Gaining Regularity of Weak Solutions by Choosing Proper Boundary Conditions

118 Views Asked by At

Abstract Question

Do you know results on consistency conditions on boundary conditions for PDEs in order to keep/gain differentiability when combining strong solutions of a PDE on different domains that share a boundary?

A Concrete Question

Because this formulation is too general and a little ambiguous, I'd like to ask this question for a specific problem. However, hints on general ways to solve such problems still are very welcome, too. Here it is:

Let $\Omega_1 = B(1) \subseteq \mathbb{R}^2$ (two-dimensional unit circle), $\Gamma_1 = \partial \Omega_1$, $\,f,g \in C^\infty(\Gamma_1)$ and $u_1 \in H^2(\Omega)$ the unique weak solution of $$ \Delta^2 u = 0, \quad u|_{\Gamma_1} = f, \quad \partial_n u|_{\Gamma_1} = g $$ where $\Delta^2 u = \Delta \Delta u$ denotes the Bi-Laplacian.

Furthermore, let $\Omega_2 = B(2) \setminus \Omega_1$ (annulus), $\Gamma_2 = \partial B(2)$ where $u_2 \in H^2(\Omega_2)$ denotes the unique weak solution of $$ \Delta^2 u = 0, \quad u|_{\Gamma_1} = f, \quad \partial_n u|_{\Gamma_1} = g, \quad u|_{\Gamma_2} = 0, \quad \partial_n u|_{\Gamma_2} = 0. $$

Define $\Omega = B(2)$ and $$ u\colon \Omega \longrightarrow \mathbb{R}~~~~~~~~~~~~~~~~~~ \\ ~~~~~~~~~~~~~~~~~x \longmapsto \begin{cases} u_1(x) \text{ for } x \in \Omega_1 \\ u_2(x) \text{ for } x\in \Omega_2.\end{cases} $$ We have $u_1 \in H^4(\Omega_1)$ and $u_2 \in H^4(\Omega_2)$. Under what conditions on $f,g$ is $u \in H^3(\Omega)$? Or, under what conditions on $f,g$ can't hold $u\in H^3(\Omega)$?

Some Thoughts on the Problem

I encountered this problem when dealing with solutions of PDEs where the respective domains share some boundary. At first, I thought that I could easily combine these two solutions to a big one, but this is only partially true. To illustrate this, in the given example above we have that $u\in H_0^2(\Omega)$ (because of the boundary conditions) and that $u$ satisfies $$ \Delta^2 u = 0, \quad u|_{\Gamma_1} = f, \quad \partial_n u|_{\Gamma_1} = g $$ in a weak sense. So combining weak solutions works. However, while you can show that $u_1 \in H^4(\Omega_1)$ and $u_2 \in H^4(\Omega_2)$ from regularity theory in order to get strong solutions (in a sense), you will in general have $u \notin H^4(\Omega)$ because the boundary conditions do not necessarily match.

If we ask for conditions on $f,g$ which imply $u\in H^4(\Omega)$ the answer maybe is kind of trivial because you probably end up with $f=g=0$ (although I did not show this yet, so I might be wrong there). But for some theory I might still want that at least $u\in H^3(\Omega)$ holds (or $u\in H^{\frac{7}{2}}(\Omega)$ would be awesome). I already experimented a little with simpler situations but I couldn't find a pattern yet. Has anybody an idea or knows how to treat this problem? Has this maybe been done already and somebody knows where to find it?