Title:Selection of Corners for the BDDC Method

Abstract: The Balancing Domain Decomposition by Constraints (BDDC) method has evolved quite fast since its introduction in 2003, as the primal counterpart to the earlier FETI-DP method. Recent results have shown close connection of these methods and theoretically supported equivalent rate of convergence. In both methods, a fundamental role is played by the coarse space. Optimal choice of constraints on continuity of the coarse space is still not a satisfactorily solved problem. The usual basic choice is a 'minimal' set of coarse nodes (sometimes called corners), that assures invertibility of local subdomain problems and also of the global coarse problem. However, this set alone does not suffice for optimal preconditioning in 3D. For this reason, continuity of some generalized degrees of freedom, such as average values on edges or faces of subdomains, have to be added. While theoretically correct, this approach does not easily offer a flexible size of desired coarse problem. In our contribution, we compare this approach with adding more coarse nodes into the coarse problem, which is technically simpler and allows flexible setting of desired approximation. Most attention is given to the basic set of corners, and a new algorithm for its selection is proposed.