UNIT 1 INTRODUCTION TO METRIC

SPACES
Structure Page No.
1.1 Introduction 7
Objectives
1.2 Definitions and Examples 8
1.3 Open Sets 16
1.4 Closed Sets 26
1.5 Continuous Functions 3ofE8
1.6 Summary 40
1.7 Hints/Solutions 40
- - -

1.1 INTRODUCTION
In this unit, we introduce you to the notion of metric spaces. As we already
pointed out in the course introduction and block introduction, the metric spaces
arose from extending the notions of continuity and convergence on the real line
to more abstract spaces.

A metric space is just a set which is equipped with a function called metric
which measures the distance between the elements of various pairs from the set.
We shall first give the definition of a metric and a metric space and consider
various examples in Section 1.2. Then we shall study various properties of these
spaces. In Section 1.3 and Section 1.4, we shall consider open and closed sets.

The structure on a metric space allows us to extend the notion of continuity to

functions in the context of these spaces. We shall define this notion in Section
1.5 and discuss several examples of continuous functions. Later you will see
that the notion of continuity is one of the most important notions for further
study of Analysis. Here we talk about two important results about continuous
functions which are called Urysohn's lemma and glueing lemma. Then we
explain the notion of uniform continuity through some examples. You will see
that the definition of continuity and uniform continuity for metric spaces are
similar to those for Euclidean spaces Rn. But extending these notions to metric
spaces, provide not only a new perspective but also a deeper insight into their
structure and properties.

Objectives
After studying this unit, you should be able to
state the properties that define a metric and apply them;

give examples of different metrics on R";

explain a discrete metric space and other metric spaces such as function spaces;

check whether
i) a subset of a metric space is open;
ii) a subset of a metric space is closed;
h lctric Spaces iii) a function defined on a metric space is continuous;
iv) a function defined on a metric space is uniformly continuous;

explain Urysohn's lemma and glueing lemma.

.+

DEFINITIONS'AND EXAMPLES
We Begin our study of metric spaces by extending those aspects of real and
complex number systems that help us in studying their analytical structure and
functions on them. What do you think are the most important concepts in
Analysis, then? You would agree that the limit and continuity are the ones
which come to your mind. Whatever you studied later had made use of these
building blocks. Now, instead of a set of real numbers, if we start with any
arbitrary set of objects, how can we introduce limit and continuity on it?
Obviously, it would have to have some extra properties. Now, if you go back
and recall the definition of a limit, we say the limit of f(x) as x approaches a is l,
if f(x) gets arbitrarily close to l as x gets sufficiently close to a. To be precise,
Vc > 0 , 3 6 > 0, s.t.

Here Ix - a1 and If(x) - C(denote the distance between x and a and between
f(x) and C, respectively.
So, it seems that if we can introduce the notion of 'distance' on our arbitrary set
of objects, we would be able to talk about limits in this new set up also.

pefinition 1: Let X be a non-empty set. A metric d on X is a function

d : X x X -+ R, such that, for x, y, z E X,
1) d(x,y) >_ Oandd(x,y) = Oifandonlyifx=y,
C

h
2) d(x,y) =d(y,x),
<
3) 4x1 Y) 4 x 1 2) + 4 2 1 Y),x, Y, z E X.
For x, y E X, d(x, y) is called the distance between x and y.

The three properties 1 , 2 and 3 listed in Definition I above mean that

0 the distance between x and y is non-negative and it is equal to zero when

and only when x and y coincide,

0 the distance between x and y is the same as the distance between y and x.
This is the property of symmetry,

0 the distance between x and y is not more than the distances between x
and z and that between z and y, added together. This is called the triangle
inequality.

Take a minute and think. Aren't these precisely the properties that we expect
from 'distance', based on our previous knowledge? And where does our
previous knowledge come from? From R or R~ of course, and so, our first
example of a metric space comes from R.
 Example 1: Let us consider R, the set of real numbers. Define a function d on Introduction to Metric
R x R by d(x, y) = Jx- yl. Then d has the three properties listed in Definition Spaces
1, and thus, is a metric on R. We refer to d as the usual metric or the
standard metric or Euclidean metric and if we refer to R as a metric space,
we assume that the metric is given by the usual metric d unless otherwise
specified. Note that the triangle inequality follows from the inequality

Definition 2: A pair (X, d) where X is a non-empty set and d is a metric on X

is called a metric space.

Elements of X are referred to as points of the metric space (X, d). When there is
no ambiguity about the metric on X, we may also denote the metric space
simply by X.
,-

Our next example identifies C also as a metric space. You need to recall the
definition of the absolute value of a complex number.
E~ample2: Consider the function d : C x C -+ R such that
4%22) = Iz1 - 221.
Now, since the absolute value function on C has the following properties,
i) JzJ>Oandlzl=Oifandonlyifz=O,
ii) 1 - zJ = lzl.
It is easy to verify that d satisfies conditions (I) and (2) in Definition 1. To show
the trianglilar inequality d(zl, z2) < d.(zl, z3) + d(z3,z2), we use the inequality
) z + wl 5 lzl + Iw1 with z = zl - z3, w = z3 - z2. Youcanverify this
inequality. This shows that d is a metric.

Note that in both Examples 1 and 2, we have defined the metric with the help of
b
the absolute value function which already existed on those sets. You will see
soon, that this is how we are going to come'up with many more metric spaces.
But now we introduce a metric, which is trivial, but throws up a very interesting
space, which serves as an example for some new concepts.
Example 3: Let X be a non-empty set. We define

0, i f x = y
1, i f x f y .
This means that the distance of a point from any other point is exactly 1. Are
you convinced that this a metric? Try this by yourself and figure out why it is
called a trivial metric (see El). This metric is called the discrete metric.

Remark 1: The above example shows that every set X has atleast one metric on
i.
it. C
C

Example 4: Let X be R2.Then consider the function d :R

" R2-+ R
 Metric Spaces defined by
i

where pl = (xl, yl) and pz = (x2, y2) represent any two elements in R2.You
recall that the expression on the Right Hand Side of Eqn(1) represents the
Euclidean distance between two points in R2. Therefore we can easily verify
the condition (I), (2) and (3) in Definition 1. Note that the condition (3) follows
from the property of triangles which states that the length of one side of a
triangle is always less than or equal to the sum of the lengths of the other two
sides(see Fig. 1). Thus the name triangle inequality is justified. This shows that
the function d defined by Eqn.(l) is a metric on R2.The metric d is called the
b
Euclidean metric or the usual metric on or the standard metric on R2.Fig. 2
Fig. 1
gives the geometrical picture of the distance measured by this metric.

Fig. 2

If we connect R2to C by the map,

then from Example 2, we'have

where zl and z2 are in @ represented by zl = X I + iyl and z2 = x2 + iy2. This

shows that the metrics given in Examples 2 and 4 are the same.
Example 5: Lit us consider R2.Consider the function dl : R2 x R2
defined by
- R

We claim that dl is another metric on R2.DO you agree with us? Convince
yourself by doing E2. The following figure gives the geometrical picture of the
distance measured by this metric.
 - -

Introduction to Metric
Spaces

Pig. 3

The metric defined by Eqn.(2) is called Taxicab metric on R2because it

models the distance a taxi-cab would have to traverse to get from one point to
another if the cab was only allowed to move in cardinal directions (north, south,
east, west) and not diagonally. You can see that the distance in this metric is
atleast as large as that in Euclidean metric.

Yet another metric on R2is explained in the next example.

Example 6: Let us define another function d, : R2x R2-+ R given by

Then d, satisfies all the properties of metric. We leave this as an exercise for
you to verify (see E2). Fig. 2 gives the geometrical picture of the distance.

