enghl/formula-to-structure-data

Materials Project 2023 structures and MPtrj

Two files, redistributed under CC BY 4.0 from the Matbench Discovery figshare record 22715158:

File	Content	sha256
mp.jsonl	142,666 Materials Project 2023-02-07 entries with at most 80 sites, one JSON object per line	9644b7e333880b0ed36573e7d7a0717853e5e39e5f23dbadee36a9da987f227d
mptrj.extxyz.zip	MPtrj, unchanged (2024-09-03-mp-trj.extxyz.zip): frames of Materials Project DFT relaxations with energies, forces, stresses and magnetic moments, one extended-XYZ file per material	0f2412e73c37c85d4f98ed01904ee210ce337596357e72cb6841ae64ff11ca3d
train_build_report.json	How mp.jsonl was built: entry counts and the sha256 of each source file

mp.jsonl

Built from 2023-02-07-mp-computed-structure-entries.json.gz (figshare file 40344436) and 2023-01-10-mp-energies.csv.gz (figshare file 49083124). Entries with more than 80 sites or partial occupancies are dropped. Fields:

• id: Materials Project material id
• formula, reduced_formula: cell formula and reduced formula
• lattice: 3x3, rows are the a, b, c vectors (Å)
• species, frac_coords: one element symbol and one fractional coordinate row per site
• energy_per_atom: DFT total energy per atom (eV/atom), uncorrected
• corrected_energy_per_atom: with the MP2020 energy corrections (eV/atom)
• formation_energy_per_atom: corrected formation energy (eV/atom)
• e_above_hull: energy above the MP 2023 convex hull (eV/atom)
• run_type: GGA or GGA+U
• hubbards: U values (eV) by element for GGA+U runs

Credits

• A. Jain et al., "Commentary: The Materials Project: A materials genome approach to accelerating materials innovation", APL Materials 1, 011002 (2013).
• J. Riebesell et al., "A framework to evaluate machine learning crystal stability predictions", Nature Machine Intelligence (2025) (Matbench Discovery).
• B. Deng et al., "CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling", Nature Machine Intelligence 5, 1031 (2023) (MPtrj).