4
$\begingroup$

Let $X,Y$ be two compact metric spaces. Suppose there is a sequence of bi-Lipschitz homeomorphisms $f_n: X\to Y$, and $c\in(0,1]$, satisfying $$c\cdot d(x_1,x_2)\le d(f(x_1),f(x_2))\le \frac{1}{c}\cdot d(x_1,x_2),$$ for any $x_1,x_2\in X$.

Is is true that one can choose a sub-sequence $f_{n_i}$ that uniformly converges to a bi-Lipschitz homeomorphism $f$ (with the same constant $c$)?

Such a statement would imply that in case the Lipschitz distance between two compact metric spaces is bounded, it is actually realised by a bi-Lipschitz homeomorphism (which is better than taking an inf over all bi-Lipschitz homeomorphisms).

Improve this question
$\endgroup$
4
  • 4
    $\begingroup$ This should follow from an appropriate version of the Arzela-Ascoli theorem for maps between metric spaces. Even if you haven't seen such a version, it would be a good exercise to adapt the proof for real-valued functions. $\endgroup$
    – Nate Eldredge
    Jan 4, 2020 at 18:53
  • 1
    $\begingroup$ Yes: all you need is a subsequence that converges uniformly (or even just pointwise). The fact that the limit function must be bi-Lipschitz is easy. $\endgroup$
    – Nik Weaver
    Jan 4, 2020 at 19:48
  • 1
    $\begingroup$ Dear Nate and Nik, I guess you are right. I wonder though why the proof of Theorem 7.2.4 in Burago, Burago, Ivanov citeseerx.ist.psu.edu/viewdoc/… doesn't use this fact. It is a tiny bit more involved than that. This theorem states, in particular, that two compact metric spaces on Lipschitz distance $0$ from each other are isometric. $\endgroup$
    – aglearner
    Jan 4, 2020 at 19:48
  • $\begingroup$ @aglearner That's a striking application ! Did not see that at first. $\endgroup$
    – M. Dus
    Jan 5, 2020 at 10:27

1 Answer 1

Reset to default
2
$\begingroup$

Consider a subsequence $\,g_n\,$ of $\,f_n\,$ which is uniformly convergent. Then consider a subsequence $\,h_n\,$ of $\,g_n\,$ such that sequence $\,h^{-1}_n\,$ is uniformly convergent. Then the limit function $\,h\,$ of $\,h_n\,$ is bilipschitz with the bi-constant equal the same $\,c.$   Great

REMARK This proves what @NikWeaver has already said in his comment under the OP's Question. (Yes, I agree with the comment by @NatEldredge).

Improve this answer
$\endgroup$
4
  • $\begingroup$ Note that the result also holds not assuming that the $f_n$ are homeomorphisms and in this case, you cannot use $h_n^{-1}$. The $f_n$ certainly are one-to-one but necessarily onto in general. However, you don't need to play this trick to obtain a bi-Lipschitz limit. Since $g_n$ converges pointwise (note that this is enough) to a limit $g$, fixing $x_1$ and $x_2$, you get $1/c d(x_1,x_2)\leq d(g_n(x_1),g_n(x_2))\leq c d(x_1,x_2)$ for every $n$, so that the same holds with $g$. $\endgroup$
    – M. Dus
    Jan 5, 2020 at 9:05
  • 1
    $\begingroup$ @aglearner Actually the answer proves that and I think that's the point of the answer (although not explicitely stated). Denote by $g$ the limit of $g_n$ and by $h$ the limite of $h_n^{-1}$. Then, $h_n$ also converges to $g$. Also, $h_n \circ h_n^{-1}$ converges pointwise to $Id$ and converges pointwise to $g\circ h$, so that $g\circ h=Id$. Similarly, $h\circ g=Id$. Since both $g$ and $h$ are Lipschitz, they are in particular continuous and so they are inverse homeomorphism. $\endgroup$
    – M. Dus
    Jan 5, 2020 at 9:56
  • $\begingroup$ You are right M. Dus, thanks! $\endgroup$
    – aglearner
    Jan 5, 2020 at 10:16
  • $\begingroup$ @Dirk, did I say "bilinear"? :) === seriously, thank you Dirk for the correction. $\endgroup$
    – Wlod AA
    Jan 5, 2020 at 11:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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