Can I develop interatomic potentials using machine learning in VASP and use them in MD simulations in LAMMPS?
I am interested in generating machine learning-based interatomic potentials (MLIPs) using VASP and then utilizing these potentials for molecular dynamics (MD) simulations in LAMMPS. Specifically, I would like to know:
Can VASP train machine learning potentials, such as those based on neural networks or Gaussian process regression?
If so, what methods (e.g., MLIP, MTP, or GAP) are compatible with VASP?
How can we create the datasets for vasp?
How can these machine-learned potentials be exported and formatted for use in LAMMPS?
Are there any known limitations or challenges in transferring these potentials between VASP and LAMMPS?
I would appreciate any insights or references related to this.
