METRIC SPACES

JAMES A. OGUNTUASE

F
Department of Mathematics
Federal University of Agriculture
Abeokuta, Ogun State, Nigeria

Being the lectures delivered at the Mathematics Graduate Improvement Pro-

gramme held at the National Mathematical Centre, Abuja, Nigeria, 11 - 22 Febru-
ary, 2013

This School is sponsored

by
National Mathematical Centre, Abuja, Nigeria

F denote author’s permanent address.

1
2 JAM ES A. OGUNTUASE

Metric Spaces
Course Outline
Metric Spaces: De…nition and examples
Open sets, closed sets, Cantor set. Neighbourhoods, Limit points,
Closure, Dense subset
Interior, exterior, frontier,
Convergence in Metric spaces
Continuity, compactness and connectedness.
Heine-Borel theorem, Bolzano–Wierstrass theorem.

1. Metric Spaces: Definition and Examples

Analysis is primarily concerned with limit processes and continuity. These con-
cepts are given by the following de…nitions.

1
De…nition 1.1. The sequence fxn gn=1 of real numbers is said to be convergent if
there exists a real number x (called the limit of the sequence) such that, given > 0;
a positive integer n0 can be found with the property that
n n0 =) jxn xj < :
That is, xn ! x as n ! 1 or limn!1 xn = x:
De…nition 1.2. A real-valued function f de…ned on a non-empty subset X of the
real line is said to be continuous at x0 in X if for each > 0 there exist a > 0
such that
x in X and jx x0 j < =) jf (x) f (x0 )j < ;
and f is said to be continuous if it is continuous at each point of X:
The above de…nitions are dependent for their meaning in the concept of the ab-
solute value of the di¤erence between two real numbers. We observe that the de…n-
itions also hold for complex sequences and complex-valued functions with complex
domain. It is therefore necessary to de…ne some notion of distance which will be
applicable to the elements of arbitrary sets.
Thus, a metric space is an initial point to start. It is a non-empty set equipped
with a concept of distance which is suitable for the treatment of convergent se-
quences in the set and continuous functions de…ned on the set. The basic facts
about metric spaces and the motivation they provide for development of topologi-
cal spaces will be well examined. First, we make the following de…nition:

De…nition 1.3. Given a non-empty set X; a distance function d on X; called a

metric for X; is a function which assigns to each pair of points a real number, (or
formally, d : X X ! R);satsifying the following properties:
For all x; y; z 2 X
M1: d(x; y) 0 (positive property)
M2: d(x; y) = 0 if and only if x = y
M3: d(x; y) = d(y; x) (symmetric property)
M4: d(x; y) d(x; z) + d(z; y); (triangle inequality).
 M ETRIC SPACES 3

A non empty set X endowed with a metric d is denoted by (X; d) and is called a
metric space. Usually, we simply say that X is a metric space, if we need to specify
the metric, we say that (X; d) is a metric space.
Remark 1.1. (i) Di¤ erent metrics could be de…ned on the same set giving rise to
di¤ erent metric spaces.
(ii) Note that if distances are generally greater going via an additional point,
then they are greater going via any number of additional points z1 ; z2 ; :::; zn : for,
by repeated use of M 4:
d(x; y) d(x; z1 ) + d(z1 ; y)
d(x; z1 ) + d(z1 ; z2 ) + d(z2 ; y)
d(x; z1 ) + d(z1 ; z2 ) + d(z2 ; z3 ) + d(z3 ; y)
:::
d(x; z1 ) + d(z1 ; z2 ) + d(z2 ; z3 ) + ::: + d(zn ; y):
Theorem 1.1. (Rearrangement of the Triangle Inequality) Suppose X is a metric
space and x; y; z 2 X: Then jd(x; y) d(y; z)j d(x:z):
Proof. The triangle inequality for d yields …rst d(x; y) d(x; z)+d(z; y) and second
d(z; y) d(z; x) + d(x; y): Using the symmetry, rearrangement of the …rst of these
inequalities gives d(x; y) d(y; z) d(x; z) and rearrangement of the second gives
d(y; z) d(x; y) d(x; z): These two together prove the theorem.

1.1. Examples of Metric Spaces. Given below are exam-

ples of metric spaces. The veri…cations that the functions de…ned are metric are,
in some cases, left as exercises, but in some cases, generous hints are also usually
given.
Example 1.1. For a non-empty set X the discrete (or trivial) metric d is de…ned
on X X by
0 if x = y
d(x; y) =
1 if x 6= y:
This is the roughest of metrics; given any x 2 X; it is simply a measure of
coincidence with x: This example shows that every non-empty set can be provided
with a metric. The veri…cation that d is a metric on X is trivial and so left as an
exercise .
Example 1.2. Let X = < - the set of real numbers: Consider the properties of
absolute value of x 2 X: That is,
(i): jxj 0;
(ii): jxj = 0 () x = 0;
(ii): j xj = jxj ;
(iii): jx + yj jxj + jyj :
Now de…ne a metric d : R R ! R by
Example 1.3.
d(x; y) = jx yj for all x; y 2 R:
Then d is a metric R often called the usual metric on R or the Euclidean metric
on R: Thus, the real line R as a metric space is always understood to have this as
4 JAM ES A. OGUNTUASE

its metric. The fact that this is a metric space follows directly from the properties
of absolute values above. To see this, we have
For M1 We have by (i) that for any x; y 2 R;
d(x; y) = jx yj 0
thus verifying M1
For M2: We have by (ii) that for any x; y 2 R;
d(x; y) = jx yj = 0 () x y = 0 () x = y:
For M2 We have by (iii),
d(x; y) = jx yj = j (y x)j = jy xj = d(y; x):
For M4, we have by (iv),
d(x; y) = jx yj
= j(x z) + (z y)j
jx zj + jz yj = d(x; z) + d(z; y):
Example 1.4. Let X = C the set of complex numbers. Note that z 2 C is of the
form z = a + ib; where a; b 2 R and i2 = 1: De…ne d : C C ! R by
d(z; w) = jz wj ; z; w 2 C:
Then (C; d) is a metric space.
Most of the metric spaces we will consider are also linear spaces. In particular,
most of these cases, the metric is generated by a simpler function called a norm
which assigns a length to each vector in the linear space.
De…nition 1.4. Given a liner space X over R (or C); a norm k:k for X is a
function on X which assigns to each element a real number, (or formally, k:k :
X ! R); satisfying the following properties:
(i): kxk 0;
(ii): kxk = 0 () x = 0;
(iii): k xk = j j kxk for any scalar
: and for all x; y 2 X;
(iv): kx + yk kxk + kyk (the triangle inequality)
A linear space X with a norm k:k is denoted by (X; k:k) is called a normed linear
space. Observe that di¤erent norms can be de…ned on the same linear space giving
rise to di¤erent normed linear spaces.
Remark 1.2. Given a normed linear space (X; k:k); it is clear that the function
d : X X ! R de…ned by
d(x; y) = kx yk
is a metric for X; and we call this the metric generated by the k:k : So, every
normed linear space is a metric space under the metric generated by its norm.
Example 1.5. (R; j:j) and (C; j:j):The set ofreal numbers R (the set of complex
numbers C) is a normed linear space with norm given by the modulus, that is,
kxk = jxj :
We call this the usual norm for R (or C) and it generates the usual metric
d(x; y) = jx yj :
 M ETRIC SPACES 5

