Optimization of Dual Field Plate AlGaN/GaN HEMTs Using Artificial Neural Networks and Particle Swarm Optimization Algorithm

Field plate technology is an effective method for improving the breakdown performance of AlGaN/GaN high electron mobility transistor (HEMT). Currently, field plate optimization relies on TCAD simulation, which is time-consuming and difficult to converge. In this study, we propose a fast and efficient method to optimize the gate-source dual field plate (dual-FP) to improve the breakdown performance of the HEMT. Specifically, an artificial neural network (ANN) model was used to fit the relationship between the dual-FP structure parameters and the breakdown voltage (BV), so that the breakdown performance could be predicted quickly and the average prediction error was only 3.06%. Furthermore, the trained ANN model was applied to the particle swarm optimization (PSO) algorithm and a dual-FP HEMT with a breakdown voltage of 1228 V was obtained by optimization. The proposed method shows significant advantages in terms of optimization efficiency and can realize automatic optimization. It also provides a reference for the optimization of other field plate structures of microelectronic devices.