Data-driven magneto-inter-atomic potentials for iron-nickel alloys

The development of new functional materials is essential for industrial applications, e.g., iron and nickel alloys for the production of strong permanent magnets. In recent years, this development has been fueled by data-driven frameworks for creating spin-aware machine-learning interatomic potentials (ML-IAP) that can be used to unravel the formation of novel phases of iron-nickel at high pressures and temperatures, which can be quenched to ambient conditions. Here we present a workflow for large-scale spin-lattice dynamics simulations using ML-IAPs coupled with a collective atomic spin model. The spin model and ML-IAP can represent a potential energy surface from which the mechanical forces on the atoms and the precession dynamics of the atomic spins can be computed with near first-principles accuracy. These are parametrized on data generated using first-principles methods—Density Functional Theory (DFT) and Spectral Neighbor Analysis Potential (SNAP) descriptors. The generated spin-aware ML-IAP can be directly used in the LAMMPS package to perform coupled spin-molecular dynamics simulations at large time and length scales not accessible with prevailing modelling techniques.