:0
100.0
100.0
100.0
100:0
100.0
100.0

AM
0.018
0.018
0.018
0.018
0.018
0.018
Q.018
0.018
0.018
0.018
0.018
0.018
0.018
'0.018

0.018
0.018
0.018
0.018
0.017
0.017
0.017

MILLIONS OF BUFFALO
YF
AF
YM
0.403
0.116
0.101.
0.115
0.402
0.101
0.402
/0.116
0.101
0.402
0.115
0.101
0.402 , 0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.402
0.115
0.101
0.401
0.115
0.101
0.401
0.115
0.101
0.401
0.115
0.101
0.401
0.115
0.}01
0.401
0:115
0.101
0.401
0.115
0.101
0.401
0.115
0.101
0.401
0.115
0.101

PERCENTAGE DISTRIBUTION'OF NERD

AM
AF
YM
YF
1.8
40.3
11.6
10.1r
1.8'
1.8
1.8

J.a
1.8

1.8

100.0;:

40.2
40.2
40.2
40.2
40.2
40.2
40.2

1,00:0,

1.8

100%0

1.8'-110.2

10

100.0

11

1.8
1.8

12

100.0
100.0

13

1,00.0

14
15
16
17

100.0
100.0
100.0
100.0
100.0
100.0
100.0

18
19

20

1.8 '21

1.8
1.8
1.8
1,8

1.8
1.7
1.7
1.7

40.2 '

40.Z
40.2
40.2
40.2
40.2
40.2
40.2
40.2
40.2
40.2 '
40.2

11.5
11.5
11.5

10.1
10.1
10.1
10.1
10.1

11.5
11.5

10.1
10.1'

11.5
111.6,

0.193
0.193
0.193
0.193
0.193
0.193
0.193
0.f93
0.193
0.193
0.193
0.193
0.193
0.192
0.192
0.192
0.1p2
0.192
0.192
0.192
0.152

CF
0.169
0.165
0.165
0.169
0.169
0.169
0.169
0.169
0.169
0.168
0.168
0.168
0.168
0.168
0.168
0.168
0.168
0.168
0.168
0.168
0.168

CM
15.3

CF
16.9

19.3,

16.9,

19.3
19.3
19.3

16.9
16.8
16.9

19.3
19.3

16.8-

CM

19.3'

11.5

10.1

19.3

-11.5

10.1

11.5
11.5

10.1
10.1

11.5
11.5
11.5
11.5
11.5'
11.5
11.5
11.5
11.6

10.1
10.1

19.3
19.3
19.3
19.3
19.3

10.1'
10.1

10.1
10.1
10.1
10.1
10.1

19.3
19.3

19.3
'19.3
19.3

16.8
16.8
108-16.8
16.8
16.8

16.8
16.8
16.8
16.9
16.9
16.9
6.9

19.3

1,9.3

169

.9

44

197

Returnta:
EDC/UMAP-

STUDENT FORM 1

55 Chapel St.
Newton, MA 02160

Request for Help

Student:
If you have trouble with a specifiC part of,)this unit, please fill
out this. form and take it to youk instructor for assistance. The information
you give will help the author to revise the unit.

Your Name

Unit No.

Page

0 Upper

Model Exam

Section
OR

()Middle

.Paragraph

Text
Problem No.

Lower

Description

nstructor:

t4

f Difficulty:. (Please be specific)

Please indicate your resolution of the difficulty in this box.

Corrected'errors/in materials.

Gav

Problem No.

OR

List corrections here:

studentltetter explanation, example., op ?rocedure than A unit.

e brief outline of your addition-here:

(2) Assisted student in aciluirini general learning and problem - solving,

skills (not using examples from this unit.)

\\--

193
el

t
structor's Signature
P1

se use reverse. if necessary.

1,

Return to
EDC/UMAP
55 Chapel St.
Newton; MA 02160

STUDENT FORM 2

Unit Questionnaire'
Name

Udit No.

Institution'

Date

Course No.

Check the Choice for each question that comes closest to your persodal opinion.,
1.

Now useful was the amount of;detail its the Unit?

Not enough detail to understand the unit

AUnit would have been clearer with more detail
ppropriate amount of detail
Unit was occasionally too detailed, but this was not distracting
Too much detail; I was often distracted
d

,2.,

How helpful were the problem answers?

Sample solutions were too brief; I could not do the intermediate 'stelis
Sufficient information was given to'solve thetproblems
Sample solutions were too detailed; I didn't peed them

3.

Except for fulfilling the prerequisites, how much did you use other sources (for
example, instructor: friends, or other-Books) in order to understand the unit?
A Lot

4.

.,

5.

'A Little

N.

Not at all

How long was this unit'in comparison to the amount of time you generally spend on
a lesson (lecture and homework assignment) in a typical math or science course?.
Much'
Longer

Somewhat

'Somewhat
Longer

About
the Same

Somewhat
Shorter

Much
Shorter

Were any of the following parts of the unit confusing or distracting? (Check
as many as apply.)
P
Prerequisites
Statement of stills and concepts (objectives)
Paragiaph headints- '
Examples
Special Assistance Supplement (if present)
Other, please explain

(1.

6.

Were any of, the folkowing parts

as apply.)

the'unit paiticularly helpful? (Check as many

i',

Prerequis.ites

Statement of skills an4 concepts (objectives)

Examples
Problems
'Paragraph he dings,
Table of C tents
Special Assistance Supplement (if present
Other,, please explain

%...

,Please describe anyttiM in theunit that you did not particularly like.
*

Please describe anything that you found particularly helpful. (Please use the back of
this sheet if you need more space.)

'H

1)

LCONOMlC EQUILIRRIIIPI1

WriaP

L, I MIT!.

I NE11: MODLLS

UNIT 208

by

Philip M. Tuchinsk,

4'

7623 Challe:;wofth
Dearborn !Its., Michigan

MODULES AND MONOGRAPHS IN UNDERGRADUATE

MA IEMATICH 'AND ITS APPLICATIONS PROJECT

TABLE OF CONTENTS

4440!

,PART'L:

ECONOMIC,EOUI LIBRIUM:
SIMPLE LINEAR MODELS

SUPPLY AND DEMAND FOR A SINGLE PRODUCT

Price Equpi6rium
2. The Purpose ea This Paper
1.

3.
4.

by Philip M. Tuchinsky

4,

PART II:

LI
1

5.

Assumptions about the Economy

Supply and Demand depend on Price
The One-Product Model

6.

Equilibrium in the One-Product. .Model

2
2

THE ANALOGOUS TWO - PRODUCT MODEL

7.

Modeling
Interrelated Proddcts
8.
The Two 'Product Model in Vector and
Matrix Notation
9.
Equilibrium Supply and Demand in the
Two-Product Model
10. Matrix"tevel vs. Entry Level Calculations

D and S,
Dollars
,

equilibrium
PART III:

demand

TART IV:

supply

9
.

GENERALIZATION TO n-PRODUCTS

PART V:

price domain

PART VI:
I

APPLICATIONS OF LINEAR ALGEBRA TO ECONOMICS

PART VII:

edC41Map/.55chapelstrinevhon.mass02160

12

13
13
15

HOW' DID WE GET THIS FAR?

16

13. Making a Start

1 4 .
Improving on Our First Effort
15. Hindsight is Perfect

Price P

11. The Model with n-Products

12. Solution of the n-Product Model

S*

si7

"-

..

16
17
18

TWO ECONOMIC INSIGHTS FROM THE MODEL

16. Total Demand
Supply
A
ARE LINEAR FUNC1IONS CRUCIAL TO THE MODEL?
18. A Job ror Taylor's Theorem
19. Discontinuous Supply and 'Demand Curves
SOLUTIONS TO EXERCISES

18
18

70
21

21'
24

26

.<

201

Intermodular Description Sheet:

Title:. ECONOMIC EQUILIBRIUM:
Authot:.

UMAP Unit 208

STAPLE LINEAR MODELS

MODULES AND MONOGRAPHS IN UNDERGRADUATE

.
.

Philip M. Tuthinsky
7623 Charlesworth
Drrborn Hts., Michigan. 48127

MATHEMATICS AND ITS,APPLICATIONS PROJECT (UMAP)

The goal of UMAP is to develop, through a communityof users

and developers, a system of instructional modules in undergraduate,
mathematics and its applications which may be used to supplement
existing courses and from which complete courses may eventually be

Dr.\Tuchinsky is a computer sci,eptiht and applied matheOati ian at Ford Motor Company s Research and Engineering
Center. He formerly t
n .the Mathematical'Stiences
Department of Ohio Wesleyan University (where earlier
editions of this.paper were written).
'

Review Stage /Date:

ITT

built.

The Prbject is guided by a National Steering Committee of

mathematicians, scientists, and educators: UMAP is funded by a
grant from the National Science Foundation to Education Development
Center, Inc., a publicly supported, nonprofit corporation engaged
:in educational research Am the U.S. and abroad.

9/20/79

Classification: APPL LIN ALG/ECON

Approximate Class Time:

Two 50- minute classes.

Intendet.Audience:, Lin r algelira studentsl'who have just learned

about calcula on
matrix inverses. To read Part VI, a
student shop). 4Have had some contact with differential calculus.
The paper is
o suitable for independent reading or seminar
presentationb'more advanced st ents.

PROJECT STAFF
).Ross L. Finne

Solomon Garf

kel

Director
Associate Director/Consortium
Coordinetot.

Prerequisite Skills:

'Felicia De
Barbara Kel zewski
Paula M.-SS tillo
Zachary Ze itas

1.

Elementary high-school algebra.

2. Graphing of straight lines.
3. .Familiarity with functions and flinction notation.
4.
Knowledge Of the domain of a function.
5. Interval notation [a,b3.
6.
Matrix and vector notation.
7.
Elementary matrix algebra including multiplication of matrices.
8.
Matrix inverses as a concept, with algebraicilaws and notation.

W.T. Martin
Steven J. Brams
Llayron Clarkson
Ernest J. Henley
William Hogan

(For Part VI only:)

9.
10.
11.
12.

NATIONAL STEERING COMMITTEE

The derivative and its notation.

Continuity and gengral eboothness concepts.
Taylor's Theorem (the equation of the tangent line),
Newton's Method (one variable) is mentioned in Exercise 17.
(Part VI fts a nice vehiclecfor motivating Newton's Method.).

Donad A. Larson
William F. Lucas

Output Skills:
'

1.

Discuss the Movementof prices due to shifts in supply and

demand, and price equilibrium.
2.
Define total demand and free supply and describe the effect on
prices of an increase in either.
3.
DesCribe an application leading to.a set of linear equations.
4.
Tell whether calculations are at matrix level or,entry level
in linear algebra.
5. (OPtimistioally) Abifity to generalize a simple ?model from one
b.

variable to two and then, many.

(Part VI) Describe an application of the tangent line.

Other Related Units:

,Unit 209:

General Equilibrium:

R. Duncan Luce
George Miller
Frederick Hosteller
Walter E. Sears
George Springer
Arnold A. Strassenburg
Alfred B. Willcox

'

M.I.T. (Chair)
New York University
Texas Southern University
Uniyersity of Houston
Haltvard University
SUNY at Buffalo
Cornell University
Harvard University
Nassau Community College
Harvard 'University
University of Michigit Press
Indiana University
SUNY at Stony Brook
MatheMatical Association of America

a
The Project would like to thank Kenneth R. Rebman of
California State University it Hayward and K.L. Huehn of California
Polytechnic State University for their reviews, and all others who
0
assisted in the production of this unit.
,

This 'material was prepared with the support of National

Science Foundation Grant No. SED76-19615' A02. Recommendations
expressed are those of the author and do not necessarily reflect'
the views of the NSF, nor of the National Steering Committee.

A Leontief Economic Mdef.

1979 EDC/Project UMAP.

All rights reserved.

202

Associgie Director for Administration

Coordinator for Materials Production
Administrative Assistant
Staff Assistant

203

r
ECONOMIC EQUILIBRIUM:

PARTiI:

--a very simple linear algeSraic introdbCtion to mathematical

economics is appropriate.

SIMPLE LINEAR MODELS

Our work here can serve as one instance of an

important phenomenon:
linear algebra is a basic tool
uted in virtually all areas of applied mathematics.

SUP2LY AND DEMAND FOR VSINGLE PRODUCT

-.
Price Equilibrium

'1.

A product is "in equilibrium" of "at its equilibrium

price" when supply equals demand for it. This means the
amount of the'product aVailable from,sellers equals the

3.

amount that purchasers want to buy.. (We include any

economy of theAdam Smith variety.

commodity, service or'manufactured product under the

general umbrella of "products" here.)
Of course, supply and demand are seldom exactly equal
for any produdt and even .ifachieved, equilibrium Is
,

Assumptions about the Economy

We will assume an economy that is grossly simplified,

from reality, a classic, competitive, capitalistic
'Prices are not
controlled by government, buyers or sellers' in this
economy--they fluctuate freely in response to supply and
demand.
There are no monopolies, no cartels, no collusion
among buyers and sellers, Inflation is not modeled; the

entire discussion is in termsof "1967 dollars" or some

other standard monetary unit of purchasing power.

momentary.

If supply exceeds demand, sellers lower their

prices to attract%buYers; i.e., prices tend to decrease.
-If demand exceeds supply, the buyers who'most want the
product bid up its price, and prices rise in response.

$1.1yers and seller

in ,our economy have "perfect

iliformafion."

This means that they all knot/ the current

supply, demand and price, as if all buying and selling
were done in one large auction room with all potential
buyers ,and sellerS participating.

It is exactly when supply equals demand that, these:two

opposite economic foi:ces are balanced, leaving the price
at a standstill.
Thgt balanced state of opposing for,ces

is exactly the usual meaning of "equilibrlup."

4.
2.

Supply and Demand depend on Price

The Purpose of This Paper

Let's 'analyze supply and demand for one product.

We will study teve;a1 versions of a Very elementary

.mathematical model of price equilibrium ih this paper.
Hopefully, the eco emit content is clear and interesting,
o
but our main goal is m
ical.
We will discbver that
mathematical economists in: itably find themselves using
linear algebra to express heir ideas.
If we went beyond

, D = current demand for the product (in dollars)

(1)
NC

eggs, etc.).

:13

S = current supply ofthe product

(in dollars)

bill

We might have expressed D and S as the amounts demanded

an supplied in production-units (boccar- loads, dozens of

differpntial/difference equations, time series forecasting,

4
linear programming, etc. All of%these use linear algebra
and linearizing methods to achieve. practical results--so

-/

P = current price of the product (dollars/item)

our simple model to some o ,the multitude of economic

models proposed in rec&nt decades we would find more
advanced mathematical tools in use;. queueing theory,

4:s 204

bLet

H6wever, we will want to compare oneiproduct

to. another later, so we'll express D and S in dolears

from the start. Once the current price P is known we

convert the amounts demanded and supplied into dollars to

calculate D and S.
(If 4 million dozen eggs are demanded

205

at a .wholesale= price of jo.s dollars dozen, we.have a

.

$2 million demand D for eggs.)

D and S,
Ddtlars

In fact, it is natural tg regard.D and S as functions

of the price P. ) This goes bandin-hand with our assumption of a purely, capitalistic economy of,vatue-conscious
.1

buyers and profit" conscious sellers.

(In reality supply

and demand depend on price as well.as such emotional
elements as style, fads, and 'the effects of fantasy-

oriented advertising.)
5.

14:

The One-Product Model

The simplest way to make D and S functions of p

is to us$ straight lines. .That is, let's- take as dur

actual domain

mathematical model
D = a + bP.
(2)

Figure 1.

S = c + dP
'

Supply and demand lines for one product.

=

where a, b, c, and d are real constants.

What can we

say about a, b, c, and d on valitative grounds As the

priceP grows, we expect demand to drop (at a highe
price there are fewer buyers), so slope b
D L'0, we know a > 0.

<

0.

Since

D and S,
Dollars.

And as P grows, the supply S

ecp

grow because more companies find it profitable to make

the product, hence slope d > 0.,.,Figure 1 sketches this

situation and shows c < 0; let's see why.

)

equilibrium

There will be

some price -.of- first - supply Ps (namely, the cost of menu-

faCtur4g) such that no supplier will make the produCt

if P < P
Thus our's'iraight'line must cross the price
s

axis at positive Ps and c, its intercept on the vertical

axis, must be negative.
Figure 1 also shows.the price-of-last-demand Pd at
which the dethand linereaches zero:

at prices.rP > Pd Ind

one is interested in buying the product.

price domain

Only non.

to

negative yalues of D and S make economic sense, of course.

1Thus we'll' consider P only in the domain (Ps,Pd), as
Figure 2. Che one-prodrt model.

shown in Figures 1 and 2.

206

-0

267

''

Equilibrium in the One-Product Model

The price equilibrium occurs when S = D (supply

equals demand). As thdtsketch shows (Figure 2), there
. is one price P* for which our model predicts equilibrium
,

Exercise 3.

Equations (2) with a > 0, b < 0,c

> 0 give

linear functions of real P.

(see Exercise 1). The corresponding dollar amounts S*

and D* are also sketched. We can calculate P* by setting
S = D in (2) end solving to get

a.

How do yoll know that these lines meet exactly once'

somewhere in the plane?

b.

How do you_know that the point of intersection

(P*,S*) = (P*,D*) satisfies Ps < P* <.Pd and S* > 0,
D* > 0?

P*

(3a)

(This requires an economic argument. Show that

it is nut true based on the mathematical facts alone.)

By plugging P*back in for P in either equation of (2)

we also find the &11,1ildium demand/supply level:
(3b)

S*

D*

da
d

PART II:

be
7.

What have we achieved with this bit o

igh:school
algebra? Under the crude assumption that simple eqUations
like (2) hold, we can predict the pri'ce P* a prOTact
should sell at and the amounts S* = D* that people should

make and will buy!

Our next goal is to extend this model

to more complicated cases of general equilibrium where
many competing.products are in equilibrium simultaneously.
Exercise le

Find P*, D*, S* if the formulas--forsupplfseddese,ifid--

1.5P

+ 5.25P.

algebraically solve for P*, D*, S.

b.

identify a, b, c, and d'and substitute them in (3a,b)

c.

graph the lines and read the equilibrium point off

can substitute for each other, this is especially - clear.

For ex
e,
s large cars have become expensixe to buy
and ope
e, people have substituted smAler. cars.
It
is logical to think of the supply and demand amounts for
large cars as functions ofboth lhige car and shall car
prices.

The demand function for cars of....41,zy size might

As a first step, let's study a two-prod4ct market.

the graph.

Put

4
Exercise 2.

In a real economy, the supply and demand for a

product depepds on its price and on the prices of other
related produc s (and on other factors). When products

prices (like that of perfume) but there are i portant

interrelationships amoung products and, to ma e our
mathematical model more realistic, we should nclude many
of those relationships.

(Ise these three methods:

a.

Modeling Two Interrelated Products

also depend on the price of labo'r for having t e car

serviced and repaired, the price of gasoline, he'price
of auto parts, etc.
It'does not depend on mo t other

are
D = 22

Repeat'Exercise 1 for
,

THE ANALOGOUS TWO-PRODUCT MODEL

D = 30

4P

S= 6P

2.

(4)

S1, Pi = demand, supply, price for the first produce

(large cars, say)
S

'

2'

demand, supply, price for the second product

(small cars? say).
,

'6

208

209

We assume Di, Si,,iir-27,S2 all 'to be functions oofsthe two

-D.1

prices P1, P2 and all are expressed it dollars:,

we assume the simplest functions (compare notation with

''. (2));

al

.111 b12

a2

-b21
b21 422

d12 -P1

Di

al + b11P1 + 1)12'32'

c2

S2

D2 = a2

b21P1

(5)

1)2

(6)

'

22

P 2
.41

and shift to the obvious matrix notation.

cl + d11P1 +

d12P2.,.
-D

d21P1

IS

.1

=,c2

21

b22P2

si

D2

d22P2'

Define

1,

S2

(7)
-

The a's, b's, c's and d's are all known real constants.
In the context of our large cars--small cars exatule, we
,

can predict the signs of these constants.

The demand for

large cars, D1, should be positive, should decrease as
P
1

increases and should increase as P

-

v..

and rewrite (6) ps

increases (i.e.,

as small carstecome more expensive and hence less '

attractive to buyers): Thus a
0, b
< 0
b
> 6..
1
,
11
12
Similarly, a2 >. 0, b21 > 0, b22 < 0.
The supply Si of
large cars should grow as P1 increases and also grow as

(&)

Compare (2).

D = a.+1)P

S = c

dP.
n.

shift-demand,to their competitive large cars and hence

Notice how naturally (2) has been generalized

through-the use of lineal 'algebra. The "supply equals

stimulate productiop of large cars).

demand" equations are now S1--

increases (because higher pries for small cars shoujd

Thus c1 < 0 (for

.....__the_vme threshold-of-manufacturing-costs reasopsas

t
before), d11 > 0 and d12 > 0. Similarly,
c2 "'0, d21 > 0,
d

22

> 0.

'c

S = D.

(.9)
.

-L''

The equilibrium price vector P* =

'

.a.

'8.

is the value
P'*
IT.I:

of P we get by substituting (8) into (9):

The Two-Product Model in'Vector and MatrixNotat,ion

Of course
We will set S1 = D1 and S2 =..D2 (supply
.

c + dP*

'equals clemand) and try to calculate the equilibrium prices

P1 *, P *

and $2 = D2, i.e.,

a + bP*.

But that will be easier to dc; after we arrange-

Elementary matrix algebra leads to

(Si 4s1

(10)'

210

(d-b) P*j= a-c,

.8

.211

to.

which is a set of linear equations for P*,'We'll assume

Exercise 5.

thgt the 2 x'2 matrix d - b has an .inverse and,we'll

Substitute your P* solution from Exercise 4 into the

I-

equations to calculate D* =

multiply through by (d-b)-1 from the left:

I,
(11a)

P* = (d-b)_r (a-c).
Of course,' D* anli.S* in (11.b) are equal, a4 we should
,expect fromthe way woatalcVlated P*.,B-*--and S.
A little
matrix algebra will show this:
.

Compare this to (3a):

multiplication by the'matrix
inverse of d-b here very naturally replaces multiplication

by the reciprocal of scalar d-b there.

*1k

Exercise 4.

D* = a + b(d-b)-1(a-c)

Finathe equilibrium prices if

1)

= (d-b)(d-b)-1; + b(d-b)-2(a-c)

= 12 - 1.5P- + P
1

D2 = 20e+ 2P

P2

(because a =_Ia = (d-b) (d-b)-1'al

= d(d-b)-1; - b(d-b)-1:

.T .S1 = -6 + 1.6P1 +, 2P2

+ Ii(d-b)

= -5 1- 4P
1

+ 5P

'

b(d-b)

-1

c.

After the.can9eLlation:

by
a.

direct caleulatilp from S1 =

b. ' identification of a, b, c, d and substitution in (11a).

Exercise 6., With thiesstart,
S* = c + d(d-b)
9.
,1

Equilibrium Supply and Demand in the Two-Product Model

Segkii4 a complete `analogy between_ (3a,b) and the

'

+ d(d-10-1(g-c),

show that
S* = d(d-b)

-1'"

a - b(d-b)

We get:

-1
c

also.

D* = a + b(d-b)

(a-c)

two- product model', we next sub.i.titu4 P* from (11a) into

the equations for D and S to find D* = S*.

-1

(a-c)

S* = c + d(d-b) -1 (a-c).

By writing (,3a) rater clumsily as

44

Hmm .

that doesn't look much like (3b)

,

in fact,

d -b

.i.t's not so obvious.that D* = S* at all.

dio4-'t
a'
e..24.

da - be =

Has our analogy

(12.-)

da

be

d-b

_1 bia

fd

d- b
_1

bjG

212

= d(d-b)

-1

b(d-b)

-1

c,

we disCover that the analogy between (3b) and' (11b1 is

not dealOt all, but who would ever write (da-bc)/(a -b)

-10. Matrix Level vs. Entry Level Calculations

.

We used exactly thg same steps to calculate P*

(see (3a) and (11a)) from the one- and two-product models.
In the two - product case, all calculations were at matrix

in so comelicated a way?!: unfortunately, the liberties

4

we enjoy with scalar arithmeticwe could use any, pf

da-bc

ad-cb
d-b

(d-b)

-1

level:

(ad-bc) = (da-bc)(d-b)-1

thinking about the individual numbers a..; c.,

bii,
J
dij,,etc., that make them up. All the calculations in

among .they formsare simply not*available when b and d

trices andpapc are,vectort, ,,,The main problem is

Sections 8 and 9 ahoge were at matrix level.

that matrix multiplication is not commutative.,00Exercise 7.

To actually calculate the components Pi* and P2*

of P* in (11a),,however, we mbst calculate the 2 x 2 matrix
inverse of d-b and multiply it by the vector a-c. Such

Prove that
d(d=b)

-1

= (d-b1 Id

calculations are at entry level (they use the entries, the

numbers that form the vectors and matrices).
This calculation is quite. a bit more complicated than the single

/)

if and only if db = bd.h

-Exercise 8.

*4

Prove that
b(d-b)

-1

= (d-b)

we thought of a, c, Pp.)), S, b, d, b-s, (b-d) -1

as, vector and matrix entities; single objects, without

division needed to compute P* in (3a).

The great beauty

and wonder ofqinear algebra .is the extertt to which we
can do useful calculations at matrix level, as if we

if and only, if db = bd.

had single "numbers" (the matrices and vectors themselves)

to work with.
Eventually, we must complete our work
with grubby arithmetic at entry level, however.

Ordinarily we must expect that matrices b and d will

not commutecommutative matrices are the'exception and
not the rule in mathematics.

It's to our advantage to seek (at matrix level) a

form of our expression' that is least painful' to work with
at entry level.
For example, we used
.

Ok

If b and d happen to commute, we would have'

D* = S* = d(d-b)-la
-1

= S* = d(d-b)

- b(d-b)

-1

to show that D* = S*.

If we actually use this to calculate D* = S* we will compute one matrix inverse and
k
four matrix multiplications.
We don't have to work that

b(d-b)-lc

(da-i5c)

hard:

as in (3b); but we would be wiser too c6nsider the "natural"

form of (3a) to be (from (12))

D *

c=

S* = a + bkd-b) -1 (a-c)

.16

ms

" D* =,S* = dl

b4
la

Only,the commUtativ&ty 1)f multiplication of real numbers

allows a simpler form like (3a).
11

214

=.2

)c,

qt)

d(d-b)-1(a-c)

each involve one matrix inversion followed by only t/9

matrix multiplications.

If

215

12

a.,
PART III:

11.

GENERALIZATLON TO n-PRODUCTS

Now suppose that products 2, 425 and 7514 (small cars,

motorcycles and rapid transit fares, perhaps) compete
with product 1. As their prices rise, product I looks

The Model with n-Products

4nd b

more attractive to buyers, so

Why stop with supply and demand functions that

interrelate two products? Suppose an economy is made up
of n products, commodities, services, etc., and let

are all positive while the other b

1,425
b1,7514
a're'all zero. A

lj

rising price for a competing product tends to increase

the supply of 'product 1, as explained in Part II.

D
(13)

Si,
1'

= demand, supply, price for thesjth product,

for j = 1, 2, ..., n.

Following the pathway from Equations (5) to (6),

rewrite (14) using matrix products:

We still assume that Dj and Sj.,depend linearly on the

prices but we permit any and all toossible interrelation-

sHips by using all the prices in each demand or supply

ra

b11 b 12 '" bin'

b21 b22

b2n

c1

d11 d12 .: a ln

function:
D2
P + b P +
11 1
12 2

D2 = a2 + b21P1 44b22P2 +

iv

2n n

+bP
nn n

(14)
+ d P + d P +
11 1
12 2

+ d

= c

+ d

n2 2

nl

n2

nn

,1

d21 d22 "' d2m

P
ln n

ln n

_Dn

D=a .11+b laP +b n2P 2 +

+ b 'P
+ b

+ dnnP

nl dn2

nn

Naturally we introduce these vectors and matrices:

_

Please compare this to (5) and (2), which are simply the
special cases n = 2'and n = '1.

Di
1

D2

For the reasons di9cuaped in Part I, all a. >-0 and

all ci < 0. Most of-the bij,and dij will be zero; they
will be,nonzero only whdh i and j are competing products.

D =

Dn

Consider product 1, which might be large cars, for

S =

n_

s-

example' Naturally b11 <0 and dii > 0:

as their prices
rise, demand foK large cars decreases.and supply increases.

P.

13

2/6 ;

Thus-

and d

1,2'
1,425
1,7514 will all be positive while the
other dij are zero.

D1 = al + b

217

we

01'

b
1

a2

21

b
b

12

...

22

level effort needed to calcula e (d-b)-1 increases

rapidly as n increases. We would need a computer to
deal with the large n we would want to use in a genuine

bin
2n
ak

b=

a =

(16)

11

,41114

economic study.

bnl bn2

bnn
PART IV:
dln

dll
d

,c

21 ,22

13.

