Indexed on: 09 May '05Published on: 09 May '05Published in: Mathematics - Numerical Analysis
Presented in this paper is a new sparse linear solver methodology motivated by multigrid principles and based around general local transformations that diagonalize a matrix while maintaining its sparsity. These transformations are approximate, but the error they introduce can be systematically reduced. The cost of each transformation is independent of matrix size but dependent on the desired accuracy and a spatial error decay rate governed by local properties of the matrix. We test our method by applying a single transformation to the 2D Helmholtz equation at various frequencies, which illustrates the success of this approach.