POMA Solution (Chapter 2)

Zhanxiong Xu
Department of Statistics, Penn State University

Summer 2015
Chapter 2

Basic Topology

1. Prove that the empty set is a subset of every set.

Proof. If the statement were not true, then there would exist some set A of which

∅ is not a subset. Thus there exists some x such that x ∈ ∅ but x ∈

/ A. But by

definition, ∅ contains no element, contradiction. Hence the empty set is a subset of

every set.

2. A complex number z is said to be algebraic if there are integers a0 , . . . , an , not all

zero, such that

a0 z n + a1 z n−1 + · · · + an−1 z + an = 0

Prove that the set of all algebraic numbers is countable. Hint: For every positive

integer N there are only finitely many equations with

n + |a0 | + |a1 | + · · · + |an | = N.

Proof. Let A be the set of all algebraic numbers. For every positive integer N , denote

by BN the set {(n, a0 , . . . , an ) ∈ N × Z × · · · × Z : n + |a0 | + · · · + |an | = N }. It is not

1
 hard to see that

∞
[ [
A= {z ∈ C : a0 z n + a1 z n−1 + · · · + an−1 z + an = 0}
N =2 (n,a0 ,...an )∈BN
∞
[ [
:= Eα .
N =2 α∈BN

By the fundamental theorem of algebra, each Eα is finite. By construction, BN is

finite for every fixed N . Finally, the countable union of finite sets is at most countable.

Therefore, A is at most countable. On the other hand, A is infinite, since A obviously

contains the set of all integers Z as a proper subset. In summary, A is countable.

3. Prove that there exist real numbers which are not algebraic.

Proof. If every real number is algebraic, we then have R ⊂ A. However, by Corollary

2.4.1., R is uncountable while the preceding problem says A is countable. This is

contradiction, since every subset of a countable set is at most countable (Theorem

2.1.1.). Therefore there exists real numbers which are not algebraic.

4. Is the set of all irrational real numbers countable?

Proof. No, it is not. If so, then R = Q ∪ (R\Q) would also be countable, contradiction.

5. Construct a bounded set of real numbers with exactly three limit points.

Proof. The answer for this problem is apparently not unique. As one answer, let

E1 = {1/2, 1/3, 1/4, . . .}, E2 = {1} + E1 := {3/2, 4/3, 5/4, . . .}, E3 = {2} + E1 . Set

E = E1 ∪ E2 ∪ E3 . E is clearly bounded with exactly three limit points {0, 1, 2}.

6. Let E 0 be the set of all limit points of a set E. Prove that E 0 is closed. Prove that E

and E have the same limit points. Do E and E 0 always have the same limit points?

2
 Proof. We show that (E 0 )c is open. For every p ∈ (E 0 )c , since p is not a limit point

of E, there exists some δ > 0, such that Nδ (p) contains no limit point of E. In other

words, Nδ (p) ⊂ (E 0 )c . Therefore, (E 0 )c is open, i.e., E 0 is closed.

To prove E 0 = E 0 , we first show for any two sets A and B, we have (A ∪ B)0 = A0 ∪ B 0 .

On one hand, if p is a limit point of A ∪ B, then for any ε > 0, ∃q ∈ A ∪ B, q 6= p,

such that d(p, q) < ε. If q ∈ A, then p ∈ A0 otherwise p ∈ B 0 , hence p ∈ A0 ∪ B 0 . On

the other hand, if p ∈ A0 ∪ B 0 , then p ∈ A0 ⊂ (A ∪ B)0 or p ∈ B 0 ⊂ (A ∪ B)0 , i.e.,

p ∈ (A ∪ B)0 . Now since E = E ∪ E 0 , we have

E 0 = (E ∪ E 0 )0 = E 0 ∪ (E 0 )0 = E 0 .

where the third equality follows from that E 0 is closed hence (E 0 )0 ⊂ E 0 .

E and E 0 don’t always have the same limit points. For example, if the metric space

X = R, E = {1, 1/2, 1/3, . . .}, then E 0 = {0}, whereas (E 0 )0 = ∅.

7. Let A1 , A2 , A3 , . . . be subsets of a metric space.

Sn Sn
(a) If Bn = i=1 Ai , prove that Bn = i=1 Ai , for n = 1, 2, 3, . . ..
S∞ S∞
(b) If B = i=1 Ai , prove that B ⊃ i=1 Ai .

Show, by an example, that this inclusion can be proper.

Proof.

(a) It suffices to show the statement for n = 2 for which

B 2 = A1 ∪ A2

= (A1 ∪ A2 ) ∪ (A1 ∪ A2 )0

= (A1 ∪ A2 ) ∪ (A01 ∪ A02 )

= (A1 ∪ A01 ) ∪ (A2 ∪ A02 )

= A1 ∪ A2 .

3
 The third equality holds following the proposition we proved in last problem.
S∞
(b) Let p ∈ i=1 Ai , then there exists i ∈ N such that p ∈ Ai , hence p ∈ Ai or p ∈ A0i .

If p ∈ Ai , then p ∈ Ai ⊂ B ⊂ B. If p ∈ A0i , ∀ε > 0, there exists q ∈ Ai , q 6= p

such that d(p, q) < ε. Since q ∈ Ai ⊂ B, it follows that p ∈ B. Therefore,

S∞
i=1 Ai ⊂ B.

