78 2.
Topology of a Metric Space
Theorem 2.1.44. (Cantor) Let (X, d) be a metric space. Then (X, d) is complete if and
only if, for every nested sequence {Fn }n $ 1 of nonempty closed subsets of X, that is,
(a) F1 F2 T. . . Fn . . . such that (b) d(Fn ) ! 0 as n ! 1,
the intersection 1 n ¼ 1 Fn contains one and only one point.
Proof. First suppose that (X, d) is complete. For each positive integer n, let xn be any
point in Fn . Then by (a),
xn , xnþ1 , xnþ2 , . . .
all lie in Fn . Given e > 0, there exists by (b) some integer n0 such that d(Fn0 ) < e. Now,
xn0 , xn0 þ1 , xn0 þ2 , . . . all lie in Fn0 . For m, n $ n0, we then have d(xm , xn ) # d(Fn0 ) < e.
This shows that the sequence {xn }n $ 1 is a Cauchy sequence in the complete metric
space X. So, it is convergent. Let x 2 X be such that limn!1 xn ¼ x. Now for any given
n, we have the sequence xn , xnþ1 , . . . Fn . In view of this,
x ¼ lim xn 2 Fn ¼ Fn
n!1
since Fn is closed. Hence,
\
1
x2 Fn :
n¼1
If y 2 X and y 6¼ x, then d(y, x) ¼ a > 0. There exists n large enough T so that
d(Fn ) < a ¼ d(y, x), which ensures that y 62 Fn . Hence, y cannot be in 1 n ¼ 1 Fn .
To prove the converse, let {xn }n $ 1 be any Cauchy sequence in X. For each natural
number n, let
Fn ¼ {xm : m $ n}:
Then {Fn }n $ 1 is a nested sequence of closed sets and since {xn }n $ 1 is a Cauchy
sequence,
lim d(Fn ) ¼ 0,
n!1
using Proposition 2.1.43.
Let
\
1
Fn ¼ {x}:
n¼1
If e > 0, then there exists a natural number n0 such that d(Fn0 ) < e. But x 2 Fn0 and
thus n $ n0 implies d(xn , x) < e. &
2.2. Relativisation and Subspaces
Let (X, d) be a metric space and Y a nonempty subset of X. If dY denotes the
restriction of the function d to the set Y Y , then dY is a metric for Y and (Y , dY ) is
2.2. Relativisation and Subspaces 79
a metric space (see Section 1.2). If Z Y X, we may speak of Z being open
(respectively, closed) relative to Y as well as open (respectively, closed) relative to X.
It may happen that Z is an open (respectively, closed) subset of Y but not of X. For
example, let X be R2 with metric d2 and Y ¼ {(x, 0): x 2 R} with the induced
metric. Then Y is a closed subset of X (for Y c ¼ {(x, y) 2 R2 : y 6¼ 0} is open in
X). If Z ¼ {(x, 0): 0 < x < 1}, then Z considered as a subset of Y is open in Y.
However, Z considered as a subset of X is not open in X. In fact, no point (x, 0) 2 Z
is an interior point of Z (Z considered as a subset of X) because any neighbourhood
of (x,0) in X is the ball S( (x, 0), r), r > 0, which is not contained in Z. Thus,
Z ¼ {(x, 0) : 0 < x < 1} is an open subset of Y ¼ {(x, 0) : x 2 R} but not of
X ¼ (R2 , d2 ).
The above examples illustrate that the property of a set being open (respectively
closed) depends on the metric space of which it is regarded a subset. The following
theorem characterises open (respectively closed) sets in a subspace Y in terms of
open (respectively closed) subsets in the space X. First we shall need a lemma.
Lemma 2.2.1. Let (X, d) be a metric space and Y a subspace of X. Let z 2 Y and
r > 0. Then
SY (z, r) ¼ SX (z, r) \ Y ,
where SY (z, r) (respectively SX (z, r)) denotes the ball with centre z and radius r in Y
(respectively X).
