Solutions to Assignment-3

September 19, 2017

1. Let (X, d) be a metric space, and let Y ⊂ X be a metric subspace with the induced metric dY . Let
E ⊂Y.
(a) Show that a set U ⊂ Y is open in Y if and only if there is a subset V ⊂ X open in X such that
U = V ∩ Y . As an example, consider X = R, Y = [0, 1]. Then U = [0, 1/2) is an open subset
of Y with the induced metric. In this case we can take V = (−1, 1/2). Then V is open in R and
U =Y ∩V.

Solution: We will denote the balls of radius r around p ∈ Y with respect to metrics dY and d
by BrY (p) and BrX (p) respectively. That is

BrY (p) = {y ∈ Y | d(p, y) < r}, BrX (p) = {x ∈ X | d(p, x) < r}.

The key observation is that

BrY (p) = BrX (p) ∩ Y.
As an illustration take X = R and Y = Q. Then B1Q (0) consists of all the rationals in the
interval (−1, 1) while B1R (0) is the whole interval (−1, 1) and clearly B1Q (0) = B1R (0) ∩ Q.
Coming back to the problem, suppose U ⊂ Y is open (with respect to the subspace metric dY ).
Then for each p ∈ U , there exists an rp > 0 such that BrYp (p) ⊂ U. So we can write

U = ∪p∈U BrYp (p).

Let V = ∪p∈U BrXp (p). Then clearly V is open in X since arbitrary union of open sets is open,
and by the above property of balls, it is clear that U = V ∩ Y .
For the converse, suppose U = V ∩ Y where V is open in X. Let p ∈ U . Then p ∈ V , and since
V is open in X, there exists rp > 0 such that BrXp (p) ⊂ V . But then BrYp (p) = BrXp (p) ∩ Y ⊂ U .
This shows that for any p ∈ U we have a ball BrYp (p) ⊂ U , and so U is open in Y .

(b) Show that E is compact subset of Y (with respect to the metric dY ) if and only if it is a compact
subset of X (with respect to the metric d).

Solution: Suppose E ⊂ Y ⊂ X is a compact subset of Y (with respect to dY ). Let {Vα } be

a cover of E by open subsets of X. By the above part, Uα = Vα ∩ Y will be a cover of E by
open subsets of (Y, dY ). Since E is compact with the subspace metric, there exists α1 , · · · αN
such that
E ⊂ ∪N N
k=1 Uαk ⊂ ∪k=1 Vαk .

So we have managed to extract a finite sub-cover of {Vα }. This shows that E is compact in
(X, d). The converse is also similar.

1
 (c) Show that E is a connected subset of Y (with respect to the metric dY ) if and only if it is a connected
subset of X (with respect to the metric d).

Solution: Again we show one direction leaving the converse as an excerise. Suppose E ⊂ Y ⊂
X is a connected subset of (Y, dY ) but not (X, d). Then there exists non-empty subsets A and
X X
B of X such that E = A ∪ B but A ∩ B = B ∩ A = φ. Here we are taking closures with
resect to the metric d.
Y Y
Claim. A ∩ B = B ∩ A = φ.
Y
Clearly A ∩ B = φ. Suppose p ∈ A ∩ B. Then p is a limit point of A (in the metric dY ). So
for any r > 0, BrY (p) ∩ A 6= φ. Since BrY (p) ⊂ BrX (p), this shows that BrX (p) ∩ A 6= A for any
r > 0. This shows that p is a limit of A even in the metric d. This is a contradiction since
X
A ∩ B = φ. This completes the proof of the claim.
But then we have written E = A ∪ B where A and B are non-empty separated sets with respect
to the metric dY , contradicting the fact that E is connected with respect to dY . Hence E must
be connected with respect to the metric d too.

2. A subset E ⊂ Rn is called convex if for any two points p, q ∈ E, the straight line
l(t) = (1 − t)p + tq, t ∈ [0, 1]
joining the two points is contained completely in E. Show that any convex set is connected. Hint.
Argue by contradiction.

Solution: If not, then we can write E = A ∪ B, where A and B are non-empty and separated, that
is, A, B 6= Φ, and A ∩ B = A ∩ B = Φ. Recall that this also means that A and B are both open and
closed in E.
Choose p ∈ A and q ∈ B. Since E is convex the straight line l joining them is contained in E.
Intuitively, there will be a first point where the line exits A and enters B. This point will lie in
A ∩ B, a contradiction. More rigorously, let

T = sup{t | l(t) ∈ A for all s ≤ t}.

