ELEMENTS

OF
POINT SET
2OPOLOGY

- 2 ae s
‘ ; |

| ! j ee »
=

John D. Baum
 ELEMENTS OF
POINT SET TOPOLOGY
by JOHN D. BAUM

Tuis Is a guide to that area of mathe-

matics that lies beyond geometry, yet
on this side of analysis and advanced
point set and algebraic topology:
namely, elementary topology.

Like all effective guides, it ap-

proaches the unfamiliar in terms of
the familiar. In this case, topology is
approached in axiomatic and geo-
metric terms: axiomatic, to harmonize
with the current treatment of modern
algebra and the reader's experience
with that treatment; and geometric,
to draw as fully as possible on the
reader's knowledge of geometry.

In early chapters, to ease the read-

ers introduction to axiomatic mathe-
matics, the author provides highly
detailed proofs. In later chapters,
proofs are less detailed and problems
are increasingly difficult. Examples
and exercises illustrate every idea.

The principal topics presented are

those which are essential to the pur-
suit of modern analysis: introductory
set theory; abstract spaces; continuous
functions and homeomorphisms; com-
pact spaces; connected spaces; and
matrix spaces.

Digressions are made, however, to

topics which are more essential to the

(Continued on back flap)

 i

|
=

fet
ool
:

:
:
—
i

= @e-

<<

a
:
” =

<=
=

eam

; ,
ne —_

*ie

yi

,
=
;
 Digitized by the Internet Archive
In 2023 with funding from
Kahle/Austin Foundation

httos://archive.org/details/elementsofpointsOO000john
ELEMENTS

OF

POINT SET

TOPOLOGY
 PRENTICE-HALL INTERNATIONAL, INC., London
PRENTICE-HALL OF AUSTRALIA, PTY., LiD., Sydney
PRENTICE-HALL OF CANADA, LTD., Toronto
PRENTICE-HALL OF INDIA (PRIVATE) LTD ,New Delhi
PRENTICE-HALL OF JAPAN, INC., Tokyo
PRENTICE-HALL DE MEXICO, 8.A., Mexico City
QA
611
.B386
1964

ELEMENTS
OF
ROUNL ord
TOPOLOGY

JOHN D. BAUM
Professor of Mathematics
Oberlin College

Laje ayy, (fs

PRENTICE-HALL, INC.
Englewood Cliffs, N.J.
 © 1964 by
PRENTICE-HALL, INC.
Englewood Cliffs, N.J.

Current printing (last digit):

12QonlieelOe
Ole Saad Om om ao

All rights reserved. No part of this book may

be reproduced in any form, by mimeograph or
any other means, without permission in writing
from the publisher.

Library of Congress Catalog Card Number 64-13253

Printed in the United States of America

C-26924
To My Mother
 PREFACE

In writing this book I have kept in mind the undergraduate at the

small college. The importance of introducing such a student to modern
algebra early in his college career has generally been emphasized. There
seems to be no similar emphasis, however, on the necessity of introducing
topology into the undergraduate curriculum. Feeling, as I do, that
functional analysis provides one of the major syntheses in mathematics,
and knowing the fundamental importance of both algebra and topology
as bases for analysis, I have tried to provide an introductory text which
in spirit parallels many of the present texts in modern algebra. I am
hopeful that in the not too distant future a one-semester course in
topology will be available to most undergraduate mathematics majors,
so that a course in modern analysis can also be offered to the under-
eraduate student.
The spirit of this book is geometric and axiomatic. The reason for
emphasizing the geometric approach is that I believe the student already
has had substantial experience with geometry. Hence his access to
topology will be easiest if he can make use of his geometric intuition.
The reason for the axiomatic approach is twofold: first, it parallels the
student’s experience in modern algebra; second, it keeps this book in
harmony with the current trends in mathematics.
Besides the principal aim enunciated above, I have had two second-
ary aims. Since point set topology in an introductory course serves as a
foundation not only for analysis but also for further work in point set
topology and in algebraic topology, I have tried to include topics which
will serve students who have interests other than analysis. The main
sequence of topics is, of course, the set of those which are of importance
in analysis. There are, however, digressions from this main sequence
into topics which are of lesser importance in analysis, but which will be
of interest to students who wish to pursue topology, either point set or
algebraic.
In another sense the intended audience has determined the form of
the book. The proofs at the beginning are presented in great detail.
They will appear overly detailed to the professional mathematician.
However, it has been my experience that the student who comes to
vu
Vill PREFACE

axiomatic mathematics for the first time through a course of this nature
requires a bit of help in the early stages. In the later chapters the student
will be more dependent on his own resources because the proofs no
longer contain quite so much detail. What has been said about proofs
applies equally to problems. At the beginning of the book they are
simple, sometimes trivial; whereas in the last chapter particularly, they
require considerably more from the student. There has been no effort
made to supply original sources for the notions presented in the text.
Instead the references have been supplied to suggest further reading to
the interested student. In a number of these reference works a detailed
list of original sources can be found, so it seemed redundant to include
them here.
Having used preliminary forms of the text with students at Oberlin
College, I find that the material in the text can be covered without
strain in one fifteen-week semester. If it appears that this is too heavy
a schedule, some sections may be omitted without disturbing the con-
tinuity of the text: Section 8 of Chapter 1, Sections 3 and 4 of Chapter 3,
Sections 4 and 5 of Chapter 4, and Section 3 of Chapter 5 may be omitted
if a shorter course is necessary. The problems contain many of the
essential examples necessary to illustrate the theory, and it is urged
that most if not all of the exercises be assigned. The few exercises
referred to as “‘term papers” are suggested as subjects for further inves-
tigation by the student and as teaching aids for the instructor who
prefers to assign a term paper rather than a final examination of the
traditional sort.
Theorems, definitions, lemmas, and corollaries are numbered con-
secutively in each chapter. Thus the tenth object of this sort in Chapter
3 is numbered 3.10. Thereafter the number 3.10 is used to refer to this
object. Exercises are numbered consecutively in each chapter beginning
with 1; for example, the twelfth exercise in Chapter 2 is simply numbered
2.12 and is thereafter referred to as Ex. 2.12. The end of the proof of a
theorem is indicated by the symbol “‘f,”’ rather than by a statement to
the effect that the proof is completed.
I acknowledge with gratitude valuable suggestions received from
Professors R. H. Bing and M. L. Curtis. I wish to thank also Ruth
Edwards and Elizabeth Carter for help in typing the manuscript and a
group of students too numerous to mention individually for finding
many a typographical error in early versions of the book. In spite of all
this, the ultimate responsibility for any errors lies with me.

Joun D. Baum
 CONTENTS

CHAPTER 0 PRELIMINARIES, 1

. Introduction, 1
Sets, 1
. The Algebra of Sets, 3
Euler-Venn Diagrams, 7
Relations, 9 -
. Infinite Sets, 11
. Miscellaneous Assumptions Regarding the Real Numbers, 16

CHAPTER | TOPOLOGICAL SPACES—BASIC DEFINITIONS

AND THEOREMS, 19

. Neighborhood Systems and Topologies, 19

. Open Sets in a Topological Space, 23
Limit Points and the Derived Set, 27
The Closure of a Set, 28
. Closed Sets, 31
Subspaces, 36
. Limits of Sequences; Hausdorff Spaces, 38
. Comparison of Topologies, 42
CONAaARWNE
. Bases, Countability Axioms, Separability, 438
—= Sub-bases, Product Spaces, 50

cHAPTER 2 CONTINUOUS FUNCTIONS (MAPPINGS) AND

HOMEOMORPHISMS, 56

by Functions, 56
2. Continuous Functions (Mappings), 58
3. Homeomorphisms, 60
4, Product Spaces, 65
ix
X CONTENTS

cHAPTER 3. VARIOUS SPECIAL TYPES OF TOPOLOGICAL

SPACES (VARIETIES OF COMPACTNESS), 69

Compact Spaces, 69
Separation Axioms, 79
Countable Compactness, 89
Bee
IN Local Compactness, 92

CHAPTER 4 FURTHER SPECIAL TYPES OF TOPOLOGICAL

SPACES (MOSTLY VARIETIES OF
CONNECTEDNESS), 98
1. Introduction, 98
2. Connected Spaces, 99
3. Components, 105
4. Local Connectedness, 107
5. Arewise Connectedness, 109

CHAPTER 5 METRIC SPACES, 115

Definitions, 115
Some Properties of Metric Spaces, 120
Metrization Theorems, 125
Complete Metric Spaces, 132
oo
orm
NoCategory Theorems, 136

REFERENCES, 145

INDEX, 147
 CHAPTER 0

PRELIMINARIES

§1 Introduction

Most mathematical fields these days begin with a collection ot

undefined objects and a set of axioms which govern the behavior of the
objects. There are many advantages to such an approach to a mathe-
matical science. Perhaps the greatest of these is that any mathematical
system we encounter that obeys the axioms of a particular mathematical
field will also obey all the theorems that are true in that field. Since any
axiomatic development begins with a set of undefined objects, it is
essential that before we start on the axiomatic development itself we
study some set theory.

§2 Sets

Set theory itself, being a field of mathematics, has an axiomatic

development, but we shall not take that view here, but rather shall
develop the theory by relying to a considerable extent on our intuition.
Following Cantor we take the word “‘set’”’ to mean “any collection into
a whole, M, of definite and separate objects, m, of our intuition or our
thought.”” The objects which are collected to form the set are called
elements or members of the set in question; the whole set, however,
being thought of as a single entity. Generally we shall designate sets by
capital letters, either Roman, A, B, etc., or script @, ®, etc., or German,
%, %, etc. The elements of the sets will generally be designated by
lower-case Roman letters.
We have two ways to indicate what the members of a set are. We
can, should the set be a small one, simply list the elements of the set.
We do this in a consistent fashion, by writing the list in curly brackets,
or braces, e.g., A = {1, 2, 3}, separating each element from the next by
acomma. If the set happens to be large this method is cumbersome, and
1
2 PRELIMINARIES cHaAP. 0

we prefer then to designate the set by giving some common property

that all the elements of the set enjoy, e.g., A = {x| is a positive
integer less than 4}. Again our notation is consistent; we use braces,
indicate a generic element of the set (in this case, x) by some letter
followed by a vertical line, and then write the common property which
the generic element as well as all the elements of the set enjoy.
We indicate that a particular element is a member of a set by
writing, for example, 2 € A. We read this as “two is an element of the
set A,” or “two is a member of A,” or simply “two belongs to A.”
Should an element not belong to a set, we write for example, 4 ¢ A, and
read this as ‘four does not belong to A”’ or any of the variants suggested
above.
A set A is a subset of, or is contained in, a set B if for each x € A,
we also have x € B. In this case we write A C B or BD A. If both
ACB and BCA, the sets A and B have exactly the same elements
and we write A = B. We reserve the symbol, @, for the empty set,
i.e., the set with no elements whatever, and observe the following
universal truths: AC A, A = A, @ CA, and A C @ if and only if
A = ©. We also note that ‘‘C” is a transitive relation, 1.e., A C B
and BC Cimply A &.¢.
The notation A C B, which means A C B and A B, is also en-
countered occasionally, though we shall have little if any use for it. To
deny that one set is contained in another, for example, to assert that
A C B is false, we write A J B.
The student is warned here for the first time, as he will be a number
of times as he proceeds through the book, that mathematical notation
is highly variable. This is the case with the notation for set inclusion,
namely “‘C,” and proper set inclusion, namely ‘“‘C.’’ There are authors
who use simply the symbol “‘C”’ for set inclusion and never introduce
a notion of proper set inclusion. In reading other works the student
should thus check on the ways in which the various notations are
defined.

EXERCISES

0.1. Prove the universal truths mentioned above, i.e., for any set A
each of the following is true:
(a) ACA.
(b) A= A;
see. 0.3 PRELIMINARIES 3

(C)ine@ SSiAn
(d) A C @ if and only if A = @.

0.2. For any sets A, B, and C, prove the following:

(a) Ii A C6 and BGC. then A € C.
(pb) MIPANG Band's.C, then AC Cc,

§3 The Algebra of Sets

We now define two operations on sets. The first is the union of two
sets, written A U B, and defined by AUB = {x|x€ A orx € B}.
The “or” is used in its nonexclusive sense, so that an element which is
in both A and B is still in A U B. The second operation on two sets is
the intersection operation, written A ( B, and defined by A (™\ B =
{x |a € A and x € B}. The following rules of operations on sets now
follow quite straightforwardly from the definitions:

Li) AS BAS OC) = (AUB),

(PAB ©) = (4B) NC.
2 (a) AWB i= Bs UA.
(Dear LB ="B-\A.
Bria Ae UO) (AYN BY OTA YO):
(Dye By Ors (AN BY TN (AV) €).
4, (a) AUA =A,
(b) ANA=A,
Db, (a) ACS AWE,
(yeas ANB,
6) @) AC Cand BCC imply AW) BCC.
(by A> Cand BD Cimply AM BDC.
ta (Gy BAIS = A
(DA)
=
As remarked earlier A = B if and only if A C Band B C A. Thus,
in order to prove most of the above statements we must prove two
inclusions. To demonstrate this we prove 2(a). Let « € A UB, then
x€ A or x € B. Consequently x € B or x € A, whence x € BUA;
and it follows that AU BCBUA. Let x € BUA, then x € B or
x € A. Consequently x € A or x € B, whence x € A U B; and it fol-
lows that B.A CA WB. From the two inclusions just proved it
follows that AU B= BUA.
4 PRELIMINARIES CHAP. 0

We shall in general assume that the sets we are discussing are all sub-
sets of some universal set, U’, and then we define the complement of A,
which we designate by A‘, by A* = {x| az € Uanda ¢ A}. The differ-
ence of two sets A and B we defineas A — B = {x |x € Aanda ¢ B}.
The further list of rules of operation which follow from these definitions
is:

8. (a) Oe =U.
(b) Us =
9. (a), AAS U,
(b) AM At = ©,
Made AP
11. If A CB, then A* D B’ and conversely.
2 Ge CAR IB \ot eA Gas
(aE GANG le =A ele
13..A,= B= Ait\ be

Rules 12(a) and (b) above are called DeMorgan’s rules; we shall
encounter more general forms of these two shortly. The student should
convince himself of the truth of the above rules by proving each of them.
Once again the student is warned that the notation used for complements
of sets, namely A’, differs from what he is likely to find in other texts.
The list we now have is probably sufficient for our needs; however,
a set of exercises follows which gives a number of further relations
which may occasionally be useful.

EXERCISES

Prove each of the following, where A, B, and C designate sets in some

universal set U.

0.3. A C B does not in general imply B C A.

O47 AGB itand only

if BB.
0.5. AC Bifandonlyif
AN B= A.

0.6.40 (B-C)=BM(A-—-C)=(ANB)-C=
(A MB)
— (ANNO).
0.7. A—-B=A—(AL\B).
sEC. 0.3 PRELIMINARIES 5

Use OVS = Ont and only i 4 Cc Be

O92, IVA OB =U and

A B= O, then B= Ae
0.10. (A— BY =BUA.
On Gob) (A =) 6) SA = (BNC).

ie OOM aay eA ewan

Og eleey Bi A eA OB)= EB).
OMe = By A B,
Oh ON Wee ons
Ole (Bi A) = oe
It is clear that by using the associative law for unions (or inter-
sections), l(a) and 1(b) above, we may take the union (or intersection)
of any finite collection of sets. It is useful, however, to be able to take
unions and intersections of arbitrary families of sets, not merely finite
families. In order to define such a notion we have recourse to an indexing
set, or as it is also called, a set of indices. We can think of an indexing
set _as a set of tags, so that each member of a family of sets has associated
with it a tag from the indexing set. Thus, let A be a set of indices, and
with each a € A associate a set B,, then define

U B, = {«|x € B. for some

a € A}
aA

and
(\) Bo= {¢\a€
B, for alla € A}.
acA bi

If A is the empty set, we define

OP Ete) and ig
aGA aA

The following rules of operation then follow:

14. Ifa € A, then 1 BeC B.C U BAD.
aECA alA

15, (@) litor cach ac 4, Bb, SC, then") Bec C.

aEA

(b) If for eacha € A, Bg DC, then 1M BDC.

acCA
6 PRELIMINARIES CHAP. 0

16. If for each a € A (AH DO), BGC, then (Ob, CeCe

acA acCA

and (ey batoe (ole

aA aEA

17: 4) SBN Ca = CO eB Dee Ca:

aA aA acA

) A GonaA
Ce COnBo
aA
Cen).
aA

18, 1), (CSB)

= CECE Boras.
aEA aA
19. OF (CAB) =O COB) for Aso
aGA aA

205-(8) WN 2Bey Sali Be

aA aA
x (DeMorgan’s rules)
(b) (A Ba)? = U Ba.
aA aGA

As an alternative notation for unions and intersections of the type

just discussed, we occasionally do the following: Let § be a family of
sets {B}, then oo B = {x|x € B for some B € &} and similarly for
intersections.
We find occasional use for the following notation: if @ is a family
of sets {A} and X is some fixed set then @(\ X means the family
{A (\X|AE @}.
Finally, let X;,7 = 1, 2,..., be sets, then the Cartesian product

of these sets XK Xi = {(a1, X2,...,%n) |vi € X;}. This amounts to say-

i=l
ing that the Cartesian product of the n sets is just the set of all possible
n-tuples where the 7th member of each n-tuple is chosen from the set X ;.
In case there are only two sets we frequently prefer to write X; X Xin
2
place of X X;. For example, if A = {1, 2} and B = {a, b, c}; then
1=1
4=

AX B= {(1, a), G, 6), G, ¢), 2, a), 2, 6), 2, ¢)}.

We shall discuss Cartesian products at greater length at a later time.

EXERCISES

0.17. Verify that DeMorgan’s rules [20(a) and (b) above] hold in case
A = ©, as well as in the case A ¥ @.
sec. 0.4 PRELIMINARIES 7

0.18. The Cartesian product of the real line, R, with itself is the real
plane. If the first quadrant is the set

{(a, y) |e > 0, y > 0, 2, y real},

write the first quadrant as a Cartesian product.

0.19. If Rt = {x |x real, x > 0} and @ is the family of all subsets of Rt,

then if B = [—1, 1] = {«| x real —1 S x S 1} describe @(/M B.

§4 Euler-Venn Diagrams

In dealing with problems in set theory a picturization of the relations

between the sets is frequently helpful. Such a picture or diagram is
usually called a Venn diagram, or Euler diagram, or even Euler-Venn
diagram. Usually the universal set is pictured as a large rectangle and
the pertinent sets are designated by means of regions inside the rectangle.
For example, an Euler-Venn diagram which pictures A (\ B is shown
in Fig. 0.1.

Figure 0.1

It should be noted that the diagram itself does not prove a particular
result in set theory; however, it frequently gives some hint as to how a
proof can be constructed. This is demonstrated, for example, in Figs. 0.2
and 0.3 below which suggest an approach to the proof of the equality
A OURO ORS (ATS) OC AG):
8 PRELIMINARIES CHAP. O

Figure 0.3 [Indicating (A (\ B) U (AN C)]

EXERCISES

0.20. Draw an Euler-Venn diagram to illustrate each of the following

situations:
(a) Al ees
(b) "AT WBs2"O-
(c) AC Be
d@)ANB#Z,BNAC#Z,CAD#B,DNA#GD,
ANC=2 BMD =o
sec. 0.5 PRELIMINARIES 9

§5 Relations

We shall not study relations in general but confine our attention to

two types of relations of particular utility for the sequel. First, however,
we define the notion of relation in general. Attelationiitsemsssinply
J ookigusmttanniaesimnrrinnnieniinetnl acme eaECRE
designat@@aelationlbymmand write aRb to indicate that the ordered
pair (a, b) is an element of the collection which determines the relation.
An equivalence relation is a relation which has the following three
properties:

(1) @R@for each a € An@eflexivity)ig

(2) Ifquithen also bRa @ymmetry).
(3) If @idgand bRc, then also akc (@ranisitivityye
If R is an equivalence relation in a set A, then we designate by R(a)
or [a, R] or [a] the set, {b| 6 € A, aRb}, and we call this set, damkgntlag
dininelginetesinceeimiemmiaiogmenit by a partition of a set, A, we mean
a family, 5, of subsets of A which are disjoint (iJeqmBgrOmenaqmBusn@s
SS ae whose union is A, then the following important
theorem results:

0.1. Theorem. Hach equivalence relation in a set A induces a partition.

Samanta
: ai Savanei : inn
fas
Proof. Let § = {{a]|a.€ A}. For each a € A, a € [a], thus

U [a] =A
[a] Es

Now suppose [a] ( [b] # @, then we can select c € [a] ( [b], whence
by definition of [a] and [b] we have cRa and cRb and by the symmetry
and transitivity of R we have akb. Let d € [a], then dRa, and since
we already have akb, it follows from the transitivity of R that dRb,
whence d € [b]. Similarly for any e € [b], we have also e € [a], whence
[a] = [b]. Consequently the equivalence classes are either disjoint or
they coincide, and § is a partition of A.
Conversely let § be a partition of A. Define aRb if and only if a
and b belong to the same set B € &. It then follows simply that F is
10 PRELIMIN ARIES cHaP. 0

an equivalence relation and the equivalence classes of R are precisely

the sets of §. The details are left to the reader. §

Order relations are the second sort of relations we shall need to

consider. In general, order relations are relations which are transitive;
however, we distinguish two particular types of order relations which
we shall need. The first of these is a partial order relation. We define a
partial order relation in a set S to be a relation which satisfies the
following conditions:
(1) For any z € S, <Rhz.
(2) For any x, y € S, xRy and yRx together imply z = y.
(3) For any z, y, z € S, cRy and yRz together imply xRz.

A typical example of a partial order relation is the relation A € B in

the family of all subsets of some universe U.
The second order relation we shall need is the simple order relation.
For our purposes we define a simple order relation as a partial order
relation in which the further condition holds:
(4) If x, y © S, and x ¥ y then either xRy or yRz.
A typical example of a simple order relation is the relation x S y in the
set of all real numbers.
The student is warned that he may encounter definitions of simple
order relations different from the one above. In particular the following
occurs occasionally as a definition of a simple order relation:
A simple order relation in a set S is a relation in which the following
conditions hold:
(1) For any x, y € S, one and only one of the following three hold:
cRy, yRu, x = y.
(2) For any x, y, 2 € S, eRy and yRz together imply xRez.
A typical example of this definition of simple order relation is the
relation x < y in the set of all real numbers. For our purposes, however,
we shall prefer the previously given definition of simple order relation.

EXERCISES

0.21. In the set Z*+ of positive integers, let xRy mean x — y is divisible
by 7 (i.e., « — y = 7k, where k is an integer). Show that R is an
equivalence relation and describe the equivalence classes.
 sec. 0.6 PRELIMINARIES 11

0.22. In the set Z+ of positive integers, let’ m <n mean m divides n

(i.e.,n = mk, where k € Z+). Show that “‘S”’ is a partial ordering
for Zt.

§6 Infinite Sets

We shall define and discuss functions at somewhat greater length at

the beginning of Chapter 2, but for the moment we require a particular
type of function. We define a one-to-one correspondence (or function)
between A and B in the following way: first think of f (the function or
correspondence) as a set of ordered pairs (a, 6), with a € A, b € B, such
that if (a, b) and (a, b’) € f, thenb = b’. This says simply thatf is single
valued, for we think of f as assigning the value 6 € B to the point (or
element) a € A. We frequently write f(a) = b, which notation should
be familiar to the student from his study of calculus. Furthermore we
insist that each a € A appear as a first element in some ordered pair
(a, b) € f. This then assures us that f assigns to each point (or element)
a € A one and only one value b € B. Under these circumstances f is
called a function. In order that f be a one-to-one correspondence we
insist further that each b € B appear as a second element in some ordered
pair (a, b) € f, and further that if (a, b) and (a’, b) € f, then a = a’.
What this amounts to is that only one element of a has as its value, f(a),
the element b € B, and that each element b € B does occur as a value
of some a € A.
The net effect of a one-to-one correspondence is that it assigns to
each a € A exactly one b € B and to each b € B exactly one a € A.
A typical example of a one-to-one correspondence is the function
f(z) = x+ 1, where A and B are both the set of real numbers. An
example of a function which is not a one-to-one correspondence is the
function f(z) = 2? in which A is the set of all real numbers, while B is
the set of nonnegative real numbers.
A set is; said to7+be infinite
of BO if there NO exists a one-to-one; correspondence
the gi

set. The set is said to be countably inf ite if there is_a_one-to-one

respondence between Z* a1 1 . A set which is not infinite
is said to be finite, and a set which is either finite or countably infinite
is said to be gountable. The word ‘“‘denumerable” is frequently seen as
a synonym for countable.
The following are easy consequences of these definitions: if A is
12 PRELIMINARIES cHAP. 0

infinite and A C B, then B is infinite; if Bis finite and A C B, then A

is finite; if A is infinite, and a € A, then A
— {a} is infinite.
We now introduce an axiom which will make possible many of the
proofs that follow. They may be possible without the assumption of
this axiom, but at present this seems most unlikely. It is known that
the assumption of this additional axiom, the so-called axiom of choice,
does not introduce into the system we already have any inconsistencies
that are not already there. The axiom is this:

Axiom of Choice. Let § = {B,| a € A, A an indexing set} be a non-

empty family of nonempty disjoint sets; then there exists a set C, such that
C(\ B, ts a single element for eacha € A.

We can think of the single element in C () B, as b,, and we can define

a function from A to C by defining f(a) = b,. Under these circumstances
f is called a choice function, for what it does for us is to choose an
element (and only one) from each B,. It is, in fact, the case that there
are many equivalent statements of the axiom of choice, but we are not
interested in pursuing these here. Later we shall need one other form
of the axiom, one that is usually called Zorn’s lemma, but we shall state
this at the appropriate place. The interested student is referred to the
references at the end of this chapter for suitable texts in which he can
pursue this topic further.
The following are now simple consequences of the definition of finite
and infinite sets:
Let F be a finite set, and let Z, be the set of integers {k |0 <k S n},
then either F = @ or there exists a one-to-one correspondence between
Z, and F for some integer n. Conversely, if either / = @ or there
exists a one-to-one correspondence between F and Z, for some integer n,
then F is finite.
From this last remark it follows easily that the union of two finite
sets is again finite, and that if A is infinite and F finite then A — F is
again infinite.
We come now to some results which we take a bit more care to prove .
in detail. Before we begin we assume that the reader is familiar with the
real number system, and in particular is aware that the set of positive
integers is well ordered, i.e., has the property that every nonempty set
of positive integers has a smallest member.
Our first result says roughly that in the family of sets the smallest
sec. 0.6 PRELIMINARIES 13

sets are the countable sets. If a finer classification is desired, we can say
that the finite sets are the smallest, and are followed by the countably
infinite sets. We prove

0.2. Theorem: Any subset of a countable set is again countable.

Proof. Let A be a countable set, and let B C A. If B is finite we are

finished, so we assume that B is infinite. Let f be the one-to-one corre-
spondence between A and Zt, the set of positive integers, where we
indicate the element of A which corresponds to the integer n by f(n) in
the usual functional notation. Let n; be the least integer such that
f(m) € B—such an integer exists because of the well ordering of
the positive integers. Then let m2 be the smallest integer such that
f(m) € B — {f(m)}, and inductively let n, be the smallest integer
aes
such that f(m) € B—U jif(n,)}. We observe that the latter set,

B— U ({f(ni)}, is nonempty for each k, otherwise we would have

constructed a one-to-one correspondence between the set of positive
integers Z, and B which would imply that B is finite. We then define
the one-to-one correspondence between Z* and B by g(k) = f(nz). The
existence of this correspondence shows that B is countable. J

It will be important for us to know that the union of a countable

family of countable sets is again countable. This result is stated in the
following:
£ .

0.3. Theorem. Let A be a countable indexing set, and with each a € A

let there be associated a countable set By, then the set U Ba is again
acA

countable.

Remarks Preliminary to the Proof. Since we know that subsets of

countable sets are countable, we might as well assume that each of the
sets B, as well as the set. A is countably infinite, for if they are not
(i.e., if they are finite in some cases) we can fill out the finite sets with
further elements so that they are countably infinite. Then the union in
which we are interested will be a subset of the union we shall prove
countable and will itself be countable as a subset of a countable set.
Similarly we may as well assume that the sets B, are disjoint (Le.,
14 PRELIMINARIES CHAP. 0

B.(\ Bs = @ for a ¥ 8B), since if they are not we may think of the
=}
elements of all the B, as distinct, form their union, keeping any like
elements distinct, Sate set countable, and then think of the union
inwhich we are interested as a subset of the union in which like elements
have been kept distinct. Again, since countable sets have at worst
countable subsets, we shall be finished.
Proof. By the remarks just made we assume A and each Bb, as
countably infinite, and the B, disjoint. Since A is countably infinite
there exists a one-to-one correspondence f between Z* and A, thus to
each a € A there corresponds a unique integer such that f(n) = a. Let
us rename the sets B, by replacing the index a by the index n, where,
of course, f(n) = a to obtain the family of sets B,,n = 1, 2,.... Now
since each B, is countably infinite there exists for each n a one-to-one
correspondence f, between Z*+ and B,, so that for each b © B, there
exists a unique integer k such that f,(k) = b. We tag the element b with
the two indices n and k, writing it as b,,. In this fashion each element
in the union \ B, has attached to it a unique pair of indices. The
aA

following diagram makes this clear:

By: by by by
Bo: ba bes bes
B3z~ bar b32 bsg

We now set up the following one-to-one correspondence:

j eee 3 4 5 6 7 8 9 10
tg we el) el Re ees
bu bie be bis bee bs1 bus bos bse ba

I LS Ole i So ee
Le ot: “pee ee
bis bag bsg, aes

and so on, in which we count first the elements whose indices sum to 2,
then those whose indices sum to 3, then those whose indices sum to 4,
etc., and for each such sum (i.e., 2, 3, 4,.. .) we arrange the elements so
that the first index occurs in the natural order for the integers. We thus
establish the existence of a one-to-one correspondence between Z+ and
sEC. 0.6 PRELIMINARIES 15

— B,, and this proves that the union of countably many countable
sets is again countable. J

It might appear that there is nothing worse than a countably infinite

set, i.e., that every infinite set is necessarily countably infinite. Such,
however, is not the case; and we give a classic proof, due to Cantor, to
show that the real numbers between zero and one are not countable.
We suppose that in fact the set of real numbers in the interval (0, 1)
is countable and assume a one-to-one correspondence has been set up
between Zt and these numbers. We indicate the correspondence by
the following diagram:
lo - 141012043014...

