LUONG HA

METRIC SPACES AND

GENERAL TOPOLOGY

HUE UNIVERSITY - COLLEGE OF EDUCATION

Hue, November, 2011

This textbook is written by Luong Ha, Senior Lec-

turer of Department of Mathematics, Hue Univer-

sity, College of Education. This material is used

to teach and study the course ”Metric spaces and

Topological spaces”, code: MATH3353 .

 PREFACE

This textbook is the first course of three courses in Modern Analysis taught at the
Department of Mathematics, College of Education, Hue University. To understand this
material, students must acquire the knowledge of Mathematical Analysis of the first year.
The core content concentrates on the notions of metric spaces and topological spaces.
Metric spaces are sets equipped with the notion of ”distance between two points”. By
means of this, we can study the notion of convergence and continuity. In metric spaces, the
meaning and content of many notions in classical analysis are clarified and many theorems
can be proved very simply and clearly. There are many things presented in metric spaces
but they do not depend directly on the metric in consideration. For instance, the basic
notion of open sets will be generalized to introduce the notion of a topological space. In
addition to familiar knowledge in metric spaces, there are many new topics as bases for a
topology, the axioms of separation, connected spaces, ...
The material consists of six chapters, and has been arranged by sections. There are
many exercises which usually appear at the end of each section. Some exercises can
be solved by applying basic properties or theorems, but some require a lot of effort.
Although there is ”Hints and Solutions” at the end of each chapter, I hope that the
reader will attempt to understand the concepts and their meanings, will do his best to
solve the exercises, and use ”Hints and Solutions” efficiently and reasonably. Reader may
find several solutions in some books in References.
The textbook is written for the standard one-semester course in analysis for the second-
year students at Department of Mathematics, HUCE. Anyway, topology is geometric in
nature, the reader have to try to draw figures illustrating for notions and proofs. I hope
that, to some extent, it is helpful for some master students.
I would like to thank my colleagues at Section of Analysis, Department of Mathematics,
HUCE for their valuable comments and corrections. Special thanks are to Dr. Nguyen
Mau Nam for a number of solutions in this textbook.

Hue, November, 2011

Luong Ha

ii
 CONTENTS

PREFACE ii

1 Preliminaries 1
1 Elementary Set Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Fundamental Concepts . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Axiom of Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1 Finite Sets - Infinite Sets . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Equivalent Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Countable Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 Real Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1 Field Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Order Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Completeness Axiom . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Metric Spaces 20
1 Metrics and Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.1 The Notion of a Metric . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.2 Distance between a Point and a Set . . . . . . . . . . . . . . . . . . 22
1.3 Convergence in a Metric Space . . . . . . . . . . . . . . . . . . . . 23
1.4 Product Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2 Open Sets and Closed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.1 Neighbourhoods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 Open sets - Closed sets . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 Adherent points - Accumulation points . . . . . . . . . . . . . . . 28
2.4 Interior and Closure . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Dense Sets - Separable Spaces . . . . . . . . . . . . . . . . . . . . 32
2.6 Open Sets and Closed Sets on the Real Line . . . . . . . . . . . . . 32
2.7 Open Sets and Closed Sets in Subspaces . . . . . . . . . . . . . . . 33

iii
 3 Continuous Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.1 Continuous mappings and basic properties . . . . . . . . . . . . . . 35
3.2 Uniformly Continuous Mappings . . . . . . . . . . . . . . . . . . . 38
3.3 Homeomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Isometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.5 Equivalent Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3 Complete Metric Spaces 49

1 Cauchy Sequences and Complete Metric Spaces . . . . . . . . . . . . . . . 49
1.1 Basic Notions and Properties . . . . . . . . . . . . . . . . . . . . . 49
1.2 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2 Contraction Mapping Principle . . . . . . . . . . . . . . . . . . . . . . . . 51
2.1 Contraction Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.2 Contraction Mapping Principle . . . . . . . . . . . . . . . . . . . . 52
3 Completion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1 Definition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4 Baire Category Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.1 Nowhere Dense Sets . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2 Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4 Compact Metric Spaces 65

1 Compact Sets - Compact Spaces . . . . . . . . . . . . . . . . . . . . . . . . 65
1.1 Bounded sets - Totally Bounded Sets . . . . . . . . . . . . . . . . . 65
1.2 Compact sets - Compact spaces . . . . . . . . . . . . . . . . . . . . 66
2 Heine-Borel’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.1 Cover - Subcover . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.2 Heine-Borel’s Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . 69
3 Compact Spaces and Continuous Mappings . . . . . . . . . . . . . . . . . . 71
4 Approximation Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.1 Approximation of a Function by a Polynomial . . . . . . . . . . . . 72
4.2 Approximation by a Trigonometric Polynomial . . . . . . . . . . . . 74
5 Ascoli-Arzela theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.1 Relatively Compact Sets . . . . . . . . . . . . . . . . . . . . . . . . 78
5.2 Ascoli-Arzela theorem . . . . . . . . . . . . . . . . . . . . . . . . . 78

5 Topological Spaces 86
1 Introduction to Topological Spaces . . . . . . . . . . . . . . . . . . . . . . 86
1.1 The concept of a Topological Space . . . . . . . . . . . . . . . . . . 86
1.2 Neighbourhoods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
1.3 Closed sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

iv
 1.4 Interior and Closure . . . . . . . . . . . . . . . . . . . . . . . . . . 88
1.5 Adherent points - Accumulation points . . . . . . . . . . . . . . . 89
1.6 Dense sets - Separable topological spaces . . . . . . . . . . . . . . 90
1.7 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2 Basis for a Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.1 Basis for a topology and properties . . . . . . . . . . . . . . . . . . 93
2.2 Lindelöf spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3 Continuous mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.1 Continuity of a mapping . . . . . . . . . . . . . . . . . . . . . . . . 96
3.2 Homeomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4 Product Space - Quotient Space . . . . . . . . . . . . . . . . . . . . . . . . 101
4.1 Product space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2 Quotient space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5 Axioms of Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.1 Axioms of separation and basic properties . . . . . . . . . . . . . . 104
5.2 Extension of a continuous function . . . . . . . . . . . . . . . . . . 108
6 The Metrizable Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.1 Metrizable spaces and Axioms of countability . . . . . . . . . . . . 112
6.2 Metrization theorems . . . . . . . . . . . . . . . . . . . . . . . . . . 113

6 Compactness and Connectedness 122

1 Compact Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
1.1 Basic Notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
1.2 Compactness and Axioms of Separation . . . . . . . . . . . . . . . . 124
1.3 Locally compact spaces . . . . . . . . . . . . . . . . . . . . . . . . . 126
1.4 Compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
2 Connected Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.1 Connected Sets and Connected Spaces . . . . . . . . . . . . . . . . 129
2.2 Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
2.3 Path-Connected Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 133
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

v
Chapter 1

Preliminaries

§1 ELEMENTARY SET THEORY

1.1 FUNDAMENTAL CONCEPTS

1.1.1 Basic Notations.

Throughout this textbook, the following commonly used mathematical symbols will
be employed:
∀ means ”for all” (or for each)
∃ means ”there exists” (or there is)
⇒ means ”implies that” (or simply, implies)
⇔ means ”if and only if”
The basic notions of set theory will be discussed briefly in this section. The interested
reader can find details treatments on the foundation of the theory of sets in references.
The concept of a set plays an important role in every branch of mathematics. Roughly
speaking, a set is considered to be a collection of objects, viewed as a single entity. In
general, the words ”set”, ”collection”, ”family” are all synonymous.
Sets will be denoted by capital letters. The objects of a set A are called elements (or
the members, or the points) of A. To denote that an objext x belongs to a set A, the
notation ∈ is used, and we write x ∈ A. Similarly, the symbol x ∈ / A means that the
element x does not belong to A.
Two sets A and B are called equal, denoted by A = B, they have the same elements.
A set A is called a subset of a set B if every element of A ia also an element of B, denoted
by A ⊂ B or B ⊃ A. Obviously, A = B if and only if A ⊂ B and B ⊂ A. If A ⊂ B and
A 6= B, then we say that A is a proper subset of B. A set without any elements is called
the empty set and is denoted by ∅.
The following sets are often used in this textbook:
N = {1, 2, ..., n, ...} the set of natural numbers or the set of positive integers;
Z the sets of integers;
Q the set of rational numbers;

1
2 Chapter 1. Preliminaries

I the set of irrational numbers;

R the set of real numbers;
C the set of complex numbers.
The basic operations in set theory are the process by which new sets can be defined
in terms of given sets. Basically, a set is defined by specifying a property of its elements;
for example, we can speak of the set of real numbers or the set of prime numbers. If P (x)
is a proposition concerning x, we shall denote by {x : P (x)} the set of all x for which the
proposition P (x) is true.
Let us now return to the description of the elementary operations that can be per-
formed with sets. If A and B are two sets, the we define
i) the union A ∪ B of A and B to be the set

A ∪ B = {x : x ∈ A or x ∈ B};

ii) the intersection A ∩ B of A and B to be the set

A ∩ B = {x : x ∈ A and x ∈ B};

iii) the difference A \ B of A and B to be the set

A \ B = {x : x ∈ A and x ∈
/ B};

iii) the symmetric difference A∆B of A and B to be the set

A∆B = (A \ B) ∪ (B \ A).

Two sets A and B are called disjoint if A ∩ B = ∅. The set A \ B is also called the
complement of B relative to A.
For three sets A, B, C, we have the identities:
1) (A ∪ B) ∩ C = (A ∩ C) ∪ (B ∩ C)
2) (A ∩ B) ∪ C = (A ∪ C) ∩ (B ∪ C)
2) (A ∩ B) \ C = (A \ C) ∪ (B \ C)
3) (A ∪ B) \ C = (A \ C) ∩ (B \ C).
The identities 1) and 2) beteween unions and intersections are called the distributive
laws.
We also consider sets whose members are themselves sets. In such situation, we shall
use the phrase family of sets. Families of sets will be denoted by capital script letters:
A, B, F, .... There is a standard way for denoting a family of sets. If for each element i of
a nonempty set I, a subset Ai of a fixed set X is assigned, then {Ai : i ∈ I} denotes the
family whose members are the sets Ai . The nonempty set I is called the index set, and
its members are called indices. Conversely, if F is a family of sets, then by setting I = F
and Ai = i for each i ∈ I, we can express F in the form {Ai : i ∈ I}.
Now let A be a set. Then the set of all subsets of A is called the power set of A, and
is denoted by P(A).

. Luong Ha - HUCE
§ 1. Elementary Set Theory 3

The concepts of union and intersection of two sets can be generalized to unions and
intersections of arbitrary families of sets. If {Ai : i ∈ I} is a family of sets, then the union
of the family is defined to be the set
[
Ai = {x : ∃i ∈ I such that x ∈ Ai },
i∈I

and the intersection of the family is defined to be the set

\
Ai = {x : x ∈ Ai ∀i ∈ I}.
i∈I

∞
[
If I = N, then the union and the intersection the family will be denoted by Ai and
i=1
∞
\
Ai , respectively.
i=1
For A ⊂ X, denote by Ac the complement of A relative to X. The following properties
of the complement operation is very useful.
1.1.2 Theorem. Let {Ai : i ∈ I} be a family of subsets of a set X. The following
identities hold ³[ ´c \ ³\ ´c [
Ai = Aci and Ai = Aci .
i∈I i∈I i∈I i∈I

The distributive laws for general families of sets now are of the form
³[ ´ [ ³\ ´ \
Ai ∩ B = (Ai ∩ B) and Ai ∪ B = (Ai ∪ B).
i∈I i∈I i∈I i∈I

There is yet another way of constructing new sets from given ones; it involves the
notion of an ordered pair. Let A and B be two sets. We define the cartesian product
A × B of A and B to be the set of all ordered pairs (a, b) for which a ∈ A and b ∈ B.
Two ordered pairs (a, b) and c, d) are called equal if a = c and b = d.
Note that in general A × B 6= B × A, and if A × B 6= ∅, then

A × B ⊂ C × D ⇔ A ⊂ C, B ⊂ D.

It is easy to prove the following properties:

1) (A × B) ∩ (C × D) = (A ∩ C) × (B ∩ D).
2) If A ⊂ X and B ⊂ Y , then

(X × Y ) \ (A × B) = [(X \ A) × Y ] ∪ [(A × (Y \ B)].

1.2 MAPPINGS

1.2.1 Definition. A mapping f from a set X into a set Y is defined to be a specific rule
that assigns to each element x of X a unique element y in Y , denoted by f : X → Y .

. Luong Ha - HUCE
4 Chapter 1. Preliminaries

The element y is called the value of the mapping f at x, (or the image of x under f ),
and is denoted by f (x). The set X is called the domain of f , and the set

{y ∈ Y : ∃x ∈ X, y = f (x)}

is called the range of f . It is understood that both X and Y are nonempty.

Two mappings f, g from X into Y is called equal, denoted by f = g, if f (x) = g(x)
for all x ∈ X.
The identity mapping of a set X, denoted by IX , is defined by I(x) = x for each x ∈ X.
Let f : X → Y be a mapping, and let A ⊂ X, B ⊂ Y . Denote
f (A) = {y ∈ Y : ∃x ∈ A, y = f (x)} and f −1 (B) = {x ∈ X : f (x) ∈ B}.
The set f (A) is called the image of A under f , and f −1 (B) is called the preimage (or the
inverse image of B under f ).
For any subsets A, C of X and any subsets B, D of Y , the following identities hold:
1) A ⊂ C ⇒ f (A) ⊂ f (C), B ⊂ D ⇒ f −1 (B) ⊂ f −1 (D).
2) f (A ∪ C) = f (A) ∪ f (C), f −1 (B ∪ D) = f −1 (B) ∪ f −1 (D).
3) f (A ∩ C) ⊂ f (A) ∩ f (C), f −1 (B ∩ D) = f −1 (B) ∩ f −1 (D).
4) f (A \ C) ⊃ f (A) \ f (C), f −1 (B \ D) = f −1 (B) \ f −1 (D).
¡ ¢c
5) f −1 (B c ) = f −1 (B) .
6) A ⊂ f −1 (f (A)) and f (f −1 (B)) ⊂ B.
Identities 1) and 2) still hold for any family of subsets of X and any family of subsets
of Y .
A mapping ϕ : N → X is called a sequence in X. If we retun to the notion of a family
of sets, then a family {Ai : i ∈ I} of subsets of X is precisely a mapping from I into the
power set P(X). If I = N, then the family {Ai : i ∈ N} is a sequence of subsets of X.
1.2.2 Compositions - Restrictions - Extensions.
Let f : X → Y and g : Y → Z be mappings. We define a mapping g ◦ f : X → Z by
(g ◦ f )(x) = g(f (x)) for every x ∈ X. Then g ◦ f is called the composition of the mappings
f and g. If (xn )n is a sequence in X, and (nk )k is a sequence in N with nk < nk+1 for all
k ∈ N, then the composition of the mapping k 7→ nk and n 7→ xn is called a subsequence
of (xn )n , denoted by (xnk )k .
We have (g ◦ f )(A) = g(f (A)) for every A ⊂ X and (g ◦ f )−1 (B) = f −1 (g −1 )(B) for
every B ⊂ Z.
Now let f : X → Y and let A ⊂ X. We can define a mapping f|A from A into Y by
setting
f|A (x) = f (x) ∀x ∈ A.

The mapping f|A is called the restriction of f to A.

Let f : A ⊂ X → Y be a mapping. If g : X → Y is a mapping such that g(x) = f (x)
for all x ∈ A, then g is called an extension of f .
1.2.3 Injection - Surjection - Bijection - Inverse mapping.

. Luong Ha - HUCE
§ 1. Elementary Set Theory 5

Let f : X → Y be a mapping.
The mapping f is called injective or one-to-one if f (x) 6= f (y) whenever x 6= y.
The mapping f is called surjective or onto if f (X) = Y , that is, for each y ∈ Y , there
exists x ∈ X such that y = f (x).
The mapping f is called bijective if f is both injective and surjective. A mapping f
which is injective, or surjective, or bijective is also called an injection, a surjection and a
bijection, respectively.
If f is a bijection from X onto Y , then for each y ∈ Y , there is a unique element
x ∈ X such that y = f (x). Hence, we can define a mapping f −1 : Y → X by letting
f −1 (y) = x if f (x) = y. The mapping f −1 is called the inverse of f . Note that f −1 also
is a bijection and f ◦ f −1 = IY , f −1 ◦ f = IX .
Let f : X → Y and g : Y → Z be mappings. It is easy to show that if f and g are
injective (or surjective), then the composition g ◦ f is injective (or surjective). Therefore,
g ◦ f is bijective if both f and g are bijective, and, in addition, (g ◦ f )−1 = f −1 ◦ g −1 .
1.2.4. Theorem. Let f : X → Y be a mapping. Then
1) f is injective if and only if A = f −1 (f (A)) for each A ⊂ X.
2) f is surjective if and only if f (f −1 (B)) = B for each B ⊂ Y .

1.3 RELATIONS

1.3.1 Relations.
A subset R of X × X is called a relation on X. If (x, y) ∈ R, we say that x is in the
relation R with y, and this is denoted by xRy.
1.3.2 Equivalence relations.
A relation R on a set X is called an equivalence relation if it satisfies the following
three properties:
1) xRx for each x ∈ X (reflexivity);
2) If xRy, then yRx (symmetry);
3) If xRy and yRz, then xRz (transitivity).
Let R be an equivalence relation on X and let x ∈ X. The equivalence class of x is
defined to be the set
x̃ = {y ∈ X : yRx}.

Note that we always have x ∈ x̃.

Denote by X/R the collection of equivalence classes determined by R. This collection
is called the quotient set of the equivalence relation R. Clearly, the union of the elements
of X/R equals X. Then
1.3.3 Proposition. Two equivalence classes are either equal or disjoint.
Therefore, the collection X/R is a particular example of what is called a partition of
X.

. Luong Ha - HUCE
6 Chapter 1. Preliminaries

1.3.4 Partitions.
A partition of a set X is a collection of disjoint nonempty subsets of X whose union
equals X.
Given a partition C of X, there is only one equivalence relation R on X such that
C = X/R. Hence studying equivalence relations on a set X and studying partitions of X
are the same thing.
1.3.5 Order relations.
A relation ≤ on a set X is called a partial order if it satisfies the following three
properties:
1) x ≤ x for each x ∈ X ((reflexivity);
2) If x ≤ y and y ≤ x, then x = y (antisymmetry);
3) If x ≤ y and y ≤ z, then x ≤ z (transitivity).
Another notation for x ≤ y is y ≥ x. The notation x < y (or y > x) means that x ≤ y
and x 6= y. A set equipped with a partial order is called a partially ordered set. If for any
x, y ∈ X, either x ≤ y or else y ≤ x, then ≤ is called a total order or a linear order on X.
Now let X be a partially ordered set. A subset Y of X is called a chain or a totally
ordered set if for any x, y in X, either x ≤ y or else y ≤ x.
Let Y be a subset of X. If there is an element u ∈ X such that x ≤ u for all x ∈ Y ,
then u is said to be an upper bound for Y , and Y is called bounded above. Similarly, Y is
said to be bounded below if there is an element v ∈ X such that v ≤ x for all x ∈ Y , and
v is called a lower bound for Y .
An element u ∈ Y is called the largest element of Y if x ≤ u for all x ∈ Y . An
element v ∈ Y is called the smallest element of Y if v ≤ x for all x ∈ Y . An element
m ∈ Y is called a maximal element of Y if for any x ∈ Y , the relation m ≤ x implies
x = m. An element m ∈ Y is called a minimal element of Y if for any x ∈ Y , the
relation x ≤ m implies x = m. Note that a partially ordered set may have more than
one maximal element or more than one minimal element, but the largest element and the
smallest element, if any, are unique.
If the set of all upper bounds for Y has the smallest element, then this element is
called the least upper bound or the supremum of Y , and it is denoted by sup Y . If the set
of all lower bounds for Y has the largest element, then this element is called the greatest
lower bound or the infimum of Y , and it is denoted by inf Y . The supremum and the
infimum of a set Y may or may not belong to Y . If they does, they are the largest and
the smallest element of Y .

1.4 AXIOM OF CHOICE

1.4.1 Cartesian products.

We have defined the notion of a cartesian product of two sets. Similarly, the cartesian
of n sets A1 , A2 , ..., An , denoted by A1 ×A2 ×...×An , is the set of all n-tuples (a1 , a2 , ..., an ),

. Luong Ha - HUCE
§ 1. Elementary Set Theory 7

with ai ∈ Ai for each i = 1, 2, ..., n. As usual, (a1 , a2 , ..., an ) = (b1 , b2 , ..., an ) if and only
if ai = bi for each i = 1, 2, ..., n. If A1 = A2 = · · · = An = A, then the cartesian product
will be written as An .
Now let (Ai )i∈I be a family of sets. The cartesian product of the family (Ai )i∈I ,
Q S
denoted by Ai , is defined as the sets of all mappings x : I → Ai such that x(i) ∈ Ai
i∈I i∈I
for each i ∈ I.
Let xi = x(i) for each i ∈ I, then we write the element x in the form x = (xi )i∈I ,
and call it an I-tuple. Each xi is called the i-th coordinate of x. Clearly, if the cartesian
product is nonempty, then each Ai is nonempty. The question is
Q
When is the cartesian product of a family of sets Ai is nonempty?
Q i∈I
One hopes that if each Ai is nonempty, then Ai is nonempty. But such a statement
i∈I
cannot be proven with the usual axioms of set theory. The answer for the question is
known as ”the axiom of choice”. From now on, this axiom will be assumed without
further explanation.
1.4.2 The Axiom of Choice. If (Ai )i∈I is a nonempty family of sets such that each Ai
Q
is nonempty, then Ai is nonempty.
i∈I
This axiom has an equivalent formulation as follows:
If (Ai )i∈I is a nonempty family of pairwise disjoint sets such that each Ai is nonempty,
S
then there exists a set E ⊂ Ai such that E ∩ Ai consists of precisely one element for
i∈I
each i ∈ I.
Q
Let j ∈ I. The mapping pj : Ai → Aj defined by
i∈I
pj ((xi )i ) = xj
Q
is surjective, and it is called the j-th projection of Ai onto its j-factor Aj .
i∈I
An useful and commonly used equivalent statement of the axiom of choice is known
as Zorn’s lemma, which reads as follows:
1.4.3 Zorn’s lemma. If every chain of a partially ordered set X has an upper bound in
X, then X has a maximal element.
For further details, please refer to [9].
We have explicitly mentioned the axiom of choice because it has excited more con-
troversy than any of the other axioms of set theory since it was first formulated by E.
Zermelo in 1908. A few mathematicians go so far as to refuse to use the axiom of choice.
However, analysis cannot be developed without this axiom and we shall use it, explicitly
or implicitly, without comment.

. Luong Ha - HUCE
8 Chapter 1. Preliminaries

§2 CARDINALITY

2.1 FINITE SETS - INFINITE SETS

2.1.1 Definition.
A set A is called finite if either A is the empty set or there is n ∈ N and a bijection
from {1, 2, ..., n} onto A. In the former case, we say that A has cardinality 0; in the latter
case, we say that A has cardinality n.
A set A is called infinite if it is not finite.
2.1.2 Remark.
1) Each subset of a finite set is finite.
2) The union and the intersection of a finite family of finite sets are finite.
3) A set which contains an infinite set is infinite.

2.2 EQUIVALENT SETS

2.2.1 Definition.
Two sets A and B are called equivalent or to have the same cardinality if there exists
a bijection from A onto B. The equivalence of the two sets A and B is denoted by A ∼ B
or card A = card B.
Note that the relation ∼ is an equivalence relation on some collection U of sets.
2.2.2 Example.
1) Let A and B be finite sets with the numbers of element are m, n, respectively. Then
A and B are equivalent if and only if m = n. Hence, if A is finite, there is no bijection of
A with a proper subset of itself.
2) Let A = {2, 4, ..., 2n, ...} ⊂ N. Then the mapping f : N → A defined by f (n) = 2n
for each n ∈ N is bijective. Hence, A is equivalent to N.
3) Let a, b ∈ R with a < b. Let f : (0; 1) → (a; b) be defined by f (t) = (1 − t)a + tb.
Then f is a bijection so that (0; 1) and (a; b) are equivalent.
π π
4) The open interval (− ; ) and the set of real numbers R are equivalent because of
2 2
the bijection
π π
f : (− ; ) → R f (x) = tan x.
2 2
5) The interval (0; +∞) is equivalent to the set R because the mapping f : (0; +∞) →
R defined by f (x) = ln x is a bijection.
From examples 3), 4), and 5), we can conclude that all open intervals of R are equiva-
lent to R, and hence, they are equivalent. According to example 2, we see that an infinite
set can be equivalent to some of its proper subsets. This cannot happen for finite sets.
Hence, one can define that a set A is infinite if A is equivalent to some of its proper
subsets.

. Luong Ha - HUCE
§ 2. Cardinality 9

2.2.3 Theorem. Let (An )n and (Bn )n be two sequences of pairwise disjoint sets such
that An ∼ Bn for all n ∈ N. Then
∞
[ ∞
[
An ∼ Bn .
n=1 n=1

Proof. By hypothesis, for each n ∈ N, there is a bijection fn from An onto Bn . Let

∞
[ ∞
[ S
f : An → Bn be defined as follows: For each x ∈ ∞ n=1 An , there exists a unique
n=1 n=1
no ∈ N such that x ∈ Ano , let f (x) = fno (x). We can verify that f is a bijection. Hence,
the two sets in view are equivalent.¥
Let A and B be two sets. We say that A has greater cardinality than B (or B has
lesser cardinality than A) if A is equivalent to a subset of B and A is not equivalent to
B, and we denote it by card A < card B. Using the axiom of choice, one can prove that
for any two sets A and B, there is one and only one of the following cases happening:
1) card A < card B;
2) card B < card A;
3) card A = card B.
But what happens if A is equivalent to a subset of B, and B is equivalent to a subset
of A? Schröder - Bernstein theorem will answer this question.
2.2.4 Theorem. Let X ⊃ Y ⊃ X1 . If X ∼ X1 , then X ∼ Y .
Proof. Since X ∼ X1 , there exists a bijection f : X → X1 . We have X ⊃ Y ,
hence the restriction of f to Y , which we shall also denote by f , is injective. Thus, Y is
equivalent to a subset Y1 of X1 , where

X ⊃ Y ⊃ X1 ⊃ Y1 ,

and f : Y → Y1 is bijective. Now Y ⊃ X1 , hence for similar reasons, X ∼ X2 , where

X ⊃ Y ⊃ X1 ⊃ Y1 ⊃ X2 ,

and f : X1 → X2 is bijective. Consequently, there exist equivalent sets X1 , X2 , X3 , ... and

equivalent sets Y1 , Y2 , Y3 , ... such that

X ⊃ Y ⊃ X1 ⊃ Y1 ⊃ X2 ⊃ Y2 ⊃ · · ·

Let B = X ∩ Y ∩ X1 ∩ Y1 ∩ X2 ∩ Y2 · · · Then
X = (X \ Y ) ∪ (Y \ X1 ) ∪ (X1 \ Y1 ) ∪ (Y1 \ X2 ) ∪ · · · ∪ B,
Y = (Y \ X1 ) ∪ (X1 \ Y1 ) ∪ (Y1 \ X2 ) ∪ · · · ∪ B.
Note further that X \ Y ∼ (X1 \ Y1 ) ∼ (X2 \ Y2 ) ∼ · · · . Hence, X ∼ Y by Theorem 2.2.3.
¥
The next result shows that the power set of a given set has ”more elements” than the
set; it is due to G. Cantor.

. Luong Ha - HUCE
10 Chapter 1. Preliminaries

2.2.5 Theorem. For every set A, the set P(A) has greater cardinality than A.
Proof. If A is empty, then the result is trivial. So we assume that A is nonempty.
Define f : A → P(A) by f (x) = x for all x ∈ A. Clearly, f is injective. Now assume by
way of contradiction that A and P(A) are equivalent. Hence, there exists g : A → P(A)
which is bijective. Consider the subset B of A defined by B = {x ∈ A : x ∈ / g(x)}, and
pick a ∈ A such that g(a) = B. But then a ∈ B if and only if a ∈
/ B, which is impossible.
¥

2.3 COUNTABLE SETS

An important class of infinite sets is the class of countable sets: these are sets which
are equivalent to the set of natural numbers.
2.3.1 Definition.
A set A is called countable if A is equivalent to the set N of natural numbers. An
infinite set which is not countable is called uncountable.
More concretely, a set A is countable if and only if there is a bijection f from N onto
A. If we let xn = f (n), then A = {x1 , x2 , ..., xn , ...}. In other words, a set A is countable
if and only if we can enumerate A into an infinite sequence. A set which is either finite
or countable is called at most countable.
2.3.2 Example.
1) Let A be the set of all odd natural numbers. The mapping f : N → A defined by
f (n) = 2n − 1 is bijective, hence A is countable.
2) The set Z of integers is countable. We can show that the mapping f : N → Z
defined by

n/2 if n is even
f (n) = 1 − n (1.1)
 if n is odd.
2
is a bijection.
2.3.3 Theorem. Every infinite sets contains a countable set.
Proof. Let A be an infinite set. It is obvious that A is nonempty. Pick a1 ∈ A,
and consider the set A1 = A \ {a1 }. This set is infinite, so we can pick a2 ∈ A1 , and
consider the set A2 = A \ {a1 , a2 }. By the same arguments, there exists an element
a3 ∈ A2 . Proceeding by this way, we obtain a set {a1 , a2 , ..., an , ...}. By definition, this
set is countable, and clearly is contained in A.¥
2.3.4 Theorem. Every subset of a countable set is either finite or else countable.
Proof. Let A be a subset of countable set B. Assume that A is infinite. By definition,
there is a bijection f from N onto B. Then the restriction of f to f −1 (A) is a bijection
between the subset f −1 (A) of N onto A. Hence it suffices to show that each infinite subset
of N is countable.

. Luong Ha - HUCE
§ 2. Cardinality 11

Let X be an infinite subset of N. We define a mapping g : N → X as follows: g(1) be

the least element of X, and then if g(1), ..., g(n) have been defined, set g(n + 1) to be the
least element of X \ {g(1), g(2), ..., g(n)}. Showing the fact that g is a bijection is left for
the reader as a problem.
The following theorem presents some useful characterizations to ensure that an infinite
set is countable.
2.3.5 Theorem. For an infinite set A the following statements are equivalent:
1) A is countable.
2) There exists a subset B of N and a mapping f : B → A that is surjective.
3) There exits a mapping g : A → N that is injective.
Proof.
1)⇒ 2) By definition, there is a bijection from N to A, and we choose B = N.
2)⇒ 3) Suppose that there exist a subset B of N and a surjection f : B → A. Then the
set f −1 (a) = {n ∈ B : f (n) = a} is nonempty. We define a mapping g : A → N as follows:
g(a) = the least element of f −1 (a). If g(a) = g(b), then a = f (g(a)) = f (g(b)) = b. Hence,
g is injective.
3)⇒ 1) Suppose that g : A → N is injective. Then g(A) is equivalent to A, hence g(A)
is infinite. By Theorem 2.3.4, g(A) is countable, and so is A.¥
2.3.6 Example.
1) Let f : N → N × N defined by f (n, m) = 2n 3m .
Suppose that f (n, m) = f (p, q), that is, 2n 3m = 2p 3q . If n 6= p, namely n < p, then
3m = 2p−n 3q , contradicting the fact that 3m is odd for all m ∈ N. Therefore, n = p. As
a result, 3m = 3q , and hence m = q. This means that f is injective. By Theorem 2.3.5,
N × N is countable.
2) If A and B are two countable sets, then A × B is countable. By hypothesis,
we can express A = {a1 , a2 , ..., am , ...} and B = {b1 , b2 , ..., bn , ...}. Define a mapping
f : N × N → A × B by f (m, n) = (am , bn ). Then f is surjective. Clearly, A × B is infinite,
and N × N is countable, by Theorem 2.3.5, the set A × B is countable.
By induction, it is easy to show that if the product of a finite collection (nonempty,
of course) of countable sets is countable.
3) Denote by A be the set of positive rational numbers. Define f : N × N → A by
m
g(n, m) = . Since N × N is countable, there is a bijection f : N → N × N. Then the
n
composition g ◦ f : N → A is a surjection. By Theorem 2.3.5, A is countable. It is easy
to see that the set of nonpositive rational numbers also is countable.
[∞
2.3.7 Theorem. Let (An )n∈N be a countable family of countable sets. Then An is
n=1
countable.
∞
[
Proof. Let A = An . Since each An is countable, we can let An = {an1 , an2 , ..., ani , ...},
n=1

. Luong Ha - HUCE
12 Chapter 1. Preliminaries

and let B = {2k 3n : k, n ∈ N}. We define f : B → A by f (2k 3n ) = ank . Then f is surjec-

tive, and hence A is countable by Theorem 2.3.4.¥
n
[
Note that if A1 , A2 , ..., An are countable, then Ai is countable. If there is at least
i=1
∞
[
one set An is countable, and the other sets are finite, then An is countable.
n=1
From Example 2.3.6 (3), and Theorem 2.3.7, it is easy to see that the set Q of rational
numbers is countable.
2.3.8 Theorem. Let M be an infinite set, and let A be an at most countable set. Then
M ∪ A is equivalent to M .
This means that, when adding an at most countable set to an infinite set, we do not
change the cardinality of that infinite set.
Proof. Let N = A ∪ M . By Theorem 2.3.4, M has a countable subset B. Let
C = M \ B, then M = B ∪ C, and N = C ∪ A ∪ B. By Theorem 2.3.7, the set
A ∪ B is countable. Therefore, there exists a bijection f from B onto A ∪ B. Let
g : M = B ∪ C → N = C ∪ (A ∪ B) be defined by

x if x ∈ C
g(x) =
f (x) if x ∈ B.

Then g is bijective, and hence, M and N are equivalent.¥

Thus, for any a, b ∈ R with a < b, then the intervals (a; b), (a; b], [a; b), [a; b] are
equivalent and they are equivalent to R.
2.3.9 Theorem. The set of all finite sequences constructed from a countable set is
countable.
Proof. Let A = {a1 , a2 , ...} be a countable set and let S(A) be the set of all finite
sequences which can be formed from elements of A. For each m ∈ N, let Sm be the
set of all finite sequence of the form (ai1 , ai2 , . . . , aim ), where aik ∈ A, k = 1, 2, . . . , m.
∞
[
Then S(A) = Sm . In view of Theorem 2.3.6, we need only to show that each Sm is
m=1
countable.
S1 is countable since S1 = A. Suppose that S1 , S2 , ..., Sm are countable. Pick an
k
element ak ∈ A. Let Sm+1 be the set of sequences of the form (ai1 , ai2 , . . . , aim , ak ). The
k
mapping ϕ : Sm+1 → Sm defined by

ϕ(ai1 , ai2 , · · · , aim , ak ) = (ai1 , ai2 , · · · , aim )

∞
[
k k
is a bijection, hence Sm+1 is countable. Since Sm+1 = Sm+1 , Sm+1 is countable. Thus,
k=1
Sm is countable for each m, and the proof is complete. ¥.
2.3.10 Corollary. The family of all finite subsets of a countable set is countable.

. Luong Ha - HUCE
§ 3. Real Numbers 13

We conclude this section by an example of an uncountable set. Let E be the set of all
real sequences (xn )n with either xn = 0 or xn = 1. Suppose that E is countable. Then
E = {s1 , s2 , . . . , sn , . . . }. We form a sequence x = (xn )n as follows: If the n-th term of
sn equals 1, then xn = 0, and if the n-th term of sn equals 0, then xn = 1. Clearly, the
element x = (xn )n belongs to E, but x 6= sn for all n ∈ N. This contradiction shows that
E is uncountable.

§3 REAL NUMBERS

The most important and familiar set used in this textbook is the set of real numbers
R. We have already used real numbers in an informal way in the examples. We wish to
study them more formally.
Here we consider exactly what axioms characterize the real numbers. They consist of
the field axioms, the order axioms, and the completeness axiom. In algebraic terminology,
the set of real numbers is referred to as the only ”complete ordered field”. The real
numbers are the members of a nonempty set R equipped with two operations, + and ·,
(that is, two mappings from R × R into R), called addition and multiplication, which
satisfy the following axioms.

3.1 FIELD AXIOMS

In this section, unless otherwise stated, the letters x, y and z will denote arbitrary real
numbers.
Axiom 1. x + y = y + x and xy = yx (commutative laws).
Axiom 2. x + (y + z) = (x + y) + z and x(yz) = (xy)z (associative laws).
Axiom 3. x(y + x) = xy + xz (distributive laws).
Axiom 4. There exists an element 0 ∈ R with x + 0 = x for all x ∈ R.
Axiom 5. For each x ∈ R,there exists an element in R (denoted by −x) such that
x + (−x) = 0.
Axiom 6. There exists an element 1 ∈ R with 1 6= 0 and 1 · x = x for all x ∈ R.
Axiom 7. For each x 6= 0, there exists some element in R (denoted by x−1 ) such that
xx−1 = 1.
One can show that the zero element of Axiom 4 is uniquely determined. Also, the
element −x is uniquely determined, and −x = (−1)x holds. Similarly, the element x−1
of Axiom 7 is uniquely determined.
From the field axioms, one can derive the familiar properties of addition and multi-
plication. For example, 0 · x = 0, −(−x) = x, (−x)(−y) = xy, x − y = x + (−y) − (y −
x), (x−1 )−1 = x.
The next requirement is that R must be not merely a field but also an ”ordered field”.
This means that R is equipped with an order relation ≥ compatible with the algebraic

. Luong Ha - HUCE
14 Chapter 1. Preliminaries

operations via following axioms:

3.2 ORDER AXIOMS

Axiom 8. For any x, y ∈ R, either x ≥ y or y ≥ x holds.

Axiom 9. If x ≥ y, then x + z ≥ y + z holds for each z ∈ R.
Axiom 10. If x ≥ y and z ≥ 0, then xz ≥ yz.
We also the notation x ≤ y instead of y ≥ x. The notation x < y (or y > x) means
that x ≤ y and x 6= y. Any number x ∈ R satisfying x > 0 is called a positive number,
and any number x with x < 0 is called a negative number. From the order axioms, one
can derive the ordinary inequality properties of the real numbers. We pay attention to
one useful property dealing with inequalities: If x + ε ≥ y holds for all ε > 0, then x ≥ y
also holds.
Indeed, if the conclusion is not true, then y−x > 0, hence the number ε = 21 (y−x) > 0.
It follows from the hypothesis that 12 (x + y) = x + 12 (y − x) ≥ y. In turn this implies
y − x ≤ 0, which is a contradiction.
The absolute value |a| of a real number a is defined as follows: |a| = a if a ≥ 0 and
|a| = −a if a < 0. Clearly, |a| = | − a| for each a ∈ R. the absolute value satisfies the
following properties:
1) |a| ≥ 0 for each a ∈ R, and |a| = 0 if and only if a = 0;
2) |ab| = |a| · |b| for any a, b ∈ R;
3) |a + b| ≤ |a| + |b| for any a, b ∈ R.
The nonnegative number |a − b| can be viewed geometrically as the distance between
the numbers a and b.

3.3 COMPLETENESS AXIOM

Axiom 11. Every nonempty subset of real numbers that is bounded above has a
least upper bound.
From this axiom, it follows that every nonempty set of real numbers that is bounded
below has a greatest lower bound. Recall that a real number is called the least upper
bound or the supremum of a set A if it is an upper bound for A, and it is less than or
equal to any upper bound for A. Hence a real number a is the supremum of A if and only
if
i) x ≤ a for all x ∈ A.
ii) ∀b < a, ∃x ∈ A : x > b.
The two conditions are rewritten as follows:
i) x ≤ a for all x ∈ A.
ii) ∀ε > 0, ∃x ∈ A : x > a − ε.

. Luong Ha - HUCE
§ 3. Real Numbers 15

If the supremum of A exists and sup A ∈ A, then sup A is the largest element (the
maximum element) of A. In other words, the supremum of a set generalizes the concept
of the maximum element of a set. Similarly, we have the corresponding result for the
infimum.
It can be shown that if A ⊂ B ⊂ R, then

sup A ≤ sup B and inf A ≥ inf B.

For any two subsets A and B of R, let

−A = {−x : x ∈ A}, A + B = {x + y : x ∈ A, y ∈ B}.

Then
inf(−A) = − sup A, sup(−A) = − inf A
sup(A + B) = sup A + sup B, inf(A + B) = inf A + inf B.
3.3.1 Theorem (The Archimedean Property). Let a, b ∈ R with a > 0 and b ≥ 0.
Then there exists some natural number n such that na > b.
Thus, the Archimedean property is equivalent to saying that the set of natural numbers
is not bounded above in R.
b
Proof. Suppose by way of contradiction that n ≤ for all n ∈ N, and hence N is
a
bounded above. By the completeness axiom, k = sup N exists. Then k − 1 is not an
upper bound for N and so there is m ∈ N with k − 1 < m. But then k < m + 1 which
contradicts the definition of k. ¥
The subset Q of R has a special property which is described in the next theorem.
3.3.2 Theorem. Let a, b be two real numbers with a < b. Then then there exists a
rational number r such that a < r < b.
Proof. We need only to prove for positive numbers. Hence we suppose that 0 < a < b.
1
Let A = {n ∈ N : n > max( b−a , 1b )}. Since N is not bounded above, the set A is
nonempty. Fix an element q ∈ A. Then 0 < 1q < b − a. Set B = {n ∈ N : n < bq}. We
have B 6= ∅ since 1 ∈ B, and B is finite. Let p = max B, then p ∈ B and p + 1 ∈ / B.
p
We shall show that the rational number is the desired number. By construction, we
q
p p+1
have < b. In addition, since b ≤ , so
q q
p+1 1 p
a = b − (b − a) < − = ,
q q q
and the proof is complete. ¥
A sequence (xn )n in R is called convergent to an element a ∈ R if for each ε > 0, there
exists no ∈ N such that |xn − a| < ε for all n ≥ no . The number a is called the limit of
the sequence (xn )n , and we write lim xn = a or xn → a.
n
By using convergent sequences, we can reformulate the concept of supremum and
infimum as follows:

. Luong Ha - HUCE
16 Chapter 1. Preliminaries

A real number a is the supremum of a set A ⊂ R if and only if

i) x ≤ a for all x ∈ A.
ii) There exists a sequence (xn )n in A such that xn → a.
Similarly, a real number b is the infimum of a set A ⊂ R if and only if
i) b ≤ x for all x ∈ A.
ii) There exists a sequence (xn )n in A such that xn → b.
The following properties of the set of real numbers are formulated without proof
because they are presented in any first course in calculus.
3.3.3. Theorem. A sequence of real numbers has at most one limit.
Recall that a sequence (xn )n is said to be bounded if there is a real number M > 0
such that |xn | ≤ M for all n ∈ N.
A sequence (xn )n is said to be increasing if xn ≤ xn+1 for all n ∈ N, and decreasing if
xn+1 ≤ xn for all n ∈ N. A sequence (xn )n is said to be monotone if it is either increasing
or decreasing.
3.3.3 Theorem (Weierstrass). Every monotone bounded sequence of R is convergent.
More precisely, every increasing sequence which is bounded above is convergent, and
every decreasing sequence which is bounded below is convergent. Note that in the former
case, we have lim xn = sup xn , and lim xn = inf xn in the latter case.
n n n n
Next is a list of basic properties of convergent sequences.
1) Every convergent sequence is bounded.
2) If xn → a, then every subsequence of (xn )n is also convergent to a.
3) If there is k ∈ N such that xn = a for all n ≥ k, then xn → a.
4) If the three sequences (xn )n , (yn )n , (zn )n of R satisfy xn ≤ zn ≤ yn for all n, and
lim xn = lim yn = a, then (zn )n converges, and lim zn = a.
n n n
Assume that lim xn = a, and lim yn = b for the statements below.
n n
5) For all α, β ∈ R, the sequence (αxn + βyn )n converges, and

lim (αxn + βyn ) = αa + βb.

n

6) The sequence (xn yn )n converges to ab.