Proof. We have
SX (z, r) \ Y ¼ {x 2 X : d(x, z) < r} \ Y
¼ {x 2 Y : d(x, z) < r}
¼ SY (z, r) since Y X: &
Let X ¼ R2 and Y ¼ {(x1 , x2 ): 0 < x1 # 1, pffiffiffi0 # x2 < 1, x1 þ x2 $ 1}. Here, the
2 2
open ball in Y with centre (1, 0) and radius 2 is the entire space Y. (See Figure 2.6.)
Figure 2.6
80 2. Topology of a Metric Space
Theorem 2.2.2. Let (X, d) be a metric space and Y a subspace of X. Let Z be a subset
of Y. Then
(i) Z is open in Y if and only if there exists an open set G X such that
Z ¼ G \ Y;
(ii) Z is closed in Y if and only if there exists a closed set F X such that
Z ¼ F \ Y.
Proof. (i) Let Z be open in Y. Then if z is any point of Z, there exists an open ball
SY (z, r) contained in Z. Observe that the radius r of the ball SY (z, r) depends on the
point z 2 Z. We then have
[
Z¼ SY (z, r)
z2Z
[
¼ (SX (z, r) \ Y ) using Lemma 2:2:1
z2Z
!
[
¼ SX (z, r) \Y
z2Z
¼ G \ Y,
S
where G ¼ z2Z SX (z, r) is open in X.
On the other hand, suppose that Z ¼ G \ Y , where G is open in X. If z 2 Z, then
z is a point of G and so there exists an open ball SX (z, r) such that SX (z, r) G.
Hence,
SY (z, r) ¼ SX (z, r) \ Y by Lemma 2:2:1
G \ Y ¼ Z,
so that z is an interior point of the subset Z of Y. As z is an arbitrary point of Z, it
follows that Z is open in Y.
(ii) Z is closed in Y if and only if (X\Z) \ Y is open in Y. Hence, Z is closed in Y if
and only if there exists an open set G in X such that
(X\Z) \ Y ¼ G \ Y using (i) above:
On taking complements in X on both sides, we have
Z [ (X\Y ) ¼ (X\G) [ (X\Y ):
Hence
Z ¼ Z \ Y ¼ ðZ [ (X\Y )Þ \ Y
¼ ð(X\G) [ (X\Y )Þ \ Y :
¼ (X\G) \ Y
So, Z is the intersection of the closed set X\G and Y.
2.2. Relativisation and Subspaces 81
Conversely, let Z ¼ F \ Y , where F is closed in X. Then X\Z ¼ (X\F) [ (X\Y )
and so
(X\Z) \ Y ¼ ð(X\F) [ (X\Y )Þ \ Y ¼ (X\F) \ Y ,
where X\F is open in X. Hence (X\Z) \ Y is open in Y, i.e., Z is closed in Y. &
Proposition 2.2.3. Let Y be a subspace of a metric space (X, d).
(i) Every subset of Y that is open in Y is also open in X if and only if Y is open in X.
(ii) Every subset of Y that is closed in Y is also closed in X if and only if Y is closed
in X.
Proof. (i) Suppose every subset of Y open in Y is also open in X. We want to
show that Y is open in X. Since Y is an open subset of Y, it must be open in
X. Conversely, suppose Y is open in X. Let Z be an open subset of Y. By
Theorem 2.2.2(i), there exists an open subset G of X such that Z ¼ G \ Y . Since
G and Y are both open subsets of X, their intersection must be open in X, i.e., Z
must be open in X.
