Projection method for eigenvalue problems of linear nonsquare matrix pencils

TL;DR

This paper extends a projection method using complex moments to nonsquare matrix pencils, enabling efficient and robust eigenvalue computation in large, complex problems, especially in parallel computing environments.

Contribution

It introduces a novel projection method for singular nonsquare matrix pencils using pseudoinverses and contour integrals, improving robustness and efficiency over previous approaches.

Findings

Method accurately finds all finite eigenvalues in a region.

Numerical experiments demonstrate robustness and efficiency.

Approach is suitable for large-scale, parallel computations.

Abstract

Eigensolvers involving complex moments can determine all the eigenvalues in a given region in the complex plane and the corresponding eigenvectors of a regular linear matrix pencil. The complex moment acts as a filter for extracting eigencomponents of interest from random vectors or matrices. This study extends a projection method for regular eigenproblems to the singular nonsquare case, thus replacing the standard matrix inverse in the resolvent with the pseudoinverse. The extended method involves complex moments given by the contour integrals of generalized resolvents associated with nonsquare matrices. We establish conditions such that the method gives all finite eigenvalues in a prescribed region in the complex plane. In numerical computations, the contour integrals are approximated using numerical quadratures. The primary cost lies in the solutions of linear least squares problems that arise from quadrature points, and they can be readily parallelized in practice. Numerical experiments on large matrix pencils illustrate this method. The new method is more robust and efficient than previous methods, and based on experimental results, it is conjectured to be more efficient in parallelized settings. Notably, the proposed method does not fail in cases involving pairs of extremely close eigenvalues, and it overcomes the issue of problem size.