¡ xn ¢ xn a
7) If yn =
6 0 for all n, and b 6= 0, then the sequence n
converges, and lim = .
yn n yn b
8) If xn ≤ yn for all n ≥ k, then a ≤ b.
Let (∆n )n be a sequence of segments in R, with ∆n = [an ; bn ]. This sequence is said
to be nested if ∆n ⊃ ∆n+1 for each n ∈ N and the limit lim (bn − an ) = 0.
n
3.3.4 Theorem (Cantor). Each nested sequence of segments has only one point in
common.
From the definition of convergent sequences, it follows that any convergent sequence
is bounded. The converse, in general, is false. However, we have the following property:
3.3.5 Theorem (Bolzano - Weierstrass). Each bounded sequence in R has a conver-
gent subsequence.

. Luong Ha - HUCE
§ 3. Real Numbers 17

A sequence (xn )n in R is called a Cauchy sequence if for each ε > 0, there exists a
natural number no such that |xn − xm | < ε whenever n ≥ no and m ≥ no .
3.3.6 Theorem (Cauchy). Each Cauchy sequence in R is convergent.
We have observe that the subsets N, Z, Q of R are countable. So is there any subset
of R which is uncountable? The next theorem will answer this question.
3.3.7 Theorem. The set R of real numbers is uncountable.
Proof. Since R and the segment ∆ = [0; 1] are equivalent, we need only to show that
∆ is uncountable.
Suppose that ∆ is countable, then we can write ∆ = {x1 , x2 , ..., xn , ...}. Trisect ∆,
then at least one of the three segments does not contain the element x1 , and denote this
segment by ∆1 . Next trisect ∆1 , then there is an segment not containing x2 , denoted by
∆2 . Proceeding by this way, we obtain a sequence of segments (∆n )n having the following
properties:
i) ∆n ⊃ ∆n+1 for each n ∈ N;
ii) The length of ∆n is equal to 1/3n ;
iii) xn ∈
/ ∆n for each n ∈ N.
Thus, (∆n )n is a nested sequence of segments, by Theorem 3.3.4, there exists a number
T
a∈ ∞ n=1 ∆n . Obviously, a ∈ ∆, hence a = xno for some no ∈ N. This contradicts iii),
and the proof is complete.¥
Theorem 3.3.7 shows that the set of real numbers has greater cardinality than the set
of natural numbers. The cardinality of R is called R the continuum cardinality, denoted
by c, and the cardinality of N is called the countable cardinality, denoted by a. Thus,
a < c.
Now it is easy to show that the set of irrational numbers is dense in R, that is,
between two real numbers, there exists an irrational number. Furthermore, the set of
irrational numbers is infinite and uncountable. More precisely, this set has the continuum
cardinality.
The countable cardinality is the ”smallest cardinality” of infinite sets. A question
arises: Is the continuum cardinality the smallest cardinality of uncountable sets? In other
words, is there any cardinality between the countable cardinality and the continuum
cardinality? A famous conjecture of set theory, called the continuum hypothesis, asserts
that there exists no set having greater cardinality than N and lesser cardinality than R.
In 1959, Gödel showed that the continuum hypothesis does not contradicts the accepted
axioms of set theory. In 1963, Cohen closed the problem by showing that the continuum
hypothesis is unprovable and undeniable by using the accepted axioms of set theory.
Hence, the continuum hypothesis becomes the continuum axiom.

. Luong Ha - HUCE
18 Chapter 1. Preliminaries

PROBLEMS
1.1 Let (An )n∈N be a sequence of sets. Let
¡ n−1
[ ¢
B1 = A1 , B2 = A2 \ A1 , . . . , Bn = An \ Ai , . . . .
i=1

Show that the sets Bn are pairwise disjoint and

∞
[ ∞
[
Bn = An .
n=1 n=1

1.2 Show that any family of disjoint open intervals of R is at most countable.
1.3 Show that the family of open intervals with two rational endpoints is countable.
1.4 Let f : A → B be a surjection. Show that B is equivalent to a subset of A.
1.5 Let A be a subset of R such that |x − y| > 1 for any distinct elements x, y of A. Show
that A is at most countable.
1.6 Show that the set of polynomials with rational coefficients is countable.
1.7 A real number r is called an algebraic number if r is a root of a polynomial

p(x) = ao + a1 x + · · · + am xm

with integral coefficients. Show that the set of algebraic numbers is countable.
1.8 Let X be any set and let C(X) be the collection of functions f : X → [0; 1]. Show
that the power set of X is equivalent to C(X).

HINTS and SOLUTIONS

¡m−1
S ¢
1.1 Consider m, n ∈ N with n < m. We have Bm = Am \ Ai . Since n belongs to the
i=1
set{1, 2, . . . , m − 1}, if x ∈ Bm , then x ∈
/ An . Hence, x ∈
/ Bn because Bn ⊂ An . Therefore
Bn ∩ Bm = ∅.
S∞ S
∞ S
∞
Since Bn ⊂ An for each n ∈ N, so Bn ⊂ An . Let x ∈ An . Then there is
n=1 n=1 n=1
n ∈ N with x ∈ An . Let no be the smallest natural numbers such that x ∈ Ano . Then
S
∞ S
∞
x∈/ Ai for all i = 1, 2, . . . , no − 1. Hence, x ∈ Bno . Consequently, An ⊂ Bn , and
n=1 n=1
we have the desired result.
1.2. Let B be a family of disjoint open intervals of R. For each I ∈ B, pick a rational
number ri ∈ I. The mapping ϕ : B → Q defined by ϕ(I) = ri is injective. Thus, B is
equivalent to a subset of Q. Hence, B is at most counatble.
1.3 Let B be the family of open intervals with rational endpoints. Each I ∈ B is of the
form I = (rI ; sI ) with rI , sI ∈ Q. Consider ϕ : B → Q2 defined by ϕ(I) = (rI , sI ). Then
ϕ is injective. Since Q2 is countable, B is at most countable. But B is infinite, hence, it
is countable.

. Luong Ha - HUCE
§ 3. Real Numbers 19

1.4 Hints.
For each z ∈ B, there exists xz ∈ A such that f (xz ) = z. Define the mapping
g : B → A by g(z) = xz with f (xz ) = z. Show that g is an injection.
1.5 Hints.
S
+∞
Note that R = [n, n + 1). By hypothesis, it follows that A ∩ [n, n + 1) has at
n=−∞
S
+∞
most one point. Since A = (A ∩ [n, n + 1)), it is at most countable.
n=−∞

. Luong Ha - HUCE
Chapter 2

Metric Spaces

§1 METRICS AND CONVERGENCE

In the elementary geometry, we get acquainted with the notion of distance between
two points in the plane or in the space. If we denote the distance between two points A
and B by AB, then AB ≥ 0, this distance is equal to zero if and only if A = B, and
for any three points A, B, C, the inequality AC ≤ AB + BC holds. These properties are
generalized to form the notion of metric on an arbitrary set.

1.1 THE NOTION OF A METRIC

2.1.1 Definition.
Let X be a non-empty set . A function d : X × X → R is called a metric on X if it
satisfies the following properties:
i) d(x, y) ≥ 0 for all x, y ∈ X,
d(x, y) = 0 ⇔ x = y.
ii) d(x, y) = d(y, x) for all x, y ∈ X.
iii) d(x, z) ≤ d(x, y) + d(y, z) for all x, y, z ∈ X (triangle inequality).
Then the pair (X, d) is called a metric space and every element of X is called a point.
We note that we can define many different metrics on a set X. When no confusion arises,
we omit the metric d from the notation and write X instead of (X, d).
Let (X, d) be a metric space. Then for x, y ∈ X, the number d(x, y) is called the
distance between x and y.
Remark.
1) For n points x1 , x2 , ..., xn of X, we have
d(x1 , xn ) ≤ d(x1 , x2 ) + d(x2 , x3 ) + · · · + d(xn−1 , xn ).
2) For 4 points x, y, u, v of X, we have :
|d(x, y) − d(u, v)| ≤ d(x, u) + d(y, v).

20
§ 1. Metrics and Convergence 21

The first inequality is easily proved by mathematical induction. For the second, we
write
d(x, y) ≤ d(x, u) + d(u, v) + d(v, y)

Therefore
d(x, y) − d(u, v) ≤ d(x, u) + d(y, v).

Similarly, we get

d(u, v) − d(x, y) ≤ d(x, u) + d(y, v).

Thus, the second inequality is proved.

2.1.2 Example.
1) The formula d(x, y) = |x − y| defines a metric on the set of real numbers R.
2) For two points x = (x1 , x2 , . . . , xk ) and y = (y1 , y2 , . . . , yk ) of Rk , we define d(x, y) =
hX
k i 21
(xi − yi )2 . Then d is a metric on Rk .
i=1
Proving the properties i) and ii) is trivial. The triangle inequality can be deduced
from Bunhiakowski’s inequality.
The metrics in examples 1) and 2) are called the ordinary (usual, Euclidean) metrics
on R and Rk . From now on, if otherwise stated, the metrics on Rk are always the ordinary
metrics.
3) Let X be the set of all real functions continuous on the segment [a; b]. For x, y ∈ X,
we define
d(x, y) = max |x(t) − y(t)|.
t∈[a;b]

It is easily seen that the greatest value of the function |x − y| does exist since this
function is continuous on the segment [a; b]. Moreover, |x(t) − y(t)| ≤ d(x, y) for all
t ∈ [a; b].
It is obvious that d(x, y) ≥ 0 for all x, y ∈ X. The equality d(x, y) = 0 means that
|x(t) − y(t)| = 0 for all t ∈ [a; b], that is, x = y.
In addition, max |x(t) − y(t)| = max |y(t) − x(t)|, so we get d(x, y) = d(y, x) for all
t∈[a;b] t∈[a;b]
x, y ∈ X.
Now we will prove the triangle inequality. For three elements x, y, z of X, we have
|x(t) − y(t)| ≤ |x(t) − y(t)| + |y(t) − z(t)| for all t ∈ [a; b].
Therefore, for all t ∈ [a; b], |x(t) − z(t)| ≤ d(x, y) + d(y, z).
Thus, d(x, z) = max |x(t) − y(t)| ≤ d(x, y) + d(y, z).
t∈[a;b]
The metric space (X, d) is denoted by C[a;b] .
4) Consider the set X of all real functions continuous on the segment [a; b]. . For

. Luong Ha - HUCE
22 Chapter 2. Metric Spaces

x, y ∈ X, we define
Zb
d(x, y) = |x(t) − y(t)|dt.
a

Clearly d(x, y) ≥ 0 for any x, y ∈ X.

If x = y, then x − y = 0 and d(x, y) = 0.
Rb
On the contrary, if d(x, y) = 0, then |x(t) − y(t)|dt = 0.
a
Since the function |x − y| is non-negative and continuous on [a; b], it follows that
|x − y| = 0, that is, x = y.
The verification of the properties ii) and iii) is left for the reader.
L
In this case, the corresponding metric space is denoted by C[a,b] .
5) On a non-empty
( set X, the formula
1 if x 6= y
d(x, y) =
0 if x = y,
defines a metric d on X.
Properties i) are ii) are clear. We prove the triangle inequality

d(x, z) ≤ d(x, y) + d(y, z) (1).

If x = z, then d(x, z) = 0 ≤ d(x, y) + d(y, z).

Suppose x 6= z. Then we must have either x 6= y or y 6= z, for instance, x 6= y. Hence
d(x, z) = 1 and d(x, y) + d(y, z) ≥ 1 = d(x, z). Thus (1) holds.
The metric d is called the discrete metric on X and the space (X, d) is called a discrete
metric space.

1.2 DISTANCE BETWEEN A POINT AND A SET

2.2.1 Definition.
Let (X, d) be a metric space. Suppose A is a non-empty subset of X and x ∈ X. The
distance from the point x to the set A, denoted by d(x, A), is defined as

d(x, A) = inf {d(x, y) : y ∈ A}.

By definition, if x ∈ A, then d(x, A) = 0. But the converse, in general, is not true. For
example, choose A = (0; 1) ⊂ R and x = 0, we get d(0, A) = 0 but 0 ∈ / A.
2.2.2 Proposition. For any x, x0 ∈ X, the following inequality holds
¯ ¯
¯ ¯
¯d(x, A) − d(x , A)¯ ≤ d(x, x0 ).
0

Proof. Consider y ∈ A. Then

d(x, y) ≤ d(x, x0 ) + d(x0 , y).

. Luong Ha - HUCE
§ 1. Metrics and Convergence 23

Therefore h i
inf d(x, y) ≤ inf d(x, x0 ) + d(x0 , y) = d(x, x0 ) + inf d(x0 , y).
y∈A y∈A y∈A

Hence d(x, A) − d(x0 , A) ≤ d(x, x0 ).

Similarly, we obtain d(x0 , A) − d(x, A) ≤ d(x, x0 ) and the equality is proved.

1.3 CONVERGENCE IN A METRIC SPACE

2.3.1 Definition.
1) A map ϕ : N −→ X is called a sequence in X. By setting xn = ϕ(n), n ∈ N, then
the sequence ϕ is designated as (xn )n∈N or (xn )n .
If h : N −→ N is a strictly increasing function, then the composite ϕ ◦ h : N → X
is called a subsequence of the sequence ϕ. Let nk = h(k). We have (ϕ ◦ h)(k) = xnk
and the subsequence is denoted by (xnk )k . Note that nk ≥ k for every k ∈ N and
n1 < n2 < · · · < nk < · · · .
2) Now let (xn )n be a sequence in a metric space X and xo ∈ X. The sequence (xn )n
is said to be convergent to the point xo if lim d(xn , xo ) = 0, denoted by lim xn = xo or
n→∞ n→∞
xn → xo . Thus

lim xn = xo ⇔ (∀ε > 0)(∃no ∈ N)(∀n ∈ N)(n ≥ no ⇒ d(xn , xo ) < ε)

n

The point xo is then called the limit of the sequence (xn )n .

2.3.2 Remark.
1) The limit of a sequence (if any) is unique.
2) If (xn )n converges to xo , then every subsequence of (xn )n is also convergent to xo .
3) If xn → xo and yn → yo , then d(xn , yn ) → d(xo , yo ).
Proof.
1) Suppose that xn → a and xn → b. Then for any n ∈ N,

d(a, b) ≤ d(a, xn ) + d(xn , b).

By definition, d(xn , a) → 0 and d(xn , b) → 0, we get d(a, b) = 0. This means that

a = b.
2) Let (xnk )k be any subseqence of (xn )n . For ε > 0, there exists ko ∈ N such that
d(xn , xo ) < ε for all n ≥ ko . Then for any k ≥ ko , we have nk ≥ ko , and hence,
d(xnk , xo ) < ε, that is, xnk → xo .
3) The result follows from the inequality
¯ ¯
¯ ¯
¯d(xn , yn ) − d(xo , yo )¯ ≤ d(xn , xo ) + d(yn , yo ).

2.3.3. Example.

. Luong Ha - HUCE
24 Chapter 2. Metric Spaces

1) On the set R, the sequence xn → xo means that for every ε > 0, there exists no ∈ N
such that |xn − xo | < ε for all n ≥ no .
k
2) Let (xn )n ⊂ R³ and xo ∈ Rk . ´ ³ ´
(n) (n) (n) o o o
We denote xn = x1 , x2 , . . . , xk and xo = x1 , x2 , . . . , xk . Then

hX
k i 12
(n)
lim xn = xo ⇔ lim (xi − xoi )2 .
n n
i=1
(n) (n)
This is equivalent to lim (xi −xoi )2 = 0 for all i = 1, 2, . . . , k. In other words, lim xi = xoi
n n
for all i = 1, 2, . . . , k.
We say that the convergence in Rk is the coordinate convergence.
3) Let (xn )n ⊂ C[a;b] , xn → x ∈ C[a;b] . By definition, we have
³ ´
lim xn = x ⇔ lim max |xn (t) − x(t)| = 0.
n n t∈[a;b]

This means that for all ε > 0, there exists no ∈ N such that for all n ∈ N, if n ≥ no , then
|xn (t) − x(t)| < ε for all t ∈ [a; b]. That is, the sequence of functions (xn )n is uniformly
convergent to the function x on [a; b].
4) Let X be a discrete metric space and xn → xo in X. The for ε = 1, there exists
no ∈ N such that d(xn , xo ) < 1 for all n ≥ no . Hence d(xn , xo ) = 0 for all n ≥ no or
xn = xo for all n ≥ no .
A sequence (xn )n having this property is called a stationary sequence. Note that in
any metric space, each stationary sequence is always convergent. Thus a sequence in a
discrete space is convergent if and only if it is a stationary sequence.

1.4 PRODUCT SPACE

2.4.1 Definition.
Let (X, d1 ) and (Y, d2 ) be two metric spaces. For z = (x, y), z 0 = (x0 , y 0 ) in the
cartesian product X × Y , we set d(z, z 0 ) = d1 (x, x0 ) + d2 (y, y 0 ).
It is easy to verify that d is a metric on X × Y . This metric space (X × Y, d) is said
to be the product of two metric spaces (X, d1 ) and (Y, d2 ).
2.4.2 Remark. The convergence in the product space X × Y is the coordinate
convergence.
Indeed, let (zn )n ⊂ X × Y, zo ∈ X × Y . With zn = (xn , yn ) and zo = (xo , yo ), we have

lim zn = zo ⇔ lim d(zn , zo ) = 0.

n n
h i
This is equivalent to lim d1 (xn , xo ) + d2 (yn , yo ) = 0. This means that lim d1 (xn , xo ) = 0
n x
and lim d2 (yn , yo ) = 0, that is, xn → xo and yn → yo .
n
Example. The set R2 equipped with the metric d(x, y) = |x1 − y1 | + |x2 − y2 |, where
x = (x1 , x2 ), and y = (y1 , y2 ), is the product of R with R with the usual metric.

. Luong Ha - HUCE
§ 1. Metrics and Convergence 25

1.5 SUBSPACES

2.5.1 Definition.
Let (X, d) be a metric space and Y be a non-empty subset of X. For x, y ∈ Y , we
define
d0 (x, y) = d(x, y).

Then d0 is a metric on Y and the space (Y, d0 ) is called a metric subspace of (X, d) or
a subspace of X.
It follows from the definition that if Y is a subspace of X, and Z is a subspace of Y ,
then Z is a subspace of X.
We conclude this section by a simple but useful property of the convergence in the
subspace.
2.5.2 Proposition. Let Y be a subspace of X and (yn )n is a sequence in Y . Then
a) If (yn )n converges to y in Y , then (yn )n converges to y in X.
b) If (yn )n converges in X to an element y ∈ Y , then (yn )n converges to y in Y.
These properties are straightforward by the definition of the convergence and the fact
that the metric on Y is merely the restriction of the metric on X to the set Y × Y .

PROBLEMS
2.1. Prove that the given formulas given below define a metric on the corresponding set
1) `∞ = {x = (xn )n ⊂ R : (xn )n is bounded } and d(x, y) = sup |xn − yn |.
n
∞
X hX
∞ i 12
2
2) `2 = {x = (xn )n ⊂ R : xn < +∞} and d(x, y) = (xn − yn )2 .
n=1 n=1
3) X = {x : [a; b] → R : x is continuous} and
h Rb i 12
d(x, y) = (x(t) − y(t))2 dt .
a
1
4) X = C[a;b] is the set of continuously differentiable functions on [a; b] and
¯ ¯ ¯ ¯
d(x, y) = max ¯ x0 (t) − y 0 (t)¯ + max ¯ x(t) − y(t)¯.
t∈[a,b] t∈[a,b]

2.2 Let X be a non-empty set. Denote by B(X) the set of bounded functions on X. For
f, g ∈ B(X), define
¯ ¯
d(f, g) = sup ¯f (x) − g(x)¯.
x∈X
Prove that d is a metric on B(X) and investigate the convergence in the space
(B(X), d).
2.3 Investigate the convergence of the following sequences (xn )n in the space C[0,1] .
a) xn (t) = tn − t2n .
tn+1 tn+2
b) xn (t) = − .
n+1 n+2

. Luong Ha - HUCE
26 Chapter 2. Metric Spaces

c) xn (t) = tn − tn+1 .
2.4 Let (X, d) be a metric space. For x, y ∈ X, we define
d(x, y)
d1 (x, y) =
1 + d(x, y)
d2 (x, y) = ln[1 + d(x, y)]
d3 (x, y) = min(1, d(x, y)).
Prove that d1 , d2 , d3 are metrics on X. In addition, a sequence xn → x in (X, d) if and
only if xn → x in each space (X, di ), where i = 1, 2, 3.

§2 OPEN SETS AND CLOSED SETS

2.1 NEIGHBOURHOODS

Let (X, d) be a metric space, xo ∈ X and r > 0.

The set B(xo , r) = {x ∈ X : d(x, xo ) < r} is called the open ball with center xo and
radius r. The set B 0 (xo , r) = {x ∈ X : d(x, xo ) ≤ r} is called the closed ball with center
xo and radius r. The set B(xo , r) \ {xo } is denoted by B ∗ (xo , r).
By definition, it is easy to see that B ∗ (xo , r) ⊂ B(xo , r) ⊂ B 0 (xo , r).
2.1.1 Definition.
A set A ⊂ X is called a neighbourhood of a point x if there exists r > 0 such that
B(xo , r) ⊂ A.
Intuitively, if the set A is a neighbourhood of x, then A contains all the points of X
which are ”sufficiently close” to x.
2.1.2 Remark.
a) Every open ball B(x, r) is a neighbourhood of x.
b) If A is a neighbourhood of x, and B ⊃ A, then B is also a neighbourhood of x.
T
n
c) If A1 , A2 , . . . , An are neighbourhoods of x, then Ai is also a neighbourhood of x.
i=1
The statements a) and b) are clear by definition. If Ai (i = 1, 2, . . . , n) are neighbour-
hoods of x, then there are positive numbers ri (i = 1, 2, . . . , n) such that B(xi , ri ) ⊂ Ai .
Set r = min ri , we get B(x, r) ⊂ B(x, ri ) for each i = 1, 2, . . . , n. Hence, B(x, r) ⊂
1≤i≤n
Tn Tn T
n
B(x, ri ) ⊂ Ai . Thus, Ai is a neighbourhood of x.
i=1 i=1 i=1
2.1.3 Interior points and Boundary points.
A point x is called an interior point of a set A ⊂ X if A is a neighbourhood of x.
A point x is called a boundary point of A if every neighbourhood of x contains points
from A and Ac (the complement of A).
The set of all boundary points of a set A is denoted by ∂A, and is called the boundary
of A. By the symmetry of the definition, the equality ∂A = ∂Ac holds for every subset A

. Luong Ha - HUCE
§ 2. Open Sets and Closed Sets 27

of X.
Note that for a set A ⊂ X and a point x ∈ X, one and only one of the following cases
happens:
1) There is a neighbourhood V of x such that V ⊂ A (that is, V ∩ (X \ A) = ∅).
Then x is an interior point of A.
2) There is a neighbourhood V of x such that V ⊂ X \ A (that is, V ∩ A = ∅).
Then x is an interior point of Ac .
3) Every neighbourhood of x contains points from A and Ac .
Then x is a boundary point of A.
On the set R, the intervals [a; b], (a; b), [a; b), (a; b] only have two boundary points.
They are a and b.

2.2 OPEN SETS - CLOSED SETS

2.2.1 Definition.
a) A set A in a metric space X is called open if A is a neighbourhood of every point
of A. This means that each x ∈ A is an interior point of A.
b) A set A is called closed if the complement X \ A = Ac is open.
If the set A is open, then each x ∈ A cannot be a boundary point of A, that is,
A ∩ ∂A = ∅.
The set A is closed if and only if Ac is open, that is, Ac ∩ ∂Ac = ∅. But ∂Ac = ∂A,
hence Ac ∩ ∂A = ∅. This means that ∂A ⊂ A.
The following statement expresses the notions of open sets and closed set via boundary
points.
2.2.2 Proposition.
1) A set A is open if and only if A ∩ ∂A = ∅.
2) A set A is closed if and only if ∂A ⊂ A.
2.2.3 Example.
1) In any metric space X, the sets X and ∅ are both open and closed. Indeed, since
∂X = ∅, so X ∩ ∂X = ∅ and ∂X ⊂ X. Hence the set X is both open and closed, and so
is the empty set.
2) Each open ball is an open set, and each closed ball is a closed set.
Consider B(xo , r) and x ∈ B(xo , r). Then ε = r − d(xo , x) > 0. We will show that
B(x, ε) ⊂ B(xo , r). Indeed, for every y ∈ B(x, ε), we get d(y, x) < ε. Therefore

d(y, xo ) ≤ d(y, x) + d(x1 , xo ) < ε + d(x, d(y, xo ) < r,

that is, y ∈ B(xo , r).

By definition, the open ball B(xo , r) is open . Similarly, we can prove that the set

. Luong Ha - HUCE
28 Chapter 2. Metric Spaces

X \ B 0 (xo , r) is open, so the closed ball B 0 (xo , r) is closed.

3) Every finite set of a metric space is closed.
Let A = {a1 , a2 , . . . , an } ⊂ X and x ∈
/ A. Set r = min d(x, ai ). Then r > 0 and the
1≤i≤n
open ball B(x, r) does not have any elements in common with A. Hence B(x, r) ⊂ X \ A.
It follows that X \ A is open, hence, A is closed.
4) On the metric space R, the interval [a, b] is closed and not open, the interval (a; b)
is open and not closed, and two intervals [a; b) and (a; b] are neither open nor closed.
2.2.4 Theorem. For any metric space, the following statements hold:
1) Arbitrary unions of open sets are open sets.
2) Finite intersections of open sets are open sets.
3) Arbitrary intersections of closed ets are closed sets.
4) Finite unions of closed sets are closed sets.
Proof.
1) Let (Gα )α∈I is an arbitrary family of open sets of X.
S
Set G = Gα and take x ∈ G. Then there exists αo ∈ I such that x ∈ Gαo . Since
α∈I
Gαo is open, Gαo is a neighbourhood of x. Hence G is open.
3) Let (Fα )α∈I be a family of closed sets of X. Then
\ [
X \ ( Fα ) = (X \ Fα ).
α∈I α∈I
S T
Since each set X \ Fα is open, the set (X \ Fα ) is open by 1) . Thus Fα is closed.
α∈I α∈I
The statements 2) and 4) are proved similarly and left to the reader.
Note that the statements 2) and 4), in general, do not hold in the case of infinite
families.

2.3 ADHERENT POINTS - ACCUMULATION POINTS

2.3.1 Definition.
A point x is called an adherent point (a closure point) of A if every neighbourhood of
x has a non-empty intersection with A or it contains at least one point of A. This means
that B(x, r) ∩ A 6= ∅ for all r > 0.
In other words, x is an adherent point of A if and only if d(x, A) = 0.
A point x is called an accumulation point of A if every neighborhood of x has a non-
empty intersection with A \ {x} or it contains at least one point of A different from x.
This means that B(x, r) ∩ (A \ {x}) 6= ∅ for all r > 0.
Hence, x is an adherent point of A if and only if for all r > 0, there exists an element
a ∈ A such that d(x, a) < r, and x is an accumulation point of A if and only if for all
r > 0, there exists an element a ∈ A such that 0 < d(x, a) < r.

. Luong Ha - HUCE
§ 2. Open Sets and Closed Sets 29

2.3.2 Remark.
1) Every x ∈ A is an adherent point of A.
2) If x is an accumulation point of A, then x is an adherent point of A.
3) If x is an adherent point of A and x ∈
/ A, then x is an accumulation point of A.
Example.
1 1
1) Let A = {1, , . . . , , . . . } ⊂ R. Then number 0 is an accumulation point of A and
2 n
the only accumulation point of A.
2) Every element x ∈ [a; b] is an adherent point (an accumulation point) of the interval
(a; b).
2.3.3 Theorem. A point x ∈ X is an accumulation point of A if and only if each
neighbourhood of x contains infinitely many points of A.
Proof. Suppose that x is an accumulation point of A and V is an arbitray neighbour-
hood of x. Then there is r1 > 0 such that B(x1 , r1 ) ⊂ V . By definition, there exists an
1
element x1 ∈ B ∗ (x1 , r1 ) ∩ A. Choose 0 < r2 < min{r1 , }, there exists x2 ∈ B ∗ (x, r2 ) ∩ A.
2
It is evident that x1 6= x2 . Proceeding this process by induction, we obtain a sequence
of positive numbers (rn )n and a sequence (xn )n ⊂ X such that xn ∈ A ∩ B ∗ (x, rn ) for
1
each n ∈ N, where 0 < rn < min{d(x, xn−1 ), }. By this contruction, xn 6= xm whenever
n
n 6= m. Hence, the set V contains infinitely many elements xn of A.
Obviously, if each neighbourhood of x contains infinitely many points of A, then x is
an accumulation point of A .
From Definition 2.3.1 and Theorem 2.3.3, we get the following remark.
2.3.4 Remark.
1) A point x ∈ X is an accumulation point of A if and only if there exists a sequence
(xn )n ⊂ A, where xn 6= xm if m 6= n, such that xn → x.
2) A point x ∈ X is an adherent point of A if and only if there exists a sequence
(xn )n ⊂ A such that xn → x.
2.3.5 Theorem. A set A is closed if and only if A contains all the accumulation points
(or all adherent points) of A.
Proof. Suppose that A is a closed set. Let x ∈ X be anu accumulation point of A.
If x ∈
/ A, then x ∈ X \ A. The the set X \ A is an open set containing x and it does
not contain any point of A. This is a contradiction because x is an accumulation of A.
Hence, x ∈ A.
Conversely, suppose that A contains all of its accumulation points. Let x ∈
/ A. Then
x is not an accumulation point of A, there must exist i a neighbourhood V of x such that
V ∩ A = ∅, that is, V ⊂ X \ A. Thus, X \ A is open, or A is closed.
From Remark 3.3.4 and Theorem 3.3.5, we obtain a very useful criterion to prove

. Luong Ha - HUCE
30 Chapter 2. Metric Spaces

closedness of a set.
2.3.6 Corollary. A set A is closed if and only if any sequence (xn )n ⊂ A which xn → x,
then x ∈ A.
2.3.7 Example.
1) Let F = {(x, y) ∈ R2 : xy = 1}. Consider (zn )n ⊂ A and zn → zo ∈ R2 . Denote
zn = (xn , yn ) and zo = (xo , yo ). It follows that xn → xo and yn → yo . Therefore,
xn y n → xo y o .
Since (zn )n ⊂ F , the equality xn yn = 1 holds for all n ∈ N. Hence, xo yo = 1, that is,
zo ∈ F . Thus, F is closed.
2) Let a, b be two real number such that a < b.
Set M = {x ∈ C[0;1] : a ≤ x(t) ≤ b, ∀t ∈ [0; 1]}. Consider (xn )n ⊂ M such that
xn → x ∈ C[0;1] . Then xn (t) → x(t) for all t ∈ [0; 1]. Since a ≤ xn (t) ≤ b for all t ∈ [0; 1],
it follows that a ≤ x(t) ≤ b for all t ∈ [0; 1]. Hence, x ∈ M , and by the above corollary,
the set M is closed.

2.4 INTERIOR AND CLOSURE

2.4.1 Definition.
o
Let A be a subset of a metric space X. The interior of A, denoted by A, is the largest
open set contained in A.
The closure of A, denoted by A, is the smallest closed set containing A.
The interior and the closure of A are also denoted by int(A) and cl(A), respectively.
2.4.2 Remark.
o
1) For any A ⊂ X, the following inclusions holds: A ⊂ A ⊂ A.
o
2) The interior A is the union of all open sets contained in A and the closure A is the
intersection of all closed sets containing A.
o
3) By definition, it follows that a set A is open if and only if A = A, and a set A is
closed if and only if A = A.
The following theorem shows clearly the structure of the interior and the closure.
2.4.3. Theorem. The interior of a set A is the set of all interior points of A. The
closure of a set A is the set of all adherent points of A.
Proof.
1) If a point x is an interior point of A, then there is an open ball B(x, r) ⊂ A. By
o o
the definition of the interior, B(x, r) ⊂ A, hence x ∈ A.
o o o
Conversely, if x ∈ A, then since A is open, and A ⊂ A, is an interior point of A.
2) Suppose that x is not an adherent point of A. Then there is an open ball B(x, r)
such that B(x, r) ∩ A = ∅. Hence A ⊂ X \ B(x, r). But X \ B(x, r) is closed, it follows

. Luong Ha - HUCE
§ 2. Open Sets and Closed Sets 31

that A ⊂ X \ B(x, r). Thus x ∈

/ A.
On the contrary, if x ∈
/ A, then x ∈ X \ A. The set X \ A is open and (X \ A) ∩ A = ∅,
so x cannot be an adherent point of A.
2.4.4 Corollary.
The closure of A is the union of A and its boundary ∂A. This closure is also the union
of A and the set of all accumulations point of A.
2.4.5 Example.
1) In R, the intervals [a; b], [a; b), (a; b], (a; b) have the same interior (a; b) and the same
closure [a; b].
2) Take x ∈ R and ε > 0. The interval (x − ε, x + ε) contains at least one rational
number and one irrational number. Hence, the closure of Q and the closure of I are the
set of real numbers R.
3) Every finite or countable (at most countable) subset of R has empty interior. Indeed,
if the set A has at least one interior point x, then there exists an ε > 0 such that
(x − ε, x + ε) ⊂ A. This is a contradiction, since the interval (x − ε, x + ε) is uncountable.
However, even uncountable sets may have empty interior. For instance, the set of
irrational numbers is uncountable, but its interior is empty.
4) In any metric space, the inclusion B(x, r) ⊂ B 0 (x, r) holds. In general, the equality
is not true. The reader can find a counterexample in a discrete metric space.
Some basic properties of the interior and the closure are expressed in the following
theorem.
2.4.6 Theorem. Let A and B be subsets of a metric space X.
o o
1) If A ⊂ B, then A ⊂ B, and A ⊂ B.
o o
2) int (A) = A, cl(A) = A.
o o
3) int (A ∩ B) = A ∩ B, A ∩ B ⊂ A ∩ B.
o
4) int (A ∪ B) ⊃ A ∪ oB, A ∪ B = A ∪ B.
Proof. We will prove the properties related to the closure, and the rest will be left
for the reader.
1) Since B ⊂ B, we get A ⊂ B. Hence A ⊂ B.
2) Of course cl(A) = A since A is a closed set.
3) From the inclusions A ∩ B ⊂ A and A ∩ B ⊂ B, it follows that A ∩ B ⊂ A and
A ∩ B ⊂ B. Consequently, A ∩ B ⊂ A ∩ B.
4) From A ⊂ A ∪ B and B ⊂ A ∪ B, we get A ⊂ A ∪ B and B ⊂ A ∪ B. Therefore
A ∪ B ⊂ A ∪ B. On the other hands, A ∪ B is a closed set containing A ∪ B, so
A ∪ B ⊂ A ∪ B. Thus A ∪ B = A ∪ B.
Note that the equality in 3) does not necessarily happen. For example, choose A = Q

. Luong Ha - HUCE
32 Chapter 2. Metric Spaces

and B = I. Then A ∩ B = R ∩ R = R, but A ∩ B = ∅ = ∅.

2.5 DENSE SETS - SEPARABLE SPACES

2.5.1 Definition.
1) Let A, B be subsets of a metric space X. We call that A is dense in B if A ⊃ B.
If A = X, we say that A is dense (everywhere) in X.
2) A metric space X is called separable if X has a dense subset which is at most
countable.
2.5.2 Remark.
1) Let A, B, C be subsets of X. If A is dense in B, and B is dense in C, then A is
dense in C.
2) A set A is dense in X if and only if for all x ∈ X, and for all ε > 0, there exists an
element a ∈ A such that d(x, a) < ε, or equivalently, for all x ∈ X, there exists a sequence
(xn )n ⊂ A such that xn → x.
It is an important, and sometimes a difficult question to determine whether or not a
given subset of a metric space is dense in the space. It follows from the definition that
a subset A of X is dense in X if and only if A has non-empty intersection with each
open ball in X, or equivalently, if and only if A has non-empty intersection with each
non-empty open subset of X.
2.5.3 Example.
1) The metric space R is separable since the set of rational numbers Q is dense in R,
and Q is countable, and it follows easily that the Euclidean space Rk is separable.
2) The metric space C[a;b] is separable because the set of polynomials with rational
coefficients is countable and dense (Chapter 4).

2.6 OPEN SETS AND CLOSED SETS ON THE REAL LINE

If X is a metric space and A ⊂ X is an open set, then for each x ∈ A, there exists
S
rx > 0 such that B(x, rx ) ⊂ A. Hence, A = B(x, rx ), that is, every open set is a union
x∈A
of a family of open balls. In addition, the open sets of the real line have a more special
structure.
2.6.1 Theorem. Every open set of the real line is a union of an at most countable family
of pairwise disjoint open intervals.
Proof. Let G ⊂ R be an open set. For each x ∈ G, denote by Fx the collection of all
open intervals I such that x ∈ I ⊂ G. This collection is nonempty since G is open. We
can check that the union of any family of open intervals containing a point in common is
S
again an open interval. Hence Jx = I is an open interval: it is the largest element of
I∈Fx

. Luong Ha - HUCE
§ 2. Open Sets and Closed Sets 33

Fx .
If x, y ∈ G, then either Jx = Jy or Jx ∩ Jy = ∅, for otherwise Jx ∪ Jy would be a larger
open interval than Jx in Fx . Thus if we let F = {Jx : x ∈ G}, then the distinct members
S
of F are disjoint, and G = J. Now for each J ∈ F , pick a rational number f (J) ∈ J.
J∈F
Then the map f : F → Q defined as above is injective, since if J 6= J 0 , then J ∩ J 0 = ∅.
Therefore J is at most countable.
Note that these open intervals may be of the form (a; b); (−∞; b), (a; +∞) and (−∞; +∞).
2.6.2 Corollary. Each closed set of the real line R can be obtained by taking out from R
a finite or countable family of disjoint open intervals.

2.7 OPEN SETS AND CLOSED SETS IN SUBSPACES

Let Y be a subspace of a metric space X. Then an open set (or a closed set) in Y is
not necessarily open (or closed) in X. For xo ∈ Y and r > 0, we have

BY (xo , r) = {x ∈ Y | d(x, xo ) < r} = Y ∩ B(xo , r),

where B(xo , r), BY (xo , r) are open balls centered at xo and radius r in X and Y , respec-
tively.
The following theorem indicates the structure of open sets in a subspace.
2.7.1 Theorem. Let Y be a subspace of X and A ⊂ Y . The set A is open in Y if and
only if there exists an open set G in X such that A = G ∩ Y .
Proof. Suppose that the set A is open in Y . Then
³ ´
A = ∪ BY (x, rx ) = ∪ (Y ∩ B(x, rx )) = Y ∩ ∪ B(x, rx ) .
x∈A x∈A x∈A
S
Set G = B(x, rx ). It is clear that G is an open set in X and A = Y ∩ G.
x∈A
Conversely, suppose that A = Y ∩G, where G is an open set in X. If x ∈ A, then x ∈ G,
so there is a number r > 0 such that B(x, r) ⊂ G. Hence BY (x, r) = Y ∩ B(x, r) ⊂ Y ∩ G,
that is, BY (x, r) ⊂ A. We conclude that the set A is open in Y .
2.7.2 Corollary. A set A ⊂ Y is closed in Y if and only if there exists a closed set F in
X such that A = F ∩ Y .
The proof of this corollary is based on Theorem 3.7.1 and the fact that the complement
of a closed set is an open set.
2.7.3 Corollary. Every open (closed) subset of Y is open (closed) in X if and only if Y
is open (closed) in X.
The proof follows directly from Theorem 3.7.1 and Corollary 3.7.2.
Example. Consider the subspace Y = [0; 1) ∪ {2} of R. Choose A = [0; 1) ⊂ Y .
Then the set A is neither open nor closed in R. However since

A = (−1; 1) ∩ Y = [−1; 1] ∩ Y,

. Luong Ha - HUCE
34 Chapter 2. Metric Spaces

it follows that A is both open and closed in Y .

PROBLEMS
2.5 Let (X, d) be a metric space, ∅ 6= A ⊂ X and r > 0. Set

U = {x ∈ X : d(x, A) < r}

V = {x ∈ X : d(x, A) ≤ r}.
Prove that the set U is open and the set V is closed.
2.6 Let xo ∈ C[0;1] and M = {x ∈ C[0;1] : x(t) < xo (t) ∀t ∈ [0; 1]}.
Prove that M is an open set and the set

N = {x ∈ C[0;1] : a < x(t) < b ∀t ∈ [0; 1]}

is open (a, b ∈ R).

1
2.7 Let A = {m + : m, n ∈ N}.
n
Find the set of accumulation points, the interior and the closure of A.
2.8 Let F be a closed subset of R. Prove that there exists a finite or countable subset E
of F such that E = F .
2.9 Show that
A = A ∪ ∂A
o
A = A \ ∂A.

2.10 In a metric space X, a subset H is called an Fσ -set if H is the union of a countable

family of closed sets. A subset K is a Gδ -set if K is the intersection of a countable family
of open sets.
1) Let ∅ 6= A ⊂ X. Show that
∞
\ 1
A= {x ∈ X : d(x, A) < }.
n=1
n

2) Show that every closed set is a Gδ -set, and every open set is an Fσ -set. Give an
example to show that a Gδ -set may not be closed.
3) Show that every countable subset of X is a Fσ -set, and the complement of an Fσ -set
is a Gδ -set, and vice versa.
2.11 Let (Aα )α∈I be a family of subsets of a metric space X. Prove that

∪ Aα ⊂ ∪ Aα and ∩ Aα ⊂ ∩ Aα .
α∈I α∈I α∈I α∈I

Give an example of a sequence of subsets (An )n of R such that

∞ ∞
∪ An 6= ∪ An .
n=1 n=1

. Luong Ha - HUCE
§ 3. Continuous Mappings 35

2.12 Let E be a dense subset of R, and let A be a finite subset of E. Prove that the set
E \ A is dense in R.
2.13 Let G be an open dense subset of R. Suppose that I is an interval of R. Prove that
there exists an interval Io such that I o ⊂ I ∩ G.
2.14 Let A and B be subsets of a metric space X. Show that if B is open, then

A ∩ B ⊂ A ∩ B.

2.15 Show that in any Euclidean space Rk , the closure of any open ball B(a, r) is the
closed ball B 0 (a, r).
2.16 Prove that in any metric space X, the intersection of two dense subsets of X is also
dense in X, provided that there is at least one set is open. Give an example for showing
that if both sets are not open, then the result above is false.
2.17 Prove that any subspace of a separable metric space is separable.
2.18 Prove that in a discrete metric space, every subset is both open and closed. Hence
a discrete space X is separable if and only if the set X is at most countable.
2.19 Prove that the metric spaces Rk and `2 are separable.
2.20 Let A and B be two nonempty subsets of a metric space X. Prove that
1. The set {x ∈ X : d(x, A) < d(x, B)} is open.
2. If A and B are closed and disjoint, then there exist two open sets U, V in X such
that A ⊂ U, B ⊂ V and U ∩ V = U ∩ V = ∅.

§3 CONTINUOUS MAPPINGS

3.1 CONTINUOUS MAPPINGS AND BASIC PROPERTIES

3.1.1 Definition.
Let f : X → Y be a mapping between two metric spaces. The mapping f is called
continuous at a point xo ∈ X if for all ε > 0, there exists δ > 0 (depending on ε and xo )
such that d(f (x), f (xo ) < ε whenever d(x, xo ) < δ.
Using the notion of open balls, this definition is equivalent to the following statement.
³ ´
(∀ε > 0)(∃δ > 0) f (B(xo , δ)) ⊂ B(f (xo ), ε) .

A mapping f is called continuous on a set A ⊂ X if f is continuous at every point

x ∈ A. When f is continuous on X, we will say that ”f is continuous” for simplicity.
3.1.2 Example.
1) Let f : C[a;b] → R be defined by f (x) = x(a). Pick xo ∈ C[a;b] . Then

|f (x) − f (xo )| = |x(a) − xo (a)| ≤ d(x, xo ).

