Graph Neural Network Force Fields for Spin Dynamics in Metallic Magnets
A GNN force-field model is used to replace repeated electronic solves in metallic magnet spin-dynamics simulations.
The paper says the model learns the effective magnetic energy functional from electronic calculations, then evaluates spin torques efficiently. It is framed as analogous to machine-learned interatomic potentials, but for itinerant spin dynamics. Benchmarks cover collinear, noncollinear, and noncoplanar magnetic order, with learned torques and nonequilibrium dynamics matching direct electronic simulations. Source: ArXiv · AI/CL/LG's note
The paper says the model learns the effective magnetic energy functional from electronic calculations, then evaluates spin torques efficiently. It is framed as analogous to machine-learned interatomic potentials, but for itinerant spin dynamics. Benchmarks cover collinear, noncollinear, and noncoplanar magnetic order, with learned torques and nonequilibrium dynamics matching direct electronic simulations. Source: ArXiv · AI/CL/LG's note
score 4