Optimal quotients for solving large eigenvalue problems (Q1731607)

This paper is devoted to find an optimal projection and resulting iterative methods to solve eigenvalue problems and generalized eigenvalue problems. At first, the notion of the optimal quotient is derived and basic properties of optimal quotients are proved. Then, several optimal quotient iterations are introduced and optimality conditions are shown. Finally, the idea of the optimal projection is extended to cover subspaces of a dimension larger than one. Significant numerical examples are discussed in details. Moreover, numerical experiments are presented which refer to an eigenvalue problem in magnetohydrodynamics (MHD) to illustrate convergence properties and computational cost of the presented algorithms.