0
$\begingroup$

As the title suggests, I am investigating whether all open sets in locally compact metric spaces are relatively compact. To me, this seems intuitive but I appear to be unable to produce a formal proof, that is, for an open set $U$ I cannot show that $\bar{U}$ is compact. I would therefore appreciate some help.

EDIT: The exact phrasing of the book from which this claim comes is : "If (X,d) is locally compact, we can even assume that $U$ is relatively compact.". Does this make sense?

$\endgroup$
  • $\begingroup$ No. Pick any topological space that is locally compact, but not compact (e.g. $\mathbb{R}$ under the usual topology). Then the space itself is open, but its closure is not compact. $\endgroup$ – Theo Bendit Sep 5 '18 at 8:28
  • $\begingroup$ @TheoBendit Great example, thank you. Do you perhaps know of any conditions that will make this true? $\endgroup$ – JohnK Sep 5 '18 at 8:37
  • $\begingroup$ Unfortunately, no I don't. Even in metric spaces, it's not that easy; it's possible for bounded sets in a locally compact space to not be relatively compact. $\endgroup$ – Theo Bendit Sep 5 '18 at 8:40
  • $\begingroup$ @TheoBendit Please see the edit for the exact statement. $\endgroup$ – JohnK Sep 5 '18 at 9:07
  • 2
    $\begingroup$ The phrase "we can even assume" makes it context-dependent. It depends on what you're using $U$ for. If you're using it in a situation where another smaller neighbourhood would suffice (i.e. one where we can always use some neighbourhood $V \subseteq U$ instead of $U$), then yes this make sense. If your neighbourhood $U$ is not relatively compact, then replace it with $U \cap V$, where $V$ is a relatively compact neighbourhood (which must exist by local compactness). $\endgroup$ – Theo Bendit Sep 5 '18 at 9:11
1
$\begingroup$

For your first question, no, the open sets in a locally compact space are not relatively compact. The entire topological space is a clopen set, and is relatively compact if and only if the entire space is compact.

For example, $\mathbb{R}$ with the usual topology, is locally compact, but not compact. Therefore, the subset $\mathbb{R}$ is open, but not relatively compact.

Another example: any infinite set equipped with the discrete metric. The space is locally compact, since for any point $x$ in the space, the open ball $B(x; 1/2) = \lbrace x \rbrace$ is open and compact. But, the open ball $B(x; 2)$ is the entire space, which is non-compact (all the singletons form an infinite open cover with no proper subcover). This illustrates how, in a metric space (unlike in a normed linear space), it's possible to have certain balls be relatively compact, and some not. It also shows an example of a bounded set in a locally compact space that isn't relatively compact.


As for your second question, in the edit, it may make sense depending on context. While not every open neighbourhood $\mathcal{U}$ of a point $x$ in a locally compact space is relatively compact, it is true that we can always find a smaller relatively compact neighbourhood $\mathcal{V}$ of $x$ contained in $\mathcal{U}$. All we do is use the definition of local compactness to find an open neighbourhood $\mathcal{W}$ of $x$ with compact closure. Then $\mathcal{U} \cap \mathcal{W}$ is an open set containing $x$. It has compact closure, because $$\mathcal{U} \cap \mathcal{W} \subseteq \mathcal{W} \implies \overline{\mathcal{U} \cap \mathcal{W}} \subseteq \overline{\mathcal{W}},$$ and closed subsets of compact sets are compact.

So, if the text uses $\mathcal{U}$ in such a way that we may always substitute $\mathcal{U}$ for a sub-neighbourhood $\mathcal{V}$, then we can safely replace $\mathcal{U}$ with a relatively compact neighbourhood $\mathcal{V}$. In most usages of neighbourhoods in topology, restricting neighbourhoods like this is acceptable (for example, showing convergence of a sequence/net).

$\endgroup$

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.