3
$\begingroup$

Let $1 \le p < n$ and let $p^*$ be the Sobolev conjugate of $p$, i.e. $p^* = np/(n - p)$. Let $(\Omega_m)$ be an increasing sequence of bounded, convex and open sets such that $$ \lim_{m \to \infty} \sup \{r : B(0, r) \subset \Omega_m \} = \infty. $$ Assume $u_m \in W^{1, p}(\Omega_m)$, $0 \le u \le 1$ and $$ \lim_{m \to \infty} \frac{1}{|\Omega_m|} \int_{\Omega_m} |u_m|^{p^*} \, dx = 0, \quad \|\nabla u_m\|_{L^p(\Omega_m, \mathbb{R}^n)} \le C, $$ where $C$ does not depend on $m$. Since $(u_m)$ is uniformly bounded in $L^\infty(\mathbb{R}^n)$, we can assume without loss of generality that there exists $u \in \dot{W}^{1, p}(\mathbb{R}^n) \cap L^\infty(\mathbb{R}^n)$ such that $u_m \overset{\ast}\rightharpoonup u $ in $L^\infty(\mathbb{R}^n)$. Is it true that $$\lim_{m \to \infty} \frac{1}{|\Omega_m|} \int_{\Omega_m} |u|^{p^*} \, dx = 0 $$ as well?

EDIT: We actually have more than weak$^*$ convergence. By the Rellich-Kondrachov theorem and the uniform boundedness of $(u_m)$ in $L^\infty$, it is true that $u_m \to u$ in $L^q(K)$ for all compact $K$ and all $1 \le q < \infty$.

$\endgroup$
2
  • $\begingroup$ Do you need general $\Omega_m$ or just balls? In the last case seems to be true. $\endgroup$ Commented Aug 31 at 9:53
  • $\begingroup$ @GiorgioMetafune It is enough for me if I can show it for an increasing sequence of rectangles with the ratio of the sides tending to 0 (or, equivalently, infinity). I would still be interested to see your proof! $\endgroup$ Commented Aug 31 at 10:14

1 Answer 1

3
$\begingroup$

Let me do for balls $\Omega_m$ and I use your Edit. Note that $\bar u_m=\frac{1}{|\Omega_m|} \int_{\Omega_m} u_m \to 0$ by H"older inequality and your assumption. Next I use Poincarè-Wirtinger inequality $$ \|u_m-\bar u_m\|_{p^*} \leq C\|\nabla u_m\|_p $$ with a constant independent of the radius of the ball. This is where I need balls or, more general, dilated domains, since Sobolev inequality is scale invariant for these exponents. If you let $m \to \infty$, you get $\|u\|_{p^*} \leq C$ and then the result follows since $|\Omega_m| \to \infty$. Of course this generalizes to the case where the biggest ball $B_m$ satisfies $|B_m| \geq c|\Omega_m|$ but not to the rectangles you are interested in, so that this is only a partial answer.

EDIT Let me complete the proof in the general case. Let $B_m \subset \Omega_m$ be a ball or radius $r_m \to \infty$ and $\bar u_m$ the mean of $u_m$ over $B_m$. If $\bar u_m \to 0$ the argument above gives $u \in L^{p*}$ and the result holds.

If $u_m \to \ell$ (up to a subsequence) then, as above, $u-\ell \in L^{p^*}$ and then $\frac{1}{|\Omega_m|}\int_{\Omega_m}|v|^{p*} \to 0$ with $v=u$ and $v=u-l$. This yields $\ell=0$.

$\endgroup$
5
  • $\begingroup$ The Poincaré constant isn't scale invariant. It scales like $r$, with $r$ being the radius. $\endgroup$ Commented Aug 31 at 11:53
  • 1
    $\begingroup$ This is not Poincarè $p \to p$ but $p \to p^*$. $\endgroup$ Commented Aug 31 at 12:04
  • 1
    $\begingroup$ Oh sorry, I didn't notice the $p^*$. So you actually prove that the limit is in $L^{p^*}$. That's interesting. Thanks! Do you expect the statement to hold more generally? $\endgroup$ Commented Aug 31 at 12:56
  • $\begingroup$ Can you explain how you deduce that $u$ is in $L^{p^*}$? I don't think one can simply take the limit in the Poincaré inequality but perhaps I am missing something. $\endgroup$ Commented Sep 13 at 22:16
  • 1
    $\begingroup$ You may apply Fatou lemma to the sequence $|u_m-\bar u_m|^{p^*} \chi_{B_m}$ or fix a compact $K$ and apply Fatou to $|u_m-\bar u_m|^{p^*}$ in $K$ and then take the supremum on all $K$. $\endgroup$ Commented Sep 14 at 6:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .