0% found this document useful (0 votes)
84 views67 pages

Ilovepdf - Merged (2) - Removed - Compressed

The document provides a study material on topology for M.Sc Mathematics students. It includes an introduction to topological spaces and examples of topologies such as the indiscrete topology, discrete topology, cofinite topology, usual topology on real numbers, and order topology on linearly ordered sets. The first chapter covers topics like open sets, bases, sub-bases, closed sets, neighbourhoods, interior, accumulation points, and continuity in topological spaces. Subsequent chapters discuss spaces with special properties, separation axioms, and properties like compactness and normality.

Uploaded by

ARYAN SRIVASTAVA
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
84 views67 pages

Ilovepdf - Merged (2) - Removed - Compressed

The document provides a study material on topology for M.Sc Mathematics students. It includes an introduction to topological spaces and examples of topologies such as the indiscrete topology, discrete topology, cofinite topology, usual topology on real numbers, and order topology on linearly ordered sets. The first chapter covers topics like open sets, bases, sub-bases, closed sets, neighbourhoods, interior, accumulation points, and continuity in topological spaces. Subsequent chapters discuss spaces with special properties, separation axioms, and properties like compactness and normality.

Uploaded by

ARYAN SRIVASTAVA
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 67

School of Distance Education

MTH2C08-TOPOLOGY

STUDY MATERIAL

II SEMESTER
CORE COURSE

M.Sc Mathematics
(2019 Admission ONWARDS)

UNIVERSITY OF CALICUT
SCHOOL OF DISTANCE EDUCATION
Calicut University- P.O, Malappuram- 673635,Kerala.

190558

MTH2c08-Topology
School of Distance Education

SCHOOL OF DISTANCE EDUCATION


UNIVERSITY OF CALICUT

STUDY MATERIAL

SECOND SEMESTER

M.Sc Mathematics (2019 ADMISSION)

CORE COURSE : MTH2C08-TOPOLOGY

Prepared by:

SRI. SACHIN CHANDRAN,


ASSISTANT PROFESSOR ON CONTRACT (MATHEMATICS),
SCHOOL OF DISTANCE EDUCATION,
UNIVERSITY OF CALICUT.

Scrutinized By:

SRI. LIBEESHKUMAR. K.B,


DEPARTMENT OF MATHEMATICS,
C.K.G.MEMORIAL GOVERNMENT COLLEGE,
PERAMBRA, KOZHIKODE.

MTH2c08-Topology
Contents

1 Topological spaces 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Bases and Sub-bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Closed sets and closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Neighbourhoods, Interior and Accumulation Points . . . . . . . . . . . . . 15
1.6 Continuity and Related Concepts . . . . . . . . . . . . . . . . . . . . . . . 19

2 Spaces with special properties 23


2.1 Making Functions Continuous, Quotient Spaces . . . . . . . . . . . . . . . 23
2.2 Spaces with Special Properties . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2.1 Smallness Conditions on a Space . . . . . . . . . . . . . . . . . . 28
2.2.2 Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.3 Local Connectedness and Paths . . . . . . . . . . . . . . . . . . . 40

3 Separation Axioms 45
3.1 Hierarchy of Separation Axioms . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Compactness and Separation Axioms . . . . . . . . . . . . . . . . . . . . 51
3.3 The Urysohn Characterisation of Normality . . . . . . . . . . . . . . . . . 58
3.4 Tietze Characterisation of Normality . . . . . . . . . . . . . . . . . . . . 62
Module 1

Topological spaces

(Reference text: ‘INTRODUCTION TO GENERAL TOPOLOGY’ by K. D. Joshi)

1.1 Introduction
A metric space is a set endowed with an additional structure, namely, the metric or the
distance function. It is clear, therefore, that any concept in the theory of metric spaces,
unless it is a purely set-theoretic concept, will have to be defined in terms of the metric.
Throughout, (X, d) will be a metric space.

Definition 1. Let x0 ∈ X and r be a positive real number. Then the open ball with centre
x0 and radius r is defined to be the set {x ∈ X : d(x, x0 ) < r}. It is denoted either by
Br (x0 ) or by B(x0 , r). It is also called the open r-ball around x0 . When we want to stress
the metric d, we denote it by Bd (x0 , r).

It is obvious that the open ball Bd (x0 , r) depends not only on x0 and r, but on the metric
d as well. Let d be the discrete metric on a set X. Then for any x0 ∈ X, B(x0 , r) consists
of X or {x0 } depending upon whether r > 1 or r ≤ 1. A subset A of X is said to be
bounded if the function d is bounded over A × A. If A is a non-empty bounded set, its
diameter, denoted by δ(A) is the number sup{d(x, y) : x ∈ A, y ∈ A}. Note that every
open ball is bounded but its diameter may be less than twice its radius.

Proposition 1.1.1. Let {xn } be a sequence in a metric space (X, d). Then {xn } converges
to y in X iff for every open set U containing y, there exists a positive integer N such that
for every integer n ≥ N , xn ∈ U .

Proposition 1.1.2. Let f : X → Y be a function where X, Y are metric spaces and let
x0 ∈ X. Then f is continuous at x0 iff for every open set V in Y containing f (x0 ), there
exists an open set U in Y containing x0 such that f (U ) ⊂ V .

1
Theorem 1.1.3. Let (X, d) be a metric space. Then,
(i) the empty set φ and the entire set X are open,
(ii) the union of any family of open sets is open,
(iii) the intersection of any finite number of open sets is open,
(iv) given distinct points x, y ∈ X there exist open sets U, V such that x ∈ U, y ∈ V and
U ∩ V = φ.

Proof. (i) and (ii) are trivial consequences of the definition of open sets. For (iii) first
consider the case of the intersection of two open sets say A1 and A2 . Let x ∈ A1 ∩ A2 .
Then x ∈ A1 and x ∈ A2 . Since A1 is open, there exists r1 > 0 such that B(x, r1 ) ⊂ A1 .
Similarly since A2 is open there exists r2 > 0 such, that B(x, r2 ) ⊂ A2 . Now let r =
min{r1 , r2 }. Then clearly B(x, r) ⊂ B(x, r1 ) ∩ B(x, r2 ) ⊂ A1 ∩ A2 . Thus A2 ∩ A2 is
open. One can either generalise this argument or use induction to settle the general case.The
exceptional case of the intersection of an empty family of open sets is already covered under
(i). For (iv) let x, y ∈ X and x 6= y. Then d(x, y) > 0. Choose r so that 0 < r < d(x,y)
2
and
let U = B(x, r), V = B(y, r). Then clearly U, V are open sets containing x, y respectively.
Also they are mutually disjoint.

Theorem 1.1.4. Let J be the collection of all open sets in a metric space X. Then J has
the following properties:
(i) φ and X belongs to J ,
(ii) J is closed under arbitrary unions,
(iii) J is closed under finite intersections,
(iv) Given distinct points x, y ∈ X, there exist U, V ∈ J such that x ∈ J , y ∈ V and
U ∪ V = φ.

So, in order to define a topological space we take a set X, a certain family J of its
subsets and require that J satisfy some of the properties listed in the above theorem. We
will give the definition of a topological space.

Definition 2. A topological space is a pair (X, J ) where X is a set and J is a family of


subsets of X satisfying:
(i) φ ∈ J and X ∈ J ,
(ii) J is closed under arbitrary unions,
(iii) J is closed under finite intersections.

The family J is said to be a topology on the set X. Members of J are said to be open in
X or open subsets of X. A sequence {xn } in a topological space (X, J ) is said to converge
to a point y of X if for every open set U containing y, there exists a positive integer N such
that for every integer n > N , xn ∈ U . A topological space is said to be metrisable if its
topology can be obtained from a suitable metric on the underlying set. It may happen that
two distinct metrics on a set yield the same topology. A trivial case of this occurs when the

2
two metrics are scalar multiples of each by a constant factor. Of course not all topological
spaces are metrisable.

Exercise :

1. Prove that the open balls in a metric space are open sets.
2. Prove that a subset A of a metric space X is open iff no sequence in X \ A converges to
a point of A.

Rough space :

3
1.2 Examples
1. It may happen that the topology J on the set X consists only of φ and X. It is called
the indiscrete topology on X.

2. The other extreme is the so-called discrete topology on X. Here every set is open; in
other words the topology coincides with the power set P (X).

3. Let X be any set. A subset A of X is said to be cofinite, if its complement, X \ A,


is finite. Let J consist of all cofinite subsets of X and the empty set. Then J is a
topology on X and this topology is called cofinite topology.

4. The usual topology on R is defined as the topology induced by the euclidean met-
ric. Note that with this metric the open balls are just bounded open intervals. So all
bounded open intervals are indeed open sets in the usual topology on R. Since un-
bounded open intervals can be expressed as unions of bounded open intervals, they
are also open in the usual topology. Thus all open intervals are open sets of R in the
usual topology.

5. There is another, and a stronger topology on R, called the semi-open interval topol-
ogy. A subset U is said to be open in this topology if for every x ∈ R, there exists
r > 0 such that the semi-open interval [x, x + r) is contained in U . The verification
that this indeed defines a topology on R and that the topology so defined is stronger
than the usual topology is left to the reader.

6. Let a set X be linearly ordered by ≤. Declare a subset A of X to be open if for each


x ∈ A there exist a, b ∈ X such that a < x < b and the interval (a, b) (i.e. the set
{y ∈ X : a < y < b}) is contained in A. Assume for the moment that X has no
smallest and no largest element. It is easy to show that the collection of open sets is
indeed a topology. It is called the order topology induced by the order ≤. For the real
line R, the topology induced by the usual ordering coincides with the usual topology.
However by selecting the set and the ordering ≤ suitably one can construct many
strange spaces. As an example, compare the usual topology on the plane R2 with the
topology induced by the lexicographic ordering on it.

7. Let X be a set. A subset A of X is said to be cocountable if its complement X \ A is


countable. Let J consist of all cocuntable subsets of X and the empty set. Then J
is a topology on X. This topology is called cocountable topology on X.

8. Let N be set of all positive integers. Consider the cartesian product N × N. Let ∞ be
any point not in N × N and let X = (N × N) ∪ {∞}. We will define a topology on
X. Let I1 be the power set P (N × N). Clearly I1 ⊂ P (X). Let I2 be the collection

4
of those subsets A of X such that ∞ ∈ A and A contains almost all points in almost
all rows (the term almost means exception of finitely few). Now, let I = I1 ∪ I2 .
Then I is a topology on X.

Definition 3. The topology J1 is said to be weaker (or coarser) than the topology J2 (on
the same set) if J1 ⊂ J2 as subsets of the power set. In this case we also say that J2 is
stronger (or finer) than J1 .

Remark : The indiscrete topology is the smallest or weakest or coarsest of all while the
discrete topology is the largest or the finest or the strongest of all topologies on the same
set.

Theorem 1.2.1. Let X be a set and {Ji |i ∈ I} be a indexed family of topologies on X.


Then J = ∩i Ji is a topology on X. Also J is weaker than each Ji , i ∈ I.

Proof. Let us first verify that J is a topology on X. Clearly the empty set φ belongs to
each J since each Ji is topology on X and so φ ∈ ∩i Ji , i.e. φ ∈ J . Similarly X ∈ J .
Next we show that J is closed under finite intersections. For this let A1 , A2 , ...., An ∈ J
and suppose A = ∩i Ai . To show A ∈ J . Now, since J = ∩i Ji , each Ai ∈ Ji . But Ji
being a topology on X, is closed under finite intersections. So, A ∈ Ji for each i. Hence
A ∈ J . The proof that J is closed under arbitrary unions is similar and left to the reader.
The rest of the theorem is just a general property of intersections.

Corollary 1.2.1.1. Let X be a set and D a family of subsets of X. Then there exists a
unique topology J on X, such that it is the smallest topology on X containing D.

Proof. Consider the collection of all topologies on X which contain D. This family is
non-empty, for the discrete topology surely contains D. Now let J be the intersection of
the members of this collection. Applying the theorem above, J is a topology on X, it
contains D and clearly it is the smallest topology containing D, for any such topology will
be a member of the collection of topologies just considered and hence stronger than its
intersection.

Remark : The topology J so obtained is said to be generated by the family D.

Exercise :

1. Prove that in the co-countable topology, the only convergent sequences are those which
are eventually constant.
2. Prove that the usual topology on the euclidean plane R2 is strictly weaker than the topol-
ogy induced on it by the lexicographic ordering.
3. What difficulty would arise in defining the order topology if either had a smallest or a

5
largest element ?

Rough space :

6
1.3 Bases and Sub-bases
Definition 4. Let (X, J ) be a topological space. A subfamily B of J is said to be a base
for J if every member of J can be expressed as the union of some members of B.

In a metric space every open set can be expressed as a union of open balls and conse-
quently the family of all open balls is a base for the topology induced by the metric. It is
not necessary to take all open ball, rather the family of all open balls of rational radii is also
a base. Indeed it suffices to take balls of radius 1/n, n ∈ N.

Proposition 1.3.1. Let (X, J ) be a topological space and B ⊂ J . Then B is a base for J
iff for any x ∈ X and any open set G containing x, there exists B ∈ B such that x ∈ B and
B ⊂ G.

Proof. First suppose B is a base for J . Let x ∈ X and let an open set G containing x
be given. Then G can be written as the union of some members of B, say, G = ∪i∈I Bi ,
where I is an index set and Bi ∈ B for all i ∈ I. Since x ∈ G, there exists j ∈ I such
that x ∈ Bj . We take this Bj as the set B required in the assertion. Conversely suppose
the given condition holds. Let H be an open set in X, i.e. H ∈ J . For each x ∈ H, there
exists Bx ∈ B such that x ∈ Bx and Bx ⊂ H. Clearly H = ∪x∈H Bx . Thus every member
of J can be expressed as the union of some members of B. So B is a base for J .

Definition 5. A space is said to satisfy the second axiom of countability or is said to be


second countable if its topology has countable base.

Definition 6. A family U of sets is said to be a cover(or covering) of a set A if A is contained


in the union of members of U. A subcover of U is subfamily V of U which itself is a cover
of A. If we are in a topological space then a cover is said to be open if all its members are
open.

Theorem 1.3.2. If a space is second countable then every open cover of it has a countable
subcover.

Proof. Let (X, J ) be a space with a countable base B and let U be a given open cover of
X. First enumerate B as {B1 , B2 , B3 , ...}. Now let

S = {n ∈ N : Bn is contained in some member of U}.

For each n ∈ S, fix Un ∈ U such that Bn ⊂ Un . Now let C = {Bn : n ∈ S} and


V = {Un : n ∈ S}. Clearly V a countable sub-family of J and covers X if V does. So the
theorem will be proved if we show C is a cover of X. For this let x ∈ X. Then x ∈ U for
some U ∈ U. By proposition above, there is some k ∈ N such that x ∈ Bk and Bk ⊂ U .
Clearly then, k ∈ S and so Bk ∈ C. So C and consequently V is a cover of X.

7
Proposition 1.3.3. Let J1 , J2 be two topologies for a set having bases B1 and B2 respec-
tively. Then J1 is weaker than J2 iff every member of B1 can be expressed as a union of
some members of B2 .

If B is a base for a topology J on a set X, then B generates J , i.e. J is the smallest


topology containing B. It is natural to inquire whether we can start with an arbitrary family
B of subsets of a set X and find a topology J on X for which B will be a base. Of course
in case such a topology exists, it must be unique. But in general no such topology exists.
The following proposition tells precisely which families can be bases for topologies.

Proposition 1.3.4. Let X be a set and B a family of its subsets covering X. Then the
following statements are equivalent:
1. There exists a topology on X with B as a base.
2. For any B1 , B2 ∈ B, B1 ∩ B2 can be expressed as the union of some members of B.
3. For any B1 , B2 ∈ B and x ∈ B1 ∩ B2 , there exists B3 ∈ B such that x ∈ B3 and
B3 ⊂ B1 ∩ B2 .

Proof. (1) =⇒ (2). Suppose there exists a topology J on X for which B is a base. Let
B1 , B2 ∈ B. Then B1 , B2 ∈ J and so B1 ∩ B2 ∈ J since J is closed under finite intersec-
tions. So, by definition of a base, B1 ∩ B2 can be expressed as the union of some members
of B.
The proof of the equivalence of (2) and (3) resembles that of 1.3.1 and is left as an exercise.

It only remains to show (3) =⇒ (1). Assume that the condition in (3) holds and define
I = {G ⊂ X : for all x ∈ G, there exists B ∈ B such that x ∈ B and B ⊂ G}. We assert
that J is a topology on X. Clearly φ ∈ J while X ∈ J since B is given to be a cover of
X. That J is closed under arbitrary unions is self-evident. It only remains to verify that
whenever G, H ∈ J , G ∩ H ∈ J . Let x ∈ G ∩ H. Then x ∈ G and x ∈ H. So there exist
B1 , B2 ∈ B such that x ∈ B1 , B1 ⊂ G, x ∈ B2 and B2 ⊂ H. Then x ∈ B1 ∩ B2 . So by
(3), there exists B3 ∈ B such that x ∈ B3 and B3 ⊂ B1 ∩ B2 . But B1 ∩ B2 ⊂ G ∩ H. So
G ∩ H ∈ I. Thus I is a topology on X and it follows that B is a base for J .

