Data for Message-Passing Neural Network (MPNN) Simulations of Magnetic Phase Transitions in CrX3(X = I, Br, Cl)

This study introduces a Magnetic Message-Passing Neural Network (MMPNN) to precisely model the potential energy surfaces of magnetic materials. The MMPNN captures interactions between atomic magnetic moments via the message-passing mechanism, decomposing the total energy of magnetic systems into magnetic and non-magnetic contributions to facilitate Hamiltonian modeling. Employing two-dimensional chrominium trihalides CrX3 (X = I, Br, Cl) as case studies, the MMPNN is integrated with the Landau-Lifshitz-Gilbert (LLG) equation to simulate magnetic phase transitions. Furthermore, this work illustrates the influence of strain on the magnetic ground state and phase transition of CrCl3 under 5% biaxial compressive strain, highlighting the promising applications of MPNNs in magnetic systems.