MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have a certain sequence $x_n$ in a complete and bounded metric space and I would like to prove that it has a convergent subnet (not necessarily subsequence). The best that I was able to do, until now, is to prove that $x_n$ verifies the following property:

for all $K\in\mathbb N$ and for all $\epsilon>0$, there is $n_{\epsilon,K}$ such that $d(x_n,x_m)<\varepsilon$, for all $n,m\geq n_{\epsilon,K}$ such that $|m-n|\leq K$.

Question: Does this sequence have necessarily a convergent subnet.

I really would like this is true, even if I suspect it is not. Nevertheless, I am not able to find a counterexample. I have also tried to prove that this is indeed true: since $X$ is a complete and bounded metric space, it is paracompact and by a theorem of Howes (th. 6.20 of his book "Modern Analysis and Topology") it would suffice to prove that $x_n$ is almost Cauchy. But it seems to me that my hypotheses do not imply that $x_n$ is almost Cauchy, since none of my sets where $x_n$ is Cauchy is cofinal.

Thanks in advance for any help,


share|cite|improve this question
Well, to exclude existence of convergent subsequences, it should be enough to take $x_n$ such that $d(x_n,x_m)=\frac{m−n}{n}$, for all $m\geq n$. It verifies the hypotheses above but has no Cauchy subsequences... subnets? – Valerio Capraro Jan 18 '12 at 15:49
Unless I am missing something, in a metric space (or any first countable space), shouldn't any sequence with a convergent subnet also have a convergent subsequence? So "subnet" here seems to be a red herring. – Nate Eldredge Sep 5 '13 at 17:09

Let the complete bounded metric space be the unit ball of the Hilbert space $l_2$, and let $\{e_i:i\in\mathbb N\}$ be an orthonormal basis. Consider the following sequence $(x_n)$, which goes from each $e_i$ to the next, but slower and slower. That is, $x_1=e_1$ and $x_2=e_2$ (so we went from $e_1$ to $e_2$ in one step), $x_3=(e_2+e_3)/2$ and $x_4=e_3$ (so we went from $e_2$ to $e_3$ in two equal steps), and so forth, going from $e_k$ to $e_{k+1}$ in $k$ equal steps. Because the steps get smaller and smaller, the sequence $(x_n)$ has the property specified in the question. But no subnet converges. Indeed, for each $i$, the inner products $\langle x_n,e_i\rangle$ are eventually zero, so the only possible limit for a subnet would be the zero vector. But each $x_n$ has norm at least $1/\sqrt2$, so convergence to zero is impossible.

share|cite|improve this answer
Yes, indeed! Thank you very much. – Valerio Capraro Jan 18 '12 at 15:54

The property you mention does not imply the existence of a convergent subnet: take e.g. the sequence $x_n:=\sqrt n$ on $\mathbb{R}$, as a metric space with the truncated standard distance, $d(x,y):=\min\{|x-y|, 1\}$, which is uniformly equivalent to the standard metric, and makes therefore $\mathbb{R}$ a bounded complete metric space. The property holds with $n_{\epsilon, K}:=\frac{K^2}{4\epsilon^2}\, ,$ although of course $(x_n)$ has no converging sub-net.

As a side remark, note that the property you wrote is equivalent to: $d(x_n,x_{n+1})\to0$ as $n\to\infty$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.