Structure and strain state evolution under nanoindentation of Ag nanospheres by machine learning interatomic potential

Unraveling the atomic-scale mechanisms governing dislocation nucleation and evolution is paramount to our fundamental understanding of deformation behavior and mechanical properties in crystalline materials. However, detailed quantification of such atomic scale deformation mechanisms is constrained by over-simplification of interatomic potentials for computational simulations. In this study, a Deep Potential (DP) model for Ag was trained and the nanoindentation on Ag nanospheres was conducted by MD simulations using both DP and empirical potentials to investigate the deformation mechanisms. During compression processing the plasticity initiates from the nucleation of Shockley partial dislocations (SPDs) on multiple slip planes at a strain of 3.8 %, followed by the appearance of sessile Hirth dislocations and stacking fault pyramid (SFP). At 24 % of compression strain, SPDs leave behind extrinsic stacking faults (ESFs), the formation of twin boundaries was observed. The atomistic simulation system comprised 30,765 Ag atoms, exhibiting Young's modulus of 79.39 GPa, Yield strength of 1,420 MPa, and shear modulus of 29.85 GPa at 300 K. This work contributes to developing efficient and accurate computational methods for studying mechanical deformation behavior and demonstrates that the DP is a viable alternative to empirical potentials in studying mechanical properties.