Baby Compactification

Previously, we have seen that \mathbb R can be compactified into the set \mathbb R^* := \mathbb R \cup \{ \infty \}, which is topologically equivalent to the unit circle (or rather, 1-sphere) S^1. We extended the usual topology on \mathbb R with sets of the form \mathbb R \backslash C \cup \{ \infty \}, where C \subseteq \mathbb R is compact and Hausdorff.

Can we accomplish such a compactification for general topological spaces (K, \mathcal T)? Can we append a point \infty to K, forming K^* := K \cup \{ \infty \}, and equip this set with the topology \mathcal T' := \mathcal T \cup \{ K^* \backslash C  : C \subseteq K\ \text{compact}\}? Well certainly, but the real question is whether K^* is a compact Hausdorff space or not.

Let’s suppose K^* is compact and Hausdorff. It is then not hard to verify that K is Hausdorff, if the subspace topology agrees with the topology on K.

Lemma 1. If the subspace topology of K \subseteq K^* agrees with the topology on K, then K is Hausdorff.

How does compactness come into play? Well, if K is compact then we don’t need to compactify anything, so clearly it’s only meaningful to discuss the case when K is not compact. Yet, ideally, K should be somewhat compact in order to be compactifiable.

Lemma 2. For any x \in K, there exists a compact set C \subseteq K that contains some neighbourhood U of x. In this case, we say that K is locally compact.

Proof. Fix x \in K and any neighbourhood U of x. Since \bar U is a closed subset of the compact space K^*, \bar U is the required compact set.

Thus, if K^* is a one-point compactification of K that contains K, then K is Hausdorff and locally compact, but not compact. It turns out that these conditions are not just necessary, but sufficient for K^* to be a one-point compactification of K.

Theorem 1. K is Hausdorff, locally compact, and not compact if and only if K^* = K \cup \{ \infty \} is a compact Hausdorff space that contains K as a subspace such that \bar K = K^*.

Proof. For the subspace claim, we claim that \mathcal T = \{V \cap K : V \in \mathcal T'\}, and it suffices to verify (\supseteq ) under the assumption V \notin \mathcal T. Since V \in \mathcal T' \backslash \mathcal T, V = K^* \backslash C for some compact C \subseteq K. Since K is Hausdorff, C is closed. Since \infty \notin K, V \cap K = (K^* \backslash \{ \infty \}) \backslash C = K \backslash C \in \mathcal T, since C is closed and thus K \backslash C is open.

For the Hausdorff claim, fix distinct x, y \in K^*, so that x \in K and y = \infty without loss of generality. Since K is locally compact, find a compact set C \subseteq K that contains some neighbourhood U \subseteq C of x. Since y \notin C, we have y \in K \backslash C. Since U \cap K \backslash C = \emptyset, we have that K^* is Hausdorff.

Finally, for the compactness claim, let \mathcal U = \{ U_x : x \in K \} be an open cover of K^*. Let U_\infty := K^* \backslash C for some compact C \subseteq K. For each x \in C, by the subspace claim we can assume U_x \subseteq K without loss of generality. Then \{ U_x : x \in C \} forms an open cover for C. Since C is compact, we can extract a finite sub-cover \{U_{x_1},\dots,U_{x_n}\}. Then \{U_{x_1},\dots,U_{x_n}, U_\infty \} forms a finite sub-cover of K^*, yielding compactness.

The preceding arguments prove (\Rightarrow), while Lemmas 1 and 2 yield the direction (\Leftarrow).

Now suppose that K is Hausdorff, locally compact, and not compact. Is K^* the only way to create a compactification of K? What if there exists another compactification L of K? Here, we mean that L \backslash K is a single point (in the context L = K^* we called this point \infty ), contains K as a subspace, and is compact and Hausdorff. It turns out that L \cong K^*, i.e. L and K^* are homeomorphic to each other.

Theorem 2. L \cong K^*.

Proof. Denote \{ \infty' \} = L \backslash K. Check via bookkeeping that the map f : K^* \to L defined by f|_K = \mathrm{id}_K and f(\infty) = \infty' is a homemorphism.

Definition 1. A compactification of a Hausdorff space K is a topological space L that is compact Hausdorff and contains K as a dense subspace, i.e. \bar K = L. Let K be Hausdorff, locally compact, and not compact. Define K^* to be the one-point compactification of K.

Hence, without loss of generality, we can call K^* the one point compactification of K. It is, in a sense, the smallest compactification of K, since we add at most one point to K. A natural question to ask would be: what is the largest compactification we could impose onto K? The answer lies in the Stone–Čech compactification of K.

But first, we need to take a detour and explore the Tychonoff theorem, which will come in handy when constructing Stone-Čech. What is this theorem, you might ask? It’s a simple question. We know that [0, 1] \subseteq \mathbb R is compact. Is it true that [0, 1]^J \equiv \prod_{j \in J} [0, 1] is compact, for even uncountable index sets J? Under the product topology, the Tychonoff theorem remarkably answers in the affirmative! We will explore this idea in the next post.

—Joel Kindiak, 2 May 25, 1739H

,

Published by


Leave a comment