4

In the book Munkres: Topology §30 I met the following problem:

Show that if $X$ has a countable basis $\{B_n\}_{n \in \mathbb{Z}_+}$, then every basis $\mathscr{C}$ for $X$ contains a countable basis for $X$. [Hint: For every pair of indeces $n$, $m$ for which it is possible, choose $C_{n, m} \in \mathscr{C}$ such that $B_n \subset C_{n, m} \subset B_m$].

I understand that $\{C_{n, m}\}_{n, m \in \mathbb{Z}_+}$ is countable collection of sets, but construction of this collection is not obvious for me. Can you give me some other hint, please?

mcihak
  • 614
  • 4
  • 12

2 Answers2

6

The suggestion is this: whenever the ordered pair $\langle n,m\rangle\in\Bbb Z_+\times\Bbb Z_+$ is such that $B_n\subseteq B_m$ and there is a $C\in\mathscr{C}$ such that $B_n\subseteq C\subseteq B_m$, let $C_{n,m}$ be that set $C$. (If there’s more than one $C\in\mathscr{C}$ such that $B_n\subseteq C\subseteq B_m$, just pick one of them to be $C_{n,m}$.) If there is no such $C\in\mathscr{C}$, let $C_{n,m}=\varnothing$. Then let $\mathscr{C}_0=\{C_{n,m}:\langle n,m\rangle\in\Bbb Z_+\times\Bbb Z_+\}$. Of course some members of $\mathscr{C}_0$ may be empty, but at any rate $\mathscr{C}_0$ is definitely just a countable subset of $\mathscr{C}$. Let $\mathscr{C}_1=\{C\in\mathscr{C}_0:C\ne\varnothing\}$; $\mathscr{C}_1$ is of course still a countable subset of $\mathscr{C}$, and the problem is to prove that $\mathscr{C}_1$ is a base for the topology.

To prove this, let $x\in X$ be arbitrary, and let $U$ be any open nbhd of $x$; we must prove that there is a $C\in\mathscr{C}_1$ such that $x\in C\subseteq U$. Since $\mathscr{B}$ is a base for the topology, there is some $B_m\in\mathscr{B}$ such that $x\in B_m\subseteq U$. Since $\mathscr{C}$ is a base for the topology, there is a $C\in\mathscr{C}$ such that $x\in C\subseteq B_m$. And since (again) $\mathscr{B}$ is a base for the topology, there is a $B_n\in\mathscr{B}$ such that $x\in B_n\subseteq C$. Can you finish it from here?

Brian M. Scott
  • 616,228
  • 1
    Yes, now it is clear! Thank you very much for your answer. Especially first paragraph of your answer helps me very much! And I have also one question. You wrote: "If there’s more than one $C$ ... just pick one of them." Do we need axiom of choice to do this? – mcihak Jul 10 '13 at 11:03
  • 1
    @mcihak: You’re very welcome. Yes, in general one does need the axiom of choice to do that. – Brian M. Scott Jul 10 '13 at 11:04
  • @BrianM.Scott:Hi Brian, there's something in your proof that makes me confused, because I'm wondering about the role of lower bound $B_{n}$. In your second paparagraph, even when you can conclude that for every $x$, there exists $B_{n} \in C \in B_{m}$, you can't say that $C \in \mathscr{C}{0}$, because in your selection process, for every $m,n$, you only choose ONE OF THEM(otherwise, you can't conclude that $\mathscr{C}{0}$ is countable). Please explain for me. Thanks – le duc quang Aug 14 '13 at 01:35
  • 3
    @leducquang: You know that there is at least one member, $C$, of $\mathscr{C}$ such that $B_n\subseteq C\subseteq B_m$, so you know that $C_{n,m}\ne\varnothing$. It's true that $C_{n,m}$ may be different from $C$, but it's also true that $x\in B_n\subseteq C_{n,m}\subseteq B_m\subseteq U$, which is what you need to show. – Brian M. Scott Aug 14 '13 at 12:01
  • @BrianM.Scott: I'm clear right now. Thanks so much :-) – le duc quang Aug 14 '13 at 13:19
  • @leducquang: You're welcome! – Brian M. Scott Aug 14 '13 at 13:20
0

I think the problem here is to describe the algorithm of finding $C_{n_1,m_1}$'s and showing that it terminates. Suppose $B_{m_1}$ is in $B$, then since $C$ is a basis, there is a $C_{n,m}$ s.t. $C_{n_1,m_1} \subset B_{m_1}$. Then again this set is open and $B$ is a basis so there exists $B_{n_1}$ s.t. $B_{n_1} \subset C_{n_1,m_1}$. Interesting enough this process can go further. We can find a $C_{n_2,n_1} \subset B_{n_1}$ since $B_{n_1}$ is open.

Now, when this process terminates? This process terminates when $C_{n_a,m_b} = B_{m_b} = B_{m_a}$. Therefore, for each $B_{m_i}$ in we can have at max $Z_+$ $C$'s. This is my particular opinion on this problem. Hope it helps.

foobar
  • 91