Let the metric space X = R, Ai = {1/i}, i = 1, 2, 3, . . ., B = {1, 1/2, 1/3, . . .}.

S∞ S∞
Thus B = {0, 1, 1/2, 1/3, . . .}, however, i=1 Ai = i=1 Ai = B $ B.

8. Is every point of every open set E ⊂ R2 a limit point of E? Answer the same question

for closed sets in R2 .

Proof. By the definition of open set, the first statement is true. For closed sets, the

statement is not true. For example, all singletons in R2 are closed, but they are not

limit points of themselves.

9. Let E ◦ denote the set of all interior points of a set E. E ◦ is called the interior of E.

(a) Prove that E ◦ is always open.

(b) Prove that E is open if and only if E ◦ = E.

(c) If G ⊂ E and G is open, prove that G ⊂ E ◦ .

(d) Prove that the complement of E ◦ is the closure of the complement of E.

(e) Do E and E always have the same interiors?

(f) Do E and E ◦ always have the same closures?

Proof.

(a) Let p ∈ E ◦ , then there exists δ > 0 such that Nδ (p) ⊂ E. For every q ∈ Nδ/2 (p), if

d(x, q) < δ/2, then by triangle inequality, d(x, p) ≤ d(x, q)+d(p, q) < δ/2+δ/2 =

δ, implying x ∈ E, hence Nδ/2 (q) ⊂ E, that is, q ∈ E ◦ . Therefore, Nδ/2 (p) ⊂ E ◦ ,

showing that E ◦ is an open set.

4
 (b) If E is open, then by definition, E ⊂ E ◦ . On the other hand, E ◦ ⊂ E, hence

E = E ◦ . Conversely, if E = E ◦ , since we showed in (a) that E ◦ is open, therefore

E is open.

(c) Since G ⊂ E, it follows that G◦ ⊂ E ◦ . By (b), G = G◦ provided G is open.

Therefore, G ⊂ E ◦ .

(d) We need to show (E ◦ )c = E c . Since E ◦ is open, (E ◦ )c is closed. Hence (E ◦ )c =

(E ◦ )c . In addition, E ◦ ⊂ E, hence E c ⊂ (E ◦ )c = (E ◦ )c . For the contrary,

if p ∈ (E ◦ )c , then it is not an interior point of E, that is, for every ε > 0,

∃q ∈ E c , such that d(p, q) < ε. If such q = p, then p ∈ E c ⊂ E c . If q 6= p, then

p ∈ (E c )0 ⊂ E c . Thus (E ◦ )c ⊂ E c . In summary, (E ◦ )c = E c .

(e) No. For X = R1 , E = Q. E has no interior while E = R1 has the whole real line

as its interior.

(f) No. Use the same counterexample as in (e). The closure of E is R1 , while E ◦ = ∅

with its closure ∅.

10. Let X be an infinite set. For p ∈ X and q ∈ X, define


 1 if p 6= q

d(p, q) = .
 0

if p = q

Prove that this is a metric. Which subsets of the resulting metric space are open?

Which are closed? Which are compact?

Proof. Requirements (a), (b) in the definition of metric space are readily checked.

For p ∈ X, q ∈ X, if p = q, then d(p, q) = 0 ≤ d(p, r) + d(r, q) for any r ∈ X. If

p 6= q, then d(p, q) = 1. For any r in X, if it is distinct from both p and q, then

d(p, r) + d(r, q) = 1 + 1 = 2 > d(p, q). If r = p or r = q, then one and only one of

d(p, r) and d(r, q) equals to 0, hence d(p, q) = d(p, r) + d(r, q). The triangle inequality

holds. Therefore X with the distance d is a metric space.

5
 Every subset E of X is both open and closed. First note that every singleton {p} is

open, as N0.5 (p) = {q ∈ X : d(p, q) < 0.5} = {p} ⊂ {p}. Since the union of open sets

is open, E = p∈E {p} is open. Consequently, E = (E c )c is closed, in view of E c is

S

open and E c is a subset of X. Only finite subsets of E are compact. If E were infinite

and compact, then there exists a limit point p in E (cf, the equivalance between (b)

and (c) in Theorem 2.3.7., which actually applies to any metric space), which means,

for every ε ∈ (0, 1), ∃q 6= p, q ∈ E, such that d(p, q) < ε, however, d(p, q) = 1 by

definition, contradiction. Hence any infinite subset of X is not compact.

11. For x ∈ R1 , define

d1 (x, y) = (x − y)2 ,
p
d2 (x, y) = |x − y|,

d3 (x, y) = |x2 − y 2 |,

d4 (x, y) = |x − 2y|,
|x − y|
d5 (x, y) = .
1 + |x − y|

Determine, for each of these, whether it is a metric or not.

