Let $(M^n,g)$ be a smooth compact Riemannian manifold. It is well known that any sequence $\{X_i\}$ of compact subsets of $M$ has a subsequence which converges in the Hausdorff metric to a compact subset $X\subset M$.

Assume now that $\{X_i\}$ are, in addition, smooth connected submanifolds with a uniform lower bound on the sectional curvature and a uniform upper bound on the diameter with respect to the induced intrinsic (!) metric.

Question. What is known about a limit space $X$? E.g. should $X$ have an integer Hausdorff dimension?

Remark. Under the above assumptions, the Gromov compactness theorem implies that there is a subsequence converging to a compact Alexandrov space $Y$ in the Gromov-Hausdorff sense (thus $Y$ is not a subset of $M$ a priori). Is there any relation between $X$ and $Y$? (I believe there is a 1-Lipschitz map onto $Y\to X$.)


1 Answer 1


You are right about 1-Lipschitz map, but that is about all you can expect. In particular the dimension of $X$ might be not integer.

Assume your $X_i$ all isometric to to a flat torus $\mathbb{T}$ and $M=\mathbb{E}^N$; that is your manifold is the Euclidean space of high enough dimensions $N$.

By Nash and Whitney, any short map $\mathbb{T}\to \mathbb{E}^N$ can be arbitrary well approximated by an isometric embedding.

It reamins to construct a short map $\mathbb{T}\to \mathbb{E}^N$ with non-integer Hausdorff dimension. An image of that type is shown on the picture. It can be arranged that the Cantor set on the top has Hausdorff higher than 1 and lower than 2. FRACTAL-3d-FLOWER,

  • $\begingroup$ But by Nash and Whitney, the isometric approximation of the short map is only $C^1$-smooth. That probably means that $X_i$ are not infinitely smooth submanifolds. Can one make them to be infinitely smooth? $\endgroup$
    – asv
    Commented Jan 21, 2016 at 13:56
  • 1
    $\begingroup$ @sva No, I use the other theorem of Nash --- if $N$ is large, you can do the same for $C^\infty$ embeddings. $\endgroup$ Commented Jan 21, 2016 at 14:42

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.