Assume that $\Omega$ is a bounded connected domain and $\partial \Omega \in C^{\infty}$. Denote $\Gamma_1,\Gamma_2,\cdots,\Gamma_n$ are $n$ connected components of $\partial\Omega$. This notation leads to $\partial \Omega=\cup^n_{i=1}\Gamma_i$. Consider the following problem. \begin{cases} \Delta \phi&=0\\ \phi|_{\partial \Omega}&=g(x),\quad g\in C^{\infty} \end{cases} Denote $\mathcal{S}$ is subset of $\{1,2,\cdots,n\}$ and $\mathcal{S}^c$ is $\{1,2,\cdots,n\}\setminus \mathcal{S} $. If we know that $$ \min_{i\in \mathcal{S}}\inf_{x\in \Gamma_i}\phi(x)\ge \max_{i\in \mathcal{S}^c} \sup_{x\in \Gamma_i} \phi(x)+\delta, $$ where $\delta$ is a positive constant, how to prove that \begin{align*} \sum_{i\in \mathcal{S}}\int_{\Gamma_i}\nabla \phi\cdot \vec{n}d\sigma>0? \end{align*} where $\vec{n}$ denotes the outer normal of $\partial \Omega$.

My effort: I meet this question when I read a paper. This paper say that it is a standard comparison principle exercise, but I still don't know how to solve this question. When $g(x)$ is a step function, we may consider Hopf Lemma or strong maximum principle. However, $g$ is a function. Any comments will be welcome. the equation (2.51) of page 2992 in this paper


1 Answer 1


Let $\psi$ be harmonic in $\Omega$, with $\psi=\phi$ on $\cup _{i \in S} \Gamma_i$, $\psi=m$ on $\cup_{i \not \in S}\Gamma_i$, where $m=\max_{i \not \in S} \max_{\Gamma_i} \phi$. By comparison, $\psi \geq \phi$ and then $\nabla \psi \cdot n \leq \nabla \phi \cdot n$ on $\cup _{i \in S} \Gamma_i$. Then $$ 0=\int_{\partial \Omega}\nabla \psi\cdot n=\sum_{i \in S}\int_{\partial \Gamma_i} \nabla \psi\cdot n+\sum_{i \not \in S}\int_{\partial \Gamma_i} \nabla \psi\cdot n. $$ The function $\psi$ attains its minimum at any point of $\cup_{i \not \in S} \Gamma_i$, hence Hopf's lemma gives $\nabla \psi\cdot n<0$ and each integral on $\partial \Gamma_i, i \not \in S$ is negative. Then $$ \sum_{i \in S}\int_{\partial \Gamma_i} \nabla \phi\cdot n \geq \sum_{i \in S}\int_{\partial \Gamma_i} \nabla \psi\cdot n >0. $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.