Proof. d1 is not a metric since d(1, −1) = 4, whereas d(1, 0) + d(0, −1) = 1 + 1 = 2 <

d(1, −1). The triangle inequality fails to hold.

d3 is not a metric since d3 (1, −1) = 0 despite of 1 6= 1.

d4 is not a metric since d4 (1, 1/2) = 0 despite of 1 6= 1/2.

d2 is a metric. (a), (b) in the definition of metric hold trivially. For (c),

p p qp p p p
|x − y| ≤ |x − z| + |y − z| ≤ ( |x − z| + |y − z|)2 = |x − z| + |y − z|.

d5 is a metric, as above, it suffices to show d5 satisfies the triangle inequality, to prove

x
it, notice the function f (x) = 1+x , x ≥ 0 is monotonically increasing on [0, +∞). Since

6
 |x − y| ≤ |x − z| + |z − y| holds for all x, y, z ∈ R1 , we have

|x − z| + |y − z|
d5 (x, y) = f (|x − y|) ≤ f (|x − z| + |y − z|) =
1 + |x − z| + |y − z|
|x − z| |z − y|
= +
1 + |x − z| + |y − z| 1 + |x − z| + |y − z|
|x − z| |z − y|
≤ + = d5 (x, z) + d5 (z, y).
1 + |x − z| 1 + |z − y|

12. Let K ⊂ R1 consist of 0 and the numbers 1/n, for n = 1, 2, 3, . . .. Prove that K is

compact directly from the definition (without using the Heine-Borel theorem).

Proof. Let {Gα } be an open cover of K. Suppose 0 is covered by Gα0 for some α0 .

Since Gα0 is open, 0 ∈ Gα0 , there exists δ > 0 such that Nδ (0) ⊂ Gα0 . Let N be the

largest integer such that 1/N ≥ δ. Then for all n > N , |1/n − 0| = 1/n < δ, i.e.,

1/n ∈ Nδ (0) ∈ Gα0 . Let Gαi , i = 1, . . . , N be the open sets that cover 1, 1/2, . . . , 1/N ,

respectively. It then follows that K ⊂ Gα0 ∪ Gα1 ∪ · · · ∪ GαN . Therefore, K is

compact.

13. Construct a compact set of real numbers whose limit points form a countable set.

Proof. Notation: For two subsets A and B of R1 , denote by A + B the set {a + b : a ∈

A, b ∈ B}.

Set E1 = {1/2} + {1/n, n > 2, n ∈ N}, E2 = {1/22 } + {1/n : n > 22 , n ∈ N}, . . . , Ek =

{1/2k } + {1/n : n > 2k , n ∈ N}. Note that these Ek ’s are disjoint.

S∞
Let K = {0} ∪ k=1 Ek .

Clearly, K is bounded, since for every x ∈ K, 0 ≤ x ≤ 1. Moreover, it is easily seen

K 0 = {0, 1/2, 1/22 , . . .}, which is a countable set. Meanwhile, K 0 ⊂ K, hence K is

closed. Since K is bounded and closed, K is compact.

14. Give an example of an open cover of the segment (0, 1) which has no finite subcover.

7
 Proof. Let O = {Nx/2 (x) : x ∈ (0, 1)}. It is easy to check O is an open cover of (0, 1). If

it had a finite subcover of (0, 1), we can denote it by U = {Nx1 /2 (x1 ), . . . , Nxm /2 (xm )}.

Denote xk = min{x1 , . . . , xm }. Consider xk /3 ∈ (0, 1), it is not contained in any set

of U , contradiction. Hence O doesn’t admit any finite subcover of (0, 1).

15. Show that Theorem 2.36 and its Corollary become false (in R1 , for example) if the

word “compact” is replaced by “closed” or “bounded”.

Proof. Let X = R1 .

If “compact” is replaced by “closed”, we can set Kn = [n, +∞), n = 1, 2, . . .. Then

Kn = ∅ (if we restricted ourselves to R1 !).

T

If “compact” is replaced by “bounded”, we can set Kn = (0, 1/n), n = 1, 2, . . .. Then

T
Kn = ∅.

16. Regard Q, the set of all rational numbers, as a metric space, with d(p, q) = |p − q|. Let

E be the set of all p ∈ Q such that 2 < p2 < 3. Show that E is closed and bounded

in Q, but E is not compact. Is E open in Q?

Proof. Obviously E is bounded. To show E is closed in Q, let p0 ∈ Q be a limit point

of E, we need to show p0 ∈ E. Otherwise, p0 ∈ / E, hence p20 ≥ 3 or p20 ≤ 2. However,

√ √ √ √
since p0 ∈ Q, p0 6= 3.p0 6= 2. Thus we have p0 > 3 or p0 < 2, under which case
√ √
it cannot be a limit point of E. For example, if p0 > 3. Setting ε = (p0 − 3)/2 > 0,

then for every q ∈ Nε (p0 ), |q − p0 | < ε. Consequently,

√
p0 3 √
|q| = |q − p0 + p0 | ≥ |p0 | − |q − p0 | > p0 − ε = + > 3
2 2

hence q 2 > 3, i.e., q ∈

/ E. In other words, Nε (p0 ) doesn’t contain any point in E,

contradiction. Therefore p0 ∈ E, and E is closed.

√
Let F = {qn } ⊂ E be an strictly increasing sequence converges to 3, for example,

{qn } can be taken as 1.7, 1.73, 1.732, 1.73205, . . .. If E is compact, by Theorem 2.37,

8
 √ √
F has a limit point in E, denote it by q0 . Since q0 ∈ E, q0 < 3. Since qn ↑ 3
√
as n → ∞. For ε = 3 − q0 > 0, there exists N0 ∈ N, such that for all n > N0 ,
√
0 < 3 − qn < 12 ε. Let δ = min{|q1 − q0 |, . . . , |qN0 − q0 |, 12 ε} > 0 (if q0 = qi for some

i ∈ {1, . . . , N0 }, q0 clearly can’t be a limit point of F ). It can be seen Nδ (q0 ) contains

no point in F , since for every qi with i ≤ N0 , |qi − q0 | ≥ δ. For every qi with i > N0 ,
√ √ √
|qi − q0 | ≥ | 3 − q0 | − |qi − 3| > 3 − q0 − 12 ε = 12 ε ≥ δ. Thus q0 is not a limit point

of F , contradiction. Therefore E is not compact.

