You are on page 1of 137

Mathematical

Analysis
a straightforward
approach

K.G. BIN

MORE

an unf ussy account of mathematical


first year courses at universities and
polytechnics that places little reliance on the
reader's previous mathematical background.

This

is

analysis for

The concepts involved


and

are profound

and subtle

PRESTON POLYTECHNIC

their application in particular instances

comes

requires a mastery of technique which

only with time and experience.

It is

introduce the concepts at a relatively high level of


abstraction;

it then becomes

LIBRARY & LEARNING RESOURCES SERVICE

tempting to
This

book

the elegance of the underlying theory. But such

treatments

make great demands on a

student's

At the other extreme is the


danger of overwhelming students with technical

technical

rrust

be returned on or before the date

last

stamped

possible to expose

nc

i^~*^ *
9

facility.

detail.

Professor Binmore's approach

is

to confine

attention to 'bread and butter' analysis


line

31.

on the

SEP.

198!

>A/<5X-

real

and, within this framework, to discuss only

those technicalities which are immediately


relevant to the basic concepts, the object being to

promote confidence and understanding rather


than

awe or mindless computational

p\n

efficiency.

Emphasis is placed on the importance of tackling


problems and the exercises are intended as an
integral part of the text. Their worked solutions,

-1

which may be consulted when difficulty arises in


solving a problem, occupy approximately a third
of the book. This abundance of exercises and
solutions

makes the text

who want to learn the

HK

1996

highly suitable for those

subject on their

own

as

well as for those attending formal classes.

515 BIN
PLI

A/C 095640

:J43

Chorley
Lancaste
Poulton
Preston

C.U.P.
NET IN

UX

9-75

30107

000

5:

5 408

3T

MATHEMATICAL
ANALYSIS
A STRAIGHTFORWARD APPROACH

K.G.BINMORE

dr^

Professor of Mathematics L.S.E.

JpL,
.

MCM tcvn

^y
II

CAMBRIDGE UNIVERSITY PRESS


CAMBRIDGE
LONDON NEW YORK
-

MELBOURNE

CONTENTS

Published by the Syndics of the Cambridge University Press

Trumpington Street, Cambridge CB2 RP


Bentley House, 200 Euston Road, London NW1 2DB
32 East 57th Street, New York, NY 10022, USA
296 Beaconsfield Parade, Middle Park, Melbourne 3206, Australia

The

Pitt Building,

Cambridge University Press 1977

First published

1977
IX

Preface
Printed in Great Britain at the University Press, Cambridge

Library of Congress Cataloguing in Publication Data

1.1

I.

Arithmetic

Inequalities

set of real

numbers

Roots

10

Quadratic equations

13-

Irrational

14

Modulus

10

Continuum Property

12

2.1

Achilles and the tortoise

12

2.2

13

2.6

The continuum property


Supremum and infimum

2.7

Maximum and minimum

2.9

Intervals

2.11

Manipulations with sup and inf

18

Natural numbers

20

3.1

Introduction

3.2

Archimedean property

20
20

3.7

Principle of induction

22

Convergent sequences

26

4.1

The

4.2

Sequences

Title.

76-28006
QA300.B536
515
ISBN 521 21480 7 hard covers
ISBN 521 29167 4 paperback

Mathemalical analysis.

Mathematical analysis.

Set notation

The

Binmore,KG 1940-

1.

Includes index.

Real numbers

numbers

15
15
:>

bulldozers and the bee

16

26

4.4

Definition of convergence

27
27

4.7

Criteria for convergence

30

4.15

Monotone sequences

4.21

Some

33
37

4.26

simple properties of convergent sequences


Divergent sequences

38

Subsequences

41

5.1

Subsequences

41

5.8

Bolzano-Weierstrass theorem

46

VI

Contents

5.12

Lim sup and

5.16

Cauchy sequences

lim inf

Contents

47
49

vii

12

Monotone functions

108

12.1

Definitions

12.3

Limits of

Series

53

12.6

monotone functions
Differentiable monotone functions

6.1

Definitions

53

12.9

Inverse functions

6.4

Series of positive terms

54

12.11

Roots

6.7

Elementary properties of series


Series and Cauchy sequences
Absolute and conditional convergence

56

12.13

Convex functions

108
108
110
110
112
114

13

Integration

119

13.1

Area

119

13.2

The

64

13.3

Some

properties of the integral

121

13.9

Differentiation and integration

13.16

Riemann integral
More properties of the

6.12

6.20
6.23

Manipulations with series

Functions

57

60
61

7.11

Inverse functions

64
66
68
68

7.13

Bounded functions

70

Limits of functions

74

7.1

Notation

7.6

Polynomial and rational functions

7.9

Combining functions

13.31

Euler-Maclaurin summation formula

124
127
129
132
134

14

Exponential and logarithm

137
137
140

14.6

Logarithm
Exponential
Powers

15

Power series

143

15.1

Interval of convergence

143

15.4

Taylor

15.7

Continuity and differentiation

145
147

16

Trigonometric functions

150

16.1

Introduction

150

13.19
13.27

14.1
8.1

Limits from the left

74
74

120

integral

14.4

Improper

integral

integrals

8.2

Limits from the right

8.3

/(x)^/as:c-*

8.6

Continuity at a point

8.8

Connexion with convergent sequences

8.11

Properties of limits

78
79

8.16

Limits of composite functions

81

8.18

Divergence

82

Continuity

84

9.1

84

16.2

Sine and cosine

9.7

Continuity on an interval
Continuity property

86

16.4

Periodicity

151
153

10

Differentiation

91

17

The gamma function

156

10.1

Derivatives

91

17.1

Stirling's

10.2

Higher derivatives

92

17.3

The gamma function

10.4

More notation

93

17.5

Properties of the

10.5

Properties of differentiable functions

95

10.12

Composite functions

98

18

Appendix

75
77

141

series

formula

gamma

function

156
157
158
161

This contains the proofs of 'propositions' left unproved in the main


11

Mean value theorems


maxima and minima

100

11.1

Local

11.3

Stationary points

101

11.5

Mean

11.9

Taylor's theorem

102
105

value theorem

100

body of the

text.

Solutions to exercises

172

Suggested further reading

251

Notation

253

Index

255

PREFACE

and unfussy introduction to mathematical


analysis. Little formal reliance is made on the reader's previous mathematical
background, but those with no training at all in the elementary techniques of

This

book is intended

calculus

as an easy

would do better

to turn to

some other book.

been made to lay bare the bones of the theory by eliminating as


much unnecessary detail as is feasible. To achieve this end and to ensure that all
results can be readily illustrated with concrete examples, the book deals only

An

effort has

with 'bread and butter' analysis on the real line, the temptation to discuss generalisations in more abstract spaces having been reluctantly suppressed. However,
the need to prepare the

way

for these generalisations has been kept well in

mind.

It is vital to adopt a systematic approach when studying mathematical analysis. In particular, one should always be aware at any stage of what may be

assumed and what has to be proved. Otherwise confusion is inevitable. For this
reason, the early chapters go rather slowly and contain a considerable amount of
material with which

many

readers

may

already be familiar.

To neglect

these

chapters would, however, be unwise.

The

exercises should be regarded as an integral part of the book. There

is

from attempting the exercises than can be obtained


from a passive reading of the text. This is particularly the case when, as may frequently happen, the attempt to solve a problem is unsuccessful and it is necessgreat deal

more

to be learned

ary to turn to the solutions provided at the end of the book.

To

help those with insufficient time at their disposal to attempt

all

the exer-

have been marked with the symbol t- (The same


notation has been used to mark one or two passages in the text which can be
omitted without great loss at a first reading.) The symbol * has been used to
cises, the less vital exercises

mark

exercises which are

more demanding than most but which

are well

worth

attempting.

The

final

of exercises

few chapters contain very


set.

little

theory compared with the number

These exercises are intended to

niques introduced earlier in the

book and

illustrate

the power of the tech-

to provide the opportunity of

some

re-

vision of these ideas.

This book arises from a course of lectures in analysis which is given at the
London School of Economics. The students who attend this course are mostly

not

IX

specialist

mathematicians and there

is little

uniformity in their previous

Preface

mathematical training. They are, however, quite well-motivated. The course is a


'one unit' course of approximately forty lectures supplemented by twenty informal problem

classes.

have found

it

possible to cover the material of this

Time is then left for some discussion of point set


topology in simple spaces. The content of the book provides an ample source of
examples for this purpose while the more general theorems serve as reinforcement for the theorems of the text.
Other teachers may prefer to go through the material of the book at a more
leisurely pace or else to move on to a different topic. An obvious candidate for
further discussion is the algebraic foundation of the real number system and the
proof of the Continuum Property. Other alternatives are partial differentiation,
the complex number system or even Lebesgue measure on the line.
would like to express my gratitude to Elizabeth Boardman and Richard
Holmes for reading the text for me so carefully. My thanks are also due to
book

in

some

REAL NUMBERS

1.1

Set notation

thirty lectures.

A set is a collection of objects which


element of the

so

would

like to

mention M.C. Austin and H. Kestelman from

whom

learned

much of what I know.

we

say that

x belongs

to

are called its elements. If x

S and

is

an

write

xes.

'Buffy' Fennelly for her patience and accuracy in preparing the typescript. Finally, I

set S,

Ify does not belong to 5, we write y S.


The simplest way of specifying a set is by

elements.

listing its

We

use the

notation

July 1976

K.G.B.

A = &1, >,,*}
whose elements

to denote the set

are the real

numbers

\, 1

\Jl, e

and

it.

Similarly

B =

{Romeo,

Juliet}

denotes the set whose elements are


This notation

is,

Romeo and

Juliet.

of course, no use in specifying a set which has an infinite

number of elements. Such

may be specified by naming

sets

the property which

from objects which are not

distinguishes elements of the set

in the set.

For

example, the notation

C =

{x

> 0}

(which should be read 'the set of


positive real

D =

real

If

is

> 0')

denotes the set of all

who love Romeo.

convenient to have a notation for the empty set 0. This


if

x denotes

a variable

is

the set which

which ranges over the

numbers, then

{x
This

that

has no elements. For example,

of all

x such

{y y loves Romeo}

denotes the set of all people


It is

all

numbers. Similarly

x2 +

0}

because there are no

0.

real

5 and T are two sets, we

numbers x such

say that

5 is

that

x2 =

a subset of

T and

write

set

Real numbers

Real numbers

SCT

number system which we propose

properties of the real

to assume

and to some

of their immediate consequences.


every element of S is also an element of T.
As an example, consider the sets P = {1 , 2, 3 4} and Q = {2, 4}. Then
Note that this is not the same thing as writing Q e P, which means that Q

if

ofP are

element of P. The elements

simply 1,2,3 and 4. But

is

QCP.
is

an

these.

The

sets

A C Cand

B,

C and D

(presumably)

given above also provide

Arithmetic

1.3

not one of

some examples. We have

BCD.

The

first

assumption

is

that the real

numbers

satisfy

all

the usual laws

of addition, subtraction, multiplication and division.


The rules of arithmetic, of course, include the proviso that division by zero

is

not allowed. Thus, for example, the expression


2

The

1 .2

set

of

real

numbers

be adequate for these notes to think of the

It will

points along a straight line


line

may

which extends

indefinitely in

then be regarded as an ideal ruler with which

real

numbers

as being

both directions. The

we may measure

makes no

sense at

We

have

In particular,

all.

it is

not true that

the

lengths of line segments in Euclidean geometry.


shall

understood
n/2

-2

The

of

set

all

real

numbers

will

The

table

below

The next assumptions concern


assume that, given any two

1,2,3,4,5,.
.

.-2,-1,0, 1,2,3,

real

numbers

are rational.

v2, e and ir.


While we do not go back
be rigorous

to

first

in so far as it goes.

It is

0,1,2.-1.*, If. -I, -I

Some examples of irrational numbers

as

between

real

numbers and

numbers a and

b, there are three mutually

>b

(a is greater than b)

(ii)

=b

(a equals b)

(iii)

<b

(a

<b

is

than b).

less

means the same thing

as

> a. We have, for example, the

-K0; -3<-2.

>0; 3>2; 2<3;


often

is

some confusion about

the statements

are

principles in these notes, the treatment will

(iv)

a~>b

(a

is

greater than or equal to b)

(v)

<b

(a

is

less

up

this

than or equal to b).

therefore important to be clear, at every

about what our assumptions are. We shall then know what has to be
proved and what may be taken for granted. Our most vital assumptions are concerned with the properties of the real number system. The rest of this chapter
and the following two chapters are consequently devoted to a description of the
stage,

clearly be

be treated

following inequalities.

There
all

real

Not

inequalities

(i)

Observe that a

number. Nor can

it

..

Rational numbers
(or fractions)

must

exclusive possibilities:

Elements

Natural numbers

Integers

it

manipulation.

We

(or whole numbers)

real

but

distin-

U.

Notation

does not represent a

Inequalities

1 .4
.

their

Subset

of use for the symbol

such except in very special circumstances.

be denoted by

guishes three important subsets of

a great deal

that

To

clear

We

confusion,

>0\ 3>2;

we note

>

1;

that the following are

2<3; -1 <0;

all

true statements.

-3 3.

assume four basic rules for the manipulation of

inequalities.

From

these

Real numbers

Real numbers

the other rules


(I) If a

>b

may be deduced.
and b >c, then a >c.

Since

> 0,

x+y

(x

+y)~ 1

>

(example

+ y)'

through inequality (1) by (x

We

.6).

can therefore multiply

to obtain

y-x>0
x<y.

i.e.

we could prove

(Alternatively,
is

If

<b +

reverses the inequality).

Example
/Voo/

It

If a

> 0, prove

_1

that a

-1

> 0.

We argue by contradiction. Suppose

cannot be true that a"

= O.a =

(since then

This
that a
1).

>

but that a

-1

< 0.

But
a

_I

<

is

that, for

that

any

>y

which

e>0, a <b +

e in the particular case

e.

<y 2 but that

is

a contradiction.)

Thena

<fc.

when

= b a. Thus

a<a.
a contradiction.

is

this inequality

through by a (since a

> 0).

which

<

was

Therefore

is

the

is

the Greek letter epsilon.

from the 'belongs

to'

symbol

G and also

It

should be

from the symbol

Greek letter*/'.)

Exercise

1.8

a contradiction. Therefore the assumption a

false.

Hence

0.
-1

Hence our assumption a > b must be

(Note: The symbol e in this example

Hence

<0.

= a"'.a<0.a =

Proof Assume
Hence a < b +

e.

carefully distinguished

we can multiply

rule III

(i))

a<b.

By

follows (as in

a<b + (a-b)
and so

7.5

Assume that* 2

as follows.

that a > b. Then a b > 0. But, for any e > 0,

> b and c < 0, then ac < be (i.e. multiplication by a negative factor

it

Example Suppose

1.7

through by a positive factor).


(IV)

>y

x > y. From x

> b and c any real number, then


a+c>b+c.
(III) If a > b and c > 0, then ac > 6c (i.e. inequalities can be multiplied
(II) If a

(ii)

false.

Hence

(1)

>0.

If*

is

any

real

number, prove that* 2

>Q.

If

0<a<

andb

> 1,

prove that

(i)0<a 2
Example

1.6

If

<y if and only if x <y


2
First, that x < y implies x < y

and y are positive, thenar

Proof We have

to

show two

things.

(2)

<a<l

(ii)

>b>

I.

IfZ>>0and>0and

< y 2 implies x < y.


2
(i) We begin by assuming that* <y and try to deduce that x <y
the inequality x < y through by x >
(rule III). We obtain

and secondly, that x 2

B'

Multiply

prove that

x <xy.

Similarly

xy<y 2
But

We now

(3) If a

now* 2 <y 2

(ii)

to

<

~b

>

+A
+B
ft

added).
follows from rule

assume that

<y
2
y -x >0

both sides of*

x2

<y

(rule II),

I.

and c

< bA

(y~x)(y + x)>0.

Deduce

that

A
B'

> d, prove

If, also,

< y. Adding x 2

obtain

(4)

Show

that a

+c>b+d(i.e. inequalities

> and d > 0, prove

that ac

that each of the following inequalities

though

a>Z>andc>d.

i.e.

can be

> bd (i.e. inequalities

between positive numbers can be multiplied).

and try and deduce that*

we

ali

(i)a-c>b-d
(1)

may

fail

to hold even

Real numbers

Real numbers
there

<>1

these symbols and that y/y simply

no ambiguity about

is

means

'y/y or

-y/y.
The genera] quadratic equation has the form

ac>bd.

(iii)

(5)
(6)

ax

we impose the extra condition that b > and d > 0?


Suppose that, for any e>0,a e<b<a + e. Prove that a = b
Suppose that a < b. Show that there exists a real number x satisfying

What happens

where a

Roots
It

Let n be

a natural

For example, x

y =x".

number. The reader

y>0 there
y =

is

be familiar with the notation

will

=x.x and* 3 = x.x.x.

Our next assumption about the


any

+ bx + c =

real

>

one
b

number system

exactly one value of x

is

Multiply through by 4a.

= 0.

We

obtain

4a 2 x 2 + Aabx + 4ac =

a<x<b.

1.9

if

{lax

+b) 2 -b 2 +4ac =

(2ax

+ b) 2 =

-4ac.

follows that the quadratic equation has no real solutions

real solution if

4ac =

and two

solutions if A

real

if

4ac < 0,

4ac > 0.

- 4ac)

If

- 4ac > 0,

the following. Given

2ax+b =

such that

y/(b

-4ac)

2
y/(b -4ac)

-b

x =

x".

2a
(Later on

we

shall see

how

this

property

may be deduced from

the theory of
2
The roots of the equation ax

continuous functions.)
If

> 0, the value of x >

nth root of y and

x = y Un

is

which
denoted by

satisfies the

equation

y =x

n
is

called the

vention,

2,

we

also use the notation y/y

= y xn

Note

> 0. If y > 0, there

are,

If r
If

is

= (y m y ,n

we can

Y =

u
v

-b +

is

also write

defined for

real

all

then

r is

therefore define

simple matter.to check that, for

ax 2

of course, two num-

+bx +

With the help of

y -ax 2 + bx +

values of x,

= a(x-a)(x-p).

formula,

this

all

we can

sketch the graph of the equation

c.

a positive rational

r
and hence y~

is

defined.

'

y = ax 2 + bx +c

y by

rational

With these conventions

numbers

r.

it

follows that,

(The definition of y

if

when x

y>
is

0, then

a<0

an irrational

wait until a later chapter.)

1.11

1.10

is

asserts that, if a,,

> 0, the equation x =y has two solutions. We denote

solution by y/y.

The negative

solution

is

therefore

A nice

Example

described above

Quadratic equations
If j>

= axJ - bx + e

7^

number must

2a

fl>0

We

y/(b
~

a negative rational,

y>

that, with this con-

whose square is y. The positive one is y/y and the negative one fcy/y.
The notation y/y means 'y/y or y/y'.
If r = m/n is a positive rational number and y > 0, we define
r

and

bers

-b-y/(b -4ac)

a =

2a

always true that y/y

it is

are therefore

It is

When n =

+ bx + c =

work on quadratic equations


Cauchy Schwarz inequality. This

application of the

the proof of the important

a2

. .

,a andZ>i,Z 2

,fe are

any

real

numbers, then

the positive

y/y. We note again that

(a,6,

+a 2 b?

Bnh? <

(i

a\

...

a n ){b

+b\

b n ).

/?ea/

Proof For any x,

For the case

<

(a

(a?

x+b

(a 2

a n )x

= Ax 2 + 2Bx +
Since

Ax 2 + 2Bx + C =

x+b
+

2)

2ia.br

...

a n h n )x

(b\

is

1.13
for all values of x,

it

follows that the equation

cannot have two (distinct) roots. Hence

1.12

Irrational

3) this inequality amounts to the assertion


triangle is less than or equal to the

sum

this.

numbers

we mentioned the existence of irrational real numbers. That


such numbers exist is by no means obvious. For example, one may imagine the
In

.2

all the rational numbers on a straight line. First one would


Then one would move on to the multiples of j and then to
the multiples of i and so on. Assuming that this program could ever be completed, one might very well be forgiven for supposing that there would be no
room left for any more points on the line.

process of marking

mark

what we had

one side of a

+ b 2n)

B 2 <AC

which

2 (or

of the lengths of the other two sides. Explain

-4AC<0

(2B) 2

that the length of

+ (a n x+b n ) 2

...

C.

y = Ax + 2Bx +

i.e.

numbers

Real numbers

to prove.

the integers.

Exercise
(1)

Suppose that n
a-"

=y

has

solutions

no solutions
y > 0.

if

< 0, one solution if y =

Draw graphs of y = x and y = x

iUtlLM

and two

that the equation

2'3

(ii)27"

> 0, z >

4'3

and r and

to illustrate these results.

assumption requires us to accept the existence of

ber* (namely \J2) which


be expressible

(iii)32

s are

But our assumption about the existence of nth roots renders this view untenable. This

(2) Simplify the following expressions:

(3) If y

-1

if

Suppose that n is an odd natural number. Prove


x" = y always has one and only one solution.

(OS

tt

an even natural number. Prove that the equation

is

satisfies

x2 =

2.

a positive real

If* were a rational number

it

num-

would

form

in the

6/s
.

x =

any rational numbers, prove the follow-

ing:

where

()

r+s

= ff

(ii)

>

(4) Suppose that a

rs

= (yj

and that a and

(yzj

(iii)

= fz\

m and n are natural numbers with no common divisor (other than


m2 =

are the roots of the quadratic

+ bx + c = (in which b 2 4ac > 0). Prove that


2
=
ax
bx
+ c is negative when a < x < and positive when x < a
+
y
orx >j3. Show also (without the use of calculus) that y = ax 2 + bx + c
achieves a minimum value of c b 2/4a when* = bj2a.
Leta,,a 2
a be positive real numbers. Their arithmetic meanA n
equation ax 2

(5)

and so

m2

// are defined

g,+g 2 + ...+a

Deduce from

the

andb

Minkowski's inequality,

,b 2

a2

<

,b n

beany

that
real

>,al
k

=i

Hn < A.
numbers. Prove

then

1/2

*:#
t=i

is

even. (If

= 4k 2 + 4& +

Thus n
This

is

is a

= 2k 2

which

m were odd, we should have


odd.)

is

We may

therefore write

even.

We

have therefore shown that both

contradiction and

it

follows that

x cannot be

and n

are divisible

rational,

i.e.

by

2.

\/2 must be

an irrational real number.

Of course, \/2

is

not the only irrational number and the ability to extract th

roots allows us to construct

i.e.

1/2

(^ + b k y

n\ai
Cauchy-Schwarz inequality

(6) Letai,a 2) ... ,a n

--

4k 2 = 2n 2

by

III.
+ -+... +
H-' =

even. This implies that

is

n
_

In 2

m = 2k + But
m = 2k. Hence

and harmonic mean

1). It

follows that

many

others. But

irrational

numbers can be obtained

irrational

number is

in this

should not be supposed that

it

way.

It is

a root of an equation of the

not even true that every

form

all

Real numbers

10

Real numbers

+ a x x + a2 x 2 +

+ a n x" =

\a+b\ 2

11

+ bf =

{a

,o n ate rational numbers. Real numbers


which the coefficients a ,a ,
which are not the roots of such an equation are called transcendental, notable
examples being e and it. This fascinating topic, however, lies outside the scope
in

of this book.

\a\

<

\a\

\a\

Modulus

It

Suppose that x

a real

is

number. Its modulus

(or absolute value) \x\

follows (from example

is

+ b\ <

\a

|o|+

=
1.14

+ b2

lab

+ 2\ab\+

2ab

\b\

\b\

2\a\-\b\

(\a\+\b\)

(theorem 1.15)

\b\

(theorem 1.16)

2
.

.6) that

\b\.

defined by

Theorem For any

1.18

Ay

>

\c-d\

1*1

-x

Proof Take a
a + b =c and so

< 0).

(x

|c|

Thus

|3

It is

3,

=6

141

and

= 0. Obviously

|0|

sometimes useful to note that

x =
|

>

|*|

for

all

s/x

(i)
.

(ii)7

1.15

Theorem For any


-|x|

the second case,

\ab\

or x

< 0. In the first case, |x| <

<* =

(1)
|x|.

In

real

(2)

Prove that

(3)

Proof The most elegant proof is the following.


\ab\

s/((abf)

y/(a b

2
)

sj(a ).yj(b )

\a\.\b\.
1

The next theorem

is

of great importance and

inequality. This nomenclature


is

may

is

>
=

|*

y\. Show

d(x,y)>0

(ii)

d(x,y)=d(y,x)

.12(6)) with

We

is

.6.]

that, for

numbers x undy

real

is

defined by

any x,y andz,

= 0ifand only if x=y


(iv)d(x,y)<d(x,z) + d(z,y).
(iii)d(;c,.y)

numbers, prove that

sy/2

(5)

For any

real

numbers a and b

irrational unless^

Suppose
2

0.

that the coefficients

a=0,b and

|6|.

c of the quadratic equation

+ bx+c =

are

\a\+

things to prove as in example

\\c\-\d\\.

(i)

>

(triangle inequality )

+ b\ <

+3=4.'

<b if and only if b <a<b.

you have two

The distance d(x,y) between two

ax

\a

be justified by observing that the theorem

give a separate proof.

Theorem

|a|

(4) If r and s are rational

usually called the triangle

the special case of Minkowski's inequality (exercise

1.1 7

+ \c-d\.

\d\

Prove that

\c-d\

numbers a and b

\a\.\b\

= |6-(-l)|> |6|-|-1| = 6-1 =5.

d(x,y)

\b\

= |-1+3|<|-1|+|3| =

[Recall that

1*| X <0 < |*|.

Tlieorem For any

|a|+

Then

Exercise

1.20

>

<

b\

the triangle inequality.

numbers:,

<x<|x|.

Proof Either x

1.16

real

= d and b = c d in

Example

1.19

values of*.

\a

numbers c and d,

|c|-|d|.

(x>0)

real

all

where

rational
r

and s

numbers and

that

are also rational

a=

+ s\/2 is

root,

Proof We have
t (6)

Prove that

3"2 and 2 1/3

root of this equation,

numbers. Prove that

are irrational

numbers.

(3

= r s\/2

is

also a

13

Continuum property

A similar lesson is

to be learned from Archimedes'

method of exhaustion.

This was invented by Archimedes as a means of evaluating the area of regions

with curved boundaries.

CONTINUUM PROPERTY

In the case of a circle, the

and

larger regular

Ai,A 2 ,A 3

Achilles

2.1

and the

The following is one of the famous paradoxes of Zeno.


of

When

feet.

tortoise will have

Achilles reaches the point

advanced a

bit,

say

where the

Achilles

is

is

to

tortoise started, the

little

bit

more, say

Achilles can never

catch the tortoise.

The

area

circle to

be

of the circle

is

then identified with the 'smallest real

**

and calculating the

A<

The discussion of

larger than all of the

numbers x ,x

paradox

is

feet,

to say that Achilles catches the

where x

+ x u x + x x2
y

is

'the smallest real


.

.'.

number

Zeno's argument

then simply reduces to subdividing a line segment of length x into an infinite

number of smaller

segments of respective lengths x ,x lt x 2

line

By taking

for a large

number

the radius of the

enough value of n,

approximation to

-n

as

he chose.

the previous section indicates the necessity of

a further assumption about the real


if

number system.

It

would, after

Achilles never caught the tortoise or if a circle

new assumption

all,

had no

making

be most
area.

We

Continuum Property of the real numbers.


The Continuum Property is not usually mentioned explicitly in a school
algebra course, but it is very important for what follows in these notes. The rest
of this chapter is therefore devoted to its consideration. We begin with some
our

shall call

to resolve this

The Continuum Property

2.2

unsatisfactory

he has run a distance of x

as close an

.'.
.

m*
>

x,

A
The simplest way

Note that this method again depends strongly on the existence of a smallest
real number larger than all of the real numbers Ai,A 2 ,A-$,...

>

A 2 ,A 3
value of A n

each of the numbers/^,

Archimedes was able to obtain

tortoise after

and then calculating their areas

given a

larger than

-*

circle

soon reaches the tortoise's

feet. Achilles

new position, but, by then, the tortoise will have advanced a


x 2 feet. This argument may be continued indefinitely and so

involves inscribing a sequence of larger

tortoise

race a tortoise. Since Achilles runs faster than the tortoise, the tortoise
start

method

polygons inside the

the

terminology.

A set S of real numbers is bounded above

which

is

if

there exists a real

greater than or equal to every element of the set,

i.e. if,

number H
some //,

for

x<H
Formulated

in this

way, the paradox

loses its sting.

This solution, of course, depends very strongly on the existence of the real

number x, i.e. the smallest


x +*!, jc +x$ +x 2
,

12

real

number

larger than

all

of the numbers

for

any x

e S.

The number H (if such a number exists) is called an upper bound of the set
A set S of real numbers is bounded below if there exists a real number h

S.

which

is

than or equal to every element of the set,

less

15

Continuum property

Continuum property

14

for

i.e. if,

B<H.

x>h

Similarly, if S

any

non-empty set which is bounded below, then S has


any other lower bound h,

is

a lower

that, given

bound b such
for

\H

some h.

xGS.
number

lower bound of the

The number h

(if

A set

both bounded above and bounded below

which

is

such

exists)

is

called a

is

just said to

b>h.

set S.

be

bounded.

A set S of real numbers is

Proposition

2.3

exists a real

number K such

any

if

and only

if

there

Examples

2.5

<K

|jc|

for

bounded

that

xG S. (This

is

easily

proved with the help of exercise

.20(1).)

bound

smallest upper

bound of the

of the

(i) The set {1 2, 3} is bounded above. Some upper bounds are 100, 10,
The
4 and 3.
set is also bounded below. Some lower bounds are 27,
and
(ii) The set {x
< x < 2} is bounded above. Some upper bounds are 100,
10, 4 and 2. The set is also bounded below. Some lower bounds are 27,
and

set

set {1,2, 3} is 3.

{x

< x < 2}

is

The

2.

largest

The

lower

largest

lower

is 1

The

(iii)

Examples

2.4

The

(i)

bound of the set {1,2, 3} is 1.


(ii) The smallest upper bound of the
set

{x

> 0} has

no upper bounds

at all.

The

largest

lower bound

{x:x>0}is0.

set

1.

(iii)

The

{x

set

> 0} is unbounded above.

bound, one has only to point to

H+

If

H>

Some

lower bounds

and

Property,
is

proposed as an upper

non-empty

If a
it

times called the

:x> 0}

is

B =

bounded below.

Continuum Property which

will

be fundamental for the

bound

is

B such

bounded above, then S has an upper


any other upper bound //of S,

non-empty

that, given

We

write

sup

is

some-

or

sup x.

which
set

is

bounded below has

and write b

inf

a largest lower

bound

b.

We

call

or

inf x.

Sometimes you may encounter the notation sup S = + . This simply means
that 5 is unbounded above. Similarly, inf 5 = means that S is unbounded
below.

2.7

Maximum and minimum


If a set

set S.

xGS

Continuum Property
Every non-empty set of real numbers which is bounded above has a
smallest upper bound. Every non-empty set of real numbers which is
bounded below has a largest lower bound.

if

bounded above, then, by the Continuum

0.

remainder of these notes.

Thus,

is

i6=S

b the infimum of the


state the

supremum of the

Similarly, a set

We now

set

has a smallest upper bound B. This smallest upper bound

to obtain an element of the set larger

than the supposed upper bound. However, the set {x


are 27

Supremum and infimum

2.6

set

which

is

has a largest element

and write M = max


1

S and write

S. If

= min

S.

M, we

call

has a smallest element

M the maximum of the set S

m, we

call

m the minimum

of

Continuum property

16

It is fairly

and

its

obvious that,

bound

smallest upper

A common

error

is

M.

Tlius,

maximum M, then it is bounded above


in this case, sup S = max S.

to suppose that the smallest upper

is

maximum

always the

S has

if a set

of the

However, some

set.

sets

bound of a set S is
are bounded above

These are the bounded intervals (classified by whether or not they have a
maximum and whether or not they have a minimum). We also need to consider
the unbounded intervals. For these we use the notation below.

which

(and hence have a smallest upper bound) do not have a

maximum.

(See example

2.8(a).)

Similar remarks, of course, apply to largest lower

bounds and minima of sets.

Examples

2.8

The set {l
upper bound. The set
(i)

maximum
3} has minimum

2, 3} has a

{l

2,

3 and

this

is

equal to

its

and

this

is

equal to

its largest

The

largest

{x

set

hand, any x

set

because

in the set satisfies

it

does not belong to the

< x < 2. But

set.

On

the other

is

The

set

{x

The

set

The
{x

set

is

However, {x

larger than x.
:

:x>d)

(-,b)

= {x:x<b}

(-,b]

= {x:x<b}

(,)

W//W////7W
/y////////////7\

this

of the empty

Hence x cannot be the

(why?). This

is

equal to

its

2.10

largest

> 0} has no maximum, nor does

In

> 0} has

no minimum (why not?).

number between them.


describing intervals we use

(1)

it

Its largest

lower bound

is

0.

Which of the following statements are true?


(iii)
(ii)2e(-,3]
0)36(1,2)
(v)3G[l,3].
(iv)2e(0,2)

(2) If %

is

a real

[a,b]

the intervals

real

16

[1,2)

> 0, prove that

number and 5

5).

does

down

the following notation:

= {x:a<x<b}
= {x:a<x<b}

wmm
(a,.'

[a,b)

= {x:a<x<b}

(a,b]

= {x:a<x<b}.

is

bounded above. For those sets which are bounded above, write down
three different upper bounds including the smallest upper bound.
Decide which of the sets have a maximum and, where this exists, write
its

value.

(iv)(3,)

b)

of all

or not the given set


(3) Decide in each of the following cases whether

(i)(0,l)

(a,

We

Exercise

have any upper bounds.

An interval I is a set of real numbers with the property that, if x E


y Eland x<z^y, then zEI, i.e. if two numbers belong to /, then so

every

set
call

and the

largest

Intervals

and

set

such an interval compact.

{x:|-x|<S} = a-S, +
2.9

are real num-

notation into supposing that or

numbers) fall
(a,6),(a,)and( ,b)open and the intervals [a, b], [a, ) and ( ,fe]
closed. (The intervals [a, b) and (6, a] are sometimes called half-open.)
Of particular importance are the closed, bounded intervals [a, b] We shall

then

< x < 2} has smallest upper bound 2.

< x < 2} has minimum

{x

{x

V7/////////'y/7//7/7////y,

All intervals (with the exception

lower bound.
(iii)

x >a}

an element of the set which


set.

[a,)

bers or that they can be treated as such.)

call

element of the

{x

into one of the categories described above.

x+2

y =

(Do not be misled by

< x < 2} has no maximum. The number 2 cannot be the

element of the

(a, )

smallest

lower bound.
(ii)

17

Continuum property

(ii)(-,2]

(iii)

{-1,0,

(4) Repeat the above question but with the

mum'

2, 5}

(v) L0, 1].

words 'above, upper and maxi-

replaced by 'below, lower and minimum'.

(5) Give an example of a set which has smallest upper


x 4.
no element x satisfying 3

bound 4 but contains

< <

(6)

Show

that, given

any element x of the

element y 6 (0, ) with


has no

minimum.

set (0, ), there is

the property that

;'

another

< x. Deduce that (0, )

Continuum property

2.11

Manipulations with sup and inf

We

prove one result and

(3)

some others

list

as exercises.

Suppose that 5

Theorem Suppose that S is


bounded above and % > 0. Then

2.12

Proof Let

non-empty

set

of real numbers which

Then

B is

the smallest

number such

that, for

bounded above. Prove that

sup x.
.*es

Hence obtain analogies of theorem 2.12 and exercises 2.13(1 and 2) in


which inf replaces sup.
non-empty set 5 of
(4) The distance d{%, S) between a real number and a
real numbers is defined by

is

% sup x.

B = sup S.

is

( x)

inf

is

sup %x

any

d(S,S)

inf

|$-*|

iS

x<ES,

(see exercise 1.20(3)). Find the distance

x<B.

and the
Let 2"- {#*

* S}. Since | > 0,

(i)S

(5)

any* G 5. Hence Tis bounded above by /?. By the Continuum Property,


rhas a smallest upper bound (or supremum) C. We have to prove that C = %B.
Since %B is an upper bound for 7* and C is the smallest upper bound for 7

(ii)

repeat the argument with the roles of S

number such

that, for

any

and

T reversed. We know that C

y&T,

(6)

y<C.
Since

> 0,

it

follows that

any

for

that

y G T. But S = {|"V ? G 7% Hence ?"' C is an


:

the smallest upper

bound

for S.

upper bound for

S.

But

5<C.
We

have shown that

2.13

C< |5 and

also that fi

< C. Hence \B = C.

Exercise
(1)

Suppose that S
sup

is

bounded above and

(2) Suppose that

S is bounded

that

sup (x

*S

that

S C

S. Prove that

< sup 5.

+ ) =

sup x

iS

%.

above and that

is

any

real

number. Prove

S, prove

is

for

Give an example of a real

which d(, S)

prove that d{%, /)

an open interval (other than

S and Tare

bers in

Thus

num-

implies that \

R or 0), show that


= 0.

j-

G /.

^/can

always be found for which d(, /)


Suppose that every element of an interval / belongs to one or the other
of two non-empty subsets 5 and T. Show that one of the two sets S or
T contains an element at zero distance from the other. (Do not assume
consist of

# is

= (0,l)
= (2,3).
that d(|, 5) = 0.

If /is a closed interval,

(iii)

If/

Now

(iv)S

ber ? and a non-empty set

the smallest

number

but G" S.
If 5 is bounded above and % = sup S, prove that d(|, 5) = 0.
Deduce that the same is true if 5 is bounded below and = inf S.

for

"I"

real

(ii)5

1,2}

{0,

If

(i)

between the

sets

(iii)S=[l,2]

is

19

Continuum property

18

all

necessarily intervals.

rational

numbers

in /

The

set S, for

and the

set

example, could

T of all

irrational

num-

/.)

[Hint Suppose that $ < t, where s G S and t G T. Let T = {x x G T


and x > s}. Show that To ^ and is bounded below. Write b = inf T
and consider the two cases b Tand b ^ T.\
:

21

Natural numbers
for

all

Hence the
bound

foJ.

set

is

bounded below by

We

0.

have to prove that

is

the largest lower

Suppose that

NATURAL NUMBERS

n'

/;

ft"

S.

Introduction

3.1

we explore some

In this chapter

implications of the

It

Thus

bound

for S.

for

each n

M,

an upper bound for the set

is

the largest lower

is

Then,

fol

follows from this contradiction that

Note that

Continuum

a lower

But then h'


above.

is

>h.

Hence, for each n


n

>

bound

which we know is unbounded


no h > can be a lower bound for

for S.

0^5 and hence S has no minimum.

Property for sets of natural numbers. As always, arithmetical properties are

taken for granted. But where


will

proof hinges on such a property, the property

Theorem Every

3.5

be explicitly stated.

set

of natural numbers which

is

not empty has

mini-

mum.
Archimedean property

3.2

Proof For this proof we need to assume that the distance between
numbers is at least 1
non-empty
set of natural numbers. Given that all natural numbers
be
a
Let S
are positive,
is a lower bound for S. By the Continuum Property, it follows
that S has a largest lower bound b. Since b is the largest lower bound, b + 1 is
not a lower bound. Therefore, for some SS,
distinct natural

The Archimedean property

Theorem The

3.3

set

is

the assertion of the following theorem.

M of natural

numbers

is

unbounded above.

Proof For this proof we need to assume that, if n is a natural number,


n + 1
Suppose that the theorem is false. Then N is bounded above. By the Continuum Property it therefore has a smallest upper bound B. Since B is the
smallest upper bound for N, B 1 is not an upper bound for N. Thus, for some
then so

is

n
If n

wj

is

the

e S,m<

N,

>B-\

n+
But then n

is

is

minimum of S, there
n.

M. This is
unbounded above.
set

Example Prove

5 =
is

{-'

fsj

which

is

greater than B. But

a contradiction.

is

3.6

not, then, for

some

< n m <

Does S have

maxi-

1.

Exercise

Prove that the set

nl :e

is

largest

bounded above with

>0

smallest

upper bound

mum?
lower bound

(2) Let

0.

are

all

positive,

X>

Prove that the set

S = {X :n&U}

o
is

unbounded above.

BX~
20

If

that the set

Proof Since the natural numbers


_1

nothing to prove.

an

(1)

:neN}

bounded below with

is

obtain the inequality

<m< n < b +

s =
3.4

We

from which follows the contradiction

an element of the set

upper bound of the

Hence

>B.

<b +

[Hint: if B were the smallest

could not be an upper bound.] Show

upper bound, then

that, if

0<x <

1,

then the

Natural numbers

22

set

T =
is

{x

:nGN}
bound

0.

smallest integer satisfying


(5) Let
that

last part,

(6) Let a
a < %

<
<

ment S n

< r<

and smallest upper bound

Show

bounded and

[Hint: for the

= 2 =1

k. Let P(n) be the state-

= Sn + (n +

l)

\n(n

1)

l)(n

+n+

(n

1)

1)

2)

satisfying
as required.

b. [Hint: exercise 1.20(4).]

The

result

now

Example The

3.10

Suppose that a line of dominoes is arranged so that, if the th one falls,


knock over the (n + l)th. If the first domino is pushed over, most people
would agree that all the dominoes would then fall down.

real

follows by induction.

number

An =

(at
n

+a 2 +

an)

will

The
below

principle of induction

how it may
Theorem

3.8

Suppose

is

an idealisation of

this

simple notion.

We show

is

the arithmetic

positive, their

be deduced from theorem 3.5.


(principle
that, for

number n. Suppose

of induction)

each n

The

P(n)

is

P(n)

Then P(n)

Gn <A n

1) is
.

P{n)

is

We want

false}.

not empty. Then from theorem 3.5

m
the minimum of S. Thus m

P(l)

is

true,

(fc

S.

,a 2

... ,a. If 01,22, ... ,a n are

1/n
.

if

is

a natural

number

LslP(n)be the statement 'For any


< A'. We shall show that
n
(i) P(2 ) is true for each nGN.

S=

defined by

induction' (see exercise 3.1 1(5)).

true.

Proof For this proof we need to assume that,


other than 1 then 1 is a natural number as well.
Let

is

Proof This inequality has an entertaining proof using 'backwards


P(n

true for every n ftj

is

(a,a 2 ...fl n )

Gn

'inequality of the arithmetic and geometric means' asserts that

a statement about the natural

also that

true, then

is

mean of the numbers a

geometric mean

Gn =

(i)P(l)is true
(ii) if

is

= & +
number

...

1).

have

use the previous question.]

b. Prove that there exists an irrational

+2 + 3 +

1).

that

Sn +

is

\n{n

\n(n

Principle of induction

3.7

it

.+ =

S 1 = l,and \. 1(1 + 1) = 1. Hence P(l) is true. We now wish to


if P{n) is true, then P(n + 1) is true'. To do this, we assume that P(ri)
=
is true and try and deduce that P{n + 1) is true. Thus we assume that S
=
\n(n + 1) and try and deduce that 5 n+1
k(n + 1)(" + 2). But
We

show

m>an.]

S be the set of all rational numbers r which satisfy


S has no maximum and no minimum. Prove that S

has largest lower bound

Proof Let S

non-empty set of integers which is bounded above has a


that
any non-empty set of integers which is bounded
maximum and
minimum.
below has a
Let a < b. Prove that there exists a rational number r satisfying
a<r<b. [Hint: justify the existence of a natural number n satisfying
l
n~>(b a)~ and consider the rational number r = tn/n where m is the

(3) Prove that any

(4)

+2 + 3 +

bounded below with largest lower

nN

Example For each

3.9
n

23

Natural numbers

Hence

true (because of hypothesis

=fc 1
1

(ii)

S is empty. Suppose that S


that S has a minimum m. Since

to prove that
it

follows

Therefore

S and

so

P(m

number. But

But then P(m)

is

is

a natural

1)

is

of the theorem). Hence

true.

m $ S which

is

is

(ii) If P(n) is

The

We

< (V^i - y/a 2


(fl|a2 )

diction.

This

2+i
true,

is

1/2

,a n

2
)

Now,

-at 2yJ(a

a2 )

a2

<Ki+^)-

simply P(2).

_ 2m. We

P(n 1) is true.
by backwards induction.

begin by proving P(2).

a contra-

numbers at, a 2

true, then

result then follows

(i)

positive

We now show

that

P(2

n
)

implies P(2"

').

Let

m = 2". Then

have to assume P(m) and try and deduce P(2m). Since P(m)

is

24

Natural numbers

Natural numbers

...a m ) l "n

(a 1 a 2

<-(a +a
1

Deduce

25
any

that, for

real

numbers a and

a"-b" m (a-bXa"' + a"'


1

a m*\ a m+2

and

<

a 2m)

~~

(a

m + 1 + "m + 2 +

+ 2m)-

(3)

A polynomial P(x) is an
=

P(x)

But we know that P(2)

true.

is

If

(a,

+ am

1/m

}" 2

P{)

..a 2m )

V2m

<

(a,

<z 2

4-

We

is

is

have shown that P(2

We now show

true

) is

that, if P(n)

is

true for

is

.+ ab"' 2 + b"' 1 ).

Prove that P(x)

0.

all

A: is

n and that, for some if,


Q(x) is a polynomial of

= (x - %)Q(x) where

[Hint: use the previous question.] If gj,

numbers and P{%\) = P( 2 )

2-

2,

are

= P(% n ) = 0, prove

that

constant.

