Fe–Ni MACE-sqs: model, DFT training data and analysis files

This dataset accompanies the manuscript "Capabilities and Limitations of MACE Interatomic Potentials for Fe–Ni Alloys: A Systematic Benchmark" by Kushal Ramakrishna, Mani Lokamani and Attila Cangi. It contains the MACE-sqs machine-learned interatomic potential for the Fe–Ni binary system and the spin-polarized DFT-PBE training and validation sets it was fitted to: 5115 training and 1274 validation configurations, in cells of 54, 108, 128 and 250 atoms. It also includes the special quasi-random structures (SQS) used in the paper, with their short-range-order analysis, DFT-PBE and experimental magnetic moments across the composition range, a complete VASP input example, and the scripts and notebooks used for the analyses and figures. The DFT reference data were computed with VASP 6.5.1. Simulations used ASE 3.25.0, mace-torch 0.3.13 and torch-sim-atomistic 0.2.1. See README.md for the file layout, data formats, DFT settings and instructions for use.