2 <> .A1A220230r4. . «
30 - 131032033034. . «

4a - 41042043044...

where each a,; represents a digit, i.e., 0 S a;; S$ 9, and where it is

assumed that where we have two alternate choices for the decimal
expression of the real number, as for example in the case where 2, could
be written either as .2000. . . or as .1999. . . , we always choose the one
that ends in a string of zeros. Now this one-to-one correspondence is
such that to every positive integer there corresponds some real number
in (0, 1) and conversely to each real number in (0, 1) there corresponds
some integer. Consequently the infinite list of decimals given above is
complete in the sense that every real number of (0, 1) occurs somewhere
in the list. If, then, we can produce a real number between zero and
one which is not in this list we shall have a contradiction, and this is
precisely what we set out to do. We define b = .bibob;. . . as follows: if
a;,is 5 let b; = 6, if ai; ¥ 5 let b; = 5. Now it is clear that b is not equal
to any one of the decimals in our list for it differs from the nth one at
the nth place. Also it is clear that 3 < b S 3, so that b € (0, 1). This
contradiction then shows that there cannot exist such a one-to-one
correspondence between Z and (0, 1), and since (0, 1) is clearly infinite,
1
containing as it does the set a ln = 2,3,.. +,the set of real numbers

in (0, 1) is uncountable.
16 PRELIMINARIES cHap. 0

EXERCISES

0.23. Prove that if A is infinite, and a € A, then A — {a} is infinite.

0.24. Criticize the following “proof” that the set of positive integers,
Z+, is uncountable: Write each positive integer in the usual way,
but precede the first digit by an infinite string of zeros stretching
off to the left, e.g., 17 is written . . .00017. Assume the set Z* is
countable and set up the obvious one-to-one correspondence
n<...000n, e.g., 124 corresponds with .. .000124. Write the
list of these down in the usual order, thus:

.. .0OO1
. . 0002
» . -0003
ete.

Now construct a new number as follows: Reading down the

diagonal entries in the above list, if the digit at the nth place in
the nth number of the list is 5, let the entry in the new number at
the nth place be 6, if the nth digit in the nth number of the list is
different from 5, let the entry at the nth place of the new number
be 5. The number thus constructed is different from the nth
number in the list at the nth place, thus is different from all of
them, and the one-to-one correspondence above constructed is
not as claimed a one-to-one correspondence, whence the set Zt is
uncountable.

§7 Miscellaneous Assumptions Regarding the Real Numbers

We assume that from his previous work the student is familiar with
the real number system. In particular we assume that he is familiar
with the principle of mathematical induction in either of the following
two forms:

1. If S is a set of positive integers such that

(2) SIRE ys.
(b) k € Simpliesk +1 € S,
then S = Zt, the set of all positive integers.
2. If S is a set of positive integers such that
sEc. 0.7 PRELIMIN ARIS ily

(a) LES,
(b) for each k < n, k € S implies n € S,
then S = Z*, the set of all positive integers.

We also assume that the set of positive integers iswell ordered, i.e.,
each nonempty subset of Z+ contains a smallest number—smallest in the
sense of the usual ordering of the real numbers.
If T is a set of real numbers we say that ¢ is an upper bound for T
if s St for each s € T, and wu is a lower bound for T if uw S s for each
s € T. We say that ais a least upper bound (supremum) for 7 if a is an
upper bound, and if 6 is any upper bound for 7, then a S b; and we
write sup 7’ = a. Similarly c is a greatest lower bound (infimum) for 7
if c is a lower bound, and if d is any lower bound, then d S c; and we
write inf 7 = c. We assume that any set of real numbers that has an
upper bound has a supremum, and any set of real numbers that has a
lower bound has an infimum.
We use the usual designation for intervals of real numbers, as
follows:
(a,b) = {z.|.a real, a <-a2 < 5}
[a, b) = {x |xreal,as
(Ul Ae lee realegs< 7s) by
le Dir. real na cs

where a and 6 are real numbers. Also we feel free to use the notation
(a, ©), (©, b), [a, ©), and (, b] with the obvious meaning, e.g.,

(G@yo:) =42 |x real, a. < x}.

REFERENCES

Although there is a general list of references at the end of the book,

there is listed here a set of books dealing principally with set theory and
related matters.
LOGIC
1. Christian, R. R., Introduction to Logic and Sets (Preliminary ed.;
Boston: Ginn, 1958). |
2. Rosser, J. B., Logic for Mathematicians (New York: McGraw, 1953).

oo Kleene, S. C., Introduction to Metamathematics (Princeton, N.J.:

Van Nostrand, 1952).
18 PRELIMINARIES CHAP. 0

SETS

. Bourbaki, N., Théorie des Ensembles (Paris: Actualités Scientifiques

et Industrielles, Herman et Cie., 846 = 1141, 1951).

. Halmos, P. R., Naive Set Theory (Princeton, N.J.: Van Nostrand,

1960).
. Kamke, E., Theory of Sets (New York: Dover, 1950).

. Suppes, P., Axzomatic Set Theory (Princeton, N.J.: Van Nostrand,

1960).

REAL NUMBER SYSTEM

. Landau, E., Foundations of Analysis (New York: Chelsea, 1951).

HISTORICAL

. Cantor, G., Contributions to the Founding of the Theory of Transfinite

Numbers (New York: Dover, n.d.).
 CHAPTER 1

TOPOLOGICAL SPACES—BASIC DEFINITIONS

AND THEOREMS

$1 Neighborhood Systems and Topologies

As we look back over the development of mathematics through the

ages, we seem to discern a dichotomy. On the one hand there is algebra,
which has studied the arithmetic structure of various mathematical
systems, as for example in algebra proper, both classical and modern,
number theory, and the like. On the other hand there is geometry,
which has studied the geometric or spatial characteristics of mathe-
matical systems, as for example in geometry proper, both Euclidean and
non-Euclidean, differential geometry, and topology. This dichotomy is
more apparent than real, for in fact, all mathematics is related, and the
two fields mentioned above are inextricably intertwined, so that as one
advances in the study of either field, notions from the other become
essential. In fact, it is probably the case that the greatest synthesis in
mathematics takes place in the field of analysis, in which the notions of
both algebra and topology are simultaneously brought to bear on the
study of mathematical systems.
Our purpose here is to study what is ordinarily called point set
topology. Point set topology forms the background for a number of
more advanced topics. Of these advanced topics we single out three,
and we have in mind laying a foundation that will enable the student
to go on in any of these three. The three we have in mind are analysis,
which we have already mentioned, advanced point set topology, which
will extend the ideas we study here, and algebraic topology, which utilizes
some of the techniques of algebra to further the study of topological
spaces.
Since our concern in point set topology will be with the spatial
characteristics of a set, we shall be thinking largely in terms of geometric
notions, and we shall thus call the elements of the set we are studying,
19
20 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

points. We introduce axioms into this set of points via definitions. Now
one of our principal concerns in geometry (or topology) is the notion of
“closeness.’’ We have, in some way, to say which points lie “close” to
other points. We can introduce a notion of closeness in a variety of
different ways, but we seek one that is quite general, so that by speciali-
zation we can obtain many of the systems we already know well. The
method we select defines first the notion of a neighborhood system of a
point x of the set we are studying. In a certain sense, a neighborhood
of a point x is a g@ipOfpointsewhichsliesiclosesstosbhespem. Let us get
under way and define a neighborhood system, via the following:

1.1. Definition. Let X be a set, and for each point x € X, let Uz =

{U(a)} be a nonempty family of subsets of X associated with x, such that

(1) x € U(x) for each U(x) € Uz.

(2) If Vv> UG) jor some UG), then V € Us
(By 0 and Ve, tiene) Ve
(4) If U € Uz, then there exists V € Uz such that if y € V then
Ue Uy,

Then U; is called a system of neighborhoods at x.

It is useful to observe at this point that the set V defined in (4) of

the above definition is such that V C U. This is easily seen, for let
y € V, then U € W,, hence by (1) of 1.1 y € U, whence V C U.
We now have some idea how we want points “close” to a given point
(i.e., neighborhoods) to behave. Of course, our notion of neighborhood
is very general and includes all sorts of strange cases; in particular we
shall find (Ex. 1.6 below) that the whole set X is a neighborhood of each
of its points and in fact need be the only neighborhood of each point.
These pathological cases need not disturb us, for we shall make our
future definitions in such a way that a given property holds for every
neighborhood of a point, and thus most of the pathology disappears.
Now we stick all the neighborhood systems of a set together to form a
topological space. We do this via the following:

1.2. Definition. Let X be a set, and let 3 = {U,| x € X} be an assign-

ment of neighborhood systems for each point x € X, then the pair (X, 3) is
SECs TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS Di

called a topological space. 3 7s called the topology for the space (X, 3),
and as mentioned before the elements of X are called points. Further, if
3 = {u,} and = {Uz} are two topologies for a set X, then 3 = 5! if and
only if Us = Uz for each x € X.

Actually the last sentence in the above definition is unnecessary, for

since a topology is sumply a collection of families of subsets of a set X,
two topologies will be the same if and only if they are equal as sets,
and this will be so if and only if WU, = U; for each x € X. It is well,
however, to make this quite explicit, for it is quite possible that two
topologies may be alike in that they specify the same subsets of X as
neighborhoods of points of X, yet they will not be equal as topologies,
since the assignment of individual neighborhoods to points of the space
may be different. The following example points up this distinction: let
X = {x, y, z} and let 3 be determined by

Ue = or, {Z; y} {z, z}, {z, Y, 2h};

U, = {{z, y, hf
and
Uz 1185 Y, Bp,

while 3’ is determined by
eee all 1 A YX
Uz, = aah, 1, Ys 12, 2, 12, Y, ss
at Sf \
Uy a 1 y}; A, Y; z}}
and
pss - ita
UU; = Ue, Y, eye

Now 5 and 3’ are alike in that they specify exactly the same sets of X
as neighborhoods, but they are not the same topology, since in 3, {z, y}
is a neighborhood of x but not of y, while in 5’ {x, y} is a neighborhood
of both x and y.
It is frequently convenient to say simply that X is a topological
space, rather than having to specify the topology, 3, and having to
write (X, 3), since, more often than not, we are interested not so much
in a particular topology, but rather in properties that any topology will
possess. We shall thus feel free in what follows to omit any specific
mention of the topology unless it is important to the context to empha-
size a particular topology, or to distinguish between different topologies.
22 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

EXAMPLES (AND EXERCISES)

Every Example is an Exercise, and the statements made in the Examples

should be verified as Exercises.

1.1. Let R be the real line, define (a, b) = {«|a <a < b} and define

U, = {U|xz € (a,b) CU for some

a, b € Ria < DY},

then U, is a neighborhood system at x and generates a topology

for R according to 1.2. The topology so generated is called the
usual topology for R.

NOTE: The function p of the following Ex. 1.2 and the function

(fa) = fp lf — gl de
1

implicitly defined in Ex. 1.4 are essentially metrics, which we shall

discuss at length in Chapter 5. For the time being, however, the student
should note and feel free to use these properties of metrics:

(a) p(x, x) = 0 for every point x in the space.

(b) p(x, y) = ply, x) for every x, y in the space.
(c) p(x, y) S p(x, z) + plz, y) for every x, y, 2 in the space.

1.2. Let H = R X R, where BR is the real line, i.e., H is the real plane.
Let x, y € E, where x = (a, b), y = (c, d) and define

alt, y) = [a= 6) 4 (Ody,

p(x, y) is called the distance from z to y. Define

S.@) = {yl e(a,y) < 4,

called an open e-sphere about z, and finally define

Uz = {U| UD S.(z) for some ¢ > 0}.

Then U, is a neighborhood system at z, and generates a topology

according to 1.2. The topology so generated is called the usual
topology for the plane.
sec. 1.2 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 23

123; Let X be a partially ordered set under some ordering, S, i.e.,

(a) «Sy and y Sz imply x Sz, (6) «Sy and y S x imply
x = y, and (y) x Sz for all z € X. Define S,(x) = {y|z S y},
and define u, = {U| U DS,(x)}, then Uz is a neighborhood
system at x. The topology, 3, defined (via 1.2) for X is called the
right order topology for X. A left order topology may also be
defined by starting instead with the sets S:(z) = {y|y S 2}.
1.4. Let X be the set of all real-valued integrable functions whose
common domain is the interval [0, 1]. For f € X define

Sf) = lg X, [lf -glax <4,

1

and define
W = {U |.U 2 S.(f) for some.e > 0},

then Uy is a neighborhood system at f, and such neighborhood

aD
BAL
systems generate a topology by 1.2.

1.5. Let X be a set, and let u. = {U |x € U} for each x € X, then U,

is a neighborhood system at x, and the topology thus generated
/ is called the discrete topology for X.

1.6, Let X be a set, and let U, = {X} for each x € X, then U, is a

neighborhood system for x, and the topology thus generated is
called the trivial topology for X.

Lis Let X be a set, 3 and 3’ two topologies for X, then 3 = 0’ if

and only if for each « € X and for each U € U, € 3 there is
a U'’e€uz€ VW such that U’ CU, and conversely for each
V' € uz € 0 there isa V € Uz € Jsuch that V CV’.

§2 Topological
Open Sets in a Space
The next definition and two theorems introduce the class of so-called
open sets of a topological space. These distinguished subsets of a space
play an important role in the topology of the space.

ginBe Definition. A set O C X, where X 1s a topological space, is said

to be open provided O € Uz for each x € O. The family of open sets of a
space X is designated by 0.
24 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS cHaAP, |

detu@heorenm Let X be a topological space, then U is a neighborhood

of x € X (te, U € Ux) af and only if there exists O € © (4.e., O 1s an
open set of X) such thatx € OCU.

Proof. (1) Let U be a neighborhood of x, and let

O = {y| there exists W € U,, W C U}.

We observe first that since U € U, and U C U, we have x € O. Now

let y € O, then there exists W € U, such that W C U, by definition
of O. Also, by 1.1(4), there exists a V € U, such that z € V implies
W € %U,, and as already remarked V C W. We have thus that for each
z€V, z€ WCU and WE, whence z€ O, and consequently
V CO. We have, however, V € U, and V CO, hence by 1.1(2),
O € U,, and thus by 1.3, O is open.
(2) Letz € X,0€ O,andzec OCU. By1.8,0€ U,, and by1.1(@2),
OS ere |

1.5. Theorem. Let X be a topological space, then

(1) The union of any number of open sets is open.

(2) The intersection of any two (consequently of any finite number) of
open sets 1s open.
(3) X is open.
(4) &@ ts open.

Proof. (1) Let A be an indexing set, and let O. be open for each
a € A. Further letO = U O,, and let x € O, then x € O, for some a
aA

Since O, is open O, € Uz, and since O. C O, O € Uz by 1.1(2), whence

O is open by 1.8.
(2) Let O, and O, be open, O = O, (\ Oo, and let z € O. Now x € 0,
and x € Oz, and since O, and QO, are open O; € Uz and Oz € Uz, con-
sequently by 1.1(3), O = O, (\ O2 € Uz, whence by 1.38, O is open.
(3) Let x € X, then U, ¥ @; there exists U € U,. Since UC X,
X € U, by 1.1(2), and X is open by 1.3.
(4) If @ were not open, there would exist « € @ such that @ ¢ U..
Since, however, there is no such x (@ being empty), this is clearly false,
hence @ is open. J

In introducing a topology into a set X, we started with the notion

of neighborhood. This is only one of many avenues to the same goal;
smc. 1.2 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 25

in particular we might just as well have started with the notion of open
set, and introduced the topology in a somewhat different way. That we
reach the same goal is shown by the following:

1.6. Theorem. Let X be a set, Gee let there be specified a family, 0, of

subsets of X, such that Br pdrle bp ty Age

(1) The union of any number of sets of © is again in ©. ny

(2) The intersection of any two (consequently of any finite number) of
sets of O is again in O.
(3) X € ©.
(Bin Oxe~0.

Then there is one and only one way in which to specify a topology, 3
in X such that the sets ofOmMillabeMheMOpenmsersm,s the topology 5. We say
that the family 0 Wemeratesesen ;
Proof. Define for each x € X,

= {U|z2¢ OC U for some

0 € 0}.

We verify that U, is a neighborhood system for z.

(1) Uz ¥ @ since by (3) X € Uz. For each U € Uz, cE OCU,

hence x € U.
(2) Let U € Uz, and U CY, then there exists O € 0, such that
po OS UC YVvand Ve.
(3) If U, V € U,, then there exist O1, O2 € 0, such that x € O, CU,
eee a Bye (2) Ou)VO> 60, tas2 C,0iwi \0s CU TY V,
and U(\ V € Uz.
(4) Let U € Uz, then there is an O € 0, such that x€ OCU.
We remark that O € u,. Let V = O, and let y € V, then
weV =0 CU; and Ure uU,.

Thus Definition 1.1 is satisfied, and ‘quistammeighborhood=ssystemmm

fore
We now verify that 0 is the set of open sets generated by the topology
5 = {U,|x © X}. Let O € 0, and let x € O, then O € U, by definition
of Uz, and consequently O € U, for each x € O, whence O is open in the
topology 3. Conversely, let U be open in the topology 3, then U € U,
for each x € U. By definition of U,, there exists O, € O such that
26 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

x € O, C Uforeachaz € U. Clearly U= U O,, whence by hypothesis

xz€U

(Ly, Leo:
Finally we must show that the topology 3 is uniquely determined.
Since, however, by 1.4 U can be a neighborhood of « if and only if there
exists an open set O, such that x € O C U, it is clear that U, can be
defined in no other way than the way chosen. J

Notice that © generates 3 in a quite explicit fashion; namely, if we

are given the family 0, we define for each x € X

={U|UCX,7€
OCU for some O'C OF:

Any family © which satisfies the four conditions of 1.6 thus wniqueky*
determines a topology 3. We could if we wished have defined the
topology of a space to be simply thestamaby=Oeandethisss=vhat=seusually=
one» The student is alerted to this fact, so that when he reads other
works on topology he will not be confused by alternate but equivalent
definitions of a topology for a space. We have chosen the approach to a
topology via neighborhood systems in order to retain as much of the
geometric flavor as possible.

EXAMPLES (AND EXERCISES)

1.8.) On the real line specify © to be the collection of all sets which are
arbitrary unions of open intervals (a, b), a < b, then O generates
the usual topology for the reals.

1.9. In the real plane specify © to be the collection of all sets which
are arbitrary unions of open spheres, then © generates the usual
topology for the plane.

1.10. In a partially ordered set X, specify 0 to be the family of all sets

which are arbitrary unions of the sets S,(x), « € X (Cf. Ex. 1.3),
then 0 generates the right order topology for X.

1.11. In the set of all integrable real-valued functions with common

domain [0, 1], specify © to be the collection of all sets which are
arbitrary unions of sets of the form S.(f), « > 0, f an integrable
real-valued function on [0,1], then © generates the topology of
Ex. 1.4.
 suc. 1.3 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS PATE

> 1.12. In any nonempty set X, specify © to be the family of all subsets
of X, then © generates the discrete topology for X.

' 1.15. In any nonempty set X, specify 0 to be the family {Y, @}, then
‘© generates the trivial topology for X.

$3 Limit Points and the Derived Set

Since the study of analysis depends to some extent upon the structure
of the space in which the functions being studied have their arguments
and their values, and in this way depends upon some topological
considerations, it is not surprising that topology should concern itself
with some of the notions of analysis, in particular with some of the
limiting processes. When we study limits of real functions, or limits of
sequences, the limit of the function (or sequence) lies in a certain sense
arbitrarily close to the values of the function (or sequence). We gen-
eralize this notion here by means of the notion of limit point, in the
following:

“tefeeDefinition? Let X be a topological space, and let A € X, the point

x € X is said to be a limit point of A, provided that i cach Cg
U (\ A contains a point y A x.

EXAMPLES (AND EXERCISES)

£1.14. On the real line with the usual topology, a as well as b is a limit
point of the open interval (a, b).
a4
1.15. In the real plane with the usual topology, any point of the form
(0, y) is a limit point of the set D = {(x, y) |x > O}.
oe
1.16. In the right order topology for a partially ordered set X, if x € X -
is not the least element of X (i.e., if it is false that x S y for all
y € X), then any y < x isa limit point of the set S,(z).

1.17. In the set of all integrable functions on [0, 1] with the topology
defined as in Ex. 1.4 let

Y= {f| [, fax = 0} .,
28 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

and let g(z) = 0 for # 1, and g(1) = 1. Then g is a limit point

of Y. In fact any function which has values different from zero for
only finitely many values of x € [0, 1] is a limit point of Y.

1.18. Let X be a nonempty set with the discrete topology, let A CX

y and let x € X, then z is not a limit point of A.

‘1.19. Let X be a set with at least two elements, and let X have the
trivial topology, let A C X and let x € X, then z is a limit point
of A unless A = @ or {zt}.

hkGeeDefinition. Let X be a topological space and let A C X. The

derivedusewmofmAy@ritien A’, is the set of all x € X such that x is a limit
point of A.

EXAMPLES (AND EXERCISES)

1.20. On the real line with the usual topology, let A = (a, 6), then
A® = [a,b). Leb b= ie) 0 <a = lore = 2), then 5) —0-ee

1.21. In the real plane with the usual topology, let D = {(a, y) |x > O},
then D’ = {(a, y)|2 = 0}. Let # = {(z, y)| a and y integers},
then E’ = @.

1.22. Let X be a partially ordered set with the right order topology, and
let A = {x}, then A’ = {y | < z}.

1.23. In any topological space @’ = &.

1.24. In any topological space, if A C B, then A’ C B’.

§4 The Closure of a Set

The situation is now this: we begin with an arbitrary set A, in a

topological space X, and derive from this set a new set, A’, the set of
all the limit points of A. It seems perhaps natural at this point to
construct a further set, by throwing together into one set all the points
both of A and of A’, and this is precisely what we do in the following:

nOuDefinitionr Lei X be a topological space, and let A CX. The

closunemofany written A, is the set A\U A’,
 sec. 1.4 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 29

EXAMPLES (AND EXERCISES)

1.25. On the real line with the usual topology, let A = (a, b), then
A = fa, 6]. LetB = {2 |0 <2 S 1 ore = 2}, then

Bese (oi aes a =. | Orap == 2,

1.26. In the real plane with the usual topology, let D = {(z, y) |x > O},
then D = {(z, y)|z = 0}. Let H = {(x, y)|az and y integers},
then =’ 2B.
1.27. Let X be a simply ordered set with the right order topology, and
let: Av= S(x)- then A= XxX"

Y 1.28. Let X be a nonempty set,with the discrete topology, and let

A CX, then A = A. OU

1.29. Let X bea nonempty set with the trivial topology, and let A C X,
then A = X unless A = @.
{)

iw
1.30. In any topological space 7 = @. aes
y

1.31. In any topological space, X, X = X.

1.32. In any topological space, X, let A C X, then A C A. dir

1.33. In any topological space, X, x € A if and only if for each U € UW,

UDNA A S.

The closure of a set A in a topological space X has, of course, a

number of interesting properties. Among others is the property that it
is maximal in the sense that no larger set is obtained from A by adding
further limit points of A to it. This is shown by the following:

F 1.10. Theorem. Let X be a topological space, and let A € X, then

A = A2. 2 wh ee
633 d
Proof. By 1.9A = A UA’. Weshow thatA’ C A. Let € A’, and ut
suppose a ¢ A = A U A’, then x ¢ A and further, since also x ¢ A’, !
there exists some U € U, such that U (\ A = ©. Select O, open, such
that z € O C U, thenO € U,, and further sinceeO 1A CUNMA = @,
we have O()\ A = @. Now since x € A’, OM A contains some point
ia
: Oe jy
) 4
ae
Wr Dfae: aa
V7
ae0
aaa
pe Wa
 30 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

y =x. Thus y € A, and since OM A = @, y € A’. Since O is open,

O€ U, by 1.3, thus there exists z ~ y such that 2 € O(\ A. This,
however, contradicts O(\ A = @, consequently x € A. This completes
the proof that A’ C A. Finally A= AU A’ = 4, since A’ CA.

There are also a number of relations between unions and intersections

of sets and their closures, as well as certain special properties of a similar
nature which depend not only upon the properties of intersections,
unions, and closure, but also upon whether the sets under consideration
are open. Several of the more important and useful results of this kind
are found in the following two theorems.

1.11. Theorem. Let X be a topological space and let AC X and

BCX, then |

(1) If ACB, thn ACB.

Wp eae SG Aue:
CG) AWB=A 2B.
Proof. (1) By Ex. 1.24 if A CB, then A’ C B’, consequently
A =A WAGER B= B:
(2) Let x € ACVB, and let UE u,, then UM (ANB) # @,
consequently neither U(\ A nor U(\ B is empty, and xz € A and
z € B, whencex € ACVB. ThusdA
MV BCANB.
(3) Since A C AU Band BCA UB, we have by part (1) of this
theorem that d C AU Band BC AUB, whence AUBCAUB.
Now let x € A UB, and suppose « ¢ A and x ¢ B, then there exist U,
Ve, such tht UNA= OG andVOAB=GO. Now UNV Eu,
and

UNV
OVE UB) = (UAV A OG Vin
S (UPN A) OY Bi,
but this contradicts « € A U B. Consequently either z € A or z € B,
whence « € AUB. Thus AUBCAUB, and finally AUB =
AUB. J
4

* V 142. Theorem. Let X be a topological space, and let A € X be open

Wand let BX, thn ANB CAB.
sec. 1.5 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS ol

Proof. Let

bce (\. Bem A) (BAO B) = (A CYB) (A OB").

If x € A) B, then

cee (AY
B) MIA B) = Ai Ve.
(earn re? ofa
ipee 427 V5", then since Ae U,, if Ue Uy AU € ate by 1.178), ey,
Now since z € B’, (A(\ U) (OB contains a point y ¥ x. Thus for
each UE WU, UM(AN this (A (\ U) C\ B contains a point y ¥ z,
whence pl Gyr ring

Pi ANG) “CLAY. Wb:

In either case x € A(\ B, whence AOV.BCAMYB. §

EXERCISE

1.34. Give an example on the real line in which A is an open set, and
the sets A(\ B, ACB, and A(VB are all distinct. Give an ,
-
example in which A is not open and A (\B Z AY B. L4

§5 Closed Sets

It is perhaps reasonable to ask at this point if there are sets which

already contain all their limit points, and if such sets have any inter-
esting and distinctive properties. Of course, we already know that there
are sets which contain all their limit points, since by 1.10, A is such a
set. We now set about investigating some of the properties of such sets,
but in order to have a name for such sets we first introduce the following:

1.13. Definition. Let X be a topological space, let AC X, then A is_

said to be closed provided A = A. n)> QUA!

As innocuous as this definition looks, it is nonetheless the source of

some difficulty. The choice of the word ‘‘closed” for the type of sets
described in the definition is perhaps a poor one, for one intuitively has
the notion that open and closed are antithetical notions, in that, for
 82 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

example, doors are either open or closed. Unfortunately this is not the
case with sets in a topological space. It might be better to think of
open and closed sets as being akin to the right to own a car. This is a
right that is open to anyone, yet there are certain colleges that prohibit
students from owning cars, so that to these students this right is closed,
so that this right is at one and the same time open and closed. It is
important to remember that this is the case with sets, namely that open
and closed are not antithetical, that a set may be both open and closed
at the same time, that it may be open but not closed, that it may be
closed but not open, and that it may be neither open nor closed. Each
of these possibilities is in fact realized on the real line with the usual
topology, for the whole space, R, is both open and closed (as is also the
empty set), the interval (a, b) a < bis open but not closed, the interval
[a, b] is closed but not open, and the interval [a, b) is neither open nor
closed—these remarks should, of course, be verified by the reader.
We might say a few more words about closed sets before we go on
to investigate their properties. First, pursuing the ideas in the preceding
paragraph, if we are in a position to want to prove a set closed, it will
do us no good whatever to prove it is not open, since we have already
seen by example that these two properties can be possessed by the same
set at the same time. It is, of course, equally true that it does no good
to prove a set is not closed if our object is to prove that it is open.
It is also useful to observe that in any space the sets @ and X (the
whole space) are invariably closed, and for that matter also invariably
open. It is interesting to speculate at this point on what character a
space must have if these two are the only sets which are at one and the
same time open and closed—interesting, yes, but perhaps premature.
(Cis
It is, of course, quite pleasant that the closure of a set is in fact
closed, as is shown by 1.10; and it is further rather nice that even though
open and closed are not antithetical notions, nonetheless, open and
closed sets are related as is shown in the following:

YW 1.14. Theorem. A set A, in a topological space_X, is closed if and only

uf its complement, A‘, is open. 4 =Ave' Kthvad ht’
Proof. Let A be closed, and let x € A*. Since xz ¢ A = A, there
exists a neighborhood U € U, such that U ( A = @, as is shown in
Ix. 1.33. Consequently, U € A’, whence by 1.1 (2), A* € W,, and by
1.3, Ais open,
 >) per PO
SECeedeo TOPOLOGICAL SPACES—BASIC DEFINITIONS AND his J 33 Da
Oke Ae
Conversely, let A* be open, and let x € A.ianthat(a € A° oA
then by 1. 3; A* € U,, and by ieee a AZ 2 his. is-clearly a

A simple application of this theorem, together with DeMorgan’s

rule, establishes the following:

1.15. Corollary. In any topological space,

(1) The intersection ofany number ofclosed sets isclosed, and

(2) The union of any two (hence any finite mumber) of closed sets 1s
closed.

Proof. (1) Let A be an indexing set, and for each a € A let Cy be

closed. Further let C = ( Cz, then
aA

aie COV Cae,

aca aGA

and since by 1.14 each C% is open, so also is U C% by 1.5. Thus C is

aA

open, whence again by 1.14, C is closed.

