site logo
LIAM AXON

Uniform Spaces 4 - The Bolzano-Weierstrass Theorem

Feb 8, 2022
tags: math, topology

Pre-requisites: point-set topology, metric spaces, previous posts on uniform spaces.

This is the fourth post in my series on uniform spaces. See my other posts here:

  1. Nets
  2. Introduction to Uniform Spaces
  3. The Completion of a Uniform Space
  4. The Bolzano-Weierstrass Theorem
  5. Countably Uniform Spaces

The Bolzano-Weierstrass Theorem

So far in this series on uniform spaces, we have talked about uniform spaces primarily as generalizations of metric spaces, and used ideas from the study of metric spaces to motivate the theory of uniform spaces. However, I haven't yet explained how the study of uniform spaces can be applied to other areas of math. In this post, I will discuss the connection between uniform spaces and compact topological spaces, and end with an application of uniform spaces to a purely topological result (Tychonoff's Theorem).

The key to this proof is the Bolzano-Weierstrass Theorem for uniform spaces which characterizes compactness in terms of uniform space properties that are easier to check.

Theorem (Bolzano-Weierstrass in a Uniform Space): [Assuming choice] A uniform space is compact if and only if it is complete and totally bounded.

To explain the terminology and intuition behind this theorem, we once again turn to metric spaces.

The Bolzano-Weierstrass Theorem is an extremely important theorem in analysis and topology. It is, in a sense, the canonical link between metric spaces and compact spaces.

Theorem (Bolzano-Weierstrass for Metric Spaces): A metric space is compact if and only if it is complete and totally bounded.

Recall that a metric space MM is totally bounded if for all ϵ>0\epsilon > 0, there exists a finite list of points x1xmx_1 \dots x_m such that every point in MM is within distance ϵ\epsilon of some xix_i. I tend to think of compact topological spaces as "the small topological spaces". As every totally bounded metric space can be embedded into its completion, a complete totally bounded metric space (and hence, a compact metric space), I view totally bounded metric spaces as being "the small metric spaces". For example, a subset of Rn\mathbb{R}^n is totally bounded if and only if it is bounded.

In terms of entourages, a metric space MM is totally bounded if and only if for all entourages AϵA_\epsilon, there exists a finite collection {x1,x2,xn}\{x_1, x_2, \dots x_n\} of points in XX such that

X=i=1nAϵ[xi]X = \bigcup_{i=1}^n A_\epsilon[x_i]

In this form, we can clearly see how to generalize the notion of total boundedness to uniform spaces.

Definition: A uniform space (X,Φ)(X, \Phi) is totally bounded if for all AΦA \in \Phi, there exists a finite collection {x1,x2xn}\{x_1, x_2 \dots x_n\} such that

X=i=1nA[xi]X = \bigcup_{i=1}^n A[x_i]

Just like metric spaces, I tend think of totally bounded uniform spaces as "the small uniform spaces". For example: every subspace of a totally bounded space is totally bounded, and a union of two totally bounded spaces is totally bounded.

Importantly for the next section, the completion of a totally bounded space is still totally bounded.

Lemma: Let (X,Φ)(X, \Phi) be a totally bounded uniform space. Then, X\overline{X} is also totally bounded.

Proof: Let Φ\overline{\Phi} be the uniform structure on X\overline{X}. Fix an entourage AΦA \in \overline{\Phi}. Let BΦB \in \overline{\Phi} such that BBAB \circ B \subset A. Let CC be the restriction of BB to X×XX \times X. Since XX is totally bounded, there exist x1xnXx_1 \dots x_n \in X such that

i=1nC[xi]=X\bigcup_{i=1}^n C[x_i] = X

Hence,

i=1nB[xi]X\bigcup_{i=1}^n B[x_i] \supset X

Since XX is dense in X\overline{X}: for all xXx \in \overline{X}, there exists yXy \in X such that (x,y)B(x, y) \in B. Thus,

i=1nBB[xi]=X\bigcup_{i=1}^n B \circ B[x_i] = \overline{X}

\square.

There is a nice characterization of totally bounded spaces in terms of Cauchy nets that is similar to the Bolzano-Weierstrass Theorem for uniform spaces.

Theorem: [Assuming choice] Let (X,Φ)(X, \Phi) be a uniform space. Then XX is totally bounded if and only if every net in XX has a Cauchy subnet.

Proof: This can be proven similarly to the Bolzano-Weierstrass Theorem for uniform spaces. \square.

We can now prove the Bolzano-Weierstrass Theorem for uniform spaces:

Theorem (Bolzano-Weierstrass for Uniform Spaces): [Assuming choice] A uniform space is compact if and only if it is complete and totally bounded.

Proof: Let (X,Φ)(X, \Phi) be a compact uniform space. In the first blog post of this series, we proved that a topological space is compact if and only if every net has a convergent subnet. If a Cauchy net aa_\bullet has a convergent subnet bxb_\bullet \rightarrow x, then axa_\bullet \rightarrow x. Thus, every Cauchy net in XX converges. In other words, XX is complete.

To prove that XX is totally bounded, fix an entourage AA. For every xXx \in X, A[x]A[x] contains an open neighborhood UxU_x of xx.

xXUx=X\bigcup_{x \in X} U_x = X

As XX is compact, this open cover has a finite subcover (Ux1Uxn)(U_{x_1} \dots U_{x_n}). Thus,

i=1nA[xi]=X\bigcup_{i=1}^n A[x_i] = X

Hence, XX is totally bounded.

Now suppose that XX is complete and totally bounded. We need to prove that XX is compact. We know from the first blog post of this series that a space is compact if and only if every net has a convergent subnet. Suppose for contradiction that a:JXa_\bullet: J \rightarrow X is a net in XX with no convergent subnet.

