Optimization of the Hermitian and skew-Hermitian splitting iteration for saddle-point problems (Q1431640)

The authors study the asymptotic convergence rate of the alternating directions implicit method for solving saddle-point problems arising in the discretization of elliptic partial differential equations. They focus on the simple model problem of the Poisson equation, for which the Fourier transforms can be used in order to analyze the iteration operator at the continuous level. However, the same technique can be used to study more complicated problems with constant coefficients (e.g. some anisotropic problems and the Stokes problem). Optimal convergence parameters for the model problem of the Poisson equation written in \textit{div-grad} form are determined. It is shown that if a splitting is used as a preconditioner for the Krylov method, a different optimization leading to two clusters in the spectrum gives an optimal, mesh independent, convergence rate. The theoretical analysis is supported by numerical experiments.