Parallelization of Particle Transport Algorithms in Semiconductor Device Physics

9312240 Goodnick这是一个更新项目，旨在继续开发使用粒子方法进行半导体器件模拟的并行算法。将对原型多处理器代码PMC-3D进行改进，以包括最先进的物理效果，如全能带结构效果，并将代码推广到各种不同的设备技术和任意几何形状。这些代码将用高级并行语言(数据并行C语言)开发，以增加它们在机器之间的可移植性。此外，为了开发高频微波和光电子器件应用的模拟工具，在粒子力计算中将耦合完整的电磁解而不是泊松方程。将开发包括分子动力学和细胞自动机技术的半导体器件的并行粒子传输算法。在分子动力学的情况下，将研究粒子-粒子-网格混合方法，以寻找高效的并行算法，从而大大降低该技术的计算要求。将开发三维的并行元胞自动机器件模拟算法，并将其与蒙特卡罗方法进行比较，作为后者的经济有效的替代技术。***

9312240 Goodnick This is a renewal project to continue the development of parallelized algorithms for semiconductor device simulation using particle methods. Improvements will be made to the prototype multiprocessor code PMC-3D to include state of the art physics such as full band structure effects, and generalize the code for a variety of different device technologies and arbitrary geometries. The codes will be developed in a high level parallel language (Dataparallel C) to increase their portability from machine to machine. Furthermore, full electromagnetic solutions rather than Poission's equation will be coupled in the particle-force calculation, in order to develop simulation tools for high frequency microwave and optoelectronic device applications. Parallel particle transport algorithms for semiconductor devices to include molecular dynamics and cellular automata techniques will developed. In the molecular dynamics case, hybrid particle-particle particle-mesh methods will be investigated to find efficient parallel algorithms which would greatly reduce the computational requirements of this technique. Parallel cellular automata device simulation algorithms in three dimensions will be developed and compared with Monte Carlo methods as a cost effective alternative to the latter technique. ***