A parametric Keldysh decomposition

TL;DR

This paper introduces a parametric extension of the Keldysh decomposition to efficiently compute eigenvalues of matrix-valued functions depending on an extra parameter, advancing contour integral eigenvalue methods.

Contribution

It develops a novel parametric Keldysh decomposition and an associated algorithm for solving parametric nonlinear eigenvalue problems, expanding the applicability of contour integral methods.

Findings

Established properties of the parametric Keldysh decomposition.

Proposed an algorithm for parametric nonlinear eigenvalue problems.

Enhanced eigenvalue computation for parameter-dependent matrices.

Abstract

Contour integral algorithms seek to compute a small number of eigenvalues located within a bounded region of the complex plane. These methods can be applied to both linear and nonlinear matrix eigenvalue problems. In the latter case, the foundation of these methods comes from the Keldysh decomposition, which breaks the nonlinear matrix-valued function into two parts: a rational function whose poles match the desired eigenvalues, and a remainder term that is analytic within the target region. Under contour integration this analytic part vanishes (via Cauchy's theorem), leaving only the component containing the desired eigenvalues. We propose an extension of the Keldysh decomposition for matrix-valued functions that depend analytically on an additional parameter. We establish key properties of this parametric Keldysh decomposition, and introduce an algorithm for solving parametric nonlinear eigenvalue problems that is based upon it.