The proof of (2) is left as a trivial exercise. J

The following two definitions are needed in the exercises which

follow. Some ambient topological space is assumed in each definition.

1.16. Definition. (1) The interior of A, designated bn A or A°) 4 Ms

. uA’ AK vee Ww Be
_the set of all x, such that A_€ Uz.
(2) The frontier of A, designated by Fr (A) is the set eae In

1.17. Definition. The set A is said to be dense in the set B if A > B.

If A is dense in the whole space, X, we say that_A 1s everywhere dense, or
sometimes, if there is no chance for misunderstanding, simply dense.

EXERCISES aa
1.35. Show that A is an open set, and is in fact the largest open set dur
4
contained in A, in the sense that if B is open, B C A, then B C A,
 34 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS cap. 1
Z

ye 1.36. Show that A is the smallest closed set which contains A, in the
Kr i . . .

f\
} sense that if B is closed and B D A, then B D A.

The preceding two exercises show that one might have defined the
interior of A as the union of all open sets contained in A, and the closure
_of A as the intersection of all closed sets which contain A.

1.37. Give an example of a set A on the real line with the usual topology,
such that from A 14 distinct sets may be formed, using only the
two operations of complementation and closure. More generally,
show that in any space the maximum number of distinct sets
which may be formed from a given set using only complementation
and closure is 14.

1.38. Show that

(a) Fr (A) = Fr (A),

(b) Fr (A) € Fr (A) and Fr (A) € Fr (A),
(c) Fr (A U B) C Fr (A) VU Fr (B),
and in (b) and (c) give examples to show that proper containment
can in fact occur.

1.39, If A and B are open show that

[A \ Fr (B)] U [BO Fr (A)] GC Fr (ANB)

C [AN Fr (B)] U [BO Fr (A)] U [Fr (A) A Fr (B)].
1.40. If A is closed, show that B ™ Fr (A) € Fr (AB),

1.41. Show that

Fr (A) = {|x ¢Aanda

¢ Ao
= {x |x is in neither the interior of A
nor the interior of A*}.

1.42. Prove 1.15(2).

1.43. Show that if X has the discrete topology, then {x} € U,. Further,
if X has the discrete topology, A C X, what form do the sets A ’

A, and Fr (A) have?

smc. 1.5 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 35

1.44. Let X be a set and let

goes) e| oe eX} and iN) mer lea GueX }

be two topologies for X. Further let © and 0’ be the families of

open sets determined by 3 and 3’ respectively.
(a) Show that 5 = 3’ if and only if for each 2 € X and each
U € u, there exists O’ € ©’ such that x € O’ C U, and for
each U’ € Uz there exists O € ©, such that x € O C U’.
(b) Show that 3 = W if and only if © = 0’.

1.45. Let X be a set, and let there be assigned to each set A C X

another set A C X, such that
7 For any 4,-5 Cx, AVE = AUB.
Il. If A = © or A is a set consisting of a single point, then
A= A.
III. A = J for any set A CX.
Prove the following:
GN) UT nse eee
eA NB CAINS.
(3) d-BCA—B.
(4) (\ A, © CX A, where C is some indexing set.
y¥EC hele
(Gs) rd ids.
Vee vyEC
(6) If A is finite, A = A.
(7) A CA for every AC X.
(Sox =X,
(9) Define A C X to be closed provided that A = A, then X isa
topological space in which the family of open sets,

© = {0O|O = A*, where A is closed, ie., A = A}

and that with the topology so defined (using 1.6) the closure
of a set A C X is in fact A.

1.46. Show by suitable example that the derived set, A’, is not neces-
sarily closed. [Hint. Let

ee fon NN faite 7 Se SEM

xX = ee Y, a), ee {@, Ue, US) VES) Xs A= Canal
36 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS cHAP. 1

§ 6 Subspaces

1.18. Definition. Let X be a topological space, and let YC X. For

each y € Y define a neighborhood system Vy in Y by Vy = Y (\ Uy, 2.€.,
V € VU, if and only if V= YC\U for some U € U,, where U, ts the
neighborhood system for y in X. The topology J! = {U,|y € Y} ts called
the topology induced on Y by the topology 3 of X, or more commonly the
relative topology of Y in X, and (Y, 3’) is called a subspace of X.

It is important to note in this definition that a subset Y of a space X

is not necessarily a subspace. Only if the topology of Y agrees with the
induced (or relative) topology is Y called a subspace. The above
definition is incomplete in that it is only asserted (without proof) that
U, is a neighborhood system; by now the reader is inured to the fact
that he is expected to prove that, in fact, U, satisfies the conditions
OLdc le

EXAMPLES (AND EXERCISES)

1.47. Let R be the real line, and # the real plane with the usual topol-
ogies, then F is a subspace of F, 1.e., the usual topology in #
induces the usual topology in R = {(a, y) | (a, y) € EL, y = 0}.
1.48. Let X and Y be nonempty sets, then the discrete topology in X
induces the discrete topology in Y, and the trivial topology in X
induces the trivial topology in Y.

1.49. Let N = {(2, y) |2, y real, y = 0}, i.e., N is the closed upper half
of the real plane, and let

N° = {(x, y) |2, y real, y > O}.

For (a, y) € N, define Vey = Uey ON if y > 0, where Ue»
is the neighborhood system for (x, y) in the usual topology for the
real plane, and define

Ven ={IVIVDIUAN*) U{G@, }}

for U € Wren ayy if — 0. Let

3 = {Vew | ayy) GN},

sec. 1.6 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS Sih

with Ui) as so defined, then (N, 3) is a topological space. N is

not a subspace of the real plane with the usual topology, nor is
the real line with its usual topology.a subspace of N. The real
line with the discrete topology is a subspace of NV, where we think
of the real line here as the set R = {(x, y) |y = O}.

If X is a topological space, and Y a subspace of X, there is a con-

siderable distinction between OC Y being open in Y (i.e., in the
relative topology) and being open in X. For example, if X is the real
line with the usual topology, and Y = [0, 1], the closed unit interval,
then the set (3, 1] is an open subset of Y, since it is a neighborhood of
each of its points, yet (2, 1] is not open in X, since it is not_a neighbor-
hood of the point 1. We must, consequently, use some discretion when
we talk about open sets in the context of spaces and subspaces, and must
specify whether a given set is open in the whole space XY, or in the
subspace Y. A similar degree of discretion must of course be exercised
in regard to closed sets, as well as other topological features, such as
closure, limit point, and the like.
It would be gratifying if we had some criterion by which we could
recognize the open sets of a subspace, and by a happy circumstance
just such a criterion is given by the following:

1.19. Theorem. Let X be a topological space, Y a subspace of X, then

O CY ts open in Y if and only afthere exists a set O' open in X_ such
0= ONY.
that
Proof. Let 3 = {U,|a € X} be the topology for X, and let

Ij
\o v= {0.|2€ Y} = (WAY (ee ¥}
\

be the relative topology for Y. Suppose O = O’ ()\ Y, where O’ is open

in X, and let x € O, then x € O’, and O’ € U, by 1.3. Thus

O SO OEY 6 UO Y,
and O is open in Y.
Conversely, suppose O is openin Y, thenO € U, Y foreach x € O,
that is O = U,(\ Y for some U, € U,z for each x € O. Now for each
x € O, and each U,, there exists O,, open in X, such that x € O, € U,,
by 1.4. Define
 38 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS cHAP. 1

OV AOS
x€O

then O’ is open in X by 1.5. We now show that O = O’() Y. Let

y € OCY, then y € O,, whence y € O', and y € O’ \ Y. On the other
hand, let y € O’ (\ Y, then y € O, for some x € O, and since O, € Uz,
y € UzC\ Y = O, whence y € O. ThusO=O'OY. J

EXERCISES

1.50. Prove that U,, as defined in 1.18, is a neighborhood system for

yey.

1.51. Let X be a topological space, and let Y be an open subset and a

subspace of X. Show that O is open in Y if and only if O is open
in X.

1.52. Show that if X is a topological space, Y a subspace of X, and

if O is open in Y if and only if O is open in X, then Y is an open
subset of X.

1.53. Show that if X is a topological space, Y a subspace of X, then

C C Y¥ is closed in Y if and only if C = C’ (\ Y for some closed
subset C’ of X.

"87 Lifaits ofSequetiterWiUsdoneui meae

1.20. Definition. Let X be a topological space, let x € X, and let
VU" {z,|n = 1,2,...} be a sequence of points in X. The sequence {x,} is
said to converge to x, and x is said to be a limit
sequence
of the {x,} if
and only uf for each U € Uz, there exists an integer N such that n => N
implies x, € U.

EXAMPLES (AND EXERCISES)

1.54. Let R be the real line with its usual topology, and let {z,} be a
sequence in FR, with x a limit of the sequence {z,}. Show that
lim z, = x in the usual sense of analysis, i.e., for each ¢ > 0,
there is an integer N such that n = N implies |x, — a| < «.
1.55. Let X be a nonempty set with the discrete topology, let {x,} be
sucn 17 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 39

a sequence in X, and let x be a limit of the sequence {z,}, then

there is an integer N such that for n = N, x, = a.
1.56. Let EF be the real plane, and define for « > 0,

S.(z, y) = {(u, ») | (u, ») € #, |z — ul < ¢

and
Ue. = AU | U DS; y) for some < > 0}.

Verify that 3 = {Uc | (2, y) € EZ} is a topology for EL. Let

{(@a,» Yn)} be a sequence in E, with the topology, 3, and let
(x0, Yo) be a limit of the sequence {(xp, yn)}, then (2, z) for any z
is also a limit of the sequence {(&n, yn)}, and observe thus that
limits of sequences need not be unique.

1.57. An alternate and somewhat more general approach to the notion of

the limit of a sequence. Let {A,|n = 1, 2,...} be a sequence of
sets in some topological space X. Define

lim sup A, = {y| UC. A, # @ for each U € 4, and for

: infinitely many indices n} ;

lim sup A, is called the limit superior of the sequence

ENE
teica eae a
Also define

lim inf A, = {y| UC\\ A, ¥ © for each U € u, and for all

but a finite number of the indices 7} ;

lim inf A, is called the limit inferior of the sequence

AAs dee lt yee fre

If lim inf A, = lim sup A,, we define this common set to be

n n

lim A,. Prove the following:

n

(a) lim inf A, C lim sup Ap.

(b) If A, = {x,} for each n, and if lim A, = {x} in the sense

 40 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

of the above definition, then lim x, = x in the sense of Defi-

nition 1.20.
(c) Let A, = {Gn yn)}, n = 1,2,... be the sequence of Ex. 1.56,
then lim A, = {(2o, 2) |2 € R} in the sense of the above defi-

nition.

For reasons that may not be apparent at this juncture, we now

introduce a new kind of space, with a rather stronger structure. The
motivation for the introduction of these spaces will appear after the
definition and a few examples.

re; 1.21. Definition. Let (X, 3) be a topological space, then 3 is said to be

& Hausdorff (or Ts) topology for X, provided that for each pair, x, y, with
1U2~4 « X y, of points of X, there exist U € Us, V € Uy such that U (\ V = &.
-¢ If this condition is satisfied we call X a Hausdorff (or T>) space.
bern

EXAMPLES (AND EXERCISES)

— 08. (a) The real line with the usual topology is a Hausdorff space.
(b) The real plane with the usual topology is a Hausdorff space.

—" Hos) The real plane with the topology defined in Ex. 1.56 is not a
Hausdorff space.

1.60. Let
X = {f|fisa real-valued bounded function
with domain [0, 1]},
define
Sf) = {9|¢ € X, |f(z) — g(x)| < ¢ for all

and define
Urs {UU > SG) torsome< > 0,

The topology, 3 = {u,|f € X}, is then a Hausdorff topology for

X. Observe that if {f,} is a sequence in XY which converges to
the limit f in the topology, 3, then lim f,(7) = f(x) uniformly on
[0, 1] in the ordinary sense of analysis, i.e., for each e > 0 there
exists an integer NV, such that n = N implies
 sec. 1.7 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 41

\fn(x) — f(x)| < for all x € [0, 1].

As we have seen, in an arbitrary topological space, limits of sequences

need not be unique. Since we are so used to having limits of sequences
unique, it would be pleasant to have some property that would guarantee
such uniqueness. This is precisely what motivates the introduction of
Hausdorff spaces, for such spaces have the agreeable property that
limits of sequences are unique, as in fact is shown by the following:

1.22. Theorem, be io be a Hausdorff me and let ay bea1 sequence V

t.e., the sequence {x,} has a uniqueo lini af itoe a limit.

Proof. Suppose {xn} converges to both x and y, with x # y. Let
Ue wu, V€U,, such that UAV = @, by 1.21. Now since {Tn}
converges to x, there is an integer N, such that n = N, implies fin (ea
€ 10)
and since {z,} converges to y, there is an integer N2 such that n = N2
implies x, € V. Select n sufficiently large so that n > N; and n > No,
then simultaneously 2, € U and x, € V, whence 7,€ UN\\V = @.
This is a palpable contradiction, and shows thatz = y. J

It is entirely possible that a sequence, {x,}, can converge to a point 5 ae

x, without x being a limit point of the set {v,|m = 1,2,...}. For |
example, on the real line if 2, = 1 for all n, clearly {x,} converges to 1, 2////
ye pO eC EE a Py
occur that a sequence {z,} has a limit point, without having {z,} wk ae
converge to that limit point, or for that matter to any point whatever. {,,
For example, on the real line, if z, = (—1)"+ 1/n, ther 1 is a limid d Sy
point of {z,}, yet {z,} does not converge to any point. Theo .
examples make things look rather black, but we can be assuréd a bit Om) ox
by the following: (Bede Me eae
‘ be ped of If’ age ys tal nm
[ 7

J 1.23. Theorem. Jf X is a Hausdorff space, {xn} a seguence in X such

that {xn} converges to some point x € X, and af further y is a limit point
of theset {a,|n = 1,2,...}, thena
=zy. Yur
Proof. Suppose « # y. Then there, eSoeist U € Uz, V € WU, such that
Ges ae “Also, since {2,4; on thoy aL
tox fierce exists an integer NV,
such that n = N implies x, € U. Now by 1.21 there exists for each 2,
1 Stay, Wi E Uz, Vi € U, such that W;1 V; = ©, unless y =
ea ¢
 r 4 v

40
‘
of, . af7 Om
’

"4

42 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

for some 7. Define Vj = V; in case x; ¥ y, and V; = V in case x = y,

and define

Vsvn (0 vi)
then y € V’ € U, by 1.1(3). Nowifzissuch that z € V’andz # y, then
z * x; for any 1, sinceif 7 => N,a,€ U, and UNV’ CUNV=Q,
while if 7 < N, then either 1; = y Xz, or x; € Wi, and W;M V’ ©
W;:0\V;= @. We have thus constructed a neighborhood V’ of y,
such that no point z ¥ y of the set {z,|n = 1, 2,...} belongs to V’.
This clearly contradicts the fact that y is a limit point of the set
1p,tt = 1.2) |, hence we 7,
Nx

58 3
Comparison of wees
Topologi | EOE
yr
s. aefi
AL
o {

1.24. Definition. Let X be a set and let 3, and 3, be two topologies for
X; further for each x € X let U;,7 be the family of neighborhoods of x in
the topology 3;, 71= 1, 2, then 3, zs finer than 3, provided that U2,. © VU1,2.
In this case we write 3. S .

In essence what the preceding definition says is that one topology is

finer than another if each point has at least as many neighborhoods in
the finer topology. The finest topology a space can have is clearly the
discrete topology, whereas the coarsest (1.e., least fine) topology a space
can have is the trivial topology. The finer the topology of a space, the
richer the space is in neighborhoods; and so one might expect that it
would also be richer in other structural features. This is, in fact, the
case, as is shown in the following:

1.25. Theorem. Let X be a set, and let 5,, 3, be two topologies for X;
further let ©, and ©, be the families of open sets generated by 3, and 3,
respectwely. Then 32 S 3; of and only if O2 € Oy. Wo e

Proof. Let O € 0. If O = @, then O € © by 1.5. If O = @, let

x € O, then by 1.3, O € Us, where U;,z,7 = 1, 2 has the meaning of
1.24, Since 3, S i, we have by 1.24 that U:,,C U2, consequently
O € U1,2, whence, since x was an arbitrary point of O, we have by 1.3,
O € 0;. Thus ©, € ©;. The proof of the converse is left as an exercise. J
 sEc. 1.9 TOPOLOGICAL ACES—BASIC DEFINITIONS AND THEOREMS 43

(1.26. )Corollary. Let X be a set, 3, and 3, two topologies for X such that
32 S hi, and let C; and @; be the family of closed sets generated by 5, and 5,
respectively, then C2 € @y.

Proof. Let C € @,, then X —C € © by 1.14. Thus by 1.25, ,

X —C€ OO, and again by 1.14, C = X — (X — C) € G, and conse-
quently @,C @;.

EXERCISES

1.61. Let X be a set, 3, and 3 two topologies for X, then 3, < 3, if

and only if for each x € X and each U € UW, € 5, there is a
We € WD € Se such that We Se GE

1.62. Let X be a set, 5, and 3, two topologies for X, then 3; = 3, if and

only if both 3, S 3, and 3, S 4.

1.63. Prove the converse portion of 1.25.

§9 Bases,Countability
Axioms, Separabilit Wdroy

We have seen at least two ways to define a topology for a set X¥. We
can simply define a neighborhood system for each point and take the
family of all such neighborhood systems as the topology, or we can
define a family of open sets, and define a neighborhood of a point as any
set which contains an open set which contains the point. There is
another way in which we can define a topology, namely by defining a
base, as follows:

_/4 1.27. Definition. A base (or basis) for a topology, 5, of a space X is a

set,.® € O, the family of open sets of X, such that for each x € X and each
U € Uz, there exists V € @ such thatx € V CU. The sets of & are called
basic sets.

Note that if @ is a basis for X and Y C X, then

BOY = {B'|B =BOY,BE SB

is a basis for the relative topology of Y. @ () Y is said to be the basis
for Y induced by the basis ® for X. [Cf. Ex. 1.64(b) and Ex. 1.72.]
 44 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. |

EXAMPLES (AND EXERCISES)

1.64. (a) Let R be the reals and let ® = {(a, b) |a <b, a,b € Rj, ie,
the set of all open intervals, then @ is a basis for the usual
topology of R.
(b) What is the basis for [0, 1] in the relative topology induced by
the basis in (a) above?

1.65. Let E be the real plane, and let S.(x, y) be an open e-sphere about
(xz, y) € H. Let

B= {S.(z, y) |e > 0, (x, y) € BH,

then @ is a base for the usual topology of EF.

1.66. Let X be a partially ordered set under “S,’’ let

S-(z) = {yl y 2 3},

and let ® = {S,(x) |x € X}, then @ is a base for the right order
topology of X.

1.67; Let X be the set of all real-valued integrable functions on [0, 1], let

Sf) = {glg€ X, frlf—glax <4,

es ie : 1

and let ® = {S.(f) | f € X, « > 0}, then @ is a base for the top-
ology of X described in Ex. 1.4.

Oe 1.68. Let X be a nonempty set, and let ® = {{x} |x € X}, then Bisa
basis for the discrete topology of X.

Ao 1.69. Let X be a nonempty set, and let ® = {X}, then @ is a basis for
the trivial topology of X.

The student should prove that the bases mentioned in the preceding
examples are in fact bases for the topologies described.

1.28. Theorem. Let X be a set, 3 a topology for X, then & is a base

for 3 if and only tf the set
sEc. 1.9 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 45

@={U B,|A an arbitrary indexing set,

aA

B. € B for each a}

as in fact the family © of open sets of X generated by 3.

Proof. Suppose @& is a base for the topology 3. Clearly, since each B,
is open, by 1.5(1) each element of @ is open, whence @ € ©. On the other
hand let O € 0, and let x € O, then by 1.3, O € U,, and by 1.27, there
exists B, € @ such that « € B, C O. Clearly

OnE.
x€O

whence O € @. Consequently © C @, and 0 = @.

Conversely, let @ have the required property, and let x € X, U € Uz,
then by 1.4 there exists O € © such that x € OC U. Now

Oa B.
acA

where B, € &. Thus since x € O, x € B, for some a, and x € Bz ©

O C U, whence by 1.27, Bisa base.

EXERCISE

E7Omletc x. = 40.42) eA 01) Be At 2) and. let

Ce EX; A, Bb, D3,

then @ is not a basis for any topology for X. [Hint. Observe that
any union of sets of @ is again in @, so that if @ were a basis for
a topology 3, @ = 0, the class of open sets of X generated by 3.
Show that @ cannot be a family of open sets.|

The viewpoint of Definition 1.27 is this: given a topology there

exists a family, 0, of open sets, and some subset of this is a basis. Now
we might start from the other end, and ask: if we start with some family
of subsets of a set X, can this family serve as a basis for some topology?
The preceding exercise makes it clear that not every family of subsets
of a given set, XY, can serve as a basis for some topology or other. We
might naturally ask, “Is there some property that distinguishes bases
from other families of subsets of a given set?” It certainly would be
 46 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

pleasant if there were such a property, since it would enable us to

distinguish bases from other families of subsets. It is, in fact, the case
that such a property exists, as is shown in the following:

A\p 1.29. Theorem. A family, ®, of subsets of a set X is a base for some

topology, 3, of X if and only uf both

(1) X= U B,, and

BaEB

(2) For each x € X and each pair U, V € 8, for which « € U and

x € V, there exists W € @ such thatn EW CUNY.

Proof. Let ® be a basis for some topology, 3, for X, and let x € X,

then there exists U € Uz € 3, such that x € U, and by 1.27 there exists
B, € @, such that. x € B, © U. Clearly

er UA psa. Wapsema
zEx BacB

hence X = ; cs B,, and condition (1) is met.

Let U, V € @,2 € Uandz E JV, define U’ = U (7 V, then since by
1.27, U and V are both open, so also is U’, whence U’ € U;. Conse-
quently, by 1.27, there exists W € @ such thata €E WCU'=UNVV,
and condition (2) is met.
Conversely, suppose ® satisfies both conditions (1) and (2). Define

© = {U B,| A an arbitrary indexing set, By € B}.

aGA

Now we wish to invoke 1.6 to show that 0 generates a topology 3, so

we need only show that 0 satisfies the hypotheses of 1.6. Since the union
of members of © is again a union of sets of ®, such a union is again a
member of 0, thus hypothesis (1) of 1.6 is satisfied. Condition (1) assures
us that X € 0, so that hypothesis (3) of 1.6 is satisfied. Let the indexing
set A = ©, then sy Bz = ©, so that @ € O, and hypothesis (4) of 1.6

is satisfied. Finally let U, V € 0. If U(\ V = @, then since @ € 0,

U(\V Eo. IfUC\V # @, then for each x € UM) V, selectBz € ®,
B; € 8, such that

ake 10PAGS 6 a CB Ve

By condition (2) there exists W, € @, such that

ee WeSC BNE
Say ye
sEc. 1.9 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 47

Finally let W = UU W,, thenclearly U \ V = W,andW € 0, since

xeGUunv
each W, € ®. Thus hypothesis (2) of 1.6 is satisfied, and © does indeed
induce a topology in X. §

An important class of spaces are those for which there exists a basis
with countably many elements—more formally:
r
1.30. Definition. A space X, which has a base, ®, which is a countable :
family, t.c.,® = {B;|7 = 1, 2,.. .} is said to satisfy the second axiom of
countability or sometimes more simply to have a countable basis.

We sometimes also speak of spaces being second countable, meaning

thereby that they have a countable basis, or satisfy the second axiom
of countability.

EXAMPLES (AND EXERCISES)

1.71. (a) Let R be the reals with the usual topology, and let

= {(a, b) |a, b rational, a < Bb},

then @ is a countable basis for R.

(b) Let # be the real plane with the usual topology and let

@ = {S.(x, y) |x, y, € all rational, « > 0},

then ®& is a countable basis for LE.

1.72. Show that any subspace of a space with a countable basis again
has a countable basis. [Hint. If Y C X and @ is the basis for X,
show first that 8B\ Y = {BO Y |B € B} isa basis for Y.]

Closely related to second countable spaces are separable spaces, defined

by:

1.31. Definition. A space X is said to be ie rrr at i a a

countable dense subset (Cf. 1.17), 7.e., af there exists A C X, A countable
such that A = X.

The relation between second countable spaces and separable spaces

is given by:
 48 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

\)‘ 1.32. Theorem. Let X be a topological space with a countable basis,

~ then X is separable.

Proof. Let 8 = {B;|7 = 1, 2, ...} be a countable base for X, and

define A ="{z;|.2, € B74 = 1,2). 3,116, A (NB, = Ye We prove
now that A = X. Let x € X;if «= x, for some7, thenz € A C A, so
assume that « ¥ 2; for each 7. Let U € Uz, then there exists B; € ®,
such that 2 € B;C U. Now x«;€ B;CU, and xz; 2, thus every
neighborhood U of x contains a point of A distinct from x, whence x € A.
Thus X G A,-and since A CX, A GC X = X, we have A = X, andA
is the required countable dense subset. ff

The converse of the preceding theorem is false, as is seen in the

following:

EXAMPLES (AND EXERCISES)

1.73. Let R be the reals (or for that matter any uncountable set), and
define the family, ©, of open sets by

© = {0O|0O = @, or O = R — F, where F is a finite set},

then © generates a topology for R by 1.6. Now let S be any infinite

set, then S = R; thus in particular if S is countable, we see that R
is separable in this topology.
Suppose F# has a countable base, ®, and let x € R, then the set
R — {x} is open by definition of ©. Define

@.= {8|BE@B,x eB) and ©; —{O|0eC ove OF

It is easily shown that

{CSG BIG ONO

BEB: 00,

whence {x} = uel B. Now since each B € @, is open, Be = F,

a finite set, thus

Ra) Sh =O pea ae
BE&: BEB;

and this last set is a countable union (since ®,, a subfamily of @, is

sec. 1.9 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 49

countable) of finite sets. Consequently R — {x} is countable. This

palpable contradiction shows that R, in the topology defined by 0,
is not second countable.

1.74. Let J be the unit square in the plane, i.e.,

=4@,9)(0Sm@5h,05
yS Uf;
and let 3 be defined as follows: if

(jy es = Gs Men) <ceeLO <ay < 1

define
UG OS Sana). \J) stor-some e > 0, where
S.(z, y) is an open e-sphere about (2, y)},

whereas if

ai ecolnl) Sania) cae, fas a=.

or yal >

define

Wen = {U|U D(a, y) OJ) U {(e, y}, for some € > 0}.
Then (J, 3) is separable, but has no countable basis.

The student may wonder why we have a second countability axiom,

without a first. There is in fact a first; namely, a space is said to obey
the first axiom of countability, be first countable, or have a countable
basis at each point, provided that for each x € X there exists a countable
family ®, C O such that for each U € U, there exists B € ®, such that
x € BCU. First countability is less demanding on the structure of a
space than second countability, for clearly any second countable space
is also first countable, while the following example shows the converse
to be false.

EXAMPLES (AND EXERCISES)

1.75. Let X be an uncountable set with the discrete topology, let x © NX,
then we may let ®@, = {B| B = {x}}, whence X is first countable,
while clearly X is not second countable.
50 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS cHaP. 1

1.76. What is the relation between first countability and separability?

That is to say, does either of these imply the other, or do there
exist spaces which are separable but not first countable, and spaces
which are first countable but not separable?

§10 Sub-bases, Product Spaces i +"

We have seen how we can construct a topology for a space by begin-

ning with either a neighborhood system (1.2) or a family of open sets
(1.6) or a family of basic sets (1.29). Each of these in turn is a step
removed from the original notion of neighborhood system. We can take
another step backward without falling over the brink of the precipice
and losing any chance of recovering a topology for a set. We take this
last gingerly step as follows:

1.33. Definition. Let X be a space, then § 1s called a sub-base for the

topology, 3, of X provided the family

k
{BIB = 85 8: € 3,4 = 1,2...)
t=1

ts a basis for 3. That is to say, the family of sets obtained by taking finite
intersections of members of § forms a basis for 3.

Our freedom of choice as we move backward through the families 0

(1.6), ® (1.29), § (1.33) becomes ever greater, as the hypotheses of 1.6,
1.29, and of the following theorem show.

1.34. Theorem. Lei X be a set, and let § be a family of subsets of X

such that

1) §#2,
(2) X = Wes,
SES

then § 1s a sub-base for some topology for X.

Proof. Let

k
B= {BIB = 1=1
8,565 b=1,2,..}.
smc. 20 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND TUEOREMS 51

Then clearly @ D §, whence

SSAA Day SIE @

BEB SES

whence X =U B. Nowlet2 € X, and let U, V € @, witha € Uand

BEB
x € V. Since U and V are finite intersections of elements of §, so also
is their intersection W = U ( V, consequently W € @, and clearly

LRA
SEN CEN AN ae

Thus the hypotheses of 1.29 are satisfied, and $ generates ® which in

turn generates 3, a topology for X. J

The sub-base which gives rise to a topology according to the theorem

and definition immediately preceding characterizes the topology it
generates in a unique way. This is shown by the following:

EXAMPLE (AND EXERCISE)

1.77. Let § be a family of subsets of a set X, such that § ¥ @ and

U S = X, let 3 be the topology generated by § according to 1.34,
SES
and let © be the family of open sets of X under 3. Then 3 is the
coarsest topology for which § C 0, in the sense that if 3’ is any
other topology whose family of open sets is 0’, then if § C ©’, it
follows that 3 S Jv’.

It is rather a nuisance to have to write each time we want to specify

some open set in a space, ‘‘O € 0, the class of open sets of the space X,”’
and we can avoid all this verbal fruit salad by agreeing that whenever
we write ‘‘O’’ we shall invariably mean the class of open sets of a space.
Of course, similar remarks apply to the other families of subsets of a
space which we have discussed, so let us make the following agreement:
Whenever we write

UW, we mean the family of neighborhoods of x € X,

3 we mean the topology of X,
© we mean the family of open sets of X,
© we mean the family of closed sets of X,
52 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

® we mean some basis for the topology of X,

S we mean some sub-basis for the topology of X.

