Trained MACE machine-learned interatomic potential and DFT reference dataset for Janus BrClM (M = Ti, Zr, Hf) monolayers

This deposit contains the trained MACE machine-learned interatomic potential (MLIP) for Janus BrClM (M = Ti, Zr, Hf) monolayers and its DFT reference dataset: the production model checkpoint, 592 training and 65 validation structures with per-composition breakdown, held-out test sets (finite-temperature AIMD, strain, Cl vacancy), and the training script. Reference energies and forces were computed with VASP (PBE, PAW, 550 eV cutoff); training used AMSGrad, learning rate 5 × 10⁻⁴, 80 epochs, loss weights E/F/S = 1.5/10/5, random seed 123. Companion dataset to the PCCP manuscript on BrClM monolayers.