2n

Makdng'a Start

Let's take on the role of the applied mathematician

HOW DID WE GET THIS FAR?

who first developed this model.

How do we start? What

brainstorms along the way lead to progress and why do
they occur? What have we learned from earlier modeling
work that we put to use here?

nn

So, we must now imagine that we do not know about

this model. An economist comes to us with a question:
-

and write (15) compactly:

4 a + bP

"Supply, demand and price have these clear intuitive

relationships. Can mathematics help us understand the
relationship more accurately? Can we predict the price

(17)

S= c + dP.

and supply/demand at which a product will/should sell?"

This is an exact copy of (8):

We do some prelidinary reading and thinking and talk with

the economist until we understand the main mechanism:

12.

&caution of the n-Product Model

when supply/demand is in excess, this causes a shift in

The matrix .level calculations that led us from the

two-product model (8) to its soutions (11a,c),are not

price" where the forces of supply and demand are in

limited to 2-vectors and, 2 x2 matrices.

balance.

the price downwafds /u1wards towards a "fair market value

One of the

In that wording, it seems that price is influenced by supply and demand:

3

gieat advantages of matrix level work is that it applies

to n-vectors and n xn matrices for any n.
Exactly the'

price = f(supply,demand).

same reasoning and algebraic operations that led us from

(8) to (lla,c) work On (17) to give' us its equilibrium

We also turn around the language, however:

as the price
increases/decreases, the supply should inorease/decrease
while the demand decreases/increases. This wording
suggests that supply and demand are influenced by price:

solution:.
.1
*

= (d-b)

6-

--C.

lk
)

(18)

S*

d(d-b)-1a

supply = g(price) and demand = h(pTice).

* and thecomponents
of D* and S* does depend on the dimension n:
the entry
T,Areally calculatei P'* *P
'

218

'

15

As experienCed applied mathitaticians,' we prefer to

work with the latter approach:,,,we have more equations
16

219

4.4

and can easily express supply = demand. Thus we make

tht basic decisions that lead to the model of Part I:
We'll think about the simplest conceivable economy (one
product) by expressing supply and demand s functions
of the price. We hope to writesdown
cor rete functions:

for both products depend on the two prices and

we
specialize to the easiest concrete functions, in (5).
Aha!

A mathematical brainstorM--we can write (5)

using matrices as in (6). Our skills with linear
algebra
take over
introduce the vectors and matrices of (7),
reach the "same" model in (8) that we had in (2),
set
supply = demand and use matrix algebra to reach P*, D*,
S* in (lla,c). AlMost nothing is
new here: based on
our skills with linear algebra we have transformed the
success of Part I into results for a
more complex economy
410
in Part II.

S = g(P) and D = h(P)

4

and to solve S = D, a single equation in the one

valuable P:
g(P)

h(P) = 0

for the equilibrium price P*.

Now the jump to n-products is easy--we follow the

path that linear algebra points out to us, exPanring
two-vectors and 2 x 2 matrices to n-vectors and n x n
matrices.
It works again!

The details of Part I now follow when we decide to

make g and h very simple (Equations (2)) as.a first
thort. And we,are successful:
we predict I;*, D*, S*
in (3a,b)
.,

15.

*14. ,Improving on Our First Effort

Hindsight is Perfect

Now that we have the Model of Part III and see that
the modelt of Parts I and II are just the special
cases,
n = 1 and n = 2, we know that'the.n-product
model (17)
and its solutions (18) are what we were'after when
'

'-

The answer to one question leads to the asking of

many more! Here are two reasonable- ones:
A.

Can we chodse functions g. and h more

.realistically? How can we know and measure
that.we achieve better realism?

B.

Can we include more of the complexity of a

real, interrelated economy in the model?

we

began!

We didn't know then'that matrix inverses would

be involved or that we would find 200 interrelated
products just as easy to handle (at matrix 1.del,
anyway)
as 20, or 2,000, but now that all seems clear, natural
and inevitable!

Both tplestions have received lots of attention from

applied mathematicians.

PART V:

TWO ECONOMIC INSIGHTS FROM THE MODEL

eo

Since owe can't do many4things at once and'want

to

proceed by small steps, we choose arbitrarily to attack

(B):
What factorsof a complex economy should we include?
The emotional elements like fads look difficult to

16,

Total Demand
In the one-product model (2):

get

D i= a + bP

a handle on. We decide to considel: two competing

produdts: Copying as much of
our successful model ion
Part I as 'we can, we decide to make su-gply and demand

S =c + dP

we might call a the total demand because it is the amount

17

'

la
.

221_

of demand if the product were free(1).= 0) and thus

S* + AS* = D* + AD* =

d(a+Aa)-bc
d-b

the largest conceivable demand.

Suppose the total demand shifts in otir economy from

a to a+Aa, i.e.; thp economy grows and is able to absorb

.more of our product. The shift in -total demand causes

8alculate AS* = AD*.

Explain why its sign is reasonable, based on

economic good sense.

Exercise 12.

Repeat Exercise 11 for

a shift in the equilibrium price'from (see (3a)

P*

a+Aa )-c

a-c

TT

d -,
b

Thus the resulting

a.

the two-product model

b.

the n-product model.

change in equilibrium price is

AP*

(a+Aa)-c

a-c

d -b'

a7F

Aa
UTE'

Free Supply

17.

This change is positive when Aa > 0, as we should expect:,

Again, in (2),

a larger total demand implies larger demand at any price

D = a + bP,

level and thus upward pressure on prices. Our model

agrees with economic common sense.

4ut it lends quantita-:

tive detail to that common sense, too:

the amount of the price increase.

we have predicted

Common sense alone

does not do that.

we c -en -call c the 'free supply or supply in nature because

it is the supply when P = _0.

For most products or

commodities, c > 0 makes no sense because no product can
economic illy be given away for free. 'However, in many
places oil

Exercise 9.

S = c + dP

In the two-product model, let. the total demand change

American frontier ?'n the 1800's, fresh water

was a free commodity in 'plen;iful supply; until redently,

from

road maps were given away fret by gas station owners.

11
2

[44:1 ],

Suppose a product has a free supply c and this

supply changes to c + Ac.,

This causes a change in the

.

causing an equilibrium

rom P* to P.* + 4*.

Calculate

equilibrium price of the product from P* = (a-c)/(d-b) to

AP *.

a-(c+Ac).
Exercise 10.

Repeat

raise 9 for the n-product model.

thus AP* =

Ac

rg.

Exercise 11.

In the one-product model:

The sign of AP* again corresponds to economic intuition:

as the total demand a

as the free supply increases,(Aq > 0), the demand, the

amount of the product people Vint, buy,4ould decrease

changes to a +.Pa there is a change in the equilibrium price, as

we've analyzed above.
supply

There is also a change in the equilibrium

demandlevel from 9* = D* fo

+ AS* = D* + AD*. -Starting

(since more of the product is supplied free) and thus

its'price should decrease:

with (3b)
E* = D* =

da - bc
d - b '

drop.,

we have

222

AP* < 0.' As with total

demand,, we are able to predict the amount of the price

ig,
.47

223

ff%

Exercise 13.
s

Repeat the free supply discussion for the two-product

model:- what change AP* in equilibril price occurs when the free
supply changes from c to c + Ac?

D and S,
Dollars

.6

Exercise 14.

equilibrium

Repeat Exercise 13 for die. n-Rroduct model.

demand
Exercise 15.

As the free supply c changeg'to c + Act the equilibrium

price changes by AP* (above) and the'equilibrium amount changes from
S* - D* to S* +AS* = D* + AD*. Calculate AS* = QD* for
a.

the one-product model

b.

thik two - product model

c.

the n-product model.

S* = 1)*
--,

supply
Price P

price domain

PART VI:

ARE LINEAR FUNCTIONS CRUCIAL TO THE MODEL?'

Figure 3.

Smooth' nonlinear suppy and demand functions.

6

18.

N.1

. A Job for Taylor's Theorem

Of course; it is unrealistic to take supply and

demand as linear functions of a product's price and the
prices of, its competitors.
Yet all our use of linear
algebra--our whole ability to. calculate equilibrium

nonlinear curves D(P) and S(P).

But recall our go4l:
we .want to calculate P*, so we only have to think about

pricesseen to depend on having such linear functions.

P*.

values of P close to P'-*-7.,Probably we know (or'can guess

Qn economic grounds) /14 price Po that is fairly close-to

We could'replaceD(P) by the'tangent line to D(P)

at P0, getting

How can we resolve this dilemma?

First of all, in the one-product model, how might

'more realistic functions D(P)imkn(P) look?
Since the
supply increases and demand 'decreases as prides rite,
we take curves with the 'appropriate monotonicity for
D(P) and S(P). When we put such
curves (choosing them,
as'a first exarupld, to be continuous and differentiable)
into Figure 2:14e arrive at Figu're 3. Both curves have
[Ps,Pdf as domain, as in Sectio

S.

B(P) 4D(P0).+ D'(P0)(P-P0)

(19)

= (D(P0 n PoD'(Po)f + DI(P0)P.

ft

We have written D(P) as a +.1DT above, with constants-5r---."

and b that we can calculate once we know P and D(P)-.
Recall that D,(P) ii'a close approximation to D(P4 for P
close to PoL

From Figure 3 it,'

is clear, that there' is still a un que equilibrium pace

-We can similatly-take the tungen

ine at P
0

1""Ps'Pd]'
af we needed to now_S and a for all P in the full
price domain (Ps,Pdl, we woul be stuck with these

2,24

to '$(P), ,
,

S(P,' = S(P0) + S'(P0)(P-P0),

(20)

= (S(P,0).- 5S'(30)] + 5`'(p0)P

21

\ 22
f-)

225.
a

as a close approximation to S(P) for P near P0.

4 shows these two tangent lines.

Exercise 16.

Substitute a, b, c, d above into (3a) and thus derive

Also calculate, from (3b),

(21).

S'OWD(12

D and S,
:Dollars

6(P1) = S(Y1) 4

Exercise 17.

)S(P )
0

u )-D'(P )
'(P
S,)-01(P
0
0

(For readers who know Newton's Method of approximately.

solving f(x)

0 for a root x given an iniqal guess x

close to the

root.)

Equation (21) clearly has a relationship to Newton's Method.

;

What is that relationship?

Wliat function f is involved?

Figure 4 suggests that P

.better approximation
of P* than our initial approximation Po was. Theory (we
1

is

omit' it here) proves this true if Po is sufficiently

close to P.

0111o

Figure 4.
curves.

Tangent lines approximate the supply and demand

We can of course repeat the process:

taking

as our new guess, we write down equations of the

tangent lines to D(P) and S(P) at P1 and use them to

calculate P2.

Aftgr a few rounds of this, we will get a

very good approximation of P. The method does generaliie
to Multi-product cases.
Equations (19),(20) are a linearized version of the
-nonlinear one-product model.

These equations are exactly

(2), with

So, when D(P) and S(P) are smooth functions of the

price, with more work we can still approximate P* (and

thus, D* = S*),closely. .The crucial assumption about D

-

a = D(P0)

- PolP(P0)5

and S seems now to be that they-change smoothly as ''.P

b = Di(P0)

changes.
c = S(P0) - PoS1(P0)

1(1)0)*

19.

When we solve for-the `price equilibrium of the approximate

.

linearizet equations we get.

S(p0)-D(P0)
P1

PO

It is not crucial that they be linear.

Discontinuous Supply and Demand Curves

And is it realistic to expect that supply and demand

curves will be smooth? Unfortunately, no. The supply
curve, especially, may have jump discontinuities, as
shown in Figure 5.

S.1(P0).7D'(P0)'

For there will be threshold values of P (such as P

This is of,course the price where the tangent -lines cross
in Figure 4, and P1 11-13*.
23

and P

shown) where it becomes economical to open a new

b
factory or put another shift on an assembly line, causing
24

226

227

supply to jump dramatically: However, in this example,

the equilibrium we seek happens to fall in a part of the
price domain, namely (PapPb], where both S(P) and 1(P)

D and S,
Dollars

are smooth; we chn apply bur methods after restricting

the 'pfice domain to (Pa,PbI.
We can draw other examples,
like Figure6, where the method doe not apply.

D(P)

PART VII:
1.

P* - 4 dollars/item,

2.

3.

SOLUTIONS TO EXERCISES
S* = D* =46 dollars.

= 3.2 dollars/item,
a.

S* = D* = 17.2 dollars.

In a plane, two straight lines with unequal slopes always

have exactly one intersection.

The slopes here are b< 0

and d > 0.
Figure 5.

A discontinuous supply curve.

b.

Twq lines with a > 0, b < 0, c < 0, d > 0 yet S*= D* < 0
can be easily drawn:

D and S,
Dollars

D(P)

S(P)

on economic grounds, if there is any market for a product,

1
1

its.demand must"be positive at P = Ps, the minimal price.

In that case

Price P
Ps

110

Pd

D(Pl.= a + bP
S(Ps) = c + dP

Figure'6.

A more difficult case of ismooih

> 0

= 0

supply --anti demand.

25

223

=t>

d'

a -

> O p

4*

ad 2 bc > 0..

26

Since d-b > 0 we have S* = D* = lEtS

0 from (3b).

havel*r.<4.!, .Pd and S* > 0 implies p* > Ps.

Since D* > 0 we

This I Is the same sign as Aa, as we should expect:

i',

when total

dema d goes up (Aa > 0), there is naturally an increased demand

7,

at

:.

4,5.

P * '=

. t
12.
6.

S* = c + d(d-b)

-1

11 price levels, including price P*.

rl

S*7= D* = [2;7.5].

[25 5],

From (llc) we have

(a-c)

S* + AS* = (D* + AD*)

= (d-b)(d-b)-1C + d(d-b)-1(a-c)

= d(d-b)
= d(d-b)

-1

c-- b(d-b)

-1'
-1'
c + dZd-b)
a - d(d-b)
c
and

= d(d-b)

7.

-1

d(d-b)

-1a
'

-1
- b(d-b), c.

= (d-b)

-1

-1

--"'

(a+Aa) - b(d-b)

-1'

S* = D* = d(d-b)

-1'

a -yd-b)

-1
c.

Subtraction gives

AS* ; XD* = d(d-b)-1Aa.

,

4>

4>

(d-b) [d(d-b)-3](d-b)

(a-b) [(d-b)

-1'
d] (d-b).

13,
14.

Equating supply and demand we_get

(d-b)d = d(d-b)

+ b(P* + AP*) = c + Ac + d(P* + AP)

t>

- bd = d

r'N
- db
P*

AP* = (d-b)

Ac).

(a

4> bd = db.
subtraction gives

Since P* = (d -b) -1(a -c),

8.

Handle as in solution to 7, above..

AP*

-b) -1Ac.

41

1,010. The supply equals demand equation is

15.

From (llc),

a + Aa + b(P* +.AP*) = c * d(P + AP*)

S* + AS* = D* + AP*

P* 1-'AP* = (d- b) -1(a + Aa - c).

=$d(d-b)
Since

- b(d-b)

S* = D* = gd-b)
Aa.

---

(c+Ac)

while

P* = (d-b)(a-c),
P* = (d-b)

-1

a - b(d-b)

-1-'-

c.

Thus
.

AS* = 5D* = -b(d-b)

11.

-1-'

Ac.

AS* = AD* = (S* + AS*) - S*

This may be applied for 1, 2 or n.

d(a+Aa)-bc' da-bc
d-b

dAa

d-b

at

230)

231

27
.16

te.

28

17.

We really want to solve S(P) = D(P), i.e.,

D(P) = 0,

S(P)

for P.

Thus f is the supply function minus the demand

function and the usual Newton's Method forimula

= x

x
1

f (x)

r(x)

is exactly (21).

'a

232
S.

29

Return to:
EDC/UMAI
55 Chapel St.
Newton, MA 02160

STUDENT FORM 1

Request for Htlp

Student: If you have trouble with a specific part of this unit, please fill
out 'this form and take it to your instructor for assistance. The information
you give willehelp the author to revise the unit.

Your Name

Unit N

Page
Section
1.11313cr

OR

OR

()Middle

Paragraph.

Text
Problem No.

Q Lower
Description of Difficulty:

Model Exam
'Problem No..,

(Please be specific).

Instructor: -Please indicate your resolution of the difficulty in this box.

(2)Corrected errors in material's.

List corrections here:

I

Gave student better explanation, example, or procedute than in unit.

Give brief outline of your addition here:

Assisted student in acquiring general learning and problem-solving

skills (not using examples from this unit.)
A

233
InstrUctor's Signature
Please use reverse if necessary.

Return el):

EDC/UMAP
55 Chapel St.
Newton, MA 02160

STUDENT FORM 2

Unit Questionnaire
Name
.

Unit Ho.'

Institution
Check the
1.

Date

Course No.

.121)!.ce for each question that comes closest to your personal opinion.

How:useful was the amOunt'Of detail in the unit?

,Not enough detail to understand the unit
Unit would have been clearer with moreAdetail
Appropriate amount of detail
Unit was occasionally too detailed, but this was not distracting
'Too much detail; I was often distracted

.2. /kow helpful "were the problem answers?

Sample solutions were too brief; I could not do the intermediate steps
'Sufficient information was given to solve the problems
Sample solutions were too detailed; I didn't need them
3.

Except for fulfilling the prerequisites, how much did you use other sources (for
example, instructo;t friends, or other books) in order to understand the unit?
'A Lot

4.

Nat at all

A Little

'

How long was this uiit in comparison to the amount of time you generally spend on
a lessdn (lecture and homework assignient) in a typical math or science course?

Much
Longer
5.

Somewhat

Somewhat.,

About

_the Same

Longer

Somewhat
Shorter'

Much
Shorter

Were any of the following parts of the unit confusing'or distracting? (Chea
as many as apply.)

Prerequisites
Statement of skills and concepts (objectives)
Paragraph headings
Examples
Special Assistahpe Supplement (if present)
Other, please explain
6.

Were any of .the following parteof the unit particularly helpful? (Check as many
as apply.)
Prerequisites
'Statement of skills and concepts.(objectiVes)
Examples

Troblems1
ParagEaph headings

TablePf Contents
Special' Assistance Supplement (if present)
ther, please explain

Please describe anything in the -unit'that you did not particularly like.

sr

Please describe anything that you found particularly helpful. (Please use the back of
this sheet if you need more space.).

234

GENERAL EQUILIBRIUM:' A LEaTIEF ECONOMIC MODEL

UNIT 209

by

Philip*M. Tuchinsky
Department of Mathematical Sciences
Ohio Wesleyan University
'Delaware, Ohio 43015

MODULES AND MONOORS IN UNDERGRADUATE

MATHEMATICS AND XI'S

P'LLICATIONS PROJECT

'

TABLE OF CONTENTS
f

111

1.

INTRODUCTION
1.1

1:2

GENERAL EQUILIBRIUM:

-1:3

A LEONTIEF ECONOMIC MODEL

2.

1
1

LEONTIEF'S MODEL
2.1

Ay Philip M. Tuchinsky

One-Product Companies as Input-Output Machines

Consumers and Companies Form a Complex Economy
The Problem: To Balance Supply and Demand

2.2
2.3
2.4
2.5
2.6
2.7

Notation for Production and Final Demand Levels

Leontief's Input-Output Coefficients
, ,,
Concise Summary of the Notation
Equating Supply and Demand
The Model in Matrix Rotation
Solving for the Production Levels
Exercises

.......

2
3
3

"Like Adam Smith and The Wealth of Nations,

3.

HOW TO CALCULATE (I-A)-' EASILY

Marshall and Principles of Economics, and

3.1
3.2

Keynes and The General Theory, Leontief and

3.3
3.4

`Input-Output' are becoming .permanent words

3.5
3.6

In the economics vocabulary.'

4.

Walter Isard and Phyllis Kaniss

Science

4.2

p. 571..

MODELING LABOR IN LEONTIEF'S ECONOMY .

4.1

Vol. 182, Nov. 9,

An Old Acquaintance Returns

How the Series Aids Calculation of (I-A-1
An Example
Why Geometric Series Extends to Matrix Cases
(1-A)-1 Will NaVe Nonnegative Entries ,
Exercises

8
9

9
10

10

The Value of Labor

Labor's Earnings and Consumption are Equal

10
11

5 ABOUT THE MODEL AND ITS USES

12

5.1

5.2
5.3
I

edo/umap/55chai)el st./,newton,mass.02160

235

12
13
13

5.4
5.5
6.

APPLICATIONS OF LINEAR ALGEBRA TO ECONOMICS

Open and Closed Leontief EConomies

Profit and Savings Have Beerl Included
UsiRg Linear Algebra in Economics -- Benefits
and Difficulties
The Model is Widely Used as a Planning Aid
The Model's Great Impact on Economics

15

REFERENCES

15
17

7.

EXERCISES: THE YUGOSLAVIAN ECONOMY IN 1962 AND 1958

18

8.

ANSWERS TO-SOME EXERCISES

22

236

Intermodular DescrIptionSheet: UMAP Unit 209

Title:
Author:
.

GENERAL EQUILIBRIUM: A LEONTIEF ECONOMIC MODEL

MODULES AND MONOGRAPHS IN UNDERGRADUATE

Philip M. Tuchinsky
Department of Mathematical Sciences
Ohio Wesleyan University
Delaware, Ohio 43015

Review Stage/Dte: III (peer reviewed 6 revised)

Classification:. ECONOMICS/ELEM LIN ALG

MATHEMATICS AND ITS APPLICATIONS PROJECT (UMAP)

The goal of UMAP is to develop, through a community of users

and developers, a system of instructional modules in undergraduate
mathematics and its applications which may be used to supplement
existing courses and from which complete courses may eventually
be built.

12/28/77

Computer Projects are natural to this application but not essential.

Realistic data for Yugoslavian economy is included.

The Project is guided by a National Steering Committee of

mathematicians, scientists and educators. UMAP is funded by a
grant from the National Science Foundation to Education
Development Center, Inc., a publicly supported, nonprofit
corporation engaged in educational research in the U.S. and abroad.

Estimated Teaching Time: A totarof 1 hour including discussion of

computer project output.
Prerequisites: Solution of a nonsingular system of linear equations;
matrix multiplication; matrix inverse.
No economic preparation
is assumeli.

PROJECT STAFF

Output Skills:
1.
Define/explain the one-product company as an nput-output
process unit.
2.
Define meaning of entry aii of Leontief 'matrix.
3.
Dilcuss an application leaaing to large sets o(linear
sirailtaneous equations.
4.
Calculatelequitibeium production.levels for a multi - sector
Leontief economy.
-5. Calculate (I-A)l approximately using geometric)
for
matrices.
-1
6.
Explain why (I-A)
will have entries > 0 if matrix A has
.
entries > O.

,/
fr

*/

'

Ross L. Finney
Solomon Garfunkel
Felicia Weitzel
Barbara Kelczewski
Dianne Lally
Paula M. Santill'o
Jack Alexander
Edwina Michener

inevitabtlity of latter in _a linear model.

Convert real economic data (input-output flow by sectors) into
Leontief matrix entries.

Structure of,the Module:' Sections. 1;2 and 5 form a satisfactory

. `applied unit about the basic Leontief open model.
Section 3
is optional;
discussds use of geometric series to approximate
(I -A)-1, which arose in Section 2.
Section 4 can follow Section
2 or 3 and extends the model so that labor costs are included;
it can be omitted also.
4
,

UMAP'editor for this module(

e3.4.,
Coordinator
Associate Director for Administration
Coordinator for Materials Production
Pallect Secretary
Financial Assistant/Secretary
Editorial Consultant
Editorial COhsultant

NATIONAL STEERING COMMITTEE

7.
Dli6ss money as simply one more product in an economy.
.8,
Discuss "economics of scale" vs. "constant returns to scale" and
.

43.

Director

Associate Director/Consortium

W.T. Martin
Steven J. Brams
Llayron Clarkson
James D. Forman
Ernest J. Henley
66nald A. Larson
William F. Lucas,
Frederick Hosteller
Walter E. Sears
George Springer
Arnold A. Strassenburg
Alfred B. Willcox

MIT (Chairman)
New York University
Texas Southern University
Rochester Institute of Technology
University of Houston
SUNY at Buffalo
Cornell University
Harvard University
University of Michigan Press
Indiana University
SUNY at Stony Brook.
Mathematical As'sociation of Atherica

Solomon Gatlibnkel

The Project would like to thank Kenneth R.Rebman for ht's

review and all others who assisted in the prod
ion of this unit.

....Other
Related Units:
a

This material was prepared with. the sup rt of National

Science Foundalion Grant No. SED76-19615.
cbmmendations
expressed are those of the author nd do notnecd5sanily reflect
the views of the NSF nor of the Natrona
rig Committee.

237

(D 1978 EDC Projes; UMAP.

All Right6Resdrued.

238

. so that the dnput products are not really consumed but

1.

are "replaced" on the market by the output products)

INTRODUCTION T.

and the public, considefed as the "final consumers" of

The amount of a.Ooduct demanded by the
finished goods.

With his famous 1941 book (31, Professor Wassify

public is the "final demand" for that product; suppliers

(companies) must satisfy,that final demand in addition

Leontief began the study of the economy as.an input-output

system.

For this work, he received the Nobel Prize in

to providing input matefials to other companies.' For

some products (like steel), Linardemand is almost zero,
while for others (like blenders), final demand is a very

Economics in 1973.

His method has been applied in more than fifty

natibns and international agencies as a predictive tool

large fraction of the tot

In Section 5 we will discuss the

for economic planning.
uses and shortcomings 9f the method; first, let's examine

it in some detail.

The Problem: To Balanc

A sensible economic questipn: how much ofeach product should be produced to closely satisfy the total demand
for the product by all users? That is, how can we match

One-Product Companies as Input-Output Machines

Imagine an economy made up bf companies that each,

make one product.

(We can, at least in,theory, mentally

split up a multi-product company tb satisfy this.) The

manufacture of one, unit of the product requires g known
recipe of input products, commodities And services. For
example, the manufacture of a ton of writing paper re-

ood that will simultaneously

Leonti'ef applied linear algebra to this problem.

We'll look at his simplest model in this paper. From a
kn9wledge of the final demand f f each product (that is,

A-paper-Making company can tie thought

the market basket of all goods th t the public is to buy),

of as,an "input-output machine" that conyerts' arecip

of inputs into a uJIit

economy?" To answer 'this guestionois-lo find a "generp

ay); that is, we seek the
equilibrium" (as economy'
make supply equal demand for them all.

water, capital investment in the form of machiiery, labor,

electricity, etc.

outputs to inputs throughout an elaborately interconnected

production amount for ea,

quires specific amounts of wood fib,r, recyclable paper,.

outbut many times a day.

we can calculate the production a ounts that will supply

that final demand.

will think of all the companies in this way, each with

Some restrict ve economic assumptions

are involved.

.its own recipe for making its one product.

1.2

upply and Demand

1.1

1.3

demand.,

"
Consumers 'and Conihnies 'Bor*Til"-a COMplex ,Econc10

The

companies' are

e
;
4
elaborately intetitIatath, Whe

2.

LEONTIEF'S MODEL

output of a steel company' is an input to a.vistAiiiipe

Of other companies whiCh make autos: applianceh

2Pf Notation for Production andFinal Demand Levels

steef!'

We will look At an economy made up of'n companies,

nutsand bolts, steel alloys, and the thousands of'ottipr,

steel-using products.
turer is a
stered f
direc

input to the manufacture of cloAing:uphol.

rpet, etc., and cloth is also sold
niture

to the Oh ic.

each creating one product, commodity or service from a

The output of a textile manufac

We draw a distinction between

companies (which convert a product into other products

230

filed recipe of input "ingredients.."

We assume that

..prices are constant and known for each produc t; we will

P

'441/..:that the ith company makes x

woducZ.

dollars worth of
i

Let di dollars of this be the "final demand"

-'240

(sales to households) of pibduct i

while the rest of xi
is used as inputs by, other companies.
We take the di,

2.4

as known (thus assuming

The,supply of.prol,V i will be xi dollars. The

demand for product i w ill be'di dollars of
final demand,
plus ailxi dollars (of product i) used in making the
x dollars of product 1,--plus a x
1
dollars (of product

at there is some known mix of

y
products that the public is ready to buy) and
hope to,.

calculatethexvalues
needed to produce a "final demand
i
i

vector" (d1, d2...idn) containing the desired final

amounts of all the products..Xhis
vector is really just
the total public "market basket" of all products
con-sumed, in dollar amounts.
2.2

Equating Supply and Demand

i2

N.

