Machine learning force-field model for phase separation dynamics in Jahn-Teller models for manganites

We present a scalable machine learning (ML) force-field model to explore the large-scale dynamics of Jahn-Teller coupling between the electronic e_g orbitals and the molecular tetrahedral distortions in manganites. While the equilibrium properties of the electronic JT model have been extensively studied before, to the best of our knowledge, the phase-ordering dynamics of this important model remains unexplored. As the driving forces on the local lattice distortions come from the electron degrees of freedom, their calculation requires solving the electron band structure at every time-step of the simulation, which could be prohibitively expensive for large systems. Assuming the principle of locality for electron systems, a neural-network model is developed to accurately and efficiently predict local electronic forces with input from lattice degrees of freedom in the immediate neighborhood. Our ML-enabled large-scale thermal quench simulations of hole-doped systems uncover anomalous phase separation phenomenon intertwined with unusual coarsening dynamics of antiferromagnetic lattice distortion. Our work opens new avenues for multi-scale dynamical modeling of correlated electron systems.