3
$\begingroup$

The classic Hahn-Mazurkiewicz theorem has the following consequence: Let $X$ be a compact, connected topological manifold. Then there is a continuous surjective map $f: [0,1] \rightarrow X$.

It is also true that such a $f$ cannot be injective unless $X=[0,1]$ (since then $f$ would be a homeomorphism). I was wondering if some weaker notions of this is possible, where injectivity holds (this would be very useful for my research). Specifically, suppose we have a compact, connected Riemannian manifold $(M,g)$, without any boundary. Let $d$ be the metric induced on $M$ by $g$. Then I want to know the following:

For every $\epsilon > 0$, does there exist a map $f_{\epsilon}: [0,1] \rightarrow M$ that is continuous and injective, such that for any $x \in M$ there exists $y \in [0,1]$ such that $d(x,f_{\epsilon}(y)) < \epsilon$?

Does anything change if $M$ is non-compact?

$\endgroup$
3
  • $\begingroup$ I am interested; what application do you have in mind? $\endgroup$
    – Sam Nead
    Commented Aug 7, 2022 at 9:53
  • 1
    $\begingroup$ @Sam Nead The application is in geometric machine learning. Without going into too many details, it is easy to learn a function on [0,1], which can be used to approximately learn a function on a manifold $X$. But one needs to be able to control the approximation errors. We need some constructions for this, and the question I asked above is part of our construction (which now seems to be true)! $\endgroup$ Commented Aug 8, 2022 at 20:22
  • $\begingroup$ Thank you for the explanation. $\endgroup$
    – Sam Nead
    Commented Aug 8, 2022 at 20:27

1 Answer 1

4
$\begingroup$

I would call such $f_\epsilon$ coarsely $\epsilon$-dense, or more simply $\epsilon$-dense. These exist in the compact case as follows. Tile $M$ by $n$-balls (these may overlap, but only finitely). Each ball is homeomorphic to $[0, 1]^n$, so admits an $\epsilon/2$-dense smooth Hilbert arc. A small perturbation of these makes them pairwise disjoint. Choose an ordering of the $n$-balls. Now connect the Hilbert arcs together, in order. Tubular neighbourhoods allow us to do this smoothly and injectively.

In the complete, non-compact case this is not possible. This is because the continuous image of a compact set is compact, and thus contained in a metric ball.

$\endgroup$
2
  • $\begingroup$ Many thanks for the nomenclature suggestion. Also can you explain a little what you mean by the usual proof? The usual proof of which statement are you referring to here? $\endgroup$ Commented Aug 8, 2022 at 5:08
  • 1
    $\begingroup$ I have edited to add a few details. $\endgroup$
    – Sam Nead
    Commented Aug 8, 2022 at 11:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .