Investigation of Different Compact Modeling Methods for AlGaN/GaN HEMTs under Radiation Environment

This article explores two compact modeling methods for AlGaN/gallium nitride (GaN) high electron mobility transistors (HEMTs) under radiation environments, based on artificial neural network (ANN) techniques. The first method is a hybrid compact modeling approach that employs ANN techniques to correlate radiation-sensitive parameters with irradiation conditions and integrates these relationships into the original compact model, thereby incorporating radiation effects. This approach significantly reduces the complexity of model expression construction and is currently compatible with process design kits (PDKs). This article illustrates the method using a gate current model under 3-MeV proton irradiation and introduces the radiation-sensitive parameters in the model. The second method is a fully data-driven ANN-based compact modeling approach. This study establishes a complete modeling framework and provides a full pseudocode prototype for converting the ANN model to Verilog-A language for subsequent circuit simulation. The method is demonstrated using gate and drain current models under 3-MeV proton irradiation. The verification results confirm that both methods are suitable for constructing compact models in radiation environments with high accuracy, allowing users to select the most appropriate method based on the specific application scenario.