1
$\begingroup$

We denote this property by $(*)$.

Of course, if we can isometrically embed $X\hookrightarrow\mathbb{R}^n$, then $X$ is $(*)$. Is this sufficient to characterize all $X$? I've tried to construct counterexamples, but they all seem to have the problem of open balls being too big, so to speak. Obviously, the infinite-dimension normed vector spaces from functional analysis are not $(*)$. However, I suspect that the techniques necessary to prove that $(*)$ implies the existence of an isometric embedding are way beyond me.

For example, if $X$ is discrete, then it is $(*)$ iff it is finite, iff it can be isometrically embedded in $\mathbb{R}^n$ for some $n$.

Even if there exists some $(*)$ $X$ that cannot be embedded in $\mathbb{R}^n$, then my intuition tells me that we nonetheless have some constraint on the "dimension" of $X$, suitably defined.

So, what would be an intrinsic property of a metric space that is equivalent to $(*)$?

$\endgroup$
  • 1
    $\begingroup$ Look up the necessary and sufficient conditions for compactness (of metric spaces). $\endgroup$ – avs Dec 2 '16 at 22:38
  • $\begingroup$ So, is it necessary and sufficient that a closed ball be compact in this metric space? $\endgroup$ – Monstrous Moonshine Dec 2 '16 at 22:54
2
$\begingroup$

Let $H$ be the Hilbert cube. For another description see this question. In either version $H$ is a compact metric space, so every subset of $H$ is totally bounded. Every separable metric space can be embedded in $H$. In particular, $\Bbb R^{n+1}$ can be embedded in $H$. If $H$ could be embedded in $\Bbb R^n$, then so could $\Bbb R^{n+1}$, but in fact it cannot: $\Bbb R^m$ embeds in $\Bbb R^n$ if and only if $m\le n$. Thus, $H$ is a counterexample to the conjecture that a metric space with $(*)$ can be embedded in some $\Bbb R^n$.

$\endgroup$
  • $\begingroup$ So, can we just replace $\mathbb{R}^n$ with $H^\infty$? Is the conjecture true in that case? $\endgroup$ – Monstrous Moonshine Dec 2 '16 at 23:19
  • $\begingroup$ @MonstrousMoonshine: You can simply use $H$ itself: a metric space satisfying $(*)$ is separable, and every separable metric space embeds in $H$. $\endgroup$ – Brian M. Scott Dec 2 '16 at 23:24

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.