For separable metric spaces, three fundamental notions of dimension are equivalent: $$ \text{dim }X = \text{Ind }X = \text{ind }X ,$$

Where does the doubling dimension fit into the picture?

  • 2
    $\begingroup$ First the first fact is topological (rather to state in term of metrizable spaces). Doubling dimension is a metric property (bilipschitz invariant, not topological). Now the question is certainly too broad: every possible rant about doubling dimension seems to answer the question. $\endgroup$ – YCor Apr 8 '18 at 12:08
  • $\begingroup$ So there is no relation between the doubling dimension ddim(X) and the toplogical dimension dim(X)? One can be finite and the other infinite? $\endgroup$ – Aryeh Kontorovich Apr 8 '18 at 12:57
  • 3
    $\begingroup$ It sounds obvious to construct metric spaces with finite (covering/inductive) dimension and infinite doubling dimension. In the other direction, finite doubling dimension implies finite dimension if I'm correct (e.g., by Assouad's embedding), but I don't know about explicit bounds. $\endgroup$ – YCor Apr 8 '18 at 13:24
  • $\begingroup$ I'll be happy to accept the obvious answer -- for both directions. $\endgroup$ – Aryeh Kontorovich Apr 8 '18 at 13:32
  • 1
    $\begingroup$ It may happen that the topological dimension is equal to the infimum of the doubling dimensions over all metrics generating the topology of the space. $\endgroup$ – Taras Banakh Apr 9 '18 at 7:37

In one direction, a rapidly branching tree will have very high doubling dimension, while having topological dimension $0$ (or $1$, if you include the edges). In another direction there is a bound, and this is discussed in the nice paper below (on the first page):

Le Donne, Enrico; Rajala, Tapio, Assouad dimension, Nagata dimension, and uniformly close metric tangents, Indiana Univ. Math. J. 64, No. 1, 21-54 (2015). ZBL1321.54059.


Let ${\rm dim}_H X$ and ${\rm dim}_d X$ denote the Hausdorff and the doubling dimension respectively. It is easy so see that ${\rm dim}_H X\leq {\rm dim}_d X$. Indeed, if ${\rm dim}_d X=s$, then we can cover a ball of radius $r$ by at most $2^{s}$ balls of radius $r/2.$ Therefore we can cover a ball of radius $1$ by $2^{ks}$ balls of radius $2^{-k}$ so we can estimate the Hausdorff measure $H^s(B(x,1))\leq C 2^{ks}2^{-ks}= C<\infty$ and hence ${\rm dim}_H X\leq s={\rm dim}_d X$. On the other hand by Theorem 8.14 in [1] (the proof is very short) we have that the Hausdorff dimension is greater than equal to the topological dimension so finally we get $$ \text{dim }X = \text{Ind }X = \text{ind }X\leq \text{dim}_H X\leq \text{dim}_d X. $$ As pointed by Igor Rivin the Hausdorff dimension of a rapidly branching tree can be arbitrarily large.

[1] J. Heinonen, Lectures on analysis on metric spaces. Universitext. Springer-Verlag, New York, 2001. MathSciNet review.


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.