Corollary 1.3.4.1. If B is a cover of X and B is closed under finite intersections then B is


a base for a (unique) topology J on X. Moreover, J consists precisely of those subsets of
X which can be expressed as unions of subfamilies of B.

Definition 7. A family S of subsets of X is said to be a sub-base for a topology J on X if


the family of all finite intersections of members of S is a base for J .

Any base for a topology is also a sub-base for the same. In general, however, a sub-base
can be chosen to be much smaller than a base. For example, for the usual topology on R,
the family of all open intervals of the form (a, ∞) or (−∞, b) for a, b ∈ R (or Q) is a
sub-base.

8
Theorem 1.3.5. Let X be a set, J a topology on X and S a family of subsets of X. Then
S is a sub-base for J iff S generates J .

Proof. Let B be the family of finite intersections of members of S. Suppose first that S is
a sub-base for J . We want to show that J is the smallest topology on X containing S.
Now since S ⊂ B and B ⊂ J we at least have that J contains S. Suppose U is some
other topology on X such that S ⊂ U. We have to show that J ⊂ U. Now since U is
closed under finite intersections and S ⊂ U, U contains all finite intersections of members
of S i.e. B ⊂ U. But again since U is closed under arbitrary unions and each member of
J can be written as union of some members of B (by definition of a base), it follows that
J ⊂ U. Conversely suppose J is the smallest topology containing S. We have to show
that S is a sub-base for J , i.e. that B is a base for J . Clearly B ⊂ J is closed under finite
intersections and S ⊂ J . Since B is closed under finite intersections we know by corollary
1.3.4.1 above that there is a topology U on X such that B is a base for U. Every member
of U can be expressed as a union of a sub-family of B and so is in J since B ⊂ J . This
means U ⊂ J and consequently U = J since J is the smallest topology containing S.
Thus B is a base for J and S is a sub-base for J .

Theorem 1.3.6. Given any family S of subsets of X, there is a unique topology J on X


having S as a sub-base. Further, every member of J can be expressed as the union of sets
each of which can be expressed as the intersection of finitely many members of S.

Proof. The first assertion follows from the last theorem. To prove that every member of J
has the desired form, let U consist of all subsets of X which can be expressed as unions of
members of B where B is the family of finite intersections of members of S. By corollary
1.3.4.1 above, U is the unique topology having B as a base. Hence J = U.

Let n be a positive integer, {(X, Ji ) : i = 1, 2, 3, ..., n} be a family of topological


spaces and X be the cartesian product X1 × X2 × ... × Xn . We shall define a certain topol-
ogy, called the product topology on X. By an open box in X we mean a set of the form
V1 × V2 × ... × Vn where Vi ∈ Ji for 1 = 1, 2, ..., n. The entire set X is clearly an open box.
It is also evident that the intersection of two open boxes is again an open box, although their
union need not be. By corollary 1.3.4.1, the family of open boxes is a base for a unique
topology J on X. This topology is called the product topology on X, the space (X, J ) is
called the topological product of the spaces (X1 , J1 ), (X2 , J2 ), ......, (Xn , Jn , ). The space
(Xi , Ji ) is called the ith co-ordinate space or the ith factor of (X, J ).
Let J1 denote the product topology on Rn while let J2 denote the usual topology on it
(induced by the euclidean metric on Rn ). Then we have J1 = J2 . The idea behind the
proof is to show that J1 ⊂ J2 and J2 ⊂ J1 , which will imply that J1 = J2 . Students are
encouraged to go through the details of the proof, which is given in the textbook.

9
Exercise :

1. Prove that if a space (X, J ) has a base B of cardinality a then the cardinality of J
cannot exceed 2a .
2. Prove that a space is second countable if and only if it has a countable sub-base.
3. Let X1 and X2 be two topologies. Prove that the metric topology in X1 × X2 coincides
with the product topology on it.
4. Prove that if each space (Xi , Ji ) is second countable, for i = 1, 2, ..., n then so is their
topological product.

Rough space :

10
1.4 Closed sets and closure
Definition 8. Let (X, J ) be a topological space. Then a subset A of X is said to be closed
in X if its complement X \ A is open in X.

Note : The empty set and the whole set are always open as well as closed in every space.
On the other hand, the set of rationals is neither open nor closed in the usual topology on
the real line. A set which is both open and closed is sometimes called a clopen set.

Theorem 1.4.1. Let C be the family of all closed sets in a topological space (X, J ). Then
C has the following properties:
i. φ ∈ C, X ∈ C.
ii. C is closed under arbitrary intersections.
iii. C is closed under finite unions.
Conversely, given any set X and a family C of its subsets which satisfies these three proper-
ties, there exists a unique topology J on X such that C coincides with the family of closed
subsets of (X, J ).

Proof. The first part follows trivially from the definition of a topology and De Morgan’s
laws. The converse part is equally trivial once it is clearly understood what it says. Here we
are given a set X and some collection C of its subsets. We are given that properties (i) to (iii)
hold for C. The theorem says that given such a family C ⊂ P (X) we can define a suitable
topology J on X such that members of C are precisely the closed subsets of X (w.r.t. the
topology J ), and that such a topology is unique. Having understood what the theorem says,
the proof itself is trivial as we have no choice but to let J consist of complements (in X)
of members of C, i.e. J = {B ⊂ X : X \ B ∈ C}. That J is a topology on X follows
by applying De Morgan’s laws. The open subsets of X are precisely the complements
of members of C, and hence the closed subsets of X are precisely the members of C as
asserted. Also this condition determines J uniquely.

Definition 9. The closure of a subset of a topological space is defined as the intersection


of all closed subsets containing it. In symbols, if A is a subset of a space (X, J ), then its
closure is the set ∩{C ⊂ X : C closed in X, A ⊂ C}. It is denoted by A.

Proposition 1.4.2. Let A, B be subsets of a topological space (X, J ).


i. A is a closed subset of X. Moreover it is the smallest closed subset of X containing A
i.e. if C is closed in X and A ⊂ C then A ⊂ C.
ii. φ = φ.
iii. A is closed in X iff A = A.
iv.A = A.
v. A ∪ B = A ∪ B.

11
It is instructive to reformulate some of the properties in the last proposition using the
terminology of operators. An operator is just another name for a function, except that the
term is generally reserved for those functions whose domains are sets of sets or of functions.
For example, if (X, J ) is a topological space, then the closure operator associated with it
is defined as the function c : P (X) → P (X) such that C(A) = A for each A ∈ P (X). In
terms of the closure operator, (ii) to (v) assume the following forms respectively:
ii0 . φ is a fixed point of C.
iii0 . The fixed points of C are precisely the closed subsets of X.
iv 0 . C is idempotent, i.e. C ◦ C = C.
v 0 . C commutes with finite unions.

Now comes an important question. Suppose we have an abstract operator θ : P (X) →


P (X). Under what conditions can we find a topology on X whose closure operator will
coincide with the given operator θ ?

Theorem 1.4.3. Let X be a set, θ : P (X) → P (X) a function such that


1. for every A ∈ P (X), A ⊂ θ(A) (this condition is sometimes expressed by saying that θ
is an expansive operator),
2. φ is a fixed point of θ,
3. θ is idempotent, and
4. θ commutes with finite unions.
Then there exists a unique topology J on X such that J coincides with the closure operator
associated with J . Conversely, any closure operator satisfies these properties.

Proof. The converse part is already established. For the direct implication, suppose θ :
P (X) → P (X) satisfies (1) to (4). We want to find a topology J on X such that for every
A ⊂ X, θ(A) = A. If at all such a topology exists then its closed subsets must be precisely
the fixed points of θ. This gives us a clue to the construction of J . We let C = {A ⊂ X :
θ(A) = A} and contend that C has properties (i) to (iii) of Theorem 1.4.1. Condition (2)
shows that φ ∈ C while condition (4) implies that C is closed under finite unions. To prove
that X ∈ C, we merely note that by (1), X ⊂ θ(X) and hence X = θ(X) since θ(X) ⊂ X
anyway. It only remains to verify that C is closed under arbitrary intersections. For this
we first note that θ is monotonic, i.e., whenever A ⊂ B, θ(A) ⊂ θ(B), which follows by
writing B as A ∪ (B \ A) and applying (4). Now let A = ∩i∈I Ai where I is an index set
and Ai ∈ C for each i ∈ I. We want to show that A ∈ C, i.e. θ(A) = A. By (1) we
already know A ⊂ θ(A). Also θ(A) ⊂ θ(Ai ) for each i ∈ I since θ is monotonic, and so
θ(A) ⊂ ∩i∈I (Ai ). But θ(Ai ) = Ai , since Ai ∈ C for all i ∈ I. Consequently, θ(A) ⊂ A
and hence θ(A) = A as desired. So by Theorem 1.4.1, the family J of complements of
members of C is a topology on X.
It remains to be verified that the closure operator associated with J coincides with θ. Let

12
A ⊂ X. Then A w.r.t. J is the intersection of all closed subsets of X containing A. But by
construction, closed subsets of X are precisely the fixed points of θ. Hence A = ∩{B ⊂
X : A ⊂ B; θ(B) = B}. Now, whenever A ⊂ B, θ(A) ⊂ θ(B) by monotonicity of θ.
So if A ⊂ B and θ(B) = B then θ(A) ⊂ B. But A is the intersection of such B’s and so
θ(A) ⊂ A. For the other way inclusion we note that by condition (3), θ(A) ∈ C while by
(1) A ⊂ θ(A), hence A ⊂ θ(A), A being the smallest member of C containing A. Hence
for all A ⊂ X, θ(A) = A completing the proof.

Definition 10. Let A be a subset of a space X. The A is said to be dense in X if A = X.


Trivially, the entire set X is always dense in itself.

Proposition 1.4.4. A subset A of a space X is dense in X iff for every nonempty open
subset B of X, A ∩ B 6= φ.

Proof. Suppose A is dense in X and B is a non-empty open set in X. If A ∩ B = φ then


A ⊂ X \ B. Hence A ⊂ X \ B since X \ B is closed. But then X \ B ⊂ X contradicting
that A = X. So A ∩ B 6= φ. Conversely assume that A meets every non-empty open subset
of X. This clearly means that the only closed set containing A is X and consequently
A = X.

Remark : In the topological space of real line with the usual topology the set Q of all
rational numbers, as well as its complement R \ Q are both dense.

Exercise :

1. Let (X, J ) be a topological space and C be the family of all closed subsets of X. Prove
or disprove that C is the complement of J in P (X).
J1
2. Let J1 , J2 be topologies on a set X and A ⊂ X. If J1 is weaker than J1 how are A
J2
and A related to each other?
Y X
3. If X is a space, Y ⊂ X and A ⊂ Y , prove that A = A ∩ Y .
4. Prove that a second countable space always contains a countable dense subset.
5. Find out the dense subsets of discrete, indiscrete and cofinite spaces.

Rough space :

13
14
1.5 Neighbourhoods, Interior and Accumulation Points
Definition 11. Let (X, J ) be a topological space, x0 ∈ X and N ⊂ X. Then N is said to
be a neighbourhood of x0 or x0 is said to be an interior point of N (each w.r.t. J ) if there
is an open set V such that x0 ∈ V and V ⊂ N . The word ‘neighbourhood’ is sometimes
abbreviated to ’nbd’.

Proposition 1.5.1. A subset of a topological space is open iff it is a neighbourhood of each


of its points.

Proof. Let X be a topological space and G ⊂ X. First suppose G is open. Then evidently
G is a nbd of each of its points. Conversely suppose G is a nbd of each of its points. Then
for each x ∈ G, there is an open set Vx such that x ∈ Vx and Vx ⊂ G. Clearly then,
G = ∪x∈G Vx , since each Vx is open so is G.

Note : Trivially if N is a nbd of a point x then so is any superset of N . It is also easy to


show that the intersection of any two (and hence finitely many) neighbourhoods of a point
is again a neighbourhood of that point.

Definition 12. Let (X, J ) be a space and A ⊂ X. Then the interior (or more precisely the
J -interior) of A is defined to be the set of all interior points of A, i.e. the set {x ∈ A :
A is a nbd of x}. It is denoted by A0 or int(A), or intJ (A) when we want to emphasise
its dependence upon J .

Proposition 1.5.2. Let X be a space and A ⊂ X. Then int(A) is the union of all open sets
contained in A. It is also the largest open subset of X contained in A.

Proof. Let U be the family of all open sets contained in A (U is nonempty since φ ∈ U).
Let V = ∪G∈U G. We have to show V = int(A). Now if x ∈ V then x ∈ G, for some
G ∈ U. This means A is nbd of x and so x ∈ int(A). Conversely, let x ∈ int(A). Then
there is an open set H such that x ∈ H and H ⊂ A. But then, H ∈ U and so H ⊂ V .
So x ∈ V . This proves the first assertion of the proposition, and also shows that int(A)
is an open set contained in A. To see it is the largest such set, supposed G is an open set
contained in A. Then G ∈ U and so G ⊂ int(A) by the first assertion.

Definition 13. The exterior of a set is defined as the interior of its complement. It is not
hard to show that it always coincides with the complement of the closure of the original set.
Consequently, it is not an independent concept and does not appear often in topology.

Definition 14. Let X be a space and x ∈ X. Let Nx be the set of all neighbourhoods of x
in X (w.r.t. the given topology on X). The family Nx is called the neighbourhood system
at x.

15
Proposition 1.5.3. Let X be a space and for x ∈ X, let Nx be the neighbourhood system
at x. Then,
i. If U ∈ Nx , then x ∈ U .
ii. For any U, V ∈ Nx , then U ∩ V ∈ Nx .
iii. If V ∈ Nx and V ⊂ U , then U ∈ Nx .
iv. A set G is open in X iff G ∈ Nx for all x ∈ G.
v. If U ∈ Nx then there exists V ∈ Nx such that U ⊂ V and V ∈ Ny for all y ∈ V .

Theorem 1.5.4. Let X be a set and suppose for each x ∈ X, a non-empty family Nx of
subsets of X is given satisfying (i), (ii), (iii) and (v) in the proposition above. Then there is
a unique topology J on X such that for each x ∈ X, Nx coincides with the family of all
neighbourhoods of x w.r.t. J .

Proof. If at all such a topology exists, then property (iv) gives us a clue for its construction.
We let J = {U ⊂ X : U ∈ Nx f or all x ∈ U } and claim J is a topology on X. Clearly
φ ∈ J . To show that X ∈ J note that for any x ∈ X, X ∈ Nx by (iii) as X is a superset
of any member of Nx . Property (ii) shows that J is closed under finite intersections while
using (iii) it follows easily that J is closed under arbitrary unions. So J is a topology for
X. Note that so far we did not use (i) and (v). With our definition of J , (v) means that
for any x ∈ X and U ∈ Nx , there exists an open set V (i.e. a member of J ) such that
V ∈ J and V ⊂ U . From (i) we now get that x ∈ V . Thus U contains a member of J
containing x and is therefore a J -neighbourhood of the point x. Hence every member of
Nx is neighbourhood of x w.r.t. J . Conversely let U be a neighbourhood of x w.r.t. J .
Then U contains a member V of J such that x ∈ V and V ⊂ U . But V ∈ Nx by the
definition of J and so by (iii) U ∈ Nx . Thus neighbourhoods of x w.r.t. J are precisely
the members of Nx for each x ∈ X and this completes the proof.

Definition 15. Let A be a subset of a topological space X and y ∈ X. Then y is said to be


an accumulation point of A if every open set containing y contains at least one point of A
other than y.

In a discrete space no point is an accumulation point of any set while at the other ex-
treme, in an indiscrete space, a point y is an accumulation point of any set A provided only
that A contains atleast one point besides y. In the usual topology on the real line, every real
number is an accumulation point of the set of rational numbers while the set of integers has
no point of accumulation.

Definition 16. Let A be a subset of a space X. Then the derived set of A, denoted by A0 , is
the set of all accumulation points of A in X. Obviously A0 depends not only on A but also
on the topology under consideration.

Theorem 1.5.5. For a subset A of a space X, A = A ∪ A0 .

16
Proof. First we claim that A ∪ A0 is closed or that X \ (A ∪ A0 ) is open. We do so by
showing that X \ (A ∩ A0 ) is a nbd of each of its points. Let y ∈ X \ (A ∪ A0 ). Then since
y is not a point of accumulation of A, there exists an open set V containing y such that V
contains no point of A except possibly y. But y ∈ / A, so we have A ∩ V = φ. We claim
0 0
A ∩ V is also empty. For, let z ∈ A ∩ V . Then V is an open set containing z which is an
accumulation point of A. So V ∩ A is nonempty, a contradiction. So A ∩ V = φ and hence
V ⊂ X \ (A ∪ A0 ). This proves that A ∩ A0 is closed and since it obviously contains A, it
also contains A i.e. A ⊂ A ∪ A0 .
For the other way inclusion, it suffices to show that A0 ⊂ A since we already have A ⊂ A.
So let y ∈ A0 . If y ∈
/ A then y ∈ X \ A which is an open set since A is always a closed set.
But y is an accumulation point of A. So (X \ A) ∩ A 6= φ which is a contradiction since
(X \ A) ⊂ (X \ A) .So y ∈ A0 . This completes the proof.

