Impact of point defects on the thermal conductivity of GaN studied using machine-learned potentials

Gallium nitride (GaN) is a crucial material for high power density electronic devices, and its self-heating effects under high power conditions have attracted considerable attention. The study of intrinsic point defects in GaN and their impact on thermal conductivity is a crucial research focus. The influence of these defects on the thermal conductivity is dependent on their type, concentration, and charge state. However, existing experimental and computational methods, such as density functional theory and molecular dynamics, encounter limitations, particularly when dealing with defects of low concentrations, various types, and charged states. To overcome these challenges, we have developed machine-learning potentials for point defects at various concentrations, charge states, and temperatures. Predictive capabilities for the mechanical properties and migration pathways of point defects are demonstrated. We investigate the temperature dependence of thermal conductivity in defect-free structures and analyze how point defect concentration affects thermal conductivity. These studies provide insights for optimizing the performance of GaN materials.