The Riccati algorithm for eigenvalues and invariant subspaces of matrices with inexpensive action (Q1855446)

This paper concentrates on numerical methods for computing Lipschitz-stable invariant subspaces. This is achieved through corrections to subspaces, corrections which satisfy a generalised algebraic Riccati equation. This is a tensor or Kronecker product operator, and the formulation is in tensor products of Krylov subspaces, and the algorithms can be interpreted as Riccati versions of subspace iteration and block Rayleigh quotient iteration. The paper ends with some numerical examples and comparisons with the Jacobi-Davidson method.