2
$\begingroup$

In the same way that we can say manifolds are made of pieces that look like $\mathbb{R}^n$, is there any way to say that spaces with the same hausdorff dimension are made up of pieces that look the same? Formally, I mean something like:

For $\alpha \in \mathbb{R}^+$, consider the category of (compact, ...) spaces with Hausdorff dimension $\alpha$ where the maps are bi-lipshitz. Is there an object in this category with a bi-lipshitz map to every other?

The reason for the maps being bi-lipshitz is that if we have a bi-lipshitz map $f : X \to Y$, then the hausdorff dimension of the domain is the same as its image $\text{dim}_H (X) = \text{dim}_H (f(X))$. In particular, this statement is saying that for every $\alpha$ there is a 'prototypical' space of dimension $\alpha$ such that all spaces with dimension $\alpha$ have a piece that looks like the prototype.

A (further) condition that I'm not sure is necessary but might help is to only consider spaces which are 'everywhere dimension $\alpha$', in that all balls in the space have dimension $\alpha$. Even in the $\alpha = 1$ case, (where you would hope that if the spaces are all compact and everywhere dimension $1$ that the initial object is a real interval?) I'm finding it hard to prove/find counterexample

$\endgroup$
5
  • $\begingroup$ For $\alpha=1$ there is no initial object - if $X$ where one, then there should be a unique map $X\to [0,1]$. But since $X$ is nonempty, we can compose this with a map $[0,1]\to [0,1/2]\subseteq [0,1]$ and get a different such map. Same idea should work for any $\alpha>0$, just find a fractal of a suitable dimension. $\endgroup$
    – Wojowu
    Commented May 3, 2019 at 17:50
  • $\begingroup$ Of course there is a huge variety of possible topological local types (or bi-lipschitz if you prefer), even if you assume local connectedness $\endgroup$ Commented May 3, 2019 at 17:51
  • 1
    $\begingroup$ @Wojowu sorry I didn't mean an object with unique morphisms to every object - I got confused with terminology. I just meant an object with a morphism to every object $\endgroup$ Commented May 3, 2019 at 17:52
  • $\begingroup$ There are instead structure theorems available for spaces of given Hausdorff dimension. You may check Federer's book. $\endgroup$ Commented May 3, 2019 at 17:55
  • 1
    $\begingroup$ For $\alpha=1$, you have fat Cantor sets which have Hausdorff-dimension $1$ and the interval doesn't map to it nontrivially, so $[0,1]$ won't work. I think by considering a suitable family of fat Cantor sets we might be able to prove such a set doesn't exist. $\endgroup$
    – Wojowu
    Commented May 3, 2019 at 17:59

0

You must log in to answer this question.

Browse other questions tagged .