Enhanced error estimates for augmented subspace method (Q6101881)

This article is concerned with enhanced error estimates for the augmented subspace methods which are designed for solving eigenvalue problems. It is shown that the augmented subspace methods have the second order convergence rate between the two iteration steps which is an improvement of the existent results in the literature and more consistent with the results of numerical tests performed by the authors. Numerical exxperiments are presented to validate these new estimate results and the efficiency of the proposed algorithms.