Restarted Krylov-space spectral filtering

Abstract

A recently developed spectral-filter algorithm, which is based on the iterative application of a Green operator approximated in a finite-order Krylov subspace, has been applied to the problem of computing vibrational eigenstates of a medium-sized molecule. A six-dimensional model Hamiltonian for formaldehyde, H ₂ CO, is used to illustrate the efficiency of the method by comparison with restarted Fourier filtering. The Krylov-space algorithm is shown to produce substantial savings in computation time in comparison with the Fourier method, indicating that it may prove to be of considerable utility for the investigation of quantum dynamics in polyatomic unimolecular reactions.