Diameter bounded from above is usually needed in the finiteness theorem or other convergence theorems in Riemannian Geometry. Let $M^n$ be a closed manifold and {$g_i$} be a family of smooth Riemannian metrics on it with $Inj_{g_i}\geq \alpha>0$ and $Vol_{g_i}\leq \beta$. Are those two conditions enough to imply $Diam_{g_i}\leq \gamma$? If so, does the $\gamma$ depend only on $\alpha,\beta$ and $n$?

  • $\begingroup$ Could you please accept an answer, by clicking the check mark next to it, or explain what more you want to see from an answer? Doing this helps the site run properly in a number of ways. $\endgroup$
    – Will Sawin
    Jan 23, 2020 at 23:16

1 Answer 1


By Croke, Some isoperimetric inequalities and eigenvalue estimates, Proposition 14, on an $n$-dimensional Riemannian manifold with injectivity radius $\alpha$, a ball of radius $\alpha/2$ has volume at least $C\alpha^n/ (2n)^n$ for a constant $C$. Specifically, by their theorem 11, $C=2^{n-1}V_{n-1}^{n\phantom{1}} / V_{n\phantom{1}}^{n-1}$, where $V_n$ is the volume of the $n$-dimensional sphere.

Let $x_1,\dots,x_m$ be a maximal collection of points such that the balls of radius $\alpha/2$ centered at $x_1,\dots,x_m$ do not overlap. Then clearly $$m\frac{ C\alpha^n }{ (2n)^n} \leq \beta.$$ By maximality, the balls of radius $\alpha$ centered at $x_1,\dots, x_m$ cover $M$. Hence because $M^n$ is connected, the graph with vertices $x_1,\dots,x_m$ and edges connecting $x_i$ to $x_j$ if the distance from $x_i$ to $x_j$ is at most $2\alpha$ is connected and thus has diameter at most $m-1$.

Thus the diameter is at most $$\alpha + (m-1) 2\alpha+ \alpha = 2m\alpha \leq 2 \beta \frac{ (2n)^n}{ C \alpha^n} \alpha = \frac{ \beta}{\alpha^{n-1} } \frac{2^{n+1} n^n} {C} $$ as for the distance between $x$ and $y$, it takes a distance of $\alpha$ to get from $x$ to one of the $x_i$, then $(m-1)$ distances of $2\alpha$ to travel to an appropriate $x_j$, and then $\alpha$ to get from $x_j$ to $y$.

This is sharp up to a constant depending on $n$, as demonstrated by a suitable long thin torus.


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.