Proposition: If metric space $M$ has a countable open base, then any open covering of $M$ admits a countable subcovering.

Definition: A collection of open subsets $\{U_i\}$ of $M$ is called an open base of $M$, if every of open set $O\subseteq M$ is expressible as $O=\bigcup_i U_i$.

Proof Attempt:

Let $\{\Omega_i\}_{i\in I}$, where $I$ is an arbitrary indexing set, be an open cover of $M$; that is $M \subseteq \bigcup_{i=1}^\infty \Omega_i$ and for all $i \in I$, $\Omega_i$ is an open set. First, we show that $\forall i\in I$, $\Omega_i \cap M$ is open relative to $M$. Let $x\in \Omega_i \cap M$, then there exists $\epsilon \gt 0$ such that $N_\epsilon (x)\subseteq \Omega_i$. But, if $N_\epsilon (x)\nsubseteq \Omega_i \cap M$, then it must be that there exists $n\in N_\epsilon (x)$ such that $n \notin M$. Relative to $M$, it does not affect whether $\Omega_i \cap M$ is open in $M$. What cannot be is $n \in M$ and $n \notin \Omega_i$. So, at any rate, $\Omega_i \cap M$ is open in $M$. Now, suppose $M$ has a countable open base $\{\mathscr{O}_k\}_{k=1}^\infty$. Thus, $\Omega_i \cap M=\bigcup_{k=1}^\infty \mathscr{O}_k$. Consider that $\{\Omega_i\cap M\cap\mathscr{O}_k\}_{k=1}^\infty$ is countable $\forall i \in I$, it covers $M$, and $\forall i\in I$, $\Omega_i\cap M\cap\mathscr{O}_k\subseteq\Omega_k$. Therefore, $\bigcup_{i=1}^\infty(\bigcup_{k=1}^\infty\Omega_i\cap M\cap\mathscr{O}_k)$ is a countable subcovering of an arbitrary open covering of $M$, because the union of countable set is countable.


1 Answer 1


I noticed some inaccuracies in your attempt that you should think about:

  1. You 'waste' a lot of time proving things related to relative open vs. open. This is not necessary here, since $M$ is the only metric space you are considering and everything happens inside $M$. In particular $\Omega_i \cap M = \Omega_i$ etc.

  2. Also $$ \Omega_i \cap M=\bigcup_{k=1}^\infty \mathscr{O}_k $$ will not be correct in general, because the right side is equal to $M$.

  3. Moreover, $\{\Omega_i \cap M \cap \mathscr{O}_k\}_k$ does not cover $M$ but $\Omega_i$.

  4. You correctly noted in the beginning that $I$ is arbitrary, but in the end you index the $\Omega_i$ by $i \in \Bbb{N}$, in which case there was nothing to prove from the start.

  5. $\{\Omega_i \cap M \cap \mathscr{O}_k\}_{k,i}$ is not a subcover of $\{\Omega_i\}_i$ in general. Recall that a subcover of $\{\Omega_i\}_{i \in I}$ is of the form $\{\Omega_{l}\}_{l \in L}$ for some $L \subseteq I$.

A better way to proceed is to prove the following statements:

  1. Every $\Omega_i$ is the union of some of the $\{\mathscr{O}_k\}_k$, say $\Omega_i = \bigcup_{k \in J_i}\mathscr{O}_k$ where $J_i \subseteq \Bbb{N}$.
  2. The set $J = \bigcup_{i \in I} J_i$ is countable.
  3. The set $\{\Omega_{k} \:|\: k \in J\}$ is a countable cover of $M$.
  4. For each $k \in J$ there exists $i_k \in I$ such that $\mathscr{O}_{k} \subseteq \Omega_{i_k}$. Choose one $i_k$ for each $k \in J$.
  5. The set $\{\Omega_{i_k} \:|\: k \in J\}$ is a countable subcover of $\{\Omega_i\}_i$.
  • $\begingroup$ at (1), I thought about assuming everything happens in $M$, but decided not to. at (2),(3),(4),(5).. I don't know what I was thinking. Thanks for pointing it out. I'll try the outline you provided. $\endgroup$ Aug 4, 2018 at 8:30

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.