Theorem 1.5.6. For a subset A of a space X, A = {y ∈ X : every nbd of y meets A non−


vacuously}.

Proof. Let B = {y ∈ X : U ∈ Ny =⇒ U ∩ A 6= φ}. We have to show that B = A.


By the theorem above, this amounts to showing that A ∪ A0 = B. First let y ∈ A ∪ A0 . If
y ∈ A then certainly every nbd of y meets A at least at the point y and so y ∈ B. If y ∈ A0
then too, by the definition of an accumulation point, every nbd of y contains a point of A
and so y ∈ B. Thus A ∪ A0 ⊂ B. Conversely let y ∈ B. If y ∈ / A ∪ A0 , then y ∈
/ A and
so X \ A is a nbd of y which does not meet A, contradicting that y ∈ B. So B ⊂ A ∪ A0 .
Thus B = A.

Definition 17. Let A be a subset of a space X. Then its boundary or frontier is the set
A ∩ (X \ A). It is immediate from the definition that the boundary is always a closed set
and that the boundary of a set is the same as the boundary of its complement. The boundary
of a set A is generally denoted by ∂A or F (A).

Exercise :

1. Prove that the interior of a set is the same as the complement of the closure of the
complement of the set, i.e. for a subset A of a space X, int(A) = X \ (X \ A).
2. Prove that in a metric space X, a point y is in the closure of a set A iff there exists a
sequence {xn } such that xn ∈ A for all n and {xn } converges to y in X.
3. For a set A in a space X, prove that A is the disjoint union of int(A) with the boundary
of A.
4. Prove that a set is closed iff it contains its boundary and that it is open iff it is disjoint
from its boundary.
5. Let X1 , X2 be topological spaces and X their topological product. Let A1 ⊂ X1 ,
A2 ⊂ X2 . Prove that

17
∂(A1 × A2 ) = (A1 × ∂A2 ) ∪ (∂A1 × A2 )

6. Prove or disprove that in a metric space, a closed ball is the closure of the open ball with
the same centre and radius.

Rough space :

18
1.6 Continuity and Related Concepts
Definition 18. Let f : X → Y be a function and x0 ∈ X and J , U be topologies on X, Y
respectively. Then f is said to be continuous (or more precisely J − U continuous) at x0 if
for every V ∈ U such that f (x0 ) ∈ V ,there exists U ∈ J such that x0 ∈ U and f (U ) ⊂ V .

Proposition 1.6.1. With the notation above, the following statements are equivalent.
1. f is continuous at x0 .
2. The inverse image (under f ) of every neighbourhood of f (x0 ) in Y is a neighbourhood
of x0 in X.
3. For every subset A ⊂ X, x0 ∈ A implies f (x0 ) ∈ f (A).

Definition 19. Let f : X → Y be a function and J , U be topologies on X, Y respectively.


Then f is said to be continuous (or J − U continuous) if it is continuous at each point of
X.

Theorem 1.6.2. Let (X, J ), (Y, U) be spaces and f : X → Y a function. Then the
following statements are equivalent:
1. f is continuous (i.e. J − U continuous).
2. For all V ∈ U, f −1 (V ) ∈ J .
3. There exists a sub-base S for U such that f −1 (V ) ∈ J for all V ∈ S.
4. For any closed subset A of Y , f −1 (A) is closed in X.
5. For all A ⊂ X, f (A) ⊂ f (A).

Proof. (1) ⇐⇒ (2): Assume f is continuous at each point of X and V is an open subset
of Y . If x0 ∈ f −1 (V ) then f (x0 ) ∈ V . But V is then a neighbourhood of f (x0 ) and
so, by continuity at x0 , f −1 (V ) is a neighbourhood of x0 . Thus f −1 (V ) is a nbd of each
of its points and so is an open set in X. Conversely suppose (2) holds and let x0 ∈ X.
Given any open set V containing f (x0 ). f −1 (V ) is an open set containing x0 and moreover
f (f −1 (V )) ⊂ V . This shows that f is continuous at x0 .
(2) ⇐⇒ (3): (2) clearly implies (3). For the converse, we note that the inverse image
preserves intersections and unions. Let V ∈ U. Then V can be written as ∪i∈I Vi where
I is an index set and for each i ∈ I, Vi can be written as the intersection of finitely many
members of S say S1i ∩ S2i ∩ .... ∩ Srii , where ri ∈ N and Sji ∈ S for 1 ≤ j ≤ ri . Then
f −1 (Sji ) ∈ J . But J is closed under finite intersections and so f −1 (Vi ) ∈ J . Further
f −1 (V ) ∈ J since f −1 (V ) = ∪i∈I f −1 (Vi ) and J is closed under arbitrary unions. Thus
(2) holds.
The equivalence of (2) with (4) follows from the fact that the inverse image preserves com-
plements, i.e. that for any B ⊂ Y , f −1 (Y \ B) = X \ f −1 (B).

Proposition 1.6.3. Compositions of continuous functions are continuous. More specifically,


if f : X → Y, g : Y → Z are functions, with f continuous at x0 and g continuous at f (x0 )

19
then g ◦ f is continuous at x0 . The identity function on any space is continuous. More
generally, if J , U are topologies on a set X then the identity function idx : X → X is
J − U continuous iff the topology J is stronger than U. Any function from a discrete space
and any function into an indiscrete space is continuous. Finally if (X, J ) is a space and
(Y, J /Y ) a subspace then the inclusion function i : Y → X is J − J /Y continuous. A
restriction of a continuous function is continuous.

Definition 20. Let X1 , X2 , ..., Xn be sets and let X = X1 × X2 × ... × Xn . For each
i = 1, 2, ..., n, define πi : X → Xi by πi (x1 , x2 , ..., xn ) = xi . Then πi is called the
projection on Xi , or the i-th projection. It is a surjective function except in the case where
some other Xj and hence X is empty. If x ∈ X then πi (x) is called the i-th coordinate of
x.

Theorem 1.6.4. Let {(Xi , Ji ) : i = 1, 2, ..., n} be a collection of topological spaces and


(X, J ) their topological product. Then each projection πi is continuous. Moreover, if Z
is any space then a function f : Z → X is continuous if and only if πi ◦ f : Z → Xi , is
continuous for all i = 1, 2, , .., n.

Proof. In order to prove that πi is continuous, let V ∈ Ji . We have to show that πi−1 (V ) ∈
J . But it is easy to check that πi−1 (V ) = X1 × X2 × ... × Xi−1 × V × Xi+i × ... × Xn
because to say that πi (x) ∈ V , for x ∈ X puts no restriction on other coordinates of x,
only the i-th coordinate is required to lie in V . Recalling the definition of product topology
πi−1 (V ) is a member of the defining base and so is open in X. Hence each projection is a
continuous function.
For the second part, suppose f : Z → X is continuous. Then for each i, πi ◦f is continuous
as compositions of continuous functions are continuous. Conversely, suppose each of π1 ◦
f : Z → Xi is continuous. Denote πi ◦ f by fi . To prove that f is continuous, it suffices
to prove that the inverse image of any member of a base for (X, J ) is an open subset of Z.
Now, by definition of the product topology, a base for J consists of all sets V of the form
V1 × V2 × ... × Vn where Vi ∈ Ji for i = 1, 2, ..., n. It is then immediate that for z ∈ Z,
f (z) ∈ V iff πi (f (z)) ∈ Vi , for all i = 1, 2, ..., n. In other words, f −1 (V ) = ∩ni=1 fi−1 (Vi ).
But each fi−1 (Vi ) is open since fi is assumed to be continuous. Hence f −1 (V ), being the
intersection of finitely many open sets, is open in Z. This proves that f is continuous and
completes the proof.

The preceding theorem is of immense use in checking the continuity of functions into
euclidean spaces. As a euclidean space is the topological product of a finite number of
copies of the real line (with usual topology), the problem reduces to checking the continuity
of real-valued functions.

20
Definition 21. Let X, Y by spaces. A function f : X → Y is said to be open (respectively
closed) if whenever A is an open (resp. a closed) subset of X, f (A) is open (resp. a closed)
subset of Y .

Definition 22. A homeomorphsim from a space X to space Y is a bijection f : X → Y


such that both f and f −1 are continuous. When such a homeomorphism exists, X is said to
be homeomorphic to Y .

Proposition 1.6.5. Let X, Y be topological spaces and f : X → Y a function. Then the


following statements are equivalent:
1. f is a homeomorphism,
2. f is a continuous bijection and f is open,
3. f is a bijection and f −1 , f are both open map,
4. there exists a function g : Y → X such that f, g are continuous, g ◦ f = idx and
f ◦ g = idy .

Definition 23. Let (X, J ), (Y, U) be topological spaces. An embedding (or imbedding) of
X into Y is a function e : X → Y which is a homeomorphism when regarded as a function
from (X, J ) onto (e(X), U/e(X)).

Proposition 1.6.6. A function e : X → Y is an embedding iff it is continuous and one-


to-one and for every open set V ∈ X, there exists an open subset W of Y such that
e(V ) = W ∩ e(X).

Inclusion maps are the most immediate examples of embeddings and as the definition
implies, these are the only examples upto homeomorphisms. An important problem in
topology is to decide when a space X can be embedded in another space Y i.e. when there
exists an embedding from X into Y . This is called the embedding problem. Theorems
asserting the embeddability of a space into some other space which is more manageable
than the original space are known as embedding theorems.

Exercise :

1. Let (X, J ), (Y, U) be topological spaces such that every function from X to y is J − U
continuous. Prove that either J is the discrete topology or U is the indiscrete topology.
2. Let (X, J1 ), (X2 , J2 ) be topological spaces. Prove that X1 × X2 is homeomorphic to
X2 × X1 each being given the product topology.
3. Let f : X → Y be a function. The graph of f is defined to be the set G = {(x, f (x)) :
x ∈ X}. Prove that if f is continuous then G is homeomorphic to X.
4. Let S 1 be the unit circle in the plane, with the relative topology on it. Define p : R → S 1
by p(x) = (cos(x), sin(x)). Prove that p is continuous.
5. Prove that the set of all homeomorphisms of a space (X, J ) onto itself is a subgroup of

21
the permutation group of X.

Rough space :

22
Module 2

Spaces with special properties

2.1 Making Functions Continuous, Quotient Spaces


Problem 1 : Let {(Yi , Ji ) : i ∈ I} be an indexed family of topological spaces, X any set
and {fi : i ∈ I} an indexed collection of functions such that for each i ∈ I, fi is a function
from X to Yi . What topology J on X will make each fi J − Ji continuous?

Problem 2 : Let {(Yi , Ji ) : i ∈ I} be an index family of topological spaces, X any


set and {fi : i ∈ I} an indexed family of functions from Y , into X. What topology U on
X will make each fi Ji − U continuous?

Either problem can be obtained from the other by reversing the directions in which the
functions go. For this reason, the two problems are said to be dual to each other. Their du-
ality will be reflected through their solutions. A trivial solution to the first problem would
be to let J be the discrete topology on X. This is hardly satisfactory. We therefore look for
a better solution. Note that if a topology J on X renders each fi continuous, then so will
any topology stronger than J . The idea then is to find as small a topology on X as possible
which will make each fi continuous. Dually, a trivial solution to the second problem would
be to let U be the indiscrete topology on X. If a topology U on X renders each fi contin-
uous then so will any topology weaker than U. The idea then is to find as large a topology
on X as possible which will make each function fi continuous.

Theorem 2.1.1. With the notation of Problem 1, there exists a unique smallest topology J
on X which makes each fi continuous.

Proof. First note that if D is any topology on X for which fi : X → Yi is D−Ji continuous,
then D must contain all sets of the form fi−1 (Vi ) as Vi ranges over all open subsets of Yi .
This gives us a clue to the construction of the smallest such topology. We let S = {fi−1 (Vi ) :
Vi ∈ Ji , i ∈ I}. Then S is a collection of subsets of X. S itself need not be a topology for
X. However we let J be the topology generated by S, that is, the topology having S as a

23
sub-base. Then clearly each fi is J − Ji continuous. Further if D is any other topology on
X which makes each fi continuous, then as we just observed S ⊂ U and so J ⊂ U. Thus
J is the smallest topology on X which makes each fi continuous.

Definition 24. With the notations above, the weakest topology on X making each fi con-
tinuous is called the weak topology determined by the family of functions {fi : i ∈ I}.

Theorem 2.1.2. With the notation of Problem 2, there exists a unique largest topology U
on X which makes each fi continuous.

Proof. Let D be a topology on X which makes each fi continuous. Then for any D ∈ D,
and i ∈ I, fi−1 (D) is open in Yi . This puts a restriction on how large D can be and gives us
a clue to the construction of the best possible solution. We let U = {A ⊂ X : fi−1 (A) ∈
Ji f or all i ∈ I}. Since the inverse images commute with intersections and unions, it
is easy to show that U is a topology on X and by what we said earlier, it is the strongest
topology on Y making each fi continuous.

Definition 25. With the notations above, the strongest topology on X making each fi con-
tinuous is called the strong topology determined by the family of functions {fi : i ∈ I}.

We shall now discuss few classic situations in which the problem of finding the weak
topology is important.

Theorem 2.1.3. The product topology is the weak topology determined by the projection
functions.

Proof. First, let {(X, Ji ) : i = 1, 2, ..., n} be a collection of topological spaces. Let X =


X1 × X2 × ..... × Xn , and let πi : X → Xi be the i-th projection function for i = 1, 2, ..., n.
We claim that J is in fact the smallest topology which makes each πi continuous. For
suppose U is any other topology on X such that each πi is U − Ji continuous. Let V be
an open box in X, say, V = V1 × V2 × .... × Vn where Vi ∈ Ji for i = 1, 2, ..., n. Then
clearly V = ∩ni=1 πi−1 (Vi ) . But each πi−1 (Vi ) ∈ U since πi is U − Ji continuous. So V ∈ U
since U is closed under finite intersections. Now the family of all such open boxes is, by
definition, a base for the product topology J on X. Hence J ⊂ U. We have thus proved
the following theorem in the case of finite products.

Theorem 2.1.4. Let X have the weak topology determined by a family {fi : X → Yi :
i ∈ I} of functions where each Yi is a topological space, I being an index set. Then
for any space Z, a function g : Z → X is continuous iff for each i ∈ I, the composite
fi ◦ g : Z → Yi is continuous.

Proof. If g is continuous then so is each fi ◦ g, as the compositions of continuous functions


are continuous. Conversely, recall that the family S of all subsets of the form fi−1 (Vi ),

24
where Vi is open in Yi , i ∈ I is a sub-base for the weak topology on X. So, in order to
show that g : Z → X is continuous, it suffices to show that the inverse image under g of
every member of S is open in Z. Now if i ∈ I and Vi is open in Yi , then g −1 (fi−1 (Vi )) =
(fi ◦ g)−1 (Vi ) which is open by continuity of fi ◦ g. This shows that g is continuous.

Theorem 2.1.5. Let X have the strong topology determined by a family {fi : Yi → X :
i ∈ I} of functions where each Yi is a topological space, I being an index set. Then for any
space Z, a function g : X → Z is continuous iff for each i ∈ I the composite g◦fi : Yi → Z
is continuous.

Proof. The proof is dual to the previous one.

Definition 26. Let (X, J ), (Y, U) be topological spaces and f : X → Y be an onto


function. Then f is said to be a quotient map or Y is said to have the quotient topology
w.r.t. X and f if U is the strong topology generated by the singleton family {f }. In such a
case we also say that Y is a quotient of X.

Each factor space is indeed a quotient space of the topological product of spaces. Let
{Xi , Ji : i = 1, 2, ..., n} be a family of non-empty topological spaces. Let (X, J ) be their
topological product. Then for any i, the projection πi : X → Xi is onto. We claim Ji is the
quotient topology w.r.t. X and πi . Let V ⊂ Xi . If V is open in Xi , then πi−1 (V ) is open in
X by continuity of πi . On the other hand, if πi−1 (V ) is open in X then V = πi (πi−1 (V )) is
open in Xi because the projection functions are open. Thus V is open in Xi iff πi−1 (V ) is
open in X and so J the quotient topology.

Proposition 2.1.6. Every open, surjective map is a quotient map.

Proof. Use the same argument as above.

Proposition 2.1.7. Every closed, surjective map is a quotient map.

Proof. Let f : (X, J ) → (Y, U) be closed, continuous and onto. Let V ⊂ Y . If V is open
in Y , then f −1 (V ) is open in X by continuity of f . On the other hand suppose f −1 (V ) is
open in X. Then X \ f −1 (V ) is closed in J and f (X \ f −1 (V )) is closed in Y because
f is a closed function. Since f is onto, X \ f −1 (V )) equals Y \ V . Thus V is open in Y .
This shows that U is the strong topology determined by J and f . Therefore f is a quotient
map.

It is not true that every quotient map is either open or closed. The class of quotient
maps is quite large inasmuch as any surjective function on any topological space becomes
a quotient map if we put the quotient topology on its codomain.

Definition 27. A topological property is said to be divisible if whenever a space has it, so
does every quotient space of it.

