From [Grisvard, Thm., p. 126], the BVP \begin{align} -\Delta u &= 0 & \text{in}\ \Omega\\ -\frac{\partial u}{\partial n} + au &= g & \text{on}\ \Gamma\\ \end{align} where $a>0$ and $g \in H^{1/2}(\Gamma)$ has unique solution $u$ with $H^2$ regularity for open bounded domain $\Omega \subset \mathbb{R}^2$ with boundary $\Gamma$ of class $C^{1,1}$.

Question Let $a_1, a_2 >0$, $g_1 \in H^{1/2}(\Gamma_1)$ and $g_2\in H^{1/2}(\Gamma_2)$ where $\Gamma_1 \subset \Gamma$ and $\Gamma_2 = \Gamma\setminus \bar{\Gamma}_1$. Does the unique solution $u \in H^1(\Omega)$ of the BVP \begin{align} -\Delta u &= 0 & \text{in}\ \Omega\\ -\frac{\partial u}{\partial n} + a_1u &= g_1 & \text{on}\ \Gamma_1\\ -\frac{\partial u}{\partial n} + a_2u &= g_2 & \text{on}\ \Gamma_2 \end{align} have higher regularity (in particular, is $u\in H^2(\Omega)$) for domain $\Omega$ of class $C^{1,1}$?

I know $u \in H^2(\Omega)$ if $\Gamma_1$ and $\Gamma_2$ are both closed, i.e., $\Omega$ is an annular domain. But, I do not know if I still have an $H^2$ regularity for the case given above.


1 Answer 1


Without further assumptions, no. Consider the case $a_1=a_2=0$ and $\Omega$ smooth. If $u\in H^2(\Omega)$, then $\partial u/\partial n$ is in $H^{1/2}(\Gamma)$ by the trace theorem. However, $g_1\in H^{1/2}(\Gamma_1)$, $g_2\in H^{1/2}(\Gamma_2)$ does not imply $g\in H^{1/2}(\Gamma)$ (a step function is not in $H^{1/2}$).

  • $\begingroup$ Thank you for you answer. Can you please state what must be the additional assumption for $u$ to be in $H^2$, and would it be possible for the case $a_1, a_2 >0$? $\endgroup$ May 25, 2018 at 1:00

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.