A Course in Modern Analysis

and Its Applications

Graeme L. Cohen
University of Technology, Sydney

p u b l i sh e d b y t h e p r e ss sy n di c a t e o f t h e u n i v e r si t y o f c a m b r i dg e
The Pitt Building, Trumpington Street, Cambridge, United Kingdom
c a m b r i dg e u n i v e r si t y p r e ss
The Edinburgh Building, Cambridge CB2 2RU, UK
40 West 20th Street, New York, NY 10011-4211, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
Ruiz de Alarc
on 13, 28014 Madrid, Spain
Dock House, The Waterfront, Cape Town 8001, South Africa
http://www.cambridge.org
c Cambridge University Press 2003


This book is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the
written permission of Cambridge University Press
First published 2003
Printed in the United Kingdom at the University Press, Cambridge
Typeface Computer Modern 10/13.

System LATEX 2 [Typeset by the author]

A catalogue record for this book is available from the British Library
ISBN 0 521 81996 2 hardback
ISBN 0 521 52627 2 paperback

Contents

Preface

page ix

Prelude to Modern Analysis

1.1 Introduction
1.2 Sets and numbers
1.3 Functions or mappings
1.4 Countability
1.5 Point sets
1.6 Open and closed sets
1.7 Sequences
1.8 Series
1.9 Functions of a real variable
1.10 Uniform convergence
1.11 Some linear algebra
1.12 Setting off

Metric Spaces
2.1 Definition of a metric space
2.2 Examples of metric spaces
2.3 Solved problems
2.4 Exercises
2.5 Convergence in a metric space
2.6 Examples on completeness
2.7 Subspace of a metric space
2.8 Solved problems
2.9 Exercises

84
84
85
94
96
98
102
107
109
111

The Fixed Point Theorem and its Applications

3.1 Mappings between metric spaces
3.2 The fixed point theorem

113
113
115

1
1
3
10
14
20
28
32
44
52
59
69
83

Contents

vi
3.3
3.4
3.5

Applications
Perturbation mappings
Exercises

118
130
135

Compactness
4.1 Compact sets
4.2 Ascolis theorem
4.3 Application to approximation theory
4.4 Solved problems
4.5 Exercises

140
140
145
149
152
153

Topological Spaces
5.1 Definitions and examples
5.2 Closed sets
5.3 Compact sets
5.4 Continuity in topological spaces
5.5 Homeomorphisms; connectedness
5.6 Solved problems
5.7 Exercises

155
155
158
160
164
167
170
171

Normed Vector Spaces

6.1 Definition of a normed vector space; examples
6.2 Convergence in normed spaces
6.3 Solved problems
6.4 Exercises
6.5 Finite-dimensional normed vector spaces
6.6 Some approximation theory
6.7 Chebyshev theory
6.8 The Weierstrass approximation theorem
6.9 Solved problems
6.10 Exercises

174
174
178
181
185
187
192
195
199
205
208

Mappings on Normed Spaces

7.1 Bounded linear mappings
7.2 Norm of an operator
7.3 Functionals
7.4 Solved problems
7.5 Exercises
7.6 Inverse mappings
7.7 Application to integral equations
7.8 Application to numerical analysis
7.9 Exercises
7.10 Unbounded mappings

210
210
214
218
221
223
225
229
235
243
246

Contents
8

Inner Product Spaces

8.1 Definitions; simple consequences
8.2 Orthonormal vectors
8.3 Least squares approximation
8.4 The Riesz representation theorem
8.5 Solved problems
8.6 Exercises

vii
251
251
259
267
273
275
278

Hilbert Space
281
9.1 Definition of Hilbert space
281
9.2 The adjoint operator
282
9.3 Separability
288
291
9.4 Solved problems
9.5 Exercises
292
9.6 Complete orthonormal sets; generalised Fourier series 294
9.7 Hilbert space isomorphism
303
9.8 Exercises
307
Bibliography
309
Selected Solutions
312
Index
330

1
Prelude to Modern Analysis

1.1 Introduction
The primary purpose of this chapter is to review a number of topics
from analysis, and some from algebra, that will be called upon in the
following chapters. These are topics of a classical nature, such as appear
in books on advanced calculus and linear algebra. For our treatment of
modern analysis, we can distinguish four fundamental notions which will
be particularly stressed in this chapter. These are
(a)
(b)
(c)
(d)

set theory, of an elementary nature;

the concept of a function;
convergence of sequences; and
some theory of vector spaces.

