Compactness

Definition 1. A cover or a covering of a topological space X is a family C of

subsets of X whose union is X. A subcover of a cover C is a subfamily of C
which is a cover of X. An open cover of X is a cover consisting of open sets.

Definition 2. A topological space X is said to be compact if every open cover

of X has a finite subcover. A subset S of X is said to be compact if S is
compact with respect to the subspace topology.

Theorem 3. A subset S of a topological space X is compact if and only if

every open cover of S by open sets in X has a finite subcover.

Proof. (⇒) Assume that S is a compact subset of a topology space X. Let

{Gα | α ∈ Λ} be a collection of open subset of X such that S ⊆ ∪α∈Λ Gα . Let
Oα = Gα ∩ S for each α ∈ Λ. Then Oα is open in S for each α ∈ Λ and

∪α∈Λ Oα = ∪α∈Λ (Gα ∩ S) = (∪α∈Λ Gα ) ∩ S = S.

Thus {Oα | α ∈ Λ} is an open cover of S. It follows that {Oα | α ∈ Λ}

contains a finite subcover {Oα1 , Oα2 , . . . , Oαn }. Hence

S = ∪ni=1 Oαi = ∪ni=1 (Oαi ∩ S) = (∪ni=1 Gαi ) ∩ S.

Thus S ⊆ ∪ni=1 Gαi .

(⇐) Let {Oα | α ∈ Λ} be an open cover of (S, τs ). Then, for each α ∈ Λ,
there exists an open set Gα in X such that Oα = Gα ∩ S. Thus

S = ∪α∈Λ Oα = ∪α∈Λ (Gα ∩ S) = (∪α∈Λ Gα ) ∩ S.

Thus S ⊆ ∪α∈Λ Gα . Then there is a finite subset {Gα1 , Gα2 , . . . , Gαn } such
that S ⊆ ∪ni=1 Gαi . It follows that

S = (∪ni=1 Gαi ) ∩ S = ∪ni=1 (Gαi ∩ S) = ∪ni=1 Oαi .

Hence (S, τs ) is compact.

Examples.
1. Any finite set is compact. In general, (X, τ ), where τ is finite, is compact.
In particular, an indiscrete space is compact.
2. Any infinite discrete space is not compact. In fact, if X is an infinite
discrete space, then {{x} | x ∈ X} is an open cover of X which has no finite
subcover.
3. R is not compact. The class {(−n, n) | n ∈ N} is an open cover of R
which contains no finite subcover.

1
Theorem 4. A closed subset of a compact space is compact.

Proof. Let (X, τ ) be a compact space and F a closed subset of X. Let C =

{Gα | α ∈ Λ} be a family of open subsets of X such that F ⊆ ∪α∈Λ Gα . Then

X = F ∪ F c = (∪α∈Λ Gα ) ∪ F c .

Thus C ∪ {F c } is an open cover of X and hence it has a finite subcover

{Gα1 , Gα2 , . . . , Gαn , F c }, i.e. X = (∪ni=1 Gαi ) ∪ F c . Since F ⊆ X, it follows
that F ⊆ ∪ni=1 Gαi . Thus {Gα1 , Gα2 , . . . , Gαn } is a finite subfamily of C such
that F ⊆ ∪ni=1 Gαi . Hence, F is compact.

Theorem 5. If A is a compact subset of a Hausdorff space X and x ∈

/ A,
then x and A have disjoint neighborhoods.

Proof. By the Hausdorffness of X, for each y ∈ A, there are open neighbor-

hoods Uy and Vy of x and y, respectively, such that Uy ∩ Vy = ∅. Then
{Vy | y ∈ A} is an open cover for A, which is compact. Hence there are y1 , y2 ,
. . . , yn ∈ A such that A ⊆ ∪ni=1 Vyi . Let V = ∪ni=1 Vyi and U = ∩ni=1 Uyi . Then
U and V are neighborhoods of x and A, respectively, and U ∩ V = ∅.

Theorem 6. Any compact subset of a Hausdorff space is closed.

Proof. Let A be a compact subset of a Hausdorff space X. To show that

Ac is open, let x ∈ Ac . By the previous theorem, there are neighborhoods
U and V of x and A, respectively, such that U ∩ V = ∅. It follows that
x ∈ U ⊆ V c ⊆ Ac . This shows that Ac is a neighborhood of x. Hence, Ac is
a neighborhood of all its elements. It follows that Ac is open and that A is
closed.

Theorem 7. A continuous image of a compact space is compact.

Proof. Let f : X → Y be a continuous function from a compact space X into

a space Y . By Theorem 3, it is sufficient to assume that f maps X onto Y
and we show that Y is compact. To see this, let {Gα | α ∈ Λ} be an open
cover for Y . By continuity of f , it follows that f −1 [Gα ] is open in X for each
α and
∪α∈Λ f −1 [Gα ] = f −1 [∪α∈Λ Gα ] = f −1 [Y ] = X.
Hence, {f −1 [Gα ] | α ∈ Λ} is an open cover for X. By the compactness of X,
there are α1 , α2 , . . . , αn ∈ Λ such that ∪ni=1 f −1 [Gαi ] = X. Thus

Y = f [X] = f [∪ni=1 f −1 [Gαi ]] = f [f −1 [∪ni=1 Gαi ]] ⊆ ∪ni=1 Gαi .

This shows that Y is compact.

2
Corollary 8. Let f : X → Y is a bijective continuous function. If X is
compact and Y is Hausdorff, then f is a homeomorphism.

Proof. Let f : X → Y be a bijective continuous function. Assume that X is

compact and Y is Hausdorff. To show that f is a homeomorphism, it suffices
to show that f is a closed map. Let F be a closed subset of X. Then F
is compact by Theorem 4. By Theorem 7, f [F ] is compact in the Hausdorff
space Y . Hence, f [F ] is closed in Y by Theorem 6.

Theorem 9. A continuous function of a compact metric space into a metric

space is uniformly continuous.

Proof. Let f : (X, d) → (Y, ρ) be a continuous function between metric spaces.

Assume that X is compact. To show that f is uniformly continuous, let ε > 0.
By continuity of f , for each x ∈ X, there is a δx > 0 such that
ε
d(y, x) < δx =⇒ ρ(f (y), f (x)) < .
2
Then {Bd (x, δ2x ) | x ∈ X} is an open cover for a compact space X. Hence,
δ
we can choose x1 , x2 , . . . , xn ∈ X such that X = ∪ni=1 Bd (xi , 2xi ). Choose
δ = 21 min{dx1 , dx2 , . . . , dxn }. Now, let x, y ∈ X be such that d(x, y) < δ.
δ
Then x ∈ Bd (xi , 2xi ) for some i. Hence, ρ(f (x), f (xi )) < 2ε . Moreover,

δxi
d(y, xi ) ≤ d(y, x) + d(x, xi ) ≤ δ + ≤ δxi .
2
It follows that ρ(f (y), f (xi )) < 2ε . By triangle inequality,

ρ(f (x), f (y)) ≤ ρ(f (x), f (xi )) + ρ(f (xi ), f (y)) < ε.

This shows that f is uniformly continuous.

Next we introduce a property that measures “boundedness” of a subset of

a metric space. We say that a non-empty set A is bounded if diam(A) < ∞,
i.e., there is a constant M > 0 such that d(x, y) ≤ M for any x, y ∈ A. This
definition of boundedness depends on the metric rather than the set itself. For
example, R with the discrete metric is bounded even though in some sense R
is “large”. The following definition of total boundedness captures this spirit.

Definition 10. A metric space (X, d) is said to be totally bounded or precom-

pact if for any ε > 0, there is a finite cover of X by sets of diameter less than
ε.

Theorem 11. A metric space (X, d) is totally bounded if and only if for each
ε > 0, X can be covered by finitely many ε-balls.

3
Proof. (⇒) Let ε > 0. Then there exist subsets A1 , A1 , . . . , An of X such that
diam Ai < ε for all i ∈ {1, 2, . . . , n} and ∪ni=1 Ai = X. We may assume that
each Ai is non-empty and choose xi ∈ Ai . If x ∈ X, then x ∈ Ai for some i
and hence, d(x, xi ) ≤ diam(Ai ) < ε. This show that X = ∪ni=1 Bd (xi , ε).
(⇐) Let ε > 0. Then there is a finite subset {x1 , x2 , . . . , xn } of X such that
X = ∪ni=1 Bd (xi , 4ε ). Let Ai = Bd (xi , 4ε ) for each i ∈ {1, 2, . . . , n}. For each
a, b ∈ Ai , d(a, b) ≤ d(a, xi ) + d(xi , b) < 4ε + 4ε = 2ε . Hence, diam(Ai ) < ε.

