Acceleration of nonequilibrium Green’s function simulation for nanoscale FETs by applying convolutional neural network model

Abstract

We investigate the application of convolutional neural networks (CNNs) to accelerate quantum mechanical transport simulations (based on the nonequilibrium Green's function (NEGF) method) of double-gate MOSFETS. In particular, given a potential distribution as input data, we implement the convolutional autoencoder to train and predict the carrier density and local quantum capacitance distributions. The results indicate that the use of a single trained CNN model in the NEGF self-consistent calculation along with Poisson's equation produces accurate potentials for a wide range of the gate lengths, and all within a significantly shorter computational time than the conventional NEGF calculations.

Device simulations with A U-Net model predicting physical quantities in two-dimensional landscapes

“…There are two main steps when developing a deep learning model for device simulations, which in sequence are training data generation and model training. The training data may be collected systematically by self-consistent simulations [1][2][3][4][5][6][7][8][9][10][11] or measurements 12,13 . In this work, we choose the former approach based on a TCAD simulator 18 to easily access the physical quantities.…”

“…The data category for the two models is summarized in Table 1. Compared to prior literatures [1][2][3][4][5][6][7][8][9][10][11][12][13] , thanks to the unique advantage of U-Net 17 , hundreds or even thousands of data are not needed to train the model.…”