On a number of occasions in this chapter, we will also take the time

to discuss the relationship of modern analysis to classical analysis. We
begin this now, assuming some knowledge of the points (a) to (d) just
mentioned.
Modern analysis is not a new brand of mathematics that replaces the
old brand. It is totally dependent on the time-honoured concepts of
classical analysis, although in parts it can be given without reference to
the specifics of classical analysis. For example, whereas classical analysis
is largely concerned with functions of a real or complex variable, modern
analysis is concerned with functions whose domains and ranges are far
more general than just sets of real or complex numbers. In fact, these
functions can have domains and ranges which are themselves sets of
functions. A function of this more general type will be called an operator
or mapping. Importantly, very often any set will do as the domain of a
mapping, with no specific reference to the nature of its elements.
1

1 Prelude to Modern Analysis

This illustrates how modern analysis generalises the ideas of classical

analysis. At the same time, in many ways modern analysis simplifies
classical analysis because it uses a basic notation which is not cluttered
with the symbolism that characterises many topics of a classical nature.
Through this, the unifying aspect of modern analysis appears because
when the symbolism of those classical topics is removed a surprising
similarity becomes apparent in the treatments formerly thought to be
peculiar to those topics.
Here is an example:

b
k(s, t)x(t) dt = f (s), a
s
b,
a

is an integral equation; f and k are continuous functions and we want

to solve this to find the continuous function x. The left-hand side shows
that we have operated on the function x to give the function f , on the
right. We can write the whole thing as
Kx = f,
where K is an operator of the type we just mentioned. Now the essence of
the problem is clear. It has the same form as a matrix equation Ax = b,
for which the solution (sometimes) is x = A1 b. In the same way, we
would like the solution of the integral equation to be given simply as
x = K 1 f . The two problems, stripped of their classical symbolism,
appear to be two aspects of a more general study.
The process can be reversed, showing the strong applicability of modern analysis: when the symbolism of a particular branch of classical
analysis is restored to results often obtained only because of the manipulative ease of the simplified notation, there arise results not formerly
obtained in the earlier theory. In other cases, this procedure gives rise
to results in one field which had not been recognised as essentially the
same as well-known results in another field. The notations of the two
branches had fully disguised the similarity of the results.
When this occurs, it can only be because there is some underlying
structure which makes the two (or more) branches of classical analysis
appear just as examples of some work in modern analysis. The basic entities in these branches, when extracted, are apparently combined
together in a precisely corresponding manner in the several branches.
This takes us back to our first point of the generalising nature of modern analysis and of the benefit of working with quite arbitrary sets. To
combine the elements of these sets together requires some basic ground

1.2 Sets and numbers

rules and this is why, very often and predominantly in this book, the
sets are assumed to be vector spaces: simply because vector spaces are
sets with certain rules attached allowing their elements to be combined
together in a particular fashion.
We have indicated the relevance of set theory, functions and vector
spaces in our work. The other point, of the four given above, is the
springboard that takes us from algebra into analysis. In this book,
we use in a very direct fashion the notion of a convergent sequence to
generate virtually every result.
We might mention now, since we have been comparing classical and
modern analysis, that another area of study, called functional analysis,
may today be taken as identical with modern analysis. A functional is a
mapping whose range is a set of real or complex numbers and functional
analysis had a fairly specific meaning (the analysis of functionals) when
the term was first introduced early in the 20th century. Other writers
may make technical distinctions between the two terms but we will not.
In the review which follows, it is the aim at least to mention all topics
required for an understanding of the subsequent chapters. Some topics,
notably those connected with the points (a) to (d) above, are discussed
in considerable detail, while others might receive little more than a definition and a few relevant properties.

1.2 Sets and numbers

A set is a concept so basic to modern mathematics that it is not possible
to give it a precise definition without going deeply into the study of
mathematical logic. Commonly, a set is described as any collection of
objects but no attempt is made to say what a collection is or what
an object is. We are forced in books of this type to accept sets as
fundamental entities and to rely on an intuitive feeling for what a set is.
The objects that together make up a particular set are called elements
or members of that set. The list of possible sets is as long as the imagination is vivid, or even longer (we are hardly being precise here) since,
importantly, the elements of a set may themselves be sets.
Later in this chapter we will be looking with some detail into the properties of certain sets of numbers. We are going to rely on the readers
experience with numbers and not spend a great deal of time on the development of the real number system. In particular, we assume familiarity
with

1 Prelude to Modern Analysis

(a) the integers, or whole numbers, such as 79, 3, 0, 12, 4,063,180;
(b) the rational numbers, such as 35 , 11
17 , which are numbers expressible as a ratio of integers (the integers themselves also being
examples);
(c) those numbers
are not rational, known as irrational num which
bers, such as 2, 3 15, ;
(d) the real numbers, which are numbers that are either rational or
irrational;
(e) the ordering of the real numbers, using the inequality signs <
and > (and the use of the signs
and );
(f) the representation of the real numbers as points along a line; and
(g) the fact, in (f), that the real numbers fill the line, leaving no
holes: to every point on the line there corresponds a real number.

The final point is a crucial one and may not appear to be so familiar.
On reflection however, it will be seen to accord with experience, even
when expressed in such a vague way. This is a crude formulation of
what is known as the completeness of the real number system, and will
be referred to again in some detail subsequently.
By way of review, we remark that we assume the ability to perform simple manipulations with inequalities. In particular, the following
should be known. If a and b are real numbers and a < b, then
a > b;
1
1
> , if also a > 0 or b < 0;
a
b

a < b, if also a  0.
With regard to the third property, we stress that the use of the radi
) always implies that the nonnegative root is to be taken.
cal sign (
Bearing this comment in mind, we may define the absolute value |a| of
any real number a by

|a| = a2 .
More commonly, and equivalently of course, we say that |a| is a whenever
a > 0 and |a| is a whenever a < 0, while |0| = 0. For any real numbers
a and b, we have
|a + b|
|a| + |b|,

|ab| = |a| |b|.

