Convolutional Machine Learning Method for Accelerating Nonequilibrium Green’s Function Simulations in Nanosheet Transistor

Abstract

This work describes a novel simulation approach that combines machine learning and device modelling simulations. The device simulations are based on the quantum mechanical non-equilibrium Green's function (NEGF) approach, and the machine learning (ML) method is an extension of a convolutional generative network. We have named our new simulation approach ML-NEGF. It is implemented in our in-house simulator called NESS (Nano-Electronics Simulation Software). The reported results demonstrate the improved convergence speed of the ML-NEGF method in comparison to the 'standard' NEGF approach. The trained ML model effectively learns the underlying physics of nano-sheet transistor behaviour, resulting in faster convergence of the coupled Poisson-NEGF self-consistency simulations. Quantitatively, our ML-NEGF approach achieves an average convergence speedup of 60%, substantially reducing the computational time while maintaining the same accuracy.

Numerical investigation on the convergence of self-consistent Schrödinger-Poisson equations in semiconductor device transport simulation

“…Recently, in efforts to enhance the convergence, both the machine learning-assisted acceleration methods [18,19] and the incorporation of the Anderson method for potential postprocessing [20] are based on the conventional Gummel scheme. Therefore, stabilizing the self-consistent iteration between the NEGF and the Poisson equation within this scheme remains fundamental and crucial.…”