. Luong Ha - HUCE
36 Chapter 2. Metric Spaces

Therefore, for any ε > 0, choose δ = ε. If d(x, xo ) < δ, then |f (x) − f (xo )| < ε. Thus, f
is continuous at xo . It follows that f is continuous on C[a;b] .
2) Let X, Y be metric spaces and let f : X → Y . Suppose that there is a number k > 0
such that d(f (x), f (y)) ≤ kd(x, y) for all x, y ∈ X. Then the mapping f is continuous.
ε
Indeed, let xo ∈ X, we have d(f (x), f (xo ) ≤ kd(x, xo ). Hence, for any ε > 0, take δ = ,
k
then d(f (x), f (xo )) < ε for all x ∈ X, d(x, xo ) < δ.
3.1.3 Theorem. Let f : X → Y, xo ∈ X. Then the mapping f is continuous at xo if
and only if for all sequences (xn )n ⊂ X, xn → xo , then f (xn ) → f (xo ).
Proof. Suppose f is continuous at xo . Take (xn )n ⊂ X, xn → xo . By definition, for
any ε > 0, there exists δ > 0 such that if d(x, xo ) < δ, then d(f (x), f (xo )) < ε. Since
xn → xo , there is no ∈ N such that d(xn , xo ) < δ for all n ≥ no . Thus if n ≥ no , then
d(f (xn ), f (xo )) < ε. This means that f (xn ) → f (xo ).
Coversely, suppose that f (xn ) → f (xo ) for any sequence (xn )n ⊂ X, xn → xo . We
will show that f is continuous at xo . If not so, then there is an ε > 0 such that for all
δ > 0, there exists x ∈ X with d(x, xo ) < δ and d(f (x), f (xo )) ≥ ε. Consequently, for
1
every n ∈ N, there is an element xn ∈ X satisfying d(xn , xo ) < and d(f (xn ), f (xo )) ≥ ε.
n
Clearly, xn → xo but f (xn ) 6−→ f (xo ). This is a contradiction and the theorem is proved.¥
For instance, the mapping f : R → R defined by f (x) = x2 is continuous. Indeed, if
xn → xo , then f (xn ) = x2n will converge to x2o = f (xo ).
The following result follows easily from Theorem 3.1.3.
3.1.4 Theorem. Let f : X → Y, g : Y → Z be continuous mappings between metric
spaces. If f and g are continuous, then the composition g ◦ f is continuous.
The next theorem describes the most useful characterizations of the continuous map-
pings.
3.1.5 Theorem. For a mapping f : X → Y between metric spaces, the following state-
ments are equivalent:
1) f is continuous.
2) The preimage f −1 (G) is an open subset of X whenever G is an open subset of Y .
3) The preimage f −1 (F ) is a closed subset of X whenever F is a closed subset of Y .
Proof.
1) ⇒ 2). Let G be an open subset of Y and xo ∈ f −1 (G). Since f (xo ) ∈ G, there exists
some ε > 0 such that B(f (xo ), ε) ⊂ G. By the continuity of f at xo there exists δ > 0
such that f (B(xo , δ)) ⊂ B(f (xo ), ε). This shows that B(xo , δ) ⊂ f −1 (G). Therefore, xo
is an interior point of f −1 (G), and hence f −1 (G) is open.
2) ⇒ 1) Let xo ∈ X and ε > 0. Since B(f (xo ), ε) is an open set of Y , by hypothesis
the set U = f −1 (B(xo ), ε) is an open set of X. Since xo ∈ U , there exists δ > 0 such that
B(xo , δ) ⊂ U . But this shows that f (B(xo , δ)) ⊂ B(f (xo ), ε). Therefore, f is continuous

. Luong Ha - HUCE
§ 3. Continuous Mappings 37

at xo . Since xo is arbitrary, f is continuous on X.

Note that for any B ⊂ Y , the following equality holds:
f −1 (Y \ B) = X \ f −1 (B).
Since the complement of an open set is a closed set and vice versa, it is easy to show
that 2) ⇔ 3). The proof of the theorem is now complete.
Now we give an example of applying Theorem 3.1.5.
Let A be a nonempty subset of a metric space X. Consider f : X → R defined by
f (x) = d(x, A). It follows easily from the equality |d(x, A) − d(xo , A)| ≤ d(x, xo ) that f
is continuous on X.
Set U = {x ∈ X | d(x, A) < r}, V = {x ∈ X | d(x, A) ≤ r}, where r > 0. Then
U = f −1 ((−∞; r)) and V = f −1 ((−∞; r]). Therefore, U is open and V is closed in X.
3.1.6 Theorem. Let f and g be continuous mappings of a metric space X into R and
let k ∈ R. Then the mappings f + g, kf, f g, |f | and max(f, g), min(f, g) are continuous
1
on X. If is defined, then it is continuous.
f
As usual, the mappings f + g, kf, f g, |f | and max(f, g), min(f, g) are defined as
follows:
(f + g)(x) = f (x) + g(x)
(kf )(x) = k f (x)
(f g)(x) = f (x)g(x)
|f |(x) = |f (x)|
max(f, g)(x) = max(f (x), g(x))
min(f, g)(x) = min(f (x), g(x)).
1
Moreover, if f (x) 6= 0 for every x ∈ X, then the mapping is defined by
f
³1´ 1
(x) = .
f f (x)
Note that
1
max(f, g) = (f + g + |f − g|)
2
1
min(f, g) = (f + g − |f − g|).
2
Proof. Let xo be an arbitrary point of X. First we prove that f + g is continuous
ε
at xo . For ε > 0, by continuity there exists δ > 0 such that |f (x) − f (xo )| < and
2
ε
|g(x) − g(xo )| < whenever d(x, xo ) < δ. Then, if d(x, xo ) < δ,
2
|(f + g)(x) − (f + g)(xo )| = |(f (x) − f (xo )) + (g(x) − g(xo ))|
≤ |f (x)f (xo )| + |g(x)g (xo )|
< ε.

. Luong Ha - HUCE
38 Chapter 2. Metric Spaces

This shows that f + g is continuous at xo . Similarly, we can prove that kf is continuous

at xo . The continuity of the mapping |f | follows from the inequality
¯ ¯
¯ ¯
¯|f (x)| − |f (xo )|¯ ≤ |f (x) − f (xo )|.

We use the following equalities to prove the continuity of the mappings f g, max(f, g),
and min(f, g):
1¡ ¢2
fg =(f + g)2 − (f − g)
4
1
max(f, g) = (f + g + |f − g|)
2
1
min(f, g) = (f + g − |f − g|).
2
1
Finally, proving the continuity ofis left for the reader.
f
Note that if we replace R by C, then the results above still hold for the mappings
1
f + g, kf, f g, |f | and .
f

3.2 UNIFORMLY CONTINUOUS MAPPINGS

3.2.1 Definition.
Let X and Y be two metric spaces and a mapping f : X → Y . The mapping f is
called uniformly continuous if for every ε > 0, there exists δ > 0 (depending only ε) such
that d(f (x), f (y)) < ε whenever d(x, y) < δ.
Clearly, every uniformly continuous mapping is continuous. Note that in the case of
uniform continuity, the number δ only depends on ε.
3.2.2 Example.
1) The mappings in Example 4.1.2 are uniformly continuous.
2) The function f (x) = x2 from R to R is continuous but not uniformly continuous.
1
Indeed, choose (xn )n and (x0n )n in R with xn = n and x0n = n + .
n
0 0 1
Then d(xn , xn ) = |xn − xn | = → 0, but
n
1 1
|f (xn ) − f (x0n )| = |(n + ) − n2 | = 2 + 2 > 2
n n
for all n ∈ N, so that f is not uniformly continuous on R.

3.3 HOMEOMORPHISMS

3.3.1 Definition.
Let X and Y be two metric spaces. A mapping f : X → Y is called a homeomorphism
if f is bijective, continuous and the inverse f −1 : Y → X is continuous. Then we call that
X is homeomorphic to Y .

. Luong Ha - HUCE
§ 3. Continuous Mappings 39

3.3.2 Remark.
1) The relation ”homeomorphic to” is an equivalence relation on the class of metric
spaces.
2) Let f : X → Y be a homeomorphism and A ⊂ X. If A is open (or closed), then
f (A) is open (or closed, respectively) in Y . Therefore, if x is an interior point, or an
adherent point, or a boundary point of A, then the image f (x) will be an interior point,
or an adherent point, or a boundary point of f (A), respectively.
3.3.3 Example.
1) The mapping f : R → (0; +∞) defined by f (x) = ex is a homeomorphism.
2) The mapping ϕ : (0; 1) → (a; b), where ϕ(t) = (1 − t)a + tb is a homeomorphism.
π π
3) The mapping f : (− ; ) → R, where f (x) = tan x is a homeomorphism.
2 2
Hence, on the real line, all open intervals are homeomorphic, and they also homeo-
morphic to R.

3.4 ISOMETRY

3.4.1 Definition.
Let X and Y be two metric spaces. A mapping f : X → Y is called an isometry if f
is a bijection and d(f (x), f (y)) = d(x, y) holds for all x, y ∈ X.
A space X is called isometric to the space Y if there is an isometry from X onto Y .
In fact, in this definition, we only need that f is surjective, since from the above
equality, it follows that f always is injective.
3.4.2 Example.
1) Consider the set R equipped with the metric d(x, y) = |ex − ey |. Then the space
(R, d) is isometric with the space X = (0; +∞) with the usual metric.
Set f : (R, d) → X with f (x) = ex . It is evident that f is surjective. Moreover, for all
x, y ∈ R,
|f (x) − f (y)| = |ex − ey | = d(x, y).

Hence f is an isometry from (R, d) onto X.

t−a
2) Let ϕ : C[0;1] → C[a;b] be defined by ϕ(x) = x∗ , where x∗ (t) = x( ) for each
b−a
t ∈ [a; b].
Then ϕ is an isometry, and the proof will be left as an exercise for the reader.
3.4.3 Remark.
1) The relation ”isometric to” is an equivalence relation on the class of metric spaces.
2) If f is an isometry from X onto Y , then f is a homeomorphism.
3) If X is isometric to Y , we can identify these spaces. This means that these two

. Luong Ha - HUCE
40 Chapter 2. Metric Spaces

metric spaces are undistinguishable in so far as their properties as metric spaces are
concerned.

3.5 EQUIVALENT METRICS

3.5.1 Definition.
Let X be a nonempty set. Suppose that d1 and d2 are two metrics on X. We say
that d1 is topologically equivalent to d2 if the identity mapping id : (X, d1 ) → (X, d2 ) is a
homeomorphism.
Two metrics d1 and d2 are called uniformly equivalent if there are two positive numbers
M, m such that for all x, y ∈ X,
md2 (x, y) ≤ d1 (x, y) ≤ M d2 (x, y).
3.5.2 Remark.
1) The relations ”topologically equivalent” and ”uniformly equivalent” are equivalence
relations on the set of metrics defined on the set X.
2) If the metrics d1 and d2 are uniformly equivalent, then they are topologically equiv-
alent.
3) The identity mapping id : (X, d1 ) → (X, d2 ) is continuous if and only if for each
open set G in (X, d2 ), id−1 (G) is open in (X, d1 ), or if and only if for each open set G in
(X, d2 ), G is open in (X, d1 ).
Hence the mapping id is a homeomorphism if and only if every G is open in (X, d1 ) is
open in (X, d2 ), and vice versa. This means that two metrics d1 and d2 are topologically
equivalent when and only when they generate the same class of open sets. By Theorem
3.1.3, we can use sequences to express the notion of topological equivalence.
Indeed, the following statements are equivalent:
a) d1 and d2 are topologically equivalent.
b) For all sequences (xn )n ⊂ X, and for all x ∈ X:
d1 (xn , x) → 0 ⇔ d2 (xn , x) → 0.
4) Suppose that d1 and d2 are uniformly equivalent, then for all x, y ∈ X,
1 d2 (x, y) 1
≤ ≤ ,
M d1 (x, y) m
n d (x, y) o
2
that is, the set : x 6= y is bounded with a positive lower bound.
d1 (x, y)
3.5.3 Example.
1) The usual metric d on Rk is uniformly equivalent to the metrics:
k
X
d1 (x, y) = |xi − yi |, d2 (x, y) = max |xi − yi |.
1≤i≤k
i=1

. Luong Ha - HUCE
§ 3. Continuous Mappings 41

Indeed, we can verify that

√
d2 (x, y) ≤ d(x, y) ≤ n d2 (x, y)

d2 (x, y) ≤ d1 (x, y) ≤ n d2 (x, y).

These inequalities show that the metrics are uniformly equivalent.
2) The usual metric d on R is topologically equivalent to the metric d∗ defined by
d∗ (x, y) = |x3 − y 3 |, but they are not uniformly equivalent. First we can prove that d
and d∗ are topologically equivalent by using convergent sequences as in Remark 3.5.2.
Moreover, for any x, y ∈ R, x 6= y,
d∗ (x, y)
= x2 + xy + y 2 .
d(x, y)
d∗ (n, 0) n d∗ (x, y) o
Therefore, sup = sup n2 = +∞, hence, the set : x 6= y is not bounded.
n d(n, 0) n∈N d(x, y)
∗
Thus, d and d are not uniformly equivalent.

PROBLEMS
2.21 Prove that the following functions are continuous on the metric space C[0;1] :
R1
1) f (x) = x(t)dt;
0
R1
2) f (x) = ϕ(t)x(t) dt, where ϕ is a given element of C[0;1] .
0
2.22 Let X and Y be metric spaces and let f be a mapping from X into Y . Prove that
the following statements are equivalent:
1) f is continuous;
2) f (A) ⊂ f (A) for any A ⊂ X;
o
3) f −1 (B) ⊂ int(f −1 (B)) for any B ⊂ Y .
2.23 Let X and Y be metric spaces and let f be a continuous mapping from X onto Y .
Suppose that A is a dense subset of X.
Prove that f (A) in dense in Y , and if X is separable, then Y is separable.
x
2.24 Prove that the mapping f : R → (−1; 1) defined by f (x) = is a homeomor-
1 + |x|
phism.
2.25 Let A = {(x, y) ∈ R2 : x2 + y 2 = 1} and B = [0; 2π). Prove that the mapping
f : B → A defined by f (x) = (cos x, sin x) is a continuous bijection, but that f −1 is not
continuous.
2.26 Let A = {(x, y) ∈ R2 : x2 + y 2 < 1}. Show that A is homeomorphic to R2 .
2.27 Let A and B be two nonempty closed subsets of a metric space X and A ∩ B = ∅.
For every x ∈ X, set f (x) = d(x, A) − d(x, B). Prove that

. Luong Ha - HUCE
42 Chapter 2. Metric Spaces

1) f is uniformly continuous on X.
2) there exist two open subsets U, V of X such that A ⊂ U, B ⊂ V and U ∩ V = ∅.
2.28 Let f : X → R be a mapping. Prove that f is continuous if and only if for all a ∈ R,
the sets f −1 ((−∞; a)) and f −1 ((a; +∞)) are open in X.
2.29 Let X and Y be two metric spaces, f and g be continuous mappings from X into
Y . Set A = {x ∈ X : f (x) = g(x)}.
Prove that the set A is closed in X. What conclusion can we draw if the set A is dense
in X?
2.30 Let F1 and F2 be closed subsets of a metric space X such that X = F1 ∪ F2 . Suppose
f : X → Y is a mapping such that the restrictions f|F1 and f|F2 are continuous. Prove
that f is continuous on X.
2.31 Let d1 , d2 be two metrics on X. Prove that d1 and d2 are topologically equivalent if
and only if for all A ⊂ X and for all x ∈ X,
d1 (x, A) = 0 ⇔ d2 (x, A) = 0.
2.32 Let X be a metric space.
1) A set F ⊂ X is closed if and only if there exists a continuous function f : X → R
such that F = {x ∈ X : f (x) ≤ 0}.
2) A set G ⊂ X is open if and only if there exists a continuous function f : X → R
such that G = {x ∈ X : f (x) < 0}.
2.33 Let A, B be two nonempty disjoint closed subsets of a metric X and consider the
function f : X → R defined by
d(x, A)
f (x) = .
d(x, A) + d(x, B)
Prove that
1) the function f is continuous, and f (A) = {0}, f (B) = {1}.
2) there exist two open sets U, V in X such that A ⊂ U, B ⊂ V and U ∩ V = ∅.
3) if d(A, B) = inf {d(a, b) : a ∈ A, b ∈ B} > 0, then f is uniformly continuous.
2.34 Prove that on the set X consisting of all the real functions continuous on the interval
[0; 1], two metrics
Z1
¯ ¯ ¯ ¯
d1 (x, y) = max ¯ x(t) − y(t)¯ and d2 (x, y) = ¯ x(t) − y(t)¯ dt
t∈[0;1]
0

are not topologically equivalent.

2.35 Let f be an increasing function defined on [0; +∞) such that
f (0) = 0, f (x) > 0 whenever x > 0 and f (x + y) ≤ f (x) + f (y) for all x ≥ 0, y ≥ 0.
1) Let X be a metric space and let d1 (x, y) = f (d(x, y)) for all x, y ∈ X. Show that
d1 is a metric on X.

. Luong Ha - HUCE
§ 3. Continuous Mappings 43

2) Show that if f is continuous at x = 0, then the metrics d and d1 are uniformly

equivalent.
3) Verify that the functions
x
xr (0 < r ≤ 1), ln(1 + x), , min(1, x)
1+x
satisfy the above conditions.
Show that each metric on X is uniformly equivalent to a bounded metric.
(A metric d on X is called a bounded if there is M > 0 such that d(x, y) ≤ M for all
x, y ∈ X.)

. Luong Ha - HUCE
44 Chapter 2. Metric Spaces

HINTS and SOLUTIONS

2.1 Hints. ∞
X
2) Note that (xn − yn ) ≤ 2
2(x2n + yn2 ), hence, the series (xn − yn )2 is convergent for
n=1
any x, y ∈ `2 . For each n ∈ N, the triangle inequality on Rn gives
n
X n
X n
X
2 2
(xk − zk ) ≤ (xk − yk ) + (yk − zk )2 .
k=1 k=1 k=1

Let n → ∞, we get the triangle inequality for d.

3) To prove the triangle inequality, we use the Cauchy-Schwarz inequality for two
continuous functions on [a; b]:
¯ Rb ¯ ³ Rb ´ 21 ³ Rb ´ 12
¯ ¯ 2 2
¯ f (x)g(x)dx¯ ≤ f (x)dx g (x)dx .
a a a
2.2 Hints.
∗ The verification of three axioms of a metric for d is similar to that for the metric
space C[a;b] .
∗ Let (xn )n ⊂ B(X) and let x ∈ B(X). Then
lim d(xn , x) = 0 ⇔ lim sup |xn (t) − x(t)| = 0.
n n t∈X
This means that for each ε > 0, there exists no ∈ N such that |xn (t) − x(t)| < ε for
all n ≥ no and for all x ∈ X. In other words, the sequence of functions (xn )n converges
uniformly to the function x on [a; b].
2.3 Hints.
1
a) It is easy to see that xn (t) → 0 for each t ∈ [0; 1]. In addition, d(xn , 0) = for all
4
n ∈ N. Hence, lim d(xn , 0) 6= 0.
n
Thus, the sequence (xn )n does not converge in C[0;1] .
b) and c): Each sequence (xn )n converges to the point 0 (the constant function 0) in
C[0;1] .
2.4 Hints.
Since d is a metric on X, the triangle inequality holds for any elements of X:
d(x, z) ≤ d(x, y) + d(y, z).
Note that if a and b are nonnegative real numbers, then
a b
a≤b⇔ ≤ .
1+a 1+b
Hence
d(x, z) d(x, y) + d(y, z)
≤
1 + d(x, z) 1 + d(x, y) + d(y, z)
d(x, y) d(y, z)
= +
1 + d(x, y) + d(y, z) 1 + d(x, y) + d(y, z)
d(x, y) d(x, z)
≤ + .
1 + d(x, y) 1 + d(y, z)

. Luong Ha - HUCE
§ 3. Continuous Mappings 45

Thus, d1 (x, z) ≤ d1 (x, y) + d1 (y, z).

d(x, y)
For any x, y ∈ X, it follows from the equality d1 (x, y) = that
1 + d(x, y)
d1 (x, y)
d(x, y) = .
1 − d1 (x, y)
Hence, lim d1 (xn , xo ) = 0 ⇔ lim d(xn , xo ) = 0.
n n
The above results still hold true for the metrics d2 and d3 .
2.5 Hints.
Pick xo ∈ U . Then d(xo , A) < r, hence, r − d(xo , A) = ε > 0. We shall show that
B(xo , ε) ⊂ U . Indeed, let x ∈ B(xo , ε). Then d(x, xo ) < ε, and
d(x, A) ≤ d(x, xo ) + d(xo , A) < ε + d(xo , A) = r,
so x ∈ U . Thus, the set U is open.
Similarly, by taking the complement, we can show that V is closed.
2.6 Hints.
Let x ∈ M . The function xo − x is in C[0;1] , and xo − x > 0. Therefore, there exists
to ∈ [0; 1] such that
min (xo − x) = xo (to ) − x(to ) = r > 0.
t∈[0;1]

We show that the open ball B(x, 2r ) ⊂ M .

r r
Let y ∈ B(x, 2r ). Then |y(t) − x(t)| < for all t ∈ [0; 1]. Thus, x(t) − y(t) ≥ − and
2 2
r
xo (t) − x(t) ≥ r for all t ∈ [0; 1]. It follows that xo (t) − y(t) ≥ > 0 for all t ∈ [0; 1].
2
This means that y ∈ M , and hence, B(x, 2r ) ⊂ M . We have the desired result.
Similarly, the set {x ∈ C[0;1] : xo (t) < x(t) ∀t ∈ [0; 1]} also is open. Hence,
N = {x ∈ C[0;1] : a < x(t) < b ∀t ∈ [0; 1]}
is open.
2.7 Hints.
o
Since A is countable, it cannot contain any open interval. Hence A = ∅.
The set of adherent points of A is N (why?). Hence, A = A ∪ N.
2.9 Hints.
By definition, A ⊂ A, and ∂(A) ⊂ A. Hence, A ∪ ∂(A) ⊂ A.
Let x ∈ A. Then x is an adherent point of A. If x ∈/ A, then x is an adherent point
of X \ A. Hence, x ∈ ∂(A). Thus, A ⊂ A ∪ ∂(A). It follows that A = A ∪ ∂(A).
Note that if x ∈ A, then either x is an interior point of A or x is a boundary point of
o
A. Therefore, A = A \ ∂(A).
2.10 Hints.
1) Let
1
Gn = {x ∈ X : d(x, A) < }.
n

. Luong Ha - HUCE
46 Chapter 2. Metric Spaces

Then Gn is open. Clearly,

∞
\ 1
x∈ Gn ⇔ d(x, A) < for all n ∈ N.
i=1
n

This is equivalent to d(x, A) = 0, or x ∈ A.

2) If A is closed, then A = A. The result follows directly from 1).
∗ The set of irrational numbers is a Gδ -set, but it is not closed in R.
2.11 Hints.
Consider an enumeration of Q = {r1 , r2 , . . . , rn , . . . } ⊂ R.
Let An = {rn }, then An = An for each n ∈ N. Thus,
S∞ S∞ S
∞ S
∞ S
∞
An = An = Q, but An = R, hence An 6= An .
n=1 n=1 n=1 n=1 n=1
2.13
Without loss of generality, we may assume that I is an open interval of R. Since G is
open and dense in R, G ∩ I is a nonempty open set. Therefore, there is x ∈ G ∩ I and
ε ε ε ε
ε > 0 such that (x − ε, x + ε) ⊂ I ∩ G. Let Io = (x − , x + ). Then I o = [x − , x + ] ⊂
2 2 2 2
(x − ε, x + ε), hence, I o ⊂ I ∩ G.
2.16
Let G1 and G2 be two dense subsets of X, where G1 is open. Let V be a nonempty
open subset of X. We shall show that V ∩ (G1 ∩ G2 ) is nonempty.
Since G1 is dense in X, we have G1 ∩ V 6= ∅. This set is open, and G2 is dense in X,
hence, (G1 ∩ V ) ∩ G2 6= ∅ or (G1 ∩ G2 ) ∩ V 6= ∅. Thus, G1 ∩ G2 is dense in X.
∗ The set of rational numbers and the set of irrational numbers are dense in R, but
their intersection is empty, hence this intersection cannot be dense in R.
2.17 Hints.
By hypothesis, X has an at most countable subset A which is dense in X. Consider
1 1
the collection of open balls (B(x, )), with x ∈ A and n ∈ N. If Y ∩ B(x, ) 6= ∅, then
n n
1
choose a point xn ∈ Y ∩ B(x, ). The set of all such xn is at most countable and dense
n
in Y .
2.19
1) We show that Qk is dense in Rk . Consider x ∈ Rk and ε > 0. Then x =
(x1 , x2 , . . . , xk ), and for each i ∈ {1, 2, . . . , k}, there exists a rational number ri so that
ε
|xi − ri | < √ . Set r = (r1 , r2 , . . . , rn ). Clearly, r ∈ Qk , and
k
Xk
2 ε2
d (x, r) = (xi − ri )2 < k. = ε,
i=1
k

hence, d(x, r) < ε. Thus, Qk is countable and dense in Rk .

2) Let x ∈ `2 and let ε > 0.

. Luong Ha - HUCE
§ 3. Continuous Mappings 47

Set A = {r1 , r2 , . . . , rn , 0, 0, . . . ) : n ∈ N, ri ∈ Q}. Then A is a countable subset of `2 .

P
The element x = (xn )n ∈ `2 , ∞ 2
n=1 xn < +∞. Hence, for any given ε > 0, there exists
P ε2
no ∈ N such that ∞ n=no +1 x 2
n < . By part 1), there are rational numbers r1 , r2 , . . . , rn
2
Pno 2
2 ε
such that n=1 (xi − ri ) < . Let r = (r1 , r2 , . . . , rn , 0, 0, . . . ). Then r ∈ A, and
P 2
d2 (x, r) = ∞ 2 2
i=1 (xi − ri ) < ε . Thus, d(x, r) < ε. This shows that `2 is separable.

2.20 Hints.
Let G = {x ∈ X : d(x, A) < d(x, B)}, and pick xo ∈ G. Then d(xo , B) − d(xo , A) =
ε
ε > 0. Show that B(xo , ) ⊂ G.
2
Note that if A is closed, then d(x, A) > 0 if and only if x ∈
/ A. Hence, if A and B are
disjoint closed sets, then for any x ∈ A, d(x, A) = 0 < d(x, B) , that is, A ⊂ G.
If there exists x ∈ G ∩ B, then d(x, A) < d(x, B) = 0, which is impossible. Hence,
G ∩ B = ∅. Let U = G, V = X \ G. Show that U and V are open sets we need.
2.21 Hints.
1) Pick xo ∈ C[0;1] . Then for any x ∈ C[0;1]

¯ Z1 ¯ Z1
¯ ¯
|f (x) − f (xo )| = ¯ (x(t) − xo (t))dt¯ ≤ |x(t) − xo (t)| dt.
0 0

R1
Hence |f (x) − f (xo ) ≤ d(x, xo )dt = d(x, xo ).
0
2) By hypothesis, the function ϕ ∈ C[0;1] , there is k > 0 such that |ϕ(t)| ≤ k for all
t ∈ [0; 1]. Then

¯ Z1 ¯ Z1
¯ ¯
|f (x) − f (xo )| = ¯ ϕ(t)(x(t) − xo (t))dt¯ ≤ |ϕ(t)||x(t) − xo (t)|dt.
0 0

R1
Therefore, |f (x) − f (xo )| ≤ k d(x, xo )dt = kd(x, xo ).
0
From the above inequalities, it is easy to follow that the given functions are continuous
at any xo of C[0;1] .
2.23. Let y ∈ Y . Since f is surjective, there exists x ∈ X such that f (x) = y. The set
A is dense in X, hence there is a sequence (xn )n ⊂ X so that xn → x. By continuity,
f (xn ) → f (x) = y. Clearly, (f (xn ))n ⊂ f (A). Thus, f (A) is dense in Y .
If A is at most countable, so is f (A). Hence, if X is separable, then Y is also separable.
2.24 Hints.
x
First show that f is bijective. Its inverse is defined by f −1 (x) = . It is easy to
1 − |x|
see that both f and f −1 are continuous. Hence f is a homeomorphism.
2.27. Hints.
1) For any x, y ∈ X, we have

. Luong Ha - HUCE
48 Chapter 2. Metric Spaces

¯ ¯
|f (x) − f (y)| = ¯(d(x, A) − d(y, A)) − (d(x, B) − d(y, B))¯
≤ |d(x, A) − d(y, A)| + |d(x, B) − d(y, B)|
≤ 2 d(x, y).

It is easy to follow that f is uniformly continuous on X.

2) Let U = {x ∈ X : f (x) < 0}, V = {x ∈ X : f (x) > 0}. Then U and V are the
desired open sets.
2.28 Hints.
The necessary condition is clear. Conversely, it suffices to show that f −1 ((a; b)) is open
for any a, b ∈ R. This follows from the equality

f −1 ((a; b)) = f −1 ((−∞; b)) ∩ f −1 ((a; +∞)).

2.30 Let F be an arbitrary closed set in Y . Since two mappings f|F1 and f|F2 are contiuous,
−1 −1
f|F1
(F ) and f|F 2
(F ) are closed in F1 and F2 , respectively. Hence, they are closed in X
because F1 and F2 are closed in X. Moreover, F1 ∪ F2 = X implies that f −1 (F ) =
−1 −1
f|F1
(F ) ∪ f|F 2
(F ). Therefore, f −1 (F ) is closed in Y . This means that f is continuous.
2.31 Hints.
Show that the closures of any A ⊂ X in (X, d1 ) and (X, d2 ) are equal. Thus, A is
closed in (X, d1 ) if and only if A is closed in (X, d2 ). Hence, the metrics d1 and d2 define
the same family of open sets.
2.32 Hints.
Show that the function f : X → R defined by

d(x, A)
f (x) =
d(x, A) + d(x, B)

is continuous on X.
1 1
Let U = {x ∈ X : f (x) < }, V = {x ∈ X : f (x) > }.
2 2
Then the function f and the sets U, V meet the requirements of the problem.

. Luong Ha - HUCE
Chapter 3

Complete Metric Spaces

On the real line R, every Cauchy sequence is convergent. But if we consider the sequence
1
(xn )n ⊂ (0; 1], where xn = , then it is a Cauchy sequence but not convergent to any
n
element of (0; 1]. The metric spaces in which the property is satisfied as in R play an
important role in the class of metric spaces. That is the class of complete metric spaces.

§1 CAUCHY SEQUENCES AND COMPLETE METRIC SPACES

1.1 BASIC NOTIONS AND PROPERTIES

1.1.1 Definition. A sequence (xn )n in a metric space X is called a Cauchy sequence (or
a fundamental sequence) if lim d(xn , xm ) = 0. This means that for each ε > 0, there
n,m
exists no ∈ N such that d(xn , xm ) < ε for all m ≥ no , n ≥ no .
A metric space X is called complete if every Cauchy sequence in X is convergent (to
a point of X).
1.1.2 Remark.
1) If (xn )n is an convergent sequence, then it is a Cauchy sequence.
2) If (xn )n is a Cauchy sequence and there is a subsequence (xnk )k converging to xo ,
then xn → xo .
Indeed, for any k ∈ N,

0 ≤ d(xk , xo ) ≤ d(xk , xnk ) + d(xnk , xo ).

By definition, the righthand side tends to zero, hence xk → xo .

1.1.3 Example.
1) The real line R equipped with the usual metric is complete.
2) The subspace X = (0; 1] of R is not complete.
3) Every Euclidean space Rk is complete.

49
50 Chapter 3. Complete Metric Spaces

Let (xn )n be any Cauchy sequence in Rk . Then

hX
k i 21 ¯ ¯
(n) (m) ¯ (n) (m) ¯
d(xn , xm ) = (xi − xi )2 and ¯xi − xi ¯ ≤ d(xn , xm )
i=1

for each i = 1, 2, ..., k.

(n) (n)
Clearly, (xi )n is a Cauchy sequence in R for each i = 1, 2, . . . , k. Thus, xi → xoi ∈ R.
Set xo = (xo1 , xo2 , . . . , xok ), then it is evident that xn → xo .
4) The metric space C[a;b] is complete.
Let (xn )n be any Cauchy sequence in C[a;b] . Then for each ε > 0, there exists no ∈ N
such that d(xn , xm ) < ε for all m, n ≥ no , that is, |xn (t) − xm (t)| < ε for all t ∈ [a; b] and
for all m ≥ no , n ≥ no . Therefore, the sequence of functions (xn )n is uniformly convergent
on [a; b]. Set lim xn (t) = x(t) for each t ∈ [a; b]. It follows that the function x is continuous
n
on [a; b]. Hence, x ∈ C[a;b] and xn → x.
L
5) The space C[a;b] is not complete. This will be left for the reader as an exercise.
6) Every discrete metric space is complete.
Indeed, let (xn )n be a Cauchy sequence in a discrete metric space X. Then for ε = 1,
there exists no ∈ N such that d(xn , xm ) < 1 for all n, m ≥ no . Hence xn = xno for all
n ≥ no . Obviously, xn → xno .

1.2 BASIC PROPERTIES

1.2.1 Theorem. Let X be a metric space and Y be a subspace of X.

1) If the space Y is complete, then Y is closed in X.
2) Suppose that X is complete and Y is closed in X. Then Y is a complete space.
Proof.
1) Let (xn )n ⊂ Y and xn → x ∈ X. Then (xn )n is Cauchy sequence in Y . Since Y
is complete, the sequence (xn )n must converge to some point x0 ∈ Y . Hence, x = x0 , so
that x ∈ Y . Thus, Y is a closed set.
2) Let (xn )n ⊂ Y be any Cauchy sequence. Then (xn )n is a Cauchy sequence in X.
Since X is complete, xn → x ∈ X. It follows from the closedness of the set Y that x ∈ Y .
Hence, the space Y is complete.¥
This theorem allows us to construct many complete or incomplete metric spaces from
a given metric space. For instance, the subspace N of R is complete, but the subspace Q
is not.
In a complete metric space, we also have a property similar to Cantor’s principle on
the real line. First we introduce the following definition.
A sequence of closed balls (Bn )n in a metric space X is called nested if Bn ⊃ Bn+1 for
any n ∈ N and lim rn = 0, where rn is the radius of the ball Bn .
n

. Luong Ha - HUCE
§ 2. Contraction Mapping Principle 51

1.2.2 Theorem (Cantor). In a complete metric space every nested sequence of closed
balls has only one point in common.
Proof. Let (Bn )n be a nested sequence of closed balls, centered at xn with radius
rn in a complete metric space X. For m ≥ n, Bm ⊂ Bn , so that d(xn , xm ) ≤ rn . By
definition, rn → 0, (xn )n is a Cauchy sequence. Since X is complete, xn → x ∈ X. For
any k ∈ N, the elements xk , xk+1 , xk+2 , . . . belong to Bk , and the sequence (xn+k )k is a
subsequence of (xn )n . It follows that lim xn+k = x. Every ball Bk is closed, hence x ∈ Bk
n
T
∞
for all k ∈ N, that is, x ∈ Bn .
n=1
T
∞
Suppose now that there is x0 ∈ Bn . Then x, x0 ∈ Bn for all n ∈ N. Hence,
n=1
0 ≤ d(x, x0 ) ≤ 2 rn for all n ∈ N. Since rn → 0, we must have d(x, x0 ) = 0 or x = x0 .
T
∞
Therefore, Bn consists of only one point.¥
n=1

§2 CONTRACTION MAPPING PRINCIPLE

2.1 CONTRACTION MAPPING

2.1.1 Definition.
Let X be a set and f : X → X be a mapping. A point x ∈ X is called a fixed point
of the mapping f if f (x) = x.
Now we suppose that X is a metric space.
The mapping f : X → X is called a contraction mapping if there is α ∈ [0; 1) such
that d(f (x), f (y) ≤ α d(x, y) for all x, y ∈ X.
The number α is called the contraction constant of f .
Clearly, if f is a contraction mapping, then f is uniformly continuous on X, and hence,
is continuous on X.
2.1.2 Example.
1) Let X = (0; 14 ) ⊂ R and f : X → X defined by f (x) = x2 . Then f is a contraction
mapping.
2) Let X = C[0; 1 ] with the metric ”max”. For each x ∈ X, we define a function f (x)
2
by (f (x))(t) = t(x(t) + 1) for all t ∈ [0; 12 ]. Clearly, f ∈ X. Then the mapping f : X → X
is a contraction mapping.
Indeed, for any x, y ∈ X and t ∈ [0; 12 ], we have

|(f (x))(t) − (f (y))(t)| = |t(x(t) − y(t)|

1 1
≤ |x(t) − y(t)| ≤ d(x, y).
2 2
1
Hence d(f (x), f (y)) ≤ d(x, y) for all x, y ∈ X. Thus, f is a contraction mapping with
2
1
the contraction constant α = .
2

. Luong Ha - HUCE
52 Chapter 3. Complete Metric Spaces

3) The mapping f : [0; 1] → [0; 1] defined by f (x) = 1−x is not a contraction mapping
since |f (x − f (y)| = |x − y| for all x, y ∈ [0; 1].

2.2 CONTRACTION MAPPING PRINCIPLE

2.2.1 Theorem. Let X be a complete metric space and f : X → X be a contraction

mapping. Then f has one and only one fixed point.
This theorem is also called the contraction mapping principle or the fixed point theorem.
Proof. Pick any point xo ∈ X and set x1 = f (xo ), x2 = f (x1 ), ...,
xn = f (xn−1 ), .... Then (xn )n is a Cauchy sequence in X. Indeed, for any two natural
numbers m, n, say n ≤ m,

d(xn , xm ) ≤ d(xn , xn+1 ) + d(xn+1 , xn+2 ) + · · · + d(xm−1 , xm ).

Since f is a contraction mapping, for any n ∈ N,

d(xn , xn+1 ) = d(f (xn−1 ), f (xn )) ≤ α d(xn−1 , xn )

= α d(f (xn−2 ), f (xn−1 )) ≤ α2 d(xn−2 , xn−1 )
≤ · · · ≤ αn d(xo , x1 ).

Consequently,

d(xn , xm ) ≤ (αn + αn+1 + · · · + αm−1 )d(xo , x1 )

αn
≤ d(xo , x1 ).
1−α
αn
Now since 0 ≤ α < 1, we get lim d(xo , x1 ) = 0. Thus,, d(xn , xm ) → 0 as n, m →
n 1−α
∞ or (xn )n is a Cauchy sequence. By hypothesis, X is complete, so that xn → x ∈ X.
From the equality xn+1 = f (xn ) and the fact that f is continuous, it follows that f (x) = x.
This means that the point x is a fixed point of f .
Suppose that x0 also is a fixed point of f . Then

d(x, x0 ) = d(f (x), f (x0 )) ≤ α d(x, x0 ).

Hence (1 − α)d(x, x0 ) ≤ 0 and we obtain d(x, x0 ) = 0, that is, x = x0 . Thus,, the fixed
point of f is unique.
2.2.2 Remark. The Contraction Mapping Principle can be used to prove existence and
uniqueness of theorems for solutions of equations of various types. Besides proving that
an equation of the form f (x) = x has a unique solution, this theorem also gives a practical
method for finding the solution, i.e., calculation of the ”successive approximations” xn . In
fact, in the proof above, the approximation sequence (xn )n does converge to the solution
of the equation f (x) = x. For this reason, this principle (the fixed point theorem) is often
called the method of successive approximations.

. Luong Ha - HUCE
§ 2. Contraction Mapping Principle 53

2.2.3 Example.
1) One of the classical application of the Contraction Mapping Principle is that of
proving the existence and the uniqueness of Cauchy problem for the differential equation
dy
= f (x, y)
dx
with the initial condition y(xo ) = yo .
Concretely, suppose that f (x, y) is defined and continuous on a plane domain G con-
taining the point (xo , yo ), and

|f (x, y) − f (x, y)| ≤ M |y − y|

for all (x, y) and (x, y) in G. Then there is an interval [xo − δ; xo + δ] in which the
differential equation
dy
= f (x, y)
dx
has a unique solution
y = ϕ(x)

satisfying the initial condition

ϕ(xo ) = yo .

The reader may refer the proof of this result, for instance, in [10], pp.71-72.
2) Consider the equation 2 sin x − 4 cos x − 7x = 0 (1)
This equation is equivalent to the following:
1
x= (2 sin x − 4 cos x) (2)
7
1 1
Set f (x) = (2 sin x − 4 cos x), then f 0 (x) = (2 cos x + 4 sin x).
7 7
6
Hence |f 0 (x)| ≤ for all x ∈ R. Let x, y ∈ R be two distinct points. By using
7
Lagrange’s theorem, we obtain

|f (x) − f (y)| = |f 0 (c)| |x − y|,

where c is some point between x and y. Therefore,

6
|f (x) − f (y)| ≤ |x − y|
7
for all x, y ∈ R, and so that f is a contraction mapping. The real line R is complete, f
must have only one fixed point. Thus, the equation (2) has only one root. In other words,
the equation (1) also has a unique real root.
3) Consider the contraction mapping f : X → X, where X = C[0; 1 ] with the metric
2
”max” and (f (x))(t) = t(x(t) + 1) for each x ∈ X.

. Luong Ha - HUCE
54 Chapter 3. Complete Metric Spaces

Choose xo = 0 ∈ X. Then x1 (t) = (f (xo ))(t) = t. Iterating with f yields the sequence
(xn )n , where
1
xn (t) = t = t2 + · · · + tn , ∀ t ∈ [0; ].
2
It is easy to see that
t
lim xn (t) = .
n 1−t
By the fixed point theorem, f has a unique fixed point. That is the function x ∈ X
t 1
defined by x(t) = for all t ∈ [0; ].
1−t 2
π
4) Let f : R → R given by f (x) = + x − arctan x. Then for all x, y ∈ R, x 6= y,
2
|f (x) − f (y)| < |x − y|,

but f does not have any fixed point. Note that in this case α = 1.

§3 COMPLETION

3.1 DEFINITION.

Let (X, d) be a metric space. A complete metric space X ∗ is called a completion of X

if X is isometric to a dense subset of X ∗ .
In this case, the space X is isometric to the dense subspace f (X) of X ∗ . If we think
of X and f (X) as identical, we may considered X as a dense subspace of X ∗ .

3.2 THEOREM.

Every metric space has a unique (up to an iometry) completion.

Proof.
Let (X, d) be a metric space. Fix an element a ∈ X. For each x ∈ X, consider the
function fx : X → R defined by fx (y) = d(x, y) − d(a, y) for all y ∈ X. If follows from
the triangle inequality that |fx (y)| ≤ d(x, a) for all y ∈ X. This means that for each
x ∈ X, the function fx is bounded, that is, fx ∈ B(X) (Problem 2.2, Chapter 1). Let
f : X → B(X) be the mapping defined by f (x) = fx . Recall that the metric on B(X) is
defined as follows
ρ(f, g) = sup |f (x) − g(x)|.
x∈X
For any x, z ∈ X, then
¯ ¯ ¯ ¯
¯fx (y) − fz (y)¯ = ¯d(x, y) − d(a, y) − (d(z, y) − d(a, y))¯
¯ ¯
= ¯d(x, y) − d(z, y)¯ ≤ d(x, z).
¯ ¯
¯ ¯
holds for any y ∈ X. Moreover, ¯fx (z) − fz (z)¯ = d(x, z). Therefore
¯ ¯
¯ ¯
ρ(fx , fz ) = sup ¯fx (y) − fz (y)¯ = d(x, z).
x∈X

. Luong Ha - HUCE
§ 4. Baire Category Theorem 55

Since B(X) is a complete metric space, the subspace f (X) is complete. Hence, the metric
space (f (X), ρ) is a completion of X.
Now suppose that X ∗ and Y ∗ are two completions of the metric space X. We consider
X as a dense subspace of X ∗ and as a dense subspace of Y ∗ . Pick x ∈ X ∗ , then there
exists (xn )n ⊂ X such that xn → x∗ in X ∗ . The sequence (xn )n now is a Cauchy sequence
in Y ∗ , so xn → y ∗ in Y ∗ .
Let ϕ : X ∗ → Y ∗ be the mapping given by ϕ(x∗ ) = y ∗ . We denote the metrics on X ∗
and Y ∗ by d1 and d2 respectively. For any y ∗ ∈ Y ∗ , there is a sequence (xn )n ⊂ X such
that lim d2 (xn , y) = 0. Then (xn )n is a Cauchy sequence in X ∗ , hence xn → x∗ ∈ X ∗ and
n
ϕ(x∗ ) = y ∗ . This means that the mapping ϕ is surjective.
Now pick any two elements x∗ , y ∗ of X ∗ . Then there are two sequences (xn )n , (yn )n in
X so that lim d1 (xn , x∗ ) = lim d1 (yn , y ∗ ).
n n
Consequently, we get
¡ ¢
d1 (x∗ , y ∗ ) = lim d1 (xn , yn ) = lim d2 (xn , yn ) = d2 ϕ(x∗ ), ϕ(y ∗ ) .
n n
Hence, ϕ is an isometry from X ∗ onto Y ∗ . The proof is now complete.
Example.
1) We have known that the subspace Q of R is not complete. The completion of Q is
precisely R since Q is dense in R and R is complete.
2) The real line R equipped with d(x, y) = |ex − ey | is isometric to the subspace
X = (0; +∞) of R with the usual metric. The completion of X is X ∗ = [0; +∞). Hence,
the completion of (R, d) is X ∗ = [0; +∞) with the usual metric.

