Investigating of Ductility of Silver Sulfide using Artificial Neural Network Potentials

Silver sulfide is a semiconductor that exhibits remarkable metallic-like ductility under room temperature. We have investigated the mechanism underlying this unusual ductility using first-principles molecular dynamics (FPMD) simulations of simple shear deformation in six directions: (100)[010], (100)[001], (010)[100], (010)[001], (001)[100], and (001)[010] ((KLM)[klm]: sliding the (KLM) plane in the [klm] direction). However, the number of atoms (192) in the FPMD simulation precludes large-scale deformation mechanisms.

To overcome this limitation, we have trained an Artificial Neural Network (ANN) potential using FPMD data for shear deformation data, which achieves quantum-mechanical accuracy with orders-of-magnitude less computational cost, thus allowing the study of larger-scale deformation mechanisms. In a 1,536-atom system, we found a different structural-recovery mechanism by in-plane movement with a shorter distance of sulfur atom movement in the (100)[010]. In a 98,304-atom system, two grains appeared when the sulfur sublattice is recovered in (001)[010] and (010)[100], while in the largest 786,432-atom system, the sulfur sublattice is recovered with the generation of multiple grains.