# HME21: High-Temperature Multi-Element DFT Training Set for the PFP Universal Potential, ColabFit Conversion

## Facts

- Provider: Takamoto et al. (Preferred Networks and ENEOS); ColabFit conversion
- License: CC-BY-4.0 (https://figshare.com/articles/dataset/High-temperature_multi-element_2021_HME21_dataset/19658538)
- Size: 4 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/colabfit/HME21_train
- Access: public, readable without an account
- DOI: https://doi.org/10.60732/845cc1b5
- Category: Chemistry & Materials
- Source: https://doi.org/10.1038/s41467-022-30687-9
- Page: https://scigantic.com/archive/c0ce2818-8e5e-5072-b5c2-4aa5bded3b3c
- Keywords: Chemistry & Materials, DFT, Parquet, Hugging Face

## About

The training split of HME21 (high-temperature multi-element 2021), the dataset Preferred Networks and ENEOS released with PreFerred Potential (PFP), a universal neural network potential. HME21 deliberately samples disordered and unstable structures, irregular substitutions, and a range of temperatures and densities across 37 elements, from alkali and alkaline earth metals through 3d, 4d and 5d transition metals to C, N, O, F, P, S and Cl. Each configuration is labeled with a DFT total energy and per-atom forces. The ColabFit card counts 19,954 configurations and about 555,000 atoms; the repository is about 70 MB (measured 2026-11-19). Validation and test splits are separate ColabFit repositories (HME21_validation, HME21_test). This is the ColabFit Exchange conversion: configurations are Parquet files under co/ (structure, positions, cell, computed properties and metadata per row) with dataset-level information in ds.parquet; read them with pyarrow, pandas, polars or Hugging Face datasets, following the ColabFit configuration schema at materials.colabfit.org/docs. The original Figshare record is released under CC BY 4.0; cite Takamoto et al., Nature Communications 2022.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("colabfit/HME21_train")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `co/co_0.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/c0ce2818-8e5e-5072-b5c2-4aa5bded3b3c
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/c0ce2818-8e5e-5072-b5c2-4aa5bded3b3c
- Over MCP: `get_schema_card` with id `c0ce2818-8e5e-5072-b5c2-4aa5bded3b3c` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Chemistry & Materials datasets](https://scigantic.com/topics/chemistry-and-materials)
- [DFT datasets](https://scigantic.com/topics/dft)
