Adaptive finite volume simulation of semiconductor devices on cluster architecture

A new scientific computing algorithm for numerical solutions of a set of fundamental semiconductor device equations is presented. This simulation based on adaptive mesh, finite volume, monotone iterative, and a posteriori error estimation methods, is proposed and implemented successfully on a Linux-cluster computing environment. The new approach fully exploits the inherent parallelism of the monotone finite volume formulation of the basic semiconductor device equations and the cluster architecture in message passing interface (MPI) library. Parallel results demonstrate an excellent speedup with respect to the number of processor. Benchmarks and numerical results for a typical MOSFET device are also presented to show the accuracy and efficiency of the method.