Let $X$ be a non-compact, locally compact space. Also suppose that there is a sequence of compact non-empty sets $\{K_n\}_{n\in N}$ such that $$X=\bigcup_{n\in N} K_n,\quad K_n\subset K_{n+1}.$$ Now could we say that for any compact subset $K\subset X$ there is a number $n\in N$ such that $K\subset K_n$? If not, which conditions can help? or, is it possible if we choose sequence of compact sets? How about topological groups?


No, not as stated.

Let $X = [0,1)$ be the half-open unit interval, which is locally compact but not compact. Let $K_n = \{0\} \cup [\frac{1}{n}, 1-\frac{1}{n}]$ which is compact. Then we have $K_n \subset K_{n+1}$ and $X = \bigcup_n K_n$. But the compact set $K = [0, \frac{1}{2}]$ is not contained in any of the $K_n$.

However, under these assumptions here is something we can prove.

There exists a sequence of compact sets $K_n'$ such that $\bigcup_n K_n' = X$ and $K_{n-1}' \subset (K_{n}')^\circ$ for each $n$. In particular, for any compact set $K$, there is some $n$ with $K \subset K_n'$.

Proof. By local compactness, for each $x \in X$ there is an open set $U_x$ such that $x \in U_x$ and $\overline{U_x}$ is compact. We construct $K_n'$ recursively. To get started, let $K_0'=\emptyset$. Now to construct $K_n'$, suppose $K_{n-1}'$ is already constructed. Since $K_n \cup K_{n-1}'$ is compact, there exist $x_1, \dots, x_r$ such that $K_n \cup K_{n-1}' \subset U_{x_1} \cup \dots \cup U_{x_r}$. Set $K_n' = \overline{U_{x_1}} \cup \dots \cup \overline{U_{x_r}}$ which is compact. Then $K_{n-1}' \cup K_n \subset (K_n')^\circ$.

In particular, $\bigcup_n K_n' \supset \bigcup (K_n')^\circ \supset \bigcup_n K_n = X$. So the $K_n'$ cover $X$.

Moreover, the sets $(K_n')^\circ$ are an increasing open covering of $X$. So if $K$ is any compact set, we must have $K \subset (K_n')^\circ \subset K_n'$ for some $n$.

  • $\begingroup$ Yeah! You are right. $\endgroup$ – Hamid Shafie Asl Mar 14 '15 at 16:27
  • $\begingroup$ Thanks! Now I can make a bounded approximate identity for Banach algebra $C_0(X)$ of bound $1$ when $X$ is $\sigma$-compact. $\endgroup$ – Hamid Shafie Asl Mar 14 '15 at 17:28

I can't comment due to lack of reputation.

In the argument of @Nate Eldredge, do we need the space $X$ to be a Haussdorf space, in order to have closed compact neighbourhoods of each point?

  • $\begingroup$ Yes! It is Haussdorf space! $\endgroup$ – Hamid Shafie Asl Aug 5 '18 at 18:03

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.