Consider $ E(x)$ some smooth function on $ \Omega$ (some smooth bounded domain in $ R^N$) and suppose $E=0$ on $ \partial \Omega$.

Suppose one knows that there is some $C_1,C_2 \in R$ such that $ x \mapsto C_1 | \nabla E(x)|^2 + C_2 \Delta E(x)$ is constant on $ \partial \Omega$. I am interested in what one can conclude about $\Omega$ and (or) $E$.

Suppose $ \Omega$ is a ball centered at the origin and $E$ is radial. Then is satisfies the above. (of course one can take a ball and extend in one more direction and let $E$ be independent of this extra dimension and still arrive at the same thing). Is this the only option?

  • $\begingroup$ Probably one can get the boundary Laplacian of $E$ if one knows the mean curvature of $\partial\Omega$. $\endgroup$
    – Fan Zheng
    Oct 17, 2015 at 17:38

1 Answer 1


For any smooth $\Omega$ and $E$ one can perturb $E$ so that the desired conditions hold. Indeed, let $d$ be a global smooth function agreeing with the distance from $\partial\Omega$ in a neighborhood of the boundary. Let $F$ be any smooth function on the boundary, globally extended so that near the boundary it is constant on normal lines. Finally, consider $G = F(x)d^2(x)/2$. Then $G = |\nabla G| = 0$ on the boundary. Tangential to the boundary, $G$ separates at most quartically from zero so its tangential Laplace is zero, so on the boundary $\Delta G = F$. Choosing $F$ appropriately one sees that $E + G$ satisfies the desired conditions.

  • $\begingroup$ @Fan Zheng, yes I played around with writing the Laplacian on boundary involving the curvature of the boundary but wasn't able to get anything useful. $\endgroup$
    – Math604
    Oct 18, 2015 at 18:18
  • $\begingroup$ @ Connor Mooney. Thank you very much for the answer. Let me adjust the questions slightly (and maybe your answer answers this too... i too dumb to tell). Let $ 0<E$ is smooth and lets assume $E=0$ on the boundary of $ \Omega$. I am curious if one can have a situation like $ C_1 | \nabla E(x)|^2 + C_2 \Delta E(x)$ constant on every level set of $E$ (there the $C_i$ can depend on which level set but no more). thanks $\endgroup$
    – Math604
    Oct 18, 2015 at 18:26
  • $\begingroup$ @Math604: For any $\Omega$, the solution to the eigenvalue problem $\Delta E = -\lambda E$, $E|_{\partial \Omega} = 0$ where $\lambda$ is the smallest eigenvalue would satisfy these conditions with $C_1 = 0$. More generally, for your second question, $E$ would have to solve a PDE of the form $\Delta E + f(E)|\nabla E|^2 = g(E)$. $\endgroup$ Oct 18, 2015 at 18:48
  • $\begingroup$ @ Connor Mooney, Thanks for the answer. I will think more about it. $\endgroup$
    – Math604
    Oct 22, 2015 at 2:47

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.