$.
Pig. 4

If you closely look at the examples above, you will see that similar metrics can
be defined for the higher dimensional spaces Rn.In the next example we
consider this.
Example 7: Let Rndenote the set of elements x = (xl,.... x,) where xi E R
f o r i = 1,. . . . n.
Now define d, dl, d, on Rnx Rn -+ R,as follows:
Metric Spaces

Then dl dl and d, define metrics on Rn.

The metric d is called the usual metric (or the Euclidean metric) and (R", d) is
called the Euclidean space. The metric dl is called the Taxicab metric on R n .
Remark 2: To verify the triangle inequality for the metric d in (4) we need the
following two inequalities which are useful in many other contexts also.

where xi, yi, i = 1,.. . , n, are real numbers.

The inequality in (5) is called Cauchy-Schwarz inequality and in the
inequality in (6)is called Minkowski's inequality.
In the following example we consider a metric on the set of bounded real
sequences.
Example 8: Let X be the set of all bounded real sequences x = (XI,x2, x3 . . .).
Define a metric d, on X by
d m ( ( ~~ ~2,~, 3 ,. ..), (YI,~ 2~, 3 ,. .))
. =~ ~ ~ (-
1 Yx I1~1x2 . ~~ 3 1- , -1.
, - ~ 2 1 , I x-

This space is denoted by tW(R).The verification that d, is a metric is left as

an exercise (see E3).

Next we show an important inequality.

Proposition 1: In any metric space (X, d), we have

Proof : Let x, y, z E X. Since d(x, z) 5 d(x; y) + d(y, z) it follows that

d(x, z) - d(Y, 2 ) Id(x,,Y)
Interchanging x and y in this inequality, we see that
i
I ~ h u we
s get Introduction to Metric
I
d(x,2) - d(y,2) 2 -d(x, Y)
Therefore

I
from which it follows.that

ld(x, 2) - d(y1z) I L d(x1 Y).

We want you to try some exercises.

-

El) Show that the function d in Example 3 is a metric.

E2) Show that the functions d, dl and d, given in Example 5,6 and 7 satisfy
all the three properties of a metric.

E3) Show that the function d, given in Example 8 is a metric.

E4) Which of the following functions d : R x R -+ R are metrics on R?
i) 4x1 Y)= 5lx - yl
ii) d(x, y) = xZ y2+
iii) d(x, y) = (x - y) 2
E5) Check whether the function d : R2 x R2 R given by
-+

~ )1x1 - X Z ~ ) Y I- ~ 2 where
~ ( P I , P= ) pl = ( x l , y ~ and is a
) pz = (x2,y~)
metric or not.

E6) Let (X, d) be a metric space. Show that the following functions give
metrics on X.
d(x1y)
i) D(x, y) =
1+ d(x1Y)

So far we have been considering sets whose elements are finite or infinite
sequences of real or complex numbers. But quite often we need to consider sets
whose elements are either functions or matrices. In the next few examples we
consider such sets and discuss certain metrics on these sets.
Example 9: Let T = [a,b] and let B(T) denote the set of all real-valued
bouilded functions defined on T. For f, g E B(T), let

Note that {(f(x)- g(x)(: x E T) is a non-empty set of non-negative real

numbers which is bounded above. Therefore the set has its supremum in R.
Thus d, is a function from B(T) x B(T) 4 R. We can verify that d, satisfies
all the conditions 1 , 2 and 3 of Definition 1. Therefore dm defines a metric on
B(T).
5,: ctric Spaces B(T) with the metric d, is called the metric space of bounded real functions
on T.
Note: A metric space in which the elements are functions is called a function
space. Thus B(T) is a function space.
In the next example we consider another function space.
Example 10: For T = [a,b], let C(T) denote the set of all real-valued functions
on T which are continuous. From your undergraduate Real Analysis course you
know that a continuous function defined on a closed and bounded interval is
bounded. Therefore each functions in C(T) is bounded; that means C(T) is a
subset of B(T).
From the algebra of continuous functions you know if f, g € C(T), then
f + g E C(T) and (-f) E C(T). We can consider the function d, as in Example
10 for C(T) as well. Accordingly, for f, g E C (T), we have

Hence (C(T),d,) is a metric space. This metric space is usually denoted by

C(T) only. It is called the metric space of continuous functions.

Remark 3: Inthe two examples above the only difficulty is to verify the
triangle inequality. For that you only need to show the following inequality for
f, g, h E C(T) or BQ.

sup Jf(t)- g(t)l 5 sup If(t) - h(t)(+ sup lh(t) - g(t)l

t€ [a,b] t€[a,bl . t€[a,b]
The following figure illustrates the geometrical meaning of the distance defined
by dm for B(T) or C(T).

Fig. 5

Example 11: LetX = C[O, 11. For f,g E C[O, 11 defined : X x X -+ R by

where the integral is the Riemann integral. We first of all note that the integral
on the R.H.S (8) exists since 1 (f - g)1 is a continuous function. We claim that dl
is a metric. All conditions except the condition (1) follows from the properties
of Riemann integral. The condition (1) follows from a result which we state Introduction to Metric
below. Spaces

Proposition 2: Let f : [O, 111 -t R be continuous with f(t) 2 0 for t E [0, 11.
Then J,' f(t)dt = 0 if and only if f(t) = 0 for all t E [0, 11.
You might be already familiar with this result from your undergraduate Real
Analysis course.
The figure below illustrates the geometric meaning of dl (f, g).

d , g g) is the shaded area

Fig. 6

Remark 4: Note that if we assume that f is non-negative, Riemann integrable

on [0, 11 and such that
6'
f(t)dt = 0, not necessarily continuous then we cannot
conclude that f = 0 on [0, 11. For instance consider the function g(t) = 0 if
t/
2
and g (4) = 0. Then i' g(t)dt = 0, whereas g is non-zero So, the
extra condition of continuity is put to avoid this situation.
Remark 5: In the example above, we have considered sets of functions on the
closed interval, [O, 11. There is no special status attached to [0, I.]. We could
take any bounded closed interval in its place. We could also try to define a
similar metric for larger spaces of functions (see E6).
Example 12: Let M(n7R ) denote the set of all n x n real matrices. We identify
a matrix A = (aij&=, in M(n, R ) with the element
2
(all,a12,...,aln,...,anl,-..,%n) E Rn
Then the Euclidean metric d on R" defines a metric on M(n, R ) which makes
M(n7R ) a metric space.

Try this exercise now

E7) Let X = R[a, b] denote the set of all Riemann integrable functions on [a,
b]. Define a function d on X x X -+ R by
d(f>g) = Jb, lf(t) - g(t1ldt
kt
I Metric Spaces where f, g E X. Check whether d defines a metric on X.

Definition 3: Let (X, d) be a metric space and let A c X be a non-empty

subset of X. For x, y E A, define 6(x, y) = d(x, y). Then S is a metric on A and
it is called the induced metric on the subset A.

Note: We adopt the convention that if a subset A of metric space X is referred

to as a metric space, then it is assumed that the metric on A is the one induced
by the metric on X, unless otherwise stated.

Definition 4: (Product Metric):. Let (XI,dl) and (X2, d2)be two metric
spaces. Let X = X1 x X2, i.e. the Cartesian product of X1 and X2. Then define
a function d on X x X by

where xl, yl E X1 and x2, y2 E X2.

Then d defines a metric on the product set X = X1 x X2. We refer to this metric
as the product metric.

Example 13: Let X = S1 x [ O , l ] where S1 = {(XI,x2) E R 2 : x: + xi = 1)

Then X = {(xl, x2,x3) : X: + xi = 1 , O 5 XQ 5 1) is a subset of R2 x R = R3.
You can note that both S1 and [O, 11 are metric spaces with Euclidean metric
induced from R2and R respectively. Then the product mehc defined on X is
the metric defined by

Can you think of.other metrics on X1 x X2 coming from the original metrics on
X1 and X2? What about the D and p given by
..