These are the spaces we are familiar with in real and complex analysis respectively.
Example 1.6. The n dimensional Eucleadean space Rn whose points are the
n tuples x = (x1 ::::; xn ) of real numbers and
h i1
2 2 2
d(x; y) = (x1 y1 ) + ::: + (x ; yn ) :
Then ( Rn ; d) is a metric space
Example 1.7. The set of all n tuples of real numbers R with metric given by
(x; y) = jx1 y1 j + ::: + jxn yn j :
Then (R; ) is a metric space.
Example 1.8. If (X; d) and (Y; ) are metric spaces, then thier Cartesian product
X Y = f(x; y) : x 2 X; y 2 Y g whose metric is given by
h i1
2 2 2
((x1 ; y1 ) ; (x2 ; y2 )) = d (x1 ; x2 ) + (y1 ; y2 ) :
Then (X Y; ) is a metric space.
Example 1.9. Let B[0; 1] = ff : [0; 1] !! R jf is bounded g : Note that f : [0; 1] !
R is bounded if 9k > 0 such that jf (x)j k 8x 2 [0; 1]: Now de…ne f g by
(f g) (x) = f (x) g(x) whenever f; g 2 B[0; 1] and x 2 [0; 1]:
Also let Z 1
kf k = jf (x)j dx
0
be the norm of f; then k:k induces a metric d on B[0; 1] where
Z 1
d(f; g) = jf (x) g(x)j dx:
0
Then (B[0; 1]; d) is a metric space.
Example 1.10. Let C[0; 1] be the set of all bounded continuous real functions in
[0; 1]. De…ne
kf k = sup fjf (x)j : x 2 [0; 1] g ;
where f 2 C[0; 1]: Then the k:kinduces a metric d on C[0; 1] de…ned by
d(f; g) = sup fjf (x) g(x)j : f; g 2 C[0; 1] and x 2 [0; 1] g :
Then (C[0; 1]; d) is a metric space.
The veri…cation that the above are metric spaces are easy and left as exercises
to the readers.
Remark 1.3. If we relax the condition that d(x; y) = 0 if and only if x = y with
the condition d(x; y) = 0 for some x 6= y; we call d a pseudometric. Thus the Lp
norms are pseudometrics on the spaces of measurable functions whose pth power
are integrable.

Exercises
(1) Given a metric space (X; d); prove that for all x; y; z 2 X; jd(x; z) d(y; z)j
d(x; y):
(2) Given a non-empty set X and a function d : X X ! R with properties
6 JAM ES A. OGUNTUASE

(i): d(x; y) = 0 ; x = y;
(ii): d(x; y) d(x; z) + d(y; z):
Prove that d is a metric for X:
(3) Use the fact that d(x; x) d(x; y) + d(y; x) for any x; y in a metric space
(X; d) to deduce that d(x; y) 0:
(4) Let X = R2 and for x; y 2 X de…ne d(x; y) by
jx1 y1 j if x2 = y2
d(x; y) = d((x1 ; x2 ); (y1 ; y2 )) =
jx1 j + jx2 y2 j + jy1 j if x2 =
6 y2 :
Then, prove that (X; d) is a metric space.

2. Open sets, Closed sets and Neighbourhoods

2.1. Open Sets.
De…nition 2.1. Let (X; d) be a metric space and let x0 be a point of X and r > 0
is a real number. We de…ne the open sphere (ball) Sr (x0 ) with centre x0 and radius
r as the subset of X given by
Sr (x0 ) = fx : d(x; x0 ) < rg :
Remark 2.1. (i) Sr (x0 ) 6= ; since x0 2 Sr (x0 ):
(ii) For the metric space (X; d) where
0 if x = y
d(x; y) =
1 if x 6= y;
we have S1 (x0 ) = fx0 g :
(iii) For Sr (x0 ) R; Sr (x0 ) = (x0 r; x0 + r) which is a bounded open interval
in R: Hence, any bounded open interval on the real line is an open sphere.
Example 2.1. Let X = R with the usual metric d de…ned by
0 if x = y
d(x; y) =
1 if x 6= y;
for arbitrary points x; y 2 R: Describe the open sphers (balls)
(i) S 21 (1) (ii) S1 (1) (iii) S2 (1) (iv) S 12 (2)

(v) S1 (5) (vi) S 32 (4):

Solution 1. (i) By de…nition
1
S 21 (1) = y 2 R : d(y; 1) < :
2
We know fron the de…nition of d that for all x; y 2 R; d(x; y) has two values
namely 0 or 1 Thus, d(y; 1) < 1 implies that it cannot be equal to 1: Hence, d(y; 1)
must be equal to 0: Therefore, it follows that
S 12 (1) = fy 2 R : d(y; 1) = 0g :
Since d is a metric, it thus follows that d(x; y) = 0 () x = y: In particular,
d(y; 1) = 0 implies that y = 1: Hence,
S 21 (1) = fy 2 R : d(y; 1) = 0g = fy 2 R : y =g = f1g :
Hence, the open ball S 21 (1) = f1g is a singleton set f1g :
 M ETRIC SPACES 7

