9
$\begingroup$

Is there a nice characterization for the complete metric subspaces of $\mathbb{Q}$ (with the usual metric)?

It seems like a such a subspace must have empty interior; if it contained an open interval then it would clearly contain a non-converging Cauchy sequence. But this isn't enough, because consider the subset $\left\{\frac{1}{n} : n \in \mathbb{N} \right\} \subseteq \mathbb{Q}$. This has empty interior but is not complete because it doesn't contain $0$. In other words, a complete subspace of $\mathbb{Q}$ must be closed (which I think is true about metric spaces in general).

So a complete subspace of $\mathbb{Q}$ must be closed and have empty interior, but are these two sufficient to imply completeness? If so, how do I prove this, and if not, what am I missing? I'm not necessarily looking for a complete answer, just some pointers in the right direction.

$\endgroup$
  • $\begingroup$ Ah right, the Cantor set. $\endgroup$ – Vikram Saraph Jun 2 '14 at 19:27
  • $\begingroup$ So is there some nice property $P$ so that $A \subseteq \mathbb{Q}$ is complete if and only if $A$ is closed, $A$ has empty interior, and $P$ is true of $A$? $\endgroup$ – Vikram Saraph Jun 2 '14 at 19:33
  • $\begingroup$ Yeah, still thinking about that. :) $\endgroup$ – Thomas Andrews Jun 2 '14 at 19:34
6
$\begingroup$

There is a complete characterization of all countable complete metric spaces. Ordinals.

More specifically, one can show that if $\alpha$ is a countable ordinal then the order topology on $\alpha$ is completely metrizable.

Of course that there are complete subspaces of $\Bbb Q$ which are not order isomorphic to an ordinal; however they are homeomorphic to an ordinal (once we choose the metric, and we forget about the order).

[In my previous post I wrote something which was false; that successor ordinals were the complete ones, however I neglected the fact that a sequence whose distances from the first element approach $\infty$ need not have a limit. Successor ordinals are the compact metric spaces, though. So a subset of $\Bbb Q$ is compact if and only if it is homeomorphic to a successor ordinal.]

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks! Regarding your last point, are you saying that a subset of $\mathbb{Q}$ is complete if and only if it's compact, or do you mean something else? $\endgroup$ – Vikram Saraph Jun 3 '14 at 18:02
  • $\begingroup$ Hmmm, I notice that something is a wee bit off here. $\Bbb N$ is complete, but not a successor ordinal (nor compact). There might be a slightly more complicated answer afoot. I'll think about it and if I won't be able to find a decent solution I'll remove my answer. $\endgroup$ – Asaf Karagila Jun 3 '14 at 21:00
  • 1
    $\begingroup$ Vikram, I couldn't find an elegant solution. (The only thing I managed to think of is that there is a sequence of intervals $[a_n,b_n]$ such that $|a_n|,|b_n|\to\infty$ and $X\cap[a_n,b_n]$ is compact; what I wrote is also false, every countable ordinal is completely metrizable; the successor ordinals are the compact ones.) $\endgroup$ – Asaf Karagila Jun 4 '14 at 17:46
  • $\begingroup$ Right, I think this is much better now. Sorry for my previous mistake. :-) $\endgroup$ – Asaf Karagila Jun 5 '14 at 1:05

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.