√ √ √ √
Yes, E is open in Q. Suppose p ∈ E, then 2<p<
3. Let r = min{ 3−p, p− 2},
√ √
then Nr (p) = {q ∈ E : |p − q| < r} ⊂ E, as |q| ≤ |p| + |p − q| < p + 3 − p = 3 and
√ √
|q| ≥ |p| − |p − q| > p − (p − 2) = 2.

17. Let E be the set of all x ∈ [0, 1] whose decimal expansion contains only the digits 4

and 7. Is E countable? Is E dense in [0, 1]? Is E compact? Is E perfect?

Proof. There is no essential change if we replace 0 and 1 in Theorem 2.14 by 4 and 7,

hence E is uncountable.

E is not dense in [0, 1], in fact, inf E = 0.444 · · · = 94 , sup E = 0.777 · · · = 79 . Therefore,

for example, there is no point of E is in the neighborhood N0.1 (0), hence it is not dense

in [0, 1].

E is compact. First, E is bounded, as stated above. Second, E is closed. Since

if p is a limit point of E, assume p = 0.n1 n2 n3 · · · . If p ∈

/ E, then there exists

some k ∈ N such that nk ∈

/ {4, 7}. Let k0 be the smallest positive integer that

nk ∈ / {4, 7}. Let r = 10−(k0 +1) , we claim

/ {4, 7}, i.e., ni ∈ {4, 7}, i ≤ k0 − 1, nk0 ∈

that the r-neighborhood of p contains no point of E. More thoroughly, we have

p − r = 0.n1 · · · nk0 −1 nk0 · · · − r ≥ 0.n1 · · · nk0 −1 − r = 0.n1 · · · (nk0 −1 − 1)99, p + r ≤

0.n1 · · · nk0 −1 999 · · · + r = 0.n1 · · · (nk0 −1 + 1)01, i.e., (p − r, p + r) ⊂ (0.n1 · · · (nk0 −1 −

1)99, 0.n1 · · · (nk0 −1 + 1)01) := I. If nk0 ∈

/ {4, 7}, it is impossible to have a point p falls

in I, hence in (p − r, p + r). Therefore p cannot be a limit point of E, contradition.

Hence p ∈ E, E is closed.

9
 E is perfect. First, we have shown E is closed, so it remains to show every point

of E is also the limit point of E. Let p = 0.n1 n2 · · · be any number of E and

ε > 0 be any given positive. By Archimedean property of R, there exists k ∈ N

such that 10−k < ε, implying p − ε < p − 10−k = 0.n1 · · · (nk − 1)nk+1 nk+2 · · · ,

p + ε > p + 10−k = 0.n1 · · · (nk + 1)nk+1 nk+2 · · · . In other words, (p − ε, p + ε) ⊃

(0.n1 · · · (nk − 1)nk+1 nk+2 · · · , 0.n1 · · · (nk + 1)nk+1 nk+2 · · · ) := J. It is then easily

seen that the number 0.n1 · · · nk n0k+1 n0k+2 · · · , which belongs to E and distinct from

p (where for i ≥ k + 1, n0i = 4, if ni = 7; n0i = 7, if ni = 4), falls in J, hence in

(p − ε, p + ε), proving that p is a limit point of E.

18. Is there a nonempty perfect set in R1 which contains no rational number?

Proof. Yes. We just need to modify the construction of the Cantor set a little bit.
√ √
For example, replace the original interval [0, 1] by [ 3, 3 3], and then proceeding in

exactly the same way.

19. (a) If A and B are disjoint closed sets in some metric space X, prove that they are

separated.

(b) Prove the same for disjoint open sets.

(c) Fix p ∈ X, δ > 0, define A to be the set of all q ∈ X for which d(p, q) < δ, define

B similarly, with > in place of <. Prove that A and B are separated.

(d) Prove that every connected metric space with at least two points is uncountable.

Hint: Use (c).

Proof.

(a) Since A and B are closed, A = A, B = B. From A ∩ B = ∅, we have A ∩ B =

A ∩ B = ∅, i.e., A and B are separated.

(b) It suffices to show A ∩ B 0 = ∅. If p ∈ A ∩ B 0 , then since A is open, there exists

some δ > 0 such that Nδ (p) ⊂ A. For such δ, since p ∈ B 0 , there exists q ∈ B,

10
 such that q 6= p, q ∈ Nδ (p) ⊂ A. Thus we have q ∈ A ∩ B, contradiction. Hence

A ∩ B 0 = ∅. Similarly, B ∩ A0 = ∅. Namely, A and B are separated.

(c) Clearly, A and B such defined are disjoint. In addition, A = Nδ (p) is open, by

Theorem 2.19. Similarly, we can show B is open. By (b), A and B are separated.

