A priori error bounds on invariant subspace approximations by block Krylov subspaces (Q1611877)

The authors derive a priori error bounds for block Krylov subspace methods in terms of ``the sine'' between the desired invariant subspace and the block Krylov subspace. The obtained results can be seen as the block analogue of the standard projection methods. The main results are obtained without assuming that the coefficient matrix is diagonalizable. If it is the case, the authors obtain extensions of the minimal problems known in the standard case. The last estimates show that the invariant subspace corresponding to a tight cluster of eigenvalues well separated from the rest of the spectrum is well approximated by block Krylov subspace methods, which is not the case with standard Krylov methods. These estimates are actually upper bounds and not necessarily optimal. The obtained results in this paper are mainly of theoretical type involving quantities which are not always readily computable.