= 1.2.3...n

and 0!

1,

we

define

n+1

that, if P(2 ) is true, then P(2


) is
by straightforward induction.
true, then P(n 1) is true. Suppose that

and

said to be of degree n.

is

a polynomial of degree

is

= k(x-^)(x-$ 2 )...(x-U

where

/'(2/m).

true.

polynomial

(4) Assuming that w!

Hence P(2")

(ii)

P(n)

P(x)

+a 2m )

2ffi

which

a"' b

+ aix+a

= 0, the

degree n

Thus

true.

an

distinct real

a2

expression of the form

+ a^iX"' +

Suppose that P(x)

a
+...+a 2m
+ m+1

m
(ai

b,

Hence

...0 ,/m (^^m + 2---2 m

{(flia2

anx

(r=0,l,2,...,ii).

|^J

r\(n-r)\

Then

Prove that

{(hai ...an .t

Gn -

^ai+a 2 +
<

yvnn

+ a_, + G_i

...

+l
(r=l,2,...,n).
r

--]

MM / (H-lMn-l+Cn-1

Using

this result

and the principle of induction, obtain the binomial

ft

theorem

Gn _,
ThusP(n

{a

</!.,.

in the

+ bf =

a"

y.

form

+ na"'

^^

a"- b

+ b"

l)is true.

This completes the proof.

a"- b

r
.

V
3.11

(5)

Exercise

Let P(n) be a statement about the natural


(i)

by induction that

(1) Prove

&2 =

2
l

+ n 2 = in(n + l)(2n+I)

(6)

nGM
then P(n 1)

true,
is

unbounded above
Suppose that <

true for every

A3

3
l

+ n3

l)

(2)

Prove by induction that,

*
fe=o

fe

1+x+x

if

l-x n+1
1-as

true.

[Hint: the set {2"

nGM,

jc

-y

SM

X< Y and that x and y belong to the interval


^TK| x

= 1,

+ ...+*"

is

nN.

AT""

that

is

(exercise 3. 6(2)).

Prove that, for each


(ii)

number n and suppose

true for each

Prove that P(n)

I=

fe

is

(ii) if P(n) is

n
(i)

P(2")

[Hint: exercise 3.1 1(2).]

1/n

-/'" < YX
|

1 '"

\x-y\.

[X, Y].

27

Convergent sequences

CONVERGENT SEQUENCES
It

infinity'

it, it is

Two

as n

-(-!)"

_ _,)1 =^a-(-r).
(

xn

as n gets larger

from this formula?


and larger, ( )" gets

gets closer

and closer to

\.

We

closer

and

say that 'x n tends to as

closer to

n tends

to

or, in symbols, x n -> \ as n -* \


is

of the greatest importance, but, before we can make proper use of

necessary to give a precise formulation of what

it

means

to say that

xn

-> I

-* .

bulldozers are moving towards each other at a speed of one mile

per hour on a collision course.

bee was perched on

When

they started they were one mile apart and a

the front of one of them.

The bee began to

fly

back and

for-

ward between the bulldozers at a constant speed of two miles per hour, vainly
seeking to avoid his fate. How far from his starting point will the unfortunate
insect

+ (-|f-*}

the answer \ to be extracted

This idea
bulldozers and the bee

)+(-)*+...

("

seems clear that,

zero and so

The

li

3 ll

How is

4.1

!{1

4.2

Sequences

A sequence may
More

precisely,

to each natural

be crushed?

be regarded as

a list

of numbers Xi,X2,x 3 ,x 4 ,

we can say that a sequence is determined by a rule which assigns


number n a unique real number x n We call x n the n th term of
.

the sequence.

The notation

fef"

means the sequence whose th term isx.


The set {x n n N} is called the range of the sequence. We say that a
sequence is bounded above (or below) if its range is bounded above (or below).
Thus (x) is bounded above by //if and only if-xr <//( = 1, 2, 3,
.).
:

fe#

fr^S
4.3

This riddle

usually posed in the

is

hope that the victim

complicated calculations involving the


that the bee will be crushed

when

flight

will

of the bee. But

it is

quite obvious

xr

(iii)<l>

(v)

1
oo<- >=

i,...

1,1,1,1,...

(iv)(2

The sequence

(x) defined inductively

The

first

2Y n

l+ x

few terms

(n

by *i

= 2,3,4,

n
>

i,|,

M..

2,4,8,16,

2 and

...).

n. r

are 2, i. 12.

I>

2
3

4.4

Definition of convergence

A sequence (x n ) is said to converge


following criterion

in general.

= -!+f7 -... + (-iy-

26

X "~

n th landing. Then

xv 2 - 2_2
3
9
and,

= -i,+ i,-i, +

the bulldozers collide. Since the bulldozers

same speed, this will happen halfway between their starting points.
The answer is therefore one half mile.
Let us, however, take up the role of the riddler's victim and examine the
flight of the bee. Letx denote the distance the bee is from his starting point
his

Examples
(i)<(-ir>

embark on some

travel at the

when he makes

2(i)"

is

satisfied.

to the limit I if

and only

if

the

Given any e
\x n

> 0, we can find an N such that, for any n>N,

write

-l\<e.

xn

<e.

'"

But

We

29

Convergent sequences

Convergent sequences

28

lim

* or

as n

'"

and so we simply need to find an

N such that, for any n>N,

/.

-<6
It

x 2 ,x 3

be helpful to think o{x h

may

... as successive approximations to

number /. The distance \x n l\ between x n and / is then the error involved in


/ by x. The definition of convergence then simply asserts that we
can make this error as small as we choose by taking n large enough.
The diagram is supposed to represent a sequence <x> with the property that
the

approximating

-*

as

i.e.

But

this solves the

problem.

We

simply chose A'

1/e.

Then, for any

n>N,

-* <*>.

n>N

= e

-<e

i.e.

l+ p.

i.e.

We

<e.

have shown that, given any e

such that, for any

> 0, we can find an N (namely N=l/e)

n>N,

n>6

<e.
For the value of e indicated in the diagram,
definition

isN=

\x n

6.

For each value of

a suitable value for

>6,

It is

-l\<e.
l\

N depends on e,

i.e.

for each e

values of e and the corresponding values of

>

we

use a

N are entered in

has been noted in the diagram.

> 0, we can

on the very small values of e > 0. It is clear


from the diagram above that, if we had chosen to look at a very much smaller
value of e > 0, then we should have had to have picked out a very much larger
value of N. In general, the smaller the value of e > 0, the bigger must be the
'

Some

the table below.

Notice that the definition of convergence begins 'Given any e


find an

important to observe that

different N.

In particular, the value of \x 9

N in the

Here the emphasis

0-4

2-5

0-1

10

0000 001

is

corresponding value of iV.

4.5

Example
1

+ -*

as n -* .

Proof Let e

>

be given.

Some

authors

insist that A'

nition of convergence a little

n>N,

be a natural number. This makes the defimore elegant but renders examples like that above
marginally more complicated. If we wanted A' to be a natural number in the
example above, we could not simply write A' = 1/e. Instead we should have to
choose
to be some natural number larger than 1/e.)

(Note:

We must

find a value

of ./V such

that, for

any

000 000

Convergent sequences

30

Let Af be whichever of

Exercise

4.6

Prove that

(1)

31

Convergent sequences

>N, both

and Af2 is the larger, i.e. A' = max {A^, A^}- Then,
and (2) are true simultaneously. Thus, for any

inequalities (1)

if

n>N,
r-fc4i
n-"~ \n 2 +
1

\(x+yn) -(/

+ )! =

<

Let r be any positive rational number. Prove that

(2)

-*

as n * .

Given any e

Let X be any real number. If x n -*

as n -> , prove that

Xx n

(triangle inequality)

e.

4.9

Example Prove

-> X/ as

what value
of A' is appropriate to any given value of e > 0. But this is by no means always
the case. It is therefore natural to look around for some shortcuts which will
enable us to determine whether a sequence converges (and what its limit is) withfairly easy to decide

it is

m\

\y n

(I

.?-

example and exercises above

any n

In the

l\

\x n

+ (y n -m)\

> we have found a value of N (namely N = max {N\, N }) such


> N, \(x n +y) + m)\< e. Hence x n +j-; + masn^.

that, for

Criteria for convergence

4.7

-l)

<\e + \e =

n
(3)

\(x n

that

- 3ra
2
5 3 + 4 -2
2n

Proof We write

2n 3 -2n

2-3/T2

5n + 4n 2 -2

+ 4-'-2 -3

2-0
2
= as n -* .
+ 0-0
5

out our having to indulge in the painful process of appealing to the definition.

and the next section we give some useful

In this

Proposition (combination theorem) Let x

4. 8

n *

results

and

let

X and

/j

be any

Xx n + ny n -*Xl +

(i)

x n y n -* Im

(ii)

-*

(iii)

as

yn

as

real

of

The supporting argument is as follows. It is obvious from the definition of convergence that 2 -* 2 as n * . From exercise 4.6(2), n' 2 -*
as n * . Hence,

this sort.

* / as n -* and y n -*m

as

[xm as n -*

5+4n- -2M-3 ^-5+4.0-2.0 =


1

The

at this stage.

Readers

who

Proof of 4.8(ij. After

Let e

and

>

prefer to omit the proofs of

Similarly,

be given. Then |e

(ii)

and

any n

we need only show


+ m asn -.

4.8(iii).

(i)

as n

* ,

can find an

-m\<\e.

it

-*

as

Uy n <x n <z n (n =

<^: n > is

Proof For

we can

\x

1\

<eis

this

true if

as

and

then

'sandwiched' between the two sequences

proof it

and only
i

->-

.),

<>> n >

l-e

n>N

- just as the bee of 4.1 was 'sandwiched' between the bulldozers.)

>N U

such that, for any

* .

(Here the sequence

(1)

A^

1, 2,

once

that, if x n -*

follows that

(the sandwich theorem) Suppose that y n *

thatz n -Z as-*.

->l

> 0. Since x n -*

Theorem

4.10

(iii)

may be deduced from

-l\<\e.

we

\y n

exercise 4.6(3)

-*m asn-*, thenx n + y n

find an A' 1 such that, for


\x n

from proposition

(provided that in i= 0).

the theory of the exponential and logarithm functions has been developed.

-+

result then follows

5 as ->">=.

rn

altogether will not suffer greatly as these results

aSH

2as-".

Similarly,

-*

4.8(i),

2-3w"2 ^-2-3.0 =

numbers. Then

The proofs of (ii) and (iii) can be found in the appendix. These proofs are
somewhat technical and we prefer not to hold up the discussion by presenting
them

by proposition

7
I.V-/I

(2)

if/

is

necessary to note that the inequality

e<x</+e.
/

(See exercise 1.20(1).)

and

<z n>

32

Convergent sequences
Let e

>

be given.

We

have to find an

N such that, for any n>N,


\x

-l\<e.

\x n

Since y

-* I as
n

we know

-> ,

that there exists an

But

-I\<e.

Similarly, since z -*

as

>N, both

and

inequalities (3)

A^

is

the larger,

>N2

such that, for any n

i.e.

TV

= max {JVj.A'j}.

and (4) are true simultaneously. Thus,

for

e<z n </ +

Then,

Hence

<x <z

(n

Hence, for any n

.).

xm *

Hence x n

-*

have found

a value

of A' (namely JV

= max {N%, N2}) such

x 1,n

Hence

as

Then x n

-l\<y n
n

->

<

* , by

0<x

\x n

< yn

y n ->

1\

equivalent to

is

as n - 0

and

+ _y + /

y n < x n < + yn
/

as

->.

a n _, and a

xr =

as

= .*.

Then G

and geometric means, take

= x 1/n and>4 n =

(n

+ *)/
1

< _A _iy
X =y =

that

|.v|

< 1. Prove

in exercise 3.1 1(6).)

>

Then

1/n

-l <-(x-l)

+ by the sandwich theorem. If < x <


'"-*as w - oand hence
1. Buty

and hence *"" -* 1


x =y~ where y >

as n

Hence
1

4.8(iii).

that

- .

Monotone sequences

A sequence
+ nh + 2 n(n *

<

Cc>

(!+/?)"

Thus

* .

+1

4.15

..

TVoo/ Write
.

- \ as

-> .

to begin with that

theorem 4.10.

Example Suppose
x"

and hence x

by proposition
4.72

that

-*

and that

+ .

But, using proposition 4.8,


-> / as

as

(=l,2,...).

Proof The inequality

xn

> 0. Prove

-* .

Corollary Suppose that v -

-* I as

as

1.

Assume

\x n

( J)"

->

4. 1 2,

(Alternatively, take

4.11

problem of the bulldozers and the bee, we obtained

In the

as

i/"_!

i.e.
I

fl:2

-l\<e.

\x n

-* .

In the inequality of the arithmetic

/>oo/
1

\xl\<e.

that,

as

Example Let x

>N,

(x-l)/n +

> 0, we
for any n>N,

.).

any

Given any e

1,2,..

the sequence

!-e<y n <x n ^z n <l + e.


e<x n <l + e

i.e.

By example

1, 2,

(=

(by exercise 4.6(2)). Thus, by corollary 4.1 1

-*

e.

<

if

4.74
But_y

Example

4.13

l-e<y n <l + e
/

\x\

n>N,
and

-Q\ =
as

x"

(4)

Let A' be whichever of

-* , there exists an

-l\<e.

\z n

1/rt ->

(3)

such that, for any

n>N u
\)' n

33

Convergent sequences

\)h

.+h"

nh

x n+l >x

(n

is

increasing

l,2, ...).

if

then

34

Convergent sequences

Similarly, <x n >

is

decreasing if

Xn+l <X

*S

Convergent sequences
is an upper bound for the sequence. Hence x
n>/V,
any
Thus, for

Also

(rt=l,2

).

< B (n =

1 ,

2,

.).

B-e<x n <B.
B-e<x n <B + e
\x n

i.e.

-B\<e.
have found an Af such that, for any n

e>0, we
Hence x n -* B as n -*
Given any

>N,

\x n

B\<e.

<*>.

(.v., >

sequence Oc>

decreasing

if

xn +

strictly increasing if x n +

is

< x n (n =

sequence which

(v

increases

is

2,

decreases

> x n (n =

2,

.).

It is

strictly

.).

either increasing or else decreasing

is

called

monotone.

Example

4.16

The
The
The
The

(i)
(ii)
(iii)

(iv)

sequence (2">
sequences (n~

sequence

(1)

is

strictly increasing.

and ( n)

is

are

both

strictly decreasing.

both increasing and decreasing.

sequence {( l)

n
> is

not monotone.

neither increasing nor

It is

decreasing.

Examples

4.18

(i)

exercise 3.6(1) that

4.17

theorem 4.17

Theorem
(i)

If <.x)

is

increasing and

bounded above, then

it

converges to

it

-l

its

If <*>

is

decreasing and bounded below, then

it

converges to

its

(ii)

Proof If (x) is decreasing and bounded below, then < x n ) is increasing


and bounded above. Hence it is only necessary to prove (i).
Suppose that (,x n ) increases and is bounded above with smallest upper bound

We must show
Let e

>

that

xn

-*

be given. Since

B as n -* .
B e is not an

upper bound, there exists at least

largest

x>x N >B e.

<x<

If

as

-*

<*>.

the sequence Ct"> decreases


(exercise 3.6(2)).

and

is

bounded below with

Thus

Example The examples above

4.19

((]

>N,x ~>x^. Thus,

lower bound

>B-e.

But (x) increases. Hence, for any n

are nothing new. But consider the

sequence

one term x N of the sequence such that

xN

as

largest

lower bound.

B.

follows that

smallest upper bound.


(ii)

it

The sequence <( - l)/) increases and we know from


By
is bounded above with smallest upper bound 1

for

any

n>N,

We

first

+-')">.

show

that this sequence increases. In the inequality for the geometric

and arithmetic means, take a^

Then

a2

a n -\

= +
1

0"

and a n

36

y"- i

i
x

^
<

-in +

(^-ixi + (/2

"

[j

(4)

+i

Hence
1

<ll+4l

n-1

Let x be a positive number, and let


satisfying N>x. Prove that

1-2,3,...)

n(n-\)(\ U
SI-

b "l?

-2i-h+

'

(5)

1
x"\<\N a +* liV|
x

M;

<I + 1
<

2!

3!

+ aT 1 )"}

that of example 4.12. If n

ial

n\

1/n
>

+ /*)", then it follows from

the binom-

-^Hl<n.

..

+ -^r,

_1
(because 2"

Therefore h n

as

-* .)

<!)

+2(1- an < 3.

is

simple properties of convergent sequences

4.21

Some

4.22

Theorem

.)

A sequence

Proof Suppose
Then

that

can have at most one limit.

xn

-*

as n -

and x n

= \l-x n +x n -m\<\l-x n +
\

m as -> . Let e >

->

-m\<e +

\x n

be

2e

(triangle inequality)

Prove that
provided that n

n3

Sn

2n3

<

-in*--.

Decide for what values of x the limit

||/

4.23

[x+x \
+ xn

lim

Draw

of*.

limit against the value

we need
Let e

Use the sandwich theorem to prove that


y/(n

1)

y/n ->

as

i.e.

.7 it

follows that, if

x n <b(n = \,2,..

Kx n <b(n=

only prove

), then

1,2,.

.),

< b.

then

-x >-b (n =

(i).

> 0. Then there exists an N such


\x-l\<e

-* .

i.e. l

= m.

Theorem Suppose that x + / as n * .


If x > a (n = 1 2,
.), then / > a.

(ii) If

which plots the value of the

from example

|/

Proof
a graph

sufficiently large. But,

(i)

exists.

is

- m\< e for every e > 0, then - m\ = 0,

(3)

example

theorem that

|f |

(2)

[Hint:

can also be established by a method like

(1

Exercise

(1)

B>3.

given.

4.20

for

that the sequence converges.

(The convergence of <n

bounded above by 3. From theorem 4.17 it follows


converges to a limit B and B < 3. (In fact, B = e = 2-718

the sequence

Hence prove

4.19.]

as n -* .
Vn decreases
)

r!

+-+-^+

tz? =

that ((I

Prove that the sequence (n

'(6)

+ +...+l

= 1+:
Hence

nl

a n
Hence show that n x +

H H)h1

"

(n>N).

- -l\

> N).

Deduce that x /n -* as n -* .
Let a be any positive rational number and let \x\ < 1 Show that there
a+1
|l < 1. Deduce that
exists a natural number N such that (1 + 1/A0

'l\

(n

(N-l)l \N

n\

N be the smallest natural number

n-N+l

x"-

^<

and so the sequence increases. To be able to deduce the convergence of the


sequence from theorem 4.17, we also need to prove that the sequence is
bounded above By the binomial theorem,

,+

37

Convergent sequences

Convergent sequences

l-e<x n <l+e.

that, for

any

n>N,

1, 2,

.).

Hence

38

Convergent sequences

> a (n = 1 2,
Hence, given any e

But x n

.)

and

39

Convergent sequences
so, for

> 0, a < +
/

> N, a < x n < + e.

any n

x n >H.

From example

e.

1 .7 it

follows that a <.

I.

Similarly, a sequence <x n ) diverges

to and x n

as n -* if,

for

any

H > 0, we can find an iV such that, for any n > N,


* I as n -* . If x n > a ( =

Example Suppose that x n

4.24

tempting to conclude that


as n -*

and l/n

>

(n

l> a.
1

2,

.),

2,

.).

xn

it is

may not be true. For example, l/n


But we certainly cannot conclude that

<-H

-*

But this

0>0.

4.25

Tiieorem

Any

Proof Let x n

(n

1, 2,

.).

< e.
1\

that, for

any

-* I as

-*

<*>.

We

II

bounded.

is

have to find a

^ such that

|jc|

<K

(See proposition 2.3.)

true that, for

It is

\x

convergent sequence

any e > 0, there

In particular, this

true

is

exists an

when

N such that, for any n>N,

1, i.e.

there exists an

such

n> JVj,
N

Ix-/I<l.
From theorem
|jc

.18

it

The

follows that, for any

>A

now

result

K =

A
*,,->

divergent sequence

ooas->is

max{|.r,|, \x 2 \, ..., |jf|,

|/|

1}.

Example Let a be

4.28

Proof Let H

A divergent sequence
is

which does not converge. Any untherefore divergent (theorem 4.25). However bounded
a sequence

is

divergent sequences exist as well.

>H
n>H ila
n

i.e.

Proof The sequence


n

-* .

l(~

so
so

1 )"
1

Example The sequence <(


Then, given any e

< e.

is

l)

>

find an

1)"

as

N such that, for any n> N,


N and

the contradiction

and

>

be given.

We

and

that, for

any

= H Ua

> N, n >Hva and hence n a > H.

Then, for any n

A common

error

is

to suppose that proposition 4.8 (about

example,
then

xn

it is

tempting to suppose that,

if

xn

-* as

y n - = 0asn->. But is not a

treated as such. Indeed, one should immediately be

real

as

n-*

if,

combining con-

+ or .
and y n

For

* as

* ,

number and cannot be

warned that something

is

To press
2
the point home, we consider the example x n = n (n = 1, 2,
.) andy = n
(n = 1,2,...). Then x n -* as n * and y n -* as n -> , but x n - .y n "* as
->. Or, again, takex = (n + 1) (n = 1, 2,
.) andy n = n (n = 1, 2,
.).
the appearance of the totally meaningless expression
.

say that a sequence (x n ) diverges to + and write x n -*


for any
0, we can find an IV such that, for any n>N,

H>

as

simply choose A'

wrong by

We

if -* +

that

have to find an A' such

vergent sequences) applies to sequences which diverge to

But there are both even and odd values of n greater than
/|< e and | 1 /| < e. These inequalities must be true for any e >
/|

we obtain

number. Prove

diverges.

obviously bounded. Suppose that

> 0, we can

a positive rational

n>N,

We
4.27

which does not satisfy x n -> + as n * or


An example is the sequence <( 1)">.

-*.

Divergent sequences

bounded sequence

<A" n >

said to oscillate.

follows if we take

n
4.26

x* asn

--

!,

(n>iV,).

:i< r.

b I-|/|<I*-/|<i

M<|l| +

i.e.

This time

jr n

y n -*

as

-* .

40

Convergent sequences

4.29

Exercise

(1)

Suppose that x n
(i) \x
(ii) \x n *

1\

(2)

->

|/|

-* I as n -> . Prove that

as

as

SUBSEQUENCES

-*

-*>.

Suppose that x n -* /
that, for any n >N,

as

-+ . If /

> 0, prove

that there exists an

N such

xn >\l.
(3) Prove that
(i)

2"

-*

+ as n -*

(ii)

V"

""*

~~ as

"

"*

Subsequences

5.1

(iii)<(-l)"n> oscillates.
(4)

Let

jc

>0 ( =

2,

.).

Prove that x -

as

* if and only if

Suppose that < x n )

is a

sequence and that

<

nr

>

is

a strictly increasing

sequence of natural numbers. Then the sequence

l/x n -*-as/! ^.
(5) Suppose that {x n ) increases and is unbounded above. Prove that
x n -* + as n * . If <x n > decreases and is unbounded below, prove

x n -> as * .
Let 5 be a non-empty set of real numbers and suppose that d{%, S) =
(see exercise 2.13(4)). Show that, for each n N, we can find an
x n GS such that | x n < l/. Deduce that n -> as -> .
If S is bounded above, show that a sequence of points of S can be
found which converges to its supremum. If S is unbounded above, show
that a sequence of points of S can be found which diverges to + .
that

(6)

is

called a

The

subsequence of < x n ).

few terms of some subsequences of < x n >

first

\X r+i ) = ^2,^3,^4,

are listed below.

i-

'*3r

'

'*3r+S'

(X^r)

= x 2> x 4> x 6>


= x S> x ll> x 14i x n>
=

X2, Xn,

-Xg,

Xi6,

if we think of a sequence as a list of its terms, then we


by crossing out some of the terms. For example, < x 2n )
may be obtained from <x n >by crossing out terms as below.

Roughly speaking,

obtain a subsequence

Al>

The

x 2> A3. 4>/S' *6>


JC

insistence that

<

n r > is strictly increasing is of some importance.


whose first few terms are

It

means,

for example, that the sequence


-*3>

is

X \, X S, X \> X <)>

not a subsequence of < x n ).

The

fact that

nr
This

is

easily

5.2
of (x n

>r

).

(r

is

strictly increasing has the

1,2,

..

consequence that

.).

proved by induction.

Theorem Suppose
Then

xn

that, for

(n r )

-*

that

xn

-/ as n-* and that

<x r > is

subsequence

as r * .

Proof Let e
any n >N,

>

be given. Since

jc

as n -*

<*>,

there exists an iVsuch

41

'

42

*,-/!<.

any e

> 0, we can find an R

(namely

R = N) such

= Kx n + OX?)

Obviously

x n >0(n =

1,2,

\xF -l\<e.

Hence x

We know
Thus

Example We have seen

5. J

A simpler proof can

terms tends to
converged,

-* .

We

be based on theorem

example 4.4 we showed

In

prove this again by

>

1 ,

x lln

then

bounded below by

"/*

and therefore (x'


verges to a limit

By theorem

1.

But,

if

<( 1)"

Since

>

in

-*

(n

> 0, then x vn *

We
1

>

1 ,

2,

.)

and hence the sequence < x lln )

From theorem 4.17

What

we know

that

We

is

it

follows that

it

<*""> con-

we omit

subsequences of (x

11 ")

must

also tend to

its first

x n -* I

/.

>

\Ja.

By theorem

term).

<jc> decreases

and

From theorem

4.1 7

is
it

bounded below by

Va (provided

follows that

as

-*

What
5.2

the value of/?

is

we

also have

x n+1

-*

as

-* .

But

follows that

limit.

Thus

(because

>

\Jd).

xn +i = i(* n +

or

^n

Hence, by proposition 4.8,


1

)- > K'

+ '" )asn^ 00
,

Since a sequence can have at most one limit,


/

But

>

and hence

have shown that


1

i.e.

as n -*<

A warning is appropriate here. Do

you thoughtlessly

looking at

2,3,...).

the value of /?

all

value of a limit until


if

(n

Aln -*
Note:

1,2,...).

x > decreases and so we consider

<

>

(n

1,2,...)

Now ^

and so

follows that

it

is

where

But a sequence can only have one

real solution.

* n -*n+i = x-i(x+ ax- 1 )

as n -* .

Using proposition 4.8,

> 0,

Hence the sequence

and

0.

>a.

wish to prove that

as

example 4.4 we need

_
= X (n+l-n)/n(n+l) _
= x 1/n(n+1) >l

decreases.

jc n+1

But further

i.e.

x n+l >s/a

Hence
,1/2"

+l

1,2,...).

Moreover

,n

5.2,

.).

x>l.

>

l/(n+l)

I/n;

limit.

that, if x

method. As

a different

'b

diverges.

Obviously the subsequence of odd

and the subsequence of even terms to

only consider the case


If x

5.2.

subsequences would tend to the same

all its

Example

5.4

bounded sequence (( 1)">

that the

in

advance that this quadratic equation in x n has a

in

- 4ac > 0'

+a =

x n2 -2x Ml x n

* .

as /

Define the rest of the sequence <x n )

>0.

jc,

inductively

* +1

We have shown that, given


that, for any r>R,

>0 and let

Example Let a
by

5.5

R=N. Then, for any r>R,n r >r> R = N and so w r > N. Thus

Lei

43

Subsequences

Subsequences

its

not use this method of working out the

you have shown

that the sequence converges.

tried to evaluate

the limit of the sequence <(-- 1)"> by

subsequences,

you would obtain

=+

1.

For example,

= kQ + al"
2
= a.

yJaoxl =
therefore shown that

Thus

x n -*

\Ja.

\Ja as n -*

But we

know

that

>

\/a.

Hence

\/a.

We have

44

Subsequences

5.6

Example We attack the previous example by a different method. Again


but this time let Xi be unspecified for the moment. Define the rest of

let

>

the sequence inductively

by

x n +i = ttxn+ax-

Now

45

Subsequences

(n

argument given here is interesting because it allows us to estimate how 'good' an


approximation x n is to \/a.
For example, suppose we want to estimate the value of y/2. It is obvious that
< sj2 < 2 (because l 2 = 1< 2 and 2 2 = 4 > 2). For a better estimate take a =

1,2,...).

2 and*!

2 in the argument above. Then

x n+l ~\/a = \(x n + ax~ l )-s/a


2

2x,

(x n

|x 4

-2x n y/a + a)

-V2|2V2
2V2

'

2x n

2jc n _,

2x n (2* n _,)

{x

with Xj

5.

P5<*i-vr.

(n=

1,2,

.).

it

xn

follows, as in example 5.5, that

(2)

> y/a

Given that
-> 1

|<2

_6

p< 0-016.

xs

or

by

starting

(see exercise 4.20(6)),

'

(3)

If

>

and

xx

lies

'

between the roots a and b of the equation

<xm 1 <Xn < b {n =

1,

verges and determine

(*i-vfc)*

that

"

Prove that a
(2"-l)/(2-l)

show

by considering the subsequence ((2m) "}.


A sequence < x n > is defined by x x = h and

where

c^-vr

2%Ja

most 0016.

asH-* 00

< k < \ and h


x 2_ x+k=Q

_1
l + 2+2 +...+2"
2

<

n u " > converges

<

x n*l = x n

Hence

l* n +i-\/l

at

> \/a, then

from y/2 by

\/2

<

/i

2x n (2x n 1 ) 2 ---(2x y
that Xj

obtain |x4

estimate for \J2 may be obtained by evaluating


.)
\ and using the estimate \J2

l/n

we assume

differs

Exercise

7
(1)

If

1)

2V2

(A better

M - Va)

V2(V2 -

= 2\/2

(V2-1)

\J2< 2 we

Using the crude estimate

Hence x 4 = gg = 414

(2-V2
2V2

>

its

and

2,

.).

Show that (x n )

con-

limit.
(

xn >

defined inductively by

is

2y/a

i~ Va
2Va

= 2Va
We know
thaty

that, if

<

theny"

1 ,

->

as

that

xn

show

that

By theorem

5.2

it

follows

\Jaa.sn^> provided that

Two
y-i

2\/

xx

>

one of the sequences

<,x 2n

>

and (x 2n -i

> is

increasing and

concerning <*>?
(4)

<1.
we have shown that x n -* V<z as
-* provided that \]a <x y < 3 \Ja. Of course we already know from example
5.5 that this was true under the weaker hypothesis that x > 0. However, the
already assumed that

'

Prove that both sequences converge to the limit / which is the posi2
tive root of the equation x + x = A:. What conclusion may be drawn

'

We have

l+xn

the other decreasing.


.

-*0asn-*.
Hence our argument shows

*n+1

\Ja and so

sequences

=
=

<

x n > and (y >

are defined inductively by

and
VCXn-lJ'n-l)

2,3,...)

x =
l

and

46

Subsequences

J.

(n

We

2,3,...)-

<x <^ n <>>_,

sequences converge to the same limit

/,

2, 3,

where

.)

be any

<(l+yn-

real

< <
/

(Actually

number. Prove that the sequence

is increasing for those values of n which satisfy n > 1 y. [Hint:


example 4.19.] Show also that the sequence is bounded above and
hence converges. [Hint: From exercise 4.20(4), (2y)"/n\ ->0 as n ->

'"

limits

of the two sequences <(1

+ xn' 1 )")

Theorem {Bolzano-Weierstrass theorem). Every bounded sequence of


real numbers has a convergent subsequence.
Proof Let < x n > be a bounded sequence. By theorem 5.9, (x n > has a
monotone subsequence <*,.>. Since (x n ) is bounded, so is <x >. Hence, by

5.10

Examples

5.11
(i)

The sequence <( 1)" >

sequence
even terms.

1.

(ii)

i-$

V~7i

A more

bounded.

is

... of

sophisticated example

Bolzano-Weierstrass theorem

5.9

Theorem Every sequence has

Every

set

{x n n
:

> N}

has a

maximum.

In this case

we can

3.

5.12*

section.

= maxx n

the

Sj

rn

>

whose

first

of

few terms

is

a rational

number between

and

Hence the

has many convergent subsequences. For example

shall

However,

lim inf

not have a great deal of use for the subject matter of


it is material which has to be learned eventually and

natural to deal with

n>n,

it

this
it is

here.

Suppose that <jc> is a bounded sequence and let L denote the set of all real
numbers which are the limit of some subsequence of < x n >. We know from the
Bolzano-Weierstrass theorem that L is not empty (i.e. L # 0). Of course, if

xn

and so on. Obviously n t

It

Lim sup and

We

n>l

= max

uu...

x n = maxjc n

x 3

the sequence

find a sequence

of natural numbers such that

*,,

bounded.

ill!
4>
2>

cases.

is

5, 4> 6> S.

monotone subsequence.

Proof Let (xn > be any sequence of real numbers. We must construct a
subsequence {x ) which is either increasing or decreasing. We distinguish two
r
(i)

is

sequence

convergent subsequences are the

are

Every term of the sequence


5.8

Two

odd terms and the sequence

2' 3> 1' 4> 4i $> S< ?i 5> 5> Sj

(n r

first

n 2 \l/n-i
1

We

so on.

x 2 be the

let

be the

theorem 4.17, <x r > converges.

N such that, for any n>N,y"/nl< ()".]

and therefore we can find an


(6) Show that the product of the
and <(1 -xn'')") is equal to

'

first term following *, for


term following x 2 for which
obtain an increasing subsequence <* r > of (x n ).

and then

Now let x nj

And

'.

and deduce that both

= xNi+1

x ni

>x n

>x n

xn

ff/4.)

(5) Let

define

which x n

y n -i)

Prove that x_,

47

Subsequences

<n 2 <n 3 <

maximum of a smaller set

decreasing subsequence of < xn

and

we are taking
Hence < x n > is a

so, at each stage,

than at the previous stage.


>

Suppose that it is not true that all of the sets {x n n > N} have a maximum. Then, for some jV, the set {x n :n>N t } has no maximum. It follows
that, given any x
can find an x n following x such that
m with
lt
(ii)

(x) converges,
5.13^

will just consist

Proposition Let

<

x n > be

numbers which are the


maximum and a minimum.
all real

of

limit

a single

point (theorem 5.2).

bounded sequence and let L be the set of


of some subsequence of < x n >. Then L has

xn

>x m

m>N w

(Otherwise the biggest of x N

{*:>AM).

+1

,xm would be

m
maximum

for

That

is

a bounded set follows from theorem 4.23. Hence

certainly has a

supremum and infimum. Proposition 5.13 asserts that these actually belong to
the set L. The proof is relegated to the appendix. Readers who prefer to omit

48

Subsequences

the proof altogether will encounter

it

again at a later stage since

it is

and show that these are not the same

a special

case of a general theorem concerning 'closed sets'.

bounded sequence, let us denote the maximum of L by 7 and the


minimum by /. Then / is the largest number to which a subsequence of <x>
converges and / is the smallest number to which a subsequence of < x > conn
(x n >

If

is

(-1 )"(! +

sup

(2)

Let

<

rn

|,4 ..

verges.

We

call /

the limit superior of

x n > and

write
t

lim sup

xn

(3)

/ is

the limit inferior of <x

and

>

and we write

(4)

lim inf x.

inf

be the sequence whose

>
.

_1

Show

(-

first

as the

numbers

1)"(1

-).

hhk, $t*h 5* I*

few terms ate ,

that every point in the interval [0, 1]

the limit of a

is

subsequence of <r>. [Hint: see exercise 3.6(4).]


Suppose that (x n ) is a bounded sequence and that, for any N, we can
b. Show that < x n > has a subsequence
such that x n
find an n >

>

n*oo

Similarly,

49

Subsequences

which converges to a limit / > b.


Let < x ) be a bounded sequence with limit superior / and limit inferior
that, for any
/. Show that, given any e > 0, we can find an N such
=
+
n>N,x n <l + e. [Hint use question 3 with b / e] . What is the
:

(Occasionally

you

means that < x n


(x n

) is

such expressions as lim sup x n

will see

> is

unbounded above.

=+

Similarly lim inf x n

corresponding result for /?


<*>.

This simply

t (5)

= - means that

unbounded below.)

Deduce from question 4 that


limit / if and only if

lim sup x n

Hence show
5.14 *

Examples

its

The sequence

(i)

<

1/n

sequences converge to

lim sup

0.

converges with limit 0.

>

Hence L contains the

lim inf

*
5.2,

all its

sub-

(6)

and

single point

lim

By theorem

that a

The sequence <( 1)">

(ii)

(i.e.Z,

= {+

lim sup

1,

1}).

Let

<

x > be

Show

(-1)" e

1;

is

bounded. The

set

consists

of two points

bounded sequence and

sup x k

lim inf

that

<

Mn

decreases and

lim sup*,,

be the sequence whose first few terms are \,


J, , ,
Obviously <r n < 1 (n = 1, 2,
I, \,
.). From theorem 4.23 it follows that
L is a subset of [0, 1 ] (i.e. L C [0, 1 J). But the subsequence 1,1,1,... con\,
.

We

Thus

let

bounded below. Deduce that (Mn )


If / is the limit of some subse_
> I (n = 1 2, .). Deduce that > I.
is

by M.

M<

[Hint: use question 4] and

lim

{sup x k }.

= 0.

(This explains the choice of notation for the limit superior.

Also the subsequence

, \, |, f,

converges to

Thus
lim inf

lim inf rn

x n = lim

{ inf

x k } .)

=
5.16

Cauchy sequences

Suppose that we want to prove that a sequence < x n


have no idea in advance what its limit / might be. Then there

Exercise

*(!) Calculate

appealing directly to the definition because


(i)lim sup

{(-!)"(!+

-)}

We also

have, of course,

have shown that

lim sup rn

5.15*

if all

M,U,

verges to 0.
/

and only

hence show that

(hi) Let < rn >


.

its limit

Obtain also the reverse inequality

(-1)" = -1.

if

quence of < x n >, show that

We have

I.

bounded sequence < x n > converges

converges and denote

and-1

xn =

convergent subsequences have the same limit.

,Mn =

0.

lim inf

bounded sequence < x n ) converges with

(ii)

lim inf {(-1)"(1

+ "')}

prove that \x
large if

can be made as small as

we do not know

the value of /.

>

converges but

is

no point

we

in

we shall certainly not be able to


we choose by taking n sufficiently

50

Subsequences

Theorem 4.17

gives us

one way of resolving

increasing sequence which

sequence which

this difficulty. It tells us that

an

Example

5.20

bounded above converges and that a decreasing


bounded below converges. Thus we are able to deduce the

is

is

which may not be monotone, we introduce the

We say that < x n )

Xn

\xm

is a

Cauchy sequence
any n>N,

if,

given any e

can find an

\<e.

Any convergent sequence

Proposition

and

Hence,
is

Any Cauchy sequence

Proposition

is

-x

x n+\\

if

Cauchy sequence.

are easy.

is

X "-l
"-'

\x n

ri

i=i+=i+.-- + 5^:

'4 + ? + - + 5^'

Proof Let (x n ) be a Cauchy sequence. By proposition 5.18, (x > is


n
bounded. Hence, by the Bolzano-Weierstrass theorem, it has a convergent subsequence (x nr >. Suppose that x -* / as r -* . We shall show that x -* 1 as
n

Let e

<

> 0. Then

xn >

and any

> 0.

\e

-/Ke.

is

Hence there

exists an

0)

Cauchy sequence, there

also exists an

n>N,

Now suppose that n >

is

satisfied

r so large that

with

m =n

r.

Given any

Thus x n

-> / as

xn

^e

> 0, we

x +
|

\x

+ ie=

be
1

,N-2

Then,

Lft

any n

for

N so large that

l<.

> N and any m > N.

(triangle inequality)

> is

they have the same


A''

Choose

Cauchy sequence. Therefore, by theorem 5.19, it converges.


(An alternative method would be to show that one of the two sequences
the other decreasing and then to show that
<*2n-i> and <*2*> is increasing and
Thus < x n

given.

rt",-,Kil-,l.
-

> N,

e.

have found an

>

n r >Nandr>R. Then

Thus, for any n

\xn -l\ = l**-* + ;c nr -/|


r

<
<

Let e

N such that, for any m>N


(2)

and choose

and also (2)

satisfied

m2

R such that, for any r >R,

\Xm-Xn \<y.
is

|A
X"
'

Theorem Every Cauchy sequence converges.

5.79

(1)

^,2
2

what makes Cauchy sequences

important.

Since

~ x n+l)-

-x n ^+x n _,-x_ 2 -.. .+x m+i -x


< X n -X_ + X n _, -X n -2 + + x m+l ~ x n

=
\XmX\
mi

bounded.

The next theorem

h( x n

>m,

The proofs

and

=
l

closer' together.

*tl

5.18

a l*

\x
'closer

= a, x 2 = b

1,2,3,...).

= iOn+l +X n) ~ x n-H =

X n *2~ x n+\

("

by x,

Thus

\X+2~ x n*l

Very roughly, the terms of a Cauchy sequence get


5.17

> 0, we

defined

Proof We have

idea of a

m >iVand

is

Prove that <x n > converges.

Cauchy sequence.
such that, for any

A sequence <x n )

x n + 2 = |(Xn+l + *n)

convergence of the sequence without necessarily knowing the value of its limit.
But, of course, theorem 4.17 deals only with monotone sequences. In order to
deal with sequences

51

Subsequences

such that, for any n

> N, \x

limit.)

< e.

-> .

Exercise

5.21
(1 )

<a<
~
a
x n+i x n < " (" ~

Suppose that
I

and hence converges.

and

1
1

2-

that

<

xn >

is

a sequence

Rmw that

<

x" )

is

which
a

satisfies

Cauch y sequence

52

Subsequences
Give an example of a sequence < y

such that y n

>

* + oo as n _ oe

lj' n+1 ->' n |-*0asM-*oo. [Hint: see exercise 4.20].


sequence <x> satisfies 0<a <x, <x 2 < b and

but
(2)

* n+2 = {*n + i*}" 2


Prove that a

***!

<x n < b

(
(it

=
1

2,

.).

-*!<

-*n

SERIES

1,2,...).

-**3

Hence or otherwise show that

2,3,...).

Deduce
(3)

that <*> is a Cauchy sequence and hence converges.


For the sequence of example 5.20, prove that

x n+i + hx n = x n + \x n _ =
i

Deduce

that

clusion

may

...

= x2 +

\x {

Given a sequence <c n > of

of

hf

is

called the

[a, b].

S of real

numbers,

5j

= {x:xGSandx=t$.

We

say that |

point') of

is

if

element of S).

N)

defined by

It is

an

fl l

fl 2

We

+ "N
series

an

is

said to converge to the

sum

write

important to remember that

formula can only make sense when the

this

series converges.

of accumulation or

'limit

at zero distance

from S. (Note that % need not be an


form of the Bolzano- Weierstrass theorem asserts that

every bounded set with an infinite

<s

n=

let

a cluster point (or point


is

an

4.29(6).]
(6) Given a set

* s as N - , the series

If s N

by assum4.29(6). Then prove that sup /

and appealing to exercise


and inf /are both elements of /with another appeal to exercise
T

numbers the sequence

sequence of partial sums of the

n=l

(5) Let /be an interval which has the property that every sequence of
points of /contains a subsequence which converges to
a point of/.
Prove that /is compact. [Hint: First prove that
/is bounded
ing otherwise

real

n+i -l\
l\xn -l\, where / = |(jc 2 + *i> What conbe drawn about the convergence of the sequence
<*>?
Tackle exercise 5.21(2) by a similar method.
(4) Let [a, b] be a compact interval (see 2.9). Prove that every sequence
of points of [a, b] contains a subsequence which converges
to a Doint
|

Definitions

6.1

number of elements

has at least

Example Consider

6.2

the series

one

cluster point. Prove this.

n=o
If x J=

sums

the partial

Sfl

xn =

satisfy

+ x + x2 +

+ xN

n=
.\'+

-x
-x

If

|x|<

l,.r

N+1 -*0

as

A'-* and hence

53

54

Series

-*

It

N -*

as

follows that the series

>

Proof Since the partial sums of this


show that they are unbounded above. But

2~=0 x" converges if

|jc|

<

and we

may

write

s 2N

(W<1).

l-x

Example The

partial

sums of the

n(n+

iN-l

series

1)

Thus the

partial

sums

unbounded above and

are

n(n

n =1

-l|i-n
n=

1)

\n

GO

N+

-->

asAf-* 00

rational

>

Then

converges.

they are

Proof Since the partial sums


bounded above. For TV > 1,

are increasing,

we need

only

and we may write

1.

1)

T(JV-l)a

1
.

.+

N
(2 -lfj

Series of positive terms


Series

whose terms

are

all

positive (or non-negative) are particularly

<

;<*

easy to deal with. This

because the sequence of partial sums of such a series is


wish to show that a series of positive terms converges, we

increasing. Thus, if

we

only need to

that

show

\4

4"

4"

4a !

is

its

sequence of partial sums

is

sequence of partial sums

is

bounded above.

unbounded above, then the

series diverges to

2<fi-i)u

If the

+ ...+ 2(N-l)Of
9JV-I

2
a

(see exercise 4.29(5)).

2
6.5

Theorem The

N-l

series

=
n=

diverges to

the series

\n(n

that

the series converges

n=

number such

N N+

Theorem Let a be a

4- 1

4WBWH+

6.4

the theorem follows.

by
6.6

Hence

I
n=
are given

the series diverges.

CO

to

I+^ + t+.-. + ^v

4
6.3

we only need

series increase,

!>" =
If \x

55

Series

+ -l + ^or2
al

.+

- 1\JV

l-(l/2a

)^l-(l/2 a

show

that

56

Series

provided that l/2

Hence

<s^>

a_1

<

But

>

the theorem follows.

Elementary properties of

6.7

from the assumption that a

this follows

bounded above and

is

tend to zero, but the series diverges.

do not tend

and

respectively.

+ pb n )

2 = (ka n
i

number

numbers, the

real

if

fl

Y,

n=N

tends to zero.

is

often referred to by saying that the

The proof is

of a convergent

series

+ p@ asiV-x*

6.12

Series

and Cauchy sequences

we have

studied so far,

we have

either been blessed with a

sums (examples 6.2 and 6.3) or else the partial sums


increased and so we only had to consider whether or not the partial sums were
bounded above (theorems 6.5 and 6.6). What should we do in the absence of
such favourable conditions? We ask the question: is the sequence of partial sums
nice formula for the partial

Theorem Suppose
an

'tail'

easy.

by proposition 4.8.
that the series

2 = a n converges. Then
i

as * oo.

->

Then, for each

n -*0as7v'--.

In the series

6.9

= /v

S"=1 a converges.

a n converges and

la
n +n >
=
n=

,.

-+\a

the series

CO

series

This result

IMA) =

TV,

series

Proof We have

n=

One might say that the terms of the


make the series converge.

2 = a and S"=1 b n converge to a

that the series

X and p. are any


converges to A a + ju0.
Then,

to zero 'fast enough' to

Proposition Suppose that the series

6.11

series

natural

Theorem Suppose

6.8

57

Series

Proof Let the sum of the

series

be

x.

Then

Cauchy sequence*! This question has some

fruitful

answers as

we

shall see be-

low.

sn

a--sas?j->.

6.13

n=l

Also

x.v _

-* s as w -*

Theorem Suppose that <a> is a decreasing sequence of positive numas n -* . Then the series

bers such that a ->


<*>.

oo

But then

flw

(fli

+a 2 +

%) -

= s N s N - -*s s =
l

6.70

Examples The

(fli

a2

OasTV-*- 00

+ %-j)

converges.

Proof We show that the sequence <s> of partial sums of the series is a
Cauchy sequence. From theorem 5.19, it then follows that the series converges.
The proof depends on the fact that, for each n>m,we have the inequality

0<a m + A m + 2 +a m +

series

This follows easily from the fact thata*,

X(-D"
n=

<(,>

diverges. This can

be deduced from theorem 6.9 by observing that

its

terms do

.<2<fl m

aj, +

always non-negative because

is

decreases.

as n
Let e > be given. Since a n -*
n>N,an <e. But for any n>m>N,

-*

we can

find an

N such

that, for

not tend to zero.

Note
series

that the converse of

tend to zero

it

provides an example.

theorem 6.9

is

false. Just

\$n~ *ml = lfai-2

because the terms of a

does not follow that the series converges. Theorem 6.5

The terms of the

series

la

.a n )-(a,a 2

m + a m+2 +a m+3
i

< m + i<e
fl

OS

+a 3 -.

(because

.a\

m >N).

I1
n=i n

Thus

<s> is a

Cauchy sequence and

the

theorem follows.

.a m )l

any

58

Series

6.14

Example We saw

it

in

theorem 6.5 that the

series

=1

l/ diverges. But

follows from theorem 6.13 that the series

converges.

59

Series

Of course, theorem

no

6.13 yields

that, if

any positive rational number and

is

n=

clue as to

what the sum

is.

(It is, in

a+2x" -+0zsn-*.

exercise 4.20(5) we know that n


a+2 n
the sequence (n
x ) is bounded (theorem 4.25).

From

ft-oo/

Hence
some H,
hypotheses are rather

its

restrictive.

\nV\<H/n 2

tive real

Theorem (comparison
numbers.
\a n

Let

a convergent series of posi-

2 = a n
i

bk

ly

how

6.17

fln

lim

Z"=

a n be

+ ai + ..+a n )-(ai + a 2 +

\(ai

<

<s>.

Then,

if

If

>

|a

m + il +

< I

+ m )l

If

fl

|fl

m +2 +
|

4- \a\

/<

(triangle inequality)

*fe<e.

1,

we may
otN,

=N,N+

test is

1,

\a\
.

.).

<

The

series

2"=

i (/

a series

n
.

If

b n (n

lim sup
1

2,

.)

(Why?) Note further

which

satisfies

the series converges.

l.Then, for

N -n\<Q + ey-

N-

a sufficiently

\a

N ^\.

"n-2

then converges by comparison with the geometric series

>

a similar

argument shows that the terms of

2=

a n do

series diverges.

i,n

\a n

I.

can be replaced by
that the hypothesis of the

If /

>

the series diverges and, if /

< 1, the series converges.

not

<
is

\a

S"=1 a n
e)

<

jV + 2

<

we may

take e

>

I**.

satisfied

1)" diverges.

by a k

so small that

<

Then, for a sufficiently

large value of TV,

fe=i

2"=1 (-

Proposition (nth root test) Let ~L" =i a n be a series which satisfies

6.18

follows.

If /

This condition

S = a be

e>0 so small that /+ e<

\a\

take

not tend to zero and so the

Cauchy sequence and the theorem

Note that the hypothesis

\^Hb (n

brief-

I.

the series diverges and, if /

"n-l

\a n

sometimes helpful. We indicate only very

a*

large value

comparison

are

Proposition (ratio test) Let

<e.

! %l

is

.).

they are proved.

Let the sequence of partial sums of the series

(s)

to zero (prop-

n>m>N,

Thus

converges (theorem 6.6), the result follows from

The next two propositions

converges.

Proof Let e > 0. Since 2~=i b converges, its tail tends


Hence we can find an A' such that, for any n >N,

1,2,

follows that, for

the comparison test.

1,2,...)

osition 6.1 1).

ft=n +

l/

It

If

\<b

then the series

test)

2=1 b n be

(m

S~ =1

Since the series

6.15

< 1,

converges.

fact, logg2.)

Theorem 6.13 is sometimes useful, but


more useful theorem is the following.

\x\

then the series

Example Prove

6.16

= (

1)''

and &

ft

l/k

2
,

but we

know

\a n

<(l+ ef

(n

>N)

(exercise 5.15(4))

that

and the convergence of 2"=i a n follows from the comparison

test. If /

> 1, e

is

60

chosen so that
'

e>

and so the terms of

Note that

> Q ~ e)"

"fe'

if

Then, for some subsequence

The

(a nk \

first

6.5).
fe

k "*

"* as

S"=1 a do not tend

then the

/),

We

of these

however the

+
is

conditionally convergent.

results

or divergence of the series at

(ii)

On

the other

hand the

=l

Example Prove

is

and hence the

1)!

/ n\

as

test. Alternatively,

n=l n

It

series converges

by the

should be noted that the comparison

test all

ratio test.

test, the ratio test

and the nth root

demonstrate absolute convergence. The only criterion we have given

is

establish the convergence

of a

series

which

is

only conditionally con-

theorem 6.13

Absolute and conditional convergence

A
converges.

series

2~=1 a

series

is said to converge absolutely if the series 2"=i \a n


which converges but does not converge absolutely is said to
\

be conditionally convergent.

Theorem Every absolutely convergent

Let an

Series are 'infinite sums'.

\a n

in

the comparison test.

'

(n

=1,2,...). Then

may be

It

would therefore be optimistic

like 'finite

expected

freely manipulated. If

one

tries to

the following argument.

..

= (1-1) + (1-1) + (1-1)+...


= 1-1 + 1-1 + 1-1 +
-i

n =,

^1

111

+(-1 +

+0 + + 0+

1.

l)

+ (...

..

l)

+ (-l +

obtain results by

series,

in general.

Example Consider

to expect to

sums'. Indeed, only absolutely con-

manipulating divergent or conditionally convergent

6.24

Manipulations with series

vergent series
series is convergent.

Examples

(i)

6.23

be able to manipulate them just

Proof Simply take b n =

6.22

I k

n -*

which can

6.21

which converges.

if* = 0,

vergent

6.20

fl

values of x.

Proof We could use the comparison

(n

(-1)"

absolutely convergent because

n=\
all

n=

that the series

T converges for

series

all.

I*n=I
n
6.19

converges (example 6.14). The second diverges (theorem

(-1)n-l

series diverge. If /

no information about the convergence

series

conclude that the series

to zero.

the expressions of propositions 6.17 and 6.18 diverge to

(instead of converging to
yield

61

Series

Series

1)

only nonsense can be

62

Series

The error

is

63

Series
find. The series 11 +
do not tend to zero.)

not hard to

divergent. (Its terms

+ 11

...

is
/

where

Example We know

6.25

6^7

converges conditionally. Denote

sum by

its

We

s.

0, prove that the series

a and

both converge or both diverge.

either

Discuss the convergence or divergence of the series

_1_i4.1_1_i4.1_1_i4.l_
4^3
8^5 T5 15^7
2

/ <

rearrange the order in which

the terms of the series appear and obtain


1i

-*

that the series

4^5

2^3

as

bn

sum of this

Let the Mth partial

tn .

Then

be

series

In -

An

_1_

-2

4m

(3) Suppose that <a>

_ = !

-1

2m
1

+-+

(4) Let

</>

1/11
- + -+

2 \2

+
2m

first

few terms are

-- +

(5) Determine whether or not the following series converge or diverge.

+.

4,4,4,... Prove that the

diverges.

\, 3, ,

(0, 1) arranged in the

_/

-1

In

Hint: consider

series

2/1-1

= -

+ &]

..

denote the rational numbers from

sequence whose

%*i+*+a +

-2

as n - .

An

4m

\4

- - + -+.

-|I + i+...

a decreasing sequence of positive terms such that

is

a n converges. Prove that na -*

..

2m

ft

_?

2m

Cm'

(iii)

n = i(2n)!

M +

\s as n

Notice that the finite


Since

t 3n

sum

3n *

can validly be rearranged in any

for

t 3n

-*

as

the rearranged series converges to

way we

n=

(6)

3n

m(m

If the

~2

If <a>

(v)

_.
n=l

\j

-{V(+0-V}
M

_ ir -.
v

sum of the

conditionally convergent series

1)(m

-1,

iss, prove that the

and <b n )

are

sum of the

rearranged series

1+1-1 + 1 + 1-J + J + X-1 +

...

2)
is

(2)

I-i + 1-i + l-...

(1) Using partial fractions, prove that

Mf

\n

n=l

Exercise

4"
l

and r 3n+2 3n ->0 aSM -*,it follows that


\s which is not the same as the sum s of our

original series (because s > 0).

6.26

(iv)

like.

two sequences of positive terms and

Is.

Hint:r 3n

= 1+

_
J

+ '"

4n

4m

2m

65

Functions
If

/ is

defined on a set S,

we

use the notation

AS) = {f(x):xeS}

FUNCTIONS

and say that f(S) is the image of the set S under the function /.
The set f(A) is called the range off. Note that f(A) need not be the whole
offi.

Example Consider the equation y

7.2

Notation

7.1

A function f from

a set

to a set fl (write /:

A-*B)

defines a rule

which assigns to each x A a unique element y G.B. The element y is called the
image of the element x and we writer = f(x).
When A and B are sets of real numbers we can draw the graph of the function
as in the

to itself.

y x2

jR there exists a

For each

Observe that,

one and only one

in

= x2

unique

This defines a function from

yU which satisfies the rule

the diagram below, each vertical line cuts the graph in

place.

diagram below. The defining property of a function ensures that each

vertical line

drawn through

point of

cuts the graph in one and only one

place.

/(-*)

The domain of this function is R. The range


1] is [0, 4]. (Why?)

is

[0, ).

The image,

for

example,

of the set [-2,

7.3
tion

Example Consider
from

to

itself. In

the equation

/ is

a function

from

to

B and S C A we
,

say that

/ is

defined on the set

ated with

x. This does not define a func-

drawn through the


no value of y associ-

the diagram below, the vertical line

point x does not meet the graph


If

2
y =

at

all,

and hence there

is

The largest set on which / is defined is, of course, the set A We call A the
domain of/. For example, a sequence is a function whose domain is the set M
S.

of natural numbers.

/f

4w&$

)
7.4
[0, ) to

64

Example Again consider y 2

R? Again

the answer

is

= x.

Does

no. This time

this define a function

it is

from

certainly true that every

66

vertical line

drawn through

but one of these vertical


gram, there

is

67

Functions

Functions
a point of [0, ) meets the graph (see

lines

meets the graph

in

two

.9).

But

all

points. Thus, in the dia-

this function a

A polynomial

polynomial of degree n.

of degree

is

called a

constant.

not a unique value of y associated with Xj.

Suppose that
all

the values

P and Q

of x

from exercise

3.1

polynomial functions. Let S denote the set R with


(If Q is of degree m, it follows

are

which Q(x) = removed.


1(3) that there can be at most
for

m such values.) Then the equa-

tion

y =

P(x)

<m

defines a function from

7.

U. Such

to

Example The function from

2
y =x 3x + 2x
3

is

Example Again consider y

7.5

x. This does define a function

function

is

called a rational function.

to itself defined

by the equation

called a polynomial function of degree 3 (or, more loosely,

a 'cubic polynomial')- Its graph


2

is

sketched below.

from

[0,~)to[0,).

= v'-ix 1 + 2.Y

10.)

7.8

Example

tion

Since

y must be

in [0, )

we omit from

the diagram the part of our previous

Then every vertical line drawn through a


one and only one point (see 1 .9). Thus,
given any x [0, ), there is a unique y G [0, ) which satisfies 2 = x. Thus a
y
function /is defined from [0, ) to [0, ). Recalling the content of 1.9, we
graphs which

lies

below the

xz

+4

-4

*-axis.

point of [0, ) meets the graph

observe that, for each

y =

Let

defines a function

S be

the set

with 2 and

-2

removed. Then the equa-

(x*2)

from S

to

R.

Its

graph

is

sketched below.

in

> 0,
!
.v

7.6

Polynomial and rational functions


If

a ,a u a 2 ,

y =

hand

a n are

from R

side generates a

real

all

+ a x + a2 x 2 +

defines a function
right

numbers, then the equation

a n x"

Any value of x which is substituted on the


unique corresponding value of y. If a n # 0, we call

to itself.

+4

68

Functions

7.9

Combining functions

f A

-*

We

begin with some almost obvious notation.

two functions from S


from S

tion

to

which

which

any

is

then

we

If S C

define the function

IR

/+ g to

and

/ and g are

be that func-

we

which has

unique

property

this

is

said to

In geometric terms, a function

f(x)+g(x)

real

an inverse function /"' B -> A if and only if each b B is the


a^A. (Otherwise /"' could not be a function). A function

B has

image of

satisfies

and

(xes).

number, we define

X/ to

be the function from

to IR

B (and hence has


through

for

all

f(x)lg(x)

latter definition to

/:

l
:

-*

A)

if

and only if each vertical

a function).

and f/g by

(xGS).

6= /(a)

make

sense,

of course,

it is

essential

thatg(x)

=/')

=h

x&S.

somewhat

less trivial

ation of composition. Let

and

correspondence between A and B.


1 correspondence between
is a 1

an inverse function f~

For the

(xes)

f(x).g(x)

(f/g){x)

(xes).

x/(x)

define the functions fg

(.fg)(x)

/ A ->B

only one point (which makes f'

be a

meets the graph of / in one and only one point (which makes / a
function) and each horizontal line through B meets the graph of / in one and

line

satisfies

(x/)(x)
Again,

to

(f+g)M =
Similarly, if

69

Functions

way of combining functions is to employ the oper5 and T be subsets of R and suppose that#: S-*T

T-* R Then we define the composite Junction fg:S->.

/*(*) = /(*(*))

R by

(*es).
7.

Sometimes fa g

is

Example Let

/2

called a 'function of a function'.

x-l
x+

/(*)

/: (1

(*

>

-* (0,

be defined by

1).

7.10

Example Let/:

f<*)-$ + \
and

let g:

R -* R

be defined by

(*

SR)

be defined by
(0,1)

*(*)

Then fog:

3
-

R -+ R

fogQc)

is

given by the formula

x 6 -1

f(g(x)) =
\g(x)}

X6 +

(I.-)

r
It

7.11

(0, 1).

that, given

any

Inverse functions

Suppose that A and B are sets and that / is a function from A to B.


This means that each element a S A has a unique image 6 = /(a) S 5.
We say that /"' is the inverse function to / if f" 1 is a function from B toA
which has the property that x = f~\y) if and only if y = f(x).

Not

seems clear from the diagram that

(1, ) and

all

functions have inverse functions. In fact,

it

is

clear that a function

y =
This

is

easily

Thus

it

/ is a

correspondence between

To prove this we must show


a unique x > which satisfies

has an inverse function.

satisfying

<y <

there

is

x-l
(1)

accomplished by solving (1) for*.

y(x+

1)

= x-\

We

obtain

Functions

70

xiy-1) =

-\-y

1+2
x =

(2)

1-y-

Thus,

if

71

Functions

satisfies

<y < 1, there

is

a unique x

>

which

satisfies (1)

namely, that given by (2). We have established the existence of an inverse function /"' (0, 1) -* (1 ). It is given by the formula
:

i+y
f\y) = l-y

Cve(o,i)).
It

may

mum

(!.)<

is

may not be true that, for some % G S, /() = B. If such a value of %


we say that B is the maximum of / on the set S and that this maxi-

or

does exist,

attained at the point

/attains a maximum of B at
the point on the set S.

Similar remarks apply to lower

bounds and minima.

bounded above and below on the set S, then


(O.n
It is

the set S.

instructive to observe

how

the graph of x

=f

(y)

is

related to that

of

5 if and only

y=m-

Let

upper bound

is

the

H if and only

same

f{S)

7.14

be defined on S.

m<
This

if,

l/(x)l

Bounded functions

7.13

From

if,

We

for

say that

any x

/ is bounded above on 5 by

the

some K,

for

<

Example

h.
This function
(0,

1 ]

is

is

/ is bounded on

bounded on

a set

S,

be defined by

(*>o).

and attains a minimum of

bounded above by H.
If / is bounded above on 5, then it follows from the continuum property
that it has a smallest upper bound (or supremum) on S. Suppose that
sup /(*)

unbounded above on (0,1].

{f(x)-.xes}

xBS

is

any x G

function /is both

say that

follows that a function

true that, for

Let /: (0, ) -*

is

B =

it

it

If a

K.

m-\

E S,

as saying that the set

proposition 2.3

we simply

sup f(S).

ymfflk

It is,

at the point

however, bounded below on

x =

72

Functions

7.15

Example

R*U

Let /:

bounded above on both the

be defined by f(x)

sets (0, 1)

and

x=

on

= x.

This function

Explain

is

y = x2

and in both cases its supremum


There is no satisfying < | < 1

fO, 1]

is 1 But / has no maximum on the set (0. 1 ).


for which /() = 1 On the other hand, / attains
1

73

Functions

maximum

of

why /has no

[0, ) exists

at the point

g' (y)
,

inverse function.

If,

and

= \/y

we use the formula


show that g' [0, ] ->

instead,

to define a function g: [0, ) -> [0, ),

that

(y>o).

the set [0,1].


(5)

A function g: A

-*

B is

correspondence between

and B. Prove

that

(0*

_1

*(*)

=*

(xGA)

(ii)gog- 1 (y)=y
(6)

(y&B).
To what do these formulae reduce when g is as in question 4?
Let / and g be bounded above on S. Let c be a constant. Prove
(i)

sup {/(x)

c)

=c+

iGS

'//7///////////A
^fy(ii)

(0,1)

sup {/(x)

xSS

sup f(x).

xGS

+ g(x)} <

sup f(x)

sup g(x).

xes

Give an example to show that equality need not hold in


Exercise

7.16
(1)

Draw

y =

diagram

illustrating the set

[5

if

x> I

if

Explain

why

<

this

of

all (x,

y) such

that

1.

is

a graph of a function

range of this function?

What

is

from

R to

itself.

What

is

the

the image of the set [1, 2] under this

function?
(2)

Draw

\x\

a diagram illustrating the equation

\y\

1.

[Hint: consider each quadrant separately.] Explain

the equation does not define a function from

(i)

the equation does not define a function

(ii)

(hi) the equation

(3)

does define

why

to

from [

function from [1,

itself;

1 ,

1]

1]

to

itself;

to [0, 1].

Let /: [0, 1] - [0, 1] be defined by


1

/(*)
1

-x
+x

(0<*<1)

let g: [0, 1] -* [0, 1]

and
g(x)

4x(l

-x)

Find formulae for

be defined by

(0<x<l).
fa g

and

go f and

hence show that these func-

tions are not the same.

Show

r
(4)

1
that /"' exists but that g' does not exist. Find a formula for

1
.

The formula y

= x 2 may

be employed to define a function /:

U * R

(ii).

that

1
75

Limits offunctions

Given any e

LIMITS OF FUNCTIONS

> 0, we

8.3

Limits from the

Suppose

(or converges) to a limit

f(x) -* I as

f{x)

as

* b

on an

such that

* I as

<x < a+5

*->

Suppose that /is defined on an

left

that /is defined

>

|/(;t)-/|<e
provided that a

8.1

can find a 5

We

interval (a, b).

say that/(x) tends

% G (a, b). We

point

interval (a, b)

except possibly for some

say that fix) tends (or converges) to a limit

/ as

x tends

to %

and write

tends to b from the left and write

f(pc)

-*lasx-+%

or, alternatively,
or, alternatively,

lim

x -* b-

f(x)

=1
if

if

the following criterion

Given any e

the following criterion

> 0, we

can find

i1

>

Given any e
-

-8<x <b.

> 0, we

\f(x)

1\

is

the distance

as the error in approximating to

between /(x) and

by f(x). The

/.

We

as

>

such that

< \x -IKS.
-

can think of it

definition of the statement

b then amounts to the assertion that we can make the error in


approximating to / by f(x) as small as we like by taking x sufficiently close to b
on the left.

fix) -*

can find

\fix)-l\'<e
provided that

The number

satisfied.

such that

\m -/|<e
provided that b

is

is satisfied.

If a

-*

<

\a\

(3

and
is

are real numbers,

it is

equivalent to < a <

often useful to note that the inequality

(3,

or,

what

is

the

same

- <a<

< e can be replaced


or, alternatively, by e < fix) < e.
< \x < 5 in the last definition equivalent to
Similarly, the condition
< \x - 1| < 5 to
the assertion -S<j:-?<6andx^|. Thus to say that

8<x<%or%<x<%
+ 8.
say that x satisfies one of the two inequalities
Thus, in the definitions above, the condition |/(x)

/|

is

%\

Limits from the right

(3

(see exercise 1.20(1)).

throughout by e < fix) / < e

8.2

thing,

is

Suppose again that /is defined on an


tends (or converges) to a limit

f(x) *

as

-> a

as

interval (a, b).

We

say that/(X)

tends to a from the right and write

With the help of the

point

if

-*

a+

f{x)

74

e (a, b). Then

/(x)->/ as *-*+.

the following criterion

it is

easy to prove the following result.

Proposition Let /be defined on an interval

8.4

lim

remark

+,

or, alternatively,

last

is satisfied.

fix) -> / as

-> % if

and only

(a,

b) except possibly at a

if fix) -* I as

-*

and

76

Limits offunctions

8.5

Example Let /be

the function from

to itself defined

Obviously, the choice 5

by

proof

-x

(x<\)

that f(x) -* 2 as

Then

f(x)

lim

(ii)

f(x)

it is

though f(x)

2.

is

Having said
ton f(x)

does not exist even though /(l)

Proof
find a 5

>

(i)

f(x)

as

is

perfectly well defined (/(l)

-*

Given any e

>

= 0).

as

* % given in

8.3, consideration of

%.

it is

not necessarily true that

this,

we can now

If /is defined
1

Since

as* tends

to % even

/=/(?)

And, even

(see exercise 8.15(2)

turn to the definition of continuity at a point.

as

on an interval

continuous on the
f(x) -*f(a) as x

->

(a, b).

Then we say that

*->.

graph of/ does not have a 'break' at the point

we are only concerned with values of x


satisfying x <
we can replace f(x) by 1 -*. The condition \f(x) - 0|< e
then becomes 1 x < e which is equivalent to e <x
< e. Adding 1
throughout, we see that f{x) - |< e is the same as 1 - e < x < 1 + e.
1

Roughly speaking, to say that /is continuous

5 <x <

completes the

not defined at the point % (see exercise 8.15(3) below).

/(*)-/(?)

we must show how to

|/(x)-0|<e
1

this

+.

Suppose that /is defined on an interval (a, b) and %


/is continuous at the point % if and only if

such that

provided that

and

below).

follows from proposition 8.4 that

it

this true

quite possible for/(x) to tend to a limit

if f(X) ' s defined,

Note that

^e suffices to make

what happens when x actually equals % was carefully excluded. When taking
limits we are only interested in the behaviour of/(x) asx tends to \ not in what
Thus

lim

In the definition of/(x) ->

happens when x equals

(i)

Continuity at a point

8.6

(x>\).

,2x

77

Limits of functions

left at the

a +, then

(a,

at the point %

ff.

b] and f(x) * f(b)

x-*bwe say

point b. If /is defined on

we

means that the

say that /is continuous

an interval [a, b)

on

that /is

and

the right at the point

a.

The problem

is

now

reduced to the following. Given any e

>. 0, find a

>

such that
1

provided that

6 < x <

Obviously, the choice 6

proof that f(x)


(ii)

f(b)---/-\-z-

-e<x<\ +e

f(x)

-*

-*

as

as

->

x *

/(a)--

=
1

e suffices to make this true and this completes the

+. Given any e

> 0, we must show how to

find a 5

continuous

at

/continuouson

/ continuous on

j-.

>

the left at

the right

b.

atfl.

such that

l/W-2|<e
provided that < x < + 6. Since we are only concerned with values of x
satisfying x > 1 we can replace f(x) by 2x. The condition \f(x) - 2|< e then
becomes \2x - 2| < e which is equivalent to e < 2(x 1) < e. Thus
l/(x) - 2|< e is the same as 1 - ^e < x < 1 + |e.
The problem now reduced to the following. Given any e > 0, find a 5 >
1

8.

Example For any

/(*)

real

numbers a and

0,

the function

/:

defined by

= ax +

is

continuous at every

is

such that
1

-|e<*<l +\e

provided that

1< x < + 6.
1

given.

real

number

|,

Proof Assume a # 0. (If a = 0, the proof is even


Choose 5 = e/|a|. Then, provided that \x %\ < 5
I/(X)-/(!)I

= N*-f)l =

easier.)

|a|.|x-$|<|a|.8

e.

Let e

>

be

78

> 0, we have found a 5 > such that |/(x) /()| < e provided
1| < 5 Hence f(x) -> /() asx -g and so /is continuous at the point
that \x
Note that we showed that /(.x) /() < e 'provided that x g < 5'
instead of 'provided that 0<|.x g|<5as appears in the definition of f(x) -*
automatically true that /(jc) /(g) < e
because, when x = g,
as x
g- This
only when we are considering a limit =/(g)
because then f(x) /(g) = 0.
Given any e

g.

it is

is

that

important to exclude the possibility that x

it is

It is

79

Limits of functions

Limits ofJunctions

Example We show how theorem 8.9 can be used to prove that the
function /: R -* R defined by f(x) = ax + (3 is continuous at every point g (see

8.10

example 8.7).
* <*>. By
Let (jc n ) be any sequence of real numbers such that x n * g as n
proposition 4.8, ax n + * | + P as n - . From theorem 8.9 it follows that
ax + * ag + p as x -* g and this concludes the proof.

g.

y=/te>=X+/

Connexion with convergent sequences

8.8

Theorem Let /be defined on

8.9

f(x)

that

xn

as

* g if

= | (n

Proof

we can

\x n

2,

(i)

.)

b) except possibly for g

(a,

(a, b).

and

+ 1 as ->

>

first that f(x)


I as x * g. Let e
be given. Then
such that \f(x) l{<e provided that < |x
6 Now
g|
a sequence of points of (a, b) such that x n = g (n = 1, 2,
.)

<

is

Then

for each sequence Cx n> of points of (a, b) such


* -> g as n * , it is true that f(x n) -* I as n -*

if,

Suppose

>

find a 5

suppose that <x n >

and x n

and only

Since 5

>

- %\< 5. But x n = g (n =

1 ,

we can
2,

.)

find an iV such that for any

n>N,

< \x n - g|<5. But this implies

and so

that

l/W-'Ke.
Given any e > 0, we have found an N such
|/(x)

- /|< e.

Thus f(x) -* / as

Now suppose
x n & % ( = 2,
(ii)

>

We now suppose

that, for each


.)

that

that, for

any

Properties of limits

8.11

The following propositions are analogues of the combination theorem


sandwich
theorem of chapter 4. They are easily deduced from these
and the
help
of theorem 8.9.
results with the

n>N,

-* .

Proposition Let /and g be defined on an interval (a, b) except possibly


as x -* g and suppose
I as x
that f(x)
g and g(x)

8.12

sequence (x n ) of points of (a, b) such that

and jc -* g as -* oo it is true
not true that/(x) * / as x
>

it is

that
- g

f(x n ) * I as n -* .
and seek a contra-

G (a, b). Suppose

at g

X and p

that

are

any

real

*m

numbers. Then

pg(x)

^\l + pmasx^H

(i)

\f(x)

(iii)

f(x)lg(x) -* l/m as

diction.
If it is

not true that /(at)

>

that for each value of 5

+ / as x * g
we can

then, for

find an

some

satisfying

> 0,

<

|.x

it

must be true

g| < 6 such

An

\f(x)-l\>e.
In particular, if S

then

we can

find an

xn

satisfying

But then U) is a sequence of points of (a, b) such that xn


1, 2, . . .)
| (
and x n -> I as n -* but for which it is not true that f(x) -* I as n -* . This

is

Hence,

(a,

as

6) such that x n

Z>

if and only

+ 6 as - ,

it is

from

the left or right. For

for each sequence Cx n > of points of


true that /(x n ) -> / as n -* .
if,

A polynomial is continuous

at every point.

rational func-

if

P(x)

the following result which

continuous

contradicts our assumption above.

Similar results to theorem 8.9 hold for convergence

is

at every point at which it is defined.


Proof Example 8.7 shows that x - g as x * g. Repeated application of
h
proposition 8.12(ii) then shows that x -> g* as x * g for each keN.

tion

)-H>e.

example, fix) ->

m = 0).

as a theorem.

Theorem

8.13

(provided

< \x n g| < 1/n

such that
\f(x n

->

important consequence of proposition 8.12

we quote
1/ra,

that

a nx

a n - 1 x"-

..

+ a,x + a

then a repeated application of proposition 8.1 2(i) yields


i.e.

Pis continuous at

To show

P(x)^P(& asx -> g,

g.

that a rational function

appeal to proposition 8.12(iii).

is

continuous wherever

it is

defined

we

80

Limits offunctions

Limits of functions

8.14

Proposition (sandwich theorem) Let f,g and /; be defined on (a, b)


e (a, b). Suppose that g(x) -* / as x * f , h(x) * I as x -* %

f(x)>0

except possibly at |

provided that

and that

81

- <x < +

/;

/i

and x

# .

What happens

if (i)

< 0,

- 0?

(ii) I

g(x)<f(x)<h(x)
when x =

except possibly

%.

-* I as

Then/(x)

-* .

Note The analogues of propositions 8.12 and 8.14


from the

left

Limits of composite functions

8.16
for convergence

or right are both true.

Suppose that f(y) -* / asy -*


to write y = g(x) and conclude that

But some care

15

Exercise
(1 )

and

ix

+ -

,73

4)

lim (-=
x-*i [x 2

'*-*o

Let/:

R- R

as

- . It is

necessary. Consider, for example, the functions /:

is

tempting

R -> R

defined by

3x

5x*

+9

0*1)'

1)

g(x)

1.

be defined by

3-x

(*>1)

(x

2x

(x<l).

lim

(ii)
J

Then /(j>)-- 2 as y -* 1 andgXx)-* 1 as x -+0. But it is not


because, for all values of x,f(g(x)) =/(l) = 3.
as x -*

1)

true that f(g(x))->- 2

Theorem Suppose that f(y) -* I as y - tj and that (x) -> 77


Then either of the two conditions below is sufficient to ensure that

&7 7

Show

that /(at) -* 2 as

->

- and f(x) -* 2 as x -> 1 + using only

definitions of 8.1 and 8.2.

Deduce that the

the

/&(*))-/

(i)/is continuous at

exists but

(ii)

is

not equal to/(l). Draw

Let /be defined for

all

x except x =

.fl=I

For some open

x S 7, except
a

graph.

possibly

77 (i.e. /

<^0).

* |.

= /(t?)).

interval / containing ,

x=

it is

true that g(x)

77

for

any

/Voo/ Let e > be given. Since f(y) -> / as y * tj, we can


such that I/O) - /|< e provided that < \y - 77I < A.
Writing v =g(x), we obtain

by

as

as *-$.

limit

lim f{x)
X -* 1

(3)

and that g{x) * 77

Calculate the following limits

(i)

(2)

g:

-*

7?

find a

A>

(0

|/fe(x))-/Ke
Prove that f(x)

2 asx

even though /(0)

is

noj defined.

Use the definition to show that |x | -* as x -> . A function /is


defined on an open interval which contains % and, for each x in this
interval, /(x) lies between and x. Prove that/(x) -> as x -* %.
(5) Let n N Prove that the function / [0, ) - R defined by
(4)

= x Vn

continuous

(6)

But (x)

\g(x)

-*

7?

as

- 7?|< A
x

-*

and

(2)

A > 0. Hence we

can find a 6

>

such that

at

each

(3 )

\g(x)-v\<A

(x>0)

>

and continuous on the right at 0. [Hint:


Use the sandwich theorem and exercise 3.1 1(2) to prove that/is continuous at > 0. For continuity on the right at 0, appeal to the defiis

<

f(x)

provided that

nition of f(x) -*

as

Suppose that f(x)

-*

as

+.]
-> . If

> 0, show

that, for

some h

> 0,

provided that

0<|.x-|<5.
We would now

like to say that (4) implies (3),

which implies

(2),

which

in

does not
turn, implies (1). This would complete the proof. But notice that (3)

4)

82

We

imply (2) in general.


the theorem.

we assume
when y = r\ (see
If

may

therefore need to use one of the hypotheses

(i)

or

(ii)

of

120

Exercise
(1) Prove the following

that/is continuous at
the

77,

then

= /(rj) and so

\f(y)

remarks at the end of example 8.7). Thus, in

replace condition (2)

by condition

(3)

l\<e even

this case,

x'
(ii) x

we

and the argument described above

(i)

7,

-*

asx->0

-*

as

that/(g(jc)) -

we assume hypothesis

(ii),

we can be

then

sure that(je)

vided that x satisfies (4) for a sufficiently small value of 5


dition (3) can be replaced

by condition

(2)

>0. But

and the argument goes through again.

-*

/ as

+ .

* I as y * + a nd g(x) * + as x * + .
x

* + .

theorem 8.17 do not occur

^17 pro-

then con-

(2) Suppose that f(y)

goes through.
If instead

83

Limits of functions

Limits offunctions

Explain

why

Prove

the difficulties encountered in

in this case.

1
-+ 1 as y *
+. Prove that fix' ) -+l asx-> + .
(3) Suppose that f(y)
on (a, b) except possibly for | G (a, b). Let
(4) Suppose that /is defined
.)
(x n ) be a sequence of points of (a, b) such that x n i= % (n = 1 2,
and x n -* % as n -* . Let (y) be a sequence with the same properties.
,

Divergence

8.18

We

write

down two sample

definitions.

The reader

will

have

and /(*)

-*

as n

* but /(y n

-*

m as M -* , show that

little diffi-

culty in supplying the definitions in other cases.

We

say that/(x)->--l-

as*->

if,

given any /f

> 0, we can

find a 5

>0

does not exist. [Hint: theorem 8.9]

such that

(5) Let/:

U^U

be defined by
(x rational)

<x < g + 6.
We say that f(x) * / as x -*

(1

provided that |

if,

given

any

> 0, we can

find

an

(x irrational).

X such
Show

that

that

|/(x)-/|<e
lim f(x)

provided that

<

X.

-*

does not exist.


(6)

If /is

defined as in the previous question, prove that

iim {*/(*)}

I-0

f{x)+->*sx^'i +

8.19

Example Prove
Proof Let

H>

that

f(x)-'l as

-*

be given.

x-*-

+ asx->0+.
We

have to find a 5

>

such that

~>H
x
1

provided that

<x < 8

Obviously the choice 8

= H'

satisfies the

requirement.

0.

If

85

Continuity

Note

both diagrams, the graph of /is unbroken above the interval

that, in

question.

CONTINUITY

In general, a function

interval / provided that:

/ is continuous on an

is not an endpoint;
hand endpoint of/, //

(i)

it is

continuous at each point of/ which

(ii)

it is

continuous on the right at the

and belongs to /;
(iii) it is continuous on the
and belongs to

left at

left

the right hand endpoint of/,

this exists

//"this exists

/.

Continuity on an interval

9.1

A
x

in

function

/ is

continuous at a point %

| (see 8.6). Geometrically, this

'break' or 'jump' above the point

To

say that

/ is

if

and only if f(x) -*/()

that the graph of

/ does

interval /,
lifting

i.e.

as

(i) All

polynomials are continuous on every interval. All rational func-

tions are continuous

that

we can draw

its

graph

the part of the graph

is

is

on any

(ii)

Let /: (0, )-

interval (0, ) because

required. This depends

interval not containing a zero

of the denominator

(theorem 8.13).

which

our pencil from the paper.

mathematical definition

Examples

9.2

not have a

continuous on an interval / should mean that

unbroken above the


lies above /without

A precise

means

on the type of

(iii)

U -+R

Let /:

R be
it

is

defined

by f(x) =

continuous

at

1/jc.

Then / is continuous on

each point % satisfying

the

> 0.

be defined by

interval in question.

A function / is continuous
is

continuous

at

on an open interval / if and only

if

(x 2

(jc<1)

\ 3/2

(*>1).

it

Then / is continuous at every point except x = 1 But / is continuous on the


x = 1 and thus /is continuous on the interval [0,1]. It is not continuous

each point of/.

left at

on [1,2].

Where an endpoint of an
plicated definition

is

interval belongs to the interval, a slightly

required.

The most important

case

is

more comthat of a compact

interval [a,b].

A
only

function
if it is

right at a

/ is

continuous on a compact interval

continuous at each point of

and on the

(a,

[a, b] if

and

b) and continuous on the

left at b.

9.3
/

-^mmt&i
/continuous on

84

ia.b)

/continuous on

\a,b)

if

Proposition Let

and only

if,

given any

/ be
xG

defined on an interval
/

and any

> 0, we

\f(x)-f(y)\<e
provided that

y /and

satisfies \x

y\ <8.

/.

Then / is continuous on

can find a 5

>

such that

86

Continuity

9.4

Proposition Let X and

n be

numbers and suppose that the func-

real

tions /and g are continuous on an interval

/.

Then

We split the content of the continuity property into three lesser theorems and
prove these separately. Given that /is continuous on 1 = [a, b] , it follows from

so are the functions

(i)X/+W?

theorem 9.9 that /=/(/) is an interval. From theorem 9.1 1


interval /is bounded and from theorem 9.12 that J includes

PDA
(iii)

(provided g(x)

fig

87

Continuity

for

any x

9.5

Theorem Let /be continuous on an


under/is also an interval.

J be continuous on the interval / and let


interval /. Then / o g is continuous on /.

/ -*

/: J-*R be continuous on the

Proposition Let /be continuous

9.6

sequence of points of / such that x n -*

if

on the

interval

as n -* ,

/.

If

6/ and (x n

Proof To show that J

show
) is

that, if y x

Since y i

then

The first is proved by an appeal to


the limit definitions of 8.1, 8.2 and 8.3. The second follows from proposition
8.2, the third from theorem 8.17 and the fourth from theorem 8.9. The last of
these results can be remembered in the form
easily proved.

i.e.

= /(lim x n )

for a continuous function the limit

these sets.

It

< tn )

f(t n

and function symbols can be validly

9.10
to a theorem which

It is,

then X

EJ (see

2.9).

is

one or other of

at zero distance

E 5 is at zero distance from

from the

T.

Then

T can be found such that t n -* s as n -* (exercise


continuous on / it follows fhat/(r) -/(s) as n -* (pro-

of points of

)>\

and therefore /(s)

Continuity property

worth mentioning.

T = {x:f(x)>\}

other (exercise 2.13 (6)). Suppose that

sequence

the image of /

an interval we need to

is

< X <y z

follows that a point of one of the sets

(n

may seem

so obvious as to be hardly

however, of a fundamental importance comparable to

Continuum Property of 2.2 and will serve as

1,2,...)

> X (theorem 4.23). We already know that f(s) < X and so

= X.

A similar argument

that of the

j>,

Then

are non-empty. Also, every point of the interval /belongs to

follows that /(s)

We now come

/.

position 9.6). But

exchanged. (We say that the symbols 'commute'.)

9.7

endpoints.

J and y 2 e J, the subsets of / defined by

4.29(6)). Since/is

lim /(*)

interval

=f(I)={ f(x) x / }

are elements of J and

andy 2

S = {x:f(x)<\};
These four propositions are

its

e /).
9.9

Proposition Let

follows that the

it

applies if a point of

T is

at zero distance

Corollary (intermediate value theorem) Let

interval / containing a and b

it

Hence X EJ.

If

between a and b such that X

lies

from S.

/be continuous on an

between f(a) and f(b), then we can find a

= /().

foundation stone in

our development of the calculus in the next few chapters.


fib)

9.8

Theorem (continuity property)


Let

[a, b]

/be continuous on

under /is also

compact

compact

interval [a, b]

Then the image of

interval.

\c.d

Proof The corollary

simply a somewhat clumsy restatement of

9.9.

9.11

Theorem Let /be continuous on the compact

bounded on
\a.b\

is

theorem

interval [a, b].

Then /is

[a, 6].

Proof Suppose that /is unbounded on

[a, b].

Then we can

find a

88

Continuity

sequence (x n

Continuity

of points of [a,b] such that

|/(xB)I-*-

+ 0O as-+

(1)

(see exercise 4.29(6)). Since [a, b]

exercise 5.21(4)).

is

compact, there exists a subsequence <x

>

by the Bolzano-Weierstrass theorem (see


Because /is continuous on [a, b\ ,/(x r ) -*/() as r -> (pro-

.which converges to a point

b]

[a,

corollary 9.10, that/()


| between

position 9.6). But this contradicts (1).

Theorem Let /be continuous on the compact


achieves a maximum value d and a minimum value c on
9.72

Proof We know from


Let

be the

<2

[a, b]

pact,

supremum of /on
d as

such that f(x n ) -*

(x n

interval [a, b].


[a,

<

it

follows that /(x

-*/() as/--*

com-

[a, &],

<*>.

Hence

Example Let/:

9.16

some

-*

[a, b]

= t

continuous function from

point of the interval. (This

defined by

b\ be continuous on [a, b] Then, for

[a,

[a, ft],

m)
Thus

some

[a, b].

= ^ and thus the supremum d is actually a maximum.


^(H)
A similar argument shows the infimum to be a minimum.

Example The function/:

for

0.

X n ) of points of

-* (see exercise 4.29(6)). Since [a, b] is

[a, b],

and

Then/

bounded on

b] and consider a sequence

[a,

b]

the previous theorem that /is

contains a subsequence <x r > which converges to a point

Because /is continuous on

9.13

89

compact

interval to itself 'fixes'

some

the one-dimensional version of Brouwer's fixed

is

point theorem.)

/(*)

(x>l)

(*<1)

fib)

not continuous on [0,2]. Observe that 2

is

no value of between

and 2 such

= /(0) < 4 </(2) =

that /()

4.

Note

that, for

5 but there

is

Proof The image of [a, b] is a subset of [a,b]. Thus f(a) > a and
The function g [a, b] * U defined by g(x) = f(x) x is continuous
b] by proposition 9.4. But g(a)>0 and g(b) <0. By corollary 9.10, for

< b.

on [a,
some

[a, b]

it is

true that()

= 0.

Thus/() =

%.

satisfying

2<X<5,
{x:f(x)<\} =

{x:f{x)>\} =

{2};

{5}.

Example The function/: (0, ) * R defined by /(x) = 1/x is continuous on the open interval (0, 1) but is not bounded on (0, 1). The function
9.14

g R * R defined by g(x) = x is continuous everywhere and happens to be


bounded on the open interval (0, 1). But g does not attain a maximum value or
a minimum value on (0, 1).
:

9.15

Example Show
17x 7 -19a- 5

9.17

that the equation

(1)

-1 =

has a solution % which satisfies

< | < 0.

Each of the following expressions defines a function on ( 2, 2).


Decide in each case whether or not the function is continuous on
( 2, 2) and (b) [0, 1 ] In each case draw a graph.
.

Proof The function/: R-*R defined by /(x) = 17x - 19.v - 1 is a


polynomial and hence is continuous everywhere. In particular it is continuous on
[-1,0]. Now/(- 1) = 1 and /(0) = - 1 Since - 1< < 1 it follows from
7

Exercise

(i)

/(*)

2x4-3

2x-5

(ii)

fix)

= \x-l\

(a)

Continuity

90

(*<1)

(1

f(x)

(iii)

(iv)
{

fix)

iO

+4
= *2
2

/(*)

(v)

(x>l)

\2

(0<x<l)

(1

(vi)

-4

/(*)

(otherwise)

10

D IFFERENTIATION

10.1

Derivatives

(0<*<1)

x(x-\)

(otherwise)

Which of these functions are bounded on (- 2, 2)1 Which attain a


maximum on ( 2, 2)? Which attain a minimum on ( 2, 2)?
for each rational number
(2) A function /is continuous on an interval / and
r

/ it

is

true that

Suppose that /is defined on an open

= r\

f(r)

(4)

Show

that

all

= x1

any xE.1. [Hint: proposition 9.6.]


polynomials of odd degree have at least one (real) root.