These may be proved by considering the various combinations of positive

and negative values for a and b.

1.2 Sets and numbers

We also assume a knowledge of complex numbers: numbers of the

form a + ib where a and b are real numbers and i is an imaginary unit,
satisfying i2 = 1.
This is a good place to review a number of definitions and properties
connected with complex numbers. If z
= a+ib is a complex number, then
we call the numbers a, b, a ib and a2 + b2 the real part, imaginary
part, conjugate and modulus, respectively, of z, and denote these by
Re z, Im z, z and |z|, respectively. The following are some of the simple
properties of complex numbers that we use. If z, z1 and z2 are complex
numbers, then
z = z,
z1 + z2 = z 1 + z 2 ,
z1 z2 = z 1 z 2 ,
| Re z|
|z|, | Im z|
|z|,
zz = |z|2 ,
|z1 + z2 |
|z1 | + |z2 |,
|z1 z2 | = |z1 | |z2 |.
It is essential to remember that, although z is a complex number, the
numbers Re z, Im z and |z| are real. The final two properties in the
above list are important generalisations of the corresponding properties
just given for real numbers. They can be generalised further, in the
natural way, to the sum or product of three or four or more complex
numbers.
Real numbers, complex numbers, and other sets of numbers, all occur
so frequently in our work that it is worth using special symbols to denote
them.
Definition 1.2.1 The following symbols denote the stated sets:
N, the set of all positive integers;
Z, the set of all integers (positive, negative and zero);
Q, the set of all rational numbers;
R, the set of all real numbers;
R+ , the set of all nonnegative real numbers;
C, the set of all complex numbers.
Other sets will generally be denoted by ordinary capital letters and their
elements by lower case letters; the same letter will not always refer to
the same set or element. To indicate that an object x is an element

1 Prelude to Modern Analysis

of a set X, we will write x X; if x is not

an element of X, we will
/ Z. A statement such
write x
/ X. For example, 2 R but 2
as x, y X will be used as an abbreviation for the two statements
x X and y X. To show the elements of a set we always enclose
them in braces and give either a complete listing (for example, {1, 2, 3}
is the set consisting of the integers 1, 2 and 3), or an indication of a
pattern (for example, {1, 2, 3, . . . } is the set N), or a description of a
rule of formation following a colon (for example, {x : x R, x  0} is
the set R+ ). Sometimes we use an abbreviated notation (for example,
{n : n = 2m, m N} and {2n : n N} both denote the set of all even
positive integers).
An important aspect in the understanding of sets is that the order
in which their elements are listed is irrelevant. For example, {1, 2, 3},
{3, 1, 2}, {2, 1, 3} are different ways of writing the same set. However,
on many occasions we need to be able to specify the first position, the
second position, and so on, and for this we need a new notion. We speak
of ordered pairs of two elements, ordered triples of three elements, and,
generally, ordered n-tuples of n elements with this property that each
requires for its full determination a list of its elements and the order
in which they are to be listed. The elements, in their right order, are
enclosed in parentheses (rather than braces, as for sets). For example,
(1, 2, 3), (3, 1, 2), (2, 1, 3) are different ordered triples. This is not an
unfamiliar notion. In ordinary three-dimensional coordinate geometry,
the coordinates of a point provide an example of an ordered triple: the
three ordered triples just given would refer to three different points in
space.
We give now a number of definitions which help us describe various
manipulations to be performed with sets.
Definition 1.2.2
(a) A set S is called a subset of a set X, and this is denoted by S X
or X S, if every element of S is also an element of X.
(b) Two sets X and Y are called equal, and this is denoted by X = Y ,
if each is a subset of the other; that is, if both X Y and Y X.
Otherwise, we write X = Y .
(c) A set which is a subset of any other set is called a null set or
empty set, and is denoted by .
(d) A set S is called a proper subset of a set X if S X, but S = X.
(e) The union of two sets X and Y , denoted by X Y , is the set of

1.2 Sets and numbers

elements belonging to at least one of X and Y ; that is,

X Y = {x : x X or x Y (or both)}.
(f) The intersection of two sets X and Y , denoted by X Y , is the
set of elements belonging to both X and Y ; that is,
X Y = {x : x X and x Y }.
(g) The cartesian product of two sets X and Y , denoted by X Y ,
is the set of all ordered pairs, the first elements of which belong
to X and the second elements to Y ; that is,
X Y = {(x, y) : x X, y Y }.
(h) The complement of a set X, denoted by X, is the set of elements
that do not belong to X; that is, X = {x : x
/ X}. The
complement of X relative to a set Y is the set Y X; this is
denoted by Y \X.
For some simple examples illustrating parts of this definition, we let
X = {1, 3, 5} and Y = {1, 4}. Then
X Y = {1, 3, 4, 5},

