Lecture 22 - Intro to Baire's Theorem

We'll need to get another definitions:

Open Cover

A⊆R. An open subcover of A is a collection {Uλ|λ∈Λ} of open sets such that:
A⊆⋃λ∈ΛUλ

See also Finite Subcover.

Example

Consider the open interval (0,1). For each point x∈(0,1) let Ox=(x2,1). Taken together {Ox:x∈(0,1)} forms an open cover for the interval (0,1). Notice that it is impossible to find a Finite Subcover. Given any proposed finite subcollection:
{Ox1,Ox2,…,Oxn}
then let x′=min{x1,x2,…,xn} and observe that we can construct a number y satisfying 0<y≤x′2. Then y∉⋃i=1∞Oxi.

Now consider a similar subcover for the closed interval [0,1]. For x∈(0,1) the set Ox defined above works covering (0,1) but in order to have an open cover of the closed [0,1] interval then we need the endpoints. Let ε>0 and let O0=(−ε,ε) and O1=(1−ε,1+ε). Then the collection:
{O0,O1,Ox:x∈(0,1)}
is an open cover for [0,1]. But this time notice that there is a finite subcover. This is because the addition of O0 is here, we can choose x′ so that x′2<ε. It follows that our collection
{O0,Ox′,O1}is a finite subcover for [0,1].

Finite subcover

A finite subcover is a finite subcollection {Uλ1,Uλ2,…,Uλn} such that:
A⊆⋃i=1nUλi

See also Open Cover.

For examples see Open Cover#Example.

Some examples include:

  1. A=[0,1] for x∈A let Ux=V110(x). Then:
A⊆⋃0≤x≤1Ux

But ∃ finite # which still covers A. So ∃ a finite subcover.

  1. A=(0,1). For 0<x<1 let Ux=(0,x). Then A⊆⋃0<x<1Ux (in fact there's equality here). But no finite subcover exists.

There's something fundamentally different to (1) and (2). In some sense (1) is smaller than (2). This "smallest quality" is called compactness:

Heine-Burel Theorem

Let K⊆R. The following are equivalent:

  1. K is compact.
  2. K is closed and bounded.
  3. Every Open Cover of K has a Finite Subcover.

Proof

Since Characterization of Compactness in R gives (1)⟺(2) then all we have to show is (3)⇒(2) and then (3)⇐(2).

(⇒): Suppose every open cover of K has a finite subcover. To show that K is bounded, we construct an open cover for K by defining Ox to be an open interval of radius 1 around each point x∈K. In the language of Epsilon-Neighborhoods, Ox=V1(x). The open cover {Ox:x∈K} then must have a finite subcover (see the supposition) {Ox1,…,Oxn}. Because K is contained in a finite union of bounded sets (ie: K⊆⋃i=1nOxi) then K itself must be bounded (just take the right boundary of the largest xi).

To show that K is closed, there's more work. We'll argue it by contradiction. Let (yn) be a Cauchy Sequence contained in K where yn→y. To show that K is closed it is enough to show that y∈K.

Assume for contradiction that y∉K. Then ∀x∈K, then each x is some positive distance away from y (otherwise then x=y⇒y∈K). Thus, we can try to construct an open cover by taking Ox to be an interval of radius |x−y|2 around each point x∈K. Because we suppose (3), the resulting open cover {Ox:x∈K} must have some finite subcover {Ox1,…,Oxn}. The contradiction comes from noticing that this finite subcover cannot contain all of the elements of the sequence (yn). To show this, set:
ε0=min{|xi−y|2:1≤i≤n}
Because (yn)→y and ε0>0 (remember, all the distances are positive), then ∃N∈N such that:
|yn−y|<ε0=min{|xi−y|2:1≤i≤n}
for n≥N. Namely just choosing n=N here works, because notice that:
x∈Oxi⟺|xi−x|2<|xi−y|2
Notice the inequality never holds for y so then y∉Oxi for all 1≤i≤n. Thus, yN∉Ox1,…,Oxn, so then:
yN∉⋃i=1nOxi
So our cover doesn't actually cover all of K, which is a contradiction. Thus y∈K and thus K is closed and bounded.

