On the convergence of multi-grid methods with transforming smoothers (Q912583)

The author presents a general convergence theory for multi-grid methods that can be applied to a variety of systems of partial differential equations such as Stokes and Navier-Stokes equations. After a short outline of the general multi-grid technique together with a convergence theorem by \textit{W. Hackbusch} [Multi-grid methods and applications (1985; Zbl 0595.65106)], the model problems are described. Then the notion of r- transforming iterations is introduced that can be used to construct smoothers for indefinite systems and that include many of the known and widely used iteration schemes. In a first step Hackbusch's convergence theorem is generalized in such a way that the smoothing property of the r-transforming iteration is replaced by a smoothing property of a product iteration, and in a second step a criterion for a two-grid method to converge is derived. The discussion of the smoothing and convergence properties of block-triangular systems gives the indication of how to construct appropriate r-transformations. The theory is finally applied to the Stokes and Navier-Stokes equations.