Dataset and trained potential used for: "Chemo-mechanical coupling stabilizes mixed Ag$_{x}$Cu$_{1-x}$GaSe$_2$ solar-cell absorbers: Insights from Monte-Carlo simulations assisted by ab initio informed machine-learning potentials"

This Zenodo record contains the data and files used to develop an ACE potential for Ag$_x$Cu$_{1−x}$GaSe$_2$ and to investigate its thermodynamic properties in the submitted manuscript "Chemo-mechanical coupling stabilizes mixed Ag$_{x}$Cu$_{1-x}$GaSe$_2$ solar-cell absorbers: Insights from Monte-Carlo simulations assisted by ab initio informed machine-learning potentials". The deposited files include: (i) the DFT training dataset (input_data.pckl.gzip), (ii) the fitting input file (input.yaml), and (iii) the final ACE potential files (output_potential.yace and output_potential.yaml). During fitting, the relative force weight in the loss function was set to kappa = 0.95 for 600 iterations and then to kappa = 0.05 for 1500 iterations.