§4 BAIRE CATEGORY THEOREM

We know that a subset A of X is dense in X, if the closure of A is X, that is A is dense

in every nonempty open subsets of X. Here we consider an extremely contrary case.

4.1 NOWHERE DENSE SETS

4.1.1 Definition. A subset A of a metric space X is called nowhere dense if A is not

dense in any nonempty open subset of X. This means that the closure of A has empty
interior, or int(A) = ∅.
Example. The subsets N, Z of R are nowhere dense in R, and every finite subset of
R is nowhere dense.
Since X \ int (A) = X \ A, it should be noted that A is nowhere dense in X if and
only if X \ A is dense in X.
4.1.2 Theorem. Let X be a metric space and A ⊂ X. Then A is nowhere dense if and
only if for any nonempty open set V ⊂ X, there exists a nonempty open set V1 such that

. Luong Ha - HUCE
56 Chapter 3. Complete Metric Spaces

V1 ⊂ V and V1 ∩ A = ∅.
Proof. Suppose that A is nowhere dense in X and V is a nonempty open set V of
X. Since V 6⊂ A, there is an element x ∈ V so that x ∈ / A. Hence, there exists an open
neighbourhood V ∗ of x such that V ∗ ∩ A = ∅.
Set V1 = V ∩ V ∗ . It is clear that V1 is a nonempty open set, and V1 ⊂ V, V1 ∩ A = ∅.
Conversely, suppose that there is an open set V 6= ∅ such that A ⊃ V . Then V1 ∩A 6= ∅
for any open set V1 ⊂ V, V1 6= ∅. This contradicts the hypothesis. Thus, A is nowhere
dense in X.
4.1.3 Corollary. Let A ⊂ (X, d). Then A is nowhere dense if and only if for any ball
B ⊂ X, there is a ball B1 ⊂ B such that B1 ∩ A = ∅.

4.2 CATEGORY

4.2.1 Definition. A subset A of a metric space X is called to be of the first category

if A is the union of a countable family of nowhere dense subsets of X. The subset A is
called to be of the second category if it is not of the first category.
4.2.2 Theorem. Each complete metric space is of the second category.
S
∞
Proof. Let X = An , where each set An is nowhere dense. Since A1 is nowhere
n=1
dense, there is a closed ball B1 , radius r1 < 1 such that B1 ∩ A1 = ∅. The set A2 is
1
nowhere dense, the ball B1 contains a closed ball B2 , radius r2 < such that B2 ∩ A2 = ∅.
2
Proceeding this process by induction, we contruct a sequence of closed balls (Bn )n in X,
1
radius rn < satisfying the followings:
n
i) Bn ∩ An = ∅ for each n ∈ N;
ii) Bn ⊃ Bn+1 for all n ∈ N;
1
iii) rn < for each n ∈ N.
n
Hence, the sequence of closed balls (Bn )n is nested. Since X is complete, there exists an
T
∞
element x ∈ Bn . Then by iii), x ∈
/ An for all n ∈ N. This contradicts the assumption
n=1
S
∞
X= An . Therefore, X is of the second category.
n=1
This theorem is often called Baire Category Theorem. Even the following corollary is
often called Baire Category Theorem.
S
∞
4.2.3 Corollary. Let X be a complete metric spacel and let X = An . Then there
n=1
exists no ∈ N such that the set Ano is not nowhere dense, that is, Ano contains some ball.

. Luong Ha - HUCE
§ 4. Baire Category Theorem 57

PROBLEMS
3.1 Prove that the metric space B(X) is complete.

(Recall that ρ(f, g) = sup |f (x) − g(x)|)

x∈X

3.2 For each natural number n define fn by



 1
1
 if 0 ≤ t < 2
fn (t) = 1 − 2n (t − 12 ) if 1
≤t< 1
+ 2−n


2 2

0 1
if 2
+ 2−n ≤ t ≤ 1.

L
Prove that (fn )n ⊂ C[0;1] , and in this metric space, the sequence (fn )n is a Cauchy
sequence which does not converge.
3.3. Let
hX
k i 21
d(x, y) = |xi − yi |2
i=1

for all x = (x1 , x2 , ..., xk ) and y = (y1 , y2 , ..., yk ) of Ck . Prove that d is a metric on Ck ,
and the space (Ck , d) is complete.
3.4 Consider two metrics on the set R

d1 (x, y) = | arctan x − arctan y|; d2 (x, y) = |ex − ey |.

Sho that R with each metric is not complete.

3.5 Let X, Y be two isometric metric spaces. Show that if X is complete, then Y is
complete.
3.6 Prove that the metric spaces (Ck , d) (Problem 2.3) and (R2k , d) are isometric.
3.7 Let X be the set of functions continuous on [−1; 1] equipped with the metric

n Z1 o 21
2
d(f, g) = (f (x) − g(x)) dx .
−1

Prove that (X, d) is not complete.

3.8 Consider two subsets of the metric space C[−a;a]
M = {x ∈ C[−a;a] : x is even },
N = {x ∈ C[−a;a] : x is odd }.
Prove that the subspaces M and N are complete.
3.9 Let `∞ be the set of bounded real sequences. For x = (xn )n , y = (yn )n in `∞ , define
¯ ¯
d(x, y) = sup¯xn − yn ¯.
x∈N

1) Prove that the space (`∞ , d) is complete.

. Luong Ha - HUCE
58 Chapter 3. Complete Metric Spaces

2) Let c be the subspace of `∞ consisting of convergent sequences. Prove that c is

complete.
3) Let so be the subspace of `∞ consisting of the sequences x = (xn )n such that all
xn = 0 except a finitely many n ∈ N. Prove that so is not complete.
3.10 Let (X, d) be a metric space. Consider the metrics on X defined as follows:

d(x, y)
d1 (x, y) =
1 + d(x, y)
d2 (x, y) = ln[1 + d(x, y)]
d3 (x, y) = min(1, d(x, y)).

Prove that (X, d) is complete if and only if each (X, di ), i = 1, 2, 3 is complete.

3.11 On the set of natural numbers N, set

1 + 1 if m 6= n
d(m, n) = m+n
0 if m = n.

1) Prove that d is a metric on N.

2) Prove that the metric space (N, d) is complete.
1
3) Let Bn be the closed ball centered at n, and radius rn = 1 + . Prove that
2n
T
∞
Bn = ∅. Does this result contradict Cantor’s theorem?
n=1
3.12 Let X be the real line with the usual metric and Y be the real line with the metric
d defined by ¯ x
¯ y ¯¯
d(x, y) = ¯ − ¯.
1 + |x| 1 + |y|
Show that X and Y are homeomorphic, but Y is not complete.
3.13 Consider the subspace X = [1; +∞) of R and the mapping f : X → X defined by
1
f (x) = x + . Prove that
x ¯ ¯ ¯ ¯
¯f (x) − f (y)¯ < ¯x − y ¯

for all x, y ∈ X, x 6= y, but f does not have any fixed point.

3.14 Let X = {x ∈ Q : x ≥ 1} with the usual metric and let f : X → Xbe given by
x 1
f (x) = + .
2 x
Show that f is a contraction mapping and that f has no fixed point in X.
3.15 Let X = [1; 2] ∩ Q with the usual metric and let f : X → X be the mapping defined
by
1
f (x) = − (x2 − 2) + x.
4
Show that f is a contraction mapping and that f has no fixed point in X.

. Luong Ha - HUCE
§ 4. Baire Category Theorem 59

3.16 Let X = [1; +∞) ⊂ R and let f : X → X be defined by

25 1
f (x) = (x + ).
26 x
Prove that f is a contraction mapping and find the fixed point of f .
3.17 Let f : R2 → R2 be defined by f (x, y) = ( 12 cos y, 12 sin x + 1). Show that f has only
one fixed point.
3.18 Let X = C[0;1] . Define F : X → X by

Zx
(F (f ))(x) = f (t) dt (f ∈ X).
0

Show that
1) (F (f ))(x) − (F (g))(x) ≤ x d(f, g),
x2
2) (F 2 (f ))(x) − (F 2 (g))(x) ≤
d(f, g), for all f, g ∈ X, all x ∈ [0; 1], and deduce
2
that F 2 = F ◦ F is a contraction mapping. Show, however, that F is not a contraction
mapping.
3.19 Let f : [a; b] → [a; b] be a differentiable function. Prove that f is a contraction
mapping if and only if there exists a number K ∈ [0; 1) such that |f 0 (x)| ≤ K for all
x ∈ (a; b).
3.20 Let X be a complete metric space and f : X → X is a contractrion mapping with
the contraction constant α. Suppose that there is an element a ∈ X and a number r > 0
such that d(f (a), a) < r(1 − α). Show that the unique fixed point of f belongs to the
open ball B(a, r).
3.21 Let X be a complete metric space and f : X → X is a surjection. Suppose that
there is k > 1 such that d(f (x), f (y)) = kd(x, y) for all x, y ∈ X. Prove that f has only
one fixed point.
3.22 Let X be a complete metric space and f : X → X is a mapping. Suppose that there
is n ∈ N such that f n is a contraction mapping. Prove that f has only one fixed point.
3.23 Let A be a nonempty subset of a metric space X. We define the diameter of A as
the number sup d(x, y), denoted by d(A).
x,y∈A
Prove that if X is complete, then for any sequence of nonempty closed subsets (An )n
T
∞
of X satisfying An ⊃ An+1 for all n ∈ N and lim d(An ) = 0, then An contains a unique
n n=1
point.
3.24 Prove that in a complete metric space X, the intersection of a sequence of dense
open subsets of X is also dense in X.
3.25 Prove that the product of two complete metric spaces is complete.
Q
3.26 Let (Xn , dn )n be a sequence of metric spaces, and let X = ∞ n=1 Xn . For each

. Luong Ha - HUCE
60 Chapter 3. Complete Metric Spaces

x = (xn )n and y = (yn )n in X, define

X∞
1 dn (xn , yn )
d(x, y) = n
· ·
n=1
2 1 + dn (xn , yn )

1) Prove that d is a metric on X.

2) Prove that (X, d) is complete if and only if each (Xn , dn ) is complete.
3.27 Show that the boundary of a closed or an open set in a metric space is nowhere
dense. Is this statement true for an arbitrary set?
3.28 Let f : R → R be a continuous function. Prove that the graph of f is nowhere dense
in R2 .
3.29 Let F be a family of continuous functions from R into R. Suppose that for each
x ∈ R, there exists Mx > 0 such that |f (x)| ≤ Mx for all f ∈ F.
Prove that there are a nonempty open set U ⊂ R and a number M > 0 such that
|f (x)| ≤ M for all x ∈ U and for all f ∈ F .
3.30 Show that the subset Q of R cannot be the intersection of a countable family of
open sets.
3.31 Let X, Y be metric spaces and f : X → Y is a surjection. Suppose that there are
two positive numbers M, m such that for any x, y ∈ X

m d(x, y) ≤ d(f (x), f (y)) ≤ M d(x, y).

1. Show that f is a homeomorphism.

2. Show that X is complete if and only if Y is complete.

HINTS and SOLUTIONS

3.1 Let (fn )n be a Cauchy sequence in B(X). Then for each ε > 0, there exists no ∈ N
such that d(fn , fm ) < ε for all m, n ≥ no . Hence, for each x ∈ X, |fn (x) − fm (x)| ≤
d(fn , fm ). It follows that (fn (x))n is a Cauchy sequence in R, so (fn (x))n is convergent.
Let f (x) = lim fn (x). From the inequality |fn (x)−fm (x)| < ε for all x ∈ X and m, n ≥ no ,
n
we get |fn (x) − f (x)| ≤ ε for all x ∈ X and n ≥ no .
Now we can show that f ∈ B(X), and then d(fn , f ) ≤ ε for all n ≥ no . Thus,
lim fn = f in B(X). Consequently, B(X) is a complete space.
n
3.4 Hints.
π 1
∗ Let xn = tan( − ). Show that (xn )n is a Cauchy sequence in (R, d1 ), but it does
2 n
not convereges.
1
∗ Let xn = ln . Show that (xn )n is a Cauchy sequence in (R, d2 ), but it does not
n
convereges.
3.5 Let f : X → Y be an isometry. Let (yn )n be any Cauchy sequence in Y . Then there

. Luong Ha - HUCE
§ 4. Baire Category Theorem 61

exists a sequence (xn )n in X such that yn = f (xn ) for each n ∈ N. Since f is isometric,
for any m, n, we have
d(xn , xm ) = d(f (xn ), f (xm )) = d(yn , ym ).
This shows that (xn )n is a Cauchy sequence in X. Hence, xn → x ∈ X. By continuity, it
follows that f (xn ) → f (x), or yn → f (x) ∈ Y . Thus,, Y is a complete space.
3.7 Hints.
We find a Cauchy sequence in X which is not convergent. Consider the sequence
(fn ) ⊂ X to be defined as follows:


 0 if −1 ≤ x ≤ 0

 1
fn (x) = nx if 0 ≤ x ≤

 n

1 1
if ≤ x ≤ 1.
n
Show that (fn )n is a Cauchy sequence in X.
If there is f ∈ X so that fn → f in X, then

0 if −1 ≤ x ≤ 0
f (x) =
1 if 0 < x ≤ 1

which is impossible since f ∈

/ X.
3.8 Hints.
Show that M and N are two nonempty closed sets in C[−a;a] . It should be noted that
if xn → x in C[−a;a] , then xn (t) → x(t) for any t ∈ [−a; a].
3.9
¡ (n) (n) ¢
1) Let (xn )n be any Cauchy sequence `∞ , and let xn = x1 , x2 , . . . . Each element
(n)
xn is a bounded sequence, there exists a positive number Kn such that |xi | ≤ Kn for all
i ∈ N. On the other hand, (xn )n is a Cauchy sequence, for each ε > 0, there is no ∈ N
(n) (k) (n)
such that |xi − xi | < ε for all i ∈ N, and all n, k ≥ no . Hence, the sequence (xi )n is
(n)
a Cauchy sequence for each i ∈ N. It follows that xi → xoi ∈ R for each i ∈ N.
(n) (k) (n)
Since |xi − xi | < ε for all n, k ≥ no , letting k → ∞ yields |xi − xoi | < ε for all
(n ) (n )
n ≥ no , and each i ∈ N. Now |xoi | = |xi o − xoi | + |xi o | < ε + Kno for all i ∈ N, hence, the
(n
element xo = (xoi )i∈N belongs to `∞ . Furthermore, sup |xi ) − xoi | ≤ ε whenever n ≥ no ,
i∈N
and this shows that lim d(xn , xo ) = 0. Thus, `∞ is complete.
n
2) It suffices to show that c is a closed subspace of `∞ .
3) Find a sequence in so which converges to an element not belonging so , that is, so is
not closed in `∞ .
3.11 Hints.
2) Let (kn )n ⊂ (N, d) be any Cauchy sequence. Then for ε = 1, there exists no ∈ N
such that d(kn , km ) < 1 for all m, n ≥ no . Hence, d(kn , km ) = 0 with m, n ≥ no . Thus,,
kn = kno for all n ≥ no , and hence kn → kno . This means that (N, d) is a complete space.

. Luong Ha - HUCE
62 Chapter 3. Complete Metric Spaces

3) By definition,
k ∈ Bn ⇔ d(k, n) ≤ rn ⇔ k ≥ n.
T
∞
Hence, Bn = {n, n + 1, . . . }. We get Bn = ∅.
n=1
We observe that (Bn )n is a decreasing sequence of closed balls in the space (N, d),
and lim rn = 1. Thus, it is not a nested sequence. Therefore, the above result does not
n
contradict Cantor’s theorem.
π
3.13 First we observe that arctan x 6= for any x ∈ R. Hence,
2
π
f (x) = + x − arctan x 6= x
2
for all x ∈ R. This means that f has no fixed point.
Let x, y be distinct elements of R. By Lagrange’s theorem, there is c between x and
y such that
|f (x) − f (y)| = |x − y| · |f 0 (c)|.

On the other hand,

1 x2
f 0 (x) = 1 − = ,
1 + x2 1 + x2
c2
so |f (x) − f (y)| = |x − y| < |x − y|.
1 + c2
3.16 Hints.
25
Show that |f (x) − f (y)| ≤ d(x, y) for any x, y ∈ X. Hence, f is a contraction
26
mapping. X is complete, f has a unique fixed point. Solve the equation f (x) = x, we get
x = 5.
3.20. Hints.
Show that f (x) ∈ B(a, r) for any x ∈ B 0 (a, r). Hence f maps x ∈ B 0 (a, r) into
x ∈ B 0 (a, r). The subspace B 0 (a, r) is complete, hence the unique fixed point of f belongs
to B(a, r).
3.21 Hints.
∗ Show that f is injective, hence, f is bijective.
∗ Show that the inverse f −1 is a contraction mapping. Then f −1 has a unique fixed
point, say u.
∗ Show that u is the unique fixed point of f .
3.22 Let α be the contraction constant of f n . By definition,

d(f n (x), f n (y)) ≤ α d(x, y)

for all x, y ∈ X.
Let xo the unique fixed point of f n . We have

d(f (xo ), xo ) = d(f n+1 (xo ), f n (xo ) = d(f n (f (xo ), f n (xo )) ≤ αd(f (xo ), xo ).

. Luong Ha - HUCE
§ 4. Baire Category Theorem 63

Thus,, f (xo ) = xo . It is easy to see that each fixed point of f is also a fixed point of f n ,
hence, xo is the unique fixed point of f .
3.23 Hints.
Choose xn ∈ An for each n ∈ N. Then (xn )n is a Cauchy sequence in X. By hypothesis,
xn → x ∈ X. Making similar arguments as in the proof of Cantor’s theorem, x is the
only point in common of the sequence (An )n .
3.24 Let X be a complete space and let (Gn )n be a sequence of dense open subsets of X.
T∞
We shall show that G = Gn is dense in X.
n=1
Consider x ∈ X and ε > 0. Since G1 is dense in X, there exists x1 ∈ G1 such that
ε ε
d(x, x1 ) < . The set G1 is open, we can find r1 such that 0 < r1 < and B 0 (x1 , r1 ) ⊂ G1 .
2 2
The set G2 is dense in X, B(x1 , r1 )∩ G2 6= ∅. Pick x2 ∈ B(x1 , r1 )∩ G2 , we can find r2 with
1
0 < r2 < and B 0 (x2 , r2 ) ⊂ B(x1 , r1 )∩G2 . Proceeding by this way, we oobtain a sequence
2
¡ ¢ 1
of closed balls B 0 (xn , rn ) n so that 0 < rn < and B 0 (xn+1 , rn+1 ) ⊂ B(xn , rn ) ∩ Gn+1
n
for each n ∈ N.
¡ ¢
Thus,, the sequence of closed balls B 0 (xn , rn ) n is nested, and X is complete, hence,
T∞
there exists an element a ∈ B 0 (xn , rn ). Since B 0 (xn , rn ) ⊂ Gn for each n ∈ N, there
n=1
T
∞ T
∞
exists a ∈ Gn . We have d(x, a) ≤ d(x, x1 ) + d(x1 , a) < ε. Therefore, G = Gn is
n=1 n=1
dense in X.
3.25 Let X and Y be complete metric spaces, and let (zn )n be any Cauchy sequence in
the product space X × Y . For zn = (xn , yn ) ∈ X × Y , we have

d(zn , zm ) = d(xn , xm ) + d(yn , ym ),

hence, d(xn , xm ) ≤ d(zn , zm ), and d(yn , ym ) ≤ d(zn , zm ). It follows that (xn )n and (yn )n
are Cauchy sequences in X and Y , respectively. By completeness, xn → xo ∈ X and
yn → yo ∈ Y .
Set zo = (xo , yo ). Clearly, zo ∈ X × Y , and

d(zn , zo ) = d(xn , xo ) + d(yn , yo ).

Hence, lim d(zn , zo ) = 0, that is, zn → zo .

n
Thus, X × Y is complete. The proof of the converse is left for the reader.
3.28 Let G = {(x, f (x)) : x ∈ R}. The function f is continuous, so G is closed, and hence,
G = G. Suppose that int(G) 6= ∅, that is, there is a point zo = (xo , f (xo )) ∈ int(G). Then
r r
there exists r > 0 so that B(zo , r) ⊂ G. Pick z1 = (xo , f (xo ) + ), then d(z1 , zo ) = ,
2 2
r
hence, z1 ∈ G. This implies that f (xo ) = f (xo ) + with r > 0, which is impossible.
2
Therefore, int(G) = ∅ or G is nowhere dense.
3.29 For each n ∈ N and f ∈ F, let Anf = {x ∈ R : |f (x)| ≤ n}. Since f is continuous,
T
each Anf is closed, hence, An = Anf is closed. For each x ∈ R, there exists by
f ∈F

. Luong Ha - HUCE
64 Chapter 3. Complete Metric Spaces

hypothesis Mx > 0 so that |f (x)| ≤ Mx for all f ∈ F. We can find no ∈ N with Mx ≤ no ,

S
∞
hence, |f (x)| ≤ no for all f ∈ F. Thus, x ∈ Ano , and R = An . By Baire theorem, there
n=1
must be k ∈ N such that Ak contains an open interval U . Since Ak is closed, U ⊂ Ak . On
the set Ak we have |f (x)| ≤ k for all f ∈ F .
Thus,, |f (x)| ≤ k for all x ∈ U and all f ∈ F .

. Luong Ha - HUCE
Chapter 4

Compact Metric Spaces

One of the important properties of the segment [a; b] ⊂ R is that every sequence in [a; b]
has a subsequence which converges to a point of [a; b]. In addition, a real function which
is continuous on this interval, attains its least value and greatest value. In this chapter,
we concern about a class of subsets of a metric space possessing these properties.

§1 COMPACT SETS - COMPACT SPACES

1.1 BOUNDED SETS - TOTALLY BOUNDED SETS

1.1.1 Definition. Let A be a subset of a metric space X.

i) We say that A is bounded if A is contained in some ball of X.
Note that since B(x, r) ⊂ B 0 (x, r) ⊂ B(x, 2r), in the definition above, we can use the
open ball or closed ball.
ii) The set A is called totally bounded if for every ε > 0, A is contained in the union
of a finite number of balls with radius ε. From the definition, it follows that
1.1.2 Proposition.
1) The set A is bounded if and only if its diameter d(A) = sup d(x, y) is finite.
x,y∈A
2) Any subset of a bounded set (totally bounded set) is bounded (totally bounded).
3) The closure of a bounded set (totally bounded set) is bounded (totally bounded).
4) The union of a finite family of bounded (totally bounded) sets is bounded (totally
bounded).
5) Every totally bounded set is bounded.
For example, we prove 4). Let A1 , A2 , ..., An be bounded subsets of a metric space
Sn
X. Set A = Ai . Since each Ai is bounded, there are closed balls Bi0 (xi , ri ) such that
i=1
Ai ⊂ B 0 (xi , ri ) for each i = 1, 2, ..., n. Pick x ∈ A, there is io ∈ {1, 2, ..., n} such that

65
66 Chapter 4. Compact Metric Spaces

x ∈ B 0 (xio , rio ). It follows that

d(x, x1 ) ≤ d(x, xio ) + d(xio , x1 )

≤ rio + d(xio , x1 ) ≤ max ri + max d(xi , x1 ).
1≤i≤n 1≤i≤n

Let r = max ri + max d(xi , x1 ). Then A ⊂ B 0 (x1 , r), and so A is bounded.

1≤i≤n 1≤i≤n
Moreover, suppose that A1 , A2 , ..., An are totally bounded. Then for ε = 1, A is
contained in the union of a finite number of balls with radius 1. Since this union is
bounded, it follows that A is bounded.
In general, the converse of last statement does not hold. But the following statement
is true.
1.1.3 Theorem. On the Euclidean space Rk , any bounded set is totally bounded.
Thus, the boundedness and the total boundedness are equivalent on Rk equipped with
the usual metric.
Proof. ([3], p. 138) Let A ⊂ Rk be a bounded set. Then there is a positive N such
that
A ⊂ AN = {(ξ1 , ξ2 , ..., ξk ) ∈ Rk : |ξj | ≤ N
for j = 1, 2, ..., k}.
√
Let ε > 0, let m be the least positive integer with mε ≥ n and let Em be the set of
p1 p2 pk
k-tuples ( , , ..., ) where pj is an integer with |pj | ≤ mN for j = 1, 2, ..., k. There
m m m
are exactly (1 + 2mN )k distinct points in the set Em . Pick u = (ξ1 , ξ2 , ..., ξk ) ∈ AN . Then
for j = 1, 2, ..., k, we have |ξj | ≤ N and hence, we can find an integer pj with |pj | ≤ mN
and |mξj − pj | < 1. Since

p1 p2 pk ³X k
pj 2 ´1/2
d(u, ( , , ..., )) = (ξj − )
m m m j=1
m
√
n
< ≤ ε.
m
Thus, AN is covered by a finite number of closed balls of centers in Em and of radius ε.
It follows that AN is totally bounded and hence, A is totally bounded.

1.2 COMPACT SETS - COMPACT SPACES

1.2.1 Definition. Let A be a subset of a metric space X.

The set A is called compact if every sequence in A has subsequence converging to an
element of A.
X is called a compact metric space if the set X is compact.
1.2.2 Example.
1) Every closed interval [a; b] ⊂ R is a compact set. More generally, every closed
bounded subset of R is compact.

. Luong Ha - HUCE
§ 1. Compact Sets - Compact Spaces 67

This directly follows from the Bolzano-Weierstrass principle and the closedness of the
set in view.
2) The real line R is not compact.
Indeed, let (xn )n ⊂ R with xn = n. Then |xn − xm | ≥ 1 if n 6= m. Hence each
subsequence of (xn )n cannot be a Cauchy sequence, and so it is not convergent.
3) Every finite subset of a metric space is compact.
Let A = {a1 , a2 , ..., ak } ⊂ X and (xn )n ⊂ A. There exists io ∈ {1, 2, ..., k} such that
xn = aio for infinitely many n ∈ N. Hence we can choose a subsequence (xnk )k of (xn )n
in order to xnk = xio for all k ∈ N. It is clear that xnk → xio ∈ A. Therefore, the set A is
compact.
1.2.3 Remark.
1) A nonempty subset A of X is compact if and only if the subapce A is a compact
space.
2) In the above definition, we only need to consider an arbitrary infinite sequence in
A which is pairwise distinct.
1.2.4 Proposition. Let A and B be two subsets of a metric space X and A ⊂ B. If B
is compact and A is closed, then A is compact.
Proof. Let (xn )n ⊂ A. Then (xn )n ⊂ B. Since B is compact, there exists a sub-
sequence (xnk )k of (xn )n such that xnk → xo ∈ B. But (xnk )k ⊂ A and A is closed, it
follows that xo ∈ A. Hence, the set A is compact.
1.2.5 Hausdorff ’s Theorem. In any metric space X, the following statements hold:
1) Every compact subset is closed and totally bounded.
2) If X is complete, then every closed and totally bounded subset of X is compact.
Proof.
1) Let A ⊂ X be a compact set. Consider any sequence (xn )n ⊂ A which converges
to some point x ∈ X. By hypothesis, there is a subsequence (xnk )k of (xn )n such that
xnk → x0 ∈ A. Since (xnk )k also converges to x, it follows that x = x0 . Hence, x ∈ A, and
the set A is closed.
Suppose that A is not totally bounded. By definition, there exists ε > 0 such that
any finite family of balls of radius ε cannot cover A. Pick x1 ∈ A, then A 6⊂ B(x1 , ε).
So there is x2 ∈ A such that d(x1 , x2 ) ≥ ε. Two balls B(x1 , ε), B(x2 , ε) cannot cover A,
there is x3 ∈ A satisfying d(x1 , x3 ) ≥ ε and d(x2 , x3 ) ≥ ε. Proceeding this process, we
obtain a sequence (xn )n ⊂ A such that d(xn , xm ) ≥ ε whenever n 6= m. It is evident
that any subsequence of (xn )n is not a Cauchy sequence, hence, it is not convergent. This
contradicts the compactness of A. Consequently, A is totally bounded.
2) Now suppose that X is complete and A ⊂ X is closed and totally bounded. Let
(xn )n ⊂ A. Then A is covered by a finite number of balls of radius 1, there must be a
ball, denoted by B1 , which contains infinitely many terms of the sequence (xn )n . So A

. Luong Ha - HUCE
68 Chapter 4. Compact Metric Spaces

will contain a subsequence (x1n )n of (xn )n extracted from these terms.

1
The set A is also covered by a finite number of balls of radius , there must be a ball,
2
denoted by B2 , which contains infinitely many terms of the sequence (x1n )n .
Let (x2n )n be the subsequence of (x1n )n contained in B2 . By the mathematical induction,
there exist the subsequences (x1n )n , (x2n )n , ... of the sequence (xn )n such that for each
k ∈ N, the sequence (xk+1 k k
n )n is a subsequence of (xn )n . Moreover, (xn )n ⊂ Bk (the ball
1
of radius ). Pick an element xn1 of the sequence (x1n )n . From infinitely many terms of
k
(x2n )n , choose xn2 with n2 > n1 . Next we can choose xn3 belonging to (x3n )n with n3 > n2 .
Hence, by proceeding this process, we get a subsequence (xnk )k of (xn )n .
Since (xkn )n ⊂ (xln )n for any k ≥ l, the elements xnk and xnl belong to the ball Bl .
2
It follows that d(xnk , xnl ) ≤ . This shows that (xnk )k is a Cauchy sequence. By the
l
completeness of X, xnk → xo ∈ X. But A is closed, we must have xo ∈ A. This means
that A is compact and the proof of the theorem is now complete.
The following important proposition can be easily derived from Hausdorff’s theorem.
1.2.6 Corollary. On the space Rk with the usual metric, a subset A is compact if and
only if A is closed and bounded.
Example.
1) The space R is not compact since it is not bounded.
2) Let A = {(x, y) ∈ R2 : x2 + 4y 2 ≤ 9}.
If (x, y) ∈ A, then x2 +y 2 ≤ 9. It follows that A is contained in the closed ball of center
O and radius 3. Hence A is a bounded set. Let (xn , yn )n ⊂ A and (xn , yn ) → (xo , yo ).
Then xn2 → x2o and yn2 → yo2 . In addition, x2n + 4yn2 ≤ 9 for all n ∈ N, we get x2o + 4yo2 ≤ 9.
This means that (xo , yo ) ∈ A, or A is closed. By Corollary 1.2.6, A is compact.
1.2.7 Theorem. A metric space is compact if and only if it is complete and totally
bounded.
Proof. By Theorem 4.2.4, if X is a compact metric space, then X is totally bounded.
Let (xn )n be an arbitrary Cauchy sequence in X. Then there is a subsequence (xnk )k of
(xn )n such that xnk → xo . By the property of a Cauchy sequence, xn → xo . Hence X is
complete.
The converse is clear by Theorem 4.2.4.
1.2.8 Theorem. Every compact metric space is separable.
Proof. Let X be a compact metric space. Then X is totally bounded. Hence, for
1
each n ∈ N, the set X is covered by a finite number of open balls of radius , say
n
S
kn 1
X= B(xnj , ).
j=1 n
Let A = {xnj : j = 1, 2, ..., kn , n ∈ N}. Then A is at most countable. For any x ∈ X
1 1
and ε > 0, there is no ∈ N so that < ε. With the integer no , there is a ball B(xnj , )
no no

. Luong Ha - HUCE
§ 2. Heine-Borel’s Theorem 69

1
containing x, that is, d(x, xnj ) < < ε. Thus, the set A is dense in X. The proof is
no
complete.¥
From the proof above, we note that if X is a totally bounded metric space, then X is
separable.

§2 HEINE-BOREL’S THEOREM

2.1 COVER - SUBCOVER

2.1.1 Definition. Let A be a subset of a metric space X. A family (Gα )α∈I of subsets
S
of X is called a cover of A (or to cover A) if A ⊂ Gα .
α∈I
A subfamily of a cover of A which is a cover of A is called a subcover of A.
A finite family of subsets of X which is a cover of A is called a finite cover of A. A
cover of A each member of which is an open subset of X is called an open cover of A.
2.1.2 Example.
1) In any metric space X, for each r > 0 the family of open balls {B(x, r) : x ∈ X}
is an open cover of X.
2) The family of all open intervals (−n; n) with n ∈ N is an open cover of R.

2.2 HEINE-BOREL’S THEOREM.

2.2.1 Theorem. A metric space X is compact if and only if every open cover of X has
a finite subcover.
Proof.([3], pp.136-137) Let X be a compact metric space. Suppose that there is an
open cover (Gα )α∈I of X which has no finite subcover. We shall define inductively a
sequence (xn )n in X with the following properties: for n=1, 2, ...
1
(1) d(xn , xn+1 ) ≤ n−1 ,
2
1
(2) the open ball B(xn , n−1 ) cannot be covered by any finite subfamily of (Gα )α∈I .
2
S
m
By hypothesis, there is a finite number of elements y1 , y2 , ..., ym so that X = B(yj , 1).
j=1
Then there is at least one integer j, with 1 ≤ j ≤ m, such that B(yj , 1) cannot be covered
by a finite subfamily of (Gα )α∈I . Let x1 = yj , where j is the least such integer. Then x1
satisfies conditions (1) and (2).
Now we suppose that the points x1 , x2 , ..., xn in X have been chosen to satisfy condi-
1
tions (1) and (2). Since X is totally bounded, the open ball B(xn , n−1 ) also is totally
2
1
bounded. Hence, there is a finite number of open balls of radius n and centers, say
2
1 1 1
y1 , y2 , ..., ym in B(xn , n−1 ) , which cover the ball B(xn , n−1 ). Since B(xn , n−1 ) satisfies
2 2 2

. Luong Ha - HUCE
70 Chapter 4. Compact Metric Spaces

1
condition (2), there is at least one integer j, with 1 ≤ j ≤ m, such that B(yj , n ) cannot
2
be covered by a finite subfamily of (Gα )α∈I ; let xn+1 = yj , where j is the least such in-
1
teger. Since xn+1 ∈ B(xn , n−1 ), conditions (1) and (2) are fulfilled. This completes the
2
definition of the sequence (xn )n .
If m and n are two positive integers with m > n, then

d(xn , xm ) ≤ d(xn , xn+1 ) + d(xn+1 , xn+2 ) + · · · + d(xm−1 , xm )

1 1 1
≤ n−1 + n−2 + · · · + m−2
2 2 2
1
≤ n−2 .
2

Consequently, (xn )n is a Cauchy sequence and converges because X is complete. Let

x = lim xn . Then there exists α ∈ I such that x ∈ Gα . Since Gα is open, there is
n
r 1 r
B(x, r) ⊂ Gα for some r > 0. There is an integer k such that d(xn , x) < and n−1 <
2 2 2
1
for all n ≥ k. If y ∈ B(xk , k−1 ), then
2
1 r
d(y, x) ≤ d(y, xk ) + d(xk , x) < k−1 + < r.
2 2
1
Thus, B(xk , k−1 ) ⊂ B(x, r) ⊂ Gα which contradicts condition (2). In short, every
2
open cover of X does have a finite subcover.
Conversely, suppose that every open cover of X has a finite subcover but X is not
compact. By the definition of the compactness, there exists an infinite sequence (xn )n in
X which is pairwise distinct and has no convergent subsequence. Hence, every x ∈ X
cannot be the limit of any subsequence of (xn )n . Therefore, there is an open ball B(x, rx )
containing only a finite number of elements of (xn )n . Clearly, the family B(x, rx )x∈X is
an open cover of X, by hypothesis, X is covered by finitely many balls B(xi , rxi ), i =
k
1, 2, ..., k. Hence, (xn )n ⊂ ∪ B(xi , rxi ). This is a contradiction because each of these
i=1
open balls contains only a finite number of elements of (xn )n .
2.2.2 Example.
1) The open cover {(−n; n) : n ∈ N} of R cannot have any finite subcover, therefore,
R is not compact.
2) Let A1 , A2 be two compact subsets of X. Pick any open cover (Gα )α∈I of the set
A1 ∪ A2 . It is evident that the family (Gα )α∈I covers A1 and covers A2 . Hence, there are
S S
two finite subsets J1 , J2 of I such that A1 ⊂ Gα , and A2 ⊂ Gα . Consequently,
S α∈J1 α∈J2
A1 ∪ A2 ⊂ Gα , where J = J1 ∪ J2 being a finite subset of I. Thus, A1 ∪ A2 is compact.
α∈J
Note that this result holds for a finite number of compact subsets of X.

. Luong Ha - HUCE
§ 3. Compact Spaces and Continuous Mappings 71

§3 COMPACT SPACES AND CONTINUOUS MAPPINGS

3.1 Theorem. Let X and Y be metric spaces and let f : X → Y be a continuous

mapping. If A ⊂ X is compact, then f (A) is compact in Y .
Proof. Let (yn )n be an arbitrary sequence in f (A). Then there exists a sequence (xn )n
in A such that yn = f (xn ) for each n ∈ N. Since A is compact, there is a subsequence
(xnk )k of (xn )n so that xnk → xo ∈ A. The continuity of the mapping f at the point xo
implies f (xnk ) → f (xo ). Let yo = f (xo ), then yo ∈ f (A) and ynk → yo . Hence, the set
f (A) is compact.
3.2 Theorem. Let X and Y be metric spaces and let f : X → Y be a continuous
mapping. If X is compact, then the mapping f is uniformly continuous.
Proof. Let ε > 0. Since f is continuous, for x ∈ X, there is rx > 0 such that
ε
d(f (y), f (x)) < whenever d(y, x) < 2rx . The family of open balls B(x, rx ), x ∈ X
2
is an open cover of X. The space X is compact, there is a finite number of elements
S
n
x1 , x2 , ..., xn of X so that X = B(xi , rxi ). Let δ = min{rx1 , ..., rxn } > 0.
i=1
Now pick any x, y ∈ X with d(x, y) < δ. Then there exists a natural number i with
ε
1 ≤ i ≤ n satisfying d(x, xi ) < rxi . Therefore, d(f (x), f (xi )) < . So
2
d(y, xi ) ≤ d(y, x) + d(x, xi ) < δ + rxi < 2rxi ,
ε
hence, d(f (y), f (xi )) < . Consequently,
2
ε ε
d(f (x), f (y)) ≤ d(f (x), f (xi )) + d(f (xi ), f (y)) < + = ε.
2 2
This completes the proof.¥
3.3 Theorem. Let X be a compact metric space and f : X → R is a continuous function.
Then f is bounded. Moreover, f attains its greatest value and least value on X.
Proof. By Theorem 3.1, it follows that the set f (X) ⊂ R is compact. Hence f (X) is
bounded, that is, the function f is bounded on X. Let α = inf f (X) and β = sup f (X),
then α and β are two adherent points of f (X). Since f (X) is closed, it contains the points
α and β. This means that the function f attains the least value and the greatest value
on X.
In other words, there exist two elements xo , x1 ∈ X such that

f (xo ) = α, f (x1 ) = β and f (xo ) ≤ f (x) ≤ f (x1 ) for all x ∈ X.

Example. Let A be a nonempty compact subset of a metric space X and xo ∈ X.

Then there exists yo ∈ A such that d(xo , yo ) = d(xo , A).
Indeed, consider the function f : A → R to be defined by f (x) = d(xo , x). It is easy
to see that f is continuous on A. Since A is compact, f attains it least value at some
point yo ∈ A, that is,

. Luong Ha - HUCE
72 Chapter 4. Compact Metric Spaces

f (yo ) = inf f (x) = inf d(xo , x) = d(xo , A).

x∈A x∈A
Hence, d(xo , yo ) = d(xo , A).
Note that the element yo , in general, is not unique.

§4 APPROXIMATION THEOREMS

4.1 APPROXIMATION OF A FUNCTION BY A POLYNOMIAL

Let C[0;1] be the metric space consisting of all continuous functions on [0; 1] with the
metric
d(f, g) = max |f (x) − g(x)|.
x∈[0;1]

Denote by W the subset of C[0;1] whose elements are polynomial functions. Thus, each
element of W is a function of the form
p(x) = ao + a1 x + · · · + an xn ,
where n = 0, 1, 2, . . . , and ao , a1 , . . . , an are real numbers.
4.1.1 Weierstrass Theorem I. Each continuous function on [0; 1] is the uniform limit
of a sequence of polynomial functions.

With the above notations, we may reformulate the theorem as follows:

The set W of polynomial functions is dense in the metric space C[0;1] .
Proof. First we will show that it is sufficient to prove the theorem in the case f (0) =
f (1) = 0. Indeed, if not so, let

g(x) = f (x) − f (0) − [f (1) − f (0)] x.

Then g ∈ C[0;1] and g(1) = g(0) = 0. If we have proved the theorem for such functions,
then there is a polynomial function q such that |g(x) − q(x)| < ε for 0 ≤ x ≤ 1. Let

p(x) = q(x) + (f (1) − f (0))x + f (0)

for all x ∈ R. Clearly, p is a polynomial function and |f (x) − p(x)| < ε for 0 ≤ x ≤ 1.
Thus, we may suppose that f (0) = f (1) = 0.
Let f (x) = 0 if x ∈/ [0; 1], then f is uniformly continuous on R. Let qn (x) = cn (1−x2 )n
R1
for n = 1, 2, . . . and cn ≥ 0 such that qn (x)dx = 1 for each n ∈ N.
−1

By Bernoulli’s inequality, we have (1 − x2 )n ≥ 1 − nx2 ,

√1
Z1 Z1 Zn
(1 − x2 )n dx = 2 2 n
(1 − x ) dx ≥ 2 (1 − nx2 )dx
−1 0 0
4 1
= √ >√ .
3 n n

. Luong Ha - HUCE
§ 4. Approximation Theorems 73

1 √
Therefore √ cn < 1, so cn < n. For β ∈ (0, 1),
n
√ √
qn (x) = cn (1 − x2 )n ≤ n(1 − x2 )n ≤ n(1 − β 2 )n

for any x satisfying β ≤ |x| ≤ 1. Thus, the sequence of polynomial functions qn is

uniformly convergent to the function 0 on [−1, β] ∪ [β, 1]. Now set
Z1
pn (x) = f (x + t)qn (t)dt for x ∈ [0, 1].
−1

Using the change of variable y = x + t and noting that f (x) = 0 if x ∈

/ [0, 1], we obtain
Z1−x Z1
pn (x) = f (x + t)qn (t)dt = f (y)qn (y − x)dy
−x 0
Z1
£ ¤n
= f (y) 1 − (y − x)2 dy.
0

Develop the expression [1 − (y − x)2 ]n , arrange it into the increasing powers of x, and
use the linearity of integrals, we see that pn (x) is a polynomial function of the variable x.
Now we only need to prove that the sequence (pn )n converges uniformly to the function
f on [0; 1].
Let ε > 0. Since f is uniformly continuous on R, there is β ∈ (0; 1) such that
¯ ¯
¯f (x) − f (y)¯ < ε for all x, y ∈ R satisfying |x − y| < β . Set M = sup |f (x) |. Then for
2 2 x∈[0,1]
all x ∈ [0; 1],
Z1
¯ ¡ ¢ ¯
|pn (x) − f (x)| = ¯ f (x + t) − f (x) qn (t)dt¯
−1
Z1
¯ ¯
≤ ¯f (x + t) − f (x)¯ qn (t) dt
−1
Z−β Zβ Z1
ε
≤ 2M qn (t)dt + qn (t)dt + 2M qn (t) dt
2
−1 −β β
√ ε
≤ 4M n(1 − β 2 )n + .
2
√ √ ε
Since lim n(1 − β 2 )n = 0, there exists no ∈ N such that 4M n(1 − β 2 )n < for all
n 2
n ≥ no .
Hence, there is no ∈ N such that |pn (x) − f (x)| < ε for all x ∈ [0; 1] and for all n ≥ no .
Consequently, the sequence (pn )n is uniformly convergent to f on [0; 1]. The proof of the
theorem is complete.¥

. Luong Ha - HUCE
74 Chapter 4. Compact Metric Spaces

4.1.2 Corollary. The metric space C[0;1] is separable.

Proof. Let f ∈ C[0;1] and ε > 0. By Weierstrass theorem I, there exists a polynomial
ε
function p(x) = ao + a1 x + · · · + an xn such that d(f, p) < . Since the set of rational
2
ε
numbers is dense in R, for each i = 0, 1, . . . , n, there is bi ∈ Q such that |bi −ai | < .
2(n + 1)
Let q(x) = bo +b1 x+· · ·+bn xn , then q ∈ C[0,1] and q is a polynomial function with rational
coefficients. For all x ∈ [0; 1],

|q(x) − p(x)| = |(bo − ao ) + (b1 − a1 )x + · · · + (bn − an )xn |

X n Xn
i ε
≤ |bi − ai | x ≤ |bi − ai | < .
i=0 i=0
2

ε
It follows that d(q, p) < , hence,
2
d(f, q) ≤ d(f, p) + d(p, q) < ε.