We shall be careful not to use these symbols to stand for anything other
than the sets indicated in order, of course, to avoid any confusion.
One of the principal uses of sub-bases is in the introduction of a
topology into Cartesian products of spaces. With the tools presently
available to us it is difficult to describe the usual topology for a Cartesian
product of infinitely many topological spaces; in fact, the very product
itself is awkward to define until we have the notion of function. We thus
restrict ourselves, for the time being, to Cartesian products with finitely
many factors; and we put off until a later date, after we have had a
chance to examine functions and continuous functions, the description
of the standard method for defining and introducing topologies into
Cartesian products of infinitely many topological spaces.

1.35. Definition. Let X;,7 = 1,2,..., n, be topological spaces and let

©; be the associated families of open sets for each 1. Let

De wove
i=l

be their Cartesian product, and let

§=.S|S= X Yu, where Y; = X;for alli 4 J, and

i=1

Y;= Uzk C= lf 2) ea

then § 1s a sub-base for the usual topology for X. The topology so defined
is called the product topology.

The student should check (via 1.34) that the sub-base, §, of 1.35 is,

in fact, a sub-base for some topology. The points of X = x X; are of

i=1
the forma = (2, %,...,2%,) where x; € X; and the z; are called coordi-
nates of x. We may briefly describe the sub-base, §, of 1.35 by saying
that its members are simply collections of points of X in which all
coordinates but one are free (i.e., may be any point in the associated
sec. 1.10 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS 53

space X,) while one coordinate is restricted so as to lie in an open set

(of X;).

EXAMPLES (AND EXERCISES)

1.78. Let H = R X R, R the real line, E consequently the real plane,

and let O C R be an open set of the reals in the usual topology,
then O = U JI,, where J, is some open interval (Proof?) and we
ala

may assume that the 7, are disjoint. A sub-basic set for # in the
product topology is a set of the form R X O (or O X R) and isa
collection of “rectangles” of infinite extent over the intervals of O.
Basic sets of # in the product topology are certain collections of
rectangles in #, while open sets of # are any collections of rec-
tangles in H, where by a rectangle we mean a set of the form

Cr Waa © Oy, Ga << 02, Oy, 01, do, Os © hh OF — to}.

It is important for the student to prove that the product topology

for Fis the same as the usual topology—essentially what must be
proved is that for each (a, y) € # and each e > 0 there is a
rectangle R C EF such that (a, y) € RC S.(a, y) and for each
rectangle R C EF such that (a, y) € R there is a 6 > 0 such that
pice y) Gh (Ct Ex. 17).

Lo. Let X;, X2 be topological spaces, 5;, 32, their respective topologies,
©, O the associated families of open sets. Let @ be the basis for
the product topology generated by § according to 1.35.
(a) Prove that @ = {U, Us | U; e O1, Uz € Oo}.

(b) Prove that if X; and X» are Hausdorff spaces so also is

Xi X Xo.

1.80. Let X be a topological space and let Y C X be a subspace.

(a) Prove that if X is second countable so also is Y.
(b) Show by counterexample that X may be separable, yet Y may
not be separable.

ADDITIONAL EXERCISES FOR CHAPTER 1

ot Let %1, 32 be two topologies for a set X, let ®1, ®. be bases for 5, 32
respectively, then 3, = 5, if and only if for each B; € ®, and for
 TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS CHAP. 1

each x € By, there exists B, € ®: such that « € By, C B,, and also
for each B3€ @. and for each x € Bs there exists Bi € ®: such
that « € Bi C By.

1.82. Let X be any infinite set. Show that if we insist that every infinite
subset of X be open then X must have the discrete topology.

1.83. Show that the set

S = {(a, b)|a = —o and b real and finite,

or a real and finite and b = «}

is a sub-basis for the usual topology of the reals.

1.84. In a Hausdorff space X, the set {x} consisting of a single point

is a closed set.
1.85. A lattice is a partially ordered set in which each pair of elements
has a least upper bound and a greatest lower bound; more formally,
if S is a set and ‘‘S”’ a relation in S such that
(a) x Sx for allz € S,
(b) 2 S yandy S zimplyz = yforallz, y € S,
(c) eS yandy Szimplyz Sz for allz, y,z€ S,
then S is said to be partially ordered under ‘‘S.”’ A pair of ele-
ments, xz, y, in S is said to have a least upper bound in S provided
there exists 2 € Ssuch that z S zand y S z, and further if for any
u,x S uandy S u, thenalsoz S wu. A similar definition, with the
inequalities reversed, yields the definition for greatest lower bound.
Now let § be the family of all topologies on some set X, and
let 5, 32 € F. Further let 0; and 2 be the family of open sets
generated by 3, and 3; respectively. In the set F let ‘‘3, < 5,’’ mean
that the topology 3; is finer than the topology 3. Further, define

Ox {0|O € O, and O € 03} = 0,1)

©;

and define § = 0, U ©. Answer the following, giving proofs

where appropriate:
(i) Does 0, satisfy the hypotheses of 1.6 and hence generate a
topology, 3,.?
(ii) Is “‘S” a partial ordering for $?
(iii) Is 5, a greatest lower bound for 3; and 3,?
she, Lalo TOPOLOGICAL SPACES—BASIC DEFINITIONS AND THEOREMS Ye

(iv) Does § satisfy the hypotheses of 1.34, and is it hence a

sub-base for some topology 3*?
(vy) Is it so that 3.8 3° and 5, = 5*?
(vi) Is 3* a least upper bound for 3, and 3?
(vii) Is S a lattice under “‘S’’?
 CHAPTER 2

CONTINUOUS FUNCTIONS (MAPPINGS)

AND HOMEOMORPHISMS

$1
Functions _
It might seem most natural at this point, when we have the general
notion of a topology for a space pretty well under control, to try to
classify topological spaces according to various properties we might
conceive. We have already seen one such property, that of being a
Hausdorff space. Natural though this path may seem, it is somewhat
rocky, until we have a few further tools at our disposal; in particular
we need the notion of continuous function (or mapping) on one topo-
logical space to another.
We take it for granted that the student is familiar with the notion
of function, and list here simply some of the useful properties of func-
tions. If the reader has never proved these for himself, as indeed he
should sometime in his mathematical career, this is as good a time as
any for him to verify them.
Let f be a function on X to Y; we write then simply f : X — Y, and
designate by f(x), for « € X, the value of f at the point (or place) z.
We also call f(x) the image of x under f. We say that two functions
fi: X— Y and f.: X — Y are the same if fi(x) = fo(x) for all x € X,
and write then fi = fo. A constant function f : X — Y isa function such
that for some fixed y € Y, f(a)= y for all a € X. The identity function
on X,1: X — X, is the function 7, such that 7(7)= x for all x € Xx.
ve imsseotAC X under f is the set

oe FCA) = CM eve = f(x) for some x € A}.

Tig tesietion of fi X>Y to ACX, written f|Aor fa isthe,

function f : A—Y such that (f|'A)(2) = f(x) for spel ae As
circumstances
under may? fis called an extension of f | A to X
56
sec. 2.1 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS Wr

We observe the following properties for f : X — Y, where A C X,

BCX, A, CX for each y € G, an indexing set:

iO) 2.
2, f(ix})= {f@)}.
3. A C B implies f(A) C f(B).
4. A # © implies f(A) # @.
5. f(A U B) = f(A) U f(B) and more generally

nea thin heele

6. f(A \ B) C f(A) O f(B) and more generally

(QA) S O,flan
-» The inverse image of a set C C Y under f, designated by f-1(C), is
the set {w |a € X, f(x) € C}, and we observe the following properties,
Were > 1,0 GY. CY, A CX, C,Ce Y for each y € G, an
indexing set:

1. f(x) = y is equivalent to a € f(y). [Note that we write f(y)

in place of what is really correct, namely f—!({y}).]
Zo OD) oO):
BCs
(oN) ares fatC) & f-0D),
4. f7(C UD) =f7(C) Uf) and more generally

= uty wee
5. fC /N D) (=)f(C) CO f(D) and more generally

OD)
6. f(C*) = [f-(C)]*, where the superscript c indicates complemen-
tation with seeloe: to X or Y as appropriate.
GAS flay.
pnb (eae
The function f : X — Y is said to be onto if f(X)= Y,Fal f is said
to be one-to-one (or simply 1-1)if fer»calazimplies fy) = fay).ae note_
the following properties, where A C X, BCX, andC CY:

1. f is 1-1 if and only if f(A OB) = f(A) C\f (8) for all A and B.
tf tH them
: Coe tes
 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS CHAP. 2
58

2. f is onto if and only if f7(C) # @ for all C 4 @.

3. If f is 1-1 and onto, then f~ is in fact a function on Y to X.
4, If f is 1-1 and onto, then f(A‘) = [f(A)]*.
5. f is onto if and only if f[f(C)] = C, for all C.
Ww 6. f is 1-1 if and only if f“[f(A)] = A, for all A.
joie. f is 1-1 and onto if and only if for each y € Y, f(y) is a single
point of X.
If f: X > Y andg: Y —Z are functions, then gf : X — Z is the
function defined by gf(xz) = g(f(x)) for allz € X. gf is sometimes written
gef. We note that (gf) = f“g—, and thatfis 1-1 and onto if and only
if ff = cy and ff = zy, where zx and zy are the identity functions on
X to X and Y to Y respectively. Grote KIM.

§2 Continuous Functions (Mappings)

We are now in a position to define a continuous function or mapping

of one space into another. The motivating notion, intuitively at least,
is the idea that points which are close together in _X stay relatively close
together in Y after f has acted on them. We have, via the notion of
neighborhood, an idea of closeness; thus we make the following:

, )
d

yi
(
a
2.1. Definition. Let (X,3) and (Y,%) be topological spaces, let
f:X—Y be a function, and let x € X, then f is continuous at x if for
each V € Up) € V there exists U € Uz € 3 such that f(U) CV. The
Function
ftssaid tobecontinuous if it iscontinuous at each point x € X.

EXAMPLES (AND EXERCISES)

2.1. Show that the identity function 7 : X — X is continuous.

2.2. Show that a constant function c : X — Y is continuous.

2.3. (a) Show that f : X — Y is continuous at x € X if and only if for

each“ V€ Ania, f“(V) 6
(b) Show that if f: X > Y and g: Y—Z are continuous, so
also is gf : X — Z.

As we have just seen in the preceding exercises there are alternate

ways to define continuity of a function at a point. Also there are
 SEC. 2.2 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS 59

alternate ways to define continuity of a function on a space X, the more

important of which are given in the following: 4

2.2. Theorem. Let X and Y be topological spaces, f: X > Y ;

function, then the following are logically equivalent statements: ,.- eo

() fizs continuous. fe ar tag

(2) Forr every open
open set O Ci (0) 4s, openin xc
(3) For every closed set C C Y, f“(C) ts closed in X.
(4) If A CX, then fA) CfA).
2 HB it

NoTE: The symbol ‘=>”’ used in the following proof is read ‘‘implies,”’
and we prove the theorem by showing that each statement implies the
next and that the last statement implies the first.

Proof. (1) => (2). Let x € f4(O), so that f(z) € O. Since O is open
( it is a neighborhood ofu(x), thus by;2,1 there exists V a neighborhood
\ of e such that f(V) C 6! Whence0Ei (0) and f(O) € W.. Thus by
4 Ge 13
YL)1.3, f-1(0) is open, Ni =C = hai ene o&c
(2) => (6). Let O-= Y — C, then O is open dn 'Y and by (27 (0) ah
is open in X WHY x — (2100)= (PO) ao 66-10)jg
is closed. 0)
_3) => (4). We bE eeive Ghat ACS (hice Ff, and since he
<i0.0) 8 plonea) by (3), f2[f(A)] is also closed. us by Pia AG |
if F-[F(A)], whence f(A) C f(A). yy « Uw (i); oS
(A) ==(1); Let aC X, nd U bea pee ees Select
msO open such that f(x) € OCU, and let V = [f*(09]*, then since lav
1a) € O, fla) Z C Diao) an cee Ve We need only show f“(0°) | G
a closed, for then Vvil be open, hence a neighborhood of x, and

HS fV) =HEHONY FF MOID =FFVO) SG g SU

cee ced
We now show f-!(0°) ae Since in any case fa(O°) C f7(0*), and
since by (4)

GeO) GTO) Oe a0,

we have also

FO) SfO%,

thus f“(O*%) = f7(09, and f+(O% is closed. J

 60 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS CHAP. 2

The preceding theorem is important, since it is often necessary to

show that a function is continuous, and this theorem gives us several
methods for proving functions continuous. The name mapping is a
useful term for continuous function and from now on we’ll use it in that
sense; that is, any function that we call a mapping will automatically
be continuous.
Many of the useful and important properties of topological spaces
are preserved under mappings. By this is meant the following: If X is a
topological space which possesses some property P andif f: X — Yisa
mapping into a topological space Y, then it is frequently the case
(depending upon the property P as well as to some extent upon the
structure of the spaces X and Y) that Y also possesses property P.
Several examples of such properties occur in the following theorems.
i,
2.3. Theorem. Let X and Y be topological spaces, f: X — Y a 1-1
“mapping, lee A C X and let x € A’, then f(x) € f(A)’.
Proof. Let U be a neighborhood of f(x), then by 2.1 there exists V,
a neighborhood of x, such that f(V) GC U. Since x € A’, there exists
y ~x,y € Vsuch that y € A, whence f(y) € f(V) C Uandf(y) € f(A).
Further, since f is 1-1, f(y) ¥ f(w). Thus in each neighborhood U of f(z)
there is a point f(y) ¥ f(x) such that f(y) € f(A), whence f(x) € f(A)’. Jj
/
NW 2.4. Theorem. Let X be a separable space, f : X — Y a mapping onto
Y, then Y ts separable.
Proof. Let {x,} be a countable dense subset of X, then {f(2,)} CY
is again countable, and further we claim that {f(z,)} is dense in Y; for
let y € Y, and let U be any neighborhood of y, then there exists x € X
such that f(z) = y and there exists V € U, such that f(V) C U. Since
{tn} 1s dense in X there exists an integer 7 such that x; € V, whence
f(x.) € f(V) € U, and {f(z,)} isdensein Y. J

NOTE: In the above theorem, since f : X — Y is a mapping onto, we say

that Y is a continuous image of X.

§3 Homeomorphisms

Two spaces, one of which is a continuous image of the other, are, of

course, related to some extent; however, there exists an even more
intimate relation between spaces, which we now define.
SEC. 2.3 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS 61
ae
2.5. Definition. Let X and Y be topological spaces, and letf : X — Y
be a one-to-one function from X onto Y (so that f-! is also a function).
Furthermore let both f and f- be continuous, then f is said to be a
homeomorphism.

We observe that the relation between spaces, X ~ Y, defined by

X ~ Y if and only if there exists a homeomorphism f : X — Y, is an
equivalence relation, for the identity function on X to X shows the
relation to be reflexive. Also, if f : X — Y is a homeomorphism then
fi: Y —X is one as well, and finally if f: X —~ Y andg: Y > Z are
homeomorphisms, then so also is gf : X — Z. We thus see that the
relation ““~”’ is symmetric and transitive as well. We may thus say that
if a homeomorphism exists from X to Y (hence by the symmetry of the
relation ““~,”’ between X and Y) then X and Y are homeomorphic; and
we may also say that Yas a homeomorphic image of XY.
Let us make a slight digression here. We are talking about an
equivalence relation in the class of topological spaces. Now, as the
reader knows, every equivalence relation in a set partitions the set into
equivalence classes,

[X] = {Y | Y is equivalent to X}.

In the case we are examining, [X] is the set of all spaces homeomorphic
to X. One of the problems of topology is to characterize the elements
of [X] for particularly well-known spaces X in terms other than their
being simply homeomorphic to X. For example, if Y = [0, 1] with the
relative topology inherited from the reals with the usual topology, is it
possible to give certain characteristic properties of a space Y to guar-
antee that Y € [X] if and only if Y has these properties? The answer
is “Yes,” although we shall not explore far enough in this book to
achieve this result. The interested reader can pursue these matters in
more advanced texts on topology.* Now let us return to the main path
of our discussion. 4
Many properties of spaces we shall eventually study are properties
which are preserved by homeomorphisms. Specifically, if, when X has —

Wey tatthepropertyPeprervelhyhoneaorhi, andwe

* See, for example, J. G. Hocking and G. 8. Young, Topology (Reading, Mass.:
Addison-Wesley, 1961), p. 129.
 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS CHAP. 2
62

eall such a property a topological property. We remark at this juncture

that any property that is preserved by continuous functions must also
be preserved by homeomorphisms. Thus, among other things, 2.4 proves
that separability is a topological property.
Before we continue with our study of homeomorphisms, it will be
useful to have a few more terms to describe functions. These terms are
given in the following:

2.6. Definition. Let f: X — Y be a function, and let X and Y be

topological spaces. If for each open set O © X, f(O) is open in Y, fis saad
io be an open function (or sometimes an interior function). If for each
closed set C C X, f(C) is closed in Y, then f is said to be a closed function.

EXAMPLES (AND EXERCISES)

2.4. Show that f: X — Y is a homeomorphism if and only if f is a’

one-to-one open mapping onto Y.

2.5. Show that f: X — Y is a homeomorphism if and only if f isa a

one-to-one closed mapping onto Y.
2.6. Let f : X — Y be a mapping, and let Z C X be a subspace of X. 4
Show that the restriction of f to Z, f | Z : Z — Y is also a mapping.

2.7. Let & be the real line with the usual topology, and let

Ko= 40, Yt, Y 6 Basa?

yy? 1
be the unit circle with the relative topology inherited from the
plane. Let f: R > K be defined by f(x) = (cos ra, sin 72), ie., f
just wraps the real line round and round the unit circle. Show
that f is not closed by showing the set

S20 lio ee

is a closed subset of R, but that f(S) is not closed in K. Note,

however, that f is open and continuous.
2.8. Let J; = [—3, 3] with the relative topology inherited from the real
line. Definef : 11 > l; by f(x) = $(x? — 32). By considering the
set S = {x|2 € h,0 < x S 3} show that fis not open. It is clear
that f is continuous and, as we shall see later, f is also closed.
 SHC. 2:5 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS 63

Since the requirement that a function be a homeomorphism is more

demanding than that it be merely continuous, we should expect that
more properties are preserved under homeomorphisms than are pre-
served under mappings. Such is, in fact, the case, and the following
theorems point this out.

2.7. Theorem. Let X and Y be topological spaces, f: X > Y a fe

homeomorphism, and let X be a Hausdorff space, then Y is a Hausdorff
space.

Proof. Let y1, y2 € Y, with y; ¥ ye. Since f is one-to-one there exists

a unique pair of points x, and ze such that f(a) = y, and f(a) = yp.
Since X is Hausdorff there exist U; and U3 open and disjoint such that
x1 € Uj, and x, € Us. Sincef is open f(U;) and f(U2) are open in Y, and
since f is one-to-one f(U;) and f(U2) are disjoint. Finally we note that
~ yr € f(U1) and ye € f(U2), whence Y is Hausdorff.
Na
2.8. Theorem. Let X and Y be topological spaces, f: X ~Y a
homeomorphism, and let X have a countable basis, then Y has a countable
basis.

Proof. Let ® be a countable basis for X, and let @’ = {f(B)| B € B},

then since f is open ®&’ is a family of open sets. Also, since & is countable,
so also is &’. Now let y € Y, U € WU, then there exist x € X, V € Uz,
B € @ such that B C V, and such that

y = f(z) € fB) Sf) SU.

Letting B’ = f(B) we see that there exists B’ € @’ such that y € B’ CU,
and that consequently @’ isa basis for Y. J

The preceding theorems might just as well be stated more concisely

by saying that the property of being a Hausdorff space (or possession
of a countable basis) is a topological property.

EXAMPLES (AND EXERCISES)

2.9. Let f: X — Y be a function, let X have the discrete topology,

then f is continuous.
 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS CHAP. 2

. Let f : X — Y be a function, let Y have the trivial topology, then

f is continuous.

. Show by example that if f: X — Y is continuous and X is

Hausdorff, Y need not be Hausdorff.
. Show that f : X — Y is continuous if and only if for each B € 8,
a basis for Y, f~!(B) is open in X.

. Let f : X ~ Y be continuous, let {x,} be a sequence in X such

that lim z, = 2, show that lim f(z,) = f(z).

. Let f : X — Y bea function; show that f is continuous if and only

if for each B C Y, f(B) C f-(B), and show by example that
proper containment can occur.
5. Let X and Y be topological spaces, X X Y their product space,
i) on Or

define the functions Py : X XK Y— X and Py: X XK Y—Y by

Px((z, y)) = xand Py((z, y)) = y for each (x, y) € X X Y. Show
that Px and Py are continuous and open. Px and Py are called
projections of X X Y into the coordinate (or factor) spaces of
AY 2
2.16. Let R be the reals with the usual topology and let X be the reals
in which the topology is defined (via 1.6) by defining the family of
open sets to be the complements of finite sets or the empty set,
1.e.,
= {0|0 = © or O¢ is finite}.

Define f : R — X by f(x) = x, and show that f is one-to-one and

continuous, but f—! is not continuous.

2 Nie (a) Let X and Y be topological spaces, Y a Hausdorff space, and

let f: X — Y and g : X — Y be mappings. Let

E = {x|x € X, f(x) = g(a)},

then # is closed.
(b) Further let D C X be a dense subset of X, then if D C £,
f=g [ie., f(x) = g(a) for all x € X].
(c) Let f be a real-valued continuous function of a real variable
defined on the rational numbers. Show that there exists at
SEC. 2:4 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS 65

most one function g defined for all real numbers with real
values and continuous, such that f(x) = g(x) for x rational.

2.18. Let f : X — Y bea function, show that f iscontinuous if and only

if f-1(S) is open in X for each S € 8, a sub-base for the topology
of LY.

2.19. Let f: X — Y be a mapping. Let A C X, and let A have the

relative topology, then f | A : A — Y is continuous.
2.20. Any two open intervals of the real line (with the usual relative
topology) are homeomorphic. [A similar statement holds if ‘open’
is replaced by “‘closed.”’|

2.21. Is the continuous image of a second countable space again second

countable?

$4 Product Spaces : Shel

Exercise 2.15 immediately preceding gives us the device for extending

the notion of product topology to Cartesian products with arbitrarily
many factors. Let us begin with a product of two spaces X; and Xe, then
a point (a, x2) in X; X Xe can be thought of as a function on the set
{1, 2} whose value at 1 € {1, 2} is x1, and whose value at 2 € {1, 2} is
Xo. We can thus think of X, X X2 as the class of all functions (2),
4 = 1, 2 on the set {1, 2} such that x(1) € X, and x(2) € X2; however,
instead of writing x(1) and (2) we prefer to write x, and x22. We now
introduce a topology into the set X; X X2 by insisting that each pro-
jection Px,, 7 = 1, 2 be continuous and we can do this by insisting that
Px) (U;) be open where U; € X;,7 = 1, 2 is an open set of X;. Thus, in
all, the product space X; X X2 can be thought of as the set of all func-
tions x on the set {1, 2! whose value at 7 € {1, 2} is z,, with the topology
generated by the sub-base

{Px3(U;) | Us € ©;:, ©; the family of open sets of X;, 7 = 1, 2}.

The generalization to arbitrary indexing sets A, rather than the set

{1, 2}, is now pretty open to us, and we make the following:

2.9. Definition. Let A be an indexing set, Xa, a € A a topological

space, then the product space X X, 1s defined to be the set of all functions
aA
 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS CHAP. 2
66

x on A such that xq (the value of the function x at a € A) ts an element of

X., whose topology is generated by the sub-base

{Pz}(Ua) | Ua € Oa, the family of open sets of Xa, a € A},

where Px,(v) = tq is the projection of X Xz. into the “ath” coordinate

aca

space Xa.

Since we were motivated in making our definition by the requirement

that each Px, be continuous, it is not surprising that at least half of the
statement of the following theorem is correct.

2.10. Theorem. Let X = X X, be the product space of the spaces Xa,

acA

where A is some indexing set. Let Px,(x) = xg be the projection of X onto

Xa, then Px, is both open and continuous.

Proof. First we prove continuity. Let x € X and let U be a neigh-

borhood of xg € Xz. Select V open such that zg € V € U, then by 2.9,
Px3(V) is open in X, andz € Px,(V), whence Px, (V) is a neighborhood
of x. Finally Px,(Px,(V)) = V © U, whence Px, is continuous.
We now prove Px, open. Let U be open in X, and let x € U (since
if U = ©, there is clearly nothing to prove), then U is a neighborhood
of x. Since U is a union of basic neighborhoods, we may as well assume
that U is itself a basic neighborhood, whence U is a finite intersection
of sub-basic neighborhoods, i.e.,

Une (\ Pz Us),
acCF

where F C A is finite and U, is open in Xq.

Now if 6 € F, let y € U, then since y € Px, (Us),

Px,(y) € Px,(Pxp(Us)) = Us, and Px,(U) = Us,

thus if 6 € F, Px, is open.

If 6 ¢ F, we wish to show that Px,(U)= To this end let
zp € Xz, and construct the point z€ X such that 2. = wa € U.fora € F,
Zp = 2s for 8 ¢ F. Then since x € Ua, 2 € Px/(U.) for each a € F,
whence z € U, and Px,(z) = zg. Consequently Py,(U) = Xg, which is,
SEC. 2.4 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS 67

of course, an open set in Xs. Thus if 6 ¢ F’, Px, is also open, whence in
any case Px, is open. §

Now we ought to point out a few things here. First of all, it is not
the case that Px, is closed. For example, in R X R, the set

CU) ye — ey Orme 1)
is closed, yet the projection of this set on the z-axis is not closed, being
the open unit interval. Secondly, it should be emphasized that if U, is
open in XQ, it does not necessarily follow that x U, is open in X, since
aGA

in general it is not possible to write XK U, as a union of finite inter-

aEA

sections of sub-basic sets. Of course, if A happens to be finite, then

x U, is open, but if A is infinite this need not be the case.
acA

We now prove two theorems which are of general utility in many of

the later theorems.

2.11. Theorem. Letf: Y— X X,beafunction, then f is continuous

aA

uf and only «f the composite function Px,f : Y — Xq 1s continuous for each

ae A.

Proof. If f is continuous, then, since by 2.10 Px, is continuous, so

also is Py,f. On the other hand, let Px,f be continuous for each a, and
let U, be an open set in X,, then

Pee Og = 7 lex (U.)|

is open in Y. However, Px,(Ua) is an arbitrary sub-basic set of the
topology for X, hence by Ex. 2.18 f is continuous. J

2.12. Theorem. Let X and Y be topological spaces, and let f : X > Y

be a function. Define F: X ~>X XY by F(x) = (2, f(x)), then f is
continuous if and only if F is a homeomorphism of X with the subspace
G = {(a,f(a))
|a € X} of
X X.¥.
Proof. Let F be a homeomorphism of X with G, then by 2.11, Py
is continuous. Pyf : X — Y is defined by (PyF)(x) = f(x), whence
PyF = f, and f is continuous.
Conversely, suppose that f is continuous. Now we know from 2.10
that Py : X X Y — X is continuous, consequently Px |G is continuous;
68 CONTINUOUS FUNCTIONS AND HOMEOMORPHISMS CHAP. 2

however, observe that (Px |G)(z, f(z)) = xz, whence Px |G = F—, and
F~ jis continuous. We also remark that F is clearly one-to-one and
onto G, thus we need only show that F is continuous.
Now PxF : X — X is defined by PxF (x) = Px[(a, f(x))] = 2, thus
PxF is the identity function on X, hence is certainly continuous. Also
PyF : X — Y is defined by PyF(x) = Py[(z, f(z))] = f(x), whence
PyF = f, and f is continuous by hypothesis. Therefore, since both PxF
and PyF are continuous, F is continuous by 2.11. This then establishes
that F isa homeomorphism. Jf

The use of the letter G to describe the set mentioned in the previous
theorem is not fortuitous, for the set G may be thought of as the ‘‘graph”’
of the function f. Thus 2.12 says in somewhat altered language that
f:X-—Y is continuous if and only if X is homeomorphic with the
graph of f under the homeomorphism F(x) = (a, f(zx)).

EXAMPLES (AND EXERCISES)

2.22. Let A be an indexing set, X, a topological space for each a € A,

and let X X. be the product space. Let 8 € A be fixed, and for
aA

each a ¥ 8 let x, € Xq be some fixed point. Define

Xp = {x |rte = x), for a ¥ B, and zz arbitrary},

then X¥; © X X, and Xz is homeomorphic to Xz.

aGA

2.23. Use the preceding exercise to show that the set of points

Ri = {(a,
1) |x € R}
is a subset of the plane and is homeomorphic to the reals.
 CHAPTER.3

VARIOUS SPECIAL TYPES OF

TOPOLOGICAL SPACES
(VARIETIES OF COMPACTNESS)

In this chapter we shall study a variety of special restrictions one

can put on a space so that a number of interesting and desirable prop-
erties result. Although we call these restrictions ‘special’ in the sense
that not every topological space need have them, they are nonetheless
quite general in that many of the important and interesting topological
spaces that we encounter frequently in analysis, algebraic topology, and
elsewhere possess one or more of these properties.

§1 Compact Spaces

Before we begin the study of our first special property we need a

preliminary notion, introduced in the following:

3.1. Definition. Let X be a set, Y C X, and let Ww Hl

{Da |a € A, an indexing set}

be a family of subsets of X, then {D,} ts called a cover or covering for Y

provided Y Dz > Y.
aGA

If all the sets, D., of a covering have some common property, P, we

call the covering (or cover) a P-covering (or P-cover). For example, if
each of the sets D, is an open set, we call {D,} an open cover. Further,
if (Es | Po BES Dela € At and Eg > Y then {E,| 6 € B} is
called a subcovering of {D,|a€ A}. We may also characterize the
coverings of a set Y by properties of the indexing set 4; for example, if
69
 70 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

A is finite, we say that {D.| a € A} is a finite cov ering, or if A is

countable we say that {D,| a € A} is a countable covering. .
There is, of course, some ambiguity here, since the adjective pre-
ceding the word “covering” can refer either to some common property
of the sets D, or to some property of the indexing set A; however, the
context will usually allow only one interpretation, and if more than one
is possible, we will make it explicit whether the adjective refers to a
property of A or to some property common to the D..
The first property of spaces we investigate is called compactness and
is defined in the following:
Z
Kor 3.2. Definition. Let X be a topological space, then X ts said to be
compact provided each open cover of X contains a finite cover. (Here ‘‘open”’
refers to a property of the D., while ‘‘finite” refers to a property of the
indexing set A.)

