Estimates for multigrid methods based on red-black Gauss-Seidel smoothings (Q1096358)

The MGR[\(\nu\) ] algorithms of \textit{M. Ries}, \textit{U. Trottenberg} and \textit{G. Winter} [Linear Algebra Appl. 49, 1-26 (1983; Zbl 0515.65070)], the algorithms 2.1 and 6.1 of \textit{D. Braess} [Math. Comput. 42, 505-519 (1984; Zbl 0539.65075)] and the algorithm 4.1 of \textit{R. Verfürth} [Linear Algebra Appl. 60, 113-128 (1984; Zbl 0548.65074)] are all multigrid algorithms for the solution of the discrete Poisson equation (with Dirichlet boundary conditions) based on red-black Gauss-Seidel smoothing. Both Braess and Verfürth give explicit numerical upper bounds on the rate of convergence of their methods in convex polygonal domains. In this work we reconsider these problems and obtain improved estimates for the h-2h Algorithm 4.1 as well as W-cycle estimates for both schemes in non-convex polygonal domains. The proofs do not depend on the strengthened Cauchy inequality.