Neural-Network Potential Simulation of Defect Formation Induced by Knock-On Irradiation Damage in GaN

Understanding the irradiation damage mechanisms of GaN is of great importance for improving irradiation resistance and the ion implantation processes of GaN-based devices. The understanding of the damage mechanisms, which are normally simulated using molecular dynamics (MD), is bottlenecked by the dilemma that ab initio MD is limited by relatively high computational cost, whereas classical MD suffers from low accuracy. In this paper, a global neural network (G-NN) potential is constructed for GaN using random stochastic surface walking global optimization combined with global neural network potential (SSW-NN). By benchmarking the properties of intrinsic defects and defect-pairs, as well as the threshold displacement energies along different crystallographic directions, this potential is found to provide accuracy similar to ab initio calculations. Furthermore, based on the large-scale simulations of the knock-on process, Ga and N vacancies, as well as N interstitials are found to be the major generated defects, whereas only Ga vacancies are predicted in the previous device simulation studies that have been widely recognized. This potential provides an efficient and accurate tool to gain a fundamental understanding of the irradiation damage mechanisms of GaN and refine the parameters in the related device simulations.