The following picture of the notion may help. Suppose a large crowd
of people (possibly infinite) is standing out in the rain, and suppose each
of these people puts up his umbrella, then they will all stay dry. It is,
of course, possible that they are all crowded so compactly together that
not all, but merely a finite number of them need put up their umbrellas,
and still they will all stay dry. We could then think of them as forming
some sort of compact space. It is, of course, assumed in all this that the
umbrellas are open.

EXAMPLES (AND EXERCISES)

3.1. (a) Any closed interval, [a,b], -~7 <a b < = of the real line
with the relative topology induced by the usual topology of
the real line is a compact space. (This is one formulation of
the Heine-Borel theorem.)
(b) More generally, any closed bounded set of real numbers is
compact in the relative topology induced by the usual
topology of the reals.

3.2. Let X be a partially ordered set with the right order topology, and
let X have a smallest element (i.e., there isan z € X such that for
ally € X,x Sy), then X is compact.

It should be noted at this point in the development that the notion

sec 3.1 SPECIAL TYPES OF TOPOLOGICAL SPACES wl

of compactness for topological spaces is motivated by the Heine-Borel

theorem, which asserts (in a version somewhat different from that given
in the foregoing problems) for the real line, that if K is a closed bounded
set of real numbers and {Z,| a € A} is a family of open intervals such
that U I, > K, then a finite number of the 7, cover K.
ala

A simple, but quite useful, consequence of 3.2 is the following:

3.3. Theorem. Let X be a topological space, then X is compact if and (ur

only uf for each family of closed sets {Ca|a€ A} of X, ey Cy = Oo
umplies that there exists F C A, F finite, such that (\\ Ca ay
acer

Proof. Let X be compact, and let {C.| a € A} be a family of closed

sets with vacuous intersection. Define O, = C%, then O, is open, and

U 0.= U C= hay aes

aGA aGA

hence by 3.2 there exists F C A, F finite, such that U O, = X, whence

aCF

= penne
Sheree ar
wees
ac€P

The proof of the converse is left as an exercise. ff

We wish to obtain some conditions equivalent to compactness beyond

the one obtained in 3.3, and in order to do so it will be useful to have
one further property of families of subsets of a space X, namely

3.4. Definition. Let X be a set, {D.| a € A} a family of subsets of X, p*

then {D,} is said to have the finite intersection property provided that for
any finite, nonempty subset F of A, (\ Da FD.
acer

The following two equivalent formulations for compactness will be

of considerable utility to us in what follows.
ur
3.5. Theorem. A space, X, 1s compact if and only if for any family
{D.| a € A} of closed sets with the finite intersection property (\ Da # ©.
acA
“Ibo SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

Proof. Let X be compact, and let {D. |a € A} be a family of closed

sets with the finite intersection property. Suppose By D,= ©, then

by 3.3 there exists F C A, F finite, such that \ D. = @, and this

acer

contradicts the fact that {D,} has the finite intersection property. Thus
(“EDA
aA
Conversely, let {O. | a € A} be an open covering of X, and suppose
that for each F C A, F finite, U O. 2 X,ie., (1S) Ee #~ @. Define
aeF aEPr

Da = O%, whence D, is closed, and for any F C A, F finite,

(\ D, = Pere ae 2k
aEr aEr aEF

Thus {D,} is a family of closed sets with the finite intersection property,
and by hypothesis

\ Da = Oy Oo (ee alae
aA aA alA

or U 0, £ X. This last statement, however, contradicts the fact that

acA

{O.} is a cover for X, and this contradiction shows that there does exist
a finite set F C A such that U O, = X, whence X is compact. ff
a€r

3.6. Theorem. Let X be a topological space, then X is compact if and

only of any family {D, |a € A} of subsets of X with the finite intersection
property has the further property that (\ Dz # @.
acA

Proof. Let X be compact, then since for each a, Dz D Da, it is clear

that the family {D,| a € A} is a family of closed sets with the finite
intersection property. Consequently by 3.5 © D. ¥ @.
aA

Conversely, let any family {D. |a € A} with the finite intersection

property have the further property that the intersection of the closures
of all the sets of the family is nonvacuous. Now let {C.|a€ A} bea
family of closed sets with the finite intersection property, thenCz = Ca,
and

(NS Cg nN RO
eae
aGA acA

and again Ly 3.5, X is compact. J

SEC. 3.1 SPECIAL TYPES OF TOPOLOGICAL SPACES 73

EXERCISE

3.3. Prove the converse portion of Theorem 3.3.

We note now that compactness is preserved under mappings. he

3.7. Theorem. Let f : X — Y be a mapping onto and let X be compact,

then Y 1s compact.

Proof. Let {O.|a€ A} be an open covering of Y, then since f is

continuous f—1(O.) is open in X for each a by 2.2(2), whence

{f"(Oa) | € A}
is an open covering of X. Thus, since X is compact there exists F C A,
F finite, such that {f-'(O.) |a € F} is a cover of X. Finally, since f
is onto,

(fG7(«)) |a € F} = {On| a € F}
is a cover for Y, and therefore Y is compact. §

We now observe that as a simple consequence of the preceding

theorem, compactness is a topological property, that is to say

YL
3.8. Theorem. Let X and Y be homeomorphic spaces, and let X be
compact, then Y 1s compact.

Let us now investigate briefly some relations between the properties

of being compact and being closed.

3.9. Theorem. Let X be a Hausdorff space, and let Y be a compact

subspace of X, then Y is a closed subset of X.
Proof. We shall show that Y° is open. To this end, let x € Y*, then
for each y € Y there exist neighborhoods O,,, O, which by 1.4 may be
assumed to be open such that x € O,, y € O,, and On (\ O, = ©,
since X is Hausdorff. Now {0O, (\ Y |y € Y} is an open cover for Y,
thus since Y is compact in the relative topology, there exists a finite
set F such that
EON Y):
yer
 74 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

Define O = () O,,, then x € O, O is open and hence a neighborhood

yer

of x. Further

ERG noSENG RA LON NSAI I 12

yr

for if there were a z € O1) ( U Ov); we would have z € O, for some

yer
y’ € F, yet
BC Ove GN One GeO),
yer

whence O, \\ Ory # @, and this contradicts the choice of O,, and Oy.
Finally, since OM. Y = @, x € OC Y*, whence Y° is open, and
from this it follows that Y is closed. §
AN
\ 3.10. Theorem. Let X be a compact space, and let Y € X, Y closed,
then Y is compact (in the relative topology).
Proof. Let {O.| a € A} be an open cover of Y in its relative topology,
i.e.,O. is openin Y. Then by 1.18, 0. = Y ( O4, where OZ is open in X.
Observe also that Y° is open, thus

{Do De = Os or Da Vo

is an open cover for X, since if x € X,z € Y°or

Ld oat Gal ih eoREGon SOY

aGA aA

Now since X is compact, there exists F C A, F finite, such that

(De) U Ye = X,

whence U D, > Y, where each of the D, is of the form 04, consequently

a€F

Yow DL = 05 and finally

a€Pr acer

Y = U0 Yee
ar aGPr

from this it follows that Y is compact. ff

SEC. 3:1 SPECIAL TYPES OF TOPOLOGICAL SPACES 75

Up till this point the questions and answers (via theorems) have
been relatively easy. We come now to a problem of an order of difficulty
not encountered before. The question we want to answer is this:
Suppose we have a family of spaces X., a € A, an indexing set, such
that each X, is compact, then is it the case that a Xq is compact?

The answer to this question is in the affirmative, and is given by a

theorem due to Tychonoff, but we shall need a bit of machinery to get
it. First we shall assume a lemma due to Zorn, which is in fact equivalent
to the axiom of choice. The statement of this lemma is as follows:

3.11. Lemma (Zorn). Let X be a partially ordered set, and let each V1
simply ordered subset of X have an upper bound in X, then X has a maximal
element. [That is, if for each Y C X, such that Y is simply ordered, there
exists 2 € X such that y € Y implies y S z, then there exists m € X such
that for any x € X, either x and m are incomparable (i.e., neither « S m
norm & x is true) or x S m.]| foo ok

We make no attempt to derive this lemma from the axiom of choice.

The interested student can find such derivations in any of a number of
texts on set theory.
In preparation for proving Tychonoff’s theorem we prove first a
lemma which is more set-theoretical than topological, but which will
considerably shorten the proof of Tychonoff’s theorem. We have in
mind the use of the finite intersection property for the proof that
X = X X, is compact where each X, is compact, and that is why the
alA

following lemma concentrates on sets with the finite intersection

property.

3.12. Lemma. Let X be a set, § = {F} a family of subsets of X such

that F has the finite intersection property, then there exists a family
mm = {M} of subsets of X with the following properties:

(1) MM has the finite intersection property.

(2) ENG aes:
(3) MN is maximal with respect to properties 1 and 2 (1.e., if M CN,
NAM, NMIF, then MN does not have the finite intersection
property).
76 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

(4) Any finite intersection of elements of IN is again an element of IM.

Proof. Define

8 = {¢|¢= {1G}, @ CX, g DS, g has the

finite intersection property},

that is, % is the class of all families of subsets of X which contain ¥ and
also have the finite intersection property. We introduce a partial
ordering into B as follows: let @1, GB. € B, then @: S B®, means Bi € Re.
We now show that the hypotheses of Zorn’s lemma are satisfied in S
with the partial ordering ‘“<.”’ To this end, let @., a € A, where A is
some indexing set, be a simply ordered subset of B. Define ® = U Ba,
acA

i.e., ® contains as elements any set which occurs in any ®,. We now
verify that @ € B. First of all, since each ®, 25, it follows that
@® > F; also, since each B € @ is obtained as some element of some ®a,
each such B C X. Finally we must see that @ has the finite intersection
property. Thus let B,, Bo, ..., B, € ®. Now B; € @,, for each 7, by
definition of @, and since the ®, are simply ordered, there exists a J,
137 =k, such that Gz. S @,, for all7, 1 S2s k% This means) that
Ba; © Ba; for all 14, 1 Sik. Now @,, has the finite intersection
property, and each B; € ®,,, 1 S27 k, thus

1 Bi ¥ ,
k

v=1

whence @ has the finite intersection property. Finally it is clear from

the definition of @ that @® D @, for each a, hence ®, S B, and con-
sequently each simply ordered subset of 8 has an upper bound in &.
We are now in a position to apply Zorn’s lemma to 8. We have seen
that % with the partial ordering introduced into % satisfies the hy-
potheses of Zorn’s lemma, thus let 9% be a maximal element of %,
whose existence is guaranteed by 3.11; then, since Iv € %, we have

(1) ov has the finite intersection property.

Qi ss.
(3) If XY DOM, then N = M, and if NX DF and MN has the finite
intersection property, then % € B, consequently if 2 ¥ 9M,
Xt > WM, and this would contradict the maximality of sv. We
see thus that 1% is maximal with respect to the finite intersection
property.
SECReor SPECIAL TYPES OF TOPOLOGICAL SPACES a!“I

k
(4) Let Mi, Mo, ..., Mi € MM, and let My = \ M; # @, since
i=

has the finite intersection property. Now let sv’ = ot U {My},

and let Ni, No,...,N,€ WM’. If N;, 1S 7S pare all elements
of 917, then

p
1\N, #O
i=l

since I has the finite intersection property, and if for some J,

N; = Mp), then

N= (0 Ni)la (A M:) wae)

i=l iAj i=l

where N; € M,7 # j, whence the intersection is nonempty since

MM has the finite intersection property. Also clearly sv’ D MN and
mM >], thus 9 = IM’, and since MN is maximal IN = I’, 1.e.,
My, € MM, whence any finite intersection of elements of 9 is
again an element of Sv. §

We now prove the Tychonoff theorem.

3.13. Theorem. Let X, be a compact space for each a € A, an indexing

set, then X = X Xa, with the product topology, 1s also compact.
aCA

Proof. Let 5 = {F} be any family of subsets of X, such that & has
the finite intersection property. Let 9W be the family whose existence
is guaranteed by 3.12 such that

(1) 9 has the finite intersection property.

Gens.
(3) 9 is maximal with respect to the finite intersection property.
(4) Finite intersections of elements of SW are again in I.

Define It. = Px,M, ie., M. € M. if and only if M, = Px,M for some

M © om. Now the finite intersection property is clearly inherited from
MM to W., and since X, is compact and MN, is a family of subsets of Xa,
we have by 3.6 that
i) wie Oo:
MaMa
78 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

Now for each a select t2 € (\ Mag, and define x = (xz), ie., x is the
MaGMea

point whose ath coordinate is %q.

We now prove thatz € (© WM. Let U bea neighborhood of x, then
Mem
there exists a basic set B such that x € BC U. Now

where S; are sub-basic sets for the product topology, 1.e.,

See x Ve Yo
aGA

with at most one exception, a’, and Yy = O,, an open set of Xv. Now
since x € B C §; for each 7, x. € Px,S; = Ya, and Y, is an open set of
X., whether Y, = Xq or Og. Then rz € M,. for each M, € IM, hence
Vat Vice.
Thus M1) 8; ¥ @ for each and for each M € 9M, and consequently
the adjunction of S; to 9% will not destroy the finite intersection property
of ov. Thus since 2 is maximal with respect to the finite intersection
property, S; € mM, for each 7. Further since finite intersections of
elements of 9% still belong to 9M,

k
B= (\8;€0.
i=1

Let M € ON, then since B € ON, and M has the finite intersec-
tion property, M1 B # @, whence x € M for each M € MM, and
x€ (\ M. Thus
MEST

(NOB RGN AE
FES Mest

since § C M and again by 3.6, X is compact. J

EXAMPLES (AND EXERCISES)

3.4. Any closed bounded set of the real plane with the usual topology
(i.e., that topology induced by the usual topology of the plane)
is compact.
 SHCL-3.2 SPECIAL TYPES OF TOPOLOGICAL SPACES 79

3.5. Show that if X is compact, Y Hausdorff, f : X¥— Y a mapping,

then f(X) is a closed subset of Y.

3.6. Verify that the partial ordering introduced into % in 3.12 is indeed
a partial ordering.

3.7. Let X = X X, be a compact space in the product topology.

aF&A

Prove that X, is compact for each a. [Hint. Use 3.7.]

§2 Separation Axioms

We now take a moment off from considering compact spaces to

consider the extent to which points or closed sets can be separated from
one another by means of open sets. We introduce a sequence of axioms,
called separation axioms, into the definition of a space as follows:

3.14. Definition. Let X be a topological space, then X is said to be a

T; space, provided it satisfies Axiom T;,7 = 0, 1, 2, 3, 4, where the axioms
fy
are as follows:

Axiom Ty: For each x and y € X, x # y, etther there exists U € Uz

such that y € U, or there exists V € U, such that x g V.

Axiom T,: For each x and y € X, x # y, there exist U € Uz and

V € WU, such that y € U anda € V.

Axiom T,: For each x« and y€ X, «x¥ y there exist U € Uz and

V € UU, such that U(1'\ V = @,
Axiom T;: For each x € X and each closed set C € X, x ¢ C, there
exist U € Uz, and O € 0, the class of open sets of X, such
hitCe Cand O11 0 =o.
Axiom T,: For each pair of closed disjoint sets, C, D © X, there exists
a pair O,, O.€ 0, such that CCO:1, DEO, and
(OF (ay O» SS S.

Before we make any remarks about this definition, we prove a simple

theorem.

3.15. Theorem. A space X is a T; space if and only if each point is

ye
closed.
SO SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

Proof. Let x € X, then for each y € X, y ¥ 2, select U € U, such

that « ¢ U. Then X — {x} D U, whence X — {x} € U, for each
y € X — {x}, whence by 1.3, X — {x} is open, and {z} is closed.
Conversely, let « € X, y € X, and x # y, then since {z} is closed,
X — {zx} is open, and since y € X — {x}, X — {x} € U,, and by an
identical argument X — {fy} € U,, whence X isa T; space. §

The language in which the above theorem is couched is open to

criticism, since points cannot really be closed. It is rather the set {zx}
that is closed. We will feel free, however, to say “points are closed”
since this is much more expressive and so much simpler to say, and have
it understood that what we really mean is that sets of the form {x} are
closed sets.
Now let us examine 3.14 more closely. First of all, we observe that
Axiom T) is just the statement that a space is a Hausdorff space, so that
the class of Hausdorff spaces and the class of 72 spaces are identical.
We leave the rest of the discussion to the following:

EXAMPLES (AND EXERCISES)

3.8. Let X be a partially ordered set with the right order topology,
then X is a Ty space, but not necessarily a TJ; space.

3.9. Let X be any infinite set, and define

0 = {0|0 = © or O = F*, where F is finite},

then X with the topology induced by 0 via 1.6 is a T; space, but

not a Ts space.

3.10. Let X have the trivial topology, then X is a T3 space, but not
necessarily a J. space, since a single point need not be a closed
set.

3.11. Let X be a JT; space, x a limit point of A C X, then for each

U € u., UC’ A is an infinite set.

3.12. Let X be a T; space, x a limit point of A C X, F a finite set in X,

then z is a limit point of A — F.

3.13. Let X bea T: space, F a finite set in X, then F has no limit points.
SEC. 3.2 SPECIAL TYPES OF TOPOLOGICAL SPACES 81

As is shown in the preceding exercises, it is not the case that each

axiom 7’; is stronger than the preceding axiom 7’;;, for we have seen
that there exist spaces which are 7’; but not necessarily 7. It would be
pleasant if we had a sequence of successively stronger axioms, such that
each one implied the preceding one. To the end of obtaining such
axioms we make the following:

3.16. Definition. (1) A space which ts at one and the same time a?
T, and a T; space is called a regular space. f
(2) A space which is at one and the same time a T; and a Ts space is
called a normal space.

Before we go one step further, an apology is due for the use of the
word “normal” to describe any property of topological spaces, since
there is probably no more overworked word in the mathematical
lexicon. The word ‘‘normal,”’ however, has been hallowed by traditional
use for spaces which are both 7; and 7, and rather than introduce new
terminology, along with new confusion, let’s stick to the conventional
term.
One more note on terminology: the notion of 7’; spaces is due to
Alexandroff and Hopf and differs somewhat from what we have done
here. The difference is slight and amounts to the following: what we
call a regular space, they call a 7; space, what we call a normal space
they call a 7, space. Some authors follow Alexandroff and Hopf, some
do as we have done here; consequently, the reader should be on the
lookout, when reading other books, for alternative meanings of the terms
“T, space” and ‘7's space.”

3.17. Theorem. Lach of the following properties of topological spaces

as stronger than the next: Normality, Regularity, T, (Hausdorff), T1, To,
in the sense that if a space satisfies the axioms for any one of these properties,
at also satisfies the axioms for all of the following ones as well.

Proof. Left as an exercise. ff

We now give some alternative characterizations for regular and

normal spaces.
 82 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3
/
iN 3.18. Theorem. A 7; space X is regular if and only if for each x € X
and each U € U, there isa V € Uz, such that V C U.
Proof. Let X be regular, whence axiom T’; is satisfied. Let x € X,
U € U,, then by 1.4 there exists O € 0, such that x € OC U. O° is
closed and x ¢ O*, thus there exist 01, O2 € © such that x € Oy, O° € Os,
and 0, (\ O. = @. Let V = O§, then V is closed, andz € O; € O3 = V,
since O, (\ O2 = @. Thus since O, € 0, V € U, by 1.4. Finally since
O° CO, V = OF COC U, and since V is closed V = V CU.
Conversely, let X satisfy the given condition, then since X is T; we
need only prove that X is T3. Let x € X,C C X, C closed, x ¢ C. Then
x € O = C¢, Ois open, hence O € U;z. Select V € U, and by 1.4 we may
assume V open, so that x € VC V CO. Let U = V*, then U is open,
UAV = @, and OC = 0- € V-= UU, dhus there east .U,.V oper
ze V,C C.U,and Uf) V = @,whence Xis 75. §

Nd 3.19. Theorem. A 7; space X is normal if and only if for each closed

set C and each open set U such that C © U, there exists an open set V such
that Cc VAY Ce U.
Proof. Left as an exercise in imitating proofs (viz. 3.18). J

EXAMPLES (AND EXERCISES)

3.14. Let J be the unit square in the plane with the topology described
in Ex. 1.74, then J is a Hausdorff space but not regular, since for

the point x = (3, 0) and the open set U = (Sip(z) A J) ae

there does not exist a neighborhood V of x such that V C U.

3.15. The space given in Ex. 3.14 above is first countable, but not
second countable.

3.16. Let N = {(z, y) |x, yreal, y = 0}. Define

3as follows:If (2, y) € N
= {(z, y) |y > 0}, define
Uev = {U|UDS.(4, y), S.(a, y) an esphere
about (a; 4), y > «>.0):

if (x,y) € Fr (VY) = {(a, y) |y = 0} define

Uew = tU|U DIS.G, 6) U {@, y)} }.
SEC. 3.2 SPECIAL TYPES OF TOPOLOGICAL SPACES 83

Then (N, 3) is a regular space but is not normal.

3.17. Show that a subspace of a 7’; space is a T; space if 7 = 1, 2, 3.

3.18. Prove Theorem 3.17.

3.19. Prove Theorem 3.19.

3.20. Prove that the homeomorphic image of a T; space is a 7’; space

ifi = 1, 2,3, 4.
The properties of being Hausdorff, normal, and compact are inter-
related as is shown in the following:
ive
3.20. Theorem. A compact Hausdorff space, X, is normal (hence also
regular).

Proof. Since a Hausdorff (i.e., T2) space is a T; space, we need only

prove that axiom 7’; is satisfied. We do this in two steps; first we prove
axiom T’; is satisfied. To this end let x € X, CC X, x ¢ C, and C be
closed. C, considered as a subspace of X, is compact by 3.10. For each
y € C, there exists a pair of neighborhoods, which we may assume to
be open, O,..0,,, suck that 0, C\0;, = 0,-2-€ O.,,.y¥ © O,, since X is
Hausdorff. Now U 0, > C; thus since C is compact, there exists a
yEC

finite set {y;|7 = 1, 2,...,m} such that CEU O,,. Let O,,, be the
t=1

open set associated with O,,, 7 = 1, 2,..., m, and define

anon
i=
60° 9.0,,
then O and O’ are open, C C O’, x € O. AlsoO MO’ = Q@, for if

then z € O,, for some 2, and since Oy, \ Ory, = DW,

2 € Oxy, 2 O,

whence z € Oand O(\ O’ = @. Thus axiom 7; is satisfied.

We now prove axiom 1’; is satisfied. Let C, D C X; C, D closed and
disjoint. Since X is 7, there exists for each x € C a pair Uz, U; of open
 84 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

sets such that « € U,, D C Uj, and U,M\ Uz = ©. Since C is a closed

subset of a compact Hausdorff space it is compact, and since C € — Uz

there exists a finite set {7;|7 = 1,2,...,n} such that C © y Ce

Let

{7 abo,
i
Oe 4=1

where U/, is the set associated with U,,. Then as in the first part of the
proof, U and V are open and disjoint and C C U, D CV, whence
axiom 7’, is satisfied. J

The theorem which follows is of a rather special character; it com-

bines conditions on a space and on a mapping sufficient to guarantee a
homeomorphism. It is a theorem of occasional utility.

ny 3.21. Theorem. Let X be a compact space, Y a Hausdorff space, and

let f : X — Y be a one-to-one mapping, then f is a homeomorphism.

Proof. By Ex. 2.5 we need only prove f is a closed mapping. Let

CCX, C closed, then by 3.10 C is compact, whence by 3.7 f(C) is
compact and by 3.9 f(C) is closed, since Y is a Hausdorff space, con-
sequently f is a closed mapping. J

We have observed (Cf. Ex. 3.17) that every subspace of a 7’; space
is again a 7’; space if 7 = 1, 2, or 3. Thus a subspace of a regular space
will also be regular. Unfortunately, however, this will not hold for
normal spaces; that is to say, there do exist normal spaces with subspaces
that are not normal. It is the object of the following example to demon-
strate this. The example is lengthy and somewhat complicated, and
is hence worked out in some detail, although a few of the details are
left for the reader. The object examined in the example is usually
called the Tychonoff plank, since it is “longer” than it is ‘‘wide.”’

EXAMPLES (AND EXERCISES)

3.21. (Tychonoff plank) Let S be any uncountable set, and let “‘<” be
a well ordering for S, ie., if 7 CS, T ¥ @, then there isat € T
such that for all s € T, t < s or t = s. The existence of such a
 ~
a aioe SPECIAL TYPES OF TOPOLOGICAL SPACES 85

well ordering is guaranteed by the axiom of choice, though we do

not prove this here. Let Z+ be the positive integers with the usual
ordering and let Y = S X Z*+ be ordered lexicographically by
P<? i.e.) <b, m) i eithers <7, ori s = i and n <' m in
the usual ordering, <’, of the integers]. We observe that Y is well
ordered by “‘<.” Let 2 € Y be the first element of Y which is
preceded by uncountably many elements under the lexicographical
ordering, let s. be the second element of S, and let w = (s:, 1),
whence w is the first element of Y preceded by an infinite number
of elements. Finally let

A= ayiyvol ged, XM = {ylye

Y,y Sa},
and provide Xo and X, with the interval topology, i.e., if « € X,,
4 = 0, 1, let

Laz = 1] |a < Y <S Zo}

and let
Us {U
Ly ie
UG 27, tor some & <2;

with suitable modification if x happens to be either the first or last

point of X,;, e.g., if a is the last point of Y;, a neighborhood of x
is a set which contains an interval of the type {y|zi<y S x}.

Now we may prove that both Xo and X; are compact and

Hausdorff. We leave it to the reader to verify that Xo and X, are
Hausdorff and that X, is a closed subspace of Xo, and content
ourselves here with showing that Xo is compact. Let {O.} be an
open covering of Xo, then there exists Op € {O.} such that 2 € Op,
and there exists x» € Oo such that the interval J,.@ GC Oo. Now
define ‘‘x is an accessible point” if there exists a finite subcollection
of {O.' which contains in its union J,,, where 2; is the first element
of Xo, and let

A = {x \% S x S 4%, x is not accessible}.

Assume A # @, then since Xo is well ordered, A has a first

element a, and a € O, € {O,} for suitable choice of O,. Further
there exists Ix. GCOu, where b < a < c, and since a is the least
element of A, b ¢ A, and 6 is accessible. There thus exists a finite
86 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

subfamily of {O,' which contains J,,, in its union, consequently

this finite subfamily along with O, contains J,,, in its union, and
this implies that a is accessible and hence a ¢ A. This contra-
diction shows that A = @, whence every 2, with z, S x S po, is
accessible; in particular x» is accessible. Thus the finite subfamily
which contains J,,,, in its union along with the set Oo covers Xo,
and Xo is compact.
We now consider the space Z = Xo X Xy. Since Xo and X, are
compact, so also is Z by 3.13, and since each of the factor spaces
is Hausdorff so also is Z. Thus by 3.20 Z, is normal. It is the
space Z that is called the Tychonoff plank. Consider the subspace
Z' = Z — {(Q, w)}, then Z’ is not normal, for if

Z = {z|z2€ Z, 2 = (2, w), where x is arbitrary}

(i.e., the top edge of the plank) and

Zo = {2|2€ Z,z2 = (Q, x), where z is arbitrary}

(i.e., the right-hand edge of the plank), then Z; and Z», are closed
subsets of Z, being the inverse image of points under projec-
tion mappings. Thus Z} = Z, ( Z’ and Z; = Z.\ Z’ are closed
subsets of Z’ and are furthermore disjoint, since Z,;(\ Z. =
(Q,w) ¢ Z’. Now let Oj and O3 be arbitrary open sets of Z’ such
that O; D Zj and O; D Z3. Consider O04, the set which covers the
right-hand end of the plank; 02 is the union of basic sets, i.e.,
rectangles of the form J,.0 * Ina, where %_ <Q, x1 < 2X. Now
the set of x, where x. is the initial point of the interval J,,9 is
countable, hence has an upper bound x < Q, by Ex. 3.22 below.
Consider the point (x, w) € Zi, then (x, w) € O{ and there exists
a rectangle

al Oe, where 173 < 4% < 2%, Le,

such that (z,w) € R € Oj. It follows (Cf. Ex. 3.23 below) that

Tay Of 22: whence Oj 1)’ 03; ¥ @.

From this it follows that Z’ is not normal.

i) . Prove that any countable set of z. € Xo, with x. < 2 for each a,
has a least upper bound x < Q.
SEC. 3.2 SPECIAL TYPES OF TOPOLOGICAL SPACES 87

3.23. Prove that R (\\ 02 ¥ @ in Ex. 3.21 above.

3.24. In the proof of compactness of Xo in Ex. 3.21 above, why need we

only consider points a € A such that 1 <a < Q?

note: The preceding description of the Tychonoff plank is really much

more complicated than it need be. The reason is that we have tried to
avoid introducing the notion of ordinal numbers. To the student
familiar with ordinal numbers it should be clear that the space Xo above
is simply the set of all ordinal numbers not greater than the first un-
countable ordinal, 2, while X, is the set of all ordinal numbers not
greater than the first infinite ordinal, w.

The Tychonoff plank motivates the definition of a sort of space

whose every subspace is normal, namely

3.22. Definition. A space is said to be completely normal if and only

yur
af every one of its subspaces 1s normal.

We would now like to relate completely normal spaces to some other

properties, but we first require a result, which is usually called the
Lindeléf theorem: ,
it
UrNe p)
3.23. Theorem. Let X be a space with a countable basis, let lw

{Ow | a € A} yh

be an open covering of Y C X, then there exists a countable subcovering

105 SEOs}.

Proof. With each x € Y and each O, associate Bz,, € ® such that

x € Baz G Oa. Define {B.} = {Baz € B|x € Y, a € A}, thensince @
is countable and {B.} C ®, {Ba} is likewise countable, ie., {Ba} =
{B;|i = 1,2,...}, and we reindex the B, in this way using the positive
integers. With each B; associate O; such that B; C O; € {O.}, then
{O; |i = 1, 2,.. .} is the desired countable covering, for clearly {07} is
countable, {Oj} C {O,}, and if « € Y, x € O. for some a, whence
xz € Ba = By = B; for some 7. Thusz € B; GCOjandU OF DY. §
t=1
 88 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

Having proved the Lindeléf theorem we are now in a position to

establish the following:
Le

\Y 3.24. Theorem. Every regular space with a countable basis 1s com-

pletely normal.

