Predictive TCAD Modeling of Normally-Off p-GaN HEMTs for Dynamic RDS(on), Leakage, Breakdown, and ML-Based Design Exploration

Abstract

This article presents a practical simulation workflow for predictive simulation of normally-off p-GaN high-electron-mobility transistors (HEMTs), with emphasis on device-physics calibration, leakage modeling, dynamic RDS(on), breakdown behavior, and machine-learning-assisted design exploration. The workflow is based on a 100 V p-GaN HEMT reference structure and combines process-informed device definition, careful meshing, calibrated Victory Device physics models, mixed-mode switching simulation, Design of Experiments (DOE), and Victory Analytics surrogate modeling. The goal is to establish a repeatable methodology that helps engineers understand how traps, polarization, Mg diffusion, electric-field distribution, and switching conditions influence GaN power-device performance and reliability.