I’m trying to learn and reproduce the examples from this paper and had several questions. The example jupyter notebooks are also on matgenb (“Data-driven First Principles Methods…”).
Following the example in part 1: I found the primitive cell, enumerated the structures with enumlib, generated VASP input files and ran the structure relaxations in VASP.
So the conventional cell is cubic. But since the structure relaxation is done on the primitive cell, the lattice params and angles are all changed after the relaxation. Which means the symmetry is broken right? Also, how would you use pymatgen to convert back to the conventional cell? And if you try to convert back to the conventional cell now, wouldn’t that result in a non-cubic structure?
For part 3, would I essentially take the CONTCAR from the lowest energy structure in part 1 and use it as the POSCAR for the AIMD simulation? Is there any additional cell ordering required in this case? And I guess I could generate the remaining VASP input files for this step using the “MPMDSet”?