Robust Mode Space Approach for Atomistic Modeling of Realistically Large Nanowire Transistors

TL;DR

This paper introduces a robust, parallel mode space basis optimization algorithm enabling efficient atomistic quantum transport simulations of large nanowire transistors, facilitating realistic device modeling with full-band accuracy.

Contribution

A new parallel algorithm for optimizing mode space basis in atomistic nanowire simulations, improving computational efficiency and enabling realistic device analysis.

Findings

InGaAs nanowire nMOSFETs do not outperform Si counterparts for cross sections up to 10nm x 10nm.

The approach allows NEGF simulations of large nanowires on small clusters.

Full-band accuracy is maintained in the simulations.

Abstract

Atomistic quantum transport simulation of realistically large devices is computationally very demanding. The widely used mode space (MS) approach can significantly reduce the numerical cost but good MS basis is usually very hard to obtain for atomistic full-band models. In this work, a robust and parallel algorithm is developed to optimize the MS basis for atomistic nanowires. This enables tight binding non-equilibrium Green's function (NEGF) simulation of nanowire MOSFET with realistic cross section of $\rm 10nm\times10nm$ using a small computer cluster. This approach is applied to compare the performance of InGaAs and Si nanowire nMOSFETs with various channel lengths and cross sections. Simulation results with full-band accuracy indicate that InGaAs nanowire nMOSFETs have no drive current advantage over their Si counterparts for cross sections up to about $\rm 10nm\times10nm$.