Let $K$ be a compact set in a metric space $X$. Show that, if $K\subset U_1 \cup U_2$ for some open set $U_1$ and $U_2$, then there exist compact set $K_1$ and $K_2$ such that $K_1\subset U_1$, $K_2\subset U_2$ and $K=K_1\cup K_2.$

I think that, in general, it's not possible to find disjoint compact sets $K_1$ and $K_2$, and I think this is a counterexample: Question about compact set contained in an union of open sets.

I define $K_1$ and $K_2$ as follow.

$$K_1=K\cap ((U_1-\partial(U_2))\cup \partial(U_1))$$

$$K_2=K\cap ((U_1-\partial(U_1))\cup \partial(U_2))$$

By $\partial(U_1)$ I mean the boundary of $U_1$. To show $K_1$ is compact, I will show it's sequentially compact.

Let $(x_n)\in K\cap (U_1-\partial(U_2)$, as $K $ is compact, $(x_n)$ has a convergent sequence $(x_{n_k})\to \bar{x}\in{K} $.

If $\bar{x}\notin U_1$, then it's in $\partial({U_1})$. (it's a limit point of $U_1)$.

If $(x_n)$ be seqeuence in $K\cap\partial(U_1) \subset K$, as $ K\cap\partial(U_1) $ and it's close set of compact set $K$, it's compact. Hence $K_1$ is compact.

Is my argument valid? Thank you.

  • $\begingroup$ Is your definition of $K_2$ correct? $\endgroup$ – Lord Shark the Unknown Aug 11 '17 at 4:40
  • $\begingroup$ Also, should it be specified that $U_1\cap U_2 = \emptyset$? As an aside, we use \partial to specify the boundary of a set, e.g. $\partial U_1$. $\endgroup$ – Michael Lee Aug 11 '17 at 4:47
  • 1
    $\begingroup$ Your $K_1$ and $K_2$ are not subsets of $U_1$ and $U_2$. $K_1$ may contain part of the boundary of $U_1$, which is not in $U_1$. $\endgroup$ – Robert Israel Aug 11 '17 at 4:47
  • $\begingroup$ @RobertIsrael Yes. I just added the boundaries to make them closed, but they are not subsets of $U_1$ and $U_2$ anymore. $\endgroup$ – Parisina Aug 11 '17 at 5:20
  • $\begingroup$ @MichaelLee No they are not disjoint here. $\endgroup$ – Parisina Aug 11 '17 at 5:22


If $Y$ is a normal space and $\{O_1 ,O_2\}$ is an open cover of $Y$, then there is a closed cover $\{F_1, F_2\}$ of $Y$ such that $F_i \subseteq O_i$ for $i=1,2$.

This is part of a proof that point-finite open covers of normal spaces have so-called closed shrinkings.

Proof of the lemma: $C_1 =Y\setminus O_1$, and $C_2 = Y\setminus O_2$. Then $C_1$ and $C_2$ are disjoint closed sets in $Y$ so they have disjoint open neighbourhoods $N_1$ resp. $N_2$. Then for $i=1,2$, $F_i = Y \setminus N_i$ are as required.

Now your question follows almost immediately: if $X$ is metric and $K \subseteq U_2 \cup U_2$ is compact, then $Y = U_1 \cup U_2$ is a normal space (all I need of a metric space is that it is hereditarily normal), and so we find $F_i \subseteq Y$ as required. Then $K_i = K \cap F_i$ are compact (this is certainly true in $Y$ but compactness is absolute) and as the $F_i$ cover $Y$, these sets are as required.

  • $\begingroup$ Very nice! I was working on a more (unnecessarily) complicated proof using the normality of $X$. I like yours much better. $\endgroup$ – Michael Lee Aug 11 '17 at 6:13
  • $\begingroup$ @MichaelLee I know the closed shrinking thing (standard fact) so this seemed a good fit. $\endgroup$ – Henno Brandsma Aug 11 '17 at 6:23

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.