(2)

x. = a

+a 12 x 2 + ... +a x + d
in n

x
il

Leontief's Input-Output Coefficients

used in making x2 dollars of product 2, and

so on.
"supply equals demand" equation for product i is '
1

i)

The

o express the recipe that4lie ith

company converts in o one dolar's worth of its output.
'(That recipe, multiplied through by xi, will yield
xi
dollars of output for the i th company. This is an

We have such an equation for each i.= 1,2,...,n, thus

n equations in all. Our goal, once again, is
to calculate
all the xi from given di and known technological
constants
aij.
The model. provides for the use of each product as
an input to every other product including,ttself; of/)

assumption called "constant returns to scale;" more about

course, many of the a

itlater.)Letaijbe the dollar amount of'product i

2.5

Next, we need

The Model 1n Matrix Notation

that-is used to make one dollar's worth of product j.

Thus a47 = .23 would mean that, to make a dollar's worth
of product 7, we use 23* worth of product 4.
The full

mix of Prodcts 1, 2, 3, ...,n used to make one dollar's

worth of product j is aij dtllars of product 1, a2j of
product 2,

We are ready to switch to matrix notation: put

'di
d
(3)

all

2
a21

1?

22

These numbers form the

column of the matrix A = ,(aij). Since less than a
dollar's worth of inputs are used in making
a dollar's
worth,of the output (or company 5 would be out of busij

will be zero.

ij

th

do

ant

n2

nn

ness), we know that

110101.

a
1=1

ij

< 1 and of course a

ij

Then the n equations of (2) may be compactly written

> 0.

A
(4)

Thus. the columns of A have sums of less than one. (The

rows of A have no coMparable economic interpretation.)
2.3

Concise Summary of the Notation

d
i

A)--Ix

In (5), we know the

nxn matrix I - A and the n-vector J.: Then (5) .is simply
a set of n non - homogenous linear. equations with the
sumed

= amount in dollars of product 1 used in making one

dollar's worth of product j,
3

241

O ',

The problem is now almost solved.

= total dollars produced of product 1,

= total' dollars worth of product 1 that is
by households = "final demand ".

Or
( 5 )

For i and j, each running

x.

A ki
x = Ax + d

wanted xi as the unknowns.

2.6

'

Solving for the Production Levels

You recall that a, set of equations like (5) may

242-

have no solution, exactly one solution,-or infinitely

many solutions.

This array should be read as follows: there is a total fl9w

In our case; although we will not prove

it, 'there must be exactly one solution.

of 70 (hundred-billiondollars) among two "companies," agri-

In fact (I - A) -1

culture and manufacturing, and onp "open sector," households.

must exist for our given matrix A. (This is true for any
matrix where a.,
ii

Ei a.

ij

<

1.)

>
--

We may use the inverse,to solve for

a/

manufacturing production (fer tilizer, machines, etc.) and 8

-.

_.%

( 6)'.

Agriculture uses 4 units of itsown production, 8 units of

0 and the column sums satisfy

(I- A)-1

in (5):,

units of household production (labor), 20 units in all, to

produce 20 units whichbre distributed as follows: 4 to agri-

di

V',

'

411I

culture, 6-tompufaCtuPing, 10 to households.

We have achieved our goal': to produce a market basket t

of final consumer goods, we should ,p4oduce the amolSnts
x given in (6).
.

The input of

6 units of blsehold production (labor) to household consumption

is domestic labor -- the labor of housewives, for example.

Two questions arise at once. Oni is economic: can

the consu
afford to pay fur the slarlvet basket a? Coresumers u wally pay for goods and services by .exchatiging
/

their labor.

The data above is not the Leontief input output array we have
studied, butillNe Can calculate the Leontief matrix from it

440

easily. The recipe

of inputs to agriculture is 4/20 from

agriculture and 8/20 from manufacturing.

Can we fit the cost of labo.r.into the model,

The recipe of inputs.

to'the manufacturing sector is 6/30 from agriculture and 18/30

from manufacturing. Thus the technical matrix and final

whet-1,e has not been mentioned so far? -We'll discuss

thiscn Section 4'

demand vector are

The other question 'is mathematical: in (6) ye are

askedto calculate (J A)-1 for.a matrix that may well

be SOO x SOb or even 10,000 x 10,.,000: we mustinclude
many companies to treat the economy wit

any realism.
there some way to calculate (I,- A)I 1 easily? See

A = (181//;(0)

1/3360]

1.

and t =
.26]

x of the set of linear equations

'

Exercises

b.

Although we have considered individual companies laking specific

(I - A)x' d

Calculate (I - A)

1_,.

and then find f:again fr:om x = (I - A)

from the table in the exeCtise?

sectors of an economy. TA "two- company" fictional exampleis

(Continued in Exercise 5.)

In hundreds of 6illion's of dollars;

d.

c. *How could you have predicted your answer to a. aJ d b.

products like stoves; the'model can 'be applied to broadly-drawn

taken from 181, page 61;

(11

a. /'Using the A and d just above, hand-calculate the solution

Is

Section 3.
2.7

(.121

2.:

in this exe rcise,

alter Exercise 1. so that agriculture,

-let'the flow be:

,manufacturing and households are the three sectors or "companies"

CONSUMPTION

+FROM/TO 4
Agriculture
Manufacturin g
Households
Total

AGR1CULT.
4

20

243

involved, while savings is the open sector.

HO.

TOTAL

10

-20

18

30

20

30

20

10

6,

'8

MFG.'

Each "company"

produces its product (which is still labor in the case of thehouseholds) so as to supply the other two companies and create
t.

a'fInal product called investment, while invested funds, called`

savings, are inpmiked only to the household sector (say, to

build houses. This example, again in hundreds of
billions of
dollars, is 'from 181, page 182:

244

I
1

CONSUMPTION
,

AGRICULT.

+FROM / TO 4'

MFG.

INVESTMENT

HO.

Agriculture
Manufacturing
Households

18

8 `

Savings

17

20

30

.20

12

Total

norms "), but first let's see the usefulness of (8). If

A4, A5, A6 and all the higher power terms are "negligibly

20
30
20

TOTAL

small," then the 4-term partial- sum I + As+ A 2 + A 3 is a

goad approximation of the hard-to-compute matrix inverse

12

(r-A)

needed for (6)':

(The inverse is nasty to compute:

think of the methods of matrix-inversion you know and

82

Convert this data into a 3-company Leohtief model by f.indine

Consider applying them to a 30 x 30 or 2000*x 2000 xatrix

A and d by the method explained in Exercise 1.

I-A.)

Predict g from the table above without any use of the

b.

intended level of this paper because it requires "matrix

In fact, a partial sum of quite a few terms

from (8) is cheap and convenient to compute by comparison

leontief model.

to direct computation of (I-A)-1.

c.

Solve (I

A); = CI for ..;C:

Show your calculations'tn4etall,

An Example*

3.3\
0airtaate-iI

A) -1 -anct:ttretr-get-X -from x

Just to see how the calculation goese put

Show your,calculations.1
Answers to b, c, d shoUld alt by the same.

.2

Thig exercise is
A =

ccptinued In Exercise .6.

3.

HOW TO CALCULATE- (I

A)

- 1

EASILY

.2

.2

.1

so that I -A =

%1

1.1392
(I-A)-1

A as thcligh it were a single number (say a) 'and of I

4-a
1-should make you think of" -- geometric serie,p!7-You recall

the geometric series fOrmula

I+A =

+ a + a2 t a3 +

(if 4a1 <

17

(8)

I.+ A + A21. + A30+

.1

1.2
0

1.1

.007
0

A3 =

.010

= (1 - A)-1,
A

.0017,

4..

We'll consider a plausibility argument for (8)

+A +A 4

.1392

.7-

'

=2

/0005

1.13
0
.24

r thus

.0017

0
.2524 .

246

.283

1.2496
.0612

.12621
0
1.1387

.171

.04

14.13

.25

.0012
0

:26
1.24

1.137
3

thus I+A+A +A =

.007

shortly (an ironclad proofisjust a little beyond the

(

thqs,I+A+A =

.03

E1.1387

.005
.0016
:0052

0
.0024

I +OA

.02

.018
.008
.016

3,2 How the SeTiesAids Calculation of (IA)-1

-245

0
1

.06
.04
.04

0
.04

and "column sums

a complete analogy to (7)...

.2

..03

1).

1" take the place of I al < 1 and it is true that

1.1
.2

'

For our matrix.A, a0e.mondilionsai. >

E.1

.1266

You shoUilcheck all the calculations here. Use an

electronic calculator.
Let's look at some partial sums
of the geometric series:

A
1

.9

Then (I -A) -1 becomes analogous to

and

(7)

'-:1

-1

In the back of your mind, you should think of the matrix

as though it were 1.

.8

.2848
1.25
.0633

.2532

There is an elegant way to calculate (I-A)

-.2

0
-.2

to four decimal places,

and,

An Old AcquaintariCe Returns

,3./

.9

.125
0
.0'56 1.137
.278

1.248

./1

This five-term partial sum'is convincingly close to

-1
(I-A)
This example was fabricated so that the

the multiplication A-A = A-,.and so on.) Then their sum

I
+ A + A +
(I-A)-1 'als.o has non negative entries.

series would converge within a few terms: entries like

3.6

.1 and .2 bec..mm.e_ rapidly smaller when multiplied by one

anothe,in matrix products.

However, in a large matrix A

the eak'ies would mostly be small and many would he zero.
Remember, all al
are > 0 and the column sums are less
than one.
The geometric series is a practical way to

Using A =

3.

'

(.3
.6

.4 1

.4)

a.

approximate. (I-A)

Exercises

'

show that (I - A)-1

5/2

b.

-1

10/31

Write and run a'short computer program that calculates and

I

p r i n t s

A 2

3.4'' Why Geometric Series Extends'to Matrix Cases

etc.

Print part;a1 sums until you have (1(- A)-1 well

approximated.

A plansibility argument for the truth of (8) was

This matrix calculation closely mimics
the
usual proof of the scalar case (7): notice that, for

3c

promised.

c.

any finite partial sum,

This will take quite "few Items.

How many terms must you include in the partial sum in b.

before you have approximated (I - A)-1'within .5 in each

entry? Within .05?

(9)

(I + A + A2 +

+ Ak'l) (I -A) -I - Ak.

4.

All 9ther`term cancel out. For matrices like' our A

with smalrpOsitive entries, the powers A k
approach the

lim A

Verify the matrix calculatio,9,..ln (5).

State each law of Matrix

clgebra you use (e.g., the "left di'stributiye. law").

15.

zero n x n matrix 0 as k increases,

because products
of small%positive numbers get smaller. To say

Within .005?

'

(Exercise 1, continued)

Foethe matrix'A and demand vector d

of Exercise 1, calculate fly computer successive approximate
solutions

= 0
(I + A)

means that all n

of the matrix entries approach

as.k increases, and this is true 'for
the matrices we'
are studying. Now let +co in (9):

(I + A + A2) CI

(I + A + A2 + A3) a
etc.

These will converge slowly to your solution x in Exercise 1.

$,,

(I+ A + A2 + A3 + ...)(l- A) =I- 0 = I.

But this exactly says that ,(I-A)

3,S

-1

6
=

+ A + A2 + A

(I-A)-1.Will Have Nonnegative Entries

'14

(Exercise 2, continued)

For the matrix A and'demand vector

' d of Exercise 2, Calculate
successive approxiMations of x by
-1
using partial sums of the series for (I
A)
.

From (8) we can conclude that all the entries of

(I-A)-1 will be > 0. (This means that'negative
production
levels xy cannot arise in (6), which
comforting: we
would throw away a model that, failed to yielld
all the
x. > 0.) To see that (I-A) :1
cannot have negative entries,
simply recall that a.. > 0 for all i,j.
Thus I, A, A 2
13
,

3
,

... all contain entries that are > 0 (think about

247

4.

4.1

MODELING LABOR IN LEONTIEF'S ECONOMY

The Value of-Labor

Now let's turn to the economic question we raised in

Section .2.6:
can the public contribute enough labor to
the economy to pay for
thefinal-demand market basket d
it'has ordered? It'is easy to calculate the
value of

11,

248

10

labor in,our economic model.

14'

To'make one dollar's worth

The proof involves more matrix algebra,

involves a,1

dollars of

tice tpat,"by inttoducing an n-vector contain4ng'all,ones,

product 1, a2j dollars of product 2, ...,.and anj dollars

of product n; in all the dollar's worth of product j
contains

-.,

a2i

+ anj <

we can write the right side as a matrix prodgtf

d

+d 2

=,(1,1,. .,1)

dn

dollars worth of input materials made by the n companies.

The maximum amoit that can be paid for labor is
a

03

a..
tj

i=1

are made,,labox receives aojxj dollars in pay.

Thus the

(13)

(a01,a02,...00n)

(1,1.7.,1) -

(1,1,1,...,1) A = u

(I-A) t
/1

The_(1,1,...,1) A term here gives the dolumn gums that ,

'appear in (10).
Now
/
s
(1

a0

(a01,. ,a0n)

=,

(I- A)

'/

x-

lz

ao, xi

(a01' a02, a03'

jEl

"" aft)

x.2

1.1

3
4,=

xn

4.2

=u

dnJ

total economy-wide earnings of labor are at most

.1.,

1Now we can write the n equations of (10) cOmpaOtly als

dollars per dollar's worth of product j that is manufactured.

The new constant a
(for j
1, 2,
n) are labor's
Maximal slide of the/pie. When x dollars of product j

Here a

d
d

=1

1:.

u = (1,1,1,...,1)

alj

First, no-,

of the 3jth product, we recall

denotes the row-vector (a

(a01, a02,

5.

a0n)''

Labor's Earnings and Consumption are Equal

5.1

u,'

(I -A) il:T) 1
cancels .
C

c1

f 4* d; +

+d

-"t

ABOUT THE MODELAND iTS USES

Open. SndClesed Leontief.aconomiet

The total worth of the final demarikOvector'd is

d

+ d

+ d

The, model we bave.looked at is knoWn as Leon

n dollars.

Thus the final demand vector

is feasib/0((can be paid for by the public) if

(12)

x > d.+ d
1

+.

and public.

dn

a0

AP
We will now prove that equality must \told in (12),
x = di + d2 +
+ dn., if we use production levels

x calculated from the Leontief model,from (6), and pay

labor its.maximal earnings, the aoj from (10).

ef's

open model because of the eprate-treatment of ompanie

In a closed' model, 'theepublfc (or

abor

force or households) istreated as one more ompany to

,
-which the k2put recipe is the market-basket dtwhile the
output is the tabor rngredient in the input recipe elf mope

traditional companies.

As we have-just seen, 44re ditalar-

."

worth of inputs to the household sector Will equal the

We will

dollar-worth of its output (labor) in the same way that-

be proving that labor's earnings exactly payfor the

"market basket" that households consume.
This turns
out to be true because we have build "conservation of

the inputs of-goods and labor /(o a manufIcturer equal

the value of its output.

equivalent.

value" into the model: the value of output is equal'to

The open and Lased models are

The distinc,tion between "final consumption

goods" in our open modei and inputs that the household

the value of input products and labor if we use (10).

11

/20

12

sector "processes" into an outp'ut product canoed labor

can hOpe to actually know. Other models that attempt

in a closed ilioderis of economic,interest, but makes no

mathematical difference:
5.2

to equate supply and demand (i.e., to study general equilibrium) tend to )e so theoretical that no useful numbers
__can be calculated from them; one can instead use them-to

Profit and Savings Have Been Included

prove that one or more general equilibria must exist!

We have emphasized so strongly the equal value of

the inputs and outputs of each company that you may

In fact, several Leontief models have been fully researched

and are in use as planning delfices.

wonder how a company can make any profit. In fact,

profit is one of the input ingredients to'each company.

But there is a price paid for the use of linear al:

gebra; the models are subject to a key criticism.
We

One of the products or commodities that flows through

the economy we have modeled is money: The paper manu-

have assumed "constant returns to scale," as economists

say.
This means that, if a specific recipe of inputs

facturer mentioned at the beginning.of this paper

really receives a few pennies of money along with the
physical inputs (like wood
and labor-tiMe in
exchange for the dollar's '4,orth
North of output (paper) made
froM these inputs. The public receives some money as

cost-per-unit-produced by enlarging their production.

For example, once an assembly line has been purchased and
installed, it can be used for one, two or three eight-hour
shifty daills.
When used for three shifts, the capital

'part of its market'basket -- this is "avings. Money

is simply one of the n 'products "manufactuied" by n compariips: one company in this economy is a,commercial bank
Certainly the role played by money is unrealistically,

investment in themachinery is spread over three times

more output than is the case If Tne shift is used. The
input-of capital to any one unit of production is much

simplified -- we have not built an investment or credit

..sCructure into -the model. That can be done, however.

less when the machines are used to capacity.

(There are
extra exiienses involved in running machinery a'roupd the
clock -- repair and mdintenance expenses, extra pay for

This model is only concerned with tile Complex

_fl

to

of goods among the companies and consumer/labor sec

of the economy. No risk is modeled.-- each
compan'

nightwork done on

ig4 ht shift, etc.

-- but these expenses are
easily ot,ercome hy the three-to. ne savings.)
It is
generally less costly (per unit of produCtion),to.mass-

knows how much of its product it can sell to the public;

prices do not change. We are modeling the distribution

process of the economy, n9t'Its other aspects,

,
,

5.3

e.

makes one dollar''s worth of output for a.given company.

then N copies of that recipe will make exactlyN dollar's
worth of output.
In'reality, companies can redUce the

Using Lineay,Aliebra in Economics -- Benefits

and Difficulties

--

Leontief hSs chosen linear algebra as his' mathemati cal tool.

He benefits from that -- to find x in terms

produce more of any itroduct than less; that is, there-are

"economies of scale." This phenomenon is an important
reason for the clear tendency toward large corporations in

our economy.

. Linear equations like (5) cannot deal with economie40s

of scale.
Inde'ed, doubling d'in (5) leads to a new 'solu,
tion x that is double the old x o "Constant returns to

we-simplysolve a-llarte-set-of-Tinear equatiens,

which we know how to do. The great contribution of
t

Leontief's models is that they permit actual calculation

of-general equilibria in terms of input data (the techarological constants
17 and final demands d.) which we

scale" isan inevitable assumption if linear algebya's

calculation advantages are to be exploited,

25.1

13

'p"

25

14-

Theuseofconstanttechnologicalciata,theaihas
j,

also been widely criticized.

The input-output process in

each company is excessively rigid in the model.
In reality,
a furniture manufacturer might very casually switch from

WASSILY W. LEONINE was born\n Leningrad in 1906.

He fled C9mmunist rule ip Russia in the early 1920s

one upholstery cloth to another.

However, that amounts

to creating a whole new economy in our model! The recipe

with his family.

for the furnitur04 maker must be alteree.(changing a column

of A) and new production levels must be calculated for all

to 1931 he was economic advi5or to the Chinese

companies.

of Economic Research in New York.

This is another prize for the use of linear

algebra -- all the companies are rigidly interconnected.
5.4

At the age of 22, he completed a

doctorate at the Unive;sity of Berlin.

Froth 1929

government; in 1931 heSjoined the National Bureau

His pain input-

output methodology matured during the '30s.

He was
chief of the Russian Economic Subdivision of the

The Model is Widely Used as a Planning Aid

Offlice of Strategic Services ddring World War II.

When a nation, a region or a city needs to know

theimpact that alter ative development projects -- a

Lecntief has been a professor at Harvard since

1946.

Sources (9) and (10).

steel mill, a cultural center, an auto assembly line, a food

processing plant -- will have if built, input-output
analysis is of great help.

The model can predict the flow

of.goods and services, including transportation needs,
new employment and pollution problems (such factors may be
added to the model we haye discussed) and point to serious
Alaortfalls or Oversupplies in tie current economy. ,Its

e World Bank use Leontief

The Bureau of Labor

atistics of to U.S. fed-

eral government hgs been a ma %r sponsor of Leontief's research and employs a massiv
Government agencies of mo
including the Scandin

odel of the U.S. economy.

than fifty other countries,

5.5

-- one product can sometimes be substituted for another

in our economy; Leontief does not include this possibility in his models.
Aside from planning and predictive uses, Isard and

an nations, Western Europe, East-

Kaniss report-a major impact upon economics.

ern Europe, the USSR and Many developing nations use such
models.

-- constant coefficients in the matrix A make the "recipes"

of inputs used by companies inflexible;

-- the extensive and precise data needed for the model

is often unavailable, "borrowed" from another region
or nation, etc, This has been a problem in developing
nations, especially.

crucial insight into a very complex problem.

The United Nations and

Such

errors an arise, they,point out, in these key ways:

-- the effects of inevitable changes in technology are

gnat included;

answers are only approximate, of course, but the, -give

models.

the method has been used by inexperienced planners.

Since the

4del requires d'011ifiete, consistent data,.it has forced

many nations to take economic data gathering more 'seriou'sly.

Uniform definitions of products and sectors of an economy

The Model's Great Impact on Economics

and uniform accounting procedures have been needed; thus

In Sdience magazine, Walter IsarXand Phyllis

planning and data collection agencies in many nations

Kaniss (10) reviewed Leontief's contributions at the time

have coordinated their programs.

10:

Hof his winning the Nobel Prize.

They highlight the power

Much easier comparative

study of related national economies has resulted.

of input-output analysis for planning, but concede that

Writing in Newsweek [9),

the model's predictipns have contained large errors'when

Paul Samuelson (himself a

famous doctoral student of Leontief's at Harvard) ment

253

45

254,

16

these uses of input-output analysis:

The author

As the Vietnam War wound down, Leontief predicted the

results of the shift of a billion dollars in gross
national product from War to peacetime production.
He conclided that there would be- an expansion in

--

mission to draw exercises and data from this source:,

S.

Leontief dicovered that exports from the United

States are more labor intensive than our imports,k'
confounding those who decry thee use of "cheap foreign
His conclulabor" as a source ofunemployMent here.
sion is that th net result of importation and'exporttation is to in1\rease use of U.S. labor.

found the three magazine articles listed below to be

(There are
particularly understandable and horthvhile.
I

The U.S. Congress discovered the greats impact of

steel-price raises on inflation in the Unitgd States.

--

Andrei Rogers, Matrix Methods in Urban and Regional

Analysis, Holden-Day, San Francisco, 1971, pp. 59-77.

employmeri.
--

wishes to thank Holden-Day, Inc., for per-

many articles'by and about, Leontief in periodicals that

almost all college librarie:=. will-have.

Look up "Leontief"

in the Reader's Guide to Periodical Literatyre.)

"Nobel Laureate Leontief," Paul Samuelson, Newsweek,
9.
Vol. 82, Nov. 5, 1973, p. 94.

6.

10.

46

REFERENCES
v-

" Walter

Isard and Phyllis Kanisc, Sciamce,_Vol-18_2, Nov.

'

9,

1973, pp. 568-591.

Advanced References:
1. 1.

"The 1973 Nobel Prize for Economic Scienc

'Input-Output Economics

R.G.D. Allen, Mathematical Economics, 2nd ed.,

Scientific American,

St. Martin's Press, New York, 1959.

' Wassily W. Leontief,

ctober, 1951.

David, Gale, Theory of Linear Economic Models,

2.

7.

Prentice-Hall, Englewood Cliffs, N.J.,=1961

Exercises:

Wassily W. Leontief, The Structure of the American

3.

Th6bYugoslavian Economy,

in 1962 and 1958'

Economy, 1919-1929, Har?ard University Press,

7.

Cambridge, Mass., 1941.

Wassily W. Leontief, The Structure of the American

4.

Economy, 1919-1929, 2nd

d., Oxford University Press,

Wassily W. LeAtief, Input-Output Economics, Oxfoyd

.

in column 16 to represent a single open sector.

Ben Noble, "Application of Matrices to Economic Models_

input-output matrix A is given also.

and Social Science Relationships," a lecture in

d as in Exercise 1.

Proceedings,..Summer Conference for College Teachers

Your assignment, should you choose to accept it:

on Applied Matematics, University of Missouri - Rolla,

..

a.

Among many elementary presentations of Lepntief models the

You All have to construct

k

Use a standard linear-equations solving program,' arready

b.' Write a linear-equations solving progrS'm that, say, uses

A.C. Ciang, Fundamental Methods of Mathematical

Gauss-elimination, to solve the equations (I -A) x = d for

Economics, McGraw Hill, New York, 1967.

this model. (This is a fairly large project.)

17

255

The

available for your computer, to find the production vector

x for this model.

author's favorite is:

through 8.

A variety of open sectors are given in columns 1'0-14; use the

1971, published by C.U.P.M., Berkeley, 1973, pp. 111-117.

7.

The data is reproduced,

by permission of Holden-Day, Inc.The closed sectors or

total

University Press, Fairlawn, N.J 1966.

6.

page 69 ffc there is given an eight-"company" model

"companies" are given in rows/columns numbered

Fairlawn, Ne.J., 1951.

5.

In [8]

of the Yugos,lavian, economy as of T962.

C.

Have the computer print out successive approximate solutions

ov,

.25c

1'

for this model, as required in prOblem 6.'

Convergence wil l

Tables for Exercises 7 and 8, reproduced from NJ, by permission,

of golden-Day,'Inc.

not be inmldiate but will occur by about the twentieth round.

8.

Consolidate the data in the tables used in Exercise 7 so that the

production and consumption "industries'! are.
1.

Input (WM Table fog the I uppttarbin Economy 1961 (In Million: of Owns)
htanu

Desionalrun

"manufacturing," made up of old manufacturing (1) and

"agriculture," vide up'of the old agriculture (2)

Orlon
1.64duulasliams
2 Agrisuliute
3 I ',vestry
4 Construstton
S Transpori and .

and fOrestr;

ReepatExercise 7 for this consolidated model.

Compar4Z.X0 the results of Exercise 7.

Comparable data (to that of-1962 used in Exercises 7 and 8) for

1958 appear on p.21. You should regai'd rows/columns 1-8 as the
"companies" and subtotals in column 14S3-4.14 single open sector,
as-in Exercise 7:

3.566
5)0
1.235

6.656
137.391

220
26,189

39,900
20,508

32.946
6.407

466

166.311

11,114
14.292

2.453

31.624
39.256

9.028

959

237

130

2.423.135

641 .106

54.222-

149.(.66

4ei'.748

42.677
525.599

11.459
60.257

1,060.309

203.281

30,673

4.036.258 1.412.663

116.610

CM

746

10 Osprwalion
Personal !maims:

12 As untolation

isamast

13 Subtotal (9 121
14 IAA fCJN: tn./

8.939
2,100

Others

37.904
3.997
1,299

11

9 Sibtutal 4101

43.704

Subtotal
(1 -8)

9 9.

9,326 2.397,665

31113

44
370
403

12.299
5.579

10.714

894

207,375
130,783

7,069
2.849

1,613

614
277

68.406
46.290

8,561
1.063.-s

'

76
703

1.25'3

859

378

760,834
88.759
196.447

469.021

193,833

74.009

58.484

12.749 3.880.'59

21,300

16,112
120.266

2,096

173,067

54.785
94.313

2,106
300,200
16,090 1,430,961

202.9.60

134.187

340,115

34.764

37,$7V2.045.179

866.34

477.1111

551.122

133.965

68,815 7.662.899

38,621

....,

'

16 7owl (11 ti)

a. 'Calculate the appropriate matrix A and final demand, vector

A

'

118.369

1.522

71.643

Slost,
I5 Imports

,9

52

and

(falls

253.527

16.086

6 Thine

II
9.

230.1941

/....,8 Others
8,

Sawa

Trade

4,5114

.79,122

../.1/7 Sersices and

Crafts

Tganipen sector is. the subtotal row/column-16 used in

81 371

1,849.873

Communisal ..... s

ser4kces, made up of the'bid sectors (5), (6),(7), (8).

Exercise 7.

1 (3) ;

3.

and
cahoots

construction (4);
2.

Transport

Con

struchon conununo

fasturatig

37.598

.2.090

146.113

1.122

4.716 412 1,596.394

119,822

.,

1).9'13
669,171

20.247

866,348

9.588

497,365

551,122

133,965

51,671
845,261

78.403 8,539.931

Input Output Table for the Yugoslavian Economy 1961 (Continued)

Q.
s.

b.

c.

Solve

Destination

he linear equations (I-A)7( = d.

........

Ns

ApprolOmate x by thing successive partial sums of the,series

se to
Gross
Stos.1,
Insestment

eonsumptton

J.
Expoits

r higin,
10

. .

for (I -A) -1, asrequirgd in Exercise 5. #

1

hlanufaoiirini

AgIliulture

forest'',

'

1782252
6
9

4 Constiuction
5

Transport and'-......)

Conunu nilations
Trade

Services and

Crafts
8. Others

9 Subtotal (1.10

681

8,835

655,773

9103

. 6.733
39,485

2,645..

196,945

Consutnpttsin

13

552,553
72.637.

