Numerical computation of inner eigenvalues using the Dunford Cauchy integral

Wendy Kress

Contact the author: Please use for correspondence this email.
Submission date: 01. Jun. 2007
Pages: 13
Bibtex
MSC-Numbers: 34L16
Keywords and phrases: eigenvalues, projection, Dunford Cauchy
Download full preprint: PDF (686 kB)

Abstract:
When computing the eigenvalues of a matrix using iterative Krylov subspace methods, convergence is usually best for the extreme eigenvalues. We present a projection technique that enables us to efficiently compute the eigenvalues that lie in a specified interval in the interior of the spectrum. To obtain such a projection, the Dunford Cauchy integral is used. The technique requires fast inversion algorithms which are available for some classes of matrices like -matrices.