Modeling of inversion layer capacitance of III-V double gate MOSFETs using a neural network-based regression technique

Abstract

This work presents a data-driven regression model of inversion layer capacitance of double gate III-V channel MOSFETs implemented using an artificial neural network. The training dataset is generated using a Schroedinger-Poisson solver for different channel thicknesses, carrier effective masses, oxide thickness, barrier height, and a wide range of gate bias voltages. The neural network predicted capacitance value is compared with Schroedinger-Poisson solver data and a physics-based analytical model result. The model effectively captures the variation in channel thickness, barrier height, carrier effective mass, and oxide thickness. Furthermore, extensive error analysis has been performed to demonstrate the correctness and degree of accuracy of the predicted result.