While reading about topologies on continuous function spaces, I've seen remarks that core-compact and locally compact are equivalent for Hausdorff spaces.

Now I can clearly see that locally compact always implies core-compact, so the Hausdorff condition comes into the proof of the converse.

Let $K$ be a core-compact Hausdorff space. Let $x\in U\subseteq K$ with $U$ open. I need to show that $U$ contains a compact neighborhood of $x$. Since $K$ is core-compact there is an open neighborhood $V$ such that $x\in V\subseteq U$ with $V\ll U$.

I don't know where to go from here.


Here are the two papers which make make me believe this is a theorem:

Core Compactness and Diagonality in Spaces of Open Sets

Topologies on Spaces of Continuous Functions

I have found another source which claims something better: every sober core-compact space is locally compact (although I can't see the proof of the theorem; I might buy it).

Non-Hausdorff Topology

This hints that the property of Hausdorff spaces that we want to exploit is the fact that the intersection of all closed neighborhoods of a point is precisely that point.


I am now asking for help in completing the proof. This has been bothering me for too long.

  • $\begingroup$ What does core-compact mean? $\endgroup$ – Stefan Hamcke Jan 27 '14 at 23:22
  • 1
    $\begingroup$ Every neighborhood $U$ of every point $x$ contains a neighborhood $V$ of $x$ such that every open cover of $U$ contains a finite subcover of $V$. $\endgroup$ – Robert Wolfe Jan 27 '14 at 23:23
  • $\begingroup$ Hmm... Wouldn't the closure of this $V$ be compact? Since if $(W_i)_I$ is an open cover of $\bar V$, then $(W_i)_I\cup\{U-\bar V\}$ is an open cover of $U$. Then it contains a finite subcover of $V$, but $U-\bar V$ is not part of this subcover, so the subvocer must also cover $\bar V$. $\endgroup$ – Stefan Hamcke Jan 27 '14 at 23:29
  • 1
    $\begingroup$ In the comment above, I agree that the set $U-\overline{V}$ is not part of the subcover which covers $V$, but how do you conclude then that $\overline{V}$ is still contained in that same subcover though? $\endgroup$ – Robert Wolfe Jan 28 '14 at 2:25
  • 1
    $\begingroup$ If we start with everything in the original post, regularity would imply there is a closed set $x\in C\subseteq V$ with nonempty interior. Take an open cover of $C$ and add $C^c$. That's an open cover of $U$ containing a finite subcover of $V$. And we can remove $C^c$. $\endgroup$ – Robert Wolfe Jan 28 '14 at 3:18

This is probably not the quickest but here is a suggestion for core-compact $\Rightarrow$ local regularity.

Let $z \in X$. Let $z \in V \subset U$ as in the definition of core compact. Let $x \in V$ and $E \subset V$ a closed subset not containing $x$. For each $y \in F=\overline{E}^U$ choose an open neighborhood $U_y$ with $x \notin \overline{U_y}$ which we may do because $X$ is Hausdorff. The cover $\{U_{y}\}_{y \in F} \cup \{ F^c =U \setminus F \}$ is an open cover of $U$ and so has a finite subcover $U_{y_1},...,U_{y_n}, F^c$ which covers $V$.

Set $U' = U_{y_1} \cup U_{y_2} \cup ... \cup U_{y_n}$. The open set $U'$ covers $E$.

$x \notin \overline{U_{y_1}},...,\overline{U_{y_n}}$ by construction of the open sets $U_y$ so $x \notin \overline{U'}$. That is to say, $V$ is regular.

Rest of proof from the comment section added for completeness: First use local regularity to reduce to the case when $U$ is regular. Let $z \in X$ and $z \in V \subset U$ as in the definition of core compact. $U$ is regular by assumption so there exists a closed neighborhood $C \subset V$ containing $z$. Any open cover of $C$ extends to an open cover of $U$ by adding in $U \setminus C$. Any such open cover has a finite subcover covering $V$ which covers $C$. Removing the set $U \setminus C$ from this finite subcover yields a finite subcover of $C$.

  • $\begingroup$ Why couldn't $x\in\overline{U''}$? In addition to that, I'm confused why you introduced sequences. We do not know that our space is even sequential. $\endgroup$ – Robert Wolfe Feb 5 '14 at 8:38
  • $\begingroup$ Actually, you only need local regularity so it simplifies. $\endgroup$ – SomeEE Feb 5 '14 at 16:20
  • $\begingroup$ One final thing. I follow how you got that points and closed sets (closed in $X$) in $V$ can be separated by open neighborhoods, but how does that get you a closed set (closed in $X$) $C\subseteq V$ with $z\in C$? I know that regularity can be formulated in two different ways (as neighborhoods or closed sets), but does that extend to local regularity? $\endgroup$ – Robert Wolfe Feb 5 '14 at 17:28
  • $\begingroup$ "Every neighborhood U of z contains a V ...." is the definition of core-compact. Therefore we may assume that U is regular and C is closed in U which is all that we need. $\endgroup$ – SomeEE Feb 5 '14 at 18:31

This can be done easily with ultrafilters. Similarly to the ultrafilter characterisation of compactness, $V\ll U$ is equivalent to saying that each ultrafilter containing $V$ has a limit in $U$.

For the proof, let $x\in X$. By core-compactness, we can choose open neighborhoods $U$ and $V$ of $x$ such that $V\ll U$ and $U\ll X$. All ultrafilters containing $V$ have no limit outside of $U$, since they already have a limit in $U$, and they can only have one limit thanks to Hausdorffness. But this means that $\overline{V}\subseteq U$. Hence every ultrafilter containing $\overline{V}$ contains $U$, and since $U\ll X$, these ultrafilters converge. As $\overline{V}$ is closed, they converge to something in $\overline{V}$. So $\overline{V}$ is compact. So $x$ has a compact neighborhood.


This would be better suited as a comment but I cannot yet do that.

You would probably find a proof online quite easily if you searched for "exponentiability" but for a more specific proof go for "A Compendium of Continous Lattices" ( You can find this online ) by Gierz, Hofmann, Keimel, Lawson, Mislove, Scott. I believe there is shown straightforwardly something like given Hausdorff space and $U, V$ open then $U \ll V$ implies existence of compact $W$ so that $U \subseteq W \subseteq V$.


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.