Prove that/(x)
(3)

Suppose that / is

continuous

minimum

as

x -
on

value

/ be continuous on

x /,

said to be differentiable at the point if

for

+ oo and/(jc) -*

value or a
(5) Let

Then /is

there exists a y

at

every point and that f(x) ->

/.

Suppose that, for each

For a function /which

/ such that
/'(I)

/ for which /() = 0.

Equivalently,
(6) Let / be a closed interval

and

<a<

let

Let /: / -* / satisfy the

fm

inequality

\ax)-f(y)\<a\x-y\
x / and y /. (Such

From an
a function

ping). Prove that /is continuous

Let x,

/ and

define

on

map-

(n

its limit

1
/

2,

and denoted

differentiable at %

is

we

therefore have

m-rm

Hm,

means

we may

- Hm

write

ffi

.).

satisfies

Prove that the

+ h)-f
(1)

intuitive point of view, to say that a function /is differentiable at %

that a tangent can be

or gradient of this tangent

/.

x n+l - /(*)

sequence <x n > converges and that

called a contraction

is

called the derivative of /at the point %

[Hint: Bolzano-

Weierstrass theorem.]

for each

i
a %

is

by/'() or >/).

l/OOKil/001Prove the existence of

x-%z

exists. If the limit exists, it

the compact interval

if the limit

lim

. Prove that /attains a maximum


the set U.

/containing the point

and only

/(*)-/()

,.

*-**

as

interval

is

drawn

to the curve

y = fix) where x =

The

slope

then equal to /'() This geometric interpretation

is

based on the diagram below.

/=/(/). [Hint:

exercise 5.21(1).]
/(A)

r-/(> + /(.v

{)

91

92

y = /(?) + l(x |) be tangent to y = fix) where x %


slope of the chord PQ approach the slope / of this line as

In order that the line

we require
x -* . The

that the

slope of the chord PQ

qr
_
PR

is

(ii)

Let/:

(0,

<*>)

-*

0.)

be defined by f(x)

1/x (x

> 0). Then

is

fXx)=D -=-
l

for

all

values of x

and hence our requirement reduces to

f^lM .

lim

when h

quite meaningless

m-m
x-%

> 0.

Proof We have

+ h)-f(x)

f(x

X |

***

93

Differentiation

Differen tiatiun

\_

h x

+h

-h
+ ft)

h \x(x

as in the definition of differentiability.

(Note:

follows immediately from our definition that a function/is differen-

It

tiable at

if

ij

and only

m+

there exists an

if

-f -

1')

if'

o(h)

itself

more

(h -* o).

We

shall

avoid the notation introduced in this section because,

this level

from / to

on

/.

whose value

at

second derivative /"() or

fm

open

= I

each point x

we

interval /,

In this case it is natural to define/' or

/ is f'(x). We

Df to be

say that/is

difficult to

it is

for our purposes. In

If /is differentiable at each point of an

-* 0.

More notation

10.4

readily to generalisation.)

Higher derivatives

differentiable

-a as

x2

(2)

using this notation,


10.2

x(x + h)

Here o(h) denotes a quantity which tends to zero when divided by h. This form
of the definition lends

such that

many

applications, however,

it is

when

a precision adequate

unnecessary to maintain

of precision. In such cases, the notation described below can be a very

and

useful and powerful tool

the function

make statements with

it

would be pedantic

to

deny oneself the use of it.

Replace by x and h by 5x in formula (1) of 0.1 Then


1

can then define the

D 2f(g) at the point % by

/'(*)

f(x

+ 8x)-f(x)

lim
&x - o

lim
Sx

8x

(3)

- o S*

/'(*)-/'(*)

x-%

If>>

=f(x), the quantity 5^ =f(x

+ 8x) f(x) is usually

said to

be the 'small

change in y consequent on the small change 8x in*'.


provided that the limit exists. Similarly for third derivatives and so on.

From formula

(3)

it is

only a short step to the

much

used (and

commonly

abused) notation

Examples

10.3

(i)

Let/:

/'(*)
for

all

R->R

= Dx 2 =

be defined by f(x)

= x 2 Then
.

symbol x ambiguously. It is used


simultaneously as the point at which the derivative is to be evaluated and the
variable with respect to which one is differentiating. In applications where this
distinction is unimportant, the ambiguity creates no problems. But, for the
This notation has the drawback that

2x

values of x.

Proof We have

material of this book, the distinction

f(x

+ h) -f(x)
h

(x

+ hf -x 2
h

2xh+ h 2

is

2x

2x

as

Further confusion

-* 0.

(Note that we studiously ignore what happens when h


that this

h -*

ential

- see

df of a

+ h)-f(x)

possible

is

when

important.

'differentials' are introduced.

The

differ-

function /may be regarded as the function of two variables given

by
8.4.

It is

as well

df(x;h)

permissible, for the expression

f(x

is

uses the

it

Formula

/'(*)//.

(2) can then be rewritten as

(4)

94

Differentiation

f{x

+h)- f{x) -dfix-h) =

o in)

ih

+ 0)

(5)

a fixed value

Newton

of*, the equation

and abandon

as the quotient of these quantities

all

the apparatus of limits. This

admittedly an attractive one and it is on this idea that


based his 'Theory of Fluxions'. Unfortunately it cannot stand up to

metaphysical notion

provided that /is differentiable at*.

For

95

Differen tia tion

is

close logical scrutiny.

k
is

dfix-h)

= f\x)h

the equation of a straight line. If the h and

(6)

axes are drawn as below, (6)

the equation of the tangent to the graph of the function

A"

(Note:

A common

derivative

is

is

error

is

to confuse the terms derivative and differential. A


as a 'differential coefficient' (because of (7)

sometimes referred to

above) but should not be called a differential.)

/at the point*.

10.5

Properties of differentiable functions

10.6

Theorem Let /be defined on an open

=/'(.v)/i

point

. If /is

interval /

which contains the

differentiable at , then /is continuous at %.

Proof We have
fix)

-m)

Hence /(*) -*/()

The reader may justifiably feel that little or nothing has been added to our
understanding by the introduction of the idea of a differential, the idea being
more useful when functions of several variables are being considered. Our only
reason for discussing differentials is to explain the meaning of the equation
dy
This

is

simply equation (6) with the variable h replaced by dx.and the variable

k replaced by dy. Thus dx and dy should be thought of as


one chooses this notation is to make the formula

variables.

= -fdx
dx

x -* | and

-1

the

U * U

proof is complete.

be defined by

> 1)

(x

[1

(*-W(S)-0as*^.

x-%

(*<1).

can see immediately that /is not differentiable at the point

not continuous there.

The reason

Example Let/:

10.8

dy

/(*)-/()

Example Let /:

10. 7

We

f'ix)dx.

as

R -> R

be defined by

(7)

a valid one.

dy denotes 'the small change in y consequent on a small change


dx
Formula (5) tells us that this is nearly true for small values of dx but it
will only be exactly true for functions /with a straight line graph. Should it be
necessary to discuss 'small changes in * and >>', the notation 5* and 8y is
It is false that

/(*)

in *'.

and one can say that 8y is approximately equal to f'ix)bx.


Even more false (if that is possible) is the idea that dy and dx are somehow
'infinitesimal quantities' obtained by allowing 5* and 8y to tend to zero. One
would then regard
available

ix>\)
2

Then /is continuous

ix<\).
at 1

but not

differentiable (The graph has a 'corner').

We have

Iim

h->0+

dy
dx

r
lim

h-o-

/0+^)-/(l)
;

lim

h--0+
,.

(1+/Q 2 -1

lim

h->o-

Thus the converse of theorem 10.6

h
is

false.

.
2.

because

it is

96

Differentiation

10.9
taining

Theorem Suppose that /and g are defined on an open interval /conthe point %. Let X and \x be any real numbers. Then, if /and g are

differentiable at ,

(ii)

+ liDg
>{X/+WT} =
D{fg) = fDg + gDf

(iii)

D j4= gJ^Zgte

(i)

97

Differentiation

Now assume

D(x h
and the

(provided *(g)

its

truth

when =

fe
Thus Z>.x "fee*"1 . Using theorem

k.

= D(x h .x) = x

.l

+x.kx

k -

= (A+l)x

0.9 (ii),

fc

by induction.

result follows

10.11
at the point f.

(1)

Exercise

Prove that, for any x.

/Voo/
1

0)

(X/(?

m+h)-m

+H

+x'

g(M+m-g

(2)

If

(3)

Let/:

RHR

am*

{/(i

+ a) -/ct)r(i

h)

jf

(Note that(i;

+m)ga + h) -/)*)}

9* 0, prove that jDx"

= nx n

~1
.

(x>l)

x 2 +1

(x<

1).

at the

point

and has derivative

R * R be defined by g(x) =\x\. Prove

tiable at the point 0. [Hint:

that

Consider the right and

2.

not differen-

is

left

hand

limits

separately.]

m+h)-m)

*ci + /o

+ A) ->-()

as

+/)

g (S+*)-g(l)
h

-*

because, by theorem 10.6,

g is continuous

(4)

A polynomial P of degree n
/>'(g)

= 0. Prove

P{x)

for

all

has the property that />()

and

that

(pc-tfQ(x)

x, where Q(x)

is

a polynomial of degree

2. (See exercise

3.11(3).)

at?.)

.v

be defined by

2x

Let g:

and

Prove that /is differentiable

+xy

(1

a negative integer

is

/(*)
()

-2x

+ A) + ?(! + h) - \f - WG)}

ftm _

flt+*)
(

R-*R be n

m g(.i+h)-m

pw

{*)/'(*) -/)*'(S)}

as A

-0.

(*-)"
(

fed)}
has the property

g.

Show

that the

defined by

f + ^rfm + ^=rVa) +
+

i
5

times differentiable at the point

'Taylor polynomial'

L m m+h)-m _

/g+ft)gg)-g(s+ft)/(S)

(5) Let/:

fCn-iy
-/-(|)

1)!

P*\|) = /

w(g) (* = 0,

2,

).

(Note that

<fc)

70.70

Example

If is a natural

number, then, for any x,

/ (l) denotes the derivative of order k at the point jf.)


^(6) Let/and,? be n times differentiable at the point %. Prove 'Leibniz's
rule',

Dx" = nx"-\
Proof This
(x

+ h)-x

is

obvious

-*

when n =
as

* 0.

i.e.
1

since

Dnfg = t ("W/EP-'f
;=o

\/

at the point g. [Hint: Recall the

proof of the binomial theorem.]

98

Differentiation

10.12

Composite functions

The next theorem

is

dz

dx

dy'dx

usually

remembered

in the

form

+ * 97)".97x 96

=y 100 andg(x) = l+x 97 By


- DfXgfx)) = f'(g(x))g'(x)

Proof Set/0)

+x 97} 100

theorem 10.13,

= 100fe(x))".97x 96

differcntiable at the point

(fg)'(x)

100
= l0O
Example D{\ + a-"}

D{\

Theorem Suppose

10.13

10.14

dy

dz

99

Differentiation

that

g is

differentiable at the point

x and that/is

+x 97}".97x 96

100{1

y = g(x). Then

f'(y)g'(x).

Proof Recall thatfog(x) =f(g(x)). We therefore have

to consider

10.15
(1)

f(g(x+h))-f(g(x))

Exercise

If*

>0 and SN, prove that

D{x vn \ =

-x lln x-K
n

= g(x +h) g(x). Since g is

Write k

and so k

-*

as

-* 0.

differentiable atx,

it is

continuous there

[Hint: exercise 3.1 1(6). Recall that the continuity of the nth root func-

Suppose that k=0. Then

tion was considered in exercise 8.15(5).]

f(g(x

+ h))-f(g(x)) _ f(gix+h))-f(g(x)) g(x+h)-g(x)


h
g(x + h)-g(x)
h

(2)

function (theorem 10.13) to

'

+ k)-f(y) g(x+h)-g(x)

f(y

__

round

get

this difficulty

we

rx

provided

thatx>0.

F(k)

(**0)

that, for

any rational number

r,

exist.

introduce the function

+ k)~f(y)

show

r ~'

D{x r ) =

(i)

(fjy

the rule for differentiating a composite

values of x for which they


(3) Evaluate the following derivatives for those

tempting to deduce the conclusion of the theorem immediately. But care


necessary. What happens if, for some values of h, k = g(x +h)g(x) = 0? To
It is

is

Use the previous question and

'

(4)

D{ 1 + x V2

s
(ii)

Dy/(X + V(*

+ V*)).

Suppose that

(*

f'(y)

= 0).

Since /is differentiable sty, F(k) -+f'(y) as fc ->- 0. Because F(0) =f'(y), it
F is continuous at the point 0. From theorem 8.1 7(i) we deduce

follows that

(5)

that

whenx=l. Prove that/'(l) = or /(l) = 1.


Let/: R->R be differentiable at x. Suppose that/admits an inverse
function/" U-*U which is differentiable atj =f(x). Prove that
1

F(k)^f'(y)ash^0

= g(x + h) g(x).
For k^Ov/e have shown

where k

DT (y)
l

Df(x)

that

f(g(x+h))-f(s(x))

g(x

F(k)

[Hint: f'

+ h)-g(x)

f(x)

= x\

The function g: R^-R defined by g(x)

admits an inverse function

#""':

U-*U.

= x3

Isg' differentiable at the point

0?

But

this

equation

is

also true

when k =

since then

both

sides are zero. It

t(6)

function/:

R^-R

is

defined by

follows that

f((x

+ h))-f(g(x))
f'(y)g'(x) a

and

this

is

what had to be proved.

sh^Q

(x rational)
(x irrational).

Show that/o/(x) = x
theorem 10.13?

for

all

values of x.

What may be deduced from

Mean

101

value theorems

Proof By definition

MEAN VALUE THEOREMS

11

/(*)-/()

/'(S)asx^?.

*-?
Suppose
interval

Let/be defined on
that

/has

x-%

maxima and minima

Local

11.1

a local

maximum

an open interval (a, b) and let

{a, b).

We

say

provided that

at if

values of x in

all

> 0.

From

Exercise 8.15(6),

it

follows that for

some open

h)

>0

(1)

xG 1 and xl=%.

h, %). Then x % < and hence


< /(). Thus /cannot have a local minimum at . Let
follows
be any number in the interval (, + h). Then x 2 % > and so
Let x, be any number in the interval (

m<m
for

that /'()

/=(-/?,? +

it

follows from (1) that/(x,)

some open

interval /

which contains

%.

Similarly for a local

x2

it

from

minimum.

(1) that/(x 2) >/(!). Thus/cannot have a local

maximum

at %.

/'()>

/ has beat max

Very roughly, /has


point

%.

f/(?)

is

/has

local

min

a local

'b

al {.

maximum at if its graph has a 'little hill' above the


Similarly,/ has a local minimum at | if its graph has a 'little valley' above

the point
J

/has

al ?.

*i

a local

maximum value of/ on the whole interval (a, b), then obviously
maximum at . But the converse need not be true (see the diagrams

similar

argument applied to /deals with the case when /'(?)

remaining possibility is/

(if)

the

1 1 .3

Stationary points

A point |

above).

at

which /'()

stationary points give rise to a local

11.2

< 0. The only

= 0.

Theorem Suppose that/is

If /has a local

/'a)

maximum

or

differentiable

minimum

on

(a, ft)

at |, then

and that

G (a,

b).

will

is

Not all
minimum. The reader

called a stationary point of/.

maximum

be familiar with the case when /'(?)

or local

and /has a point of inflexion at %

in the diagrams below.

o.

f\i) = o

100

heal min

as

102

Mean

value theorems

But worse behaviour than this is possible. The diagram below illustrates a case
where /'(?) = but /has no local maximum, no local minimum, and no point of
inflexion at

Mean

value theorems

/'()

= Kb)

103

-m

b-a

(See the solution to exercise 16.5(6).)

In the diagram, the slope of the

PQ is

chord

f(b)-Ka)

b-a
Thus, for some
to

Theorem (Rolle's theorem) Suppose that /is continuous on


differentiable on (a, b). If /(a) = f(b), then, for some % G (a, b),
11.4

f'Q) =

[a,

/at

G (a, b), the

parallel to

F(x)

where h
(a, b).

is

/()

h
Since

= -

by

+ hx
Then F is continuous on

the constant

+ ha =

[a, b]

f(b)

/(

so that F(a)

[a, b]

and differentiable on

F(b). Then

+ hb

mb-a-m

F satisfies the conditions of Rolle's theorem, F '() =

% G (a, b).

for

some

But then

F'{%)

Proof Since /is continuous on the compact interval [a, b],it follows
from the continuity property (theorem 9.12) that/attains a maximum
at

f(x)

a constant.

We choose

and so
Xa)=f(b)

tangent

PQ.

Proof LetFbe defined on

b] and

0.

/'(?)

is

= f'm + h -

and the theorem follows.

some point
the interval

i in the interval [a, b]

Suppose
follows that

[a,

minimum

attains a

m at some point g

in

and

% 2 are

both endpoints of

m = M and hence /

at ,.

Theorem Suppose

11.7
(a, b).

b].

Suppose that | t

maximum

and

is

is

[a, b]

constant on

not an endpoint of

[a,

b]

Thus, by theorem 11.4, /'(])

[a, b]

Then

= f(b) it then
then /'({) =
for all

Because f(a)

But

G (a, b) and /has

0. Similarly if | 2 is

U-6

Theorem {mean value theorem) Suppose that /is continuous on


(a, b). Then, for some % G (a, b),

(a, b),

on [a,b] and

then /is constant on

differentiable

f'CO

m-m
y-a

But /'() = and so f(y) = /(a) for any y G [a, b\


(Note: This theorem will be invaluable in chapter 13, about integration.
intuitively obvious but,

it

would be very hard

to prove.)

[a, b]

11.8
(1)

It

may

without the mean value theorem (which depends

ultimately on the continuity property),

value theorem

on

[a, b]

seem

Mean

that /is continuous

xG

not an end-

1 1 .5

for each

Proof Let y G (a, b] Then /satisfies the conditions of the mean value
theorem on [a,y] Hence, for some G (a,y),

a local

point of [a,b].
[Note the importance of the continuity property in this proof.]

and differentiable on

lff'(x)

Exercise

Find the stationary points of the function

/:

U ~*

defined by

Mean

104

f(x)

= x(x

of/on
(2)

Mean

value theorems

2). Determine the maximum and minimum values

l)(x

Taylor's theorem

1 1 .9

the interval [0, 3]

Suppose that /and g are continuous on [a, b] differentiable on (a, b)


and that g'(x) = for any x G (a, b). Prove that, for some | S (a, b),
,

Suppose that /is

s'y =

differentiable at the point .

This

is

g(a)

?)/')

at the point %

y=f(x)
is

=/ + (* -)/')

hypotheses of question 2, suppose that /(a)

Deduce that

0.

[Hint:

consider F = f+ hg

a constant.)

is

In addition to the

y = fix)

Cauchy mean value theorem.

the

where h

M) + U-

Then

the equation of the tangent to

g{b)-g{a)-

g'm

(3)

105

value theorems

Thus, for values of x which are

as

to

/() ""(*"" )/'()

is

'close' to , it is reasonable to

a 'good' approximation to/(x).

But

how

expect that

may

large

the

error in this approximation be?

provided that the second limit exists. This

is

L'Hopital's rule. [Hint:

now

Suppose

exercise 8.15(4).]

that /is

times differentiable at the point

Use L'Hopital's rule to evaluate

{y-y/V+y

lira

=x l
that /: U

[Hint: put_y
(4)

Suppose

p(x)

)}

is

/'(*)

= x2

for

x. Prove that f(x)

all

/(?)

+ yr(*

differentiable at every point and that

are the

|ar

4- c,

where c

is

Then the

%.

first

derivatives at the point % of the polynomial

same

- tr'cu +
(x

"

T)i

r"

- ?) /"od +
2

""

/<

1>(l)

as those of/. (See exercise 10.1 1(5).)

a constant.
,->'

(5)

Let /:

U * U be n

times differentiable at every point. Let

polynomial of degree n

<n)

(?)

>

i>

0.

for

We

which

CF-o. !,...)
,

some

|,

a 'very good' approximation to

again,

how

large

may

But

'close' to .

equation

~y

the error be in this

'
i

These questions are answered


Taylor's theorem. If n

value theorem).

dy

Prove that/(x)

mean

1.

for

any x satisfying

<x<

theorem to show that any open subinterval of

which /(|)

/'(.v)

y = fix)

y =

>-

=
g(x)f+y
dx

at

be

1/

approximation?

*(6) Letg-: U~*-R satisfy ^(0) =^(1) = and suppose that/: R->R is
differentiable at every point and y = f(x) is a solution of the differential

will

f(x) for values of x which are

% n are a 'l distinct points. Prove that, for

11

might therefore reasonably expect that

P(x)

mi) = pai)
where |

P be

exercise 9.17(2).]

Then appeal

[Hint:

Use

(0, 1) contains a

to the continuity of /in the

Rolle's

point

One may

to

some extent by

the following version of

theorem reduces to theorem


therefore regard Taylor's theorem

1, the

1 1

.6

(the

as the

mean

Kth order

value theorem.

11.10
Theorem (Taylor's theorem) Suppose that /is n times differentiable on
open
interval
/ which contains the point %. Given any x G I,
an

manner of
fix)

m) + ^ (X - !)/*<D + i (x - oY'm +
+

(-l)!

&-&

-1

f *-%) + E
i

Mean

106

Mean

value theorems

where the error B

but/ (n)(g)

satisfies

107

value theorems

# 0. In

each of the following cases decide whether or not /

maximum or a local minimum at g. In those cases where/


neither a local maximum nor a local minimum discuss its behaviour

has a local
has

close to the point .

n\

/ (g) > and n even.


(
/ "%)>0andHodd.
<n>

for

some

value of 7? between

(i)

andx.

if

(ii)

Proof liy

lies

between x and

g,

put

(iii)/

(iv)/

^(y)

fix) -f(y)

- (x -^)/'(y) - ... - (*~_^"

'

"CO-

t?

between

and

.*

Discuss the behaviour of the functions given below at the point 0.

g,

(v)/(x)

m)

^(x-m

in

\n)

(2)

n\

The function

F has the simplifying property

F'O) =
(

that

/<>(y)

- 0!

The function

= G(x) = 0. It follows from


between x and % for which

has the property that G()


there exists an

77

G'cn)

= nq)+

irtr

(n-l)!

theorem that

*TS)

Rolle's

((x-|) n

rw

Identity (2) follows.

72.77
(1)

Exercise

IfnENandxE
theorem

(1

(2)

in

+ *)" =

R,

use Taylor's theorem to prove the binomial

the form

+nx+

n(n-\)
K

x2

Use Taylor's theorem with n = 3 and = 4 to obtain an approximation


9
to y/5 for which the error is at most 2"
.

(3)

Let /be n times differentiable on an open interval 7 which contains the


point

if

and suppose that D"/is continuous

given

fm

= /" =...=/ <n

"

l)

()

(S)<0 and n even.


()<0andodd.

[Hint: exercise 8.15(6).]

some

(n)

'

Taylor's theorem asserts that, for

<n)

at the point g.

You

are

= x 93

(vi)g(x)

= x 94

(vu)A(x)

= -x 96

109

Monotone functions
Theorem

12.4

Let /be increasing and bounded above on (a, b) with smallest upper
bound L. Then f{x) -* L as x -* b
largest lower bound
(ii) Let /be increasing and bounded below on (a, b) with
-> / as x -*a +.
/. Then f{x)
(i)

MONOTONE FUNCTIONS

Proof We only prove (i). Let


such that, given any x satisfying b 8

>

be given.

We

have to find a 5

>

< x < b.

l/<X>-/.Ke
i.e.

Definitions

12.1

Let /be defined on


only

if,

for each

a set S.

We

say that f increases on the set

x G S and y G S with x <y,

it is

S if and

true that

L-e<f(x)<L + e.
The inequality f(x) < L + 6 is automatically satisfied because L an upper
bound for/on (a, b). Since L - e is not an upper bound for/on (a, b), there
exists a y G (a, b) such that /( y) >L-e. But /increases on (a, b).
is

Therefore, for any


If strict inequality

always holds,

we

say that /is strictly increasing on the set S.

Similar definitions hold for decreasing and strictly decreasing.

A function which is either increasing or decreasing


function which

is

is

called

monotone.

satisfying

y<x<b,
L-e<f(y)<f(x).

both increasing and decreasing must be a constant.

The choice

= b y then

completes

the proof.

Examples The function/:

12.2

increasing on

The function/:

decreasing on (0,

by f(x) =x 3 is strictly
->R defined by f(x) = l/x is strictly

K -*U

(0, )

defined

).

12.5

Corollary Let /be increasing on (a, b). If

G (a, b), then /( ) and

/( +) both exist and

/(*) </(i -) </(?) </(? +) </(y)


provided that a

<x<%<y<b.

/(+>

12.3

Limits of

monotone functions

We introduce
/($-) =

the notation

lim f{x);
x->$-

provided that these limits exist.

which

108

Do not

the value of the function/at

is

/(l +)

/(?+) -

2.

lim /(*)
x-*i+

) and /( +) with/(),
= f(\)but
In example 8.5, /(l ) =

confuse /(

f.

Ptoo/ The function /is bounded above on the interval


By theorem 12.4, the smallest upper bound is/( -). It follows

(a, %)

by /().
any

that, for

fix)

</(*-) </(*).

A similar argument

111

Monotone functions

Monotone functions

110

its

domain / and

its

strictly increasing

is

"' and this


range /. Hence / always has an inverse function /

on J. (Proof?)

for the interval (, 6) yields the other inequalities.

/-?/
monotone functions

12.6

Differentiable

12. 7

Theorem Suppose

thai /is continuous

on

fa, b]

and differentiable on

(a, b).
(i) If f'{x)

>

x G (a,

for each

(ii) If f'(x)

for each

xG

<

for each

xE(a,b), then/is

for each

(a, b),

then /is increasing on

b), then /is strictly increasing

on

x G (a, b), then /is


strictly decreasing

[a,

b] If fix)
.

>

[a, b]

decreasing on [a,b]. If f\x)

<

on [a,b].

Proof Let c and d be


/satisfies the conditions

some

if

a/ty numbers in [a, b] which satisfy c <d. Then


of the mean value theorem on [c,d\ and hence, for

If /'(*)

If fix)
is

>

on

when / is an

interval

and / is continuous on

/,

then J

= /(/) is also

_1

fm = f(d)-m
d-c
increases

In the case

J -* / is continuous on /. This is a
an interval (theorem 9 .9) Observe that /
1
(If/"
had a discontinuity at % G/, then
corollary
12.5.
of
simple consequence

G(c,rf)>

for each

we could

>

and hence f(d) >f(c). Thus /

G (a, b), then /'() >

and hence fid) >/(c). Thus /

find a

X G / for which /(X) would be undefined.)

'

(a, b),

then /'()

[a, b].

>

for each x

strictly increasing

on

[a, b].

Similarly, for the other cases.

'

discontinuous at

if

r (-)<x </-'(+)
1

12.8

We
It

Example Consider the function/:

have f\x)

follows that

R-*-|R defined

Hence f(x) > when x < \ and /'(*)


/ increases on ( , j] and decreases on \\, <).

=\-2x.

/(X) undefined

by f(x) =x(l x).

<

when x

>\

The next theorem

dx

this

usually

remembered

in the

form

(dy\
\dx)

dy
although

is

notation begs a lot of questions.

Theorem Let / and / be intervals and let 7 and J be the correspond/:/->/ is continuous
ing open intervals with the same endpoints. Suppose that
on I and that /=/(/).
If /is differentiable on 1 and
12.10

12.9

Inverse functions

>/(*)

A strictly

increasing function /defines a

1 :1

correspondence between

then

>0

ixEI")

v
f' :J^I exists and

is

"'
continuous on /. Also /

is

differentiable

on J" and

Monotone functions

112

Df~\y) =

>

=y

Df(x)

in

calculated the derivative of the nth root function.


in this case.
can therefore check the validity of the formula of theorem 12.10

We

remains the consideration of its derivative.


/"'(;'

y G/. Then x =/"'(j) S 1. Put k =f~\y + h) -f~\y). Then


+ h) =f-'(y) + k=x + k. Thusj> + h = f(x + k) and it follows that

Since/"
and so k

continuous on J, k

is

&

unless h

We may

as h

* 0.

If

is

1/n

= sr-

n{y Un f-'

_ i y i/ y -i
y y
n

'

function
an arbitrary integer, the rule for differentiating a composite

Also,/"

is

D(y m

strictly increasing

'

= D(y m ) iln = ^()

n
)

Un -m my m-l

^^m/n^-1

=0.

therefore appeal to theorem 8.17(ii) to obtain

+ fQ-r

(.y

(y)

Note that

all

12.12

i
|

f{x

+ k)-f{_xff\x)

(1) Suppose that /increases

Roots

As an application of our observations on

inverse functions,

we now

dis-

Let n be a natural

number and consider

the function /defined

on

[0, )

by

x".

nx"-

>0

Thus/is strictly increasing on


function/" We write
1

f(X)

(x

on

\iln
1

l)

[0,

T (5)

Let /:

* R defined by

_ x l/n

).

be defined by f(x)=

R -* U

function

true that,

(x>0).
and, as in theorem 12.10,

Prove that /

at a.

Prove that the function /: [0,

1
:

R-*R

+ x + x 3 Show
.

that

/ has an
when

Calculate the value of >/"( y)

the interval / and for each a and b in / it is


between f(a) and fib), then a % G / can be found

/ increases on
X

if

such that /(?)

[0, )

minimum

G N.

inverse function/"

y=-\.

Since /(0)

and

interval [a, b]

increasing on an open interval /.


(3) Suppose that /is differentiable and
Prove that f\x) > for each x G /. If /is strictly increasing on /, does
for each x G /? Justify your answer.
it follow that /'(*) >
(4)

= and f(x) -* + oo as x -> + , it follows from the continuity


property (theorem 9.9) that the range of /is also [0, ).
Observe that

at b

on the compact

Let n

decreases

cuss the theory of nth roots.

/'(*)

maximum

attains a

the above results apply equally well to strictly decreasing func(2)

fix)

Exercise

as/z->0.

tions.

12.11

we

(theorem 10.13) then yields

=f( x + k)-y =f(x + k)-f(x).

made

1.9 has therefore been justified.


In exercise 10.15(1),

Proof It follows from theorem 12.7 that /strictly increasing on /and


hence/"' :/-*/ exists. The continuity of/" 1 has already been discussed. There
Let

>

there exists exactly one x


This argument shows that, given any y
u
=
") such that.y
x". Our basic assumption about nth roots
(namely x

(ye J")

= fix).

provided that y

113

Monotone functions

lies

Does the same conclusion hold if the hypothesis that /increases


it

on /.

(see corollary 9.10). Prove that /is continuous


is

abandoned?

has an inverse

t(6)

Let/: [0,

1] -*-(0, )

be continuous on

[0,

l].LetM:

be defined by

M{x) =

sup

^y <x

f(y)

(0<x<l).

Prove that the function

<*>(x)

lim

Mix)
is

continuous

if

and only if/ increases on

[0, 1].

[0, 1] ->(0,)

Mono tone functions

1 14

1 2. 1

Convex functions

Let

/be

12.14

defined on an interval

true that, for each

a>

Monotone functions

and

(3

>

/.

with a

We say that /is convex on / if it


+ =1

Tlieorem Suppose that /is convex

(1)

both exist for each


cave on

/. It is

The geometric
of

we say that /is conconcave on /if and only if /is convex on /.

E.I. If the inequality sign is reversed,

obvious that /is

interpretation

convex function

lies

is

any point on a chord drawn to the graph

that

/.

Then the

is

h _>o-

G/andy

interval

(3

f(ax+py)<af(x)+0f(y)
whenever x

on the open

x/.

<

Proof Suppose that


;c 3 = x + ft 2 in (3). Then

/(x

on or above the graph.

>>-o+

ft,

/(x +

+ ft,) -/(*)

x = x,, x 2 = x +

<

ft

ft

)-/(x)

ft

Put

ft,

and

Hence the function


a/(.v)+i3/0-)

F(ft)

/(O.Y+0V)

increases in

/(x

some

+ ft)-/(x)

interval (0, 6).

The existence of

lim FQi)
h-*a+

/convex

/ concave

then foUows from theorem 12.4.

A similar argument establishes the

existence of

the left hand limit.

sometimes useful to rewrite the definition of a convex function in the


A function /is convex on an interval / if and only if, for all
pointsx,,x 2 and 3 on the interval / satisfying x, <x 2 <x 3
It is

following way.

/(*i)-/(*i)

12.15

/(* 3 )-/(* 2)
(2)

Example The function/:

(Tlie chords

lie

Note however that the two


Equivalently, a function /is convex

points

x lt x 2 andx 3

on an

in the interval satisfying

f(X2)-f(x 1 )

< f(x 3 )-f(x

Xo-X,

interval / if

and only

Xj<x 2 <x 3

if

for

(4) are not

all

U^U

defined by /(x)

\x

is

convex on

above the graph).


limits of

equal when x = 0.

J-

1x1

(3)

x*x.

Theorem Suppose
continuous on /.
12.16

that /is

convex on the open

interval

/.

Then /is

Proof From theorem 12.14,

-v 3

(2) slope

P,P,<

slope

lim

.v,

P,P3

(3) slope

P,P,

<

slope P,P S

{/(* +

ft)

-/(*)}

h->o-

hm

f(x

+ )-/00

ft

'
"h-*0-

lim

ft)

=0.

ft|

= 0.

h->-0-

Similarly,

That (2) and (3) have the same content


substitutions x,

=x,x 2 = ax + (Sy

and x 3

as (1) is easily seen

=y

(see exercise

by making the

2.21(2)).

lim {/(x
h-*0 +

+ ft)-/(x)} -

,.

lim
h-0 +

f(x + h)-f(x )
:

lim
h-*0+

116

Monotone functions

12.17

Example The function/: R->

117

Monotone functions

Examples The functions/,^ and h defined on (0, ) by /(x) -x\


12.20
=
x~ l ate all convex on (0, ) because their second
1/x and h(x) =
g(x)

defined by

x+

(|x|<I)

/(*)

derivatives are

The function

(|x|>l)

non-negative on (0,

4>

Let /:

Example 12.15 shows that

a function

convex on an open interval /


differentiable on /, then the
is

Theorem Let / be an open interval and suppose that /is


Then /is convex on /if and only if Df increases on /.

12.18

Proof (i) Suppose


mean value theorem

first

that

Df increases on /.

- /()>

X-iX-,

< % <x 2 <r\<x 3

Since

Let x,

(2)

(ii)

/(*a)~/(*i)

/(*a)-/(*i )
x 2 -x x

differentiable

<x 2 <x 3

By

the

x->

x,

X,Xi

We

twice differentiable on

/,

by f(x) = (x - l)(x - 2)(x - 3). Find the


which /is (a) increasing, (b) decreasing, (c)
a

graph.

the interval /

and

< /(*3)-/(*i)

<

By

x x x 2 and x 3 be points of /
,

fix 3 )-f(x 2 )

x 3 -x 2

X 3 -Xi

on

G /, prove

/,

that

(xe/).

Interpret this result geometrically.

< x 2 < x 3 < x4

let

Prove that

f(x)-m)>f'(%\x-%)

and therefore

inequality

cave

/(x 4 )-/(x 3)

on

What

result is

obtained

if /is

con-

on/?

Let 0:

U -* R

be a

strictly increasing,

and suppose that 0()

0. If x,

convex, differentiable function

> % and

Xa

x 3 -> x 4

We

= x

12.19

Theorem Let / be an open interval and suppose that /is twice


rentiable on /. Then / is convex on / if and only if

$.

Let

/be

(a

1,2,...)

* so. (This method of obtaining an approxima-

as n

tion to the zero of


(5)

is

differentiable,

called the

Newton-Raphson

process.)

convex and bounded on R. Show that /is a

constant.
diffe"I"

(6)

Let /be continuous on an interval /and satisfy

'x+y

f"(x)>0

/
for each

m+f(y

x e I.
Proof The theorem follows from theorems 12.7 and 12.18.

0(*n)

prove that x

an even simpler characterisation of convexity


+

all

concave on (0, ).

on the open interval /and let J = /(/). If /is strictly


show that/" is concave on J. What happens if /is
strictly decreasing on 11
*(4) Suppose that /is convex and differentiable on the open interval /. If
increasing

results.

for

is

(3) Let /be convex

Ignore the middle term in this inequality and let x -> x , + and
2
obtain /'(x,) </'(* 4 ) and it follows that Df increases on /.
If /is

Let /be convex on

(t?)

Df increases, /'(?) </'(t?)

< /(* ) -f(x 2)

for

which satisfy x, <x 2 <x 3

inequality (2) holds.

Thus /is convex on /.


Suppose next that / is convex on /. Let x ,

of x

convex, (d) concave. Draw

/which

need not be differentiable on /. However, iff is


following theorem provides a useful characterisation of convexity.

where x,

s/x

(*>0).

R -* R be defined

ranges of values

X 2 ~X,

4>(x)

Exercise

(1 )

/.

).

by

(0, )

E%%% <//////&

1 )

12.21

on

on

defined

= -^' 3/2 <0

<p"(x)

convex on the compact interval


[1,1]. Observe that it is not

is

continuous on

all

x E I and y E I. Prove

that, for any

, ,

x2

in the

interval /,

/(

"I < ~ ifiXx)+f{X2)

+*)>.

118

Monotone functions

[Hint: recall the induction argument of example 3.10.] Deduce that/


convex on /. [Hint: establish inequality (1) of 12.13. Begin with the

is

case

when a and

are rational.]

INTEGRATION

13

Area

13.1

Suppose that /is continuous on the compact

Does

it

make

sense to discuss the 'area' under

the graph of/? And, if so,

pute

its

interval [a, b]

how can we com-

value?

makes sense at all to talk about the area under the graph, then presumhave been
ably this area must be at least as big as the areas S] and S 2 which
If it

shaded

in the

diagrams below.

sums of
Let S denote the set of all numbers S which can be obtained as the
under
the
'area'
Then
the
above.
diagrams
the areas of little rectangles as in the
reasonable
to
seems
It
element
of
S.
as
every
graph of /must be at least as big
identify the 'area' under the graph of /with the smallest

number

larger than

i.e. with sup S.


These remarks are intended to motivate the formal mathematical definition

every element of S,

given in the next section.

119

120

Integration

13.2

The

integral

P= [vo, y\,
"

interval [a, h].

We

pro-

f(x)dx

[a, b].

defining a partition

m, =

<

{y Q ,y u

bounded on
.

This

is

a finite set

,y n } of

[a,

all partitions

of

[a, b]

under the graph off. But notice that any area below the x-axis has

as the 'area

b.

to be counted as negative for this interpretation.


[a,

b].It therefore makes

Note

also that, if b

b],to define

f(x)

f(x)dx

> a, we define

= -f*f(x)dx.

ft*)

and so on

Some

3 .3

For each partition P, we can then form the sum

properties of the integral

Much of the power of the


differentiation.

= E^feOft-^-i).
k =

(1)

J>> dx

y,

^"t

interval [a, b]

the property that

-<y n -i<y n =

[a, 6], it is

P=

inf.
yS

P of the

numbers with

real

Since /is continuous on


sense, given a partition

sup5(P)

where the supremum extends over


We wish to think of

= yo<yi<yi<

S(P)

dx

,y n ] of

define
f
Ja

of /over the interval

We begin by

numbers S(P), where P is a partition of [a, b\ is


bounded above by H(b a). Hence it has a smallest upper bound and we may
follows that the set of all

It

Suppose that /is continuous on the compact


pose to give a definition of the integral

f>

121

Integration

to be

worked out

directly

idea of integration

lies in its

connexion with

would be very painful indeed if all integrals had


from the definition. Our first priority is therefore to

For example,

it

we can

prove the theorems which link integration with differentiation. Before

The diagram below

illustrates the case

4.

The value of S(P) is

the area of

do

this,

however, we need to derive a few basic results directly from the

definition.

the shaded region.

13.4

Proposition Let /be continuous on

[a,

b] with

maximum

M and

minimum m. Then

m (b-a)<{

f(x)

dx<M(b-a).

The proof of this proposition

is

essentially contained in 13.2.

We

quote two

simple corollaries.
>'i

>'2

j'j

y*

Suppose that / is bounded above on


r

fa, ft]

13.5
[a,

b]

Corollary If c

by H. Thus /(/) <Hfor any

"b

Then

S(f)

a constant, then

cdx = c(ba).

WfcO'fc-J'fc-i)

13.6
t

<^

is

ZOfe-^-i) = -W{Oi-^o) +
fe

[a, b]

Then, for any and x

+ 0' n -/-:)}
JlfiOdt <k|x-||

= H(y n -y

...

Corollary Let /be continuous on

= H(b-a).

[a, b]

and

in the interval [a, b]

(x*&.

satisfy |/(r)l
,

<k

for

any

122

Integration

The next proposition


ties

123

Integration

of sup and

is

also very simple, provided that

recalls the proper-

fit)dt< j

[a,

b\ and let c be a constant.

Then

s(p*)<j
a

+ c}dt = \
a

{f(t)

The next and

f(t)dt

+ c(b-a).
A

of this section

final result

is

similar

not quite so easy and so we give the

Theorem Let /be continuous on

f(t)dt-S(P2)

[a,

b] and let

mdt-\

f(t)dt.

argument now yields

"
J"c

proof.

13.8

and so

Proposition Let /be continuous on

13. 7

one

2.13(2) and exercise 7.16(6i).)

inf. (See exercise

fit)

dt < " fit) dt

fit) dt

jl

and therefore

a<c< b. Then

b
b

ja

f(t) dt

+j

f(f) dt

a
ja f(t)dt>f

fit) dt.

Proof We use the notation of 13.2. Suppose, in the first place, that
and
Pi
Pi are any partitions of [a, c] and [c, b] respectively. Then the set P of
points in at least one of the sets P, and P2

is

Now let P be any


P by inserting

"

p,

It is

ii

= S(P ) + S(f>2).
t

f(t)dt.

and

partition of [a, b]

the extra point c (if

let

(3)

Q be

the partition obtained from

does not already belong to P).

it

It is easily

seen that

< 5(0.

Let Pi be the partition of

/>,

obvious that S(P)

+j

a partition of [a, b]

S{P)

f(t)dt

[a, c]

and

lie in

[c,b].

let

P2

Moreover,

[a, b]

consisting of those points of

be the partition of

HI

II

11

[c, b]

lie

in

which

],,

Q which

consisting of those points of

~b

S(P)<j

f(t)dt

because of definition (1) of 13.2. Thus, given any partition

P x of

[a, c]

and any

partition/^ of [c,b],

We

have

SiP)<S(Q)
siPi)

+ s(p2) <[/) dt

=SiPd+Sm
<j fit)dt + j*
a

S(Pd<j*f(t)dt-S(P2 ).

(2)

fit)dt.

The right hand side of this inequality is therefore an upper bound for the
Hence
all numbers S(P) where P is a partition of [a, b]

set

of

Hence, given any partition

bound

for the set of

[a, c]

Therefore

sup5(P,)<

all

P2

of

hand side of (2) is an upper


numbers of the form S(P{) where />, is a partition of
[c,

b], the right

f(t)dt-S(P2)

where the supremum extends over all partitions


definition of an integral, we obtain

sup S(P)

.e

ja

fit) dt

Combining
t

of

[a, c]

Recalling the

< fa f(t) dt + jc

(3)

<

and

f(t) dt

ja

fit) dt

(4),

we

fit) dt.

obtain the conclusion of the theorem.

(4)

124

Integration

13.9

Differentiation

Since /is continuous, one feels that, if h

and integration

Let /be defined on (a, b). Suppose that


differentiable

on

(a,

F\x) =

/on

[a,

and

b]

F(x

fix)

shall say that

= x3

then the function

[a,

F(x)

F defined

by F(x)

= Jx" +

+c

very small, the shaded region will

Thus

hf(x)

(5)

fix).

dubious on a number of counts - particularly at the stage


where we divide by the small number h to obtain (5). In the next theorem we
provide a precise proof of the result.

on

[a, b]

+ c is a
[a,

b) and, for each

b]

primitive for /on [a, 6]


.

Then

F-G

is

follows that

Suppose

continuous on

[a,

FG

b],

Then F is
able

constant on

[a,

F is defined

{a<x< b).

f{t) dt

on

(a,

a primitive of /on [a, b],

b) and

F\x) = f(x)

i.e.

for each

Proof Since /is continuous on

0.

that \f{t)\<K
is

and that

there-

G (a, b),

D{F(x)-G(x)} = F'(x)-G'(x) = /(*)-/(*) =


1 1 .7 it

[a, b]

by
F(x) m f*

(a,

From theorem

is

Tlieorem Suppose that /is continuous on

75.72

b]

G is a primitive for /on

differentiable

is

F'(x)=f(x).
This argument

on

Proof Obviously F(x)


fore that

is

area will be approximately hf{x).

only approximate, but the smaller h becomes the better the


approximation should be. Allowing h -* 0, we might therefore expect that
This equation

G(x)

xG

its

a primitive (or anti-derivative) of

is

Theorem Let fbe a primitive for/on [a, b] and let G be defined on


Then G is a primitive for/on [a, b] if and only if, for some constant c,

for each

+ h)-F(x) =

F(x+h)-F(x)

Example lff(x)
primitive for/

[a, b]

continuous on

6].