3.501

26.4

General

Consurnoton

1;

II

' 463.62d

6.106
1,204

Personal

-..--

1.168,940

109.758.
51.571

'
1,008
796.742

960,817
731,979
20.423

14

15

162.802

24,0*
1,124

150,151

19,1

306,870

10,694
10,190

Ata,

2,231.814

v..,

16

1.123.619
756,005
21,547
23.748

23,748
:'s

Subtotal
-(10:14)

Total Consuntpoon
(13-14)

2,318,747
835.560
31.063
679,901
.

169,998
.323.177

Total
Output
(9+16)
17

1.596.30
119.822
866.348

289,990
420,339

497.365
551,122
133,965

11.626
20.651

61,710

65,559

30.841
1.----......

12.113

711.403

278,834

2,510,645

4,673,272

8,559,831

'

Za..,J ra Stutot hem (1966j 'Itautobno Odndu Provrednoh Detainetio Juandavoe u 1992 GodonCrIntersndustry
turns of the 1 usorlas k onum) on 1902") Beograd

0 3920 0 0510' 0 0313 0:2926 0.2310 0 06118 0,3262 0 1190'

0.0489 0 3277 0.0298
0
11,0002 0.061 0 0003
0
0 0168 0 0003 0 0046 0'0077 0.0004 0.0024 0.0028 0.0010-

As 0 004 0 0008 0.0103 0.1516 0.0527 0 0036 0.0030 0.0090

0 0225

25

'

4,716.412

0 0071, 0 0205',0.0461 0.0662 0.0221 0.0094 0.0110

0 0152' 0 0090 0.0062 0 0237 0.0129 0.0101 0,0800 0.0114
0 0067 0 0057-0-0080 0 0103 0.0171 '0.0128 0.0120 0.0078
0 0083 0 0001 0 0011 0 0024 0.0021 0.0052 0.0028 0 0035
litgramaporl by Ervin lett

Technical Ceeffickat Alteryix for thejagleloman Economy

Data for Exercise 9,

reproduced from m18], pp. 73-74 by
permission of Holden-Day, Inc.

8.

(J4

I.

The 1958 Yugoslavian Economy

Transport

Destinaiiiin

Mani,.
factunng

Agriculture

forestry

Origin

I. Manufactut
2. Agriculture

299

47,216
33,143

3.654
7,680

1.393

992

104,452

85.882

4,222

50
195

86,083

2.316

1.320

6
498

13.481
16,428

2,635

301.

93,935
63,040

73

79

7.102

88

56.352

2.829

414

370

.1.849

662

1,379,403

337.640

24.183

191.970

113.764

40.901

101,306

13,362

20,729
388,407

1.971
45,332

7.933
55,435

34,500
41,358

7.24,9

il'54,695

4,391
56,380

77,250

28,967

53,622

50.085

139,197

45,267

21;'676"1.046.196

2,238,915

824,086

100.453

307,960

241,707

242,042

-207,344

38,392 4,260,899

1,851
443.691

91.566

683

30

8.086

630,132
...___.

8,208

9,"38

1.158

4.912

84,275
40

1,349
8,446

915,652

101 136

Increase in
Gross
Stocks
Insestmen
10

Exports

12

119,406
10.549

4. Construction
S. Transport and

B. Subtotal (1-8).,

276.257
70.250
7,544

4,946

$8,688
25,628

1,986

18.989

2,301

281,797
5,953
1,113
272,000

1,010 ,

7. Services and
1
Crafts,
S. Others

242.012

2.558

'20',344

546,614

135,732

.2,

.2

.4

.6

.4

.2

.15
.05
4-.2

d ='

,.

2
b.

"Product on totals" predict x =

.8

-.2

- .4

.4

-.05
.8

aI

sumptson

sumption
-14)

13

14

15

General

Con

Total Con

438,903
419,447
17,830

132,724

8,625
1.168
19.136

428,0%1
18.998.

74,599%

16.615

120,135

30]
[20

={7313;:i
2

12,294_

Subtotal
(10-14)

Total
Output
(9 +16)

16

17

1,249,08
514,824

2.684.457
915,652

19,1361

27,655
291,562

161,136
307,990

91,214
132.429

115.85
179,002

<242,042

'15.810
6,238

118,184
10,397

123.069
11,119

207,344
40,950

584,768

441.61$

1,177,447

212,610

1.390.057

2.552.176

4.749.364

7rednih Ditipn'usts iuso.10.4.

30
20

as expected.'

1 e arithmetic isnasty if a methodical Gaussian approach

is used, but easy If one tiptoes through .the equations
using a little fc4esighi.

I
11.

249.793

102.374
4.159

Mr). Beograd.

3/20
1/20
4/20

40.950 4,749.364

242

'Some: Sewn, Zara to &animal, (1962). "hiedusobni Odnesi

ot th4 Yer1111 E4641C4117

and

c as the solution x =

Persona:Consumption

426

2,584

6/30
18/30
6/30

151,444
806,071

%-

'

480 o'
4

A =

a,

249 793

307,993

Origin

6, Trade

the table give us x, as we should'

20

Consumption

Communication*

4/20
8/20
8/20

"

2.

29.831

InpurOutpur Table for he Yugoslavian Economs 1958 (Continued)

3. Forestry

results.

expect from the definition of x.

1,851

2.6114.457

I. Manufaciunng
2. Agriculture

[2300)

The "production.totals" in

6.021 2,197,188

4'

Niko=

x =

The *same solution

"

19.843,

156,018

16 Tout (13-15)

has solution

11

Precisely.

(I-AT1)= (552 ;0%61)

5.166 1.435,370
10
400,828

j168

1427

b.

Subtotal

21,919'

10,437

26,

Sto.14
Imports

Others

Crafts

14, I/wax in
IS

and

15.685
10,845

10 13epreetation
1f Personal Income

II Subtotal (9-121'

199

50,030
4.922

9.5ub4ul (1-8)

12 Accumtilation
(saingsl

1,485
5,377
15.462

123,354

Trade

=
SerRa

and

49.133
268,853
1,019

1.081.250

.
3. Forestiy
4-Gonstruction
S. Transport and
Communications
6. Trade
7 Semites and ,
Crafts
8 Others

Con

strut:lion CommuOi
cattuns

(-1 -1

input-01door Table for the Yugoslavian Economy 1958 (ot Minions of Dinars)

ANSWERS TO SOME EXERCISES

19240941191-

Interindustryltelatsows

Abe
I

.260

21

259

22
4

3.

b.

About 50 iterations are needed to get noticeable convergence.

5.

Reproduction

of computerl results are Just below, giving the

matrix,sums and results after multiplication by

Res'ults:

.THE 5 TERM APPROXIMATION IS

12.838399 *
1.458999]
11.091999
2.110399

THE 5 TERM APPAIXIMATION IS

(1.473599
8.569599)
2.612799
1039199

d:

AND LEADS.TO OUTPUTS

x =

117.014399j
21.843199

THE 10 TERM APPROXIMATION IS

2.447314
3.979849
1.835486
2.756191
THE 15 TERM APPROXIMATION IS
4.518542
2.915179
'
2.186384
3.060952
THE 20 TERM APPROXIMATION IS
3.135986
4.772777'
2.351990
3.204784
THE,2? TERM APP OXIMRTION
4.892762
\3.240196
2.430147
'''333.272664

THE 10 TERM APPROXIMATION IS

.1.621937
0.772231
1.544463
3.166400

AND LEADS TO OUTPUTS

19.308298
28.110235

THE 15 TERM APPROXIMATION IS

1.656303
0.819177
1.638354
3.294658

AND LEADS TO OUTPUTS

19.839746
29.562180

THE 20 TERM APPROXIMATION IS

AND LEADS TO OUTPUTS

19.962872
29.898565

e830053

1:664265
1.660107

3.324373'

IS

THE 25 TERM APPROXIMATI.ON. IS

1.666110
1.6461.6

TIE 30 TERM APPROXIMATION IS

4.949389
3.289377,
2.467033
3.304700

AND LEADS TO OUTPUTS

19.991398

0.832573
3.331257

29.876499

THE30 TERM APPROXIMATION IS

1.666537
1.666314

THE 35 TERM APPROXIMATION IS

4.976114
'' 34312588
2:.484441
3.319820
THE 40 TERM APPROXIMATION IS
4.988/27
x.323542
2.492657
3.326955

'No

.'THE 45 TERM APPROXIMATION IS

'4.994679
3.328712
'2.496534
3 330323
THE, 50 TERM APPROXIMATION IS
4.997489
3.331152
'2.498364
3:331912

AND LEADS TO OUTPUTS

0.833157

3.33282

19.99800-,

29.994555

THE 35 TERM APPROXIMATION 15

1.666636
0.833292
1.666585.
3.333221

AND LEADS TO OUTPUTS

19.999538
29.998738

THE 40 TERM APPROXIMATION IS

1.666659 0.833323
1.666647
1.333307.

AND LEADS TO OUTPUTS

19.999892

THE 45 TERM APPROXIMATION IS

1.666665 ,
0.833331
1.666662
3.333327.

AND LEADS TO dUTPUTS

THE 50 TERM APPROXIMATION IS

(1.666666
0.833332) A
1.666665
3:333331

AND LEADS TO OUTPUTS

(19.999993)
x

2.999707

19.999974,
?9)999931

..

261

29..999983

262

-p

24
231

6.

Reproduct*n of computer printbuts of successive matrix

7.

approximations and the x they yield from multiplication by d:

THE 5 TERM APPROXIMATION IS
[1.633799
0.690499
1.266199
2.699099
0.323299
0.979199
0.833999
1.432899
,

ANO LEAOS TO OUTPUTS

17.811995
25.381104

fp 17.102386
ANO LEAOS TO OUTPUTS
19.184608
28.278700
18.920162

THE 20 TERM APPROXIMATION IS

2.125927
1.286879
0.477001'
2.304429
0.675213.
3.959032
1.631108
1.623539
1.654602

ANO LEAOS TO OUTPUTS

19.696132
29.358532
10.597582

THE 25 TERM APPROXIMATION IS

2.142771
1.307308
0.482716
2.339987
4.002158
0.687277
1.653415
1.650594
1.662170

ANO LEAOS TO OUTPUTS

194886759
29:760947
19.850033

THE 30 TERM APPROXIMATION IS

2.149048
1.314921
0.484845
2.353219
4.018230
0.691773
1.661728
1.660677
,1.664991

ANO LEAOS TO OUTPUTS

THE 35 TERM APPROXIMATION IS

2.151388
1.317759
0.485639
2.358177
4.0242)9
0.6934149
1.664826
1.664434
1.666042

ANO LEADS TO OUTPUTS

Total Output alter

Manufacturing

Agriculture

Forestry

4. Construction
5. Tyansport and Communications
6 Trade
7.

Services and Crafts

8. Others
Total

Total Output

Final Demand

20th Round

1st Round

2nd Round

3rd Round

10:h Round

2,318.747
835,560
31.063
679,901
289.990
420,339
65,559
32,113

1c91,c9e

4.201.567
1.418,674
100,006
848,359

4,483,704

4.115.489
1.596,030

4,673,272

6,775,199

1,226.465

76.903
814,776
409.701

492930
104,519
56,310

458,548
524,941
120,686

68,182

7,742,963

1.512.349
110,812
859,071
479,915

539,248
127,933
73,751

8,186,783

119,786
866,321
497,296
551,074
133,941
78,385

8,558,322

(Reproduced by permission of Holden-Oay, inc.)

19.997815
/9.995388

19.99710

264

29.998281
19.998921

26
25

133,965
78,403

8,559,831

ANO LEADS TO OUTPUTS

19.994138
29.987627
19.992238

AND LEADS TO OUTPUTS

19.999185

4,7i .412
1,5
,394
119,822
866,148
497,365
551,122

Programmed by Ervin Bell

19.984272
29.966800
19.979172

THE 50 TERM APPROXIMATION IS

2.152705
1.319357
0.486086
2.360959
4.027593
0.694392

263

Yugoslavia, l962

19.957799'

ANO LEADS TO OUTPUTS

1,.666634

have not

29.910913
19.944112

THE 45 TERM APPROXIMATION IS

.2.152584
1.319210
0.486045
2.360703
4.027283
0.694306
1.666411
1.666356
1.666579

1.666551

Iterative Solution of Me lnputOutput Model

114.1276991
[17.607299
12.222199)

THE 40 TERM APPROXtMATION IS

7.152259
1.318816
0.485935
2.360017
4.026451
0.694073
1.665980
1.665834
1.666433

Only d and five x vectors are given.

Industrial Sector

THE 15 TERM APPROXIMATION IS

2.080723
1.232059
0.461666
2.209013
3.843307
0.642840
1.571250
1.550941
1.634293

gives these results which

ANO LEAOS TO OUTPUTS

THE 10 TERM APPROXIMATION IS

1.959441
1.084958
0.420515
1.952976
3.532773
0.555972
1.410627
1.356132
1.579797

1.666571

Rogers, in [3], page

confirmed.

k
R turh to:
/UMAP
55 Chapel St.
New n, MA 02160

STUDENT FOgii 1

Request for Help

Student:
If you have troublt with a specific part of this unit, please fill
out this form and take it to your instructor for assistance. The information
you give will help the author to revise the unit.

Your Name

Unit No.

Page

0 Upper

Model Exam
Problen No.

Section
OR

OR__

()Middle

Paragraph

Text
Problem No.

0 Lower
Description of Difficulty:

(Please be specific)

;r

VO.

Instructor:

Please indicate your resolution of the difficulty in this box.

A.

Corrected errors in materials.

List corrections here:

0 Gave student ,better explanation, example, or procedure than in unit.

, Give brief outline of your addition here:
F

Assisted student in acquiring general learning and problem-solving

skills (not using examples from this unit.)
a

26';
1
Instructor's Signature

Please use reverse if necessary.

Return to:
EDC/UMAP
55 Chapel St.
Newton, MA 02160

STUDENT FORM 2

Unit Questionnaire
.1%

Date

`Unit No.

Name

Course No.

Institution

Check the choice for each question that cbmes closest to your personal opinion.
How, useful was the amount of detail in the unit?

1.

Not enough,detail to understand the unit

Unit would'hive been clearet with more detail
Appropriateamount of detail
Unit was occasionally too detailed, but this was not distracting
Too much detail; I was often distracted
.....

How helpful were the problem answers?

2.

Sample solutions were too brief; I could not'do the intermedtate steps
Sufficient information was given to solve the problems
Sample solutions were too detailed;_Ididn't need them.
(for
Except for fulfilling the prerequisites, how much did'you use other sources
unit?
example, instructor, friends, or other books) in order to understand the
Not at all
A Little
,Somewhat
A Lot

3.

generally spend on
How long was this unit in comparison to the amount of time you
'sciende course?
a lesson (lecture and homework,assignment) in a typical.math or
Much
Sociewhat
Abaut
Somewhat
Much
Shorter'
Shorter
theSame
Longer
Longer

4.

'

(Check
Were any of the following parts'of the unit confusing or distracting?
as many as apply.)

5.

Prerequisites
Statement of skills and concepts (objectives)
Paragraph headings
Examples
Special Assistance Supplement (if present)
Other, please explain

particularl,helpful?(Check as many
Were any of the following parts, of the unit
as apply.)
Prerequisites
Statement of skills and concepts (objectives)
Examples
Problems
-Paragraph headings
Table of Contents
Special Assistance Supplement (if present)
Other, please explain

6.

particularly like.
Please describe anything in the unit that you did not

helpful. (Please use the back of

'Please describe anything that you found particularly
this sheet if you need more space.)
)

266

1,

umap

VISCOUS FLUID FLOW AND THE INTEGRAL CALCULUS

UNIT 210

ti

by

Philip Tuchinsky

MODULES AND MONOGRAPHS IN UNDERGRADUATE

MATHEMATICS AND ITS APPLICATIONS PROJECT

TABLE OF CONTENTS

VISCOUS FLUID FLOW

AND THE INTEGRAL CALCULUS

1.

LAMINAR FLOW
1

by Philip Tuchinsky
2.

POISEUILLE'S LAW
2

3.

WHEN DOES NIS, LAW HOLD?

4

4.

top of fluid t

5.

fi
6.

THE VELOCITY OF FLOW AND THE AMOUNT OF FLOW

THE TOTAL FLOW THROUGH A PIPE'OF RADIUS R.

THE RIEMANN INTEGRAL

8

7.* THE RIERANN-STIELTJES INTEGRAL

area
vR2

10
aka

8.

DISCRETE SUMMATION
11

9.

INTEGRATION: LOCAL DATA YIELDS GLOBAL RESULT1

(,

10.

collect
flow
here

CALCULATION OF VISCOSITY
14

11.

EXERCISES
15

,e

12.

REFERENCE

13.

SOLUTIONS OR HINTS TOWiXERSISES

17

APPLNSI'IONS OF CALCULUS TO ENGINEERING

edc/umap /65chapel s)t./newton, m4gs. 02160

* This section may be omitted without

affecting the readability
of later sections.
,

ti

268

Intermodular Description Sheet:

Title:
Author:

UMAP Unit 210

VISCOUS.FLUID FLOW AND THE INTEGRAL CALCULUS

Philip i'uehinsky

MODULES AND MONOGRAPHS IN UNDERGRADUATE,

7623 Charlesworth
Dearborn Heights, MI 48127

MATHEMATICS AND ITS APPLICATIONS PROJECT (UMAP)

Dr. Tuchinsky is ,a member of the Computer Science Department

of Fbrd Motor Company's Research and Engineering Center.
He
formerly taught in the Mathematical Sciences Department at
Ohio Wesleyan University (where earlier editions of this
paper were written).
Review Stage/Date:
Classification:

III

be built.

9/1/78

APP(. CALC/eNGINEERING

Suggested Support Materials: Norie are essential.

A lab set up like
that shown in Section 10 would make an interesting display.
Exercise
4 calls for use of a computer or programmable calculator.
Approximate Class Time Needed:

One 50 minute class.

The Project is guided by a National Steering Committee of

mathematicians, scientists and educators.
UMAP is funded by a
grant from the National Science Foundation to Education
Development Center, Inc., a publicly supported, nonprofit
corporation engaged in educational research in the U.S. and

abroad.

Intended Audience: Calculus students lelining how to integrate

polynomials. The paper is suitable for independent
reading and
seminar presentation by more advanced students as well.
References:

The goal of UMAP is to develop, through a community of users

and developers, a system of instructional modules in undergraduate
mathematics and its applications which may be used t9 supplement
existing courses and from which complete coursellay eventually

PROJECT STAFI

Ross L. Finney
Solomon Garfunkel

See Section 12 of the paper.

Prerequisite Skills:
1.
Calculation of the integrals fxdx and fx3 dx
2.
Knowledge that lc f(x) dx.= cff(x)dx
3.
Recognition of an integral as a limit of Riemann sums.
4.
Comfort with summation results like 1 + 2 + 3 +
+ n
n(n+1)/2.

Felicia WeitzelZ
Barbara Kelczews1ti

5.

Elementary computer programming (for Exercise 4 only).

Output Skills:
1.
Replace'a simple integral by a discrete sum, calculate both and
compare results.
2.
Average a function over an interval.
3.
Reduce simple Riemann-5tieltjes integrals to Riemann integrals
and calculate the latter (if the optional Section 7 in included).
Discuss how well Poiseuille's Law models a spetified viscous
fluid flow situation.
5.
Describe a laboratory procedure for finding the coefficient of
viscosity of a fluid.
6.
Identify local vs. global information.
'

UMAP Editor for this module:

Solomon Garfunkel

Other Related Units:

The Human Cough (forthcoming as.UMAP Unit 211) Starts wit0 the
result of this paper that total flow is proportional to R and goes
on to discbss maximizing the speed.of air flow during a cough.
Diffdrential calculus is its method.

Dianne Lally
Paula M. Santillo

Project Secreta7

Financial Assistant/Secretary

NATIONAL STEERING COMMITTEE

W.T. Martin
Steven
Brams
Llayron Clarkson
Ernest J. Henley
Donald A. Larson
William F. Lucas'
Frederick Mosteller
Walter E. Sears
George Springer'
Arnold A. Strassenburg
Alfred B. Willcox

MIT (Chair.man)

New York University

Texas Southern University
University oT Houston
SUNY at
Cornell Unive sity
Harvard University
University of Michigan Press
Indiana University
SUNY at Stony Brook.
'
Mathematical Association of America
,

The Project would like to'thank Melvin A. Nyman, Peter

Signet' and L.M. Larsen for their reviews, and' all others
w
assisted in the production of this unit.
This material was prepared with the supp9rt of National
Science-Foundation Grant No. SED76-19615. Recommenditionsexpressed acethose of the author and do not necessarily reflect
the views of the NSF nor of the National SteeringoCommi4tee.

0'1978 EDCiProject
All rights reserived.

259

Director
Associate Director/Consortium.
Coordinator
AssOciate Director for Administration.
Coordinator for Maerials Production

270

r
.

.
.

Poiseuille.
result about viscous fluid flew has
many other applications.
We can use it to study!the
flow'of air in the windpipe, oi,l in a pipeline,
water
in a pipe system, grain flowing by.pipe,into the hold'
of a ship, etc. The assumptions involved .in
the result
make it more applicable to some of these problems than

VISCWS FLUID FLOW AND THE INTEGRAL CALCULUS

LAMINAR FLOW
4

-When a 'thick, sticky (viscous) fluid

a pike, it does no; all flow at the same

through--

ad

.others (see Sedtion;".5),'but it provides a gbod first

the fluid Closest to the wall of the pipe, suffer

much friction with the wall that 'it hardly moves tat all,
while fluid closer to-othe central:a'Xis,of the'pipe
moves
more rapidly- The fluid's speed increases
steadily as',
the distance from the wall increases. Because of
.
circular symmetry, the effectis that,of concentric
""
,tubes of fluid sliding over pne another (see Figure 1).

"

Another important use of.,Poiseuille'st'LaW'is to

measure the relative viscosity Of-.fluids. More-about

this rater, in Section-ID:.

We will use Poiseuille's Law to calculate total

7
4

%pproximation to them all.

flow *through a pipe using a. -finite sum and the "continuous

summation" proces calldd integration.' The two!Tesuts

;will deserve comparison.

4

-flow

:_
Figure

r.

Laminar Flow in a Cylindrical Pipe.

2.

.Y

poiseuille discovered and others later deduced

from theory (see
ction 12) that the velocity of the
particas of fl d at a distance r centimeters
out fpom
the center axis of,the pipe is

We call ihis'iaminar flow:

each lamina or layer of

fluid moves at its own speed.
Different laminae move
at different speeds.

(1)

The exact way in witich.latinar flow happens was

found by a French scientist, named Poiseuille
more than
a century ago. 'He was studying.blood
pressure, which
had just been accurately measured for the first time.

v(r) = TEEP ,(R 2

- r2)

(cm7Sec.;)
:

where Irefer to Figure 2)

R = radius' of the pipe, id cm. (Thus 0 < r < R)

L e'length of the pipe (em.)

P

He wanted to know how much blood' flows through

a blood
vessel in a given time.
From that information and
analysis of ,blood aMples one can Sailz.how much oxygen,

pressure change PI
P2 down the length of
' the 0.pet'(dyneicie)
=

'k = coefficie4 of viscosity (poise)

and nutrients are tieing de- livered to the cells

ser-

.
a*with their
Variables will ie.given
cm-gram-second (cgs) Units
to help us understand their physical

viced by that blood,vesiel.'

Knowlede.of blood flow
is a basic part of understanding the body

as ephysical

system.

POISEUILLE' S LAW

units could be usedeoff course.

meaning.

.,.

4ny system of

11.

r'

.5

27

0-1
I.

(Let Me remind you that pressure is force per unit cross

sectional area.) One can prove that the pressure will
decrease_st*adily [as a straight line (linear) function]
as the fluid moves through the pipe.
It. is the difference

JEAN LEONARD MARIE POISEUILLE (1797-1869) was a

well-known physiologist and physicist. Hp
invented the mercury m'anomeer to raelasure blood

in final vs.

initial pressure that enters the equation.

The cgs unit of viscosity, the poise, is named after

pressure, improving the pioneering work of

Stephan Hales. The law consideted here appeared
.in a paper of 1840 and was found through labora-

Poiseuille.
fluid in at higher
pressure PI

tory experiments with distilled water, either and

Mercury.
The mathemgical derivation was first

fluid speed
is v(r)

found in 1860 by F. 1+iimann and J. E. Wagenbach,

o-

who named the result Poiseuille's Law.

fluid'out at
lower pressure P2

'name is disputed:

But the

G. H. L. Hagen found the same.

aw independeuly in 1839; his workiwent un-,

noticed for decades.

Reference:
Dictionary
of Scientific Biography, 1975 edition, vol.II,
-T- p. 62.

t
Figure' 2.
4.1

3.

WHEN DOESTHIS LAW,..44ALD?

d)

T1lemajor assumptions that must be true to have

,equation (1) vatid are theser
_
a)

out thrdugji the, pipe wall and no feeder

.

,TWre must be no.turbidity in the fluid.

pipes are pouring fluid in or out.

This, means that there is no swirling;

The tube, is hofizontal and the (very slight,

'particles of tluidMroVe-in_straight lines

down the pipe.
b)

downward pulling effects Of gravity are

igpored.

For'a vertical tube this minor

variation on (f) is true:
,

The speed of flow v is assumed to depend oil

r only.
Ives

Thus vdoesnot change as fluid

down the length of the 14pec and it

not. change with time;

'

xer)

1.) 7'

/ (R2

2j

the flow is
where g'4. 980 cm /sec/sec is the gravitational

peither.speeding up norslowing down, it

is steadycstute.
c)

Fluid is conserved,,i.e. neither creited nor

I
lost, in the pipe. Thus no fluid is leaking

corstarrt and p isfhe density of the fluid,

i.e., its mass per unAt'volume. 1 For .4 slanted

'Thefluicris incompressable, i.e.,,male up

of particles that cannot'be crushed or Racked.

Rn closer t,getherl(by the forces presenW

pipe, thesqphorAon'tal and vertical velocities

must be vector-added., For simplicity we will

use- (1) .

f)

The pipe is a right-circular cylinder with

4 constant dimensions Lrend R.

273
J

4r

There is 'so much friction at the wall that

fluid there does not move at all.

4.

THE VELOCITY OF FLOW AND THE AMOUNT OF FLOW

We

ant to use Poiseuillei.S. 'Law to calculate the

(Notice

that r = R reads to v(R) = 0.)

'h)

One assumption that is not present:

'

in

other classes you may study so-called

.

"-ideal fluids" in which particles slip

-

frictionlessly by each other.

We are

'assuming thateach layer exerts a drag

on the layer next-further-in. Ours is'

'First, we need a preliminary result. Consider,

in Figure 3, any tnical small piece of cross-sectional
aree'of the piper consistiong of M square centimeters,
located r cm out from the center. Hi/ much' fluid will

, not an ideal fluid.

These assumptions are satisfied to varaous degrees

by the applications mentioned earlier. Swirling,
turbid effects are bound to occur in any large diameter
pipe.

This 'limits yhe usefulnessof'oUelaW41

total f ow through a pipe of radius R.

The flow'F
is the total volume o fluid p'assing through the pipe
each second; in units of (cm)3 /sec.

leave the pipeAhrough this J;At of area in one fecond?

The fluid po'vesfIr(r) cm in the one .second; ,thus a
stack of, fluid (r) cm longpf constant cross section

-.-

AA(cm) 2 '(shd

studying water pipes, oil pip..0ines, grain chutes,

etc.
Blood vessels flex: their dimensions change a
little,
Blood surges because of the heart'spumping

'

) will flow out of the pipe through AAin the.one:stcond. This stack has, volume v(r)..(AA).
A

action; thus the flow is not steady-state. Oxygen

and nutrients leave a blbod'vesselyby osmos4s through

area AA
0

the pipe's wall and wastes are added to the blood

flow, so that fluid lisonly approximately Conserved.
a

Despite these and other practical short-comings,

Poiseuille's Law'is a valid simplification ofvistous
fluid flow.
It is the right sort of. law: v(r) is

Figure 3.

'zero at the pipe wall and increases steadily as r

decreases and we apprOach the pipe's center. It has

Thus fluid leaves AAat 4 steady rate'of

v(r)AA(cm)3/sec.

a solid, well:understood theoretical basis. Weran

really calculate with it, as we shall shortly see,,.
And in the laboratory, the assumed conditions 'Can be
k

Summ ar

(2)

made almost true, giving a practical' way to measure

the-viscos.i.* coefficient k for any fluid.' This
coeffidlent is a fundamental property of the f141C.
'

5.

,important in.design and

t

If AA is, any area through which

of i
lows. at alconetant veloc.ity v,
dren vAA is the total flow' through the
.area AA, per second.
4
:

THE TOTAL FLOW THROUGH A PIPE OF RADIUS R

engineering

,In the pipe's cross-sectional circle of radio's R,

4

the velocity v(r) given by Poiseuille's Law is the same

4
at all points located r cm from the center.
If we
6

ot.

