6

Question 1. Let $\epsilon > 0$ and $V > 0$. Is there always a complete connected Riemannian manifold $M$ with $$ \operatorname{diam} M < \epsilon\quad\text{ (small diameter)} \quad \text{and} \quad \operatorname{vol}M > V\quad\text{(large volume)}? $$ In other words, can we construct worlds with room for arbitrarily many planets, but where any two planets are arbitrarily close to each other?

Note that $M$ has to be compact by Hopf-Rinow. I'm almost certain that the answer is 'yes', but I'm having a hard time finding an explicit sequence $M_1, M_2, M_3,\ldots$ such that $\operatorname{diam}(M_n)$ goes to zero while $\operatorname{vol}(M_n)$ goes to infinity.

Question 2. (Fixed dimension) Is there a sequence of complete connected Riemannian manifolds $M_1, M_2, M_3, \ldots$ with $$ \dim M_n = m, \quad \lim_{n \to \infty}\operatorname{diam} M_n = 0, \quad \text{and} \quad \lim_{n \to \infty}\operatorname{vol}M_n = \infty $$ for some fixed $m \in \mathbb{N}$.

For this one, I have less intuition, but I'm more inclined towards 'no'. Note that for $m = 0, 1$, this is clearly impossible. For $m = 2$ my intuition is very strongly in favour of 'no'.

Simon Parker
  • 1,363
  • 2
    By itself, diameter does not control volume. See e.g. https://mathoverflow.net/a/314653/1573. – Igor Belegradek Aug 22 '21 at 16:14
  • @Igor So a the connected sum of very many spheres of ever smaller radius would answer the question? – Harry Wilson Aug 22 '21 at 17:43
  • @HarryWilson: yes, the union of a large number of unit spheres connected by thin nearly cylindrical tubes answers the question, which is a special case (not even a duplicate) of the one I linked to. – Igor Belegradek Aug 22 '21 at 17:48

1 Answers1

6

The answer to both questions is positive, even in dimension 2. Take a round sphere of diameter $\epsilon$, and make many, say $N$ little holes in it. Then take $N$ spheres of diameter $\epsilon$ and make one little hole in each. Then glue these $N$ spheres to the first sphere along the boundaries of the holes. The diameter of the resulting surface is about $3\epsilon$ while the volume (area) is approximately $(N+1)\epsilon$. Now make $\epsilon$ as small as you wish, while $N>V/\epsilon$.

Here is another construction. Take a round sphere $S$ of diameter $\epsilon$, think of it as the Riemann sphere, and consider the $N$-fold ramified covering $S_1\to S,\;z\mapsto z^n$. Here $S_2$ is also a sphere, and equip it with the pullback of the metric from $S$. Then the solume of $S_1$ will be $N\epsilon$ while diameter is at most $2\epsilon$. If a smooth metric is desirable, approximate it with a smooth one.

Alexandre Eremenko
  • 88,753