Facilitation of pyramidal dislocation slip in Mg-Bi alloys: A molecular dynamics study via machine learning potential

Magnesium (Mg) is a lightweight structural metal with high specific strength, but its limited room-temperature ductility---stemming from its hcp crystal structure---remains a major barrier to its broader engineering applications. Enhancing the plastic deformability of Mg through alloying has emerged as a promising strategy, yet the atomistic mechanisms underlying these improvements remain poorly understood. In the present work, we develop a machine learning potential (MLP) to accurately model Mg-Bi alloys and elemental Mg with quantum-level fidelity and computational efficiency. Using this MLP, we perform large-scale molecular dynamics (MD) simulations to investigate the mechanical behavior of pure Mg and Mg-5%Bi nanopillars under uniaxial compression and tension along the [0001] direction. The simulation results reveal that Bi addition (5 at %) significantly enhances Mg's plasticity by promoting the nucleation and glide of $1/2\ensuremath{\langle}\mathrm{c}+\mathrm{a}\ensuremath{\rangle}$ dislocations. This leads to a transition in the dominant deformation mechanism---from transformation-assisted plasticity in pure Mg to dislocation-dominated plasticity in Mg-5%Bi. Moreover, our simulations quantitatively reproduce the experimentally observed asymmetry in yield strength, where compressive strength exceeds tensile strength along the $c$ axis. This work not only elucidates the atomic-scale origin of ductility enhancement via Bi addition, but also reveals the importance of atomistic modeling using MLP for predicting the mechanical behaviors of complex alloy systems.