(d) By assumption, we have two distinct points p and q in the connected metric space

X. Let δ be any positive number such that d(p, q) > δ and set A = {r ∈ X :

d(p, r) < δ}, B = {r ∈ X : d(p, r) > δ}. Clearly, A 6= ∅ as p ∈ A, B 6= ∅

as q ∈ B. Additionally, both A and B are open sets. By (c), A and B are

separated. Since X is connected. X is not the union of A and B, therefore

{r ∈ X : d(p, r) = δ} is not empty. Therefore we can construct a one-to-one

mapping of E = (0, d(p, q)) into X by assigning δ ∈ E to an element r ∈ X such

that d(p, r) = δ. Since E is uncountable, X must be uncountable.

20. Are closures and interiors of connected sets always connected? (Look at subsets of

R2 .)

Proof. No. For example E = (−∞, +∞) × {0} is connected in R2 . Clearly, E = E,

E ◦ = ∅. Hence E and E ◦ are separated.

21. Let A and B be separated subsets of some Rk , suppose a ∈ A, b ∈ B, and define

p(t) = (1 − t)a + tb

for t ∈ R1 . Put A0 = p−1 (A), B0 = p−1 (B). [Thus t ∈ A0 if and only if p(t) ∈ A.]

(a) Prove that A0 and B0 are separated subsets of R1 .

(b) Prove that there exists t0 ∈ (0, 1) such that p(t0 ) ∈

/ A ∪ B.

(c) Prove that every convex subset of Rk is connected.

Proof.

11
(a) If A0 ∩ B0 6= ∅, assume t0 ∈ A0 ∩ B0 . Since A and B are separated, a ∈ A, b ∈ B,

we have a 6= b, or |b − a| > 0. If t0 ∈ B0 , (1 − t0 )a + t0 b ∈ A ∩ B ⊂ A ∩ B. If t0 ∈ B00 ,

then for arbitrary ε > 0, there exists t ∈ B0 such that |t − t0 | < ε/|b − a|, thus for

this t,

ε
|[(1 − t)a + tb] − [(1 − t0 )a + t0 b]| = |t − t0 ||b − a| < |b − a| = ε
|b − a|

that is (1 − t0 )a + t0 b ∈ B 0 . Therefore A and B can’t be separated. Similarly, if

A0 ∩ B0 6= ∅, we can also derive A and B are not separated. Therefore, A0 and B0

are separated.

(b) Assume that for every t ∈ (0, 1), p(t) ∈ A ∪ B. Let A1 = {t ∈ (0, 1) : p(t) ∈ A},

B1 = {t ∈ (0, 1) : p(t) ∈ B}. Since A ∩ B = ∅, A1 ∩ B1 = ∅. Additionally, by

assumption, A1 ∪ B1 = (0, 1). Notice first that A1 6= ∅, B1 6= ∅. For if one of them,

for instance, A1 is empty, then (0, 1) = B1 , which implies a ∈ B 0 , contradicts with the

assumption that A ∩ B = ∅. Next, we show both A1 and B1 are open. Take t0 ∈ A1 ,

then a0 = p(t) ∈ A. Since A ∩ B = ∅, a0 ∈ B c . Since B is closed, B c is open, hence

∃δ > 0 such that Nδ (a0 ) ∈ B c . Now for every t such that |t − t0 | < δ/|b − a|,

|p(t) − a0 | ≤ |t − t0 ||b − a| < δ

Hence p(t) ∈ Nδ (a0 ), p(t) ∈

/ B, i.e., p(t) ∈ A. Therefore A1 is open. Similarly, B1

is open. Now we express (0, 1) as the union of two nonempty disjoint open sets (by

Exercise 19 (b), these two sets are separated), which contradicts with the fact that

(0, 1) is connected. Therefore, the assumption cannot hold, and there exists t0 ∈ (0, 1)

such that p(t) ∈ A ∪ B.

(c) Suppose C is a convex set in Rk , if it is not connected, then there exist two

nonempty separated sets A and B, such that C = A ∪ B. Take a ∈ A, b ∈ B. By

convexity of C, since A and B are separated, for every t ∈ (0, 1), p(t) = (1 − t)a + tb ∈

C = A ∪ B. On the other hand, by the result of (b), there exists t0 ∈ (0, 1) such that

12
 p(t0 ) ∈
/ A ∪ B, contradiction. Therefore, C must be connected.

22. A metric space is called separable if it contains a countable dense subset. Show that

Rk is separable. Hint: Consider the set of points which have only rational coordinates.

Proof. Let E be the set of points which have only rational coordinates, namely, E =

{(x1 , . . . , xk ) ∈ Rk : xi ∈ Q, 1 ≤ i ≤ k}. By Theorem 2.13, E is countable. To show

it is dense in Rk , take any point p = (p1 , . . . , pk ) ∈ Rk , we need to show p is a limit

point of E or a point of E. If p ∈ E, there is nothing to prove. If p ∈ E c , let ε be

any given positive. Since Q is dense in R1 (which follows from the construction of R
√
from Q), there exist qi ∈ Q such that |pi − qi | < ε/ k, for i = 1, 2, . . . , k. Clearly,

q := (q1 , . . . , qk ) ∈ E, and since p ∈ E c , p 6= q, in addition,

v v
u k u k
uX uX ε2
|p − q| = t (pi − qi )2 < t =ε
i=1 i=1
k

showing that q ∈ Nε (p), hence p ∈ E 0 .