25
Proposition 2.1.8. The property of being a discrete space is divisible. In other words, every
quotient space of a discrete space is discrete.

Proof. Let f : X → Y be a quotient map where X is a discrete space. Let U be the


quotient topology on Y . We have to show that (Y, U) is discrete. Recall that U is the
strongest topology on Y which makes f continuous. But since the domain X is discrete,
any topology on Y renders f continuous. So U is the strongest topology on Y , that is the
discrete topology.

Quotient spaces provide an important tool for the construction of new topological spaces
from old ones. For example, let f : X → Y be a surjective function. Then f determines an
equivalence relation R on X defined by xRy iff f (x) = f (y). The equivalence classes of R
are precisely the inverse images of singleton subsets of Y . Now let D be the collection of all
equivalence classes under R. D is called the quotient set of X by R and is also denoted by
X/R. There is a canonical function p : X− → X/R, called the projection which assigns
to each x ∈ X its equivalence class under R. The function θ : Y → X/R defined by
θ(y) = p(x) for any x ∈ f −1 ({y}) is obviously a well-defined bijection and the following
diagram is commutative.

So far, X was merely a set and f a set-theoretic surjection. Suppose now that X, Y are
topological spaces and that f is a quotient map. On X/R, we put the strong topology gener-
ated by the projection function p. The function θ then becomes continuous as its composite
with f viz., θ ◦ f is continuous. Similarly θ−1 is continuous. Thus θ is now not merely a
bijection but a homeomorphism. Thus, upto a topological equivalence, we may identify the
quotient space Y with the quotient space X/R and the quotient map f with the projection
p.
It follows that given a space X, any quotient space of X can be obtained starting from
an equivalence relation R on X and putting the strong topology on the set of equivalence
classes X/R. This gives us an important tool to generate new topological spaces because
there is absolutely no restriction on the relation R on X other than that it be an equivalence
relation.

26
Exercise :

1. Prove that for a metric space, the weak topology determined by the family of all contin-
uous real-valued functions on it coincides with the metric topology.
2. Let X1 be the set of rational numbers and X2 the set of all irrational numbers, each with
the usual topology. Let i1 , i2 be the inclusion functions of X1 , X2 respectively into R. Find
the strong topology on R which makes i1 , i2 continuous.
3. Let X be a space and U be an open cover of it. Prove that a function f from X into some
space Y is continuous iff for each U ∈ U, the restriction fU is continuous.
4. Prove that the composite of two quotient maps is a quotient map.
5. Define a relation R on R by xRy iff x − y ∈ Q. Prove that R is an equivalence relation
on R and that the quotient space R/R is indiscrete.
6. Obtain the torus surface as a quotient space of the unit square.

Rough space :

27
2.2 Spaces with Special Properties

2.2.1 Smallness Conditions on a Space


Definition 28. A subset A of a space X is said to be a compact (Lindeloff) subset of X
if every cover of A by open subsets of X has a finite (respectively countable) subcover. A
space X is said to be compact (Lindeloff) if X is a compact (resp. Lindeloff) subset of itself.

Proposition 2.2.1. Let (X, J ) be a topological space and A ⊂ X. Then A is a compact


(Lindeloff) subset of X if and only if the subspace (A, J /A) is compact (resp. Lindeloff).

Proof. We give the argument for compactness. Replacing ‘finite’ by ‘countable’ in it, we
could use it for Lindeloff subsets. Suppose first that A is a compact subset of X. Let G
be an open cover of the space (A, J /A). Each member G of G is of the form H ∩ A for
some H ∈ J . For each G ∈ G, fix D(G) ∈ J such that G = D(G) ∩ A. Then the family
{D(G) : G ∈ G} is a cover of A by open subsets of X. Since A is a compact subset of
X, this cover has a finite subcover, say, {D(Gi ) : i = 1, 2, ..., n} where Gi ∈ G for all
i = 1, 2, ..., n. Clearly then, {G1 , G2 , ..., Gn } is a finite sub-cover of G. This shows that the
subspace (A, J /A) is compact.
Conversely suppose the subspace (A, J /A) is compact. Let G be a cover of A by open
subsets of X. Then {G ∩ A : G ∈ G} is an open cover of the space (A, J /A). By
compactness of the space (A, J /A), this cover has a finite subcover, say {Gi ∩ A : i =
1, 2, ..., n} where Gi ∈ G for i = 1, 2, ..., n. Clearly then, {G1 , G2 , ..., Gn } is a finite
subfamily of G covering the set A. Thus A is a compact subset of X.

The above proposition leads to an interesting concept. Suppose we have a set X, a


subset A and two topologies J1 , J2 on X such that J1 /A = J2 /A. If we know that A is a
compact subset of (X, J1 ), nn view of the last proposition, it follow that A is also a compact
subset of (X, J2 ). The proposition above shows that as long as J1 and J2 relativise to the
same topology on A, A is a compact subset of (X, J1 ) iff it is a compact subset of (X, J2 ).
In other words, compactness of a subset depends only on the topology induced on it, the
topology on the entire space is not directly relevant. To stress this point further we note that
a similar statement no longer holds for denseness of a subset. It may very well happen that
two distinct topologies J1 , J2 on a set X induce the same topology on A and still A is a
dense subset of X w.r.t. J1 but not w.r.t. J2 . A trivial example of this occurs when J1 is
the indiscrete topology, J2 is the discrete topology and A is a singleton subset of X.

Definition 29. A property of a subset of a topological space is said to be an absolute


property if it depends only on the relativised topology on that set, otherwise it is called
a relative property. Thus compactness and the property of being a Lindeloff subset are
absolute properties while denseness is a relative property. Similarly being an open set is a
relative property.

28
Definition 30. A space is said to be separable if it contains a countable dense subset.

Theorem 2.2.2. Every second countable space is Lindeloff.

Proof. Straightforward.

Theorem 2.2.3. Every second countable space is separable.

Proof. We already proved this as part of an exercise.

Note :

• The converses of both these theorems hold for metric spaces. In general, however,
the converses fails.

• Every compact space is Lindeloff. The converse is false. The real line with the usual
topology is not compact because the cover by open intervals of the form (−n, n)
where n varies overall positive integers, has no finite subcover. However, it is second
countable and hence Lindeloff.

• Every cofinite space is compact. For, let X be such a space and let U be an open
cover of X. Take any non-empty member U of U. Then X \ U has only finitely many
points say x1 , x2 , ...., xn . For each such xi , choose Ui ∈ U such that xi ∈ Ui . Then
{U, U1 , U2 , ...., Un ) is a finite subcover of U.

Theorem 2.2.4. Every continuous real-valued function on a compact space is bounded and
attains its extrema.

Proof. Let X be a compact space and suppose f : X → R is continuous. First we show


f is bounded. For each x ∈ X, let Jx be the open interval (f (x) − 1, f (x) + 1) and let
Vx = f −1 (Jx ). By continuity of f , Vx is an open set containing x. Note that f is bounded on
each Vx . Now the family {Vx : x ∈ X} is an open cover of X and by compactness, admits
a finite sub-cover say {Vx1 , Vx2 , ..., Vxn }. Let M = max{(f (x1 ), f (x2 ), ..., f (xn )} + 1 and
let m = min{f (x1 ), f (x2 ), ..., f (xn )} − 1. Now for any x ∈ X there is some i such that
x ∈ Vxi . Then f (xi ) − 1 < f (x) < f (xi ) + 1 and so m < f (x) < M showing that
f is bounded. It remains to show that f attains its bounds. Let L, λ be respectively the
supremum and infimum of f over X. If there is no point x in X for which f (x) = L, then
we define a new function g : X → R by g(x) = L−f1 (x) for all x ∈ X. Then g is continuous.
However g is unbounded, for given any r > 0, there exists x such that f (x) > L − 1r and
hence g(x) > r. This contradicts the earlier part of the theorem and shows that f attains
the value L. Similarly f attains the infimum A.

Theorem 2.2.5. (Lebesgue covering lemma) Let (X, d) be a compact metric space and let
U be an open cover of X. Then there exists a positive real number r such that for any
x ∈ X there exists V ∈ U such that B(x, r) ⊂ V .

29
Proof. By compactness of the space X, we may suppose that the cover U is finite say
U = {U1 , U2 , ..., Un }. We may also assume that each Ui is nonempty and proper (if Ui = X
for some i, then any r will work). Let Ai = X \ Ui . Then each Ai is a non-empty
closed set. Define fi : Ai → R by fi (x) = d(x, Ai ) = inf {d(x, y) : y ∈ Ai }. Then
fi is continuous and is positive on Ui for each i = 1, 2, ..., n. Define f : X → R by
f (x) = max{fi (x) : i = 1, 2, ..., n} and f is continuous. Note that f is positive everywhere
since for any x ∈ X there is some i such that fi (x) > 0. So by the last theorem, f has
a minimum on X and this minimum is positive. Let r be the minimum. Now let x ∈ X.
Then for some i, f (x) = fi x. We contend B(x, r) ⊂ Ui . For suppose d(x, y) < r. If
y∈ / Ui then y ∈ Ai whence, r > d(x, y) ≥ d(x, Ai ) = fi (x) = f (x) contradicting that r is
the minimum of f . So B(x, r) ⊂ Ui and we can let Ui be the desired set V ∈ U.

A number r which satisfies the conclusion of the theorem is called a Lebesgue number of
the cover U.

Theorem 2.2.6. Let X be a compact space and suppose f : X → Y is continuous and


onto. Then Y is compact. In other words, every continuous image of a compact space is
compact.

Proof. Let V be any open cover of Y , Let U = {f −1 (V ) : V ∈ V}. Then V is a cover of X


and since f is continuous, it is an open cover of X. Since X is compact some finitely many
members of U, say {f −1 (V1 ), f −1 (V2 ), ..., f −1 (Vn )} where V1 , V2 , ..., Vn ∈ V cover X. But
then V1 , V2 , ..., Vn cover Y since f is onto. So Y is compact.

Definition 31. A topological property is said to be preserved under continuous functions


if whenever a space has it so does every continuous image of it, that is if f : X → Y is
continuous and onto and X has the given property, then so does Y .

Properties such as compactness, Lindeloff and separability are preserved under contin-
uous function. However, second countability is not preserved under continuous functions.

Definition 32. A topological property is hereditary if whenever a space has it, so does every
subspace of it. Second countability is a hereditary property. On the other hand separability
is not hereditary. The properties of compactness and of being a Lindeloff space are not
hereditary.

Proposition 2.2.7. If X is a compact (Lindeloff) space and A ⊂ X is closed in X then A


in its relative topology, is also compact

Proof. Suppose X is compact and A ⊂ X is closed. Let U be an open cover of A in the


relative topology on A. For each U ∈ U, fix an open set V (U ) in X such that A ∩ V (U ) =
U . Then the family V = {V (U ) : U ∈ U} ∪ {X \ A} is an open cover of X and hence
admits a finite subcover consisting of, say, V (U1 ), V (U2 ), ..., V (Un ) and possibly X \ A.

30
But then {U1 , U2 , ...., Un } covers A and is a finite subcover of U. This shows that A in its
relative topology, is compact. The proof that A is Lindeloff in case X is, is similar.

Definition 33. A topological property is said to be weakly hereditary if whenever a space


has it, so does every closed subspace of it. With this terminology, the last proposition shows
that compactness and the Lindeloff property are weakly hereditary.

Definition 34. Let X be a space and x ∈ X. Then a local base at x is a collection L


of neighbourhoods of x such that given any neighbourhood N of x there exists L ∈ L
satisfying L ⊂ N .

The collection of all open neighbourhoods of a point is a local base at that point. If X
is a metric space, then the collection of all open balls (closed balls will do as well) centred
at a point constitutes a local base at that point. Note that it suffices to take open balls of
rational radii, thereby giving a local base which is countable.

Definition 35. A space is said to be first countable at a point if there exists a countable
local base at that point and a space is said to be first countable or to satisfy the first axiom
of countability if it is first countable at each point.

Theorem 2.2.8. Every second countable space is first countable.

Proof. Let X be a second countable space and x ∈ X. Suppose B is a countable base for
X. Let L = {B ∈ B : x ∈ B}. Clearly L is countable and the proof will be complete if we
show that it is a local base at x. For this, let V be a neighbourhood of x. Then there exists
an open set V such that x ∈ V ⊂ N . Since B is base, there exists B such that x ∈ B ⊂ V .
But then B ∈ L, showing that X is a local base at x.

Remark : The converse of the proposition is false. The real line with the semi-open interval
topology is not second countable. However it is first countable.

Theorem 2.2.9. Let X, Y be spaces, x ∈ X and f : X → Y a function. Suppose X is first


countable at x. Then f is continuous at x iff for every sequence {xn } which converges to x
in X, the sequence {f (xn )} converges to f (x) in Y .

Proof. The direct implication is easy and does not require that X be first countable at x.
For the other way implication suppose {V1 , V2 , ..., Vn ...} is a countable local base at x. We
let W1 = V1 , W2 = V1 ∩ V2 , W3 = V1 ∩ V2 ∩ V3 ,...., Wn = V1 ∩ V2 ∩ ... ∩ Vn ,... etc. Then
the collection {Wn : n = 1, 2, 3, ...} is also a local base at x. Its advantage over the given
base is that it is a nested local base, that is, Wm ⊂ Wn for m ≥ n. The reason for switching
to Wi ’s from Vi ’s will be clear in the course of the proof.
Suppose f is not continuous at x. Then there exists a subset A, of X such that x ∈ A but
f (x) ∈
/ f (A). Now for each n, Wn is a neighbourhood of x and since x ∩ A, Wn ∩ A 6= φ.

31
Choose xn ∈ Wn ∩ A, for n ∈ N. Note that for m ≥ n, xm ∈ Wn since Wm ⊂ Wn . We
claim that the sequence {xn } converges to x in X. For, let G be an open set containing x.
Then, since {W1 , W2 , ...., Wn , ...} is a local base at x, there is some n such that Wn ⊂ G.
But then for all m > n, xm ∈ Wm and so xm ∈ G. So {xn } converges to x in X. However,
for each n, f (xn ) ∈ f (A) ⊂ f (A). We are assuming that f (x) ∈ / f (A). Then Y \f (A) is an
open set which contains f (x) but contains no term of the sequence {f (xn )}. Consequently
{f (xn )} does not converge to f (x) in Y . This contradicts the hypothesis and shows that f
is continuous at X.

Consider now the space X = N × N ∪ {∞} discussed previously. This space is first
countable at each (x, y) ∈ N × N. Indeed at any such point the singleton family {{(x, y)}}
is a local base. However, X is not first countable at the point ∞. It is not hard to show this
directly, but an indirect argument using the last theorem can be given as follows. Define
f : X → R by f (x, y) = 0 for (x, y) ∈ N × N and f (∞) = 1. Then f is not continuous
at ∞ although the condition in the last theorem is satisfied because all the sequences in X
which converge to ∞ are eventually constant. Hence X cannot be first countable at ∞, for
otherwise f would be continuous at ∞ by the last theorem. We thus have an example of a
countable space (i.e. a space whose underlying set is countable) which is not first countable.

Definition 36. A space is said to satisfy the countable chain condition if any family of
mutually disjoint open sets in it is countable.

Proposition 2.2.10. Every separable (and hence every second countable) space satisfies
the countable chain condition.

Proof. Suppose D is a countable dense subset of a space X. Let Q be a family of open sets
in X such that for G, H ∈ Q, G ∩ H = φ unless G = H. We have to show Q is countable.
Suppose first that φ ∈/ Q. For each G ∈ Q, G ∩ D is nonempty since D is dense in X.
Choose a point xG ∈ G ∩ D and define the function f : Q → D by f (G) = xG . Then f
is one-to-one since xG = xH would imply G ∩ H 6= φ, and hence G = H. Since the set
D is countable, it follows that Q is countable. In case the empty set φ ∈ Q, we apply the
argument to Q \ {φ} and find that it is countable. But then Q is also countable.

Exercise :

1. Prove that the co-countable topology on a set (defined analogously to the cofinite topol-
ogy, by letting the whole set and all countable subsets be closed) makes it into a Lindeloff
space.
2. Let x be an uncountable set with the cocountable topology on it. Prove that X is not
separable, although it satisfies the countable chain condition.

32
3. Let X be an uncountable set with the cofinite topology on it. Prove that X is separable
but not first countable at any point.
4. Prove that every infinite subset A of a compact space X has at least one accumulation
point in X.
5. Using the Lebesgue covering lemma, prove that every continuous function from a com-
pact metric space into another metric space is uniformly continuous.
6. Prove that first countability is a hereditary property.

Rough space :

33
2.2.2 Connectedness
Definition 37. A space X is said to be connected if it is impossible to find non-empty
subsets A and B of it such that X = A ∪ B and A ∩ B = φ. A space which is not connected
is called disconnected.

Proposition 2.2.11. Let X be a space and A, B subsets of X. Then the following statements
are equivalent:
1. A ∪ B = X and A ∩ B = φ.
2. A ∪ B = X, A ∩ B = φ and A, B are both closed in X.
3. B = X \ A and A is clopen (i.e. closed as well as open) in X.
4. B = X \ A and ∂A (that is, the boundary of A) is empty.
5. A ∪ B = X, A ∩ B = φ and A, B are both open in X.

Proposition 2.2.12. Let X be a space. Then the following are equivalent:


1. X is connected.
2. X cannot be written as the disjoint union of two nonempty closed subsets.
3. The only clopen subsets of X are φ and X.
4. Every nonempty proper subset of X has a nonempty boundary.
5. X cannot be written as the disjoint union of two nonempty open subsets.

Every indiscrete space is connected and that the only connected discrete spaces are
those which consist of at most one point. The space of rational numbers is disconnected;
given any irrational number a the sets {x ∈ Q : x < a} and {x ∈ Q : x > a} are both
open in the relative topology on Q and Q is clearly their disjoint union. Similarly the set
of irrational numbers is disconnected.

Proposition 2.2.13. Let f : X → Y be a continuous surjection. Then if X is connected,


so is Y .

Proof. Suppose Y is not connected. Then we can write Y = A ∪ B where A, B are


disjoint, nonempty and open subsets of Y . But then X = f −1 (A) ∪ f −1 (B). The sets
f −1 (A), f −1 (B) are mutually disjoint, and open since f is continuous. Further each is
nonempty since f is onto. This contradicts that X is connected. Hence Y is connected.

Theorem 2.2.14. A subset of R is connected iff it is an interval.

Proof. First note that a subset I ⊂ R is an interval iff it has the property that for any
a, b ∈ X, (a, b) ⊂ X. Now if X is not an interval then there exist real numbers a, b, c
such that a < c < b; a, b ∈ X and c ∈ / X. Let A = {x ∈ X : x < c} and B = {x ∈
X : x > c}. Clearly A, B are disjoint, open subsets of X (in the relative topology) since
A = X ∩ (−∞, c) and B = (c, ∞) ∩ X and A ∪ B = X. Further a ∈ A, b ∈ B and hence
A, B are nonempty. This shows that X is not connected.

34
Conversely suppose X is an interval and that X = A ∪ B where A ∩ B = φ, A 6= φ,
B 6= φ where the closure is relative to X. Let a0 ∈ A, b0 ∈ B. Without loss of generality
we may suppose that a0 < b0 . Now let x be the mid-point of the interval from a0 to b0 ,
i.e. x = a0 +b2
0
. Then x ∈ X and so x is exactly in one of the sets A and B. If x ∈ A
we rename it as a1 and rename b0 as b1 . If x ∈ B, we rename a0 as a1 and x as b1 . In
any case [a1 , b1 ] is an interval with its left end-point in A and the right end-point in B. We
can now take the mid-point of [a1 , b1 ] and get an interval [a2 , b2 ] of half the length with
a2 ∈ A, b2 ∈ B. Repeating this process ad infinitum, we get a nested sequence of intervals
{[an , bn ] : n = 0, 1, 2, 3, ...} such that an ∈ A and bn ∈ B for all n. Note that {an }
is a bounded monotonically increasing sequence while {bn } is a bounded monotonically
decreasing sequence and that (bn − an ) → 0 as n → ∞. By the order completeness of R,
both sequences converge to a common limit, say c. Note that c ∈ X since a0 ≤ c ≤ b0 .
Also every neighbourhood of c intersects A as well as B. So c ∈ A ∩ B a contradiction.
Hence X is connected.

Theorem 2.2.15. Every closed and bounded interval is compact.

Proof. Since any closed and bounded interval [a, b] (with a < b) is homeomorphic to the
unit interval [0, 1], it suffices to prove that [0, 1] is compact. Let U be an open cover of [0, 1].
An element U of U is open relative to [0, 1] and hence is of the form V ∩[0, 1] where V is an
open subset of R. Replacing such U ’s by the corresponding V ’s, we get a cover V of [0, 1]
by sets which are open in R and evidently it suffices to show that V has a finite subcover.
Now let S = {t ∈ [0, 1] : the interval [0, t] can be covered by finitely many members of
V}. We have to show that 1 ∈ S. Evidently 0 ∈ S and so S 6= φ. We claim S is both open
and closed in [0, 1]. First, let t ∈ S. Then [0, t] can be covered by, say, V1 , V2 , ..., Vn with
t ∈ Vn (say). Since Vn is open, there exists δ > 0 such that (t − δ, t + δ) ⊂ Vn . Now for
any t0 ∈ (t − δ, t + δ) ∩ [0, 1], the interval [0, t0 ] is also covered by V1 , V2 , ..., Vn showing
that t0 ∈ S. Hence (t − δ, t + δ) ∩ [0, 1] ⊂ S and hence S is open, being a neighbourhood
of each of its points.
On the other hand suppose t ∈ [0, 1] \ S. Choose V ∈ V such that t ∈ V . V being open,
there is δ > 0 such that (t−δ, t+δ) ⊂ V . We claim that (t−δ, t+δ)∩[0, 1] ⊂ [0, 1]\S. For
let t0 ∈ (t − δ, t + δ) ∩ [0, 1]. If t0 ∈ S then [0, t0 ] can be covered by, say, V1 V2 , ..., Vn ∈ V.
But then V1 , V2 , ..., Vn and V together cover [0, t] contradicting that t ∈ / S. Thus we have
shown that [0, 1] \ S is also open.
Putting it all together, S is a non-empty clopen subset of [0, 1]. But by the theorem above
[0, 1] is connected. We are thus forced to conclude that S is the entire interval [0, 1] and
hence, in particular, 1 ∈ S as was to be proved.

Definition 38. Two subsets A and B of a space X are said to be (mutually) separated if
A ∩ B = φ and A ∩ B = φ.

35
Proposition 2.2.16. Let X be a space and C be a connected subset of X (that is, C with
the relative topology is a connected space). Suppose C ⊂ A ∪ B where A, B are mutually
separated subsets of X. Then either C ⊂ A or C ⊂ B.

Proof. Let G = C ∩ A and H = C ∩ B. Then G, H are closed subsets of C since, A, B


are closed in A ∪ B. Also G ∩ H = φ. But C is connected. So either G = φ or H = φ. In
the first case C ⊂ B while in the second, C ⊂ A.

Theorem 2.2.17. Let C be a collection of connected subsets of a space X such that no two
members of C are mutually separated. Then ∪C∈C C is also connected.

Proof. Let M = ∪C∈C C. If M is not connected then we could write M as A ∪ B where


A, B are nonempty and mutually separated subsets of X. By the proposition above, for
each C ∈ C either C ⊂ A or C ⊂ B. We contend that the same possibility holds for all
C ∈ C, i.e. either C ⊂ A for all C ∈ C or C ⊂ B for all C ∈ C. If this is not the case, then
there exist C, D ∈ C such that C ⊂ A and D ⊂ B. But, A, B are mutually separated and
hence their subsets C, D are also mutually separated contradicting the hypothesis. Thus
all members of C are contained in A or all are contained in B. Accordingly M = A or
M = B, contradicting that A, B are both non-empty.

Corollary 2.2.17.1. Let C be a collection of connected subsets of a space X and suppose


K is a connected subset of X (not necessarily a member of C) such that C ∩ K 6= φ for all
C ∈ C. Then (∪C∈C C) ∪ K is connected.

Proof. Let M = (∪C∈C C) ∪ K. Let D = {K ∪ C : C ∈ C}. Clearly M = ∪D∈D D. By the


theorem above, each member of D is connected since it is a union of two connected sets
which intersect (and which are therefore not separated). Now any two members of D have
at least points of K in common and so are not mutually separated. So again by the theorem
above, M is connected.

Proposition 2.2.18. Let X1 , X2 be connected topological spaces and X = X1 × X2 with


the product topology. Then X is connected.

Proof. If either X1 or X2 is empty then so is X and the result holds trivially. So as-
sume X1 , X2 are both non-empty. Fix a point y1 ∈ X1 . Then the subset {y1 } × X2
is homeomorphic to X2 and hence is connected. Call it K. For each x ∈ X2 , the set
X1 × {x} is similarly connected and its intersection with K is non-empty. Also note that
X1 × X2 = (∪x∈X2 X1 × {x}) ∪ K. So by the corollary above X1 × X2 is connected.

Corollary 2.2.18.1. The topological product of any finite number of connected spaces is
connected.

Proof. If X1 , X2 , ..., Xn−1 , Xn are spaces (with n ≥ 2) then X1 × X2 × ... × Xn is home-


omorphic to (X1 × ... × Xn−1 ) × Xn .The result follows by induction on n and the last
proposition.

36
Proposition 2.2.19. The closure of a connected subset is connected. More generally if C is
a connected subset of a space X then any set D such that C ⊂ D ⊂ C is connected.

Proof. Suppose C is connected and C ⊂ D. If D is not connected then we can write


D = A ∪ B where A, B are nonempty, disjoint and closed relative to D. Then C ∩ A,
C ∩ B are disjoint closed subsets of C whose union is C. But C is connected. So one of
D
them, say, C ∩ B is empty. This means C ⊂ A, and hence C ⊂ A where the closure is
D X X
w.r.t. D. But C = C ∩ D = D since D ⊂ C . Hence A = D contradicting that B is
nonempty. So D is connected.

Definition 39. A component of a space is a maximally connected subset, that is, a connected
subset which is not properly contained in any connected subset of that space.

For example, the space (−1, 0) ∪ (0, 1) with the usual topology has two components,
(−1, 0) and (0, 1). A space is connected iff it has only one component, namely, the whole
space itself. On the other hand in a discrete space the only connected subsets are the empty
set and the singleton subsets and hence all components are singleton sets. Such a space is
said to be totally disconnected.

Theorem 2.2.20. a. Components are closed sets,


b. Any two distinct components are mutually disjoint,
c. Every nonempty connected subset is contained in a unique component,
d. Every space is the disjoint union of its components.

Proof. (a) Let C be a component of a space X. Then C is connected. Then, C is also


connected. Now C ⊂ C. But C is maximal w.r.t. the property of being connected, that is,
no proper superset of C can be connected. Hence C = C and so C is closed.
(b) Let C, C 0 be two components. If C ∩ C 0 is nonempty then C ∪ C 0 would be connected.
But C ⊂ C ∪ C 0 and C 0 ⊂ C ∪ C 0 . So again by maximally of C, C 0 , we get C ∩ C 0 = φ.
Thus two distinct components are disjoint.
(c) Let A be a nonempty connected subset of X. Let C be the collection of all connected
subsets of X containing A and let M = ∪C∈C C. Then any two members of C intersect
and so M is connected. Clearly A ⊂ M . We claim M is a component. For suppose V is
a connected subset of X containing M . Then N ∈ C and so N ⊂ M , whence M = N .
In other words, M is a maximally connected subset of X. Thus every nonempty subset is
contained in a component. Such a component is unique since two distinct components are
disjoint.
d. This assertion follows from the fact that for any x ∈ X, {x} is a connected set and hence
there is a unique component C of X such that x ∈ C.

The number of components of a space is evidently a topological invariant. A direct or


an indirect application of this fact is often useful in showing that certain spaces cannot be

37
homeomorphic to each other. As an example, consider [0, 1] and the unit circle S 1 each
with the usual topology. Both have one component each. However in [0, 1] there are points
x such that [0, 1] \ {x} is not connected (such points are called cut points). It is clear that if
h : [0, 1] → S 1 is a homeomorphism then h would map a cut point of [0, 1] to a cut point of
S 1 . But it is easy to show that S 1 has no cut points. Hence [0, 1] cannot be homeomorphic
to S 1 . Clearly the number of cut points is also a topological invariant.

Exercise :

1. Prove that a subset X of R is an interval iff it has the property that for all a, b ∈ X,
(a, b) ⊂ X.
2. Prove that the complement of Q × Q in the plane R2 is connected.
3. Give rigorous arguments to show that no two spaces in the following list are mutually
homemorphic:
i. the closed interval [0, 1]
ii. the open interval (0, 1)
iii. the semi-open interval [0, 1)
iv. a triod, that is a space homeomorphic to the figure Y
v. the unit circle S 1
vi. a space homeomorphic to the figure X
vii. a figure eight curve.
4. Prove or disprove that the interior and the boundary of a connected set are connected.
5. A space X is said to have the fixed point property if every map f : X → X has a fixed
point, i.e. a point x0 ∈ X such that f (x0 ) = x0 .
a. Prove that a space having the fixed point property must be connected.
b. Prove that the unit interval [0, 1] has the fixed point property.
6. Prove that a continuous bijection from R to R must be a homeomorphism.
7. Let A = {(x, sin1/x) : 0 < x < 1} ⊂ R2 and let X = A where the closure is w.r.t. the
usual topology on R2 . Prove that X is connected.

Rough space :

38
39
2.2.3 Local Connectedness and Paths
Definition 40. A space X is said to be locally connected at a point x in it if for every
neighbourhood, N of x (in X) there exists a connected neighbourhood M of x such that
M ⊂ N . X is said to be locally connected if it is locally connected at each of its points.

Proposition 2.2.21. A space X is locally connected at a point x ∈ X iff for every neigh-
bourhood N of x, the component of N containing x is a neighbourhood of x.

Proof. Suppose X is locally connected at x. Let N be a nbd of x and let C be the compo-
nent of N containing x (caution: C may not be a component of X). We are given that there
exists a connected nbd M of x such that M ⊂ N . Then M is contained in a unique com-
ponent of N and this component must be C since M intersects C. So M ⊂ C and hence
C is a neighbourhood of x. Conversely let N be a nbd of x and let C be the component of
N containing x. Then C is a connected neighbourhood of x contained in N and so X is
locally connected at x.

Note : All indiscrete and all discrete spaces are locally connected. Also real line with the
usual topology is locally connected since we proved in the last section that all intervals are
connected.

Proposition 2.2.22. For a topological space X the following statements are equivalent:
i. X is locally connected.
ii. Components of open subsets of X are open (in X).
iii. X has a base consisting of connected subsets.
iv. For every x ∈ X and every neighbourhood N of x there exists a connected open
neighbourhood M of x such that M ⊂ N .

Proof. (i) =⇒ (ii). Suppose X is locally connected, G is open in X and C is a component


of G. We have to show that C is open in X. Let x ∈ C. Then G is a neighbourhood of
x and C is the component of G containing x. So by local connectedness at x and the last
proposition, C is a neighbourhood of x. Thus C is a neighbourhood of each of its points
and therefore C is open in X.
(ii) =⇒ (iii). Let G be any open subset of X. Write G as the union of its components.
All these components are open, connected subsets of X by (ii). This means that every open
set can be expressed as the union of some open, connected subsets of X, or in other words
that the family of such sets forms a base for X.
(iii) =⇒ (iv). Let x ∈ X and N be a neighbourhood of x. Then there exists an open set
V such that x ∈ V ⊂ N . By (iii) and the characterisation of a base, there exists a connected
open set M such that x ∈ M and M ⊂ V . Hence M ⊂ N .
(iv) =⇒ (i). This is immediate.

40
Now we will see an example of a space which is connected but not locally connected.
Let X be the set B ∪ A where B = {(x, 0) ∈ R2 : 0 < x ≤ 1} and A = {(x, y) ∈
R2 : 0 ≤ y ≤ 1, x = 0 or x = 1/n for some n ∈ N}. The set X is pictured below.
It consists of infinitely many vertical segments of unit length, including a segment on the
y-axis and a horizontal segment along the x-axis. Give X the relative topology induced by
the usual topology on the plane. X is then called a ‘comb space’. It is easy to show that
X is connected, for each of the vertical segments is connected and meets the horizontal
segment B which is also connected. However X is not locally connected. For, let V
be the open ball (in X, with usual metric) centred at (0, 1/2) and with radius 1/4. The
components of V will be portions of the vertical segments. They will all be open (in X)
except the one along the y-axis, namely the component {(0, y) ∈ R2 : 1/4 < y < 3/4}.
Another example is the famous ‘topologist’s sine curve’ pictured below, which is defined
as A = {(x, sin(1/x)) : 0 < x < 1} ∩ {0, 0}.

Theorem 2.2.23. Every open subset of the real line (in the usual topology) can be expressed
as the union of mutually disjoint open intervals.

Proof. Let V be an open subset of R. Write V as the disjoint union of its components.
Each such component is a connected subset of R and hence an interval. But by the last
proposition, each component of V is open in R. So V is the disjoint union of open intervals.

We know that the collection of open intervals is a base for the usual topology on R and
hence that any open subset of R can be expressed as a union of open intervals. The crux
of the preceding theorem is that these intervals could be chosen to be mutually disjoint.
There is no analogue of this theorem which holds good for the plane or higher dimensional
euclidean spaces.

Proposition 2.2.24. Every quotient space of a locally connected space is locally connected.

