Homework 5

Exercise 5.1. Let A and B denote subsets of a topological space X = (X , T). Prove the following:
(i) If A ⊂ B, then cl(A) ⊂ cl(B).
(ii) cl(A ∪ B) = cl(A) ∪ cl(B).
Solution.
(i) Note that cl(B) is a closed set and A ⊂ B ⊂ cl(B). But cl(A) is the intersection of all closed sets
containing A and cl(B) being one of them we conclude cl(A) ⊂ cl(B).
(ii) Since finite unions of closed sets remain closed, the set cl(A) ∪ cl(B) is closed in X . Furthermore
A ⊂ cl(A) ⊂ cl(A) ∪ cl(B) and B ⊂ cl(B) ⊂ cl(A) ∪ cl(B), so A ∪ B ⊂ cl(A) ∪ cl(B). Again, cl(A ∪ B)
is the intersection of all closed sets containing A ∪ B and cl(A) ∪ cl(B) is such a set, so we conclude
cl(A ∪ B) ⊂ cl(A) ∪ cl(B).
On the other hand, from part (i) and A ⊂ A ∪ B we can conclude cl(A) ⊂ cl(A ∪ B) and similarly
cl(B) ⊂ cl(A ∪ B). Therefore, cl(A) ∪ cl(B) ⊂ cl(A ∪ B) and combining this with the previous paragraph
we conclude that cl(A ∪ B) = cl(A) ∪ cl(B).

Exercise 5.2.
(i) Show that in a metric space (X , d ), the closure of an open ball B(x, r ) is contained in the closed ball
B(x, r ).
(ii) Give an example (with proof) where B(x, r ) is different from the closure cl(B(x, r )).
Solution.
(i) First, observe that the closed ball B(x, r ) is a closed set: Consider the function f : X R with
f (y) = d (x, y). This function is continuous1 and B(x, r ) = f −1 ([0, r ]) is the preimage of a closed set in
R.
Now, B(x, r ) is a closed set containing B(x, r ) and cl(B(x, r )) is the intersection of all such sets. So
cl(B(x, r )) ⊂ B(x, r ).
(ii) Consider any set X containing at least 2 elements and equip it with the discrete metric. Let x ∈ X . Then
B(x, 1) = {x } is closed in X and therefore cl(B(x, 1)) = B(x, 1) = {x }. But B(x, 1) = X , {x } because
we assumed X contains at least 2 elements.

Exercise 5.3. If A is a subset of a topological space X , define the boundary of A to be the set

∂A = cl(A) r int(A).

That is, the boundary of A is the difference between the closure of A and the interior of A. Prove that
(i) The boundary ∂A is closed for any set A ⊂ X .
(ii) A ∪ ∂A = cl(A) for any A.
(iii) A r ∂A = int(A) for any A.
Solution.
(i) Note that ∂A = cl(A) ∩ (X r int(A)) and cl(A) is closed in X . Also, int(A), being a union of open sets,
is open and therefore X r int(A) is closed in X . Hence, ∂A is an intersection of two closed sets, and
therefore closed itself.
(ii) First, by definition A ⊂ cl(A) and ∂A ⊂ cl(A) and therefore A ∪ ∂A ⊂ cl(A). On the other hand,
because int(A) ⊂ A, there is an inclusion cl(A) r A ⊂ cl(A) r int(A) and because A ⊂ cl(A) we have
cl(A) = (cl(A) r A) ∪ A ⊂ (cl(A) r int(A)) ∪ A = A ∪ ∂A.
(iii) Writing C c = X rC, we have Ar ∂A = A∩ (cl(A) ∩int(A) c ) c = (A∩cl(A) c ) ∪ (A∩int(A)). But A ⊂ cl(A)
and int(A) ⊂ A, so A ∩ cl(A) c = ∅ and A ∩ int(A) = int(A). Therefore, A r ∂A = ∅ ∪ int(A) = int(A).
1 For every ε > 0 if d (y, y 0 ) < ε, then f (y 0 ) − f (y) = d (x, y 0 ) − d (x, y) ≤ d (y, y 0 ) < ε and f (y) − f (y 0 ) = d (x, y) − d (x, y 0 ) ≤
d (y, y 0 ) < ε; i. e. | f (y 0 ) − f (y)| < ε.

1
Exercise 5.4. Let A = (Q ∩ (0, 1)) ∪ {2} ∪ (3, 5], thought of as a subset of R with its standard topology.
Compute with proof the sets cl(A), int(A) and ∂A.
Solution. From problem 1 we know that

cl(A) = cl((Q ∩ (0, 1)) ∪ {2} ∪ (3, 5]) = cl(Q ∩ (0, 1)) ∪ cl({2}) ∪ cl((3, 5]).

