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
NMC622 — LiNi₀.₆Mn₀.₂Co₀.₂O₂ Cathode (DeepMD)
What this is
DeepMD-kit potentials for NMC622 (LiNi₀.₆Mn₀.₂Co₀.₂O₂), a layered-oxide Li-ion cathode material — one of the most widely used commercial cathode chemistries, here used to study Li-ion diffusion in the layered structure.
Transport properties
| AIMD | MLMD (DeepMD) | |
|---|---|---|
| D (Li⁺, cm²/s) | 9.0×10⁻¹² | 2.58×10⁻¹¹ (300K) / 5.12×10⁻¹¹ (600K) |
| Eₐ (eV) | 0.34 | 0.035 |
Mechanical: E = 200 GPa, ν = 0.30, ρ = 4.7 g/cm³.
Files & Validation
| File | RMSE energy (eV/atom) | RMSE force (eV/Å) | Training steps | Size |
|---|---|---|---|---|
model/nmc622_production.pb |
0.00239 | 0.166 | 1,000,000 | 450.6 MB |
model/nmc622_model1.pb |
0.00226 | 0.178 | 1,000,000 | 407.6 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.
2 of 3 current production potentials, the pair with the lowest validation force RMSE; several earlier iterations (6k/11k/22k-atom cell-size scans, memory-optimization variants) exist in the source tree as superseded intermediates and aren't included here.
Training pipeline
AIMD random-structure sampling (VASP, PBE, 500K) → DeepMD-kit train/freeze/ compress. Produced by HPCA (github.com/selvachandrasekaranselvaraj/hpca).
Citation
Selva Chandrasekaran Selvaraj, University of Illinois Chicago. Continuum transport model parameters per Ncube et al. 2026.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support