Domain decomposition for quasi-periodic scattering by layered media via robust boundary-integral equations at all frequencies
Carlos P\'erez-Arancibia, Stephen Shipman, Catalin Turc, Stephanos, Venakides

TL;DR
This paper introduces a robust domain decomposition method for scalar wave scattering in layered media, effectively handling all frequencies including Wood frequencies through advanced boundary-integral formulations and shifted Green functions.
Contribution
The paper presents a novel DDM approach using stable boundary-integral equations and shifted Green functions to ensure convergence at all frequencies, including problematic Wood frequencies.
Findings
Stable computations of RtR operators across all frequencies.
Successful application to multi-layered media in 2D and 3D.
Efficient solution of large-layer systems with recursive Schur complements.
Abstract
We develop a non-overlapping domain decomposition method (DDM) for scalar wave scattering by periodic layered media. Our approach relies on robust boundary-integral equation formulations of Robin-to-Robin (RtR) maps throughout the frequency spectrum, including cutoff (or Wood) frequencies. We overcome the obstacle of non-convergent quasi-periodic Green functions at these frequencies by incorporating newly introduced shifted Green functions. Using the latter in the definition of quasi-periodic boundary-integral operators leads to rigorously stable computations of RtR operators. We develop Nystr\"om discretizations of the RtR maps that rely on trigonometric interpolation, singularity resolution, and fast convergent windowed quasi-periodic Green functions. We solve the tridiagonal DDM system via recursive Schur complements and establish rigorously that this procedure is always completed…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