Proof. Let f : X → Y be a quotient map and suppose X is locally connected. We have to


show that Y is also locally connected. It suffices to prove that components of open subsets
of Y are open in Y . So let V be an open subset of Y and let C be a component of V . Then

41
f −1 (V ) is an open subset of X and we assert that f −1 (C) is a union of some components of
f −1 (V ). This amounts to showing that if x is in f −1 (C) and D is the component of f −1 (V )
containing x, then D ⊂ f −1 (C). Since D is connected and f is continuous, we know that
f (D) is connected. Also f (D) ⊂ V and f (D)∩C 6= φ since x ∈ f −1 (C)∩D. So f (D)∪C
is connected. But C is a maximally connected subset of f −1 (V ). So f (D) ∪ C = C or
f (D) ⊂ C. Hence D ⊂ f −1 (C) and we have thus shown that f −1 (C) is the union of
some components of V . But since X is locally connected and f −1 (V ) is open, each of its
components is open in X. So f −1 (C)is open in X. This shows that C is open in Y as Y
has the strong topology generated by {f }. Thus we have shown that components of open
subsets of Y are open and so Y is locally connected.

Definition 41. A path in a topological space X is a continuous function α from the unit
interval [0, 1] into X. The points α(0) and α(1) are called respectively the initial and the
terminal points or the beginning and the end of α. We say α is a path from α(0) to α(1) or
that it joins α(0) to α(1).
We could of course replace the unit interval by any closed, bounded interval [a, b] where
a < b. However, topologically the generality gained is illusory since [a, b] is homeomorphic
to [0, 1]. A path is sometimes also called a curve; however, this term is generally reserved
for paths in a euclidean space. A path a is said to be simple if the function α is injective
and it is said to be closed if α(0) = α(1). A simple closed path is a closed path α which is
injective except for α(0) = α(1).

Definition 42. A space X is said to be path-connected if for every two points x, y ∈ X,


there exists a path α in X such that α(0) = x and α(1) = y.

Clearly the real line, all euclidean spaces and any convex subsets of them are all path-
connected. The unit sphere S n is also path-connected for n > 1 since given any two distinct
points on S n there is an arc of a great circle joining them and it can be parametrised as a path
(in fact as a simple path). It is easy to show that the topological product of path-connected
spaces is path-connected.

Proposition 2.2.25. Every path-connected space is connected.

Proof. Let X be a path-connected space. If X is empty then certainly it is connected.


Suppose X is non-empty. Fix some point x0 ∈ X. For each x ∈ X, we are given a path
αx in X such that αx (0) = x0 and αx (1) = x. Let Cx be the range of αx . Then Cx is
connected since it is a continuous image of the unit interval which is connected. Clearly
X = ∪x∈X Cx . For any two x, y ∈ X, x0 ∈ Cx ∩ Cy and so Cx , Cy are not mutually
separated. So X is connected.

The converse of Proposition is false. A classic counterexample is the topologists’s sine


curve.

42
Proposition 2.2.26. In any space X, the binary relation defined by letting x ∼ y for x, y ∈
X iff there exists a path in X from x to y, is an equivalence relation.

Proof. Reflexivity of the relation follows trivially by considering constant paths. For sym-
metry, suppose α is a path in X from x to y. Define β : [0, 1] → X by β(t) = α(1 − t) for
t ∈ [0, 1]. Now β is continuous since it is the composition α ◦ f where f : [0, 1] → [0, 1] is
the map f (t) = (1 − t), t ∈ [0, 1]. Then β is a path from y to x. For transitivity, suppose α
is a path from x to y and β is a path from y to z. Define γ : [0, 1] → X by

α(2t), if 0 ≤ t ≤ 1/2
γ(t) =
β(2t − 1), if 1/2 ≤ t ≤ 1

Then γ is well-defined since α(1) = y = β(0). Also β is continuous since its restriction
to each of the two closed subsets [0, 1/2] and [1/2, 1] of [0, 1] is continuous. Clearly γ is a
path in X from x to z and so x ∼ z. Hence ∼ is an equivalence relation on X.

Definition 43. The equivalence classes under the equivalence relation defined above are
called the path-components of the space X.

Proposition 2.2.27. A subset C is a path-component of a space X iff C is a maximal subset


of X w.r.t. the property of being path-connected.

Proof. Suppose C is a path-component of a space X. We claim C is path-connected. Let


x, y ∈ C. We certainly know that there is a path a in X from x to y. We assert that the
range of a is contained in C. For otherwise, there exists s ∈ (0, 1) such that α(s) ∈
/ C. Now
define β : [0, 1] → X by β(t) = α(st). Then β is a path in X from x to α(s). This means
that x and α(s) are in the same equivalence class under ∼, contradicting that α(s) ∈ / C.
Hence α can be considered as a path in C, showing that C is path-connected. Moreover if
D is a proper superset of C, then a point of D \ C cannot be joined to a point in C by a path
in X, and certainly not by a path in D, showing that D is not path connected. Hence C is a
maximally path-connected subset of X.
Conversely suppose C is path-connected and is not a proper subset of any path-connected
subset of X. Then any two points of C can be joined by a path in C and hence, a fortiori,
by a path in X. Hence C is contained in some equivalence class under ∼, say C ⊂ C 0 .
But as we just saw, every path-component of X is path-connected. In particular C 0 is path-
connected and so C = C 0 by maximality of C. Thus C is a path-component of X.

Exercise :

1. Let X1 , X2 , ..., Xn be topological spaces and X be their topological product. Suppose


Xi is locally connected at a point x1 for i = 1, 2, ..., n. Let x = (x1 , x2 , ..., xn ) ∈ X. Prove
that X is locally connected at x.

43
2. Prove that an open subspace of a locally connected space is locally connected.
3. Prove that path-connectedness is preserved under continuous functions.
4. Prove that the topological product of a finite number of non-empty spaces is path-
connected iff each coordinate space is so.
5. Prove that a connected, locally path-connected space is path-connected.

Rough space :

44
Module 3

Separation Axioms

Every concept in topology is defined in terms of open sets, in order to make non-trivial
and interesting statements about a space, it is necessary that the space possess a fairly rich
collection of open sets. In this section we shall study various degrees of such richness. We
shall define a number of related conditions all of which assert the existence of open sets
which will contain something but which will also exclude something else. For this reason,
the conditions are known as separation axioms.

3.1 Hierarchy of Separation Axioms


The separation axioms are of various degrees of strengths and they are called T0 , T1 , T2 , T3
and T4 axioms in ascending order of strength.

Definition 44. A topological space X is said to satisfy the T0 -axiom, or is said to be a T0


space if given any two distinct points in X, there exists an open set which contains one of
them but not the other.

T0 axiom is the weakest separation axiom. For if a space X is not T0 , then there would
exist two distinct points x, y in X such that every open set in X either contains both x and
y or else contains neither of them. In such a case, x and y may as well be regarded as
topologically identical.
Examples of T0 spaces include all metric spaces. As another example, let J be the topology
on R whose members are φ, R and all sets of the form (a, ∞) for a ∈ R. Note that in this
space, for x, y ∈ R with x < y, there exists an open set containing y but not x although
there exists no open set which contains x but not y.

Definition 45. A space X is said to satisfy the T1 -axiom or is said to be a T1 -space if for
every two distinct points x and y ∈ X, there exists an open set containing x but not y (and
hence also another open set containing y but not x).

45
Every T1 space is also T0 and the space (R, J ) above shows that the converse is false.
Thus the T1 -axiom is strictly stronger than T0 . The essential point is that given two distinct
points, the T0 -axiom merely requires that at least one of them can be separated from the
other by an open set whereas the T1 -axiom requires that each one of them can be separated
from the other.

Proposition 3.1.1. For a topological space (X, J ) the following are equivalent:
1. The space X is a T1 space.
2. For any x ∈ X, the singleton set {x} is closed.
3. Every finite subset of X is closed.
4. The topology J is stronger than the cofinite topology on X.

Proof. Only the equivalence of (1) with (2) needs to be established. The rest follow from
properties of closed sets and the definition of the cofinite topology. Assume (1) holds and
let x ∈ X. We claim that X \ {x} is a neighbourhood of each of its points. For, let
y ∈ X \ {x}. Then x, y are distinct points of X and so by the T1 -axiom, there exists an
open set, say, V such that y ∈ V and x ∈ / V . But this means V ⊂ X \ {x} and hence
X \ {x} is a neighbourhood of y. So X \ {x} is open and therefore {x} is closed in X.
Conversely assume (2). Let x, y be distinct points of X. Then X \ {y} is an open set which
contains x but not y. Thus the T1 -axiom holds in X.

Proposition 3.1.2. Suppose y is an accumulation point of a subset A of a T1 space X. Then


every neighbourhood of y contains infinitely many points of A.

Proof. Let N be a neighbourhood of y and let F = A ∩ (N \ {y}). We claim that the set F
is infinite. For, if not, X \ F will be an open set containing y and so N ∩ (X \ F ) will be a
neighbourhood of y. Evidently this neighbourhood contains no point of A, except possibly
y, contradicting that y is an accumulation point of A. So F is infinite, showing that every
neighbourhood of y contains infinitely many points of A.

Definition 46. A space X is said to satisfy the T2 axiom (or the Hausdorff property) or is
said to be a T2 (or Hausdorff) space if for every distinct point x, y ∈ X there exist disjoint
open sets U, V in X such that x ∈ U and y ∈ V .

Every T2 space is T1 since the condition in the definition implies that U contains x but
not y and that V contains y but not x. The converse is false. An infinite set with the cofinite
topology is T1 but not T2 , in fact no two open sets in it are disjoint unless one of them is
empty. All metric spaces are T2 . However, there are spaces which are T2 but which are not
metrisable, the real line with the semi-open interval topology is such a space.

Proposition 3.1.3. In a Hausdorff space, limits of sequences are unique.

46
Proof. Let {xn } be a sequence in a Hausdorff space X and suppose xn → x and xn → y
as n → ∞. We have to show that x = y. If not, then there exist open sets, U, V in X such
that x ∈ U , y ∈ V and U ∩ V = φ. Then there exist N1 , N2 ∈ N such that xn ∈ U for
all n ≥ N1 and xn ∈ V for all n ≥ N2 . Let m be an integer greater than both N1 and N2 .
Then xm ∈ U ∩ V contradicting that U ∩ V = φ. So x = y and thus limits of sequences in
X when they exist, are unique.

Definition 47. A space X is said to be regular at a point x ∈ X if for every closed subset
C of X not containing x, there exist disjoint open sets U, V such that x ∈ U and C ⊂ V .
X is said to be regular if it is regular at each of its points.

Definition 48. A space X is said to be normal if for every two disjoint closed subsets C
and D of X there exist two disjoint open sets U and V suchthat C ⊂ U and D ⊂ V .

Informally, a space is regular if every point can be separated from every closed subset
(not containing it) and it is normal if every two mutually disjoint closed subsets can be
separated from each other. These conditions can be satisfied in a vacous manner if the
space fails to have very many closed sets. For example, an indiscrete space is regular
because if x is a point not in a closed set C then C must be the empty set and hence we can
take (actually we have to take) U and V to be respectively the whole space and the empty
set. Similarly an indiscrete space is normal because the only way two closed subsets of it
can be mutually disjoint is when one of them is empty. This argument also shows that the
space (R, J ) considered above is normal. Note however that it is not regular.

Definition 49. A topological space is said to satisfy the T3 -axiom or is said to be a T3 -space
if it is regular and T1 .

Definition 50. A topological space is said to satisfy the T4 -axiom or is said to be a T4 -space
if it is normal and T1 .

Theorem 3.1.4. The axioms T0 , T1 , T2 , T3 and T4 form a hierarchy of progressively stronger


conditions.

Proof. We have to show that each axiom from T1 onwards implies the preceding one. We
have already seen that T1 implies T0 and T2 implies T1 . The implication that T3 implies
T2 follows easily from the definition of regularity by taking the closed set to be a sigleton
set. Note that normality does not, by itself, imply regularity as we saw in an example
above. However, in presence of the T1 -axiom, we can apply the condition of normality to a
singleton set and a given closed set and see that regularity holds. Hence T4 implies T3 .

An example of a space which is Hausdorff but not regular is provided by the following
topology on the real line. Let U be the usual topology on R and let C be the set {1/n : n ∈
N}. Let S be the smallest topology on R containing U ∪ {R \ C}. Then S makes R a T2

47
space since S is stronger than U and (R, U) is a T2 -space. Note that C is closed in R w.r.t.
S (although not with respect to U) and that 0 ∈
/ C. Thus (R, S) is not a regular space.

Theorem 3.1.5. All metric spaces are T4 (and hence T3 as well).

Proof. We already know that metric spaces are Hausdorff and hence T1 . It only remains
to show that they are normal. Let (X, d) be a metric space and let C, D be disjoint, closed
subsets of X. If either C or D is empty then we could separate them by the empty set
and the set X. Suppose then that both C, D are non-empty. Define f : X → R by
f (x) = d(x, C) − d(x, D) for x ∈ X. Then f is continuous, being the difference of
two continuous real valued functions. Note that d(x, C) = 0 iff x ∈ C and d(x, D) = 0
iff x ∈ D. So f is positive on D and negative on C. Now let U = f −1 (−∞, 0) and
V = f −1 (0, ∞). Then U, V are open subsets of X since they are the inverse images of
open subsets of R. Also C ⊂ U and D ⊂ V as noted above. Finally U ∩ V = φ
completing the proof that X is normal.

Definition 51. A space X is said to be completely regular if for any point x ∈ X and
closed set C not containing x, there exists a continuous function f : X → [0, 1] such that
f (x) = 0 and f (y) = 1 for all y ∈ C, where the continuity is w.r.t. the usual topology on
the unit interval [0, 1]. A space is said to be a Tychonoff space if it is completely regular
and T1 .

In a completely regular space, a point can be separated (or distinguished) from a closed
set by means of a continuous real-valued function. This condition is significantly different
from the earlier separation axioms which assert separation by means of mutually disjoint
open sets.

Theorem 3.1.6. Every completely regular space is regular. Every Tychonoff space is T3 .

Proof. Suppose X is a completely regular space. Let x ∈ X and C be a closed subset of


X not containing x. We are given a continuous function f : X → [0, 1] which assumes the
value 0 at x and the value 1 at all points of C. Now [0, 1] is a Hausdorff space (in fact it
is a metric space). Let G, H be disjoint open sets in [0, 1] containing 0 and 1 respectively.
Let U = f −1 (G) and V = f −1 (H). Then U, V are mutually disjoint, open sets in X and
clearly x ∈ U , C ⊂ V . This shows that X is a regular space whenever it is completely
regular. If in addition X is also T1 then X is T3 whenever it is a Tychonoff space.

The converse of this proposition is false. Normality does not imply complete regularity,
in fact, it does not even imply regularity as we saw above. However, in presence of the T1
condition, a normal space is a Tychonoff space. This is a non-trivial result which we shall
prove later. It shows that T4 is stronger than Tychonoff while the proposition above shows
that Tychonoff itself is stronger than T3 . For this reason, Tychonoff spaces are sometimes
called T3 1 -spaces.
2

48
Proposition 3.1.7. For a topological space X the following statements are equivalent.
1. X is regular.
2. For any x ∈ X and any open set G containing x there exists an open set H containing x
such that H ⊂ G.
3. The family of all closed neighbourhoods of any point of X forms a local base at that
point.

Proof. (1) =⇒ (2). Suppose X is regular, x ∈ X and G is an open set containing x. Then
X \ G is a closed set not containing x. So by regularity of X there exist open sets U, V
such that x ∈ U , (X \ G) ⊂ V and U ∩ V = φ. Then U ⊂ X \ V and hence U ⊂ X \ V
since X \ V is a closed set. But X \ V ⊂ G and thus U ⊂ G. So we can let H = U and
(2) holds.
(2) =⇒ (3). Let x ∈ X and N be a neighbourhood of x ∈ X. Let G be the interior of
N in X. Then G is an open set containing x and so by (2) there exists an open set H such
that x ∈ H and H ⊂ G. Then H is a closed neighbourhood of x contained in N . Hence
the family of all closed neighbourhoods of x is a local base at x.
(3) =⇒ (1). Suppose x ∈ X and C is a closed subset not containing x. Then X \ C
is a neighbourhood of x. So by (3), there exists a closed neighbourhood M of x such that
M ⊂ X \ C. Let U = int(M ) and V = X \ M . Then U, V are mutually disjoint open sets
and clearly x ∈ U . Also C ⊂ V . This shows that X is a regular space,

Proposition 3.1.8. For a topological space X the following are equivalent:


1. X is normal.
2. For any closed set C and any open set G containing C, there exists an open set H such
that C ⊂ H and H ⊂ G.
3. For any closed set C and any open set G containing C, there exists an open set H and a
closed set K such that C ⊂ H ⊂ K ⊂ G.

Proof. The argument is similar to that of the last proposition and is left to the reader. Note
that (3) can be informally stated as, in any closed-open inclusion we can insert an open-
closed inclusion.

Note: Every discrete space satisfies all the separation axioms mentioned so far. Since any
space whatsoever can be expressed as the continuous image of a discrete space, it is clear
that none of the properties defined in this section is preserved under continuous functions.
However, most of them are hereditary, with the exception of normality.

Proposition 3.1.9. Regularity is a hereditary property.

Proof. Suppose X is a regular space and Y is a subspace of X. Let y ∈ Y and D be a


closed subset of Y not containing y. Then D is of the form C ∩ Y where C is a closed
subset of X. Note that y ∈
/ C for otherwise y ∈ D. Hence by regularity of X, there exist

49
open sets U, V (in X) such that y ∈ U , C ⊂ V and U ∩V = φ. Let G = U ∩Y , H = V ∩Y .
Then G, H are open in the relative topology on Y . Also y ∈ G,D ⊂ H and G ∩ H = φ.
Thus the space Y (with the relative topology) is regular.

