Atomistic insight into dislocation behaviors in ultra-lightweight Mg-Li-Al alloys.

CONTEXT: Mg-Li alloys are among the lightest structural materials and exhibit excellent ductility, but their relatively low strength limits broader applications. Alloying, such as with Al, is an effective approach for strengthening Mg-Li alloys. However, the dislocation behavior in BCC Mg-Li alloys remains poorly understood. In this study, we first developed a highly accurate machine-learning potential (MLP) for Mg-Li-Al alloys. The developed MLP was subsequently applied to investigate the dislocation behavior of three representative alloys, Mg-38Li, Mg-38Li-2Al, and Mg-38Li-4Al (a.t.%). The results show that both alloys exhibit split screw dislocation core structures. Al addition is associated with more tortuous, intermittent edge dislocation motion and lower edge dislocation mobility. It also reduces screw dislocation velocity, although the reduction is smaller than that for edge dislocations. Screw dislocations are substantially less mobile than edge dislocations over the investigated stress range. The simulations further indicate that kink-pair activity on multiple planes and the resulting cross-kink configurations may contribute to the low screw dislocation mobility. These findings provide atomistic evidence for how minor Al additions modify dislocation motion in the modeled BCC Mg-Li alloys. METHODS: DFT calculations were performed using the VASP package to generate the training dataset for MLP development. The exchange-correlation interactions were described using the PBE form of the GGA. The MLP was developed and trained using the GRACE framework. All molecular dynamics simulations were subsequently carried out in LAMMPS employing the trained MLP. Atomistic models for the dislocation dynamics simulations were generated using Atomsk, while dislocation structures and their evolution were characterized using the dislocation extraction algorithm (DXA) implemented in OVITO.