Parallel dynamic load balancing for semiconductor device simulations on a linux cluster

A parallel dynamic load balancing for 2-D and 3-D semiconductor device simulations is proposed. The hydrodynamic and drift diffusion models are discretized and solved with finite volume (box) and monotone iterative methods. Dynamic load balancing for parallel domain decomposition as well as parallel I-V point simulations are demonstrated to be efficient methods for multidimensional device simulations. Compared with the measured data, numerical results and benchmarks on a realistic NMOSFET device are presented to show the accuracy and efficiency of the method. The code has been successfully implemented on a Linux-cluster with message passing interface (MPI) library.