Finally, as regards divisibility, it turns out that none of the separation axioms is divis-
ible. If on R we define an equivalence relation R by letting xRy iff x − y ∈ Q, then
the quotient space X/R is indiscrete. Thus we see that none of the T0 , T1 , T2 , T3 , T4 and
Tychonoff conditions is divisible.

Exercise :

1. Prove that the properties T0 , T1 , T2 and complete regularity are all hereditary.
2. Prove that the co-countable topology on an uncountable set does not make it a Hausdorff
space although limits of sequences in it are unique.
3. For a set Y , the diagonal ∆Y is defined to be the set {(y, y) ∈ Y × Y : y ∈ Y }. Prove
that a space Y is T2 iff the diagonal ∆Y is a closed subset of Y × Y in the product topology.
4. Let Y be a Hausdorff space. Prove that for any space X and any two maps f, g : X → Y
the set {x ∈ X : f (x) = g(x)} is closed in X.
5. Let R be an equivalence relation on a space X. Prove that the quotient space X/R is T1
if and only if the equivalence classes of R are closed in X.

Rough space :

50
3.2 Compactness and Separation Axioms
Suppose we have a Hausdorff space X. This means that any two distinct singleton subsets
say {x}, {y} of X can be separated from each other by disjoint open sets. Suppose we
replace the singleton set {y} by a finite subset F . It is then easy to show that {x} and F can
be separated from each other by disjoint open sets. Indeed, suppose the distinct elements
of F are y1 , y2 , ..., yn . For i = 1, 2, ..., n let Ui , Vi be open sets such that x ∈ Ui , yi ∈ Vi
and Ui ∩ Vi = φ. Now let U = ∩ni=1 Ui and V = ∪ni=1 Vi . Then, clearly U, V are open sets,
x ∈ U , F ⊂ V and U ∩ V = φ. We have used here that the intersection of finitely many
open sets is open. The argument will break down precisely at this point in case the set F is
infinite. If, however, F is compact, things are not so bad, as we will see in the proposition
below.

Proposition 3.2.1. Let X be a Hausdorff space, x ∈ X and F a compact subset of X not


containing x. Then there exist open sets U, V such that x ∈ U , F ⊂ V and U ∩ V = φ.

Proof. For each y ∈ F , there exist open sets Uy , Vy such that x ∈ Uy ,y ∈ Vy and Uy ∩ Vy =
φ. The family {Vy : y ∈ F } is an open cover of F . Since F is compact, there is a finite
subcover, say {Vy1 , Vy2 , ..., Vyn }. Let U = ∩ni=1 Uyi , and V = ∪ni=1 Vyi . Then U, V are
disjoint open subsets, x ∈ U and F ⊂ V .

Corollary 3.2.1.1. A compact subset in a Hausdorff space is closed.

Proof. Suppose X is a T2 space and F is a compact subset of X. Then by the proposition


above, for any x ∈ X \F there exist open sets U, V such that x ∈ U , F ⊂ V and U ∩V = φ.
In particular, U ∩ F = φ and hence U ⊂ X \ F . Thus X \ F is a neighbourhood of each
of its points. So X \ F is open and F closed.

Corollary 3.2.1.2. Every map from a compact space into a T2 space is closed. The range
of such a map is a quotient space of the domain.

Proof. Suppose f : X → Y is continuous where X is compact and Y is Hausdorff. Let


C be a closed subset of X. Then C is compact and so f (C) is compact since continuous
image of a compact set is compact. But then f (C) is closed in Y by the corollary above.
Hence images of closed sets in X are closed in Y , i.e. the map f is closed. Let Z be the
range of f . Then f is a map from X onto Z is a quotient map since f is a closed surjective
map. Consequently Z is a quotient space of X.

Corollary 3.2.1.3. A continuous bijection from a compact space onto a Hausdorff space is
a homeomorphism.

Proof. Let f : X → Y be a continuous bijection where X is compact and Y is Hausdorff.


We claim f is open. Let G be an open subset of X. Then X \ G is closed and hence

51
f (X \ G) is closed in Y by the corollary above. But f (X \ G) = Y \ f (G) because f
is a bijection. So f (G) is open in Y . Thus f is a continuous, open bijection and hence a
homeomorphism.

Corollary 3.2.1.4. Every continuous, one-to-one function from a compact space into a
Hausdorff space is an embedding.

Proof. This is immediate from the last corollary.

An application of the above corollaries is to prove that there can be no continuous one-
to-one map from the unit circle S 1 , into the real line. For if f were such a map then f (S 1 )
would be homeomorphic to S 1 by the corollary above since S 1 is a compact space. But
then f (S 1 ) would be a compact, connected subspace of R, whence f (S 1 ) must be a closed,
bounded interval. But such an interval cannot be homeomorphic to S 1 since it has a cut
point whereas S 1 has no cut points.
Another interesting consequence is that if X is any set then a compact topology on X
(i.e. topology on X which makes it a compact space) cannot be properly stronger than a
Hausdorff topology. For, suppose J1 , J2 are,topologies on X such that (X, J1 ) is compact
and (X, J2 ) is Hausdorff. If J2 ⊂ J1 then the identity function idx : X → X is J1 − J2
continuous and hence a homeomorphism. So J1 = J2 . It follows that in the class of all
compact topologies on a set, every Hausdorff topology is maximal and that in the class of
all Hausdorff topologies on a set, every compact topology is a minimal one.

Theorem 3.2.2. Every compact Hausdorff space is a T3 space.

Proof. Let X be a compact, Hausdorff space. Then every closed subset of X is compact
and so the space X is regular by Proposition (3.2.1). Since X is also T1 (being T2 ), the
result follows.

Proposition 3.2.3. Let X be a regular space, C a closed subset of X and F compact subset
of X, such that C ∩ F = φ. Then there exist open sets U, V such that C ⊂ U , F ⊂ V and
U ∩ V = φ.

Proof. Use arguments analogous to those of Proposition (3.2.1)

Theorem 3.2.4. Every regular, Lindeloff space is normal.

Proof. Let X be regular, Lindeloff space and let C, D be disjoint, closed subsets of X.
By the characterisation of regularity, we get for each x ∈ C, an open set Ux containing x
such that Ux ⊂ X \ D and similarly for each y ∈ D an open set Vy containing y such that
Vy ⊂ X \ C. The sets C, D are closed subsets of a Lindeloff space and hence they are
themselves Lindeloff spaces . So the open covers {Ux : x ∈ C} and {Vy : y ∈ D} of C, D
respectively have countable subcovers say {Un : n = 1, 2, ...} and {Vn : n = 1, 2, ...}. It

52
is now tempting to let U = ∪n Un and V = ∪n Vn . Unfortunately although Un ⊂ X \ D
for each n, we cannot deduce from it that U ⊂ X \ D because the closure of U may be
larger than ∪n Un (however, this argument would be valid in case X were compact, for then
the subcover could be chosen to be finite and the closure operator does commute with finite
unions).
For each n ∈ N, let Gn = Un \ ∪ni=1 Vi and Hn = Vn \ ∪ni=1 Ui . Note that Gn , Hn are open
sets for all n. Let G = ∪∞ ∞
i=1 Gi and H = ∪i=1 Hi We contend C ⊂ G. For, let x ∈ C. Then
x ∈ Un for some n ∈ N. Also x ∈ / Vm for all m since Vm ⊂ X ⊂ C for all m. Hence
x ∈ Gn and so x ∈ G. Similarly D ⊂ H. Thus G, H are open sets in X containing C and
D respectively and to complete the proof we need only show that G ∩ H = φ. If this is not
so, then there exist m, n ∈ N such that Gm ∩ Hn 6= φ. Without loss of generality we may
suppose m ≤ n. Let x ∈ Gm ∩ Hn . Then x ∈ Um ⊂ Um which contradicts that x ∈ Hn .
So G ∩ H = φ and X is normal.

Corollary 3.2.4.1. Every regular, second countable space is normal.

Proof. This is immediate since every second countable space is a Lindeloff space

Corollary 3.2.4.2. Every compact Hausdorff space is T4 .

Proof. We have already seen that every compact Hausdorff space is regular and hence by
previous theorem it is also normal. Moreover, it is a T1 -space since it is T2 -space. Putting
together, every compact Hausdorff space is T4 .

Theorem 3.2.5. Let A, B be compact subsets of topological spaces X, Y respectively. Let


W be an open subset of X × Y containing the rectangle A × B. Then there exist open sets
U, V in X, Y respectively such that A ⊂ U , B ⊂ V and U × V ⊂ W .

Proof. The result is trivial if either A or B is empty. So assume A, B are both nonempty.
Fix b ∈ B. For each a ∈ A, W is an open neighbourhood of the point (a, b) ∈ X × Y . So
by the definition of the product topology, there exist open sets Ga , Ha in X, Y respectively
such that a ∈ Ga , b ∈ Ha and Ga × Ha ⊂ W . The family {Ga : a ∈ A} is an open cover
of the compact set A. Let {Ga1 , Ga2 , ..., Gam } be a finite subcover. Let Gb = ∪ni=1 Gai and
Hb = ∩ni=1 Hai . Then Gb , Hb are open sets in X, Y respectively such that A ⊂ Gb , b ∈ Hb
and Gb × Hb ⊂ W . These sets depend on the point b ∈ B.
We now let b vary over B. For each b ∈ B we find open sets Gb , Hb as above. The family
{Hb : b ∈ B} is an open cover of the compact set B. Let {Hb1 , Hb2 , ..., Hbm } be a finite
subcover. Let U = ∩m i=1 Gbi , and V = ∪{Hbi : i = 1, ..., m}. Then U, V clearly have the
desired properties.

Informally, the says that any neighbourhood of a compact rectangle contains an open
rectangular neighbourhood.

53
Proposition 3.2.6. Let X be a completely regular space. Suppose F is a compact subset
of X, C is a closed subset of X and F ∩ C = φ. Then there exists a continuous function
from X into the unit interval which takes the value 0 at all points of F and the value 1 at
all points of C.

Proof. If F is the empty set, the function which is identically 1 will work. Similarly if C
is empty, the identically zero function will work. Let us assume then that C and F are both
non-empty. For each x ∈ F , there exists a map fx : X → [0, 1] such that fx (x) = 0 and
fx (y) = 1 for all y ∈ C. If the set F were finite we could easily get the result by taking
the minimum (or the product) of the finite family of functions {fx : x ∈ F }. Since F is
compact (and not necessarily finite), we do the next best thing, namely, to apply a standard
compactness argument. For each x ∈ F , let Ux be the set fx−1 ([0, 1/2)). Then Ux is an
open set containing x and so the family {Ux : x ∈ F } is an open cover of the set F . By
compactness of F , there exists a finite subcover, say, {Ux1 , Ux2 , ..., Uxn }. Now define f :
X → [0, 1] by f (x) = min{fx1 (x), fx2 (x), ...., fxn (x)} for x ∈ X. Then f is continuous.
Also f assumes the value 1 at all points of C since each fxi does so for i = 1, 2, ..., n.
However, f may not vanish identically on the set F . This difficulty can be corrected as
follows. We certainly know that f (F ) is a subset of the semi-open interval [0, 1/2) since
for x ∈ F , there exists i such that 0 ≤ fxi (x) < 1/2 and f (x) ≤ fxi (x). Now let g be a
continuous function from the unit interval [0, 1] into itself such that g([1, 1/2)) = {0} and
g(1) = 1. The composite g ◦ f vanishes identically on F and assumes the value 1 at all
points of C.

When the members of a decomposition of a space X are compact and the projection map
p is closed, the quotient space shares many of the nice properties of the original space. We
now study a few results regarding this. Let X will be a space and D be some decomposition
of X. We shall regard D as a quotient space of X and p : X → D will denote the projection
map. Note that for a subset S of X, p−1 (p(S)) is in general large than S. If R denotes the
equivalence relation on X corresponding to the decomposition D then p−1 (p(S)) is the set
{x ∈ X : xRy for some y ∈ S}. It is clear that p−1 (p(S)) = S if and only if S is the union
of some members of D and that this is the case iff every member of D is either completely
contained in S or completely contained in X \ S. There is a name for such sets.

Definition 52. A subset S of X is said to saturated (w.r.t. the decomposition D) if p−1 (p(S)) =
S, equivalently S is saturated if there exists a subfamily C of D such that S = ∪C∈C C.

Note : For any A ⊂ X, the set p−1 (p(A)) is always saturated. Also the complement of
a saturated subset is saturated. If A, B are mutually disjoint and at least one of them is
saturated then p(A) and p(B) are mutually disjoint.

Proposition 3.2.7. With the notation above, the quotient map p : X → D is closed if and
only if for any D ∈ D and any open set G (in X) containing D, there exists a saturated

54
open set H such that D ⊂ H ⊂ G.
Proof. Assume first that p is closed. Let D ∈ D and an open subset G (of X) containing
D be given. Let K = p−1 (p(X \ G)) and H = X \ K. Then K is closed since p is closed
and continuous. So H is open in X and clearly it is saturated since K is so. Also H ⊂ G
since X \ G ⊂ K. It remains to show that D ⊂ H. For this, let x ∈ D. Since D ⊂ G, it
follows that p(x) = D 6= p(X \ G). So x ∈ / K and hence x ∈ H as desired.
Conversely assume that the given condition holds and suppose C is a closed subset of X.
We have to show that p(C) is closed in D. In view of the fact that D has the quotient
topology on it, this amounts to showing that p−1 (p(C)) is closed in X. So let V = X \
p−1 (p(C)). We claim that V is a neighbourhood of each of its points and hence is open.
Note first that V is the union of those members of D which are disjoint from C. Hence
V is saturated and does not intersect C. Now let x ∈ V . Let D be the unique member of
D containing x. Then D ⊂ X \ C which is open. By the given condition there exists a
saturated open set H such that D ⊂ H ⊂ X \ C. Then H is the union of some members of
D. None of these members intersects C. So H ⊂ V . Thus x ∈ D ⊂ H ⊂ V , showing that
V is a neighbourhood of x.
Theorem 3.2.8. Suppose D is a decomposition of a space X each of whose members is
compact and suppose the projection p : X → D is closed. Then the quotient space D is
Hausdorff or regular according as X is Hausdorff or regular.
Proof. Assume first that X is Hausdorff. Let C, D be distinct elements of D. Then C, D
are compact subsets of X and X is T2 . we can get open subsets U, V of X such that C ⊂ U ,
D ⊂ V and U ∩ V = φ. By the last proposition, there exist saturated open sets G, H such
that C ⊂ G ⊂ U and D ⊂ H ⊂ V . Clearly p(G), p(H) are mutually disjoint subsets of D,
containing C, D respectively. Also p−1 (p(G)) = G is open in X and so p(G) is open in D
by definition of the quotient topology. Similarly p(H) is open. It thus follows that D is a
Hausdorff space.
Next, suppose X is regular. Let A ∈ D and suppose C is a closed subset of D not containing
A. Then p−1 (C) is a closed subset of X which is disjoint from A. Now, there exist open
subsets U, V containing A and p−1 (C) respectively such that U ∩ V = φ. Note that p−1 (C)
is the union of some members of D. For each of these we apply the last proposition and get
a saturated open subset contained in V . The union of all such open saturated sets gives an
open, saturated subset H such that p−1 (C) ⊂ H ⊂ V . Also there exists a saturated open
subset G such that A ⊂ G ⊂ U . The rest of the argument is now similar to that given in the
last paragraph.

Before moving on to the next theorem, we introduce some notation. For a subset V of
X let K(V ) denote the union of those members of D which are contained in V . Evidently
K(V ) ⊂ V and X \ K(V ) = p−1 (p(X \ V )). It thus follows that if p is closed and V is
open then K(V ) is open. Moreover, p(K(V )) is open since p−1 (p(K(V ))) = K(V ).

55
Theorem 3.2.9. With the hypothesis of the last theorem, if X is second countable, so is D.

Proof. Let B be a countable base for X. Let U be the family of all finite unions of members
of B. Then U is also countable. Let L = {p(K(U )) : U ∈ U}. Then L is a countable family
of open sets in D. We contend that X is a base for the quotient topology on D. For this,
let A ∈ D and G be an open subset of D containing A. Then p−1 (G) is an open subset
of X containing A. Since A is compact, we can find a finite number of members of B
covering A whose union is contained in p−1 (G). This means, there exists U ∈ U such that
A ⊂ U ⊂ p−1 (G). Then p(K(U )) is an open subset of D containing A and contained in G.
Since p(K(U )) ∈ L by definition, it follows that L is a base for D.

Exercise :

1. Prove that the unit circle S 1 is compact.


2. For any map f : S 1 → R prove that there exists a point x0 ∈ S 1 such that f (x0 ) =
f (−x0 ). (Hint: Consider the sets {x ∈ S 1 : f (x) > f (−x)} and {x ∈ S 1 : f (x) < f (−x)}
and use connectedness of S 1 . Note that this result is stronger than saying that S 1 cannot be
embedded in R.)
3. Let A, B be closed subsets of S 1 such that S 1 = A ∪ B. Prove that at least one of A and
B contains a pair of mutually antipodal points.
4. Prove that the closure of a compact subset of a regular space is compact.
5. Prove that the real line with the semi-open interval topology is normal.

