II.4.

Compactness 1

II.4. Compactness

Note. Conway states on page 20 that the concept of compactness is an extension

of benefits of finiteness to infinite sets. I often state this idea as: “Compact sets
allow us to transition from the infinite to the finite.”

Definition II.4.1. A subset K of a metric space X is compact if for every collection

G of open sets in X with the property K ⊂ ∪G∈G G, there is a finite number of sets
G1 , G2 , . . . , Gn in G such that K ⊂ ∪nk=1Gk . The collection G is called an open
cover of K.

Note. We can paraphrase the definition of compact as: “Every open cover has a
finite subcover.” You may recall from senior level analysis that a set of real numbers
is compact if and only if it is closed and bounded (the Heine-Borel Theorem). The
Heine-Borel Theorem holds in C (and Rn and Cn ), but it does not hold in all metric
spaces. BEWARE! The important property of a compact set is that every open
cover has a finite subcover, not something else (like “closed and bounded” which,
in general, is not true for compact sets). In fact, let’s violate Heine-Borel in `2!

Example. In `2, consider the set

H = {(1, 0, 0, . . .), (0, 1, 0, 0, . . .), (0, 0, 1, 0, 0, . . .), . . .}.

√
Then the distance between any two distinct points in H is 2. So all points of
H are isolated and hence H is closed (if you like, consider the complement of H).
II.4. Compactness 2

Since each element of H is distance 1 from the “origin” (0, 0, 0, . . .), then H is
bounded. However, H is not compact! Consider the open cover G = {B(x; 1/2) |
√
x ∈ H}. Since the distance between any two distinct points of H is 2, then
B(x; 1/2) ∩ B(y; 1/2) = ∅ for x 6= y. So no set in G can be eliminated from G and
the result still cover H. Therefore there is no finite subcover of G, and H is not
compact. The reason this example works in `2 (and not in Rn or Cn ) is because
of the infinite number of “directions” (i.e., axes) in `2. Consider the first three
elements of H and G:

Imagine this example extended along infinitely many axes and you will get the idea
of how the construction works in `2.

Note. In fact, in a normed linear space, the closed unit ball is compact if and only
if the dimension of the normed linear space is finite. See Theorem 2.34 of A First
Course in Functional Analysis, by S. David Promislow, John Wiley & Sons Publi-
cations (2008). Also see my class notes at http://faculty.etsu.edu/gardnerr/
Func/notes/2-8.pdf.
II.4. Compactness 3

Note. Set H is also an example of an infinite bounded set with no limit point.
Recall that Weierstrass’ Theorem says that an infinite bounded set of real numbers
(or elements of Rn or Cn ) has a limit point.

Proposition II.4.3. Let K be a compact subset of X. Then

(a) K is closed, and

(b) if F is closed and F ⊂ K then F is compact.

Note. In fact, we can show that, in a metric space, a compact set is closed AND
bounded (by “bounded” we mean that a set A satisfies A ⊂ B(x; K) for some x ∈ X
and some K ∈ R). If A is not bounded then the open cover {B(x; N ) | N ∈ N}
of A (where x is some element of X) has no finite subcover. That is, A is not
compact. Hence, if A is compact then it is bounded.

Definition. If F is a collection of subsets of X such that whenever {F1 , F2 , . . . , Fn } ⊂

F we have F1 ∩ F2 ∩ · · · ∩ Fn 6= ∅, then F has the finite intersection property (de-
noted “f.i.p.”).

Proposition II.4.4. A set K ⊂ X is compact if and only if every collection F of

closed subsets of K with the finite intersection property satisfies ∩F ∈F F 6= ∅.
II.4. Compactness 4

Note. The following two results are useful corollaries which follow from the finite
intersection property result and they illustrate the use of Proposition 4.4

Corollary II.4.5. Every compact metric space is complete.

Note. The following may remind you of Weierstrass’ Theorem.

Corollary II.4.6. If X is a compact set in a metric space, then every infinite set
has a limit point in X.

Note. The following definition makes use of sequences and defines a sequentially
compact metric space. We will see that this definition is equivalent to the standard
compactness.

Definition II.4.7. A metric space (X, d) is sequentially compact if every sequence

in X has a convergent subsequence.

Lemma II.4.8. Lebesgue’s Covering Lemma.

If (X, d) is sequentially compact and G is an open cover of X then there is an ε > 0
such that if x ∈ X, there is a set G ∈ G with B(x; ε) ⊂ G.

Note. We now give special conditions equivalent to the compactness of X.

II.4. Compactness 5

Proposition II.4.9. Let (X, d) be a metric space. The following are equivalent:

(a) X is compact,

(b) every infinite subset of X has a limit point,

(c) X is sequentially compact, and

(d) X is complete and for all ε > 0 there are a finite number of points x1 , x2 , . . . , xn ∈
X such that X = ∪nk=1B(xk ; ε). This property is called total boundedness.