276

Consider concentric rings of area (Figure 4), he

fluid's velocity-will be approximately constant
each ring. We can then use (2) to calCulate.the total

If we take n large and the r.'s close to each other,

J,
the velocity of fluid flowing through any one region
will be almost constant, although different from ring

flow through each ring; the sum of these ring-by-ring

flows will be the total flow through the pipe, which

to ring.
What value will approximate the constant
velocity in the jth
:
Pick any point in that ring;
say, pick a point that is tj
its out from the cebter

we set out to find.

4

with r.

Them vil
is a typical speed
3-it, 3
)
for the )
ring and (2) says that

To clearly identify these rings, partition the

interval

<

<

.1

(0 < r < R)

the flow through the jth

v(t.)[ir.2-.Iir. 2]
J
J
J-1

into n pieces using partition points

We callyti an evanption point for the j th subinterval

0 = r

< r1 < r2
1

Lo

< r

n-1

< r

= R

[r

j-1'

ri.
)

Thus the total flow through ,all n rings

(p2rhaps not equally spaced).'

..(3)

F =

v(ti) [Irrj2

j=1

is

Irj_1].

'

We write "approximately" instead of equality because

we have replaced all the various values of v(r) in

the j!11 ring by the single value, v(tj).

In fact,

we have a vast family of app-roximations of F in

Equation (3).
For any choice of a partition r0, ri,

rn and any choice Of evaluation points ti, t2,

to (such that r.
)-1 < t j < r j forseach j) we get
an approximation Of
As we take farger values of
n and more closely spaced r 's And t 's, the theoryjigf

r.

'integration tells us that such sums approach a limiTfing

value more and moirlosely, and that(limit is an integral'.

r
7

6.

The first, second, '... regions are then chosen as

.th
Sketched. ,For j =
n, the
region is a
ring with inner and outer radii rand .,r and-thus

has area

THE RIEMANN INTEGRAL

We must 4o s hit more work on Equation (3) before

it is recognizable as a Riemann sum. Let the width pf

,
th.
the )
subinterval be

IT(r)2

Ar. = r.

rj_i

7
8

es*

7 "1
,Co

278

Then
7.

n r it- nr.
33=t,

= n(r.
.3-1 + Dr.);
3

rj_i2

(At.)

The integral usually studied by calculus students

is the Riemann ifitegraA

4'

As, nincreases, rj and rjf approach each other and

f(x) dx.

An important generalization its the Riemann-Stieltjes

negligibly small' by comparison to the first term,

'and becomes more negligible as n grows larger.
AV
Thus,
from-(3),

integral wherehe "dx" representing change

in x can
be repltced by "d g(x)", the change in
a function of x
between one partition point and the next.
That is,
the RiemanWlkums and the limits they approach
have
the forms

hr.).
3

= 1.
3v(t.)(2nr.

P Ar.becoTessmall.'nleule(Dr.)2 erm above

THE RIEMANN-STIELTJES INTEGRAL

6
n -0- m, and all' subinterval widths 66.r.

shrink to

=
j1

ero', this Riemann 'sum becomes

f(t.)[x.
3

j-1

f(x) dx
a

while the comparable Riemann

v(r)(2nr)dr

-Stieltjes forms are

b

I 'f(t.1Cg(x.) -: g(x.

j=1

R
f

r--P (R2

2
- r2)

2r14 dr
t

You are asked to calculate the 4ntegial in kercise

1.Another conversion of (3) into a Riemann sum:Since
nr.

i= n(r. +r.
3

3-1

3-1.

)(r.'3

3-1

f(x) dg(x)
a

8k.1,

)1

nR

j"

J'

'

In each case a = xo,< x <

< x = b is a partition
1
n
of [a, b] and e.'is an evaluation point in the j th
subinterval: xj_l t
<j <

We can now tecognize (3) as a Riemann-Stieltjes

sum with this in,tegral as its limit
.4

r3 -1)

we havefroil (3)

v(r) d(ler,2)
0

(4)

v(t.3 ) n(r. +r.

-1

j=1
As .n

we get

co, t j

,.

)(r. -r.
).
3-1

P'

(R 2

-'r2)

and rj-1 all approadh'each other and

* This section can be omitted without affecting

readability of later
sections.

/.

n(r + r) dr
o

,-

_ 1R v(r) (2nr) dr

as before.
9.

279

ig

280

10

ti

We cast convert this integral to a Riemann integral

by using this theorem:

To
develop this idea, we should let all the rings
*
have that fixed finite thickness Or. Thus r0
0 = 0.Ar,
ri
r2 = r1 + tr = 2tr, etc.; the n+1 partition
points are r = jAr, j= 0, 1, 2, ..., n.
Let's

If f

continpoUs and g has a continuous

first derivative for a <....x*< bthen
.

simplify by taking the evaluation points to be

t, = jAr also. Then, from (4),

b
j

'

t(x) g'(x) dx.

We get (since g(r) =,,nr2 has derivative g'(r) = 2nr)

4kL (R2- (jAr) 2) n[jAr +(j-1)Ar][Ar]

( 5)

j=1

F =

--

.*

2nr dr,
Pn
4kL

(R2

ri

- j2(Or)2)(2j

1)(Ar)2:

j=1

the same Riemann integral as in Section 6.

Plug in R = nAr and simplify to:

Why should we be interested.in to RiemannStieltjes integral if it simply leads us back

Riemann integral we derived twice in Section 6? The

124W1

j2)(2j

(n2

j=1

1)

Stieltjds case,betomesteresting wheli g is not a

smooth function, when g'(x) dcies not exist.

Then

Plr,(4g).4

[ 2Et;

3) + z 0.246 2r12

Riewn-Stieltjes theory must be used directly;

we cannot escape to th'e easier Riemann case.

n2E 1].

We can prove by mathematical induction that

There

are important applications, especially in theoretical

A

economi.cs; wftereg must' be taken-as a step function,

for example.

(j 3)

= 13 + 23 +

() 2) =
-

8.

+n3 -

j =1

DJSCRETE SUMMATION

1 2

2 2

+n2

n(n+1)(2n + 1)

j=1

Is it valid-to len =5- c

taking rings of-atbdtrarily

smaller and smaller Nidfh? That is, should we convert

j= 1+ 2+ 3+

+ n

n(n +l)
2

'

j=1

(3) into an-integral ?. The fact that you art learning

1

In fact, we often shoufil:nOt,take the limit.

After

1 =

= n.

j=1

blOod'is made Lip of red blood cells and other

Particles.

et

n-

calculus is'norSufftoient to.make the answer "yes"!

,pcstimes

They hate e certain non-zero. thickness

Plug these in and dp the algebra to reach

*bir an4 no layer:Of blood can. lid thinner than that

thickness.
The same is true of aLl fluids, in fact'.

12
11

.281
of

282

41.

F . 141_
(6)

pipe's total'flow, to the pipe as am entity in itself:*

Integration/(or discrete summation, which is used less),

112(n+1)(n-1)

htil*:Pn(1

n+1

.11;11

cbnverts,loca .11y vary=ing information into the global.

We are reasoning from the'mor

T il
As,n + co,

Do we lose information through that

proceS>G.an
we reason back to the local.if, we know the global
result? You might immediately answer
"no" or "sure,
just differentiate." Can you justify eithert.wwer
\
carefully? My question is

+ 0 and this does approach the integral's

-1
2

value, as lif should.

\
When we Want to compute a sum, we often
use the
integral to approximatein
a prOblem (like our curyent
f
one) where n + co does not make sense.
If n is in fact
4
very large, only.a small error is made. To do the ,
actual sum for large n would'be cumbersome; by
letting
n
0, we gainiall theNcalcuIational
power of the integral
calculus and save the algebra that led to (6).
I

Suppose we knoW that

',1

.detailedto the less

detailed when we integrate.

..

R
v(r) d (Tri2)

--4

13.11

SR L

R4

for any R > 0.

Can we deduce Poiseuilles Law, that'

There are other problems in, which it is

an integral
we want but we, are forced to use a sum. (Many integrals

v(r) =

can't be calc9.ated by aniti-differentiation).

By
taking n suffiCiently large, a high accuracyhapproxi-

Pn

T" (R

- r 2 )?

leave it'unanswered here.

'nation of the, integral can be gotten with the help

of
a computer.

10.

CALCULATION pF VISCOSITY

(a' 4.

Integration and discrete summation are associates.

Each can fie'{ as a replacement for the other,
in
appropriate circumstances.

'To calculge k for a specific liquid, set up a. tank je

and pipe in the laboratory as in Figure SA Get a steady
flow going, then collect (say) ten seconds flow in
,

,a beaker.

Measure that Volume of 'fluid.

9.

INTEGRATION;

LOCAL DATA YIELDS GLOBAL RESULTS

LaW,contains'lobal information:
the
speed of fluid flow at a specific spot in
the pipe
is v(r).
Our result (21 that v,AA is the -total flow
through a bit of area AA where v is the (almost)

I-

10F = 10

8kL

In this. equation we know every constant except k,.which

We know R and L by measurement.

To
find P we take the-difference between the pressures,
a
PI and p2, at the beginning and end 1
of the flowpipe..
The outlet pressure P2 is simply atmospheric
pressure.

When we sum that local data overall parts of

the pipe's cross- sectional circle, we gather
the

result, referring t7O the

ITR413

we calculate.

constant speed of flow is still local infordation.

local data into arglobal"

According 'to our integratipn, in tep seconds the

.
volume of fluid ,flowing out should be

13

14

231

:284

.10

c)

3.

area

Use v(r) from' (b) to fiad the total flow,

through the pipe of radius R.

The velocity v(r) varies from place to place in

the pipe's cross-section, but has some average

nR2

value V.
a)

collect,
'flow
here

,b)

. Figure

Explain how to find 7 from the total flow

F and the principle in (2)
The .definition of the average value of
the function v(r). Is
R

5.

f v(r) 2yr dr
v

If the fluid has weight dtnsity (weightper

unit volume)p
and the fluid depth is h as shown, the inlet
pressure
P
is pgh, where g is the gravitational constant.
1
-'11.

R
1.

Calculate this and check against your work

in (a).
The two answers should agree.
-

c)

Show that f --..-6 41,

Notice

(112

r2) 2nr dr =

yll`P

,,
.
8KI.,

The largest velocity is V and occurs at

r = 0.
Check-that V 1 217
.

4.

k, L and R are simply constants.

2.

1.2 nr dr
0

EXERCISES

a)

We have assumed that the fluid's velocity at the

pipe wadl is,zero. There's no need to do that:
T1
advanced derivation (see Section 12) that we
have omitted in tki#Ppaper in fact shows that the

Use a computer program to calculate tie

sum (5) for, reasonable values of n, R, L,
'etc.
Check the computer results against
Arle:algebyl'aic result (6).

Repeat with

larger valUes of n.
b)

veloLty fs

How large must n be to have the discrete

sum within 1% 'of the integral result?

V (y

, 4

-r4)rr

4.

+ b

.."(L

12.

where b is a .constant we may choose.

Show that v(R) = 0 leads te the formula (1)

we have used.
b)

REFERENCE

to'

Suppose the velocity at the wg11 is one -half

of the velocity at the center (r = 0). Find
the function' v(r) for s.this case.

If you know multivariable calculus and a little

mathematical phy;ics, you can read a clear derivation
of Poiseuille's Law from basic ideas in elasticity and
fluid flow:.

Slater, J.C. and' Frank N. H. Introduction to

Theoretical Phyosics, McGraW-Hill, 1933.

15
o

ft

Mbre'recent book; with similar titlts.

285

8(3
a

Or
.

16

.13.

1.

VOW

SOLUTIONS OR HINTS TC EXERCISES

First convert

IR

to -2274-ra

ZnP

r dr
0

2:

b)

v(r) = 2PR2 4kL

c)

(R 2PR2

Kr,

Prz'

r2 dr.

44.

a)

27r dr = 3nPR"

727

If th
vipg at the same speed;
at air ivants in t e cross
- sectional circ14
of radius R, that constant.speed
would of
course be the average of the Poispuille's
Law speeds:
From (2), using AA .=47R2,,

the full circular area,

Totl flow = Ve(nR2)

4
7

c).

:21( 7E1:

At r = 0, v(0)

= V .=

R2 . 2V

..

=f

("\f

17

287;

.%

Return to:
EDC/UMAP
55 Chapel St.
Newtoni, MA 02160,

STUDENT FORM 1

Request for Help

/

Student: If you have trouble with

a specific part of this unit, please fill
out this form and take it to your instructor for assistance. The information
you give will Help the author to revise the unit.

Your )ame
^

Unit No.

Page

Upper

Model Exam
Problei No.

Section

OR

OR

()Middle

Paragraph

Text
Problem No.

Q Lower
Description of Difficulty:

Instructor:

(Please be specific)

Please indicate your resoIutipaofthe difficulty in thisqbme.

Corrected errors in mater als.

List corrections here:

Gave student better explanation, example, or procedure than in unit.

Give brief outline of your addition here:

ft

o,
11^

d student in acquiring general learning

and problem-solving
i
skills (not using examples from this u nit.)
.

)1
3
'Instructor's Signature
Y.

(lease use reverse if necessary._

Return to:
EDC/UMAP
55 Chapel St.'
Newton,\MA 02160

STUDENT FORM 2
.

Unit Questionnaire

Name

Unit No.

Institution

Date

Course No.'s..

Check the choice for each question that comes closest to your personal opinion.
1.

How useful was the amount of detail in the unit?

f Not enangh detdil to understand the unit
1 Unit would have been clearer with more detail
Appropriate amount of detail
Unit was occasionally too detailed; but this was not distracting
Too much detail; I was often distracted

2.

How helpful were the problem answers?

Sample solutions were too brief; I-could not do the intermediate steps
Sufficient information, was given to solve the problems
Sample solutions were too detailed; I didn't need-..them
3.

Except for fulfilling the prerequisites, how Mich did you use other sources (for
example, instructor, friends, or other books) inarder to understand the unit?
(A Lot

4.

Somewhat

1 Little

Not at all

How long wad:ths unit. in ,comparison to the amount of time you'generally spend on
alesson (lecture and homework assignmt in'a typical math or science course?
%

..,

Much

4
it

Longerifi

Somewhat
Longer

5.

III

About
the Same
._...

Somewhat
Shorter

Much
,..

Shorter

Were any ofithe following parts.of the unit confusing or diatracting? (Check
as many ablapply.)
,

Prefequll!tes
Stat'tient of skills and concepts (objectives)
Paragraph headings
o
Examples
SpeLial Adsistance Supplement (if present)
Other, please explain
6.

.#

Were any of the following parts of the unit particularly helpful? (Check as litany
as apply.)
_____
Prerequisite*
Statement of kills and concepts, (objectives)
Examples
.

Problems'

/
,

Paragra

headings'
74able of ontents
rSRecial Adsistance Supplement (if present)
Other, please explain

Please describe anything in the unit that you'did not particularly like.

Please dfecrihe anything that you found particularly helpful. (Please use the back of
this sheet if" you need, more space.)

2q9

.()UMAP MODULE?"
MODULES AND
MONOGRAPHS IN
UNDERGRADUATE
MATHEMATICS
AND ITS
APPLICATIONS
"tr

"Co

'CA

ed

CC`

er

:3-.

r
The Human
Cough
by Philip Tuchinsky

,o-r

trj

CT1

N m N ct.
>,

>,

r'r

err

JYr

,
-0

4
.

.0

QS

.
.

sL

err

0
911.

.0

Applications of Calculus to Physics,

Biological and Medical Sciences

ts1

fZ

ts4

Co
.3
co. --3 Birthauser Boston Inc.

381/Green Street
Cambridge, MA 02139

Intermodular Description Sheet: ,UMAP Unit 211

Title:
Author:

THE HUMAN COUGH

Philip Tuchinsky
7623 Gharlesworth
Dearbo
Heights, MI 48127

Di. Tuchins y is a computer scientist and mathematician at Ford

Motor Company's Research and Engineering Center. Hekformerly taught
in the Mathematical Sciences Department at Ohio Wesleyan University
(where earlier editions of this paper were written).
Review Stage/Date:
Classification:

IV

7/30/80

APPL CALC/PHYSICS, BIO & MED SCI

Approximate Class Time:

Less than one 50-minute class.

Intended Audience: Calculus students learning to use the derivative

to compute extreme values of functions. The paper is suitable for
indepentent reading or seminar presentation by more advanced students
as well.

Prerequisite Skills:
1.
Differentiation of polynomials.
Interpretation of dy/dx = 0 and the Second Derivative Test for
identifying maxima and minima.
3.
Operations on inequalities.
4.
Basic curve sketching as taught in calculus.

'Educational Objectives:
1.
To see how a physical assumption may lead to a, choice of domain
for a function.
' 2.
To see an application of maximization of a unction on a closed
interval domain.
3.
To interrelate biology, physics, and calculus.
Related Units:
Viscous.Fluid Flow and theIntegral CalculdS (Unit 210)
UMAP Editor for this Module:

ti

Solomon Garfunkel

The Project would, like to thnk L.M. Larsen of Kearney State

College,for his review, and all others who assisted in the production
'of- this unit,

This material was field-tested and/or student reviewed in class

by,8imon Cohen of New Jersey Institute of Technology, and T.R. Hamlett
of Arkansas Tech University, and has:been revised on the basis of
data-received frob these sitos.
This material was prepared with the partial support of National
Science Foundation Grant No. SED76-19615 A02.
Recommendations expressed are those of the author and do not necessarily reflect the,
views\of the NSF or the copyright holder.

(f) 1980 EDC/Project UMAP

All rights reserved.
N't

:2 9 1

A*

THE HUMAN COUGH

by

Philip Tuchinsky
7623 Charlesworth
Dearborn Heights, MI 48127

TABLE OF CONTENTS

1.

WHEN YOU COUGH

mve

2.

NOTATION FOR A CALCULUS MODEL OF COUGHING

3.

LAMINAR FLOW

4.

AVERAGE VELOCITY AND TOTAL FLOW

5.

PERFECT ELASTICITY

6.

WtAT RADIUS R MAKES V THE LARGEST?

'

7.

ACKNOWLEDGEMEAT

8.j

SOLUTIONS TO EXERCISES

A
,

41

fl

292

Ii
THE HUMAN COUGH

.1:

WHEN YOU COUGH

ei.

a foreign object in your trachea (windpipe)

Wh

leads yOu'to cough, your diaphragm thrusts sharply ipward,:

As a result, the air in your.lungs is suddenl,compressed

to a higher pressure than the air outside your body.

high-Speed stiam of air shoots upward through the

trachea equalizing these pressures arid, it is to be hoped,

clearing the passage.

BY Newton's law, the force exerted on the object to

be cleared is due to the sudden acceleration 'of the ai,r

,flowing through the trachea.

The greater the velocity of

the airstream during the cough, the greater the force on

the foreigner and the more effective the cough.

To

increase the speed of the airflow, your body also contracts the.windpipe during a cough, making a nafrarer-,

channel
for the air to flow through.
-

For a given*amount,

of air to escape in a fixed amount of time, it must move

faster through a narrower channel than a wi1er one,'just,

a\a river flows rapidly where it is narrow but placidly

where it is wide.
In fact, x -rays show that the radius
of the tracheal tube reduces to about two-thirds its usual
radius duping a cough.

2.

NOTATION FOR A CALCULUS MODEL OF COUGHING

We can relate the speed of the airflow during a

cough to the body's contraction of theetrachea amazingly
well by studying a simple mathematical model of the
situation.

We think of the trachea as a pipe with a

circular cross section, andapply.the differential calcu4, using the following notation:
1,

293.

Ro = the "rest radius's of the trachea (its usual

radius when you are relaxed and not coughing)
in' centimeters;

= the contracted r dius of the trachea during

cough (thus R < t,0),

V = the average velocity of the air in the trache

when it is contracted to R cr. This depends
on'R and we wish Xo calculate R such that V(R)
is maximal;
P = the

xtra pressure in the lungs during a cough,

i.e.:the difference P -P between the pressure
1

P
1

`P

in

otir ,lungs and the atmospheric. pressure

outside your mouth, measured in dyne /cm2.

F = the total volume Of air flowing through the

trachea per second, in cm 3 /sec.
We will make two physical assumptions, one about the
airflow, the other about the flexibility of the trachea's
wall.

3.

LAMINAR FLOW

First, we assume th&.'t the airflow is laminar. This

means that layers'of air move at different speeds in the
trachea.
The thip layer of air right next to the pipe
wall hardly
moves at all because of friction-with the
\,
wall.

The layer, or lamina, just inside that one moves a

,little faster, and ,so on until the fastest airflow is

ford along the central axis of the trachea. It is as if

th4. airstream were made of thin concentric tubes of
air
sliding fiver one another.

See Figure 1.

Laminar flow is an appropriate model for the motion of

any fluid through a confining pipe.

Tn 1840, French physiologist Jean Poiseulle* established that the speed of the
*1797-1869.
arteries.

He was studying the flow of blood thrtiligh veins and

.

. -t

central axis

Figure 1.
The air in the trachea is assumed to flow in
thin concentric cylindrical layers called Laminae.
Inner
layers move'faster than outer ones, which are slowed by
friction with the tracheal wall.

fluid (of air in the trachea in our. case) at a point x

ctn out from the center axis of the pipe of radius R cm
is
A

T1)

v(x) = kP(R -x 2 ) 41r/sec for 0 < x < R.

.

Here k is a constant depencling on the length of the pipe

-and.the particular fluid involved. We defined P and R
earlier.
The average speed V.is the average of these
v(x) values over all points in the' pipe.

Formula (1) is usuaTly called Pois'euilte's Law.of

.viscous fluid flow.

By using integral calculus, it is

easy to deduce from (1) that the total flow per second

'through the trachea (when it' is contracted to a.radius

of R cm) is
(2)

F = cPR

cm /sec,

The'constant c again dependt on the length of the

pipe an4 the fluid involved.
Formula (2) is derived
from (1) in several ways in A companibn paper to this
one, Viscous Fluid Flow and the Integral Caiculus, UMAP
Unit 210. Laminar flW is discussed in more detail
there, too.
Dti

9:1

295

4.

AVERAGE, VELOCITY AND TOTAL FLOW

We mentioned above that we'could compute the

average airspeed V in the trachea by using integral cal
culus to average the speeds v(x)
However, we can relke
V to the total flow, per second F in a much simpler way.
Imagine air flowing through the trachea at a
steady velocity of V cm/set.

In t seconds, each particle

Now, the cross-sectiOnal,area
of the contracted tracheal tube is nR 2 cm2'.' Therefore, a

ef air would travel Vt cm.

cylinder of air Vt cm long by nR 2 cm 2 would leavethe

tube during those t seconds. :The flow of air through the

tube, measured in volumeper second, would be

P =

(3)

(Vt)(nR )

nR V

cm /sec:

We can now write V in terms of P and the contracted

.

radius R by using. (Z) and.(3):

4

'V =

(4)

F
--7
-t-TR
T.

PR

nR

riR

2
,

wfitYt ci = c/n.

-...

5.

PERFECT ELASTICITY

_The second assumption, about the'flexi8ilityzor

elasticity of the trachea's wall- tissue, is needed. next..

We assume that these tissue's are "perfectly elastic."

This means that the tissues contraZt so as to reduce

the radius. of the windpipe in proportion to the pressurechange P between the two ends of tle pipe.
(5)

Ra -

That ig,

R = aP,

for some constant a > OA

pressure, changes P,

Thisis valid for fairly small

fact for
4

296

Rn
(6)

< P < 7i.

If larger values of P occur, the tracheal wall stiffens

and the contracted radius R would be larger than the
value predicted by (V).

(This is fortunate--if the

trachea were to contract too much, we would suffocate.)

Exercise 1.- Use (5) to prove that the, inequality
R

0
2a

0 < P < --

Is equivalent to the inequality

R <

Thus, by assuming perfect elasticity, we are also assuming that the

contracted radius R is at least 50 percent of the rest radius R0.

You may be familiar with Hooke's Law, which says that

the change x-x0 in a spring's length when a pull, or force,
of magnitude f is applied is proportional to f.

Natural Length

a.

Unstretched:

b.

Stretched

Figure 2.
A spring stretched beyond its natural (unstressed)
length by a force of magriitude f.

ti

297
4

'114

That is,
f

k(x-x0),

for some constant k.

This Is really the principle behind

)le pressure change sucks in the
4
tracheal wall with pressure P and the wall behaves as
perfect elasticity.

though it were made up of small Springs, which, stretch

(Figure 3).

tracheal wall

Figurg 3. The tracheal wall is sssumed to behave elastically

as though it were made up of small springs which stretch as-the
trachea contracts,

As (5) says, the amount of stretch, R0 -R,

to the magnitude of the force.

is proportional

Although this is a rather

simplified explanation, it leads to a good working model,

as you will see in the next section.

298

r
6. WHAT RADIUS R MAKES V THE LARGEST?
We'can use '(5), the f5rmulqfot-.perfect elasticity,

.
..

to ekpress P in terms of.11.:

t
'

110-R

Inserting this in (4) gives us V in terms ofR'llone:

.

(1).,

c i(

= c (R -R)12
0

cm/sec.
.

Here c

= c /a and R

are constants.

Equatipn (7) tells

us that airspeed V is produced when the trachea contracts
from R to ,R cm.
2.

Our original goal was to discover what valUe of R

gives the largest value of V. Since V is a differentiable
function of R, for''R in the domain ['IR R ] , V apst assume
.

0'

its maximum at one of the endpoints 11R0 or R0, or at'an

interior point where dV/dR
O.

"N

Exercise 2.
Show that V = c (R

a.

R)R2

satisfies dV/dR = 0 (has horizontal

s.

,tangents) for R = 2R0/3 anaR = 0 but no other values.

b.

'41

result:
c.

Show that R = 2R0/3 leads to d V/dR

.
< O.

Inperpret this

what sort of horizontal tangent is R = 2R0/32

Carefully explain how you know that V has 'its absolute Maximum

at R = 2R0/3 when
when R is restricted to the domain [IR0,R0].
O

'.

As Exercise 2c shows, our model leads us'to predict that

our body can maximize the cough's effectiveness` by Con-

tracting about 33 percent, from R0 to 2/3R0.

,

This' agrees

with experimental evidence as to how the body actually

behaved It is as though "Mother Nature" Tsed calculus
in designing the complex muscle- actions of coughing to
maximize the'airflow speed produced!
7

:a

Exercise 3..

Sketch the.graph of f(R) = (R -R)R2

0

a.

for 0 JK?12 < R0

b.

for all real B.

Results from Exercise 2 will help, because V is just a constant

multiple of the fuation f here.

7.

ACKNOWLEDGEMENT
O

I first learned of this application fr,om Alfred B.

Willcox, Executive Dire?tor of the Mathematical Association

of America, whom I wish to azi4owledge and thank.
Dr. Willcox presentedrmost of this material under the title
"Coughing with Calculus" as part of a talk at the Spring
1975 meeting'of the Ohio Section of.the MAA at Bowling
Green State University.
SOLUTIONS TO EXERCISES

8.

R
1.

0 <P <

0 #

-'<.=> 0 < aP <

2a

(multiplication by a)

0
<-.> 0 < R0 - <R T (substitution from (5)).
The left ha'f, 0 < R0 - R,

is equivalent to R < R0 and the right

half, R

0,

Ro

R < -2

is equivalent to T. < R.

'

Together they give

a
0

T < R < 11.0*

2.

a.

By the product rule.(there aretther,Ways)

dV
dR

b.

dV
wif

= 10-1)R 2

+,(R0 -R)2R) = c2R(2R0-3R).

d2v

= 0 and --T < 0 at a particular R indicates a local

dR

maximum.

300

The abf;olute maximum needed here must-occur at atlendpoint

c.

of the domain or at an interior point where dy/dR '7 0.

Thus the candidates are

Corresponding
Value of V

i.
0

eridpo into -

c R
2

3
0

endpoint R0

local maximum 2/3R0

3
c2 Ro

41--- the largest V

We ignore the horizontal tangent at R = 0 because it is

outside the domain of our function.

ke
3.. the polynomial f(R) = (120-R)R2 has a double root at R = Q, and
a ingle root at s12 = R0.
a
f(R)

Max

inflection

min

R0/3

R0/2. 2R0/3

(Vertical axis not to scale)

301

.
1

UMAP

ODULE 215

MODULES AND
MONOGRAPHS IN
UNDERGRADUATE
MATHEMATICS
AND ITS
APPLICATIONS

',
...

tb
..2

..2

-i

,..

0,

-i

u-.

u...e

..1

>, 0 >,

-)*

err

0 > 0

the numbers of words appearing once,

:0.

'7. Aca

,1

'C--

-C-"17)

I =Il

c4

'words in that novel!

Applications of Calculus
to Social Science

[t ]

*6

'Cr ,

NJ
14

O
V)

LA

.41

"V

E.
51

":1

Ulysses,split up the total number of

7.

