Cu(100) validation
Cu adatom hop on Cu(100). Each row is one model's as-predicted barrier from the live catalog, set next to the references. Models are listed from lowest to highest barrier. We report the deviation and make no ranking claim.
| Model | Barrier (eV) | vs experiment 0.40 (eV) | vs GGA DFT 0.52 (eV) |
|---|---|---|---|
| chgnet 0.3.8 | 0.497 | +0.097 | -0.023 |
| allegro-oam-l 0.8.3 | 0.531 | +0.131 | +0.011 |
| mattersim-5m 1.2.5 | 0.561 | +0.161 | +0.041 |
| mace-mpa-0 0.3.16 | 0.573 | +0.173 | +0.053 |
| alphanet-v1-oam 0.0.1 | 0.595 | +0.195 | +0.075 |
| orb-v3 0.7.0 | 0.604 | +0.204 | +0.084 |
| mace-mp-0 0.3.16 | 0.621 | +0.221 | +0.101 |
References
- Experiment: 0.40 +/- 0.05 eV, experimental consensus, four determinations, reviewed in arXiv:cond-mat/9704009
- GGA DFT: 0.52 +/- 0.05 eV, docs/validation.md reference set
- Source file: derived/cu100/experimental_overlay.json
Caveats
- Dimer mobility matters near room temperature; an i=1 KMC is a documented approximation (PRB 73, 245405).
- Island-density N(T) points from Pomeroy and Brock are not yet digitized (paywalled source), so there is no KMC-level comparison yet.
- One system only. More systems need the runner generalized past Cu.