In summary, the countable set E is dense in Rk , thus Rk is separable.

23. A collection {Vα } of open subsets of X is said to be a base for X if the following is true:

For every x ∈ X and every open set G ⊂ X such that x ∈ G, we have x ∈ Vα ⊂ G for

some α. In other words, every open set in X is the union of a subcollection of {Vα }.

Prove that every separable metric space has a countable base. Hint: Take all neigh-

borhoods with rational radius and center in some countable dense subset of X.

Proof. Since X is separable, there exists a countable subset E of X that is dense in X.

By hint, let V be the collection of all neighborhoods with rational radius and center

in E. Clearly, V is countable. We shall show V is a base for X.

Fix x ∈ X, for every open set G ⊂ X such that x ∈ G. Since G is open, there

exists δ > 0 such that Nδ (x) ⊂ G. Because E is dense in X, there exists p ∈ E such

that d(x, p) < δ/3. Take r ∈ Q such that δ/3 < r < 2δ/3, then Nr (p) ∈ V , and

13
 d(x, p) < δ/3 < r implies x ∈ Nr (p). Finally we shall show Nr (p) ⊂ Nδ (x), hence

Nr (p) ⊂ G. This follows from that for every q ∈ Nr (p), d(q, x) ≤ d(q, p) + d(p, x) <

r + δ/3 < 2δ/3 + δ/3 = δ. Therefore, x ∈ Nr (p) ⊂ G, and thus V is a base for X.

24. Let X be a metric space in which every infinite subset has a limit point. Prove that

X is separable. Hint: Fix δ > 0, and pick x1 ∈ X. Having chosen x1 , . . . , xj ∈ X,

choose xj+1 ∈ X, if possible, so that d(xi , xj+1 ) ≥ δ for i = 1, . . . , j. Show that

this process must stop after a finite number of steps, and that X can therefore be

covered by finitely many neighborhoods of radius δ. Take δ = 1/n (n = 1, 2, 3, . . .),

and consider the centers of the corresponding neighborhoods.

Proof. The hint almost says everything, let me just fill in the gaps.

If the process described in the hint can be proceeded infinitely many times, we can

get an infinite subset E = {x1 , x2 , x3 , . . .} of X, where xi are the points accumulated

during the construction process. Note by construction, d(p, q) ≥ δ for any two distinct

points in E. By assumption, E has a limit point, which we shall denote by p0 .

By defnition, there exists p1 ∈ E such that p1 6= p0 and d(p0 , p1 ) < δ/2. Now

since d(p0 , p1 ) > 0, there exists p2 ∈ E such that p2 6= p0 and d(p0 , p2 ) < d(p0 , p1 ).

Since d(p1 , p2 ) ≥ d(p1 , p0 ) − d(p0 , p2 ) > 0, p1 and p2 are distinct points in E, hence

d(p1 , p2 ) ≥ δ. On the other hand,

δ
d(p1 , p2 ) ≤ d(p1 , p0 ) + d(p0 , p2 ) < 2d(p0 , p1 ) < 2 × = δ.
2

This contradiction shows that the process must stop after a finite number of steps, say

N steps. In other words, there doesn’t exist any x ∈ X such that d(x, xi ) ≥ δ for all

i = 1, . . . , N , or equivalently, for every x ∈ X, there exists i ∈ {1, . . . , N }, such that

d(x, xi ) < δ, which proves that X can be covered by finitely many neighborhoods of

radius δ.
(n) (n)
For every n ∈ N , let N1/n (x1 ), . . . , N1/n (xmn ) be the neighborhoods of radius 1/n

such that X is covered by them (this is possible based on the preceding argument),

14
 (n)
then set V = {N1/n (xj ), n = 1, 2, . . . , j = 1, . . . , mn }. Clearly V is countable. In

addition, for any x in X and ε > 0, choose n0 sufficiently large so that 1/n0 < ε.
(n0 ) (n )
Since X can be covered by N1/n0 (x1 ), . . . , N1/n0 (xmn00 ), there exists some m ∈
(n ) (n )
{1, . . . , mn0 } such that x ∈ N1/n0 (xm 0 ), i.e., d(x, xm 0 ) < 1/n0 < ε, proving that V

is dense in X.

In summary, X is separable.

25. Prove that every compact metric space K has a countable base, and that K is sep-

arable. Hint: For every positive integer n, there are finitely many neighborhoods of

radius 1/n whose union covers K.

Proof. For every positive integer n, {N1/n (p) : p ∈ K} is an open cover of K.

(n) (n)
Since K is compact, it has a finite subcover {N1/n (p1 ), . . . , N1/n (pmn )}. Thus
(n)
V = {N1/n (pj ), n = 1, 2, . . . , j = 1, . . . , mn } is a countable base for K. That V

contains countably infinite open sets is obvious. In addition, for every p in K and

every open set G ⊂ K such that p ∈ G, since G is open, there exists δ > 0 such that

Nδ (p) ⊂ G. For this δ > 0, we can choose n0 sufficiently large such that n10 < 2δ , since
Smn0 (n ) (n )
K ⊂ j=1 N1/n0 (pj 0 ), there exists j0 ∈ {1, . . . , mn0 } such that p ∈ N1/n0 (pj0 0 ).
(n ) (n ) (n )
Since for every q ∈ N1/n0 (pj0 0 ), d(q, p) ≤ d(q, pj0 0 ) + d(pj0 0 , p) < 1/n0 + 1/n0 =
(n )
2/n0 < δ, i.e., N1/n0 (pj0 0 ) ⊂ Nδ (p) ⊂ G.