Here is an exercise for you.

E8) ~ e(Xl,
t dl) and (X2,d2)be two discrete metric spaces. Then verify that
the product metric on X1 x X2 is discrete.

We hope that you have understood the new concept of distance. In what follows,
we shall try to see what other new concepts can be developed inthis context.

1.3 OPEN SETS

In this section y e study certain sbbbskts of metric spaces called open sets. We
b

first introduce two related terms: 'the distance of a point from a set', and 'the Introduction to Metric
diameter of a set'. Spaces

Let us consider a metric space (X, d). Let A be a non-empty subset of X. Then
for any x E X, consider the set A, = {d(x, y) : y E A). Then what can we say
about the set A,?
>
From the condition 2 in the definition of a metric, we have d(x, y) 0 for all
x, y E X. That means Ax is a non-empty set of non-negative real numbers which
is bounded below by 0. Then this set has a greatest lower bound (infimum).

Definition 5: Let (X, d) be a metric space. If x <X and A C X, A # 4, then

the distance of x from A, d(x, A) is taken as the greatest lower bound of the
set A, that is,
d(x, A) = inf{d(x, y) : y E A).

Remark 6: If x E A, then clearly, O E {d(x, y) : y E A), and hence d(x, A) = 0.

Let us see an example.

TT 7 r
[
Example 14: Let us consider the function space C 0, - ~ e t i].
g(t)=0,O<t<;andfn(t)=sinnt,O<t<;.LetA=[fn:n~N].Then
L L
<
you can observe that d(g, A) = 1. This follows from the fact that - 1 fn(t) <1
for all n E N and t E 0 and also that for any n, there exists some t,
O <t
m-

< r,
[ 7 5 1

such that fn(t)= 1.

2

For any non-empty subset A of a metric space (X,d), we consider another set
D(A) given by
D(A) = {d(al, a2) : all a2 E A).
Then D(A) is a non-empty set of non-negative real numbers and therefore has a
supremum ( or least upper bound) in the extended real number system. The
supremum of this set is called the diameter of A.
\

Definition 6: Let (X, d) be a metric space. The diameter of a non-empty

subset A of X which is denoted by d (A) is defined as the supremum of the set
D(A). That is,
d(A) = sup {d(al, a2) : all an E A).

Remark 7: Note that d(A) > 0 since D(A) is non-empty and consists of
non-negative numbers.
>
We have observed here that if A # 4, then d(A) 0. The diameter of a set may
not be a finite number. It is possible that d(A) = co.You can easily see that for
the set Z in R with the usual metric, d(Z) = oo.
So, what are the possible value of the diameter of a non-empty subset of X? It
could be a non-negative real number or it could be +oo.
If the diameter of a set is not infinity, it is called a 'bounded set'.
d
Metric Spaces Note that any non-empty subset of a bounded set is bounded. Also a non-empty
finite set is bounded. It is for these reasons that we term the empty set q5 also to
be a bounded set even though

There are metric spaces where any set is bounded.' Can you give an example?
Try these exercises now.

E9) If (X, d) is a metric space with the metric p given by

then any non-empty subset of X is bounded.

E10) Find d(A) in the context of R with the standard metric when
i) A = (O,2)
ii) A = the set of rationals in [I, 101.
El 1) Give an example of a set A together with an x $! A in a metric space (X, d)
satisfying d(x, A) = 0

E12) Give an example of an unbounded set in R2.

Now we shall discuss an important class of sets in a metric space (X, d).

You are already familiar with the concept of a ball in Euclidean space Rn.Let
us for example, consider R3. Then the open ball of radius r with centre
a = (al, a2,as) is the set B (a, r) given by

Now we extend this definition to a metric space (X, d).

Definition 7: Let (X, d) be a metric space. Let x E X and r > 0. The subsets

Bd(x, r) := {y E X : d(x,y ) < r} and Bd[x,r] := {y E X : d(x,y) 5 r}

are respectively called the open and closed balls centred at x with radius r with
respect to the metric d. We use the notation Bd(x,r) or Bd[x,r] only when we
want to emphasize that the metric under consideration is d. If no confusion
arises, we denote Bd(x,r) by B(x, r) and B ~ [ xS), by B[x, s].

Example 15: Consider X = R with the standard metric. Then we note that
B(x, r) = (X - r, x + r).

Example 16: Let X = R2 with the standard metric d. Then

B(0, r) = { ( X I , x2) E R2 : X: + X; < i2}. This is because
p := ( X I , XP) fi B(0, r) iff d(p, 0) < r iff {d(p, 0)j2 < i2iff X: + X; < ?. More
generally, if q = (al, aa), then B(q, r) is the set of points inside the circle of
radius r with centre at (al, a2).
 Introduction to Metric
Spaces

If we look at the shape of the unit balls defined by the three metrics on R2
considered in Examples 4, 5 and 6, we will be able to find the difference
between the metrics.
Note that in R2, the unit balls with respect to the three metrics are given

The following figure illustrates the balls.

CL2.
Fig. 7

Example 17: Consider C[O, 11, the set of conunuous rea valued functions on
[O, 11, with supremum metric,

How will you visualize B(f, a)for an f E C[O, 11 and a > O? For instance, let us
consider B(0, a).Note that here 0 denotes the function which is identically 0 on
ro, 11.
We first note that an element g E C[O, 11 lies in B(0, a)if and only if (or iff)its
graph lies in the region bounded by the lines y = fa and x = 0 and x = 1
excluding the horozontal lines y = fa.Similarly, g lies in B(f, a)iff its graph
lies in the region bounded by the curves y = f f cr and x = 0 and x = 1
excluding the curves y = f fa (see Fig. 8).

(a) (h)
Fig. 8
integral

How do we visualize open balls in this situation? For instance, let us consider
B(O, 1).
We shall look at the geometric meaning of the integral illustrated by Fig.9.
Then, g E B(0,l) iff the area 'under the graph' of (gJis less than I . So, we can
say that the function g given in Fig. 9(a) lies in B(0,l) whereas 4 in Fig. 9(b)
does not.

1- g is not in B(O.1)
y =-1

(a)
Fig. 9
***
Proposition 3: Let (X, d) be a metric space. Then given two distinct points
x, y E X, there exists r > 0 such that B(x, r) n B(y, r) = q5 (See Fig.10). (This
property is called (Hausdoe Property)). '

1
Proof: Since x and y are distinct d(x, y) > 0. Let 0 5 r 5 -d(x, y).
2
Let V = B(x, r) and W = B(y, r) (see Fig. 10). Then V and W are open in X
and x E V, y E W. Also V n W,= 4. Hence the result.
Try these exercises now. Lntroduction to Metric
Spaces

,B(2,3) in (R,p) where p is the metric given by

p(x, y) = mint17 lx - yll.
E14) Let (X, d) be a metric space with X # (0) and, x E X and 0 < r < s.
Show that B [x,r] c B[x, s]. Give examples to show that it is possible to
have
i) B[x,r]=B[x,s]
5
ii) BP,r] B[x,s]

.We now define open sets.

Definition 8: A subset U c X of a metric space (X, d) is said to be d-open if

for each x E U, there exists r > 0 such that B(x, r) c U. See Fig. 11.

If there is no confusion about which metric is being used, we shall simply refer
to as an open set rather than as a d-open set.

