Optimal operator preconditioning for pseudodifferential boundary problems (Q2038413): Difference between revisions

In this paper, a preconditioner is constructed for an elliptic pseudodifferential operator. The authors consider a Dirichlet problem with a pseudodifferential operator \(A\) of order \(2s\) either in \(\mathbb{R}^n\) or in a Riemannian manifold. The bilinear form associated with \(A\) is nonlocal, therefore Galerkin discretization of \(A\) results in dense matrices. The condition number of the Galerkin matrices is of order \({\mathcal O}(h^{-2|s|})\) when low-order piecewise polynomial basis functions are used, where \(h\) is the size of the smallest cell of the mesh. Therefore, the solution of the resulting linear system on fine meshes by iterative solvers becomes prohibitively slow. The proposed preconditioner is optimal such that the bound for the condition number neither depends on the mesh refinement, nor on the choice of bases for trial and test spaces. The authors prove that the required mesh assumptions hold for adaptively refined meshes. In this way, an operator preconditioning with standard low-order primal-dual finite element discretization is on adaptive meshes is developed. The performance of the proposed preconditioner is demonstrated for numerical examples on quasi-uniform, graded, and adaptively generated meshes.