1
$\begingroup$

Suppose $f$ is a Lipschitz continuous real-valued function over a bounded domain $\Omega \subset \mathbb{R}^d$ with smooth boundary, and let $\overline{f} := \frac{1}{|\Omega|}\int_\Omega f(x) dx$. Is it possible to upper bound $\|f-\overline{f}\|_{L^\infty(\Omega)}$ in terms of $\|\nabla f\|_{L^2(\Omega)}$ using a generalized Poincare inequality or a related inequality? Of course, standard Poincare gives: $$\|f-\overline{f}\|_{L^2(\Omega)} \leq C\|\nabla f \|_{L^2(\Omega)}$$ but can this be "upgraded" to an $L^\infty$-norm bound in terms of $\|\nabla f\|_{L^2(\Omega)}$ using the fact that $f$ is Lipschitz?

This question asks something similar: Bounding supremum norm of Lipschitz function by L1 norm and I think it might be possible to string together an answer from the replies there. But I wonder if there is a more direct approach, using, for example, the Gagliardo-Nirenberg interpolation inequality. However, all versions of Gagliardo-Nirenberg for bounded domains that I have found do not allow for $p=\infty$ norm control (see, e.g., https://arxiv.org/abs/2110.12967).

$\endgroup$
2
  • $\begingroup$ Could you precise a little bit what type of dependence you allow for $C$ ? I mean, if $\nabla f \neq 0$, obviously $C:= \text{diam}(\Omega) \frac{\|\nabla f\|_\infty}{\|\nabla f\|_2}$ does the job .... $\endgroup$ Commented Apr 3 at 20:11
  • $\begingroup$ Yes, I am hoping for a bound of the form: $\|f-\overline{f}\|_\infty \leq C L^p \|\nabla f\|_2^q$ where L is the Lipschitz constant of $f$, $p,q>0$, and $C$ is a constant independent of $f$ but could depend on $\Omega$ and $d$. Also, it doesn't have to be a power of $L$ and $\|\nabla f\|_2$. Any increasing functions of these quantities would suffice, too. $\endgroup$
    – Greg O.
    Commented Apr 3 at 21:12

1 Answer 1

3
$\begingroup$

For $p>\max(2,d)$ you have by Sobolev embedding $$\|f-\overline{f}\|_\infty \lesssim_{\Omega,p} \|f-\overline{f}\|_p + \|\nabla f\|_p.$$ The interpolation $L^p(\Omega) = [L^2(\Omega),L^\infty(\Omega)]_\theta$ with $\theta\in(0,1)$ entails $$\|f-\overline{f}\|_\infty \lesssim_{\Omega,p} \|f-\overline{f}\|_2^\theta \|f-\overline{f}\|_\infty^{1-\theta} + \|\nabla f\|_2^\theta\|\nabla f\|_\infty^{1-\theta}.$$ Now, you can absorb part of the first term of r.h.s. in the l.h.s. with Young's inequality $a b \leq \theta a^{1/\theta} + (1-\theta) b^{1/(1-\theta)}$ : $$\|f-\overline{f}\|_\infty \lesssim_{\Omega,p} \|f-\overline{f}\|_2 + \|\nabla f\|_2^\theta \|\nabla f\|_\infty^{1-\theta}.$$ The usual Poincaré-Wirtinger inequality allows to write $$\|f-\overline{f}\|_\infty \lesssim_{\Omega,p} \|\nabla f\|_2 + \|\nabla f\|_2^\theta \|\nabla f\|_\infty^{1-\theta},$$ and this implies what you hope for, because you can write (but this is rough ...) $$\|\nabla f\|_2 = \|\nabla f\|_2^\theta \|\nabla f\|_2^{1-\theta}\leq \|\nabla f\|_2^\theta \|\nabla f\|_\infty^{1-\theta}.$$

$\endgroup$
2
  • $\begingroup$ I wrote it using the hilbertian Poincaré-Wirtinger inequality as you referred to it, but you could directly use it on $W^{1,p}(\Omega)$ for $p>\max(2,d)$ right after the Sobolev embedding, and then directly interpolate $\nabla f$ between $L^2(\Omega)$ and $L^\infty(\Omega)$ avoiding the absorption argument with Young's inequality and probably getting a better constant than this rough estimate ... $\endgroup$ Commented Apr 4 at 14:55
  • $\begingroup$ Looks good, thanks! Not too worried about constants, but yes, I see what you mean about using P-W inequality to simplify the argument. $\endgroup$
    – Greg O.
    Commented Apr 4 at 17:52

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.