Then T is the maximum time such that the line is always in A for any time smaller than T .
Claim-1. l(T ) ∈
/ A.
Proof. Now suppose l(T ) ∈ A. Since A is open in E, there is an ε > 0 such that

{x ∈ E | |x − l(T )| < ε} ⊂ A.

Now note that

|l(s) − l(T )| = |(T − s)p + (s − T )q| ≤ |s − T |||p| + |s − T ||q| = |s − T |(|p| + |q|).

ε
So if s ∈ [T, T + |p|+|q| ), then |s − T | < ε/(|p| + |q|), and so

|l(s) − l(T )| < ε.

ε
By our choice of ε and T , this shows that l(s) ∈ A for all s ∈ [0, T + |p|+|q| ), contradicting the
maximality of T . This proves the claim.
Now, since E = A ∪ B, the claim implies that l(T ) ∈ B. In particular T > 0, since l(0) = p ∈ A.
Claim-2. l(T ) is a limit point of A.

2
 Proof. Let ε > 0. We need to show that there is some x ∈ Bε (l(T )) ∩ A. Now let
ε
s=T − .
2(|p| + |q|)

Then by the estimate above,

ε
|l(s) − l(T )| ≤ < ε,
2
and so l(s) ∈ Bε (l(T )). But by definition of T , since s < T , automatically l(s) ∈ A, hence proving
the claim.
So we have now shown that l(T ) ∈ A ∩ B, a contradiction.

3. A metric space (X, d) is called separable is it has a countable dense subset. A collection of open sets
{Uα } is called a basis for X if for any p ∈ X and any open set G containing p, p ∈ Uα ⊂ G for some
α ∈ I. The basis is said to be countable if the indexing set I is countable.
(a) Show that Rn is countable. Hint. Q is dense in R.

Solution: Let Qn ⊂ Rn be the set of all points with all rational coordinates. Then Qn is dense
in Rn , and is countable. So Rn is separable.

(b) Prove that a metric is separable if and only if it has a countable basis of open sets. Hint. One
direction is not hard (which one?). For the other direction, think of smaller and smaller balls of
rational radii.

Solution: Proof of ⇐= . Suppose (X, d) has a countable basis {Uk }. We can assume without
loss of generality that none of the basis elements are identical open sets. Pick a point pk ∈ Uk .
Let P be the collection of points {pk }. Note that unless X is finite (in which case the proof
is trivial anyway), we can always pick points which are not repeated. We claim that X = P .
To see this, let x ∈ X \ P . We have to show that x is a limit point of P . Let G be any open
set containing x. Then by definition of basis, there is a k such that x ∈ Uk ⊂ G. But then
pk ∈ G ∩ P . So every open set around x intersects P at a point different from x itself (since
x∈ / P ). This shows that x is a limit point of P .
Proof of =⇒ . Let S = {x1 , x2 , · · · } be the countable dense subset, and let Un,j = B1/n (xj ).
We claim that {Un,j } forms a basis, and since there are clearly countable number of sets, this
completes the proof. To see that it forms a basis, let x ∈ X and G be an open set containing
x. Since G is open, there is an integer N such that the ball B4/N (x) ⊂ G. Since S is dense,
xj ∈ B1/N (x) for some j. Choose an integer N/2 < M < N Then we claim that

x ∈ B1/M (xj ) ⊂ B4/N (x).

The first inclusion follows since M < N and d(x, xj ) < 1/N < 1/M . For the second inclusion,
let y ∈ B1/M (xj ). Then

1 1 3
d(x, y) ≤ d(x, xj ) + d(y, xj ) < + < .
N M N
In particular,
x ∈ UM,j ⊂ G,
completing the proof.

(c) Prove that every compact metric space is separable. Hint. First, cover the metric space by balls of

3
 radius 1. By compactness only a finitely many such balls are needed (see problem 1). Continue to
do this for balls of radius 1/n for larger and larger integers. Does this give you a countable dense
subset?

Solution: Let X be a compact metric space. By problem 1, for each integer n there is a
number K(n) and points xn,1 , · · · , xn,K(n) such that
K(n)
X ⊂ ∪k=1 B1/n (xn,k ).

We let Un,k = B1/n (xn,k ).