Fig. 11

Thus, we observe that to show that a set is open, we have to show that every
point in it is the centre of an open ball fully contained in the set.
Since the empty set 4 does not contain any point, we term the empty set as an
open set.
Now if x E X, B(x,r) C X for any r > 0. ~ e n cisiopen.
~
This means that in any metric space (X, d), 4 and X are open sets.
What are the other open sets in a metric space (X, d)? As the name suggests, we
expect that each open ball is an open set. How do we prove this? The next
proposition shows this.
Proposition 4: In anymetric space (X, d), each open ball is an open set.
Proof : Consider the open ball B( q , r) in X, and let x E B(xo,r). Then
d(x, q ) = rl < r.
i
Metric Spaces Now take r, = r - rl. Then r, is a positive real number.
Now let y E B(x, r,). Then d(y, x) < rx = r - rl.

... d(y, xo) 5 d(y, x) + d(x, xo)

< (r - rl) + rl
= r.
So y E B(x0, r). Thus we have shown that B (xo, r,) C B(Xo,r) (SeeFig. 12). So
we have proved that around every point of B(xo,r) we have an open ball, fully
contained in B(xo,r). This implies that B(xo,r) is open.

Fig. 12 Now we shall prove a theorem which brings out the strong connection between
open sets and open balls.
Theorem 1: Let X be a metric space. A subset of X is open if and only if it is
a union of open balls.
Proof : Suppose A is a subset of X, which is the union of a class U of open balls
of X. Now if U is empty, then A = 4, and is thus open.
Next, if U is non-empty, then A is also non-empty.
Let a E A. Then a belongs to some open ball B(xo,r) in U.As in the proof of
the proposition above we can choose r, s.t.

B(a,ra) c B(xo,r) c A
Thus, around every point of A we can find an open ball fully contained in A.
This shows that A is open.
On the other hand, suppose A C X, and A is open. If A is empty, then it is the
union of the empty class of open balls. If A is non-empty, then for every a E A
there is an open ball B(a, r,) which is fully contained in A. That is,
a E B(a,r,) c A.
So A = UB(x, r,) Let U be the class of all such opem balls B(x, rx),x E A.
xEA
Then A = Ux c UB(x, r.) c A. Thus A is the union of open balls in U.
xEA XEA

Next we shall consider an interesting aspect of openness. Let us consider

X = [O, 11 with the standard metric. In this metric space [0, I] is an 'open' set
being the whole set. But we know that the set [O, l] as a subset of R is not open
in R. Do not let this confuse you. Note that when we consider [O, 11 as a metric
space, our attention is focussed only on the points in [0, J] and thus we do not
take into consideration the other points of R. Using a similar arguement, we get
that B(0, 112) in this space is [O,lM[and not]-1/2,1/2[ as in R. In fact,
B(0,r) = X = [O,l], i f r > 1.
= [O,r[, ifr 5 1

This means that the openness of a set is a relative concept. It depends on the
metric space to which it belongs. We fuaher note that if r > 1then
D(B(0, r)) = 1 < 2r and if 0 < r 5 1then D(B(0, r)) = r < 2r. This means that
the diameter of a set and diameter of a ball in the usual sense axe not the same
 Next we shall see how open balls w.r.t. to the induced metric differ from the Introduction to Metric
open balls in the original space. Let A be a nonempty subset of a metric space Spaces
(X, d). Let us continue to denote by the same letter d the induced metric on A.
Let BA(x,r) denote the open ball in the metric space (A, d) with centre a and
radius r. Then BA(x,r) = B(x, r) n A, where B(x, r) is the open ball in X
centred at x with radius r.

Example 19: Let X be a discrete space. Then for x E X, {x) = B

therefore is open. By the above theorem every subset is open.

Before proceeding further we want to make an important point.

Let us consider R with the standtrd metric. Then, it is easy to check that {x) is
not an open set in this metric space. Whereas if we consider R with the discrete
metric, then {x) is open.
Once we have understood what an open set is, we are interested in the collection
of open sets. More precisely we would like to know whether the union or
intersection of open sets are open or not. Our next theorem explains this.
Theorem 2: In a metric space X, the following are true ,
i) An arbitrary union of open sets is open, and
ii) a finite intersection of open sets is open.
Proof : Let [A,],,1 be a class of open sets in X where I denotes an index set and
put A = UA,
and^ =
I€ I
n~, IEI
An index set is a set
which is used to make
i) If A = $ then it is open. Suppose A # 4. Let x E A. Then there exist some indices.
io E I such that x E Aio.Since A,, is open, there exists a b > 0 such that

This shows that A is open. By convention, if I is

ii) If I = $, then B = X, which is open. Also if B = 4, then it is open. U
empty, then Ai = $
I
i
Now suppose I # 4 and B # 4. We may take i = I, . . . , n. Let x E B.
Then x E Ai for each i. Since each Ai is open, we can find r, > 0, such that and n
iEI
i€I
= X.

I
I
B(x, r,) c A,.
Let r = min{rl, r2,r3,. . . ,rn). Then r > 0
Further, B (x, r) C B(x, ri) C Ai for each i = I, . . . n.
Therefore B (x, r) c nn

i=l
Ai = B.

So, around every element of B we can find an op'en ball fully contained in
B. Thus, B is open.

According to the theorem above, union of any arbitrary class (finite or infinite)
of open sets is open. But for intersections, the class is taken to be finite. In fact,
you will see in the example below that an infinite intersection of open sets need
not be open.
Example 20: Consider R with the standard metric. Let us consider the class
,?,f = (1 -A,![.
n n
E N) of open in R Then] -!,![is
n n
open for alln E N
But n]
n€N
- 1/n, 1/n[= {O) and is not open in R, since any

B(0, r) =] - r, r[,r > 0, will contain 1-12,and so B(0, r) @ {O).

Next we shall see another interesting feature of Ln open set.

Observe that an interval (a, b) is open if and only if all its points are
interior points. You can observe the same feature for discs in the plane also.
How do we express this for metric spaces? Let us see.

Definition 9: Let X be a metric space and A be a subset of X. A point x E A is

called an interior point if there exists an open ball B(x, r), fully contained in A.

The set of all interior points of A is called the 'Interior of A', and is denoted by
Int(A). Thus,
Int(A)= {x E A : B(x,r) c Aforsomer > 0).
We can prove that Int(A) is clearly an open subset of A (see E16). Infact Int A
is the largest open set in X contained in A. How do you show this? For that we
have to show that Int A contains all open subsets of X that are contained in A.
To see that, let us take an open subset A. of X contained in A. Then we have to
show that A. c Int(A). Let x E Ao. Since A. is open, there is an open ball
B(x, r), such that
x€B(x,r) CA.

This means that x is an interior point of A, i.e., x E Int(A). So, A. C Int(A).

This shows that Int A is the largest open set. ,

This leads us to the following result.

Proposition 5: A set A in a metric space (X, d) is open if and only if A = Int A.
Next we shall restate the Hausdroff in term of open sets rather than open ball.

Remark 8: If x and y are distinct points of a metric space X, then there exists
disjoint open sets V and W in X such that x E V and y E W.

So far we have been discussing open sets in a metric space. As we stated earlier,
the concept of openness changes from one metric to another. In this connection
we introduce you to another concept called equivalence of metrics

Definition 10: Two metrics dl and d2 on a set X are called equivalent if a set
A is open in X w.r.t. dl if and only if it is open w.r.t. d2.

Before considering examples we shall give a characterisation of the concept of

equivalence of metrics.
Theorem 3: TWOmetrics dl and d2 are equivalent if and only if given any
x E X and r > 0, there exist positive real numbers rl and r2 such that
 Introduction to Metric
Spaces

(b) .
Fig. 13

The proof of this follows from the definition of open sets.

Let us see an1 example.
Example 21: Let us check whether the Euclidean metric d on R and another
metric dl defined on R by

are equivalent or not.

