1.

3 Open Set, Closed Set, Neighborhood 17

in agreement with Prob. 8. Show that for any x, y EX,

ID(x, B) - D(y, B)I;;; d(x, y).

11. If (X, d) is any metric space, show that another metric on X is defined
by

d(x y)= d(x,y)

, 1+d(x,y)

and X is bounded in the metric d.

12. Show that the union of two bounded sets A and B in a metric space is
a bounded set. (Definition in Prob. 6.)
,
13. (Product of metric spaces) The Cartesian product X = Xl X X 2 of two
metric spaces (Xl> d l ) and (X2 , dz) can be made into a metric space
(X, d) in many ways. For instance, show that a metric d is defined by

14. Show that another metric on X in Prob. 13 is defined by

15. Show that a third metric on X in Prob. 13 is defined by

(The metrics in Probs. 13 to 15 are of practical importance, and other metrics

on X are possible.)

I . :J Open Set, Closed Set, Neighborhood

There is a considerable number of auxiliary concepts which playa role

in connection with metric spaces. Those which we shall need are
included in this section. Hence the section contains many concepts
(more than any other section of the book), but the reader will notice
18 Metric Spaces

that several of them become quite familiar when applied to Euclidean

space. Of course this is a great convenience and shows the advantage
of the terminology which is inspired by classical geometry.
We first consider important types of subsets of a given metric
space X = (X, d).

1.3-1 Definition (Ball and sphere). Given a point XoE X and a real
number r> 0, we define 6 three types of sets:
(a) B(xo; r) = {x E X I d(x, xo) < r} (Open ball)
(1) (b) B(xo; r) = {x E X I d(x, xo) ~ r} (Closed ball)
(c) S(xo; r) = {x E X I d(x, xo) = r} (Sphere)
In all three cases, Xo is called the center and r the radius. •

We see that an open ball of radius r is the set of all points in X

whose distance from the center of the ball is less than r. Furthermore,
the definition immediately implies that

(2) S(xo; r)=B(xo; r)-B(xo; r).

Warning. In working with metric spaces, it is a great advantage

that we use a terminology which is analogous to that of Euclidean
geometry. However, we should beware of a danger, namely, of assum-
ing that balls and spheres in an arbitrary metric space enjoy the same
properties as balls and spheres in R3. This is not so. An unusual
property is that a sphere can be empty. For example, in a discrete
metric space 1.1-8 we have S(xo; r) = 0 if reF-I. (What about spheres of
radius 1 in this case?) Another unusual property will be mentioned
later.
Let us proceed to the next two concepts, which are related.
1.3-2 Definition (Open set, closed set). A subset M of a metric space
X is said to be open if it contains a ball about each of its points. A
subset K of X is said to be closed if its complement (in X) is open, that
is, K C = X - K is open. •
The reader will easily see from this definition that an open ball is
an open set and a closed ball is a closed set.

6 Some familiarity with the usual set-theoretic notations is assumed, but a review is
included in Appendix 1.
1.3 Open Set, Closed Set, Neighborhood 19

An open ball B(xo; e) of radius e is often called an e-

neighborhood of Xo. (Here, e > 0, by Def. 1.3-1.) By a neighborhood 7
of Xo we mean any subset of X which contains an e-neighborhood of
Xo·
We see directly from the definition that every neighborhood of Xo
contains Xo; in other words, Xo is a point of each of its neighborhoods.
And if N is a neighborhood of Xo and N eM, then M is also a
neighborhood of Xo.
We call Xo an interior point of a set Me X if M is a neighborhood
of Xo. The interior of M is the set of all interior points of M and may
be denoted by ~ or Int (M), but there is no generally accepted
notation. Int (M) is open and is the largest open set contained in M.
It is not difficult to show that the collection of all open subsets of
X, call it fT, has the follpwing properties:

(Tl) 0 E <Y, XE <Yo

(T2) The union of any members of fJ is a member of fl:

(T3) The intersection of finitely many members of fJ is a member

of fl:

Proof (Tl) follows by noting that' 0 is open since 0 has no

elements and, obviously, X is open. We prove (T2). Any point x of the
union U of open sets belongs to (at least) one of these sets, call it M,
and M contains a ball B about x since M is open. Then B c U, by the
definition of a union. This proves (T2). Finally, if y is any point of the
intersection of open sets M b ' •• ,Mm then each ~ contains a ball
about y and a smallest of these balls is contained in that intersection.
This proves (T3). •

We mention that the properties (Tl) to (T3) are so fundamental

that one wants to retain them in a more general setting. Accordingly,
one defines a topological space (X, fJ) to be a set X and a collection fJ
of sUbsets of X such that fJ satisfies the axioms (Tl) to (T3). The set fJ
is called a topology for X. From this definition we have:

A metric space is a topological space.