(ii)
S1 (1) = fy 2 R : d(y; 1) < 1g = fy 2 R : d(y; 1) = 0g
Again, since d is a metric, we have that for each pair x:y 2 R; d(x; y) = 0 ()
x = y: In particular, d(y; 1) = 0 =) y = 1: Hence
S1 (1) = fy 2 R : d(y; 1) = 0g = fy 2 R : y = 1g = f1g :
(vi)
3
S 23 (4) = :
y 2 R : d(y; 4) <
2
3
But d(x; y) = 0 or 1 for all x; y 2 R: In either case, d(x; y) is always less than 2
for all x; y 2 R: Hence,
3
S 23 (4) = y 2 R : d(y; 4) < = fy 2 Rg = R:
2
(iii) (v) are left as exercises.
De…nition 2.2. A subset G of a metric space (X; d) is open if given any point
x 2 G; 9 r > 0 such that Sr (x) G:
Remark 2.2. On R = ( 1; 1) :
(i) fxg is not open since for r > 0; Sr (x) * fx0 g
(ii) [0; 1] is not open since 0 2 [0; 1] and r > 0; Sr (0) * [0; 1]: Also for r >
0; Sr (1) * [0; 1]:
(iii) (0; 1) is open.
(iv) any open interval in R; bounded or not is an open set in R: Furthermore,
the open intervals are the only intervals of R which are open sets.
Theorem 2.1. In any metric space (X; d); ; and X are open.
Proof. Since ; has no point, each point in ; is the centre of an open sphere which
is contained in ;: This implies that ; is open. Clearly X is open since every open
sphere centred at each points of X is contained in X:
Remark 2.3. A set is open or not only with respect to a speci…c metric space
containing it and not on its own. For example, with respect to the metric space R;
[0; 1] is not open. However, with respect to the metric space ([0; 1]; d) it is open.
Theorem 2.2. In any metric space (X; d), each open sphere is an open set.
Proof. Let Sr (x0 ) be an open sphere in X; and let x 2 Sr (x0 ): Then we shall
produce an open sphere centred at x and contained in Sr (x0 ): Since d(x; x0 ) < r;
set r1 = r d(x; x0 ) > 0: Then, we shall show that Sr1 (x) Sr (x0 ): Now, let
y 2 Sr1 (x); then d(y; x) < r1: Then by triangle inequality we obtain that

d(y; x0 ) d(y; x) + d(x; x0 )

< r1 + d(x; x0 )
= [r d(x; x0 )] + d(x; x0 )
= r:
This implies that y 2 Sr (x0 ): We have shown that
Sr1 (x) Sr (x0 ): This implies that Sr (x0 ) is open.
8 JAM ES A. OGUNTUASE

Theorem 2.3. Let (X; d) be a metric space. A subset G of X is open i¤ G is a

union of open spheres.
Proof. First, suppose G is open, we shall show that G is a union of open spheres.
If G = ;; then G is a union of empty class of open spheres and so the conclusion is
trivial. Now, suppose G 6= ;; then each of its points is the centre of an open sphere
which is contained in G: Hence, G is the union of all the open sphere contained in
it.
Assume that G is the union of a class S of open spheres. We shall show that G
is open. If S = ;; then G is also empty, and by Theorem 2.1 , G is open. Suppose
S= 6 ;; then it follows that G 6= ;: Let x 2 G; since G is the union of open spheres
in S, x belongs to an open sphere Sr (x0 ) in S. By Theoren 2.2, x is the centre of
an open sphere Sr1 (x) v Sr (x0 ): Since Sr (x0 ) v G; Sr1 (x) v G and we have an
open sphere centred on x and contained in G: G is therefore open.

Theorem 2.4. Let X be a metric space. Then

(i) any union of open sets is open;
(ii) any …nite intersection of open sets is open.
Proof. (i) Suppose fGi g is an arbitrary class of open sets. We shall show that
G = [i Gi is open. If fGi g = ;; then G = [i Gi = ; and so G is open. Next,
suppose fGi g is not empty. Then Gi being open is the union of open sphere. Hence
G is the union of all such open spheres and so G is open.
(ii) Let fGi g be a …nite class of open sets. we must show that G = \i Gi is open.
Suppose fGi g = ;: Then G = \i Gci = [i Gci = X . Hence G is open for X is open.
Now assume that fGi g is not empty. Let fGi g = fG1 ; G2 ; :::; Gn g ;where n 2 N:
If G = \ni=1 Gi = ;; then G is open. Suppose G 6= ;; then 9x 2 G and so x 2 Gi
8i = 1; 2; :::; n:
Since each Gi is open, 9ri ; i = 1; :::; n such that Sri (x) Gi ; where ri > 0:
Let r = min fr1 ; :::; rn g : Then Sr (x) Sri (x) Gi ; i = 1; :::; n: This implies that
Sr (x) \ni=1 Gi = G: Hence G is open and the proof is complete.

Remark 2.4. In any metric space, the class of open sets is closed under the
formation of arbitrary unions and …nite intersections However, if we take arbi-
trary intersection of open sets, the set obtained is not necessarily open. To see
1 1
this, consider the class n ; n : n = 1; 2; ::: of open intervals in R: Observe that
1 1
\1n=1 n ; n = f0g ; is closed since a singleton set is closed.

2.2. Closed Sets.

De…nition 2.3. Let (X; d) be a metric space and let A X: A point x 2 X is
called a limit point of A if each open sphere centred at x contains at least one point
of A di¤ erent from x:
Example 2.2. Let X = R:
(1) If A = 1; 12 ; 13 ; ::: : Then, the point 0 2 R is a limit point of A:
(2) If A = [0; 1]: Then 0 and 1 are limit points of A: Indeed, every element x;
such that 0 x 1 are limit points.
(3) If A = (0; 1): Then 0 and 1 are limit points.
(4) If A = f:::; 2; 1; 0; 1; 2; 3; :::g :Then A has no limit point. To see this, take
x = 14 2 R; then S 16 1
1 ( ) does not contain any element of A:
4
 M ETRIC SPACES 9

De…nition 2.4. A set F in a metric space (X; d) is said to be closed if F contains

each of its limit points.
Example 2.3. In the example above, (2) is closed while (1) and (3) are not closed.
Theorem 2.5. Let (X; d) be a metric space, then ; and X are closed sets.
Proof. The empty set ; has no limit points, so it contains them all and is therefore
closed.
Since X contains all points, it automatically contains its own limit points and
thus closed.
Theorem 2.6. Let (X; d) be a metric space. A subset F of X is closed if and only
if its complement X F is open.
Proof. Suppose X F is open. We shall show that F is closed. We know that F
fails to be closed if it does not contain all its limit points. That is, if F has a limit
point in X F: Since X F is open, each of its points is the centre of an open
sphere which is completely contained in X F: Hence, no point of X F is a
limit point of F: This implies that F contains all its limit points and so F is closed.
Conversely, suppose F is closed. We shall show that X F is open. If X F
= ;; then we know that it is open since ; is open. So, assume X F 6= ;: Let
x 2 X F: Since F is closed, x 2 = F and so x is not a limit point of F: Hence, 9 an
open sphere Sr (x) such that Sr (x) \ F = ;: This says that Sr (x) X F: This
implies that X F is open.
De…nition 2.5. Let x0 be a point in a metric space (X; d) and r is a nonnegative
real number, the closed sphere Sr [x0 ] with centre x0 and radius r is the subset of
X de…ned by

