Engineered Atomic Structures — MLIP Zoo
Collection
DeepMD/MACE potentials and training data for SSE, Na-ion, Li-S, and cathode battery materials research. GPU-trained today! • 10 items • Updated • 1
Na₂O / Na₂O₂ / NaO₂ — Na-Air Battery Discharge Products (DeepMD)
What this is
A single DeepMD-kit potential covering Na-air battery discharge-product phases: Na₂O, Na₂O₂, NaO₂, and their mixed-phase interfaces (Na₂O–Na₂O₂, Na₂O₂–NaO₂, and the three-phase Na₂O–Na₂O₂–NaO₂ junction), trained across a shared composition space so the same model covers pure phases and their interfaces together (7 temperatures, 300–380K, per phase/interface).
Na-air (Na-O₂) batteries discharge by depositing one of these three sodium oxide phases at the cathode; which phase (and how it interconverts at phase boundaries) governs rechargeability and overpotential — hence training one potential across all of them together rather than phase-by-phase.
Transport properties
D (Na, AIMD) = 2.3×10⁻¹⁰ cm²/s, Eₐ = 0.21 eV · D (MLMD) = 2.5×10⁻¹⁰ cm²/s, Eₐ = 0.22 eV. Mechanical: E = 15 GPa, ν = 0.30, ρ = 2.2 g/cm³.
Files & Validation
| File | RMSE energy (eV/atom) | RMSE force (eV/Å) | Training steps | Size |
|---|---|---|---|---|
model/na_air_production.pb |
— | — | — | 27.5 MB |
RMSE values are validation-set (held-out) energy/force error, read directly from each run's DeepMD-kit lcurve.out at its final training step — not re-derived or estimated.
One or more files have no lcurve.out preserved on disk (training log not kept) — size/provenance is real, but no quantitative accuracy number can be reported for those.
Training pipeline
AIMD (VASP, PBE) across all phase/interface configurations → DeepMD-kit train/freeze/compress. Produced by HPCA (github.com/selvachandrasekaranselvaraj/hpca).
Citation
Selva Chandrasekaran Selvaraj, University of Illinois Chicago.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support