Multiplicative And Additive Schwarz' Methods: Convergence In The 2-Domain Case

We consider the classical Schwarz alternating algorithm and an additive version more suitable for parallel processing. The two methods are compared and analyzed in the case of two domains. We show that the rate of convergence for both methods, can be directly related to a generalized eigenvalue problem, derived from subdomain contributions to the global stiiness matrix. Analytical expressions are given for a model case.