Sr [x0 ] = fx : d(x; x0 ) rg :
Remark 2.5. (i) Sr [x0 ] is not empty because it contains only its centre when r = 0:
(ii) Closed spheres on R are precisely the closed intervals.
(iii) Observed that though open spheres on R are open intervals, there are open
intervals which are not open spheres, e.g., ( 1; +1):
Theorem 2.7. In any metric space (X; d); each closed sphere is a closed set.
Proof. Let Sr [x0 ] be a closed sphere in X: By Theorem 2.6, it su¢ ces to show that
its complement X Sr [x0 ] is open. If X Sr [x0 ] = ;; it is clearly open since ;
is open. Next, assume X Sr [x0 ] 6= ;: Let x 2 X Sr [x0 ]: Since d(x; x0 ) > r;
set r1 = d(x; x0 ) r > 0: Take r1 as the radius of an open sphere Sr1 [x] centred
on x; and we show that X Sr [x0 ] is open by showing that Sr1 [x] X Sr [x0 ]:
Let y 2 Sr1 [x] so that d(y; x) < r1 : Because of this and the fact that d(x0 ; x)
d(x0 ; y) + d(y; x), we see that
d(y; x0) d(x; x0 ) d(y; x)
> d(x; x0 ) r1
= d(; x0 ) [d(x; x0 ) r]
= r;
so that y 2 X Sr [x0 ]:
10 JAM ES A. OGUNTUASE

Theorem 2.8. Let (X; d) be a metric space. Then

(i) any intersection of closed sets in X is closed; and
(ii) any …nite union of closed sets in X is closed.
1
Proof. (i) Let fFi gi=1 be a collection of closed sets in X: Consider the \1
i=1 Fi : It
will be closed if and only if X \1 i=1 F i is open. But
X \1 1
i=1 Fi = [i=1 (X Fi ) (by De Morgans law).
Hence X \1
i=1 Fi
is the union of a collection of open sets since each X Fi ;
i = 1; 2; ::: is open. By Theorem 2.4 (i) [1 i=1 (X Fi ) is open and so \1i=1 Fi is
closed.
k
(ii) Let fFi gi=1 be any …nite member of closed sets in X: Consider the [ki=1 Fi :
It will be closed if and only if X [ki=1 Fi is open. But
X [ki=1 Fi = \ki=1 (X Fi ) (by De Morgans law).
Hence X [ki=1 Fi is the intersection of a …nite member of open sets, since each
X Fi ; i = 1; 2; :::; k is open. Then, by Theorem 2.4 (ii) \ki=1 (X Fi ) is open
and so [ki=1 Fi is closed.

Remark 2.6. Observe that in Theorem 2.8 (ii) the …niteness condition is important
because the union of an in…nite family of closed sets need not necessarily be closed:
In R with the usual metric, consider the family of closed intervals
1 k k 1 1 k k 1
; :k2N : Now [1
k=1 ; = ( 1; 1)
k k k k
an open interval which is not a closed set.
Example 2.4. Closed intervals in R are closed sets and closed spheres (balls) in
Rn are closed sets in Rn : To prove the latter assertion, consider the closed sphere
Sr [x0 ]: Let x be an arbitrary point in Rn Sr [x0 ]: We shall construct an open sphere
with x as centre which is entirely contained in Rn Sr [x0 ]: Now d(x0 ; x) > r and
hence r0 = d(x; x0 ) r is a positive real number. We construct the open sphere with
x as the centre and r0 as radius. We shall show that Sr0 (x) is contained entirely in
Rn Sr [x0 ]: In fact, let y 2 Sr0 (x): Then we have
d(x; y) < d(x0 ; x) r
and hence
d(x0 ; y) = d(x0 ; y) + d(y; x) d(x; y)
> d(x0 ; x) d(x; y)
> r:
Thus y 2 Rn Sr [x0 ] and hence Sr [x0 ] is closed.
2.3. The Cantor Set. One important example of a closed set on R is the
famous set …rst studied by George Cantor (1845-1918)
Example 2.5. We denote by F0 the closed interval [0; 1]: From F0 ; we remove the
open middle interval ( 13 ; 23 ); that is the open middle thierd of F0 : Let the remaining
closed set be denoted by F1 ; thus F1 = [0; 31 ] [ [ 23 ; 1]: From each of the two disjoint
 M ETRIC SPACES 11

closed sets [0; 13 ]; [ 23 ; 1]; we next remove the open middle third i.e. ( 19 ; 29 ) and ( 97 ; 89 )
and denote the remaining closed set by F2 ; i.e.
1 2 1 2 7 8
F2 = [0; ] [ [ ; ] [ [ ; ] [ [ ; 1]:
9 9 3 3 9 9
We continue in this way in…nitum, at each stage, removing the open middle
third of each of the closed intervals remainig from the previous stage. In this way
we obtain an in…nite sequence of closed sets fFn g such that Fn Fn+1 for all
\1
n = 0; 1:2; :::: Write F = Fn : Then F is known as Cantor’s set, Since each Fn
n=1
is closed, it follows from Theorem 2.8 that F is closed.
Note that F is the set that remains when we have removed the open intervals
1 2 1 2 7 8
( ; ); ( ; ); ( ; ); ::::
3 3 9 9 9 9
To see this, denote by Mn the union of the open middle thirds removed at the
nth stage; thus
1 2 1 2 7 8
M1 = ( ; ); M2 = ( ; ); ( ; ); etc.
3 3 9 9 9 9
and
F1 = F0 M1 ; F2 = F0 M2 ; :::; Fn = F0 Mn ; ::::
Then
1
\ [1 1
[
F = Fn = (F0 Mn ) = F0 Mn :
n=1 n=1 n=1