X Y = {1},

X Y = {(1, 1), (1, 4), (3, 1), (3, 4), (5, 1), (5, 4)},
Y X = {(1, 1), (1, 3), (1, 5), (4, 1), (4, 3), (4, 5)}.
We see that in general X Y = Y X. The set Y \X is the set of
elements of Y that do not belong to X, so here Y \X = {4}.
The definitions of union, intersection and cartesian product of sets
can be extended to more than two sets. Suppose we have n sets X1 , X2 ,
. . . , Xn . Their union, intersection and cartesian product are defined as
X1 X2 Xn =

n


Xk

k=1

= {x : x Xk for at least one k = 1, 2, . . . , n},

n

Xk
X1 X2 Xn =
k=1

= {x : x Xk for all k = 1, 2, . . . , n},

n

Xk
X1 X2 Xn =
k=1

= {(x1 , x2 , . . . , xn ) : xk Xk for k = 1, 2, . . . , n},

1 Prelude to Modern Analysis

respectively (the cartesian product being a set of ordered n-tuples). The

notations in the middle are similar to the familiar sigma notation for
addition, where we write
x1 + x2 + + xn =

n


xk ,

k=1

when x1 , x2 , . . . , xn are numbers.

For cartesian products only, there is a further simplification of notation when all the sets are equal. If X1 = X2 = = Xn = X, then


in place of nk=1 Xk or nk=1 X we write simply X n , as suggested by
the notation, but note that there is no suggestion of multiplication:
X n is a set of n-tuples. In particular, it is common to write Rn for the
set of all n-tuples of real numbers and Cn for the set of all n-tuples of
complex numbers.
It is necessary to make some comments regarding the definition of
an empty set in Definition 1.2.2(c). These are gathered together as a
theorem.
Theorem 1.2.3
(a) All empty sets are equal.
(b) The empty set has no elements.
(c) The only set with no elements is the empty set.
To prove (a), we suppose that 1 and 2 are any two empty sets.
Since an empty set is a subset of any other set, we must have both
1 2 and 2 1 . By the definition of equality of sets, it follows
that 1 = 2 . This proves (a) and justifies our speaking of the empty
set in the remainder of the theorem. We prove (b) by contradiction.
Suppose x . Since for any set X we have X and X,
we must have both x X and x X. This surely contradicts the
existence of x, proving (b). Finally, we prove (c), again by contradiction.
Suppose X is a set with no elements and suppose X = . Since X,
this means that X is not a subset of . Then there must be an element
of X which is not in . But X has no elements so this is the contradiction
we need.
All this must seem a bit peculiar if it has not been met before. In
defence, it may be pointed out that sets were only introduced intuitively
in the first place and that the inclusion in the concept of a set with no
elements is a necessary addition (possibly beyond intuition) to provide

1.2 Sets and numbers

consistency elsewhere. For example, if two sets X and Y have no elements in common and we wish to speak of their intersection, we can
now happily say X Y = . (Two such sets are called disjoint.)
Manipulations with sets often make use of the following basic results.
Theorem 1.2.4 Let X, Y and Z be sets. Then
(a) (X) = X,
(b) X Y = Y X and X Y = Y X (commutative rules),
(c) X (Y Z) = (X Y ) Z and X (Y Z) = (X Y ) Z
(associative rules),
(d) X(Y Z) = (XY )(XZ) and X(Y Z) = (XY )(XZ)
(distributive rules).
We will prove only the second distributive rule. To show that two
sets are equal we must make use of the definition of equality in Definition 1.2.2(b).
First, suppose x X (Y Z). Then x X and x Y Z. That
is, x is a member of X and of either Y or Z (or both). If x Y then
x X Y ; if x Z then x X Z. At least one of these must be true, so
x (X Y )(X Z). This proves that X (Y Z) (X Y )(X Z).
Next, suppose x (X Y ) (X Z). Then x X Y or x X Z
(or both). In both cases, x X (Y Z) since in both cases x X,
and Y Y Z and Z Y Z. Thus X (Y Z) (X Y ) (X Z).
Then it follows that X (Y Z) = (X Y ) (X Z), completing
this part of the proof.
The following theorem gives two of the more important relationships
between sets.
Theorem 1.2.5 (De Morgans Laws) Let X, Y and Z be sets. Then
Z\(X Y ) = Z\X Z\Y

and

Z\(X Y ) = Z\X Z\Y.

There is a simpler form of de Morgans laws for ordinary complements:

(X Y ) = X Y

and (X Y ) = X Y.

To prove the first of these, suppose x (X Y ). Then x

/ X Y so
either x
/ X or x
/ Y . That is, x X or x Y , so x X Y .
This proves that (XY ) XY . Suppose next that x XY .
If x X then x
/ X so x
/ X Y , since X Y X. That is,
x (X Y ). The same is true if x Y . Thus X Y (X Y ),
so we have proved that (X Y ) = X Y .

10

1 Prelude to Modern Analysis

We can use this, the definition of relative complement, and a distributive rule from Theorem 1.2.4 to prove the first result of Theorem 1.2.5:
Z\(X Y ) = Z (X Y ) = Z (X Y )
= (Z X) (Z Y ) = Z\X Z\Y.
The second of de Morgans laws is proved similarly.

