It’s known that every closed subset of a compact topological space is also compact. However, it’s not always true that compact subspaces are closed(by taking the cofinite topology on $\mathbb{Z}$ the set of positive integers $\mathbb{Z}+$ is compact but is not closed). The following is my attempt to replace “closed” with a weaker condition so that the converse actually holds.

Let $(X, \mathcal{T})$ be a compact space. The idea for the following arguments is that in the proof of “closed subset $A$ of compact space is compact”, we show the complement $X-A$ can be removed so that the rest of the open sets also form a cover:

For a subset $A$ of $X$, define $$K(A)=\bigcap\{U\in \mathcal{T}: U\subset A\}.$$ If there exists a closed set $C$ in $X$ such that $A\subset C\subset K(A)$, we say $A$ is $K$-closed. Apparently every closed set is also $K$-closed.

Now if $A$ is $K$-closed, then by taking $X-C$ together with an open cover of $A$, we have an open cover, thus a finite subcover $\{U_i\}$of $X$. It can then be shown that if $X-C=U_i$ for some $U_i$, then it is removable so that the rest of the finite cover also covers $A$.

Now I wonder if the converse holds:

If $A\subset X$ is compact, is it true that $A$ must be $K$-closed?

Is there an alternate way to modify the result so that the converse holds?


$\mathbb{Z}+$ is compact in $\mathbb{Z}$ with the cofinite topology, but $K(\mathbb{Z}+)=\mathbb{Z}+$ so it’s not $K$-closed. My attempt does not solve the problem. Does anyone have a better idea?

  • $\begingroup$ $K(A) = \bigcap \{O \in \mathcal{T}: A \subseteq O\}$ of course. $\endgroup$ – Henno Brandsma Dec 12 '18 at 23:03
  • $\begingroup$ Thank you for the comment. I have re-edited the question. $\endgroup$ – William Sun Dec 12 '18 at 23:11

No. For instance, let $X=\{0,1\}$ with $\{0\}$ open but $\{1\}$ not open. Then $A=\{0\}$ is compact but $K(A)=A=\{0\}$ is not closed so $A$ is not $K$-closed.

Ultimately, the issue is that the "reason" $A$ is compact could be totally unrelated to the "reason" $X$ is compact. In particular, $X$ might have a point $x$ (like $1$ in the example above) whose only neighborhood is $X$ itself, which automatically guarantees that $X$ is compact. If $x\not\in A$, there's never going to be any nice way to relate open covers of $A$ to open covers of $X$ to learn anything interesting using the compactness of $A$, since open covers of $X$ are all trivial ($X$ itself must always be one of the sets).

| cite | improve this answer | |

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.