Theorem 12. A subspace of a totally bounded metric space is totally

bounded.

Proof. Let (X, d) be a totally bounded metric space and Y a subspace of X.

Let ε > 0. Then there exist A1 , A2 ,. . . , An ⊆ X such that diam Ai < ε for all
i ∈ {1, 2, . . . , n} and ∪ni=1 Ai = X. Then diam(Ai ∩ Y ) ≤ diam Ai < ε for each
i ∈ {1, 2, . . . , n} and ∪ni=1 (Ai ∩ Y ) = (∪ni=1 Ai ) ∩ Y = X ∩ Y = Y.

Theorem 13. Every totally bounded subset of a metric space is bounded.

Proof. Let (X, d) be a totally bounded metric space. Then there exists a finite
subset {x1 , x2 , . . . , xn } of X such that ∪ni=1 Bd (xi , 1) = X. Let

K = max{d(xi , xj ) | i, j ∈ {1, 2, . . . , n}} + 2.

Let x, y ∈ X. Then x ∈ Bd (xi , 1) and y ∈ Bd (xj , 1) for some i, j. Thus

d(x, y) ≤ d(x, xi ) + d(xi , xj ) + d(xj , y) < 1 + d(xi , xj ) + 1 ≤ K,

i.e. diam X ≤ K, so X is bounded.

In general, the converse is not true. The space R with the discrete metric
is bounded because d(x, y) ≤ 1 for all x, y ∈ R, but it cannot be covered by
a finitely many balls of radius 12 . However, it is true for Rn with the usual
metric.

Theorem 14. A bounded subset of Rn is totally bounded.

Proof. We will prove this theorem for the case n = 1. The proof for the case
n > 1 is similar but slightly more complicated.
By Theorem 12, it suffices to prove that a closed interval [a, b] is totally
bounded. Let ε > 0. Choose an n ∈ N such that (b − a)/n < ε. For i =
0, 1, . . . , n, let
b−a
xi = a + i .
n
Then [a, b] = ∪ni=1 [xi−1 , xi ] and diam([xi−1 , xi ]) = xi − xi−1 < ε.

4
Theorem 15. A metric space is totally bounded if and only if every sequence
in it has a Cauchy subsequence.

Proof. (⇐) Assume that (X, d) is not tally bounded. Then there is an ε > 0
such that X cannot be covered by finitely many balls of radius ε. Let x1 ∈ X.
Then Bd (x1 , ε) 6= X, so we can choose x2 ∈ X − Bd (x1 , ε). In general, for each
n ∈ N, we can choose xn+1 ∈ X −∪ni=1 Bd (xi , ε). If m > n, then xm ∈ / Bd (xn , ε),
and thus d(xm , xn ) ≥ ε. It is easy to see that (xn ) has no Cauchy subsequence.
(⇒) Assume that (X, d) is totally bounded and let (xn ) be a sequence in
(X, d). Set B0 = X. There exist A11 , A12 , . . . , A1n1 ⊆ X such that
n1
[
diam A1i < 1 for all i ∈ {1, 2, . . . , n1 } and A1i = X.
i=1

At least one of these A1i ’s, called B1 , must contain infinitely many terms of
(xn ). Let (x11 , x12 , . . . ) be a subsequence of (xn ) which lies entirely in B1 .
Since B1 ⊆ X, it is totally bounded. There are A21 , . . . , A2n2 ⊆ B1 such that
n2
1 [
diam A2i < for all i ∈ {1, 2, . . . , n2 } and A2i = B1 .
2 i=1

At least one of these A2i ’s, called B2 , must contain infinitely many terms of
(x1n ). Then diam B2 < 21 and we can choose a subsequence (x21 , x22 , x23 , . . . ) of
(x11 , x12 , x13 , . . . ) in B2 . We can continue this process. To make this argument
more precise, we will give an inductive construction.
Assume that we can choose Bi ⊆ Bi−1 with diam Bi < 1i and a subsequence
(xi1 , xi2 , xi3 , . . . ) of (x(i−1)1 , x(i−1)2 , x(i−1)3 , . . . ) in Bi for all i ∈ {1, 2, . . . , k−1}.
Since Bk−1 ⊆ Bk−2 ⊆ . . . B1 ⊆ B0 = X, Bk−1 is totally bounded. Then there
exist Ak1 , Ak2 , . . . , Aknk ⊆ Bk−1 such that
nk
1 [
diam Akj < for all j ∈ {1, 2, . . . , nk } and Aki = Bk−1 .
k i=1

