Robust algebraic multilevel methods and algorithms. (Q1034018)

Classical (geometric) multigrid methods are known to be a powerful tool for solving large linear systems which arise from the discretization of partial differential equations. The method requires the information how the system matrices are calculated for a hierarchy of finite dimensional spaces. Algebraic multigrid methods intend to take all information from the given system matrix for the finest mesh and require almost no extra information. The difficulty is the appropriate construction of the ``coarse subspaces''. A crucial point is that the condition numbers of the matrices for the smaller systems are reduced as it is done by the coarsening in geometric multigrid algorithms. A lemma (Lemma 2.1) that is considered as a basis of various preconditioning techniques in the book, can also be taken as a connecting link between algebraic and geometric multigrid. At each level the variables are split into two groups namely the coarse grid variables and the fine grid variables. If \(A\) is a positive definite matrix, then \[ \min_{x_1} \begin{pmatrix} x_1\\ x_2\end{pmatrix}^T \begin{pmatrix} A_{11} & A_{12}\\ A_{21} & A_{22}\end{pmatrix} \begin{pmatrix} x_1\\ x_2\end{pmatrix} =\,x_2^T\,Sx_2\,, \] where \(S\) in the Schur complement of \(A_{11}\). Therefore the Schur complement and its approximations paves the way for the construction of matrices with appropriate condition numbers. This and the constants in strengthened Cauchy inequalities are the main tools in the book. The titles of the chapters are: 1. Introduction -- 2. Algebraic multilevel iteration (AMLI) methods -- 3. Robust AMLI algorithms: Conforming linear finite elements (FEs) -- 4. Robust AMLI algorithms: Nonconforming linear finite elements -- 5. Schur complement based multilevel preconditioners -- 6. Algebraic multigrid (AMG) -- 7. Preconditioning of Rannacher--Turek nonconforming FE systems -- 8. AMLI algorithms for discontinuous Galerkin FE problems -- 9 AMLI methods for coupled problems -- 10. Practical issues. The book will find its readers among those who are interested in (algebraic) multigrid.