Particle‐based Full‐band Approach for Fast Simulation of Charge Transport in Si, GaAs, and InP

Marco Saraniti,Yibing Hu,Stephen M Goodnick

doi:10.1080/1065514021000012354

Marco Saraniti, Yibing Hu + Show 1 more

Open Access

https://doi.org/10.1080/1065514021000012354

Copy DOI

Abstract

We discuss the application of the fullband cellular automaton (CA) method for the simulation of charge transport in several semiconductors. Basing the selection of the state after scattering on simple look‐up tables, the approach is physically equivalent to the full band Monte Carlo (MC) approach but is much faster. Furthermore, the structure of the pre‐tabulated transition probabilities naturally allows for an extension of the model to fully anisotropic scattering without additional computational burden. Simulation results of transport of electrons and holes in several materials are discussed, with particular emphasis on the transient response of photo‐generated carriers in InP and GaAs. Finally, a discussion on parallel algorithms is presented, for the implementation of the code on workstation clusters.

Highlights

Particle-based methods [1] are well established for the simulation of charge transport in semiconductors
The computational burden continues to limit its use in commercial device modeling, which has not been alleviated by the evolution of faster computers, when sophisticated physical models are implemented
We attempt to use the same physical models in terms of empirical pseudopotential method (EPM) band structure, valence shell phonon dispersion, and scattering mechanisms as used in DAMOCLES to compare the accuracy of the present code

Summary

Introduction

Particle-based methods [1] are well established for the simulation of charge transport in semiconductors. Their relative popularity is based on both physical and numeric considerations including: (i) the possibility of implementing sophisticated models without affecting algorithmic stability, (ii) the capability of microscopic modeling, and (iii) the numerical robustness. The computational burden continues to limit its use in commercial device modeling, which has not been alleviated by the evolution of faster computers, when sophisticated (and computationally demanding) physical models are implemented. The tremendous speedup obtained by the CA approach allows to simulate slower phenomena, and/or to implement more sophisticated physical models

Objectives

Results

Conclusion