I In the older literature, neighborhoods used to be open sets, but this requirement
hus heen dropped from the definition.
20 Metric Spaces

Open sets also play a role in connection with continuous map-

pings, where continuity is :;I. natural generalization of the continuity
known from calculus and is defined as follows.

1.3-3 Definition (Continnous mapping). Let X = (X, d) and Y = (Y, d)

be metric spaces. A mapping T: X ~ Y is said to be continuous at
a point Xo E X if for every E > 0 there is a 8> 0 such that 8 (see Fig. 6)

d(Tx, Txo) < E for all x satisfying d(x, xo)< 8.

T is said to be continuous if it is continuous at every point of X. •

Fig. 6. Inequalities in Del. 1.3-3 illustrated in the case of Euclidean planes X = R2 and
¥=R2 -

It is important and interesting that continuous mappings can be

characterized in terms of open sets as follows.

1.3-4 Theorem (Continuous mapping). A mapping T of a metric

space X into a metric space Y is continuous if and only if the inverse
image of any open subset of Y is an open subset of X.

Proof (a) Suppose that T is continuous. Let S c Y be open and

So the inverse image of S. If So = 0, it is open. Let So ¥- 0. For any
Xo E So let Yo = Txo. Since S is open, it contains an E -neighborhood N
of Yo; see Fig. 7. Since T is continuous, Xo has a 8-neighborhood No
which is mapped into N. Since N c S, we have No c So, so that So is
open because XoE So was arbitrary.
(b) Conversely, assume that the inverse image of every
open set in Y is an open set in X. Then for every Xo E X and any

8 In calculus we usually write y = [(x). A corresponding notation for the image of x

under T would be T(x}. However, to simplify formulas in functional analysis, it is
customary to omit the parentheses and write Tx. A review of the definition of a mapping
is included in A1.2; cf. Appendix 1.
1.3 Open Set, Closed Set, Neighborhood 21

(Space Xl (Space Yl

Fig. 7. Notation in part (a) of the proof of Theorem 1.3-4

e -neighborhood N of Txo, the inverse image No of N is open, since N

is open, and No contains Xo. Hence No also contains a 5-neighborhood
of xo, which is mapped into N because No is mapped into N. Conse-
quently, by the definitjon, T is continuous at Xo. Since XoE X was
arbitrary, T is continuous. •

We shall now introduce two more concepts, which are related. Let
M be a subset of a metric space X. Then a point Xo of X (which mayor
may not be a point of M) is called an accumulation point of M (or limit
point of M) if every neighborhood of Xo contains at least one point
Y E M distinct from Xo. The set consisting of the points of M and the
accumulation points of M is called the closure of M and is denoted by

M.
It is the smallest closed set containing M.
Before we go on, we mention another unusual property of balls in
a metric space. Whereas in R3 the closure B(xo; r) of an open ball
B(xo; r) is the closed ball B(xo; r), this may not hold in a general
metric space. We invite the reader to illustrate this with an example.
Using the concept of the closure, let us give a definition which will
be of particular importance in our further work:

1.3-5 Definition (Dense set, separable space). A subset M of a

metric space X is said to be dense in X if

M=X.
X is said to be separable if it has a countable subset which is dense in
X. (For the definition of a countable set, see A1.1 in Appendix I if
necessary.) •
22 Metric Spaces

Hence if M is dense in X, then every ball in X, no matter how

small, will contain points of M; or, in other words, in this case there is
no point x E X which has a neighborhood that does not contain points
of M.
We shall see later that separable metric spaces are somewhat
simpler than nonseparable ones. For the time being, let us consider
some important examples of separable and nonseparable spaces, so
that we may become familiar with these basic concepts.

Examples

1.3-6 Real line R. The real line R is separable.

Proof. The set Q of all rational numbers is countable and IS
dense in R.

1.3-7 Complex plane C. The complex plane C is separable.

Proof. A countable dense subset of C is the set of all complex

numbers whose real and imaginary parts are both rational._

1.3-8 Discrete metric space. A discrete metric space X is separable if

and only if X is countable. (Cf. 1.1-8.)

Proof. The kind of metric implies that no proper subset of X can

be dense in X. Hence the only dense set in X is X itself, and the
statement follows.

1.3-9 Space l"". The space I"" is not separable. (Cf. 1.1-6.)

