On conjugate gradient type methods and polynomial preconditioners for a class of complex non-Hermitian matrices (Q915367): Difference between revisions

The author presents a detailed study, with the emphasis on practical aspects, of conjugate gradient type methods for arbitrary complex matrices of the form \(A=e^{i\theta}(T+i\sigma I)\), T Hermitian matrix, \(\sigma\) and \(\theta\) real. Three approaches based on a minimal residual property, a Galerkin condition and an Euclidean error minimization are considered. In particular it is shown how SYMMLQ and MINRES can be extended to numerically stable implementations of all three approaches and derive error bounds for all three methods. Also it is shown how the special shift structure of A can be preserved by using polynomial preconditioning, and results on the optimal choice of the polynomial preconditioner are given. Finally, some numerical experiments for matrices arising from finite difference approximations to the complex Helmholtz equation with constant coefficients are reported.