Interfacial Mechanical Properties and Optimization of GaN/Diamond Heterostructures via Machine Learning Potential

Gallium nitride (GaN)-based electronic devices underpin high-power and high-frequency applications owing to their exceptional performance characteristics. However, their potential is often constrained by the inherent low thermal conductivity of GaN. Integrating GaN with diamond heat sinks improves thermal management but often compromises reliability due to mechanical failures arising from lattice and thermal expansion coefficient mismatches. Herein, the interfacial mechanical properties and failure mechanisms of GaN/diamond heterostructures are investigated utilizing large-scale molecular dynamics (MD) simulations enabled by a high-precision, machine-learning-based Neuroevolution Potential (NEP). The developed NEP model exhibits superior accuracy in capturing the anisotropic fracture behaviors and experimental cleavage planes of the constituent materials. Simulation results reveal that the mechanical integrity of the heterostructure is critically governed by crystallographic orientation and interfacial chemistry. Specifically, the GaN(0001)/diamond(100) interface, dominated by strong C–N covalent bonds, displays adhesion superior to that of C–Ga-bonded or diamond (110)-oriented interfaces. Furthermore, thermal annealing and surface morphology are identified as pivotal factors for interface optimization. While geometric interlocking induced by surface roughness enhances damage tolerance, it also reduces ultimate tensile strength due to stress concentration. Conversely, thermal annealing emerges as a universally effective reinforcement strategy, strengthening interfacial bonds by relaxing atomic potential energy. Combining an atomically planar interface with high-temperature annealing (1100 K) yields optimal mechanical performance, achieving a fracture stress 69.74% higher than rough interfaces and 7.54% higher than the unannealed planar state. These findings provide a theoretical foundation and quantitative optimization strategies for the interface engineering of GaN-on-diamond devices.