Machine-Learning Laser-Driven Transition Dynamics in Itinerant Spin Systems

Nonequilibrium phase transitions driven by external fields have become a central theme in condensed-matter physics, offering new routes for controlling correlated quantum materials. We investigate a laser-induced Néel-to-stripe transition in an itinerant spin system on a square lattice, where continuous-wave optical driving dynamically reshapes the underlying exchange interactions. Within the adiabatic approximation, Landau-Lifshitz-Gilbert dynamics is coupled to a Lindblad formulation of the driven electronic subsystem to capture the time evolution of coupled spin–electron systems. A generalized machine-learning force-field framework is employed to learn the nonconservative torques generated by out-of-equilibrium electrons, enabling large-scale simulations of laser-driven spin dynamics. The simulations reveal rich transient inhomogeneous states during the transition, demonstrating the potential of machine-learning approaches for multiscale modeling of light-controlled correlated systems.