Proof. We remark that by Ex. 1.72 and by Ex. 3.17 any subspace of
a regular space with a countable basis is again such a space. Thus all
we need to prove is that any regular space with a countable basis is
normal.
To this end let C and D be closed disjoint subsets of X. For each
x € C, there exists an open set U, such that x € U,C Uz, C Ds ie.,
U,0D = @. Then U U, D Cand by 3.23 we may select a countable
xzEC
subset {U; [7 = 1) 2). 7 GU, are Ce such that

u SC.
Similarly select {V;|7 = 1, 2,...} such that

Vain cr U VOD.

Now define O, = Ui, Wi = Vi, and inductively thereafter define

On = Ua (A v?) and Wa= Val (A Tw’).

Clearly O, and W, are open, whence

Om ll
Os eanda Wee se
3 n=1

are open. Further since C C V;° for each 7, and C C U U; we have

i=1
C CO and similarly D C W. Also we observe that if n = m,

OnWm= Un VA TE) OV (A Tr) Se AVn = B,

t=1

and if nm S m,
SHOmo so SPECIAL TYPES OF TOPOLOGICAL SPACES 89

ORI
ESUL NS =o
Thus for each m, 0, (\ Wn = @ whence On \ W = @, and since this
last holds for each n, we have further that O(\ W = @. We have thus
constructed the required disjoint open sets which contain D and C, and
X isnormal. ff

EXAMPLES (AND EXERCISES)

3.25. Show that every completely normal space is normal.

3.26. Show that every regular space with a countable basis is normal.

3.27. Show that every subspace of a completely normal space is again

completely normal.

§3 Countable Compactness

We observed some time ago that the notion of compactness was

motivated by the Heine-Borel theorem for the real line. Another
property of the reals, the so-called Bolzano-Weierstrass property, asserts,
in one version, that any closed bounded infinite set of reals has a limit
point which belongs to the set. This property is the motivating one for
countable compactness, according to the following:

3.25. Definition. Let X be a topological space, then X 1s countably

compact if and only if every infinite subset of X has a limit point in X.

We immediately remark that the notion of countable compactness

is no stronger than the notion of compactness. This is shown by the
following:

3.26. Theorem. Let X be a compact space, then X 1s countably compact.

Proof. Let A C X, A infinite, and suppose that A has no limit point

in XY. Select from A an infinite sequence {z,} of distinct points, then
by assumption {z,} has no limit point in X. Thus for each n, there
exists an open V, € U,, such that

Va OV{tn}
= Bn.
90 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

Further, for each y ¢ {z,}, since y is not a limit point of {xn}, there
exists an open V, € U, such that V,M {z,} = @. Let

Vom Vi,
yElzn)}
then Vo is open. Now
oO

WO Vip = 44
i=0

yet clearly no finite subcollection of this open covering covers X, since

the omission of any V;, 7 = 1 implies the omission of z;. This, then,
implies that X is not compact, and this contradiction proves that X is
countably compact. §

It is in fact the case that countable compactness is strictly weaker

than compactness, in the sense that there exist countably compact
spaces which are not compact. This is shown in the following examples.

EXAMPLES (AND EXERCISES)

3.28. Let X be the set of positive integers. Let B; = {27, 2c — 1} and

let 8 = {B;|7 = 1, 2,...} be a basis for a topology for X, then
X is countably compact but not compact.

3.29. If X isa space such that every countable open cover of X contains
a finite subcover of X, then X is countably compact. (This result
motivates the name ‘‘countable compactness.’’)

3.30. Show that although the space X of Ex. 3.28 above is countably
compact it does not have the property that every countable open
cover has a finite subcover.

3.31. Show that any closed subspace of a countably compact space is

again countably compact.

There is an unfortunate state of affairs exhibited in the preceding

exercises (Ex. 3.29, 3.30), namely that though every space which has the
property that every countable open cover has a finite subcover is
countably compact, the converse of this statement does not hold. It
may consequently seem strange that we motivate the nomenclature of
SEC. 3.3 SPECIAL TYPES OF TOPOLOGICAL SPACES 91

countably compact spaces as we do (Cf. note after Ex. 3.29 above). Our
reason for doing so is that the simple addition of the hypothesis that X
be T, makes the two properties equivalent, as is shown by the following:

3.27. Theorem. Let X be a T; space, then X is countably compact if

and only if every countable open cover of X contains a finite subcover.

Proof. Exercise 3.29 above proves this in one direction. We must

thus only prove that countable compactness implies the given condition.
To this end, let {0,} be a countable family of open sets which cover X.

Suppose no finite subcollection covers

X, then if C, = X — UY 0;,
i=1
C, # © for any n. Furthermore C,,, as the complement of an open set,
is closed, and also C,, D Ci41. We can thus, for each n, select a point
t, € C,. If the set S ="{z, |\n = 1, 2,.. .} is infinite, then since X is
countably compact, S has a limit point 2 in X. By Ex. 3.12, zo is a
limit point of

Op Se hha on) SC,

Since C;, is closed, x» € Cy for each k, whence x ¢ O, for any k. This

contradicts the fact that {O,} covers X. It thus follows that the set
S = {z,|n = 1,2,...} is finite. Thus there exist 7’ and N > 0 such
that for all n 2 N, z, = 2’, and as before-a’ € C, for n > N, and
because the C,, are a decreasing sequence of sets (i.e., Cn D Cn41), 2’ € Cr
for all n. Once again this means that x’ ¢ O, for all n, and {O,} is not
a covering. This contradiction establishes the theorem. J

We show now that under suitable conditions it is in fact the case

that countable compactness implies compactness.

3.28. Theorem. A second countable T; space is countably compact if

and only if it 1s compact.

Proof. Compactness implies countable compactness by 3.26. Con-

versely if X is countably compact, and {O.} is an open covering of X,
then by 3.23 there is a countable subfamily {0,} of {O.} that also
covers X. Then by 3.27 {0;} contains a finite subcover, whence X is
compact. §
92 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

EXAMPLES (AND EXERCISES)

3.32. Show that a 7, space is countably compact if and only if every

countable family of closed sets that has the finite intersection
property has a nonvacuous intersection.

3.33. A T; space is countably compact if and only if every infinite open

cover (where infinite refers to the indexing set) has a proper
subcover. [Hint. Imitate 3.27.|

$4. Local Compactness

3.29. Definition. A space X is said to be locally compact provided that

for each x € X there exists U € Uz such that U is compact.

EXAMPLES (AND EXERCISES)

3.34. The real line with the usual topology is locally compact but not
compact.

3.35. The real plane with the usual topology is locally compact but not
compact.

3.36. Every compact space is locally compact.

3.37. Every closed subspace of a locally compact space is again locally

compact.

The motivation for introducing the notion of local compactness is

twofold. First of all, as has already been pointed out (Cf. Ex. 3.34, 3.35),
the real line and the real plane are locally compact, as is, in fact,

Er = x Ris,
t=1

where each FR; is just the reals with the usual topology. Since such
spaces are of general interest to the analyst, the notion of local compact-
ness is of some concern in analysis. Second of all, there is a very pretty
topological analog of the process of stereographic projection. To refresh
the memories of those of us who may have forgotten, stereographic
projection consists of the following:
SEC. 3.4 SPECIAL TYPES OF TOPOLOGICAL SPACES 93

Consider the real plane with coordinate x and y axes. Construct a

sphere of radius 1/2 tangent to the plane at (0,0), and let the real
three-dimensional space with coordinate axes be so located that the
origin is at the origin of the real plane, the 2’ and y’ axes of the real
three-dimensional space coincide with the x and y axes of the real plane
respectively, and the positive z’-axis passes through the north pole of
the sphere, the south pole being at (0, 0). Now to each point P = (z, y)
of the real plane we can make correspond a point Q = (a’, y’, 2’) of the
sphere S, whose equation is given by x” + y” + 2” = 2’, in the following
fashion: Draw a straight line from N = (0,0, 1), the north pole, to
P = (x, y, 0). This line meets the sphere in some point; let this point
be Q. A simple calculation shows that

where 72 = 2? + y?.
It is not difficult to see that the mapping f: R—S — {(0, 0, 1)}
that carries the real plane into the sphere whose north pole has been
removed, which is defined by

{GoPipe ee eee
(an y le

where r? = x? + y?, is one-to-one and onto, for another calculation shows

that

ee) Ga
ie eZ

and further from this it is seen that both f and f~ are continuous.
Thus f is a homeomorphism.
Now what happens is this: We have embedded the locally compact
real plane in the compact sphere S in a homeomorphic fashion so that
S — f(R) isasingle point N = (0, 0, 1). In some sense, then, it is rather
natural to ask, ‘‘If X is a locally compact space, which is not compact, is
it possible to find a compact space X’ and a homeomorphism f : X — X’
such that X’ — f(X) is a single point?” The answer to this question is,
in fact, “Yes,” and we shall set out in the next few theorems to prove
exactly this result. All that this shows is that stereographic projection
is really only one instance of a rather general phenomenon.
94 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

It might be pointed out here that the utility of this result lies in the
fact that we can view a locally compact space as a subspace of a compact
space X’ and, in general, compact spaces are more pleasant to deal with
than noncompact spaces.

3.30. Definition. Let X be a noncompact T, space, let K be the family

of closed compact subsets of X, and let « be an ideal element such that
ug X. Define a space *X = X U {i} with basis

@= {U|U€ 0,o0r U =X — K, K€ x},

then °X 7s called the one-point compactification of X.

The insistence that X be a JT; space is motivated by the desire to

have K # @. For if K = @, . would have no neighborhood system.
The most important properties of the one-point compactification, °X,
are exhibited in the following theorems.

3.31. Theorem. Let X be a noncompact T, space, then °X is a T; space.

Proof. Clearly U B= °X. Leta € °X, and let U, V € @ such that

BEB
x€ Vanda € U. Ifa € X, U and V are open and consequently there
exists W = U ()\ V open in X, hence again in ® such that

£EWLG UNV.
If « = 1, then

U=°X — K,, V=°X — K, Ky, Ke € &.

Let

W=UQV = (X — Ki) (X — BK) = X — (Ki U EK),

and K, U Ky is again closed and compact, whence W € ®, anda € WC

U(\ V. Consequently, by 1.29, «X is a topological space.
We show now that °X is T;. Let 2, y € *X,2 ¥ y. Ifa, y € X then
since X is 7), there exists U € Uz (or U,) such that y ¢ U (or a ¢ U).
On the other hand, if one of x and y (say 2) isu, then since {y} is a closed
compact set in X, U = *X — {y} € U,and y ¢ U. In either case ¢X is
aT, space. J
sec. 3.4 SPECIAL TYPES OF TOPOLOGICAL SPACES 95

3.32. Theorem. Let X be a noncompact T, space, then °X is compact.

Proof. Let {O.} be an open covering of “X, then some one of the
covering sets contains c , let us call it Oo. There exists then a basic set
BC QO), and B = «X — K, K closed and compact in X¥. Now

Re,0)
aA

thus there exists a finite subcollection {0;|7 = 1, 2,...,n} such that

‘0.
I
:. ran

Since -X — K C Oy, we have

VE UO 0i,
i=0

and we have found a finite subcovering of the covering {O,}, whence «X

is compact. ff

3.33. Theorem. Let X be a noncompact T, space, *X its one-point

compactification, then X 1s locally compact and Hausdorff if and only if
‘X «s Hausdorff.

Proof. Let ‘X be Hausdorff, then X as a subspace of *X (Cf. Ex. 3.39

below) is again Hausdorff. Let « € X, then there exist open sets U
and V of *X. such that ¢ U,.€ V, UNV = 2, VD°X — K for
some K € K. Now

CX=—KAUGVOAU=o
so that U C K. Since K is closed and compact, U C K, and U is
compact. Clearly U € Uz, whence X is locally compact.
Conversely let X be locally compact and Hausdorff. Let x, y € °X,
x xy. If x, y € X, there exist U, V open in X, hence also open in °X
such that z € U,y€ V, U(\ V = ©. If one of x or y = 1, say = 4,
then since X is locally compact we may select U € U, such that U is
compact, hence also closed. Then

Voie and CK UT = oy.

Thus in any case °*X is a Hausdorff space. §
96 SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 3

EXAMPLES (AND EXERCISES)

3.38. (a) Show that the mapping f : X — °X defined by f(z) =z isa

homeomorphism of XY with f(X).
(b) Infer from (a) above that every locally compact Hausdorff
space is homeomorphic to a subspace of a compact Hausdorff
space.

3.39. It is clear from 3.30 that X is a subset of “X. Show further that
it is also a subspace.

3.40. Show that every locally compact Hausdorff space is regular.

(Hint. Use 3.33, 3.20, Ex. 3.17, and Ex. 3.39 above.)

3.41, Let X = {n|n > 0, n an integer}, define

Sae= {m||I/n — 1/mise

and: let’ @ = {S;,.| m7 = 1/2, ..5,¢> 0}. Show that with Gas
basis, X is a topological space, X is Hausdorff, and that *X is
obtained by adjoining the point « = © and defining the basis
elements at © by S,,. = {m||1/m| < ¢}.

3.42. (Term paper) Paracompactness.

Definition. A family {F.} of subsets of a topological space X is
locally finite if and only if for each x € X there exists U € Uz such
that U(\ F. ¥ © for only finitely many indices a.
Definition. Let {U.} be a covering of a topological space, then a
covering {Vg} ts said to be a refinement of {U.\ if for each B there
exists an a such that Vg € Ug.
Definition. A topological space is said to be paracompact if it is
a Hausdorff space, and if each open cover {O.} has a locally finite
refinement.
Prove at least the following:
(1) Every compact Hausdorff space is paracompact, but not
conversely.
(2) Paracompactness is a topological property.
(3) Every paracompact space is regular.
(4) Every paracompact space is normal.
sec. 3.4 SPECIAL TYPES OF TOPOLOGICAL SPACES 97

(5) Every closed subspace of a paracompact space is para-

compact.
(6) The topological product of a paracompact space and a
compact space is paracompact.
(7) Show by counterexample that the topological product of
two paracompact spaces need not be paracompact.

NoTE: A result owing to A. H. Stone assures us that every metric

space (Cf. Chapter 5) is paracompact. This result appears in Bull.
Amer. Math. Soc. (1948), pp. 977-982.
 CHAPTER 4

FURTHER SPECIAL TYPES OF

TOPOLOGICAL SPACES
(MOSTLY VARIETIES OF CONNECTEDNESS)

§1 Introduction

In this chapter we continue to explore various restrictions we can

place on the topology of a space. The major notion we shall explore is
that of connectedness. Intuitively, a connected space is one which
consists of one piece. The problem then is to write a formal definition
of the intuitive notion of ‘‘one-pieceness.’’ Let us first consider a few
examples. Certainly we would think of R, the real line, and #, the real
plane, as being of one piece. Suppose we consider R — {0}; surely this
is not of one piece, since without doing any serious violence to the
topological (i.e., geometric) structure we can separate R — {0} into
the two pieces

B= {leo
Roe > 0} sand ho = 42) ae Roe <0

Let us consider one further example, namely

X= {(e.)|2,y€ Ry = sin} for0 <a

ale

orz =0,and -Il sys i}.

Certainly the infinite are A defined by

Va for Wee
x T

is all of one piece, and certainly the line segment

98
SEC. 4.2 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 99

L={0,y|-lsys}
is of one piece. Do we do violence to the topological structure of X by
separating these two pieces, A and L? It would certainly appear so,
since each point of L is a limit point of points of A, e.g., the point (0, 4
is a limit point of the set

1
Ain = {(@,) |# = y=sinon = 1,2... 604.
1/6 + 2nx’

It would thus appear that we don’t want a connected space to be such

that we can write it as the union of two sets, neither of which has a limit
point of the other. With this in mind we make the following two
definitions.

4.1. Definition. Let X be a topological space, A and B nonempty sub-

sets of X, then A and B are said to be separated if and only if A(\ B = @
and A1\B = @.

4.2. Definition. Let X be a topological space, then X is said to be

connected provided that X cannot be written as the union of two nonempty
separated sets. A set A C X is said to be connected if A is a connected
subspace of X.

§2 Connected Spaces

There are a number of other ways in which a connected space may

be defined. Several of these are given in the following:

4.3. Theorem. A space X is connected if and only if any one of the

following conditions is met: Ye

Vv (1) X is not the disjoint union of two nonempty open sets.

/ (2) X is not the disjoint union of two nonempty closed sets.
(3) The only sets in X which are both open and closed are @ and X.
(4) For every continuous f: X > R, f(X) does not consist of two
distinct real numbers.

Proof. We leave the proofs of (1), (2), and (3) as exercises. We

 100 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

prove (4) contrapositively. Let f be a continuous function of X into R

such that f(X)= {a} U {b},a 4b. Let « = 3|a — 5] and define

=(a—¢at+e), V=(6-—¢6+
68),

then U (\ V = @, and U and V are open in R. Thus f-!(U) and f7(V)

are open in X since f is continuous, and

oe CL) aea
and thus by part (1), X is not connected.
Conversely, suppose X is not connected, then there exist open sets
U, V, such that X = UUV, U(\\V =, by part (1): Define
f{: XR by f(z) = 0 if x € U,7 f(z) = 1 if « € V, thenf is clearly
continuous. J

Connectedness is a topological property, in fact even more is true,

namely:

V 4.4. Theorem. Let X be connected, let f : X — Y be continuous and

onto, then Y is connected.

Proof. We proceed contrapositively, by proving that if Y is not

connected, then X is not connected. If Y is not connected, we can by
4.3(1) write Y = A U B, where A (\ B = @ and A and B are open and
nonempty in Y. Then

Xe Jaa se 8)
where, since f is continuous, f(A) and f-(B) are open in X, and
f(A) AO f7(B) = ©. Thus X is not connected. J

EXAMPLES (AND EXERCISES)

4.1. Let C C R, the reals, and let C be connected. Let a, b € C, and

leta <c <b, thene € C. [Hint. Supposec ¢ C, define f : CR
by f(x) = a if x < c and f(x) = b if x > c, then prove f is con-
tinuous and use 4.3(4).|

4.2. The nonempty connected subsets of R are of the following forms:

SEC. 4.2 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 101

(a) Intervals or points, i.e., (a,b), [a, 6), (a, b], a < b, or [a, b],
@ sb:

(b) Rays, ie., (a, ©), [a, ©), (—%, a), or (—~, a].
(c) FR itself.
4.3. Let X be a space, then X is connected if and only if for each
f:X—R, f continuous, f has the intermediate value property,
ie., if f(vi) = a, f(x2) = b, and c is between a and b, then there
exists x € X such that f(x) = c.
4.4. Prove Theorem 4.3(1), (2), (3).

4.5. Theorem. Let X be a space, X = X, U Xo, where X; and X2 are

separated. Let C C X be connected, then either C C X, or C € Xz.

Proof. Define CQ, = CC) X1, Co = CO) Xo, then C = C; \ C2. Then
C, € X, and C2 € Xo, whence

Ciegy CS Wan Ne = sand. Oi) CS Ny \ Xe =

Thus if both C; ¥ @ and C,; ¥ @, C is not connected, consequently
either C; = @ in which case C C X.or Cy = GW inwhichcaseC CX. J

4.6. Definition. Let ® = {D,} be a family of sets in some set X. A

finite subset
D, = {D;|i=1,2,...,n}
CD
is a chain provided that D; (1\ Din 4 O, 274 = 1, 2,...,n—1. Sucha
chain is said to be simple if D; (\ D; = @ unlessj7 = 71+ 1. The family
of subsets @ = {Ag} of X is said to be chained by D 7f for each Ag, Ay € @,
there 1s a chain Dn CD, such that Ag (\ D, ¥ O, Ay (\ Dn ¥ OD. Bis
said to be simply chained by D cf for each Ag, A, € @ there is a simple
chain Dn € D such that Ag (\ Di ¥ O, Ay \ Dn ¥ S.

4.7. Theorem. Let X be a connected topological space, {Oa} an open

covering of X, then {Oa} ts chained by {0}.

Proof. We proceed contrapositively. Let {O.} be an open covering

of X which is not chained by {0,}. Let O: € {O.} and let

©: = {0|0 € {0,}, such that {O, O,} is chained by {0.}}.

102 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

0: X @ since O; € O;. Let O: = {O.} — O1, then ©: is not empty, since

{0.} is not chained by {0,}. Let
aA 50; v= Oe
OC: OE:
then U(\ V = @, U and V are each nonempty, whence X = UU V
is not connected. J

4.8. Theorem. Let X be a topological space, then X is connected if and

only if X is simply chained by any open covering of X, 2.e., any two points
of X can be joined by a finite set of open sets of the given covering.

Proof. Let X be connected, {O,} an open covering of X, and let

xz € X. Let C, = {y| {z, y} is simply chained by {0,}}. C. ¥ @, since
xz € C,. We show that C, is both open and closed. Let y € Cz, then
there exists a simple chain

0, = LOS temee Lr ce, the

such that z € O1, y € O,. Now for any z € O,, On is a simple chain
joining z to z, thus z € C,, and O, C C,, whence C, is open, since O, is
a neighborhood of y.
Now let y be a limit point of C., then y € O, for some a and
Oa. (\ Cz # O. Thus let z € On (1) Cz, and let

O;.= {074 = 152. . <2, a

be a simple chain from z to z. Further let k be the smallest index

1<k <n such that 0.110; ¥ @. Such a & surely exists since
On (\ O. ¥ GO, then

Orn = {0;|¢= 1,2,...,

k, a}

is a simple chain from z to y, and y € C,. Thus C, is closed. The proof

is completed by invoking 4.3(3).
Conversely, suppose X is not connected, then

X=AUB, A¥Z, B¥Z, ANB=G,

A and B open. Let a € A, b € B, then {A, B} is an open covering and
there is no simple chain (in fact no chain of any kind) from a to b
SEC. 4.2 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 103

consisting of sets of the open covering {A, B}. This is a contradiction,

hence X is connected. J

4.9. Theorem. Let X be a space, let N be a connected set in X, and let

N CA CN, then A ts connected, 1.e., under the adjunction of any number
of limit points a connected set remains connected.

Proof. Suppose A is not connected, then

A = A, UA), A, # @, Ar, ¥ @,

A, and A: separated. Since N is connected, and N C A, either N C A;

or N C A» by 4.5. We assume N C A). Since N is connected and A is
not, A — N ¥ @, and further A, ¥ @, thus we may select x € A — N
such that € A» Then since A C N,

H03 € N os Ai,

whence A; () Ay ¥ @, and this contradicts the fact that A, and A: are

separated. This contradiction establishes that A is connected. Jf

EXAMPLES (AND EXERCISES)

4.5. A space X is said to be dense in itself if each x € X is a limit point

of X — {x}. Show that every connected T; space which consists
of more than one point is dense in itself.

4.6. Show that in Ex. 4.5 above X must be a 7 space for the conclusion
to hold by considering

X = {a,b} — where 0 = {{a, B}, {b}, D}.

4.7. Show that in a connected 7, space every connected set which
consists of more than one point is infinite.

4.8. Prove that if X is a connected space,

a finite covering of X by closed sets C;, then € is chained by @.

104 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

4.9. Let X be a space, M, N C X, M, N connected, then if either

a) MON=o or (b) MOAN#D

then M U N is connected.

4.10. If X is a space, NC MCX, M and N connected, and if

M —N=AUB, where A and B are separated, then N U A is
connected.

4.11. Suppose we had proved the following in place of 4.9:

Theorem 4.9(a). If N is a connected set in X, then so also is N.

Prove 4.9 on the basis of 4.9(a) by considering A with the relative

topology.

. Let {C,} be a family of connected sets in some space indexed by

a€ A. If

\ Ca # OD,
acA

then U C, is connected.
alA

4.13. Let {A;|7 = 1,2,...,n} be a finite family of connected sets

such that

Ail) Ain # D, = 1 2.24 — AP

Then U 4A; is connected.

t=1

4.14. Let A C X be a connected set, and let B C X. Suppose BM A

~ O, and Be(\ A # O. Show that A (Fr (B) ¥ @, where
Fr (B) is the frontier of B (Cf. 1.16). Infer that in a connected
space, if Bb# @, B # X, then Fr (B) ¥ @.

4.15. let A and B be closed subsets of the space X. If A U B and

A (\ B are connected, so also are A and B. Show by counter-
example that both A and B must be closed, i.e., if one of A and
B is not closed then in general the theorem fails.
SEC. 4.3 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 105

§3 Components

In some topological spaces it is occasionally interesting to examine

the maximal connected subsets. We can do this via certain subsets of
the space called components, as defined in the following:

4.10. Definition. Let X be a space, M C X and x € M, then C,, the

component of x in M, is the union of all connected subsets of M to which x
belongs, or more formally

C, = {y| x and y lie in a connected subset of M}.

4.11. Theorem. Let X be a space, C, a component of M in X, then

(1) C, ts connected.
(2) C, 1s maximal with respect to the properties:
(a) x € C,,
(b) C, is a connected subset of M, 7.c.,if AC M,x € A and A ts
connected, then A € C,.
Proof. (1) Let y € C;, define A, to be a connected subset of M such
that z, y € A,. Now U A, is connected by Ex. 4.12, since x € A, for
yECz
any y € C,. Also clearly

CAN xAT.
yECz

On the other hand if 2€ WU A,, then z € A, for some y and since

yECz

x € A,, and A, is connected, x and z belong to some connected subset

of M, whence z € C,. Hence

A SCs whence C, = U Ay,

yECz y EC,

and C, is connected.
(2) Let A C M, x € A, and let A be connected. For any z € A,
xz and z lie in a connected subset of M, namely A, hence z € C,, and
ACC. J
Until Theorem 4.11 was established we had a rather cumbersome
mode of talking about components of a set. We had to refer to the
component of z in M. Now that we know that components are merely
106 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

maximal connected subsets of M, we no longer have to refer to “the

component of z in M” but can simply talk about components of M. In
this connection see also Ex. 4.18.

4.12. Theorem. Let X be a space, M a closed subset of X, x € M, then

C,, the component of x in M, 1s also closed in X.

Proof. Since C, € C;, Cz is connected by 4.9. Since M is closed,

C, CM implies C, C M = M. Finally by 4.11(2) C, = Cz, and C, is
closed. J

EXAMPLES (AND EXERCISES)

4.16. If X is a topological space, M C X is open, then there is little

one can say about the components of M beyond what has already
been said about components in general. In particular we can
conclude nothing about components being either open or closed
or neither. Consider the space

x= {G,y)|0S251,y=72,0=1,2,...,0ry =o}

with the relative topology inherited from the real plane. Let
M = X — {(0,0)}, and let z = (1,0), then M is open and the
component of z in M,

C.= {@,y|0<2S1,y
=O}.
Is C, closed in X? Surely not,
since (0,0) ¢ C,. Is C, open in
X? Surely not, since any neigh-
borhood of z contains points
of the form (1,1/n) for n
sufficiently large. For a picture
of X see Fig. 4.1.

4.17. Show that if X is a_ space,

Figure 4.1
xe M CX and ih ayesce
then C, = C,.

4.18. Let X be a space, M C X,2, y € M. Define x ~ y if and only if

there exists a connected subset NV of M such that z, y € N. Show
sec. 4.4 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 107

that ‘““~”’ is an equivalence relation in M, and show further that

the equivalence classes induced by ““~” in M are exactly the
components of M.

$4 Local Connectedness

We would like to introduce a property of spaces which would allow

us to draw some conclusions about components of open sets. By analogy
to closed sets, it would certainly be pleasant if in some spaces, at least,
components of open sets were open. Let us proceed boldly, and merely
give spaces that have this desired property a new name.

4.13. Definition. A space X is said to be locally connected if and only

uf components of open sets are open.

The name ‘“‘locally connected” for such spaces may seem ill chosen;
however, as will shortly appear, it is quite in harmony with our definition
of another local property, namely, local compactness. The next theorem
exhibits exactly the analogy with Definition 3.29 that we want.

4.14. Theorem. Let X be a space, then X is locally connected if and

only if X has a basis of connected sets, t.e., for each x € X and each
U € U, there exists V, open, connected, V € Uz such that V C U.
Proof. Let X be locally connected. Define

® = {C|C CX, C connected and open}.

We show that @ is a basis. Clearly ® C 0, further let x € X, U € Uz,

then there is an O € © such that rx € OCU. Let V = C,, the com-
ponent of x in O, then V is open by 4.13 and V is connected by 4.11(1),
whence V € @. Alsoz € VCOCU, thus @ is a basis.
Conversely suppose & is a basis of connected sets for X. Let M C X,
M open, let x € M, and let C, be the component of x in M. Let
y € C, C M, then there isa U € ® such that y €C UC M. Sincey € U
and U is connected, U € C,, the component of y in M, by 4.11; but by
Ex. 4.17, C, = C,, thus for each y € C,, there exists U € U,(\ B such
that y € U C C,, whence C, is open. J

The student should be warned that Definition 4.13 is rather atypical

in that, should he look in other texts on topology, he will not find locally
108 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

connected spaces so defined. Rather he will find them defined as

Theorem 4.14 suggests, ie., so that each neighborhood of each point
contains a connected subneighborhood. Once we have proved 4.14,
however, it makes no difference how we define “locally connected,” and
certainly our definition has more intuitive appeal.

EXAMPLES (AND EXERCISES)

4.19. Show that a space X is locally connected if and only if for each
U € Uz, the component C, of x in U also belongs to Uz.

4.20. Let

x={wy)|2=5,05y 81,forn=1,2,...}

U {(0,0), ©, 1)}
have the relative topology inherited from the usual topology of
the plane. Show that {(0, 0)} and {(0, 1)} are components in X.
Show further that if D C X is both open and closed, then either
both {(0, 0)} and {(0, 1)} lie in D or neither of them does.