Review exercises 1.2

(1) Let a and b be real numbers. Show that
(a) ||a| |b||
|a b|,
(b) |a b| <  if and only if b  < a < b + ,
(c) if a < b +  for every  > 0 then a
b.
(2) Suppose A B = X. Show that X Y = (A Y ) (B Y ), for
any set Y .
(3) For any sets A and B, show that
(a) A\B = A if and only if A B = ,
(b) A\B = if and only if A B.

1.3 Functions or mappings

We indicated in Section 1.1 how fundamental the concept of a function is
in modern analysis. (It is equally important in classical analysis but may
be given a restricted meaning there, as we remark below.) A function
is often described as a rule which associates with an element in one
set a unique element in another set; we will give a definition which
avoids the undefined term rule. In this definition we will include all
associated terms and notations that will be required. Examples and
general discussion will follow.
Definition 1.3.1 Let X and Y be any two nonempty sets (which
may be equal).
(a) A function f from X into Y is a subset of X Y with the property
that for each x X there is precisely one element (x, y) in the
subset f . We write f : X Y to indicate that f is a function
from X into Y .
(b) The set X is called the domain of the function f : X Y .

1.3 Functions or mappings

11

(c) If (x, y) f for some function f : X Y and some x X, then

we call y the image of x under f , and we write y = f (x).
(d) Let S be a subset of X. The set
{y : y Y, y = f (x) for some x S},
which is a subset of Y , is called the image of the set S under
f : X Y , and is denoted by f (S). The subset f (X) of Y is
called the range of f .
(e) When f (X) = Y , we say that the function f is from X onto Y
(rather than into Y ) and we call f an onto function.
(f) If, for x1 , x2 X, we have f (x1 ) = f (x2 ) only when x1 = x2 ,
then we call the function f : X Y one-to-one.
(g) An onto function is also said to be surjective, or a surjection. A
one-to-one function is also said to be injective, or an injection. A
function that is both injective and surjective is called bijective, or
a bijection.
Enlarging on the definition in (a), we see that a function f from a set X
into a set Y is itself a set, namely a set of ordered pairs chosen from
X Y in such a way that distinct elements of f cannot have distinct
second elements with the same first element. In (c), we see that the
common method of denoting a function as y = f (x) is no more than an
alternative, and more convenient, way of writing (x, y) f . Notice the
different roles played by the sets X and Y . The set X is fully used up
in that every x X has an image f (x) Y , but the set Y need not be
used up in that there may be a y Y , or many such, which is not the
image of any x X. Paraphrasing (e), when in fact each y Y is the
image of some x X, then the function is called onto. Notice that the
same term image is used slightly differently in (c) and (d), but this will
not cause any confusion.
It follows from Definition 1.2.2(b) that two functions f and g from X
into Y are equal if and only if f (x) = g(x) for all x X.
In Figure 1, four functions
fk : X Yk ,

k = 1, 2, 3, 4,

are illustrated. Each has domain X = {1, 2, 3, 4, 5}. The function

f1 : X Y1 has Y1 = {1, 2, 3, 4, 5, 6} and the function is the subset {(1, 3), (2, 3), (3, 4), (4, 1), (5, 6)} of X Y1 , as indicated by arrows
giving the images of the elements of X. The range of f1 is the set
f1 (X) = {1, 3, 4, 6}. The other functions may be similarly described.

12

1 Prelude to Modern Analysis

1
1
1

1
2

2
3

3
4

4
5

5
6

7
f1 : X Y1

f2 : X Y2

f3 : X Y3

f4 : X Y4

Figure 1

For all four functions, each element of X is the tail of an arrow and of
only one arrow, while the elements of the Y s may be at the head of
more than one arrow or perhaps not at the head of any arrow. This
situation is typical of any function. The elements of Y2 and Y4 are all at
heads of arrows, so the functions f2 and f4 are both onto. Observe that
f1 (1) = 3 and f1 (2) = 3. Also, f2 (1) = 3 and f2 (5) = 3. This situation
does not apply to the functions f3 and f4 : each element of Y3 and Y4 is
at the head of at most one arrow, so the functions f3 and f4 are both
one-to-one.
Only the function f4 is both one-to-one and onto: it is a bijection.
This is a highly desirable situation which we pursue further in Chapters 5 and 7, though we briefly mention the reason now. Only for the
function f4 of the four functions can we simply reverse the directions of
the arrows to give another function from a Y into X. We will denote
this function temporarily by g : Y4 X. In full:
f4 = {(1, 2), (2, 3), (3, 1), (4, 5), (5, 4)},
g = {(1, 3), (2, 1), (3, 2), (4, 5), (5, 4)}.
The function g is also a bijection, and has the characteristic properties
g(f4 (x)) = x for each x X,
f4 (g(y)) = y

for each y Y4 .