(⇐): We'll use the following lemma:

DeMorgan's For Infinite Sets

Given a collection of sets {Eλ:λ∈Λ}:
(⋃λ∈ΛEλ)c=⋂λ∈ΛEλc,(⋂λ∈ΛEλ)c=⋃λ∈ΛEλc

Proof

(1) If x∈(⋃λ∈ΛEλ)c then x∉⋃λ∈ΛEλ, meaning x∉Eλ for any λ∈Λ. This implies that x∈Eλc for all λ. So then x∈⋂λ∈ΛEλc Thus giving ⊆ for (1).

For ⊇, let x∈⋂λ∈ΛEλc. Then x∉Eλ for all λ meaning x∉⋃λ∈ΛEλ so then ⊇.

(2) Comes from doing a very similar argument to (1).

☐

We'll show closure inductively. Let F=F1∪F2. Let xn∈F and x=limn→∞xn. Let (xnk) be a subsequence of (xn) fully contained in F1 or F2. The subsequence must also converge to x, so x∈F1∪F2=F. This is the base case.

For the inductive step, let F=⋂λ∈ΛFλ. Then:
Fc=⋃λ∈ΛFλc
Each Fλc is closed, thus Fc is open, so then F=(Fc)c is closed.

☐

Lecture Proof

We did the proof of this in class, but I prefer to use the textbook's proof of the idea (it's just easier to follow). If you want to view the proof in another context though see the following.

Scratch:

For (2→3) it's kind-of a mess. But the idea is that we can get a suprema s and then actually say that we could go even futher past the suprema to indicate a finite subcover.

Proof

We'll show (1)⇒(2)⇒(3)⇒(1).

(1 → 2): See Characterization of Compactness in R.

(2 → 3): Let's prove a useful lemma really quick:

Let A⊆R be nonempty and bounded. Then ∃(an)⊆A and ∃(bn)⊆A such that an→supA and bn→infA.

Proof

For n∈N choose an∈A such that supA−1n<an≤supA. Using the Squeeze Theorem then we can show an→supA. The same can be done for the inf by using infA≤bn<infA+1n.

☐

Let's start the proof. Suppose K is closed and bounded. Then all limit points of K are in K and any sequence (xn)⊆K is bounded. If K is empty then we are done so suppose K≠∅. Because of this then supK and infK exist, so then let a=supK and b=infK. Therefore K⊆[a,b]. Note that a,b∈K since K is closed.

Now let {Uλ|λ∈Λ} be an open cover of K. Let S={c∈[a,b]|K∩[a,c] has a finite subcover}. Notice that since S⊆[a,b] then S is bounded. Further S≠∅ since a∈S for example. Thus we can take s=supS. Since K is closed, then s∈K (see the Lemma we just proved). Thus then ∃Uλ0 such that s∈Uλ0. Therefore ∃ϵ>0 such that s∈Vϵ(s)⊆Uλ0.

Since s−ϵ<s then ∃s1∈S such that s−ϵ<s1≤s. Therefore ∃ a finite subcover of K∩[a,s1]. Let's call that subcover Uλ1,…,Uλn. Then taking:
Uλ0,Uλ1,…,Uλn
certainly still covers K∩[a,s]. Therefore s∈S.

Now assume for contradiction that s<b. Then ∃s2 such that s<s2≤b and s2<s+ϵ and K∩[a,s2] is covered by Uλ0,…,Uλn. Therefore s2∈S is a contradiction s2>s which contradicts s=supS.

(3 → 1): Suppose every open cover of K has a finite subcover. Now if you consider:
⋃n∈N(−n,n)=R⊇K
Thus we found a finite subcover (−n1,n1),…,(−nℓ,nℓ). Let M=max(ni|i=1,…,ℓ). Then M is a bound for K, so ∀x∈K then |x|≤M.

Now let (xn)⊆K. By the Balzano-Weierstrass Theorem then ∃ a subsequence xnk where xnk→x.

We need to show that x∈K to get (1), so assume for contradiction that x∉K. For y∈K let:
Uy=V|y−x|2(y)
forms an open over of K, so then ∃ a finite subcover Uy1,…,Uyn. Now let ϵ=min{|y−x|2|i=1,…,n}. None of the Uy epsilon-neighborhoods get into distance of x, since every y∈K is a distance ≥ϵ from x. This contradicts x being a limit.

☐