Friction Mechanisms of γ-Al2O3 Under Nanoindentation–Scratch Coupling: A Deep Neural Network Potential Approach

Abstract

This study constructs and validates a deep neural network potential with quantum-level accuracy for γ-Al₂O₃, achieving root-mean-square errors of 1.2 meV/atom for energy and 19 meV/Å for force. Based on this potential, nanoindentation-scratch simulations at three indentation depths are performed to investigate the evolution of elasticity, creep, and viscoplasticity. Hertzian fitting yields an elastic modulus of 317.4–322.4 GPa, which closely agrees with experimental measurements. Frictional analysis reveals that the normal force stabilizes into a plateau, while the friction coefficient exhibits a nonlinear increase, governed respectively by the real contact area and the atomic-scale shear strength spectrum. The normalized load Π = F_f/N_c indicates that the mean interfacial shear strength remains nearly constant with depth, whereas the amplitude of its fluctuations increases by a factor of 3. These findings elucidate the synergistic roles of contact area control and shear spectrum broadening in ceramic interfacial friction, providing quantitative guidance for friction reduction strategies that integrate surface texturing and interfacial chemistry, and establishing a computational foundation for large-scale data-driven materials design.