I am reading the following paper: Takáč, Peter On the Fredholm alternative for the p-Laplacian at the first eigenvalue. Indiana Univ. Math. J. 51 (2002), no. 1, 187–237.

I need help to understand the following argument (page 193 section 2.1):

$\Omega\subset\mathbb{R}^N$ is a bounded regular domain, $p\in (2,\infty)$. Let $\phi_1$ be the first eigenfunction associated with the problem $-\Delta_p u=f$ and $u\in W_0^{1,p}(\Omega)$, i.e. $$\int|\nabla\phi_1|^{p-2}\nabla\phi_1\nabla v=\lambda_1\int|\phi_1|^{p-2}\phi_1v,\ \forall\ v\in W_0^{1,p}$$

where $\lambda_1>0$ os the first eigenvalue. We can assume that $\phi_1\in C^1(\overline{\Omega})$, $\phi_1>0$ in $\Omega$ and $\frac{\partial\phi_1}{\partial\eta}<0$ in $\partial\Omega$, where $\frac{\partial\phi_1}{\partial\eta}$ represents derivative in the normal direction.

Define in $W_0^{1,p}$ the semi-norm $$\|u\|_{\phi_1}=\Big(\int|\nabla\phi_1|^{p-2}|\nabla u|^2\Big)^{\frac{1}{2}}$$

The author says that in fact, $\|\cdot\|_{\phi_1}$ is an norm because of the following argument: if $v\in W_0^{1,p}$, then

\begin{eqnarray} \lambda_1\int\phi_1^{p-1}v^2 &=& \int|\nabla\phi_1|^{p-2}\nabla\phi_1\nabla(v^2) \nonumber \\ &\leq& 2\int|\nabla\phi_1|^{p-1}|\nabla v||v| \nonumber \\ &\le& 2\|v\|_{\phi_1}\Big(\int|\nabla\phi_1|^pv^2\Big)^{\frac{1}{2}} \end{eqnarray}

I can understand the last two inequalities, and I can use it to prove that $\|\cdot\|_{\phi_1}$ is a norm. The problem is the first equality. To use the characterization of the eigenvalue, we have to take $u\in W_0^{1,p}$. Why does $v^2\in W_0^{1,p}$? I think that he is using this fact, is this true?

  • $\begingroup$ When referring to a 51-page paper, it's not a bad idea to give the page on which the argument is presented. $\endgroup$ – user53153 Feb 18 '13 at 18:23
  • $\begingroup$ Sorry Pavel, I forgot this detail. Take a look in page 193 section 2.1. $\endgroup$ – Tomás Feb 18 '13 at 18:27
  • $\begingroup$ OK, I see that the inequality (4.7) assumes that $v$ is in the domain of some quadratic form, not just an arbitrary $W^{1,p}_0$ function. In general, $v\in W^{1,p}_0$ does not imply $v^2\in W_0^{1,p}$. For example, consider $v(x)=|x|^{-\alpha}-1$ on the unit ball. This function belongs to $W^{1,p}_0$ as long as $p(\alpha+1)<n$. Squaring $v$ will take it out of the Sobolev space if $p(2\alpha+1)\ge n$. $\endgroup$ – user53153 Feb 18 '13 at 18:31
  • $\begingroup$ But the domain of the quadratic form contains $W_0^{1,p}$. $\endgroup$ – Tomás Feb 18 '13 at 18:34
  • $\begingroup$ I see. Then I guess that the validity of the formula $$\int|\nabla\phi_1|^{p-2}\nabla\phi_1\nabla v=\lambda_1\int|\phi_1|^{p-2}\phi_1v$$ extends to $v\in \mathcal D_{\varphi_1}$ by continuity. $\endgroup$ – user53153 Feb 18 '13 at 19:27

To prove that $\Vert u\Vert_{\phi_1} = \left(\int |\nabla \phi_1|^{p-2}|\nabla u |^2\right)^{\frac 12}$ is a norm on $W_0^{1,p}$ rather than a semi-norm, we have to show that $\Vert u \Vert_{\phi_1} = 0$ implies $u=0$.

Assume that $u\in W^{1,p}_0$ satisfies $\Vert u\Vert_{\phi_1}=0$. Define $u_M := \min(|u|, M)$ for $M>0$. Then $\Vert u_M\Vert_{\phi_1}=0$ and $u_M \in W^{1,p}_0\cap L^\infty$. In particular, the latter implies that $u_M^2\in W^{1,p}_0$ (using $|u_M|^p \le |u|^p$ and $ |\nabla u_M^2|^p = 2^p|u_M|^p|\nabla u_M|^p\le 2^pM^p |\nabla u|^p$) The argument at the end of the original post (and our assumption on $u$) now shows that $$\lambda_1 \int \phi_1^{p-1} u_M^2 \le2 \Vert u_M \Vert_{\phi_1} \left(\int |\nabla \phi_1|^p u_M^2\right)^{\frac 12} = 0.$$ This holds for any $M> 0$. Since $u_M^2\nearrow u^2$ as $M\to +\infty$, the dominated convergence theorem implies that $\lambda_1 \int \phi_1^{p-1} u^2 \le 0$. It follows that $u=0$, so $\Vert \cdot \Vert_{\phi_1}$ is a norm.

  • $\begingroup$ I think it is right. Let me think a little bit more about it, then I accept it. Thank you. $\endgroup$ – Tomás Jul 22 '13 at 22:23

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.