0
$\begingroup$

I am reading Partial differential equations written by Jurgen Just (Third edition) chapter 4, and I want to do some exercises. I think that I need some hints to solve the problem 4.2:

Let $x_0 \in \Omega_h$ have neighbours $x_1,\ldots ,x_{2d}$.We consider a difference operator Lu for u:$\bar\Omega_h\rightarrow\Bbb R$,

$\hspace{6cm}$$Lu(x_0):=\sum_{\alpha=0}^{2d}b_{\alpha}u(x_{\alpha}),$

satisfying following assumptions:

$\hspace{4cm}b_{\alpha}\geq0 \hspace{0.2cm}for \hspace{0.2cm}{\alpha=1,\ldots,2d}, \hspace{0.2cm}\sum_{\alpha=1}^{2d}b_{\alpha}>0,\hspace{0.2cm}\sum_{\alpha=0}^{2d}b_{\alpha}\leq0.$

Prove the weak maximum principle: $Lu\geq0 \hspace{0.2cm}in\hspace{0.2cm} \Omega_h$ implies,

$$\max_{\Omega_h} u\leq \max_{\Gamma_h} u$$.

Note that in the above ${\Omega_h}$ is discretization of ${\Omega}$ and ${\Gamma_h}$ is the boundary of ${\bar\Omega_h}$. (c.f. pages 59, 60 of Partial differential equations written by Jurgen Just, Third edition)

Thanks in advance...

$\endgroup$
1
$\begingroup$

Let $x_0$ be an interior point, i.e. $x_0\in\Omega_h$. By hypothesis

$$Lu(x_0)=\sum_{\alpha=0}^{2d} b_{\alpha}u(x_{\alpha})\geq0,$$ hence \begin{equation} -b_0u(x_0)\leq\sum_{\alpha=1}^{2d} b_{\alpha}{u(x_{\alpha})}.\quad{(1)} \end{equation}

On the other hand by hypothesis $$\sum_{\alpha=0}^{2d} b_{\alpha}\leq0,\, \sum_{\alpha=1}^{2d} b_{\alpha}>0.$$

hence $$\left\{0<\sum_{\alpha=1}^{2d} b_{\alpha}\leq-b_0\atop {1\over-b_0}\leq{1\over \sum_{\alpha=1}^{2d} b_{\alpha}}\quad (2)\right.$$

From (1) and (2) we conclude $$u(x_0)\leq {1\over -b_0}{\sum_{\alpha=1}^{2d} b_{\alpha}u(x_{\alpha})}\leq {\sum_{\alpha=1}^{2d} b_{\alpha}u(x_{\alpha})\over \sum_{\alpha=1}^{2d} b_{\alpha}}$$therefore $$u(x_0)\leq max_{\alpha=1,\ldots,2d} u(x_{\alpha})$$

with equality only if $$u(x_0)=u(x_\alpha)\quad for\, all\quad \alpha\in\{1,\ldots,2d\}.$$

Thus, if u assumes an interior maximum at a vertex $x_0$, it does so at all neighbors of $x_0$ as well, and repeating this reasoning, then also at all neighbors of neighbors, etc. Since $\Omega_{h}$ is discretely connected by assumption, $u_h$ has to be constant in $\bar\Omega_{h}$. Otherwise the problem is trivial.

$\endgroup$

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.