Thus, the set of polynomials with rational coefficients is dense in C[0;1] . This set is
countable, so C[0;1] is separable.
4.1.3. Remark. For a, b ∈ R and a < b, let ϕ : C[0;1] → C[a;b] defined by ϕ(f ) = g, where
x−a
g(x) = f ( ), ∀x ∈ [a; b].
b−a
Then we can prove that ϕ is an isometry from C[0;1] onto C[a;b] . In addition, ϕ transforms
each polynomial function into a polynomial function. Therefore, the set of polynomial
functions is dense in C[a;b] and the metric space C[a;b] is separable. In other words, Weier-
strass theorem I holds true for any continuous function on an arbitrary segment [a; b].

4.2 APPROXIMATION BY A TRIGONOMETRIC POLYNOMIAL

4.2.1 Weierstrass Theorem II. Let f be a continuous and 2π-periodic function on R.

Then for each ε > 0, there exists a trigonometric polynomial
n
ao X
sn (x) = + (ak cos kx + bk sin kx)
2 k=1

such that |f (x) − sn (x)| < ε for all x ∈ R.

Proof. For each x ∈ R, set
f (x) + f (−x) f (x) − f (−x)
ϕ(x) = , Ψ(x) = sin x.
2 2
Then ϕ and Ψ are continuous on R, even and periodic with the period 2π. For x ∈ [0; π],
let x = arccos t, then the functions

α(t) = ϕ(arccos t), β(t) = Ψ(arccos t)

. Luong Ha - HUCE
§ 4. Approximation Theorems 75

are continuous on [−1; 1]. By Weierstrass theorem I, there exist two polynomials g, h
defined on [−1; 1] such that for all t ∈ [−1; 1]
ε ε
|α(t) − g(t)| < , |β(t) − h(t)| < .
4 4
Hence, for all x ∈ [0; π], we have
ε ε
|ϕ(x) − g(cos x)| < and |ϕ(x) − h(cos x)| < .
4 4
Moreover, the functions ϕ(x), Ψ(x) and cos x are even and 2π-periodic, these inequalities
hold for all x ∈ R.
By our construction, f (x) sin x = ϕ(x) sin x + Ψ(x). Hence

|f (x) sin x − g(cos x) sin x − h(cos x)| ≤ |ϕ(x) − g(cos x)| | sin x| + ϕ(x) − h(cos x)|
ε
< .
2
Let u(x) = g(cos x) sin x + h(cos x). The function u(x) is a polynomial of the variable x
and
ε
|f (x) sin x − u(x) | < ∀x ∈ R. (1)
2
π
For the function f ( − x), the similar result holds
2
π ε
|f ( − x) sin x − v(x)| < ∀x ∈ R, (2)
2 2
where v(x) is also a polynomial of the variable x.
π
Replace x by − x in (2), we get
2
π ε
|f (x) cos x − v( − x)| < ∀x ∈ R. (3)
2 2
It follows from (1) and (3) that
¯ ¯ ¯ ¯
¯f (x) − u(x) sin x − v( π − x) cos x¯ = ¯f (x)(sin2 x + cos2 x) − u(x) sin x − v( π − x) cos x¯
2 2
≤ |f (x) sin x − u(x)|| sin x|
π
+ |f (x) cos x − v( − x)|| cos x|
2
ε ε
< + = ε.
2 2
π
Now let s(x) = u(x) sin x + v( − x) cos x. Then s(x) is a trigonometric polynomial
2
and |f (x) − s(x)| < ε for all x ∈ R.¥

. Luong Ha - HUCE
76 Chapter 4. Compact Metric Spaces

PROBLEMS
4.1 Prove that the intersection of an arbitrary family of compact subsets of a metric space
is compact.
4.2 Let A and B be two nonempty subsets of a metric space X. The distance between
the sets A and B is defined as

d(A, B) = inf{d(x, y) : x ∈ A, y ∈ B}.

1) Show that d(A, B) = d(A, B) = d(A, B) = d(A, B.

2) Show that if A and B are compact, then there exist elements xo ∈ A, yo ∈ B such
that d(A, B) = d(xo , yo ).
3) Suppose that A is closed, B is compact and A ∩ B = ∅. Prove that d(A, B) > 0.
4) Give an example for showing that if the sets A are B closed and both are not
compact then the result in 2) does not hold.
4.3 Prove that the closed unit balls in the metric spaces C[0;1] and `2 are not compact.
4.4 Show that the product of two metric compact spaces is compact.
4.5 Let X be a metric space and f : X → X is a continuous mapping.
1) Show that the set of fixed points of f is closed.
2) Suppose that X is compact and for each c > 0, there exists x ∈ X such that
d(x, f (x)) < c. Show that the mapping f has at least one fixed point.
4.6 Let X and Y be metric spaces and f : X → Y is a mapping. Prove that f is
continuous on X if and only if f is continuous on every compact subset of X.
4.7 Let X be a compact metric space and f : X → X such that d(f (x), f (y)) < d(x, y)
holds for x, y ∈ X, x 6= y. Prove that f has a unique fixed point.
4.8 Let X be a compact metric space and f : X → X is a mapping such that d(f (x), f (y)) ≥
d(x, y) for any x, y ∈ X. Show that f is an isometry from X onto X.
4.9 Prove that a metric space X is compact if and only if each infinite subset of X has
at least one accumulation point.
4.10 Prove that a discrete metric space X is compact if and only if the set X is finite.
4.11 Let X be a compact metric space and f : X → X is a continuous mapping. Suppose
that (Kn )n is a decreasing sequence of nonempty closed subsets of X. Prove that
∞
\ ∞
\
f( Kn ) = f (Kn ).
n=1 n=1

L
4.12 Show that the metric space C[a;b] is separable.
R1
4.13 Let f be a function continuous on [0; 1]. Suppose that xn f (x) dx = 0 for all
0
n = 0, 1, 2, .... Show that f = 0.

. Luong Ha - HUCE
§ 4. Approximation Theorems 77

4.14 Show that a continuous function f : (0; 1) → R is the uniform limit of a sequence of
polynomials on (0; 1) if and only if it admits a continuous extension.
4.15 Assume that a function f : [0; +∞) → R is either a polynomial or else a continuous
R∞
bounded function. Then show that f is identical to zero if and only if f (x)e−nx dx = 0
0
for all n = 0, 1, 2, ....
4.16 Let (fn )n be a sequence of real-valued functions defined on a compact metric space
X such that xn → x in X implies fn (xn ) → f (x). If f is continuous, then show that the
sequence (fn )n is uniformly convergent to f .
4.17 Let M = {x ∈ C[0;1] : x(1) = 1, 0 ≤ x(t) ≤ 1 for all t ∈ [0; 1]}.
1) Prove that M is a nonempty closed and bounded set in C[0;1] .
R1
2) Consider the function f defined on C[0;1] by f (x) = x2 (t) dt. Prove that f is
0
continuous on M but it does not attain the least value on M . Is the set M compact?
4.18 Let X and Y be two metric spaces and f : X → Y is a mapping. The graph of f ,
denoted by G(f ), is defined by
G(f ) = {(x, y) ∈ X × Y : y = f (x)}.
1) Show that if f is continuous, then G(f ) is closed in the product space X × Y .
2) Suppose that Y is compact and G(f ) is closed in X × Y . Show that f is continuous.
Does this result hold true if Y is not assumed to be compact?
4.19 Let X be a compact metric space and (fn )n is a monotone sequence of continuous
functions on X. Show that if the sequence (fn )n is pointwise convergent to a continuous
function f on X, then (fn )n is uniformly convergent.
4.20 Let X be a metric space, A is a nonempty closed subset of X and xo ∈ X.
1) Suppose xo ∈
/ A. Show that there exists r > 0 such that d(xo , x) ≥ r for all x ∈ A.
2) Let A be compact. Show that yo ∈ A such that d(xo , A) = d(xo , yo ). Give an
example for showing that the point yo may not be unique.
3) Suppose X = Rk with the usual metric. Show that there exists yo ∈ A such that
d(xo , A) = d(xo , yo ).
4.21 Let X and Y be two metric spaces and K ⊂ X is a compact set. Let f : X →
Y is a continuous mapping. Show that for each ε > 0, there exists δ > 0 such that
d(f (x), f (y)) < ε for all x ∈ K, y ∈ X with d(x, y) < δ.
4.22 Let X be a compact metric space and (Gα )α is an open cover of X. Show that there
exists a number r > 0 such that each open ball of radius r is contained in some Gα . (The
number r is called a Lebesgue number for the open cover (Gα )α .)

4.23 Let X be a metric space. Suppose that there is r > 0 such that all closed balls
B 0 (x, r), x ∈ X is compact. Show that X is complete.

. Luong Ha - HUCE
78 Chapter 4. Compact Metric Spaces

§5 ASCOLI-ARZELA THEOREM

5.1 RELATIVELY COMPACT SETS

5.1.1 Definition. Let X be a metric space. A set A ⊂ X is called relatively compact if

the closure A is compact.
5.1.2 Remark.
1) Any compact set is relatively compact.
2) Every subset of a relatively compact set is relatively compact.
3) If A is relatively compact, then A is totally bounded. Conversely any totally
bounded subset of a complete metric space is relatively compact.
5.1.3 Theorem. A set A ⊂ X is relatively compact if and only if every sequence in A
has a convergent subsequence.
Proof. Let A be relatively compact and (xn )n ⊂ A. Then (xn )n ⊂ A and since A is
compact, there exists a subsequence (xnk )k of (xn )n such that (xnk )k converges to some
point of A.
Conversely let (xn )n ⊂ A. Then for each n ∈ N, there exists yn ∈ A such that
1
d(xn , yn ) < . By hypothesis, there is a subsequence (ynk )k of (yn )n so that (ynk )k →
n
yo ∈ X. But (ynk )k ⊂ A, it follows that yo ∈ A. On the other hands
1
d(xnk , yo ) ≤ d(xnk , ynk ) + d(ynk , yo ) < + d(ynk , yo ),
nk

we get d(xnk , yo ) → 0 as k → ∞, that is, xnk → yo . Thus, A is compact, hence A is

relatively compact.

5.2 ASCOLI-ARZELA THEOREM

The general criteria for compactness obtained in Section 4.2 are seldom easy to verify.
In this section we will establish an important and often readily verifiable, characterization
of the compact subsets the metric C(X) when X is itself a compact metric space.
Let X be a compact metric space. Denote by C(X) the set of all continuous functions
from X into R with the metric

d(f, g) = max |f (x) − g(x)|.

x∈X

Then the metric space C(X) is complete (the proof is similar to the proof in the case
X = [a; b]).
5.2.1 Definition. Let A be a nonempty subset of C(X). We call A is equicontinuous
at a point x ∈ X if for each ε > 0, there exists δ > 0 such that |f (y) − f (x)| < ε for all
y ∈ X with d(x, y) < δ and all f ∈ A.

. Luong Ha - HUCE
§ 5. Ascoli-Arzela theorem 79

The set A is called equicontinuous on X if it is equicontinuous at each point of X.

Note that each f ∈ A is continuous at the point x, there exists δ(f ) > 0, which may
depend on x, ε and f , such that |f (y) − f (x)| < ε for all y ∈ X with d(y, x) < δ(f ). In
general, δ(f ) varies with f ; the condition that A is equicontinuous at x requires that a
single δ > 0 can be chosen independently of f ∈ A.
5.2.2 Ascoli-Arzela theorem. Let X be a compact metric space. Then a nonempty
subset of C(X) is relatively compact if and only if it is bounded and equicontinuous on X.
Proof. Let A ⊂ C(X) be relatively compact. Then A is totally bounded, so it is
bounded.
ε
Let ε > 0. There exists a finite number of open balls of radius and centered
3
at f1 , f2 , . . . , fn which cover A. Each function fi is continuous on X, it is uniformly
ε
continuous. Hence, there is δi > 0 such that |fi (x) − fi (y)| < for all x, y ∈ X with
3
d(x, y) < δi . Set δ = min{δ1 , δ2 , . . . , δn }. Then for all f ∈ A, there exists i ∈ {1, 2, . . . , n}
ε
such that d(f, fi ) < . Pick any x, y ∈ X with d(x, y) < δ, we have
3

|f (x) − f (y)| ≤ |f (x) − fi (x)| + |fi (x) − fi (y)| + |fi (y) − f (y)|
≤ 2d(f, fi ) + |fi (x) − fi (y)|
2ε ε
< + = ε.
3 3
Thus, the set A is equicontinuous at any x ∈ X.
Conversely, suppose that A is bounded and equicontinuous on X. Since C(X) is
complete, we only need to prove that A is totally bounded. Let ε > 0. By the hypothesis
of the equicontinuity of A, for each x ∈ X there exists δ > 0 such that
ε
|f (y) − f (x)| < (1)
3
for all y ∈ X with d(y, x) < δx and for all f ∈ A. X is compact, the open cover
{B(x, δx ) : x ∈ X} of X has a finite subcover, say {B(x1 , δ1 ), B(x2 , δ2 ), . . . , B(xn , δn )},
where δi = δxi , i = 1, 2, . . . , n.
Consider the mapping Φ : C(X) → Rn given by

Φ(f ) = (f (x1 ), f (x2 ), . . . , f (xn )).

The set A is bounded, there is K > 0 such that |f (x)| ≤ K for all x ∈ X and all
n
£X ¤1 √
f ∈ A. Then for each f ∈ A, d(Φ(f ), 0) = f (xi )2 2 ≤ nK. Thus, the set
√ i=1
Φ(A) ⊂ B 0 (0, nK), that is, Φ(A) is bounded in Rn . It follows that Φ(A) is totally
ε
bounded. Therefore, Φ(A) is covered by a finite number of open balls of radius centered
6
at ξj = (ξj1 , ξj2 , . . . , ξjn ), j = 1, 2, . . . , m. If there is an open ball which does not intersect A,
we will remove it, then we may suppose that Φ(A) has nonempty intersections with these

. Luong Ha - HUCE
80 Chapter 4. Compact Metric Spaces

ε
open balls. For each j ∈ {1, 2, . . . , m}, there exists fj ∈ A such that Φ(fj ) ∈ B(ξj , ).
6
Thus, for any f ∈ A
|f (xi ) − fj (xi )| ≤ |f (xi ) − ξji | + |ξji − fj (xi )|
ε ε ε
< + = (2)
6 6 3
for all i = 1, 2, . . . , n.
S
m
Now we are in position to show that A ⊂ B 0 (fj , ε). (3)
j=1

Let f ∈ A. Then there is j ∈ {1, 2, . . . , m} such that (2) is satisfied for all i =
1, 2, . . . , n. Pick x ∈ X, there is i ∈ {1, 2, . . . , n} such that x ∈ B(xi , δi ). It follows from
(2) that
|f (x) − fj (x)| ≤ |f (x) − f (xi )| + |f (xi ) − fj (xi )| + |fj (xi ) − fj (x)|
ε ε ε
< + + = ε.
3 3 3
Hence, d(f, fj ) ≤ ε, that is, the set A is totally bounded. The proof of the theorem is
now complete.¥
5.2.3 Remark. In the above proof, if we replace the assumption that A is bounded by
that A is bounded at every point of X, that is, for each x ∈ X, there is Kx > 0 such
that |f (x)| ≤ Kx for all f ∈ A, then the set Φ(A) is still bounded. Hence A is relatively
compact.
Example. Let M be a bounded set of the metric space C[0;1] . Then
Zt
A = {y ∈ C[0,1] : y(t) = x(s)ds, x ∈ M}
0

is relatively compact.
Since M is bounded, there exists K > 0 such that for all x ∈ M and t ∈ [0; 1], we
have |x(t)| ≤ K. Then for any y ∈ A and t ∈ [0; 1]
Zt Zt Zt
¯ ¯ ¯
|y(t)| = ¯ x(s)ds¯ ≤ ¯x(s)ds ≤ Kds ≤ K,
0 0 0

hence A is bounded.
Let ε > 0 and to ∈ [0, 1].
¯ Zt Zto ¯ ¯ Zt ¯
¯ ¯ ¯ ¯
|y(t) − y(to )| = ¯ x(s)ds − x(s)ds¯ = ¯ x(s)ds¯
0 0 to

≤ |t − to |K.
ε
Choose δ = , then |y(t) − y(to )| < ε for all t ∈ [0; 1] with |t − to | < δ and all y ∈ A.
K
This means that A is equicontinuous on [0; 1]. By Ascoli-Arzela theorem, A is relatively
compact.

. Luong Ha - HUCE
§ 5. Ascoli-Arzela theorem 81

PROBLEMS
4.24 Let X be a compact metric space and A ⊂ X is a finite set. By using definition,
show that A is equicontinuous on X.
4.25 Let X be a compact metric space. Suppose that A is a subset of X which is bounded
at every point of X. Show that A is bounded.
4.26 Let X be a compact metric space and A ⊂ C(X) is equicontinuous on X. Show
that A is uniformly equicontinuous on X, that is, for each ε > 0, there exists δ > 0 such
that |f (x) − f (y)| < ε for all x, y ∈ X with d(x, y) < δ and for all f ∈ A.
4.27 Let f ∈ C(X), where X = [0; 1] × [0; 1]. For every y ∈ [0; 1], define fy : [0; 1] → R
by fy (x) = f (x, y). Show that the set F = {fy : y ∈ [0; 1]} is uniformly equicontinuous
on [0; 1].
4.28 Let α, β, γ be 3 positive numbers and a, b ∈ R with a < b. Let F be the subset of
C[a;b] consisting of the functions f satisfying
|f (a)| ≤ γ, |f (x) − f (y)| ≤ β|x − y|α
for all x, y ∈ [a; b]. Show that F is a compact set in C[a;b] .
4.29 Which of the following sets is compact in C[0;1] ?
a) {xn ∈ C[0;1] : xn (t) = tn , n ∈ N}.
b) {xα ∈ C[0;1] : xα (t) = sin αt, α ∈ [1; 2]}.

HINTS and SOLUTIONS

T
4.1 Let (Aα )α be a family of compacts sets in a metric space X. Set A = Aα . Each
α∈I
set Aα is closed, A is closed. Pick αo ∈ I, then Aαo is compact, and A ⊂ Aαo . Thus,, A
is compact.
4.2
1) By definition, there exists a sequence (xn )n in A and a sequence (yn )n in B such that
lim d(xn , yn ) = d(A, B). Since A is compact, there exists (xnk )k ⊂ (xn )n so that xnk →
n
xo ∈ A. The set B is compact, the subsequence (ynk )k of (yn )n has a subsequence (ynkj )
³ ´
with ynkj → yo ∈ B. Then xnkj → xo , hence, d(xnkj , ynkj ) → d(xo , yo ). But d(xnkj ) is
j
a subsequence of (d(xn , yn ))n , so d(xnkj , ynkj ) → d(A, B). Thus, d(A, B) = d(xo , yo ).
2) Suppose that d(A, B) = 0. Then there exist (xn )n ⊂ A and (yn )n ⊂ B such that
d(xn , yn ) → 0. The sequence (yn )n belongs to the compact set B, there is a subsequence
(ynk )k with ynk → xo ∈ B. Since d(xnk , xo ) ≤ d(xnk , ynk )+d(ynk , xo ), we have d(xnk , xo ) →
0 as k → ∞, or xnk → xo . Hence, xo ∈ A since A is closed. Thus, xo ∈ A ∩ B. This
contradicts the hypothesis that A ∩ B = ∅. Therefore, d(A, B) > 0.
3) In R2 , consider two sets

A = {(x, y) : xy = 1}, B = {(x, y) : y = 0}.

. Luong Ha - HUCE
82 Chapter 4. Compact Metric Spaces

These sets are closed, but not bounded, hence, they are not compact. Furthermore,
A ∩ B = ∅.
1
Let (un )n ⊂ A with un = (n, 1n) and (vn )n ⊂ B with vn = (n, 0). Then d(un , vn ) =
n
1
for all n ∈ N. Hence, d(A, B) ≤ for all n ∈ N. Thus, d(A, B) = 0.
n
4.3
a) The sequence (xn )n ⊂ B 0 (0, 1) with xn (t) = tn is pointwise convergent to the
function x, where x(t) = 1 if t = 1, and x(t) = 0 if t ∈ [0; 1). Clearly, every subsequence
of (xn )n is pointwise convergent to x on [0; 1]. The function x ∈ / C[0;1] , hence, each
0
subsequence of (xn )n does not converges. This means that the ball B (0, 1) is not compact.
b) Consider the sequence (en )n ⊂ B 0 (0, 1) ⊂ `2 with en = (0, . . . , 0, 1[n] , 0, . . . ). Then
√
d(en , em ) = 2 for n 6= m. Hence, each subsequence of (en )n does not converge. Conse-
quently, B 0 (0, 1) is not compact.
4.4 Hints.
Show that each sequence in X × Y has a convergent subsequence.
4.5 1) Let A = {x ∈ X : x = f (x)}. If (xn )n ⊂ A, and xn → xo , then xo = f (xo ), hence,
xo ∈ A. Thus, A is closed.
1
2) By hypothesis, for each n ∈ N, there exists xn ∈ X such that d(xn , f (xn )) < .
n
Since X is compact, there is a subsequence (xnk )k of (xn )n so that xnk → xo ∈ X. Then
1
f (xnk ) → f (xo ) since f is continuous. We have d(xnk , f (xnk )) < for each k ∈ N. Let
nk
k → ∞, we get d(xo , f (xo )) = 0. This means that xo is a fixed point of f .
4.6 Clearly, if f is continuous, then it is continuous on any compact subset of X. Con-
versely, suppose that f is continuous on any compact set of X. We shall show that f is
continuous at any point xo of X.
Consider (xn )n ⊂ X with xn → xo . Set A = {xn : n ∈ N} ∪ {xo }. Let (Gα )α∈I be any
open cover of A. Then there is αo ∈ I so that xo ∈ Gαo . The set Gαo is open , hence, there
exists ε > 0 with B(xo , ε) ⊂ Gαo . Since xn → xo , there is no ∈ N such that xn ∈ B(xo , ε)
for all n > no . For each i ∈ {1, 2, . . . , no }, there exists Gαi which contains xi . Hence,
³S no ´ S
no
A⊂ Gαi ∪ Gαo = Gαi . Therefore, A is compact. By hypothesis, f is continuous
i=1 i=0
on A. From the fact that (xn )n ⊂ A and xn → xo ∈ A, it follows that f (xn ) → f (xo ).
Consequently, f is continuous on X.
4.7 Suppose that f has no fixed point, that is, d(f (x), x) > 0 for all x ∈ X. Let g : X → R
be defined by g(x) = d(f (x), x). Then g is continuos. Since X is compact, there is xo ∈ X
with g(xo ) ≤ g(x) for all x ∈ X. Set u = f (xo ), then d(f (u), u) = d(f (f (xo )), f (xo )) <
d(f (xo ), xo ), that is, g(u) < g(xo ), which is a contradiction.
Thus, f has at least one fixed point. It is easy to show that if f has two fixed points
x1 and x2 , then x1 = x2 .
4.8 Hints.

. Luong Ha - HUCE
§ 5. Ascoli-Arzela theorem 83

Let a1 = f (a), a2 = f (a1 ), . . . , an = f (an−1 ), . . . , and b1 = f (b), b2 = f (b1 ), . . . ,

bn = f (bn−1 ), . . . for any a, b ∈ X. Since X is compact, there exist two subsequences
(aln )n , (bln )n of (an )n , (bn )n respectively, which are convergent.
Show that for each ε > 0, there exists p ∈ N such that d(a, ap ) < ε, and d(b, bp ) < ε.
It follows that the set f (X) is dense in X, and d(f (a), f (b)) < 2ε + d(a, b). Hence,
d(f (a), f (b)) = d(a, b) for any a, b ∈ X. Thus, f is injective and continuous. Therefore,
f (X) is compact, and hence, f (X) = f (X) = X. This means that f is an isometry from
X onto X.
4.10 Hints.
Apply Heine-Borel theorem.
4.12 Consider P (x) = ao + a1 x + · · · + an xn . We have

Z1 Z1 Z1 Z1
P (x)f (x)dx = ao f (x)dx + a1 xf (x)dx + · · · + an xn f (x)dx.
0 0 0 0

R1
By hypothesis, P (x)f (x)dx = 0 for any polynomial P . The function f is continuous on
0
[0; 1], there exists K > 0 so that |f (x)| ≤ K for all x ∈ [0; 1]. By Weierstrass theorem I,
ε
for each ε > 0, there is a polynomial P such that max |f (x) − P (x)| < . Hence,
x∈[0;1] K

Z1 Z1 Z1
£ ¤
f 2 (x)dx = f 2 (x) − P (x)f (x) dx ≤ |f (x)||f (x) − P (x)| dx.
0 0 0

Z1 Z1 Z1
ε
It follows that f 2 (x)dx ≤ K dx = ε. Thus, f 2 (x)dx = 0. Since f 2 ≥ 0 and is
K
0 0 0
continuous [0; 1], we must have f 2 = 0 or f = 0.
4.17
1) The constant function x = 1 belongs to M , so M 6= ∅. If (xn )n ⊂ M , and
xn → xo ∈ C[0;1] , then xo (1) = 1, and 0 ≤ xo (t) ≤ 1 for all t ∈ [0; 1]. Hence, xo ∈ M ,
and M is closed. On the other hand, for any x ∈ M , d(x, 0) = max |x(t)| ≤ 1, hence,
t∈[0;1]
0
M ⊂ B (0, 1). Thus, M is bounded.
2) Consider xo ∈ M . For any x ∈ M , we have:

Z1
|f (x) − f (xo )| ≤ |x(t) − xo (t)||x(t) + xo (t)|dt.
0

Note that
|x(t) + xo (t)| ≤ |x(t) − xo (t)| + 2|xo (t)|,

. Luong Ha - HUCE
84 Chapter 4. Compact Metric Spaces

and |x(t) − xo (t)| ≤ 1, |xo (t) ≤ 1 for all t ∈ [0; 1]. It follows that
Z1
|f (x) − f (xo )| ≤ 3 d(x, xo ) dt = 3dt(x, xo ).
0

It is easy to see that f is continuous on M . Clearly, f (x) ≥ 0 for any x ∈ M , so

Z1
1
n
inf f (x) ≥ 0. Consider (xn )n ⊂ M with xn (t) = t . Then f (xn ) = t2n dt = ,
x∈M 2n + 1
0
and f (xn ) → 0 as n → ∞. Thus, inf f (x) = 0. By definition, x(1) = 1 and x2 ≥ 0 for
x∈M
Z1
all x ∈ M , hence, x2 (t)dt > 0. This shows that f does not attain the least value on
0
M . Therefore, M is not compact.
4.18
1) Let (zn )n ⊂ G(f ), and zn → zo in X × Y . With zn = (xn , f (xn )), and zo = (xo , yo ),
it follows that xn → xo , and f (xn )) → yo . By the continuity of f , f (xn ) → f (xo ). Hence,
yo = f (xo ), that is, zo ∈ G(f ). This shows that the graph of f is closed in X × Y .
2) Suppose that f is not continuous at some point xo ∈ X. Then there exists a sequence
(xn )n ⊂ X, xn → xo but f (xn ) 6−→ f (xo ). Hence, there is ε > 0, and a subsequence (xnk )n
of (xn )n such that d(f (xnk ), f (xo )) ≥ ε for all k ∈ N. Since Y is compact, there exists a
subsequence (xnkj )j of (xnk )k so that f (xnkj ) → yo ∈ Y .
Then xnkj → xo , and we have (xnkj , f (xnkj )) → (xo , yo ). By hypothesis, G(f ) is
closed, so yo = f (xo ). This is a contradiction since d(f (xnkj ), f (xo )) ≥ ε for all j ∈ N.
Consequently, f is continuous.
4.19 Without loss of generality, we may assume that the sequence (fn )n is increasing. Let
ε > 0 and x ∈ X. Since lim fn (x) = f (x), there exists n(x) ∈ N such that for n ≥ n(x)
n

0 ≤ f (x) − fn (x) < ε.

The functions f and fn(x) are continuous at x, there is rx > 0 so that if y ∈ B(x, rx ), then
ε ε
|f (y) − f (x)| < , and |fn(x) (y) − fn(x) (x)| < . Hence, for each y ∈ B(x, rx ),
3 3
|f (y) − fn(x) (y)| < ε.

The family of open balls (B(x, rx ))x∈X is an open cover of the compact space X, there
S
is a finite number of elements x1 , x2 , ..., xn in X such that X = ki=1 B(xi , rxi ). Set
no = max {(n(x1 ), n(x2 ), ..., n(xk )}. Then for all n ≥ no , and for any x ∈ X, there exists
i ∈ {1, 2, ..., k} such that

0 ≤ f (x) − fn (x) ≤ f (x) − fno (x) ≤ f (x) − fn(xi ) < ε.

Thus, for each ε > 0, there exists no ∈ N so that 0 ≤ f (x) − fn (x) < ε for all n ≥ no and
all x ∈ X. This means that the sequence (fn )n converges uniformly to f on X.

. Luong Ha - HUCE
§ 5. Ascoli-Arzela theorem 85

4.20 Hints.
2) Let A = {−1, 1}. Then A is compact.
Choose xo = 0, then d(xo , A) = 1 = |1 − 0| = | − 1 − 0|.
3) Choose yo = xo if xo ∈ A. Suppose that xo ∈ / A. Then d(xo , A) > 0. Pick
0
b ∈ A, and let r = d(xo , b). Set C = A ∩ B (xo , r). By 2), there is yo ∈ C such that
d(xo , yo ) = d(xo , C). Show that d(xo , C) = d(xo , A), and we get d(xo , yo ) = d(xo , C).
4.24 Consider A = {f1 , f2 , . . . , fn } ⊂ C(X), xo ∈ X, and ε > 0. Then for each i =
1, 2, . . . , n, there exists δi > 0 such that |fi (x) − fi (xo )| < ε whenever d(x, xo ) < δi .
Let δ = min δi . For any x ∈ X with d(x, xo ) < δ, we have |fi (x) − fi (xo )| < ε for all
1≤i≤n
i = 1, 2, . . . , n. Hence, the set A is equicontinuous at any xo ∈ X.
4.26 Let ε > 0. Since A is equicontinuous on X, for each x ∈ X, there is δx > 0 such
ε ¡ 1 ¢
that if d(y, x) < δx , then |f (x) − f (y)| < for all f ∈ A. The collection B(x, δx ) x∈X
2 2
is an open cover of the compact space X, so there exist x1 , x2 , . . . , xn ∈ X such that
Sn 1
X= B(xi , δi ), where δi = δxi .
i=1 2
1
Set δ = min δi . Pick x, y ∈ X with d(x, y) < δ. Then there is i ∈ {1, 2, . . . , n} so
2 1≤i≤n
that x ∈ B(xi , δi ). Therefore,
1
d(y, xi ) ≤ d(y, x) + d(x, xi ) < δ + δi ≤ δi .
2
Thus, for all f ∈ A
ε ε
|f (x) − f (y)| ≤ |f (x) − f (xi )| + |f (xi ) − f (y)| < + = ε.
2 2
Consequently, A is uniformly equicontinuous on X.
4.28 Hints.
1) Show that F is closed in C[a;b] .
2) Show that F is bounded and equicontinuous on [a; b].
4.29
a) The sequence of functions (xn )n is pointwise convergent to the function x with
x(t) = 1 if t = 1, and x(t) = 0 if t ∈ [0; 1). Then every subsequence of (xn )n is pointwise
convergent to x on [0; 1]. Since x is not continuous on [0; 1], each subsequence of (xn )n
does not converge in this space. Thus, the given set is not compact.
b) Show that this set is closed and relatively compact (by using Ascoli-Arzela theorem).

. Luong Ha - HUCE
Chapter 5

Topological Spaces

§1 INTRODUCTION TO TOPOLOGICAL SPACES

In the previous chapter, the basic properties of metric spaces were discussed. The
open and closed sets played an important role in that study. In this chapter the concepts
used in a metric spaces will be generalized by introducing the notion of a topological
space. Then we will sudy the properties of fundamental concepts of a topological space.

1.1 THE CONCEPT OF A TOPOLOGICAL SPACE

1.1.1 Definition.
Let X be a nonempty set. A family τ of subsets of X is called a topology on X if
i) ∅ and X belong to τ ;
ii) The union of an arbitrary subfamily of τ is an element of τ ;
iii) The intersection of any two elements of τ is an element of τ .
Then the pair (X, τ ) is called a topological space. Each element of τ is called an open
set, and each element of X is called a point. When no confusions arise, we may not
mention to the topology τ and (X, τ ) is denoted by X. Note that the property iii) still
holds for a finite subfamily of τ . Then in a topological space X, we have
i) The empty set and the set X are open sets.
ii) The union of an arbitrary family of open sets is an open set.
iii) The intersection of any finite family of open sets is an open set.
1.1.2 Example.
1) Let X be a nonempty set and let τ = {∅, X}. Then τ is a topology on X, called
the indiscrete topology.
2) Let X be a nonempty set. Then the family τ = P(X) is a topology on X, called
the discrete topology.
3) Let (X, d) be a metric space. Then the family of all open sets of X satisfy the

86
§ 1. Introduction to Topological Spaces 87

properties for a topology. This topology is called the topology generated by the metric d.
Thus, each metric space is a topological space.
4) The metric spaces Rk (with the usual metric) are topological spaces, and the cor-
responding topologies are called the usual topologies on Rk .
5) Let X be a nonempty set and A be a nonempty proper subset of X. The family
τ = {∅, A, X} is a topology on X.
6) Let X be an infinite set and let τ = {U ⊂ X : X \ U is finite} ∪ {∅}. Then τ is a
topology on X and it is called the cofinite topology. This space is often called a cofinite
topological space.
We verify the axioms of a topology. It is clear that ∅ ∈ τ and X \ X = ∅ is finite,
hence, X ∈ τ .
S
Let(Gα )α∈I be a subfamily of τ . Set G = Gα . If G = ∅, then G ∈ τ . Suppose that
α∈I
G 6= ∅. Then there exists α0 ∈ I with Gα0 6= ∅. By definition, X \ Gα0 is finite, and since
X \ G ⊂ X \ Gαo , the set X \ G is finite. Hence, G ∈ τ .
Let A, B belong to τ . If A ∩ B = ∅, then A ∩ B belongs to τ . Suppose that A ∩ B 6= ∅.
Thus, A 6= ∅ and B 6= ∅, and hence, X \ A and X \ B are finite. Since

X \ (A ∩ B) = (X \ A) ∪ (X \ B),

it follows that X \ (A ∩ B) is finite, that is, A ∩ B ∈ τ . Therefore, τ is a topology.

1.1.3 Comparison of topologies.
Let τ1 and τ2 be two topologies on a set X. The topology τ1 is called weaker than the
topology τ2 (or τ2 is stronger than τ1 ) if τ1 ⊂ τ2 . It is evident that if τ1 is weaker than τ2
and τ2 is weaker than τ1 , then τ1 = τ2 .
Obviously, the indiscrete topology is the weakest topology, and the discrete topology
is the strongest topology on the set X. If τ1 and τ2 are arbitrary topologies for X, it may
happen that τ1 is neither stronger nor weaker than τ2 . Then we say that τ1 and τ2 are not
comparable. For instance, if A and B are two nonempty proper subsets of X, and A 6= B,
then the topologies
τA = {X, A, ∅}; τB = {X, B, ∅}
are not comparable.

1.2 NEIGHBOURHOODS

1.2.1 Definition. Let (X, τ ) be a topological space and x ∈ X. A set A ⊂ X is called a

neighbourhood of x if there is U ∈ τ such that x ∈ U ⊂ A.
Thus, each open set is a neighbourhood of its points. Moreover, if A a neighbourhood
of x, and B ⊃ A, then B is also a neighbourhood of x, and the intersection of a finite
number of neighbourhoods of x is a neighbourhood of x.

. Luong Ha - HUCE
88 Chapter 5. Topological Spaces

1.2.2. Interior points and Boundary points.

Let A ⊂ (X, τ ) and x ∈ X.
i) The point x ∈ X is called an interior point of A if A is a neighbourhood of x.
ii) The point x is called a boundary point of A if each neighbourhood of x has the
nonempty intersections with A and with the complement of A.
Note that it is similar to the case of metric spaces, for a point x ∈ X and a set A ⊂ X,
there is one and only one situation happening.
1) There exists a neighbourhood U of x with U ⊂ A (that is, U ∩ (X \ A) = ∅). This
means that x is an interior point of A.
2) There exists a neighbourhood U of x with U ⊂ X \ A (that is, U ∩ A = ∅). This
means that x is an interior point of X \ A.
3) Each neighbourhood U of x has the nonempty intersections with A and with the
complement of A (that is, U ∩ (X \ A) 6= ∅ and U ∩ A 6= ∅). In this case, x is a boundary
point of A.
1.2.3 Theorem. A set A ⊂ X is open if and only if A is a neighbourhood of each point
of A, that is, each x ∈ A is an interior point of A.
Proof. Clearly, if A is open, then A is a neighbourhood of each point of A. Conversely,
for each x ∈ A, there exists an open set Ux such that x ∈ Ux ⊂ A. It follows that
S
A= Ux , hence, A is open.¥
x∈A

1.3 CLOSED SETS

1.3.1 Definition.
A set F in a topological X is called closed if the complement X \ F is open.
It follows from the definition that a set G ⊂ X is open if and only if the complement
of G is closed. Moreover, by taking complements, the following properties of closed sets
follow directly from the definition:
1.3.2 Theorem. Let X be a topological space. Then
1) The sets ∅ and X are closed.
2) The intersection of an arbitrary family of closed sets is closed.
3) The union of any finite family of closed sets is closed.

1.4 INTERIOR AND CLOSURE

1.4.1 Interior. Let X be a topological space and A ⊂ X. The interior of A is defined

o
as the largest open set contained in A, denoted by A or int(A).
Thus, the interior of A is the union of all open sets contained in A. Clearly, a set A

. Luong Ha - HUCE
§ 1. Introduction to Topological Spaces 89

o
is open if and only if A = A.
In the next sections, several theorems and properties are formulated without proof,
because the proof is very similar to the case of metric spaces and we just note that the
role of open balls will be replaced by open sets.
1.4.2 Theorem. Let A and B be two subsets of a topological space X. Then
o o
1) if A ⊂ B, then A ⊂ B;
o o
2) int(A) = A;
o o
3) int(A ∩ B) = A ∩ B;
o o
4) int(A ∪ B) ⊃ A ∪ B
e) The interior of A is the set of all interior points of A.
1.4.3 Closure.
Let X be a topological space and A ⊂ X. The closure of A is defined as the smallest
closed set containing A, denoted by A or cl(A).
By definition, the closure of A is the intersection of all closed sets containing A, and
a set A is closed if and only if A = A.
1.4.4 Theorem. Let A and B be two subsets of a topological space X. Then
1) If A ⊂ B, then A ⊂ B;
2) cl(A) = A;
3) A ∩ B ⊂ A ∩ B;
4) A ∪ B = A ∪ B.

1.5 ADHERENT POINTS - ACCUMULATION POINTS

1.5.1 Definition. Let X be a topological space and A ⊂ X.

i) A point x ∈ X is called an adherent point of A if each neighbourhood of x has
nonempty intersection with A.
ii) A point x ∈ X is called an accumulation point (or a limit point) of A if each
neighbourhood of x has nonempty intersection with A \ {x}. In other words, a point
x is an adherent point of A if and only if each neighbourhood of x contains at least one
point of A; and x is an accumulation point of A if and only if each neighbourhood of x
contains at least one point of A different from x.
1.5.2 Theorem. Let A be a subset of a topological space X.
1) The closure of A is the set of all adherent points of A.
2) The set A is closed if and only if it contains all of its adherent points (or all of its
accumulation points).

. Luong Ha - HUCE
90 Chapter 5. Topological Spaces

If we denote by ∂A the set of all boundary points of A, then by definition

∂A = ∂(X \ A) = A ∩ X \ A.
In addition, we may verify that
o
A = A ∪ ∂(A); A = A \ ∂(A).
1.5.3 Example.
1) In an indiscrete topological space X, if A is a nonempty proper subset of X, then
o
A = ∅ and A = X.
o
2) If X is a discrete topological space and A ⊂ X, then A = A = A.
3) Let X be the set of real numbers with the cofinite topology, and let A = N and
B = {1, 2}. In this space, if G is a nonempty open set, then X \ G is finite, and hence,
the set G must be infinite. The family of closed sets in X consists of X and all finite
o
subsets of X. Thus, B = ∅ and B = B.
Since A = N is an infinite set, there is only closed set which contains A, it is X itself.
Thus, A = X. On the other hand, suppose that there is a nonempty open set U ⊂ A,
then X \ U ⊃ X \ A. Hence, X \ U is infinite. This contradicts the definition of open sets
o
in X. Consequently, A = ∅.

1.6 DENSE SETS - SEPARABLE TOPOLOGICAL SPACES

1.6.1 Definition. Let A be a subset of a topological space. The set A is called dense in
X if the closure of A is X.
The topological X is called separable if X has an at most countable dense subset.
Thus, a set A ⊂ X is dense in X if V ∩ A 6= ∅ for all nonempty open sets V of X.
1.6.2 Example.
1) Let X be a discrete topological space. Suppose that X is uncountable. If A ⊂ X
is at most countable, then A = A 6= X. Thus, X is not separable.
2) Let X be an infinite set equiped with the cofinite topology. Then X has a countable
subset A. Then A = X, hence, X is separable.

1.7 SUBSPACES

1.7.1 Definition. Let (X, τ ) be a topological space and Y is a nonempty subset of X.

Let τY = {G ∩ Y : G ∈ τ }. Then τY is a topology in Y and (Y, τY ) is called a subspace
of (X, τ ). The topology τY is called the induced topology of τ on Y .
1.7.2 Proposition.
1) A set A ⊂ Y is open in Y if and only if there is an open set G in X such that
A=G∩Y.

. Luong Ha - HUCE
§ 1. Introduction to Topological Spaces 91

2) If Y is a subspace of X, and Z is a subspace of Y , then Z is a subspace of X.

3) Each open set in Y is open in X if and only if the set Y is open in X.
The proof is straightforward from the definition of a subspace.
1.7.3 Proposition.
1) A set A ⊂ Y is closed in Y if and only if there is a closed set F in X such that
A=F ∩Y.
2) Each closed set in Y is closed in X if and only if the set Y is closed in X.
Proof.
1) Let A ⊂ Y be a closed set in Y . The set Y \ A is open in Y , there exists an open
set G in X such that Y \ A = Y ∩ G. Hence
A = Y \ (Y \ A) = Y \ (Y ∩ G)
= Y \ G = Y ∩ (X \ G).
Let F = X \ G. We have F is a closed set in X and A = F ∩ Y .
Conversely suppose A = Y ∩ F , where F is closed in X. Then

Y \ A = Y \ (Y ∩ F ) = Y \ F = Y ∩ (X \ F ).