Let us take an x E R and r > 0. Let rl = min{l, r}.
Since dl (x, y) Id(x, y) for all y E R, we get that Bd(x, rl) S Bdl (x, rl).
Now let y E Bdl(x, rl). Then dl (x, y) < rl. So dl (x, y) < 1. Thus
dl(x, y) = d(x, y) and therefore d(x, y) < tl 2 r. This shows that y E B ~ ( xr),

Hence the two metrics are equivalent.

***
Try some exercises now.

E15) Let (X, d) be a metric space and let a E X and r > 0.Can B[a, r] be an
open set? Justify your answer.

E16) Show that Int A is an open set.

E17)Let A1 and A2 be open sets in metric spaces (XI, dl) and (X,,dz)
respectively. Show that Al x A2 is open in the metric space (XI x Xz), d)
where d is the product metric defined in Section 1.2.Use this result to
show that a rectangle ]a, b[x]c, d[ is open in R x R.

E18) Show that d, D, p defined on the product space X = X1 x X2 (see Section

1.2)are equivalent metrics on X.

In the next section we shall consider sets which are complements of open sets.
* Metric Spaces
1.4 CLOSED SETS
A notion which is related to "openness" is "closedness". Here we shall define
closed sets in metric spaces. .
Before defining closed sets, we talk about 'limit points' of a set. We have the
following definition.

Definition 11: Let A be a subset of a non-empty metric space X. A point x E X

is called a limit point of A if every open ball centred at x contains a point of A
other than x. Limits points are also called "cluster points" or "accumulation
points"

Example 22: Let A = [O, 11 C R. Then 1 is a limit point of A, that is, every
open interval around 1 contains points of A, apart from 1. In fact, you will
realize that every point of [0, 11 is its limit point.

Example 23: Consider the set F = {1,1/2,1/3, . . .) in R. Then 0 is a limit

point of this set. In fact, 0 is its only limit point and it does not belong to F. No
other points of F are its limit point. For example, take 1/5 € F. The points of F
close to 115 on right and left sides are 114 and 116 respectively.
1 1
Then l j 4 - 1/5 = - and 1/5 - 1/6 = -.
20 30
Now, B (1/5, k) (1/5 - -,I15
1
30
+ - is an open interval around 115
30
I)
and it does not contain any point of F other than 115. So 115 cannot be a limit
point of F.

Example 24: Let us consider B = [3,4[.All points of B are its limit points. In
addition, 4 is also a limit point of B.

Example 25: Let X = R2 with the standard metric.

Let A be the union of x and y-axes; i.e: A = {(x, y) E R2 : xy = 0). Then all
the points of A are its limit points.
Let B be the set given by B = {(x, y) E R2 : x2 + y2 L 1,x # 0, y # 0). Then
(0,O) is a limit of B which do not lie in B.
***
From these examples, we see that a point in a set may or may not be its limit
point. On the other hand, a limit point of a set may or may not belong to it.
.We now define closed sets.

Definition 12: A subset A of a metric space X is said to be a closed set if it

contains all its limit points.

From the examples we d i s c u m ~ i e rwe , get that [I ,2] is a closed set in R.

But {1,1/2,1/3, . . .) and p, 4[ are not closed in R. What about the set
 P = {1,2,3,4)? Why don't you try to show that no real number can be its limit Introduction to Metric
point. Since it does not have a limit point, it is a closed set. Spaces

Consider a discrete metric space X. In this space, any open ball with radius less
than or equal to 1 contains only its centre. It cannot contain any other point of
X. This means no subset of X can have a limit point. So, a discrete space is a
metric space in which every subset is a closed set. When we compare this with
E(17),in which you have been asked to show that every subset of a discrete
space X is open, we get that in a discrete space every set is both open and
closed in a discrete space.
Here we want to ask a question. What are the sets which are both open and
closed in R?
You will find an answer to this in Unit 4.

k Unbelievable in the real sense, isn't it? This is what makes a set different from
real objects like "Doors"!

rn Theorem 4: A subset A of a metric space is closed, if and only if its

complement A' is open.

Proof : Suppose A is closed. This means it contains all its limit points. That is,
no point of A" can be its limit point. Thus, if x E A", then we can find an open
ball B(x,r) which does not contain any point of A, which means that it is fully
contained in A'. This proves that A" is open.
Now suppose A" is open. Let x be a limit point of A such that x 4 A. Then
x E Ac and since A" is open, we can find an open ball B(x, r) contained in A".
That means B(x, r) does not contain any point of A. This is not possible since x
is a limit of A. Thus all the limit points of A lie in A. Hence A is closed.

Example 26: Let A = {(xl, x2,0) : (xl, x2) E R 2 ) . Then A is closed in R3.

Let x =(xl, x2,O) where (xl, x2) E R2.Then x E A. For r > 0, put

r i).
y = ( x I , x ~ , Then y E B(x,r). But y $! A. SoB(x,r) A. Thusx $! IntA
and therefore A is not open. Hence A is closed.
***
r Proposition 6: Let (X,d) be a metric space. Every closed ball B[x, r] in X is
closed in X.
Proof: Suppose y is a limit point of B[x, r], and that d(y,x) = rl. Let, if
possible, rl > r. This means r2 = rl - r > 0, Now consider B(y, rz). This open
ball around y must contain a point, say z, of B[x, r]. Then

d(x, z) 5 r and d(y, z) < 1-2.

Hence, rl = d(y, x) F +
d(y, z) d(x, z)
< r2+r
= (rl-r)+r
-
- r1.

This is a contradiction. Hence our assumption that rl > r is not possible. There-
fore rl < r, and hence y E B[x, r]. That is B[x,r] contains all its limit points.
Therefore B[x,r] is a closed set in X.
Metric Spaces Theorem 5: In a metric space X, the following results hold:
i) Any arbitrary intersection of closed sets is closed, and
ii) A finite union of closed sets is closed.
Proof : Let {Ai)iGlbe a collection of closed sets. Then is collection of
We will be using De U
open sets. Thus A: is open and therefore by De Morgan's law.
Morgan's law quite often
without actually referring . .
id

to it always.

is closed. Hence the result. We leave the proof of the second part of this theorem
as an exercise for you to verify (Refer E19).

We now define a related concept.

Definition 13: Let A be a subset of a metric space X. The union of A with the
set of its limit points is called the closure of A and is denoted by A.
Obviously, if A is closed, then A = x. In general, A c A.
Theorem 6: If A is a subset of a metric space X, then is closed in X.

Proof : Suppose x is a limit point of x.

If x is in A, then x E x.
Let us assume
that x $ A. Since x is a limit point of x , given r > 0, the open ball B(x, r)
contains an element y of other than x.

Let d(x, y) = rl and let r2 = r - rl . Then r2 > 0. Since y E A,the open ball
B(y, r2) contains an element z of A other than y. Then

This shows that z E B(x, r). Also z # x simply because x $ A. Thus we got that
x is a limit point of A and therefore x E x.
Since A contains all its limit points, A is closed. Hence the result.

Remark:9 A set is A closed, if and only if A = A.

Using the concept of closure of a set in a metric space, we define another

concept.

Definition 14: Let A be a subset of a metric space X. If a point x is such that

every open ball around x contains a point of A and a point of A", then x is called
a boundary point of A. The set of all boundary points of A is called the
boundary of A and is denoted by bdry(A).

Remark:lO It follows immediately from the definition that bdry(A) = ;dry(~")

For exqmple, the boundary of ]a, b[ in R consists of the points a and b.
The boundary of B(x,r) in C consists of the circle with centre x and radius r.
. - r. -.-
 I

The following figure illustrates that the definitions of boundary, closure and Introduction to Metric
interior coincides with our geometric intuition, for some specific subsets of R2 Spaces
with the standard metric.