In summary, K has a countable base, as a consequence, K is separable.

26. Let X be a metric space in which every infinite subset has a limit point. Prove that

X is compact. Hint: By Exercises 23 and 24, X has a countable base. It follows that

every open cover of X has a countable subcover {Gn }, n = 1, 2, 3, . . .. If no finite

subcollection of {Gn } covers X, then the complement Fn of G1 ∪ · · · ∪ Gn is nonempty

T
for each n, but Fn is empty. If E is a set which contains a point from each Fn ,

consider the limit point of E, and obtain a contradiction.

Proof. The hint is detailed enough except for the last sentence may need further

explanation. First we need to show the set E is infinite. In fact, we construct E

15
 in the way to make sure E is infinite. For G1 , since F1 := Gc1 6= ∅, there exists

p1 ∈ Gc1 . Note p1 must be covered by one of the other open sets, say G2 . Now since

F2 := (G1 ∪ G2 )c 6= ∅, there exists p2 ∈ F2 . By construction, p2 6= p1 . Proceeding, we

can form E = {p1 , p2 , . . .} such that the elements in E are all distinct, hence E is an
T
infinite set. By condition, E has a limit point, denote it by p. We shall show p ∈ Fn .

If p ∈
/ F1 , then p ∈ G1 , since G1 is open, there exists r > 0 such that Nr (p) ⊂ G1 , by

construction, no points of E is in G1 , hence p cannot be the limit point of E, therefore

p ∈ F1 . If p ∈
/ F2 , then p ∈ G1 ∪ G2 , as we have shown p ∈
/ G1 , p ∈ G2 . Clearly p 6= p2

otherwise p is not in E 0 , hence as before, we can argue p ∈

/ E 0 . Similarly, we can show
T T
p ∈ Fn for each n ∈ N. Namely, p ∈ Fn , in contradict with the fact that Fn = ∅.

In summary, for each open cover {Gα } of K, we can find a finite subcover of K, hence

K is compact.

27. Define a point p in a metric space X to be a condensation point of a set E ⊂ X if

every neighborhood of p contains uncountably many points of E.

Suppose E ∈ Rk , E is uncountable, and let P be the set of all condensation points of

E. Prove that P is perfect and that at most countably many points of E are not in P .

In other words, show that P c ∩ E is at most countable. Hint: Let {Vn } be a countable

base of Rk , let W be the union of those Vn for which E ∩ Vn is at most countable, and

show that P = W c .

Proof. We will go following what the hint suggests. To show P = W c , first show

P ⊂ W c . Let p ∈ P , so that every neighborhood contains uncountably many points

of E. If p ∈ W , then p ∈ Vn for some n. Since Vn is open, there exists a neighborhood

Nr (p) of p such that Nr (p) ⊂ Vn . By the definition of p, E ∩ Nr (p) is uncountable,

whereas by the definition of Vn , E ∩Vn is at most countable. Since E ∩Nr (p) ⊂ E ∩Vn ,

this leads to a contradiction. Hence p ∈ W c , P ⊂ W c .

Next we show W c ⊂ P . Let q ∈ W c . For every neighborhood G of q, by the definition

of base, there exists some Vm such that q ∈ Vm ⊂ G. Since q ∈ W c , the Vm that

16
 contains q satisfy E ∩ Vm is uncountable, implying that E ∩ G is uncountable, hence

q ∈ P, Wc ⊂ P.

In summary, we showed that P = W c , consequently, P c ∩ E = (W c )c ∩ E = W ∩ E is

at most countable, since W is the (at most countable) union of Vn for which E ∩ Vn

is at most countable.

It remains to show P is perfect. Since P = W c and W is union of open sets (hence

W is open), P is closed. Let’s now show W c ⊂ (W c )0 . For p ∈ W c and ε > 0, there

exists Vm such that p ∈ Vm ⊂ Nε (p). Since p ∈ W c , E ∩ Vm is uncountable. Note

that Vm is necessarily uncountable, allowing us to take some q ∈ Vm that differs from

p. Clearly, such q ∈ W c and d(p, q) < ε, proving that p ∈ (W c )0 . Combining that P

is closed and P ⊂ P 0 , it follows that P is perfect.

28. Prove that every closed set in a separable metric space is the union of a (possibly

empty) perfect set and a set which is at most countable. (Corollary: Every countable

closed set in Rk has isolated points.) Hint: Use Exercise 27.

Proof. Let E be a closed subset of a separable metric space X. If E is at most

countable, there is nothing to prove. So we assume E is uncountable. Since X is

separable, the results of Exercise 27 still applies. (in the proof of Exercise 27, the only

property of Rk is that it admits a countable base, which also holds for any separable

metric space, by Exercise 23.) Let P be the set of all condensation points of E,

decompose E as follows:

E = (E ∩ P ) ∪ (E ∩ P c ) := E1 ∪ E2 . (1)

By Exercise 27, E2 is at most countable while E1 is perfect, since E is closed and P

is perfect.

Proof of the Corollary: Let E ⊂ Rk be closed and countable. If E doesn’t have isolated

points, then every point of E is a limit point of E, then E is perfect. But perfect sets

in Rk are uncountable, contradiction. (I don’t see why it is a corollary...)

