## Theorem 5.2.6: Heine-Borel Theorem

**S**of real numbers is compact if and only if every open cover

**C**of

**S**can be reduced to a finite subcovering.

### Proof:

First, assume that every open cover**C**of

**S**can be reduced to a finite subcovering. We will show that

**S**must then be closed and bounded, which means by the previous result that

**S**is compact.

**S** must be bounded: Take the collection
**C** = { :
**S**},
where
=
( - 1,
+ 1).
Then this collection is an open cover of **S**, and by assumption
can be reduced to a finite subcovering of **S**. But if
*a _{j1}* is the smallest of the centers of
the sets , and

*a*is the largest one, then

_{j2}**S**is contained in the set

*( a*and is therefore bounded.

_{j1}- 1, a_{j2}+ 1)
**S** must be closed: Suppose **S** was not closed. Then there
exists an accumulation point *s* of **S** that is not
contained in **S**. Since *s* is an accumulation point of
**S** we know:

- for any
*n > 1*there exists*a*with_{n}**S***| s - a*_{n}| < 1 / n

*s*must contain elements from

**S**. The sequence

*{ a*clearly converges to

_{n}}*s*. Define the collection of sets

**C**= { comp*([s - 1/n, s + 1/n]), n > 0 }*

**C**is open, being the complement of closed sets. They also cover

**S**, because the only point that this collection is missing is the point

*s*, which is not part of

**S**. By assumption, a finite subcover already covers

**S**. If

*N*is the largest index of that finite subcovering, then

*a*is not part of that subcovering. However,

_{N+1}*a*is an element of

_{N+1}**S**, so that this subcovering can not cover

**S**. That is a contradiction, showing that if

**S**was not closed, not every covering of

**S**can be reduced to a finite subcovering. Hence,

**S**must be closed.

Now we have to prove the other direction. Assume therefore that
**S** compact. Let **C** be any open cover of **S**. We
need to reduce **C** to a finite subcover. Since **S** is
compact, we know it is closed and bounded. Then
*a = inf( S)* and

*b = sup(*are both part of

**S**)**S**( Why ?). Define the set

**A**as

- A = {
*x: x [a, b]*and a finite subcollection of**C**covers*[a, x]*}**S**

**A**is not empty (because

*a*). Define

**A***c = sup(A)*

**A**is a subset of

*[a, b]*, we know that

*a < c b*. Suppose

*c < b*. Since

**S**is closed, comp(

**S**) is open. Therefore, if

*c*comp(

**S**) then there exists an open neighborhood

**U**of

*c*that is contained in

*[a, b]*(because

*c < b*) and disjoint from

**S**. But then

*c*can not be the supremum of the set

**A**. Therefore, if

*c < b*then

*c*

**S**. Then

*c*must be contained in some set from the open cover

**C**of

**S**. Choose two points

*y*and

*z*in with

*y < c < z*. As before, there exists a finite subcollection of

**C**whose members cover

*[a, y]*. Then these sets, together with cover

**S***[a, z]*. But then

**S***z*, which means again that

**A***c*can not be the upper bound for

**A**. This means that assuming

*c < b*leads to a contradiction, so that

*c = b*. But that will be exactly what we need. If

*sup(*, then let be that member of the open cover

**A**) = c = b**C**that contains

*b*. There exists some open neighborhood

*(b - , b + )*contained in . But

*b -*is not an upper bound for

**A**, so there exists

*x*with

*x > b -*and

*x*. Then

**A***[a, x]*is covered by a finite number of members of

**S****C**. Together with the set these sets form a finite open cover for

**S**.

We have indeed reduced the open cover of **S** to a finite
subcovering of **S**, finishing the proof. I think.