At least one of these Aki ’s, called Bk , must contain infinitely many terms of
(xk−1,n ). Hence Bk ⊆ Bk−1 , diam Bk < k1 and we can choose a subsequence
(xk1 , xk2 , xk3 , . . . ) of (x(k−1)1 , x(k−1)2 , x(k−1)3 , . . . ) which lies entirely in Bk .
Now we choose the diagonal elements (x11 , x22 , x33 , . . . ) from the above
subsequences. This is to guarantee that the index of the chosen subsequence
is strictly increasing. To see that it is a Cauchy subsequence of (xn ), let ε > 0.
Choose an N ∈ N such that N1 < ε. Let m, n ∈ N be such that m, n ≥ N .
Then xmm ∈ Bm ⊆ BN and xnn ∈ Bn ⊆ BN . Thus
1
d(xmm , xnn ) ≤ diam BN < < ε.
N
Hence (xnn ) is a Cauchy subsequence of (xn ).

5
Definition 16. A space X is said to be sequentially compact if every sequence
in X has a convergent subsequence.

Theorem 17. A metric space X is sequentially compact if and only if it is

complete and totally bounded.

Proof. (⇒) Let (X, d) be a sequentially compact space. By Theorem 15, it is

totally bounded because a convergent sequence is a Cauchy sequence. To see
that it is complete, let (xn ) be a Cauchy sequence in X. Since X is sequentially
compact, it has a convergent subsequence. Hence, (xn ) is convergent. It follows
that X is complete.
(⇐) Assume that (X, d) is totally bounded and complete. To see that X
is sequentially compact, let (xn ) be a sequence in X. By Theorem 15, it has
a Cauchy subsequence (xnk ). Since (X, d) is complete, (xnk ) is convergent.
Hence, (xn ) has a convergent subsequence. This shows that X is sequentially
compact.

Definition 18. Let C be a cover of a metric space X. A Lebesgue number for

C is a positive number λ such that any subset of X of diameter less than or
equal to λ is contained in some member of C.

Remark. If λ is a Lebesgue number, then so is any λ0 > 0 such that λ0 ≤ λ.

Example. Let X = (0, 1) ⊆ R and C = {( n1 , 1) | n ≥ 2}. Then C is an

open cover for X, but it has no Lebesgue number. To see this, let λ > 0.
Choose a positive integer n such that n1 < λ. Let A = (0, n1 ) ⊆ X. Then
diam(A) = n1 < λ, but A is not contained in any member of C. Hence C has
no Lebesgue number.

Theorem 19. Every open cover of a sequentially compact metric space has a
Lebesgue number.

Proof. Let C be an open cover of a sequentially compact metric space (X, d).
Suppose that C does not have a Lebesgue number. Then for each n ∈ N there
exists a subset Bn of X such that
1
diam(Bn ) ≤ and Bn 6⊆ G for all G ∈ C.
n
For each n ∈ N, choose xn ∈ Bn . Since X is sequentially compact, the sequence
(xn ) contains a convergent subsequence (xnk ). Let x ∈ X be the limit of this
subsequence. Then x ∈ G0 for some G0 ∈ C. Since G0 is open, there exists
ε > 0 such that Bd (x, ε) ⊆ G0 . Since (xnk ) converges to x, there exists an
integer N such that
ε
d(xnk , x) < for any k ≥ N .
2

6
Choose M ∈ N such that M1 < 2ε . Let K = max{M, N }. Then nK ≥ K ≥ M .
Hence,
ε
d(xnK , x) < and xnK ∈ BnK .
2
Moreover,
1 ε
diam(BnK ) ≤ < .
nK 2
For any y ∈ BnK , we have
ε ε
d(x, y) ≤ d(x, xnK ) + d(xnK , y) < + = ε.
2 2
Hence y ∈ Bd (x, ε). Then BnK ⊆ Bd (x, ε) ⊆ G0 . This contradicts the fact
that BnK 6⊆ G for all G ∈ C.

Definition 20. A space X is said to satisfy the Bolzano-Weierstrass property

if every infinite subset has an accumulation point in X.

Theorem 21. In a metric space X, the following statements are equivalent:

(a) X is compact;

(b) X has the Bolzano-Weierstrass property;

(c) X is sequentially compact;

(d) X is complete and totally bounded.

