Parallel Computing in Semiconductor Device Simulation

Stefan Thomas and Helmut Weberpals
Research period: 1995-1999

The ability to model complex systems interactively enhances the productivity in research and development. A typical example of this kind is a semiconductor device with many parameters influencing its behaviour.

Modelling modern submicron semiconductor devices requires the full set of three-dimensional hydrodynamic equations to be solved. A transformation of variables casts these equations into an elliptic form which lends itself to fast solvers. In order to accelerate the solution process we reduce the number of unknowns by adaptive local grid refinement. Finally, parallel processing yields a speed-up which turns out to be scalable over a wide range of processors. By combining adaptivity and parallelism we achieve an interactive simulation of a semiconductor device.