Domain decomposition methods for the numerical solution of partial differential equations (Q925210)

The treatment of domain decomposition methods is split according to the four points: Schwarz formulation -- Steklov-Poincaré formulation -- Lagrange multiplier formulation -- least squares control formulation. Moreover there is the distinction of overlapping and non-overlapping subdomains. The basic methods above are combined with several techniques for the solution of the auxiliary problems, and they are also applied to various partial differential equations (PDEs). This is best described by the titles of the 18 chapters: Decomposition frameworks -- Schwarz iterative algorithms -- Schur complement and iterative substructuring algorithms -- Lagrange multiplier based substructuring: Finite element tearing and interconnecting (FETI) method -- Computational issues and parallelization -- Least-sqares control theory: iterative algorithms -- Multilevel and local grid refinement methods -- Non-self adjoint elliptic equations: iterative methods -- Parabolic equations -- Saddle point problems -- Non-matching grid discretizations -- Heterogeneous domain decomposition methods -- Fictitious domain and domain imbedding methods -- Variational inequalities and obstacle problems -- Maximum norm theory -- Eigenvalue problems -- Optimization problems --Helmholtz scattering problem. The book focuses on the handling of the formulas that are required for the computations. Complete proofs are given only in a few cases. For example, Xu's convergence theory for treating PDE's via subspace decomposition that is famous in multigrid theory, is presented in the framework of Schwarz methods. On the other hand, it is always assumed that the inf-sup condition is satisfied when saddle point problems are considered. The definition of the space \(H^{1/2}_{00}\) should be taken by the reader more as a hint at those Sobolev spaces rather than as a precise mathematical definition. The details of the items listed above make that the book contains more than 700 pages.