(ii) The proof is equally easy and is, therefore, not included. &
Proposition 2.2.4. Let (X, d) be a metric space and Z Y X. If clX Z and clY Z
denote, respectively, the closures of Z in the metric spaces X and Y, then
clY Z ¼ Y \ clX Z:
Proof. Obviously, Z Y \ clX Z. Since Y \ clX Z is closed in Y (see Theorem
2.2.2(ii)), it follows that clY Z Y \ clX Z. On the other hand, by Theorem 2.2.2(ii),
clY Z ¼ Y \ F, where F is a closed subset of X. But then
Z clY Z F,
and hence, by Corollary 2.1.27(ii),
clX Z F:
Therefore,
clY Z ¼ Y \ F Y \ clX Z:
This completes the proof. &
In contrast to the relative properties discussed above, there are some properties
that are intrinsic. In fact, the property of x being a limit point of F holds in any
subspace containing x and F as soon as it holds in the whole space, and conversely.
Another such property is that of being complete. The following propositions
describe relations between closed sets and complete sets.
Proposition 2.2.5. If Y is a nonempty subset of a metric space (X, d), and (Y , dY ) is
complete, then Y is closed in X.
82 2. Topology of a Metric Space
Proof. Let x be any limit point of Y. Then x is the limit of a sequence {yn }n $ 1 in Y. In
view of Proposition 1.4.3, the sequence {yn }n $ 1 is Cauchy, and hence, by assump-
tion, converges to a point y of Y. But by Remark 3 following Definition 1.3.2, y ¼ x.
Therefore, x 2 Y . This shows that Y is closed in X. &
Proposition 2.2.6. Let (X, d) be a complete metric space and Y a closed subset of X.
Then (Y , dY ) is a complete space.
Proof. Let {yn }n $ 1 be a Cauchy sequence in (Y , dY ). Then {yn }n $ 1 is also a Cauchy
sequence in (X, d); so there exists an x 2 X such that limn!1 yn ¼ x. If follows (see
Proposition 2.1.28) that x 2 Y , which is the same set as Y by Corollary 2.1.27(i). &
2.3. Countability Axioms and Separability
Definition 2.3.1. Let (X, d) be a metric space and x 2 X. Let {Gl }l2L be a
family of open sets, each containing x. The family {Gl }l2L is said to be a local base
at x if, for every nonempty open set G containing x, there exists a set Gm in the family
{Gl }l2L such that x 2 Gm G.
Examples 2.3.2. (i) In the metric space R2 with the Euclidean metric, let Gl ¼
S(x, l), where x ¼ (x1 ,x2 ) 2 R2 and 0 < l 2 R. The family {Gl :0 < l 2 R} ¼
{S(x, l):0 < l 2 R} is a family of balls and is a local base at x. Note that S(x, l),
where x ¼ (x1 ,x2 ), can also be described as {(y1 ,y2 ) 2 R2 : (y1 x1 )2 þ (y2 x2 )2 < l2 }.
(ii) Let x ¼ (x1 , x2 ) 2 R2 and Gl ¼ {(y1 , y2 ) 2 R2 : (y1 x1 )2 þ 2(y2 x2 )2 < l},
where 0 < l 2 R. Then the family {Gl : 0 < l 2 R} is a local base at x. To see why,
consider any open set G R2 such that x 2 G. Since G is open, there exists r > 0
such that S(x, r) G. Now S(x, r) ¼ {(y1 , y2 ) 2 R2 : (y1 x1 )2 þ (y2 x2 )2 < r 2 }.
Let l ¼ r 2 . Then y 2 Gl ) (y1 x1 )2 þ 2(y2 x2 )2 < l ) (y1 x1 )2 þ (y2 x2 )2
< l ) (y1 x1 )2 þ (y2 x2 )2 < r 2 ) y 2 S(x, r), so that Gl S(x, r) G. In this
example, the sets Gl are ellipses.
(iii) Let x 2 R. Consider the family of all open intervals (r,s) containing x and
having rational endpoints r and s. This family is a local base at x. It consists of open
balls, not necessarily centred at x. Moreover, the family is countable and thus
constitutes what is called a countable base at x.
Proposition 2.3.3. In any metric space, there is a countable base at each point.