2.4. Neighbourhoods.
De…nition 2.6. Let X be a metric space with metric d and let p 2 X: A subset N
of X is called a neighbourhood of p if N contains an open set G which contains p:
That is, 9 an open set G such that p 2 G N:
Example 2.6. Let (X; d) be any metric space and let p 2 X: Then X is a neigh-
bourhood of p:
Example 2.7. Let X = R and for x; y 2 X; de…ne d(x; y) = jx yj : Write
I = [ 1; 1] and let 0 2 R; then I is a neighbourhood of 0:
Theorem 2.9. A subset A of a metric space (X; d) is open () A contains a
neighbourhood of each of its points.
Proof. First, let us assume that A contains a neighbourhood of each of its points.
For each p 2 A; 9 Np such that p 2 Np A: Hence 9 open set Gp such that
p 2 Gp Np A:
Write [
G= fGp : p 2 Ag :
We shall show that G = A:
To see this, …rst let x 2 G; then x 2 Gp for some p: This implies that x 2 Gp
Np A: Hence, x 2 A and so G A: i.e. 9 open set Gx 3 x 2 Gx Nx A:
Therefore, x 2 G and so G = A:
Conversely, let A be open. We shall show that A contains a neighbourhood
of each of its points. But since A is open and contains itself, we take A as a
12 JAM ES A. OGUNTUASE

neighbourhood of its points. Hence, A contains a neighbourhood of each of its

points.
2.5. Accumulation (or Limit Points).
De…nition 2.7. Let (X; d) be a metric space and let A X: A point p 2 X is
called a point of accumulation (or limit point) for A if every neighbourhood of p
intersects A nontrivially. In other words, p is a point of accumulation for A if every
neighbourhood of p contains at least one point of A distinct from p:
Example 2.8. (1) Let X = R and let A = 1; 12 ; 13 ; ::: : Then 0 is a limit point
for A:
(2) Let A = [0; 1): Then 0 and 1 are limit points of A: Indeed, every element
0 x 1 is a limit point.
(3) Let A = (0; 1): Then 0 and 1 are limit points.
(4) Let A = f:::; 2; 1; 0; 1; 2; :::g : Then A has no limit point. To see this, let
1
x = 41 2 R: Then S 12 ( 16 ) does not contain an element of A: Also, if x 2 A; say
1
x = 1; then S 1 ( 2 ) does not contain an element of A:
Let us denote by A0 the set of all points of accumulation for A: Then we have
the following results:
Theorem 2.10. Let A be a subset of a metric space (X; d); then the set A is
closed () A0 A: That is, A is closed () A contains all its points of
accumulation.
Proof. First, let A be closed. We must show that A contains all its points of
accumulation. That is„ we must show that A0 A: Let p2 A0 ; and suppose
p 2 A , then p 2 X A: But X A is open since A is closed. Since X A is
open, it contains a neighbourhood of each of its points, in particular, it contains
a neighbourhood of p: This neighbourhood is lying entirely in X A which is
impossible since p is a point of accumulation for A: Hence, p 2 A:
Conversely, let A0 A: We must show that A is closed. That is, X A is open.
To see, we shall show that X A contains a neighbourhood of each of its points.
Let p 2 X A; then p 2 = A and therefore p 2 = A0 : That is, p is not a point of
accumulation for A: Hence, there exists at least one neighbourhood of p which does
not intersect A: Thus p 2 N 2 X A: Hence, we have shown that X A contains
a neighbourhood of each of its points and so X A is open. i.e. A is closed.
Theorem 2.11. Let A be any subset of a metric space (X; d) and denote by A0 the
collection of all accumulation points of A: Then A [ A0 is closed.
Proof. We shall show that X (A [ A0 ) is open. To do this, we shall show that
0
X (A [ A ) contains a neighbourhood of each of its points. Let p be any point
in X (A [ A0 ); we shall show that X (A [ A0 ) contains a neighbourhood of p:
Now,
X (A [ A0 ) = (X A) \ (X A0 ) by De Morgans.
0
So p 2 X A and p 2 X A : Hence p 2 = A and p 2
= X A0 : i.e. p is not a
point of accumulation for A: This implies that 9 at least one neighbourhood U of p
disjoint from A: i.e. U \ A = ; . Also U \ A0 = ; . To see this, let x 2 U \ A0 ; then
x will be a point of accumulation for A: Every neighbourhood of x will intersects
A: In particular, U intersects A; a contradiction, hence U \ A0 = ;: Thus U \ A = ;
 M ETRIC SPACES 13

and U \ A0 = ;: Hence U does not intersect A [ A0 : Thus U is a neighbourhood of

p and is contained entirely in X (A [ A0 ):
De…nition 2.8. Let (X; d) be a metric space and A a subset of X: By closure of
A we mean the smallest closed set containing A:
Denote the closure of A by A:
Theorem 2.12. Let A be a subset of a metric space (X; d): Then, A = A [ A0 :
Proof. Let K = f\E : E is a closed set and that E Ag : Then K is closed
since it is the intersection of a collection of closed sets. Also K A since A E
for all 2 : Let x 2 A; this implies that x 2 E for all 2 : Since A 2
for all =) x 2 \E = K: Now if F is a closed set such that F A; then we
shall show that F K: Clearly, F is one of the E 0 s ; 2 : Since F is bigger than
their intersection, then F K:
De…nition 2.9. Let (X; d) be a metric space. Then A is said to be everywhere
dense in X if A = X: More generally, a set A is dense in a set B whereA; B are
subsets of X if A = B:
Example 2.9. (1) (0; 1) is dense in [0; 1]:
(2) The set Q of rational numbers is everywhere dense in R:
2.6. Interior of a set.
De…nition 2.10. Let (X; d) be a metric space and let A X: A point x 2 A is
called an interior point of A if x is contained in an open sphere (ball) contained in
A:
Thus x is an interior point of A if A is a neighbourhood of x: We denote the set of
all interior point of A by Int(A) Thus Int(A) = fx 2 A : Sr (x) A; for some r > 0g :
Example 2.10. (1) If A = (0; 1) R; then every point of A is an interior point
of A but of course if A = [0; 1]; then either 0 nor 1 is an interior point;
(2) If A = 1; 21 ; 13 ; 14 ; 15 : Then A has no interior point
(3) If A is any open set in (X; d), then every point of A is an interior point.
The basic properties of interiors are the following:
(i) Int(A) is an open subset of A which contains every open subset of A (this is
often expressed by saying that the interior of A is the largest open subset of A);
(ii) A is open () A = Int(A);
(iii) Int(A) equals the union of all open subsets of A:
The proof of the above facts are easy and so left as exercise.
Exercise 2.1. (1) Let A and B be subsets of a metric space (X; d): Show that
(i) Int(A) [ Int(B) Int(A [ B);
(ii) Int(A) \ Int(B) = Int(A \ B):
(2) Give an example of two sets A and B of the real line such that Int(A) [
Int(B) 6= Int(A [ B):
2.7. Boundary point.
De…nition 2.11. Let (X; d) be a metric space and A X: We say that a point
x 2 X is a frontier point (i.e. a boundary point) of A if every neighbourhood Nx of
x satis…es Nx \ Int(A) = ;; Nx \ A = ;: That is, x is a frontier point if x A \ Ac :
14 JAM ES A. OGUNTUASE

The set of all frontier points of A X is denoted by F r(A) or @A and is called the
frontier or boundary of A:
The following theoren is left as exercise
Theorem 2.13. Let (X; d) be a metric space and let A; B; C be subsets of X; then
1. @A = A Int(A)
2. B is closed () F r(B) B
3 C is both open and closed () @C = ;
4 @A is a closed set.

