Symbol Based Convergence Analysis in Block Multigrid Methods with applications for Stokes problems
The main focus of this paper is the study of efficient multigrid methods for large linear system with a particular saddle-point structure. In particular, we propose a symbol based convergence analysis for problems that have a hidden block Toeplitz structure. Then, they can be investigated focusing on the properties of the associated generating function 𝐟, which consequently is a matrix-valued function with dimension depending on the block of the problem. As numerical tests we focus on the matrix sequence stemming from the finite element approximation of the Stokes equation. We show the efficiency of the methods studying the hidden 9× 9 block structure of the obtained matrix sequence proposing an efficient algebraic multigrid method with convergence rate independent of the matrix size. Moreover, we present several numerical tests comparing the results with different known strategies.
READ FULL TEXT