On convergence of double splitting methods for non-Hermitian positive semidefinite linear systems (Q987715)

The results known for double splitting of monotone matrices and for double splitting of Hermitian positive definite matrices are extended to linear systems with a large sparse non-Hermitian positive semidefinite matrix. A corresponding iterative scheme is spanned by three successive iterations. The iterations converge to a solution of the original problem for any initial guess \(x^{(0)}\), \(x^{(1)}\), if and only if the spectral radius of the iteration matrix is less then unity. Some convergence conditions are given and the method is applied to a so called generalized saddle point problem. Under suitable assumptions, it is proved that the double splitting iteration process is uniquely convergent.