Datasets:
image imagewidth (px) 800 1.35k |
|---|
Polyanion Sodium-Ion Cathode Materials (DFT)
298,144 VASP structures for four polyanion Na-ion cathode families — olivine, maricite, silicate, and alluaudite — with transition metals Fe, Mn, Co, Ni.
Published in Scientific Data (2025) · Zenodo archive
| Total structures | 298,144 |
| With Bader charges | 113,532 |
| Format | Extended XYZ (ASE-compatible) |
| Method | VASP 6.4, PBE+U |
| MD temperature | 1000 K |
| Repository size | ~2.6 GB (28 files) |
Quick start
1. Download one file (no account needed):
curl -L -O "https://huggingface.co/datasets/Mahpe/polyanion-sodium-cathode-dft/resolve/main/polyanion_cathode_dataset/NaMPO4_olivine_single_optimization.xyz"
2. Read structures with ASE:
from ase.io import read
atoms_list = read("NaMPO4_olivine_single_optimization.xyz", index=":")
for atoms in atoms_list:
energy = atoms.get_potential_energy()
forces = atoms.get_forces()
magmoms = atoms.get_array("magmoms")
if "bader_charge" in atoms.arrays:
charges = atoms.get_array("bader_charge")
3. Browse or download more → see Download below, or open the Files tab.
What's in the dataset
Cathode families
| Material | Unit cell | Supercell (MD) |
|---|---|---|
| NaTMPO₄ (olivine) | 28 atoms | 1×2×2 |
| NaTMPO₄ (maricite) | 28 atoms | 1×2×2 |
| Na₂TMSiO₄ (silicate) | 32 atoms | 1×2×2 |
| Na₂.₅₆TM₁.₇₂(SO₄)₃ (alluaudite) | 77 atoms | 1×1×2 |
Properties per structure
Energy (eV) · forces (eV/Å) · stress (GPa) · magnetic moments (μB) · Bader charges (e, on optimized & MD structures)
Sampling methods
| Method | Description |
|---|---|
| Structure optimization | Lowest-energy configurations |
| AIMD | Ab initio MD at 1000 K |
| ML-driven MD | CURATOR + PaiNN active learning at 1000 K |
| Ionic steps | Optimization trajectories without Bader charges |
File guide
Data is split into 28 separate XYZ files — download only what you need.
Folders
| Folder | Structures | Bader charges |
|---|---|---|
polyanion_cathode_dataset/ |
113,532 | Yes |
polyanion_cathode_dataset_with_optimization_steps/ |
298,144 (incl. ionic steps) | On final frames only |
overview.txt |
— | Structure counts per file |
Naming convention
{MaterialPrefix}_{sampling_type}.xyz
| Filename contains | Content |
|---|---|
_single_optimization |
Single-TM, optimized structures |
_single_AIMD |
Single-TM, AIMD trajectories |
_single_MD_sampling |
Single-TM, ML-driven MD |
_multiple_optimization |
Multiple-TM, optimized structures |
_single_total |
All single-TM data combined (opt + AIMD + ML) |
_single_structural_optimization |
Single-TM ionic steps (no Bader) |
_multiple_structural_optimization |
Multiple-TM ionic steps (no Bader) |
| Prefix | Material |
|---|---|
NaMPO4_olivine_ |
NaTMPO₄ olivine |
NaMPO4_maricite_ |
NaTMPO₄ maricite |
Na2MSiO4_silicate_ |
Na₂TMSiO₄ |
Na2M2SO4_alluadite_ |
Na₂.₅₆TM₁.₇₂(SO₄)₃ |
Full file list (click to expand)
polyanion_cathode_dataset/ (with Bader charges)
| File | Material |
|---|---|
NaMPO4_olivine_single_optimization.xyz |
Olivine |
NaMPO4_olivine_single_AIMD.xyz |
Olivine |
NaMPO4_olivine_single_MD_sampling.xyz |
Olivine |
NaMPO4_olivine_single_total.xyz |
Olivine (combined) |
NaMPO4_olivine_multiple_optimization.xyz |
Olivine |
NaMPO4_maricite_single_optimization.xyz |
Maricite |
NaMPO4_maricite_single_AIMD.xyz |
Maricite |
NaMPO4_maricite_single_MD_sampling.xyz |
Maricite |
NaMPO4_maricite_single_total.xyz |
Maricite (combined) |
NaMPO4_maricite_multiple_optimization.xyz |
Maricite |
Na2MSiO4_silicate_single_optimization.xyz |
Silicate |
Na2MSiO4_silicate_single_AIMD.xyz |
Silicate |
Na2MSiO4_silicate_single_MD_sampling.xyz |
Silicate |
Na2MSiO4_silicate_single_total.xyz |
Silicate (combined) |
Na2MSiO4_silicate_multiple_optimization.xyz |
Silicate |
Na2M2SO4_alluadite_single_optimization.xyz |
Alluaudite |
Na2M2SO4_alluadite_single_AIMD.xyz |
Alluaudite |
Na2M2SO4_alluadite_single_MD_sampling.xyz |
Alluaudite |
Na2M2SO4_alluadite_single_total.xyz |
Alluaudite (combined) |
Na2M2SO4_alluadite_multiple_optimization.xyz |
Alluaudite |
polyanion_cathode_dataset_with_optimization_steps/ (incl. ionic steps)
| File | Material |
|---|---|
NaMPO4_olivine_single_structural_optimization.xyz |
Olivine |
NaMPO4_olivine_multiple_structural_optimization.xyz |
Olivine |
NaMPO4_maricite_single_structural_optimization.xyz |
Maricite |
NaMPO4_maricite_multiple_structural_optimization.xyz |
Maricite |
Na2MSiO4_silicate_single_structural_optimization.xyz |
Silicate |
Na2MSiO4_silicate_multiple_structural_optimization.xyz |
Silicate |
Na2M2SO4_alluadite_single_structural_optimization.xyz |
Alluaudite |
Na2M2SO4_alluadite_multiple_structural_optimization.xyz |
Alluaudite |
Combined.xyzis not included. Skip*_single_total.xyzif you already download the three split single-TM files.
