Generalised eigenfunction expansion and singularity expansion methods for two-dimensional acoustic time-domain wave scattering problems

TL;DR

This paper develops and compares generalized eigenfunction expansion and singularity expansion methods for solving two-dimensional acoustic wave scattering problems, providing a framework for efficient time-domain solutions in complex geometries.

Contribution

It introduces a discrete GEM approach for numerical approximation and a regularisation technique for SEM coefficients, extending applicability to arbitrary scatterer geometries.

Findings

SEM converges rapidly to GEM solution over time

Discrete GEM simplifies computations to matrix multiplications

Regularisation of SEM coefficients via analytic continuation

Abstract

Time-domain wave scattering in an unbounded two-dimensional acoustic medium by sound-hard scatterers is considered. Two canonical geometries, namely a split-ring resonator (SRR) and an array of cylinders, are used to highlight the theory, which generalises to arbitrary scatterer geometries. The problem is solved using the generalised eigenfunction expansion method (GEM), which expresses the time-domain solution in terms of the frequency-domain solutions. A discrete GEM is proposed to numerically approximate the time-domain solution. It relies on quadrature approximations of continuous integrals and can be thought of as a generalisation of the discrete Fourier transform. The solution then takes a simple form in terms of direct matrix multiplications. In parallel to the GEM, the singularity expansion method (SEM) is also presented and applied to the two aforementioned geometries. It expands the time-domain solution over a discrete set of unforced, complex resonant modes of the scatterer. Although the coefficients of this expansion are divergent integrals, we introduce a method of regularising them using analytic continuation. The results show that while the SEM is usually inaccurate at $t=0$, it converges rapidly to the GEM solution at all spatial points in the computational domain, with the most rapid convergence occurring inside the resonant cavity.