Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Suppose $\mathcal{M}$ is a Riemannian manifold with metric $g$ and with the volume measure $d\mu$ (induced by $g$). Let $f:\mathcal{M}\to \mathbb{R}$ be $C^{\infty}$. Is it true that $$\max_{\mathcal{M}}f-\min_{\mathcal{M}}f\leq\int_{\mathcal{M}}|\nabla f|d\mu \;?$$ Here $|\nabla f|$ is the norm of the gradient of $f$ with respect to the metric $g.$

Edit 1: Following the comments I'd like to know if the following inequalities hold $$(\max_{\mathcal{M}}f-\min_{\mathcal{M}}f)\operatorname{diam}{({\mathcal{M}})}\leq\int_{\mathcal{M}}|\nabla f|d\mu \;?$$ Edit 2:

$$(\max_{\mathcal{M}}f-\min_{\mathcal{M}}f)\operatorname{Vol}{(\mathcal{M})}\leq\int_{\mathcal{M}}|\nabla f|d\mu \;?$$

share|cite|improve this question
A physical intuition is helpful here: your inequality has "the wrong units". Surely you need to multiply $\lvert \nabla f \rvert$ by some kind of length, like $\operatorname{diam}(\mathcal M)$. – Rahul Aug 15 '11 at 18:41
yes I just found it I added it as a comment. – timhortons Aug 15 '11 at 18:46
Regarding your edit: You still have the wrong units ($f$ times length on the left, $f$ by length times volume on the right), and the answer is still no. Willie Wong's answer remains a counterexample to your new inequality, because $\operatorname{diam}(M) \approx 1$. I'd conjecture that $\sup_M f - \inf_M f \leq \operatorname{diam}(M) \sup_M \lvert \nabla f \rvert$ instead. – Rahul Aug 15 '11 at 19:03
I'd suggest that you stop for a moment and think about what the inequality would actually mean, and why you might or might not expect it to be true, instead of throwing in multiplicative factors willy-nilly hoping for something to stick. – Rahul Aug 15 '11 at 19:46
A bit of unsolicited advice: speaking from the point of view of someone who has been through it, I encourage you to pause and try and figure out how I came up with my counterexamples, and think about how @Rahul got his conjecture, before you write your next edit. Sometimes it is this step that helps the most when learning new things. – Willie Wong Aug 15 '11 at 20:10
up vote 5 down vote accepted


Let $M = (0,1)\times (0,\epsilon)\subset \mathbb{R}^2$ with coordinates $(x,y)$ and with the metric inherited from the Euclidean metric. Take $f(x,y) = x$. $|\nabla x| = 1$. $\sup f - \inf f = 1$ (Please note that in general on an arbitrary set $S$ given a function $f:S\to\mathbb{R}$, the notion $\max_S f$ is generally not defined, since you cannot guarantee that $f$ will attain its maximum on $S$). But $\int_M |\nabla f| d\mu = \epsilon$ which can be made arbitrarily small.

Response to edits 1 & 2

  1. My counter example above is still a counterexample for Edit 1.
  2. For a counter example to edit 2: observe that it is false even in one dimension. Let $f$ be an arbitrary non-negative bump function supported on (-1,1). Consider the manifolds $M_n = (-n,n)$. The RHS stays constant, but the LHS can grow to as large as you want.

Remark: What you are writing down is getting closer and closer to Sobolev/Poincare inequalities. For more details on the scaling/dimensional analysis argument, you may want to read Terry's answer here and his linked blog post.

share|cite|improve this answer
thanks for directing me towards Sobolev inequalities that is what i wanted indeed. – timhortons Aug 15 '11 at 20:26

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.