'S

.6
ti

--I

vf

twice, thrice, etc., in James Jpyce's

A 4A 4
.0

'

split up 1 in just about the way that

:0-

C:D

The terms of the series LL 9.--1--

by Philip Tuchinsky

a) >,

err, IX err
0'

.--.

. = . .. =
,

Zipf's Law
and His
Efforts to Use
.
Infinite Series
in Linguistics
1

0, -1

tzi -3

tO

'

tii
CD N m N CD N
t21

..1

elV

IA

ICO

tO

to

>P>P>
A

...:

.
.

Birkhauser Boston Inc.

380 Green Street
Cambridge, MA 02139

30n4

.
.

81

.Intermodu

,Descript%on Sheet:

UMAP Unit 215

ZIPF'S LAW AND HIS EFFORTS TeUSE INFINITE SERIES

IN LINGUISTICS

Authoh ,Philip Tuchinsky

.7623 Charlesworth
Dearborn Heights, MI 48127
Dr. Tuchinsky is a*member, of Engineering Computer Systems at Ford
;Motor Company's Research and Engineering Center. He formerly taught
in the Mathematical Sciences Department at Ohio Wesleyan University
(where earlier editions of this paper were written).
Fieview Stage/Date:

Classification:

IV

2/4/80

APPL CALC/SOCIAL SCIENCE

Suggested Support Material: Add one or more ,selections of English on

which to do word-count experiments.
Approximate Class Time Needed: One 50 minute lecture plus out-of-class
time for word-count experimentation and exercises.
Intended Audience:
Calculus students studying series.
By ignoring
Exercises 4-7, the paper could be used at an intuitive level in precalculus or finite math or liberal arts mathematics courses. The unit
.is appropriate for independent study or seminar presentation by more
advanced students.'

Prerequisite Skills:
1. Definition of infinite series, and its sum.
2. Partial sums.

3 Geometric series skation.

4. Algebra on inequalities (for Exercises 1,2,3,5).
5. For Exercise 4 only: comparison, ratio and integral tests of convergence of series.
6. Algebra related to the logarithm function.
7. Log-log graph paper and its uses.
(You can use this paper as a context in which to teachryour students
that y_AxB will appear as a straight line on log-log paper, with A
and B predictable geometrically or mathematically from the graph,
and y=AElx will graphas a_straight line on log-ordinary (semilog)
paper.lo my experience, many students are using these facts in
science lab work without understanding why they work. They are
delighted to have this enlightenment; their mjstaken feeling that
"none of this calculus is really useful for much" will be suBstandaily reduced.)
15.

Output Skills:
1. Use partial fraction? to explain the summation of E1 /k(k+1),,
2. Calculate relative errors to measure quality of match-up between
two sets of data.
3. Carry out a word-count study on any lengthy text in any language.
4. Convert item-count study data into rank-frequency data.,
5 Use log-log paper to graphically test whether rank- frequency data
obeys Zipf's Law.,0
6. Give an example of pure, apparently,Impractical research that has
practical implications for a sophisticated system like human language.

UMAP Editor for this module:

Solomon Garfunkel

Q 1580 EDC/Project UMAP

All rights reserved,
k

C-

82

ZIPF'SLAW AND HIS EFFORTS TO USE

INFINITE SERIES IN LINGUISTICS

by

Philip Tuchinsky

TABLE OF CONTENTS

'
1.

PARTIAL SUMS CAN HELP US ADD UP A SERIES

2.

SUMMING THE SERIES ZIPF USED

3.

WORD COUNTS IN'JOYCE'S ULYSSES

4.

HOW GOOD IS THIS SERIES MODEL?

5.

THE EXTENSIVE RESEARCH

INTO WORD-COUNTS AND

RELATED LANGUAGE PATTERNS

6.

ZIPF'S LAW (THE RANK-FREQUENCY LAW)

7.

A LOG-LOG GRAPH REVEALS OBEDIENCE TO ZIPF'S LAW

!
8.

EXERCISES;

9.

MANDELEIROT'S EXPLANATION OF THE LANGUAGE PATTERNS

15

SOURCES

17

ANSWERS TO EXERCISES

19

10.

*11.

DERIVING A NUMBER-OF -WDRDS LAW.

*This section is included in the instruckional unit but'

omitted in the UMAP Journal version for the sake of brevity.

304

10

'

83

ZIPF'S LAW ,AND HIS EFFORTS TO USE

INFINITE SERIES IN LINGUISTICS

a

1.- PARTIAL SUMS CAN HELP US ADD UP'A SERIES

The partial sums of the series
00

a.

j=1
t"

are, of course,
n
a.

j=1
for n = 1, 2,

The sum of the series is 'defined

....

3,

to be the limit ufsthese partial sums as n

Although

00.

that's a sound definition, its almost useless

hen we

want to calculate the sum of a series, because it is

impossible to simRlify the partial sums of most series
into a form where the limit can be obtained.
exception to this rule is geometric series.

partial sum a + ar
1

ar2 +

A classic
The n-team

+ arn-1, simplifies to

- rn

- 1-r
(as you should be able to prove).
,form, we can see what happens as n

In this simplified
00:

for r such

that Irl_< 1, we have rn -*0 and the series converges

td

while Irl

>

1- r

'

rn 4 o and the series diverges.

happens whey, r = 1?)

(What

This paper is about 'another exception, another

series wholse partial sums can be directly analyzed.

84

This series is not as important as geometric series (which

has dozens of significant applications). .However, our

'series played an interesting role,in the linguistics.research of deOrge Kingsley Zipf i the 1920's and 1930's.
We will examine'.that application and the later research
about artificiall languages that has made Zip#'s work obsolete.

A surprising interplay between the study of human'

languages and engineering research into communications
networks and computer languages will be discussed.
We will see that the serie'so we study is not completely
successful as a mathematical' model in.Zipf's work. Several
efforts to vary and improve the model will all lead to

difficulties--no single accepted model will emerge.

That
sort, of partial success is common uhen applied mathematicians work on actual complex problems; this deservel contrast against the experience of most students, who see Ile
successful theorem proven after another as they study 'the
established branches of mathematics.

SUMMING THE SERIEI'ZIPF USED

2.

The series we consider here is

1

k=1 k(k+1)

1.0

17/"'

/77

The key is to use partial fractions.

1

Please check that

k(k+1) V: v71.

Now 'the partial sum through n terms is

.

k=1

k(k+1)

T77
IT
IT

(n-1)n + n(n+1)
1)

fl

7)

1)

Tj

1)

0-.17T

0.7
(1

t____J
cancels
=

cancels cancels

cancels

n+1

306
I$

This partial sum is now so nicely simplified that we can

see what happens as n i co.

Of icourge

n+1
.

and thus
,74)

(1

k(k-44)

1.

oz'

:The 'original series adds up to Y. _

.

'WORD COUNTS IN ,JOYCE'S ULYSSES

3.

series gives a mathematical model ,of the occur,

nence of rare words in J4mes Joyce's novel Ulysses.

Among the 260,430 words in Ulysses there are N = 29,899
different' words-

once or twice.

Many are "rare" words appearing only

A few are common words that appear' a

thousand times orNmore.

words here.

We'll study "the rarely appearing

There.4zre 16,432 words (hat appear exactly

once each in Ulysses (about half of N); 4,776 words that

43
appear exactly twice
(about i-1+1 ;

4.3,N),

2,194 wordt that appear exactly 3 times, each,

( about A- N =

N),
.

and%so on
_

In fact, if nj is the number of words Okav appear

exactly j times in Ulysses (j = 1, 2, 3, ...), these n.

words make up a fradtiOn n./N (of the total 1 words) that
J

is rather closely given by

1

j(j+1)'

the jth telt of our series.

Thus we use the series to model 11

(1st modelL, n

I.

30

as

86

This says that the terms of the series

eries (which, you recall,
add up to 1) split up 1 in just bout the way that the
words appearipg oIrce, twice, thrice, etc. in Ulysses

split up the total ordifferent words appearing in that

novel.

HOW GOOD IS THIS SERIES MODEL?

4.

The actual/ number of words appearing once, twice, ...,

ten times in Ulysses is listed in Table 1 along with the
number predicted by the series-model.
TABLE t

Model's prediction
of number of words
that appear
exactly j times

n = actual ? of
words appearing
exactly j times

16,432

- 2,194

3
4

14,9:9c

4,776.

906
637
483
371
298

10

2.22

Source:

2,492
1,495
.997

4,285

4,9

712
534

415
332
272

Relative
error
9.0%
4.4%
13.6%
16.3%
10.0%
11.7%
10.6%
11.9%
11.5%
22:5%

IF

;AO, Human Behavior andthe Principle of

Least Effort.

-The last column provides a ;Imple measurement of the

extent to which predicted and actual values agree:
.
relative error is defined to be
--

RE

'predicted value - actual value

actual value

The

.
.

As an example, for j = 7, the ME is

153?- 4831

51

483

= .10559 = 10.6%.
\

The predicted Acalues are obtained from our series model

as-in this example: for j = 3, the model VT,edicts that
n3 =

= 29899
72--

which we round. to 2,492.

tit

308

2491.58,

87

The predicted values given by the terms of our series

do fol44 the trend of the actual data quite well, but you
*may feel that the specific numbers, (483 vs. 534 for j

= 7,

for example) are not as close'as )tou might prefer.

Shouldn't the modej. match reality better than that?

RE's in"the last column average 12.5%.

The

For most research .

in the natural sciences such relative errors would be

considered large--repeated experiments done with laboratory equipment,
for example, usually yield much more con_
sistent results.
Errors above even 5% make us wionder
about the experimenter's measuring abilities or the design
of the experiment., But we shoul not expect such hardscience accuracy- in a "law".or
del that concerns so
complicated a social-science process as the choice Of
words by one human in treating one novel. fnstead, we ask:
Is this pattern obeyed by a wide range of language samples?

'S.

THE EXTENSIVE RESEARCH INTO WORD-COUNTS-AND

RELATED LANGUAGE PATTERNS

During the 1920's and 1930's, many word-count

experiments were performed by psychologists and linguists,
led by Professor George Kin ley Zipf.of Harvard and his
students. They found striki g patterns in the frequency
. of occurrence of: rarely appearing words, the number of

*paggs between appea4nces of a word, the number of and

spacing between uses of individual letters, syllables,
preixes, suffixes, meanings, etc. Some of the language

'

texts studied (not all for rare-word frequencies) were:

Ulysses by Joyce

--- Stretches of English language newspaper text

--- the plays of Plautus in Latin
the

Iliad in Homeric Greek-

wonks in7Old English, and other, medieval

languages

.part of a Bible in Gothic German

--- traditional oral legends in. Dakota and Plains

Cree (American Indian languages) and Nootka

(an Eskimo language)

-7
6

--- works in modern languageg from German to

Hebrew to Chinese
--- the speech of children at various ageg
--- some schizophrenic speech.

This exceptionally broad selection of language samples

all yielded fiery regular patterns that astonished the
researchers. A few studies failed to support the
1.

pattern* but the evence suggested that important

cross-cultural properties of language were being found.

Linguists puisued this research in the search ear
fundamental structural properties of language.

Psychol-

ogists hoped to explain just whet peis goes on in a

human mind as it calls on its whole history of language

experiences when crafting new sentences, paragraphs, or

books.
One of Zipf's books (see Section 10) contains a
readable survey of these experiments.

It also contains.

the extens.ive,consequences for human behavior that Zipf

put forward as implications.of the research. A too-brief
review of his logic: Zipf claimed that different amounts
4P

of mental effort are exerted bra speaker or writer in

,
choosing words.
Common words, very ftequently encountered
in the writeres past experiences, "come to mind" with
little effort while words met less often -in the past
requir more effort for their up. A human selects words
fr

to exp7ress an idea using the "Rrinciple of least effort.U.

Zipf hoped- to derive the specific quantitative patterns
he had found from such a basic principle (in the same

way that Newton, starting from a few basic assumptions

tsu4ii as'the law of gravity, could derive the motion of

the planets and many other results).

Zipf offered
situations analogous' to writing or language usage where
behArior obeYing a law .of least effort did lead to the
-

patterns found, but hedid not succeed in deriving the

surprising patterns from language structure itself.
*One of the exceptions is another novel by James Joyce, Finnegan's
Wake.

31 o

89
6.

ZIPF'S LAW (THE RANK-FREQUENCY LAW)

A central result of this research'is "Zipf's Law"

also called the "rank-ffequency law." We have. looked at
the

number nj of rarely appearing words that appear with

frequency (number of occurrences) j for j = 1, 2, 3, ....
In a rank- frequency study, one looks instead at thetrank
of a word (1st, 2nd, 3rd, etc.) when the words of a book
-are listedOn 'order of decreasing frequency. Thus the

most-repeated word has rank 1 apd frequency fl, the secondmost-repeated word has rank 2 and appears.f

times, and
so on.- Zipf's Law, also found empirically, is that
'r.f = constant

i.e., that the rank and corresponding frequency are inversely related. Ai an example, Table 2 gives various

ranks, frequencies, and rf products for Ulysses.

TAB4

Actual Rank-FrequAcy Data from Ulysses

Rank

Frequency

rk products

(f)

10
20
30
40

2,653
1,311
926

50

100
200
300
400

556
265
133
84
62

500

50

26,530
26,220
27,780
28,680
27,800
26,500
26,600
25,200

717'

2"4,800

25,000
26,000
24,000'
24;000

.4,000

24000

5,000
10,000
20,000
29,899

1,000
2,000
3,000

26
12'

ir

2
1
1

ti

25,000
20,000
- 20,000
29,899

Source:

Zipil Human Behavior and the Principle of Least

Effort.'

311
c.?

90

The approximate constancy of this third column is

strikini and intuitively unexpected.

And the constant

value obtained is roughly N = 29,899, the number of

distinct words being ranked, or perhaps it is a bit less
than N.

This is di'scussed in comments following Exercise

1 in Section 8.*

7.

A LOG-LOG GRAPH REVEALS OBEDIENCE TO ZIPF'S LAW

There is an easy way to graph the (r, f) pairs from

Ulysses for r = 1, 2, 3, ..., 29,899 so that the closeness

of fitto r.f = k becomes visible.

On ordinary graph

paper, rf = k appears as a hyperbola; it is hard to look

at the graph and determine that we have f = k/r as opposed
to some other similar curve, like f = k/r2 or f = k/r3-2.
'hit these curves are easy to tell apart when gtaphed on
log-log-graph paper.
Notice that rf = k implies logr+
logf =log k.

Thus the points (r, f) fall on the curve

rf = k if and only if the points (x, y) = (log r, log f)
fall on the straight line with slope -1 x + y = log k.

On

log-log graph paper (see Figure 1), the axes are labeled
with values of r and f but, because of the special spacing
of points along these axes, we are really plotting y = leg f

vs. x = logr. Wb will have a goo&fit to r.f = k if the

data fall along a straight line with slope -1, cutting 'both axes at 450.
In Figure 1, the tendency of both curves A, and B to

follow the straight line C is very striking.

(The "steps"

at the bottom-right of both curves occur because, for high

*Zipf's Law rf = k appears to fit many kinds of ranked data beyond
our word counts.
For example, when U.S. cities are ranked by popula-,
tion (so that r = 1 for New York, etc.) then rf = k holds pretty
well, where f = fr is the population of the city with rank r. The
rule fails for cities wod-wide, or for cities in much less urbanized, industrialized societies, and the extent of fic to this law has
been proposed as a measure of a nation's urbanizatidn. Consult the
social science literature for more details and other examples.

t9

312

ranks there are many 'ties, many occurrences of the rare

frequencies 1, 2,-3, .,..)

Researchers up to this point had not explained Zipf's

.

Law, or the series model that we began with in this paper

or other patterns.

10000

41000

10

10

100

1000

10000

'RANK
e

Figure 1.ft- Data that precisely obeys Zlpf's Law would,

graph like C, having slope -1, to whicY1 curves A and B
should be compared
Curve
'consists of all the (r4f)
data pairs for Ulysses, not just the few given in Table
connected together into a curve. Curve B is a
similar rank-frequency graph for a sample of 43,98
running words of Amdrican newspaper text,,studied y
.R.C. Eldridge.
(The Ulysses data was created by Hanley
and Joos, but first graphed by Zipf. Soucte: Zipf,
Human'Behavior and the Principle of Least.Effort.)

9.

3106
9

:4,

92

8.

EXERCISES:

DERIVING A NUMBER-OF-WORDS LAW

4

We have studied two parts of Zipf's research, which

we summarize as follows:
(A)

The rank-fre.quency law fr = k/r gives the approxi-

mate frequency (number of appearances) fr of the,

r

th

-most-commonly-appearing word in the language

sample, for r = 1, 2, 3, ..., N.

(B)Thenlonber-of-wordslawn.=N/j(j+1) tells how

many words (among the N different words of the
language sample) appear exactly j times, for j = 1,
2, 3,

....

Both are empirical laws--they work quite well for a wide

variety of languge samples.
So far we have no derivation of these laws from obvious{ or widely accepted facts,
no clear explanation askto why they should be true.
\These two laws are related to each other and that
is worth our study--if one follows from the other, they
are more believable together than either is by itself.

Therefore, let's assume that (A) is true and'try to (

deduce a number-of-words law from it.

Specifically, let's
try to calculate nl, the number of words that appear
exactly once (i.e., that have f = 1).

The rank-frequency law predicts frequencies fr between

1 and 2 for all words with ranks k/2 + 1 up to.k:

1 K f< 2 a01 <

49

<

< r < K.

Thus, a total of k/2 words have theoretical frequeRcies

f in the interval [1, 2).

However, frequencies must be integers; fractional

frequencies do not make sense.

Let's decide that we will

always round f downward to the next lower integer. Then

.0'

314

411

93

ft (1, 2) becomes

= 1, and n1, the number of words

with f = 1, is
k

7- 177

nl

This looks promising--if we are going to.derive nj =

, N /j(j +l) from (A), we need that denominator 1.2 in n1.
But that k

the numerator?

Maybe the correct constant

k in the rant-frequency law i5 N? We'll have to test that
'idea later. First, extend our result for nl by doing
Exercise 1.

Exercise

k/r for r

Assume that f

a) Show that fe
re

k
(Fa

1, 2, 3, ....

j+1) occurs exactly for ranks

.

b) If we round f e (3, 3+1) downward to the Integer value f = j,

show that
ni

for any 3.

..111-

Thus we can deduce (B) from (A) if we agree to round f

downwArds and if k = N.

We sould test whether k = N'empirically by trying

it on many language amples. We can starthere with
Ulysses, which contains N =. 29,899 different words. The
r and f data in Table 2 can be used to get h comparable
value of k. Let's exclude the data for r = 10,000, 20,000,
and 29,899 because these (r, f) pairs are located in the
"steps" of the ,(r, f) graph where r changes while f does
not. arid those rf products are not very constant. When
we average the r.f products in Table 2 for 10 < r < 5,000,
et k = 25,874. Thus k # N. We have k about 13.5%

smaller than N' in thig one example.

.
V

315

94

This is no time to quit on the problem--the

Wait.

values N/j(j+,1) art also 10-15% too large for fhe actual
nj of Ulysses in Table 1 (except for j

= 1, 2).

We could

correct that by decreasing N/j(j+1) to k/j(j+1). . Thus

we propose
(2

nd

nj = j(p+1)

for all but the smallest j.

We cannot apply this model

for all j because'

N =

n.

however,
k

, r

TTITT)
But the 2

nd

j(j+1)

model may work well for all but the

mallest

few values of j, which are special cases requiring

eir

own formula.

Ulysses data comparable to that in Table 1

appears in Table 3. We must be cautious in,concluding

that the 2

nd

model will do this well fdr j > 10 or for

language samples other than Ulygses.. The second model

does not seem to appgar in the psychological literatEre,
probably because Zipf deduced yet another number -of -words
formula from tife rank - frequency law.

TABLE 3

Additional NuMber-of-Words Predictions vs. Ulysses Data.

nd
2
model*
predicted

true
n.

n.

__1_

:1

16,432
4,776
2,194
1,285

906

637,

483

3 -

8
9

10

1
,

298
222

12,937
'4,312
2,156.
1,294
862
616
462
359

rd

3
model*
pfedicted

RE

n.

21.3%
9.7%
1.7%

34,4'9
6,900
2,957
1,643
1,045

'1.6%,.

287
235

4.9%
3.3%
4.3%
3.2%
3.7%
5.9%

724
531

406
320
259
\

*All

calculations are based on k = 25,874.

31 6

RE

109.9%
44.4%
34.8%
27.9%
15.3%
13.7%
9.9%
9.4%
7.4%
16.7%
.

95.

Surely you wanted to object to the "rounding", of

fc [j, j+1) to f =

After all, would you round 3.01,

3.3, 3.49, 3.51, 3.99 all to '3?
It would also mean that
fc [0, 1), which is predicted by f = k/r for ranks k <x
N, is .rounded to f = 0, although each of the words

with these ranks appears in the language sample.at

least once.
Zkpf proppsed instead to round fc[1/2, 3/2)
to f = 1, fc (3/2, 5/2) to f = 2, etc.
1

Exercise 2.

Assume that f = k/r for r = 1, 2, 3;

....

a) Show that fe [j - 1/2, j + 1/2) occurs exactly for words'

with ranks
j
b)

< r <

+k 1/2

- 1/2i'

If we round fe [j - 1/2, j + 1/2) to f

(3rd model)

nj

(j - 1/2)(j + 1/2)

j,'show that
for any j.

This third model is the one given by Zipf.

It leads us

'

to ask:
a

Exercise 3.
n.
j -1

should equal N, the total of di

study.

Sum the series suggeste

s in the book'under
n Exercise 2,

formula,

PO

2 j+ 1

by simplifying the partial SUMS in much the way E 1 /j(j+1) was

summed early in this paper.

Eni
SinceExercis%,..3tellsos.that.

N, we know we
cannot use the 3 rd model for all j, based on k and N from
Ulysses. sAs with the 2nd model, for low j the predicted
values are far too large. Table 3 shows a very, pior fit

44

317

"sr

96

between this model and the Ulysses data; for much larger
values of j the fit may be much 'better.
So it goes!

In three tries, we have not achieved

a trouble-free model.
Exercise 4.

Without finding the sum, give more

lian_one,proof that

co
1

is a convenient series.

Mention the convergence tests you use.-

Exercise5.
Assume rf = k and decide

Suppose we decide to round upward:

toreplacefe0-1,flbyf.j.,miatruleforri.fol lows?
a better model than the ones we have discussed?
equivalent of Table 1 for this 4

th

model.

Is it

Prepar4the

How did you decide whether

or not it is better than the first 3?

The series result

T

, 1(00.1)

can be used to find the sums of other series./ Two

examples appear as Exercises 6 and 7.*

Ex/se 6.

irst show that

co
J.

k ,1001,
k

,+

2.3

1 .2

I..

"47

13

n=1 (2n +
(2n - '4

Use this result to show

c0

+
1.3

4.* 1 +
3.5
5.7

n1
=

2n - 1)(2d + 1)

*Thanks go to
Glessner of Central Washington University
for suggettIng Exercises 6 and 7.

31 8

97

Exercise 7.
we

'start with the result in Exercise 6:

1

.1

-17

35 -57

't.

and use the same "sum up two terms at a rime" method (as displayed

in Exercise 6) on it, show that we get

1

9.

(4n + 1)(4n + 5)

1.5 + 5.9 + 9.13 +

MANDELBROT'S EXPLANATION OF THE LANGUAGE PAi.tigNS

Zipf'sLaw and other striking patterns found through

word-count sorts of experiments op natural (i.e., human)

languages-were finally explained by scientists working
on very different Problems, problems related to artifici'al languages.

Zipf and his colleagues had examinedthe structure of language and the process of writing or,

speaking; now-Norbert Weiner and Claude Shannon led the

study of communications channels. Human speech and
writings, electronic signals sent over telephone lines,
a

messhges sent in Morse code, radar signals sent out and

received after bouncing back, coded data moving / from IBM
cards into a computer's electronic memory, all are examples
of information being coded and sent by a transmitter
(speaker, writer, telegraph key, user, etc.) then received,
decoded, And interpreted by a receiver (listener, reader,
etc:). The researcher asked: How colkild, ilzformation be
most efficiently coded and sent so that it would be
received at'lowest cost-and with high accuracy?

How much
repetition. ("redundancy") should be included as- a check

on the"accuracy of the message received? Their main goals

here the efficient design of high speed, high volume, highs
pa curacy man-made data channels for use i

computers,

J'ternational telephone and microwave syste s and-military

0

applications, but the linguiSts and psychol gists noticed

at once this re earch w
relevant to he study of
.
human language communications, t
1

5'

319

98

This research led to an anticlimatic completion'Of

'the project begun by Zipf and` his team.

In 1953-54,
Benoit Mandelbrot showed that the number-frequency, rank
frequency and oiler patterns .found by Zipf will always
,

arise in any language satisfying these two assumptions:

1.

The language is made up of words- -small units

of information separated by spaces.

.2.

The transmitter encodes and the receivdr,

decodes word by word- -that is, the speaker

(or writer) formulates,..a4d speakope word

at a time and the listener.(reader) listens'
aZ

and interprets one word at a time.

The main point is the presence of a space between units
of' information.

By.random processes this spacing, and

the word by word handling of messages, accounts for the
-411-

patterns\

There is no need, in'explaining'the patterns,

to claim that James Joyce, while writing Ulysses, was

choosing words using unknown "universal laws" of language
structure at, some deep almost-unconscious level of thought.

Iqstead, we simply claim that JdFce was choosing his words

one at a time to convey his meanina.
The space-betweenwords structure of English then suffices ,to produce the
patterns.
Mandelbrot showed this by.using a lot of
advanced mathematical statistics.
Zipf's ideas persisted for a while. The applicability
of Shannon's work to human languages was challenged and.
some of Mandelbrot's assumptions were questioned, by
H.S. Simon and others.
Simon, in 1955, published a/ter4.t.4-re'eNplanations o

using the idea tha

the more1likely i

Zipf's Law and other patterns,.'

the more prior usage.a word has had,

is to recur.
.

Mandlebrot has won the day, however.

My most
recent reference, in Mathematics and Psychology, edited
.

40)Y George A. Miller., John Wiley and Sons, New York, 1964,
includes 'Otis quote from Barbel Inhelder and jean Piaget
on page 249:
.
4

320

99

... during the 1930's G.K. Zipf stirred up

considerable intbrest in various statistical

regularities that he uncovered in his, analysis

of word frequencies. 'Twenty years later the

mathematician Benoit Mandelbrot was able to
demonstrate that.Zipf's laws were attributable
to random processes and implied Ao'deep lin.

guistic or psychological consequence's.

10.

SOURCES

I first met this application in the essay "The

Sites of paings" by'H.A. Simon in Statistics: A Guide
to the Unknown ed. Judith Tanur, Holden-Day, 1972',

pp .195-202. -This paperback contains many short essays

that

ow the'applicability and practical uses of star

tis i cs, especially the difficulties of statistical

experiment design. Most are only modestly mathematical.

The work of Zipf and his colleaguels well summarized in G.K. Zipf, Human Behavior and the Principle of
Least Effort, Addison-Wesley, Cambridge, Mass., 1949,,
Chapters 2, 3, and 4.

The original Ulysses data, complete, appears in

M.L. Hanley et al,-Word Index to James Joyce's

*loses, Madison, Wisconsin, 1937.

o

The mathematics used by Zipf to relate his_rankfrequency law to the number - frequency -law for rare words,

presented in Exercise 2, was.presented in G.K. Zipf,."Homogeneity and heterogeneity in language", Psychological

Record, 241918), pp.347-367. A more general argument
Martin Joos appears in a:book'review of Zipf's The PsychoBiology of Language, )-loughton Mifflin, Boston, 1935 in

Languages12 (1936)

Joos, while contributing

to Zipf's rigor, is not uncritical.

A good summary of Mandelbrot's results and their'

meaning maybe found on pp 60-69 of R.D. Luce, edc,

321
'T.

0:

100

Developments in Mathematical Psychology:

Information,
Learning and Tracking, Free Press ol Glencoe, Illinois,
1960.
Part I (by Luce) is "The Theory of Selective

Information and Some of Its Biological Implications" and

covers Shannon's work and some brief mention of Zipf.
I
did not obtain the papers of Mandelb7oi-i Miller and Simon

referenced there but relied on Luce's 'rendition

work, which I hope I have not misrepresented.
The

their...)

bibliography on 'pp 110-119 of Luce (above) will direct

you to the original literature.

The Project would like to thank William Glessner of.Central

Washington University, Ellensburg, Washington, and Mitchell Lazarus
of Edvation Development Center, Inc., Newton, Massachusetts for
their reviews, and all others who assisted in the production,of
this unit.

