The Application of Sparse Supernodal Factorization Algorithms for Structurally Symmetric Linear Systems in Semiconductor Device Simulation

Abstract

It is well known that the solution of sparse linear systems, generally expressed in the form Ax = b, is a core task of numerical simulation. In case of semiconductor device simulation the coefficient matrix A is unsymmetric, but structurally symmetric ([2]). The solution of linear systems can be achieved by iterative or direct methods. While iterative methods do not always lead to a solution due to matrix conditions, direct methods usually consume more time and memory.