The hierarchical basis multigrid method (Q1103353)

Based on earlier work of the third author and the PLTMG package of the first author, the authors describe, analyze and implement the mentioned method for the numerical solution of a standard self-adjoint twodimensional elliptic problem as a symmetric block Gauß-Seidel method with inner iterations (which can be rewritten in V-cycle manner). For the discretization, linear triangular elements are employed. The point is here that the triangulation may be generated (refined) adaptively: The described multigrid method does not assume that the numbers of unknowns on the different refinement levels grow geometrically to assure a work estimate per iteration of the overall unknown number since on any level only the new points are iterated. The analysis also does not require the usual strong ellipticity condition. The price for these advantages is that the number of iterations depends on the number of levels.