Proof. We have already proved (c) ⇔ (d) in Theorem 17.

(a) ⇒ (b). Let (X, d) be a compact metric space and S an infinite subset
of X. Suppose that S has no accumulation point. Then for each x ∈ X
there exists an open neighborhood Vx such that Vx ∩ (S − {x}) = ∅. Hence,
C = {Vx | x ∈ X} is an open cover of X. Since X is compact, there exists a
finite subcover {Vx1 , Vx2 , . . . , Vxn } of C. Since each ball contains at most one
point of S, X = ∪ni=1 Vxi contains finitely many points of S. Hence, S is finite,
contrary to the hypothesis.
(b) ⇒ (c). Assume that X has the Bolzano-Weierstrass property. To show
that X is sequentially compact, let (xn ) be a sequence in X.
Case I. The set S = {xn | n ∈ N} is finite. Then there is a ∈ X such that
xn = a for infinitely many n’s. Choose n1 = min{n ∈ N | xn = a} and for any
k ≥ 2, let
nk = min({n ∈ N | xn = a} − {n1 , n2 , . . . , nk−1 }).
Then (xnk ) is a constant subsequence of (xn ) and hence is convergent.

7
Case II. The set S = {xn | n ∈ N} is infinite. By the assumption, S has an
accumulation point x (in X). For each n ∈ N, Bd (x, n1 ) ∩ (S − {x}) 6= ∅; in
fact, Bd (x, n1 ) ∩ (S − {x}) is an infinite set. Let

n1 = min{n ∈ N | xn ∈ Bd (x, 1) ∩ (S − {x})} and

nk = min({n ∈ N | xn ∈ Bd (x, k1 ) ∩ (S − {x})} − {n1 , . . . , nk−1 }),

for any k ≥ 2. Then (xnk ) is a subsequence of (xn ) such that d(xnk , x) < k1 for
each k ∈ N. Hence (xnk ) converges to x.
(c) ⇒ (a). Assume that (X, d) is a sequentially compact metric space. Let C
be an open cover for X. Hence C has a Lebesgue number λ > 0. Moreover, X is
totally bounded. Thus there exist A1 , A2 , . . . , An ⊆ X such that X = ∪ni=1 Ai
and diam(Ai ) ≤ λ for each i. Hence for each i ∈ {1, 2, . . . , n}, there exists
Gi ∈ C such that Ai ⊆ Gi . Thus X = ∪ni=1 Gi . This shows that C has a finite
subcover. Hence X is compact.

Theorem 22 (Heine-Borel). A subset of Rn is compact if and only if it is

closed and bounded.

Proof. By Theorem 21, a metric space is compact if and only if it is complete

and totally bounded. Hence, if a subset of Rn is compact, then it is closed
and bounded. (In fact, we can prove directly that compactness implies a set
being closed and bounded without using Theorem 21.) Conversely, let A be
a closed and bounded subset of Rn . By Theorem 14, A is totally bounded.
Since A is closed subset of Rn , which is complete, A is also complete. Hence,
A is compact.

Corollary 23 (Extreme Value Theorem). A real-valued continuous func-

tion on a compact space has a maximum and a minimum.

Proof. Assume that f : X → R is a continuous function and X is compact.

By Theorem 7, f [X] is a compact subset of R. Hence, f [X] is closed and
bounded. Let a = inf f [X] and b = sup f [X]. Since f [X] is closed, a and b are
in f [X]. Thus a and b are maximum and minimum of f [X], respectively.

Theorem 24 (Bolzano-Weierstrass). Every bounded infinite subset of Rn

has at least one accumulation point (in Rn ).

Proof. Let A be a bounded infinite subset of Rn . Since A is bounded, it is

contained in some closed cube In = [−n, n] × · · · × [−n, n]. Since In is closed
and bounded, it is compact by Heine-Borel theorem. Since A is an infinite
subset of a compact set In , it must have an accumulation point (in Rn ) by
Theorem 21.

8
Theorem 25. Every bounded sequence in Rn has a convergent subsequence.

Proof. Let (xn ) be a bounded sequence in Rn . Hence it is contained in some

closed cube In = [−n, n] × · · · × [−n, n]. Since In is closed and bounded, it
is compact by Heine-Borel theorem. Then In is sequentially compact. This
implies that (xn ) has a convergent subsequence.