I'm reading a proof from my teacher about the weak principle of the maximum that is,

$$\max_{\overline{\Omega}} u = \max_{\partial\Omega} u$$

when $u$ is harmonic ($\Delta u = 0$) in $\Omega$ and $u\in C^2(\Omega)\ (\overline\Omega)$, and $\Omega$ is an open bounded.

It goes like this:

Let's name $M = \max_{\overline{\Omega}} u$ and $m=\max_{\partial\Omega} u$. It is clear that $m\le M$ because $\partial\Omega\subset\overline\Omega$. Suppose that $m<M$. Let's pick a point $x_0\in\Omega$ such that $u(x_0) = M$

Then we can construct the following function:

$$v(x) = u(x) + \frac{|x-x_0|}{2d^2}(M-m)$$

where $d$ is the diameter of $\overline\Omega$

and we see that

$$v(x_0) = u(x_0) = M$$

For $x\in\partial\Omega$:

$$v(x) \le m+\frac{1}{2}(M-m) = \frac{1}{2}(M+m) < M$$

Therefore there exists $x^*\in\Omega$ such that $v(x^*) = \max_{\overline\Omega} v$. Since it's a maximum, $$\Delta v(x^*) \le 0 \tag{1}$$

The main question is: why $v(x^*)$ is a maximum for $v$ in $\overline\Omega$?. The only thing I can conclude is that $v$

Now we'll find a contradiction to $(1)$:

$$\Delta v = \Delta u + \frac{M-m}{2d^2}\Delta(|x-x_0|^2) = \frac{M-m}{2d^2}2N >0 \mbox{ in $\Omega$}$$


I think since $v$ in $\partial\Omega$ has value less than $M$, it's only possible place to attain a maximum is is in $\Omega$. There are still some details on why. For example, what if $v$ is concave downwards, therefore has no maximum in $\Omega$?

  • 1
    $\begingroup$ Is not $v(x) = u(x) + \frac{|x-x_0|^2}{2d^2}(M-m)$? $\endgroup$ – mfl Aug 15 '18 at 21:23
  • $\begingroup$ Note that $v(x_0)=M>v(x), \forall x\in \partial \Omega.$ So $v$ has a global maximum on $\Omega.$ $\endgroup$ – mfl Aug 15 '18 at 21:26

Here's a proof that seems simpler to me. Suppose that $M>m$. Pick $x_0$ with $u(x_0)=M$, such that $x_0$ is at minimal distance to the boundary $$d(x_0,\partial\Omega)=\min\{d(x,\partial\Omega):u(x)=M\}.$$

Now, assuming we're in $\Bbb R^2$ just to simplify the notation, if $0<r<d(x_0,\partial\Omega)$ then $$M=u(x_0)=\frac1{2\pi}\int_0^{2\pi}u(x_0+re^{it})\,dt<M,$$contradiction. (That average is less than $M$ because $u(x_0+re^{it})\le M$ for all $t$, but there is an interval $I$ such that $u(x_0+re^{it})<M$ for $t\in I$, since $d(x_0+re^{it},\partial\Omega)<d(x_0,\partial\Omega)$ for $t\in I$.)


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.