Proof. Let (X, d) be a metric space and x 2 X. The family of open balls centred at x
and having rational radii, i.e., {S(x, r): r rational and positive} is a countable base at
x. In fact, if G is an open set and x 2 G, then by the definition of an open set, there
exists an e > 0 (e depending on x) such that x 2 S(x, e) G. Let r be a positive
rational number less than e. Then
x 2 S(x, r) S(x, e) G: &
2.3. Countability Axioms and Separability 83
Definition 2.3.4. A family {Gl }l2L of nonempty open sets is called a base for the
open sets of (X, d) if every open subset of X is a union of a subfamily of the family
{Gl }l2L .
The condition of the above definition can be expressed in the following equiva-
lent form: If G is an arbitrary nonempty open set and x 2 G, then there exists a set
Gm in the family such that x 2 Gm G.
Proposition 2.3.5. The collection {S(x, e): x 2 X, e > 0} of all open balls in X is a
base for the open sets of X.
Proof. Let G be a nonempty open subset of X and let x 2 G. By the definition of an
open subset, there exists a positive e(x) (depending upon x) such that
x 2 S(x, e(x) ) G:
This completes the proof. &
Generally speaking, an open base is useful if its sets are simple in form. A space
that has a countable base for the open sets has pleasant properties and goes by the
name of ‘‘second countable’’.
Definition 2.3.6. A metric space is said to be second countable (or satisfy the
second axiom of countability) if it has a countable base for its open sets.
The reason for the name second countable is that the property of having
a countable base at each point, as in Proposition 2.3.3, is usually called first
countability.
Examples 2.3.7. (i) Let (R,d) be the real line with the usual metric. The collection
{(x, y) : x, y rational} of all open intervals with rational endpoints form a countable
base for the open sets of R.
(ii) The collection
{S(x, r): x ¼ (x1 , x2 , . . . , xn ), xi rationals, 1 # i # n, and r positive rational}
of all r-balls with rational centres and rational radii is a countable base for the
open sets of the metric space (Rn , d), where d may be any of the metrics on Rn
described in Example 1.2.2(iii).
(iii) Let X have the discrete metric. Then any set {x} containing a single point x is
also the open ball S(x, 1/2) and therefore must be a union of nonempty sets of any
base. So any base has to contain each set {x} as one of the sets in it. If X is
nondenumerable, then the sets {x} are also nondenumerable, forcing every base to
be nondenumerable as well. Consequently, X does not satisfy the second axiom of
countability when it is nondenumerable.
It is easy to see that any subspace of a second countable space is also a second
countable space. In fact, the class of all intersections with the subspace of the sets of
a base form a base for the open sets of the subspace.
84 2. Topology of a Metric Space
Definition 2.3.8. Let (X, d) be a metric space and G be a collection of open sets in X.
If for each x 2 X there is a member G 2 G such that x 2 G, then G is called an open
cover (or open covering) of X. A subcollection of G which is itself an open cover of
X is called a subcover (or subcovering).
Examples 2.3.9. (i) The union of the family { . . . , ( 3, 1), ( 2, 0), ( 1, 1),
(0, 2), . . . } of open intervals is R. The family is therefore an open covering of R.
However, the family of open intervals { . . . , ( 2, 1), ( 1, 0), (0, 1), (1, 2), . . . } is
not an open covering, because the intervals’ union does not contain the integers.
The aforementioned cover contains no subcovering besides itself, because, if we
delete any interval from the family, the midpoint of the deleted interval will not
belong to the union of the remaining intervals.
(ii) Let X be the discrete metric space consisting of the five elements a, b, c, d, e.
The union of the family of subsets {{a}, {b, c}, {c, d}, {a, d, e}} is X and all subsets are
open. Therefore the family is an open cover. The family {{b, c}, {c, d}, {a, d, e}} is a
proper subcover.
(iii) Consider the set Z of all integers with the discrete metric. As in any discrete
metric space, all subsets are open. Consider the family consisting of the three subsets
{3n : n 2 Z}, {3n þ 1 : n 2 Z} and {3n þ 2 : n 2 Z}:
Since every integer must be of the form 3n, 3n þ 1 or 3n þ 2, the above three
subsets form an open cover of Z. There is no proper subcover.
