Unraveling Mg <c + a> Slip Using Neural Network Potentials

Abstract

Magnesium (Mg) activates <c + a> dislocation slip on the second-order pyramidal slip plane. Historically, under c-axis compression, there has been some debate regarding the preferred pyramidal slip plane as measured by experiments and atomistic simulations. Here we investigate the structure and motion of <c + a> dislocations using several interatomic potentials, including two recent neural network potentials. The new potential shows better agreement with density functional theory and experimental calculations than previous interatomic potentials for Mg, but agree on the structure and movement of <c + a> dislocations. Both new potentials demonstrate second-order pyramidal edge dislocations that spontaneously dissociate onto the basal plane, confirming earlier results and agreeing with experimental data. However, unlike earlier potentials, these new potentials predict that the dissociated cores are not sessile, but can move at shear stress as low as 25 MPa.KeywordsMagnesiumPyramidal slip <c + a> Machine learning