Neural network-driven molecular insights into alkaline wet etching of GaN: toward atomistic precision in nanostructure fabrication

We present large-scale molecular dynamics (MD) simulations based on a neural network potential (NNP) to investigate alkaline wet etching of GaN, a process critical to nitride-based semiconductor fabrication. A Behler–Parrinello-type NNP is trained on extensive DFT datasets to capture chemical reactions between GaN and KOH. Using temperature-accelerated dynamics, our NNP-MD simulations accurately reproduce experimentally observed structural modifications of GaN nanorods during etching. The etching simulations reveal surface-specific morphological evolution: pyramidal pits on the −c plane, truncated pyramids on the +c plane, and planar morphologies on non-polar m and a surfaces. We also identify key chemical reactions governing the etching mechanisms. Enhanced-sampling simulations provide free-energy profiles for Ga dissolution, which critically influences the etching rate. The −c, a, and m planes exhibit moderate activation barriers, confirming their etchability, while the +c surface shows a significantly higher barrier, indicating strong resistance. We also observe the formation of Ga-O-Ga bridges on etched surfaces, which may act as carrier traps. This work provides atomistic insights into the mechanisms and kinetics of GaN wet etching, offering guidance for the fabrication of nanostructures in advanced GaN-based electronic and display applications.