[a,

75.77

is

a rectangle and hence

and

13. 1
is

'nearly' be

b) and satisfies

x G (a, b). Then we

for each

125

Integration

b] Hence the
.

By theorem

F is continuous

xG

on

[a, b], differenti-

(a, b).

[a, b]

it is

bounded on

[a, b]

Suppose

(a<t<b).
1

.8,

for any

x and

% in

[a,

b]

result.

[a,

F(x)-F =

What have primitives to do with integration? Suppose that/is continuous on


b] Then we may define a function Fon fa, b] by

}*/(') dt

and therefore, by corollary 13.6,

%)

J*

f4t

In the next theorem,


first

we

we

(a<x<b).

shall

give a rough, intuitive

show

that

\F(x)-F()\<k\x-$\.
is

a primitive

of/ on

[a, 6]

But

argument to indicate why one might suppose

to be true in the first place. Given

x G (a,

b),

we

seek to explain

this

why one might

And, from this inequality, it follows that F is continuous on [a, b]


A more subtle argument is needed to show that F is differentiable on
If x and | are in (a, b) and x = %, then

suppose that F'(x) -f(x).

The number F(x) may be interpreted geometrically as the 'area under the
graph of /between a and x'. The shaded region in the diagram below should
therefore have area F(x + h)F(x).

/(.v

h)-F{x)

The
x x +h

ft

last step is justified

by proposition 13.7 withe =/().

(a, b).

126

Integration

Let

some

127

Integration

e>0 be given. If G (a, b), then /is continuous at the point % and so, for
> 0,

for

some constant

Hence

c.

provided that
1/(0 /(I)

| < 5.

\t

< e for every

and*. From corollary

/(f) dt

<

1/(0-/(91

{F(b)

Hence, provided that \x

13.6,

we conclude

compact

value of t in the

F(x)-F

[F(fl)

c}

%\

< 5,

it

= G(b)-G{a).

follows that

interval with endpoints

that

Example Since

13.15

-/(I)

x-%

-SI

I*

D -

(1

; )

2 2
r )

'

.elx-il
it

2(1

<

follows from theorem 13.14 that

:dt

< \x |< 5. But

provided that

+ c}-

what

this is

is

meant by

i;

the assertion

o.+ff

2(1

f)

F(x)-FQ)
/($)as*-*{.

*-i
We
each |

Riemann

13.16

have therefore shown that

F is

differentiable

on

(a,

b) and F'(|)

= /() for

The

(a, 6).

/.
75.

75

dt

c l+t

100
1

+x 100

'

i5.i4

Theorem Any function /which

on

[a, b]

If

|*

is

integral

/w*

(6)

was defined as the supremum of all areas like that shaded in the left hand diagram below. It would be equally sensible to define the integral as the infinum of
all areas like that shaded in the right hand diagram below.

Example

integral

is

continuous on

[a,

b] has a primitive

any primitive of/, then

/MA

Proof Let
F(x) = f

= G(b)-G(a) =
Fbe

defined on

/(r) dt

(a

<

[G(t)] a

[a, b]

a:

<

by
If /is continuous

6).

reason

is

simple.

on

The

[a, b]

these

two

same result. The


would yield the result

definitions yield the

use of the second definition

Then

-j"

= F(b) = F(b)-F(a).
Ja /(/)A
By theorem
other primitive

G(x)

13.12,

Fis

a primitive

of/ on

G satisfies
=

F(x)

+c

(a<x<b)

[a, 6]

and, by theorem 13.11, any

(7)

-f(x)dx.

But (6) and (7) are equal because

is a

primitive for

is

a primitive

for/ on

[a, b] if

and only

if

/.

If /is not continuous on [a, b]

the same result. But, //they

do

the

two

yield the

definitions

same

result,

do not necessarily

we

say that /is

yield

Riemann

128

Integration

integrable and call the

common number

obtained from the two definitions the

13.18

integral of /on [a, b]

Riemann

(1)

Example We know

13.17

Exercise

Prove that, given any natural number ,

Show
a

i.
(2)

illustrate the idea discussed in

13.16,

we

evaluate the integral without

J>" dx

that

X *dx = [W]h =
JJ
To

129

Integration

B-f

that the
1

same

result holds if??

<a <

provided that

Since the integrand

is

Jo
(3)

If

(x-l?

x-l

is

a positive rational

- -ir{l

Hffl

"

13.19

More

{0, \/n, 2/n, 3/n,

hand diagram is smaller than the


hand diagram is larger. Thus

area of the left

1}

integral

of

[0, 1]

The shaded

and the shaded area of

= -1-1 = -2.

number, prove

that

+ 3 Q + ...+"} =

a+

properties of the integral

In 13.3
is

the right

number

must be wrong.

a square, the following result

ax

Pn =

replaced by any rational

Explain why.

using the theory of differentiation.

Introduce the partition

is

b.

we mentioned some

particularly important but its proof

properties of the integral.

Theorem 13.8

which we obtained directly from the


used the properties examined in 13.3 to

was long and laborious. We


prove the theorems which connect integration with differentiation. Having
proved these theorems, we can now obtain the rest of the properties of the
definition

integral very easily.

fit \

Jo

tx

\n]

13.20

2
2
2
2
^{0 +l +... + (-l) }<^ x dx< L{ l 2 +

- .-(n-l)n(2n-l)<

f '

ju

x 2 dx

<- .-( +

no
3

Theorem Suppose that /and g are continuous on


Then

+...+*}

l)(2#i

\" {x/(o

+ ng(t)} dt =

Proof Let

F and G be

theorem 10.9(i) the function

and that X and

\a

f{t) dt

primitives

+ n j*

of/ and g

H = XF + jdG

is

dt.

respectively on [a, b]

a primitive of

X/+ fig on

[a,

By
b]

Hence, by theorem 13.14,

fe<f ll+^l ll+

"
f

Considering the limit as n

1)

(exercise 3.1 1(1 i))

[a, b]

are any real numbers.

{X/(0 + pgffi dt

* yields

+ M G(r)] b

[XF(t)

HF(t)]Z + n[G(t)] a

= \f*f(t)dt + vj
a
13.21
[a, b]

Theorem

g(t)dt.

by parts) Suppose that / and g are continuous on


and G respectively on [a, b] Then

(integration

and have primitives

130

f(t)G(,)dt

\"

Ja

[F(r)G(r)]2-

Proof Let
F(t)g(t)dt.

DH(t)

Ja

[a,

Proof By theorem 10.9(h),

D(FG) =
and thus

b] Thus Hib)
.

+ Fg on

[a,

b\

+ F(t)g(t)}dl =

< t < b).

[a, b)

By theorem

Then

2.1,11 therefore increases on

> //(a) and hence


= Hih)-Hia)>0.

Hence
that /is continuous

Theorem Suppose

13.24
{f(t)G(t)

(a

{git)- fit)} dt

Ja

of /G

H be a primitive of g /on

- g(t) f{t) >

fG+Fg

fG is a primitive

131

Integration

Integration

[F(f)G'(/)]*

on

[a,

b].

Then

fit)dt<[*

1/(01 dr.

and the theorem follows.

As

is

from

clear

its

proof, the formula for integrating

integral analogue of the

Proof Since

by

parts

is

formula for differentiating a product. The next result

the integral analogue of the formula for differentiating a composite function,


a 'function

and

is

of a function'.

usually

dt

remembered
dt

du

It is

< fit) < 1/(01 (a<t< b),

|/(/)|

[* |/(0I dt

i.e.

form

and hence the

result.

< J"*

fit) dt

(Note that

|/|

du.

<

Theorem (Cauchy-Schwarz
tinuous on [a, b] Then
13.25

it

follows that

is

the rule for changing the variable in an integral

in the

just the

1/(01 dt

2 2
{/ }" and

is

therefore continuous.)

inequality) Suppose that / and

g are

con-

Theorem Suppose that has a derivative which is continuous on [a, b]


and that /is continuous on an open interval which contains the image of [a, b]
under 0. Then
13.22

</>

Proof This
"

0<J

is

much

the

same

as that of

example 1.11. For any x,

{xfit)+git)fdt

fl

Proof Let

= x2 f
JO

{fit)} dt

= Ax 2 + 2Bx +
From theorem

10.13 and theorem 13.12,

d
F(0())

du
It

F'(<p(u))<p'(u)

f(<p(u))<p'(u)

du

is

what had

Ax 2 +2Bx + C =

cannot have two (distinct)

Exercise

to be proved.

13.23
Theorem Suppose that /and g
any te [a, b],f(t)< g(t). Then

are continuous

on

[a, b]

and that, for

Suppose that g

is

continuous on

[a, b]

and thatg(0

Prove that

\\it)dt =

f(t)dt<fg(t)dt.
Ja

{git)} dt

[F(0(W ))]

(1)

J(a

I*"

JO

as required.

13.26
and this

B2 <AC

follows from theorem 13.14 that

ja

fit)g(t) dt

JQ

C.

Thus the quadratic equation


roots. Hence

/(0(W ))0'(W ).

+ 2x\"

if

and only

if git)

for each

[a, b]

>0(<f <S).

132

Integration
(2) Suppose that /is twice differentiable

ous on

[a, b]

on

fa, b]

and that/"

is

continu-

Prove that

f(x)dx

*(3) Let /be positive and continuous on

[1

).

Suppose

provided that the limit

P~

that

\"

lim

y-*a* Jy

J-i-a

= {bf\b)-f{b)}-{af(a)-f{a)}.

xf"(x)dx

ja

133

Integration

exists.

f(x)dx

Note

f(x)dx

f(x)dx

provided that the appropriate limits exist,

+ and -

limits to recede to

Prove that /(.v)


ft {F(t)}-

>

[Hint

The

the improper

integral

g are

continuous on

[a, b]

lim

>

and that g(t)

>

some

[a,

with theorem
*(5) Suppose that
b]

and

its

(a<t<b).

1 1

b]

[Hint: continuity property.]

What has

is

[a,

We do not

define

by

f(x) dx.

[a,

b] with f'(t)

On

>

dx

f(fl)

g(t) dt

ja

+ f{b) J*

Improper

-a. -? r
1

x->>

1-

0<y< 1,
x

l-+co a sy->0 +

=
y

and hence the improper integral

does not

exist.

inte1

(ii)f

^=

lim

form

dt

Hm (2-2V>') =

lim[2x" 2 ]i=

f*-

2.

J?

provided that/ (,,)


in

hm

g(t) dt

.10) can be expressed in the

&=w fr-o*- /^)

s1 =

x->~Ji

Jy X 2

lim

the other hand, if

Prove that

term given

x
rx dx

dx

b] and that /is differentiable on

continuous on

Taylor's theorem (theorem

13.27

JX

do

this to

some G [a, b] What happens if f'(t) <0(a<t<b)l [Hint:


grate by parts.]
By integrating many times by parts, show that the error term in

DO

o>r

derivative

for

we allow the upper and lower

Examples

.6?

g is continuous on

/ f(t)g(t) dt

(6)

i.e.

respectively independently.

~r g(t)dt
13.28

{a,

P~ f{x)dx

b). Prove that

la mg<?)dt
for

1/2

*(4) Suppose that /and

(a<r<

>

(x
(x
)
F'(t) dt is relevant.]

integral

continuous everywhere, we define

that, if/is

is continuous on /. Show how the form of the error


theorem 11.10 can be obtained from that above.

,/

'

dx

V*

^777

integrals

We
on

[0, ).

down some sample


Then we define
write

definitions.

Suppose that /is continuous

example. Observe that / ~ x dx does not exist


(why not?). Hence fjT.I * <& cannot exist even though
(iii)

Note

finally the following

j;

/w^

iimjo

/w^

provided that the limit exists. Similarly,


define

if /is

|**xdx =
continuous on

(a, b]

then

U -^
2

^0asAT^ + .

we
they are not

The integrals defined in this section are called improper because


to include the
given by the definition of 13.2. It is perhaps pedantic

little

134

Integration

arrows

in the

135

Integration

notation for improper integrals and most authors would simply

who

propose to study integration more deeply, it


probably better not to do without the little arrows since this will avoid condelete them. But for those

a +1

is

zm-j**

-a =

fusion with the differing definitions of the powerful Lebesque theory of integra-

f(n

1)

fix)**)-

/<*>-JT

**

fix) dx

'

</(

4-

1)

-f(n +

1)

J""

tion.

Proposition Suppose that

13.29

negative on the open interval

/.

is

a function

Suppose

which

and so <A) decreases. Also


is

continuous and non-

also that /is continuous

on /and

satis-

fies

\m\<m
If the

This
series

on / a

(xei).

is,

of course, an analogue of the comparison

test

of monotone functions.

Since /

"*~

),

then the

series

>

reason

is

continuous and decreas-

if /is positive,

and the improper

integral

C~ ftx)dx

A) and

Jl

The theorem therefore provides

criterion for the convergence of a series or an

often referred to as the integral

improper integral and for

this

test.

+X'2^32-

13.33

Example Take fix) = x"

in

theorem 13.32.

.-22

[1

follows that,

it

either both converge or else both diverge.

to observe that, for any

dx
1

on

n=l

r + x2

we simply have

bounded below.

is

From theorem 13.32

the existence of the integral

dx

f-oo

Ji

and so <A n >

ing

Example To prove

/()> o

(theorem 6.15), for

may be proved in much the same way. If/ happens to be non-negative


simpler proof may be based on theorem 12.4 concerning the convergence

and

13.30

> /(*)- I A*) =


fe=i
fc=i

improper integral of <p over the interval /exists, then so does that of/.

dx

exists, the result

6.5, that the series S"=1 n~ diverges to +.

then follows from proposition 13.29.

We know from theorem

From theorem

13.32,

we conclude

that
r

13.31

Euler-Madaurin summation formula

We
13.32

Theorem Let /be continuous,


by

positive

and decreasing on

[1

).

is

decreasing and

f(k)-C

13.34

f(x)dx

Proof Since /decreases,


ft

/(*

Thus

zero.

for

any

it

k~ 1, 2,

improper integral /

dx

-i

converges.
.

-* .

x" 1 dx does not

,..,

(ll)

V^7)

V(l -x)

(1
*i

3/2

(1

^^*
+x

Jo

dx

x 2 dx

f-"

3 2
)

-x}

[Hint: for (v)

and

(2) Prove that, for

(vi) recall

0< y <\,

dx
(vi)

(v)
o

il-x)

+ !)< J/ f(x)dx<f(k)

exist.

Discuss the existence of the following improper integrals.

W Jor
Hence

Exercise

...

bounded below by

as

Then

(1)

An -

In particular, the

discuss only a simplified version of this useful result.

the sequence <A> defined

"dx

example 13.33.]

+ X + x1

+ x+x s

J^o X

dx

136

Integration

r(i'2)+y

J(l/2)-y

\,
X)
-dx =
2x

0.

x(l

*i
Discuss the existence of the improper integral
(3)

Let

a be

a rational

number with a >

In

2x-

J-, o

theorem 6.6

it

dx.

14

EXPONENTIAL AND LOGARITHM

14.1

Logarithm

was shown

that the series


oo

converges. Base another proof of this result on theorem 13.32.

We define

the logarithm for each

>

by

*
fr* dt

logx=J

It

follows immediately from the definition that log

and

that, for each

x>0,
Dlogx = -.
x

Also,

D 2 \ogx =
We conclude

=.

that the logarithm

is strictly

increasing and concave

on

(0, ).

Further, in view of example 13.33 and exercise 13.34 (lvi),


log x -*

and
log

+ as x -* +

as x

-*

+.

(For an alternative proof of these


information the graphs

= log x

results, see exercise

14.3(1) below.) With this

can easily be drawn.

The number e is defined by the equation

137

Exponential and logarithm

138

log e

139

Exponential and logarithm

logx^y.
The value of e

approximately 2-718 (see exercise 15.6(4)).

is

> 0, y >

Theorem Suppose that x


\ogxy = log* + log .y

14.2
(i)

logOO =

(ii)

Proof
f(x)

rational.

is

Given a positive rational number


r
(i)x" logx ->0 asx -> + <

Then

a fixed value

of y

> 0, consider the

prove that

r,

-0 +.

->-0 as>>

(These results are usually remembered by saying 'powers drown


s < r when proving (i).]
logarithms'.) [ Hint take

function

<

"(3)

logy

iii)y

r]ogx.

For

(i)

and r

lOgATJ-logJC.

Show

that

= x log x - x is a

Fix)

primitive for log x

on

(0, ).

Deduce

that

We

have

xxx

xy
for

each* >0. Thus /is constant (theorem


log

To

xy\ogx =

obtain the value of c, put

log y

-log

We

log(* r )-rlogJC.

Hence /is

is

little

that the equation log

/-log

(x>0)
and only

if

-* e as

x=

Then

...

log n -

-I

are discussing here


'In'

rather than

is

In pure mathematics,

> 0.

(2) Prove that, for each


1

as

2,

.).

Prove that

* .

{(log x)

Hence show

>

(6) Evaluate

By considering

+ as x -* + and

number and x

the 'natural logarithm'.

'log'.

Exercise

-*

(Euler's constant) with the pro-

show that

(ii)i){loglogx}

number y

[Hint: theorem 13.32.] Hence

(i) Z)

Prove that log 2

0.

familiar logarithmic tables.

log

= e.

- .

1-

1 H

occasion arises for the use of 'logarithms to the base 10' as found in the

(1)

if x

> e and x n+1 = e log x

c.

perty that

sometimes stressed by writing

14.3

x-ax has solutions if and only if ae <

*(5) Prove the existence of a real

Note The logarithm we


This

and

)-r\ogx =

log

as n -* .

that

Suppose that x x

0.

obtain the value of c, put


c

Show
Show

with equality

-'--

a constant

\ogix

To

- .rx

log

elogx<x

have

f\x) =

w!

[Hint: recall example 13.17.]

logj\

Consider the function

fix)

(2)!

"(4)
(ii)

2-1.

C
1

log-'

1. Then

2 log

Hence show that

11.7) and

(x> 0).

log(l+*)dx =

that log

log 2" and log 2~",

> 0, logy <.y

deduce that

-*
1

as x -*
.

If x

is

show

that

(JC

>

and s

rational)

(X>1).

that the series

+.

a positive rational
converges

if r

>

and diverges

if

< 1. (See theorems 6.5 and 6.6 for

comparison.) [Hint: theorem 13.32 again.]

140

Exponential and logarithm

14.4

Exponential

From theorem 12.10 we may deduce

We call

lim

and

only

x = \ogy.

if

and range

H(x)

Since the logarithm function has domain (0, ) and range R, the exponential

function has domain

....

(ii)

log(l

..

hm

is

+x)

and continuous on

positive

x
[0, )

and that

h {x)>H(x){x>Q\ where

write

y = exp* if and

x-0

Suppose that h

(4)

x-*0

the existence of an inverse func-

this inverse function the exponential function

expx-1

,.

(i)

tion to the logarithm.

141

Exponential and logarithm

h{t)dt.
Jo

(0, ).

Prove that, for any x

> 0,

h(x)>expx.
y exp

.v

Determine the range of values of x for which the function /:

(5)

by

defined

= exp{18* 3 }

/(*)

(a) convex, (b) concave.

is

continuous and increasing on [0,

If /is

(6)

Observe that the exponential function

increasing

is strictly

on R and that

f{x)dx<

k%

Jo

exp x-t

+ coasx-^ +

exp x

as

c*

show

f(x)dx.

n < log n\ < (n +

prove that

Deduce that n log n


-*

["* i

/<

),

1) log (n

1)

- n and hence

that

Also

D exp x =

D\ogy

\/y

Thus the exponential function

Powers

(1) If x andjc are real

numbers and r is rational, prove that


(i)exp(x+,y) = (exp x)(exp y)
r
(ii) exp (rx) = (exp x)
[Hint: These may be deduced from theorem 14.2. Alternatively, they
proved directly. In the case of

(i),

for

example, by differentiat-

ing

If

>
a

(2) Prove that, for any rational

x~ r exp x-^
r
(ii) x exp x -

But does

number

r,

is

-*

is

(Roughly speaking, 'exponentials drown powers'.)


(3) Use L'Hopital's rule (exercise

.8(3)) to prove that

r
.

We

exponents.

define

with our old definition in the case when x

exp {log

defined by log e

exp {x log

which explains the

we

from theorem 14.2(h)

{/-log a}

+ ooasx-*-!- 00
as

real,

this definition agree

the expression a

this definition to include non-rational

rational? It follows

we have defined

rational, then

is

exp (xloga).

Recall that e

with respect to x.]

(i)

and x
x

exp

+ y)/(exp x)

and r

propose to extend

exp (x

>

If a

Exercise

may be

derivative.

14.6

14.5

n!

n\

own

is its

< exp n <

exp*.

e}

(a )}

that, if r

is

rational, then

is

r
.

Equivalently e
,

= exp

Observe that

exp x

familiar notation for the exponential function.

142

Exponential and logarithm

14.7

Exercise
(1) If a and

(i)

are positive, prove the following.

x *y

a" =

(iii)

a ay

(ab)

(ii)

(a

(iv)

(2) Prove that the derivative of a

a b

x x

15

POWER SERIES

15.1

Interval of convergence

xy
.

with respect to

is

(log

a)a x

(3) Write

+ -|

=exP

n log

+-

A power series

about the point

% is

an expression of the form

and hence show that

lim

e"
in

(see

example 4.19). [Hint: theorem 8.9 and exercise

,/n

is a

variable.

14.5(3ii).]

Theorem The

15.2

(4) Write

which x

of values of x for which a power

set

series

exp -\ % n

and hence show

that

n l,n

*1

as

-* (see exercises

4.20(6) and

converges

5.7(1)).

is

an interval with midpoint

We

Proof

(5) Discuss the existence of the following improper integrals.

then

it

shall

converges for

all

show that, if the power series converges when x =y,


x satisfying \x$\<\y%\. The theorem then

follows.

dx
J-f-oo

J-+0

+X

;dx.
2

[Hint: proposition 13.29.]


'

(6) Obtain

all

/(x

*r

f(x) +f(y)

(X

6 U,y GR)
If the

(ii)

5-

solutions of the functional equations

f{x +y)

(i)

Ir-fl

I.v-i

<

+y)=

fix) f(y)

(iii)

/(xy)

(iv)

f(xy)

= f(x)f(y)

/(at)

+ /(y)

(X R,y S R )
(x > 0, ? > 0)

(x

power
a n (y

%)"

\a n

for

zero.

that \x

I*

(theorem 6.9).

)">

n
{y-%) \<H

Now suppose

when x =y, then

as n ->

-*

Hence the sequence <a n (j'

> 0, y > 0)

which the derivative of/ exists and is continuous wherever /is


defined. For (ii) and (iv) you may assume that /never takes the value

series converges

in

is

bounded (theorem 4.25). Thus,

for

some H,

1,2,...).

1|< y f

I.

Then

-
<1.

y-ll

Hence
\a n (x-t-)"\

= l^^-O^.P^-^P"

The convergence of the power

1,2,...).

series

143

Power series

144

fl(jf-0"

[-1,0
:Ti

now

We

call

+ R, we
is

ofx

the set of values


If the

call

convergence

gence

which

power

series

converges

endpoints of the interval of convergence are

the set

is

for

its

interval

and

of convergence of the power series. (If the interval of


U of all real numbers, we say that the radius of conver1

5.2 shows that a

power series converges absolutely

conditionally or
It is

it

may

limsup

interval / containing the point

point

Then

g.

|a|

is

given

we choose

as often as

in

some open

Taylor series expansion about the

its

is

diverge.

sometimes useful to observe that the radius of convergence

-=

series

Suppose that /can be differentiated

at

of convergence with the possible exception of the endpoints. At the endpoints the series may converge absolutely or it may converge
interval

its

Taylor

15.4

infinite.)

points of

[-1,1]

of

the radius

The proof of theorem


all

follows from the comparison test (theorem 6.15) since 2=oP" converges.

convergence.
|

145

Power series

f ^-j^/ (B) tt) =

by

fe=W*=* /+...

my

1/n

should not be automatically assumed that this power series converges with
sum/(x). Indeed there is no reason, in general, to suppose that it converges at
It

and

lim

all

by propositions 6.17 and


priate conventions must be adopted if the right hand
or

are justified

oscillate. If this

becomes

when x =

%) and, even if

6.18. In both cases, approside of the formula

In the second case, of course, the sequence <|


n+1 \j\a

happens
|>

may

it

does converge,

sum need not be f(x)


sum of its Taylor series

its

is
1 5 .6(6) below). A
expansion in some open interval containing is said to be analytic at the point .
The most natural way of showing that a function is analytic is to prove that
* <*>.
the error term in Taylor's theorem (theorem 11.10) tends to zero as n

function / which

the

paTfeularly easy to write

down

(see exercise

The formulae
to be

(except

happens, the limit does not exist and so the second formula

useless.

Example

75.5

It is

the Taylor series expan-

sion of the exponential function about the point 0. If f(x)

Example We

15.3

list some power series about the point 0, together with


of convergence. In each case, the radius of convergence can be
calculated using the second of the formulae above. In the last three cases one is

f \x) = exp x.
in

their intervals

then

with the problem of whether or not the endpoints of the interval of


convergence belong to the interval of convergence.

Since exp

i ,

the expansion

= exp x, then

is

left

Power series

Interval of convergence

From examples 1 5.3 we know that this power


sum e x l From Taylor's theorem we know that
*

I
n=o

++
1!

where

I
n=0

inx)

Z* n

{0}

77

lies

X*-1
.

(h-

2!

1)!

converges for

and x. Hence

between

x2

series

x
x

+
1!

2!

4-

.+

*~]
(

x"
n\

l)!j

(-1,1)

n\

as

-*

all

x.

Is its

147

Power series

Power series

146

(by exercise 4.20(4) or by theorem 6.9). The partial sums of the power series
x
therefore converge to e
and thus, for all values of x.
,

+ +.

++

2!

1!

and hence show


.

that 2-7083

< e < 2-7184.

*(5) Prove that e

3!

are natural

is

irrational. [Hint:

numbers and seek

assume that e

= mjn, where m

and n

contradiction using the inequality of

question 4.]

15.6

Exercise
(1)

*(6) Let/: |R-*R be defined by

Determine the intervals of convergence of the following power

series

,-l/x"

fix)

00

nix"

OH)

n=0

(-1)"

(2)

C-I)

n
:

(2

(y)

1)!

n2 n
(!)
^rr.x

&

(vi)

(2b)!

n=l

Prove that the Taylor series expansion of log

(1

+ x)

(x

[0

0).

(2>!

Show
0v)

(**0)

=
that, for

/<">(*)

V"

about the point

where

is

= Pn

Pis, a

any x J2 0,

-l/x'

polynomial of degree 3m. [Hint: use induction.] Using

exercise 14.5(2), deduce that, for each natural

I (- ir
n=l

Using Taylor's theorem


that this

power

give another

+X

X*

= x

series

...

^-*0asx--0+.

form given by exercise 13.26(6) show

converges to log (1

x) for

<x <
1

Hence show

Hence

proof of the identity

Write

that

/can be
(n)

and/
down the

the point

(0)

many

differentiated as

(n

0,

2,

times as

we choose

this

at

.).

Taylor series expansion of/ about the point

what values of jc does

= l-i + i-|+...

log2

n,

in the

number

0.

For

converge to/(x)?

(see exercise 14.3(5)).

(3)

Prove that the application of Taylor's theorem in the form of theorem


11.10 (with | = 0) to the function log (1 + x) yields a remainder term
of the form

(-1)'

15.7

Continuity and differentiation

The following proposition is useful. Its proof has been placed in the
appendix since it is somewhat involved. The reader will encounter the proof
again at a later stage when studying 'uniform convergence'.

1+7,
15.8

where

r?

lies

between

show that En -*
manner in which

and x. For what values of x

as n -*
77

<x>

is it

without further information about the

depends on x and

Proposition Suppose that the power series

possible to

f( X )

1 4&-r
n=0

;i?

(4) Prove that

has interval of convergence


1

1)!

Deduce that

...<
2)!

(n

1)!

'

+
+

on / (except

at the

/.

Then

its

endpoints). Moreover

fix) =

sum

mx-Qr*.

is

continuous on / and differentiable

148

Power series

15.9

Example Proposition 15.8 allows

us to tackle exercise 15.6(2) in an


(1

way.

alternative

We know

149

Power series

+ x) a =

a (a

+ ax

1)

..-

x2 +

2!

that the

power

series

provided that

<

(general binomial theorem).

(2) Prove that the power series

I(-l)"-'"

n =i

(
on (

has interval of convergence


1

5.8,

it is

differentiable

1 ,

(see

1
]

Hence, by proposition

5.3).

2n

n-x

n=0

and

1)

1 ,

Example

converges for

K-D"-

(-

-x+x x

+x

It

follows that, for each

x 6 (

(3)

n-ln-l
1
I)""
*

..

<

Suppose that the power

f(X)

anx

its

sum.

series

n=0

has interval of convergence

(-!<*<!).

/. If

y E I but

not an endpoint of/,

is

prove that

),

\jMdx= n=0
1 n-Shy*.

Jo

D log(l+*)- E

and determine

l+x

l+x

4-

(-!)"-

(4) Prove that

and therefore that

log

-x)

dx.

log(l+.x)=

(-I)""

n=0 n

+c (-KK1)
(5)

where c
stant c

is

a constant (theorem

1.7).

By

substituting

x =

0,

we

Suppose that the two power

see that the con-

0.

Since the power series

is

=0

n=0

continuous on (

1, 1],

converge
log 2

1ob(1+jc)

lim

lim
x ->i-

(-!)-n

n=l

(-ir

in

some open interval /containing


x G / if and only if

(6)

m bn (n m

0, 1,2,

n=o

(1)

Exercise
Let

a be

DL

converges for
a real

+x)

number. By considering

(*-. ..<a-n +

dy
l)

dx

.).

series

all real

that a n+l

< l.show that


f{x)

x and

satisfies the differential

equation

y.

Show
for \x

Prove that their sums are

On**
s

1=0

Suppose that the power

y = fix) =
15.10

\.

equal for each

an

series

a e*.

a
l

n +

n (n

= 0,

2,

.)

and deduce that

151

Trigonometric functions

16 TRIGONOMETRIC FUNCTIONS

16.1

Introduction

based our definitions of the exponential and logarithm functions on

We
the formula

Sine and cosine

16.2
* dt
f

We

Suppose that the power

could equally well have begun with the differential equation

t **"

/(*)

series

n=0

%-y
dx

converges for

and defined the exponential function as the sum of a power

In a similar

way, the definitions of the trigonometric functions can be based

on the formula

We

prefer,

-r
Jo

+y =

]+t

why

+...+

3a 3 x 2

+...+

a2x

fix) = 2a 2 +

2a 3 x

+4

3a A x 2 +

have some intuitive ideas about the sine and cosine functions from
It is

2)in

l)a n+2

This recurrence relation

is

150

+ (n + 2)(n +

= cosx =

dx
dy

l)a nti x

l)a n + 2 x"

may be

is

measured in radians and.y

sin x =
yv = sinx

0,1,2,...).

and we obtain
,,2n+l

+8
ei)
'*I<V+>'
n=0
>*e* 'S

= sin x

and z

= cos x.
we

We want sin

and

D sin = cos =

dz

sin*

x 2n+l
(-1)"

n=0

dx

(2n

^2"

dy

''

dx 2

"

dx

z.

-<-"

We

therefore define

=;t _^! +
*
3!

l)!

Similarly

d 2z
y

2n

equation.

follows that

?!

...

chosen in any

satisfy

this differential

an

readily solved

obtain

d 2y _ dz
dx 2 ~ dx

+...

therefore sensible to begin by indi-

the sort of argument considered adequate in elementary trigonometry,

it

Thus, to satisfy the differential equation, a Q and a,

the sine and cosine functions, as conceived of in trigonometry,

ich

in

0.

In the diagram, the angle

anx

way, but then the remaining coefficients must be chosen to

should be expected to satisfy

By

x+
2a 2 x+
at

!+

in

We already

o.

15.8,

fix) =

elementary geometry and trigonometry.


cating

of x and that

'

however, to base them on the differential equation

dx*

By proposition
f(x)

dt
arctan

/"(*)+/(*) =

series as indicated in

exercise 15.10(6).

real values

all

152

153

Trigonometric functions

Trigonometric functions

Both these power series converge for all values of x by comparison with the
x
power series for e
Note that we take these formulae as definitions. The definitions of elementary
trigonometry concerning the 'opposite', the 'adjacent' and the 'hypotenuse' are
not precise enough for our purposes in this book. The remarks of 16.1 are
simply intended to indicate why we chose to define sine and cosine as we did.

*(6) Use exercise 13.26(5) to prove that

c" sin

-dt

n>m>

provided that

0.

Explain

why

it is

possible to deduce the

existence of the limit

(-"sin
"sin
lim
->=oJl
t

dt.

Exercise

16.3
(1)

16.4

Prove the following.

Periodicity

Since cos

cosO

(i)

cos

(iii)

(ii)

sinO

(iv)

sin

( x) =

cos

>

But the cosine function

= sin x.

( x)

and cos ( x)

tinuous, there exists a %

cos

such that cos

not positive

all

is

x and

>

for

the cosine function

x S (

is

con-

, ).

the time. If this were the case, then

2
would follow from the formula D cos x = cos x that the cosine function
was concave. But a bounded, differentiable function cannot be concave unless

it

Show

that, for

values ofX,

all

D cos x = sin x

(v)

D sin x =

(vi)

is

cos x.

It

(2) Let

be any

number and

real

g(x)

sin

h(x)

cos(x

(x

+ y) sin x

cos y

follows that there exists a smallest positive

We define

define

+y)~ cosx cosj' + sin* sin.y.

for
sin

all

{h(x)} 2 with respect to x and hence show that,

of* andy,

values

(x+y) =

(3)

= 0.

%.

cos y

+ cos x sin y

+ y) = cos*

cosy

sin x sin y.

appeal to the formulae of exercise 16.3(2).

(x

j7r)

sin

cos (x

jit)

cos

We

have

cos \ti

+ cos x sin

cos

sin x sin \tt = sin x

\tt

jt7r

cos

Use the previous question to prove the following.

from which
(i)

cos

x+

sin

* =

(ii)

cos

<

and

sin

<

sin

2x

2 sin

cos

(iv)

cos

2x =

cos

it

follows, in turn, that

(x

4-

2n)

sin

cos (x

2n)

cosx.

sin
(iii)

(4)

sin

cos (x

which cos

We now
sin

% for

By definition, cos tt = cos ( \n) = and cos x > for - \v <x < \-n. We
2
2
It follows from the formula cos x + sin x = 1 that
show that sin \it =
2
on
s n Jtt = 1 But D sin x = cos x and so the sine function increases
[ {-n, j7rj Since sin = it follows that sin jtt > 0.
1

Differentiate {g(x)}

number

number n by

the real

\n

cos x sin y

x sin 2 x.

Use L'Hopital's rule to prove that


Because of these
, sin

(i)

...

00

^
1

trt
(5) Use the

cos

mean

>-asx

0.

value theorem to

show

that, for

all

values of x

sinx

<1.
Hence show
vergence of

that the series

S"=I

last

formulae we say that the sine and cosine functions are

periodic with period 2n,

1 asx^-0

sin (l/).

2=1

sin (l/

2
)

converges. Discuss the con-

it

constant (exercise 12.21(5)).

154

Trigonometric functions

16.5
(1)

Exercise

and hence obtain the identity

Prove thai the cosine function decreases on [0, rr] and increases on
n, 2n] Prove that it is concave on [ \-n, \tt} and convex on [2^. jt]

= !_+->+...

(2)

Trigonometric functions

We

define the tangent function by

sm x

tan*

(x^{n +

\)rt:n

.).

fix)

+ 7r) =

Prove that tan (x


1, 2,

(
(3)

Show

Show

ar| d

tnat

is

we ignore the fact


interval [ \-n, |ir], the

1, 1]. If

side the

an inverse function with domain

i\

where n
on

strictly increasing

is

tan

= 0,

as

-*

and continuous on
under the sine function

interval

that the sine function

we

function

is

defined out-

obtain therefore admits

[1,1] and range [ \ti, \-n\ We


Draw a graph of this function
.

function the arcsine function.

call this

D arcsin x

and calculate

The

function to

for

we

notation sin" x, but


inverse.

= (

strictly increasing

image of this

{ ' ie

+ as x -> \tt and

that the sine function

provided that

that the tangent function

and that tan x

?"> 5 7r ]

is

\ir, \tt)

tan

<x <
1

(Some authors use the


no

prefer not to, since the sine function has

arcsine function

[ \-n,

is

the inverse of the restriction of the sine

^n]).

Discuss the arccosine function obtained as the inverse of the restriction of the cosine function to [0,

(4)

rr\

Calculate

D arccos x

and explain

how the arcsine and arccosine functions are related.


Show that the arctangent function, obtained as the inverse of the'
restriction

range

of the tangent function to (

Draw

j7T, \tj).

\tr, \-n),

graph and calculate

has

domain

and

D arctan x. Hence show

that

dx
Jo
+

(5)

\+x 2

Suppose

that

2'

/has

on

and that

x +y

/M+/O0 =/
for each

a continuous derivative

\xy

x andy such

that

xy <\. Prove

that, for

some constant

C,

f(x) - C arctan x. By writing* =jc in the formula and considering


what happens when x -* 1 , prove that arctan 1 = \n.

Show
arctan

that, for

x = x

- <x<l.
1

K..

*U

Let/:

sin

0,l,+2,

COS

(6)

be defined by

{x

# 0)

(x

= 0).

Let#: R -> R and/i: R->R be defined by g{x) = xf(x) and


2
h(x) = x f(x). Draw graphs and prove the following.
(i)/is not continuous at the point 0.
Qi)glS continuous at the point but not differentiable there,
and h'(Q) = 0.
(iii) h is differentiable at

155

157

The gamma function

From

17

(1)

it

also follows that, for

f(x)<j{\+x

THE GAMMA FUNCTION

<

1,

+x*+...}

-x 22\")

3(1

Thus
1

-"<*+,<;

rf n

17.1

If

a n ) and

<

b > are sequences of

real

numbers, the notation a n

~ bn

means that
*

as

-+ ca

d n -{\2n)-

Proposition {Stirling's formula)


n

nl~y/(2ir)n n

i/2

e-

n\

n"n

Proof Consider the sequence < d n > defined by

dn =

We

seek to

!( +

log

show

that (,d n

>

j) log

It

-l}

(12b)"

(,d n

--h

12

(n

} is increasing. In particular,

1,2,...)

decreases and so

-d n+l = -log(+

1)

we examine

-* .

C=

V(2t)-

We

leave this as an exercise (exercise

the sign of

- ( + |) log + (n +

|) log (

1)

7 .3 *

The gamma function


The gamma function
r(x)

|l+(2n+l)_1

C as n

e- n

remains to show that

4 4-4^-'
2n +

U2

17.4(1) and (2)).

n+n.

dn

>d

l)

12(n+l)

and hence the sequence < d n > is bounded below. Since < d n > decreases, it follows
that {d n ) converges.
Suppose that d n -* d as n -* . Since the exponential function is continuous at
every point, exp d n * exp d as n *. *>, Let C = exp d then

b
17.2

follows that the sequence

It

formula

Stirling's

3{(2n

To justify
In

is

defined on (0, ) by the formula

= J~*V -l <r*a

(>o).

the existence of the improper integral,

we

appeal to proposition 13.29.

view of the inequality

But, for |jc|<1,

^io g

the existence of the integral

+* -1
-x

vided that

> 0.

Since

x+1
r

f*-*-<JSft**
2

-1
2*11
2
jc

3
In particular,

f(x)>0

xA

x6

for

U|<

Therefore <c/> decreases.

and hence the existence of

and thus d

~dn+1 >0 (n =

1, 2,

.).

/7.4 +

''e-'dt follows from that of

+ , we

as t ->

have, for

x_1

rfr,

pro-

some i/ >0,

(r>l)
r* x ~
t

e~'dt follows from that of/,"*"

2
f

dr,

Exercise

t(l) Let /n

/f"

sin"

xdx

(n

0,

2,

sequence of positive numbers which


(n

756

e"'

x
t
fl

0)
1

= 2,

3,

.).

.).

Prove that

satisfies

nl

<

/ > is a decreasing

= (n

l)/n -2

158

Tlie

gamma function

Deduce

thai

as

-> .

n n\

fix)

hn

159

The gamma function

lim
"'-

(2)

(x+nY

x(x +])...

function satisfies
follows from exercise 17.4(4, 5 and 6) that the gamma
most one limit, we
have
at
sequence
can
the hypotheses of the theorem. Since a
=
T(x)
(x
0).
that/(x)
can therefore conclude from (2)
It

^(2) With the notation of the previous question, establish the identities

(2n)\

hn
(2

("

(2n+

(3)

= (a + 0)s + = s +

From the convexity of log/, it

0,1,...).

I)!

Then

follows that

log/(r)

<

a log/(x) +

fit)

<

imrm + o}"

s^m

that the constant

1 .

= t - s.

and so

(3

C obtained in 1 7.2 satisfies C = \/(2n).


Let dn = log n\ - (n + j) log n+n. Show that the sequence
_1
-l
{d n (12)
increases but that the sequence (dn (12n + 1) >
Hence show

=0,1,...)

n\f 2

(2"!) 2
'2n*l

>

The proof of (2) uses the convexity of log / Suppose that s < f < s +
we may write t = as + (3 (x + 1) where a>O,0>Oanda + |3=l. Now

log/(s

1)

>

decreases.

Deduce

that

n\

l/(12n-H)^

\/(27r)n"n V2 e- n

<P vl2n

Since s

<t<s +

tions in (3),

Hence estimate the error on approximating to 100! by Stirling's


formula. Is this error large or small compared with the value of 100!?
+
(4) Show that r(x + l) = xr(x)(x>0). Deduce that, fom = 1,2,3, ...

r(n
'

l)

f(s)

gamma

function

is

continuous on (0,

).

[Hint: If

prove that, for some constant

not depend on x aty, \r(x)


T(y) \<ff\x
(6) Prove that the logarithm of the gamma function

is

{r(a)

H which does

Properties of the

gamma

convex on

(x

1 7.6*

Is it

From

/(x+1) =

positive
).

*/(*)

> 0,

f.

Making appropriate

and continuous on

(0, )

+ n)

fr~

s
l) "'/(0-

f(n

f(s).

is

a natural

number. We may take

Then

+ 1)< f(x +n + 1)<( + D

it

/(

(5)

follows that

+ n) xn < (x + n)(x + n !

1)

xf{x)

<(n +

l) n\

+ i)',
(,

< <

and

let its

f(x+l)=xf(x).

(x>0)
+

(x>0).
that,

substitu-

x
This completes the proof of the formula (2) in the case when
equation
functional
the
the
help
of
general case is easily deduced with

If/ satisfies the functional equation

Proof The proof consists of showing


theorem, for each x

~ 1) -

and that n

,^

17.7^

T(x)

and (4) yields the inequality

(3)

this inequality

a generalisation of the factorial function

and/(l)= l.then
/(*)

<s <

>+.x+ .-n...v M <

the only such generalisation?

Theorem Let /be

logarithm be convex on (0,

(0, ).

function

The gamma function provides


(see exercise 17.4(4)).

ir**V(*

(x
1

also have

Now suppose that <x <


and t = x + n +
s = n +

1X0)}.]

(Hint: exercise 12.21(6) and theorem 13.25.]

7.5"

we

(t-\y-sf(S )<f(t)<s'-

0<a<a<.K<_)><<|3,

(3)

obtain

< (t -

Combining

b!

(5) Prove that the

we

= s<-*m.

under the hypotheses of the

(1)

Exercise

lfx>0,

prove that

log-

=
I

dt.

The

The gamma function

1 60

T(x)

(2) Prove that

gamma

yj(2n)x x'

in e' x This
.

is Stirling's

formula for the

function. [Hint: see theorem 17.6, inequality (5).]

*(3) Use L'Hopital's rule to

log(l

show

that

18

APPENDIX

18.1

Introduction

+z)-z

lim
z-0

With the help of


that, for all

Stirling's

formula for the

gamma

function, deduce

> 0,
1

(V)/u( +*>/)

'

as u

V(2rr)

number of results were stated without


to as 'propositions'. The proofs were omitted

In the preceding chapters a

proof. These results were referred

where

from the main body of the text to avoid confusing the

AM - j^-V.
1

* (4)

The beta function

>

the function of

two

variables defined for

>

and
1

by the formula

B(x,y)

=P'
J-*o

exists a real

> 0. Prove that, for a given fixed value of y > 0, B(x, y)

is

that the improper integral exists provided that

continuous function of x on (0,

and
a positive,

whose logarithm is convex on

(0,-).
'(5) With the notation of the previous question, prove that, for a given fixed

f(& =

> 0, the

r( *
(

satisfies the

f B(x,y)

sets

number K such

that

and

< K for

is

bounded

if

and only

if there

any xES.

