Dataset, models, and tests for benchmarking InP potentials
Tests:
- Lattice Constant
- Equation of State
- Elastic Constants
- Phonon Spectrum
- Point Defect Formation Energies
- Point Defect Migration Barriers
- Stacking Fault Energies
- Dislocation Quadrupole Formation Energies
- Dislocation Quadrupole Migration Barriers
Basic requirements stored in requirements.txt. Also requires the LAMMPS Python module & executable installed with the ML-PACE (for ACE models), ML-SNAP (for EME-SNAP), and MANYBODY (for Vashishta) packages installed. Models are driven by the ASE LAMMPSlib calculator interface.
Dataset: Holds the InP Dataset in .xyz format, as well as the train and validation partitions used in MACE training
DFT_Convergence: Has details of the DFT convergence study to determine cutoff energy and k-point density
Potentials: Holds input files for each of the bespoke models tested: new ACE and MACE models trained on the dataset in Dataset/, as well as the Branicio Vashishta model and the EME-SNAP model
Test_Plots, Test_Results: Directories which hold the plots and results from the testing framework
Tests: Python files containing the tests for each potential