This material was class-tested and/or student reviewed ino

preliminary form by: Stephen Corder, Southern Baptist College,
Walnut Ridge, Arkansas; Paul Nugent, Franklin College, Franklin,
Indiana, and George C.T. Kung, University of Wisconsin, Stevens
Point, Wisconsin,-and has been revised on the basis of data
received from these sites.

This material was prepared with the Support of National

Science Foundation Grant No. SED76-19615 A02. ,Recommendations
expressed are those of the author and do not necessarily reflect
the views of the NSF. or the copyrightsholder.

322

'11.

ANSWERS TO EXERCISES
k

1 _r+f = k and j < T < j+1 4, j

< j+r

< r <

Thus a total of

.k

j(j +1)

ranks r have associated fe[j, j+1).

2.

z
,,i
+
1. "

Similarly 7

7 .1

k
_--< r<

k
1

and

3.

.2-

LJ

(j '-

I) (j

1.
+ 1-)

%..

j 7 7

1.1

r:
1.)

Using partial fractions

,

_or

nj

J2 y
4

the partial sum is

1

"n

CUT - 377)

W-2')

*L-

[(in- 1)/2

(2n + 1)/2

172

(2n +1)/2

Thus the series ,sums to 2.

But E n. = 2k >> N
makes Zipf's model also only* partially useful.
A

4.

Comparison-and integral tests are easy enough.

5.

The rule is f = j+1 for theoretical fe(j, j+1],

i.e., for ranks
k
< 3-

(using the solution method.of Exercise 1).

323.

Then

or.

JT

-j+1

j(j+1)

for j = 1, 2, 3,....

Thus nl is excluded, which

makes no sense, and the split-up'of k totals

.
r
2

.1

J(J+1)

'

also not nicely interpretable.

.

nj+1
+ .does
not h elp

nj

Shifting the terms

ps fit the Ulyss rr

"data

better, as an eyeballing of Table 1 will, show.

.

6.

All thit's missing is

.

k(k+1)

4.1

7W (7W77I
n=1

2h +'1
Nta.
ce`

4n

1
L.

n= 1
7.

2n ((2n -

1)

(2n +1)1

..n1
=

(2n - 1) (2n + 1)

First we 'show
1

1.3

375

111 411
3(1
5)

n=0

n=0

4n + 3

377
111
7(3

Ill+ 1 1
11(9

(4n + 1)

1-1)

(4n + 5))

1
8n + 6
4rTrS (4n + 1)(4n+ 5)

2
.1

The result then follows at once.

(4n +

1)

(4n +5)

al

20

324

UMAP MODULE 216

MODULES AND
MONOGRAPHS IN
UNDERGRADUATE
MATHEMATICS
AND ITS
APPaLICATIONS

>p>
lzb

170

-2

..

o,

"

"

I>

Curves'and their
Parametrization
by Nelson L. Max

VN

ce,

..

f...-...1

CD

...

). m >,
l'A.

1,1

N1

..

=
CD

er
0

ca N
...,

>,

er
'?"

e.,

0
Introductory TopolOgy

Birkhauser Bostop Inc.

380 Green Street
Cambridge, MA 02139

'Intermodular Description Sheet:

Title:

Author:

UMAP Unit 216

CURVES AND THEIR PARAMETRIZATION

Nelson L. Max
Qepartment of Mathematics and Statistics
Case Western Reserve University
Cleveland, ,Ohio 44144

Review Stage/Date:
Classification:

IV

6/30/80

INTRO TOPOLOGY

Prerequisite Skills:
1.
Understand the representation of points in the plane by
Cartesian coordinates.
2.
Understand the trigonometric functions of sine and cosine, and
the measurement of angles in radians.
3.
Understand thl,natural logarithm function.
4.
Sum the geometric and harmonic infinite series.
Output Skills:
1.
Define a parametrized curve, image of a curve, orientation of
a curve.
2.
Given the image of a curve, write the equation
one or more
parametrized curves which trace the image with a given orientation.
3.
Given the image of a curve, describe ail its possible orientations.

Other Related'Units:
The Alexander Horned Sphere (Unit.232)

The Project would like to thank Joseph Malkevitch of York

College of CONY and Anthony Phillips of SUHY at Stony Brook for
their reviews, and all others who assisted in the production of
this unit.
This unit was field-tested and/or student reviewed in pre- .
liminary form by W. Hugh Haynsworth of The College of Charleston,
Charleston, South Carolina; C.iWagner of Pennsylvania State
University, Middletown, Pennsylvania; Tom Haighiof Saint John's
University, Collegeville, Minnesota, and Phillip Lestmann of
Bryan College, Dayton, Tennessee, and has been revised on the basis
of data received from these sites.
This materiai was prepared with the'partial support of National
Science Foundation Grant No. SED76-19615 A02: Recommendations expr ssed are those of the author and do 'hot necessariiy reflect the
s of the NSF or the copyright holder.

4) 1980 EDC/Project UMAP

All rights reserved.

II
.,

CURVES AND THEIR PARAMETRIZATION

..

by
Nelson L. Max
Department of Mathematics and Statistics
,Case Western Reserve University
Cleveland, Ohio 44106

TABLE OF CONTENTS
..

1.

THE DEFINITION OF "CURVE"

2.

PARAMETRIZATIONS OF THE UNIT CIRCLE

3.

OTHER PARAMETRIZED CURVES

4.

CONTINUOUS CURVES

5.

CURVES WITH UNUSUAL PROPERTIES '

6.

EXERCISES

7.

MODEL EXAM.,...

13

15

(
8.. ANSWERS
TO QUESTIONS IN TEXT
,

17

9.

10.

ANSWERS TO EXERCISES

ANSWERS TO MODEL EXAM

17

18
C

32;

1.

THE DEFIVITION OF "CURVE"

Webster's Dictionary defines a curve as "the path

of a moving point." If the moving point were the point
of a pencil, it could trace out the curve on paper.

,For example, the point of the pencil on a compass

might trace out a circle.

/-

Webster gives another more technical definition of

a curve: "A line that may'Ue,precisely defined by, an'

'

equation in such a way that its points are functions of

a single independent variable or parameter."

We can

think of the variable or parameter *time and call !,/t.,

Then the coordinates of the moving point, x(t) and y(t),

are the functions of time.

328

a%

yr

,/
If we imagine the pencil as making
curve every second, these chits will show

of on the

ow the curve

has`beentraced. In particular, thelr'spacing will indicate the speed of the moving point. Here the point is
speeding up as it moves to tie right.
p

2.

PARAMETRIZATIONS OF THE UNIT CIRCLE

Below is a circle which is traced counter-clockwise

at a uniform speed,of 150, or 3 rr7 radians, every second.

When it.i& finished in 24 seconds, it will have 24 evenly

spaced dots.

The Coordinates.of the moving point are

given by the equations

.

x(t) = cos,(

t);

y(t) = sin, TT ti.

329
CP
1

tt

There are many other ways to trace the same circle.

In the figure below we see only twelve evenly spaced
dots, so the equations might be
x(t) = cos

(:iri tj;

y(t) = sin

t).

46

However, they might also be

x(t) = cos

t);"

y(t) = -sin (g- t)

which would trace the circle with the same constant

speed in the opposite direction.

330

We might also trace the circle by having the

coordinate move at a uniform speed from 1 to
example,
,x(;.)

= 1

1, for

0 < t < 6

t.

-y(t) =

t2

These equations work only to trace out a semi( circle.

Here the dots- are not evenly spaced.

They are
closest together at the top and bottom, indicating that

the curve is traced most slowly there.

The tracing point

actually moves infinitely fast at thd left and right
sides.

st

QUESTION A:

Can you find similar equations to trace out

the bottom semicircle, for 6 < t < 12?

\'

331

.44

All these different fu 'nctions define diff&rent

parametrizations of the circle. jie say that they define

different parametrized curves.

The set of points which

a curve passes through is called its image.

All the
different parametrizations of the circle have the same
image.

In addition to defining the speed,"a parame.trization

also defines the order in which the points in .the image
are traced.

Thus, a point traodng a clockwise circle

moves in the opposite direction from a point tracing a

counter clockwise circle, so it passes through the points

in the image in the opposite order. Thus there aretwo

orientations to the circle, clockwise and counter clockwise.

3.

OTHER PARAMETRIZED CURVES

The situation becomes more complicated if the Curve

i8 not one-to-one, i.e., if it passes through some points
more than, once.

Here is'acurve which crosses itself,

yassinghrough the point B twice. One orientation would
be to pass through the points on the image in the order
CDBE.

lb

.0

Another pethod of tracing the sage image, shown

Partly,completed here, would pass through the points in
the order ABDCBE, making two corners at B. Two more
.

332

orientations would start at. E and end at A.

What are they?

QUESTION 13:.

Although a curve can pass through certain, points

on its image more than once, it should not cover whole
sections more than once.
Thus, ABCDBDCBE would not give
a Valid orientation for the curve.
There is/not

g wrong with a corner in a curve.

A mathematial

e is not necessarily a smoothly

curving line, but may have corners, and can even consist
entirely of straight lines. Far example, a square is a
curve. Gan you find a set of equations which describe
this curve?

The trick is to find separate formulas for the

different side's of the square, just as separate formulas
could be used for thetwo,semicircles making up a circle.

$
6

333

The functions below define two sides of the.square.

0 < t < 1

(t,

x(t)

1,

1 < t <

0,

0 < t <

t -.1,

1 < t <

y(t) =

y
10,

ti

(0,0)

(1 , 0)

QUESTION C:
2

Can you continue these functions for

< t < 4 to define the other two sides?
4.

CONTINUOUS CURVES

A natural subcollection of the class of parametrized

turves are ones for which the tracing point moves dentinuously, without jumping. This condition'is equivdlent to
requiring that the curve can be drawn withou41 i lifting ,thee
pencil from the paper!"
.

is

..t

"0

ti

'334s

81,,,

It is also equivalent to requiring that the two

coordinates x(t) and y(t) be continuou*sfunctions of
the time parameter t.0

.-.

If one of the coordin'ate functions is discontinuous,

for example,

--,

0 <

<

1 <

<

x(t) =
t+1,
1

y(t) = 7 t

i
1

the resulting image, shown below, may have a gap in it.

If both x(t) and y(t) are continuous, the result will IA
a continuous parametrized curve, called simply a curve
for short.
IP.

..,

-1.

..:

...re'

..

8
--.......

II

"33-

...

I'
'4:
S.

CURVES WITH UNUSUAL PROPERTIES

There are many strange exam

es which satisfy this

definition of curve.
- Example 1
L7
of le,

For example, if lnx denotes the natural logarithm

then the equations
x = t cos(2n ln t)

y = t sin(2n ln t)

which make sense on the interval 0 < t < 1, can be eb

tended to a continuous function on O.< t < 1 by defining
x(0) '. y(0) = O.
This gives a curve, called the logarithmic spiral, which has infinitely many (similar)
turns near t = O. Nevertheless we will prove that it
has finite length.

41

Consider the first turn of the spiral, from t . 1

to t . e -1
Suppose it has length L.
.

9'

336
a

illp

The next turn of the spiral, from t =

looks exactly similar, but e
Le

-1

-1.

-1

-3

-2

is large, so its length is

Similarly the length of the next turn is Le

-Thus the length of the whole spiral is L + Le

+ Le

to t = e

-1

-2
.

+ Le

..., a geometric series which converges to

L/(1 - e - ,1 ), a finite length.

ti

Problem

Verify that the turn of the spiral from t = e

-1

to

-2

is similar to the tun from t = 1 to t = e -1 with

1
constant of proportionality .
t = e

There is also a spiral which winds toward the origin

in

uch a way that it, has infinite length.

It is the

hyperbolic spiral
= t

cosh)

y = t sin(1).

which can again be defined for 0 < t < 1 by letting x(0)

= y(0) . O.

The length of the spiral must be at least

as long as the length of the inscribed polygon ABCDE...

which we will show is infinite.

If 0 is the origin, then

AB and BC, are both longer than OB, while CD and DE are
both longer than OD...and so forth.

So the length of the

spiral is greater than twice the sum otthe lengths of

10

337

the line segments from thorigin to the "y crossings,"

the poirits where the curve crosses the y-axis, the first

three of which are-B, D, F.

,
4

How long is OB? Consider the intersections of the

t
spiral with the y-axis (x=0). Since t # 0, we have
(IT1

-fj

= 0

3n

cos

so that,
n

Sn

7n

to- I' -2' -27' T'

i.e.,
,

t = 2,

2
-s,

2-'2 '

3-,_ 7,

Since t < 1, the acceptable values for 1.!are

1,
tare -r, SIT

The length of OB is thd y value when

n

3n

2)

t21

It = 3j: y t-sj

2
= -3 , so .0B, = 3.

11

333

n
the graph of y = t tin (1.1,,with
the poiRt.
B' = [3, :T
giving the y coordinate of the oint B,
i.e., the length of OB.
Similarly D'= (5,, i gives the

Below

)
2is

y coordinate for the point D.

The length of the spiral,,

which is greater 'than the length of the polygon, is this

greater than the series
+ 2 R)

which diverges.

[.;_i

= 4 1

o,the length is infinite.

Example 3

The infinitely wiggly graph of y = ts4n (i) alio has

infinite 16112th by a similar argument.
t.

Other Examples

There are functions wItose 2'raphs have infinitely

many wiggles, and infinite length, between any.twolioints.

33j
O

The snowflhe ecuzme also has infinite length between any two of its points.
er

Among the strangest examples of curves, are the

"space filling curves," which pass through every point
in an area such as a square.

6..

1.

EXERCISES

How many different parameterized curves have the image shown

beloy?

13

340

C.'

2.

',.

How many different oriented curves have this image?

3. 'Find two different parametrized curves, defined for 0 < t < 1,

which have this piece of a parabola as an image?

oe

3.

Find the equations for the parametrized curve which traces this
equilateral triangle at uniform speed in a'counter-clockwi,se
direction, starting at the vertex (0,0), 'in the time interval

r\

0 < t < 3.6

ti

r
a

14

7;

MDDEL EXAM

State whether eaph o? the curves described below is an

oriented curve, a parametrized curve, or the image of a
curve.
a)

The contrail left by a jet plane.

b)

The script letter m, drawn from left to fight.

c)

A marathon course.

d)

The straight path of an automobile, accelerating

e)

A figure eight.

uniformly'from 0 to 60 miles per hour in ten seconds.

f)
2.

OP
The path of the tip of a ,second hand on a wall clock.

If ABCD is the curve defined by tracing the first three S'idbs

of the hexagon below, with constant speed in the time interval

0 < t.< 3, find the formulas for x(t) and y(t).

----

15

-3 4

1.

Describe all possible orientations of the figure 8, starting

/at the top point P.

1'-

4.

What is meant by the image of a parametrized curve?

5.

Give a parametrized curve whose image

{

(x,y)

ID <

x < 1, 0 <

y < 1, x2 = y31.
A

'

16
4

343

I
8.

A.x(0,...71,-t- 3

ANSWERS TO 'QUESTIONS IN TEXT

X7,7=

y(t)

(t -3)2.- -1 Xtit -t2

B.

The order EBDCBA and the order EBCDBA.

A set of the equations for all

0<t<1

72

four sides of the square are:

f
10
0 < t <
o

x(0.=

< t < 2

It3 -t,
0

2<

t < 3

3 <

t < 4

9.
i.

infinitely many.

2.

Sixteen.

3.

(Possible answers)
x(t) = 1 - t,
x(t) = t,

t-1,

y(t) =
1

4-c,

< t < 2

2<

<3

3. <t4
=it

ANSWERS IQ EXERCISES

y(t) = (1-02;

y(t) " t2;

I

x(t)

Y(t) = t4;411/t5

x(t) = iF,

At)

4.`

(2t,
x(t) =

3-c,

=t
0

0 < t < 1
1

< t < 3

y(t)

/(c-1),
3(3-0,

6 < t <
1

<

c < 2

2 < t < 3.

177

*344

ANSWERS TO MODEL EXAM

10.

1.

oriented curve

a)

image

d)

parametriFed curve

b)

c)

image

e)

oriented curve
parametrized curve

f)

2.
.

1 + 2t,

2 + t

A(c) I

b 7

0 <

<

<

<

:11 (t-1)

3.,

PABCDEBFP, PABEDCBFP41/ , PFBEDCBAP, and PFBCDEBAP.

5.

Possible answeri
x{t) = t, Y(t) = t2/3

0 < t <

x(t) =

0 < t <

y(t) = t2

<

Y(t) =

2< t< 3

9 <

<

t'<

18'

34 5
-tt

umap

UNIT 231-

THE ALEXANpER HORNED SPHERE

by

DATE

MODULESANDMONOGRAPH8INUNDE
MATHEMATICS AND ITS APPLICATIO.NS

Nelson L. Max
Department of Mathematics'and Statistics
'Case Western Reserve University

.L Cleveland, Ohio. 44106

THE ALEXANDER HO
by Nels

ED.SPHERE

L. Max

ABLE OF CONTENTS
1.

INTRODUCTION
1

2.

3.

4.

'

THE JORDAN CURVE AND SHOENFLIESS THEOREMS

THE HORNED SPHERE'

..... ..... A .

5.

THE EXTERIOR OF THE HORlitD SPHER6

PROBLEM

.4. .

SIMPLY CONNECTED SETS

,0

.
10

a
SOLUTION

b.

11

t
t

dl
41%

INTRODUCTORY TOPOWGi

ede/umap /55 chapel stinewton, mass 021

.346

347
1

/,.

Intermodular Deacription Sheet:

Titlit

7
UMAP Unit 231

THE ALEXANDER HORNED SPHERE

MODULES AND MONOGRAPHS IN UNDERGRADUATE

Author:

MATHEMATICS AND ITS APPLICATIONS PROJECT (UMAP)

Nelson
Max
Department of Mathematics and Statistics
Case Western Reserve'University

AL

Cleveland,,Ohio 44106
Review Stage/Date:

111

6/20/77

The goaf of UMAP is to develop, through a commupity of users

and developers; a system of instructional modules in undergraduate
mathemptics and its applications which may be used to supplement
exilting courses and from which complete courses may eventually
be built.

Olassification:

INTRO TOPOLOGY.

Suggested Support Material:,

Films:
,

The Alexander Horned Sphere, 21 minutes, silent, color;

Zooms, 7 minutes, color, musical track.'
Available,from:

The Project is guided by a National Steering Committee of

mathematicians, scientists and educators.
UMAP is funded by
a grant. from the National Science Foundation to Education
Development Center, Inc., a publicly supported, nonprofit
'corporation engaged in education research in the.U.S. and abroad.
PROJECT STAFF

International Film ,Bureau332 'South Michigan Avenue

Chicago Illinois 60604

)toss L. Finney
Solomon Garfunkel

Director
Associate Director /Consortium
'
Coordinator
-Associate Directdr for. Administration
Coordinator for Materials Production
Project Secretary
'
Finaneial/AdMinistrative secretary
Editorial Consultant
0
Editorial Consultant
Editorial Consultant

Prerequisite Skills;
1
1,
Parametrization of simple closed curves.
2.
TopoIogical*defimi:Fions of connec edness, o Pen .and closed.,
sets, continuous (unctions, hom
rphisms.

Felicia Weitzel
Barbara Kelczewski
Dianne Lally
P4ula M. Santillo

Output Skills:

Edwina Michener

1.

Louise Raphael,

2.

Understand the construction of the Alexander honed

here.
Discover properties oE the horned sphere as a counterexample
to Shoenfliess Theorem for the standard sphere SI.

Jack. Alexander,

..NATIONAL ST CRING AMMITTEE

tior in

Other.RelatedUnits:
.1%
Curves and Their Parametrizations (Unit 216)

t'Steven,..t Brams

O
0

Ilayron't arkson
'' James D. Forman, ,
Ernest J. Henley
Donald A. Larson
1 William F. Lucas
Frederick Mostelier
Walter E. Sears
George Springer
Arnold A. Strassenburg
Alfred B. Willcox

'4"

Turning a Sphere Inside Out'(uhlit 289)

MIT (Chairman)
*
New York University
Texas Southern University
Rochester Institute of Technology
University*of Houston
SUNY.at Buffalo
Cornell University
Harvard University
-University of Michigan Press
Indiana University
SUNY at Stony Brook
Mathematical Association of America

The Project would like to thank Joseph Malkevitch and

Solomon Garfunkel for their reviews and all others who assisted
in the production of this unit.
This material was prepared with the support of Nation
Science Foundation Grant No. 5E076-19615.
Recommendations
expressed ar$ those 0 the author and do not necessarily reflect
the views of ,the NSF nor of the National Steering Committee.

A 1978 EDC/Project UMAP

,

All Rights Reserved.

348

34.9
S

00

Suppose we have such asimple 'Closed curve g.

1.

ar)
In this unit we will describe
the image of a homeomorphism from the standard sphere into three dimensional
space, whose exterior is not homeomOrphic to.the exterior of'a standard sphere. It.is called the Alexander

non-empty.and open, and in particular, gs(S 1 ) is the complete frontier of both A and B.
The previous set, A, is
called the interior of the curve, and the unbounded
one,
B, is called the exterior.
(See Figures2.)

horned sphere because it was discovered by J.W. Alexander

in 1924, and looks as if it has grown horns. We
will start by discussing the situation for simple closed
curves in the plane.
Then we will describe tize-horned
sphere, and suggest the idea behind the proof that
it
has a non-standard exterior.

,2.

The

Jordan curve theorem skates that g(S 1 ) separates the

plane into the union Of two non-empty connected open
sets
A and B. That is; R 2' - g(S 1) = A U B, A and Bare both

INTRODUCTION

THE_JORDAN CURVE AND SHOENFLIESS THOEREMS

A simple closed curve is a closed curve which does

If it is parametrized by a continuous

riot cross itself.

function f frothe interval (0,1] to the plane R 2

then
f(a) = f(b), for a < b, if and only if a
= 0 and b = 1.
(See Figure 1.)
,

Figure 2

The Shoenfliess Theorem states in ad,dition that

the homeomorphism g, which is defined only
on the unit
circle S 1 can be extended to the whole
plane, so that
it takes the interior of S 1 .to A and the exterior
to B.
Thus A and B are homeomorphic to the standard
"round"
,

regions.

We will not prdVe either of these theorems

here.
3.

Figure 1

If S

stands for the unit circle,((x,y)LR 2 Ix 2

+y =1):
we may also think of our curve
as a homeomorphism g of
S1 into the plane.
This means that g is a homeomorphism
1

of S

onto its image g(S 1 ),

although g(S1) is not'neces-

sarily the whole plane,

350

THE HORNED SPHERE

Let R 3 denote the three

dimensional space.of
triples of real numbers (x,y,z), let
2

f(x,y,z,)eR 3 Ix 2 + y 2 + z 2 = 1] be the surface of

standard round sphere in R 3
and let g be a homeomo
ism
of S2 into R3. Then the gederalization
of the Jorda
Curve Theorem, sometimes called the Jordan
Separation
Theorem, states that R 3 - g(S 2 )
= A U B, the union of
two non-empty4connected open sets, and g(S 2
)
is the
S

351

,.
t

complete frontier of each.

Again, A represents the in

`terior of the distorted sphere g.(S 2 ) and '13 represents

the exterior.
(See Figure 3.)
-

Ficiure 3
.

Figure 4

The analogous generalization is actually true in

any number of dimensions.

need only move points which liebside the'leur discs

However, the generalization.of the Shoenfliess

Theorem is not true in the-case of g(S2) and the Alexander

Horned Sphere is ecounterexample.

In this section, we

'will const"ruct a bomeomorphism g

s2 into R3, such

00'

CO3, C
10 '

and C11. (Figure 5.)

L.

that the exterior B of g(S2) is not homeomorphic-teOe

exterior of the standard sphere'S,.
In particular, then,
it will not be`possible to extend g to a homeomorphism
Of the exterior of S2 onto B.

.Toconstruct thehorned'sphere, we start with a

'round sphere as the first approximation and push out
.
,a pair of horns to make the second approximation. We
.

can'do this by taking two pairs of cencentric discs on

the sphere,.Dgc:Co;sand Dic:Ci.' Then we k e p S 2c - (CGUCI.)
fixed, push Cg-- Dg-afid C1 - D1 to'tbe tubular sides
,

of the horns, leavink-'circulir caps made fronicT0--and.Di,

as shown' in Figure 4.)'
r,

From, the flat .ends of-these horns we push out two

new branches in the same way to gef the third'approximation. -St leeks like a pale of crab's olmws parti",ally

Interlocked but notclosed or touching,

To do this, wp

,Figure 5

We repat'again and again, gfowing, new branches

on the tops of each of the aid branches.. Since each.
new pair of clat4s. is a reduced version of the previous

pair, the total amount any point moves is dominated by a,

geometric progression. ,Therefore, the approximations

3
rf

s'

353.

converge uniformly to a continuous limit function g from"'

S2 to R3.
By the wdy the construction isarranged, g is
also one-to-one, so it can be proysd that g is a homeomorphism. (Figure 6.)1

."

,
Figure 7d

Figure 6

We can rotind'the
corners of our surface to mike
anew function g which is
smooth (Figure 7), except at
the points which belong to an infinite
number of the
discs Ci.
These exceptional points are'called utii
points.
If we take any infinite binary expansion,

say

.011'00110, we can get a. corresponding

contracting
C01= C011 C0110
a wild point P in common. ThUs there is at
east one
wild'point forde'very real number between 0
and 1, so
that the collection of wild points is

uncountable.

11C into R

Therefore the interior A ofg(*S2) js homeombrphic

to
'
la*
(

SIMPLY CONNECTED SETS

Th

demonstration tha

B is not homeomorphic to
topol gical property. A set X
'is called simply connected if every closet
curve in
X (calle,da loop in X for (short) can be shrunk
continu:
ouslyin X until its ima
is a single point.
1Jg1le

Let C = ((x,y,z) cR 3 Ix 2 +y tz 2
(<1) be the interior
of the unit sphere S2.
Then we could pull Calong as'we
pda,out S2, so tile' functiong
can be extended to C,
giving a homeomorphism of thei closed ball s2
3
the round ball C.'

D uses_:.,he following

-sequence of discs C0

But what about the exterior B of g(S 2 )?

We will
show in the next section why B is not homeomorphic
to
the exterior D of the unit sphere S 2

For, example, the

exterior

D of S

in R

is simply
connectod,'because ever loop L can be pulled off the
sphere and collapsed to a single point P.
A number of
intermediate positions are shown in Figure 8:
,/
Suppose the loop L is parametrize() by
a continibus
function f(s) from JO 1) to'X, and that
the shrinking
lihmotion takes place f r t in the
time interval (0,1)

Then ,for each fixed

354
4

we get an intermediate curve

.355

J.

Figure 9

in

Figure 8

&!rve ft(s), which is alsb closed and

continuous, and
these intermediate curve depend continuously
on t.

The

intermediate curve must agree 'with, f(s), when t =

and stay fixed at.P when ,t = 1..
Thus, a parametrization
of the,sbrinhng motion is a continuous
function
5f(s,t) = ft(s) of two variables, se(0,1), wfiich marks

\distance along each curve, and te10,1), which marks

4ifferent intermediate
curves in the motion.
.

t(s) foil

F(s,0) =

b.

f(0,t) =F(1,0 for all t, and

c.

= P for all S.
It is said

If T denots the stolid- donut, or torus, shown in

3
= R -T is not simply conifected.
The loop L, which w4ans around the. hole, cannot
`be shrunk to a point without;crossing T.

c-

Figure 9,.then its exterior

a.

5.

THE EXTERht OF.THE HORNED SPHERE

t

At first, this might seem difficult, because the

We lay simply connectedness is a topological property,

because it is preserved by homeoMbrphisms.

.We can prove similarly thatthe exterior of B of

the Alexander Horned Sphere ks not homeomorphic to D,
if we can show that it is not simply connected.

alf s,

Such a function F is called a homotopy.

to shrink the leop f id
to the point P.

is a closed loop in D.
Since D is
simply connected, there is a homotopy F which
shrinks
the loop h -1 of in D to a point P.
Then hof will shrink
the loop f in Y'to a point 4(.P),
'Since this works for
any loop4:in Y, Y is simply connected.

the.

It must fratisfyb
a.

Then h-lof

claws never 'touched, sothe 4terior of each

approxi-

mation is simply connected.

However; 'a property which

is true of each of a seclitence of approx.imations is
not
necessarily true of the limit.
In fact,we can define
the horned sphere differently, so that the
exterior
of each approximation is not simply
connected.

Suppose there were a homeomorphism h from the ex2

.

terior D of S to the.exteq.or Y of T.
Than, knowing
4c Iris simply connected, we could
prove Y to be simply
connected d2)iopIows. Let f parametrize a closed. loop
.

ODA)

Imagine you are carving theablid horned sphere

2
g(S UC) out of a piece of wood.
The first approximation
0(l will be a torus with' two bulges, one for theforiginal

sphere, and one' econtain the claws, as shown( in

Figdre 10.

