A block Rayleigh quotient iteration with local quadratic convergence (Q1279549)

The author presents an iterative eigensolver, based on a block generalization of the Rayleigh quotient iterative method, to search for the \(p\) lowest eigenpairs of the generalized matrix eigenvalue problem \(Au= \lambda Bu\). Local quadratic convergence is proved assuming \(A\) and \(B\) are such that \(B^{-1}A\) is symmetric. This scheme produces well-conditioned linear systems and has the ability to treat multiple and nearly degenerate eigenvalues.