1.3 Functions or mappings

13

We call g the inverse of the function f4 , and denote it by f41 . The

precise definition of this term follows.
Definition 1.3.2 For any bijection f : X Y , the inverse function
f 1 : Y X is the function for which
f 1 (y) = x whenever f (x) = y,
where x X and y Y .
It follows readily that if f is a function possessing an inverse function,
then f 1 also has an inverse function and in fact (f 1 )1 = f .
It is sometimes useful in other contexts to speak of the inverse of a
function when it is one-to-one but not necessarily onto. This could be
applied to the function f3 : X Y3 , above. We can reverse the arrows
there to give a function h, but the domain of h would only be f3 (X) and
not the whole of Y3 .
The following definition gives us an important method of combining
two functions together to give a third function.
Definition 1.3.3 Let f : X Y and g : Y Z be two functions.
The composition of f with g is the function g f : X Z given by
(g f )(x) = g(f (x)),

x X.

Note carefully that the composition g f is only defined when the range
of f is a subset of the domain of g. It should be clear that in general
the composition of g with f , that is, the function f g, does not exist
when g f does, and even if it does exist it need not equal g f .
For example, consider the functions f1 and f4 above. Since Y4 = X,
we may form the composition f1 f4 (but not f4 f1 ). We have
(f1 f4 )(1) = f1 (f4 (1)) = f1 (2) = 3,
and so on; in full, f1 f4 = {(1, 3), (2, 4), (3, 3), (4, 6), (5, 1)}.
There are some other terms which require mention. For a function
itself, of the general nature given here, we will prefer the terms map and
mapping. The use of the word function will be restricted to the classical
sense in which the domain and range are essentially sets of numbers.
These are the traditional real-valued or complex-valued functions of one
or more real variables. (We do not make use in this book of functions
of a complex variable.) The terms functional and operator will be used
later for special types of mappings.

14

1 Prelude to Modern Analysis

We will generally reserve the usual letters f , g, etc., for the traditional
types of functions, and also later for functionals, and we will use letters
such as A and B for mappings.

Review exercises 1.3

(1) Let f = {(2, 2), (3, 1), (4, 3)}, g = {(1, 6), (2, 8), (3, 6)}. Does f 1
exist? Does g 1 exist? If so, write out the function in full. Does
f g exist? Does g f exist? If so, write out the function in full.
(2) Define a function f : R R by f (x) = 5x 2, for x R. Show
that f is one-to-one and onto. Find f 1 .
(3) For functions f : X Y and g : Y Z, show that
(a) g f : X Z is one-to-one if f and g are both one-to-one,
(b) g f : X Z is onto if f and g are both onto.

1.4 Countability
Our aim is to make a basic distinction between finite and infinite sets
and then to show how infinite sets can be distinguished into two types,
called countable and uncountable. These are very descriptive names:
countable sets are those whose elements can be listed and then counted.
This has to be made precise of course, but essentially it means that
although in an infinite set the counting process would never end, any
particular element of the set would eventually be included in the count.
The fact that there are uncountable sets will soon be illustrated by an
important example.
Two special terms are useful here. Two sets X and Y are called
equivalent if there exists a one-to-one mapping from X onto Y . Such
a mapping is a bijection, but in this context is usually called a one-toone correspondence between X and Y . Notice that these are two-way
terms, treating the two sets interchangeably. This is because a bijection
has an inverse, so that if f : X Y is a one-to-one correspondence
between X and Y , then so is f 1 : Y X, and either serves to show
that X and Y are equivalent. Any set is equivalent to itself: the identity
mapping I : X X, where I(x) = x for each x X, gives a oneto-one correspondence between X and itself. It is also not difficult to
prove, using the notion of composition of mappings, that if X and Y are
equivalent sets and Y and Z are equivalent sets, then also X and Z are
equivalent sets. See Review Exercise 1.3(3).

1.4 Countability

15

We now define a finite set as one that is empty or is equivalent to the

set {1, 2, 3, . . . , n} for some positive integer n. A set that is not finite is
called an infinite set. Furthermore:
Definition 1.4.1 Countable sets are sets that are finite or that are
equivalent to the set N of positive integers. Sets that are not countable
are called uncountable.
It follows that the set N itself is countable.
For the remainder of this section, we will be referring only to infinite
sets. It will be easy to see that some of the results apply equally to finite
sets.
According to the definition, if X is a countable set then there is a oneto-one correspondence between N and X, that is, a mapping f : N X
which is one-to-one and onto. Thus X is the set of images, under f , of
elements of N:
X = {f (1), f (2), f (3), . . . },
and no two of these images are equal. This displays the sense in which
the elements of X may be counted: each is the image of precisely one
positive integer. It is therefore permissible, when speaking of a countable
set X, to write X = {x1 , x2 , x3 , . . . }, implying that any element of X
will eventually be included in the list x1 , x2 , x3 , . . . .
In proving below that a given set is countable, we will generally be
satisfied to indicate how the set may be counted or listed, and will not
give an actual mapping which confirms the equivalence of the set with N.
For example, the set Z of all integers is countable, since we may write
Z = {0, 1, 1, 2, 2, 3, 3, . . .}
and it is clear with this arrangement how the integers may be counted.
It now follows that any other set is countable if it can be shown to be
equivalent to Z. In fact, any countable set may be used in this way to
prove that other sets are countable.
The next theorem gives two important results which will cover most
of our applications. The second uses a further extension of the notion of
a union of sets, this time to a countable number of sets: if X1 , X2 , . . . ,
are sets, then


k=1

Xk = {x : x Xk for at least one k = 1, 2, 3, . . . }.

