Correction layers for GRACE-2L-SMAX-OMAT-large
Additive correction layers for 13 elements (Ag, Al, Au, Ca, Cu, Mg, Mo, Ni, Pa, Ti, U, V, Zr).
Each layer is a linear atomic cluster expansion (ACE) that is added to the energy of the foundation model GRACE-2L-SMAX-OMAT-large.
The layers are fitted so that the corrected model reproduces the transition temperatures of the SGTE unary assessment at zero pressure.
The base model is not modified.
Use in LAMMPS
LAMMPS must be built with the GRACE (pair_style grace) and ML-PACE (pair_style pace) packages.
The layer is combined with the base model through hybrid/overlay:
pair_style hybrid/overlay grace pace
pair_coeff * * grace /path/to/GRACE-2L-SMAX-OMAT-large Cu
pair_coeff * * pace layers/Cu_correction.yace Cu
example.lammps contains a complete input.
Each layer is valid only together with GRACE-2L-SMAX-OMAT-large and only for the element it was fitted for.
The layers are for single elements; they contain no terms for mixed-element environments.
Files
| File | Content |
|---|---|
layers/<El>_correction.yace |
layer in the format read by pair_style pace |
layers/<El>_correction.yaml |
the same layer in the python-ace potential format |
manifest.json |
basis size, cutoff and transition temperatures per element |
SHA256SUMS |
checksums of the layer files |
example.lammps |
example LAMMPS input |
All layers use a cutoff of 6.5 Å and simplified spherical Bessel radial functions. The basis contains 62 functions (Ag: 78, Au: 136).
Transition temperatures at zero pressure
Mean and standard deviation over independent reversible scaling sweeps (four sweeps; three for Ag and Au).
| Element | Transition | SGTE (K) | GRACE (K) | GRACE + layer (K) |
|---|---|---|---|---|
| Al | melting | 933 | 926 | 933 ± 2 |
| Cu | melting | 1358 | 1270 | 1359 ± 2 |
| Ni | melting | 1728 | 1681 | 1736 ± 5 |
| Ag | melting | 1235 | 991 | 1228 ± 5 |
| Au | melting | 1337 | 842 | 1331 ± 4 |
| Mg | melting | 923 | 899 | 927 ± 2 |
| Ca | fcc–bcc | 716 | 471 | 701 ± 21 |
| Ca | melting | 1115 | 1078 | 1105 ± 5 |
| Ti | hcp–bcc | 1155 | 1283 | 1149 ± 14 |
| Ti | melting | 1941 | 1788 | 1941 ± 9 |
| Zr | hcp–bcc | 1139 | 1130 | 1101 ± 8 |
| Zr | melting | 2128 | 1947 | 2131 ± 4 |
| V | melting | 2183 | 2149 | 2187 ± 3 |
| Mo | melting | 2896 | 2791 | 2891 ± 4 |
| U | melting | 1408 | 1236 | 1405 ± 8 |
| Pa | melting | 1845 | 1832 | 1842 ± 12 |
Scope and known limitations
- The layers were fitted at zero pressure. At finite pressure the transition lines follow the pressure dependence of the base model; for Ti the hcp–bcc temperature of the corrected model does not decrease with pressure as observed experimentally.
- The layers change the free energy differences between phases; they do not correct the transition entropies of the base model.
- Properties of the base model that the layers do not address remain: bcc V has a negative C44 in
GRACE-2L-SMAX-OMAT-large, bcc U is mechanically unstable below about 900 K, and fcc Pa is lower in energy than the experimental ground state α-Pa. - The layers leave the 0 K lattice parameters, bulk moduli and relative energies of competing crystal structures close to those of the base model.
Citation
If you use these layers, please cite the accompanying publication (reference to be added) and the GRACE foundation model.