0
$\begingroup$

Let $\left\lbrace X_n \right\rbrace$ be a sequence of compact subsets of a metric space $M$ with $X_1 \supset X_2 \supset X_3 \supset \dots$. Prove that if $U$ is an open set containing $\cap X_n$, then there exists $X_n \subset U$.

This proof is tricky for me because I can't use many facts about compactness beyond its definition (a metric space $M$ is compact if every open cover of $M$ has a finite subcover). For example, I can't use the fact that every sequence in a compact metric space has a convergent subsequence, or the fact that a compact subset is closed and bounded. I do "know" that a finite union of compact subsets is compact and the intersection of compact subsets is compact.

My main idea has been a proof by contradiction in which I assume that $X_n \not\subset U$ for all $n$. Then I find a point in each $X_n$ that is not in $U$ and build a sequence $x_n$ that converges to a point in $U$. Since $x_n$ is in the closed subset $M \backslash U$, I would have a contradiction. But I can't figure out how to show that such a sequence exists.

$\endgroup$
  • $\begingroup$ Well. $X_n$ are not empty and $U$ doesn't contain any $X_n$ , then you have at the very least a constant sequence. $\endgroup$ – Jacky Chong Oct 19 '16 at 4:04
  • $\begingroup$ Okay. But a constant sequence in $M \backslash U$ could not converge to a point in $U$. $\endgroup$ – mr_gasbag Oct 19 '16 at 4:08
  • $\begingroup$ If $x_1 = x_2 =x_3 = \ldots$ which belongs in all the $X_n$ then the limit is in $\bigcap X_n \subset U$. $\endgroup$ – Jacky Chong Oct 19 '16 at 4:09
  • $\begingroup$ So I argue that if $X_n \not\subset U$ for all $n$, then there must exist a point $x \in X_n \cap M \backslash U$ for all $n$. Then $x \in \cap X_n \subset U$, a contradiction. I'm not so sure about this because it doesn't make use of the openness of $U$ or the compactness of $X_n$, and I know that the result doesn't hold without those restrictions. Am I missing something? $\endgroup$ – mr_gasbag Oct 19 '16 at 4:33
1
$\begingroup$

Given the tools available to you, sequences aren’t the best way to go.

HINT: Let $K=X_1\setminus U$; $K$ is a closed subset of the compact set $X_1$, so $K$ is compact. For each $n\in\Bbb Z^+$ let $V_n=M\setminus X_n$, and let $\mathscr{V}=\{V_n:n\in\Bbb Z^+\}$.

  • Show that $\mathscr{V}$ is an open cover of $K$.
  • Conclude that there is an $n\in\Bbb Z^+$ such that $K\subseteq V_n$.
  • What does this tell you about $X_n$?
$\endgroup$
  • $\begingroup$ This looks like it would work. But in order to show that $\mathscr{V}$ is an open cover of $K$, I need to show that $V_n = M \backslash X_n$ is open in $M$ for each $n$, which means (I think) using the theorem that the $X_n$ compact implies $X_n$ closed, which is not available to me. $\endgroup$ – mr_gasbag Oct 19 '16 at 15:13
  • $\begingroup$ @mr_gasbag: Nothing keeps you from proving it as a step in the larger proof. If $x\notin X_n$, then for each $y\in X_n$ there are disjoint open sets $G_y$ and $H_y$ such that $x\in G_y$ and $y\in H_y$. $\{H_y:y\in X_n\}$ is an open cover of $X_n$, so it has a finite subcover, say $\{H_y:y\in F\}$ for some finite $F\subseteq X_n$. Then $\bigcap_{y\in F}G_y$ is an open nbhd of $x$ disjoint from $X_n$, and $X_n$ is therefore closed. $\endgroup$ – Brian M. Scott Oct 19 '16 at 15:18
  • $\begingroup$ Great, thank you very much! $\endgroup$ – mr_gasbag Oct 19 '16 at 20:51
  • $\begingroup$ @mr_gasbag: You’re very welcome. $\endgroup$ – Brian M. Scott Oct 19 '16 at 21:03
1
$\begingroup$

Suppose not. For each $n$ let $p_n\in X_n$ \ $U.$ Then $(p_n)_{n\in N}$ is a sequence in $X_1$ so there exists $q\in X_1$ and a strictly increasing $f:N\to N$ such that $(p_{f(n)})_{n\in N}$ converges to $q.$

And for $j>1, $ the sequence $(p_{f(n)})_{f(n)\geq j}$ is a sequence in $X_j,$ also converging to $q,$ so $q\in X_j.$

So $q\in \cap_{n\in N}X_n.$ So $q\in U.$

Now $(p_{f(n})_{n \in N}$ converges to $q ,$and $f$ is strictly increasing, so any nbhd of $q$ contains $p_{f(n)}$ for all but finitely many $n.$ And $U$ is a nbhd of $q$ (because $q\in U$). So $p_{f(n)}\in U$ for all but finitely many $n .$

This contradicts $\forall n\;(p_{f(n)}\not \in U).$

$\endgroup$

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.