# MSR-ACC/TAE25: 73k CCSD(T)/CBS Total Atomization Energies

## Facts

- Provider: Microsoft Research AI for Science
- License: CDLA-Permissive-2.0 (https://zenodo.org/records/15387280)
- Size: 5 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/microsoft/msr-acc-tae25
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.15387280
- Category: Chemistry & Materials
- Source: https://arxiv.org/abs/2506.14492
- Page: https://scigantic.com/archive/568dcef8-4695-55d9-8e26-a869459ccd2f
- Keywords: Chemistry & Materials, DFT, quantum chemistry, graph learning, Parquet, Hugging Face

## About

MSR-ACC/TAE25 is part of the Microsoft Research Accurate Chemistry Collection: 73,040 total atomization energies at the CCSD(T)/CBS level obtained with the W1-F12 thermochemical protocol. It exhaustively covers closed-shell, charge-neutral, covalently bound equilibrium molecules with up to five non-hydrogen atoms drawn from elements up to argon and without significant multireference character. The publishers built it to train and evaluate machine-learned exchange-correlation functionals, density functional and semi-empirical methods, and graph neural networks beyond organic chemistry. Molecules are stored in QCSchema-compatible columns (symbols, geometry, charge, multiplicity) with reference TAE labels (tae@w1-f12), component energies, singlet-triplet gaps and molecular graphs. The repository holds three Parquet files (all, train, valid), about 0.13 GB (measured 2026-11-14); open them with pandas, pyarrow or datasets. Released by Microsoft under CDLA-Permissive-2.0 (Zenodo 10.5281/zenodo.15387280).

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("microsoft/msr-acc-tae25")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `msr-acc-tae25-all.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/568dcef8-4695-55d9-8e26-a869459ccd2f
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/568dcef8-4695-55d9-8e26-a869459ccd2f
- Over MCP: `get_schema_card` with id `568dcef8-4695-55d9-8e26-a869459ccd2f` 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)
- [Quantum chemistry datasets](https://scigantic.com/topics/quantum-chemistry)
- [Graph learning datasets](https://scigantic.com/topics/graph-learning)
