I am trying to construct a phase diagrams using Calphy and the Universal Models of Atoms (UMA) potential, but UMA does not have a pair_style potential for lammps. Instead, they use fix-external command as explained in this link. This presents an issue when doing thermodynamic integration with liquids where you need pair_style hybrid/scaled and in reversible scaling. This prompted me to think about building an implementation of Calphy in ASE instead of lammps to avoid this integration issue with lammps. I know that running MD in ASE is very slow, but I am hoping that it is still feasible to calculate a phase diagram using UMA on a GPU or multiple GPUs.
I wonder if you have thought about building an implementation of Calphy in ASE and what challenges you foresee if it is even possible to do such implementation. I understand that this is a very specific question, but I appreciate any thoughts you have about this.
I am trying to construct a phase diagrams using Calphy and the Universal Models of Atoms (UMA) potential, but UMA does not have a pair_style potential for lammps. Instead, they use fix-external command as explained in this link. This presents an issue when doing thermodynamic integration with liquids where you need pair_style hybrid/scaled and in reversible scaling. This prompted me to think about building an implementation of Calphy in ASE instead of lammps to avoid this integration issue with lammps. I know that running MD in ASE is very slow, but I am hoping that it is still feasible to calculate a phase diagram using UMA on a GPU or multiple GPUs.
I wonder if you have thought about building an implementation of Calphy in ASE and what challenges you foresee if it is even possible to do such implementation. I understand that this is a very specific question, but I appreciate any thoughts you have about this.