4.21. Let X be a space which is locally connected and connected.

(a) Show that if C is a component of an open set A C X, A ¥ X,

then Fr (C).& A (Ch 116.)
(b) Let M and N be disjoint nonempty closed sets in XY. Show that
there is a component C of (MUN) such that C0 M # @
and C (\N <2.
(c) Let B be a closed set in X, B ¥ X, Ca component of B, show
that GOO Be ao:

4.22. Let f : X — Y be a closed mapping onto and let X be locally

connected. Show that Y is locally connected.

4.23. Show that any open subset of a locally connected space is again
locally connected.

4,24. (Term paper, junior grade) For each a € A, let X. be a Hausdorff

space.
sec. 4:5 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 109

(a) Show that X X, is connected if and only if each X, is

aA

connected.
(b) Show that X XX, is locally connected if and only if each X,
actA

is locally connected and each X, is connected with at most a

finite number of exceptions.

[Hint for (a). Prove first that if f : X — Y, Y a Hausdorff space,

is continuous, D C X is dense in X, and f(D) = yo € Y is a fixed
point of Y, then f(X) = yo. Assume X X, = X is not connected,
alA

then use 4.3 to define f: X — RFR so that f(X) = {a, b}, a # },

a, b € R. Use the homeomorphism obtained in Ex. 2.22 to show
that f is constant (say, f(v) = a) on sets of the form X¢% defined
in Exe 2,22. Let

Xe eat, SE Ora Ott lek. ace, Or bitrary}

as in Ex. 2.22 and show that for x € Xe... ,, f(x) = a. Then

show that sets of the form Ng,,...s, are dense in X, and finally
that f(b) = @ to obtain a contradiction. |
[Hint for (b). Use (a).]

$5 Arcwise Connectedness

We attack finally one more type of connectedness which is of

considerable utility in other branches of topology. The motivation
here is that some of the spaces that are quite familiar have the property
that points may be joined by means of arcs—for example, the real line

or the real plane or for that matter any Euclidean space H”" = x Ri,
i=1
where each RF; is the reals with the usual topology. We generalize this
notion via the following:

4.15. Definition. (1) Let X be a topological space, A € X, and let

={¢:\0 =2 S11; x real} with the relative topology inherited from the
usual topology of the reals. A is said to be an arc provided there exists a
homeomorphism h of I with A. We say that the arc A is from the point
xo = h(0) to the point x; = h(1), or simply from x to x1.
110 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

(2) Let X be a topological space, then X is said to be arcwise connected

if for each x, y € X there is an arc from x to y.

We shall not explore arcwise connected spaces to any great extent,

but shall content ourselves with the following two theorems.

4.16. Theorem. Let X be an arcwise connected space, then X 1s

connected.

Proof. Let « € X and for each y € X, let I, = h(Z) be an arc from

x toy, thenz € () I,, and this intersection is nonempty. Furthermore
yEXx
by 4.4 each I,, is connected, consequently by Ex. 4.12, U I., = X is
yEXx
connected. §

4.17. Theorem. Arcwise connectedness is a topological property.

Proof. Let f : X — Y be a homeomorphism and let yi, y2 € Y, then
there exist 21, 22 € X such that f(z1) = y1, f(z) = ye. Since X is arcwise
connected there exists a homeomorphism h : I — X such that h(O) = 1
and h(1) = 2. fh: IT— Y is a homeomorphism of J into Y such that
fh(O) = y: and fh(1) = yo, whence Y is arewise connected. J

EXAMPLES (AND EXERCISES)

4.25. The space

Sopa ki 1
xX ={@vly = sin 7 0 <7 situ {(0, 0)}

is not arcwise connected, but is connected in the relative topology

inherited from the plane.

4.26. Not every arcwise connected space is locally connected; consider

the space of Ex. 4.16.

4,27. We now use some of the results obtained so far in this chapter to
prove a fixed-point theorem. Prove that if f: I—TJ is a map-
ping where J = {x |0 S$ x S$ 1, z real}, then there is an x € J
such that f(z) = x. [Hint. If f(0) = Oorf(1) = 1 we are finished;
sEc. 4.5 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 111

if neither of these happens, f(0) > 0 and f(1) <1. n I XJ,

which is an arewise connected space (why?), let

Cea) ibe > a 2 0},

let x = (0, f(0)) and a = (1, f(1)), then

Xo € Ur U1 € Ue

and f : I > TI defines a path, i.e., a continuous image of [0, 1],

from 2 toa. Let A = fYJ), then AOU # O,andAM\ Us # @,
thus by Ex. 4.14

Af\ Fr (U) # ©.
Show that

Fr (U) a L(x, y) |# = y, 0 SYS i

and that if (vz, y) € AC Fr (U), then f(z) = z.]

As a closing topic in this chapter we investigate the other side of the

coin and look at spaces which are as badly disconnected as possible. To
do this we look at the components of the space, and if the space is to be
as badly disconnected as possible, we shall want the components to be as
small as possible, namely points. We thus make the following:

4.18. Definition. A space X is said to be totally disconnected if and

only if all of its components are points.

There are, of course, many examples of such spaces. Any space with
the discrete topology is totally disconnected; thus, for example, the
rationals as a subspace of the reals are totally disconnected. Any finite
T, space is totally disconnected. We have in mind, however, looking
at a more interesting example.

EXAMPLES (AND EXERCISES)

Let
Tae 0ts 7S 1, & real),
ta Wor <2, e realy, OC) = 1 — ty,
 SPACES CHAP. 4
112 FURTHER SPECIAL TYPES OF TOPOLOGICAL

Ih = {|1/9 <2 < 2/9, i= {x| 4/9 <x < 8/95,

Th = {a |1/8" <2 < 2/3}, Ih = {| 7/3" <2 < 8/3%,

eine
i= 2) = 2/sees = 6 — la
C= Ca (u ia) and finally
i=1

Care
n=1

What we are doing is to remove from the unit interval the

open middle third, and then at each successive step to remove
the open middle third from each interval that is left after the pre-
ceding middle thirds have been removed. See Fig. 4.2.

1 2

1 2 { 2
Oe eae dee a Ss7 8

as —— ee Rod ki
Gite ee ae 2 19 20 7 8 25 26 4
2%. 27.9 9 2% 2753 eh rePare ©) 9 27 2%,

Figure 4.2

There is a great deal known about C, called the Cantor set, or

sometimes the Cantor middle third, or ternary set. We mention
a few facts here, which we leave to the student to verify.

4.28. If it is agreed to write all numbers z, 0 < x < 1 in the scale 3

(i.e., 0, 1, 2 are the only digits which occur) and if all infinite
repeating decimals in which eventually an infinite string of zeros
preceded by a one occurs are replaced by a decimal in which the
one is replaced by a zero in the place before the infinite string of
zeros, and this place followed by an infinite string of two’s, e.g.,
001000 . . . is replaced by .000222... and if similarly .1222...
SEC. 4.5 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES 113

is replaced by .2000, then the points of C are exactly those deci-

mals written in the scale of 3 in which no one ever occurs.

4.29. From Ex. 4.28 above it is clear that there is a one-to-one corre-
spondence between the class of all functions on

Z* = {n|na positive integer}

to a set consisting of two objects (namely 0 and 2) and C. The

class of such functions, written 22° or 2% is known to be un-
countable, thus C is uncountable.

. Cis closed in R.
)
. Cis dense in itself, i.e., for each x € C, x isa limit point of C — {x}.

. Cis nowhere dense in J, i.e., if xz, y € C, x < y, then there is an

Interval| Sal — Ch = (orb <a <b < yy.

3. C is totally disconnected. [Hint. Use Ex. 4.32 above.|

. Cis compact. [Hint. Use the fact that J is compact, and Ex. 4.30
above.|

It is, in fact, the case that any compact, totally disconnected,

dense in itself, metric (to be defined in the next chapter) space
is homeomorphic to the Cantor set; however, we make no at-
tempt to prove this here. The interested student is referred to
more advanced works on topology for a proof of this, e.g., Hocking
and Young, Topology, p. 100. The reason these last remarks are
made about the Cantor set is that the stated properties are a
complete set of invariants for the Cantor set; that is to say, if we
have a topological space which possesses the given properties,
namely is compact, totally disconnected, dense in itself, and
metric, then it is topologically indistinguishable from the Cantor
set. Of course, it may be distinguishable from the Cantor set
in some other, nontopological way, say in that its points are
blue, or are named Sam; but as topologists, this is of no concern
to us.

4.35. A point x in a connected topological space X is called a cut point

of X, if X — {x} is not connected. Prove the following:
il4 FURTHER SPECIAL TYPES OF TOPOLOGICAL SPACES CHAP. 4

(a) If X is homeomorphic to Y under f and z is a cut point of X,

then f(x) is a cut point of Y.
(b) Every point of the real line is a cut point.
(c) No point of the real plane is a cut point.
(d) The real line and the real plane are not homeomorphice.
 CHAPTER 5

METRIC SPACES

§1 Definitions

We apply ourselves now to a rather special sort of topological space,

one in which there is defined a distance function, so that we can
say what the distance between points is. In a sense these spaces,
so-called metric spaces, are rather special, since, as will turn out in the
sequel, they will enjoy properties of the sorts we have already discussed,
but under less restrictive hypotheses than more general spaces. On the
other hand, metric spaces are still quite general, since all the common
spaces of analysis are metric spaces. We define a metric space in two
stages; first we make the following:

5.1. Definition. Let X be a set, let pp: X X X > R be a function (not

necessarily continuous) of X * X into R, the reals, such that
DB
(1) plz, y) 2 0 and p(z, y) = 0 af and only ifx = y, Hi
(2) p(x, y) = ply, 2), ae
(3) p(x, y) + ely, 2) 2 pla, 2), Z —>
then (X, p) is called a metric set, and p is called a metric for X.

Unfortunately 5.1 is so general that every set is a metric set, for

define p(x, y) = 1 if x ¥ y, and p(z, y) = 0 if x = y. It is a simple
verification that with this definition of p, any set X is a metric set.
What we are really after is to have the metric generate the topology
for X. To that end we make the following:

5.2. Definition. Let (X, p) be a metric set, define for x € X ye

S.(z) = {y| p(z, y) < efor y € X, «> 0 and real}

115
116 METRIC SPACES CHAP. 5

and define

@ = {S.z)
|ae xX, eS 0}.

Then X with & as basis is called a metric space. The topology so generated
is called a metric topology generated by p.

Now only the rankest of amateurs at mathematics tries to prove

definitions; however, the above definition makes some assertions which
must be verified. In particular, it is asserted that @ is a basis, and this
is perhaps not so evident without some proof. We deal with this minor
matter in the following:

5.3. Theorem. The set ®8 = {S.(x) | x € X, € > 0} is a basis for some

topology for the metric set (X, p).

Proof. We shall apply 1.29. Since x € S.(x) for each x € X clearly

have
LA Se) = OX,
Sez) EB

Now let x € X and let S.,(x1), S.,(x2) € @ so that

fe Bas) tes. es).

Let p(x, 11) = di < «and p(x, 22) = de < €, and let

€e= min (€. a ds, ese dy).

We consider S.(x); let y € S.(x), then p(y, x) < «. Now

ply, X2) s ply, x) H- p(z, £2) <(¢ ae de

IIA (eg — do) + dh = @,

so that y € S.,(a2). Also

ply, 21) PUP) 1 Olt 21) ea an

IIA
IA (a Ean dy) alc dy = aly

so that y € S.,(a1). Thus we have found S.(x) € @ so that

S.(2) Se S.,(@1) \ S.,(2),

and by 1.29, ® isa basis. J

sec. 5.1 METRIC SPACES 117

NoTE: The S,(z) defined in 5.2 and used so extensively in 5.3 are called
spherical e-neighborhoods of zx.

The sort of computation with the metric that appears in 5.3 is fairly
typical of what goes on in metric spaces. The student should familiarize
himself with the techniques of such computations, and should draw
suitable pictures to assist him. For example, the choice of ¢ in the proof
of 5.3 is motivated by the situation exhibited in Fig. 5.1.

Me=

Figure 5.1

Now, of course, it can happen that a space X is already given us,

and we may wish to know if it is possible to define a metric, p, so that
the topology generated by the metric, using 5.2, is in fact the same as
the original topology. We define a space with this desirable property in
the following:

5.4. Definition. Let X be a topological space with topology 3. If it is

possible to define a metric p so that the metric topology generated by p
coincides with 3, then X is said to be a metrizable space.
118 METRIC SPACES CHAP. 5

EXAMPLES (AND EXERCISES)

5.1. The real line, R, with p(z, y) = |x — y| is a metric space, and the
metric so defined generates a topology which is the same as the
usual topology for R.
5.2. The real plane with

p((a1, yi), (2, Y2)) = [1 — 2)? + (ys — Yy2)?]”

is a metric space, and the metric so defined generates a topology
which is the same as the usual topology for the real plane.
5.3. Let (X, p) and (X’, p’) be metric spaces and let f be a relation
on X to X’ fie., f is a (possibly) multivalued function with
domain X and range X’]. Further, for each z, y € X, let p(x, y) =
p’ (f(x), f(y)) where f(x) (or f(y)) represents any one of the values
of f at x (or y), then f is called an isometry from X to X’. Show
that if f is an isometry, then it is a homeomorphism (possibly
into X’).

5.4. Show that what is suggested by the last three words of Ex. 5.3
cannot occur if f is an isometry of X with itself and X is compact,
i.e., show that if f : X — X is an isometry and X is compact, then
f is necessarily onto.

5.5. Let X be a set, and let p(x, y) be a real-valued function on X XK X

such that
(a) p(x, y) = O if and only if z = y.
(b) p(x, y) S plz, z) + ply, 2) for all z, y, 2 € X.
Then (X, p) is a metric set.

NOTE: Before attacking the next three exercises it is useful to observe

that in every metric space X, if A C X and z is a limit point of A, then
there exists a sequence {z,} of distinct points of A, such that lim z, = 2.

5.6. Let X be a metric space with metric p, and let @ =A CX.

Define the diameter of A, 6(A) = sup p(z, y), i.e., the least upper
zyCA
bound of distances between points of A. Show that
sEc. 5.1 METRIC SPACES 119

(a) If A C B, then 6(A) S 6(B).

(b) 6(A) = 6(A).
(c) If A is compact, then there exist z, y € A such that 6(A) =
p(x, y).
(d) There exist sets A (obviously not compact) such that for all
x,y € A, p(x, y) < 6(A).
5.7. Let X be a metric space with metric p, let x € X, and let
@ AA CX; define p(x, A) = inf p(z, y). p(x, A) is called the
yEGA
distance from the point x to the set A. Show that
(a) If A is compact, there exists a point y € A such that p(x, A) =
p(x, y).
(b) There exist sets A (obviously not compact) such that for all
y € A, p(x, A)» < p(a, y).
(c) A = {x |p(x, A) = 0}.
5.8. Let X be a metric space with metric p, let 0 ~A CX, OH
BC X; define the distance from A to B,

p(A, B) a inf p(x, y).

zE€A,yEB

Show that

(a) p(A, B) = p(A, B).

(b) If A and B are compact, then there exist x € A, y € B such
that p(A, B) = p(z, y).
(c) There exist closed sets, A and B (obviously not both compact),
such that for alla € A,y¢€ B

o(A, B) < p(s, y).

(d) For any sets A, B, CC X,

o(A, C) <p(A, B) + p(A UB, C) + 6(B).

(e ~~
For any sets A, B, CC X,

Give an example in which 6(B) # 0 in which equality holds.

120 METRIC SPACES CHAP. 5

§2 Some Properties of Metric Spaces

One of the more interesting problems is to try to decide what sorts

of spaces are metrizable. Before we tackle this problem, however, let us
explore some of the properties of metric spaces. Metric spaces have a
fairly strong structure, and this is demonstrated by the following
theorems.

5.5. Theorem. Every metric space is a Hausdorff space.

Proof. Let z, y € X, with x # y, then p(z, y) = d > 0; let e = d/2

and let U = S.(x), V = S.(y), then U and V are open sets, and we need
only show UM V = @.
Suppose in fact that U(\ V # @, then there isa z€ US) V and
since xz € U, p(z,x) < e«. Similarly p(y, z) < «. Thus

p(a, y) s p(x, z) =f ply, z) <2e= d,

and d = p(x, y) < d. This palpable contradiction shows that

GiANV =,

whence X is Hausdorff. J

5.6. Theorem. Lvery metric space is a normal space.

Proof. We already know that if X is metric, then it is a Hausdorff

space, hence 7;. We need thus only verify that if A and B are closed
disjoint subsets of X, then there exist open sets U, V such that A C U,
Bey and OGY =o,
Now for each a € A, a is not a limit point of B, for if it were we
should have a € B, since B is closed; and then a € A‘) B = @ and
this is clearly impossible. Thus for each a € A, there exists an e, > 0
so that S.(a) (1) B = @. Let U. = S,.,/2(a). Similarly for each b € B,
there is an «, > 0 so that S8.,(6) -\ A = @. Let Ve = S./2(b). Finally
let
T= WS UG V=U YN,
aGA

then U and V are open.

sec. 5.2 METRIC SPACES Al

We show now that U (\ V = ©. Suppose in fact that U NV # @,

and let € UV, then x € U, for some a and x € V, for some,
whence p(x, a) < €o/2, p(x, b) < e/2, and

p(a, b) S pla, x) + p(w, b) < 3(ea + 6).

If «a S &, then

p(a, b) << Z(€a =: €) Se and a€ S.(6),

so that S.,(b) (\ A ¥ @, contrary to the choice of e&. On the other hand

if 6, S ea, then

pia, 6) =F +a)S6— and b € S.(@),

so that S.,(a) (\\ B.¥ @, contrary to the choice of €. It must be the

case that U(\ V = @, and hence that X is normal. J

Further exploration of the properties of metric spaces shows that

notions which were heretofore distinct become equivalent in metric
spaces. The following few theorems demonstrate this.

5.7. Theorem. Let X be a metric space, then X is separable if and only

uf X as second countable.

Proof. Let IK be separable, and let

DRS AV ra)Vacant ho ay |

be a countable dense subset, let

@ = {S,(z,) |x; € D, r rational, r > 0},

then & is a countable family of open sets, being the countable union of
families of the form

{S,(a;) |a; fixed, r rational, r > 0}

which are countable since the rationals are countable. We show now
that @ is a basis.
Let x € X, U € Uz, then there exists « > 0 such that S.(z) € U. If
x = x, for some 2, select r, rational, 0 <r < e, then
122 METRIC SPACES CHAP. 5

ee S,(2) GC S.a), SU.

If x ¥ 2; for any 7, then z is a limit point of D and there exists an 7 such

that x; € S.j3(x). Select 7, rational, so that «/3 <7 < «/2, then since
p(x, 2:1) < €/3 < 7, x € S,(x,). Also if y € S,(z,), then

p(a, y) S ele, 2) + ples y) <$+5 = a/R

<eé

so that y € S.(a) and we have

2e Sr) Sse) CU;

In any case we have found S,(x;) € ® such that x € S,(a,) € U,

whence & is a basis.
The converse is settled by 1.32. §

5.8. Lemma. Let X be a countably compact metric space, and let e > 0,
then there exists a finite set F, such that X = U S,(2).
rECF,.

Proof. Suppose the theorem false. Select x; € X, then X ¥ S,.(x1),

and we may choose a2 € X — S,.(a). Note that p(a, x2) 2 e. Once
again
2
KA SHe)),
i=1

and we proceed similarly; at the nth step,

X # U S.(2,)
7=1

and we may select tn41 € X — U S.(a:) and we observe that

p(ri, xj) 2 fort. ~7,. 152,97

Su i

Consider the sequence {x,} thus constructed. It is an infinite subset

of X, since x; ¥ x; for 7 ¥ Jj, thus since X is countably compact {z,} has
a limit point x) € X. There then exist infinitely many points of {z,} in
S./2(%0), hence a fortiori at least two distinct such points, say v,, and ap.
Then

P\hey in) Spm to) Olina es 5aie 5 =~6

SEC. 5.2 METRIC SPACES 123

and this contradicts

p(n, as) =e.

This contradiction proves the lemma. J

5.9. Lemma. Hvery countably compact metric space ts second countable.

Proof. For each positive integer n, let F'1,, be the finite set, whose
existence is guaranteed by 5.8 such that

2 Sin (x) = Xx.

EF in

Let D= U Fin, then D is the countable union of finite sets and is thus
n=1

again countable. We show D = X. Let x € X —D and let «> 0.

Select n > 0 so that 1/n < e«, then since

Xda — ane), Si/n(x),

zEFijn

there exists y € Fijn G D so that x € Sijn(y), whence

Wee he t2 A EON

There thus exists for each e > Oa y € D()\ S,(x), consequently x € D,

and D = X. X is thus separable, and therefore by 5.7 second count-
able. §f

5.10. Theorem. A metric space is compact if and only af it 1s countably

compact.

Proof. Compactness implies countable compactness by 3.26. Con-

versely if X is countably compact, then by 5.9, X is second countable.
Since X is metric it is also Hausdorff by 5.5 and is consequently a T;
space. The desired conclusion now follows from 3.28. ff

EXAMPLES (AND EXERCISES)

5.9. Go back and look at Definition 3.22. Now prove that a T; space
X is completely normal if and only if, whenever A and B are two
124 METRIC SPACES CHAP. 5

separated subsets of Y, then there exist disjoint open sets, U

and V, such that A CU, BCY.

. Show that every metric space is completely normal. [Hint. Show

first that every subspace of a metric space is metric, Le., if A C X,
then the metric is compatible with the relative topology. |

. Show that if X is a metric space, then the mapping p: X X

X — R defined by p[(z, y)] = p(2, y), i.e., the distance from z to y,
is a continuous function.

. Show that if X is a connected metric space, then for each x € X,

the function f,(y) = p(x, y) has the intermediate value property
fie., if f.(y1) = a<c <b = f.(y2), then there exists y € X such
that jf.) = el:

. Show that there exist metric spaces Y which are not connected
yet for which for each x € X, f,(y) = p(x, y) has the intermediate
value property.
0.14. Show that any subspace of a separable metric space is again a
separable metric space. [Hznt. Use 5.7 and Ex. 1.72.] Show fur-
ther that if XY is not metric this result may fail. [Hint. Consider
the space J of Ex. 1.74, and let P = Fr (J) in E be the subspace.|

5. Let X be a metric space with metric p. Define p(A, B) as in Ex.

5.8. Show that in general p(A, B) is not a metric for the set of all
subsets of X.
Now define S.(A) = U S.(x), and define
rEA

p(A, B) = inf {e] BC S8,(A) and A C S.(B)}.

e>0

Let X be a metric space, and let X¥* = {C| C C X, C closed}, then

pris a metric for X*. The metric p; is usually called the Hausdorff
metric for the space X%*.

5.16. Show that any metric space X is homeomorphic to a metric space

X’ with metric p’ for which p’(z’, y’) <1 for all z’, y’ € X’.
[Hint. Let the points of X’ be the same as the points of Y, and
define p’(x, y) = p(x, y)/[1 + p(x, y)], where p is the metric in X.
Then show that p’ is a metric, p’(x, y) < 1 for all x, y € X, and
that the mapping f(~) = x is a homeomorphism. |
SEC MOIS METRIC SPACES 125

$3 Metrization Theorems

The object of the following sequence of theorems and lemmas is to

establish that every regular second countable space is metrizable in the
sense of Definition 5.4. We require a good bit of machinery before we
are ready to prove this result, and it is the purpose of the following
lemmas to assemble the appropriate hardware.
First a definition:

5.11. Definition. Let

= HOP Un real for each n

5C a {y|y

such that > y2 < a

n=1 fe

1.¢e., KH ws the collection of all sequences of real numbers such that the serie:
formed from the squares of the terms of the sequence 1s a convergent series.
Define, for x, y € &,

py) =| 3 @— va],
00 1/2

then the resulting metric space is called Hilbert space.

EXAMPLES (AND EXERCISES)

5.17. Show that p is a metric for 3.

5.18. Show that the subspace of 3¢ defined by

de ae a fan, = Otorns> 1}

is homeomorphic to the real line with the usual topology.

5.19. More generally #” C 3, defined by

Er =e lee Re = 42), eF = 0 for 1 > n}

is homeomorphic to Euclidean n-space, i.e., to X R;, where each

a=1
126 METRIC SPACES CHAP. 5

R; is the reals with the usual topology.

5.20. The subspace 3¢’ of 3C defined by

a’ = {|x € H, x = {x,},0 Sz, S 1/n for eachn}

is called the Hilbert cube (or Hilbert parallelotrope). Let J, =

[0, 1] for each n, i.e., the unit interval with the relative topology

inherited from the reals, and let J2 = X I,, then 5’ is homeo-

i 1

morphic to I’.

5.21. Show that the Hilbert cube is compact.

ODL Wee
Ais Gp (BES, g2— 0,7 = 2,

where 6/, the Kronecker delta, is defined by

y= 0 if gos) eth ee
i.e., A is the set of unit points on the “coordinate axes” of %.
Show that A is closed but not compact.

5.23. Is the unit sphere

s={x| > o, = 1) where ¢ = fey bc oe

i=1

compact? [Hint. Use Ex. 5.22 above. |

Our plan of attack on the metrization theorem is to show that every

second countable regular space is homeomorphic to a subset of Hilbert
space (in fact of the Hilbert cube), and is thus metrizable, since all we
need do is use the metric of the subspace of 5¢’ to which the given space
is homeomorphic to define a compatible metric. Now in order to define
the appropriate mapping of our space X into 35’, we need to specify the
terms, yn, of the sequence y € 3’ which is to be the image point of some
preselected point x € X. We thus need some device for associating a
sequence of real numbers {y, |0 < y, S 1/n} with each point of our
space.
We wish to exploit the regularity of X, and specifically to make use
of 3.18, which tells us that for each open set, consequently for each basic
SEC. 5.3 METRIC SPACES D7

neighborhood B; of a point x € X, there is another open set, which we

may choose as a second basic neighborhood, B;, such that « € B; C
B; C B,. Then Bj and X — B; will be disjoint closed sets. If we consider
all pairs of basic neighborhoods, (B;, B;) of X such that B; C B,, which
set of pairs is countable (by the second countability of X), and if we
can associate with each such pair a real-valued function \; such that
0 S X,(x) S 1, we would at least be partially on our way. In order to
accomplish this we prove first the following:

5.12. Lemma (Urysohn). Let X be a normal space, A, B closed disjoint

subsets of X, then there exists a mapping f : X > I, I = [0,1] such that
CA =D By = 1.

NOTE: It may seem strange that we start with a normal space, since we
are interested in second countable regular spaces; however, we know
that second countable regular spaces are normal by 3.24 and the exercises
following it.

Proof. Stage 1. Let X — B = Gj, an open set, since B is closed, then

A CG; since A (\ B = ©. By 3.19, there exists Gi. open such that
AS Gij2 = Gio CG

Stage 2. Again by 3.19, there exist G1;; and Gs,4 open such that

A = Gis Ss Gays Ss Gye S Gi/2 S Gya & Gs/4 e Gi.

Stage 3. Once again by 3.19 there exist G1/s, (s/s, Gss, and G7/s open
such that

A S Gis S Giys ‘= Gis = Gays & Gsys es Gays = Gij2 e Gry S Gs;s

= Gos S Gy = Gs/4 = Grjs = Gis Ss Gy

and so forth up to
Stage V. By 3.19 for each-odd integer 22 — 1,1 S$ 27 -1 S 2" — 1,
there exists an open set Gi;—1)/2v such that A C Gv and

Geei—2)en © Gai—1yaw SG @ei-1yow | Gaon

for each 7.
By induction we construct for each dyadic fraction, t, between 0
and 1, i.e., for each fraction whose denominator is 2", n 2 0, an open
128 METRIC SPACES CHAP. 5

set G, such that if and ¢’ are two dyadic fractions then ¢ < ¢’ if and only
if G, S Gy.

Now for « € X, define

f(z) I inf t forx gB

rEG:

wl forz € B.

Observe that A C G; for all t, thus f(z) = 0 if x € A, and also note that
a 3 epee
We are left with the task of showing f continuous. Let us examine
the structure of f—([0, y)) for 0 < y S$ 1. Now f(x) € [0, y) provided
0 < f(x) < y, and since the dyadic fractions are dense in [0, 1] there
exists a dyadic fraction f such that

f(z)= inf t<h <y.

zea:

Consequently x € G;,,. On the other hand if f) < y and x € Gy,

Te) inf bahay and “Ff@)e (Ory:

zrEG:

We thus see that f—1([0, y)) = UG. Since G: is open for each 1,
t<y

f-({0, y)) is open for each y.

By a similar argument it is clear that

A TD) oe a (X — Gi)

where 0 S y < 1. Since G; C Gi, X — G, D X — G;, for each t. Thus

U(X -@) DU (X —- G).

t>y t>y

However, if « € U (X —G,), then there exists a ¢ > y such that

t>y

x € X — G,, and again by the density of the dyadic fractions in [0, 1]

we may select ¢’, a dyadic fraction, such that t > t/ > y, then Gy C Gi,
and X — Gy D X — G,, so that « € X — Gy for some t’ > y, whence

tie Xa
t>y
sEec. 5.3 METRIC SPACES 129

Thus
ORG) GUOKOG=.G)
t>y t>y

and finally
U(X —G) = U (X — @)).
t>y t>y

Consequently, since each G, is closed, each X — G; is therefore open,

and we see that f((y, 1]) = U (X — G,) is open.
t>y

Now let U be some open set in [0, 1], such that f(a) € U, then there
exists a basic set V in [0, 1] such that f(x) © V C U, and the basic set
V in the relative topology of [0, 1] has one of the following forms:

OD) Toe), etl gi Y.

(ye 20) arpa Te
(3) WH, ¥2), D<y<y <1.
(4) [0, 1).

If V = [0, y), then f-'(V) is open by what we have proved above.

If V = (y, 1], then f-(V) is open by what we have proved above,
also.
IY = Ging), then V = Vil’ Ve-where Vi = [0, 42); Vi = (i, 1);
and f—'(V) = f7(V1) (\ f(V2) is open as the intersection of open sets.
Finally if V = [0, 1], f-°(V) = X. In any case then f—!(V) is open
for any basic set-of [0, 1], whence by Ex. 2.12 f is continuous. ff

We are now in position, having assembled all the heavy hardware,

to attack the principal problem.