3. Convergence in Metric Spaces

One of our aims in studying metric spaces is to study convergent sequences
in a more general setting than that of classical analysis. We have the following
de…nition:
De…nition 3.1. Let (X; d) be a metric space and let fxn g a sequence of points in
X: The sequence fxn g is said to be convergent if 9 a point x in X such that either
(1) for each > 0; 9 a positive integer n0 such that n n0 =) d(xn ; x) < ; or
equivalently,
(2) for each open sphere S (x) centred on x; 9 a positive integer n0 such that xn
is in S (x) for all n n0 :
Remark 3.1. (i) Observe that condition (1) is a direct generalization of conver-
gence of sequences of real numbers while condition (2) is saying that each open
sphere centred on x contains all points of the sequence from some place on.
(ii) the statement that fxn gis convergent can equivalently be expressed as follows:
9 a point x in X such that d(xn ; x) ! 0: This is symbolize by writing xn ! x; and
we express it by saying that xn approaches x or that xn converges to x:
(iii) the point x is called the limit of the sequence fxn g and we sometimes write
lim xn = x or lim xn = x
De…nition 3.2. Let fxn g be a sequence in a metric space (X; d): Then xn is called
a Cauchy sequence if for any > 0; 9 n0 2 N such that d(xm ; xn ) < if m; n n0 :
De…nition 3.3. A metric space (X; d) is said to be complete if every Cauchy se-
quence in X converges to a point in X.
Example 3.1. (1) The real line R is a complete metric space
(2) The complex plane C is a complete metric space. To see this, let fzn g be a
Cauchy sequence of complex numbers, where zn = an + ibn : Then fan g and fbn g
are themselves Cauchy sequences of real numbers, since
jam an j jzm zn j
and
jbm bn j jzm zn j :
Since R is complete, then 9 real numbers a and b such that an ! a and bn ! b:
Now put z = a + ib; the we see that zn ! z by means of
jzn zj =j(an + ibn ) (a + ib)j
=j(an a) + i(bn b)j
jan aj + jbn bj
! 0:
 M ETRIC SPACES 15

Observe that the completeness of C depends directly on the completeness of R:

The following result relates these concepts to one another.

1
Theorem 3.1. A sequence fxn gn=1 in a metric space (X; d) can converge to at
most one point in X:
Proof. Let fxn g converges to x and x0 ; where x; x0 2 X; the we shall show that
x = x0 : Given > 0; 9 integers N; N 0 such that
n N =) d(xn ; x) < 2 : Also n N 0 =) d(xn ; x0 ) < 2 :Now, take
0
n max fN; N g ; we obtain that
d(x; x0 ) d(xn ; x) + d(xn ; x0 )
< + = :
2 2
Since is arbitrary, we conclude that d(x; x0 ) = 0: This implies that x = x0 and
the proof is complete.
1
Theorem 3.2. In a metric space (X; d); the sequence fxn gn=1 converges to x if
1
and only if every subsequence fxnk gk=1 converges to x:
Proof. Since xn ! x as n ! 1 implies that 9N 2 Z+ such that d(xn ; x) <
1
8n N; where > 0 is pre-assigned. Also, since fxnk gk=1 is a subsequence,
9M 2 Z+ such that kn N if n M: Hence, n M =) d(xnk ; x) < . Thus,
this implies that xnk ! x as k ! 1:
1
The converse is trivial since fxn gn=1 itself is a subsequence.
Theorem 3.3. If is a convergent sequence fxn g in a metric space (X; d) has in-
…nitely many distinct points, then its limit is a limit point of the set of points of
the sequence.
Proof. Trivial and therefore left as exercise.
Theorem 3.4. Let (X; d) be a complete metric space, and Y be a subspace of X:
Then Y is complete if and only if it is closed.
Proof. First assume that Y is complete as a subspace of X; then we shall show
that Y is closed. Let y be a limit point of Y: Then, for each positive integer n1 ,
S ni (y) contains a point yn in Y: Clearly fyn g converges to y in X and it is a Cauchy
sequence in Y; and since Y is complete, y 2 Y: Therefore, Y is closed.
Conversely, assume that Y is closed, then we shall show that it is complete. Let
fyn g be a Cauchy sequence in Y: Clearly, it is also a Cauchy sequence in X; and
since X is complete, fyn g converges to a point x in X: We shall show that x is
in Y: If fyn g has only …nitely many distinct points, then x is that point in…nitely
repeated and is thus in Y: Otherwise, if fyn g has in…nitely many distinct points,
then by Theorem 3.3, x is a limit point of the set of points of the sequence; it is
therefore a limit point of Y; and since Y is closed, x 2 Y:

4. Continuity in Metric Spaces

De…nition 4.1. Let (X; d) and (Y; ) be metric spaces, we say that a mapping
f : X ! Y is continuous at x0 2 X if either of the following equivalent conditions
is satis…ed:
(i) for each > 0 9 > 0 such that d(x; x0 ) < =) (f (x); f (x0 )) < ;
16 JAM ES A. OGUNTUASE

(ii) for each open sphere S (f (x0 )) centred on f (x0 ) 9 an open sphere S (x0 )
centred on x0 such that f (S (x0 )) S (f (x0 )):
Remark 4.1. Observe that the condition (i) is a generalization of the de…nition
of convergence ealier given while condition (ii) translates condition (i) into the
language of open spheres.
Example 4.1. Every mapping on a discrete metric space is continuous. To see
this, consider a map f from a discrete metric space (X; d) into a metric space (Y; ):
For x0 2 X; S1 (x0 ) = fx0 g so, given > 0 (f (x); f (x0 ) < when d(x; x0 ) < 1:
Theorem 4.1. Let (X; d) and (Y; ) be metric spaces and f is a mapping of X
into Y: Then f is continuous at x0 if and only if xn ! x0 =) f (xn ) ! f (x0 ):
Proof. First, assume that f is continuous at x0 : We shall show that if fxn g is a
sequence in X such that xn ! x0 ; then f (xn ) ! f (x0 ):
Let S (f (x0 )) be an open sphere centred at f (x0 ): By our assumption, 9 an open
sphere S (x0 ) centred at x0 such that f (S (x0 )) S (f (x0 )): Since xn ! x0 ; all
x0n s from some place on lie in S (x0 ) . But since f (S (x0 )) S (f (x0 )); all f (xn )0 s
from some place on lie in S (f (x0 )): From this it follows that f (xn ) ! f (x0 ):
Conversely, assume that f is not continuous at x0 : We shall show that xn ! x0 ;
f (xn ) ! f (x0 ): By assumption, 9 an open sphere S (f (x0 )) with the property
that the image under f of each open sphere centred on x0 is not contained in it.
Now consider the sequence of open spheres S1 (x0 ); S 21 (x0 ); :::; S n1 (x0 ); :::: Form a
sequence fxn g such that xn 2 S n1 (x0 ) and f (xn ) 2
= S (f (x0 )): Clearly xn ! x0 and
that f (xn ) 9 f (x0 ):
Example 4.2. (i) Consider the real function on (R2 ; kk2 ) de…ned by

