Let $M$ be a open complete manifold with Ricci curvature $\ge 0$. By a theorem of Calabi and Yau, the volume growth of $M$ is at least of linear. I am wondering whether the following statement is true:

Let $p$ be any fixed point in $M$ and $B(p, r)$ be the distance ball of radius $r$ in $M$. Then for any given $R>0$, there exists a constant $c=c(p,R)>0$ such that $Area(\partial B(p, r))\ge c(p,R)$ for any $r>R$.

  • 1
    $\begingroup$ Do you mean for $r$ sufficiently large? $\endgroup$ – Ian Agol Oct 8 '11 at 21:52
  • $\begingroup$ Oh, yes. Thanks Agol, I've corrected my statement $\endgroup$ – user16750 Oct 8 '11 at 22:42
  • 1
    $\begingroup$ Don't have time to work out the details, but can't you use the Calabi-Yau lower bound on, say, $V(B(p, 4r))$ and the lower bound on Ricci to infer an isoperimetric inequality on all domains contained in $B(p, 2r)$. That and the Calabi-Yau lower bound on $V(B(p,r))$ implies a lower bound on the area of $\partial B(p,r))$. Not sure how the lower bound depends on $r$ though. $\endgroup$ – Deane Yang Oct 9 '11 at 0:15
  • 1
    $\begingroup$ Some evidence in favour: the analogue for horospheres (rather than distance spheres) is true. Specifically, for any Busemann function $b$ on $M$, the area of $b^{-1}(r)$ is eventually nondecreasing in $r$. See Lemma 20 in C. Sormani, "Busemann functions on manifolds with lower bounds on Ricci curvature and minimal volume growth," JDG 1998. (I don't think the theorem's hypothesis of exactly-linear volume growth is needed for this part of the conclusion.) $\endgroup$ – macbeth Dec 19 '11 at 1:15

The answer is YES.

Let $b\colon M\to \mathbb R$ be a Busemann function for a ray $\gamma$ from $p$, so that $b(p)=0$ and $b(x)\le 0$ for any $x\in \gamma$.

Set $$L_t=b^{-1}(t)\ \ \text{and}\ \ L^-_t=b^{-1}(-\infty,t].$$ Note that the sublevel sets $L_t^-$ are mean curvature concave for all $t$. In particular, any area minimizing hypersurface in $L_t^-$ with the boundary in $L_t$ lies in $L_t$.

Fix $t<0$. From above, $$\mathop{\rm area}\partial B(p,r)\ge\mathop{\rm area}(\partial B(p,r)\cap L_t^-)\ge \mathop{\rm area}( B(p,r)\cap L_t)\ge \mathop{\rm area}( B(p,R)\cap L_t);$$ i.e., the inequality holds for $c(p,R)=\mathop{\rm area}( B(p,R)\cap L_t)$.

It remains to choose $t$ and $R$ so that $\mathop{\rm area}( B(p,R)\cap L_t)>0$; $R=2$ and $t=-1$ will do the job.

  • 1
    $\begingroup$ P.S. I realized that there is yet simpler proof. Take a point $q$ on the distance $2{\cdot} r$ from $p$. Denote by $W_r$ the set of all minimizing geodesics from $q$ to the points in $B_\varepsilon(p)$. The standard comparison gives the lower bound on the area of $W_r\cap\partial B_r(p)$. $\endgroup$ – Anton Petrunin Aug 18 '13 at 21:52
  • $\begingroup$ can you give more detail of how to estimate $W_r$? $\endgroup$ – J. GE Oct 6 '13 at 22:11
  • $\begingroup$ @user3922, Consider the radial map $f\colon W_r\to \partial B_{2r}(q)$. The lower Ricci urvature gives a bound for the $(n-1)$-Jacobian of $f$. It remains to apply the coarea formula. $\endgroup$ – Anton Petrunin Oct 7 '13 at 1:56

This is clearly false, just consider the cylinder

$$ R_t \times S_{\theta} $$

with the product metric

$$g_\alpha=dt^2+\alpha^2 d\theta^2.$$

This is a flat metric so $Ric_{g_\alpha} = 0$. On the other hand, for $r>>\alpha$, it is easy to see $Area(\partial B_r)<8\pi \alpha$. Since $\alpha$ is arbitrary there is no uniform lower bound.

Maybe you need a uniform lower bound on the injectivity radius? (I'm not an expert on comparison geometry so don't know off the top of my head if this would suffice) [Edit: Or maybe this can only happen if the metric splits off an isometric euclidean factor].

[As an aside I can't seem to get math blackboard fonts to work anyone else have a problem with this?]

  • 2
    $\begingroup$ The constant $c=c(p,R)$ has to depend on the manifold of course. $\endgroup$ – user16750 Oct 9 '11 at 1:14

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.