Accelerated Calder\'on preconditioning for Maxwell transmission problems

TL;DR

This paper introduces techniques to accelerate Calderón preconditioning in boundary integral methods for electromagnetic transmission, significantly reducing computational costs and memory usage in complex scattering problems.

Contribution

The paper presents novel methods for reducing the computational expense of Calderón preconditioning, including using reduced preconditioners and cheaper matrix assembly routines, applicable to Maxwell transmission problems.

Findings

Achieved 99% reduction in memory cost.

Reduced total computation time by at least 75%.

Effective in scattering by multiple dielectric particles.

Abstract

We investigate a range of techniques for the acceleration of Calder\'on (operator) preconditioning in the context of boundary integral equation methods for electromagnetic transmission problems. Our objective is to mitigate as far as possible the high computational cost of the barycentrically-refined meshes necessary for the stable discretisation of operator products. Our focus is on the well-known PMCHWT formulation, but the techniques we introduce can be applied generically. By using barycentric meshes only for the preconditioner and not for the original boundary integral operator, we achieve significant reductions in computational cost by (i) using "reduced" Calder\'on preconditioners obtained by discarding constituent boundary integral operators that are not essential for regularisation, and (ii) adopting a "bi-parametric" approach in which we use a lower quality (cheaper) $\mathcal{H}$-matrix assembly routine for the preconditioner than for the original operator, including a novel approach of discarding far-field interactions in the preconditioner. Using the boundary element software Bempp (www.bempp.com), we compare the performance of different combinations of these techniques in the context of scattering by multiple dielectric particles. Applying our accelerated implementation to 3D electromagnetic scattering by an aggregate consisting of 8 monomer ice crystals of overall diameter 1cm at 664GHz leads to a 99% reduction in memory cost and at least a 75% reduction in total computation time compared to a non-accelerated implementation.