Repo contains codes for three essential steps 1- create training data; 2- fit a ACE model to that data; 3- Evaluate ACE pot using LAMMPS (needs work)
-
(01 step) Using ASE's EMT calculator; vacancy migration path is calculated using NEB and the final structures are outputted to .extxyz file
-
(02 step)Use ACE1pack to fit a ACE model to the following structures:
vacancy( 5 images + initial state), bulk (4x4x4 reps of unit cell), surfaces: fcc111 of reps of 2, 3, 4, 5 , isolated atom, clusters: icosahedron, octahedron twist sigma17 (001) GB tilt sigma17(014), sigma5(013) GB -
(03-1 step) Check if your local LAMMPS installation has ML-PACE package[https://docs.lammps.org/Packages_details.html#pkg-ml-pace] installed properly.
-
(03-2 step) use the .yace file generated from step 02 to run a simple LAMMPS minimization of: -- monovacancy in Cu (minimization was ok) -- sigma5(012) GB (minimization was ok)
Note: need to use the latest version of LAMMPS and install 2023 version of ML-PACE library; TODO:
-
Need to think about now adding more structures in here. Maybe GBs and dislocations?
Additional info: Build LAMMPS (22 Dec 2022 version) using make and install additional packages such as ML-PACE, PYTHON and MANYBODY to run ACE and EAM in LAMMPS via Python.
Each folder has a bash script that allows one to run python or Julia codes.