Note. The following proof of the Heine-Borel Theorem may be a little different
from the proof given in your senior level analysis class. This is because we have
developed a lot of “heavy equipment” concerning compactness in this section (in
particular, the total boundedness of Theorem 4.9). First, we need a technical
lemma.

Lemma. Let F = [a1, b1 ]×[a2 , b2]×· · ·×[an , bn ] ⊂ Rn for some real a1, a2 , . . . , an , b1 ,
b2 , . . . , bn where ak < bk for k = 1, 2, . . . , n. Then F is totally bounded.

Outline of the Proof. The result follows by proving Exercises 4.2 and 4.3:
Exercise II.4.2. Let p = (p1 , p2 , . . . , pn ) and q = (q1, q2 , . . . , qn ) be points in Rn ,
with pk < qk for k = 1, 2, . . . , n. Let R = [p1 , q1 ] × [p2 , q2 ] × · · · [pn , qn ]. Then
( n )1/2
X
diam(R) = d(p, q) = (qk − pk )2 .
k=1
II.4. Compactness 6

Exercise II.4.3. Let F = [a1 , b1 ] × [a2 , b2 ] × · · · × [an , bn ] ⊂ Rn and let ε > 0. Use
Exercise 4.2 to show that there are rectangles R1 , R2, . . . Rm such that F = ∪m
k=1Rk

and diam(Rk ) < ε for k = 1, 2, . . . , m. Use this to show that F is totally bounded.

Theorem II.4.10. Heine-Borel Theorem.

A subset K of Rn (n ≥ 1) is compact if and only if K is closed and bounded.

Note. In Section II.1 of these class notes, we introduced topological spaces. Many
of the concepts of this section also hold in the topological space setting. For ex-
ample, a separation of a set is defined in terms of open sets (see the class notes for
Section II.2) and so this definition and hence the definition of connectedness is a
topological property. In the previous section (Section II.3) we explored convergence
of sequences in a metric space. We can also do this in a topological space, as the
following definition shows.

Definition. Let {xn } be a sequence in topological space (X, T ). Let x ∈ X. If for

all open sets )x ∈ T containing x, there exists N ∈ N such that for all n ≥ N we
have xn ∈)x, then x is the limit of sequence {xn }, denoted lim xn = x.

Note. In a topological space, it may not be the case that the limit of a sequence
is unique (it depends on the topology).
II.4. Compactness 7

Note. In the next section, we will define continuity of a function from one topo-
logical space to another.

Note. In this section, we approached compactness in a metric space by appealing

to Cauchy sequences. One might assume that a similar approach can be taken in a
topological space. However, this is not the case as illustrated in an example below.
First, we need another topological definition.

Definition. Let (X, T1) and (Y, T2) be topological spaces. A function f : X → Y
is a homeomorphism if f is continuous, one to one, onto, and has a continuous
inverse.

Note. Homeomorphic topological spaces are indistinguishable as topological spaces.

A set in T1 is open if and only if its image is open in T2 . So homeomorphic topo-
logical spaces share all “topological” properties.

Example. Let X = (0, 1) and Y = (1, ∞) where T1 is the usual topology on X and
T2 is the usual topology on Y . Then the function f (x) = 1/x is a homeomorphism
between (X, T1) and (Y, T2 ). Since each space has the “usual topology,” then these
spaces are metrizable (that is, there is a metric which induces the topology; it
is the usual metric d(x1 , x2 ) = |x1 − x2 | for each space). However, the sequence
{xn } = {1/n} is Cauchy in metric space (X, d) but the sequence {f (xn )} = {n}
is not Cauchy in metric space (Y, d). So we cannot define a Cauchy sequence in a
topological space in a way such that a sequence’s Cauchy-ness is preserved between
homeomorphic topological spaces!
II.4. Compactness 8

Note. John von Neumann published “On Complete Topological Spaces” in 1935
(Transactions of the American Mathematical Society 37(1), 1–20). This is available
online at: http://www.ams.org/journals/tran/1935-037-01/S0002-9947-1935
-1501776-7/S0002-9947-1935-1501776-7.pdf (accessed 10/10/2015). In this
paper he addresses the idea of Cauchy sequences in metric spaces and comments:
“The need of uniformity in [metric space] M arises from the fact that the elements
of a fundamental sequence are postulated to be ‘near to each other,’ and not near
to any fixed point. As a general topological space . . . has no property which leads
itself to the definition of such a ‘uniformity,’ it is impossible that a reasonable no-
tion of ‘completeness’ could be defined in it.” In this paper, von Neumann discusses
total boundedness and comapctness is the setting of topological linear spaces. His
definition of complete is then:
Topological linear space L is topologically complete if every
closed and totally bounded set S ⊂ L is compact.
The ‘uniformity’ concern is dealt with by ‘anchoring’ open sets at the origin of the
linear space (that is, using the zero vector 0): “However, linear spaces . . . , even if
only topological, afford a possibility of ‘uniformization’ for their topology: because
of their homogeneity everything can be discussed in the neighborhood of 0.”

Revised: 10/3/2017