Define a set SP(X)\mathcal{S} \subset \mathcal{P}(X) (a set of subsets of XX) to be a "good collection" if the following hold:

  1. S\mathcal{S} \neq \varnothing
  2. Whenever A1AnSA_1 \dots A_n \in \mathcal{S}, then a(A1An)a_\bullet \in (A_1 \cap \dots \cap A_n) infinitely often.

For example, {X}\{X\} is a good collection. Whenever (Sα)α(\mathcal{S}_\alpha)_\alpha is an increasing chain of good collections, then αSα\bigcup_\alpha \mathcal{S_\alpha} is a good collection. By Zorn's Lemma, there exists a maximal good collection S\mathcal{S}^*. It can be shown that S\mathcal{S}^* is a filter on XX, and that for any ASA \in \mathcal{S}^*, aAa_\bullet \in A infinitely often. Less obviously, it is also true that whenever ASA \in \mathcal{S}^* and A=i=1nAiA = \bigcup_{i=1}^n A_i, then AiSA_i \in \mathcal{S}^* for some ii.

We will use S\mathcal{S}^* to construct a Cauchy net without a limit. Define

T={(A,j):AS,ajA}\mathcal{T} = \{(A, j) : A \in \mathcal{S}^*, a_j \in A\}

Order T\mathcal{T} by: (A1,j1)<(A2,j2)(A_1, j_1) < (A_2, j_2) when A2A1A_2 \subset A_1. Let b:TXb_\bullet: \mathcal{T} \rightarrow X by b(A,j)=ajb_{(A, j)} = a_j. It is not hard to show that bb_\bullet is naturally a subnet of aa_\bullet.

We now show that bb_\bullet is Cauchy. Fix an entourage BΦB \in \Phi. Let CΦC \in \Phi such that CCBC \circ C \subset B and CC is symmetric. Since XX is totally bounded, there exist x1xnXx_1 \dots x_n \in X such that

X=i=1nC[xi]X = \bigcup_{i=1}^n C[x_i]

It follows from the properties of S\mathcal{S}^* that C[xi]SC[x_i] \in \mathcal{S}^* for some ii. Thus, bC[xi]b_\bullet \in C[x_i] eventually. Hence, (b,b)B(b_\bullet, b_\bullet) \in B eventually as well. It follows that bb_\bullet is Cauchy.

Since bb_\bullet is Cauchy and XX is complete, bb_\bullet converges. Thus, aa_\bullet has a convergent subnet. \square.

Compact Spaces as Uniform Spaces

In order to apply uniform space techniques to boring old topological spaces, we need to give those topological spaces uniform structures. Thankfully, every compact Hausdorff topological space has a unique uniform structure. The details are omitted, but the uniform structure can be defined follows.

Let XX be a compact Hausdorff topological space. For every finite open cover (U1Un)(U_1 \dots U_n) of XX, construct the entourage

A={(x,y):x,yUi for some i}A = \{(x, y) : x, y \in U_i \text{ for some } i\}

These entourages generate a uniform structure on XX.

Tychonoff's Theorem (Again)

After a couple lemmas about uniform spaces, we will be ready to apply uniform space theory to Tychonoff's Theorem.

Lemma: Let (Xα)αI(X_\alpha)_{\alpha \in I} be a collection of complete uniform spaces. The product X=αIXαX = \prod_{\alpha \in I} X_\alpha is complete.

Proof: Consider a Cauchy net aa_\bullet in XX. The projection of aa_\bullet into XαX_\alpha is a Cauchy net in XαX_\alpha, and has a limit xαx_\alpha. Let xx be the point (xα)αI(x_\alpha)_{\alpha \in I}. The limit of aa_\bullet is xx. \square.

Lemma: Let (Xα)αI(X_\alpha)_{\alpha \in I} be a collection of totally bounded uniform spaces. The product X=αIXαX = \prod_{\alpha \in I} X_\alpha is totally bounded.

Proof: If XX is empty, then it is trivially totally bounded. So assume that XX is non-empty. In particular, for any xXαx \in X_\alpha, there exists yXy \in X whose projection onto XαX_\alpha is xx.

Since entourages of the form E=Aα×βα(Xβ×Xβ)E = A_\alpha \times \prod_{\beta \neq \alpha} (X_\beta\times X_\beta) generate the uniform structure on XX, it suffices to check the total boundedness condition on these entourages. If

i=1nAα[xi]=Xα\bigcup_{i=1}^n A_\alpha[x_i] = X_\alpha

for some collection x1xnx_1 \dots x_n in XαX_\alpha, then

i=1nE[yi]=Xα\bigcup_{i=1}^n E[y_i] = X_\alpha

where the projection of yiy_i onto XαX_\alpha is xix_i. \square.

Theorem (Tychonoff's Theorem, Hausdorff Case): [Assuming choice] Let (Xα)αI(X_\alpha)_{\alpha \in I} be a collection of compact Hausdorff topological spaces. The product X=αIXαX = \prod_{\alpha \in I} X_\alpha is compact.

Proof: Each XαX_\alpha has a unique uniform structure, and that structure is complete and totally bounded. It suffices to show that XX is complete and totally bounded as a uniform space. This follows from the above two lemmas. \square.

The use of the Axiom of Choice in this proof (via the Bolzano-Weierstrass Theorem for uniform spaces) can not be avoided. In fact, it has been shown that the Tychonoff's Theorem is equivalent to the Axiom of Choice (in ZF). While this is arguably not the simplest proof of Tychonoff's Theorem, it still (I would argue) displays the power of uniform spaces.

Extensions: You can also define uniform spaces in terms of open covers, and if so, many of the arguments presented in this post become more natural.