Proof. Let y = (TJl. TJz, TJ3, ••• ) be a sequence of zeros and ones.
Then y E I"". With Y we associate the real number y whose binary
representation is

We now use the facts that the set of points in the interval [0,1] is
uncountable, each yE [0, 1] has a binary representation, and different
fs have different binary representations. Hence there are uncountably
many sequences of zeros and ones. The metric on I"" shows that any
two of them which are not equal must be of distance 1 apart. If we let
1.3 Open Set, Closed Set, Neighborhood 23

each of these sequences be the center of a small ball, say, of radius 1/3,
these balls do not intersect and we have uncountably many of them. If
M is any dense set in I"", each of these nonintersecting balls must
contain an element of M. Hence M cannot be countable. Since M was
an arbitrary dense set, this shows that 1 cannot have dense subsets
00

which are countable. Consequently, 1 is not separable.

00

1.3-10 Space IP. The space IP with 1 ~ P < +00 is separable. (Cf.
1.2-3.)

Proof. Let M be the set of all sequences y of the form

y = ('1/10 '1/2, ... , '1/m 0, 0, ... )

where n is any positive integer and the '1//s are rational. M is

countable. We show that M is dense in IP. Let x = (g) E lP be arbitrary.
°
Then for every 8> there is an n (depending on 8) such that

because on the left we have the remainder of a converging series. Since

the rationals are dense in R, for each ~j there is a rational '1/j close to it.
Hence we can find ayE M satisfying

It follows that

[d(x,y)]p=tl~j-'1/jIP+
j=l j=n+l
f l~jIP<8P.

We thus have d(x, y)<8 and see that M is dense in IP.

Problems
I. Justify the terms "open ball" and "closed ball" by proving that (a) any
open ball is an open set, (b) any closed ball is a closed set.

2. What is an open ball B(xo; 1) on R? In C? (a. 1.1-5.) In era, b]? (a.

I. 1-7.) Explain Fig. 8.
24 Metric Spaces

Fig. 8. Region containing the graphs of all x E C[ -1, 1] which constitute the 6-
neighborhood, with 6 ~ 1/2, of XoE C[ -1,1] given by xo{t) = t2

3. Consider C[O, 2'lT] and determine the smallest r such that y E R(x; r),
where x(t) = sin t and y(t) = cos t.

4. Show that any nonempty set A c (X, d) is open if and only if it is a

union of open balls.

5. It is important to realize that certain sets may be open and closed at

the same time. (a) Show that this is always the case for X and 0.
(b) Show that in a discrete metric space X (cf. 1.1-8), every subset is
open and closed.

6. If Xo is an accumulation point of a set A c (X, d), show that any

neighborhood of Xo contains infinitely many points of A.

7. Describe the closure of each of the following subsets. (a) The integers
on R, (b) the rational numbers on R, (c) the complex numbers with
rational real and imagin~ parts in C, (d) the disk {z Ilzl<l}cC.

8. Show that the closure B(xo; r) of an open ball B(xo; r) in a metric

space can differ from the closed ball R(xo; r).

9. Show that A c A, A = A, A U B = A U 13, A nBc A n B.

10. A point x not belonging to a closed set Me (X, d) always has a
nonzero distance from M. To prove this, show that x E A if and only if
°
vex, A) = (cf. Prob. 10, Sec. 1.2); here A is any nonempty subset of
X.

11. (Boundary) A boundary point x of a set A c (X, d) is a point of X

(which mayor may not belong to A) such that every neighborhood of x
contains points of A as well as points not belonging to A; and the
boundary (or frontier) of A is the set of all boundary points of A.
Describe the boundary of (a) the intervals (-1,1), [--1,1), [-1,1] on
1.4 Convergence, Cauchy Sequence, Completeness 2S

R; (b) the set of all rational numbers onR; (c) the disks {z Ilzl< l}cC
and {z Ilzl~ l}cC.

12. (Space B[a, b]) Show that B[a, b], a < b, is not separable. (Cf.
1.2-2.)

13. Show that a metric space X is separable if and only if X has a

countable subset Y with the following property. For every E > 0 and
every x E X there is ayE Y such that d(x, y) < E.

14. (Continuous mapping) Show that a mapping T: X ---- Y is continu-

ous if and only if the inverse image of any closed set Me Y is a closed
set in X.

15. Show that the image of an open set under a continuous mapping need
not be open.

1.4 Convergence, Cauchy Sequence, Completeness

We know that sequences of real numbers play an important role in

calculus, and it is the metric on R which enables us to define the basic
concept of convergence of such a sequence. The same holds for
sequences of complex numbers; in this case we have to use the metric
on the complex plane. In an arbitrary metric space X = (X, d) the
situation is quite similar, that is, we may consider a sequence (x,.) of
elements Xl, X2, ••• of X and use the metric d to define convergence in
a fashion analogous to that in calculus:

1.4-1 Definition (Convergence of a sequence, limit). A sequence (x,.)

in a metric space X = (X, d) is said to converge or to be convergent if
there is an X E X such that

lim d(x,., x) = o.
n~=

x is called the limit of (xn) and we write

limx,.=x
n-->=

or, simply,