Artificial Neural Network based Model for Power GaN HEMTs down to 4.2K

This paper presents an experimental investigation and an artificial neural network (ANN)-based modeling of the DC characteristics of power GaN-based high electron mobility transistors (HEMTs) operating at temperatures ranging from 300 K to 4.2 K. The proposed ANN architecture accurately captures the correlation between electrical properties. More importantly, based on the optimized ANN architecture, we further predict the electrical performance of untrained devices fabricated on the same wafer. The predictive results show nearly perfect matching with the experimental results in the DC model, particularly at cryogenic temperatures, featuring a root-mean-square (RMS) error of less than 1%.