Download
Base URL (replace <path> with any file above):
https://huggingface.co/datasets/Mahpe/polyanion-sodium-cathode-dft/resolve/main/<path>
One file
curl -L -O "https://huggingface.co/datasets/Mahpe/polyanion-sodium-cathode-dft/resolve/main/overview.txt"
Invoke-WebRequest -Uri "https://huggingface.co/datasets/Mahpe/polyanion-sodium-cathode-dft/resolve/main/overview.txt" -OutFile overview.txt
Several files (Python)
pip install huggingface_hub
from huggingface_hub import hf_hub_download
REPO = "Mahpe/polyanion-sodium-cathode-dft"
for filename in [
"overview.txt",
"polyanion_cathode_dataset/NaMPO4_olivine_single_AIMD.xyz",
]:
print(hf_hub_download(repo_id=REPO, repo_type="dataset", filename=filename))
One folder (Git sparse checkout)
git lfs install
git clone --depth 1 --filter=blob:none --sparse \
https://huggingface.co/datasets/Mahpe/polyanion-sodium-cathode-dft
cd polyanion-sodium-cathode-dft
git sparse-checkout set polyanion_cathode_dataset
git lfs pull
Full dataset (~2.6 GB)
git lfs install
git clone https://huggingface.co/datasets/Mahpe/polyanion-sodium-cathode-dft
from huggingface_hub import snapshot_download
snapshot_download(
repo_id="Mahpe/polyanion-sodium-cathode-dft",
repo_type="dataset",
local_dir="polyanion-sodium-cathode-dft",
)
Trained MLIPs
Nineteen PaiNN / cPaiNN checkpoints trained on this dataset are published on Hugging Face:
Mahpe/polyanion-pretrained-mlips (~1.0 GB)
Includes PaiNN/cPaiNN ensembles plus fine-tuned CHGNet and MACE-MP-0 foundation models (scratch + fine-tuned), with validation MAE training curves and benchmark figures from AIDi (2026).
git lfs install
git clone https://huggingface.co/Mahpe/polyanion-pretrained-mlips
Figures & structure counts
Figure 1. Phase space of the dataset — unit cells of (a) Na₂TMSiO₄, (b) Na₂.₅₆TM₁.₇₂(SO₄)₃, (c) NaTMPO₄ olivine, (d) NaTMPO₄ maricite. Scientific Data (2025).
Table 3.2. Single-TM structures (108,267 total) — Opt, AIMD, and ML counts per material.
Table 3.3. Multiple-TM structures (5,265 total).
Table 3.4. Ionic optimization steps without Bader charges (184,612 total).
Citation
If you use this dataset, please cite:
@article{Petersen2025polyanion,
title={Dataset exploring the atomic scale structure and ionic dynamics of polyanion sodium cathode materials},
author={Petersen, Martin Hoffmann and Chang, Jin Hyun and Bhowmik, Arghya and Garc{\'i}a Lastra, Juan Maria},
journal={Scientific Data},
volume={12},
pages={1440},
year={2025},
doi={10.1038/s41597-025-05799-8}
}
Links
| Resource | URL |
|---|---|
| Paper | https://www.nature.com/articles/s41597-025-05799-8 |
| Zenodo | https://doi.org/10.11583/DTU.27202446.v5 |
| Generation workflow | https://github.com/dtu-energy/cathode-generation-workflow |
| Example scripts | https://github.com/dtu-energy/cathode-generation-workflow/tree/main/extract_data |
- Downloads last month
- 51