No label defined (Q4040641)

The present monograph is devoted to the construction of asymptotically optimal algorithms with respect to the amount of arithmetical operations and to the memory demand needed to obtain a solution of a discrete elliptic problem with some desired relative accuracy \(\varepsilon\). Such algorithms are usually characterized by the fact that the computational work is proportional or, at least, nearly proportional to the number of discrete unknowns involved in the approximation scheme. To get such methods the author starts with special discretizations of the elliptic problems resulting in discrete schemes which can be usually solved by means of fast iterative methods. One of the main goals is here the construction of efficient preconditioners for several problems of interest. The author considers a wide range of problems including linear and nonlinear, second- and fourth-order elliptic PDEs and systems of PDEs, elliptic problems with linear constraints like Stokes- and Navier-Stokes systems as well as spectral problems with elliptic operators. As mentioned above, the discretization of these problems is carried out by special finite difference or finite element techniques. One chapter is devoted to the construction of grids topologically equivalent to grids on model domains. On these model domains (e.g. rectangles), one can construct preconditioners for which fast methods are available. The method of fictitious components, two-stage iterative methods, domain decomposition methods and multigrid methods are further techniques to obtain optimal algorithms. Another chapter is concerned with the Stokes and Navier-Stokes problems resulting in discrete problems with equality constraints arising from the incompressibility condition \(\operatorname{div} u=0.\) Fast algorithms constructed for the discrete Stokes problem are used to obtain similar algorithms for fourth-order problems. The final chapter is devoted to the construction of effective algorithms for the solution of discrete elliptic eigenvalue problems of the form \(Lu=\lambda Mu.\) Most of the results presented in this monograph have been obtained by the author during the last 30 years since his pioneered paper in 1961 on iterative methods [Dokl. Akad. Nauk SSSR 138, 522-525 (1961; Zbl 0102.116)]. This book can be recommanded to every numerical analyst dealing with the numerical solution of elliptic problems as standard work.