-K<x<K.

and hence

>

numbers

-K<h<x<H<K

(x>0)

conditions of theorem 17.6. Deduce that, for x

A set S of real

If
Proof By exercise 1.20(1), \x <^if and only if
is
an
for
and
K
lower
bound
is
a
5
x < K for any x e S, it follows that K
upper bound for S. Thus 5 is bounded.
Suppose, on the other hand, that S is bounded. Let H be an upper bound and
h a lower bound. Put K = max {| H I, h !}. Then

function /: (0, ) -*R defined by

much

give the proofs here.

Proposition 2.3

>

value of y

Bounded

8.2

^-'(l-ty-'dt.
x

Check

is

we

detail. Instead

issue with too

(xGS)

< K for any x G 5.

Combination theorem

1 8 .3

^>0,
Proposition 4.8 Let

rro>)
B(x,y) =
r(x+y)
(6) Use the previous question to evaluate T(j).

- / as

Hence show that

('0

Xx n + p.y n

Vn
x

-*lm

-*\l
as

e~

x2' 2

dx =

V2ir.

as n -* . Let

X and m

+ /ti asn-*<*

n-*

P~

* and y n * m

be any real numbers. Then


(i)

xn

(iii)

->

yn

as n -* (provided that

Proofs o/0'O and


(ii)

(iii)

Since <x) converges,

!*!<*( =

m i= 0).

it is

bounded (theorem

4.25).

Suppose that

1,2, 3,...). Consider

161

Appendix

162

\x n y n

-lm\ =

\x n y n

<

xn

-x n m+x n m-lm\
xn 1
y n ~~ m + m
I

< K.\y n -m\ +


=

zn

Taking (1) and (2) together, we obtain, for any


1

yn

as

-*

-o

\m\.\x n -l\

i.e.

for

\x mr -i+i-h

<
<

l* mr -/|

ke

> N, consider
yn

he

\x mr -l\<e.
This does not conclude the proof since the subsequence

collection

With
'nx n

-yj

\mx n

this

satisfies

change as e

will

exists an infinite

(3>

xn

construct inductively a subsequence (xr

ml =

as

-* .

The

Then there

in (3).

l*n r

similar

>

an , such that

>n

which

such that

*_

x 2

- 1<
/

Take

< k- In this waY we

satisfies

-'l<7

and hence x r -* 7 as

exists

there exists an n 2

Then

\ in (3).

result therefore

which

-* 1 as r * .

construct a subsequence <jc

-y n l\.

By proposition 4.8 (i), mx n - y n l -* ml


follows from corollary 4.1 1

we

information,

Take e

my

<

<x mj_)

we have shown is that, given any e > 0, there


of terms x n of the sequence < x n > which satisfy

changes. All

r -*

(sandwich theorem).

argument shows that the infimum

of

also belongs to L.

Subsequences

18.4

Any convergent sequence

Proposition 5.17

|/-7l

|x-?|<e.

X n_J_

of all

>R,

>R,

any r

\y n \>h\m\.

For n

as

\x mr -i\

(triangle inequality)

and so x l\-*Q as n -<*> (exercise 4.29(H)). Similarly


n -* . From proposition 4.8(f), z -*0 as/i -<. Hie result
then follows from corollary 4.1 1.
-> \m as -* (exercise 4.29 (lii)). Since
(iii) Since y-*-m as -***, \y
m # 0, m > 0. It follows from exercise 4.29(2) that we can find an N such
that, for any n >N,
But x

163

Appendix

Proposition 5.13 Let (x n > be a bounded sequence and let L be the set
numbers which are the limit of some subsequence of (x n ). Then L has

real

maximum and

non-empty (Bolzano-Weierstrass theorem) and


bounded (theorem 4.23). Hence L has a smallest upper bound/. We seek to show
that IGL, i.e. there exists a subsequence (x
n > such that x n -*lasr-*.
set

find an A' such that, for

> 0. Then je > 0. Since 7

is

the smallest upper

znlEL

bound of L, 1 e

is

Equally, for any

\x m
Thus,

such that

if

we can

-l\<i e.
L

m > N and n > N,

\x n

-x m =
\

<

|f-7|<ie.

G L, we can
exists an

\xm

as n -* , then

m > A^,

and therefore

> 0. If x n *

n>N,

\*n-l\<te.

not

-I,
l>l>l-\e.

Hence there

be given. Then e

any

is

an upper bound of L. Hence there exists

Because

>

minimum.

Proof The

Let e

Proof Let e

Cauchy sequence.

is a

(i)
find a

subsequence

<

xm

such that, for any r

u<y.

>

which

satisfies

xm

-* I as r

and thus ix n )

is

\x n

-l + l-x m

\x n -l\

\x m

-l\<ke +

he

Cauchy sequence.

Proposition 5.18

>R,

Any Cauchy sequence

bounded.

Cauchy sequence. It is true that, for any e > 0, we


- x m < e. In partican find an N such that, for any n > N and any m > N, x n
for any n>Ni
that,
t such
cular, this is true when e = 1 i.e. there exists an
Proof Let <x> be

(2)

is

164

Appendix

and any

m>Ni.

m =N +

Take

that, for

Then, by theorem

\-\x Ni +

\x n

>

If I

any

8, for

. 1

-x Nitl

\<\x n

1 ,

we may

|<

n>N

>

choose e

sequence such that * r ->

-x m \<\.

\x n

165

Appendix

so small that

-e>

as r -* (proposition 5.13).

Let

We

<

x r > be

can find an

a sub-

R such

r>R,

any

x>Q-e)

lt

\a nr

l*nl<l*JVll + !

\>(l-e) n r-> + asr^.

Hence a^0asn- M (theorem

and thus S*-i a n

5.2)

diverges.

Now take
= max

A'

and

it

follows that

{|*i
|

|x 2

<

1,

A" (w

\x Ni

2,

1 ,

|,

|*n,-mI

1}

Limits of functions

8.6

.).

Proposition 8.4 Let /be defined

*
g e (a, ft). Then/(*)

at a point
1

iVoo/
Proposition 6.17 {ratio test)

S~=

Let

a n be

<n+l

lim

a series

can find a 5

which

satisfies

/>

Proof The text


take e

>

so small that

if

<

1,

indicates the

e>

(i)

>

as

-*

|.

>

Let e

be given. Then we

such that

8.3). Hence

for the case

<

If /

>

1 ,

we may

Then, for a sufficiently large value of /V,

Similarly, %

implies that
(ii)

"N+2

given.

<*<g+5

Suppose that /(*)

We

implies that

<

/(*) -l\<e. Thus /(*) -* /

can find a 5j

-*

>

as

/(*)

Hence

?<*<? + 5

-* g +.

- and /(*) *

-* %

such that

- g 1< 8.

as

the series converges.

proof

0<|*-l<5.But-5<*< implies that < * - % |< 5


- 1< e. Thus /(*) * as
| - 6 < * < g implies that /(*)

provided that

/.

the series diverges and,

1,

Suppose that /(*)

interval (a, b) except possibly

on an

+ g if and only if /(x) -Was* * | -and

|/(*)-/|<e

(see
If

/(*)->/ as x->+.

Tests for convergence of series

8.5

as

as

+g

+. Let e

>

be

- < e provided that


/

Kv-h
n-l

"n-2

g-5 <*<|.
1

>Q-e) n-N*l
Hence a n j*Q

as

-*

aN+i

-*
I

+ as M
Also

and so 2~=i a n diverges.

Proposition 6.18 (nth root test) Let


lim sup |a n

1/n
|

'

>
e

2 =! a be

a series

which

satisfies

Let 5

/.

<

any n

the series diverges and, if /

<

Proof Let x n =

we may choose e >


follows that we can find an

From

\a n |"". If

exercise 5.15(4)

it

<

>

the series converges.


1,

min

{S, , 5 2 }

such that /(*)

< e provided

that

Then /(*)
|

-l\<e

provided that

-6<*<andg<*<g-6
^-5<*<? + 5and*^?

i.e.

so small that
iV such that, for

0<|*-l<5.
Thus /(*)-* /as*-*

|.

> N,
xn

Proposition 8.12 Let /and

< +e

\a n

2=1

comparison

(/

<(/ +

)".

e)" converges, the convergence of

g be

defined

on an

* g and g(*)
at G (a, ft). Suppose that f(x)-+l as *
Then
numbers.
any
real
that X and n are

i.e.

Since

find a 6 2

<*<g + 8 2

i.e.

If

we can

E~=1

a n follows from the

(i)

test.
(ii)

X/(*) + pg(x)

->

+ (WW

X/

/(*)(*) ^/ as*

as

interval except possibly

* m as * - g and suppose

Appendix

166

f(x)/g(x) -* l/m as x -* (provided

(iii)

w = 0).

(i) Let <x n > be any sequence of points of (a, b) such that x
and x -> as n -> <. By theorem 8.9, /(x) -* / as n -* and
+ . By proposition 4.8(i)

Proof
(n

g(x n )

2,

-*

m as

-*

X/(x)

X/

Integration

Proposition 13.4

/im as ->

\\ {/(<)

obtain

+ Mf (*) ~* M + A"" as* * f.

(iii)

+ c} dt =

hztP=

Proof

m\U+c)

same way.

are proved in exactly the

Proposition 8.14 Let

in 13.2.

/,

-*

Suppose that g(x)

/be continuous on

[a,

b] and let c be constant.

have

and

(a, b).

proved

Proposition 13. 7 Let

g and h be defined on (a, b) except


as x -* , h (x) -> / as x + and that

possibly at

{yo,y\,y2,

{f(x)

inf

a).

c}=c+

inf

s if+c) (p) m

. .

>a).

{/(*)}

+ m\ n

<x<y,

j>

m^ (k - 1, 2,

We

>y} denote a partition of [a,b].

mj/ +c) = c +

Similarly

g(x)<f(x)<h(x)

f{t) dt +c(h-

<x<y,

is

Then

we

Appealing to theorem 8.9 again

(ii)

8.8

.)

*/(*n) + Atf(*n)

Items

767

Appendix

It

follows that

mg* e\yk -yk_t )

fe=i

except possibly when

X %. Then fix) -* I

Proof Let < x > be


(n

1,2,

/i(*n) ->

as

that/(#)

.)

->

and x

->

-* .

x -*

a sequence of points of (a, ) such that

as n

->

-* oo. Since g-(x)


as

as

By theorem 8.9,#(x)->/

</(x)

Hence /(x)

as

< /z(x) it follows from

as

-* and

wi'^fc-^fc-J +

* by theorem 8.9.

= S in (P) + c(b-a).
It

follows that

Continuity

18.7

on / if and only

if,

given any

x G / and any

> 0, we can find

a"

>

{f(t)

\
Ja

Proposition 9.3 Let /be defined on an interval. Then /is continuous

+ c} dt =

sup {S if
p

such that

sup

y& /and \x y <5.


Proof If */ but not an endpoint of/ then the condition y G / and
x y < 5 just the same as x y |< 5 provided that 8 sufficiently small.

provided that

is

""*/(*) as y * x,

i.e.

cib

sup
p

{S\P) + cib -a)}

-a) =

l"
Ja

is a

f(t)dt

function which

is

+ c(b-a).

continuous

and non-negative on the open


and

i.e.

Suppose also that /is continuous on /

interval /.

satisfies

is

/is continuous at x.

Suppose that x / and is a left hand endpoint of /. The the condition y G I


and x y < 5 reduces tox<y<x + 8, provided that 5 is sufficiently small.
The criterion of the proposition therefore reduces to/(^) ~*f(x) asy -*x +,
|

n iP)}+

Proposition 13.29 Suppose that

(xG/).

|/(*)|<0(X)

criterion given in the proposition therefore reduces to the assertion that

f(y)

\P)}

is

{S<-

*c

l/(*)-/O0Ke

The

(n-yfe-i)

fe=l

fe=l

theorem 4.10

improper integral of (p over the

If the

interval /exists, then so does that of/.

Proof We consider only the case when /

= (0, )

and the integral

/is continuous on the right at x. Similarly

if

x G/and

is

a right

= P

hand end-

<t>ix)dx

point.
exists. Let

Propositions 9.4, 9.5 and 9.6 are entirely

trivial

consequences of the results

indicated in the text.

Then the
Also

f"
Jn-l

series

/(*)<**;

2~=i b

is

bn=r
Jn-1

convergent

series

<Kx)dx

1,2,...).

of non-negative terms (with

sum

/).

\a n

and hence

\<b n

2"=1

exists.

fying

(H

we may deduce

1,2,...)

a n converges by the comparison

test, i.e.

f(x)dx

lim f

where Xo = *o

X > 0, we may take N to be

Given any

the convergence of the series

(-!)

m =

169

Appendix

Appendix

168

the smallest natural

number

Put 5

satis-

= x
|

a*o"

For values of y satisfying

\.

A7

-*-

00

y <6, consider
|

- nX n ~ =
f(x)dx =

We

show

seek to

v""
l

lim

if.

n =i
that

A -*

r X

as

Y-*-X. From

exercise 3.11(2),

follows that

-x

^i

X = x ? and F = y

where

<&;v->-0as

(Y"-

(Y"' 1

+Y"~2 X+

-X n ~

+ YX"'2 + X"' ) - nl"


1

...

+ x(r n_2 - x"- 2 ) + ...+ X"-\Y- X).

But from each of these terms we can extract a factor of (Y X). The right
hand side then reduces to the product of {Y X) with the sum of \n(n 1)
r
s
terms of the form X Y where r + s = n 2 The number \n (n 1 ) arises from

as required.

1 8.9

<

N>X. Then

^ x

It

Continuity and differentiation of power series

the use of the formula

Proposition 15.8 Suppose that the power series

+2+3 +

..

(n

- 1) =

%n(n

- 1)

(example 3.9).

Since

fix)

(*-*)"
\X\

n=0

has interval of convergence

on / (except

at

/.

Then

its

sum

is

continuous on

and

differentiable

= |x-5l<|x -5l -

we have

the endpoints). Moreover

y n -z"
r-AT

f(x)=

X mdx-HT'1
n=l

Proof Let the

*,,

and

m - ly-|KI*-tl +

-nJC n-I <-(n-i}ijr

radius of convergence of the

power

series

nH
i

|y-jfK

-+0as

K-i-AT.

is

It

follows that /is differentiable at each

in /

which

is

not an endpoint and

M#**.

->
x

right hand continuity at the endpoints


happen to belong to /. A somewhat
deal with this question which is the subject

There remains the question of left or

-f1

n-1

x-

-/?

of the interval of convergence

more

From the formula of 15.1

subtle argument

is

/, if

these

required to

matter of Abel's theorem quoted below.


l/n

lim sup |a n l"


n~>

1-Xbl

be R.

/'(.*)

lAKir-jriJ^i^Ti

[X-f|<|ar -||<fi.

Hence
-

any point of / other than an endpoint. We can then


find another point x of / so that x lies between and x and thus

Suppose that X

n(n-\)
lim sup
n*-

Abel's theorem If the series Sfe =0 a k converges, then

171

Appendix

Appendix

70

lim
*-+l-

akx

k
\

fe=o

ak

k -

akx

fe=0

Proof Let e > be given. Since the series 2 a k converges, its sequence
such that,
of partial sums is a Cauchy sequence. We can therefore find an
whenever k>m>N,

tffe

<e.

ft=o

This concludes the proof.

k
a,

(4)

<r-

lemma (which

is

easily

proved by induction)

asserts that

+ vn
k-m

We

k=m

apply Abel's

= ak

=x

and v k

> m > N and < x <

la k x k <0-*)"f
k=m
3

We may conclude

Then

a k x'

Next observe

->

(1

x)

that

<

k=N

(5)

r-

that, for

<x<

N-i

akx

- V
k=0

fe=0

ak

< I
fe=0

\a n

\(\-x") +

This follows from (4) and (5). But the finite

\ex k + \ex n

k=m

zero as

uh

k=m

follows that, for

-*

k-m

\\l=m

lemma with u k

fe=m

It

We can
.

therefore find a 5

>

sum on

-e + -e.

the right

such that, for

hand

any x

side tends to

satisfying

\-8<x<\,
N

t\a n \(l-x")<\e.
3

k=o

> 0, we have therefore


-5<x<

Given any e
fying

Stirling's

formula

The proof of proposition 17.2

Z
Abel's

18.10

found a 6

>

such that, for any

satis-

is

completed

in exercises 17.4(1

and

2).

173

Solutions to exercises (1.8, 1.12)

5-3>4-l

Note

(exercise

inequality

There

x2 =

x =

are three possibilities:

0. In the

second case rule

Rewrite the third possibility

= 0.x<x.x = x 2
We

(i)

We are given that


= b.b>l.b = b.

(ii)

> 0, x < 0. In the

yields that

case

= x.x >0.x = 0.

we may apply

rule III.

>

6 Take

Since b

> we may apply

rule III.

Then

If

ff.fi

fi

= -.bB<-.bB = Ab

We deduce
a{b

Since (6

+ flfi <

a*

B)'

>

x =

(a

< b. Hence a = b.

b)/2.

is

an even natural number

aft

(rule III),

and 6" 1

v46

>

+ 4)4

(a

(example

= x 2h =

(x

R 2
)

all

(rule II).

1.5), it follows

tion has exactly

from

two

Next suppose

rule III

+A
+ B'

The second

.*"

half of the inequality to be demonstrated

is

obtained simi-

>b

c> d. Then, by rule U,a+c>b + c and


c>b + d (rule
If b >
and d > 0, we first observe that c > d > 0. Then, by rule
III, ac > be and 6c > W, Hence ac > M (rule I).
Suppose that a
b

+ c>b+d.

and

Hence a +

4 Substitute the values a


(ii).

7 72

We

be expressed

in the

form n

2k.

=y

<

is

odd. If y

there

has exactly one solution. Ifiy

when n

is

and one negative.


is no difficulty in showing
>0, there is exactly one

solutions, one positive

that

odd.

If

is

the only solution because

< 0, we use the

and only one solution and hence the same

larly.

may

values

that*"

it

>0

positive solution and this

4.

of* (exercise 1.8(1)). Hence, if y < 0, x" = y has no


The equation x" = has only the solution x = since x =
implies that x" =t 0. Uy > 0, our assumption about the existence of
nth roots assures of the existence of a unique x>0 such that x" = y.
But, since n is even, x" = y if and only if ( x)" =y. Hence the equa-

that

= d = 2 in

and b

>0 and
and d >

e>0,b<a + e and a e < b. Since b < a + e for


b < a (example .7). Since a < b + e for any

follows that a

it

=c=

obtain the false assertion

follows that

it

>

when b

solutions.

that

+ B)

we take a

if

true

But then

for

/4

we

(ii),

have, for any

> 0,

x"

a.

However,

.8(3)).

(iii) is

Exercise 1.12

bB >0,

2 Since

other hand inequality

Then

= a.
b

We

any e > 0,

i.e.x

>

Since a

2
a;

5
first

>x and apply rule IV. Then


2
>0 or 2 = 0.

the form

casex 2 >0,

<a<

are given that

0. a<a.a<l.a

III

iii

In each

0,

12.

= (-l)(-l)>(-2)(-2) =

Exercise 1.8

that the inequalities are false in spite of the fact that

> 0. On the

>

i.e.

SOLUTIONS TO EXERCISES

I).

5,b

= 4, c =

obtain the false assertions

3,d

in inequalities (i)

and

x"=y.

is

true

<

implies

= y has one
of ( xf = y, i.e.

fact that z"

74

Solutions to exercises

(i)

2' 3

3 2

{(2

}" 3

(iii)

32

5 6 1/5

ax

1/3

27'
6/s

-4/3

3
3
{4 }"

1/3

(ii)27

{(2 ) }

We

To

bers,

prove the results

(ii)

when

and (iii) in the case when r and


r and s are any rational num-

,/n

prove

(y

(a),

n2

y/ai

+ V<* n -j

r- +

(c)y

Un Vn

(yz)

Vn

and the

result follows.

1/

{(y

")"} m

= ym

the unique positive solution

is

m
ofx" = y

t=

fe
,

kk + Kf =
i

I 4+2 Z

k=i

= (ytoftfi-f =

{y"z >'"}"

and hence y

Vn vn
z

is

We may now

= qjn (p

and q

(iii), (ii)

(iii)

yz = ry)" m (zp ) ,/m =


=

(yzf'

1/m

{{(y )

pq/mn

>

(i)

=yz.
when r

in the general case

Consider, for example, the case n

= pjm

\fc=i

fe=i

inequality)

ov) ""
1

&zffm

]}

1/"

two

vertices

is

{l(y

Vm
)"}

l,n
}

fe=i

fe=i

lyP-y"""

rpn

qmil/mn

show

that

\a\

b implies

Since (2ax

b)

>0,

b)

< b implies -b < a <


< 6.

1.10 to obtain
(b

b and then that

\a\

Suppose that

1 .1

it

{(2ax

5, a < \a\ and \a\ >a.


\a\ < b. From theorem
>
follows
that -fc< a < 6.
6.
It
a<
i.e.
a
<6
and

b,
Hence a
each
(ii) Suppose that -b<a<b. Then a < b and -a < 6. Since, for
follows that \a\ < b.
a, |a| = a or \a\ = a,
(i)

(3)

as in

first

fc

.v

(3)

We

<a<

(a:

U/2

Exercise 1.20

y^

- a)(x - fi).Whena< x <0,jc -a>0 and


x-p<0. Hence (x - a)(x - < 0. When < a, x - a < and
- a)(x - 0) > 0. When *>/3,x-a>0
x - < 0. and therefore
and x > and therefore (* a)(x
> 0.
c

-z k +z k -y k ?
1/2
[,'

(ypnyimnf qm\Vmn

4 Write ax + bx + c = a(x

+ bx +

(x k

<

'Complete the square'

that (xj,x 2 ,x 3),

1/2

\l/2

(3

and suppose

given by

fe

yP""y<""

=3

(z

,JS8
>"

y(P" + f")/mn

U.)

trih z 2 ,z 3) are the co-ordinates of the vertices of a


(y lt y 2 ,y 3)
angle in three dimensional space. The length of the side joining the firs',

and

m =
(yzf

00 Iff =
_

and

LI

integers).

\l/2

(Cauchy-Schwarz

mn = y. To

yz

the unique positive solution of x"

prove

and

ax

6?

observe that

(c),

r+s

1/2 /
n

\fc=i

fc=i

the unique positive solution of x

is

fc=i

and hence (y Um ) Vn

ah b k

fe=i

i.e.

= (y) '". To prove (b), observe that


1/m 1/ m " =
Vm /n n m = (y" m m =
"}
{(y
{{(y
y } }
)
)
y

observe that

v ") m

(y /)"

(J)

+<(ai +a 2 + ...+a n ) \a,


-+J+...
a

(b)0 1/m )"" = y umn

(y" n ) mn

=
(y

Vn

and hence

prove

= 0, i.e. x = bjla.

"")"*}"

{(}>

helpful to have available the following preliminary results in

it is

To

which m and n denote natural numbers:

(a)O

Apply the Cauchy-Schwarz inequality using the numbers \Ja\,\Ja 2


l/y/a. We obtain
,\Jan and 1/Va,,

64.

take for granted the truth of (i),

s are integers.

81

+ bx + c>c -\b 2 a'

with equality when lax

_1_

3 4

1 75

Solutions to exercises (1.12, 1.20)

12

( 1.

- 4ac)}/4a.
2

From theorem

1.18,

Solutions to exercises (2.10, 2.13)

Solutions to exercises (1.20, 2.10)

1 76

\c-d\>\c\~\d\

equivalent to %

77

>x > g -5. But ft - 5, % + 5) =

{x:-5<;c< + 6}.

and

\c-d\

= \d-c\>\d\-\c\ =

-{|c|

Idl}.

- {kl - \d\) < k - d\ < {kl - \d\} and the

Thus,

result follows

bounded above; some upper bounds


upper bound is 1 no maximum.

(i) (0, 1):

smallest

from

exercise 1.20(1).

and

the

( ,

(ii)

are 100, 2

2]

bounded above; some upper bounds are 100, 3 and


bound is 2 and this is also a maximum.

2; the

smallest upper
3

Only

(iv)

is

We have

not immediate.

{1,0,

(iii)

d(x,y)

\x z + z y\ < \x z\ +

= \xy\ =
=

d(x,z)

\z

y\

5; the smallest

d(z,y).

4 Suppose that r + s\/2

t,

(iv) (3, ):

where

t is

rational.

Then, provided

vi - =.
4

so

right

we have

We

hand

side of this equation

is

bounded above; some upper bounds are 100, 2 and


upper bound is 1 and this is also a maximum.

a contradiction.

are given that

a=

equation ax

satisfies the

+ bx + c =

{1,0,2,

2as

br

c}

{2a

+ b}s^2 =

0.

- sy/2) + b(r - sy/2) + c =

a(r

follows that (2a

it

{ar

2as

br

+ b)s = 0.

Thus

the

10, - 1, and 0; the

bounded below; some lower bounds are 10, 2


bound is and this is also a minimum.
1

bounded below; some lower bounds


lower bound is 3; no minimum.

are

0,

2 and 3 ; the

bounded below; some lower bounds are 10,-1 and 0;


lower bound is and this is also a minimum.

(v) [0, 1]

Since a,b,c,s and r are rational,

5}:

the largest lower

1 ;

unbounded below.

and

largest
2

are

(iv) (3, ):

+ sy/2) 2 +b(r + s\/2) + c =

{ar

2\.

(iii)

0.

Hence
a(r

( <*>,

(ii)

+ s\/2

bounded below; some lower bounds


lower bound is 0; no minimum.

(0, 1):

(i)

largest

number (why?) and

a rational

unbounded above.

(v) [0,1]:

s = 0,

smallest

But the

bounded above; some upper bounds are 100,6 and


upper bound is 5 and this is also a maximum.

2, 5}:

largest

+ c)
5 {3,4}.

-(2a +
and so &

= rs\/2

is

also a root

b)s\/2

=
6 Take y

of the equation.

= \x. No m e (0, ) can

element of the
6 Suppose that

m = 3n Then m divisible by 3 and hence m


m = 2k + or m = 3k + 2.) Hence m = 3k. But
2

3n

is

i.e.

False

(ii)

true

Exercise 2.13
1

Since

last

1.20(1),

inequality

B=

sup

is

S,x<B for each x G S. But


it is true that x^S. Hence x < B

an upper bound for

S C5 means that, for each x G S


Thus B is an upper bound
for each x G S
,

as large as the smallest

(iii)

true

(iv) false

T=

(v) true.

2 By exercise

is

a smaller

set.

upper bound sup

S and
i.e. sup S

for
,

2 The proof is very similar to that of theorem 2.12. Let

Exercise 2.10
(i)

\m is

is

because

2k = n and so n is also divisible by 3.


Suppose that m = 2n 3 Then m 3 is divisible by 2 and hence m is
3
divisible by 2. (Try w = 2k + 1 .) Hence m = 2&. But then 8A: = 2n\
3
3
i.e. 4 = and so n is also divisible by 2.
then 9&

minimum

divisible by 3. (Try
2
2

be a

||

-jc|<5

equivalent to 5

if

and only

if

-5 < -x<8.

>x > 6

which

is,

This

in turn,

{x

+x

+ % x G S).
:

Since

< B for any x G S, it is

< \ + B for any x G S and hence

for T. If

C is

the smallest upper

bound

that %
for T,

+ B is
it

therefore at least

< B = sup S.
B = sup S

and

let

true that

an upper bound

follows that

C<%+B.

C %
On the other hand, y < C for any y G T and therefore y | <

Solutions to exercises (2.13)

1 78

G T.

any y

for

upper bound

We

Since

for

S=

S and

:y G

{y

7"}, it

follows that

B < C .
+ B and
+ 5.

therefore that

C<

have shown that

is

If/ is an open interval other than

an

supremum
Hence

B=

We may

= { x

< 5 for any x G 5 it is


B a lower bound
for T. If C is the largest lower bound for T,
follows that C > B.
On the other hand, y C for any y G 7" and therefore j> < C for
any 7 G T. Since 5 = {y :y&T), it follows that C an upper
Let

5 and

sup

G 5}.

x > B for any x G S and

true that

Since

hence that

then can be taken as

or

or infimum (at least one of these

its

must exist).

C = + tf
6

let 7"

179

Solutions to exercises (2.13, 3.6)

assume that the

(otherwise the problem

To

set

is

Let

it

ft

is

S and Thave no elements

sets

is trivial).

not empty because

GT

in

common

With reference to the given 'hint', the


.

The

set

is

bounded below by

s.

= inf TV

~5*

is

bound

S and

for

therefore that

OB.

We have shown that C > -B and C<B. Hence C = -5.


To obtain the required analogues, consider a non-empty set

If

(iii)

x&T

(i)l

5 (i)
is

>MT

inf x.

xGT
(iii)

{||

Let

-x| :x G

5} has

-x| =

>

is

D = { x :xGS}. If this
x G S. But
% -x > h for

all

ft

is

If
(iii)

as a

lower bound. If

G 5,

an upper bound for

then

bound

we can

is

is

at zero distance

from

by

exercise

at zero distance

have

that
all n G (^ and therefore 1 is an upper bound for S. Suppose
ft, where > 0, is a smaller upper bound. Then, for all n G f^,
ft

an interval, |

for the set

for all

>

such that

G S and hence

-\

smaller than the smallest upper bound.

T) where

$ / implies that

is

'

?<

T= {-x :x G S}.

either an

ft

upper bound

or a lower bound for /. Suppose the former. Let B be the smallest upper
bound of/. Then B G / because / is closed. Given any x G /,

||#| =

ft

example for which

find an h

<%-h

= inf 5, consider instead d{~ I,

Since

G S. But

for

then

-x for each x G 5. We therefore

a lower

is false,

ft

l--<l-ft

sup S. Then
ft

^ T, then

for D. Exercise 4(iv) provides an

have to show that no

We

(iv) 0.

ft

Exercise 3.6

= ObuttS.

tf(,S)
(ii)

set /J

minimum

(ii)2

The

from T.
and hence we have found a point of S
subset
of 5.
If ft G T, then ft > s and the interval (x, ft) is a non-empty
Hence ft is a point of T at zero distance from S.

(>0)

et

7\ then inf

( x)

sup

inf

|
x

J C

(ii) If

$x

il

2.13(5ii)

inf

which

bounded below and then apply theorem 2.12 and exercises 2.13(1)
and (2) to the set S = {- x x G T). We obtain

ir

huh

To
7"

is

(i)

and hence

is

an upper bound for N. This

is

bound for S.
therefore
no n G [^ is
maximum.
For
has
no
The set S
is

h-1 =

%~x = S-B+B-x = $-B+\B-x\.

ft"

a contradiction

and

the smallest upper

it

true that

1.

Hence, by exercise 2.13(2),

x|
x es
inf

= -fi+
%

inf
* 4s

and therefore d(|, 5) = - + d(B, S). But - B > 0, t/(Z?, S) >


and d(t, S) = 0. It follows that = 5 and hence G /. Similarly if
a lower

bound

for

/.

2 Suppose that S

\B-x\

BX'

X">BX'

Since

is

X"* l >B.

is

bounded above. Then

it

has a smallest upper

cannot be an upper bound, there exists an n

G N

bound B.

such that

180

Solutions to exercises (3.6)

Hence

It is

not an upper bound for

is

clear the

lower bound

xn

>h

l\"

(i.e.

for T.

Suppose

> 0). Then, for any S

and we have a contradiction.

>*>

bound

a lower

is

that h

is

a larger

Exercise 3.11
1

fsj,

+ n 2 and let P(n) be the proposition that


+
Theni'(l)
is true because s, = 1 and
+
1)s'= ( + 0(2n
=
assume
that P(n) is true and try and
+
We
next
+
1)
l(l
1)(2
deduce thatP^i + 1) is true.
(i)Lets=

which contradicts the

G M}

S = {(llx)":n

follows that the set

first

part

bounded above by h' 1

is

2'?

be a non-empty set of integers which

maximum of the

is

\n(n

l)(2n

l){(2n

1)(

+ (n +

1)

+ 6(n +

1)

2)(2

3)

l)

(assuming P(n) true)

1)}
l
6

+ (n+l) 2

sn

l(n

m(m +

1)(2

l)(2m

+ In +

6)

1)

is

set S.

Similarly if S

+2 2 +... + n 2 +(n+l) 2 =

= K +
= l(n +

of the question.

bounded above by B.
Since F*J is unbounded above, we can find an n G N such that n> B.
The set T~ {n x:x S} is a non-empty set of natural numbers and
hence has a minimum m (theorem 3.5). The integer n m is then the

3 Let

sn + ,

It

181

Solutions to exercises (3.11)

(m=n +
+n
Let s n = l + 2 +
= kn 2 (n+ l^.ThenPO^s
3

(ii)

sn

We

bounded below.

and

next assume that P(n)

is

let

1).

P(n) be the proposition that


2

andjl 2 (l + 1) =
and deduce that P(n + 1) is

true because s,

true and try

1.

true.

N is unbounded above, there exists a natural number n such that


n>(b a)" Otherwise (b a)~' would be an upper bound for f^. Let
m be the smallest integer satisfying m>an. Then m < an. Hence

4 Since

s n +i

+ 2 3 +... + 3 + (+l) 3 =

\n

\(n

+ (+l) 3

1)V + 4n + 4) =

l)

(n

(n

(assuming P(n) true)

m
m
a < and < a + < a
1

(b-a) =

|(

\)\n

2)

2
.

2 Let P(n) be the assertion

as required.
1

Suppose thatm &S. Then


cannot be
It is

(0

maximum.

obvious that

< H<

1)

\{tn

+ I)G5and^(m + \)>m.

Similarly

is

Hence

Suppose that

number

such that

element of S larger than//. Thus

A similar argument shows

that

H<r<\. But
is

is

Provided

this

were a smaller upper bound. By exercise 3.6(4) we could

find a rational

...+x"

cannot be a minimum.

bounded above by

+x +

+x+x 2 +

is

-x

clearly true

when n =
1

+ x n +x n * =
,

..

0.

Assuming P(n)

the largest lower

true,

-x

-x

then r would be an

l-x"

the smallest upper bound.

is

n+l

+x"

-x

l-x

bound.
n+2

We know
rational

that

all

numbers of

the

and non-zero (exercise

find a rational

number r such

form r\j2
.20(4)).

By

are irrational provided r

exercise 3.16(4)

we

is

can

that

and therefore P(n + 1) is true.


The second formula of the question
(Note that,

and hence a
rational

< r\j2 < b. (If r happens to be zero, locate a second


satisfying
<s\J7 < b.)

-x
\-x

if a

3 Suppose that

number s

P(x)

=b

P(!-)

or b

= 0, the

is

formula

obtained by writing a- =a/b.


is

obviously true.)

= 0. Then

= P(x)-PQ)
=

a n {x

-% n ) + a n .

{x"-

-% n )+
i

...+

ai {x-%).

182

By

the previous question

x*-r =

(*->(*

we have
&-2

fe-l -L

- l)
f-fe-1

Fk _

/>(*)

We

(*

(x)

a polynomial of degree

is

-WW*) +B

-1

A:

Fn _a

>c)

place,

n+

+ 6"

\s-l

-slS

It

follows that

+ ,} =

as required.
(a-

-$)8(x).

apply this result to the second part of the question and obtain, in

first

si

s=

= (x li)2i(x) where Qi(x) of degree n


Then = P(% 2) = (2 - f OGi(h). Since ? 2 # | b follows that
Qi{% 2 ) = and hence that Q (x) = (x - % 2 )Q 2 (x) where Q 2 (*) is of
degree n 2. ThusP(x) = (x i)(x % 2 )Q 2 (x). Next consider 3 and
the

li1

_n+I -sts
6*

a"

+r

l-fc-2 J.

=
where

183

Solutions to exercises (3.11, 4.6)

Solutions to exercises (3.11)

P(x)

is

number k for which P(k) is false. Since


= 2 N > k.
unbounded above, we can find an
Now put S = {n i>(") is false and n < M). This is a non-empty set of
natural numbers which is bounded above. From exercise 3.6(3) it has a
maximum element m. The assertion P(n) is true for m < n < and
hence P(m + 1) is true. But P(m + 1) implies P(m) and hence we have
Suppose there
the set {2" n
:

exists a natural

e M} is
:

it

so on.

a contradiction.

B!

r-lJ

r!(n-r)!

!
-

(r-

!)!(

H!

(r-l)!(n-r)!

-r +

n-r+1

+1
(r-l)!(n-r)! r(n-r+
n

+ 1)1

r!(-r+

From

exercise 3.11(2),

"- b " =
a

(a-b)(a"-

Hence,

>b >0,

+ an

-2

+ ...+ab"- 2

fr""

).

Ir

n\

fe

1)!

nb"

if a

<

a"

1)

-6"
<na"

a-b

Therefore,

l\

1)!

Let P{n) be the assertion that


If x

(a

+ b)"=

\a-

r r

> y, the result follows on writing a = x


,/

.v<^,writea=>' "and6

=x

'"

and b

=y 1/n

If

Vn
.

r=0
Clearly P(l)

(a+Z>)"

+1

is

true.

(a

Assuming P(n)

is

Exercise 4.6

true,

+ b)(a + b) n

s:K''
f a -v+ i(V~ *
(r=0

replace r by s in the

+ b)n

any n

first

sum and

replace r

by

s in the

-l
-1 =
2
n +
1

- n 2 -\
2

+\

and so we simply need to find an


I

+1
U""' 6'+

"J*""**

,7TT
i.e.

<e

>N,

second sum. Then

(a

a value of A' such that, for

But

r+I

r = n\>-

We

> be given. We must find


2 -\
n
-1 <e.
n 2 +l
Let e

+\

N such that, for any n>N,

184

Solutions to exercises (4.6, 4.20)


2

n +

> 2/e.

N = (2/e)

Choose

+ \>n

>N

1/2

We have shown
(2/e)

that, given

any e

> 0, we can find an /V (namely

>N,

x.x~" +

This leaves

x=

1
1

2
2

>

2 Let e

7" Q

-l
+l

+
+

The

>

that

x+x"
+x n

and

1"

as

-* .

Hence

= \

asn

giving

-1 +(-!)"
n
1 +(-l)

1"

"

0+1

and x

Then x

'

not even defined for odd n. Hence

latter expression is

1.

(x>

be given.

x'"+l

Thus
lim

Next suppose

2/e.

such that, for any n

x+x" x +
as n
+x"^ +0

> N,

-\
-1 <e.
+l

n2
n

1/2

Then, for any n

185

Solutions to exercises (4.20

We must

find a value of TV such that, for any

x+x"
lim
\l+x n

>N,

1)

< <
jc

(*<-!).

<e

i.e.

r >l/e.

We choose N = (1 /e) 1/r


n

>N

Then, for any

> TV,
limit

1/e

undefined

and the

result follows.

>

3 Let e
\\x n
If

there

are given that

find an
\x n

We must

find a value

of

N such that, for any n > N,

-\l\<e.

be given.

is

xn

We

have

(V(n+I)-V")(V(n + l)+Vn)
0<V(+ i)-V =

(V(+l)+V)

nothing to prove.
-* I as

-> .

We may

Since e/|X|

N such that, for any n>N,

X # 0. We
that we can

n+l-n

V(+l) + \/n

V"

therefore assume

>

it

follows

(see exercise 4.6(2)).

~l\<el\X\

V(h +

1)

V" "*

It

as

as

-*-

follows from the sandwich theorem that

* .

i.e.

4 Let N be

|Xx-X/|<e.

X X

x"
This completes the proof.
n~\
i!

+
2n

5n
s

+2

+9

5/T

n>N,

,v-i

\n-N+l

IX

1+0 +

2n~ 3

2 + 9n" 3

as n * .

Since

x/N

<

(x/N)"

sandwich theorem.
2 First consider the case when |x|<

for

X
1

number such thatN>x. Then,

" T'2'Nl'N'"N
2 N-l N

Exercise 4.20
1

the smallest natural

Then x"

as

-><.

Hence

-*

as

-*

and the

result follows

from the

186

Solutions to exercises (4.20, 4.29"J


5

Suppose that

(1

l/)

a+1
1*|

>

for

all

n e N. Then

But

\\x n

0| =

\x\

for

all

PiJ

and

example

3.4).

Hence, for

+UN)

If x = 0,

It

\n

a*

xn

\<:\n

a*

xn

Since this

\.

true for any e

is

conclude that, for

xn

>l~\l -

n>N,

(i)

Let //

>

be given.

as

by the sandwich theorem.

exists an

JV6 N

when e "|l (for an


we therefore have

true

that

We

have to find an

N such that, for any n>N,

G M} is unbounded above (exercise


N >H. If n >N, then

2">2 N

>H

3.6(2)), there

such that 2

wish to prove that

and the

>n

2"
This inequality

is

equivalent to

n
\)

<n

(i)

i.e.

y/n>H

i.e.

n>H 2

We may
know from example

<3
n

It

H, Hence (1) holds provided 3

Exercise 4.29

>

be given.

||x-/|-0|<e.

We

We

one

may

appeal to the inequality

proved by induction.)

have to find an

therefore take

N = H2

N such that, for any n>N,

The sequence <( l)"n) cannot converge because it is unbounded.


+ because its odd terms are all negative. It can-

cannot diverge to

have to find an

because

its

< n.
4

Let e

be given.

easily

not diverge to
for all

is

4.19 that
(iii)

> 1) which

H>
-s/n<-H

n* 1

i.e.

But we

result follows. (Alternatively

{n

Let

(ii)

(i)

it is

W = A^). For > TV,

T>H.
:

n>N,

Since the set {2"

and thus nx"

any

that, for

\.

N such

> 0, then

appropriate value of N, say

n>N,

x n \<\Na+1 x n

(n

exists an

/-e<^:</ + e.

a+i

We

the sandwich theorem.

i.e.

1/iVK-UV'Q+l^jVl

* by

as n

|/|

1*1.

follows that, for

We

\x n -*

-l|<e.

l*B

Un>N,
l)

just

|-|/ll<U-/|

||.x n

2 For any e > 0, there

consider the expression


a+l

n a+1x n

is

\x\<l.

iOr-Hn
+1
x +
(n+lf

\(n

and so what has to be proved

1\

-* .

exercise 1.20(2),

and hence

a+ 1

(1

this is a contradiction (see

We W,

some

By

(ii)

\x

/|

the definition of x n -* / as

187

Solutions to exercises (4.29)

N such that, for any n>N,

xn

* . Let H >

(i)

Suppose

an

N such that, for any n >N,

\/x n

But

that

as

even terms are

all

positive.

be given.

We

have to find

>H.

H' > 0.
1

Since

xn

-*

as n -* ,

we can

therefore find an

N such

1 88

Solu [ions to exercises

xn =

Let

H>

that

xN

that \/x n -*

be given.

> #. If (x n

follows that

But*?, x+k<

+ as n -* implies that x n -*

as

-> ,

xn

-*

Suppose that x n
x + k and so

-*

as

D=

a value of n GN for which no x G S can be found


x\< 1/h. Then 1/n a lower bound for the set

-*

i-

contradicts the assertion that

- follows

as

= sup S,

from

||

jc| <

S is unbounded above,

l/ (

dQ, S) =
2,

Suppose that 1/n ->

.)

rf(|,

5)

= 0.

It

> n.

n+

as

follows that

X n -1

then, given any n

Hence x n

->

GH, we can

+ as n

as n

We

-*

increases

x n+1

= P. Since / >

we deduce

a<x n < bby induction. It


a < x n < b, follows that
2
x% + ka>a a + k = 0.

is

that

a =

Hence

that

given that a

<

.).

Observe

x n)

has the opposite sign tox.. 2

xn

It

follows, using

-i) increases

ifx 3

>*i

k;

follows that
Ct,,) is

l+l

m + Im = k.
= m and that 2 + = k as required. The conclusion
I

that

xn

-* I as

-> .

< b. If
4

If

<a<

6, it is easily seen that the geometric

example 3.10) and the harmonic

mean//=

mean

{\(\ja

G = V(a&) (see
1

1/6)}" (see

exercise 1.12(5)) satisfy

a<H<G <b.

-b =xl+k-b<b 2 -b + k =

Hencea<x n+ <
,

both

+m'

+ Im =

It

it

b.

positive.)

0.

(Note that the condition

the quadratic equation


are

l+*- 2

and the other decreases. (In fact (x 2n

=
1

Similarly

xn + s

and decreases if -X 3 < x t .)


That both sequences converge follow from theorem 4.17. Suppose
that x 2n * I as n -* and that x 2n - 1 "* W as n -* . Then

5.2 that

(example 4.14). Thus n v2 " -> / as n -*

prove that

we assume

(n

an induction argument that one of the sequences <x 2n ) and <x 2n _j>

-> .

* . It follows from theorem

x n .

Hence x n+i

find an

about
2

l+X n

k(X-2

It

>

therefore

k > x

0.

-<-/asH-.

2n
But 2"

that

= a.

be-

Exercise 5.7

1/2n

Butx n + i =

But b cannot be the infimum of the

not a lower bound. Hence

it is

shown by induction

part of the question that a sequence Cc n > of

first

x n G S such that* n

(2rt)

a, it converges.

->-/as ^-.

that

points of S can be found such thatJ n - as - . (Note that the


terms of this sequence are not necessarily distinct.)

It is easily

then from exercise 2.13(5ii)

follows from the

If

is

cause of the sandwich theorem.


If

Thenx n +

is

{\%x\:xGS} which

That x

-+ .

= aotl = b.

sequence (x> because

satisfying

bounded below by

is

k.

This implies that

bounded below.
6 Suppose there

b (see exercise 1.12(4)). This

< x n + <x n < b.

is

increases, then, for

+ oo.

