c-bone/chili100k_strat

Dataset Card for Chili-100K XRD

Experimentally determined crystal structures from CHILI-100K paired with up to 20 of their most intense simulated XRD peaks.

Dataset Details

• Curated by: Bone et al. (University College London)
• Source data: CHILI-100K (from the Crystallography Open Database)
• Used to train: c-bone/CrystaLLM-pi_Chili100K-XRD
• Paper: Discovery and recovery of crystalline materials with property-conditioned transformers (arXiv:2511.21299)
• Repository: GitHub: CrystaLLM-pi
• License: Apache 2.0, following the source data

Dataset Structure

Column	Description
Material ID	Material identifier in the source database
Database	Source database
Reduced Formula	Reduced chemical formula
CIF	Crystal structure in CrystaLLM-π augmented CIF format
condition_vector	Up to 20 of the most intense XRD peaks (CuKα), stored as 2θ positions followed by intensities, min-max normalised and padded with -100
is_novel	True if no matching structure is present in LeMaterial
is_comp_novel	True if the reduced formula is absent from LeMaterial
token_count	Number of CrystaLLM-π tokens in the CIF

Split	Rows
train	11,091
validation	1,500
test	1,500

Dataset Creation

The structures come from CHILI-100K, a curated inorganic subset of the Crystallography Open Database. Within each group of structures with identical cell composition and space group, the structure with the lowest volume per formula unit was retained. XRD patterns were calculated with pymatgen's XRDCalculator at the CuKα wavelength. CIFs longer than 1536 tokens were removed. Novelty is measured against LeMaterial. The test set holds 500 structures present in LeMaterial, 500 unseen structures whose reduced formula is present in LeMaterial, and 500 structures whose reduced formula is absent from LeMaterial and from the training and validation sets. The validation set holds 1,500 structures present in LeMaterial, and the remaining structures form the training set.

Citation

@misc{bone2025discoveryrecoverycrystallinematerials,
      title={Discovery and recovery of crystalline materials with property-conditioned transformers},
      author={Cyprien Bone and Matthew Walker and Bradley A. A. Martin and Kuangdai Leng and Luis M. Antunes and Ricardo Grau-Crespo and Amil Aligayev and Javier Dominguez and Keith T. Butler},
      year={2025},
      eprint={2511.21299},
      archivePrefix={arXiv},
      primaryClass={cond-mat.mtrl-sci},
      url={https://arxiv.org/abs/2511.21299},
}

@inproceedings{friisjensen2024chili,
      title={CHILI: Chemically-Informed Large-scale Inorganic Nanomaterials Dataset for Advancing Graph Machine Learning},
      author={Friis-Jensen, Ulrik and Johansen, Frederik L. and Anker, Andy S. and Dam, Erik B. and Jensen, Kirsten M. {\O}. and Selvan, Raghavendra},
      booktitle={Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
      pages={4962--4973},
      year={2024},
      doi={10.1145/3637528.3671538},
}

@article{grazulis2009cod,
      title={Crystallography Open Database -- an open-access collection of crystal structures},
      author={Gra{\v{z}}ulis, Saulius and Chateigner, Daniel and Downs, Robert T. and Yokochi, A. F. T. and Quir{\'o}s, Miguel and Lutterotti, Luca and Manakova, Elena and Butkus, Justas and Moeck, Peter and Le Bail, Armel},
      journal={Journal of Applied Crystallography},
      volume={42},
      number={4},
      pages={726--729},
      year={2009},
      doi={10.1107/S0021889809016690},
}