17
29. Prove that every open set in R1 is the union of an at most countable collection of

disjoint segments. Hint: Use Exercise 22.

Proof. For the first time, I found the hint is not that useful.

Let U ⊂ R1 be a nonempty open set. Define the ∼ relation on U : x ∼ y if and only

if [min(x, y), max(x, y)] ⊂ U . We shall show ∼ is an equivalence relation on U .

∼ is reflexive: If x ∈ U , then [min(x, x), max(x, x)] = {x} ⊂ U , hence x ∼ x.

∼ is symmetric: If x ∼ y, then [min(x, y), max(x, y)] ⊂ U , implying [min(y, x), max(y, x)] ⊂

U , hence y ∼ x.

∼ is transitive: If x ∼ y and y ∼ z, without losing generality, we may assume

x < y < z. Then [x, y] ⊂ U and [y, z] ⊂ U . Hence [x, z] = [x, y] ∪ [y, z] ⊂ U , i.e.,

x ∼ z.

Since ∼ is an equivalence relation on U , we can partition U into different equivalence

∼ classes. Let I be the collection of ∼-classes. We shall show the elements in I are

pairwisely disjoint segments (including unbounded segments). That the elements in

I are disjoint follows directly from the definition of equivalence class. So it remains

to show the elements in I are segments.

Suppose I ∈ I , to show I is an segment, it suffices to show I is open and connected

(Theorem 2.47). Take x ∈ I, since I ⊂ U and U is open, there exists δ > 0 such

that (x − δ, x + δ) ⊂ U . Hence for every y ∈ (x − δ, x + δ), [min(x, y), max(x, y)] ⊂

(x − δ, x + δ) ⊂ U , that is y ∼ x. Hence (x − δ, x + δ) ⊂ I, I is open. To show

I is connected, let x ∈ I, y ∈ I and x < z < y. Since x ∼ y and x < y, we have

[x, y] ⊂ U , in conjunction with x < z < y, this implies [x, z] ⊂ U and [z, y] ⊂ U , hence

z ∼ x(∼ y), therefore z ∈ I.

Finally, we should show I is at most countable. This can be verified by observing

the map f : I → Q, where f assigns I ∈ I to a rational number rI ∈ I, is injective.

S
Now we can write U = I∈I I, with I having the asserted properties. The proof is

complete.

18
30. Imitate the proof of Theorem 2.43 to obtain the following result:

S∞
If Rk = 1 Fn , where each Fn is a closed subset of Rk , then at least one Fn has a

nonempty interior.

Equivalent statement: If Gn is a dense open subset of Rk , for n = 1, 2, 3, . . ., then

T∞ k
1 Gn is not empty (in fact, it is dense in R ).

(This is a special case of Baire’s theorem; see Exercise 22, Chap. 3, for the general

case.)

Proof. We first show the two statements in the problem are indeed equivalent. Denote

the first statement by S1 , and the second statement by S2 .

T∞ S∞
S1 ⇒ S2 : If 1 Gn = ∅, then 1 Gcn = Rk . Since every Gn is open, Gcn is closed,

hence by S1 , at least one Gcn , say Gci , has nonempty interior. That means,

∃x ∈ Gci and r > 0 such that Nr (x) ⊂ Gci , contradicts with Gi is dense in Rk .

S2 ⇒ S1 : If for every n, Fn◦ = ∅, then Fnc is open and dense in Rk (recall the
T∞
identity (E ◦ )c = E c , by Exercise 9(d)). Hence by S2 , 1 Fnc 6= ∅. Therefore,
S∞ k
1 Fn $ R , contradiction.

Having shown S1 and S2 are equivalent, we will prove S2 only. Let ε be any given

positive, and take x0 ∈ Rk . Since G1 is dense in Rk , there exists x1 ∈ G1 such that

|x1 − x0 | < ε/2. Since G1 is open, there exists r1 > 0 such that Nr1 (x1 ) ⊂ G1 ,

note we may adjust r1 so that for all x ∈ Nr1 (x1 ) such that for all x ∈ Nr1 (x1 ),

|x − x0 | ≤ |x − x1 | + |x1 − x0 | < r1 + δ/2 < δ. Denote Nr1 /2 (x1 ) by V1 so that

V 1 ⊂ G1 . Now since G2 is dense, there exists x2 ∈ G2 such that |x2 − x1 | < r1 /2,

in addition, since G2 is open, we may find a neighborhood of x2 with radius r2 such

that V 2 ⊂ V 1 and V2 ⊂ G2 . Proceeding we can get a sequence of neighborhoods

V1 ⊃ V2 ⊃ V3 ⊃ · · · satisfying: (i) V n ⊃ V n+1 for every n ≥ 1. (ii) V n ⊂ Gn for every

n ≥ 1. (iii) Each V n is nonempty.

T∞
By construction, each V n is compact, hence by Corollary of Theorem 2.36, 1 V n 6=

19
 T∞ T∞ T∞
∅. By (ii), this implies 1 Gn 6= ∅. Let p ∈ 1 V n ⊂ 1 Gn , since p ∈ V1 , we have
T∞
d(p, x0 ) ≤ d(p, x1 ) + d(x1 , x0 ) < ε, showing that 1 Gn is also dense in Rk .

20