(iv) The family of intervals {( n, n): n 2 N} is an open cover of R and the family
consisting of the open balls {z 2 C : jz þ 17j < n3=2 , n 2 N} is an open cover of C.
If we extract a subfamily by restricting n to be greater than some integer n0, the
subfamily is also an open cover. Indeed, if we delete a finite number of sets in the
family, the remaining subfamily is an open cover. Thus, there are infinitely many
open subcovers.
Definition 2.3.10. A metric space is said to be Lindelöf if each open covering of X
contains a countable subcovering.
Proposition 2.3.11. Let (X, d) be a metric space. If X satisfies the second axiom of
countability, then every open covering {Ua }a2L of X contains a countable subcover-
ing. In other words, a second countable metric space is Lindelöf.
Proof. Let {Gi : i ¼ 1, 2, . . . } be a countable base of open sets for X. Since each Ua is
a union of sets Gi , it follows that a subfamily {Gij : j ¼ 1, 2, . . . } of the base
{Gi : i ¼ 1, 2, . . . } is a covering of X. Choose Uij Gij for each j. Then
{Uij : j ¼ 1, 2, . . . } is the required countable subcovering. &
Definition 2.3.12. A subset X0 of a metric space (X, d) is said to be everywhere
dense or simply dense if X0 ¼ X, i.e., if every point of X is either a point or a limit
2.3. Countability Axioms and Separability 85
point of X0 . This means that, given any point x of X, there exists a sequence of
points of X0 that converges to x.
It follows easily from this definition and the definition of interior (see Definition
2.1.12) that a subset of X0 is dense if and only if X0c has empty interior.
It may be noted that X is always a dense subset of itself; interest centres around
what proper subsets of a metric space are dense.
Examples 2.3.13. (i) The set of rationals is a dense subset of R (usual metric) and so
is the set of irrationals. Note that the former is countable whereas the latter is not.
(ii) Consider the metric space (Rn , d) with any of the metrics described in
Example 1.2.2(iii). Within any neighbourhood of any point in Rn , there is a point
with rational coordinates. Thus,
Qn ¼ Q Q . . . Q
is dense in Rn .
(iii) In the space C[0, 1] of Example 1.2.2(ix), we consider the set C0 consisting of
all polynomials with rational coefficients. We shall check that C0 is dense in C[0, 1].
Let x(t) 2 C[0, 1]. By Weierstrass’ theorem (Theorem 0.8.4), there exists a polyno-
mial P(t) such that
e
sup{jx(t) P(t)j: 0 # t # 1} < ,
2
where e > 0 is given. Corresponding to P(t) there is a polynomial P0 (t) with
rational coefficients such that
e
sup{jP(t) P0 (t)j: 0 # t # 1} < :
2
So,
sup{jx(t) P0 (t)j: 0 # t # 1} < e:
It is easy to see that C0 is countable. In fact, if P n denotes the set of all polynomials
of degree n and having rational coefficients, then the cardinality of P n is the same as
that of Qnþ1 ¼ Q Q . . . Q, which is countable. The assertion now follows
from the fact that a countable union of countable sets is countable.
(iv) Let (X, d) be a discrete metric space. Since every subset is closed, the only
dense subset is X itself.