<x n <

because a

Since Cc> decreases and

unbounded above, there exists an N such


any n >N,x >xN > H. It
Similarly in the case when 0c) decreases and is

Cx)

If

+ k.

completes the proof that a

one simply has to reverse the above argument.


5

xn = xn

xn + l

>H.

To prove

(ii)

Next consider

fx.-OKJT

\lx n

i.e.

189

Solutions to exercises (5. 7)

n>N,

any

that, for

(4. 29, 5. 7)

x2 x + k =

We

< k <$ ensures that

has two real roots and that these

are given

h= x i<yi=

1.

On

the assumption that

x n _,

x n-l< x n<y n -l
because

xn

is

the geometric

mean ofx n _, andj'n-!. Further

<^ n _

1,

190

Solutions to exercises (5. 7)

xn

Solutions to exercises (5.

<y n <y n -i

because y n

is

mean of x n and y n -j.

tion that x_i

<x n <y n <y.

("

= 2,

3,

It

.).

and is bounded above by y% 1. The


sequence (y n ) decreases and is bounded below by Xi = \. Hence both
sequences converge. Suppose that x n -* I as n -* and y n -> m as n ->

The sequence

bounded above. Note that the


when n is sufficiently large.

follows by induc-

(x n ) increases

convergent since

limit

is

it

positive since

increases
1

+ y/n

is

and

is

positive

x2
]

is

6 From the previous question we know that the sequence

Then
I

i9/

5.73^

follows that -the sequence

It

the harmonic

7,

Im

~n 2

converges to a positive limit

/.

follows that positive numbers a and b

It

can be found such that

\l

= m.

Both of these equations yield I

for sufficiently large values of n.


5

As

example 4.19 we apply the inequality of the geometric and

in

metic means with ai Ot

. .

fl-i

= (1

+>'/("

0) an d a n =

""<H"H"

Then
.

(n-l)/n

y
II

<

n-\

The

(-l)(l+K -!)-')+! _ L Z\
+
n

result

T4+i
is

a n are non-negative numbers

that the sequence increases

when

5.

15

For the sequence <(-

(-l)

only valid if a u a 2

then follows from the sandwich theorem and example

4.14.

Exercise

This analysis

<4 '"

Hence

and hence we have only shown

Hence

arith-

(_1)

2n

(l

-1

2"

l/2n)->

1)"(1
1

l/)> the set

1, 1}.

Observe that

as >?->

+ l/(2-l ))-*-!

(1

L = {-

as-*-

and the sequence possesses no subsequences which tend to a limit other


1

n-l

>0

i.e.n>\-y.

than

1+

in

<1+w+

Since there exists an

Note

+ ... + fi.

It

follows that

l/)

00

lim inf (- 1)"(1

l/)

= -1.

2!

N such

n\
that, for

any

that

(2)

sup (-1)"(1

n>N,

>

ff<6
obtain from (2)
iJV

<C+

_/" + !

33 <C+2

where Cis a constant.

r &

1/n)

= (-1) 2 (1 +i) =

1/n)

= (-l)l (l + 1/0 = -2.

inf (-1)"(1
n > 1

we may

example 4.19,

4"
nj

or

limsup(-l)"(l

(i)

As

2 Observe that every rational number in the interval (0, 1) occurs infinitely often as a term in the given sequence.
Let x [0, 1 ] By exercise 3.6(4), a term /_ of the sequence can be
.

found such that

x-\<r<x+

1.

>i can

term

r,

with n 2

then be found such that

192

Solutions to exercises (5.15, 5.21)

x-\<rn <x + \.

x nr

Continuing in this way we construct a subsequence

<r

>

and hence r fc

-*

The hypothesis of
all of whose terms

by

Thus /"+ e

M<M

follows from theorem 4.23 that

>

find an

n>N such that x n >

can then deduce the existence of

exists an

> /"+ e. This contradicts the definition of /"(see


The corresponding

result for

Hs

>

Suppose that [

=1=

can find an

such that, for any

A'i

5.15(3)

we

that for any e

Exercise 5.21

If>m,

5.12).

> 0, we

can find an

\x n

Let e

1.

n>N

By

the previous question

< + e.

We

can find an

such that, for any

n>N2

any n

> N, inequalities (3) and (4) are

^.

first

xn +

-* implies I

= 1 = 1 is

1-a"
a"

that a

..

<

-a

\x m+l

the last part of the question one only has to note that, if
/

<m<

< xn <

The sequence <Mn ) decreases because we


and smaller

a smaller

<Af>. It

1-a

= \/n

(n

b (n

2,

.).

Hence

set.

that

Xn +

supremum at
Any lower bound for Cc n > is also

Xn+2

<

follows from theorem 4.17 that <M> con-

-* .

> n.

Xn +

are taking the

Suppose that n -+M as n


Suppose that x n -*lasr-+. Then, for n T

X n+i =

x n + 2~ x n + l ~ X n + \X n

I.

\X n *2

a lower bound for

-x m

+ am

Next observe

trivial.

the limit of a convergent subsequence, then

verges.

+ a"' +

-x n . +

-* .

-* I as

each stage of

\x n - 1

(Example 4.12).

xn

To prove

n -1

>N,

|x-/|<e.

is

result then follows.

2 Note

That x n

xn

Exercise 4.20(3) provides the example y n

- e < xn < I + e

Urn sup
n -*

> 0, choose N so large that, for any m>N,

<e

-a

The
Provided n

true simultaneously. Hence, for

-> / as

.7).

(4)

}.

follows that

_m
,

Thus x n

(3)

l-e<x n
Take N = max {#,, N2

i.e.

it

N2

M=

we
Given e

xn

e> 0, M < /"(see example

any

<,a

be given.

that

x m < \x n - X-i +

>N,x n >/ e.

A' such that, for any n

> n} and

true that for each A'

it is

convergent subsequence with limit

true for

lim I sup x k \
n->\h > n
)

N such that, for

+ e. From exercise

is

We conclude
i.e.

<l + e.

Since this

> b.

4 Suppose it is false that, for any e > 0, there


any n >jV, x < /"+ e. Then for some e >

an upper bound for the set {x k k

is

this

we can

< 1+ e.

xk

the question implies that Ce n ) contains a subsequence

>.

sandwich theorem.

the

KM

by theorem 4.23. From theorem 4.23 again it follows


This is true for each / e L and so /"< M.
that / <
From exercise 5.1 5(4) we know that, given any e > 0, we can find
an n such that, for any k > n,

are at least as large as b. By the Bolzano- Weierstrass


subsequence in turn contains a convergent subsequence
Suppose that x -* I as r
. Since x
b (r = 1 , 2,
.), it

theorem,
<x

as

<Mn

and hence

such that

fc

193

Solutions to exercises (5.151

It

follows that

+b

'

\X n
|A

JC n

XA nr + n

Xn+l

\X n

+ i(X n

-Xn+ll

X n + i)

2,

.).

194

Solutions to exercises (5.21)


,

|x n +

-*|<

\n-l
\X 2

+b

It

Exercise 6.26

Xy\

and so we may proceed

We

3rc-2
n(n
It

Put

<"

"*"

*n + l
^n +

2~X n +

= X n+ +

2~X n

l(x2

2*n

kxt).

have

as in the previous question.

x n +2 = %(X n + x n + i),

Xn + 2

195

Solutions to exercises (6.26

X2

+ jXj.

+ 2)

l)(

+
m

+2

follows that

3-2

&

Then

fft (

!)(

(-*+!-#

=
2)

2'

+ (-i+/-l)

l-^ = l(/-^ n ).

+Hf+X-A

Hence
l*n +

It

i-/l

follows that
'*

=
-^n + 2

\\x n -l\

=...=

xn

i*n + l*n/
i

Put /={*2*.}"

3
-

/|

->0

as h

-*.

= X2 Xi.

Then

follows that x n -*

it

|*i

>

~ Xn + Xn =

Xn + 2 x n+l

and

-> / as

as

-* .

4 Let <x n > be

a sequence of points in [a, Z>]. Since [a, fcj is bounded, the


sequence <*> is bounded. By the Bolzano-Weierstrass theorem, the
sequence therefore has a convergent subsequence (x >. Suppose that
n

Xn r
a

<

-*

as r -* 00. Since a

b.

Hence

< *r < 6,

it

follows from theorem 4.23 that

Suppose that /is unbounded above. By exercise 4.29(6), we can find a


sequence <x n ) of points of / such that x n -*- + as - . But no subthis

is

a contradiction.

Hence / is bounded above. Let b = sup /. By theorem 4.29(6) we can


find a sequence 0e> of points of/ such that x n * b as n * . yl// subsequences of tx> also converge to b (theorem 5.2). Hence
Similarly for a

be I.

is

a cluster point of 5.

(theorem 4.25).

It

follows that, for

<Hb n

an

1,2,...).

Thus the convergence of S"= &n implies that of S = a by the


,

comparison
Since

test.

> 0, from exercise 4.29(2), there exists an N such

that, for

any n > N,

a n >\lb n

= inf /.

6 Let (x n ) be a sequence of distinct points of 5. Since (x n ) is bounded


contains a convergent subsequence (x n >. Suppose that x
as /

Then

A convergent sequence is bounded


some H,

(ae> converges to a point of [a, 6]

sequence of such a sequence can converge and so

it

and hence the convergence of 2=i a n implies that of 2~=i


comparison test.

* .
(i)

Take a n

and b
2

*"
00

it

follows that

n=l

= -. Since

diverges.

ft

by the

- diverges (theorem 6.5),

n=l"

it

196

00 Take an =

(iii)

Take a n

m
1

it

-. Again
n
n - j ln

(v)

V0LJW

= -5. Since

and b n

and hence the

It

We

have

afc

<z

J]

->fe

as h -* .

*-*-B
=

>na 2n >0.

4 The

as n ->

because

its

s2n

We use

'n*l

The
(ii)

=~

the ratio test.

+ l)!)
(2(/i + 1))!
2

((

vergent series with

(2n)!
*

(!)

(n

(2

l)

test

may

_ ((n+l)l )\ n+l

an + l

-^TTyy.

an

2)(2n

asn

1)

(iii)

We

From

^7?^ 4

series therefore diverges,

as

"

Hence

as

S 2n

+ 2*2n "* I s

aS

-* o.

Exercise 7.16

The

series therefore converges because e

(iv)

Again the nth root

test

may

>

be applied.

5 1

Vn

n
,n

4"
1

as n

-*

The

series therefore diverges because e

< 4.

1
1

Y//////7\
1

2n

2|

sum of the

given conditionally con-

Observe that

il

An

this identity it follows that


1

n+

I-...

2k

n+2

2n

-+2

.+.
4n

-H-

2n

f3n
1

s.

"* -

use the nth root test

be applied.

The

n+

(2n)\

sum

series therefore converges.

Again the ratio

/ith partial

-3

4fl

+ 2

"*

+4

JL + _l_ + ...+i 1

Here s denotes the


(i)

-+

+2

5.11).

terms do not tend to zero (see example

series diverges

I + i-i + i

follows from the sandwich theorem that 2na 2n "*


l)a 2n ., -*
as n -* .

Similarly (2

converges by the comparison test (see theorem

Since the sequence (a> decreases,

bn

(-l,2,...)

This series converges by theorem 6.13.

(vi)

+.. + a 2n

"'

+ V)

6.6).

6
a n+

i)

converges.

2n

series

_<1

(V( +

-5 converges (theorem

=i 1+3

follows that

= I

We have
bn

We have

diverges.

6.6),

and b n

197

Solutions to exercises (6.26, 7.16)

Solutions to exercises (6.26)

2n

198

Solutions to exercises (7.1 6

Each
range

is

meets the graph in one and only one point. The

vertical line

{2, 5}.

199

Solutions to exercises (7.16)

The image of the

[1,2]

set

The

{5}.

is

and

inverse function therefore exists

1-v
r'o) = l+y
(Note that

/=

(0<y<\).

)-

The equation

y =

-x)

4x(l

<x <

does not have a unique solution x satisfying


satisfying
y 1 For example,

< <

44(1-1) =

Thusg" does not


The equation does not define a function from R to itself.
The value x = 2 is an example of an x for which no corresponding y

-f) =

4.|(1

for

each y

I.

exist.

(i)

exists.
(ii)

The equation does not define


x = is an example of an x

value

of y (namely
(iii)

and

satisfying 1 < x <

satisfying |xj

|.v|

1,

for

1) exist.

The equation does define

each x

from [ 1, 1] to R. The
which two corresponding values

a function

4 The function /has no inverse function. When^ = 1 for example, the


2
equation^ = x has no real solution at all.
The function g has an inverse function. For each>- > there exists
2
precisely one x >
such that y = x which we denote by y/y (see
1.9). Hence

i.e.y

function from [

there

is

1, 1]

precisely one

<y <

y with

We

go fix) =

me*

g(f(x))

\=&
+

(ii)g
1

Ttl :3

(o<,<i)

l+4x(l-;t)

g(x)

=g~

have j' =g(x)if and only ifx

We have
/o<fr)

O>0).

y/y

(y).

\x\.

(i)g-

g' 1 (y)

to [0, 1]. For

4/(*)(l -/(*))

\l

-x
\l

o g(x) = -%(*)) - r'OO

o g'Hy) = g(g~

When g is

Ix

+xj

yjx

+x

(y))

g(x)

as in question 4, these

= x

(x>0)

= y

(y>o).

(\/y)

= y

(xGA)
0'

6 B).

formulae reduce to

(0<x<\).
The two formulae

give different results

when x

and hence define

different functions.

that/"

In order

satisfying

(i)

We

apply exercise 2.13. Put

sup {f{x)

[0, 1] -* [0, 1] exist,

<y <

must be true

it

that, for each

c}

iES

the equation

sup {y
y

eT

sup

7= {f(x):x<= S}. Then

+ c]
y = c + sup

l-x
(ii)

has a unique solution

satisfying

<x<

This requirement

is

easily

Let

H =

y +yx =
x =

x
-y

l+y

sup fix);
es

checked by solving the equation

Then, for
fix)

all

*e S,

+ gix)<H + K

K =

sup gix).

6S

f(x).

200

Solutions to exercises (7.16, 8.15)

and hence

H+K

an upper

is

bound

Solutions to exercises
for the set

(ii)/00
5

{f(x)+g(x):xS}.
The

that equality need not hold, take

S=

[0, 1]

and

of*

).

sup

iss

es

> 0, we

must show how to

find a

i.e. 1

= e therefore

suffices.

follows from proposition 8.4 that

It

but
sup /(*)

+. Given any e

satisfying*

The choice

sup {/(*)+ #00}


X es

-* 1

-2|<e

provided that

2 as

< x < + 5. Since we are only concerned with values


> 1, we can replace /(*) by 3 *. The condition
-e<x<l + e.
1/00 -2|<e then becomes |3 -x 2| <e,

let

-x (0 <*< 1). Then


f(x) = x(0<x<l)andg(x) =
=
+
g(x)
<
x
<
Thus
(0
f(x)
1

->

201

15)

such that

1/00

result follows.

To show

>

(8.

lim
X - 1

=1+1=2.

#00

/OO
and

exists

is

continuous

equal to 2. Note that/(l)

point

at the

and so the function

is

not

Exercise 8.15
1

Since rational functions are continuous at every point at which they are

defined (theorem 8.13),

we

lf**0,

obtain
..

1+4
1-4

x +4

W AiH
,

(ii)
v

73

im

'*-o

3*

"

5x
t;
23

42

3-

lim
* -

/OO =
e

>

lim
x -*

be given.

< |*

51

l/&)-|KlS-*l

1)

(2+*) =
We must

< 5- The

(jc> 1)

and hence /(*) *

(x<\).

7x

Observe that

(*

+ x.
we
*=

deliberately exclude considera0,

it

follows that

that

/(*)

tion of the value of the function at

4 Let

3-JC

l+2*+* 2 -l =

Since in considering a limit as

0+5.0 +
=
3.0 + 7

/OO =

% as

2.

find a 5

choice 5

->

>

such that |*

%\ < e provided

= e clearly suffices.

by the sandwich theorem (proposition

8.14).

(0/00 +
5

>

2 as*

If>0,let0<Jr<$< Y and

take %

Given any e

> 0, we must show how to find a


LY

such that

1'

Y- \x-^\<\x 1 "'-^ /n \<-X lln X- \x-^


i

l/W-2Ke
provided that

of*

satisfying*

1/00

6 <x <
<

we can

- \e < x <

provided that

Since

we

are only concerned with values

by 2x. The condition


- 2|< e, i.e. 1 - |e < x < 1 +

replace /(*)

2| < e then becomes

therefore have lo find a 5


1

=y in exercise 3. 11 (6). If

X < x < Y,

\2x

>

such that

and the sandwich theorem applies.


To prove that /(*) -* as * -* +, we must show

> 0, we can find a 6 >


x Vn m |/n_ <

e.

We

provided that
1

The choice

such that

he

6 <x <

= ie clearly

suffices.

<* <5

The choice

e" suffices.

that, given

any

202

> 0, we can

6 For any e

find a 6

>

If x

such that
(6)

|/(*)-/|<e

<

provided
true for e

=I

\x

<5
|

203

Solutions to exercises (8.20, 9.1 7)

Solutions to exercises (8.15, 8.20)

Since this

is

true for any e

> 0,

it is

= h). We

(with an appropriate value of 6, say 6

is

> X,

it

follows that (7)

= g(x).

satisfied with;-

is

y = g(x) and

satisfied with

But

what we had

this is

hence that

to prove.

The problems of theorem 8.17 arise because of the possibility that


g(x) = r? for some values of x. In the above problem 77 is replaced by

therefore

+ which

then

is

a possible value for g.

not

obtain

3 Let e

>

be given.

We

have to find an

X such that, for any x > X,

\f(x)-l\<l

= l-Kf(x)<l +

i.e.O

provided that < \x - | <


h <x < +
(i) if < 0, an A >
can be found such that /"(x) <
Z-h<x<H+handx*t.

h and*

ft, i.e.

= 0, nothing of this sort

Given //

(i)

provided that

We
can be said.

x'

is

<0,

(5) is equivalent to

x2

Given

H > 0, we

=H~

must

2 Let e

Then,

if

x > X,

Then f(x n )

find an

> X.

be given.

The choice

We must

X such

is

as n ->

and f(y n )

-*

1/n (

-*

as

1, 2,

.)

and;' n =y/2/n (n

and /(y)

as

1, 2,

-> .

.).

Then

Hence

fix)

X such

We

have

< \xf(x)\ < \x\

and so the

result follows

from the sandwich

theorem.

Exercise 9.17

that

|/fe(*))-/|<e
provided that

> X. Since /(y) -*

as >> -*

+ , we

can find a

Y such

that

k.

113
! i

>

J-

l/0)-/|<e

|.v

II

(6)
2j + 3

provided that

> Y.

(7)

Since g(x)

x>X,

as

x -* + , we can

find an

X such

that, for

any
Continuous on

2, 2)

and on

[0, 1], Recall that a rational

g(x)

>

Y.

a contradiction.

does not exist by the previous question.

that

A" = yjH suffices.

find an

(theorem 8.9). This

-*

x -*

>

* \ as x -*

that f(x)

Takex n =

lim

will suffice.

>H

provided that

/(x n )

-H' <x

(ii)

>-Hx

in turn equivalent to

and hence the choice 6

such that

result follows.

4 Suppose
If x

>

< y < 5.

X = 5~

(5)

-6 <x<0.

find a 5

such that

as

which

we can

+,

H
y = X <X~* = 5

<-H

provided that

choose

and the

> 0, we must find a 5 >

as >> -*

l/0)-/|<e

Exercise 8.20
1

provided that

(ii) If /

Since /(y) ->

^ .

function is continuous wherever


is defined.

[0,
it

Continuous on (2, 2) and on


I ]. Note that the function has

a 'corner' at

204

Solutions to exercises

,l "

Hence we can

sh

and b

find a

tinuous everywhere,
1

it

between a and b such that P()

Not continuous on ( 2, 2)
and not continuous on [0, 1]. We
have/tr) -* 1 as X 1 but

is

con-

= 0.

(Similarly if a n

< 0.)

4 Suppose that g can be found with /(g) > 0. Since fix) -* as x - + ,


we can find 6 > g such that, for any x > b, \f{x)\ < /(g). Also, since
f(x) -* as x -* , we can find a < g such that, for any x < a,
\f(x)\ </(g). By theorem 9.12, /attains a maximum on [a, b].
Suppose that /(t?) > f(x) (x E [a, b ] ). Then /(t?) > /(g) > f(x)

/(l)

< < P(b). Since P

that P(a)

follows from corollary 9.10 that there exists a g

[l

205

Solutions to exercises (9.17, 10.11)

(iv)

A>

(9. 1 7)

(x

Not continuous on (2, 2) but


continuous on [0,1].

b]).

[a,

If a |

Hence/attains

maximum on

can be found for which /(g)

< 0,

R.

then a similar argument

shows the existence of a minimum on R.

2.

If f(x)

(v)

Let Xj

I.

for

all

x, the result

is trivial.

Then there exists an x 2 E I such that |/(x 2)l < |l/C*j)l and
x 3 E I such that |/(x 3)| < \\f(x 7)\ and so on. We can

there exists an

therefore find a sequence (x n ) of points of / such that

l/(*)l

It

< \ l/(*-.)l < p

follows that /(x) *

contains a subsequence

j"-*Jer#a

1/(^-2)1

as

<je

>

<

<

-* . Since / is

l/fedl.

compact, the sequence (x n )

which converges to a point g

5.21(4)). Because /is continuous on /,/(*,.)-* /(g) as


osition 9.6).

Continuous on (

Continuous on (

and on

2, 2)

[0,1].

2, 2)

and

[0,1],

(ii)

\x n +

-* R be defined by g(x) = f{x) x


Then g is continuous on /.
Given any point g E I, let </> be a sequence of rational numbers in /
such that r -* g as n - . Then #(r) * g(g) as n * (proposition 9.6).

2 Let g: 1

=*) =
3 Let

P be

/(jc)

1, 2,

/(g)-g

.). It

follows that

.i)\<a\x n

-x n .

have

1 \.

-*il

2, 3,

.).

the equation

x n 1 = f{Xn)
odd degree

+ -=rr. +
x

that /=/(/)

n. Let

+ ...+a x+a

then P(x) -*

-ayKaP-Haa

may deduce from

>0,

+a n . x n

lfa n

a polynomial of

an x

\f(x n )-f(x n

We

As in exercise 5.21(1), we may conclude that (x n > is a Cauchy sequence


and hence that 0c n > converges. Suppose that x n -* I as n -* . Because /
therefore we
is closed we have that / E I. But /is continuous on / and

(n

(prop-

Hence
|x B+1

2).

But g(r)

-x n =

I (exercise

0.

6 That /is continuous on / follows from proposition 9.3.

bounded on (2, 2): all but (i)


maximum on (2, 2): all but (i) and (v) attain a mini-

attain a

on (2,

and on

All of the functions except (v) are

mum

Hence /(g)

r -*

'

Exercise 10.11
1

-=

+ as x -* + andP(x)^ asx -* .

We

have to consider
1

+ (x + hf

\+x 2

)
\

+x 2 -\ -x 2 -2xh-h 2
2
2
(1 + (x + h) )(l +x )

206

-2x-h
+

(i

= m

2 Let n

= D

"
Z).v

-2x
(1+ x y

+ hy)(i +x 2)

(x

where

We

prove the result by induction.

10.9(iii),

then

w*"

1 .

{Di+1fD

3 lfft>0,

and

-2 =

2/;

2-2

as

->

>x. Then, by

ft

g(0

-^

g}

withX = * and Y=y,

follows from the sandwich tlieorem that

then follows from proposition 8.4.

the other hand, if h

+ h)-g(0)

> 0,

lim
y-*x+

ft-0

->

as

ft

-*

ft

<0,

but, if ft

+ ft->-2asft->0-.
It

=2

ft

*(0

ft

That/'(1)

On

g + D fD

exercise 3.1 1(6)

v I/n_ y l/n

1-2

+/!)'+

n -i

the proof continues as in the solution to exercise 3.1 1(4).

Let.y

lfh<0,
(1

("]

Exercise 10.15

f{\+h)-f{\) _

n 1. If

h rffir-ig

DJg = i

D n + fg = I

/(1+/Q-/(1) _ 2 +

certainly holds for

It

-> 0.

as

mN. Then, by theorem

xm

207

Solutions to exercises (10.11. 10.15)

Solu tions to exercises (10.11)

-A-0

ft

as

ft

->

= -x^x~\
n

argument shows that the

= m/n

where

is

= (x g)R(x) where R

is

a polynomial of

3 (i)Z){l

+x" 27} 3/5 =

left

hand

limit is the same.

an integer and n a natural number. Then

m vn
D{x m,n } = D{(x ) } = -

ft

4 By exercise 3.1 1(3), Pfr)


degree n 1 Thus

y x

similar

2 Write

+ ft)-*(0)

|{1

+x

(x

1/2

1ln

2/S

m mx m - 1

x-

^^" 26,27

-x m/B *-

1
.

(*>)"

/'(*)

(*) +

Since P\%)

(jc

it

follows that

Dfi(*) where Q

(*

(ii)5{x

- %)R \x)
is

R () =

and

therefore /?(*)

a polynomial of degree n

flfeft)

follows that

It

(/-ft

*&()

(ft

D g,(x) = /! (xGR).
= +
Hence D kg,(x) =

-'""

1)(*

(ft

= 0,1,

(ft

2/'(l)

and the
1, /

+ 2,

.).

We now

apply these results

D p$) =

= /

+o+

<fc,

<)

(ft

+ i /o^dz^cd +
= 0,

!,..., -1).

+ xv

2}- V2

{1

+ le* +

*" 2)~ 1/2 (i +

+o

ax

{/(*)}

2f(x)f\x).

2/(l)/'(l)

result follows.

By theorem

10.13,

= Dx =

/)(/-'

obtain
1

(x

= /V)2x;

{/(* )}

When x =

= 0, 2, ...,/- 1) and

P and

2,. ..,/)

to the 'Taylor polynomial'

dx

= (*-?)'. Then
= /(/-I)...

I {x

2.
4

5 Put #,(*)

1/2
+ (x+x 1,2) 1/2 }

(theorem 10.9(h)).

/)(*)

= Df-^Dfix).

K"

2
)}

> )-

208

Solutions to exercises (10.15, 11.8)

209

Solutions to exercises (11.8)

and hence the stationary points are given by

3V(9-6)
x =

=-'(>)

^VS'

the compact interval [0, 3] is not attained at an


be achieved at a stationary point (theorem
must
endpoint, then
candidates for a point at which a maximum is
possible
only
1 1.2). The
= 0,x = 1 3" 1 ' 2 and* = 3. If /is evaluated
therefore
x
achieved are
If the

maximum on
it

each of these points, one finds that/(3) = 6 gives the largest value
2/3 \/3
and hence 6 is the maximum. A similar argument shows that
3~
=
achieved
at
x
+
1
this
is
is the minimum and
at

Observe that, for h

+ /0

(0

i/3_

i/3

H
Thus^" 1

is

> 0,

^^

as

not differentiable

-*

+.

2 Choose

the constant

h so that F

Rolle's

theorem may

be applied.

at 0.

fia)

6 If*

fo

is

rational, then

is

rational

and so

= /O) = j, = /(*) = *.
irrational, then^ = f(x) = x is irrational

f{x)

If x is

y =f(x) = x

fO f(x) =

+ hgia) =

fib)

= f + hg satisfies F(ff) = Fib) and


We

thus

need

+ hgib)

i.e.

fib) -fid)

/(/(*))

f(fix))

A thoughtless

= /O) = -y =

-fix)

gib)-gia)'
and so

We

*.

obtain the existence of a

= f\%) =

G (a, b) such that

fm+hg'm

application of theorem 10.13 would yield


i.e.

Dfix)

Dfiy)

(V =/(*))

But the function /is not differentiable

h
at

any point - thus nothing can

= -'
g'm

and the

result follows.

be deduced from theorem 10.13.


3 Apply the previous result with b

Exercise 11.8
1

The stationary points


fix)

are the roots of the equation /'(x)

=x 3 -3x 2 + 2x and so

fix)

= 3x 2 -6x +

0.

But

G (a, x) such

fa) m /(*)
g&)
g'm

x.

We deduce

the existence of a

that

'

Observe that % depends on the value of x, i.e. ' is a function of x\ By


-* a as x -* a. Moreover %i= a when x > a
exercise 8.1 5(4), we have %

and so

it

follows from theorem 8.17(ii), that

'y=x(x- l)(x-2)

fi%)
hm -7 =
,.

*-* a+ gii)

..
f(y)
hm ,,
y-**g(y)
.

provided that the second limit exists. The result follows.

lim

2
=
^~V(1
* V +y
" )
'

y -too"'

lim

X->0+
.

''

\X

21

Solu tions to exercises (11.8)

-V(l +x 2)

lim

211

Solutions to exercises (11.11)

Exercise 11.11

U -> U

Let /:

be defined by fix)

= x"

and apply Taylor's theorem

in

the form

lim
x -* o+

0.

f(y)

4 Given

that /'(*)

for
for

x and hence
all x (theorem
Rolle's

J?o.ni

/c

there exists a constant c such that fix)


1 1

%x 3 = c

may deduce the existence of numbers


that | <T?o<^i<'7i<- -<g n -i <*?-,

<g n

(i

0,

l,...,n-l).
2

way we can demonstrate

But

{n

\%)

has the property that P^"\x)

'

n)m+En+v

to observe that, for

-!)( -* +l)g"-

By

X n + / (n
1

fe
i

l)

(7?)

0.

0!

Taylor's theorem
2

= 2+^4-" 2 +^-Ki-i)4- 3 2 +j? -ia-i)a-2)'?'

0+4)

1'

where

17 lies

0.

a polynomial of degree

'

the existence of a such

that

f Mi%)-P

One then has

h (y

ft!

("

f\Vi)-P\m) =

---

and that

and

Containing in this

n(n

.7).

such

-i

j^/ni) =

theorem we

,V n

= x + and g =
= 0,l,.. .,,

Put y

all

From

follows that

= f'(x)-x =

3
}

D{f(x)-\x

= x 2 for all x, it

m)+Y^-^'^ +

between 4 and

5.

s/2

Hence

for all*.

But
6 Since g(0)

=g(i)

= 0, it

=/(l)

follows that/(0)

theorem may be applied on the

interval [0, 1].

of a g (0, 1) such that /'(g) =


theorem may be applied on the

intervals [0, g]

0.

We

and so

f{x) =

A =

But then /(g) =


and

and
[g,

1 ]

a<b,A<B.

[0, 1].

with/(g n )
1

2s

result follows.

which

fix)

/(g)

xGl,

^A?) +

.-.

^9"/

(,,

- 1)

) +

Let

inf {y

:y>b and/O) =

found between

A and B

for

Since every open subinterval of

Hence, with the given information,


l}.

(8)

Also, the continuity of /implies that

exercise 4.29(6) and theorem 8.9]. But

point g for which /(g)

which /(C)

we know

(n>

and

this is a contradiction.

[0, 1]

45/2

3 By Taylor's theorem, for any

for

fix)

-v,

and so on.

x and .y

[use
fiA) =fiB) =
that a point C can be

Let*

and the

so Rolle's

= 1 we can find a point z for which /(z) =


< a < b < and (a, b) contains no point x for

sup{x:*<aand/(.x) = \}\B

Because

5/2

obtain the existence

This argument shows that between every pair of points

which f(x) f(y)


Suppose that

Rolle's

1,

we can

[0, 1]

contains a

find a sequence g of points of

such that %n -*x as^-. This implies that

because /is continuous.

some 7? between * and g. Since / is continuous at g, it follows


from exercise 8.15(6), that, if x is taken sufficiently close to g, then
(n)
(n,
(n)
/ (7?) will have the same sign as/ (g) (provided that/ (g) > 0).
for

The problem therefore reduces


hand side of (8) under the various
(n

>

to examining the sign of the right


cases listed.

and n even. Then the right hand side of (8) is positive


x sufficiently close to g and hence /has a local minimum at g.
(i)/

>(g)

for

212

(ii)/

hand

>

(,1>

()

side

If x

and odd.

of (8)

<
(n)
(iv)/ (?) <

is

We

and n even.

(iii)/*"^!)

is

positive if x

We

sufficiently close to f, then the right

> and negative if x < %.

is

similar to

(ii)

have

fix)

maximum.

obtain a local

and n odd. This case

213

Solutions to exercises (12.12)

Solutions to exercises (11.11, 12.12)

3x 2

(*eR)

>0

and hence theorem 12.10

assures us of the existence of an inverse func-

R
f' R
lie unique value of x satisfying \=y
obviously x = 1 Hence
1

tion

see the diagram

below.

= f(x) =

+x +x

is

(i)

DrH- 1) =
5

(iv)

(iii)

f(X ~)

Df(x)

we can

A slightly
three functions given, (v)
(vii)

into class

falls

into class

(ii),

(vi) falls into

(iii).

We have f(a) < f(x) < f(b)

(a<x< b).

= -(_x+]) ll "(x+\)- --x


,

because n

i+

"'x-

<0 (x>0)

>

0(x)

>]

m=

lie in

Wm
y->*

(Kx<2)

<fi(x)=

lim

each*

+ as x

+> and /(x)

more

deduce that /is increasing given that


begin with that, if is continuous on [0,

difficult to

We show to
1 (GX<1).

1.

is

])

{0, 1}. But,

strictly increasing

on

an interval (theorem 9.9). Since f(x)


it

follows that /(R)

-*

=R

is

by theorem

9.9, the

is

con-

1],

then

image of

another interval. Hence either

= {1}. But the former case is impossible


Thus 0(x) = (0 < x <
because 0(0)
Let < x <y < 1 We know that fix) = Mix) and fiy) = M(y).
follows that / increases on [0,1].
Obviously Miy)>Mix) and
0([O, 1])

that 0([O,

an interval under a continuous function

* - as x -* ,

an endpoint.

< f(x) < Mix). It follows that

<x <

satisfying

{0} or 0([O, 1])

it

is

is

with

Mix)

4 Since/is continuous, /(R)

hence

in consistent

defined by

<X<

satisfying

We know

/W^,o.
y-**+
j
lim

+ R defined by /(a:) = x

required if %

= X and

/and. y >x, then f(y) >f{x). Hence

The function/: R
Rbut/'(0) = 0.

is

is

/(*)'

It is

that /(*)

and that 0(x) = or 0(x) =


=
Mix) for each x satisfying
If /increases on [0,1], then/(x)
=
is continuous.
(0 <x < 1). Thus
< x < 1 and hence 0(x)

(x>oy

m=M
m
y

*R

X-J

tinuous.

3 If x and y

(0, 2)

(0<x<l)

exists for

and hence

)H)

argument

different

For each x

2 Observe that
/'(*)

x e / such

Since f&)

shows that the requirement that /be increasing cannot be abandoned


(although it might be replaced by some new condition).

Exercise 12.12
1

4"

>/(-!)

*f(%) this
12.5 and hence /(-) =/( +).

The function /:

and

find an

< /(*) < /$ +)

corollary

Of the

"

Suppose that % is not an endpoint of/. To show that /is continuous at


(see corollary 12.5).
, we need to prove that/(|-) =/(?+)
If /( -) < /( +) then we can find X such that /( -) < X </( +)
and X =#/(!). Choose a and ft in / so that a < % < b. Then /(a) < X <
f(b) and so

class (i)

Df-'iy)

1 ).

214

Solutions to exercises (12.21)

Exercise 12.21
1

which

We have f(x) =

(x

- l)(x - 2)(x - 3) = x - 6x +
3

f'(x)

f"(x)

= 6x-12 = 6(x-2).

3x -12x +

- 6. Hence

lx

x3

ll

immediately that / is convex for x

The roots of the quadratic equation

3.x

> 2 and concave for x < 2.

12x

are

follows that /decreases on the interval [2

increases

A similar argument establishes

x,

on the

intervals

( , 2

1/2
]

- 3~ l/2

and [2

2
3

3" 1 ' 2

/(ax + ft)

-1/2

/"'

/,

the inequality

x 3 x 2

LetZG7and Y&J.

on
it

(9).

is

X = f(x) and

Set

< a/(x) + 0/0) =

provided that

x = 2 3" 1/2
and
and

is

-/(*) ^ /(*3)-/(*2)

/(*3)

3
If follows

215

Solutions to exercises (12.21)

'

a > 0,

(3

ctX

Y=f(y). From

+ HY

and a + (3 = 1 Since /is


on J. Thus

>

(1) of 12.13,

strictly increasing

strictly increasing

{aX+m

i
ar'(X) + $r\Y) = Qtx+$y<f~

")

and hence /"'

in

concave on J.
1
on /, then/" is strictly decreasing on
the proof is replaced by

If /is strictly decreasing


this case the last step in

J. In

af\X) + PrXY) = ax+Py>r\aX + 0Y)


and so/"

is

convex on

/.

4 By the mean value theorem, there


which

/(*)-/) _
=

= x,x 3 =y andx 2 = ax + /3y


a + 0=l,

2 Take*i

in (1)

x-%

of 12.13. Since

t]

between x and for

/'(*?)

Because /is convex,

its

and/'(r?) </'() if*

x 2 = ax +(1 -a)x 3

exists an

<

derivative increases.

Thus/'O?)

>f () 'x > %

I Hence

and

a =

x3

-x 2

X3

Xi

Thus (1) of 12.13


(*3 -*l)/(*2)

Consider

X 2 ~Xi
This
(*3

is

above any tangent drawn

)f(x 3 ).

(9)

= /)+/') (-v-S)

the inequality

^ /(*3)-/(*i)
X 3 ~X
t

equivalent to

~xi)f{x 2 )

- (x

/ concave
3

~Xi)f{x{)

< (x 2 -x,)/(*3) - (x 2 -x,)/^,)


If/ is concave the inequality should be reversed.

i.e.

The geometry of
(x 3

lies

to the graph.

i-=/U)

equivalent to

< (*3 -X 2 )f(X ) + (X -X

now

/(* 2 )-/T*i)

is

Geometrically the result asserts that the graph of the function

x 2 ~Xx
X 3 X\

-x

)f(x 2 )<(x 3

-x -x +x
1

)f(x 1 )

(x 2

-x

)f(x 3 )

the

Newton- Raphson

process

is

indicated below.

216

Solutions to exercises (12.21)

>

2f

<M*b)<I

<

any x because

for

>

<t>(x n )

<t>'(x)

and hence

(p(x +

Since <x>

is

Then

-* <t>'(l

(p'(x n )

)>0.

It

follows that

<x n+

+)

<p(x n )

as

is

/jc

{/0f0

increasing.

is

We

> /)

= f(X)=f

0.

Thus

follows that /"()

for

all

a = m/n

x,=x 2 =

+ X2 +

+ X n _] + X

<- {/(xi) + /(* 2

If
.

+ /(* - 0}

follows.
rational

number

<a <

satisfying

We may

write

= a, then = (/; w)/ Apply inequality ( 0)


=x n = y. Then
.=x m = xandx m + =x m + 2 =

&
.

with

values of % and therefore /is con-

/(ax +0>O </(*) + 0/00.

1.7).

assertion that

.+x n

-) /(*)

and P(n 1 )
Let a be a

If

Xi+X 2 +

l(n-l)X + X

Hence

/(*) >/KS) + /'(* ->-* +as* -*- ~.

6 Let P(n) be the

1) holds.

<-{/X*l) + ---+/(*n-l)+/W}.

Similarly, if /'()< 0,

It

X\

= /

obtain

> 0. Then, using question 4,


+ f'(Mm ~ D - + as x * + ..

stant (theorem

continuous

5 Suppose that /'(I)


/(*)

'

+ /(*am)}-

*'(/+)
0(/)

+ /(* 2 '" )}

'
'

'

and hence

x,

+ " + /(* m)} +


{f(Xm + t) +
2m

+x 2 + ..+X n n-\

converges. Say

* 0(/) as n -* because

-* because 0'

Then

<x.

it

+f/

+x 2 + -+* n -l
K-l

xt

X =

= -0(*n)

decreasing and bounded below,

-* I as n -* .

and

1
*

Write

and hence we would have <t>'(y) =


On the other hand, by the inequality of the

*(*n + l) -0(*n) >*'(**)(*,. + l-*n)

+*

assume that P(n) holds and seek to deduce that P{n

part of the question,

first

xn

> 0. (We cannot have <p'(x) =

0' is increasing

< x)). Thusxn+1 <x n

(y

and

{fXXl)

~
2m

We now

> %,

2m

Given that x

217

Solutions to exercises (12.21, 13.18)

an

<-{/(*.) + ...+/(*)}

irrational, consider a

is

-*

f(an x

(10)

as n -* .

n y)

Then

sequence <a> of rational numbers such that

< a n f(x) +

(11)

a n - a =

as

n f(y)

(w

2,

-> .

We

have

.)

Since /is continuous, consideration of the limit shows that (1 1) holds

any x, x 2
x n in the interval /. That P(2) holds is given. We
assume that P(2") holds and seek to establish P(2" + l ). Write m = 2".
for

Then2w =

even when

2" + 1 and

^T^i-4

is

irrational.

Exercise 13.18

-(x,+

..

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

.+x 2m

If

H,

we have

for

X"

any x

)\

/J

and hence the

result follows

from theorem 13.14.

218

If

number

replaced by a rational

is

# 1, the same argument


a

The restriction < a < b is required because x need not be


0:
defined for* <0 (e.g. if a = 2,;c is not defined for* = 0).

=
If a
1, the argument fails. (Why?)
applies.

{2{F{t)Y

is

x=

not defined when

(theorem 13.14)

] i

= 2{F(x)y 2 <2f(x).
the continuity property (theorem 9.12), the function /attains a
maximum and a minimum on the set [a, b]. Hence, by theorem

4 By

2 The integrand

219

Solutions to exercises (13.26)

Solutions to exercises (13.18, 13.26)

and, in fact, has an unpleasant

13.23,

'singularity' at this point.

3 As in example 13.17, we obtain

^n

{0

+l Q +

...

+ (-lf}

<j

xdx<-^-

{\"

function

..

is

follows that

It

F(x)

}.

F defined
/(*)

on

by

[a, b]

f g(t)dt

Ja

continuous on

F{%)

(12)

g(t) dt.

appeal to the intermediate value theorem (corollary 9.10). The

We now
a

< J o" f(t)g(t) dt<M ja"

g(t) dt

j^

[a,

b] and hence a %

[a,

b] can be found such that

t\t)g(f)dt

J a

and the

because of (12).

result follows.

Ug(t)

Suppose

some

that, for

tinuous on

[a,

< <

(a

b) and Fis

/on

a primitive for

[a,

b], then the

result reduces to

Exercise 13.26
1

b]

[a,

b],g(%)

a subinterval [c,

> 0. Then, because #

is

F(b)-F(a) = F'()(b-a)

con-

d\ containing can be found for

which

is

the

mean

value theorem

(i.e.

theorem

1.6) for the function F.

which

f(*)>k(D

(c<x<d)

We
b

(proof?). But then

ja
g(t) dt

ja

>

jg

have

g(t) dt

> \g()(d - c) > 0.

f(t) g (t) dt

Since f\t)

>

=
(a

[f(0G(t)) a

We

integrate by parts.

Then
j" t\t)g{t)dt

Ja

xf"(x)dx

and the

= [xf(pe))*-f*f'(x)dx

[a,

-G)J" oV(0 df

-G(k)im] a

[/(/)G(f)L

[f(0G(t)] a

3 Observe that F'(t)=f(t)

1)

result follows.

dt

< f

> {F(t)}
{F(t)}'

xn

U2

/()

the previous question and

b] such that

(x

noG(t) dt.

jl

< < b) we may apply

deduce the existence of


2

Jj05 ^

+ /(*) j{

(theorem 13.12). Hence


6

F\t) dt

(theorem 13.23)

We

integrate the given integral

by

parts.

Then

*'

220

Solutions to exercises (13.26, 13.34)

_ \n-lf(n-W
[(x-ty-v^Xm
-

f i

,x

+ x + x3

l+x+x

(-!)!

r~

w -...-(x- a/' + f

j-i-6
r i-8
,.

Jo

But

\l

f\t)dt

/( + ( -!)/'(!) + ...

fc-8f

~y~

J&

and therefore

1/z.

dy_

Js
J6

Then

zdz _ 1/5 dz
f
~~^~ "Jl
Z

Jl'6

-as8->-0

by example 13.33.

~1

/*-*to+J$,
+ >-

'2

as required.

(v) reduces to (vi).

dy

i-i

/(*)

-x=y.Then
& dy
dx
_
f
~ Ji
-y
v
-x

.4- 2 ^ +ooas5 - o+

(vi)Put.y

=m-rm

= **-**"*** =

i-6

1
1

and so

(x>0).

\+x-

"V5f

(v)Put

- sr ~

<

13.29 (see example 13.30).

(iv)

(*

existence of the improper integral then follows from proposition

The

+ J* (n-\)(x-t)"- 2f -"- 1) (t)dt

221

Solutions to exercises (13.34)

Observe that, by question 4 above,

= i:1/2 -y

2 /

Put .v

2v

-x)

x(\

= k - i"

dx

r
f

2x-i
2*-

1/2

Ji/2-y

jc

1/2+
1
y

-x)

J"

=\ +u

in the

v(i

and

the first integral

2x-i
x(l

-x)

second

dx.

integral.

Then
i

(x-tf^dt =

V, (7
(h-1)! /^to)
'J*

some

for

^r^/

(n)
(r?)

