Prototype structure in ClusterSpace class
So I am using phonopy to find the harmonic frequencies and due to some inconsistency in my own code I found that depending on if I pass the supercell structure or primitive structure to ClusterSpace I get a non negligable difference in the end results.
From the documentation:
- prototype_structure (ase.Atoms) – prototype structure; spglib will be used to find a suitable cell based on this structure unless the cell is already a primitive cell.
I figured that using the supercell as the prototype structure should produce the more accurate end result as both hiPhive and phonopy work with the same primitive structure in that case. This is my interpretation I'd like to have your opinion on the inconsistency as well. I realize that the problem might also be related to how I generate the supercell (see code)
atoms_ideal = prim.repeat(cell_size) cell = atoms_ideal.get_cell() atom = sorted(atoms_ideal, key=lambda atom: atom.symbol) # Vasp wants the cell to be sorted atoms_out = ase.Atoms(atom, cell=cell, pbc=True) atoms_out.center()