(v) Let X ¼ ‘p of Example 1.2.2(vii). Recall that the metric is given by
!1=p
X1
p
d(x, y) ¼ jxi yi j :
i¼1
Let E denote the set of all elements of the form (r1 , r2 , . . . , rn , 0, 0 . . .), where ri are
rational numbers and n is an arbitrary natural number. We shall show that E is
86 2. Topology of a Metric Space
dense in ‘p . Let x ¼ (x1 , x2 , . . . ) be an element in ‘p and let e > 0 be given. There
exists a natural number n0 such that
X
1
ep
jxj jp < :
j ¼ n0 þ 1
2
Choose an element x0 ¼ (r1 , r2 , . . . , rn0 , 0, 0, . . . ) in E such that
X
n0
ep
jxj rj jp < :
j ¼1
2
We then obtain
X
n0 X
1
ðd(x, x0 )Þp ¼ jxj rj jp þ jxj jp < ep ,
j ¼1 j ¼ n0 þ 1
and this implies
d(x, x0 ) < e:
Thus, E is dense in (‘p , d). Also, E is countable (in fact, if En denotes the subset of all
thoseSelements x ¼ {ri }i $ 1 such that rj ¼ 0 for j $ n þ 1, then En is countable and
E¼ 1 n ¼ 1 En ).
(vi) By Definition 1.5.1, any metric space is dense in its completion.
Definition 2.3.14. The metric space X is said to be separable if there exists a
countable, everywhere dense set in X. In other words, X is said to be separable if
there exists in X a sequence
{x1 , x2 , . . . } (2:1)
such that for every x 2 X, some sequence in the range of (2.1) converges to x.
Examples 2.3.15. In Examples 2.3.13(i)–(iii) and (v), we saw dense sets that are
countable. Therefore, the spaces concerned are separable. In (iv) however, the space
is separable if and only if the set X is countable.
There are metric spaces other than the discrete metric space mentioned above
which fail to satisfy the separability criterion. The next example is one such case. Let
X denote the set of all bounded sequences of real numbers with metric
d(x, y) ¼ sup{jxi yi j : i ¼ 1, 2, 3, . . . },
as in Example 1.2.2(vi). We shall show that X is inseparable.
First we consider the set A of elements x ¼ (x1 , x2 , . . . ) of X for which each xi is
either 0 or 1 and show that it is uncountable. If E is any countable subset of A, then
the elements of E can be arranged in a sequence s1 , s2 , . . . . We construct a sequence s
as follows. If the mth element of sm is 1, then the mth element of s is 0, and vice versa.
Then the element s of X differs from each sm in the mth place and is therefore equal
2.3. Countability Axioms and Separability 87
to none of them. So, s 62 E although s 2 A. This shows that any countable subset of
A must be a proper subset of A. It follows that A is uncountable, for if it were to be
countable, then it would have to be a proper subset of itself, which is absurd. We
proceed to use the uncountability of the subset A to argue that X must be
inseparable.
The distance between two distinct elements x ¼ (x1 , x2 , . . . ) and y ¼ (y1 , y2 , . . . )
of A is d(x, y) ¼ sup{jxi yi j : i ¼ 1, 2, 3, . . . } ¼ 1. Suppose, if possible, that E0 is a
countable, everywhere dense subset of X. Consider the balls of radii 1/3 whose
centres are the points of E0 . Their union is the entire space X, because E0 is
everywhere dense, and in particular contains A. Since the balls are countable in
number while A is not, in at least one ball there must be two distinct elements x and
y of A. Let x0 denote the centre of such a ball. Then
1 1
1 ¼ d(x, y) # d(x, x0 ) þ d(x0 , y) < þ < 1,
3 3
which is, however, impossible. Consequently, (X, d) cannot be separable.
Proposition 2.3.16. Let (X, d) be a metric space and Y X. If X is separable, then Y
with the induced metric is separable, too.
Proof. Let E ¼ {xi : i ¼ 1, 2, . . . } be a countable dense subset of X. If E is contained in
Y, then there is nothing to prove. Otherwise, we construct a countable dense subset of
Y whose points are arbitrarily close to those of E. For positive integers n and m, let
Sn, m ¼ S(xn , 1=m) and choose yn, m 2 Sn, m \ Y whenever this set is nonempty. We
show that the countable set {yn, m : n and m positive integers} of Y is dense in Y.
