Computing interior eigenvalues of large matrices (Q811085)

The computation of eigenvalues from the interior of the spectrum of a large matrix is considered. The Rayleigh-Ritz procedure is a standard way for reducing it to a smaller problem, but it is not optimal for interior eigenvalues. Here a method is given that does a better job. The main idea is to apply a rectangular matrix whose columns span the desired subspace, and to avoid the explicitly inverted operator. In contrast with standard Rayleigh-Ritz, a priori bounds can be given for the accuracy of interior eigenvalue and eigenvector approximations. When applied to the Lanczos algorithm, it yields better approximations at early stages. Applied to preconditioning methods, it improves the convergence rate.