5.13. Theorem. Every second countable regular space is homeomorphic

to a subset of the Hilbert cube.

Proof. Let @ = {B;|7 = 1, 2,...} be the countable basis. We ob-

serve by 3.24 and the exercises following that our space is normal, thus
there exist pairs of elements of ® such that B; C B;. Since ®& is count-
able, the collection of all such pairs is again countable; let us call it

OSE
et errs,

where P,, = (B?, B?) and B? C B?. Now since B? M (X — B7) = ©,

130 METRIC SPACES CHAP. 5

and both 57 and X — B? are closed, we may define, by 5.12, a mapping

fn 2 X — I = [0, 1] such that f,(B”) = 0, fa(X — Bf) = 1. Finally de-
fine f : X — KH’, the Hilbert cube, by

f(z) = {falx)/n|n = 1,2,...}.

Since for each x, 0 S f,(x) S 1, f(x) € XK’.
First we show f is one-to-one. Let « ¥ y, then since X is Hausdorff
by 3.17, there exist open sets, which we may choose as basic sets, B, B’
such that « € B, y € B’,and B(\ B’ = @. Further, since X is normal
we can find B” € @, such that x € B’” C B” CB, then x € B”, y€
X — B, and the pair (B”, B) € @, i.e., for some n, (B”, B) = (Bi, BF).
Thus

fn(x) = fr(B?) = fr(B”) = 0,

while
aly) = tae = BT) fx X= Beale
Thus f(z) ¥ f(y), since f(x) differs from f(y) at the nth place.
Now we prove f is continuous. Let « € X, and let e > 0. We wish
to construct U € U, such that for any y € U,

e(f(x), fy)) < ein KR’.

First, since for any point y € X, 0 S f,(y) S 1, we have that

faz) — fay)? S —

Now the infinite series Y n~? converges, thus for N sufficiently large,
n=1

oo a 2

Pe ra SS 9
n

whence
no

ayia <s 5 ne< Ss

2 |fn(x) =
n=N i)
el

Now let k < N, then the function f;, : X — I is continuous, thus there
exists a U; € Uz such that y € U; implies
ke
[flz) — faly)| < QW — 1)”
sEc. 5.3 METRIC SPACES 131

or

ile) — AP
kb 2(N — 1)
N-1
Now let U = ()\ U;,, then if y € U,
k=1

3 frlx) — frly)? eos frlx) — fry? ios fr(a) — fry?

n=1 nv n=1 nv HN nr

<(N - Ven peas

and finally p(f(x), f(y)) < «. Thus f is continuous.

Finally we must show that f is an open mapping. Let U be open
in X, and let x € U, then there exist B;, B; € @ such that

pe B.C By GRE U

by the normality of X and the fact that @ is a basis. Thus the pair
(Bi, Bi) € ©, say (Bi, Bi) = (Bi, Bj). Then

ful) = fr(Bi) = 0,
and since X —U GX — B7,

Ii 0) = fnX — Bp) = 1,
so that for any y € X — U,

i —_ 2771/2 = 2771/2
o(f(x), f(y)) om |3, AG =e ] > (ane frly) ] =-
n

Thus if V = Sin(f(z)) C &’, y € V implies

1
(f(z), y) — n and J &) E U,

for if not,

PQ) EX-U and p(fle), (FW) 2 >

a contradiction. Thus f~1(V) G U, and « € V C f(U), whence f(U) is
132, METRIC SPACES CHAP. 5

open. Consequently we have proved that f is a one-to-one continuous

open mapping, and consequently f is a homeomorphism. Jf

5.14. Theorem. Every second countable space is metrizable if and only

if it is regular.
Proof. If X is second countable and regular, then by 5.13 X is
metrizable.
Conversely, if X is second countable and metrizable, X is a metric
space with some compatible metric, p. Hence by 5.6 X is normal, and
by 3.17 X is regular. J

EXAMPLES (AND EXERCISES)

5.24. Show that every locally compact second countable Hausdorff space
is metrizable. [Hint. Compactify the space.]

5.25. Show that if X is 7, the converse of Urysohn’s lemma (5.12) holds.

$4 Complete Metric Spaces

Our plan now is to look at another special class of metric spaces,

which possess a distinguishing property, completeness, somewhat weaker
than compactness. These spaces are of considerable utility in analysis,
and in fact the Banach spaces of analysis are essentially complete
metric spaces with some further algebraic structure. Consequently to
the student with a bent for analysis, complete metric spaces will be of
particular interest.

5.15. Definition. Let X be a metric space, {xn} C X be a sequence,

then {x,} ts called a Cauchy sequence provided that for each « > 0, there
exists an integer N > 0 so that m,n > N implies p(an, tm) < €.

5.16. Definition. Let X be a metric space, then X is complete provided

every Cauchy sequence in X converges to a point xo of X.

EXAMPLES (AND EXERCISES)

5.26. The real line is a complete metric space with metric

p(t, y) = |x a y\.
sEc. 5.4 METRIC SPACES 138

5.27. Euclidean n-space, #”, the topological product of n spaces each

of which is the real line, is a complete metric space with metric

n 1/2
p(t, y) = & a= nl | .

0.28. Hilbert space is a complete metric space.

5.29. Let
C(O, 1] = {f |f a continuous real-valued function
with domain [0, 1]},
and let

ef, g) = jauD i) 0);

then C[0, 1] is a complete metric space.

5.30. Let m be the set of all real-valued bounded sequences, i.e.,

Hee = Neha oerealifor each = 1,2) 25. ula! < KK,

for some real K, depending upon 2}.
Define

p(x, y) = sup
n
|%_ — Yn,

then m is a complete metric space.

. Let c be the set of all real convergent sequences; define p(x, y) =

sup |v, — Yn|, then cis a complete metric space, and is furthermore
n

a subspace of the space m of the preceding exercise.

5.32. Show that the definition of a complete metric space may be

weakened by insisting only that every Cauchy sequence have a
limit point in X, i.e., show that if a Cauchy sequence has a limit
point 2» in X, then it converges to 2».

. Show that every compact space is complete. [/7int. Use Ex. 5.32
above and 3.26. |

. Show that any closed subspace of a complete metric space is

complete.
134 METRIC SPACES CHAP. 5

5.30. (a) Let R be the reals, define f : R — (—1, 1) by

then f is a homeomorphism.
(b) R is a complete metric space, but the subspace (—1, 1) to
which R is homeomorphic [by part (a)] is not complete, since
the Cauchy sequence

n
feel ee.

does not converge to a point of (—1, 1).

5.36. Completeness is not a topological property. [Hint. Use Ex. 5.35

above. |

5.37. (a) Let X be a metric space, and let {z,} be a sequence in X such
that lim x, = 2 € X. Show that {z,} is a Cauchy sequence.
(b) Show that if Y C X, Y a subspace of X, then if Y is complete,
Y is closed in X.
(c) In particular if X = R, then any complete bounded subset
of X is compact. [A set A C X is said to be bounded if there
exists a point « € X and a finite real number N such that
A Sy @)s|

5.38. Let A = {x |x; = 6,7 = 1, 2,...} C 3C be the set of unit points

of Hilbert space (Cf. Ex. 5.22). Show that A is complete and
bounded but not compact.

The situation encountered in Ex. 5.37 above is somewhat more

general, namely in any Euclidean space HE" = X R,, where each R; is

1=1

the reals, a complete bounded subspace is compact. Unfortunately, this

result does not hold in general, as Ex. 5.38 above shows us. We need not
be dismayed, however, for by a slight strengthening of hypotheses we
may overcome this difficulty.
We need first
sec. 5.4 METRIC SPACES 135

5.17. Definition. A metric space is totally bounded if for each « > 0,

there exists a finite set F = {21, 22,..., Xn, such that X C U S.(:).
i=1

Then we have

5.18. Theorem. Hvery complete totally bounded metric space is compact

and conversely.

Proof. Let X be complete and totally bounded. We exploit 5.10 and

prove merely that X is countably compact. To that end, let A be an
infinite subset of X. Since 3 > 0, there exists

P, => {ru, H12,- + + 4 int

such that X C U S12(a1,) by the total boundedness of X. For some 2,

i=l
S12(a1;) (\ A is an infinite set, otherwise A would be finite. We can
assume §}/2(%) () A is infinite by suitably renaming the points of F4.
Let Ar = AQ Sip(an).
Now there exists FP, = {%2, 122, ..., 22n,} Such that

PC U S1/s(X2i).

Since A, is infinite there exists an 7 such that S1/4(@2;) ( A1 is infinite,

and again we may assume A, = S1/4(%21) (\ A; is the infinite set. Thus
in general having defined A;-1, there exists Fi, = {wm, Cia)... , Limb
such that

= A S12 (ei),

and A,-1 () S1/2(xz:) is infinite for some choice of 7, which we assume

again to be 1. We let A, = Aga O S1/2(x). We note that, in general,
A, € A;-4 for each. k > T.
Now select x1 € Ai, 2 € Az — {ai}, and in general

ket
xt, € Ap — Y {x}.

Since each A, is infinite, this selection will always be possible. The

136 METRIC SPACES CHAP. 5

sequence {z,} thus selected is a Cauchy sequence, for let « > 0, and
select k > 0 so that 1/2* < «. Then if m, n 2 k,

Xm ‘Ss Ag Se Ax and Xn e AN Ss Ax,

while A, © Sy/ox(vm). Consequently an, &m € S1/2x(em) and

PpGus tp) Ss Pry Le) Ga, Oe)

Since X is complete, {v,} converges to some point a € X, and since

{xz,' CA, a is a limit point of A. Thus X is countably compact, and
as remarked earlier 5.10 completes the argument.
The converse is relatively trivial, and we leave its proof to the
student.

$5 Category Theorems

We now investigate a sequence of theorems which lead to the

so-called Baire Category theorem (sometimes called the Baire-Moore
theorem), which finds occasional use in both topology and analysis.
We need first some definitions.

5.19. Definition. Let N C X, a topological space, then N is called

nowhere dense if (N)° = @ (i.e., if the interior of the closure of N is
empty).

EXERCISE

5.39. If N C X is nowhere dense then

(a) (N)¢ is dense in X.
(b) For each x € X and each open set O such that x € O, there
exists O;, open and nonempty, such that O, C O and O, C N°.

5.20. Definition. A set b C X, a topological space, is said to be of

Baire Categcry I (or of the first category) if B is the union of countably
many nowhere dense sets. A set is said to be of Baire Category II (or of the
Syxel, fs METRIC SPACES 137

second category) if it is not of Batre Category I. A set is said to be residual

af its complement in X ts of Baire Category 1.

5.21. Theorem. Let X be a metric space, then X is complete if and

only if for each sequence {C,,} of closed sets such that Cn CG Cp for each
n, and such that
lim 6(C,,) = 0,

C,, is a single point.

a i

NOTE: 6(A) = sup p(w, y). (Cf. Ex. 5.6.)

yA

Proof. Suppose X is complete. Construct a sequence

aE Ce ett, te ON gate

then {z,} is a Cauchy sequence, for let « > 0. Select NV. > 0 so that
i> IN. implies 6(C,)< <«_, Then for 2» 2m > iN. 2, © C,.S.G,,, and
Lm € Cm, whence
Doan) Se OOe i re

Since X is complete, {x,} converges to some point ao € X.

Now if x, = 2 for all n > N, for some N > 0, then

Lo = Ln € C,, S Cr

for 1 Sig nn and v4 © 1) C,.

n=1

On the other hand, if there exist arbitrarily large values of n such

that ¢, ¥ Xo, then let m > 0 be fixed. Then fore > 0, there isann > m,
such that
Xn ca Xo, Ga Xo) << €,

since lim x, = 2. Thus 2 is a limit point of C,, and since C,, is closed,
n

ay € Cy. Also since C, GC,, 2 € Cn, and since m was arbitrary,

xo € Cm for all m, consequently

n=1
138 METRIC SPACES cHaAP. 5

Suppose now that y € A C,. Let ¢ > 0, and select N. so that

n=1

n > N. implies 6(Cn) < ¢, then

to, y € Co and p(x, y) S 6(Cr) <e.

Since ¢ was arbitrary, p(x, y) = 0, and a = y. Thus

n=1

Conversely suppose that the condition of the theorem holds. Let

{x} be a Cauchy sequence in X, and let S.(x) = {y |p(z, y) < ¢}. For
each positive integer k, there exists a positive integer n, such that
n = nm implies p(any, Tr) < 1/2*, and we may assume n; so chosen that
it is the least integer with this property. It then follows that n, S m4.
Define

Cr = Siem), fotiiccael a2 aed

Observe that
1
Cre {y|pP@m)Y) S asi

so that 6(C,) S 1/2** and limié(C,) = 0.

k
We show that C; is a contracting sequence, i.e., Cy41 C Cy. Let
y € Cras, then

1 1
= Dk and Pitney Cae) s Dk
ply; Tg)

by choice of n,. Thus

1 1 1
ply, aan) = ply; Tn,.) st P(r Ln) s Ok ze Qk Fs 9k-V

and y € C,. Therefore Cy4; C C,.

By hypothesis 1S C;. = {xo} is a single point. Let « > 0, and select

k so that 1/2" < «, then for n > mg, since % € Cx, p(Xo, tn,) S 1/2*-,
and
SHO ID METRIC SPACES 139

1
pe, te) ee Lig) ip Cay ta)S 3 +5
ot < ora <6

Thus lim z, = x, and X is complete. J

5.22. Theorem. Let X be a complete metric space, then X is of Baire

Category II.

Proof. Let NC X, N of Category I. We need only show that

N # X, or that there is a point x € X — N.

N= UO N,, where N, is nowhere dense. Now (N,,)° is dense in X,

n=1

for each n, thus there is a point x; € (N;y)¢ and there is an e > 0 so that
Sis) PVG, =O by Hx, 5.39. Thus

S_2(%1) & S.(21) and Se2(a1) VN, = ©.

Let ¢/2 = a, then again since (N2)° is dense in X, there exists a point
te € S,(a1) CO (N2)*. As before we select ¢: > 0 so that
(1) .0 < e@ < «/2,
(2) S,(@2) A No = ©,
(3) S,@2) © 8.(),
and in general for each n, select rt, € Se ,(@n1) (\ Nn, and e, so that
CL) One, = 63/2,

(2) S..@n) Na = D,
(3) e(@,) SoBe oes).
We thus obtain a sequence S,,(z,,) of contracting closed sets such that

5 (S.,(tn)) < €/2" 0, whence by 5.21 there is a point x € a) Sana)

Let m > 0 be fixed, then

Xo € en (a) and DE (Cy) (\ Vm = gm,

thus a> ¢ Nm, and since m was arbitrary, xo ¢ N» for any m. Thus

xo ¢ LENS SNe wy € X —N,

n=1

whence X is not of Category I.

140 METRIC SPACES CHAP. 5

5.23. Theorem. Let X be a complete metric space, R € X, R residual,

then R is dense.

Proof. Let « € X — R, we need only show that 2 is a limit point

of R. Now X — R is of Category I, thus

x= R= ONG
n=1

where each N,, is nowhere dense. Let O be an open set such that x € O,
and select « > 0 so that S.(z) C O. We now proceed to construct a
sequence {z,} of points and closures of neighborhoods of these points,
S.,(v,) as in the proof of 5.22, so that
(PO en ena2,
(2) S..@n) AN, = OS,
Cy CATS SoG)
Ged,
and further so that x, € S.(x), and S.
S,(
(x1) € S.(z). This is possible,
since 2; was chosen in (Nj)*, a dense alibset of X, thus there is an
a1 € S.(x) M (Ni)*, and we may choose « so that S.
(a) C S.(x). Then
as in 5.22

Xo € A Spl a) = Se(x De
n=

and 2) ¢ N. Thus there is an a € X — N = R in each open set O

which contains x. Consequently z € R, and R is dense in X. ff

EXAMPLES (AND EXERCISES)

5.40. In the space C[0, 1] of Ex. 5.29 which we know to be complete, let

P,, = {f|for some x € [0, 1], for the positive integer m,

and for all h > 0, |(f(@ + h) — f(x))/h| S m}.

Let S = C(O, 1] — P,,, then S is the set of all continuous

functions on [0, 1] not having a finite right-hand derivative at any

point of [0, 1].

(a) Show that each P,,, is closed.

(b) Show that for each f € P, and each « > 0 there exists a
sec. 5.5 METRIC SPACES 141

g € C(O, 1] such that g ¢ Pn, g € Sf). [Hint. Approximate

f to within «/2 by a polygonal line, then approximate each
segment of the polygonal line (to within ¢/2) by a sawtooth
with teeth having slopes, both positive and negative, greater
in absolute value than m.|
(¢) Use (a) and (b) above to show that each P,,, is nowhere dense.

(d 7
Show that U P,,, is of Baire Category I.
m=1

(e) Show that S # @.

(f) Prove that there exist functions continuous on [0, 1] which do
not have a finite derivative at any point of [0, 1].

5.41. (a) let X be a complete metric space and let f : X — X bea

mapping such that p(f(x), f(y)) < ap(x, y), where0 <a <1
and a does not depend upon x or y. Show that there exists a
point 2 € X, such that f(x) = x. [Hint. Let 2 € X, define
ie thy nh tae Oy Epa), & ans DOW that yr, Isa
Cauchy sequence, and let 2x = limz,. Show 2 has the

property f(ao) = 2o.]

(b Ss Show further that a is unique, i.e., there is no other point 2
such that (a1) — 21.
(c) Consider the differential equation y’ = f(x, y), with boundary
condition y(ao) = y, where x and y are real. Let f be con-
tinuous in

D = {((z,y) | |e — a Sa, ly — wl S$b,a>0,b> 05,

let N = max |f(z, y)| and let aN < b. Further let f satisfy
zyCD

the Lipschitz condition:

Wa,on) —f@, yw) Ss Kigy — x

for some fixed K > 0 and for (a, y1), (@, yo) € D.
Now define

X = fy(x) |y a continuous real-valued function of the real

variable x, and |y(x) — yo| S 8 for |x — a| S a},

and define
142 METRIC SPACES CHAP. 5

p(Yi,y2) = iat |yi(x) — yo(z) | for m1, y2e X.

Show that X is a complete metric space with the metric p.

(d) Let X be as in part (c) above. Define T : X — X by

Ty) = yo+ [*fle, y@)) de.

Show that T(y) € X. Show further that

o(T (yr), T (y2)) = aK p(y, Y2).

Now select a so small that ak < 1.

(e) [Continuation of (d)] Show that there is a unique y € X such
that T(y) = y, 1.e.,

yo) = yo+ [*fe, y@) a,

and hence that the differential equation y’(x) = f(z, y) with
boundary condition y(x.) = yo has a unique solution in at
least that portion of D for which a < 1/K.

5.42. (Term paper) It is the purpose of this exercise to generalize the

technique that is often used to obtain the reals from the rationals.
Specifically we want to prove that any metric space (the rationals,
for example) is isometric to some subspace of a complete metric
space (the reals, for example).
Now let X be a metric space with metric p. Let

Y = {{x,} | {x,} a Cauchy sequence of points z, € X}.

Define a relation, cc
‘“~,” in Y, by

t= {Xn} ~y = {yn}
if and only if lim p(zn, yn) = 0.

(a) Show that ‘“~” is an equivalence relation in Y.

(b) Let Z = Y/~, ie., Z is the set of all equivalence classes of
elements of Y under the equivalence relation, ““~.”’ Denote
an element 7 € Z by % = [x], where x € Y, x = {z,}, {rn} a
Cauchy sequence in X. In Z define
sEc. 5.5 METRIC SPACES 143

p(z, 7) a lim p(n, Yn)-

Then
(i) Show that j is independent of the choice of representatives
{raf € [x] and {yn} € [y].
(ii) Show that j is a metric for X.
(ec) Let Z be provided with the metric topology induced by .
Show that Z is complete.
(d) Define f : X = Z by f(@) = [2] = [{2,}4,. where 2, = a for
all n, i.e., every point x € X is mapped into a constant
sequence, 7,7, 2,.... Show that fis an isometry of X witha
subset of Z.
(e) Show that f(X) is dense in Z.
(f) Show that X is separable if and only if Z is separable.
 : _

1 apie” aa
Sh : arene tee ’
NG Poa gin (lh eee se as
ale! iey bel a

Divi ae bebe
y 7 . "

aw
me! - 2s a

- = =i ane

haw

- - 7

= —
 REFERENCES

The following list of books is intended as a guide to further reading in

point set topology. Some of the books cover the same material as is
covered here; however, a number of them go considerably beyond what
is attempted here, either exploring point set topology further, or
investigating the application of point set topology to algebraic topology
or analysis. Although each of the books listed below has its merits, the
author has found the book by Kelley and the one by Hocking and Young
to be particularly useful.

1. Alexandroff, P., and H. Hopf, Topologie (Ann Arbor, Mich.: Ed-

wards, 1945).

2. Arnold, B. H., Intuitive Concepts in Elementary Topology (Englewood

Cliffs, N.J.: Prentice-Hall, 1962).

3. Bourbaki, N., Topologie Générale (Paris: Actualités Scientifiques et

Industrielles, Herman et Cie., 858 (1940) = 1152 (1951), 916
(1942) = 1148 (1951), 1029 (1947), 1045 (1948), 1084 (1949)).
4, Hall, D. W., and G. L. Spencer, Elementary Topology (New York:
Wiley, 1955).

. Hausdorff, F., Mengenlehre (Berlin: de Gruyter, 1927, 1935).

On

6. Hocking, J. G., and G. 8. Young, Topology (Reading, Mass.:

Addison-Wesley, 1961).

7. Hurewicz, W., and H. Wallman, Dimension Theory (Princeton, N.J.:

Princeton, 1941).

8. Kelley, J. L., General Topology (Princeton, N.J.: Van Nostrand,

1955).

9. Kuratowski, K., Introduction to Set Theory and Topology (New

York: Pergamon, 1961).
145
146 REFERENCES

10. Kuratowski, K., Topologie, vols. 1 and 2 (2nd ed.; Warsaw, 1948).

it Mansfield, M. J., Introduction to Topology (Princeton, N.J.: Van

Nostrand, 1963).

12. Moore, R. L., Foundations of Point Set Theory (New York: American
Mathematical Society Colloquium Publication No. 13, 1932).
13. Newman, M. H. A., Elements of the Topology of Plane Sets of Points
(New York: Cambridge U.P., 1939).

14. Patterson, E. M., Topology (New York: Interscience, 1959).

Ly Sierpinski, W., Introduction to General Topology (Toronto: Univer-

sity of Toronto Press, 1934, 1952).

16. Simmons, G. F., Introduction to Topology and Modern Analysis

(New York: McGraw, 1963).

Vi, Vaidynathaswamy, R., Treatise on Set Topology, Part 1 (Madras:

Indian Mathematical Society, 1947).

18. Whyburn, G. T., Analytic Topology (New York: American Mathe-

matical Society Colloquium Publication No. 28, 1942).

19. Wilder, R. L., Topology of Manifolds (New York: American Mathe-

matical Society Colloquium Publication No. 32, 1949).
 INDEX

Arc, 109 Connected, 99

Arewise connected, 109, 110 arewise, 109, 110
Axiom: locally, 107
of choice, 12 Constant function, 56
separation, 79 Continuous, 58
image, 60
Batre category, 136 ; at a point, 58
Baire-Moore Theorem, 136 Countable:
Base (basis): base (basis), 47
countable, 47 set, 11
for a topology, 43 Countability:
Bound: first axiom of, 49
greatest lower, 17 second axiom of, 47
least upper, 17 Countably:
lower, 17 compact, 89
upper, 17 infinite, 11
Bounded: Cover, 69
set, 134 Cut point, 113
totally, 135
Dre Morean’s rules, 4, 6
Cantor set, 112 Dense, 33
Cartesian product, 6, 65 in itself, 103, 113
Category, 136 Derived set, 28
Cauchy sequence, 132 Diameter, 118
Chain, 101 Difference, 4
simple, 101 Discrete topology, 23
Chained by, 101 Distance:
Choice: from one set to another, 119
axiom of, 12 from a point to a set, 119
function, 12
Closed: Empty set, 2
function, 62 Equivalence:
set, 31 class, 9
Closure of a set, 28 relation, 9
Compact, 70 Euler—Venn diagram, 7, 8
Complement of a set, 4 Extension of a function, 56
Complete, 132
Completely normal space, 87 Factor space, 64
Component, 105, 106 Finer topology, 42
148 INDEX

Finite intersection property, 71 Limit (cont.):

First axiom of countability, 49 of a sequence of sets, 39
Frontier of a set, 33 superior of a sequence of sets, 39
Function, 56 Lindeléf Theorem, 87
choice, 12 Line, usual topology for, 22
closed, 62 Lipschitz condition, 141
constant, 56 Locally:
continuous, 58 compact, 92
at a point, 58 connected, 107
extension of, 56 finite, 96
identity, 56 Lower bound, 17
interior, 62
one-to-one, 11, 57 Mappine, 60
onto, 57 Mathematical induction, 16
open, 62 Metric, 115
restriction of, 56 Hausdorff, 124
set, 115
GRrRaPH, 68 space, 116
Greatest lower bound, 17 topology, 116
Metrizable space, 117
HAUSDORFF:
metric, 124 NEIGHBORHOOD system, 20
space, 40, 79 Normal space, 81
Hilbert: Nowhere dense, 136
cube, 126
space, 125 ONE-TO-ONE correspondence, 11
Homeomorphism, 61 One-to-one function, 11, 57
One-point compactification, 94
Ipentity function, 56 Onto, 57
Image, 56 Open:
continuous, 60 function, 62
Inclusion, 2 set, 23
proper, 2 Order relation, 10
Infimum, 17
Infinite set, 11 PARACOMPACT, 96
Interior: Partial order relation, 10
function, 62 Partition, 9
of a set, 33 Plane, usual topology for, 22
Intersection, 3, 5 Product:
Inverse image, 57 Cartesian, 6, 65
Isometry, 118 space, 52, 65-68
topology, 52, 65-68
KRONECKER delta, 126 Projection, 64, 66
stereographic, 92, 93
Latics, 54 Property:
Least upper bound, 17 finite intersection, 71
Limit: topological, 62
inferior of a sequence of sets, 39
point, 27 REFINEMENT, 96
of a sequence of points, 38 Reflexivity, 9
 INDEX 149

Regular space, 81 Space (cont.):

Relation, 9, 10 complete metric, 132
equivalence, 9 completely normal, 87
order, 10 connected, 99-104
partial, 10 countably compact, 89
simple, 10 factor, 64
Relative topology, 36 Hausdorff, 40, 79
Residual, 137 Hilbert, 125
Restriction of a function, 56 locally compact, 92
Right order topology, 23 locally connected, 107
metric, 116
SECOND axiom of countability, 47 metrizable, 117
Separable, 47 normal, 81
Separated, 99 product, 52, 65-68
Separation axioms, 79 regular, 81
Sequence, Cauchy, 132 separable, 47
Set(s): T, 79
bounded, 134 T1, 79
Cantor, 112 T>, 79
Cartesian product of, 6, 65 T3, 79
closed, 31 ay OY
closure of, 28 topological, 21
complement of, 4 Stereographic projection, 92, 93
countable, 11 Sub-base for a topology, 50
countably infinite, 11 Subset, 2
dense, 33 Subspace, 36
derived, 28 Supremum, 17
difference of, 4 Symmetry, 9
elements of, 2 | System of neighborhoods, 20
equality of, 2
finite, 11
frontier of, 33 TOPOLOGICAL:
inclusion, 2 property, 62
indexing, 5 space, 21
infinite, 11 Topology, 21
interior of, 33 base (basis) for, 43
intersection of, 3, 5 comparison of, 42
metric, 115 discrete, 23
notation for, 1 product, 52, 65-68
open, 23 relative, 36
proper inclusion, 2 right order, 23
separated, 99 sub-base for, 50
union of, 3, 5 trivial, 23
Simple: usual:
chain, 101 for the plane, 22
order relation, 10 for the reals, 22
Simply chained by, 101 Totally bounded, 135
Space: Totally disconnected, 111
arewise connected, 109, 110 Transitivity, 9
compact, 70 Trivial topology, 23
150 INDEX

Tychonoft: Urysohn’s lemma, 127

plank, 84-87
theorem, 77 WELL-ORDERED, 12

Union, 3, 5
Upper bound, 17 Zorn’s lemma, 75
 i
Date Due

Lourdes Libr
vwynedd Merc ary
P.O. Box 904 y
aoe
WYRedd Valley
4 1O4AF_Anp
ia
(Continued from front flap)
understanding of advanced point set
or algebraic topology than to the un-
derstanding of analysis. Thus, this
treatment is useful not only to the
reader whose goal is functional analy-
sis but also to the reader whose goal
is advanced topology.

The Author
Joun D. Baum, Ph.D. Yale University,
is Professor of Mathematics at Oberlin
College. He is the author of several
research papers on topological dy-
Q
namics and a member of the Publica-
tions Committee and the Consultants’
Board of the Committee on the Un-
Cri:
dergraduate Program of the Mathe-
—/.
La»
a)
2eCw
matical Association of America.

PRENTICE-HALL, INC.
Englewood Cliffs, New Jersey
1067 © Printed in U.S. of America
Another book of interest. . .

PROBABILITY: An Introduction

by SAMUEL GOLDBERG, Oberlin College

This definitive volume deals with the essentials needed by

all who use probability concepts, apply statistical techniques,
or measure the effects of chance phenomena. Logical in its
approach, the book formulates its probabilistic theory by
employing the language and notation of sets.

By emphasizing experiments with a finite number of possible

outcomes, the study requires no previous knowledge of cal-
culus and can be readily understood by the reader with a
good background in high school algebra.

Supplying basic information first, the book opens with a

helpful treatment of the elementary mathematics of sets. It
continues with a penetrating examination of sample spaces
and probability theorems. A separate chapter is devoted to
more sophisticated counting procedures which when mas-
tered, are applied to a wide variety of complex experimental
situations.

Published 1960 322 pages

PRENTICE-HALL, INC.
Englewood Cliffs, New Jersey