Let $U$ be a bounded, open subset of $\mathbb{R}^n$. Prove that there exists a constant $C$, depending on only $U$, such that $$\max_{\bar{U}} |u|\le C(\max_{\partial U} |g|+\max_{\bar{U}} |f|)$$ wherever $u$ is a smooth solution of \begin{cases}-\Delta u=f & \text{in }U\\ \quad \, \, \, u=g & \text{on } \partial U \end{cases}

This is from PDE Evans (2nd edition), Chapter 2 Exercise 6. (Note that $\bar{U}$ is the closure of $U$.)

First, I tried proving for the boundary case, that is, suppose $x \in \partial U$. Then $u=g$. We have $$|u|=|g|\le \max_{\partial U} |g| \le \max_{\partial U}|g|+\max_{\bar{U}}|f|.$$ Hence, $|u|$ is bounded by $\max_{\partial U}|g|+\max_{\bar{U}}|f|$. So there exists a $C > 0$ such that $|u| \le C(\max_{\partial U}|g|+\max_{\bar{U}}|f|)$. Thus, $$\max_{\partial U} |u| \le C(\max_{\partial U}|g|+\max_{\bar{U}}|f|),$$ as required.

Now, how can I prove for the interior case, that is, for $x \in U$? I wanted to use strong maximum principle, but that applies only to harmonic solutions, so I can't use that here.


Here is a hint. You already noticed that we want to use Maximum principle and Maximum principle only happens on harmonic functions.

Let's divide your problem into too cases. Namely,

\begin{cases}-\Delta u_1=0 & \text{in }U\\ \quad \, \, \, u_1=g & \text{on } \partial U \end{cases}

\begin{cases}-\Delta u_2=f & \text{in }U\\ \quad \, \, \, u_2=0 & \text{on } \partial U \end{cases}

You already have the estimation of $u_1$ by M-P. Now we work on $u_2$.

Since $U$ is bounded, let's assume that $U\subset \{x\in R^N,\,\, |x\cdot \xi|\leq d\}$ where $|\xi|=1$ and $d>0$ is an constant.

Next we define $$ k(x):= e^{Lx\cdot \xi} $$ and we notice that $$ -\Delta k(x)=-L^2\xi_i^2e^{Lx\cdot\xi}=-L^2 e^{Lx\cdot\xi}\leq 1 $$ for large enough $L>0$.

Now we send $$ v(x):= (e^{Ld}-e^{Lx\cdot \xi})\sup_{U}f^+ $$

Now we compute and find out $$ -\Delta v(x)\geq \sup_{U}f^+ $$ and hence we have $$-\Delta (u_2(x)-v(x))\leq f-\sup_U f^+\leq 0 $$ hence we have $u-v$ is subharmonic and by M-P we have $u(x)\leq v(x)$ for all $x\in U$ and hence we have $$ \sup u_2\leq C\sup f^+$$

Now write $u=u_1+u_2$, I am sure you could take it from here.

  • $\begingroup$ So we can use the maximum principle, even if a solution like $v(x)$ is only subharmonic, rather than fully harmonic? $\endgroup$ – Cookie Dec 22 '14 at 4:16
  • 1
    $\begingroup$ This is right. I suggest you to work out an proof yourself, although it is almost equal to the proof of harmonic case $\endgroup$ – spatially Dec 22 '14 at 4:20
  • $\begingroup$ I worked out the proof of the maximum principle as you suggested (and this is Chapter 2, Exercise 5B), and instead of $$M=v(x_0) = \def\avint{\mathop{\,\rlap{-}\!\!\int}\nolimits} \avint_{B(x_0,r)} u \, dy \le M$$ in the harmonic case with $u(x_0)=M:=\max_{\bar{U}} u$ (see page 27 in 2nd edition), we have instead $$M=v(x_0) \color{red}{\le} \def\avint{\mathop{\,\rlap{-}\!\!\int}\nolimits} \avint_{B(x_0,r)} u \, dy \le M$$ in the subharmonic case. Both the statements force $u \equiv M$ to remain true. This is how the maximum principle still holds in both the harmonic and subharmonic cases? $\endgroup$ – Cookie Dec 23 '14 at 6:08
  • $\begingroup$ Yes. This is a nice proof $\endgroup$ – spatially Dec 23 '14 at 6:11

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.