Fig. 14

But this is not the case always. We shall illustrate this with an example. Before
that we establish the following result.
Now we prove a result relating to closure, interior and boundary.
Proposition 7: If A is a subset of a metric space X then
bdry(A) = - Int(A).

Proof : We will first prove that

bdry~cx-Int~
Let x E bdry (A). Then it is immediate from the definition that x E x.
Also x 4 Int A since every open ball B(x, r) contains an element of A".
Therefore
bdry A c -6;- Int A, (9)
Now we have to show that x-
Int A C_ bdy A. Let y E x- Int A. Let B(y,
r) be an open ball. Since y E x,
B(y, r) contains an element of A. Suppose
B(y, r) does not contain any element of A". That means B(y ,r) C_ A. So by
definition y E Int A open set and Int A is the largest open set contained in A,
we get that B(x, r) C Int A this is not possible since y $ Int A. This shows
that B(y, r) intersects with ACand A. Therefore y E bdry(A). Thus we got that
- Int A bdry(A). This together with (9) shows that

Example 27: Let us find the interior, closure a'nd boundary of the following
sets in the given spaces
i) B = {(x, y) E R2: y = 0) in R2.
We shall first show that Int B = 4. Let if, possible there exists
p = (%, yo) E Int B. Then there exists r > 0 such that B(p, r) c B. Let
Metric Spaces
?
q = (xu, i). Then

So q E B(p, r). But q $ B, which is a contradiction. Thus Int B = d.

Next we shall show that B = B. Let C E B\B. Let C = ( a ,P ) since

C E B, ( I:)'.
1 > 0. Then consider B C, - hen^ (c,?) n B = d . So
C $ B, a contradiction. Thus B = B.

From the Proposition 7 above we get that bdry B = B.

***
Here are some exercises for you.

El 9) Complete the proof of Theorem 4.

E20) Show that any finite subset of a metric space is closed.

E21) Show that A = {x E X : B(x, r) n A # for every r > 0).

E22) -
If X is a metric space and A is a non-empty subset of X, then show that
A = { x : d(x,A) = 0).

In the next section we shall discuss continuous functions in metric spaces.

1.5 CONTINUOUS FUNCTIONS

The concept of continuous functions is one of the very basic concepts in
Mathematical Analysis.

You are already familiar with this concept for the Euclidean spaces R, R2 and
so on. In this section we shall formulate a definition of continuity for general
metric spaces that will include these spaces. Then we shall study some
properties.

You might be already familiar with continuous functions from R" -+ Rmwith
the standard metrics (see IGNOU Maths Course MTE-07). Here are some
examples.
i) The function t -+ (cos t, sin t) from R to R2 is continuous on R,
ii) The function (x, y) -+ (cos x, sin x sin y, exsin y) from R2 + R3 is
continuous on R2.
Like these you must have seen many continuous functions from one Euclidean
space to another Euclidean space. You know that in the case of real line R there
are different ways of defining continuity. For instance, " E - 6 definition"
"convergent sequence definition" and so on. We shall try to generalise these
definitions to metric spaces.
Before we do this we shall familiarise you with some notations.
Suppose (X, dl) and (Y, d2) are two metric spaces and consider a function Introduction to Metric
f : X -+ Y. Let A be any subset of X. Then the image of A by f is the set Spaces
4

In other words, f(A) is the set of the values that f assigns to the elements of A.
Suppose that B is a subset of Y. Then the inverse image of B denoted by f-'(B),
is given by

) {x E X I f(x) E B).
t ' ( ~=

Now we shall try to generalise the E - 6 definition of continuity. To do this the

best way is to recall the definition of continuity of functions on R.
Let f be a real-valued function defined on R. Then f is said to be continuous at
xo E R if given F > 0, 3 a 6 > 0 such that

Also a function f is said to be continuous on a set A in R iff is continuous at all

points of A.
We can easily extend this definition to general metric spaces by replacing the
modulus Ix - y I by the distance d(x, y)
This gives the following definition.

Definition IS: Let (X, d l ) and (Y, d2) be two metric spaces and f : X + Y.
Then f is said to be continuous at a point xo E X, if given E > 0 3 a 6 > 0 such
that

Equivalently, given E > 0 there exists a 6 > 0 such that

x E B(xo, 6) * f(x) E B(f(xo), f)

A function f is said to be continuous on X if it is continuous at each point of X.

Such functions are called continuous functions.
It is immediate from the definition that the following functions are continuous
Example 28: Let X = B(T) with the metric defined in Example 9. Let
$ : X -+X be given by d(f) = f2. Then 4 is continuous.

Let fo E X. Put go = 4(fo)

Let M = slip fo(t)
t €T

Consider any f E X and put g = 4(f).

We note that for t E T,

g(t) - go(g) = (f(t) - fo(t>I2+ 2(f(t) - fo(t))fo(t),

and thus
 Choose 6 = min
{d+ l 11)
Then for d(f,fo) < 6, we have d(g, go) < c.
So 4 is continuous at fo. Since fo is arbitrary, 4 is continuous.

Many of you might have faced problems with E - 6 definitions of limit and
continuity in your earlier Calculus and Analysis courses. The following real-life
situation may help you to have a better understanding.

Situation 1: Let X be a set of ingredients or input and Y is the output and f is

the process or precedure of a recipe which takes an input x and produces an item
(or output) y. The end user or consumer needs y E Y. The manufacturer knows
c is the Greek letter for e that to get y, hetshe can put the input x and by applying the process f to x, she
and e is for error in the gets y = f(x). But now in real-life, there is no guarantee that the manufacturer
k ouput and 6 is the Greek will give precisely the output required as there is no surety that the manufacturer
6 letter for d and d is the will be able to put the input exactly xo or apply the process correctly. The
1
C "differenbe in the input" customer knows these limitations and therefore is ready to accept the output if
the output does not vary with hisher expection too much. In this case the
customer fixes some tolerance level and decides that if the manufacturers
delivers the product within that tolerance level, it will be accepted. From the
experience, the manufacturer finds a 6 such that if the input is within 6 distance
from a fixed input xo, the process will produce f(x) within the tolerance level
for f(xo). Note that in such situation the error level upto r of the output that can
be tolerated is given, then we find difference level upto 6 of the input. Now the
continuity of the process infact tells us that given E > 0 we can find such a 6,
such that the process gives the desired result within the tolerance level E.

The above situation might have helped you to understand the E - 6 definition of
continuity in a better way. By now you must be clear about which one out of E
or 6 in the definition of continuity, needs to be considered first. It is important to
know this because this is what is used by scientists and engineers.

Now we shall consider some examples.

Example 29: Let (X, d) be a metric space. For any fixed a E X, define a
function fa : X --+ R by

Theti fa is continuous on X.
To see this, let us apply E - 6 definition.

Let xo E X. Let E > 0. Then we have to show that there exists a 6 > 0 such that

d(x, xo) < 6 * Ifa(x)- fa(xo)l < E

 But, by the definition of the function fa Introduction to Metric
Spaces
Ifa(x) - fa(xo)I = Id(a, x) - d(a, xo)l 5 d(x, xo) (10)
So we may choose S = 6 . Then from Eqn.(lO) we get that

This shows that fa is continuous on X.

Another interesting continuous function is illustrated in the next example.

Example 30: Let A be a nonempty subset of a metric space (X, d). We define a
function on X to R by dA on X to R by

.dA(x)= d ( ~A)
,
.r

as given Definition 5 of Section 1.3. That is,

Then dAis continuous.

To show that dAis continuous, we observe that, for any arbitrary x, y E X and
a E A, the triangle inequality implies that

d(a, Y) 2 d(a, x) - d(y, x) (1 1)

> dA(x) - d(y, x), (by the definition of dA(x)) (12)

Taking infimum on both sides of (12) for a E A, we have

Thus,

~ A ( x-) ~ A ( Y5)d(y, x).

