The save_phonopy() method can not generate correct FORCE_CONSTANT file for phonopy when dealing multi-component systems.

[wftao1995]

The save_phonopy() method could not work correctly when multiple species exist. This is because of the supercell created by the code is inconsistent with phonopy's supercell conversion (in which atoms are sorted according to their types), and one get incorrect results when using FORCE_CONSTANTS generated directly.

[DjordjeDangic]

You can try using this hack:

import cellconstructor as CC
import cellconstructor.Phonons
import cellconstructor.ForceTensor
import cellconstructor.Methods

import phonopy
from phonopy import Phonopy
from phonopy.structure.atoms import PhonopyAtoms

dyn = CC.Phonons.Phonons(dyn_prefix, nqirr)
unitcell = PhonopyAtoms(symbols=dyn.structure.atoms, cell=dyn.structure.unit_cell, positions=dyn.structure.coords)
phonon = Phonopy(unitcell, dyn.GetSupercell(), primitive_matrix = np.eye(3).tolist())
tensor2 = CC.ForceTensor.Tensor2(dyn.structure, dyn.structure.generate_supercell(dyn.GetSupercell()), dyn.GetSupercell())
tensor2.SetupFromPhonons(dyn)
phonon.force_constants = CC.Methods.tensor2_to_phonopy_fc2(tensor2, phonon)
phonon.save(settings={'force_constants': True})

Is this what you need?

[adenchfi]

Possibly for the same reason, it appears that in my case (PHONOPY supercell with multiple species) reading PHONOPY phonons/FORCE_CONSTANTS by following the User Guide

dyn = CC.Phonons.Phonons()
dyn.load_phonopy("phonopy.yaml")
dyn.save_qe("new_dyn")
dyn.save_phonopy("./tmp/")

and then re-saving them (both in QE form and again FORCE_CONSTANTS form) produces a different result than the original FORCE_CONSTANTS. The resulting phonon frequencies from phonopy and from CellConstructor are different (in CellConstructor they are negative even after symmetrization).

What I would like to do is convert a PHONOPY output to QE dynamical matrices.