Claim. S := {xn,k } is a dense subset. Since it is clearly countable, this will prove that X is
separable.
Proof. We need to show that for any x ∈ X and any open set G, G ∩ S is non-empty. Since
G is open, there exists an integer N such that B1/N (x) ⊂ G. But since the collection of balls
K(N )
{B1/N (xN,k )}k=1 is a cover for X, there is some k such that x ∈ B1/N (xN,k ). But then
trivially xN,k ∈ B1/N (x) ⊂ G, and so G ∩ S is non-empty. This completes the proof of the
claim.

4. The aim of this exercise is to complete the proof that compactness and limit point compactness are
equivalent. Let (X, d) be a limit point compact metric space.
(a) Show for every δ > 0, X can be covered by finitely many balls of radius δ. (Note that this is easy
for a set already known to be compact; see problem 4 from the previous assignment).

Solution: Pick any point x1 . Then pick x2 such that d(x2 , x1 ) ≥ δ. If there is no such point
then already X = Bδ (x1 ) and the claim is proved with N = 1. Now inductively, having picked
x1 , x2 , · · · , xn−1 pick an xn such that d(xn , xj ) > δ for all j = 1, · · · n − 1. We claim that the
process terminates in a finitely many steps, thus proving the claim. If not, then we have a
sequence {xn }. Since X is limit point compact, there is a limit point p ∈ X. So there is an
infintely many terms of the sequence in the ball Bδ/2 (p). If xn and xm are two such points,
then d(xn , xm ) ≤ d(xn , p) + d(xm , p) ≤ δ/2 + δ/2 = δ contradicting the fact that d(xn , xm ) > δ.
So the claim is proved.

(b) If Fn is a collection of non-empty closed subsets of X such that Fn+1 ⊂ Fn for all n, then show
that ∩∞n=1 Fn is non-empty.

Solution: Choose points xn ∈ Fn . If the range of the sequence {xn } is finite, then clearly
there is a point, say xN , which is in infinitely many of the sets Fn . Then, since the sets are
decreasing, clearly xN will lie in the common intersection. So suppose the range {xn } is infinite.
Then since X is limit point compact, there is a limit point p.
Claim. p ∈ ∩∞ n=1 Fn .
Proof. If not, then there exists and N such that p ∈ / FN . Since FN is closed, there is a ball
Br (p) such that Br (p) ∩ FN = φ. Since Fn ⊂ FN for all n > N , clearly Br (p) ∩ Fn = φ for all
n > N . On the other hand, since p is a limit point, Br (p) ∩ {xn } has infinitely many points.
In particular there is an n > N such that xn ∈ Br (p). But xn ∈ Fn which is a contradiction
since Br (p) ∩ Fn = φ.

(c) Prove that limit point compactness implies compactness.

Solution: The proof is similar to the proof that closed and bounded sets in Rn are compact.
We proceed by contradiction. Suppose there is an open cover {Gα } such that no finite sub

4
collection covers X. By the proof of part (b) there exists a finite collection of points x1 , · · · xN
such that X ⊂ ∪N j=1 B1 (xj ). At least one of the balls cannot be covered by a finite sub collection
of {Gα }. Label this ball B1 . Clearly B 1 is also limit point compact (Why?), and so we can
cover B1 with finitely many balls of radius 1/2 and pick a ball which cannot be covered by a
finite sub-collection from {Gα }. We continue and obtain a sequence of balls {Bj } such that

• Bj+1 ⊂ Bj for all j.

• The radius of Bj is 1/j.
• No finite sub collection of {Gα } covers Bj . In particular, no Bj is contained in any of the
Gα s.

Now, pick a point xj ∈ Bj . If the range {xj } is finite, then one of the points, say x1 , belongs
to infinitely many of the balls Bj . If x1 ∈ Gα , then since Gα is open, there is a j big enough
so that x1 ∈ Bj ⊂ Gα . But this contradicts property 3 above.
If the range {xj } is infinite, then by limit point compactness, there exists a limit point p ∈ X.
Let α such that p ∈ Gα . Since Gα is open there is an r > 0 such that Br (p) ⊂ Gα . Moreover,
since p is a limit point of {xj }, there is subsequence {xjk } such that xjk ∈ Br/2 (p). Choose jk
big enough such that 1/jk < r/2. Then by triangle inequality, Bjk ⊂ Br (p) ⊂ Gα contradicting
property 3.