Solving an Integral Equation Eigenvalue Problem via a New Domain Decomposition Method and Hierarchical Matrices

TL;DR

This paper introduces a novel domain decomposition method combined with hierarchical matrices to efficiently solve discretized integral equation eigenvalue problems, extending the applicability of the AMLS method to dense matrices.

Contribution

The paper generalizes the AMLS domain decomposition technique to handle dense matrices from integral equations, integrating it with hierarchical matrices for improved efficiency.

Findings

Effective domain decomposition for dense matrices

Enhanced eigenvalue problem solving efficiency

Applicable to large-scale integral equations

Abstract

In this paper the author introduces a new domain decomposition method for the solution of discretised integral equation eigenvalue problems. The new domain decomposition method is motivated by the so-called automated multi-level substructuring (short AMLS) method. The AMLS method is a domain decomposition method for the solution of elliptic PDE eigenvalue problems which has been shown to be very efficient especially when a large number of eigenpairs is sought. In general the AMLS method is only applicable to these kind of continuous eigenvalue problems where the corresponding discretisation leads to an algebraic eigenvalue problem of the form $Kx=\lambda Mx$ where $K,M \in \mathbb{R}^{N\times N}$ are symmetric sparse matrices. However, the discretisation of an integral equation eigenvalue problem leads to a discrete problem where the matrix $K$ is typically dense, since a non-local integral operator is involved in the equation. In the new method, which is introduced in this paper, the domain decomposition technique of classical AMLS is generalised to eigenvalue problems $Kx=\lambda Mx$ where $K,M$ are symmetric and possibly dense, and which is therefor applicable for the solution of integral equation eigenvalue problems. To make out of this an efficient eigensolver, the new domain decomposition method is combined with a recursive approach and with the concept of hierarchical matrices.