Rough space :

56
57
3.3 The Urysohn Characterisation of Normality
We have seen that complete regularity implies regularity i.e. separation of a point from a
closed set by means of a continuous function implies separation by open sets. This also
holds for separation of two closed sets as the following proposition.
Proposition 3.3.1. Let A, B be subsets of a space X and suppose there exists a continuous
function f : X → [0, 1], such that f (x) = 0 for all x ∈ A and f (x) = 1 for all x ∈ B.
Then there exist disjoint open sets U, V such that A ⊂ U and B ⊂ V .
Proof. We simply choose any two disjoint open sets G, H in [0, 1] containing 0 and 1
respectively (for example we could let G = [0, 1/2] and H = (1/2, 1) and set U = f −1 (G)
and V = f −1 (H). The assertion follows from the given properties of f .
Corollary 3.3.1.1. If a space X has the property that for any two mutually disjoint closed
subsets A, B of it, there exists a continuous function f : X → [0, 1] taking the value 0 at
all points of A and the value 1 at all points of B, then X is normal.
Proof. This follows immediately from the last proposition and the definition of normality.

The interesting thing is that the converse of the corollary above is true. This is a non-
trivial result due to Urysohn and is known as the Urysohn characterisation of normality.
In this section, our aim is at proving this result.
Theorem 3.3.2. A topological space X is normal if and only if it has the property that for
every two mutually disjoint, closed subsets A, B of X, there exists a continuous function
f : X → [0, 1] such that f (x) = 0 for all x ∈ A and f (x) = 1 for all x ∈ B.
Before we move on with the proof of the theorem, we will introduce couple of lemmas
which we will use to prove the theorem.

We are given some information about the subsets of X and we want to construct a
continuous function f : X → [0, 1] with some special properties. Let us see how the
existence of a continuous function f : X → [0, 1] implies the existence of a certain family
of subsets of X.
Lemma 3.3.3. Let f : X → [0, 1] be continuous. For each t ∈ R let Ft = {x ∈ X :
f (x) < t)}. Then the indexed family {Ft : t ∈ R} has the following properties:
i. Ft is an open subset of X for each t ∈ R.
ii. Ft = φ for t < 0 (actually F0 is also 0 but this is not very important.)
iii. Ft = X for t > 1.
iv. For any s, t ∈ R, s < t =⇒ Fs ⊂ Ft . Moreover, for each x ∈ X, f (x) = inf {t ∈ Q :
x ∈ Ft }.

58
Proof. Note that Ft is the inverse image (under f ) of the set (−∞, t) which is open in R.
So, by continuity of f , each Ft is open, showing (i),(ii) and (iii) follow easily from the fact
that f takes values in the unit interval. For (iv) let r be any number between s and t and let
C = {x ∈ X : f (x) ≤ r}. Then Fs ⊂ C ⊂ Ft . But C is closed in X by continuity of f .
So Fs ⊂ C and FS ⊂ Fr .
For the remainder of the lemma, let x ∈ X and let Gx be the set {t ∈ Q : x ∈ Ft }. We
have to show f (x) = inf Gx . Gx is nonempty because t ∈ Gx for all rational t > f (x).
Also t ≥ 0 for all t ∈ Gx and so Gx is bounded below. Let y = inf Gx . Now, for any
t ∈ Gx , x ∈ Ft and so f (x) < t. Hence f (x) ≤ inf Gx i.e. f (x) ≤ y. Suppose f (x) < y.
Let q be a rational number between f (x) and y. Then q ∈ / Gx since q < y. Hence x ∈ / Fq .
So f (x) ≥ q, which is a contradiction. Hence we get f (x) = y.

The lemma above shows that a continuous function f : X → [0, 1] induces a certain
indexed family of open subsets of X and moreover that we can recover the function f if we
knew some of these sets, namely the sets Ft for t ∈ Q. This is exactly what the next lemma
shows us.

Lemma 3.3.4. Let X be a topological space and suppose {Ft : t ∈ Q} is a family of sets
in X such that
1. Ft is open in X for each t ∈ Q.
2. Ft = φ for t ∈ Q, t < 0.
3. Ft = X for t ∈ Q, t > 1.
4. Fs ⊂ Ft for s, t ∈ Q, s < t.
For x ∈ X, let f (x) = inf {t ∈ Q : x ∈ Ft }. Then f is a continuous real-valued function
on X and it takes values in the unit interval [0, 1].

Proof. For x ∈ X, let Gx = {t ∈ Q. : x ∈ Ft }. Condition (3) shows that Gx is non-empty


and condition (2) shows that it is bounded below (by 0) for all x ∈ X. So the function
f (x) = inf (Gx ) is certainly a well-defined, real-valued function on X. Also conditions (2)
and (3) easily imply that for each x ∈ X, 0 ≤ f (x) ≤ 1 and hence f takes values in the
unit interval. It only remains to prove that f is continuous. For this, note-that the family of
all intervals of the form (−∞, a) or (b, ∞) for a, b ∈ R is a sub-base for the usual topology
on R. Hence, continuity of f will be established if we can prove that for any s ∈ R, the
sets {x ∈ X : f (x) < s} and {x ∈ X : f (x) > s} are open in X.
Let s ∈ R. Let H be the set {x ∈ X : f (x) < s}. It is tempting to think that H is precisely
the set Fs and hence is trivially open. Unfortunately this need not be so even when s is
rational. Nevertheless, we claim H = ∪{Ft : t ∈ Q, t < s}. For, suppose first, x ∈ H.
Then f (x) < s. Since f (x) = inf (Gx ) and f (x) < s, there exists q ∈ Gx such that q < s.
This means that x ∈ Fq and hence x ∈ ∪{Ft : t ∈ Q; t < s}. Conversely, we have to show
that if t ∈ Q and t < s then Ft ⊂ H. Let x ∈ Ft . Then clearly f (x) ≤ t and so f (x) < s,
showing x ∈ H. Thus we have shown that H is the union of Ft ’s, for t ∈ Q, t < s. But

59
each Ft is an open subset of X by (1) and so H is open in X.
Next, let K = {x ∈ X : f (x) > s}. We show that K is open in X by showing that its
complement X \ K is closed. To do this, we claim that X \ K = ∩{Ft : t ∈ Q, t > s}.
For, suppose first that x ∈ X \ K. Then f (x) ≤ s . Suppose t ∈ Q and t > s. Then
f (x) < t. Since f (x) = inf (Gx ), there exists q ∈ Gx such that q < t. But then x ∈ Fq
and by (4), Fq ⊂ Ft . So x ∈ Ft for all t ∈ Q for which t > s. Conversely suppose
x ∈ ∩{Ft : t ∈ Q, t > s}. We must show that f (x) ≤ s. If not, then s < f (x). Let q, t be
rational numbers such that s < q < t < f (x). Then clearly x ∈ / Fq for otherwise x ∈ Ft
by (4), and so t ∈ Gx violating that f (x) = inf (Gx ). Thus q ∈ Q, q > s and x ∈ / Fq ,
a contradiction. This establishes that X \ K is an intersection of closed sets and therefore
is closed in X. As noted before, this completes the proof of the continuity of f and of the
lemma.

In view of the preceding lemmas, the problem of finding a continuous function on a


space reduces to the problem of constructing a family {Ft : t ∈ Q} of subsets with certain
conditions. Countability of Q allows us to apply an inductive method in the construction.
We will now complete the proof of the theorem. We are given disjoint closed subsets A
and B of a normal space X. We want a continuous real-valued function f on X such that
f (x) = 0 for all x ∈ A and f (x) = 1 for all x ∈ B. We define a family of sets {Ft : t ∈ Q}
satisfying the conditions of the last lemma. For t < 0, and t > 1 we have no choice but to
let Ft be respectively the empty set and the set X. Let F1 = X \ B. Define F0 to be any
open set containing A such that F0 ⊂ X \ B (such a set exists by the characterisation of
normality). For rational numbers between 0 and 1 we proceed as follows.

Enumerate the set of rationals in [0, 1] as {q0 , q1 , q2 , q3 , ...., qn , ...} with q0 = 0 and
q1 = 1. Now Fq0 , Fq1 are already defined. Consider q2 . Clearly q0 < q2 < q1 . Define Fq2
to be any open set such that Fq0 ⊂ Fq2 ⊂ Fq2 ⊂ Fq1 . Such a set exists by normality of X.
Now suppose n ≥ 3 and that the open sets Fq1 , Fq2 , ...., Fqn−1 have already been defined so
as to satisfy condition (4) of the lemma. Consider qn . Let qi be the largest among those of
q0 , q1 , q2 , ...., qn−1 which are less than qn , i.e. q1 = max{qr : 0 ≤ r ≤ n − 1, qr < qn }.
Similarly let qj be the smallest among those of q0 , .., qn−1 which are greater than qn . Then
qi < qn < qj . By the inductive hypothesis Fqi ⊂ Fqj . By normality of X, there exists
an open set Fqn such that Fq1 ⊂ Fqn and Fqn ⊂ Fqj . Then condition (4) continues to be
satisfied with this Fqn included in the set of the F ’s defined so far. This completes the
inductive step in the definition and also shows that the family of sets {Ft : t ∈ Q} satisfies
all the conditions in the last lemma. So the function defined by f (x) = inf {t ∈ Q : x ∈
Ft } is a continuous function from X into [0, 1]. Now, if x ∈ A then x ∈ F0 and f (x) = 0.
Similarly if x ∈ B then x ∈ / Ft for any t ∈ Q, t ≤ 1 and so f (x) = 1. Thus the proof is
complete.

60
Corollary 3.3.4.1. All T4 spaces are completely regular and hence Tychonoff.

Proof. Simply apply the theorem to the case where one of the closed sets is a singleton.

Note : It need not be true that f is 0 only on A and 1 only on B. There may be points
outside A ∪ B at which f is 0 or 1. In other words we merely claim that A ⊂ f −1 ({0}) and
B ⊂ f −1 ({1}) and not that A = f −1 ({0}) or B = f −1 ({1}). A function whose existence
is asserted by the Urysohn’s lemma is called a Urysohn function.

Exercise :

1. Suppose X is a metric space and A, B are non-empty disjoint, closed subsets of X.


Prove that there exists a continuous function f : X → [0, 1] such that A = f −1 ({0}) and
B = f −1 ({1}).
2. Prove that every continuous real-valued function on an indiscrete space is constant, even
though such a space is always normal.
3. Prove that a connected, T4 space with at least two points must be uncountable (i.e. the
underlying set must be uncountable). (Hint: A Urysohn function on such a space must be
onto).
4. Prove that there exists no countable, connected, T3 space.

Rough space :

61
3.4 Tietze Characterisation of Normality
Suppose X is a topological space, A is a subset of X and f : A → R is a continuous
function (w.r.t. the subspace topology on A and the usual topology on R), We want to find a
continuous extension of f to the space X, that is we want a continuous function F : X → R
such that for any x ∈ A, F (x) = f (x). Of course such an extension may not always exist.
For example, let X = [0, 1], A = (0, 1] and define f : A → R by f (x) = sin(1/x). Then
f cannot be continuously extended to [0.1] because there exist sequences {xn },{yn } in A
which converge to 0 in [0, 1] such that {f (xn )}, {f (yn )} converge to distinct limits in R
and this would violate the continuity of any extension of f to [0, 1].

Proposition 3.4.1. Let A be a subset of a space X and let f : A → R be continuous. Then


any two extensions of f to X agree on A. In other words, if at all an extension of f exists
its values on A are uniquely determined by values of f on A.

Proof. Suppose F, G : X → R are both extensions of f . Let C = {x ∈ X : F (x) =


G(x)}. Clearly A ⊂ C. But since R is a Hausdorff space, C is closed in X. Hence A ⊂ C
which implies the result.

The general problem of extending f from A to X can be broken into two steps: (i) to
extend f from A to A, and (ii) to extend it further from A to X. The proposition just proved
says that the first part has at most one solution and the example given above shows that
there may actually be no solution. In this section we will be concerned with the second part
of the extension problem. Let us then suppose that f has already been extended somehow
to A and inquire whether it can further be extended to X. This means that without loss of
generality, A may be regarded as a closed subset of X. Even then, an extension of f may
not always exist. In fact the existence of such extensions puts a strong condition on the
space X as we see in the following proposition.

Proposition 3.4.2. Suppose a topological space X has the property that for every closed
subset A of X, every continuous real valued function on A has a continuous extension to
X. Then X is normal.

Proof. Let B and C be disjoint closed subsets of X. Let A = B ∪ C and define f : A → R


by f (x) = 0 for x ∈ B and f (x) = 1 for x ∈ C. Then A is a closed subset of X. Also the
function f is well defined and continuous. By hypothesis, there exists a continuous function
F : X → R which extends f . Then F (x) = 0 for x ∈ B and F (x) = 1 for x ∈ C. From
this, as we have seen many times, it follows that there are disjoint open sets containing B
and C respectively. Hence the space X is normal.

The interesting point is that its converse is true. Proposition (3.4.2) along with its con-
verse is called the Tietze characterisation of normality. The proof of the converse re-

62
quires the use of Urysohn’s lemma and the notion of uniform convergence of sequences of
functions.

Definition 53. Let X be a topological space and (Y, d) a metric space. Then a sequence of
functions {fn } from X to Y is said to converge uniformly on X to a function f : X → Y
if for every  > 0, there exists N ∈ N such that for all n ≥ N , and for all x ∈ X,
d(fn (x), f (x)) < . The sequence {fn } is said to converge point wise to f if for every
x ∈ X the sequence {fn (x)} converges to f (x) in Y .

Proposition 3.4.3. Let X, (Y, d), {fn } and f be as above and suppose {fn } converges to f
uniformly. Then if each fn is continuous, so is f .

Proof. Let x0 ∈ X and let V be an open neighbourhood of f (x0 ) in Y . Choose  > 0


so that B(f (x0 ), ) ⊂ V . Let N ∈ N be such that for all n ≥ N and for all x ∈ X,
d(fn (x), f (x)) < /3. Since fN is continuous at x0 there exists an open neighbourhood W
of x0 such that for all x ∈ W , d(fN (x), fN (x0 )) < /3. Now for any x ∈ W we haved

d(f (x), f (x0 )) ≤d(f (x), fN (x)) + d(fN (x), fN (x0 )) + d(fN (x0 ), f (x0 ))
</3 + /3 + /3 = .
Hence f (W ) ⊂ B(f (x0 ), ) ⊂ V showing that f is continuous at x0 . Since x0 ∈ X was
arbitrary, f is continuous.

Proposition 3.4.4. Let ∞


P
n=1 Mn be a convergent series of non-negative real numbers.
Suppose {fn } is a sequence of real valued functions on a space X such that for each x ∈ X
and n ∈ N, |fn (x)| ≤ Mn . Then the series ∞
P
n=1 fn converges uniformly to a real valued
function on X.

Proof. By the comparison test for series, for each x ∈ X, the series ∞
P
n=1 fn (x) is abso-
lutely convergent. Denote its sum by f (x). Then f is a real valued function on X. Let {sn }
be the sequence of the partial sums of the series ∞
P
n=1 fn . We claim that {sn } converges to
f uniformly on X. Note that for each x ∈ X and n ∈ N , |sn (x) − f (x)| ≤ ∞
P
k=n+1 Mk .
P∞
Since the series n=1 Mk is given to be convergent, given  > 0, we can find N ∈ N
such that for all n ≥ N , ∞
P
k=n+1 Mk < . But then for any x ∈ and any n ≥ N ,
P∞
|sn (x) − f (x)| < . Hence n=1 fn converges uniformly to f on X.

We now have all the machinery to prove the Tietze extension theorem. First we establish
it for functions into the interval [−1, 1].

Theorem 3.4.5. Let A be a closed subset of a normal space X and suppose f : A → [−1, 1]
is a continuous function. Then there exists a continuous function F : X → [−1, 1] such
that F (x) = f (x) for all x ∈ A.

Proof. Refer the textbook.

63
Theorem 3.4.6. Let A be a closed subset of a normal space X and suppose f : A →
(−1, 1) is continuous. Then there exists a continuous function F : X → (−1, 1) such that
F (x) = f (x) for all x ∈ A.

Proof. Refer the textbook.

Since any open interval and the real line are homeomorphic to (−1, 1), any continuous
real-valued function on a closed subset of a normal space can be extended continuously to
the whole space, thus proving the theorem.

Exercise :

1. Prove that there exists a map r : R → [0, ∞) such that r(x) = x for all x ∈ [0, ∞). (in
other words r is a right inverse to the inclusion map of [0, ∞) into R. (Such a map is called
a retraction and in the present case we say that [0, ∞) is a retract of R).
2. Prove that if A is a closed subset of a normal space X then any map f : A → [0, ∞) can
be continuously extended to a map from X to [0, ∞).
3. Does the Tietze extension theorem hold for maps into R if on R we put the semi-open
interval topology?

Rough space :

64

You might also like