2+ 2 ; ( ; ) 6= (0; 0)
f( ; ) =
0 ( ; ) = (0; 0):
Now the sequence n1 ; n1 converges to (0; 0) but f n1 ; n1 = 21 8n 2 N: So f is
not continuous at (0; 0):
(ii) Consider the scalar function p0 on (C[0; 1]; kk1 ) de…ned by p0 (f ) = f (0):
Now the sequence ffn g where
1
1 nt 0 t n
fn (t) = 1
0 n <t 1:
1
satis…es kfn k1 = 2n! 0 as n ! 1; so ffn g converges to the zero function, but
p0 (fn ) = 1 8n 2 N: So p0 is not continuous at the zero function.
De…nition 4.2. A mapping of one metric space into another is said to be contin-
uous if it is continuous at each point in the domain.
Theorem 4.2. Let (X; d) and (Y; ) be metric spaces and f is a mapping of X
into Y: Then f is continuous if and only if xn ! x =) f (xn ) ! f (x):
Proof. This follows from the proof of Theorem 4.1.
Remark 4.2. Observe that the above Theorem shows that convergence is preserved
under continuous mapping.
Our next result characterizes continuous mapping in terms of open spheres.
 M ETRIC SPACES 17

Theorem 4.3. Let (X; d) and (Y; ) be metric spaces and f is a mapping of X
into Y: Then f is continuous if and only if f 1 (G) is open in X whenever G is
open inY:
Proof. First assume that f is continuous. If G is open in Y; we shall show that
f 1 (G) is open in X: If f 1 (G) = ; it is clearly open, so assume f 1 (G) 6= ;: Let
x 2 f 1 (G); then f (x) 2 G:
Since G is open, 9 an open sphere S (f (x)) centred at f (x) and contained in
G: By de…nition of continuity, 9 an open sphere S (f (x)) such that f (S (x))
S (f (x)): Since S (f (x)) G; we also have f (S (x)) G and from this we easily
see that S (x) f 1 (G): S (x) is therefore an open sphere centred on x and
contained in f (G); so f 1 (G) is open.
1

Conversely, assume that f 1 (G) is open whenever G is open. We shall show

that f is continuous. We do this by showing that f is continuous at an arbitrary
point x 2 X: Let S (f (x)) be an open sphere centred on f (x): This open sphere
is an open set, so its inverse image is an open set which contains x: By this, 9 an
open sphere S (x) which is contained in the inverse image. It is clear from this that
S (x) S (f (x)); so f is continuous. Since x was chosen to be arbitrary point in
X; f is continuous.
Remark 4.3. Observe that the above result established that continuous mappings
pull open sets back to open sts.

5. Compactness for Metric spaces

We recall the classical Bolzano-Weierstrass theorem: If X is a closed and bounded
subset of the real line, then every in…nite subset of X has a limit point in X:
In this section, we shall look for an equvalent formulation of the above result for
metric spaces. First, we give the following de…nitions
De…nition 5.1. A metric space is said to have the Bolzano-Weierstrass property
if every in…nite subset has a limit point.
De…nition 5.2. A metric space is said to be sequentially compact if every sequence
in it has a convergent subsequence.
Remark 5.1. We shall show that each of De…nitions 5.1 and 5.2 is equivalent to
compactness for metric spaces.
Theorem 5.1. A metric space is sequentially compact if and only if it has the
Bolzano-Weierstrass property.
Proof. Let X be a metric space and assume that X is sequentially compact. We
shall show that an in…nite subset A of X has a limit point. Now, since A is in…nite,
a sequence fxn g of distinct points can be extracted from A. By the sequential
compactness of X; this sequence has a subsequence which converges to a point x:
By Theorem 3.3, x is a limit point of the set of points of the subsequence, and since
this set is a subset of A; x is also a limit point of A:
Next, assume that every in…nite subset of X has a limit point, and we shall show
from this that X is sequentially compact. Let fxn g be an arbitrary sequence in X:
If fxn g has a point which is in…nitely repeated, then it has a constant subsequence,
and this subsequence is clearly convergent. If no point of fxn g is in…nitely repeated,
then the st of points of this sequence is in…nite. By assumption, the set A has a
18 JAM ES A. OGUNTUASE

limit point x; and we can extract from fxn g a subsequence which converges to
x:
Theorem 5.2. Every compact metric space has the Bolzano-Weierstrass property.
Proof. Let X be a compact metric space and A an in…nite subset of X. Assume
that A has no limit point, and from this we shall arrive at a contradiction. By
assumption, each point of X is not a limit point of A; and so each point of X is the
centre of and open sphere which contains no point of A di¤erent from its centre.
The class of all these spheres is an open cover, and by compactness there exists
a …nite subcover. Since A is contained in the set of all centres of spheres in this
subcover, A is clearly …nite. This contradicts the fact that A is in…nite, and the
proof is complete.
Theorem 5.3. (Lebesgue’s Covering Lemma) In a sequentially compact metric
space, every open cover has a Lebesgue number.
Proof. Omitted.
Theorem 5.4. Every sequentially compact metric space is totally bounded.
Proof. Let X: be a sequentially compact metric space, and let > 0 be given.
Choose a point a1 2 X and form the open sphere S (a1 ): If S (a1 ) contains every
point of X; then the single- element fa1 g is an net. If there are points outside of
S (a1 ); let a2 be such a point and form the two-element set S (a1 ) [ S (a2 ): If this
union contains every point of X, then the two=element set fa1 ; a2 g is an net. If
we continuue in this way, some union of the form
S (a1 ) [ S (a2 ) [ ::: [ S (an )
will necessarily contain every point of X; for if this process could be continued
inde…nitely, then the sequence fa1 ; a2 ; :::; an ; :::g would be a sequence with no con-
vergent subsequence, contrary to the assumed sequential compactness of X: We see
by this that some …nite set of the form fa1 ; a2 ; :::; an g is an net, so X is totally
bounded.
Finally, we shall now show that compactness is implied by sequential compact-
ness.
Theorem 5.5. Every sequentially compact metric space is compact
Proof. .Let X be a sequentially compact metric space, and let fGi g be an open
cover. By Theorem , this open cover has a Lebesgue number a: We put = a3 ; and
use Theorem 5.4 to …n an net
A = fa1 ; a2 ; :::; an g :
For each k = 1; 2; :::; n; we have d(S (ak )) 2 = 2a 3 < a: By the de…nition of
Lebesgue number, for each k we can …nd a Gik such that S (ak ) Gik : Since every
point of X belongs to one of the S (ak )0 s; the class fGi1 ; Gi2 ; :::; Gin g is a …nite
subcover of fGi g : X is therefore compact.
We have so far shown that if X is a metric space, the the following three condi-
tions are all equivalent to one another
(1) X is compact;
(2) X is sequentially compact;
 M ETRIC SPACES 19

(3) X has the Bolzano-Weierstrass property.

