Integral equation methods for scattering by general compact obstacles: wavenumber-explicit estimates

Technical Summary: Wavenumber-Explicit Estimates for Integral Equation Methods in Scattering by General Compact Obstacles

Problem Statement
The paper addresses the exterior sound-soft (Dirichlet) scattering problem for the Helmholtz equation, $\Delta u + k^2 u = 0$, in the presence of a general compact obstacle $O \subset \mathbb{R}^n$ ($n \ge 2$). While classical boundary integral equation (BIE) formulations exist for Lipschitz domains and smooth screens, recent work by Caetano et al. (2025) proposed a novel integral equation (IE) formulation applicable to arbitrary compact obstacles, including those with fractal boundaries or non-Lipschitz geometries. This formulation seeks a solution $u = A_k \phi$ where $\phi$ satisfies an equation $A_k \phi = g$ on a compact set $\Gamma$ such that $\partial O \subset \Gamma \subset O$.

The primary challenge addressed in this paper is the lack of understanding regarding the dependence of the operator norms $\|A_k\|$ and $\|A_k^{-1}\|$ on the wavenumber $k$, particularly in the high-frequency regime ($k \to \infty$). Such estimates are critical for numerical analysis, specifically for determining the condition numbers of discretized systems and ensuring the stability and convergence of numerical solvers (e.g., boundary element methods) as $k$ increases.

Methodology
The authors analyze the operator $A_k$, defined as a composition of the acoustic Newtonian potential $A_k$ (with density $\psi$), a smooth cut-off function $\chi$, and a projection operator $P$ onto a closed subspace of $H^1(\mathbb{R}^n)$. The operator maps $H^{-1}_\Gamma$ (distributions supported on $\Gamma$) to a subspace of $H^1(\mathbb{R}^n)$.

Key methodological steps include:

1. Function Space Framework: The analysis utilizes Sobolev spaces $H^s(\mathbb{R}^n)$ and their duals, equipped with both standard norms and wavenumber-dependent norms $\|\cdot\|_{H^s_k}$. The choice of the projection operator $P$ is shown to influence the specific operator norms, though the solvability of the IE is independent of this choice. The authors focus on the canonical choice of $P$ as an orthogonal projection with respect to the relevant inner products.
2. Resolvent Estimates: The bounds on the inverse operator $A_k^{-1}$ are derived by relating the norm of the inverse to the norms of cut-off resolvents of the Dirichlet Laplacian, specifically $R(k; \Omega)$ (exterior domain) and $R(k; \Omega^-)$ (interior domain $\Omega^- = O \setminus \Gamma$).
3. Complex Wavenumber Technique: To establish upper bounds for the inverse, the authors employ a technique involving complex wavenumbers $\lambda$ where $\text{Im}(\lambda) > 0$. By analyzing the operator at complex frequencies where the problem is coercive, they derive estimates that are then transferred to the real axis, avoiding the spectrum of the Laplacian.
4. d-set Generalization: The framework is extended to the case where $\Gamma$ is a $d$-set (Ahlfors-David regular), allowing for the treatment of fractal obstacles. This involves relating the operator $A_k$ to an integral operator defined via Hausdorff measure.

Key Contributions and Results

• First $k$-Explicit Bounds for First-Kind IEs on General Obstacles: The paper provides the first explicit dependence on $k$ for the norms of first-kind integral operators and their inverses for arbitrary compact obstacles. Previous results were largely restricted to second-kind equations or specific geometries (Lipschitz domains, flat screens).
• Upper Bound on $\|A_k\|$: It is shown that $\|A_k\| \le ck$ for $k \ge k_0$, where $c$ depends only on $k_0$ and the geometry. This bound is sharp when the interior of $\Gamma$ is non-empty.
• Upper Bound on $\|A_k^{-1}\|$: The norm of the inverse is bounded by the norms of cut-off resolvents. Specifically, $\|A_k^{-1}\| \le C k^2 (C_{k,R}(\Omega) + C_k(\Omega^-))$, where $C_{k,R}$ and $C_k$ are resolvent norms.
  • Star-shaped Obstacles: If $\Gamma = O$ and $O$ is star-shaped, $\|A_k^{-1}\| \le Ck$, leading to a condition number $\text{cond}(A_k) \le Ck^2$.
  • General Smooth Obstacles: If $O$ is $C^\infty$ but trapping, the bound can grow exponentially, $\|A_k^{-1}\| \le Ce^{\alpha k}$, which is shown to be sharp for certain strongly trapping geometries.
  • Arbitrary Obstacles (Measure-Theoretic Result): For any compact obstacle, the authors prove that for any $\epsilon > 0$, there exists a set $E \subset [k_0, \infty)$ of Lebesgue measure $m(E) \le \epsilon$ such that outside $E$, $\|A_k^{-1}\| \le C k^{2n+2+\delta}$. This implies that the growth of the inverse norm is at worst polynomial in $k$ if one avoids a set of arbitrarily small measure.
• Lower Bounds and Non-Uniformity: The paper demonstrates that for any increasing unbounded sequence $(a_m)$, there exists a compact obstacle such that $\|A_{k_m}^{-1}\| \ge a_m$ for a sequence of wavenumbers $(k_m)$. This highlights that without geometric restrictions (like star-shapedness) or measure-theoretic exclusions, the inverse norm can grow arbitrarily fast.
• Implications for Classical Single-Layer BIOs: As a corollary, the authors derive the first $k$-explicit bounds for the classical acoustic single-layer boundary integral operator $S_k$ on the boundary of a Lipschitz domain and for screens.
  • For Lipschitz boundaries, $\|S_k\| \le Ck$ and $\|S_k^{-1}\|$ satisfies bounds analogous to those for $A_k^{-1}$.
  • For $L^2$ norms, $\|S_k\|_{L^2} \le C k^{(n-3)/2}$ for $n=2,3$, and $\|S_k\|_{L^2} \le C k^\epsilon$ for $n \ge 4$.

Significance
The paper claims significance in several areas:

1. Generality: It extends wavenumber-explicit analysis to the most general class of compact scatterers, including fractals and non-Lipschitz sets, where previous techniques relying on smoothness or specific boundary regularity failed.
2. First-Kind Equations: It addresses the conditioning of first-kind integral equations, which are often preferred for their compact perturbation properties but are notoriously ill-conditioned. The results provide the theoretical foundation for analyzing the convergence of Galerkin methods for these equations at high frequencies.
3. Numerical Analysis: The derived bounds on condition numbers are essential for designing discretization schemes (choice of $N$ relative to $k$) that ensure accurate solutions. The results suggest that for star-shaped obstacles, polynomial growth of the condition number allows for efficient numerical solution, whereas for general trapping obstacles, the growth may be exponential or require careful selection of wavenumbers to avoid resonance peaks.
4. Fractal Scattering: By establishing the framework for $d$-sets, the paper provides the first rigorous $k$-explicit estimates for scattering by fractal obstacles, a topic of increasing interest in acoustics and electromagnetics.

The authors note that while the motivation is numerical analysis, the paper focuses on the analytical bounds, deferring the detailed numerical implementation and specific algorithmic consequences to future work. The results are presented as the first steps toward a comprehensive $k$-explicit numerical analysis for general compact scatterers.