The set X \ F is open in X, hence Y \ A is open Y . Therefore A is closed in Y .

The second statement is clear.¥
The following theorem indicates the relationship between the closure of a set A ⊂ Y
in the topological space X and that in the subspace Y .
∼
1.7.4 Theorem. Let Y be a subspace of a topological space X and A ⊂ Y . Denote by A
∼
the closure of A in the subspace Y . Then A = A ∩ Y , where A is the closure of A in X
as usual.
Proof. The set A is closed in X, hence, A ∩ Y is closed in Y and A ⊂ A ∩ Y . By
∼
definition, it follows that A ⊂ A ∩ Y .
∼
On the other hand, since A is closed in Y , there is a closed set F of X such that
∼ ∼
A = F ∩ Y . Hence A ⊂ F . Thus, A ⊂ F , and A ⊂ F = F . Therefore, A ∩ Y ⊂ F ∩ Y ,
∼ ∼
that is, A ∩ Y ⊂ A. Consequently, A = A ∩ Y .¥
Note that a similar result does not hold for the interior of A. For instance, consider
the subspace Y = [0; 1) ∪ {2} of R with the usual topology. Let A = [0; 1) ⊂ Y . Then
the interior of A in R is (0; 1), while the interior of A in Y is A since A is open in Y .

PROBLEMS
5.1 Let A and B be two subsets of a topological space X. Suppose that A is open and
A ∩ B = ∅. Show that int(A) ∩ int(B) = ∅.
5.2 Let A be a dense subset of a topological space X and U ⊂ X is open. Show that

U = U ∩ A.

. Luong Ha - HUCE
92 Chapter 5. Topological Spaces

5.3 Let A be a subset of a topological space X. Show that

o
X \A=X \A
o
∂(A) = A \ A.

5.4 Let τ = {S ⊂ R : 0 ∈ / S} ∪ {R}. Show that τ is a topology on R. Find the interior

and the closure of the sets A = (1; 2) and B = (−1; 1).
5.5 Let X be the set of natural numbers. For each n ∈ X, set Sn = {k ∈ X : k ≥ n}.
Show that the family τ = {Sn : n ∈ X} ∪ {∅} is a topology on X. Find the interior and
the closure of the sets A = {100} and B consisting of all even naturals.
5.6 Let X be the set of real numbers with the cofinite topology. Let A = N and B = {1, 2}.
Find the interior, the closure, and the boundary of the sets A and B.
5.7 Let X be the set of real numbers with the topology τ = {∅, [0; 1], X}. Let A = [−1; 1]
and B = [2; 3]. Find the interior, the closure, and the boundary of the sets A and B.
T
5.8 Let (τi )i∈I be a family of topologies on a set X. Show that τi is a topology on
i∈I
X. Take an example for showing that the union of two topologies on X may not be a
topology on X.
5.9 Let f : P(X) → P(X) be a mapping satisfying:
i) f (∅) = ∅;
ii) A ⊂ f (A) for each A ⊂ X;
iii) f (A ∪ B) = f (A) ∪ f (B) for any A, B ⊂ X;
iv) f (f (A)) = f (A) for each A ⊂ X.
Show that the family τ = {A ⊂ X : f (A) = A} is a topology on X and A = f (A) for
each A ⊂ X.
The mapping f is called the closure operator.
5.10 Let f : P(X) → P(X) be a mapping satisfying:
i) f (X) = X;
ii) f (A) ⊂ A for each A ⊂ X;
iii) f (A ∩ B) = f (A) ∩ f (B) for A, B ⊂ X;
iv) f (f (A)) = f (A) for each A ⊂ X.
o
Show that the family τ = {A ⊂ X : f (A) = A} is a topology on X and f (A) = A for
each A ⊂ X.
The mapping f is called the interior operator.
5.11 Let d1 and d2 be two metrics on a set X. Suppose that there is c > 0 such that

d1 (x, y) ≤ c d2 (x, y)

for all x, y ∈ X. Show that the topology generated by d1 is weaker than that of d2 .

. Luong Ha - HUCE
§ 2. Basis for a Topology 93

1.12 Let (X, τ ) be a topological space and A ⊂ X. Show that ∂(A) = ∅ if and only if A
is both open and closed.
5.13 Let Y be a subspace of a topological space X. Denote by int(B), intY (B) the
interior of a set B ⊂ Y in X and in Y , respectively. Show that int(B) ⊂ intY (B).
5.14 Let Y and Z be subspaces of a topological space X with X = Y ∪Z. Let M ⊂ Y ∩ Z.
Show that if M is open (closed, res.) in both spaces Y and Z, then M is open (closed,
res.) in X.

§2 BASIS FOR A TOPOLOGY

2.1 BASIS FOR A TOPOLOGY AND PROPERTIES

By definition, a topology is completely defined if we can describe the entire family

τ of open sets. However in many cases, we need only a small family of subsets of X to
determine the topology τ .
2.1.1 Definition. Let (X, τ ) be a topological space. A nonempty subfamily B of τ is
called a basis for the topology τ if each element of τ is a union of some elements of B.
In this case, each element of B is an open set. Thus, each open set G in X can be
expressed as a union of basis elements. This expression for G is not unique. Thus, the
use of the term ”basis” in topology differs radically from its use in linear algebra, where
the the representation of a given vector x is unique. From the definition, it follows that
2.1.2 Theorem. A family B ⊂ τ is a basis for the topology τ if and only if for each
A ∈ τ and each x ∈ A, there exists an element U ∈ B such that x ∈ U ⊂ A.
Proof. Let B be a basis for (X, τ ). If A ∈ τ , then there is B 0 ⊂ B such that
S
A = {B : B ∈ B 0 }. Hence, for each x ∈ A, there exists B ∈ B such that x ∈ B ⊂ A.
Conversely, suppose that for each x ∈ A there is Ux ∈ B such that x ∈ Ux ⊂ A. Then
S
A = {Ux : x ∈ A}, that is, each element of τ is a union of some elements of B. By
definition, B is a basis for τ .¥
Note that every topological space (X, τ ) has at least one basis, for instance, τ itself.
2.1.3 Example.
1) Let (X, d) be a metric space and let A be an open set in X. Then for each x ∈ A,
S
there is rx > 0 such that B(x, rx ) ⊂ A. Hence, A = B(x, rx ). This means that the
x∈A
family of all open balls of X is a basis for the topology generated by the metric d on X.
2) Each open ball on R is an open interval of the form (a; b) with a < b. Thus, the
family of open intervals of R is a basis for the usual topology on R.
3) For any set X, the family of all one-point subsets of X is a topology for the discrete
topology on X.
S
2.1.4 Theorem. A family B of sets is a basis for some topology on the set X = {B :

. Luong Ha - HUCE
94 Chapter 5. Topological Spaces

B ∈ B} if and only if for any U, V in B and for any x ∈ U ∩ V , there exists W ∈ B such
that x ∈ W ⊂ U ∩ V .
Proof. Let B be a basis for a topology τ on the set X. Consider U, V in B and
x ∈ U ∩ V . Then U ∩ V belongs to τ , by Theorem 2.1.2, there is W ∈ B such that
x∈W ⊂U ∩V.
Conversely, suppose that B be a family of subsets of X satisfying the properties in
Theorem 2.1.4. Let τ be the family of all subsets of X expressed as a union of elements
of B. Clearly, X and ∅ are in τ , and the union of any subfamily of τ is in τ . Let U, V ∈ τ .
Then there are B1 ⊂ B, B2 ⊂ B such that
[ [
U= {U ∗ : U ∗ ∈ B1 }, V = {V ∗ : V ∗ ∈ B2 }.

Consequently,
[ [
U ∩V = (U ∗ ∩ V ∗ ).
U ∗ ∈B1 V ∗ ∈B2

If x ∈ U ∩ V , then x ∈ U and x ∈ V , hence, there are elements U1 , V1 in B such that

x ∈ U1 ⊂ U , and x ∈ V1 ⊂ V . Thus, x ∈ U1 ∩ V1 ⊂ U ∩ V . By hypothesis, there is
Wx ∈ B with x ∈ Wx ⊂ U1 ∩ V1 . Hence
[
U ∩V = {Wx : x ∈ U ∩ V },

so U ∩ V ∈ τ .
In short, τ is a topology on X and it is clear that B is a basis for τ .¥
2.1.5 Example.
1) Let X be the plane, and for each real number x, let Lx be the vertical line passing
through the point (x, 0). Define B = {Lx : x ∈ R}. Since Lx ∩ Ly = ∅ for x 6= y, B is a
basis for a topology on X. If A = {z = (x, y) : |z| < 1}, then in the topology for which
o
B is a basis, we have A = {(x, y) : 0 < x < 1}, A = ∅.
2) For any two natural numbers m, n, let

B(m, n) = {km + n : k ∈ Z}.

Then the family B = {B(m, n) : m, n ∈ N} is a basis for a topology on Z. Indeed, we

have Z = ∪{B(m, n) : m, n ∈ N}. Let x ∈ B(m, n) ∩ B(p, q). We can verify that

x ∈ B(mp, x) ⊂ B(m, n) ∩ B(p, q).

2.1.6 Theorem. Let B be a countable basis for a topological space X. Then (X, τ ) is
separable.
Proof. Let B = {B1 , B2 , . . . , Bn , . . . } ⊂ τ . Choose a point xn ∈ Bn for each n ∈ N.
Then the set A = {xn : n ∈ N} is either finite or countable. Let x ∈ X and let U be an
open set containing x. There is Bn ∈ B such that x ∈ Bn ⊂ U . Hence, xn ∈ Bn ∩ U . But
xn ∈ A so A ∩ U 6= ∅. This means that A is dense in X, hence, X is separable.¥

. Luong Ha - HUCE
§ 2. Basis for a Topology 95

2.2 LINDELÖF SPACES

2.2.1 Definition.
A topological space X is called a Lindelöf space if each open cover of X contains a
subcover which is either finite or countable.
2.2.2 Theorem. Every topological space having a countable basis is a Lindelöf space.
Proof. Suppose B = {B1 , B2 , . . . } be a countable basis for X. Let U be an open
cover of X. Sine each U ∈ U is open, U is the union of a subfamily of B. Hence X is
covered by a subfamily B 0 = {Bk1 , Bk2 , . . . } of B with B 0 is finite or countable. Let Uki be
the element of U containing Bki . Then the family (Uki )i is finite or countable and covers
X.¥
Note that a Lindelöf space does not necessarily have a countable basis. Indeed, let X
be the set R with the cofinite topology. Let U be an arbitrary open cover of X and pick
Uo ∈ U. Then X \ Uo = {x1 , x2 , . . . , xn }. For each i = 1, 2, · · · , n, there exists Ui ∈ U
S
n
such that xi ∈ Ui . Hence X = Ui . Thus, X is a Lindelöf space.
i=o
Suppose that X has a countable basis B = {Bn : n ∈ N}. We will show that
T
∞ S
∞
Bn 6= ∅. If this intersection is empty, then X = (X \ Bn ). Each set X \ Bn is is
n=1 n=1
finite, the left-hand side is countable. It contradicts the fact that X is uncountable.
T
∞ T
∞
Thus, there must be an element xo ∈ Bn . Suppose that there is x ∈ Bn and
n=1 n=1
x 6= xo . The set X \ {x} is open, hence it is the union of some elements of B, say,
S S
X \ {x} = Bni , I ⊂ N. It follows that x ∈/ Bni . This is a contradiction since
i∈I i∈I
T
∞ T
∞
x∈ Bn and it shows that Bn = {xo }.
n=1 n=1
S
∞
Therefore X \ {xo } = (X \ Bn ). We have a contradiction since the cardinalities of
n=1
both sides are not equal. This means that X cannot have a countable basis.

PROBLEMS

5.15 Show that the family of all open intervals with rational endpoints is a basis for the
usual topology on R.
5.16 Let U = {[a; b) : a, b ∈ R, a < b}. Show that U is a basis for a topology τ on R.
Compare this topology with the usual topology on R. (This topological space is called
the Sorgenfrey line.)
5.17 Let τ1 = {U ⊂ R : 0 ∈ U }∪{∅}. Show that τ1 is a topology on R and the topological
space (R, τ1 ) is separable but it does not have a countable basis.
5.18 Let X be a topological space, and let B be a basis for the topology. Show that
a point x is the accumulation point of a set A ⊂ X if and only if every member of B

. Luong Ha - HUCE
96 Chapter 5. Topological Spaces

containing x meets A in a point distinct from x.

5.19 Let B be a basis for the topology on X. Then the family
BY = {B ∩ Y : B ∈ B}
is a basis for the induced topology on Y .
5.20 Show that each closed subspace of a Lindelöf space is a Lindelöf space.
5.21 Let X be a metric space. Prove that the following statements are equivalent:
1) X has a countable basis.
2) X is a Lindelöf space.
3) X is separable.
5.22 Let X be a topological space with a countable basis. Let A ⊂ X be an uncountable
set. Show that there is a point in A which is an accumulation point of A.

§3 CONTINUOUS MAPPINGS

The notion of continuous function on R and continuous mappings between metric

spaces are studied. In this section, we will state a definition of continuity that will
include all these as special cases, and we will study various properties of continuous
mappings. Many of these properties are direct generalizations of things we have known
about continuous functions (mappings).

3.1 CONTINUITY OF A MAPPING

3.1.1 Definition. Let (X, τ ) and (Y, U) be topological spaces. A mapping f : X → Y is

called continuous at a point xo ∈ X if for each neighbourhood V of f (xo ), there exists a
neighbourhood U of xo such that f (U ) ⊂ V .
We say f is continuous on X (or simply continuous) if it is continuous at any x ∈ X.
Example.
1) If X is a discrete space, then every maping f : X → Y is continuous. Indeed,
let xo ∈ X and let V be a neighbourhood of f (xo ). Then f −1 (V ) is an open set in X
containing xo . Let U = f −1 (V ), we have f (U ) ⊂ V . Hence, f is continuous at xo .
2) If Y is an indiscrete space, then any mapping f : X → Y is continuous. Pick
xo ∈ X. The only neighbourhood of f (xo ) if Y . Since f −1 (Y ) = X is a neighbourhood of
xo and f (X) ⊂ Y , f is continuous at xo .
3) Each constant mapping f : X → Y (f (x) = b ∈ Y for all x ∈ X) is continuous. In
this case, we can choose U = X for any neighbourhood of f (xo ).
4) If Y is a subspace of X, then the inclusion j : Y → X defined by j(x) = x for each
x ∈ Y , is continuous.

. Luong Ha - HUCE
§ 3. Continuous mappings 97

5) If f : X → Y is continuous, and if A is a subspace of X, then the restricted mapping

f|A : A → Y is continuous.
6) Let f : X → Y be continuous. If Z is a space having Y as a subspace, then the
mapping h : X → Z obtained by expanding the range of f is continuous.
The results of 4), 5) and 6) follow directly from the definition of continuity.
3.1.2 Theorem. Let X and Y be topological spaces and let f : X → Y be a mapping.
Then the following propositions are equivalent:
1) f is continuous.
2) f −1 (G) is open for each open set G in Y . .
3) f −1 (F ) is closed for each closed set F in Y .
Proof. 1) ⇒ 2). Let G be an open set in Y and x ∈ f −1 (G). Then f (x) ∈ G. Since f
is continuous at the point x, and G is a neighbourhood of f (x), there is a neighbourhood
of x such that f (U ) ⊂ G. Thus, x ∈ U ⊂ f −1 (G), and hence, f −1 (G) is open.
2) ⇔ 3). Note that for any mapping f , we have
X \ f −1 (F ) = f −1 (Y \ F ).
Hence, if F is closed in Y , then by 2), f −1 (Y \ F ) is open, that is, f −1 (F ) is closed.
Similarly, it is easy to show that 3) implies 2).
2) ⇒ 1). We only need to prove that f is continuous at an arbitrary point of X. Let
x ∈ X and V is a neighbourhood of f (x). Then there is an open set G in Y such that
x ∈ G ⊂ V . By hypothesis, U = f −1 (G) is open, and we have f (U ) ⊂ G ⊂ V . It is clear
that U is a neighbourhood of x. Hence, f is continuous at x.¥
3.1.3 Corollary. Let f : X → Y be a mapping between two topological spaces. Suppose
that B is a basis for the topology on Y . Then f is continuous if and only if f −1 (B) is
open for any B ∈ B.
Proof. The necessary condition follows directly from Theorem 3.1.2. Conversely, if G
S S −1
is an open set in Y , then G = {B : B ∈ B 0 , B 0 ⊂ B}. Therefore f −1 (G) = f (B).
B∈B0
−1 −1
Since each set f (B) is open, the set f (G) is open. By Theorem 3.1.2, f is continuous.¥
3.1.4 Theorem. Let X, Y, Z be topological spaces and let f : X → Y, g : Y → Z be
continuous mappings. Then the composition g ◦ f : X → Z is continuous.
Proof. Let G be an arbitrary open set in Z. Then (g ◦ f )−1 (G) = f −1 (g −1 (G)). The
mapping g is continuous, g −1 (G) is open in Y . From the continuity of f , it follows that
f −1 (g −1 (G)) is open in X. Hence, g ◦ f is continuous.¥
3.1.5 Example.
1) Let τ be the topology on R generated by the basis
B = {[a; b) : a, b ∈ R, a < b}
and let U be the usual topology on R.

. Luong Ha - HUCE
98 Chapter 5. Topological Spaces

Consider f : (R, τ ) → (R, U) defined by


x if x < 0
f (x) =
x + 1 if x > 0.

The family of open intervals is a basis for the usual topology, it suffices to prove f −1 (a; b)
is open in (R, τ ). Note that U ⊂ τ .
If a ≥ 1, then f −1 (a, b) = (a − 1; b − 1).
If 0 ≤ a < 1, then f −1 (a; b) = [0; b − 1) when b > 1 and equals ∅ when b ≤ 1.
If a < 0, then f −1 (a; b) = (a; b − 1) when b > 1. If 0 ≤ b < 1, f −1 (a; b) = (a; 0) and if
b < 0, we have f −1 (a; b) = (a; b). In any case, the set f −1 (a; b) belongs to τ . Thus, the
mapping f is continuous.
2) In the above example, if we consider the mapping of the same definition f : (R, U) →
(R, τ ), then f is not continuous. Indeed, the set A = [−2; −3) is open in (R, τ ) but
f −1 (A) = [−2; −3) is not open in (R, U).
3) Let τ1 and τ2 be topologies on a set X. Then the identity mapping id : (X, τ1 ) →
(X, τ2 ) is continuous if and only if τ2 ⊂ τ1 .
3.1.6 Theorem. Let X be a topological space. Let f and g be continuous mappings from
X into R and let k ∈ R. Then the mappings f + g, kf, f g, |f | and max(f, g), min(f, g)
1
are continuous . If the mapping is defined, then it is continuous.
f
The proof is similar to the case of a metric space.

3.2 HOMEOMORPHISMS

3.2.1 Definition. A mapping f from a topological space X into a topological space Y

is called a homeomorphism if f is bijective, continuous, and its inverse f −1 is continuous.
Then we say that X is homeomorphic to Y .
3.2.2. Remark.
1) The relation homeomorphic to is an equivalence relation between topological spaces.
If two topological spaces are homeomorphic, we can identify them.
2) Let τ1 and τ2 be topologies on a set X. Then the identity mapping id : (X, τ1 ) →
(X, τ2 ) is a homeomorphism if and only if τ1 = τ2 .
3) The function f : (−1; 1) → R defined by
x
f (x) =
1 − x2
is a homeomorphism.
First we can show that f is a bijection and its inverse g is defined by
2x
g(x) = .
1 + (1 + 4x2 )1/2

. Luong Ha - HUCE
§ 3. Continuous mappings 99

From the continuity of the elementary functions and the square-root function, it follows
that both f and g are continuous.
A property which does not change under a homeomorphism is called a topological
property. For instance, the closedness or the openness of a set is a topological property,
but the completeness of a metric space is not a topological property, since an incomplete
metric space can be homeomorphic to a complete metric space, say (−1; 1) and R with
the usual topology.

PROBLEMS
5.23 Let f : X → Y be a mapping between two topological spaces. Show that the
following propositions are equivalent:
a) f is continuous.
b) For each A ⊂ X, f (A) ⊂ f (A).
c) For each B ⊂ Y, f −1 (B) ⊂ f −1 (B).
o
d) For each B ⊂ Y, f −1 (B) ⊂ int(f −1 (B)).
5.24 A mapping f : X → Y is called an open mapping if the image under f of any open
set in X is open in Y . Similarly, f is called a closed mapping if the image under f of any
closed set in X is closed in Y .
Let f be a continuous bijection from a topological space X onto a topological space
Y . Show that the following propositions are equivalent:
1) f is a homeomorphism.
2) f is an open mapping.
3) f is a closed mapping.
5.25 Let X be a cofinite topological space. Let f : X → R be an arbitrary mapping.
Show that if A is a proper closed subset of X, then f (A) is closed in R. Take an example
to show that the set f (X) may not be closed.
5.26 Let X be a cofinite topological space. Let f : X → R be a continuous function.
Show that f is a constant function.
5.27 Let X be a topological space. Suppose that there exist two subsets A, B of X such
that A ∪ B = X, A ∩ B = ∅, A = B = X. Show that the mapping f : X → R defined by


0 if x ∈ A
f (x) =
1 if x ∈ B

is not continuous at any x ∈ X.

. Luong Ha - HUCE
100 Chapter 5. Topological Spaces

5.28 Let f : R → R be the function defined by




0 if x ≤ 0

f (x) = x if 0 < x < 1



1 if x ≥ 1.

Show that f is a closed mapping but not an open mapping.

5.29 Let X be a topological space and X can be written as the union of open sets (Ui )i∈I .
Let f : X → Y be a mapping such that f|Ui is continuous for each i ∈ I. Show that f is
continuous.
5.30 Suppose f : X → Y is a continuous mapping from one topological space onto
another. Let B be a basis for the topology on X. Show that the collection
f (B) = {f (B) : B ∈ B}
is a basis for the topology on Y if f is open.
5.31 (The pasting lemma) Let X and Y be topological spaces and X = A ∪ B, where
A and B are closed in X. Suppose that f : A → Y and g : B → Y be continuos mappings.
If f (x) = g(x) for every x ∈ A ∩ B, then the mapping h : X → Y defined by h(x) = f (x)
if x ∈ A, and h(x) = g(x) if x ∈ B, is continuous.
5.32 Let (fn )n be a sequence of real-valued functions defined on a topological space X
such that lim fn (x) exists for each x ∈ X. Then a new function f can be defined by
n
f (x) = lim fn (x) for each x ∈ X. The sequence (fn )n is said to converge pointwise to
n
f (or f is the pointwise limit of (fn )n ) and is written symbolically as fn → f . In other
words, fn → f if for each ε > 0, there exists some no (depending upon both ε and x) such
that |fn (x) − f (x)| < ε for all n ≥ no .
A sequence (fn )n is said to converge uniformly on X to a function f if for each ε > 0,
there exists some no (depending only upon ε) such that |fn (x)−f (x)| < ε for all n ≥ no and
all x ∈ X. It is easy to see that uniform convergence implies pointwise convergence. Note
that the pointwise limit of a sequence of continuous functions need not be a continuous
function.
Let (fn )n be a sequence of continuous functions on a topological space X. Show that
if (fn )n converges uniformly to f on X, then f is continuous.
5.33 Let X be a topological space. Let f : X → R be a mapping.
The function f is called lower semicontinuous on X if {x ∈ X : f (x) > a} is open for
any a ∈ R.
The function f is called upper semicontinuous on X if {x ∈ X : f (x) < a} is open for
any a ∈ R. . Show that
1) A function f is continuous if and only if it is both upper semicontinuous and lower
semicontinuous.
2) Characteristic functions of open sets are lower semicontinuous; characteristic func-
tions of closed sets are upper semicontinuous.

. Luong Ha - HUCE
§ 4. Product Space - Quotient Space 101

3) The supremum of any collection of lower semicontinuous functions is lower semi-

continuous. The infimum of any collection of upper semicontinuous functions is upper
semicontinuous.
(Let A ⊂ X. The characteristic function of the set A, denoted by 1A is defined as
1A (x) = 1 if x ∈ A and 1A (x) = 0 if x ∈ X \ A.)

§4 PRODUCT SPACE - QUOTIENT SPACE

4.1 PRODUCT SPACE

4.1.1 Product topology.

Q
Let (Xα , τα )α∈I be a family of topological spaces, and let X = α∈I Xα be the carte-
sian product of the family (Xα )α∈I . Consider the projection pα : X → Xα for each α ∈ I.
If X is equipped with the discrete topology, then all the mappings pα are continuous.
We shall show that there is a weakest topology on X such that all the mappings pα are
continuous.
T
n
For each n ∈ N, let G = p−1
αi (Gαi ), where Gαi ∈ ταi .
i=1 S
Let B be the family of the sets G constructed as above. We first have X = {G :
G ∈ B}. Clearly, the intersection of any two members of B is a member of B, hence B is
a basis for a topology τ on X. By definition, if Gα ∈ τα , then p−1 α (Gα ) belongs to B, so
each mapping pα is continuous.
Suppose that τ 0 is a topology on X such that each mapping pα is continuous. Then
T
n
for any G = p−1 −1 0 0 0
αi (Gαi ) in B, each set pαi (Gαi ) is in τ . Thus, G ∈ τ , and τ ⊂ τ .
i=1
Consequently, τ is the weakest topology on X such that all the mappings pα are
continuous. It is called the product topology on X, and with this topology, X is called a
product space.
Q
4.1.2 Theorem. Let τ be the product topology on X = α∈I Xα and let f : X → Y ,
where Y is an arbitrary topological space. Then f is continuous if and only if pα ◦ f is
continuous for each α ∈ I.
Proof. The necessary condition follows directly from the fact that the composite of
two continuous mappings is continuous. Conversely, let G be any member of the basis
B of τ as above. Then there are α1 , α2 , . . . , αn ∈ I and Gα1 , Gα2 , . . . , Gαn being open in
Xα1 , Xα2 , . . . , Xαn , respectively, such that
n
G = ∩ p−1
αi (Gαi ).
i=1

Hence
n ¡ ¢ n
f −1 (G) = ∩ f −1 p−1 −1
αi (Gαi ) = ∩ (pαi ◦ f ) (Gαi ).
i=1 i=1

From the continuity of the mappings pαi ◦ f , it follows that f −1 (G) is open in Y . Hence,
the mapping f is continuous.¥

. Luong Ha - HUCE
102 Chapter 5. Topological Spaces

Q
From now on, whenever consider the product α∈I Xα , we shall assume it is given the
product topology unless otherwise stated.
Note that the each of the form
n
\
p−1
αi (Gαi ) for n ∈ N, Gαi ∈ ταi
i=1

can be written as
n
\ Y
p−1
αi (Gαi ) = Wα ,
i=1 α∈I

where Wα = Xα for all α 6= α1 , α2 , . . . , αn and Wα = Gαi if α = αi , i = 1, 2, . . . , n.

4.1.3 Corollary. Let X = X1 × X2 × · · · × Xn be the product of a finite family of
topological spaces. Then the collection of all the sets of the form U1 × U2 × · · · × Un ,
where Ui is open in Xi , i = 1, 2, . . . , n, is a basis for the product topology.
Thus, if W is an open set in the product space X × Y , and (x, y) ∈ W , then there
exist open sets U ⊂ X and V ⊂ Y such that (x, y) ∈ U × V ⊂ W .

4.2 QUOTIENT SPACE

4.2.1 Quotient topology.

Let X a topological space and R is an equivalence relation on X. Denote by X/R the
corresponding quotient set and consider the canonical mapping π : X → X/R defined by
∼
π(x) = x (the equivalence class of x).
Let τ be the collection of all subsets G of X/R such that π −1 (G) is open in X. Then
τ is a topology on X/R. Indeed, it is clear that the sets ∅ and X/R are in τ . Since the
preimage of an intersection and of a union are equal to the intersection and the union of
the preimages, respectively, the union of an arbitrary subfamily of τ and the intersection
of two members of τ are in τ . With this topology on X/R, the mapping π is continuous.
Now let τ 0 be a topology on X such that π is continuous. If G ∈ τ 0 , then π −1 (G) is
open in X and by the definition of τ , we have G ∈ τ . Thus, τ 0 ⊂ τ .
Consequently, the topology τ is the strongest topology on X/R such that the mapping
π is continuous, and it is called the quotient topology. In this topology, X/R is called a
quotient space of X
A set G ⊂ X/R is open in the quotient topology if and only if π −1 (G) is open in X.
∼
Note that π −1 (G) = {x ∈ X : x ∈ G}
4.2.2 Theorem. A set F ⊂ X/R is closed if and only if π −1 (F ) is closed in X.
Proof. We see that F is closed if and only if (X/R) \ F is open, that is,

π −1 ((X/R) \ F ) = X \ π −1 (F )

is open in X or π −1 (F ) is closed in X.¥

. Luong Ha - HUCE
§ 5. Axioms of Separation 103

4.2.3. Theorem. Let Y be any topological space and let X/R be the quotient space.
Then a mapping f : X/R → Y is continuous if and only if f ◦ π : X → Y is continuous.
Proof. Evidently, the mapping f ◦ π is continuous if f is continuous. Conversely, let
V be any open set in Y . Then since
(f ◦ π)−1 (V ) = π −1 (f −1 (V ))
is open in X, the set f −1 (V ) ∈ τ . Thus, f is continuous.¥

PROBLEMS
5.34 Let X and Y be two topological spaces, A ⊂ X and B ⊂ Y . Show that the following
equalities hold on the product space X × Y :
o o
int(A × B) =A × B
A × B =A × B.

5.35 Denote by X, Y the set of real numbers with the usual topology and with the
discrete topology, respectively.
© ª
a) Prove that the collection B = (a; b) × {c} : a, b, c ∈ R is a basis for the product
topology on Z = X × Y .
b) Find the interior and the closure of the following sets in Z:

A ={(x, 0) ∈ Z : x ∈ [0; 1)}

B ={(0, y) ∈ Z : y ∈ [0; 1)}
C =[0; 1) × [0; 1).

5.36 Let (Xα )α∈I be a family of topological spaces, and let Y be any topological space.
Suppose that fα : Y → Xα be a mapping for each α ∈ I. Define a mapping
Q
f : Y → X = α∈I Xα
by f (x) = (fα (x))α∈I .
Show that the mapping f is continuous if and only if each fα is continuous.
Q
5.37 Show that each projection pβ : X = α∈I Xα → Xβ is an open mapping. Take an
example for showing that a projection may not be a closed mapping.

§5 AXIOMS OF SEPARATION

We have known that every metric space is a topological space. But many things
can happen in an arbitrary topological space which differ in an essential way from what
happens in a metric space, for instance, a finite subset of a topological space may not
be closed. Hence, it is desirable to specialize the notion of a topological space somewhat
by considering topological spaces more closely resembling metric spaces. This is done by

. Luong Ha - HUCE
104 Chapter 5. Topological Spaces

imposing extra conditions on a topological space. These supplementary conditions are

called axioms of separation.

5.1 AXIOMS OF SEPARATION AND BASIC PROPERTIES

5.1.1 T1 -space. A topological space X is called to satisfy the first axiom of separation if
for any distinct points x, y ∈ X, there are a neighbourhood U of x and a neeighbourhood
V of y such that x ∈/ V and y ∈ / U . Then X is called a T1 -space.
5.1.2 Theorem. A topological space X is T1 if and only if each singleton is closed.
Proof. Let X be a T1 -space and x ∈ X. For y ∈ X \ {x}, then y 6= x, so there is a
neighbourhood U of y such that x ∈ / U . It follows that U ⊂ X \ {x}, and hence, X \ {x}
is open, that is, the set {x} is closed.
Conversely, suppose that each singleton in X is closed. Let x, y ∈ X with x 6= y. Then
X \ {y}, X \ {x} are open. Hence, they are neibhourhoods of x, y, respectively. Clearly,
X \ {x} does not contain x and X \ {y} does not contain y. Thus, the space X is T1 .¥
Example.
1) Each discrete space is T1 .
2) Let X be a cofinite topological space. Then each finite set in X is closed. Hence,
X is a T1 -space.
3) If X is an indiscrete space having more than one point, then X is not a T1 -space
because two distinct points of X have only one neighbourhood X.
4) Let X = {a, b} with the topology τ = {∅, X, {a}}. Then the singleton {a} is not
closed. Thus, the space X is not T1 .

5.1.3 Hausdorff space. A topological space X is called to satisfy the second axiom of
separation if for any two disctinct points x, y ∈ X, there are a neighbourhood U of x and
a neighbourhood V of y such that U ∩ V = ∅. Then X is called a Hausdorff space or a
T2 -space.
Clearly, a T2 -space is a T1 -space.
Example.
1) Every metric space X is T2 . Indeed, let x, y ∈ X with x 6= y. Then the distance
ε ε
d(x, y) = ε > 0. Set U = B(x, ), V = B(y, ), then U is a neighbourhood of x and V is
2 2
a neighbourhood of y with U ∩ V = ∅.
2) Let X be a cofinite topological space. Let A, B be two nonempty open subsets of
X. If A ∩ B = ∅, then X = (X \ A) ∪ (X \ B). This is a contradiction because the
righthand side is finite while the lefthand side is infinite. Thus, A ∩ B 6= ∅, and hence, X
is not T2 .
Topological spaces more general than Hausdorff spaces are rarely used in analysis. In
fact, most of the topological spaces of interests in analysis satisfy a separation condition

. Luong Ha - HUCE
§ 5. Axioms of Separation 105

even stronger than the second axiom of separation.

5.1.4 T3 -space. A topological space X is called a regular space or a T3 -space if X is T1
and for each x ∈ X, for each closed set F in X not containing x, there exist open sets
U, V such that x ∈ U , F ⊂ V and U ∩ V = ∅.
We see that each T3 -space is T2 .
5.1.5 Theorem. Let X be a T1 -space. Then X is regular if and only if for each x ∈ X and
for each open set V containing x, there exists an open set U such that x ∈ U ⊂ U ⊂ V .
Proof. Let X be a regular space. Let x ∈ X and let V be an open set containing x.
Then the set F = X \ V is closed and x ∈ / F . Hence, there exist two open sets U, W so
that x ∈ U, F ⊂ W and U ∩ W = ∅. Thus, U ⊂ X\W. Therefore,
U ⊂ X \ W = X \ W ⊂ X \ F = V.
Conversely, let x ∈ X, and F is a closed set not containing x. The set V = X \ F
is an open set containing x, by hypothesis, there is an open neighbourhood U of x such
that U ⊂ V . Let W = X \ U , then W is open, W ⊃ X \ V = F and U ∩ W = ∅. Hence,
X is regular.¥
Example. Let K = {1/n : n ∈ N}. Denote by RK the set of real numbers with the
topology having as basis all open intervals (a; b) and all sets of the form (a; b) \ K. This
space is Hausdorff, since any two distinct points have disjoint open intervals containing
them.
But it is not regular. The set K is closed in RK , and it does not contain the point
0. Suppose that there exist disjoint open sets U and V containing 0 and K, respectively.
Choose a basis element containing 0 and lying in U . It must be a basis element of the form
(a; b) \ K, since each basis element of the form (a; b) containing 0 intersects K. Choose
1 1
n sufficiently large that ∈ (a; b). Then choose a basis element about contained in
n n
1
V ; it must be a basis element of the form (c; d). Finally, choose z so that z < and
n
1
z > max{c, }. Then x belongs to both U and V , so they are not disjoint. This is a
n+1
contradiction, and it shows that RK is not regular.
5.1.6 Normal space. A topological space X is called a normal space or a T4 -space if X
is T1 and for any two disjoint closed set A, B in X, there exist open sets U, V such that
A ⊂ U , B ⊂ V and U ∩ V = ∅.
Example.
1) Each subset of a discrete space is both open and closed, hence any discrete space
is normal.
2) Each metric space is normal.
Let A and B be two disjoint closed subsets of a metric space X. Then for every x ∈ A,
the distance dx = d(x, B) > 0 and dy = d(y, A) > 0 for every y ∈ B. Let
[ 1 [ 1
U= B(x, dx ), V = B(y, dy ).
x∈A
2 y∈B
2

. Luong Ha - HUCE
106 Chapter 5. Topological Spaces

It is clear that U, V are open and A ⊂ U, B ⊂ V . Suppose that there is a point z ∈ U ∩V .

Then there are points xo ∈ A, yo ∈ B such that
1 1
d(xo , z) <dxo , d(z, yo ) < dyo .
2 2
Without loss of generality, we may assume that dxo ≤ dyo . Then
1 1
d(xo , yo ) ≤ d(xo , z) + d(z, yo ) < dxo + dyo < dyo ,
2 2
that is, xo ∈ B(yo , dyo ). This contradicts the definition of dyo , and shows that U ∩ V = ∅.
Note that if X is a T4 -space, then it is also Ti for i = 1, 2, 3.
5.1.7 Theorem. Let X be a T1 -space. Then X is normal if and only if for any closed
set A ⊂ X and for any open set G ⊃ A, there is an open set U such that

A ⊂ U ⊂ U ⊂ G.

Proof. Let X be a normal space. Consider a closed set A ⊂ X and an open set G ⊃ A.
Let B = X \ G. Then B is closed and A ∩ B = ∅. Hence there exist two disjoint open
sets U, V such that A ⊂ U and B ⊂ V . Since X \ V is closed and U ∩ V = ∅, we have
A ⊂ U ⊂ X \ V . Thus, U ⊂ X \ V = X \ V ⊂ X \ B = G.
Conversely, consider two disjoint closed sets A, B in X. Let G = X \B, then G is open
and G ⊃ A. By hypothesis, there is an open set U ⊂ X such that A ⊂ U ⊂ U ⊂ G. Let
V = X\U . Clearly, V is open and U ∩V = ∅. In addition, we have B = X\G ⊂ X\U = V .
Hence, X is normal.¥
5.1.8 Theorem. Let X be a regular space. If X has a countable basis, then X is normal.
Proof. Let B be a countable basis for the topology on X. Consider two disjoint closed
sets A, B in X. The set X \ B is open and contains A, so X \ B is an open neighbourhood
of each x ∈ A. Since X is regular, by Theorem 5.1.5, for each x ∈ A, there is Vx ∈ B
S
such that x ∈ Vx ⊂ Vx ⊂ X \ B. Thus, A ⊂ Vx . The family B is countable, there
x∈A S
are a finite number or countable sets Vxi = Vi , i ∈ I ⊂ N so that A ⊂ Vi , Vi ∈ B and
i∈I
Vi ∩ B = ∅ for all i ∈ I.
Similarly, since the set X \ A is open and contains B, we can find an at most countable
open sets Wj , j ∈ J ⊂ N such that B ⊂ ∪ Wj , Wj ∈ B and Wj ∩ A = ∅ for all j ∈ J.
j∈J
Let

P1 =V1 , Q1 = W1 \ P1
P2 =V2 \ Q1 , Q2 = W2 \ (P1 ∪ P2 )
···············
³ n−1 ´ n
Pn =Vn \ ∪ Qi , Qn = Wn \ ( ∪ Pi )
i=1 i=1

···············
and P = ∪ Pn , Q = ∪ Qn .
n∈I n∈J

. Luong Ha - HUCE
§ 5. Axioms of Separation 107

Thus, the sets Pn , Qn is open, and hence, P, Q are open. Moreover, Pn ∩ Qm = ∅ for all
m, n. Indeed, for n ≤ m we have
³ m ´
Qm = Wm \ ∪ Pi ⊂ Wm \ Pn .
i=1

If n > m, then ³ n−1 ´

Pn = Vn \ ∪ Qi ⊂ Vn \ Qm ⊂ Vn \ Qm .
i=1
S S
Hence we always have Pn ∩Qm = ∅. Therefore, P ∩Q = ∅ since P ∩Q = (Pn ∩Qm ).
n∈I m∈J
Now we only need to show that A ⊂ P and B ⊂ Q. If x ∈ A, then there is io ∈ I
such that x ∈ Vio . Since Wn ⊂ X \ A, x ∈ / Wj for all j ∈ J. Hence, x ∈/ Qn for all n ∈ J.
³ioS−1 ´
Thus, x ∈ Vio \ Qn , that is, x ∈ Pio , and hence, x ∈ P . Similarly, we get B ⊂ Q.
n=1
Thus, X is normal. The proof of the theorem is complete.¥

PROBLEMS
5.38 Show that the cofinite topology on an infinite set X is the weakest topology on X
such that X is a T1 -space.
5.39 Let f and g be continuous mappings from a topological space X into a Hausdorff
space Y . Show that the set A = {x ∈ X : f (x) = g(x)} is closed.
5.40 Let X be a Hausdorff space and let f : X → X be a continuous mapping such that
f ◦ f = f . Show that f (X) is closed in X.
5.41 Let X be a topological space. Let ∆ = {(x, x) : x ∈ X}. Show that X is T2 if and
only if ∆ is closed in the product space X × X.
5.42 Let (X, τ ) be a topological space, and let R ⊂ X × X is an equivalence relation on
X.
1) Show that if the quotient space X/R is T2 , then R is closed in X × X.
2) Suppose that R is closed in X × X and the canonical mapping π : X → X/R is
open. Show that X/R is a T2 -space.
5.43 Show that the quotient space X/R is T1 if and only if every equivalence class is
closed in X.
T
5.44 Show that a topological space X is T2 if and only if for each x ∈ X, {x} = F,
where F is any closed neighbourhood of x.
5.45 Let f : X → Y be a continuous mapping between two topological spaces. Let
G = {(x, y) ∈ X × Y : y = f (x)}.
1) Show that the subspace G of X × Y is homeomorphic to X.
2) If the space Y is T2 , show that G is closed in X × Y .
5.46 Let f : X → Y , where Y is a Hausdorff space. Show that if f is injective and
continuos, then X is a Hausdorff space.

. Luong Ha - HUCE
108 Chapter 5. Topological Spaces

5.47 Let X be a Hausdorff space. Show that for n distinct points x1 , x2 , . . . , xn in X,

there are n disjoint open sets V1 , V2 , . . . , Vn such that xi ∈ Vi for each i = 1, 2, . . . , n.
Hence, it follows that a finite Hausdorff space must be a discrete space.
5.48 Show that if X is regular, every pair of distinct points of X have neighbourhoods
whose closures are disjoint.
5.49 Show that if X is normal, every pair of disjoint closed sets in X have neighbourhoods
whose closures are disjoint.
5.50 Let X be a normal space and suppose that X = G1 ∪ G2 , where G1 and G2 are open
in X. Show that there exist two closed sets F1 , F2 in X such that F1 ⊂ G1 , F2 ⊂ G2 and
X = F1 ∪ F2 .
5.51 Let f : X → Y be a closed continuous surjective mapping. Show that if X is normal,
then so is Y .
5.52
1) Show that a subspace of a Ti -space (i = 1, 2, 3) is also Ti .
2) Show that a closed subspace of a normal space is normal.
Q
5.53 Show that if each space Xα is a Hausdorff space, then the product space Xα is a
Hausdorff space.

5.2 EXTENSION OF A CONTINUOUS FUNCTION

5.2.1 Theorem (Urysohn lemma). Let X be a normal space; let A and B be disjoint
closed subsets of X. Then there exists a continuous mapping

f : X → [0; 1]

such that f (x) = 0 for every x ∈ A, and f (x) = 1 for every x ∈ B.

Proof. ([11], pp. 207-210) The proof will be divided into three steps.
Step 1. At this step, by using normality, we will construct a certain family (Up )p of
open sets in X, indexed by the rational numbers. Then we use these sets to define the
continuous function f .
Let P be the set of rational numbers in the interval [0; 1]. We shall define, for each
p ∈ P , an open set Up in X, in such a way that whenever p < q, we have U p ⊂ Uq .
Thus, the sets Up will be simply ordered by inclusion by the same way their subscripts
are ordered by the usual ordering in the real line.
Because P is countable, we can use induction to define the sets Up . Arrange the
elements of P in an infinite sequence in some way; for convenience, let us suppose that
the numbers 1 and 0 are the first two elements of the sequence.
Now define the sets Up as follows: First, define U1 = X \ B. Second, since A is a closed
set contained in the open set U1 , we may by normality of X choose an open set Uo such

. Luong Ha - HUCE
§ 5. Axioms of Separation 109

that
A ⊂ Uo and U o ⊂ U1 .
In general, let Pn denote the set consisting of the first n rational numbers in the sequence.
Suppose that Up is defined for all rational numbers p belonging to the set Pn , satisfying
condition
(∗) p < q ⇒ U p ⊂ Uq .
Let r denote the next rational number in the sequence; we wish to define Ur .
Consider the set Pn+1 = Pn ∪ {r}. It is a finite subset of the interval [0; 1], and, as
such, it has a simple ordering derived from the usual order relation < on the real line. In
a finite simply ordered set, every element (other than the smallest and the largest) has an
immediate predecessor and an immediate successor. The number 0 is the smallest and 1
is the largest element, of the simply ordered set Pn+1 , and r is neither 0 nor 1. So r has
an immediate predecessor p ∈ Pn+1 and an immediate successor q ∈ Pn+1 . The sets Up
and Uq are already defined, and U p ⊂ Uq by the induction hypothesis. Using normality
of X, we can find an open set Ur in X such that

U p ⊂ Ur and U r ⊂ Uq .

We assert that (∗) now holds for every pair of elements of Pn+1 . If both elements lie in
Pn , (∗) holds by the induction hypothesis. If one of them is r and the other is a point
s ∈ Pn , then either s ≤ p, in which case

U s ⊂ U p ⊂ Ur ,

or s ≥ q, in which case
U r ⊂ Uq ⊂ Us .
Thus, for every pair of elements of Pn , relation (∗) holds.
By induction, we have Up for all p ∈ P .
Step 2. Now we have defined Up for all rational numbers p in the interval [0; 1]. We
extend this definition to all rational numbers in R by defining

Up = ∅ if p < 0,

Up = X if p > 1.
We can check that it is still true for any pair of rational numbers p and q,

p < q ⇒ U p ⊂ Uq .

Step 3. Given a point x ∈ X, let us define Q(x) to be the set of those rational numbers
p such that the corresponding open sets Up contain x:

Q(x) = {p : x ∈ Up }.

. Luong Ha - HUCE
110 Chapter 5. Topological Spaces

This set contains no number less than 0, since no x is in Up for p < 0. And it contains
every number greater than 1, since every x is in Up for p > 1. Therefore, Q(x) is bounded
below, and its greatest lower bound is a point of the interval [0; 1]. Define

f (x) = inf Q(x) = inf {p : x ∈ Up }.

Step 4. We show that f is the desired function. If x ∈ A, then x ∈ Up for every p ≥ 0, so

that Q(x) equals the set of all nonnegative rationals, and f (x) = inf Q(x) = 0. Similarly,
if x ∈ B, then x ∈ Up for no p ≤ 1, so that Q(x) consists of all rational numbers greater
than 1, and f (x) = 1.
The only hard part is to show that f is continuous. For this purpose, we first prove
the following elementary facts:
(1) x ∈ U r ⇒ f (x) ≤ r.
(2) x ∈ / Ur ⇒ f (x) ≥ r.
To prove (1), note that if x ∈ U r , then x is in Us for any s > r. Therefore, Q(x) contains
all rational numbers greater than r, so that by definition we have

f (x) = inf Q(x) ≤ r.

To prove (2), note that if x ∈

/ Ur , then x is not in Us for any s < r. Therefore, Q(x)
contains no rational numbers less than r, so that

f (x) = inf Q(x) ≥ r.

Now we prove continuity of f . Given a point xo in X and an open interval (c; d) in R

containing the point f (xo ), we wish to find a neighbourhood U of xo such that f (U ) ⊂
(c; d). Choose rational numbers p and q such that

c < p < f (xo ) < q < d.

We assert that the open set

U = Uq \ U p

is the desired neighbourhood of xo .

First, we note that xo ∈ U . For the fact that f (xo ) < q implies by condition (2) that
xO ∈ U q , while the fact that f (xo ) > p implies by (1) that xo ∈
/ U p.
Second, we show that f (U ) ⊂ (c; d). Let x ∈ U . Then x ∈ Uq ⊂ U q , so that f (x) ≤ q,
by (1). And x ∈ / U p , so that x ∈
/ Up and f (x) ≥ p, by (2). Thus, f (x) ∈ [p; q] ⊂ (c; d), as
desired.¥
5.2.2 Theorem (Tietze - Urysohn). Let M be a closed subspace of a normal space
X. Let f : M → R is a bounded continuous function. Then there exists a continuous
function F : X → R such that
1) F (x) = f (x) for all x ∈ M ;

. Luong Ha - HUCE
§ 5. Axioms of Separation 111

2) sup|F (x)| = sup|f (x)|.

x∈X x∈M

Proof. Let c = sup |f (x)|. If c = 0, let F (x) = 0 for all x ∈ X. Then the function F
x∈M
is continuous on X and satisfies two conditions above.
Suppose that c > 0. Let
c
A = {x ∈ M : f (x) ≤ − }
3
c
B = {x ∈ M : f (x) ≥ }.
3
The function f is continuous on M , the sets A and B are closed in M . Since M is
closed in X, A and B are closed in X. By Uryson lemma, there is a continuous function
h : X → R such that h(x) = 0 for x ∈ A and h(x) = 1 for x ∈ B. Now for each x ∈ X,
let
2 1
h1 (x) = c [h(x) − ].
3 2
Then the function h1 is continuous on X and
c
(1) |h1 (x)| ≤ for all x ∈ X;
3
2
(2) |f (x) − h1 (x)| ≤ c for all x ∈ M .
3
Using (1) and (2) for the function f − h1 , we get a function h2 continuous on X such
that
1 2
(3) |h2 (x)| ≤ × c for all x ∈ X;
3 3
¡ 2 ¢2
(4) |f (x) − h1 (x) − h2 (x)| ≤ c for all x ∈ M .
3
By induction, we obtain a sequence (hn )n of continuous functions on X such that
1 ¡ 2 ¢n−1
(5) |hn (x)| ≤ c for all x ∈ X;
3 3
¯ P ¯ ¡ 2 ¢n
(6) ¯f (x) − ni=1 hi (x)¯ ≤ c for all x ∈ M .
3
X∞
It follows from (5) that the series of functions hn is uniformly convergent on X.
i=1
∞
X
Let F = hn . Then F is continuous on X since each hn is continuous on X.
n=1
¡ 2 ¢n
Since lim c = 0, we have from (6) f (x) = F (x) for all x ∈ M . Thus,
n→∞ 3

sup |F (x)| ≥ sup |F (x)| = sup |f (x)|.

x∈X x∈M x∈M

Hence
sup |F (x)| = sup |f (x)|.
x∈X x∈M

5.2.3. Corollary. Let M be a closed subspace of a normal space X and let f : M → R

be a continuous function. Then there exists a continuous function F : X → R such that
F (x) = f (x) for all x ∈ M .

. Luong Ha - HUCE
112 Chapter 5. Topological Spaces

Proof. If f is bounded, then the existence of F follows from Theorem 5.2.2.

Suppose that f is not bounded. For each x ∈ M , let g(x) = arctan f (x). The function
π
g is continuous on M and sup |g(x)| = . By Theorem 5.2.2, there is a continuous
x∈M 2
function G : X → R such that
1) G(x) = g(x) for all x ∈ M ;
π
2) sup |G(x)| = .
x∈X 2
π
Let B = {x ∈ X : |G(x)| = }. Then B is closed, and M ∩ B = ∅. By Theorem
2
5.2.1, there exists a function h continuous on X such that
3) 0 ≤ h(x) ≤ 1 for all x ∈ X;
4) h(x) = 0 if x ∈ B;
5) h(x) = 1 if x ∈ M .
The function G · h is continuous on X, and has the following properties:
i) h(x) · G(x) = g(x) for all x ∈ M ;
π π
ii) − < h(x) · G(x) < for all x ∈ X.
2 2
Let F (x) = tan[h(x) · G(x)]. Then F is the desired function.¥

§6 THE METRIZABLE SPACES

6.1 METRIZABLE SPACES AND AXIOMS OF COUNTABILITY

6.1.1 Definition. Let (X, τ ) be a topological space. If there is a metric d on X such

that the topology generated by d is τ , then we say that (X, τ ) is a metrizable space.
Example.
1) Let X be a discrete space. Then this topology is generated by the discrete metric
on X, that is, 
1 if x 6= y
d(x, y) =
0 if x = y.

2) Let τ be the topology on X = R defined as τ = {∅, X, [0; 1]}. Then (X, τ ) is not
metrizable since two elements cannot have two disjoint neighbourhoods.
3) An indiscrete space X where X consists of more than one point is not metrizable
since X is not a Hausdorff space.
6.1.2 Axioms of countability.
Let X be a topological space and let x ∈ X. A family U of neighbourhoods of x is
called a local basis at the point x if for any neighbourhood V of x, there exists U ∈ U
such that U ⊂ V .
1) A topological space X is called a first countable space if it satisfies the following

. Luong Ha - HUCE
§ 6. The Metrizable Spaces 113

axiom called the first axiom of countability.

[C1 ] For each point x ∈ X, there exists a countable local basis at x.
2) A topological space X is called a second countable space if it satisfies the following
axiom called the second axiom of countability.
[C2 ] There exists a countable basis for the topology on X.
Example.
1) Let X be a metric space and let x ∈ X. Then the countable class of open balls
B(x, n1 ), n ∈ N is a local basis at x. Hence, every metric space satisfies the first axiom of
countability.
2) Consider the discrete topology D on the real line R. A class B is a local basis for
a discrete topology if and only if it contains all singleton sets. But R and the class of
singleton subsets of R, are non-countable. Accordingly, the space (R, D) does not satisfy
the second axiom of countability.
Now if B is a countable basis for a topological space X, and if Bx consists of the
members of B which contain the point x ∈ X, then Bx is a countable local basis at x. In
other words, we have
6.1.3 Proposition. A second countable space is also first countable.

6.2 METRIZATION THEOREMS

The metrization problem in topology consists of finding necessary and sufficient topo-
logical conditions for a topological space is metrizable. An important partial solution
to this problem was given in 1924 by Urysohn as a result of his celebrated Urysohn’s
lemma. The complete solution to this problem was proven by Bing in 1951 and was an
independent discovery with the Nagata-Smirnov metrization theorem that was proven by
both J. Nagata (1950) and Y. Smirnov in 1951.
6.2.1 Theorem (Urysohn metrization theorem). Every regular space (X, τ ) with a
countable basis is metrizable.
Proof. Let B = {U1 , U2 , ..., Un , ...} be a countable basis for the space X. For each
x ∈ Um , by regularity, there is a Un such that x ∈ Un ⊂ U n ⊂ Um . Thus, there exist pairs
(Un , Um ) of elements of B such that U n ⊂ Um . It is clear that the collection of such pairs
is countable. We enumerate these pairs as
(Un1 , Um1 ), ..., (Unk , Umk ), ...
By hypothesis, X is a normal space. For each k, the sets U nk and X \ Umk are closed and
disjoint, there exists a continuous function fk : X → [0; 1] such that fk (x) = 0 if x ∈ U nk
and fk (x) = 1 if x ∈ X \ Umk .
Let g : X × X → R be given by
X∞
1
g(x, y) = k
|fk (x) − fk (y)|. (1)
k=1
2

. Luong Ha - HUCE
114 Chapter 5. Topological Spaces

Since |fk (x) − fk (y)| ≤ 1 for any x, y ∈ X and any k ∈ N, the series in the righthand side
of (1) is convergent.
Evidently, g(x, y) ≥ 0. Let x, y ∈ X with x 6= y. Then there is Um ∈ B such that
x ∈ Um and y ∈ / Um . By the above argument, there is Un ∈ B so that x ∈ Un ⊂ U n ⊂ Um .
Thus, the pair (Un , Um ) must be some pair (Unk , Umk ), where x ∈ U nk and y ∈ X \ Umk .
1
Hence, fk (x) = 0 and fk (y) = 1. Therefore, g(x, y) ≥ k > 0. The two other axioms of a
2
metric is easily verified. Hence, g is a metric on X.
Now we shall show that the topology Φ generated by the metric g is precisely τ . For
V ∈ Φ, and xo ∈ V , there exists r > 0 such that B(xo , r) ⊂ V . The series of functions
X∞
1
g(x, xo ) = |fk (x) − fk (xo )|, x∈X
k=1
2k

is uniformly convergent on X, and each term is a function which is continuous on X, its

sum is continuous on X. Since the function x 7→ g(x, xo ) is continuous at the point xo ,
there exists an element Un ∈ B such that xo ∈ Un and for all x ∈ Un , we have

g(x, xo ) = |g(x, xo ) − g(xo , xo )| < r,

that is, Un ⊂ B(xo , r) ⊂ V . Thus, V ∈ τ , and hence, Φ ⊂ τ .

Conversely, let V ∈ τ and let xo ∈ V . Then there exists Um ∈ B so that xo ∈ Um ⊂ V .
Let Un be an element of B such that xo ∈ Un ⊂ U n ⊂ Um . The pair (Un , Um ) is some pair
1 1
(Unk , Umk ). Let x ∈ B(xo , k ), then g(x, xo ) < k . Since fk (xo ) = 0, we have
2 2
1 1
g(x, xo ) ≥ k |fk (x) − fk (xo )| = k fk (x).
2 2
1
It follows that fk (x) < 1. Hence x ∈/ X \Umk , that is, x ∈ Um . Thus, B(xo , k ) ⊂ Um ⊂ V .
2
This show that V ∈ Φ, and hence, τ ⊂ Φ.
Consequently, τ = Φ. This means that the given topology on X is generated by the
metric g.¥
The Urysohn metrization theorem gives conditions under which a space X is metriz-
able: that it be regular and have a countable basis. But mathematicians hope that
they can get a stronger result. Since a metrizable is normal, the regularity hypothesis
in Urysohn metrization theorem is necessary. The only thing to do is try to replace
the countable basis condition by a something weaker. This is a difficult task since the
condition has to be strong enough to imply metrizability, and yet weak enough that all
metrizable spaces satisfy it.
This condition was formulated by J. Nagata and Y. Smirnov independently, and it
involves a new notion, that of local finiteness.
6.2.2 Definition. Let X be a topological space. A collection A of subsets of X is said
to be locally finite if every point of X has a neighbourhood that intersects only finitely
many elements of A.

. Luong Ha - HUCE
§ 6. The Metrizable Spaces 115

For example, the collection of intervals

A = {(n; n + 2) : n ∈ Z}
is locally finite in the topological space R.
6.2.3 Definition. A collection B of subsets of a topological space X is said to be countably
locally finite if B can be written as the countable union of collections Bn , each of which is
locally finite.
Now we are in position to formulate the necessary and sufficient condition of metriz-
ability.
6.2.4 Theorem (Nagata-Smirnov metrization theorem). A topological space X is
metrizable if and only if X is regular and has a basis that is countably locally finite.
For the proof of this theorem, please refer to [11], pp.250-252.

PROBLEMS
5.54 Let Bx be a countable local basis at the point x of a topological space X. Then we
can index the members of Bx by N, i.e. we can write Bx = {B1 , B2 , ...}. If, in addition,
B1 ⊃ B2 ⊃ B3 ⊃ ..., then we call Bx a nested local basis at x.
Show that we can construct a nested local basis from a countable local basis.

5.55 Let (xn )n be a sequence in a topological space X. This sequence is said to be

convergent to a point xo ∈ X if for any neighbourhood U of xo , there exists no ∈ N such
that xn ∈ U for all n ≥ no . Then we denote by xn → xo .
Let A ⊂ X. Show that if there is a sequence in A converging to x, then x is an
adherent point of A; and the converse holds if X is a first countable space.

5.56 Let X be a Hausdorff space. Show that each sequence in X converges to at most
one point.

5.57 Let f : X → Y be a mapping between topological spaces. Show that if f is

continuous at xo ∈ X, then f (xn ) → f (xo ) for any (xn )n ⊂ X, xn → xo . Show the
converse holds if X is a first countable space.

5.58 Let f : X → Y be a continuous open mapping. Show that if X satisfies the first or
the second countability axiom, then f (X) satisfies the same axiom.

HINTS and SOLUTIONS

5.1 Since A ∩ B = ∅, and A is open, each x ∈ A is not an adherent point of B. Hence,
A∩B = ∅. It follows that A∩ int(B) = ∅. Now we have int(B) is open and int(B)∩A = ∅,
so int(B) ∩ int(A) = ∅.

. Luong Ha - HUCE
116 Chapter 5. Topological Spaces

o
5.3 We have A = A \ ∂(A), so
o
X \ A = X \ (A \ ∂A) = X \ (A ∩ (X \ ∂A).

The lefthand side is equal to (X \ A) ∪ ∂(A) = (X \ A) ∪ ∂(X \ A) = X \ A. Therefore,

o o
A \ A = A ∩ (X \ A) = A ∩ (X \ A) = ∂(A).

5.5
i) By definition, ∅ ∈ τ , and S1 = N yields X ∈ τ .
S
ii) Let (Gα )α∈I be a subfamily of τ . Set G = Gα . If G = ∅, then G ∈ τ . Suppose
α∈I
that G 6= ∅. Then there exists α ∈ I so that Gα 6= ∅. Hence, there is nα ∈ N with
Gα = Snα . Let no = min{nα ∈ N : Gα 6= ∅}. Note that Sn ⊃ Sn+1 for each n ∈ N,
S
Snα ⊂ Sno for all α so that Gα 6= ∅. Thus, G = {Snα : Gα 6= ∅} = Sno , so G ∈ τ .
iii) Consider G1 and G2 to belong to τ . If G1 ∩ G2 = ∅, then G1 ∩ G2 ∈ τ . Suppose
that G1 ∩ G2 6= ∅. Then G1 6= ∅, and G2 6= ∅. Hence, there are m, n ∈ N such that
G1 = Sm , and G2 = Sn . Let k = max(m, n), we have Sm ∩ Sn = Sk . This shows that
G1 ∩ G2 ∈ τ.
Thus, τ is a topology on X. Note that if G ∈ τ , and G 6= ∅, then G = Sn for some
n ∈ N. This shows that G is infinite.
A set F ⊂ X is closed if and only if either F = X or F = X \ Sn , that is, F =
o o
{1, 2, . . . , n − 1}. Therefore A = ∅, and A = {1, 2, . . . , 100}. We also have B = ∅, and
B = X.
5.6 Hints.
o
A = ∅, A = X, ∂(A) = X.
o
B = ∅, B = {1, 2}, ∂(B) = {1, 2}.
5.8 Hints.
Let X = {a, b, c}, and τ1 = {∅, X, {a}}, τ2 = {∅, X, {b}}. Then τ1 and τ2 are two
topologies on X, but τ1 ∪ τ2 is not a topology on X.
5.10 Consider τ = {A ⊂ X : f (A) = A}, then we get X ∈ τ from (i). By ii), f (∅) ⊂ ∅,
but ∅ ⊂ f (∅), it follows that f (∅) = ∅. This implies that ∅ ∈ τ .
Let A, B ∈ τ . Then A ∩ B = f (A) ∩ f (B) = f (A ∩ B) (by (iii)), hence, A ∩ B ∈ τ.
S
Let (Aα )α∈I be a subfamily of τ . Set B = Aα . We shall show that B ∈ τ . Since
α∈I
Aα ⊂ B,
f (Aα ) = f (Aα ∩ B) = f (Aα ) ∩ f (B) ⊂ f (B).
S S
Thus, B = Aα = f (Aα ) ⊂ f (B). On the other hand, by (ii) f (B) ⊂ B. Hence,
α∈I α∈I
f (B) = B, that is, B ∈ τ.
In short, τ is a topology on X.

. Luong Ha - HUCE
§ 6. The Metrizable Spaces 117

o o
Let A ⊂ X. Denote by A the interior of A in (X, τ ). By definition, A = ∪{G :
G ∈ τ, G ⊂ A}. By iv), we have f (f (A)) = f (A), so f (A) ∈ τ . But f (A) ⊂ A, hence,
o o o o o
f (A) ⊂ A. On the other hand, A ∈ τ yields f (A) = A. Furthermore, since A ⊂ A,
o o o
f (A) ⊂ f (A). Thus, f (A) ⊃ A, and hence, A = f (A).
o
We can show that τ is the unique topology on X such that f (A) = A for each A ⊂ X.
5.12 If ∂(A) = ∅, then A ∩ ∂(A) = ∅, so A is open. Moreover, since ∂(A) = ∅ ⊂ A, A is
closed.
Conversely, suppose that A is both open and closed. If ∂(A) 6= ∅, then there is
xo ∈ ∂(A). Since A is open, xo ∈
/ A. On the other hand, since A is closed, X \ A is open.
We have xo ∈ ∂(X \ A), so xo ∈/ X \ A. This is a contradiction because A ∪ (X \ A) = X.
5.13 Let B ⊂ Y ⊂ X, and let x ∈ intX (B). Then there exists an open set V in X such
that x ∈ V ⊂ B. Thus, V ⊂ Y , so V = V ∩ Y is an open set in Y , and x ∈ V ⊂ B,
hence, x ∈ intY (B). Therefore, intX (B) ⊂ intY (B).
5.15 Hints.
Note that the open intervals (a; b) form a basis for the usual topology on R. If x ∈
(a; b), then there are r, s ∈ Q with a < r < x < s < b. Hence, x ∈ (r; s) ⊂ (a, b). We get
the desired result.
5.17
i) By definition, ∅ ∈ τ1 . Since 0 ∈ R, we also have R ∈ τ1 .
S
ii) Let (Gα )α∈I be a subfamily of τ1 . Set Gα . If G = ∅, then G ∈ τ1 . Suppose that
α∈I
G 6= ∅. Then there is αo ∈ I sao cho Gαo 6= ∅. Thus, 0 ∈ Gαo , and hence, 0 ∈ G. This
shows that G ∈ τ1 .
iii) Let G1 and G2 belong to τ1 . If G1 ∩ G2 = ∅, then G1 ∩ G2 ∈ τ1 . If G1 ∩ G2 6= ∅,
then 0 ∈ G1 , and 0 ∈ G2 , so 0 ∈ G1 ∩ G2 . Thus, G1 ∩ G2 ∈ τ1 .
Consequently, τ1 is a topology on R.
∗ Let V ∈ τ1 with V 6= ∅. Then 0 ∈ V , so 0 ∈ V ∩ Q. Thus, Q is dense in (R, τ1 ), and
(R, τ1 ) is separable.
∗ Suppose that (R, τ1 ) has a countable basis. By Lindelöf theorem, each open cover
of (R, τ1 ) has a finite or countable subcover. For each x ∈ R, x 6= 0, set Vx = {0, x}, then
S
Vx is open, and R = Vx . Thus, there are a finitely many or countably many elements
x6=0 S
x1 , x2 , . . . , xn , . . . of R so that R = Vxi . This is impossible since the righthand side is
i
at most countable, while the lefthand side is uncountable. Hence, (R, τ1 ) cannot have a
countable basis.
5.20 Suppose that U is an open cover of a closed subspace Y of X. Each element of U is
of the form G ∩ Y , where G is an open set in X. Then the family U ∗ = {G : G ∩ Y ∈
U} ∪ (X \ Y ) is an open cover of X. Since X is a Lindelöf space, there is an at most
countable subcover which consists of X \ Y , and the sets Gn ∩ Y, n ∈ I ⊂ N. Then the

. Luong Ha - HUCE
118 Chapter 5. Topological Spaces

family (Gn ∩ Y )n∈I is an at most countable subcover of U. Thus, Y is a Lindelöf space.

5.21 Hints.
Each topological space having a countable basis is Lindelöf, so 1) ⇒ 2).
1
For each n ∈ N, let Bn = {B(x, ) : x ∈ X}. Then Bn is an open cover of X.
n
1
Hence, Bn has an at most countable subcover {B(xnk , ) : k ∈ In ⊂ N}. The set
nk
{xnk : n ∈ N, k ∈ In } is dense in X. Thus, 2) ⇒ 3).
3) ⇒ 1). Suppose that A is an at most countable set which is dense in X. Let
1
B = {B(x, ) : x ∈ A, m ∈ N}, then B is a basis for the topology on X.
m
5.23 a) ⇒ b). The mapping f is continuous, so f −1 (f (A)) is closed in X. Since A ⊂
f −1 (f (A)) ⊂ f −1 (f (A)), we get A ⊂ f −1 (f (A)). Hence, f (A) ⊂ f (A).
b) ⇒ c). For any B ⊂ Y , f −1 (B) ⊂ X, by hypothesis, f (f −1 (B)) ⊂ f (f −1 (B)) ⊂ B.
Hence f −1 (B) ⊂ f −1 (B).
o o
c) ⇒ d). Let B ⊂ Y . Then Y \ B = Y \ B (Problem 1.3). Therefore, B = Y \(Y \ B).
Thus,
o
f −1 (B) = f −1 (Y \ (Y \ B) = X \ f −1 (Y \ B).

In addition, f −1 (Y \ B) ⊃ f −1 (Y \ B) = X \ f −1 (B), hence

o
f −1 (B) ⊂ X \ (X \ f −1 (B)) = intf −1 (B)).
o
d) ⇒ a). Let G be an open set in Y . By hypothesis, f −1 (G) ⊂ int(f −1 (G)). But
o
G = G, so f −1 (G) ⊂ int(f −1 (G)). Hence, f −1 (G) is open. Thus, f is continuous.
5.24 1) ⇒ 2). Let G ⊂ X be any open set. We have f (G) = (f −1 )−1 (G), where the
inverse f −1 : Y → X is continuous. Hence, f (G) is open in Y , and f is an open mapping.
2) ⇒ 3). Let F be closed in X. Since f is bijective, f (X \ F ) = Y \ f (F ). But X \ F
is open, and f is an open mapping, Y \ f (F ) is open in Y . Thus, f (F ) is closed, and
hence, f is a closed mapping.
3) ⇒ 1). Suppose that F is closed in X. We shall show that (f −1 )−1 (F ) is closed in
Y . Since f is bijective, we get (f −1 )−1 (F ) = f (F ), and f (F ) is closed since f is a closed
mapping. Thus, the inverse f −1 is continuous, and hence, f is a homeomorphism.
5.26 Recall that if X is an infinite set equipped with the cofinite topology, then there
are no nonempty disjoint open subsets of X. Suppose that f is not constant. Then
there are x1 , x2 ∈ X such that f (x1 ) < f (x2 ). Choose any positive number c so that
f (x1 ) < c < f (x2 ). Let

A = {x ∈ X : f (x) < c}, B = {x ∈ X : f (x) > c}.

Then A = f −1 ((−∞, c)), and B = f −1 ((c, +∞)). Clearly, A and B are open in X. They
are nonempty because they contain x1 and x2 , respectively. Furthermore, A ∩ B = ∅.
This contradicts the above remark. Thus, f is a constant mapping.

. Luong Ha - HUCE
§ 6. The Metrizable Spaces 119

5.28 Let A1 = (−∞; 0], A2 = [0; 1], and A3 = [1; +∞). Then A1 , A2 , A3 are closed, and
R = A1 ∪ A2 ∪ A3 . Consider A to be a closed subset of R. We have
f (A) = f (A ∩ (A1 ∪ A2 ∪ A3 )) = f (A ∩ A1 ) ∪ f (A ∩ A2 ) ∪ f (A ∩ A3 ).
Since f (A ∩ A1 ) ⊂ f (A1 ), f (A ∩ A1 ) is equal to either ∅ or {0}. On A ∩ A2 , f (x) = x,
so f (A ∩ A2 ) = A ∩ A2 is closed. Moreover, f (A ∩ A3 ) = ∅ or {1}. Thus, the sets
f (A ∩ A1 ), f (A ∩ A2 ), and f (A ∩ A3 ) are closed. It follows that f (A) is closed, and hence,
f is a closed mapping.
Choose B = (−∞; 0), then B is open, and f (B) = {0} is not open. Thus, f is not an
open mapping.
5.34
o o
1) int(A × B) = A × B.
o o o o
First we observe that A, and B are open in X and Y , respectively. Hence, A × B is
o o o o
open in the product space X × Y . Furthermore, A × B ⊂ A × B, so A × B ⊂ int(A × B).
Conversely, let (x, y) ∈ int(A × B). Then there are an open set U in X, and an open
set V in Y such that (x, y) ∈ U × V ⊂ A × B. Thus, x ∈ U ⊂ A, and y ∈ V ⊂ B.
o o o o
Therefore, x ∈ A, and y ∈ B, that is, (x, y) ∈ A × B. Consequently, we get the desired
equality.
2) A × B = A × B.
Consider (x, y) ∈ A × B. Let U be an arbitrary neighbourhood of x, and let V be
an arbitrary neighbourhood of y. Then there are an open set G1 ⊂ X, and an open set
G2 ⊂ Y so that x ∈ G1 ⊂ U, y ∈ G2 ⊂ V . Hence, G1 × G2 is a neighbourhood of (x, y).
Since (x, y) ∈ A × B, we have (G1 × G2 ) ∩ (A × B) 6= ∅. But (G1 × G2 ) ∩ (A × B) =
(G1 ∩ A) × (G2 ∩ B), so G1 ∩ A 6= ∅, and G2 ∩ B 6= ∅. Thus, U ∩ A 6= ∅, and V ∩ B 6= ∅.
This implies that x ∈ A, y ∈ B, and hence, A × B ⊂ A × B.
On the other hand, let (x, y) ∈ A × B and let W be an arbitrary neighbourhood of
(x, y). Then there are two open sets G1 ⊂ X, G2 ⊂ Y such that (x, y) ∈ G1 × G2 ⊂ W .
It follows that G1 , G2 are two neighbourhoods of x, y, respectively. Hence, G1 ∩ A 6= ∅,
and G2 ∩ B 6= ∅. This shows that (G1 × G2 ) ∩ (A × B) 6= ∅, and so W ∩ (A × B) 6= ∅.
Thus, (x, y) ∈ A × B, and hence, A × B = A × B.
If A and B are closed, then A = A and B = B. Hence, A × B = A × B, and this
means that A × B is closed in X × Y .
5.35 First we prove a general result as follows:
Let (X, T ) and (Y, U) be topological spaces. Suppose that B1 is a basis for T , and B2 is
a basis for U. Then the family B = B1 × B2 is a basis for the product topology on X × Y .
Let G be any open set in X × Y , and let (x, y) ∈ G. By definition, there are open sets
U1 ⊂ X, and U2 ⊂ Y so that (x, y) ∈ U1 × U2 ⊂ G. Then there are B1 ∈ B1 , B2 ∈ B2 with
x ∈ B1 ⊂ U1 , y ∈ B2 ⊂ U2 . Thus, there exists B1 ×B2 ∈ B such that (x, y) ∈ B1 ×B2 ⊂ G.
Therefore, the family B is a basis for the product topology on X × Y .

. Luong Ha - HUCE
120 Chapter 5. Topological Spaces

The family of open intervals (a; b) is a basis for the usual topology on R, and the
family of one-point sets {c}c∈Y is a basis for the discrete topology on Y , we have the
desired results. In the product space, we have
o
A = (0; 1) × {0}, A = [0; 1] × {0}
o
B = ∅, B = {0} × [0; 1)
o
C = (0; 1) × [0; 1), C = [0; 1] × [0; 1).
5.36 Let pα be the projection from X onto Xα . Then the mapping f is continuous if and
only if pα ◦ f is continuous for each α ∈ I. But

(pα ◦ f )(x) = pα (f (x)) = pα (fβ (x))β∈I = fα (x)

for all x ∈ Y , so f is continuous if and only if fα is continuous for each α ∈ I.

5.37 Hints.
Show that the image of each members of the basis for the product topology on X is
open.
5.38 Let τ be the cofinite topology on X. Then (X, τ ) is a T1 -space. Suppose that ξ is a
topology on X such that (X, ξ) is T1 . Consider G ∈ τ . Clearly, if G = ∅, then G ∈ ξ.
Suppose that G 6= ∅. The set X \ G must be finite. Since (X, ξ) is T1 , X \ G is closed
in (X, ξ). Hence, G ∈ ξ. This shows that τ ≤ ξ.
5.39 Consider xo ∈ X \ A. Then f (xo ) 6= g(xo ). The space X is T2 , so there are open sets
U, V in Y with f (xo ) ∈ U, g(xo ) ∈ V , and U ∩ V = ∅. Set U1 = f −1 (U ), V1 = g −1 (V ),
then U1 , V1 are open sets in X containing xo . Hence, G = U1 ∩ V1 is a neighbourhood of
xo . For each x ∈ G, we have f (x) ∈ U , and g(x) ∈ V . Since U ∩ V = ∅, so f (x) 6= g(x),
that is, x ∈ X \ A. This implies that G ⊂ X \ A. Thus, the set X \ A is open, and hence,
A is closed.
5.40 Hints.
Let M = f (X). Note that M = {y ∈ X : f (y) = y}. Now the proof is similar to that
of Problem 5.39.
5.42
∼ ∼
1) Denote Rc = (X × X) \ R, and let (x, y) ∈ Rc . Then x 6= y. By hypothesis,
∼ ∼
there exist two open neighbourhoods V (x), V (y) in the quotient space X/R such that
∼ ∼ ∼ ∼
V (x) ∩ V (y) = ∅. Let U (x) = π −1 (V (x)), V (y) = π −1 (V (y)). Then U (x), V (y) are open
neibourhoods of x and y, respectively. We will show that U (x) × U (y) ⊂ Rc . Indeed,
∼ ∼ ∼ ∼
for any (x1 , y1 ) ∈ U (x) × U (y), we have π(x1 ) = x1 ∈ V (x), π(y1 ) = y 1 ∈ V (y). Hence,
∼ ∼
x1 6= y 1 or (x1 , y1 ) ∈ Rc . Thus, Rc is an open set, and hence, R is a closed set.
∼ ∼
2) Let x, and y be two distinct elements of X/R. Then (x, y) ∈ Rc . The set Rc
is open, there exist two open sets U, V in X such that (x, y) ∈ U × V ⊂ Rc . Let
U1 = π(U ), V1 = π(V ). Since π is an open mapping, the sets U1 , V1 are open in X/R. If
∼ ∼ ∼
there is u ∈ U1 ∩ V1 , then there are u1 ∈ U, v1 ∈ V so that π(u1 ) = u, and π(v1 ) = u.

. Luong Ha - HUCE
§ 6. The Metrizable Spaces 121

Therefore, (u1 , v1 ) ∈ R. This contradicts the fact that (u1 , v1 ) ∈ U × V ⊂ Rc . Thus, we

∼ ∼
must have U1 ∩ V1 = ∅. Clearly, x ∈ U1 , y ∈ V1 , so X/R is a Hausdorff space.
5.44 Let X be a Hausdorff space. Consider x ∈ X and let Fx is the family of closed
neighbourhoods of x. We will show that {x} = ∩{F : F ∈ Fx }. Suppose that there is
y ∈ ∩{F : F ∈ Fx } with y 6= x. Then there exist two open neighbourhoods V (x), V (y) of
x, y, respectively, such that V (x) ∩ V (y) = ∅. It follows that y ∈
/ V (x). This is impossible
since V (x) ∈ Fx .
Conversely, let x, y ∈ X with x 6= y. By hypothesis, there is a closed neighbourhood
V (x) of x so that y ∈ / V (x). The set X \ V (x) is open and contains y, hence, it is a
neighbourhood of y. Let V (y) = X \ V (x), then V (x) ∩ V (y) = ∅. Thus, X is a Hausdorff
space.
5.45 Hints.
1) Consider the mapping g : X → G to be defined by g(x) = (x, f (x)). Then g is
continuous and bijective.
The inverse g −1 : G → X is defined by g −1 (x, f (x)) = x. Hence g −1 = p1|G , where
p1 : X × Y → X; p1 (x, y) = x.
The projection p1 is continuous, so is g −1 . Thus, g is a homeomorphism.
2) The proof is left for the reader.
5.50 We have X = G1 ∪ G2 , so (X \ G1 ) ∩ (X \ G2 ) = ∅. The sets X \ G1 , X \ G2 are
closed, and disjoint. The space X is normal, there exist two open sets U, V in X such
that X \ G1 ⊂ U, X \ G2 ⊂ V , and U ∩ V = ∅. Let F1 = X \ U, F2 = X \ V . Then F1 , F2
are closed, and F1 ⊂ G1 , F2 ⊂ G2 . We have F1 ∪ F2 = X \ (U ∩ V ) = X.

. Luong Ha - HUCE
Chapter 6

Compactness and Connectedness

§1 COMPACT SPACES

1.1 BASIC NOTIONS

1.1.1 Definition. Let X be a topological space. A set A ⊂ X is called compact if each

open cover of A has a finite subcover.
The space X is called a compact space if the set X is compact.
1.1.2 Example.
1) In any topological space, a finite set is compact.
2) The union of a finite number of compact subsets of X is compact.
3) A discrete space X is compact if and only if the set X is finite.
These statements follows directly from the definition.
1.1.3 Theorem. Let A be a nonempty subset of a topological space X. Then the set A
is compact if and only if the subspace A is compact.
Proof. Let A ⊂ X is a compact set. Consider an open cover (Gα )α∈I of the subspace
S
A. This means that each Gα is open in A and A = Gα . Then for each α ∈ I, there
α∈I S
exists an open set Vα in X such that Gα = A ∩ Vα . Hence, A ⊂ Vα . By hypothesis,
α∈I
S
n Sn S
n
there are α1 , α2 , . . . , αn in I so that A ⊂ Vαi . Therefore, A = (A ∩ Vαi ) = Gαi .
i=1 i=1 i=1
This means that the subspace A is compact.
The proof of the converse is left for the reader.¥
Recall that for A ⊂ Rk with the usual topology, the following statements are equiva-
lent:
1) A is compact.
2) A is closed and bounded.
3) Each sequence in A has a subsequence converging to a point of A.
1.1.4. Theorem. Every closed subset of a compact space is compact.

122
§ 1. Compact Spaces 123

Proof. Let X be a compact space and let A ⊂ X be a closed set. Consider an open
cover (Gα )α∈I of A. Then all the collection (Gα )α∈I and the set X \ A form an open cover
of X. Hence, there exist α1 , α2 , . . . , αn ∈ I such that
³Sn ´S
X= Gαi (X \ A).
i=1
S
n
Thus, A ⊂ Gαi , and hence, the set A is compact.¥
i=1
Note that a compact set in a topological space may not be closed. For instance, let
X = {a, b} with the topology τ = {∅, X, {a}}. Then the set {a} is compact but not
closed.
1.1.5 Finite intersection property. A collection F of subsets of X is said to have
the finite intersection property if for every finite subcollection {F1 , F2 , ..., Fn } of F, the
intersection F1 ∩ F2 ∩ · · · ∩ Fn is nonempty.
1.1.6 Theorem. A topological space X is compact if and only if for every collection F
T
of closed sets in X having the finite intersection property, the intersection F of all
F ∈F
elements of F is nonempty.
Proof. Let X be a compact space and let (Fα )α∈I be a collection of closed sets in X
T S
having the finite intersection property. Suppose that Fα = ∅. Then X = (X \ Fα ).
α∈I α∈I
Thus, the collection (X \ Fα )α∈I is an open cover of X. By the compactness of X, there
S
n T
n
are α1 , α2 , . . . , αn ∈ I such that X = (X \ Fαi ). Hence, Fαi = ∅. This contradict
i=1 i=1 T
the hypothesis of having the finite intersection property of (Fα )α∈I , hence, Fα 6= ∅.
α∈I
S
Conversely, let (Gα )α∈I be an open cover of X. Thus, X = Gα , so we have
T α∈I
(X \ Gα ) = ∅. Since each set X \ Gα is closed, the collection (X \ Gα )α∈I cannot
α∈I
have the finite intersection property. Therefore, there exist α1 , α2 , . . . , αn ∈ I such that
Tn S
n
(X \ Gαi ) = ∅. Thus, X = Gαi , and this shows that X is compact.¥
i=1 i=1
1.1.7 Example.
1) In the space of real numbers with the usual topology, the subspace [0; 1) is not
compact. The collection F = {[1 − n1 ) : n = 1, 2, ...} is a collection of closed subsets of
[0; 1) that has the finite intersection property but has empty intersection.
2) In the plane R2 with the usual topology, the subspace X = R2 \ {(0, 0)} is not
compact. For each positive integer n, let Fn = {(x, y) ∈ R2 : x2 + y 2 ≤ n1 }. Then (Fn )n
is a collection of closed sets in R2 that has the finite intersection property and has empty
intersection.
3) Let X be a compact space. Suppose that we have a decreasing sequence F1 ⊃ F2 ⊃
· · · ⊃ Fn ⊃ · · · of closed sets in X. If each of the sets Fn is nonempty, then the collection
T
∞
(Fn )n has the finite intersection property. Hence, the intersection Fn 6= ∅.
n=1
1.1.8 Theorem. Let f : X → Y be a continuous mapping. Then f (A) is compact for

. Luong Ha - HUCE
124 Chapter 6. Compactness and Connectedness

every compact set A ⊂ X.

S
Proof. Let (Gα )α∈I be an open cover of f (A). Then f (A) ⊂ Gα , hence,
S α∈I
A ⊂ f −1 (f (A)) ⊂ f −1 (Gα ).
α∈I ³ ´
The mapping f is continuous and each Gα is open, f −1 (Gα ) is an open cover of A.
α∈I
S
n
There exists α1 , α2 , . . . , αn ∈ I such that A ⊂ f −1 (Gαi ). It follows that
i=1
S
n ³ ´ S n
f (A) ⊂ f f −1 (Gαi ) ⊂ Gαi .
i=1 i=1
Thus, f (A) is compact.¥
1.1.9 Theorem. Let X be a compact space and let f : X → R be a continuous
function. Then f is bounded, and it attains the greatest value and the least value on X.
Proof. The proof is quite similar as the case of a compact metric space.

