Uniform preconditioners for problems of positive order (Q2192525)

The authors consider the numerical solution of partial differential equations discretizing the domain and the boundary with finite or boundary elements with the aim to solve the problem efficiently with the help of a kind of preconditioning. They use a corresponding operator, that offers a general possibility for the construction of preconditioners for a discrete operator \( A \). The basic idea of preconditioning and practical difficulty is roughly spoken to find an approximation of the inverse of \( A \). Based on a detailed analysis of the choice of finite dimensional subspaces of Sobolev spaces the authors create uniform preconditioners for elliptic operators that, compared to earlier proposals, avoid the inversion of a non-diagonal matrix and the construction of a so-called dual mesh. The theoretical results are numerically validated. Especially, nicely bounded spectral condition numbers are computed for a hypersingular system using uniform and local mesh refinements. The cost of the new preconditioner increases only by a minor cost of linear complexity in comparison with the application of an elliptic opposite order operator. The authors refer for details to their earlier publication [Math. Comp. 89, 645--674 (2020; Zbl 07146711)].