2
$\begingroup$

I have some question on global boundedness of weak solution to Neumann problems.

Let $u\in W^{1,2}(\Omega)$ is a weak solution for Neumannn problem $$ \mathrm{div} (A \nabla u )= \mathrm{div}\, F\quad \text{in } \Omega,\quad A\nabla u\cdot \nu = F \cdot \nu$$ i.e., $u$ satisfies $$ \int_{\Omega} A\nabla u\cdot \nabla \phi\, dx =\int_{\Omega} F \cdot \nabla \phi\, dx $$ for all $\phi \in W^{1,2}(\Omega)$. Here $A(x)=(a^{ij}(x))$ is uniformly elliptic, i.e., $$ \Vert A \Vert_{L^\infty (\Omega)} \leq B, \quad a^{ij}(x) \xi_i \xi_j \geq \delta |\xi|^2$$ for all $x\in \Omega$ and $\xi \in \mathbb{R}^n$.

I want to show that if $F \in L^q(\Omega)$, $q>n$, then $u\in L^\infty(\Omega)$ and $$ \Vert u \Vert_{L^\infty(\Omega)} \leq C \Vert F \Vert_{L^q(\Omega)} $$ for some constant $C>0$.

In the case of Dirichlet problem, we can prove similar conclusion via De Giorgi type argument due to Stampaccia and Moser type argument. It seems that the result is quite classical, but I cannot find exact reference on this result. I tried to prove it via Moser type argument, but I fail to conclude the desired result.

Thanks in advance.


Add : Here is my trial (not rigorous for the check whether I can use Moser's approach) Set $w=u^+$. Note that $$ w^\beta \nabla w = \frac{1}{\beta+1} \nabla(w^{\beta+1})$$ Testing $\phi=w^{\beta}$, we obtain via Young's inequality that $$ \frac{4}{(\beta+1)^2} \int_\Omega |\nabla w^{(\beta+1)/2}|^2 dx =\int_\Omega w^{\beta-1} |\nabla w|^2 dx \leq \int_\Omega w^{\beta-1} |F|^2 dx.$$

Using Sobolev's embedding theorem and Holder's inequality we get \begin{align*} \Vert w^{(\beta+1)/2} \Vert_{L^{2^*}(\Omega)} &\leq C\Vert w^{(\beta+1)/2} \Vert_{L^2(\Omega)}+C\Vert \nabla(w^{(\beta+1)/2}) \Vert_{L^2(\Omega)}\\ &\leq C\Vert w\Vert_{L^{(\beta+1)}}^{(\beta+1)/2} + C(\beta+1) \Vert{w^{\beta-1}}\Vert_{L^{p/(p-2)}(\Omega)}^{1/2} \Vert F \Vert_{L^p(\Omega)}^{1/2} \end{align*} I'm not sure to conclude the iteration due to the term $\Vert w\Vert_{L^{(\beta+1)}}^{(\beta+1)/2}$.

$\endgroup$
6
  • 1
    $\begingroup$ You are right, this is usually stated in the case of Dirichlet boundary condition. But what is the problem trying to repeat the usual proof? In the case where $\Omega$ is the whole space $R^n$ both De Giorgi's approach or Moser's work. $\endgroup$ Commented Jul 11, 2020 at 15:08
  • $\begingroup$ @GiorgioMetafune Thanks for your answer. I edit my question. In the case of Dirichlet boundary problem, our function is lie on $W^{1,2}_0$, but in the case of Neumann problem, the original function does not lie on that space. This makes me hard to use Sobolev embedding. In the case of De Giorgi approach, it seems okay to obtain the desired result, but I'm not sure whether the desired assertion can be deduced via Moser's argument. $\endgroup$
    – Will Kwon
    Commented Jul 11, 2020 at 22:08
  • 1
    $\begingroup$ What kind of regularity do you assume for the boundary of $\Omega$? (I have serious doubts whether this is true for arbitary bounded domains.) $\endgroup$ Commented Jul 11, 2020 at 22:48
  • $\begingroup$ Take $\|F\|_p=1$ and iterate your inequalities until $\beta>p-1$. Then you can absorbe the first term on the RHS into the second and go on with the usual iteration. I did not check the details but it should work. $\endgroup$ Commented Jul 11, 2020 at 23:01
  • $\begingroup$ @JochenGlueck Of course, we need to assume the regularity of $\Omega$, at least Lipschitz domain to guarantee the Sobolev embedding theorem. $\endgroup$
    – Will Kwon
    Commented Jul 12, 2020 at 7:59

0

You must log in to answer this question.

Browse other questions tagged .