For this purpose, let y 2 Y and e > 0. Let m be so large that 1=m < e=2 and find
xn 2 S(y, 1=m). Then y 2 Sn, m \ Y and
1 1 e e
d(y, yn, m ) # d(y, xn ) þ d(xn , yn, m ) < þ < þ ¼ e:
m m 2 2
Thus, yn, m 2 S(y, e). Since y 2 Y and e > 0 are arbitrary, the assertion is proved. &
The main result of this section is the following.
Theorem 2.3.17. Let (X, d) be a metric space. The following statements are equiva-
lent:
(i) (X, d) is separable;
(ii) (X, d) satisfies the second axiom of countability;
(iii) (X, d) is Lindelöf.
Proof. (i))(ii). Let E ¼ {xi : i ¼ 1, 2, . . . } be a countable, dense subset of X and let
{rj : j ¼ 1, 2, . . . } be an enumeration of positive rationals. Consider the countable
collection of balls with centres at xi , i ¼ 1, 2, . . . and radii rj , j ¼ 1, 2, . . . ; i.e.,
{S(xi , rj ): xi 2 E for i ¼ 1, 2, . . . and rj , is rational j ¼ 1, 2, . . . }:
88 2. Topology of a Metric Space
If G is any open set and x 2 G, we want to show that for some i and some
j, x 2 S(xi , rj ) G. Since G is open, there is a ball S(x, d) such that S(x, d) G.
Let rk > 0 be a rational number such that 0 < rk < d. Since x is a point of closure of
E, there is a point xi 2 E such that d(x, xi ) < 1=2rk . Hence,
1
x 2 S(xi , rk ) S(x, rk ) G:
2
In fact, if y 2 S(xi , 1=2rk ) then d(y, x) # d(y, xi ) þ d(xi , x) < 1=2rk þ 1=2rk ¼ rk .
(ii))(iii). See Proposition 2.3.11.
(iii))(i). From each open covering {S(x, e): x 2 X}, we extract a countable
subcovering
S {S(xi , e): xi 2 X, i ¼ 1, 2, . . . } and let A(e) ¼ {x1 , x2 , . . . }. Define
E¼ 1 1
n ¼ 1 A( n ). Then E is a countable, dense subset of X. &
2.4. Baire’s Category Theorem
Definition 2.4.1. Let (X, d) be a metric space. A subset Y X is said to be nowhere
dense if (Y )o is empty, i.e., (Y )o contains no interior point. A subset F X is said
to be of category I if it is a countable union of nowhere dense subsets. Subsets that
are not of category I are said to be of category II.
Remarks 2.4.2. (i) A subset Y of X is nowhere dense if and only if the complement
(Y )c is dense in X, or (X Y ) ¼ X. This follows easily from the remark immedi-
ately after Definition 2.3.12.
(ii) If d denotes the discrete metric, the only nowhere dense set is the null set.
(iii) The notion of being nowhere dense is not the opposite of being everywhere
dense, i.e., not being nowhere dense does not imply that the set is everywhere dense.
For an example of a set which is neither, let R denote the real line with the usual
metric and consider the set Y ¼ {x 2 R : 1 < x < 2}. Then
(Y )o ¼ Y 6¼ 1 and (Y )c ¼ {x 2 R : x < 1 or x > 2} ¼ ( 1, 1) [ (2, 1),
which is not dense in R.
(iv) Every subset must be either of category I or of category II.
(v) It is clear that the null set is of category I. Also, the subset Q of rationals in R
is a set of category I. Indeed, if x1 , x2 , . . . is an enumeration of the rationals, each {xi }
is closed and {xi }o ¼ 1; it follows that [{xi }, the set of all rationals in R, is of
category I.
(vi) Since a denumerable union of denumerable sets is again a denumerable S set, it
follows that, if Y1 , Y2 , . . . are each of category I, then so must be i Yi .
(vii) If X ¼ Y1 [ Y2 and it is known that Y1 is of category I while X is of category
II, then Y2 must be category II. For, if Y2 is of category I, then it follows from (vi)
above that X, too, is of category I, which is a contradiction.