16

1 Prelude to Modern Analysis

Theorem 1.4.2 If X1 , X2 , . . . are countable sets, then


(a) nk=1 Xk is countable for any integer n  2,

(b) k=1 Xk is countable.
Our proof of (a) will require mathematical induction. We show first
that X1 X2 is countable. Recall that X1 X2 is the set of all ordered
pairs (x1 , x2 ), where x1 X1 and x2 X2 . Since X1 and X2 are
countable, we may list their elements and write, using a double subscript
notation for convenience,
X1 = {x11 , x12 , x13 , . . . },

X2 = {x21 , x22 , x23 , . . . }.

(The first subscript is the set number of any element, the second subscript is the element number in that set.) Writing the elements of
X1 X2 down in the following array
(x11 , x21 ) (x11 , x22 ) (x11 , x23 ) (x11 , x24 ) . . .




(x12 , x21 ) (x12 , x22 ) (x12 , x23 ) (x12 , x24 ) . . .


(x13 , x21 ) (x13 , x22 ) (x13 , x23 ) (x13 , x24 ) . . .


(x14 , x21 ) (x14 , x22 ) (x14 , x23 ) (x14 , x24 ) . . .
..
.

..
.

..
.

..
.

and then counting them in the order indicated (those whose subscripts
total 5, then those whose subscripts total 6, then those whose subscripts
total 7, . . . ) proves that X1 X2 is countable.
Now assume that X1 X2 Xn1 is countable for n > 2 and let
this set be Y . Then Y Xn can be shown to be countable exactly as we
showed X1 X2 to be countable. Now, Y Xn is the set of ordered pairs
{((x1 , x2 , . . . , xn1 ), xn ) : xk Xk , k = 1, 2, . . . , n}. The mapping
f : Y Xn X1 X2 Xn given by
f (((x1 , x2 , . . . , xn1 ), xn )) = (x1 , x2 , . . . , xn1 , xn )
is clearly a one-to-one correspondence, and this establishes that X1
n
X2 Xn , or k=1 Xk , is countable. The induction is complete,
and (a) is proved.
The proof of (b) uses a similar method of counting. As before, we
write Xk = {xk1 , xk2 , xk3 , . . . }, for k N. We write down the elements

1.4 Countability
of

k=1

17

Xk in the array
x11 x12 x13 x14 . . .



x21 x22 x23 x24 . . .


x31 x32 x33 x34 . . .

x41 x42 x43 x44 . . .
..
..
..
..
.
.
.
.

and count them in the order indicated (those whose subscripts total 2,
then 3, then 4, . . . ), this time taking care that any xs belonging to more
than one Xk are counted only once. This proves (b), a result which is
often expressed by saying: the union of countably many countable sets
is itself a countable set.
It should be clear that the proof of (b) covers the cases where there are
only finitely many sets Xk , and where some of these are finite sets. In
particular, it implies that the union of two countable sets is countable.
We now prove two fundamental results.
Theorem 1.4.3
(a) The set Q of rational numbers is countable.
(b) The set R of real numbers is uncountable.
To prove (a), for each k N let Xk be the set of all rational numbers
that can be expressed as p/k where p Z. That is,


0 1 1 2 2
,
, ,
, ,... .
Xk =
k k k k k
Writing Xk in this way shows that Xk is countable for each k. Any

rational number belongs to Xk for some k, so
k=1 Xk = Q. Hence, Q
is countable, by Theorem 1.4.2(b).
We now prove (b), that R is uncountable, giving our first example
of an uncountable set. The proof relies on the statement that every
real number has a decimal expansion. (The following observations are
relevant to this. Any real number x has a decimal expansion which, for
nonnegative numbers, has the form
n2
n1
n3
+
+ 3 + ,
x = m.n1 n2 n3 . . . = m +
10 102
10
where m, n1 , n2 , n3 , . . . are integers with 0
nk
9 for each k. The
number is rational if and only if its decimal expansion either terminates

18

1 Prelude to Modern Analysis

0.125000 . . . terminates and

or becomes periodic: for example, 18 =
1887
=
0.38121212
.
.
.
is
periodic,
whereas
2 = 1.4142135 . . . is neither
4950
terminating nor periodic, being irrational. One problem with decimal
expansions is that they are not unique for all real numbers. For example,
we also have 18 = 0.124999 . . . .)
The proof that R is uncountable is a proof by contradiction. We
suppose that R is countable. Then the elements of R can be counted,
and all will be included in the count. In particular, all real numbers
between 0 and 1 will be counted. Let the set {x1 , x2 , x3 , . . . } serve to
list all these numbers between 0 and 1 and give these numbers their
decimal expansions, say
x1 = 0.n11 n12 n13 . . . ,
x2 = 0.n21 n22 n23 . . . ,
x3 = 0.n31 n32 n33 . . . ,
..
.
the double subscript notation again being convenient. Consider the number
y = 0.r1 r2 r3 . . . ,
where
rk =

