The square root problem for second-order, divergence form operators with mixed boundary conditions on \(L^p\) (Q2351630)

This paper contains a unifying approach to the characterization of the domain of the operator \(A_0^{1/2}:=\sqrt{-\nabla\cdot\mu\nabla+\text{id}}\) in \(L^p(\Omega)\), \(1<p\leq 2\), (this is the so-called \textit{Kato square-root problem}) for some bounded and open domain \(\Omega\subset\mathbb R^n\). The coefficient function \(\mu\) is only assumed to be real, bounded, measurable and elliptic. The aim is to identify the domain as a suitable Sobolev space \(W_{D}^{1,p}(\Omega)\); \(D\) is a closed subset of \(\partial\Omega = \overline{\Omega}\setminus\Omega^\circ\), typically denoting the `Dirichlet part' of the boundary; this means that the above set-up allows for Dirichlet \(D=\partial\Omega\), Neumann \(D=\emptyset\) and mixed boundary conditions. The Sobolev space \(W_F^{1,p}(\Omega)\), with \(F\subset\overline{\Omega}\) being a closed subset, is defined as the closure of the set \[ C_F^\infty(\Omega) := \{\psi|\Omega\,:\, \psi\in C_c^\infty(\mathbb R^n),\;\text{supp}\,\psi\cap F=\emptyset\} \] with respect to the norm \(\psi\mapsto\left(\int_\Omega (|\nabla\psi|^p+|\psi|^p)\,dx\right)^{1/p}\). The dual space is in the sense of \(L^2\)-scalar products (Lax' negative norm trick), i.e., it is the space of continuous antilinear forms. The main result of the present paper is the following. Assume that {\parindent=6mm \begin{itemize}\item[(1)] \(1<p\leq 2\); \item[(2)] \(\Omega\) is open and bounded, \(D\subset\partial\Omega\) is closed, \(\overline{\partial\Omega\setminus D}\) is Lipschitz; \item[(3)] \(D=\emptyset\) or \(D\) is a \(n-1\) set, i.e., the \(n-1\)-dimensional Hausdorff measure of \(D\cap B_{r}(x)\), \(B_r(x)\) being a ball in \(\mathbb R^n\) is comparable to \(r^{n-1}\); \item[(4)] \(\mu:\Omega\to\mathbb R^{d\times d}\) is Lebesgue measurable, bounded and a.s.\ uniformly elliptic; \item[(5)] \(A_0^{1/2}:W_D^{1,2}(\Omega)\to L^2(\Omega)\) is a topological isomorphism (i.e.,\ \(W^{1,2}_D(\Omega)\) is the \(L^2\)-domain of \(A_0^{1/2}\)). \end{itemize}} Then \(A_0^{-1/2}:L^p(\Omega)\to W^{1,p}_D(\Omega)\) and \(A_0^{1/2}:W_D^{1,p}\to L^p(\Omega)\) are continuous maps.