If we interchange x and y in this inequality, we see that
&(dA(x)- dA(y))5 d(x, y) that is, ( ~ A ( x-)d ~ ( y )<_( d(x, y). From this
inequality the continuity of dAcan easily be shown.

We give some exercises now.

E23) The identity function i : X X defined by i(x) = x V x E X is

continuous.

E24) Let (X, dl) and (Y, dZ)be two metric spaces. For any fixed c E Y the
function f : X 4 Y defined by f(x) = c for all x E X is continuous. (f is
called a constant function.)

E25) Let X = C[O, I] with sup metric. Then the following functions are
continuous
 4 : X -+R given
$(f) = / 0
1
f(t)dt, f E
is continuous.
ii) 4 :X -+ R given by 4(f) = f (i) is continuous.

Next we shall discuss a characterisation of continuous function.

Theorem 7: Let (X, dl) and (Y, d2) be metric spaces. Let c E X, Then the
following are equivalent.
(a) f is continuous at c.
(b) Given an open set V containing f(c) in Y, we can find an open set U
containing c in X such that f(U) C V.

Proof : We shall first prove that (a)=+-(b)

7
?
Let f : X 4 Y be continuous at c. We have to show that f satisfies (b).
Let V be an open set containing f(c) in Y. Since V is open and f(c) E V, there
exists an c > 0 such that the open ball B(f(c), 6 ) C V. Since f is continuous, 11
given c > 0, 3 6 > 0 such that !

1
dl(x, c) < 6 =+-dz(f(x),f(c)) < E .
Let U = B(c, 6). Then U is an open set containing c. We claim that f(U) c V. i
To show this let us consider b E f(U). Then v = f(x) for some x F-U. This
implies that

dl (x, c) < 6 =+-d2(f(x),f(c)) < E i.e. f(a) f B(f(c), E ) .

Thus we have shown that f(U) c V.
This shows that (a) + (b).
Next we have to show that (b) + (a).
4

Let f be a function such that (b) holds. We have to show that f is continuous at
c E X. Let e > 0 be given. Now we apply the condition (b) to the open set
,-
B(f(c): E).Then we get that 3 an open set U containing c such that

f(U) c B(ftc), 4.
Since U is open and c E U, 3 a 6 > 0 SUC that
the fact that f(U) 2 B(f(c), E), shows tha1

d l b , c) < 6 + dz(f(x),ftc)) < (1 3)

Thus we have shown that given E > 0 36 > 0 such that (13) holds. This shows
that f is continuous. 0
From the above theorem we can easily establish the following theorem.
Theorem 8: Let X, Y be metric spaces. Then a map f : X -+ Y is continuous
if and only if for every open set V c Y, its inverse image f-'(V) is open in X.
The proof of this theorem is left as an exercise for you.
 Remark 11: This characterisation of continuous function is very important (or Introduction to ',I:?xi$
useful) because this does not involve the notion of "distance function (metric)" SE':l"""
directly; it mainly considers sets which are open. Infact this creteria is used in
defining continuity in more general setting like topological spaces.
Now we want you to try some exercises based on the discussions above.

E26) Prove Theorem 8.

E27) Show that the result in Theorem 7 is true if we replace open sets by closed
sets.

E28) Let (X, dl), (Y, d2) and (Z, dJ) be three metric spaces. Let f : X + Y be
. continuous at x E X and g. Y + Z be continuous at y = f(x). Then
composite map gof : X + Z is continuous at x E X.
r

E29) Let f : (X, dl) + (Y, d2) be continuous. Let A c X. Show that the
restriction fJAis a continuous function from (A, dl) to (Y, d2).
1

E30) Let pl : R2 --+ R and p2 : R2 + R be the functions such that pl(x, y) = x

and pz(x,y) = y. (pl and p2 are called coordinate projections). Show that
pl and p2 are continuous on R2.

E3 1) Let (X, d) be a metric space and f : X + R be a continuous function. -

Then show that for each t E R the sets

v, = {X E X : f(x) < t)
w,= {X E X : f(x) > t)
are open sets in X.
/

Next we shall state some basic properties of continuous functions which you
can verify by yourself.
Theorem 9: Let f, g : X 4 R be continuous at a E X. Then the functions
f+g,f-g,f.gdefinedby

(f + g)(x) = f(x) + g(x), x E X

(f - g)(x) = f(x) - g(x),x E X
f . g(x) = f(x) . g(x), x E X

f
are all continuous on X at a E X. If g(x) # 0 for all x E X, then - is also
g
continuous at a E X.
Now we shall prove a theorem which shows that two mutually disjoint closed
sets in a metric spaee ean be separated by a continuous funetion. The theorem is
called Urysohn's Lemma. Lemma is a proved
statement used as a
Theorem 10: (Urysohn's Lemma): Let A, B be two disjoint non-empty stepping-stone towards
closed subsets of a metric space X. Then there exists a continuous function the proof of another
<
f : X + RsuchthatO f 5 l , f = Oon Aandf = 1 onB. statement 35
3
i4
Metric Spaces Before we start proving this we recall that dAand dBare continuous on X which
vanish on A and B respectively. So various combinations of them can work as
the required function. For example, any a > 0 and consider f given by,

P b f : Let f(x) = dA(x) . We first show that f is defined for all x E X.

+
d~ (x) dB (x)
Let x E X. For that we note that dA(x)+ dB(x) # 0. For, otherwise, each of
the terms, being nonnegative, must be zero. But then this means that x E A and
x E B (recall that by E21, d(x, A) = 0 iff x E x). This a contradiction since A
+
and B are disjoint. Thus ~ A ( x )dg (x) # 0 for each x E X and we conclude that
f(x) makes sense for each x E X.
Since dA and dB are continuous real-valued functions on X by Theorem 8, f is
continuous on X. Clearly, 0 < f < 1.
Also if x E A, then dA(x) = 0 so that f(x) = 0.If x E B,.then the denominator of
+
f(x) is dA(x) dg(x) = dA(x) # 0 and therefore f(x) = 1for x E B. Hence f is
a required continuous function.
Try this exercise now.

E32) Deduce the.following result from Urysohn's lemma.

Let A and B be nonempty disjoint closed subsets of a metric
space (X, d). Then show that there exist open sets U > A and
V 1 B s u c h t h a t U n V = 4.
.-

Next we shall consider another interesting property.

Lemma (Gluing Lemma): Let X and Y be metric spaces.
Let {Ai : i E 1) be a family of non-empty open sets such that UAi
1€1
= X. Let

{fi : i E I) be a family of functions satisfying

i) for each i E I, domain of fi = Ai and fi : Ai + Y is continuous, and
ii) foreachi,j E I,fi(x) = f j ( x ) f o r a l l x E A i n A i .
Then the function f : X + Y defined by setting f(x) := fi(x) if x E Ai is
well-defined and continuous on X.
Proof : Let Here we use Theorem 8 B c Y be open. Then we have to show that
f-'(B) is open in X. We observe that for any fixed, i, f-'(B) n Ai = f ; ' ( ~ )n Ai.
ThisisbecausexEf '(B)flAiiffx€Aiandf(x) GB,iffx €Aiandfi(x) EB,
iff x E ~i n F'(B).
Since fi : Ai + Y is continuous, the set f;-'(B) n Ai is open in Aip But, Ai is
open in X. So f-'(B) fl Ai is open in X. Now

Thus it follows that f '(B) is the union of open sets {f-'(Bi) n Ai,i E I) and
hence is open. This proves the continuity off on X.
 It is natural to ask whether such a continuous gluing is possible for closed sets. Introduction to Metric
The following example shows that it is not possible in general. Spaces

For n E N,let us take

Now let X = UA, = {O) U considered as a subspace of R.

nzo
Then each A, is closed in X.
For each n > 0. define a function f, on A, given bv

and

Then for each n, fn is a continuous functioil considered as a function on the

closed subspace A,. Since An's are mutually disjoint f,'s can be glued together
to give the function f on X given by

and

. .

