Hybrid Physics - AI Framework for Optimization of GaN-based P-MISFETs Using InGaN/AlN/AlGaN Heterostructures

Description

This SURF project focused on the optimization of CMOS-compatible p-type GaN MISFETs through a hybrid physics–AI framework. The students investigated polarization-engineered InGaN/AlN/AlGaN/GaN heterostructures using a combined approach of physics-based TCAD simulations and data-driven analysis. The work addressed the key challenge of balancing hole confinement, leakage suppression, and drive current in p-type devices, which is essential for realizing GaN CMOS technology.

The methodology integrated heterostructure design, systematic design-of-experiments (DOE), clustering analysis, and SHAP-based feature attribution. A large dataset of transfer characteristics was generated to extract statistical device archetypes ranging from ultra-low-leakage to high-drive operation. The analysis revealed that recess depth, indium composition, and aluminum composition are the dominant parameters shaping device trade-offs, while AlN layer thickness primarily influences carrier confinement at the InGaN/AlN interface.

Through this framework, the team identified design strategies that link material engineering (polarization charge, band offsets, and quantum well confinement) to device-level performance (Ion, Ioff, SS, and Vth). The results highlight a systematic path toward achieving balanced performance in GaN p-MISFETs and provide insights for future multi-objective optimization and co-design of complementary p- and n-channel devices for GaN CMOS logic.

Period	9 Jun 2025 → 28 Aug 2025