1.2 COMPACTNESS AND AXIOMS OF SEPARATION

1.2.1 Theorem. Let X be a Hausdorff space. Let A ⊂ X be compact and x ∈ / A. Then

there exists an open neighbourhood U of x and an open set V containing A such that
U ∩ V = ∅.
Proof. Let y ∈ A. Then y 6= x, so there are open neighbourhoods Vy , Vx (y) of y, x,
respectively so that Vy ∩Vx (y) = ∅. The collection (Vy )y∈A is an open cover of the compact
S
n
set A, there are a finite number of elements y1 , y2 , . . . , yn of A suct that A ⊂ Vyi . Let
i=1
S
n T
n
V = Vyi , U= Vx (yi ).
i=1 i=1
It is clear that V is an open set containing A, and U is an open neighbourhood of x. We
have
n n n
¡\ ¢ ¡[ ¢ [
U ∩V = Vx (yi ) ∩ Vyj ⊂ (Vx (yj ) ∩ Vyj ).
i=1 j=1 j=1

Since Vx (yj ) ∩ Vyj = ∅ for each j = 1, 2, ..., n, the sets U and V are disjoint.¥
1.2.2 Theorem. Every compact set in a Hausdorff space is closed.
Proof. Let X be a Hausdorff space and let A ⊂ X be compact. If x ∈ / A, then by
Theorem 1.2.1, there exist an open neighbourhood U of x and an open set V containing
A such that U ∩ V = ∅. Hence, U ⊂ X \ V ⊂ X \ A. Thus, the set X \ A is open, and
hence, A is closed.¥
1.2.3 Theorem. Let X be a Hausdorff space, and let A, B be disjoint compact sets in
X. Then there are two open sets U, V in X such that A ⊂ U, B ⊂ V and U ∩ V = ∅.
Proof. Let x ∈ A. By Theorem 1.2.1, there are two open subsets Ux , Vx in X such
that x ∈ Ux , B ⊂ Vx and Ux ∩ Vx = ∅. The collection (Ux )x∈A is an open cover of
the compact set A, there exist a finite number of elements x1 , x2 , . . . , xn of A such that

. Luong Ha - HUCE
§ 1. Compact Spaces 125

S
n
A⊂ Uxi . Let
i=1
S
n T
n
U= Uxi , V = Vxi .
i=1 i=1
Then U and V are the desired open sets.¥
1.2.4 Corollary. Every compact Hausdorff space is normal.
Proof. Evidently, X is a T1 -space. Let A and B be disjoint closed sets of X. By
Theorem 1.1.4, A and B are compact. Hence, by Theorem 1.2.3, there are open sets U, V
in X such that A ⊂ U, B ⊂ V and U ∩ V = ∅.¥
The following theorem plays an important role in the theory of topological spaces.
1.2.5 Tychonoff ’s theorem. Let X be the product of a family of topological spaces
(Xα )α∈I . Then X is compact if and only if each Xα is compact.
Proof. Suppose that X is compact. Each projection pα : X → Xα is surjective and
continuous. Thus, pα (X) = Xα , and by Theorem 1.1.8, each Xα is compact.
Conversely, suppose that each Xα is compact. Let F be a family of closed sets having
the finite intersection property in the product space X. By Zorn’s lemma, there exists a
maximal family Fo which has the finite intersection property and contains F. We have
\ \ \
A= A⊃ A.
A∈F A∈F A∈Fo

Therefore, in order to prove the intersection of the family F is nonempty, we only need
T
to prove that A is nonempty.
A∈Fo
By the maximality of Fo , we have:
T
m
(a) If A1 , A2 , . . . , Am are in Fo , then Ai ∈ Fo .
i=1
(b) If Ao ⊂ X and Ao ∩ A 6= ∅ for all A ∈ Fo , then Ao ∈ Fo .
(The verification of existence of the family Fo and two properties (a) and (b) are left
for the reader.) ³ ´
From the finite intersection property of Fo , it follows that the family pα (A) also
T A∈Fo
has the finite intersection property in each space Xα . Since Xα is compact, pα (A) 6= ∅
A∈Fo
T
for each α ∈ I. Pick xα ∈ pα (A) for each α ∈ I, and let x = (xα )α∈I .
A∈Fo
T
We shall show that x ∈ A. Let V be any neighbourhood of x. Then there exist
A∈Fo
neighbourhoods W1 , W2 , . . . , Wm of xα1 , xα2 , . . . , xαm in the spaces Xα1 , Xα2 , . . . , Xαm , re-
T
m T
spectively, such that p−1
αi (Wi ) ⊂ V . Since xαi ∈ pαi (A), we have Wi ∩ pαi (A) 6= ∅
i=1 A∈Fo
for each A ∈ Fo and for each i = 1, 2, . . . , m. Hence, p−1
αi (Wi ) ∩ A 6= ∅ for each A ∈ Fo and
T
m
for each i = 1, 2, . . . , m. From the property (b) of Fo , it follows that p−1
αi (Wi ) ∈ Fo .
i=1
T
m
Thus, p−1
αi (Wi ) ∩ A 6= ∅ for any A ∈ Fo , hence, V ∩ A 6= ∅. Therefore, x ∈ A for any
i=1
A ∈ Fo , and the proof is now complete. ¥

. Luong Ha - HUCE
126 Chapter 6. Compactness and Connectedness

1.3 LOCALLY COMPACT SPACES

1.3.1 Definition. A topological space X is said to be locally compact if each x ∈ X has

a closed and compact neighbourhood.
Note that this definition can be reformulated as follows: A topological space X is said
to be locally compact if each x ∈ X has a neighbourhood whose closure is compact.
The locally compact spaces form an important collection of topological spaces. Obvi-
ously, a compact space is locally compact. The spaces Rk with the usual topologies are
locally compact but not compact.
1.3.2 Theorem. Every closed subspace of a locally compact space is locally compact.
Proof. Let X be a locally compact space and let Y be a closed subspace of X. For
each x ∈ Y , there exists a neighbourhood U of x such that U is compact in X. Then the
set Y ∩ U is a closed neighbourhood of x in the subspace Y . But this set is also closed in
the subspace U , and since U is compact, Y ∩ U is compact. Thus, Y is locally compact.¥

1.4 COMPACTIFICATION

1.4.1 Definition.
1) Let X, Y be topological spaces. A mapping ϕ : X → Y is called a homeomorphic
imbedding if ϕ : X → ϕ(X) ⊂ Y is a homeomorphism.
2) Let X be a topological space. A pair (Y, ϕ), where Y is a compact space, and ϕ is a
homeomorphic imbedding from X into Y such that ϕ(X) = Y , is called a compactification
of X.
For example, we know that the space R with the usual topology is homeomorphic
x
to its subspace X = (1; 1) by the mapping f (x) = . Then the pair (Y, i), where
1 + |x|
Y = [−1; 1] and i : X → Y defined by i(x) = x, is a compactification of R.
1.4.2 One-point compactification.
Let (X, τ ) be locally compact, but not compact. Denote by ∞ is a point not belonging
to X. Let X∞ = X ∪ {∞}, and define the family τ∞ of subsets of X∞ as follows: G ∈ τ∞
if either G ∈ τ or G = U ∪ {∞}, where U ∈ τ and X \ U is compact in X. Denote by i
the inclusion from X into X∞ , that is, i(x) = x for all x ∈ X.
1.4.3 Theorem (Alexandrov). The pair (X∞ , i) is a compactification of the space X.
Proof. First, we can verify that the family τ∞ is a topology on X∞ and X is a
subspace of X∞ . Then the inclusion i : X → X∞ is a homeomorphic imbedding.
Now we shall prove that the space X∞ is compact. Let (Gα )α∈I be an open cover of
X∞ . Then there exists αo ∈ I such that ∞ ∈ Gαo . By definition, the set Gαo \ {∞} = Uαo
S
is open in X, and X \ Uαo is compact in X. We have X \ Uαo ⊂ (Gα \ {∞}), and each
α∈I
Gα \ {∞} is open in X. Therefore, there exist a finite number of indexes α1 , α2 , · · · , αn

. Luong Ha - HUCE
§ 1. Compact Spaces 127

S
n S
n
in I such that X \ Uαo ⊂ (Gαi \ {∞}). Hence, X∞ ⊂ Gαi . Thus, X∞ is compact.
i=1 i=0
Next we show that X is dense X∞ . Let U be a nonempty open set in X∞ . If ∞ ∈
/ U,
then U is open X, and X ∩ U = U 6= ∅. Suppose that ∞ ∈ U . Then U = {∞} ∪ V with
V ∈ τ and X \ V is compact in X. Since X is not compact, we have V 6= ∅. Hence,
U ∩ X = V 6= ∅. The theorem is proved.¥
The pair (X∞ , i) is called the one-point compactification or the Alexandrov compacti-
fication of X.
1.4.4 Example.
1) Let X = [0; 1) be the subspace of R with the usual topology. Then the one-point
compactification of X is the space [0; 1].
2) Let R∞ be the one-point compactification of R. Denote by S 1 be the unit circle in
R2 . Then S 1 is homeomorphic to R∞ .
Let h : S 1 → R∞ defined by

∞ if x = 1
h(x, y) = y
 if x 6= 1.
1−x
The mapping h is bijective, and its inverse k = h−1 is defined by

(1, 0) if t = ∞
k(t) = ¡ t2 − 1 2t ¢
 , if t ∈ R.
t2 + 1 t2 + 1
Since both h and k are continuous, the assertion is proved. In this case, we can say that
the one-point compactification of R is the unit circle S 1 .
1.4.5 Theorem. The space X∞ is a T2 -space if and only if X is T2 .
Proof. If X∞ is a T2 -space, then it is easy to see that X is T2 .
Conversely, let x, y be any distinct elements of X∞ . If both x and y belong to X, then
there are two open sets U, V in X such that x ∈ U, y ∈ V and U ∩ V = ∅. By definition,
U and V are open in X∞ .
Suppose that x ∈ X and y = ∞. Then x has a closed and compact neighbourhood U
in X. Hence the set V = {∞} ∪ (X \ U ) is open in X∞ . Clearly, U ∩ V = ∅. ¥
Next are some important properties of locally compact spaces.
1.4.6 Theorem. Let X be a locally compact Hausdorff space. Let A be a compact set in
X and U be an open set containing A. Then there exists a continuous function f : X → R
such that
a) 0 ≤ f (x) ≤ 1 for all x ∈ X;
b) f (x) = 1 if x ∈ A;
c) f (x) = 0 if x ∈
/ U.

Proof. Let X∞ be the one-point compactification of X defined as in Theorem 1.4.3.

By Theorem 1.4.4, X∞ is T2 , and hence, it is a normal space. The set A is compact in X,

. Luong Ha - HUCE
128 Chapter 6. Compactness and Connectedness

it is compact in X∞ . Therefore, A is closed in X∞ . In addition, X∞ \ U is closed in X∞

and A ∩ (X∞ \ U ) = ∅. Since X∞ is normal, there is a continuous function f ∗ : X∞ → R
such that
i) 0 ≤ f ∗ (x) ≤ 1 for all x ∈ X∞ ;
ii) f ∗ (x) = 1 if x ∈ A;
iii) f ∗ (x) = 0 if x ∈ X∞ \ U .
Let f be the restriction of f ∗ on X, then f is the desired function. ¥
1.4.7 Corollary. Let X be a locally compact Hausdorff space. Then for every closed
set F in X and for any point xo ∈
/ F , there exists a continuous function f : X → R such
that:
a) 0 ≤ f (x) ≤ 1 for all x ∈ X;
b) f (xo ) = 1;
c) f (x) = 0 for all x ∈ F .
Proof. Let A = {xo }. Then A is compact, and U = X \ F is an open set which
contains A. Using Theorem 1.4.6 yields the results.¥
1.4.8 Corollary. Every locally compact Hausdorff space is regular.
Proof. Obviously, X is T1 . Let x ∈ X and let F be a closed set in X, not containing
x. Using the function f in Corollary 1.4.7, let
1 1
U = f −1 ((−∞, )), V = f −1 (( , +∞)).
2 2
It is clear that U and V are disjoint open sets in X, and x ∈ V, F ⊂ U .¥

PROBLEMS
6.1 Let X be a compact space. Suppose that (Fn )n is a decreasing sequence of closed sets
T
∞
in X such that Fn = ∅. Show that there exists no ∈ N such that Fno = ∅.
n=1
6.2 Let X be a cofinite topological space. Show that X is compact.
6.3 Let X be a compact space and let Y be a Hausdorff space. Let f : X → Y be a
continuous mapping. Show that f is a closed mapping.
6.4 Let X, Y be topological spaces with X is compact. Show that the projection from
the product space X × Y onto Y is a closed mapping. (Recall that p : X × Y → Y is
defined by p(x, y) = y.)
6.5 Let f : X → Y be a continuous mapping, where X is a compact space and Y is a
Hausdorff space. Show that f (A) = f (A) for any A ⊂ X.
6.6 Let f : X → Y be a continuous bijection from a compact space X onto a Hausdorff
space Y . Show that f is a homeomorphism.
6.7 Let τ and Φ be two Hausdorff topologies on a set X such that Φ ⊂ τ and (X, τ ) is

. Luong Ha - HUCE
§ 2. Connected Spaces 129

compact. Show that τ = Φ.

6.8 Let K be a compact set in a Hausdorff space X. Suppose that thre are two open sets
G1 , G2 in X so that K ⊂ G1 ∪ G2 . Show that there exist two compact sets F1 and F2 in
X such that F1 ⊂ G1 , F2 ⊂ G2 and K = F1 ∪ F2 .
6.9 Let X be a topological space. Suppose that the intersection of any sequence of dense
open sets in X is dense in X. Show that X is of the second category.
6.10 Let X be a Hausdorff locally compact space. Show that the intersection of any
sequence of dense open sets in X is dense in X, and hence, X is of the second category.
6.11 Let X be a locally compact Hausdorff space. Let A be a compact set in X and let
V be an open subset of X containing A. Show that there exists an open set U such that
U is compact and A ⊂ U ⊂ U ⊂ V .
6.12 Show that every open subspace of a locally compact Hausdorff space is locally
compact.
6.13 Let X be the set of positive numbers with the discrete topology. Show that the
one-point compactification of X is the subspace of the reals with the usual topology
1
Y = { : n = 1, 2, ...} ∪ {0}.
n

§2 CONNECTED SPACES

2.1 CONNECTED SETS AND CONNECTED SPACES

2.1.1 Definition. A topological space X is called connected if only the empty set and X
are both open and closed in X.
This definition can be reformulated in many ways. In fact, the following statements
are equivalent:
1) X is connected.
2) X cannot be expressed as the union of two disjoint nonempty open subsets.
3) X cannot be expressed as the union of two disjoint nonempty closed subsets.
4) There is no nonempty proper subset of X which is both open and closed.
A set A ⊂ X is called a connected set if the subspace A is connected.
A topological space which is not connected is called disconnected.
2.1.2 Example.
1) Every indiscrete space is connected.
2) Let X be a discrete space. If X has more then one point, then X is disconnected.
3) Let X = {a, b} equipped with the topology τ = {∅, X, {a}}. Then X is connected.
4)The set of rational numbers Q is a disconnected set in R. Pick an irrational number

. Luong Ha - HUCE
130 Chapter 6. Compactness and Connectedness

c. Then
Q = [Q ∩ (−∞; c)] ∪ [Q ∩ (c; +∞)].
Since the two sets on the righthand side are nonempty, disjoint and open in Q, the set Q
is not connected.
Similarly, the set of irrational numbers is a disconnected subset of R.
5) The real line R is connected.
Suppose that there is a nonempty proper subset E of R, which is both open and closed.
Since E 6= R, there is an element c ∈ R \ E. It follows that E ⊂ R \ {c}, and hence

E = E ∩ (R \ {c}) = [E ∩ (−∞; c)] ∪ [E ∩ (c; +∞)].

Therefore, at least one set of the righthand side is nonempty, say A = E ∩ (−∞; c) 6= ∅.
The number c is an upper bound of A, there exists a = sup A. Hence, there is a sequence
(xn )n ⊂ A converging to a. On the other hand, (xn )n ⊂ E and E is closed, we must have
a ∈ E. Thus, a < c. The set E is also open, there exists β > 0 such that (a−β; a+β) ⊂ E.
Choose a real number y with a < y < min (a + β; c). Then y ∈ E and y < c. Thus, y ∈ A.
This is a contradiction since a < y and a = sup A.
2.1.3 Theorem. Let A, B be subsets of a topological space X and A ⊂ B ⊂ A. If A is
connected, so is B.
Proof. Suppose that B is not connected. Then there exist two open sets U, V in B
such that U ∩ V = ∅ and B = U ∪ V . By hypothesis, A = (A ∩ U ) ∪ (A ∩ V ). The sets
A ∩ U and A ∩ V are open in A and disjoint. Since B ⊂ A, and U, V is open in B, we
have A ∩ U and A ∩ V are nonempty. This contradicts the connectedness of the set A.
Thus, B is connected.¥
2.1.4 Corollary. The closure of a connected set is connected.
2.1.5 Theorem. Let (Aα )α∈I be a family of connected sets of a topological space X. If
T S
Aα 6= ∅, then Aα is connected.
α∈I α∈I
S
Proof. Let A = Aα . Suppose that A = B ∪ C, where B, C are two nonempty open
α∈I
sets in A and B ∩ C = ∅. Pick a ∈ A, then a ∈ B or a ∈ C. Without loss of generality,
we may assume that a ∈ B. Hence, B ∩ Aα 6= ∅ for all α ∈ I. On the other hand,
S S
C = C ∩ ( Aα ) = (C ∩ Aα ), and C 6= ∅, there exists αo ∈ I such that C ∩ Aαo 6= ∅.
α∈I α∈I
Thus, Aαo = (Aαo ∩ B) ∪ (Aαo ∩ C), with B ∩ Aαo , C ∩ Aαo are open in Aαo , nonempty
and disjoint. This contradicts the connectedness of the set Aαo . ¥
2.1.6 Corollary. Let A1 , A2 , . . . , An be connected sets in a topological space X so that
S
n
Ai ∩ Ai+1 6= ∅ for each i = 1, 2, . . . , n − 1. Then Ai is connected.
i=1
Proof. For n = 1, the result is obvious. Suppose that the result holds true for all
n = 1, 2, ..., k. Let A1 , A2 , . . . , Ak , Ak+1 be connected sets in X such that Ai ∩ Ai+1 6= ∅
Sk
for each i = 1, 2, . . . , k. Then, by induction hypothesis, the set A = Ai is connected.
i=1

. Luong Ha - HUCE
§ 2. Connected Spaces 131

Moreover, since Ak ∩ Ak+1 6= ∅, we have A ∩ Ak+1 6= ∅. By Theorem 2.1.5, A ∪ Ak+1 is

S
k+1
connected, that is, Ai is connected.¥
i=1
2.1.7 Theorem. Let X, Y be topological space. Let f : X → Y be a continuous surjection.
If X is connected, so is Y .
Proof. Suppose that Y = M ∪ N , where M, N are disjoint nonempty open sets in Y .
By the continuity of f , the sets f −1 (M ) and f −1 (N ) are open in X, and disjoint. Since
f is surjective, the two sets are nonempty. This contradicts the fact that X is connected.
Thus, Y is connected. ¥.
2.1.8 Theorem. A subset of the real line R is connected if and only if it is an interval.
Recall that the intervals of the real line R are of the following forms:
(a; b), (a; b], [a; b), [a; b] (finite intervals)
(−∞; a), (−∞; a], [a; +∞), (a; +∞), (−∞; +∞) (infinite intervals)
An interval E can be characterized by the following property:

a, b ∈ E, a < x < b ⇒ x ∈ E

Proof. Any open interval is homeomorphic to R, so they are connected by Theorem

2.1.7. The other intervals are connected by Theorem 2.1.3.
Conversely, let E be a connected subset of R. We assert that for any x, y ∈ E, if
x < z < y, then z ∈ E. Suppose that z ∈ / E. Then E ⊂ R \ {z}, and hence

E = [E ∩ (−∞; z)] ∪ [E ∩ (z; +∞)].

The sets E ∩ (−∞; z) and E ∩ (z; +∞) are open in E, nonempty and disjoint. This
contradicts the connectedness of E.¥
2.1.9 Corollary. Let f be a real-valued continuous function on an interval I and let
t1 , t2 ∈ I. Then for each number m between f (t1 ) and f (t2 ), there exists t ∈ I such that
f (t) = m.
Proof. Since I is connected, by Theorem 2.1.7, f (I) is connected in R. Hence f (I)
is an interval. Consequently, if m is between f (t1 ) and f (t2 ), then m ∈ f (I).¥
In this case, we say that the function f has the intermediate value property.
2.1.10 Theorem. The product of two connected spaces is connected.
Proof. Let (a, b) ∈ X × Y . Then {a} × Y ⊂ X × Y is homeormorphic to Y and is
therefore connected. Similarly, X × {y} is connected for any y ∈ Y . But ({x} × Y ) ∩
(X × {b}) = {(x, b)}; so by Theorem 2.1.5, the set

Zx = ({x} × Y ) ∪ (X × {b})

is connected.
The set ∪ Tx is connected because it is the union of a collection of connected sets
x∈X
that have the point (a, b) in common. Since this union is equal to X × Y , the space X × Y
is connected. ¥

. Luong Ha - HUCE
132 Chapter 6. Compactness and Connectedness

2.1.11 Example.
1) In the plane with the usual topology, the unit circle S 1 = {(x, y) : x2 + y 2 = 1} is
connected, since it is a continuous image of the interval [0; 2π]. Indeed, let f : [0; 2π] → R2
be defined by f (t) = (cos t, sin t). Then f is continuous and S 1 = f ([0; 2π]).
2) The graph of any continuous function f : [a; b] → R, is a connected subset of the
plane with the usual topology. Setting g(t) = (t, f (t)) defines a continuous function from
[a; b] into the plane, and g([a; b]) is the graph of f .
3) The plane with the usual topology is connected. Each straght line through the origin
is connected because it is homeomorphic to the space of reals with the usual topology,
and the plane is the union of all such lines.
The most direct way of showing that two given spaces are homeomorphic is to con-
struct a homeomorphism. To show that two spaces are not homeomorphic is quite a
problem, one which can be extremely difficult. It is solved if one finds a topological
property possessed by one space but not by the other, for then the two spcaes are not
homeomorphic. Connectedness is a topological property and so we can conclude that, for
example, R and R \ {0} are not homeomorphic.
Similarly, one can show that no two of the intervals (a; b), (a; b] and [a; b] are homeo-
morphic.

2.2 COMPONENTS

2.2.1 Definition. Let X be a topological space. A component E of X is defined as a

maximal connected subset of X, that is, E is connected and E is not a proper subset of
any connected subset of X.
Clearly E is nonempty. The central facts about the components of a space are con-
tained in the following theorem.
2.2.2 Theorem. Let X be a topological space. Then each component is closed. The
components of X form a partition of X, i.e. they are disjoint and their union is X.
Every connected subset of X is contained in some component.
Proof. Let E be a component of X. Since E is connected, the closure E is connected
and E ⊃ E. By definition, we have E = E, that is, E is closed.
For each x ∈ X, let Cx be the union of all connected subsets of X containing x. Then
by Theorem 2.1.5, Cx is connected. Clearly, if C is connected set in X which contains x,
then C ⊂ Cx . Suppose that D is a connected subset of X containing Cx . Then x ∈ D,
hence D ⊂ Cx . This means that Cx is a component of x.
Let C = {Cx : x ∈ X}. We claim that C consists of all the components of X. Each
element of C is a component as proved above. On the other hand, if D is a component of
X, then D contains some point xo of X and so D ⊂ Cxo . But D is a component; hence
D = Cxo .

. Luong Ha - HUCE
§ 2. Connected Spaces 133

S
We now show that C is a partition of X. Obviously, X = {Cx : x ∈ X}. Hence,
we need only show that distinct components are disjoint or, equivalently, if Cx ∩ Cy 6= ∅,
then Cx = Cy . Let a ∈ Cx ∩ Cy . Then Cx ⊂ Ca and Cy ⊂ Ca , since Cx and Cy are
connected sets containing a. But Cx and Cy are components, hence, Cx = Ca = Cy
Finally, if E is a connected subset of X, then E contains a point x ∈ X and so
E ⊂ Cx .¥.
2.2.3 Remark. Let X be a topological space. For x, y ∈ X, we define xRy if there is a
connected subset of X that contains x and y. Then R is an equivalence relation on X.
The equivalence classes determined by R are precisely the components of X.
2.2.4 Example.
1) If X is connected, then X has only one component: X itself.
2) Let X = {a, b, c, d, e} with the topology
© ª
τ = ∅, X, {a}, {c, d}, {a, c, d}, {b, c, d, e} .
Then the components of X are {a} and {b, c, d, e}.
3) Let X = [0; 1] ∪ [2; 3] be the subspace of R with the usual topology. Then X has
two components; [0; 1] and [2; 3].

2.3 PATH-CONNECTED SPACES

2.3.1 Definition. Let X be a topological space. A continuous mapping f : [0; 1] → X is

said to be a path in X. Let a = f (0), b = f (1), then a, b are called the initial point and
the terminal point of this path, and f is a path in X joining a to b .
2.3.2 Definition. A topological X is called path-connected if for every a and b in X,
there is a path in X joining a to b.
A subset A of X is called path-connected if the subspace A is path-connected.
2.3.3 Example.
1) Any set X with the indiscrete topology is path-connected since any mapping from
[0; 1] to X is continuous.
2) In the space R2 , every open disk is path-connected. Let a = (a1 , a2 ), b = (b1 , b2 )
belong to an open disk D. Consider the mapping f : [0; 1] → R2 to be defined by
f (t) = (a1 + t(b1 − a1 ), a2 + t(b2 − a2 )).
Note that f (t) ∈ D for any t ∈ [0; 1]. The mapping f is continuous since its coordinate
functions are continuous, and f (0) = a, f (1) = b.
Generally, each open ball or closed ball in Rk is path-connected.
3) Each space Rk is path-connected. Consider a = (a1 , a2 , ..., ak ), and b = (b1 , b2 , ..., bk )
in Rk . Let f : [0; 1] → Rk be defined by
f (t) = (a1 + t(b1 − a1 ), a2 + t(b2 − a2 ), ..., ak + t(bk − ak )).
Then f is a path in Rk joining a to b.

. Luong Ha - HUCE
134 Chapter 6. Compactness and Connectedness

2.3.4 Theorem. Every path-connected space is connected.

Proof. Let a ∈ X. For any x ∈ X, there exists a path fx in X from a to x. Let
T S
Ax = fx ([0; 1]). By Theorem 2.1.7, Ax is connected, and a ∈ Ax . Hence, X = Ax
x∈X x∈X
is connected by Theorem 2.1.5.¥
The converse of this theorem, in general, is not true. We can take the following
example. Consider two subsets of R2
x
A ={(x, y) : 0 ≤ x ≤ 1, y = , n ∈ N}
n
1
B ={(x, 0 : ≤ x ≤ 1}.
2
Thus, A consists of the points belonging to the line-segments joining the origin O with
the points (1, n1 ), n ∈ N, while B consists of the points on the x-axis, between 12 and 1.
The sets A and B are path-connected, and hence, connected. Then A ∪ B is connected,
but not path-connected since there is no path in A ∪ B from any point in A to any point
in B.

PROBLEMS
6.14 Let X be a topological space such that for any x, y ∈ X, there exists a connected
subset of X contains x and y. Show that X is connected.
6.15 Let X be a topological space. Let A ⊂ X and B is a connected set in X. Show that
if B has nonempty intersections with A and with the complement of A, then B ∩ ∂A 6= ∅.
6.16 Show that a cofinite topological space is connected.
6.17 Two sets A and B in a topological X is called separated if A ∩ B = ∅ and A ∩ B = ∅.
Show that X is connected if and only if X cannot be expressed as the union of two
separated sets.
6.18 Let A and B be two separated subsets of X and let Y be a connected subset of X.
Show that if Y ⊂ A ∪ B, then either Y ⊂ A or Y ⊂ B
6.19 Show that a space X is disconnected if and only if there exists a continuous mapping
from X onto a discrete space having two points.
6.20 Let A and B be closed subsets of a topological X such that A ∪ B and A ∩ B are
connected. Show that A and B are connected.
6.21 Show that a space X is connected if and only if each continuous function f : X → R
has the intermediate value property.
6.22 Let f : [a; b] → [a; b] be continuous. Show that f has at least one fixed point.
6.23 Show that the spaces Rn and R with n > 1 are not homeomorphic.
6.24 Let X be a topological space and let f : X → R be a function. Suppose that for
each x ∈ X, there is a neighbourhood U of x such that f is constant on U . Pick a ∈ X

. Luong Ha - HUCE
§ 2. Connected Spaces 135

and let b = f (a). Show that

1) f is continuous.
2) The set A = {x ∈ X : f (x) = b} is both open and closed.
3) If X is connected, then f is a constant function.
6.25 Show that each component of the set Q of rational numbers is a one-point set.
6.26 Let X be a topological space having only a finite number of components. Show that
each component is open.
6.27 Let f : X → Y be a continuous surjection. Show that if X is path-connected, then
Y is path-connected.
6.28 Let (Aα )α∈I be a family of path-connected subsets of a topological space X. Suppose
T S
that Aα 6= ∅. Show that Aα is path-connected.
α∈I α∈I
k
6.29 A nonempty set A ⊂ R is said to be convex if for any a, b in A, the segment joining
a with b, that is, the set of points of the form (1 − t)a + tb for all t ∈ [0; 1], is contained
in A.
Show that a convex subset of Rk is path-connected.
6.30 Show that every open connected set of Rk is path-connected.
6.31 In the space R2 , let
1
A = {(x, 0) : −1 ≤ x < 0} and B = {(x, sin ) : 0 < x ≤ 1}.
x
1) Show that A, B and A ∪ B are connected.
2) Show that A ∪ B is not path-connected.
6.32 Show that the product of any family of connected spaces is connected.

. Luong Ha - HUCE
136 Chapter 6. Compactness and Connectedness

HINTS and SOLUTIONS

6.1 Suppose that Fn 6= ∅ for all n ∈ N. Since F1 ⊃ F2 ⊃ · · · ⊃ Fn ⊃ · · · , the family of
T
∞
closed sets (Fn )n∈N has the property of finite intersection. Hence, Fn 6= ∅. This is a
n=1
contradiction.
6.3 Let F ⊂ X be any closed set. Then F is compact. By the continuity of f , the set
f (F ) is compact. The space Y is Hausdorff, so f (F ) is closed.
6.5 Let A ⊂ X. Then f (A) ⊂ f (A). Now the set A is compact since X is compact.
Hence f (A) is compact, and so it is closed. From the inclusion f (A) ⊂ f (A), we get
f (A) ⊂ f (A). Thus, we have the desired result.
6.8 From the hypothesis K ⊂ G1 ∪ G2 , it follows that (K \ G1 ) ∩ (K \ G2 ) = ∅. Let
A = K \ G1 , and B = K \ G2 . Note that X is T2 , the set K is closed. It follows that A
and B are closed in X, and A ⊂ K, B ⊂ K. Thus, A and B are two disjoint compact sets
of X. Hence, there exist two open sets V1 , V2 so that A ⊂ V1 , B ⊂ V2 , and V1 ∩ V2 = ∅.
Let F1 = K \ V1 , F2 = K \ V2 . Then F1 , F2 are compact in X.
We have F1 = K \ V1 ⊂ K \ A ⊂ G1 , and F2 = K \ V2 ⊂ K \ B ⊂ G2 . Moreover,
F1 ∪ F2 = (K \ V1 ) ∪ (K \ V2 ) = K \ (V1 ∩ V2 ) = K.
S
∞ S
∞
6.9 Suppose that X = An , where each An is nowhere dense. Then X = An , so
n=1 n=1
T
∞
(X \An ) = ∅. But An is nowhere dense, X \An is open and dense in X. By hypothesis,
n=1
T
∞ T
∞
(X \ An ) is dense in X. This contradicts the fact that (X \ An ) = ∅.
n=1 n=1
Thus, X is of the second category.
6.10 Let (Gn )n be a sequence of dense open sets in a Hausdorff locally compact space X.
T∞ T∞
We shall show that Gn is dense in X. Set G = Gn , and let V be a nonempty open
n=1 n=1
set in X. Then G ∩ V 6= ∅.
Indeed, since G1 is dense in X, we have G1 ∩ V 6= ∅. Pick x ∈ G1 ∩ V , then x has an
open neighbourhood C so that C is compact. The set G1 ∩V ∩C is an open set containing
x. The space X is Hausdorff locally compact, it is regular. Hence, there exists an open
set W1 with x ∈ W1 ⊂ W 1 ⊂ V ∩ G1 ∩ C. Since W 1 ⊂ C, we get W 1 is compact.
The set G2 is dense in X, so G2 ∩W1 6= ∅. Pick x ∈ G2 ∩W1 , then by a similar argument,
there is an open set W2 with x ∈ W2 ⊂ W 2 ⊂ W1 ∩ G2 and W 2 is compact. Proceeding
by this way, we find a sequence of nonempty open sets (Wn )n so that Wn ⊃ Wn+1 for each
n ∈ N, and W n is compact. Then (W n )n is a decreasing sequence of nonempty closed sets
T
∞ T
∞
in the compact space W 1 . Therefore, W n 6= ∅. Let K = W n , then K ⊂ W 1 ⊂ V .
n=1 n=1
T
∞ T
∞
Since W n ⊂ Gn for each n ∈ N, we have Wn ⊂ Gn , that is, K ⊂ G. It follows that
n=1 n=1
K ⊂ G ∩ V . Since K 6= ∅, we get G ∩ V 6= ∅.
Now by Problem 6.9, we conclude that X is of the second category.

. Luong Ha - HUCE
§ 2. Connected Spaces 137

6.11 From the hypothesis, it follows that X is regular. For each x ∈ A, there is an open
neighbourhood Vx of x such that V x is compact. By the regularity of X, there exists an
open neighbourhood Wx of x such that W x ⊂ V . Let Ux = Vx ∩ Wx . Then Ux is an open
neighbourhood of x. Since U x ⊂ V x , the set U x is compact. The family (Ux )x∈A is an
S
n
open cover of the compact set A, so there are x1 , x2 , · · · , xn in A so that A = Uxi . Let
i=1
S
n S
n
U= Uxi . Then U is an open set containing A. Note that U = U xi , so U is compact.
i=1 i=1
S
n
Furthermore, U ⊂ W xi ⊂ V .
i=1
6.14 Pick a ∈ X. Then for each x ∈ X, there exists a connected subset Cx of X
containing x and a. We have X = ∪ Cx and the family (Cx )x has a common point a, so
x∈X
X is connected.
6.16 Let X be an infinite set equipped with the cofinite topology. If X is not connected,
then X = A ∪ B, where A and B are disjoint non-empty open sets in X. Therefore, X \ A
and X \ B are finite. Moreover, (X \ A) ∪ (X \ B) = X. This is impossible since X is
infinite. Hence, X is connected.
6.18 Let Z = A ∪ B. Then
A ∩ Z = A ∩ (A ∪ B) = A ∩ A = A.
Hence, A is closed in Z. Similarly, B is closed in X. Now by hypothesis, we have
Y = Y ∩ Z = (Y ∩ A) ∪ (Y ∩ B). The rightthand side is the union of two disjoint closed
sets of Y . Since Y is connected, we must have either Y ∩ A = ∅ or Y ∩ B = ∅. This shows
that either Y ⊂ B or either Y ⊂ A.
6.20 Suppose that A is disconnected. Then there exist two disjoint closed sets U, V in A
such that A = U ∪ V . The set A is closed in X, so U, V are closed in X. Let M = A ∪ B
and N = A ∩ B. Then N = (U ∩ B) ∪ (V ∩ B). The set N is non-empty, we must have
U ∩ B 6= ∅ or V ∩ B 6= ∅, say U ∩ B 6= ∅. Thus, N = (N ∩ U ) ∪ (N ∩ V ). Note that
N ∩ U = B ∩ U 6= ∅, and N is connected, we have N ∩ V = B ∩ V = ∅.
Now M = A ∪ B = (B ∪ U ) ∪ V . We have (B ∪ U ) ∩ V = (B ∩ V ) ∪ (U ∩ V ) = ∅.
Thus, the righthand side is the union of two disjoint closed sets in M . This contradicts
the connectedness of M . Hence, A is connected. Similarly, B is connected.
6.22 Let g(x) = f (x) − x for each x ∈ [a; b]. Then g is continuous, and g(a) ≥ 0, g(b) ≤ 0.
If g(a) = 0 or g(b) = 0, then a or b is a fixed point of f . Suppose that g(a) > 0 and
g(b) < 0. The image of g is a connected subset of R, so it is an interval. Hence, there
exist x ∈ [a; b] such that g(x) = 0. This shows that x is a fixed point of f .
6.24
1) Let xo ∈ X. By hypothesis, there exists a neighbourhood U of xo such that f (x) =
f (xo ) for all x ∈ U . For any neighbourhood V of f (xo ), we have f (U ) = {f (xo )} ⊂ V .
Thus, f is continuous at xo , and hence, f is continuos on X.
2) Consider x ∈ A. Then there is a neighbourhood U of x so that f is constant on U ,

. Luong Ha - HUCE
138 Chapter 6. Compactness and Connectedness

that is, f (z) = b for any z ∈ U . Therefore, U ⊂ A, and it follows that A is open.
By 1), the mapping f is continuous, and A = f −1 (b), so A is closed.
3) By 2), the set A is both open and closed in X. Furthermore, A 6= ∅ since a ∈ A. If
X is connected, then A = X. This means that f is constant on X.
6.26 Suppose that X has a finite number of connected components C1 , C2 , · · · , Cn . We
S
n
know that each component Ci is closed, and Ci ∩ Cj = ∅ whenever i 6= j, and X = Ci .
i=1
¡S
n ¢
Thus, Ci = X \ Cj . Therefore, Ci is open.
j=1
j6=1

6.29 Let A ⊂ Rk be a convex set. For a, b ∈ Rk , consider the mapping f : [0; 1] → A to

be defined by
f (t) = (1 − t)a + tb.
We can verify that f is continuous, and f (0) = a, f (1) = b. Thus, A is path-connected,
and hence, is connected.
6.30 Let X be an open connected set in Rk . Fix a ∈ X. Let A1 be the subset of X
consisting of the elements x so that there is a path in X joining x to a. Clearly, A1 6= ∅
since a ∈ A1 . We will show that X = A1 .
Let A2 = X \ A1 . Then A1 ∩ A2 = ∅, and X = A1 ∪ A2 .
We first show that A1 is open in Rk . For each x ∈ A1 , consider C to be a path in X
joining x and a. Since X is open, there is an open ball B(x, ε) ⊂ X. Then each point
y ∈ B(x, ε) can be joined to x by a segment in X. Thus, each y ∈ B(x, ε) can be joined
to a by a path in X (a segment from y to x, then the path C). Hence, y ∈ A1 , and
B(x, ε) ⊂ A1 . Thus, A1 is open.
Now we show that A2 is open. Consider b ∈ A2 . Then there is an open ball B(b, ε) ⊂
X, and there is no path in X joining b to a. It follows that there is no path in X joining
a to any point of the ball B(b, ε). This shows that B(b, ε) ⊂ A2 , and A2 is open.
Thus, X = A1 ∪ A2 , where A1 , A2 are two disjoint open sets, and A1 6= ∅. Since X is
connected, we must have A2 = ∅, that is, X = A1 .

. Luong Ha - HUCE
 REFERENCES

[1] Aliprantis Charalambos D., Burkinshaw Owen. 1990. Principles of Real Analysis.
Academic Press, CA.

[2] Aliprantis Charalambos D., Burkinshaw Owen. 1999. Problems in Real Analysis.
Academic Press, CA.

[3] Brown A., Page A. 1970. Elements of Functional Analysis. Van Nostrand, London.

[4] Bryant Victor. 1994. Metric Spaces. Cambridge University Press, London.

[5] Cain G. 1993. Introduction to General Topology. Addison - Wesley, New York.

[6] Dugundji James. 1996. Topology. Allyn and Bacon, Boston.

[7] Folland Gerald B. 1984. Real Analysis: Modern Techniques and Their Applications.
Wiley Interscience, New York.

[8] Giaquinta Marianno, Modica Giuseppe. 2007. Mathematical Analysis: Linear and
Metric Structures and Continuity. Birkhauser, Boston.

[9] Kelley John. 1969. General Topology. Van Nostrand, New York.

[10] Kolmogorov A. N., Fomin S. V. 1975. Introductory Real Analysis. Dover Publications,
New York.

[11] Munkres James R. 2000. Topology. Prentice Hall, New York.

[12] Lipchutz Seymour. 1965. Theory and Problems of General Topology. Shaum’s Outline
Series, Singapore.

[13] Rudin Walter. 1970. Real and Complex Analysis. McGraw-Hill, New York.

[14] Seebach J. Arthur, Steen Lynn A. 1970. Counterexamples in Topology. Holt, Rinehart
and Winston, New York.

139
 INDEX

Fσ -set, 34 boundary point, 26

Gδ -set, 34 bounded, 65
I-tuple, 7 bounded above, 6
T1 -space, 104 bounded at every point, 80
T2 -space, 104 bounded metric, 43
T3 -space, 105 Bunhiakowski’s inequality, 21
T4 -space, 105
i-th coordinate, 7 cardinality, 8
n-tuple, 7 Cauchy sequence, 49
chain, 6
bounded below, 6 closed, 27, 88
characteristic function, 101 closed ball, 26
component, 132 closed mapping, 99
Hausdorff’s Theorem, 67 closure, 30, 89
Heine-Borel’s Theorem, 69 closure operator, 92
onto, 5 cofinite topological space, 87
surjective, 5 cofinite topology, 87
uniformly continuous , 38 compact, 66, 122
Zorn’s lemma, 125 compact metric space, 66
compactification, 126
accumulation point, 89
comparable, 87
adherent point, 28, 89
complete, 49
Alexandrov compactification, 127
completeness axiom, 13
algebraic number, 18
completion, 54
antisymmetry, 6
composition, 4
Approximation Theorems, 72
connected, 129
Archimedean property, 15
connected set, 129
Ascoli-Arzela theorem , 78
continuous at a point, 35, 96
at most countable, 10
continuous on a set, 35
axiom of choice, 7
continuum axiom, 17
axioms of separation, 104
continuum cardinality, 17
Baire Category Theorem, 55 continuum hypothesis, 17
basis for the topology, 93 contraction constant, 51
bijection, 5 contraction mapping, 51
boundary, 27 Contraction Mapping Principle, 51

140
 INDEX
141
converge pointwise, 100 graph, 77
converge uniformly, 100 greater cardinality, 9, 17
convergent, 23 greatest lower bound, 6
convex, 135
Hausdorff space, 104
coordinate convergence, 24
homeomorphic imbedding, 126
countable, 10
homeomorphic to, 39, 98
countable cardinality, 17
homeomorphism, 38, 98
countably locally finite, 115
cover, 69 identity mapping, 4
image, 4
decreasing sequence , 123 inclusion, 96, 126
dense, 90 index set, 2
diameter, 59 indices, 2
difference, 2 indiscrete topology, 86
disconnected, 129 induced topology, 90
discrete metric, 22 infimum, 6
discrete metric space, 22 infinite, 8
discrete topology, 86 initial point, 133
disjoint, 2 injection, 5
distance, 20 injective, 5
distance from x to the set A, 22 interior, 30, 88
distributive laws, 2 interior operator, 92
domain, 4 interior point, 26
intermediate value property, 131
elements , 1
intersection, 2
empty set, 1
inverse, 5
equicontinuous at a point, 78 inverse image, 4
equivalence relation, 5 isometric, 39
extension, 4 isometry, 39

family of sets, 2 largest element, 6

field axioms, 13 least upper bound, 6
finite, 8 Lebesgue number, 77
finite cover, 69 lesser cardinality, 9
finite intersection property, 123 limit of the sequence, 23
first axiom of countability, 113 limit point, 89
first axiom of separation, 104 Lindelöf space, 95
first category, 56 linear order, 6
first countable space, 113 local basis, 112
fixed point, 51 locally compact, 126
fixed point theorem, 52 locally finite, 114

. Luong Ha - HUCE
 INDEX
142

lower bound, 6 proper subset , 1

lower semicontinuous, 100
quotient space, 102
mapping, 4 quotient topology, 102
maximal element, 6
range, 4
members, 1
reflexivity, 5
method of successive approximations, 52
regular space, 105
metric, 20
relation, 5
metric space, 20
relatively compact, 78
metric subspace, 25
restricted mapping, 97
metrizable space, 112
restriction, 4
metrization problem, 113
minimal element, 6 Schröder - Bernstein theorem, 9
second axiom of countability, 113
Nagata-Smirnov metrization theorem, 115
second axiom of separation, 104
neighbourhood, 26, 87
second category, 56
nested, 16, 50
second countable space, 113
normal space, 105
segment, 135
nowhere dense, 55
separable, 32, 90
one-point compactification, 127 separated, 134
one-to-one, 5 sequence, 23
open, 27 smallest element, 6
open ball, 26 Sorgenfrey line, 95
open cover, 69 stationary sequence, 24
open mapping, 99 subcover, 69
order axioms, 13 subsequence, 4, 23
ordinary (usual, Euclidean) metrics, 21 subset, 1
subspace, 90
partial order, 6
supremum, 6
partially ordered, 6
surjection, 5
partition, 6
symmetric difference, 2
path, 133
symmetry, 5
points, 1
pointwise limit, 100 terminal point, 133
polynomial functions, 72 topological property, 99
power set , 2 topologically equivalent, 40
preimage, 4 topology, 86
product of two metric spaces, 24 total order, 6
product space, 101 totally bounded, 65
product topology, 101 totally ordered, 6
projection, 7 transitivity, 5

. Luong Ha - HUCE
 INDEX
143

trigonometric polynomial, 74
Tychonoff’s theorem, 125

uncountable, 10
uniformly convergent, 24
uniformly equicontinuous, 81
uniformly equivalent, 40
union, 2
upper bound , 6
upper semicontinuous, 100
Urysohn lemma, 108
Urysohn metrization theorem, 113
usual topologies, 87

Zorn’s lemma, 7

. Luong Ha - HUCE