Provided in this zip file are the inputs for Machine-learned Force Field (MLFF) and their corresponding outputs.
The two folders:
SOAP: Contains the training performed with VASP
1-train: The training inputs (ICONST, INCAR, KPOINTS, POSCAR, POTCAR) with outputs
2-refit: Refit the force field with previously generated ML_ABN with resultand force field files (ML_FFN)
3-EF: The validation of energy and forces on the same aforementioned 1,000 randomly generated K3C60 distortions.
4-phonon: The MLFF phonon calculation using finite displacement method with phonopy
5-elastic: The MLFF prediction of elastic moduli including DFT reference.
ACE: Contains the training performed with FLARE package combined with Quantem Espresso
1-train: The training inputs for FLARE (init.xyz, otf_train.yaml) with concised outputs
2-hyp_scan: The scan for radius cutoff (C-C pair from 3.7 to 5.3 A and C-K pairs from 3.3 to 5.0 A, both with step size of 0.1 A) The input given (offline_train.yaml) is for the optimal in the scan, along with its correspponding force field output files (*.flare) The model likelihood for all other scans are summarized in K3C60_flare_rcutscan.csv
3-EF: The validation of energy and forces on the same aforementioned 1,000 randomly generated K3C60 distortions.
4-phonon: The MLFF phonon calculation using finite displacement method with phonopy. PhonoLAMMPS is used for interfacing
5-elastic: The MLFF prediction of elastic moduli including DFT reference.