1
$\begingroup$

Previously, in this post I've shown the following characterization of spaces with only zero-dimensional compactifications:

Theorem. Let $X$ be strongly zero-dimensional and $\beta X\setminus X$ locally compact. The following are equivalent:

  1. Every compactification of $X$ is zero-dimensional

  2. $\beta X\setminus X$ is scattered

So the case of when $\beta X\setminus X$ is locally compact is solved.

Although I don't think this problem can be worked on much further, I am looking for an example of a Tychonoff space $X$ such that every compactification of $X$ is zero-dimensional, but $\beta X\setminus X$ is not locally compact.

$\endgroup$
3
  • $\begingroup$ Can't we just let $X = \beta \omega \setminus (A \cup \{p\})$ where $A$ is a countably infinite discrete subset of $\beta \omega \setminus \omega$ and $p$ is an accumulation point of $A$? $\endgroup$
    – Anonymous
    Commented Nov 10 at 23:33
  • $\begingroup$ @Anonymous I don't know, the same argument doesn't apply to it $\endgroup$
    – Jakobian
    Commented Nov 11 at 0:04
  • $\begingroup$ @Anonymous sorry, I do see how to refine this argument to see that it works now $\endgroup$
    – Jakobian
    Commented Nov 11 at 0:16

2 Answers 2

3
$\begingroup$

Let $X = \beta X_0 \setminus A$ where $A\subseteq \beta X_0\setminus X_0$ and $\overline{A}\setminus A$ is locally compact, $X_0$ is strongly zero-dimensional, $A$ is scattered or countable, $A\to B$ is a perfect map with $B$ Tychonoff then $B$ is zero-dimensional.

Remark. The last condition is satisfied when $A$ is countable or scattered and locally compact.

Then $X$ has only zero-dimensional compactifications:

Suppose that $K$ is a compactification of $X$, let $\beta X_0\to K$ be the induced map, where $\beta X = \beta X_0$.

If $S\subseteq K$ is connected with at least two points then $S\subseteq K\setminus (\beta X_0 \setminus \overline{A}) = \overline{A}\setminus A\cup K\setminus X$.

The map $A\to K\setminus X$ is perfect so $K\setminus X$ is zero-dimensional. By the same argument as previosuly, $S\subseteq \overline{\overline{A}\setminus A}$. The space $\overline{A}\setminus A$ is open in its closure and zero-dimensional so that $S\subseteq \overline{\overline{A}\setminus A}\setminus (\overline{A}\setminus A)\subseteq K\setminus X$.

It follows that $\overline{S}\subseteq K\setminus X$ is a compact connected set and so because $A\to K\setminus X$ is perfect, $A$ contains a compact set surjecting continuously onto $[0, 1]$, and hence not scattered. This is impossible since $A$ is scattered or countable.

$\endgroup$
1
  • $\begingroup$ In fact, I wouldn't be surprised if given any countable subset $A$ of $\beta \omega \setminus \omega$, every continuous image of $\beta \omega \setminus A$ is zero-dimensional. $\endgroup$
    – Anonymous
    Commented Nov 11 at 1:49
3
$\begingroup$

Here is an argument that if $A$ is a countable subset of $\beta \omega \setminus \omega$, then every compactification of $X = \beta \omega \setminus A$ is zero-dimensional.

Suppose $K$ is a compactification of $X$ and $f \colon \beta \omega \to K$ is the Stone extension of the injection.

First note that, since every infinite compact subset of $\beta \omega$ is uncountable, every fiber $f^{-1}(y)$ is finite. The claim is that only finitely many fibers have size greater than one. For suppose not. Let $D$ be an infinite discrete subset of $K$ such that $f^{-1}(d)$ has at least two elements for each $d \in D$. Let $w_d$ and $z_d$ be distinct elements of $f^{-1}(d)$. Then $N = \cup_{d \in D}\{w_d,z_d\}$ is a copy of $\omega$ in $\beta \omega$ and the restriction of $f$ to $N$ is two-to-one. Therefore, by continuity, the restriction of $f$ to $Cl_{\beta \omega}N$ is also (at least) two-to-one, and, in particular, there is an element $y$ of $Cl_KA \setminus A$ such that $f^{-1}(y)$ has more than one element, a contradiction. This proves the claim.

The result now follows from the easy fact that if $X$ is any zero-dimensional space and $Y$ is obtained from $X$ by collapsing each of a finite number of finite sets to a point, then $Y$ is zero-dimensional.

It is not the case that the same argument works if $A$ is assumed only to be separable. In fact, if $A$ is any infinite compact subset of $\beta \omega \setminus \omega$ and $X = \beta \omega \setminus A$, then $X$ has a compactification which is not zero-dimensional. To get such a compactification, note that $A$ contains a copy of $\beta \omega \setminus \omega$, and let $K$ be the compactification obtained from $\beta \omega$ by mapping that copy to the interval $[0,1]$.

$\endgroup$
8
  • $\begingroup$ Do there exist subspaces $A\subseteq \beta\mathbb{N}$ (equivalently $A\subseteq \beta\mathbb{N}\setminus \mathbb{N}$) such that $A$ is not separable? $\endgroup$
    – Jakobian
    Commented Nov 23 at 15:45
  • $\begingroup$ $\beta {\mathbb{N}} \setminus {\mathbb{N}}$ itself is not separable. It also contains an uncountable discrete subset. $\endgroup$
    – Anonymous
    Commented Nov 23 at 21:32
  • $\begingroup$ I've never seen this fact. Is it because if you take a maximal almost disjoint family $\mathcal{A}$ on $\mathbb{N}$, then $\{\overline{A}\setminus\mathbb{N} : A\in\mathcal{A}\}$ is a pairwise disjoint family of non-empty clopen sets of size $\mathfrak{c}$? $\endgroup$
    – Jakobian
    Commented Nov 23 at 23:58
  • $\begingroup$ Oh sorry, that $\beta\mathbb{N}\setminus \mathbb{N}$ is not separable is a corollary of this answer! $\endgroup$
    – Jakobian
    Commented Nov 24 at 0:09
  • $\begingroup$ When you say that $f$ is two-to-one on the closure of $N$, which fact do you use? If $g:\mathbb{N}\to\mathbb{N}$ is a two-to-one map for example, then it extends to $\hat g :\beta\mathbb{N}\to \mathbb{N}\cup\{\infty\}$ which cannot be two-to-one. $\endgroup$
    – Jakobian
    Commented Nov 25 at 12:00

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .