A new, self-contained proof of Shahgholian's theorem using the thickness function

1. Problem Statement

The paper addresses a geometric result concerning quadrature surfaces, originally established by Henrik Shahgholian in 1994. The core problem involves an overdetermined boundary value problem for the Laplacian equation in a bounded domain $\Omega \subset \mathbb{R}^N$:

$$\begin{cases} -\Delta u = \mu & \text{in } \Omega, \\ u = 0 & \text{on } \partial\Omega, \\ \frac{\partial u}{\partial \nu} = -1 & \text{on } \partial\Omega, \end{cases}$$

where $\mu$ is a positive measure compactly supported within $\Omega$. The goal is to characterize the geometry of the domain $\Omega$ given that such a solution $u$ exists. Specifically, the theorem asserts that $\Omega$ satisfies the C-GNP (Convex Geometric Normal Property) with respect to the convex hull of the support of the measure, denoted as $C = \text{conv}(\text{supp}\,\mu)$. This property implies that for every point $x$ on the boundary $\partial\Omega$, the inward normal ray at $x$ intersects the set $C$.

2. Methodology

The author proposes a novel, self-contained proof that replaces the original moving plane method (which relies on sophisticated comparison arguments and technical complexity) with a geometric approach based on level set analysis and the thickness function.

The methodology proceeds through the following logical steps:

• Radial Parametrization: The domain $\Omega \setminus C$ is parametrized using rays originating from the boundary of the convex set $C$. For any point $c \in \partial C$, the outward unit normal is denoted by $\nu(c)$. Any point $x \in \Omega \setminus C$ is uniquely represented as $x = c + r\nu(c)$.
• The Thickness Function: A function $d_t(c)$ is defined as the distance from $c \in \partial C$ to the level set $\Gamma_t = \{x \in \Omega : u(x) > t\}$ along the normal ray $\nu(c)$.
  $$\Gamma_t = \{c + d_t(c)\nu(c) : c \in \partial C\}$$
• Strict Monotonicity: Using the Hopf maximum principle, the author establishes that the solution $u$ strictly decreases along these normal rays outside $C$ (where $\Delta u = 0$). This ensures the existence and uniqueness of the thickness function.
• Rigidity Lemma (The Core Argument): The proof hinges on establishing that the magnitude of the gradient $|\nabla u|$ is identically equal to 1 throughout $\Omega \setminus C$. This is proven in two parts:
  1. Upper Bound ($|\nabla u| \leq 1$): By analyzing $w = |\nabla u|^2$, the author shows $w$ is subharmonic ($\Delta w \geq 0$) in $\Omega \setminus C$. Since the boundary condition imposes $|\nabla u| = 1$ on $\partial\Omega$, the maximum principle forces $|\nabla u| \leq 1$ everywhere inside.
  2. Lower Bound ($|\nabla u| \geq 1$): By considering the directional derivative $v = \nabla u \cdot \nu(c)$ along the rays, the author notes that $v$ is harmonic. Applying the minimum principle to the segment from $C$ to $\partial\Omega$, and using the boundary condition $v = -1$ at $\partial\Omega$, it is shown that $v \geq -1$, implying $|\nabla u| \geq 1$.
• Integration: Combining the bounds yields $|\nabla u| = 1$. This constant gradient magnitude leads to a simple differential relation for the thickness function: $\frac{\partial d_t}{\partial t} = -1$.

3. Key Contributions

• New Proof Technique: The paper provides a self-contained proof that avoids the moving plane method. It relies instead on elementary maximum principles and the specific structure of the level sets.
• Conceptual Clarity: The approach reveals that the overdetermined condition ($\partial u / \partial \nu = -1$) forces a rigid geometric structure where all level sets are parallel surfaces to the convex hull $C$.
• Explicit Formula: The derivation yields an explicit linear relationship for the thickness function:
  $$d_t(c) = d_0(c) - t$$
  This implies that the solution $u$ varies linearly along the normal rays: $u(c + r\nu(c)) = d(c) - r$.
• Regularity Results: The proof naturally establishes that the boundary $\partial\Omega$ (specifically $\partial\Omega \setminus C$) is Lipschitz continuous, as the thickness function $d(c)$ inherits Lipschitz continuity from the distance function to the convex set $C$.

4. Main Results

The paper confirms and refines Shahgholian's theorem with the following specific conclusions:

1. Geometric Structure: The domain $\Omega$ admits a radial parametrization relative to $C = \text{conv}(\text{supp}\,\mu)$.
2. Normal Ray Intersection: For every $x \in \partial\Omega$, the inward normal ray intersects $C$.
3. Parallel Level Sets: All level sets of the solution $u$ are parallel to $\partial C$.
4. Gradient Rigidity: The gradient magnitude is constant ($|\nabla u| = 1$) in the region $\Omega \setminus C$.

5. Significance and Implications

• Simplification: By bypassing the moving plane method, the proof reduces the technical barrier to understanding this geometric theorem, making the underlying mechanism (the rigidity of the gradient) more transparent.
• Generalizability: The framework of the "thickness function" is presented as a versatile tool. The author suggests it can be extended to:
  • Other elliptic operators (e.g., $p$-Laplacian).
  • Higher-order problems (e.g., biharmonic equations).
  • Parabolic problems (time-dependent thickness).
  • General free boundary problems.
• Connection to Optimization: The method naturally connects to isoperimetric inequalities and shape optimization, offering a new perspective on how overdetermined conditions constrain domain geometry.

In conclusion, Barkatou's work offers a more geometric and autonomous derivation of a fundamental result in potential theory, demonstrating that the overdetermined boundary condition essentially forces the domain to be a "parallel body" of the convex support of the source measure.