This function f is not continuous at 0.

This shows that continuous gluing is not possible for infinitely many functions
defined and continuous on infinitely many closed sets considered as subspaces
of X. However, it is possible for finitely many functions as the following
theorem shows.
Theorem 11: Let X and Y be metric spaces. Let Al, A2,. . . ,A, be a finite
n 2.
U
family of closed sets such that A, = X. Let {fi, i =.I, . . . , n) be a finite
i= 1
family of functions satisfying
i) for each i = 1,. . . ,n, domain of fi = Ai and f, : Ai -+ X is continuous.
ii) fori,j = 1,..., n,fi(x) =fj(x),x E A i n A j .
Then the function f : X -+ Y defined by setting f(x) = fi(x) if x E Ai is
I
well-defined and continuous on X.

Proof : The proof of this is very similar to the earlier proof, except that we use
the characterization of continuity by means of inverse images of closed sets. Let
C c Y be a closed set. We shall show that f-'(C) is closed in X. As in the
earlier case, we find that f-'(C) fl Ai = f i - ' ( ~nAi
) and that it is closed in Ai and
~ e t r i Spaces
c hence in X. We then express f--'(C) as a finite union of closed sets of the form
as follows:

This completes the proof. 0

We consider an example to illustrate the lemma.
Example 31: Let X = [0, 11 considered as a subspace of R. Let Y be any
I]--
metric space. k t f, g : (0, ntinuous functions.
f ( l ) = g(O).

10,i] Li,11 considered as closed subspaces of X. Then

I
Put Al = and A2 =
X = Al UA2.
Let fl(t) = f(2t),t E Al
1
f2(t) = g(2t - I), t E A2

Then fl, f2 are both continuous in their respective domains.

Then by the above lemma the function h defined by - .
1I
is continuous on X.
We infact note that the function h is given bv 1

and can-be called the join of fi and f2.

**
Now we define uniform continuity for a general metric space on the same lines
as for the real line. I
1

f
Definition 16: Let us consider a function f : (X, dl-) -+ (Y, d2). We say that f is
uniformly continuous on X if for a given E > 0, there exists a b > 0 such that
whenever xl, x2 E X are such that dl (xl,x2) < 6, we have d2(f(xl),.f(x2))< E .

Remark 12: It follows immediately from definition that a uniformly continuous

function is continuous. It is also immediate that if dl is the discrete meuic then
by taking 6 = 1in definition 14 f is uniformly continuous.
Remark 13: Here you may recall that a continous function is uniformly
continuous on a closed bounded interval. However a real valued continuous
function defined on other binds of subsets may or may not be uniformly
continuous. For instance you may recall the following: ;
Introduction to Metric
Spaces
Metric Spnces

40
 Introduction to Metric
ii) Conditions 1 and 2 are immediate from the definition of p. To check
the triangle inequality we take any let x, y, z E X. Since p(x,.z) 5 1 Spaces
the required inequality

is satisfied if either p(x, y) = 1or p(y, z) = 1. The only other

possibility is that p(x, y) < 1 ahd p(y, z) < 1. In this case both
p(x3 Y) = d(x, Y) and P(Y,z) = d(y7 2 ) -
So

Hence the condition 3 is satisfied. Thus p is a metric on X.

E7) Hint : Use Remark 4
E8) Try it by yourself
E9) Try it by yourself
EIO) Hint :
i) To get the answer d(A) = 2, the observation that for 0 < a < 1, helps.
ii) To get the answer d(A) = 9, the observation that for
n E N, 1 + n1 10 - -n1 E A, helps.
-:

El I) Iiint : Try the following

i) A = ( 1 , 2 ) , x = 1

E12) Hint : Try Z x Z and (3) x R for unbounded sets in R2.

E13) Hint : Try to get the answers given below:
3 5
i) [5,i]
ii) R
E 14)Try it by yourself.
El 5) Try it by yourself
E16) Let x E Int A. Then 3 r > 0 such that B(x, r) c A. Let y E B(x, r) put
rl = r - d(y, x), then B(y, rl) 2 B(x, r) c A. So y E int A. Thus we have
B (x, r) C Int A. Since x E A is arbitrary we conclude that Int A is open.
E17) To show that Al x A:! is open in X1 x X2. Let x = (xl, x2) E Al x A2.
Then xl E A1 and x2 E AZ. Since Al is open, 3 rl > 0 such that
B(xl, rl) C A l . Similarly 3 r2 > 0 such that B(x2, 1-2)C A2. Let
r = min(rl, r2). Then B(x, 6 ) is an open ball with centre x. Also for any
y = (yl, y2) E B (x, r), we have dp(x,y) < r. This implies that
d(x1, yl) < r and d(x2, y2) < r. That means yl E B(xI, rl) c A, and
y2 E B(x2, r2) C A2 and therefore (yl, y2) E A1 x A2. This shows that
B(x, r) C Al x A2. Thus Al x A2 is open.
E18) Hint : d < D 5 2d
Metric Spaces E19) To get the reverse inclusion, consider any x E X such that d(x, A) = 0.
Let r > 0 by the definition of a(x, A), then exists y E A with d(x, y) < r.
x.
Thus B(x, r) n A # 4. So x E Thus we have [x : d(x, A) = O] c A.
This completes the proof.
E20) We first show that {x) is closed. Let A = {x). Then it is enough to show
that A" is open. Let y E AC.Then d(x, y) >-0.Let d(x, y) = r, for some
r
real number r and rl = - Now consider B(y ,rl ) . Then y E B (y, rl ) and
2'
B(y, rl) C AC.This shows that ACis open and therefore {x) is closed.
Then by Theorem 4, any finite set is closed.
E21) Let x E A. We will show that d(x, A) = 0. Let, if possible, d(x, A) > 0.
a
Let r = - and consider B(x, r). Since x E A, 3 an y E B(x, r) nA.
2
So

-
This
- is a contradiction. Thus d(x, A) = 0.Thus we have
A C {x:d(x,A) = 0 )
Let B(x, r) be a balI containing x. To show that B(x, r) n A # 4. Let
B(x, r) n A =-4. This implies that d(x, y) > r for all y € A which
contradicts the fact that d(x, A) = 0.Therefore B(x, r) n A # 4. Hence
xE x.
E22) Try by yourself.
E23) Let {Ai)icIbe a finite collection of dosed sets. Then {lqe}iEI is an open
Ff. Each is operi. :. nn

i=l
is open. Hence the
result.
E24) Try by yourself
E25) Hint :
i) Use the inequality

for f E C[O,11
ii) Try it by yourself.
E26) Take x E f '(V). Then y = f(x) f V. Since V is open, it contain an open
ball B(y, E) for some E. Since f is continuous given > 0, there exists
5 > 0 such that x1 E B(x, 5) + f(xl) E B(y ,E) .This means that

Therefore f-'(V) is open in X.

E27) Hint : Use the fact and that a set is open if and only if its complement is
closed.
E28) Hint : If A is open in Z, then g-'(A) is open in Y and therefore
f'(g-' (A)) is open in X. But
 E29) Hint : The function flAequals the composite of the inclusion.rn Introduction to Metric
j : A -* X and the map f : X + Y. Therefore it is continuous. Spaces

E30) Hint : Use the fact that, for A R

E31) Kint : Note that V, = f-I{(-oo, t)) and W,= f'(t, oo)) and
(00,t), (t, oo) are open sets in R.

E32) Hint : Try

I E33) Hint : A careful look on the last line in the Example 30 helps.