R!

value of tj between

x and

'

(0

"'

Jo

X
= f-2(!-x)" 2 ]'- 5
.,,
V(l -x)

2sr

-1/2

x 2 dx

x{\

i/2-y

111
3(1+

'-^f

we can

+x+x
find an

exist.

d" =

We have

ucto

(1-Z))(1+U)

_r
Jo

lv

-3

oasy^-\

as X * +

On

-1/2+2

J 1/2

+x 7 )
s

as

the other

**,

Since

4-x+;c 3)(l

not

f2s>

x)

=2-25" ^2as5->0+.

r^r^)

dv
v\

(1

Jo

integral does

3\-llX
= [-ki+*r
io
/. g
(i + x y

(iii)

^kTht)""

However,
the improper
Ho\

Exercise 13.34

- -J.

hand

2x 1
jc(1

-x)

2y

dx

After theorem 13.32

proper integral

H such

that
dx_

s:

xa

(exercise 13. 34(1 v)).

= +2 j
it is

urfzi
-

as z

*
t

(1-0)0 +v)

enough to

establish the existence of the im-

222

Solutions to exercises (13.34, 14.3)

But

rxfa
r*dx _
x

-a

provided thai

a-1

X"- 1

a-

asZ-

<log {(l+-

a>

Exercise 14.3

We

)(

2) ... (2

I,"

1)

'<7

log

have

fa

l/n

1)

(2k)

<log

= {* ^>i(2-l)>0.

log 2

(w

l/n

n-

BH"

>

1
1

223

Solutions to exercises (14.3)

l/n

l/n

(w(2/i)!

Observe that

log (2")

unbounded above on

is

strictly increasing, it follows that

= n

Thus log a:

log*

-*

(0, ). Since

+ as x

-*

+ .

follows that

it is

0<log-

Also

* asx->0+.

If

all

s log

x=

x~ \o%x

<x
s

Kx

> 0, put s = \r in
= x-

r' 2

(x-

a<

oo

0, </>(x) -
It

equation $(x)

>

case

as

-*

somewhat more

have F'(x)

|o

the above inequality.

\ogx)<x-

Then

rn

/s--0asx-+ +

As

in

= log x

and hence

is

<p'(x)

= --a.

From

this identity

a primitive

*(*><*

log /</?

[t

log

solutions

log 2 -

M
Hence

?H*<

(|

it

follows that

of <p(x) =

log(l/a)>l

+ x) dx <

and only ifx

0is non-negative,

log (4/e).

T - lo
k=i
i

k
n

on

(0, I/a]

ea<

exist if

and

3)
=

1/a. Because

and only

if the

<j>

is

continuous on

maximum

(0,

*>),

value 0(1 /a) of

loge.
is

strictly increasing,

(14) holds if and only

14 )

if

1.

The inequality e

We prove

strictly increasing

i.e.

Since the logarithm

l/a>e,i.e.

is

Thus

-tog~l

witli equality if

t-t]t

example 13.17,

,+e

-*

c.

remains true that

(exercise 14.3(2)). Consider the derivative

By theorem 13.14,

=J

It

-> asjc

= log x + x x~

]og(\+x)dx

interesting.

but, for this case, 0(.*)

strictly decreasing on [1/a, ).

for the logarithm.

that a solution of the

exists.
is

<p(x) ^ as x -*

We

+ and (p(x) -* + as x -* + (exercise

then follows from the continuity of

The

14.2(a),

(ii)Puty=x~ m(i).
3

-* .

log* ax.

14.3(2)).

log (x )

Given r

(i)

1)

>0 (see exercise 12.21(4)).


=x s Then, using theorem

Take y

as

the function 0: (0, ) -*B8 defined by

4 Consider
<p(x)

for

concave,

is

logj>-logl<(j/-

log 4\ < log 2 *

(2)!

* as - .

log 2

2 Since the logarithm

h!

2n.n\
It

and hence the logarithm

log (2"")

-<tog - <l0|

log.,

* + as n -*

n log 2

'(2)!

\ogx<x(x> 0) is simply

that the sequence (x n )

is

(13) with

decreasing and

a=

1/e.

bounded below by

224

Solutions to exercises

We have

e.

X+i

Exercise 14.5

= elogx <x

and so <x> decreases.


jc +

X = exp x and Y = exp y. By theorem


log AT = log X + log Y = x+y.

If

> e, then

xn

But
5

equation holds

and only

if

1.1.
+ i +i +

1/x.

e.

234

+... +

Ji

(x+y)
=
expx

{log 2n

- =

2n

2n

n+\

...

If r 96 1,

s-

take x

dx

(ii)0{loglogx}

\lx

* + . If r -

logx

I.

x log*

in

theorem 14.2(H).

= Hog A =

log

exp (log

= expy.

rx.

= X =
r

(expx)

r
.

> 0,

A^0asA^ + .

use

AT A =
r

(ii).

[log log xj 2 '"*

^ + as A ^ +

exercise 14.3(2ii), for any s

log

A*

as

X -*

{(log

<==.

X = exp x (see exercise 8.20(2)).

By

> 0,

+.

X)X 1,r r *

as

A ->

+.

Put

X = exp x again and

(i)

lim

Then

expx
X

the result follows.

= hm

x-*0

expx

f- =

+ asx * + .

w
,.,,
(11)

The appropriate conclusions about


theorem 13.32.

(ii)

(log A)'"A
i(log 2)

A) A~"T

{(log

only remains to put

dx =

rlog (expx)

exercise 14.3(21), for any s

(r<l)

-r

we

By

(i)

It

Then

(log*)
1

3
1

X = exp x

(log

r-1
*

Put

shows that c

Hence

- = -r in (i).

r(logx)'"

is A"

(ii)

A~ 5

x
j2

The choice x =

Hence

* log 2 + y 7asn-*.
s

(xGR)

c.

-A n

6 {i)D{\ogx} =s{\ogx}

v)

obtain

some constant

logA"

+ A 2n }-{log + A}

= log2+A 2n

We

for

11

2n -

(exp x)

exp

Then

+ y).expx expx.exp (x +

exp (x

exp x

exp(rx)

ex<p(x+y)

for a fixed value of v.

and A -> y as /? -* .
As in the solution to exercise 6.26(6),

111
+

(expx)(exp.y).

dt
i-f^

.1

...

if

XY =

XY} =

Alternatively, consider

exp {log

continuous,

Apply theorem 13.32 with/(x)

A=l

exp(x+j<)

elog.
tliis

14.2(f),

Hence

and so the fact that <x> is bounded below by e follows by induction.


Suppose that x n -* f as n -*. Since x n+1 = e log x n and the
is

Put

(i)

= elogx>eloge =

logarithm

225

Solutions to exercises (14.5)

14. 3)

the given series follow from

..

hm
-

4 We have

log(l+*)
=
x

expO =

(1+*)''

hm
,.

* -*

1.

_
=

1+0

1.

226

Solutions lo exercises (14.5)

H\t)

227

Solutions to exercises (14.5, 14. 7)

li(t)>H(t).

< \ogk<[x\ogx-x\r

[x\ogx-x\ n

(15)

Hence
T* H'(t)

Jo

dt>

n-n< log !<( +


nlogn-log! <<(+

Udt = x

>*

(x>oy

log

Thus

log

[logH(t)]

But H(0)

and

< exp

follows that

it

/?

log

+ l)-( + 1)+

(rt

l)log(n+

log/j!

1)

R!

>x.

+ ir

<n<log

i)"

+1

log//(x)>x

i.e.

h(x)

We

have

>

f"(x)

= 18.6xexp(18x

sign

4>(x)

18.3a;

exp

of f"(x)

is

it

(1

the

8x 3)

5) involves the use of an improper integral

From

side.

same

as that

exercise 14.3(2), x

+ 8.27x =

follows that

<p is

Thus

is

3
).

of

Since

exp {(x+y) logo}

exp {x log a

exp (x

(6x)

(ii)

(abf

3
,

strictly increasing for

decreasing for 6x < 2

6x

>

'3

and

strictly

non-negative on the intervals

[3, 0] and so/is concave on

left

+.)

+ y log a}

log a)

exp (y

6 Since /increases on

( , |] and
is

[0, )

exp {x log {ah)}

exp {x

exp (x log a) exp (x log b)

x x
a b

log a

+ x log ft}

(m)a~ x

non-positive on

exp

{-x

log a}

this interval.

(exercise

4.5(1 ii))

exp (x log a)

[0, )

~
(iv)

Hence

J_
a

'

Note

that, since

= exp

(x log a),

follows that

it

log (a*). Hence


1

I* max.

f(x)dx<f(k)<

But

mdx

xy

We
Z)

ft

Apply

(exercise 14.5(H))

(k<x<k+l).

f(k)<J\x)

(theorem 14.2(i))

and so

(k-l<x<k)

f(x)<f(k)

(exercise 14.5(H))

log a)

on the

1' 3

/is convex on these intervals (theorem 12.19). But

logx

asx *

-*

+ 27x 3 ).

= <t>{\) = 0.
3

+ (18.3) x exp(iax

(i)c?* y

2x + 18.3x 4 = 2x{\

Observe that 0(0)


4>'(x)

hand

(1

Exercise 14.7

/'(*)

The

(Note that

> exp x.

H(x)

J ft _

f(x)dx;
i

(he result with/(x)

= log x.

J,

/(x)dx

k 1

exp {xy log a}

exp {y log

(a*)}

x y

(a )

have

{exp (x log a)}

log a

exp (x log a)

/(*)<&.
ft

By theorem

8.9

and exercise

14.5(3ii),

(logff)a*.

log a

228

n log(l

+ xn'

229

Solutions to exercises (14.7)

Solutions to exercises (14. 7)

= xJogO+xn-

does not exist, even though

as

-* .

xn
Since the exponential function

continuous

is

at

every point,

-x

/.
it

follows

= 0->-0asA'^+.

+*'

that

+-) =

exp log

+-

exp x

as

The trick is to differentiate the given equation with respect to x


keepings constant, and then with respect toy keeping* constant. For

(i)

we obtain

each x and y,

4 By exercise
n v"

14.3(2i),

log

log n) * exp

exp

as

as

-* .

n
n

Hence,

f'(x)

= f'(x+y) =/'<

and

follows that

it

- .

f\x)
5

for

Consider

(i)

j^~e'

some constant c and hence that

f(x)

xl/2

(xGR)

ex + d.

dx.

Since /(0)

Take

<j)(x)

e~

= e~ xn

in

proposition 13.29.

We know

that

(ii)

As

in (i)

we must have d =

2/(0),

we

obtain

/V)/00 - /0)/'0)

xn dx = [-2e~* n x ->2asX-*
]

and therefore
arid

fix)

A similar argument (with


r'

J-*-

tp(x)

=e x 2 ) establishes
'

e-^dx.

DO

the existence of

log/(x)

(xGR).

Hence

follows that the improper integral

fix)

(iii)

As

/'(*)

exists.

rf.

=/(0)

Since /(0)
It

G*

cx

in (i)

(x

2
,

d=

and hence

R).

we obtain

_f(y)
x

y
(ii)See 17.3.

and therefore
(iii)

Jo

We

have

+ AT

dx

= riog(i+x

C
f\x)= -

(*>0).

2\1.Y
2

)js

Hence

| log (1

Hence the improper

xdx
+ 2
.v

+ A" 2) ^+ as *->+>.

integral

fix)

= clog* +d

Since/(l)
(iv)

As in

(x>0).

= 2/(l),d =

(i)

we

obtain

0.

0.

230

/'(*)/(?)

f'(y)f(x)

y
and therefore

we need the nth derivative of the


tion/:(-l,)--R defined by /(*) = log (1 + x). We have

To employ

fix)

log/(jc)

+x

f(x)=x

and hence

n-1

(i)

The

interval of convergence

|<7

|""

R because

is

log(i+^)

/(o)

The

interval of

(n

(iii)

The

convergence

1)!

1 )

+ as

**

U. This

is

As

is

most

easily seen

by com-

E - ==niir
(8-1)

The

(n-l)!
interval

of convergence

(4, 4). The

is

Note that the

+ 2)(2#i+l)
series

a4"

The

equal to

=
(

2(2

>

(w

as n -*

series

2,

l)

is

[ 1,

I4J

->

as

->

<
(

.).

1+'V

1).

The

radius of convergence

for*

x=

<*>.

<

"<

We

We

+ 0"

(r

-*)""'

by theorem 6.13.

x"

It is ea: ily

seen that

dr.

are concerned with the values

therefore have to consider

'-'r '*

(
i'

is

>-

(1

ffO<x<l.

as n -* .

does not converge for

(-1)"" 1

(1+0

Jo

= 4 and

<<z4"> is strictly increas-

1)

series converges

En

If-Kx<0,

+ En

^=i;
.

(o)4-^

(H-l)

have

"- 1)

(-1)"

because

Jo

interval of convergence
1

We

/(
i

is

of x for which
l)

cannot converge at the endpoints x

r
+ J)

V( +
The
The

of the interval of convergence because

*- -w dt

These formulae hold for any x

ing and hence cannot tend to zero.

(vi)

radius of convergence

4 because

(2

(n+1) 2

x=

2 rex-ff^H-l)!

(iii).

equal to

...

deriva-

where

paring the series with the Taylor series expansion for exp x.
(iv)

+-

= x

{0} because

is

of convergence

interval

/'(0)

(ii)

(v)

(-1)
(1+x)"-

(n-l)\

(A formal proof would require an induction argument.) These


and from exercise 13.26(6)
tives are evaluated at x =

as^-.

>-0

f<"\x)

(x>0).

Exercise 15.6
1

(i+*r

+ d.

= f(\f,d =

Since /(l)

f\x) =
clogjc

ftlBO

exercise 13.26(6),

/(*)

231

Solutions to exercises (15.6)

Solu tions to exercises (14.7, 15.6)

as m -* .

232

Solutions to exercises (15.6

.tx
0<-<-x
+t

(-Kx<t<0)

\En

\Ax

- < x < 0,

that, for

(-xf

From

the

+2
+ 1"

power

series

expansion for e with X

+ ...+
0<e-l+ +
2!
H!
1

we know

that

R.

1!

= ( ^)" _1 [ log (1 + *)]-


The

+ !)!

and it follows

25.?

Solutions to exercises (15.6)

expansion for log 2

series

To obtain

now

is

as

->

We

obtained simply by setting.*

the form of the remainder as given in theorem

have to write x

= 7? in

the formula for/

{n)

.10,

have to take n

1.

Rt <

= 4.

16

16

5! 5

120 5

Observe that
1

100

we

(x) in the previous question.

Suppose that e

= m/n where m

and n are natural numbers. Then

This yields

n-l

= (-D

2!

1!

Rj

1+TJ
is a

where
1

r\ lies

between

and x.

If

<x<

there

is

natural number. But,

+ 7J>1,

(n

< < 0, we can assert no more than

2)

l)!(n

1)

+ 2 ).<1
+ l) 2

j:

provided that n

\1

+X

But

(n

x
>1
1+x

+2
+ l) 2

(r

and hence

(-1<jc<-|)

and hence we are only able to show that the remainder term tends to
zero for values of x satisfying ^ < x <

We

no natural numbers
and hence we have a contradiction. (Note that

natural number. But there are

is a

in the interval (0, )

is

l)

+
(n

'

1)

obviously decreasing.)

prove by induction that, for any x

fXx-)

the previous question,

(n

n!

0<7 = n\R<
(n

If

by

no problem. Since

= Pn

4* 0,

^\e-^

(16)

We have
where P
1

Rn =
(n

1)!

(n

1)!

(n

+ 2)

1)!

+2)
1

(n

1)!

!-( + 2)i-i

(n

+ 2)( +

<

a polynomial of degree 3n.

For n

= 0, there

is

nothing to

2)!
1

is

prove. Given that (16) holds,

(n

+ 2) 2

(n

(*)

3)
=

..

-^K^ + Mt)-^

^HH&

-y 2

P'n(y) +
Observe that the polynomial Pn + 1 defined by Pn+1 Q>) =
3
2y Pn (y) has degree 3n + 3 provided that P has degree 3n. This completes the induction argument.

234

Solutions to exercises (15.6, 15.10)

We next
is

given for n

0.

On

We show that/ <n) (0) =


the assumption that/ (

/<"-(*)-/t"-'>(0)

x
It

<n>

consider the existence of the derivatives

proceed by induction.

""'

235

Solutions to exercises (15.10)

(n

" _1>

(0).

= 0,

Again we

1, 2,

(0) exists

and

.).

is

This

1)

a(a

= + I

(a

n)

(-!)!

U-L
an
n

zero,

- tt(tt-l)...(a-/i +1) ^,1

= a i+

v<

\x

follows that

lim
x -

It

j>/ViO)e~ 5 =

lim

'

follows that, for

D{(\

+xy

\x\

<

1,

= -a(I +x)- a -'/(x) +

a
f(x)}

(l

+x)- Q/'W =

by exercise 14.5(2) (exponentials drown powers). The same argument

hand

also applies with the left

limit. It follows

that/

(n)

(0) exists and

is

zero as required.

about the point

f(x)

Since /(0)

= 0,

(n

0,

1, 2,

the Taylor series expansion

.),

for

..

when x

is

c.

But /(0)

2
(n + l)
r
= hm
;
=
n - -

0.

of the power

n2

is

and hence c

given

by

1.

We know that
series
1

("-l)--(-"+l)

I
n=0

some constant

radius of convergence

fix)

(|x|<l)

2 The radius of convergence

Exercise 15.10

The

= c(\+xf

of/

is

This converges tof(x) only

and therefore

n=

n!

(W<1).

JC

Hence, using
usi
proposition 15.8,

given by

oo

CM<1)

rt!

^
=
For

|a-n|

..

lim

|jc|

<

(1
l)|

1.

(-l)x"- 2 =

(|x|<l).

we may

Therefore,
therefore apply proposition 15.8 and obtain

- + !)

a-l)

|a(-l)...(a-/i +

l)!

-^) 2

= V

J
n=l

= X2

Hence,

{n(n-l) + n}x n

(-l)x"- 2 + X.

n=2
(1

2x.22

+*)/'(*)

- a(tt-l)...(a-w+l)

A
= a+

~ tt(a-l)...(tt-H+l)

4l

(-D!
"

a(a 1)

...

a(a

(a w)

1)

(1-x) 3

..

(a

(1-x) 3

n+ 1)

3 The power

series

BX""

(1-x) 2

2x 2 +x(l + s) _ x(l+x)
=

(-D!
.

(1-x) 3

(W<1).

236

Solutions to exercises (15.10)

_y

Ifgix)

=
has the same radius of convergence as the given power series because
lim sup

lim Lap \a n
15.8, if x

F'(x)

n%

=
/ but

is

/),

g()

-b

= g'm =
= /() =

then, using proposition 15.8,

l(i-*i)

-b 2
= 3.2(a -6

*'")

2(a 2

3)

not an endpoint,

.(n+l)x n =

(x

I)

By proposition

257

Solutions to exercises (15.10, 16.3)

f{x).

and the

1)

Hence

By

result follows.

proposition 15.8,

fo-f(o) -/&#*.
n=

Observe that F(0)

= 0.

4 The function/: ( ,
fix)

From

l)->-|R defined

a n _,

by

log(l-*)
= --

the previous question

(=

a n

as required.

1,2, ..

it

follows that

.)

But then

(17)

X
has the Taylor series expansion

/(*)

(Note that formula

(1

natural to define /(0)

<y <

If

we

7) does not

make

sense

+
1+I + J J+--- (-1<*<1).

22

when x =

0,

but

= ^1 =

it is

1.)

obtain from the previous question

and, in general,

f(x)dx=y+f2 +f3 +...

^
n\

1,2,...).

and hence

Thus
r

/{*)<&

lim

But the
15.8
val

final

we know

power

that the

s,

lim

series

V
v"

/(*)

converges for .y

sum of a power

series is

1 ,

"o

its inter-

log(l-*)

(i)cos0=

We

o**.

1--+--...-1

=i n

(ii)sinO
5

Exercise 16.3
1

'

-T

and from proposition

continuous on

of convergence. Thus

-/.

n=0"l

= 0- +

-. ..=

consider the power series

*(*)=

I
n=0

(-*)(* -8".

(iii)cos(-x)

(,-xf

(~xf

i--^r + -^r

.
.

cos

238

Solutions to exercises (16.3)

(iv)sin(-*)

(-*) 3

= (-*)-

(-*) 5

3!

We

sin x.

cos *

x-*o

5!

sin

lim

(ii)

*-o 2*
lim

appeal to proposition 15.8. Then


5

Dcosx = D

4!

2!

2*

4* 3

6x

2!

4!

6!

-x+

=
Dsin* = D{x +

By

the

sin

mean

sin

= sin x.

cos

and we know, from question 3 above, that |cos | < 1.


2
The convergence of E~=1 sin (l/ ) may be deduced from the comparison test because

5!

<4:

sin

(=1,2,...).

\n

5608 *'"i + Ij----"

we may appeal

Alternatively,

From Question

to exercise 6.26(2).

4(i)

above

We have

(x+y)

cos

cos

cos y

sin

sin

y =

sin(l/
ft

lim

(*)

2
)

sin(l/)

lim

l;

j> 4-

cos*

sin

7 = g(x).

The

latter result

"*-*""

I//1

= sin (x +7) + sin* cos

'(*)

*-0

5!

3!

'(*)

and * such that

value theorem, there exists between

3!

239

Solutions to exercises (16.3, 16.5)

shows that

S~ =

1.

1/n
sin (1/n) diverges.

Hence
6

D{g(xf + h{xf) = 2g(x)g\x) + 2h{x)h\x)

=
for

all

x.

7g(x)h(x)-2h(x)g(x)

From

exercise 13.26(5)

r" sinr
sin r

dt

Jm

all

x. Since {#(*)}

all

dt

+
,

f"

between

and n,

sin t

dt

n Jf

>

(cos m cos ?} + {cos cos n\.

and (M*)} 2

> 0, we may

conclude that, for

if

Hence
r" sin

>

Let e

(*)

0.

r" sin
f"

Ji

= cos(*-*) = cos*.cos(-*)-sin*.

< cos 2* + sin 2* =

Hence

|cos x\

<

sin

2x

it

Similarly

sin t

N = 4/e. Then for any n>m>N,


4

dt
f

we

are dealing with a

Cauchy sequence. Hence

sin t
?

cos x. sin x
exists

(iv)cos2*

rm
dt-\Ji

lim
-o Ji

= sin x. cos x +

be given and choose

follows that

1.

(iii)

sin(-x).

and

< cos
|sin x| <

(iii)

<

dt

Jm

and

(i)

1
1

sin

values of x,

m
3

rJ
r(

Jm

some

follows that

It

g(xf + h(xf = g(0) 2 + h(0) 2 =


for

follows that, for

it

(theorem 5.19).

= cos x. cos* sin*. sin*.


Exercise 16.5

...

(1)

..

hm

-*

sin*

cos*
lira

cosO

1.

From

the discussion of

Since sin (*

jn)

6.4

cos*,

it

we know

that cos

>

follows immediately that

( \ti < *

< \n

).

240

Solutions to exercises (16.5)

sinx>0(0<x<7r).

But

D (arcsin v) =

= sin*

D(cosx)

and thus the cosine function decreases on

cos x, we may also deduce


[ir,

2ir]

Because cos (x

[0, r].

that the cosine function increases

+ ir) =

where x

on

We

(jn

< x < 7r) and so the cosine function

We

have
sin (x

+ n) =

Since cos

x +

sin

x=

we have

is

convex on

[\it, jt]

+ it)
+ n)

sin

D (arcsin y) =

that the arcsine function

(-Ky<\).

-y 2 )

V(i

cos x =

we know

have

(0<x<tt).

and cos (ir)


cosine function is [ 1 1 ]
Also cos (0)

tanx

= l.

Hence the image of[0, n] under the

^ 0). Also
cos x. cos x sin x ( sin x)

(provided that cos

cos
(provided that cos
follows that tan

tions are positive


sin

cos

= 0).

is

strictly increasing

x * \it and sin x

x = y.

Thus

Dcosx = -sinx<0
cos (x

tanx

sin

i.e.

take the positive sign because

We

as

= arcsin .y,

increases.

.Dtanx

cos x

= V(1 -sin 2x) = V(1 -y 2 )-

cosx

= cosx.

(cosx)

tan (x

D sm x

Further,

Thus Z> 2 (cos x) < ( %n < x < \it) and so the cosine function is concave on [ |ff,|jr]. Because cos (x + n) = cos x, D 2 (cos x) >

241

Solutions to exericses (16.5)

on

Since cos

* 1 as

(0, |ff),

it

-+ 00

->

x does not vanish on ( \n, |jr), it


on ( jt, ^it). We have cos x -

^w Since
.

the sine

and cosine func-

follows that

asx-+57r

cosx
Similarly for tan

-*

as* * ^rr +.
>(arccosj>)

We

have

>sinx

= cosx>0

We

( |<x<Jff)

-1

Dcosx

sinx

\/(l

~y 2 )

take the negative sign because the arccosine function decreases.

Note that

and hence the sine function is strictly increasing on [ 5^, \n\. Since
sin ( \t() = sin (*) and sin (\it) = 1 it follows that the image of

cos (x

+ ^tt) = sinx =

sin( x).

If x lies in the range

[ \-n,

\ti]

we may

write

x = arcsin^. But

then
.v

arcsin

cos (x

Since

^7r)

x +\n

x + %ir =
i.e.

It

\u

;'.

lies in

the range [0,

ir) it

follows that

arccos>>.

arccos>>

arcsin y.

follows from question 2 above that the tangent function

is

strictly

242

Solutions to exercises (16.5)

on ( \n, \-n) and tan x -* + as x * \n and tan *


as x-* \tj+. Thus the image of ( \ti, \ti) is R.
increasing

3'

lan

243

Solutions to exercises (16.5)

= y, we

If*

obtain

-'(&)

.v

We

'

= arctan *. Since arctan y -* j7r

apply this formula with/(*)

y-> +
jrX =

(J<>1)

>

arctan v

arctan

{arctan*}

lim

hm

arctan

1*

(
I-

";

The power
3

x
x
+
*

Z) arctan

But cos x

sin

tan

*=

cos

(y

is

and so

tan

*=

series

1/cos *. Thus

have interval of convergence (

easily seen to

We

theorem 6.13).

D arctan y =

1+tan 2*

+ y2

"
1

arctan

dx

[arctan *]

pect

toy keeping*

f {x)

arctan X-*-

constant. If*y

x+ y\

f]

as X-* + .

to* keeping y constant and then with

5 Differentiate with respect

(i

<

res-

It

1+* 2

1+* 2

arctan

(\-xyf

* = C + x

But arctan

/',

(x+y
1

xy

+y

continuous on

'

-xy)-,2

(1

\-n

arctan

*3

proposition 15.8 that the

lim_ {arctan*}

lim

* -

Thus /'(*)(!

+*

fix) =
Therefore /(*)

d=

0.

=f'(y){\

1+* 2

+y

2
)

use

C = 0.

interval of convergence.

its

x +y
+* 2
\i -*,/
id -xy?\
2

(W<1).

'"

Also

'< =

*=

< 1,

...

+ *

and so

We know from
-

|*|

follows that

Hence
f'(x)

1,

-*y)-(x+y)(-y)

(for

4
v
11**1
i

]+x2

1 ]

** + * X +...

r- + *

'

Finally,
C

apply proposition 15.8. Then, for


3

= tan *).
2

= U-

-w
=

as

(xy

< 1). Hence

(Xi

= c arctan * + d (* E M).

(i)

= /(0)

and so

*- xT + x--...

Consider the sequence <*> defined by

*n =
But 2/(0)

t-

{i7r

Then *
/(*)

-*

27r}-'.
as

sin(|ir

-* .

sum of a power

Hence

But

+ 2n7r) =

-//(()) as

series is

244

Solutions to exercises (16.5, 1 7.4)

245

Solutions to exercises (1 7.4)

- 1)

(n

- sin 2 *) sin"- 2 * <c

(1

J"'

= (n-1 ){/--/}.
= (-l)/-a

n/

i.e.

follows that

It

An +

2 Use (18) of the previous question.

We have

\g(x)\

< |x| and hence g(x) -*

as

->

by the sandwich

r
/*-

* 1 as n

-*

x
(iii)

We

have \h(x)\

h(x)-h(p)

<x

as

-J^f

We have

< sin x <

(0

< x < tt). Hence

0</n+1 </n

(n

= 0,1,2,...).

obtain the recurrence relation nln

(h

l)/- 2

It
,

we

> 2,

(2"!)

"

"

(2"m!)

n {co)
n
J

{2

x. sin" 'jtdx

cosx.(n-\)sin

(k=0,1,2,...).

1)!

^(2"+OC" a

y
V

2n+i/

Cn"

+1/

4n"'

{(2m)!}

-n

v ~"

" "2.M
,2

7T

+n

x <&

(2m

[-cosx.sin"- x]S

(2n+l)\

(2k)!

sin

f
Jo

(2 !)

~2in

Jo

/2

follows that

l 2n

0 2
..,/.
[-cosx]5/2 =
+ 1)!

sin"*

sin

'

/2 n-3

integrate by

/a + i

- 2) 2

+ 1)(2m)(2k-1)...3.2

(2m

(2n

2(2n-2)
(2m+1)(2m-1)

'"'

(2

For n

(2h-2) ...2

r*
'lo

c-o,,. 2> ...,.

(2m) (2m

and therefore

parts.

2h(2w-2)...2
(2m + 1)(2m - 1) ...

0<sin n + 1 x<sin"x

To

by the sandwich theorem.

Exercise 17.4
1

2m
2

and hence m'(0)

i2 '- 4
)

- l)(2w - 2)(2 - 3) ... 2.1

2m(2m

(2m)

-0 < Ixj ->0 as;t->-fJ

_2

(2-l)(2n-3)...l

* .

Hence

obtain

2 (2w

2k(2m-2)...2
_

We

(2h-1)(2w-3)

2/?

g(0+x n )-g(0)
g(0-x n )-g(Q)

1)

li-a-

theorem. However,

*>

(2m

-> .

as

^2n + l

+ 1 -*

2m

hn Jin-i
l<-^<

(ii)

(18)

2,3,...).

"}

y
*

{2"m!}'

n 4

4" +

}
1

e-

4"

4n^e ^n

n -2

x.cosxdx

2m+

-*-2nC~ asn^-.

By

the previous question, 27rC

1, i.e.

C = y/(2ir).

246

Solutions to exercises (1 7.4)


3

The

fact that (dn

(12m)

'/increases was demonstrated in 17.2.

As

X-

247

When m =

will

100, this is small compared with m! (though, of course, it


be a very large number compared with those usually encountered).

in

17.2,

x
x
x
*-*,-/ -j+
7 + 7 +...

Solutions to exercises (1 7.4)

4 Consider,

f->e-<dt =

Hence

js

-dn +

dn

x*
>-

3 (2w

12m 2

1)2

12m

12(/i

+ r. - e'

-t

dt

= -{Axe- A -8 x e- s } + -

Observe that
x s -*

+ 13- 12m -1
(12m + 1)(12 + 13)
12m

12m

But
1

< 5 < A, the value of

for

+ 1)+1
"

rw

+ 1)(12m+

(12m

as 5 -*

5 e~

12

x
x
A
A e' -*

= -r(x +

+.

C fe'*

A -* + (exponentials

as

dt.

J&

drown powers) and

follows that

It

l)

'

13)
as required.

It

follows that

dn

12m

>

(12m

- *,1

+ 1)(12m+

13)

12(

+ 1)4-1

- 12(12m 2 +

3(12m+1)(12m+13)(2m +

The numerator

is

12m

Let0<a<a<jc<^<6</3andlet0<5<

{!>->'**<

+ 3)

We may

1)

24-23>0 (m = 1,2, ...)


and therefore the sequence (d - (12m + l)""') decreases.
We know that dn + d as n -> and therefore
12m

dn"

12m

x-y
some

for

(m

\t

C = exp d = s/(2ir), we

obtain

the

-/*"*!*"**.

Qogtytt-*

> 0, t~r log

(20)

it

-*

as t

we can

-t y \<H{t- +

and hence

* + and

find an

r
r

log r

as r

+.

p - x }\x-y\

follows from (19) that

deduce that

is

maximum

)< m! - y/(27T)n n + U2 e-" <

It

H such that

and the continuity of the gamma function then follows from the sand-

m!(1

-e" 1 ' 12 ").

approximately equal to x for small values of x (Why?),


possible error in approximating to n! is about

6 We show that log V{\x + \y) < \ log P(x) + \ log T(y) and appeal
exercise 12.21(6). By the Schwarz inequality (theorem 13.25), if

0<5<A,
A
j"

,U* y -2),2 e - t

|6

(^-0/2 e -

/2

)(

^-l)/2 e -(/2)A
)

n\.

12m

(19)

wich theorem.

_-l/(12n + D

Since e

I**"

TO - r(y)i < h\x -j>i{r(oo + tm)

^g-l/Oln + l)

as required.

1(1
m!(1

-1

1,2,...).

n<

We may

<J" S

between x andy.

For any r

Taking exponentials and noting that

'<-

e' dt

apply the mean value theorem and obtain

follows from (20) that

~ -T < d <

y-i -tc

r*-i_,y-i

equal to

12m(14-12) + (13-36) =

dn

A. Then

<

-t
t*-\ e~'
dt

St"-"-

dt

to

248

Solutions to exercises (1

7.4, 1 7. 7)

Thus

Solutions to exercises (1

y/u

fu (U + Xy/u)

7.

249

7)

+ Xy/uf^

<T"

"*

V"

y/u{u+Xy/uy- 1 e- u -*^u
u

and the

V(27r) w-

result follows.

Exercise 17.7
1

Make

the change

-l

1+
i

Write z
use the inequality

+ nfn\ =

U log

Cv

+ n)

(n

r(n

< TO + +

1)

1)

let

<

("

+ 1)T( +

n +

y/(2Tt)x

x- U2 e- x

<y <

<y <

Then

y
l) !

and

U2

y/(2n)x

W"

x- in e-

l)WVV"

+ +
1

1.1.

,y +

>>

ey * '

for the left

+ +
=

hand

as

(1

+ Z) -

+ *)-*

x 2 as z -*

side of inequality (21)

-\i-ty- <tx

-\\-ty- <(i-ty-

d+z)-'-i]
lim

z -

2z

-(1+z) i)

that the

improper

integral exists for

(0<r<i)

(o<r<i)

y) is a continuous function
of x (and of>) is proved in the same manner as exercise 17.4(5). The
fact that its logarithm is convex is proved in the same manner as exer-

The

fact that B{x,

cise 17.4(6).

com-

5 After the previous question,

f(x

-1

we need only show

that /(l)

and

+ l) = xf(x). Now

m
+z)-z

^ log

result follows.

suffice for this purpose.


,y + i

-* .

We have
(log(l

+y

lj

pletes the proof.

lim
*-* o

-xy/u =

4 We use proposition 13.29 to check


x > and y > 0. The inequalities
t

~J

and the

,2
(y+n+\) y+ "(y+n+iyl/2-y-n-l
e

A similar argument

M.

+ l)V(2w)n

y/(2ir)x"x
x- "' e(/!

1+
8( ^l

be the largest natural number satisfying

jy

(l

log(l

< x and write x = y + n +


<

lo

p("

(21)

Consider

r(*)

ex

^j

= x/y/u. Then

1)

y
l) w!

established in the proof of theorem 17.6 for

Given x, we

y/u

j f^e-'dt.
We

e\u-l

the integral

.A

"

V(2tt)

of variable t = log u in

1/2

^B(l,y
ro)

= yj;\l-t) y -' d t =

l.

Also,ifO<a<0<l
-0

ja

m-ty-'dt = -r*-(l-f)

lim

Z-+0
It

follows that

B(x

l,y)

+ Ja
f* xf- -{\-tf dt.
x

= xy~ B(x,y +

1).

But

Consider

B(x,y

l)

r*V-

J *

(l--/f<& =

r^-'il-ty-^l-Odt

J -*

= fi(*^)-5(x

+!,>-).

250

Solutions to exercises (1 7. 7)

Hence

B(x+l,y) = -{B(x,y)-B(x+l,y)}

+-

B(x + \,y)

SUGGESTED FURTHER READING

= -d(x,y)
y

B(x+l,y) =

x +y

B(x,y).

follows that

It

Further analysis

/.

*+-&g&** + U*

J.

r(y)

(*

+y)

C. Burkill and H. Burkill, ^l

(Cambridge University

r(x +y)

This book

x+y

T(y)

B(x,y)

W. Rudin,
Since /satisfies

f(x)

= r(x)

ix

all

the conditions of

theorem 17.6,

it

follows that

> 0) and therefore

This
T.

Vt

6 Take

Tix+y)
in the

O/l jU - r
Q,0-

&r &

On

hand

r(1)

previous question.

Thus r(|) = \AWrite t = x 2ll.

is

perhaps a

little

old-

of Mathematical Analysis (McGraw-Hill, 1964).


and popular text. The treatment is rather condensed.

Principles

well-tried

is

a useful text, especially in its

treatment of vector methods.

Complex Variables and Applications (McGraw-Hill, 1960).


books above include a discussion of complex analysis (i.e. analysis
0. b ut this book is a particularly easy introduction
involving the entity i =

dx

Now

_
-Ip/K2
).
(2

H. L. Royden, Real Analysis (Macmillan, 1968).


This is an excellent book. The choice of material and manner of presentation has
1/2

2
sin

sin

flV(l

cos 6

d6

- sin 2 0)

been carefully considered. Its treatment of the Lebesque integral is particularly


It is, however, a book which makes considerable demands on the reader.

good.
ff.

G. F. Simmons, Introduction to Topology and

Modern Analysis (McGraw-Hill,

1963).

The

Then

=
Jo

=
= --r/*
2?)"
r =
"(a/r^Vd*
(

V2

title

of the
2.

V2

book describes

Logic and

set

its

content.

It is

very well-written indeed.

theory

E. A. Maxwell, Fallacies in Mathematics (Cambridge University Press, 1963).

This

Hence

exposition but

to the topic.

&b-Loy/ua-t)\-L*o

V2

its

treatment of some topics.

R. V. Churchill,

dt

Jo

Mathematical Analysis

All the

x =y =

the other

is a

notable in the clarity of


its

in

M. Aposto], Mathematical Analysis (Addison-Wesley, 1957).

This

"

is

fashioned in

xftx).

Second Course

Press, 1970).

is

for entertainment purposes.

But

it

also contains

some very

instructive

examples.
=

2../-

V(2ff).

N.

Ya

This

is

Vilenkin, Stories about Sets (Academic Press, 1968).

another entertaining book in which the paradoxes of the infinite are

explored at length.
P.

It is

highly

recommended.

R. Halmos, Naive Set Theory (Van Nostrand, 1960).


is the standard reference for elementary set theory.

This

251

252

Suggested further reading

K. Kuratowski, Introduction to Set Theory and Topology (Pergamon Press,


1972).
This is a very good

book which covers some

difficult material.

A. Margaris, First Order Mathematical Logic (Blaisdell, 1967).


Formal mathematical logic is a fascinating and expanding field. This book

NOTATION
is

good introduction.
3.

E. Moise,

Foundations of analysis

Elementary Geometry from an Advanced Standpoint (Addison-Wesley,

1963).

N.Z.Q.R

2
2

x*| /(x) * / as x - |
/(x) -4-asx- - +
/(x) - / as = * oo

oo

fai

>,>,<,<

0(h)

number system was developed largely in response to geometric needs.


This fascinating and clearly written book describes this process and many other

e.*

topics.

The

real

E. Landau, Foundations

of Analysis (Chelsea, 1951).

This remains an excellent and simple account of the algebraic foundations of the
real

number system.

1
1

-*

f(x)

/"(?).

d(x,y)

11

8y,

15,70

15

(fi,b),(a,b),[a.b)
[a,b],

(a, ), [a, )

(-,&),(-",*]
d(l S)
x - / as w
lim

n -

oo

19
-*

xn

x n "* + as n
x n * as n

28
28

"*

38

39
48

xn
n-M
liminfx n

53

fix)

f(S)

f + g.W.fg.f/g

fg

f(x)

->lasx->Z +

64
64
65
68
68
68
74

92
93

5x

93
93

94

dx

S(P)

no-

no

f"f(x)dx
f(x)dx,\

JO

f(x)dx

132

J-*a

137

logx, Inx

137 ,141

expx
a

f.A*B

D 2m)

/+),/(?-)

1',

E n

91

48

oa

n=

WB

108

lim sup

82
82

dx
df
dy,

75

92

10

max, min

74

as

>

M
sup, inf

140

141

oo

I a n (*-)"
n=0
sin x, cos

143

151

153

7T

154

tanx
arcsin x, arccos x, arctan

an

~ bn

T(x)

154

156
157

253

INDEX

Abel's lemma, 170

at a point,

Abel's theorem, 16

on an

absolute convergence

of series, 60
of power series, 144
absolute value, 10
analytic function, 145
anti-derivative,

124

77

interval,

84

continuity property, 86

continuous, 77

on

left,

on

right,

77

77
continuum property, 141

arccosine function, 154

of functions, 74

Archimedian property, 20
arcsine function, 154
arctangent function, 154

of sequences, 27

13,119
arithmetic mean,

et seq.

convergence

of series, 53
convex functions,

14

cosine function, 151

area,

decreasing

backwards induction, 23, 25


beta function, 160
binomial theorem, 25, 106, 149

function, 108

sequence, 34
degree

of polynomial, 25, 67

bounded

95

above, 13

derivative, 91,

below, 13

differentiable, 91

function, 70

differential, 93,

set,

distance

bounds, 13
Brouwer's fixed point theorem, 89

Cauchy mean value theorem, 104


Cauchy-Schwarz inequality, 7, 13
Cauchy sequence, 50
closed interval, 17, 19
cluster point,

95

differentiation, 91 et seq.

13

52

between points, 1
between point and

set,

19

divergent
functions, 82

sequences, 38
series,

54

domain of function, 64

combination theorem
for functions, 79

element of

for sequences, 30, 161

empty

set, 1

set, 1

compact interval, 17, 52


comparison test, 58
composite function, 68
composition, 68

Euler Maclaurin formula, 134

concave functions, 1 14
conditional convergence, 60

function, 64 et seq.

continuity, 84 et seq.

gamma function, 157

Euler's constant,

39

exponential function, 140

255

Index

256

geometric mean, 23

open

gradient 9

oscillating sequence,

interval, 17,

19
39

graph, 64

sum, 53
partition, 120
periodicity, 153
partial

polynomial, 25, 66
powers, 141

of point, 64
of set, 65
improper integral, 132

power

increasing

principle of induction, 22

series,

143 et seq.

primitive, 124

function, 108

sequence, 33

quadratic equation, 6

22

induction,

radius of convergence, 144

inequalities, 3

range

infimum, 15,71
infinitesimal, 94

of function, 65
of sequence, 27
rational function, 67
rational number, 2

integer, 2
integral,

120
135

integral test,

integration,

by

parts,

ratio test, 59, 144,

19 el seq.

real

129

intermediate value theorem, 87


interval, 16

128
102

integral,

Rolle's theorem,
roots, 6,

of convergence, 144

164

number, 2

Riemann

112

inverse function, 68, 1 10


irrational

number,

sandwich theorem

2, 9

for functions, 80
for sequences, 3

Leibniz's rule, 97
limit

sequence, 27

of function, 74

of sequence, 27; inferior, 48;

48
limit point of set, 52
local maximum, 100
local minimum, 100
logarithm function, 137

series,

53

set, 1

superior,

sine function, 151

slope, 91

stationary point, 101


Stirling's

lower bound, 14

formula, 156, 160

strictly increasing or decreasing

function, 108

maximum, 15, 71
mean value theorem, 102
minimum,

sequence, 34
subset,

15, 71

Minkowski's inequality, 8,10


modulus, 10

monotone

sum of series, 53
supremum, 15, 70
tail

of series, 58

function, 108

tangent function, 154

sequence, 34

Taylor polynomial, 97
Taylor series, 145
Taylor's theorem, 105

natural logarithm, 138

natural number, 2, 20

Newton-Raphson
nth root

test,

process,

59, 144, 164

tends to a limit, 27, 74

117

50

unbounded
set,

upper bound,

point of accumulation, 52

image

uniform convergence, 147

triangle inequality, 10

trigonometric function,

harmonic mean, 8
L'Hopital's rule, 104

257

Index

term of sequence, 27
transcendental number, 10

14

function, 71

Jacket design by Ken Vail


Printed in Great Britain

Surfaces
H.B.GRIFFITHS
in terms of everyday experience.
by surfaces?' he goes on to
we
mean
'What
do
Beginning with the question,
of three-dimensional
examination
on
an
based
build up a set of definitions
the more advanced formal
these
to
relate
to
only
then
models and drawings, and
readers
to inject spatial
for
background
provides
the
mathematics. He thus

Professor Griffiths relates and explains toplogy

concepts into their mathematical thinking. The text

is

well illustrated with toned

where appropriate and there are


drawing which create a three-dimensional effect
abundant exercises with hints to their

Partial Differential

solution.

Equations

E.T.COPSON
'Copson aims
There
It

is,

at

both mathematicians and users of

again, nothing available like this

amalgam

suddenly becomes clear as one reads how

part of a surprisingly elegant

whole

...

partial differential equations.

of pure theory

the various tricks

The book

will

fall

and methods.
into place as

form a perfect advanced


There

or physics
undergraduate or early postgraduate course in mathematics
chapter.'
is a good selection of examples in each

The Times Higher Education Supplement

Cambridge University Press


521 21480 7