Solving singular generalized eigenvalue problems. III: Structure preservation (Q6936074)

In this interesting paper, solving generalized eigenvalue problems is studied. It is known, that often the QZ algorithm can be used to compute the eigenvalues of a singular matrix pencil. The algorithm produces eigenavlues that are close to the original eigenvalues and also eigenvalues which produce additional values that result from perturbations by rounding errors made to the singular part of the pencil. A natural question then is how to tell the difference between the ``true'' eigenvalues from the ``random eigenvalues'' among the values produced by the QZ algorithm. The work in this paper follows an approach given in two earlier papers by the authors which deal with three methods for nonsymmetric singular pencils: rank-completing perturbations, rank-projections, and augmentation.\N\NIn these later papers, it was observed surprisingly that a structure-preserving adaption for symmetric pencils was not possible and it was left as an open question how to address this challenge. In this paper this question is addressed. Indeed, it is shown that this issue can be circumvented by using Hermitian perturbations. This leads to structure-preserving analogs of the three techniques above for Hermitian pencils including real symmetric pencils as well as for skew-Hermitian, * -even, * -odd, and * -(anti-) palindromic pencils.\N\NThe paper is well written with a very good set of references.