Because {2} is closed already, we immediately see cl({2}) = {2}. We will argue that cl(Q ∩ (0, 1)) = [0, 1] and
cl((3, 5]) = [3, 5] so that cl(A) = [0, 1] ∪ {2} ∪ [3, 5]. First, [0, 1] and [3, 5] are closed sets containing Q ∩ (0, 1)
and (3, 5] respectively, so cl(Q ∩ (0, 1)) ⊂ [0, 1] and cl((3, 5]) ⊂ [3, 5].
Conversely, suppose first that x ∈ (0, 1) and let U ⊂ R be an open set containing x. Then (0, 1) ∩ U is open,
so there is an ε > 0 such that (x − ε, x + ε ) ⊂ (0, 1) ∩ U . But every open interval in R contains infinitely
many rational numbers. Hence, (x − ε, x + ε) ∩ Q , ∅ and we conclude that in fact U ∩ Q ∩ (0, 1) , ∅. This
implies that x ∈ cl(Q ∩ (0, 1)). Now, { n+2 1
}n ∈N is a sequence in Q ∩ (0, 1) and limn n+2 1
= 0 in R. Hence we
must have 0 ∈ cl(Q ∩ (0, 1)). Similarly, limn 1 − n+2 1
= 1 and therefore 1 ∈ cl(Q ∩ (0, 1)). In conclusion,
cl(Q ∩ (0, 1)) = [0, 1] as claimed.
As for (3, 5], we already know (3, 5] ⊂ cl((3, 5]), so we only need to prove 3 ∈ cl((3, 5]). For this, just
observe that limn 3 + n1 = 3 and 3 + n1 ∈ (3, 5] for all n ∈ N.
To compute the interior of A, let U be any open set with U ⊂ A. Suppose for contradition that some
x ∈ Q ∩ (0, 1) were contained in U . Then there would be some ε > 0 with (x − ε, x + ε ) ⊂ U . But then
(x − ε, x + ε) ∩ (0, 1) ⊂ U as well and the latter would have to contain infinitely many irrational numbers. If
y ∈ (x − ε, x + ε) ∩ (0, 1) is irrational, then y < A contradicting our assumption that U ⊂ A. We conclude
that U ∩ Q ∩ (0, 1) = ∅. Furthermore, we also have 2 < U since otherwise there would again be some
ε > 0 with (2 − ε, 2 + ε ) ⊂ U . But (2 − ε, 2 + ε) contains infinitely many points outside of A, for example
2 + n1 < A for n > 1/ε. Similarly, 5 ∈ U is impossible because otherwise there would again be some ε > 0
with (5 − ε, 5 + ε) ⊂ A. But 5 + n1 < A for n > 1/ε.
In summary, any open set U contained in A satisfies the stronger inclusion U ⊂ (3, 5). But (3, 5) is an open
set contained in A, so we conclude that in fact (3, 5) = int(A).
Now, the boundary of A is easily computed as

∂A = cl(A) r int(A) = ([0, 1] ∪ {2} ∪ [3, 5]) r (3, 5) = [0, 1] ∪ {2, 3, 5}.

Exercise 5.5. Consider Y = Q, endowed with the subspace topology for the inclusion Q ⊂ R (where R carries
its standard topology). Let A = {p ∈ Q : 2 < p 2 < 3} ⊂ Q ⊂ R.
First, we note that A is an open subset of Q. Indeed, A = U ∩ Q, where U = {p ∈ R : 2 < p 2 < 3} is an open
subset of R. Therefore, by definition, since A is the intersection of an open subset in R with Q, A is open in
the subspace topology of Q.
(i) Prove, on the other hand, that A is not open in R.
(ii) What is the closure of A in R (denoted clR (A))?
(iii) What is the closure of A in Q (denoted clQ (A))?
Solution.
(i) Suppose for contradiction that A were open in R and pick x ∈ A. Then there would be some ε > 0 such
that (x − ε, x + ε) ⊂ A ⊂ Q. But any open interval in R contains infinitely many irrational points, so
this is impossible. √ √ √ √
(ii) We first note that A = A+ ∪ A− with A+ = Q ∩ ( 2, 3) and A− = Q ∩ (−√ 3, √ − 2). Therefore,
clR (A) = clR (A+ ) ∪ clR (A− ). To compute √ √ + ), first suppose that x ∈ ( 2, 3) and let V be
clR (A
some open neigborhood
√ √ of x in R. Then ( 2, 3) ∩ V is open, so that there is some ε > 0 with
(x − ε, x + ε) ⊂ ( 2, 3) ∩V . Because√any √ open interval in R contains infinitely
√ many
√ rational numbers
we
√ conclude that A + ∩ V = Q ∩ ( 2, 3) ∩
√ √V , ∅. We conclude that
√ ( 2, 3) ⊂ clR√(A+ ). Now,
{ 2 + n }n ∈N is a sequence (eventually) in ( 2, 3) that converges to 2 and similarly { 3 − n1 }n ∈N
1

2
 √ √ √ √ √ √ √
is a sequence (eventually)
√ √ in ( 2, 3) converging to 3. Therefore,
√ √ 2, 3 ∈ clR ((
√ √ 2, 3)) and we
conclude that [ 2, 3] ⊂ clR√(A+√). On the other hand,√ √A + ⊂ [ 2, 3] and because [ 2, 3] is closed in
R we also have clR (A+ ) ⊂ [ 2, 3]. So, clR (A+ ) = [ 2, 3].√ √
An entirely analogous argument
√ √ shows √ clR (A− ) = [− 3, − 2]. Therefore, we can conclude in
√ that
summary that clR (A) = [− 3, − 2] ∪ [ 2, 3].
(iii) Quite generally, we have clQ (A) = clR (A) ∩ Q. Therefore,
√ √ √ √
clQ (A) = Q ∩ ([− 3, − 2] ∪ [ 2, 3]) = {p ∈ Q : 2 ≤ p 2 ≤ 3} = {p ∈ Q : 2 < p 2 < 3} = A
√ √
since ± 2, ± 3 < Q.