357.

If we continue, we get a 'sequence of closed

sets
2
g(S
U
C),
each
of
whose
K 1=K 2=K 3
exteriors
is non,-simply connected. (See Figure 12.)

Figure 10

The exterior K

is not simply lconnected, since the

1
loop L cannot be shrunk.

The next step will be to carve out

two claws from
the upper bulge (see Figi?re 11), leaving their
tips connected by two smaller bulges.
The result K2 has a nonsimply-connected exterior, Since the loop L
still cannot
be shrunk.
Figure 12
.

Now suppose the loop L coUld be shrunk to a point

in the exterior B of g(S2), using a homotopy F(s,t)..
Since the image of F does not meet g(S2 U C)',
it must
remain a finite distance c away
But now.re find a
solid approximation .Kn within c of g(S 2 U C), and the
image of the homotopy will also miss K.
This contradiets the fact that L cannot be shrunk,to a poibron

the exterior of K.

6.

PROBLEM

Draw a sphere g(S2) such that its Interior A is

not homeomorphic to the interior C of a round sphere.

Figure 14

358

9.

359

10

Solution: Push the horns into the inside

of the sphere.

-.A hole has. been cut away from

the surface to make them visible.

STUDENT FORM

Return to:
EDC/UMAP

55 Chapel St.

Request for Help

Newton, MA 02160

Student: If you have trouble with a specific part of this unit, please fill
out this form and take it to your instructor for assistance. The information
you give will help the author to revise the unit.

YoUr Name

Unit No.

Page
Upper

OR

OR

()Middle

Model Exam
Problem No.

Section

Paragraph

Text
Problem No.

Lower

Description of Difficulty:

(Please be specific)

Instructor:

Please indicate your resolution of the diffitulty in this box.

Corrected errors in materials.

List corrections here:

0 Gave student better explanation, examples or procedure than in unit:

Give brief .outline of.your addition here:

(2)

Assis
u
-5 learning and problem-adrAng
skills (not using examples from this unit.)
el"

361
Instructor's Signature
.Please use reverse if necessary.

Return to:
EDC/UMAF
5'
Chapel St.
Newton, MA 02160

STUDENT FORM 2

Unit Questionnaire
Name

Unit No.

Institution

Date

Course No.

Check the choice for each question that comes closesito your personal opinion.
1.

How useful was the amount of detail in the unit?

Not enough detail to uhderstand(the.unit

. Unit would have been clearer with more detail
Appropriate amount of detail
Unit was occasi nally too detailed, but this was not distracting
Too' much detail, I was often distracted
2.

Howrhelpful were the roblem answers?

Sample solutions were too brief; I could not do the intermediate steps
Sufficient information -was given v) solve' the problems
Sample solutions were too detailed; I didn't need them

3.

Except for fulfilling the prerequisites, how much did you use other sources (for
example, instructor, friends, orother books) in order to understand the unit?
A Lot

4.

Somewhat

Not at all

How long was this unit in compariicin to the amount of time you generally spend on
a lesson (lecture and homework assignment). in a typical math or science course?
Somewhat
Longer

Much
Longer
5.

A Little

About
the Same

Somewhat
Shorter

Much
Shorter

Were any of the following parts of the unit confusing or distracting? (Check
as many as apply.)
Prerequisites
Statement of skills and concepts, (objectivep)
Paragraph headings
Examples
Special Assistance Supplement (if present)
Other, please explain
)

6.

Were dny,of the following parts of the unit particularly' helpful? (Check as many
rs apply.)

Prerequisites
Statement of akills and concepts (objectives)
Examples
Problems
Paragraph headings
Table of Contents
(if present)
Special Assistance Suppl
Other, please explain
,,

Please describe anything in the unit that you did not particularly like.

Please describe anything that you found particularly helpful. (Please use the back of
this sheet if you need more space.)
0
'

3F2

UMAP MODULE "2

I-

MODULES AND
MONOGRAPHS IN
UNDERGRADUATE
MATHEMATICS
AND ITS,
APPLICATIONS

>

>

p
TA

tics of

TA

"CA

(20

>

0:1

0:1

Sirtle
.Kin
Reactant'
Reactions

-4"

fv.

,t

t<>

"*.r

trf

trl

co N co N co N
A

>,

CD

>, CD >,

.-..

'r

irr
o

trr

err

>0

:3"

by Brindell Horelick and

Sinan Koont

F
CD

1.0

il

0>
.....

./. q
1.1

s o
x

00

.-..

i-

e
O

td

.4

e
-s

',.
1.1

q
4

1.1
I

0 s- o

x=x
-c..

.)

t '-

-11

4..4 - 4 - . 4 -.4 - 4 4- 4

.r.)

.2

ti

t
i

-I

i 4

I
I

R
Co

-F.200

tA

,-i
. Co '-i
Birkhauser Boston Inc.
Co

+ ...1 I

-.....-

c
R.

.c.,

300 :000

,300 sec.--7

1500
I,

iii---345 sec.Y

K--- 400 sec .--->

1.

Applications of Calculus to Chemistry

380 Green Street

Cimbridge, MA 02139

363

KINiTacs OF SINGLE REACTANT REACTIONS

by

Brindell Horelick
Department of Mathematics
University of Maryland Baltimore County,
Baltimore, MD 21228
and

Sinan-Koont
1
Department of Economics
University of Massachusetts
Amherst, MA 01003

TABLE OF CONTENTS
1.

S1NGLE4REACTANT IRREVERSIBLE REACTIONS

Oefiniti-en and Some Examples

1.2 Graph of the Results
.
1.3 Questions
1.4 'Chemical Kinetics

REACTION ORDER

2.1 Oefinitions
2.2 Zero-order Reactions
2.3 First-order Reactions
2.4 Second-order Reactions
2.5 Statement of the Problem

1.1

2.

3.

DETERMINING THE REACTION ORDER

3.1.
3.2

3.3
3.4
4.

HALF-LIFE
4.1
4.2

4.3
4.4
4.5
5.

Solving for,a(t,
The Difficulty
Solving the Difficulty
An Example

Definition
Formulas for Half-Life
Zero-order Reactions

5
6
6
6
s

7
7

8
9
10
11
11
11

12

First -order Reactions

13

Second-order Reactions

13

MODEL EXAM

15

/ntermodutar Description Sheet:

Ti)ie:

UMAP Unit 832

KINETICS OF SINGLE REACTANT REACTIONS

Author:

Brindell Horelick

and

Sinan Koont

Departmefft, of Mathematics

Departffient of Economics

Univeesity'vf Marylind
Baltimor County
Baltimore,Jlaryland 21228

University of Massachusetts'
Amherst,'Massachusetts 01003

Review Stage/Date:
.Classification:

It 7/2/79
0

APPL OALC/CHEM-KINETICS
.

4
,

...

References:
'
*
,.
Bares, J., Cerny, C., Fried, V., and..J. Pick (1962). Collection of
emblems in Physical Chemistry. Addisop Wesley, Reaking, MA.
Capellos, C., 8nd B. Bielski (1972). ,Kinet4c Systems. Wiley..
interscience,'NY. .
i
Frost, A.A., and R.G. Pear%on (1961). Kinetilv and Mechanism. (2nd
ed.) John Wiley and Sons, Inc.NY.
A
t
Laidler, K.J. (1965). Chemical Kinetics.) Mcicaw7Aill, Inc., NY.
Stevens, B. (1970).
Chemical Kinetics.
Chat. an and Hall,:London.
Weston, R.E. Jr.,.and H.A.
C
(1972).
al Kinetics:
Prentice-Hall,;Englewood Cliffs, NY.
i

Prerequisite Skills:
1.
Be familiar with the Cuftsian coordinate sistem.
2.
Understand that aqr) describes the rate of Change
3.
Be able to integrate
...
ft a'(t)

. .

alt).'

"

-...

Jo. 1-6-cidt

for 11.... 0,i...12,3.

4.

'Be able to solve an exponential equation.

.

This unit is intended for Calculus students with an aclive&

interest in and some background knowledge of chemistry. This b4k-

ground may berepresented by concurrent regitration in a colle0

0
4
level chemistry course. '
t

'
.

t,

'

..

Output Skills:
1. Be able to describe single reactant Irceversible'teactions, including definitions of rate constant, reaction order, band half - life.
2: Be able to find explicit.formulas for a(t) and for. the
half-life
for a reaction of order n.
3.
Be able to determine the reaction order and rate constant of a
reaction, given data on a(t), provided the reaction is of order
'AI

0,

1, or 2.
1

This material was prepared with the partial support of National

Science Foundation Gi-Nnt No. 5E076-19615 A02. Recommendations expresled are Chose of the author and do not necessarily reflect the
views of the NSF or the copyright hoider.

4) 1979 EDC/Project UMAP

All rights reserved.

3 6 541

'

SINGLE :REACTANT IRREVERSIBLE REACTIONS

1.

1.1

Definition and Some Examples

Suppose we have a'chemicalr4action of a V

arly

simple sort,.one which involves only one substhnce (1

call it A) as a reattant,.:and 'wtich is irreversible,

therefore eing to compaetion.' ri may be-represented by

o.;

writing:
A- -I, Si

where B1, B2, e.. Bn

Bn

i 13

re,the pxoduct's. Suppose, at ,time

t = 0 we have a certain'concenpAtii4n a,

of ,A (measured,

for.example, in moles perliter).vIt is possible to

observe and record ttie concentration a(tp,of A at variotiss
later times t.
t

TABLE I
.

Experimental Data from Three

Reactant Ir;eversible Reactions.

51

(seconds)

-206

454

751

1132

1575

2215'

9.73

7.79

6.08

4.17

.
.

(a)
a(t)
(mm Hg)

..

14-)0

l.''''. 0

(minutes)

13.3f

11.49

50

38

10

30

(b)

40
.

a(t)
(mm Hg)

55'

21

1.5

1
'

(seconds)

120

180

.6705

5825

240

330

530'

600

...

(c)
.--.

a(t)1a,

1.

512

41'95

.310

.(

;a.

.2965

Table I gives three sets of such observations. Part

is for an experiment conducted at 280 o C'involving the
decomposition of trichloromethyl chloroformate into phos(a)

gene:

CI-C-0-C 1 3 -4- 2C1-C-Cl.

Part C6)

is for the decomposition at 500C of ethylamine

into ethylene and ammonia:
'
C H NH
/
2
Part (c)

C H
2

NH 3'

is for alkaline hydrolysis of ethylNititrobenzoate

at an Hitial concentration of 0.05 moles perliter.

`

The reactants in parts (a) and (b) are gaseous. At

constant temperature and volume, a(t)4ispioportional to
its partial pressure, and it is this figiie, in millimeters
of mercury (mm Hg),that appeIrs'in Table I.
In part (c),
a(t) fsgiven as'a fract4On of aQ.

'

' In the conversion of trichlorome thyl chloroformate to

,-(
phosgene, bstil the reactant and the w0dug4,are gaseous,

/-

and the total pressure actfiallv ilitileases as the reaction

proceeds-tsecause each trichloromethyl chlotroforM'ate molecule

gives rkse to two phosgene molecules.' The partial pressure

ot the trichloromethyl chloroformate isCleducedfrom the

total pressUre
taking the ieactionequationand the
originl pressure into Account.

In many reactions, however,

theamount of the reactant y determined by techniques

based on its absorption of 'lifht.
1.2

Graphiof the Result's

We have plotted these results in Fdgaures 1, 2, ind

In'that all of the curves decrease as. increases,
'these curves look very similar.
But there is at least
one significant difference (,aside from the differences oa
scales).
In each figure we haye selected various concentrations of A and determined graphically approximately how
.3.

lone it takes for a(t) to decrease'from the selected,con-

centFation to half of it. ?For example, Figure 1 shows us

2'
114,

s.

3.67

10

- --

- - - 4- - - - -

1....

...

.1

/-

1
1

I
I

500
'
1 170 sec

K--------

,4-.1

1500

1000 .

)1

0-,,

t (5")

12000

'-

1220 sec

--4.

1200 sec

'

Figure. 1.

4-

Decomposition of Tr ichloromethylChloroformate (from able I (a) ) .

/.
-

a (t

mm Hg)

50

.J

-40

'

1--

.>

20
A

'1

lb

'

:'

I1-I

'

t.

I.

I
, 20

4- it. II min.)) 101

.6
t

:t7 . 5 min

Figure

2.'

40

Dec

---

"
-

1 (b) )t.
-

..

j
`36

1'
06,

.
.

posit io'n of Ethylamine (from Table

V

-1

,6

'51

min

-,,.

.....d

Z..2 gi nil

'4

:*

,
..

..

,.
"

`r

. .

....
51

;
fa

.
:!*

At

... ...X

L...,-

--t
,

0.2

3I

0 '

go sy1014001560. :600 t(sec)

10

k---e)-345 sec----0

;400 sec

-Figure 3.

Alkaline Hydrolysis of Ethyl gilrobenioatd (from Table 1(c)).

hat it takes apnr6ximatAy 1170 seconds for a(t) to

decrease from 44 to 7 mmilg,.or1220.teconds'for it"to

decrease

to 5.mm Ng.

fiom

In elcIr of.thilfirst two

1)
figures the measured
time intervals are approximately

equal, but in Figufe 3 they are not,

',
.

1.3

Questions

rl
.

1.-

.'

Can we explain this difference in teimsof.the

''

...

reactions?
4._
4

*!

Or, turning the oupstion aroutd, can we draw

any conclusions bas'"d' on these observations, bao t te.
h
1
nature of ihe reactions?
.'''

1.4

themical Kinetics

'

Q.uestions such as these are part of a branch of

4

chemistry known as cA'emical ktnetics.

"'

rates.anc'

is concerned wIt
withh the

k/

Chemical kinetics,

Tiechapisms ofchemical
4

69

At

reactions..

The name reflects the fact that "kinetics"

is concerned with the changing aspects of systems, as
distinguished from "statics" which concerns systems at
equilibrium.
We should also point out here that the
rate at which a cheMical process takes place and the
mechanism of the process (i.e., what exactly happens

V$

during thq transformation of A into B1' + B2

Bn)
are two diffeient things. The study 'of ireaction meth antsms lies at a higher theoretical level than the study
of reaction rates.
In general, experimentally determined

reaction rates can he used to rule out a proposed mechanism

if they are' inconsistent
gut experimental data,
that are consistent with a proposed mechanism can only.
serve as suPpn-ting-evideoce_for.it; they cannot be used
di'rectlytrprove its correctness.
.REACTION ORDER

2.

2. Definitions
o

To make the question in Sectibn 1.3 more specific,

we shall summarize some background information about the,
reaction rates in reactions of this type.
If substance
A (in gas or'liquid form) is uniformly distributed; and
if the temperaturetand volume are kept constant, then,it
usually turns out that the rate a'(t') at which A decomposes is peoportional eb a, non-negative Lncdger.power

(0,1,2,) of the concentration a(t).

DV

In other words

a'(t) = -k(a(t))n

where k

i-s

a post(ive constant and n is a non-negative

.

integer.

We'call k the rate constant and 'n, the order ei

the reaction. Equation (1) with n established is called
the rate3CC flopt'he reaction:

We shall cotsieer reaction orders 0, 1 and 2 in

detail.. Higher, reaction-orders for reactions of tli---type

we are discussing are considerably more rare,

.

370
'or

2.2

to

Zero-order Reactions
)
J

Setting n = 0 in Equation
a'(t) =

(2)-

gives

-kt;'

where we have introduced the ,subscript to 'denote the

reaction order.
tration of A:

The rate is independent vf,ithe'concen-

It is determined by other f.46toN such as

temperature, the intensity of light in light-induced

reactions, the surface area available in surface-catalyzed
reactions, or the amount of catalyst in homogeneous
catalysis.
(A catalyst is a chemical substance that cop-.trols the
e of a reaction without undergoing any net
change in it if over the course of the reaction.)
2.3

First-order Reactions
In this cote, we have
011.

a"(t) = -kia(t).

''(3)

Most simple decomposition reactions pivolving,a single_

reactant are of first-order. This is not surprising if
we'imagine the reaction process to consist of molecules
01 A decomposing randomly.
If; for example, each moleo4le
has 1 chance in 10 of decomposipg in 'the next second, 'then
1

about nth of those present will in fact decompose in that

second.

In other words, the 'change in a(t) in that second

1

is about Tba'(t).) We describe this by writing

a
2.4

1
a '(0=---a(t)

A
'

Second-order Reactions
4

The rate law for second - order reactions is:

a'(t), = -k

II

a2(t).

In Teneral, elementary. reactions which require' the collision

-epf two molecules are good candidates for this category.

2.5. -- Statement of the Problem

Equation (1), has been confirmed for many reactions

'Sy numerous experiments, and also explained theoretically.

6

We shall not get into the theoretical explanation except

to say (as has already been indicated in Sections 2.2, 2.3
and 2.4) that different reaction orders are the result of
different underlying reaction mechanisms.
So if we have
a reaction and want to know mo re about its mechanism a
4

very useful first steis to determine its reaction order

experimentally.

Can we use data such as that given in Table I to

determine whether'a reaction has one of the orders we have
discussed, and, if so, which one?
CP

DETERMINING THA,REACTION ORDER4*

3.

3.1

Solving for.a(t)

To begin with, we can use Equations (2), (3), and

(4) to obtain explicit formulas for a(t) in the three
casds.
(a)
Zero-order reactions:
If a'(t) = -1; then '
o
a(t) = -kot + C where C is a constant of integration:

Using the'fact that a(0) = ao we see that C = ao and

a(t) = ao

(S)

kot.

First-order reactions:
Starting with Equation
(3), divide both sides by a(t) (whichAis never zero):
(b)

r.

a' (t)

at
ft

a(t

dt

Aft

= -k,t

k I dt:
t

In a(t)

.0

.1

'

In a(t),= -,kLt

In ao

_p-k t

40

(6.1

A1

d7-e

r.

*k
;*

es

:: '44"7.1

.
(c)
Second-order reactions: In Equation (4) we
diVide each sidb by a2 (t), 'and conclutle that

&

a'

t)

k
II

a2 (t,)

It

n ft)
- dt

-It k

0 a2 (t)

1)
-7

II

d, t'

0
t

it

10

-kIIt

10

aokiit +
a

t)

III

t +

ao

ac,

a
a(t)

(7)

at) k

t + ,1

Exercise 1

Find () expl,icitly fora thiA-order: reaction.

Exercise 2
,.,

Assume to reactions

are of first and second order respectively:

itt

a'-(t)

-k a (t)

b' (t)

- kI I b2 (t)

As,same they begin with the same amount of rektaat (ao

40

b0), and

their initial rates are the same [e(0) = b' (.0)]. Prove that
a(<t)"< b(t)

for all t > 0:

(flint: Note-that a(t) = 1 hen t = 0 and show that it is strictly

.4
decreasing for t > 0.)
e

3.2

The Difficulty

111,

The rate constant 1,(k0

of course' not

mown, so...we cannot get away with anything so naive as

..;

'plugging our data into -Equations (5) , (6)

to see which one checks out.

*

373

and -(.7).

It is true that the graphs

a

of these equations have three distinitive "shapes:',

whatever the constants are (for example, Equation (S)

Sp wQ,.cpuld confider graphing

is a straight line).

our experimental data and trying to deterMine which

In this unit however, we

"shape" curve 'fits it'best.

present a method of determining,the reaction order that

does not depend on graphing, and whic also gives us the
-

rate constant at 1110 extra cost.

3.3

Solving the.Difficulty
The method_starts with solving Equations (S), (6J,

and (7) for k0, kL, and lc

o' 0.2
II
a(t)

Ta0

(8)

(9)
(10)

- k0

kII

t >

kI =

-t- .1n' -a(fy

t >

4 (AT

>

0.

Now'if, fbr example, the reaction oriler is zero,

then all the data points should satisfy Equation CS)

for some constant k0.

Thus whenever we Substitute any

data point (t, a(t)) to, the right side'of Equation "p8)

we should get more or leis the same value (namely k0).

Naturally there, will be small variations due to experi-

'Similar comments apply to Equation (9) if

mental.error.

the reaction order is one, and Equation (10)

.

the

reaction order i9etwo.

So all we need to do is compute three rows of

figures -- the right sides of Equations (%), .(9), and

(10) -- for!bur'data points, and see if-any row remains

If so, that row gives us the

mdre or less constant.

zeaction,,qrder, and its constant value is the-rate

constant 0.0, kr or. kid.,

s
,9

374

3.4

An-Example

As an example, let's go back to part (a) of Table

In Table II, we hake repeated the data and also tabulated
the right sides .of Equations (8), (9), and (10). The
figures in. the row corresponding tb Equation (9) are

nearly constant

x 10 "see -') while those in the

(r..- 5.8

other rows are not.

So this reaction is apparently

first order reaction with ki.z. 5.8 x 10 "set -'.

* TABLE 11

CalcUlation of Rate Constant andReactIon

Order for Data of Table 1(a).

-t

sec

51

2,06

14.58

13.32

8,82

8.30

454

751

1132

1575

49

9.73

7.79

6.0664.17

7.80

7.06

6.40

5.68

2215

mmHg

a(t)

ao - a(t)
e

15. k

10)

mm Hq
sec

4.90

In -z-1-5:-a x 10'

sec

. 47(17

.)' 1135

mm

Hig

sec

5.96

5.86

5.9;

5.79

5.81

5.75

5.79

4.03'

4 15

4.52

4.83

5.46

4 22

.7.82

A'
9

Exercise 3
Determine the reaction order and rate constant from the data
given in part (b) of Table I.
Exercise 4
' Determine the reaction order

rate constant from the data

given in part (c) of Table I.

'

%,.

.10

11
S

4.

4.1

HALF-LIFE

Definition

The half-life t of a certain amount of a reactant

is the length of time -required for exactly half of it to
be used up.

In other words, if the amount of reactant

at time t = 0, and if,a(t) is the amount at alater
time t, then t is the solution of the equation
is ao

1/2

A..

a(t) =,

a0._

In Section1:2 we determined graphically the halflives 8f various athountg of three reactants, and "discovered
that for two of the reactants t1/2 did not seem toldtpend
uppn the initial amount,,, but for the third tedct,artAir
did.
Let us. see if this phenomenon can shed a little

more light.on the concept of reaction order,

t

4.2 ,Formulas for Fralf-Life

To start with, let us compute

for each of the
three reaction orders we are considering. All we need
- to do is -set: -aSt)--=-7T-ilrTeach of Equations (5), (6),

'and (7) and solve for t:

(11)

t, = TFT,a,
e

(12)

(Zero-order)

'

2.

TFii-s7oraer)

t1/2

(13)

(Second-order) ..
4

II

Exercise 5
,

Find t; as.a.function of.a-0 for a third-order reaction:

Exercise V
We define tv
used up.

That

3.
as the time required for.w
3,

Find t

of

a reactant tq'be

as a function of ao for
3/4

reactions of iero; first,, and second-order.

r

11

'\.
376.

. Exercise
t3

Find thp-ratio

/4

Exercise 8

for reactions of zero, first, and

secorild:order4
t

Table III gives ti and t3/4

for three initial amounts of the

reactant in the reaction

CH3CHO

CR4

acetaldehyde

CO

methane

carbon monoxide

Determine if possible whether the reaction has one of the three

or1ers discussed il this unit and, if so, which one.

TAB* III
Half-life and 3/4-Life Data for the
Reaction CH3CHO + CH4 + CO
(Exercise 8)
i&

ao(mm Hg):
(seconds)

t"

34

.1

421),.

225

184

38,5

572

665

1135

1710.

1920

(seconds)
.
.

Exercise.9
,.

Suppose, for every x between%b and41, we write tx f'or the,

time required for fraction x of a reactant to be used up.
Exercise 6

3/,

an example of

with x = 34.)

(In

Show that in

a first-order reaction "Ms independent of the initial amount no

matter what x is.

4.3

Zero-order Reactions

For a zero-order reaction, half-life is proportional

to initial{ amount.

ts

The greater the amount, the longer

12

377
01

the half-life.
To help your'self understand and remember
this, think of a very large.number of marbles, from which

we remove, say, 10 each second (k

= 10).
The more there
o
are originally, the longer, it will take to remove half of
them.

4.4

First-order Reactions
For a first-order reaction

half-life is independent

of initial amount!!

To heip"Understand and remember this,

think again of a very Large,number.of marbles.
This time
remove ona half of the pile in the first second, then one
haXf of,the remaining pile in the next second, etc.
1
(k1 = 7). No matter how many we start with, it will take
one second to remove half o,f them. ,Also, at any later
stage it will take one second to remove half of what
remains.

'4'.5 'Second-order Rea=ctions

IPor a second-order reaction, half-life is proportional
to the reciprocalN of t e initial amount. Another way of

saying this is *chat a

t,
is-a constant. The more'of A
(11
o
there is, the leis time it takes for'one half of it to
decompose! Although this May seem paradoxical we invite
you to consider the fact that second-order reactions

depend upon collisions of pairs of molecules.

Equation
(13) says that the more molecules there are, the more
likely they will collide, and the faster the reaction VIC
proceed.
Exercise 10

The following data were obtained by F. Daniels and E.H. Johnston

(J. Am. Chem. Soc., 43, 53 (1921)) for the deeompositioen of nitrogen
pentoxIde (N
2

in .solutiOn in carbOn tetr,achloride (CC1

2N 05 4 2A20, +

02.

) at 45 C:

\h.
13.

.3'78

.40

od,
t

(seconds)
concentration
of N20,

181(_

319

, 526

867

1198

1.36

1.11

1877

2-345,....3144.

2.33

2.08

1.91

1.67

(mole/ir

.72

.55

.34

Determine the reaction Oder and the rate constant, as well es the
half -life t

How long would it take for 87.5% of the reactant to

be used up?

The Project wouldelike Wthank Scott C. Mohr of

Boston University, Andrew Jprgensen of Indiana State
University at Evansville, Bernice Kastner of Montgomery
College, Rockville, Mryland, Barbara Juister of Elvin
Community College, Elgin, Iltinois,for their reviews,
and all others whusassisted intth production of this
unit.
This unit was field-tested and/or student reviewed
in preliminary form at Indiana State University,at
Evansville, Southern Oregon State College, Ashland,
-Colorado School of Mines, Golden, and Northern Illinois
University, DeKalb, and has been revised on the basis
of data received from these sites.
,

14
loo

14

5.

MODEL EXAM

.
For some reactions the reaction order is found

1.

Find a(t) explicitly (in terms

to be fractional.
of a

for a reaction with reaction order,n =

1/2.

o
.

';'

x)a

22 Define tx *as that time for which a(tx):= (1

.P'

independent of a

Is this tx

Find,tx for a second order reaction.

Determine the reaction order and rate constant from

3.

the forlowifig data for a hypothetial-reaction.

t (seconds)
a(t)

(moles/1)

2#

.10.0

3.98

2.$1

1.82

1.44

-10

1.19

41.

si

.
T5-

<.

33o

ANSWERS TO EXERCISES

1.

a (t)

= a
13

2.

[2 a2 kur t

The information given to us

1.

a'(t) = -kia(t)

2.

b'(t) = - kIlb2(t)

3.

a(0) = b(0)

4.

a'(0) = b'(0)

at
(

To see that 1-Tm-

is:

is a decreasing function of t, we show

i

that the derivative of the quotient is negative.

b(t)a'(1) - a(t) b'(t)

d (a(t) 1

dt CT51)

b2(t)

b(t)(-k a(t)) -7a(t)(-1(1ib2(0)

b2(t)

^:

ki

(14)

Now, a'(0) = b'(0) means that1(1.3(0) = 1(1'0(0)

.

and a(0),

b(0) means further,that

kib(0) = lcub2(0)
,i(I = 1(1'410).

When we substitute this value of k

d 13(0
CIF

kIlb(0) 1

TF)

tkII
(

klIa(t)(1

b(0)1.

;TOT

Since b(t) < b(0) for t > 0,

U(0)

b(0)1v

TM) c

in Equation (14) we obtain

0.

381

41.

Reaction order =

3.

min-1.

k = 9.4 X' 10

Reaction order = 2

4.

aok

5.

103sec-1 or, since ao'= 0.05

k = 8.2 );10-2

4.1 x

% mole-1sec -1

t,

2a02k3
3a0

6.

te, =

Zero order:

,First- order:

t3

Second-order:

7.

In 4.

1/4

aol,c/I

Zero-order:
First-order:

Second-order:

__Second - order.

First order, k z 6.1x

10.

104

sec., ti -.: 1120 slc, 3t

z 3360; sec'

ANSWERS TO- MODEL' EXAM

4
k2 t2

a(t) =a0 -

kt +
0

.x

1-

a k
0

ic/

- x

11

1-4
3.1
f

k =

7.5 x 101 rmole 2

sec

-1

Order = 2.
1

t
*,

382