I would like to know:

Why is the Stone-Čech compactification of a discrete space the set of ultrafilters on that space?


We can show this by checking the following universal property for the Stone-Čech compactification:

If $ X $ is a topological space, then the Stone-Čech compactification of $ X $ consists of a compact Hausdorff space $ \beta X $, together with a continuous map $ \iota: X \to \beta X $, such that given any other compact Hausdorff space $ Y $ and a continuous map $ f : X \to Y $, there exists a unique continuous map $ \tilde{f} : \beta X \to Y $ such that $ \tilde{f} \circ \iota = f $.

So let $ X $ be a discrete space. Note that an arbitrary function from $ X $ to any other space is continuous, so we don't have to worry about checking continuity for maps coming out of $ X $.

Let $ F(X) $ be the set of ultrafilters of $ X $. We claim that $ F(X) $ is the Stone-Čech compactification of $ X $. The topology on $ F(X) $ is defined as follows. For any $ S \subseteq X $, let $ {F_{S}}(X) $ be the set of ultrafilters containing $ S $. Then the sets $ {F_{S}}(X) $ are closed under finite union: $ {F_{S}}(X) \cup {F_{T}}(X) = {F_{S \cup T}}(X)$. (This is equivalent to the following property of ultrafilters: An ultrafilter contains $ S \cup T $ iff it contains either $ S $ or $ T $.) Thus the sets $ {F_{S}}(X) $ form a closed basis of a topology (closed sets are defined to be arbitrary intersections of sets of the form $ {F_{S}}(X) $).

The map $ \iota: X \to F(X) $ is defined by sending $ x $ to the principal ultrafilter associated to $ x $.

Now let $ Y $ be any compact Hausdorff space, and consider a map $ f: X \to Y $. How will we extend $ f $ to $ F(X) $? If $ U $ is an ultrafilter, then for any $ S \in U $, consider the subset of $ Y $ defined by $ f(S) \stackrel{\text{df}}{=} \{ f(s) \mid s \in S \} $. Now the properties of ultrafilters imply that for any finite number of elements $ S_{1},\ldots,S_{n} \in U $, the intersection $ \bigcap_{i} S_{i} $ is nonempty. Hence $ \bigcap_{i} f[S_{i}] $ is nonempty as well. So any finite number of $ f[S] $ have nonempty intersection. By compactness of $ Y $, the intersection of all the closures, $ \bigcap_{S \in U} \overline{f[S]} $, is nonempty.

We claim this intersection cannot have more than one element. If $ y_{1},y_{2} \in Y $ and $ y_{1} \neq y_{2} $, then the preimages (under $ f $) of $ y_{1} $ and $ y_{2} $ are disjoint subsets of $ X $, and so every ultrafilter of $ X $ contains an element meeting one of the preimages but not the other. Hence the intersection $ \bigcap_{S \in U} f[S] $ (before taking closures) does not contain more than one element. The intersection of closures $ \bigcap_{S \in U} \overline{f[S]} $ also cannot contain more than one element, since if $ y_{1} $ and $ y_{2} $ were distinct elements in $ \bigcap_{S \in U} \overline{f[S]} $, we use the Hausdorff property to find disjoint open subsets $Y_{1},Y_{2} $ of $ Y $ separating $ y_{1} $ and $ y_{2} $, and reach a similar contradiction by considering the disjoint preimages of $ Y_{1} $ and $ Y_{2} $ under $ f $.

Hence, for every ultrafilter $ U $, the intersection $ \bigcap_{S \in U} \overline{f[S]} $ contains a single element $ y \in Y $. We define $ \tilde{f}(U) \stackrel{\text{df}}{=} y $.

Now I will leave it to you to check all of the following:

  1. $ F(X) $ is a compact Hausdorff space;
  2. $ \tilde{f} $ extends $ f $ (where $ X $ is identified with its image under $ \iota $);
  3. $ \tilde{f} $ is continuous;
  4. $ \tilde{f} $ the unique extension satisfying (2) and (3).

That is just one of many ways to define the Čech-Stone compactification $\beta D$ of a discrete space $D$. The Čech-Stone compactification of a Tikhonov space is characterized by the fact that it is the maximal compactification of that space, or by the fact that it has a certain universal property. Any construction that yields the maximal compactification of a space $X$, or that yields a compactification with that universal property, is a construction of $\beta X$, the Čech-Stone compactification of $X$. However, in the case of a discrete space $D$, the ultrafilter construction is in many ways the simplest.

And the answer to your question is that $\beta D$ is the set of ultrafilter on $D$ (equipped with the appropriate topology) precisely because the space that you get from that construction can be shown to have the universal property that characterizes $\beta D$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.