2, nkk = 1,
1, nkk = 1,

for k N. This choice of values (which may be replaced by many other

choices) ensures that rk = nkk for any k. Hence, y = x1 (since these
numbers differ in their first decimal place), y = x2 (since these numbers
differ in their second decimal place), and so on. That is, y = xj for any j.
The choice of 1s and 2s in the decimal expansion of y ensures that there
is no ambiguity with 0s and 9s. But y is a number between 0 and 1
and the set {x1 , x2 , x3 , . . . } was supposed to include all such numbers.
This is the contradiction which proves that R is uncountable.
We will not prove here the very reasonable statement that a subset of
a countable set is itself a countable set, possibly finite. This result was
used already in the preceding paragraph and may now be used to prove
further that the set C of all complex numbers is uncountable: if this
were not true then the subset of C consisting of all complex numbers
with zero imaginary part would be countable, but this subset is R.

1.4 Countability

19

On the other hand, the set

X = {z : z = x + iy, x, y Q}
of all complex numbers with rational real and imaginary parts is countable. This follows using the two theorems above. For Q is countable,
so Q Q is countable, and there is a natural one-to-one correspondence
between X and Q Q, namely the mapping f : Q Q X given by
f ((x, y)) = x + iy, x, y Q.
Presumably, uncountable sets are bigger than countable sets, but is
N N bigger than N? To make this notion precise, and thus to be able
to compare the sizes of different sets, we introduce cardinality.
Definition 1.4.4 Any set X has an associated symbol called its
cardinal number, denoted by |X|. If X and Y are sets then we write
|X| = |Y | if X is equivalent to Y ; we write |X|
|Y | if X is equivalent
to a subset of Y ; and we write |X| < |Y | if |X|
|Y | but X is not
equivalent to Y . We specify that the cardinal number of a finite set is
the number of its elements (so, in particular, || = 0), and we write
|N| = 0 and |R| = c.
There is a lot in this definition. First, it defines how to use the symbols
=, < and
in connection with this object called the cardinal number
of a set. For finite sets, these turn out to be our usual uses of these
symbols. Then, for two specific infinite sets, special symbols are given
as their cardinal numbers.
Any infinite countable set is equivalent to N, by definition, so any
infinite countable set has cardinal number 0 (pronounced aleph null).
So, for example, |N N| = |Q| = 0 . It is not difficult to see that
n < 0 for any n N and that 0 < c. This is the sense in which
uncountable sets are bigger than countable sets.
The arithmetic of cardinal numbers is quite unlike ordinary arithmetic.
We will not pursue the details here but will, for interest, list some of the
main results. We define addition and multiplication of cardinal numbers
by: |X| + |Y | = |X Y | and |X| |Y | = |X Y |, where X and Y are
any sets, and we define |Y ||X| to be the cardinal number of the power
set Y X , which is the set of all functions from X into Y . Then:
1 + 0 = 0 ,

0 + 0 = 0 ,

c + c = c,

c c = c,

0
0

0 = 0 ,
= c.

The famous continuum hypothesis is that there is no cardinal num-

20

1 Prelude to Modern Analysis

ber satisfying 0 < < c. All efforts to prove this, or to disprove it by

finding a set with cardinal number strictly between those of N and R,
had been unsuccessful. In 1963, it was shown that the existence of such
a set could neither be proved nor disproved within the usual axioms of
set theory. (Those usual axioms have not been discussed here).

Review exercises 1.4

(1) Define a function f : Z N by

2n + 1, n  0,
f (n) =
2n,
n < 0.
Show that f determines a one-to-one correspondence between Z
and N.
(2) Suppose X is an uncountable set and Y is a countable set. Show
that X\Y is uncountable.
(3) Show that the set of all polynomial functions with rational coefficients is countable.

1.5 Point sets

In this section, we are concerned only with sets of real numbers. Because
real numbers can conveniently be considered as points on a line, such
sets are known as point sets and their elements as points.
The simplest point sets are intervals, for which we have special notations. Let a and b be real numbers, with a < b. The point set
{x : a
x
b} is a closed interval, denoted by [a, b], and the point
set {x : a < x < b} is an open interval, denoted by (a, b). There are also
the half-open intervals {x : a
x < b} and {x : a < x
b}, denoted by
[a, b) and (a, b], respectively. In all cases, the numbers a and b are called
endpoints of the intervals. Closed intervals contain their endpoints as
members, but open intervals do not. The following point sets are infinite intervals: {x : a < x}, denoted by (a, ); {x : a
x}, denoted
by [a, ); {x : x < b}, denoted by (, b); and {x : x
b}, denoted
by (, b]. These have only one endpoint, which may or may not be
a member of the set. The use of the signs and is purely conventional and does not imply that these things are numbers. The set R
itself is sometimes referred to as the infinite interval (, ).
A special name is given to an open interval of the form (a , a + ),
where is a positive number. This is called the -neighbourhood of a.