Theorem 5.6. Any continuous mapping of a compact metric space into a metric
space is uniformly continuous.
Proof. Let f : be a continuous mapping of a compact metric space (X; d) to a
metric space (Y; ): Let > 0 be given. For each point x 2 X; consider the image
f (x) and the open sphere S 2 (f (x)) centred on this image with radius 2 : Since f
is continuous, the inverse images of these open spheres is an open subset of X;
and the class of all such inverse images is an open cover of X: Since X is compact,
Theorem 5.3 guarantees that this open cover has a Lebesgue number : If x and
x0 are any two points in X for which d(x; x0 ) < ; then the set fx; x0 g is a set
with diametre less than ; both points belong to the inverse image of some of the
above open spheres, both f (x) and f (x0 ) belong to one of these open spheres, and
therefore (f (x); f (x0 )) < ; which shows that f is indeed uniformly continuous.

6. Connectedness
De…nition 6.1. A metric space (X; d) is said to be connected if it can not be
represented as the union of two disjoint non-empty open sets. That is, there do not
exist non=empty open sets A and B such that
X =A[B and A \ B = ;:
Also observe that there do not exist two disjoint non-empty closed sets whose
union is X: If such A and B exist, we say that X is disconnected and that A and
B are said to give a disconnection.
Theorem 6.1. A metric space (X; d) is disconnected i¤ X has a non-empty proper
subset that is both open and closed.
Proof. If A X is both open and closed, then X = A [ (X A) gives a discon-
nection. Conversely, if X = A [ B; A \ B = ; and A; B are open and non-empty,
then B = X A and A = X B: Thus X = A [ (X A): This implies that A is
both open and closed.
De…nition 6.2. A subset S of a metric space (X; d) is disconnected if there exists
two non-empty open sets A and B of X such that A \ S and B \ S are disjoint
non-empty open sets whose union is S. i.e.,
A \ S 6= ;; B \ S 6= ;

(A \ S) \ (B \ S) = ;

(A \ S) [ (B \ S) = S:
If S is disconnected, then the open sets A and B with the above properties are
said to form a disconnection of S:
De…nition 6.3. A subset S of a metric space is said to be connected if it is not
disconnected.
Example 6.1. The set N of natural numbers is disconnected in R: To see this, take
13 13
A= x2R:x< ; .....B = x 2 R : x > :
12 12
20 JAM ES A. OGUNTUASE

Example 6.2. The set Q+ of all positive rational numbers is disconnected. To see
this, take
A = fx 2 R : x < 2g ; .....B = fx 2 R : x > 2g :
We may ask, are there any connected subsets of R or Rn ?: The answer is given
in the following result:
Theorem 6.2. A subspace of the real line R is connected if and only if it is an
interval. In particular, R is connected.
Proof. Assume X is connected subspace of R: We shall show that X is an interval.
Suppose X is not an interval, then 9 x; y; z such that x < y < z with x; z 2 X but
y2= X: We shall show that this gives a disconnection of X: In fact, consider the sets
[X \ ( 1; y)] [ [X \ (y; +1)] :
Then we see that
X = [X \ ( 1; y)] [ [X \ (y; +1)] :
This gives a disconnection of X and so X is disconnected. This is a contradiction.
Hence, X is an interval.
Conversely, assume that X is an interval. Then we shall show that X is con-
nected. Suppose not, then 9 a disconnection X = A [ B: Since A and B have to
be non-empty, we can choose x 2 A; z 2 B and since A \ B = ;:
We cannot have x = z; hence we have x 6= z: There are two possibilities, either
x < z or z < x: Take x < z and so 9 y such that x < y < z:
X = A [ B =) that 9 x; z such that x 2 A; z 2 B: Since X is an interval,
[x; z] X: Now, de…ne
y = sup f[x; z] \ Ag for …xed x; z:
Then y exist as a real number. Since A is closed, we must have y 2 A because
y is a point of accumulation for A:
[Note: if K is a bounded in…nite set of real numbers and = sup K; then
is a point of accumulation for K: Similarly, if = inf K; then is a point of
accumulation for K ].
Hence, y < z (note that we cannot have equality because z 2 B). Again, by the
de…nition of y, we have the following:
y + 2 B for every > 0 such that y + z (use de…nition of sup). Thus we
must have y 2 B: This is because if we choose a sequence f n g such that n > 0
and y + n z and n ! 0; then y + n ! y and since all such must be in B since
B is closed.
But we have shown earlier that y 2 A which is disjoint from B: This is a con-
tradictiong arising out of our assumption that X is disconnected. Therefore, X is
connected.
Theorem 6.3. The image of a connected metric space under a continuous maps
is connected.
Proof. Let (X; d) and (Y; ) be metric spaces where (X; d) is connected and let
f : X ! Y: We shall show that f (X) is a connected subspace (Y; ): Suppose not,
then 9 two non-empty disjoint open sets G; H in Y such that
f (X) [f (X) \ G] [ [f (X) \ H] :
 M ETRIC SPACES 21

We choose G and H so that their union contains f (X) and their intersection
with f (X) are disjoint and non-empty. This leads to a contradiction because
1 1
X=f [G] [ f [H]
will give a disconnection of X:

References
[1] Mícheál Ó. Searcóid, Metric Spaces, Springer-Verlag, London, 2007.
[2] J. R. Giles, Introduction to the Analysis of Metric Spaces, Cambridge University Press, Cam-
bridge, Great Britain„ 1989.
[3] Victor Bryant, Metric Spaces: iteration and applications, Cambridge University Press, New
York, 1985
[4] G. F. Simmons, Introduction to Topology and Modern Analysis, McGraw-Hill, London, 1963.
[5] Charles E. Chidume, Functional Analysis: an introduction to metric spaces, Longman, Nigeria,
1989