Spin-aware quantum-accurate interatomic potentials for heavy elements

Studying matter under extreme conditions using density functional theory (DFT) is computationally expensive, since the degrees of freedom and consequently the configurational space grows rapidly with increasing temperature and pressure. Therefore, the use of DFT for such simulations is limited to fairly small simulation cells and time scales. Machine learning-based interatomic potentials (ML-IAP) provide access to much larger spatial and temporal domains, thus enabling the discovery of new and exotic magnetic materials. A majority of existing descriptors required to construct ML-IAPs neglect the spin degrees of freedom. Here, we present our preliminary ideas/workflows to construct "spin-aware" ML-IAP using the SNAP[1] descriptors and the coupled spin-molecular dynamics framework implemented in LAMMPS [2]. This modeling capability will complement upcoming experiments to magneto-structural properties in shock- compressed or laser-driven samples at elevated temperatures and pressures exposed to strong, pulsed magnetic fields, which are planned at photon sources such as within the HIBEF consortium at the European XFEL.