The Dirichlet Problem for the Laplacian in Lipschitz Domains Revisited

Technical Summary: The Dirichlet Problem for the Laplacian in Lipschitz Domains Revisited

Problem Statement
This work addresses the Dirichlet problem for the Laplacian in bounded Lipschitz domains $\Omega \subset \mathbb{R}^N$ ($N \ge 2$):
$$\begin{cases} -\Delta u = f & \text{in } \Omega, \\ u = g & \text{on } \Gamma = \partial\Omega. \end{cases}$$
The authors focus on data $f$ and $g$ belonging to appropriate fractional Sobolev spaces, specifically examining the limit cases $s=1/2$ and $s=3/2$. While the problem has been extensively studied since the 1960s (notably by Lions-Magenes for smooth domains and Grisvard for $C^{r,1}$ domains), the behavior in general Lipschitz domains remains a subject of debate. The paper specifically targets the maximal regularity of solutions, the definition of traces for non-smooth functions, and the uniqueness of solutions in $L^p$-based Sobolev spaces.

Methodology and Functional Framework
The authors employ a functional analytic approach grounded in interpolation theory, duality arguments, and the study of harmonic kernels. Key methodological components include:

1. Redefinition of Traces: The paper moves away from the non-tangential trace notion that has dominated literature since the 1980s. Instead, it rigorously defines traces for functions in $H^{1/2}(\Omega)$ and $H^{3/2}(\Omega)$ by introducing the functional space:
   $$E(\nabla; \Omega) = \{ v \in H^{1/2}(\Omega); \nabla v \in [H^{1/2}(\Omega)]' \}.$$
   This space allows for the definition of a continuous trace operator $\gamma: E(\nabla; \Omega) \to L^2(\Gamma)$.

2. Norm Equivalences and Interpolation: The authors establish new equivalences of norms involving weighted gradients and dual spaces. They utilize the interpolation of subspaces (referencing Ivanov-Kalton and Asekritova-Cobos-Kruglyak) to analyze the range of the Laplacian operator between fractional Sobolev spaces.

3. Counter-Examples and Explicit Constructions: To challenge prevailing claims, the authors construct explicit counter-examples using domains with specific geometric singularities (e.g., polygons with large interior angles or "cracked" domains). They utilize explicit harmonic functions (such as $z(r, \theta) = (r^{-a} - r^a)\sin(a\theta)$) to demonstrate the non-triviality of harmonic kernels in $W^{s,p}_0(\Omega) \cap \mathcal{H}$ for certain ranges of $p$.

4. Area Integral Estimates: The paper revisits the classical Area Integral Estimate (Dahlberg, Kenig, Pipher, Verchota) which relates the $L^2$ norm of a harmonic function on the boundary to its area integral in the interior. The authors provide a counter-example showing this estimate fails in its stated form for general Lipschitz domains.

Key Contributions and Results

• Characterization of $H^{1/2}_{00}(\Omega)$: The authors prove that the kernel of the trace operator $\gamma$ on the space $E(\nabla; \Omega)$ is precisely $H^{1/2}_{00}(\Omega)$. This provides a new characterization:
  $$H^{1/2}_{00}(\Omega) = \{ v \in H^{1/2}(\Omega); \nabla v \in [H^{1/2}(\Omega)]' \text{ and } v=0 \text{ on } \Gamma \}.$$
  Consequently, the trace operator is well-defined and continuous from $E(\nabla; \Omega)$ to $L^2(\Gamma)$.

• Maximal Regularity $H^{3/2}$: Contradicting claims in the literature since the 1990s (specifically regarding Jerison-Kenig results), the paper proves that maximal regularity $H^{3/2}$ holds for the Dirichlet problem with homogeneous boundary conditions ($g=0$) for all right-hand sides $f$ in the dual space $[H^{1/2}_{00}(\Omega)]'$. Specifically, the operator:
  $$\Delta: H^{3/2}_0(\Omega) \to [H^{1/2}_{00}(\Omega)]'$$
  is an isomorphism. This result relies on the failure of the Area Integral Estimate in general Lipschitz settings, which the authors demonstrate via a counter-example involving a sequence of domains $\Omega_\epsilon$ where the $H^1(\Gamma)$ norm of a harmonic function blows up while the weighted $L^2$ norm of the Hessian remains bounded.

• Uniqueness in $L^p$ Theory: The paper clarifies the conditions for the uniqueness of solutions in $W^{s,p}_0(\Omega) \cap \mathcal{H}$. It establishes that for a bounded Lipschitz domain, there exists a critical exponent $p_0(\Omega) < 2N/(N+1)$ such that the kernel is trivial if $p \ge p_0(\Omega)$ and non-trivial if $1 < p < p_0(\Omega)$. For polygonal domains, $p_0(\Omega)$ is explicitly determined by the largest interior angle $\omega_\star$:
  $$p_0(\Omega) = \frac{2\omega_\star}{\pi + \omega_\star} \quad (N=2).$$
  This corrects previous assertions (e.g., in Jerison-Kenig [28]) that suggested uniqueness holds under broader conditions without accounting for the non-trivial harmonic kernels in Lipschitz domains.

• Reformulation of Nečas Property: The authors extend the classical Nečas property to functions with less regular Laplacians. They prove that if $u \in H^1(\Omega)$ and $\Delta u \in [H^{1/2}(\Omega)]'$, then $u \in H^1(\Gamma)$ if and only if $\partial_n u \in L^2(\Gamma)$. Furthermore, they show that for harmonic functions in $H^{3/2}(\Omega)$, the normal derivative $\partial_n u$ belongs to $L^2(\Gamma)$ provided the domain is of class $C^{1,1}$, but this regularity fails for general Lipschitz domains.

Significance and Claims
The paper claims to resolve fundamental questions regarding the Dirichlet problem in Lipschitz domains that have been "poorly understood" or based on "partially valid" results in the literature.

1. Correction of Literature: The authors assert that the prevailing claim that $H^{3/2}$-regularity is unattainable for data in the dual of $H^{1/2}_{00}(\Omega)$ is incorrect. Their counter-example to the Area Integral Estimate invalidates the arguments used to support that claim.
2. Clarification of Traces: By replacing the non-tangential trace with a functional definition based on the gradient's dual regularity, the paper provides a rigorous framework for handling boundary values of non-smooth functions, leading to a precise characterization of $H^{1/2}_{00}(\Omega)$.
3. Refinement of Uniqueness Criteria: The work highlights that the existence of non-trivial harmonic kernels in $W^{s,p}$ spaces for Lipschitz domains necessitates compatibility conditions on the data $f$ for uniqueness, a nuance often overlooked in previous $L^p$-theory formulations.

The authors conclude that while the results for $C^{1,1}$ domains remain valid, the extension to general Lipschitz domains requires a more delicate analysis of the interplay between domain geometry, harmonic kernels, and the specific functional spaces involved. The paper does not propose new applications but aims to solidify the theoretical foundation for future studies of elliptic problems in non-smooth domains.