Manually enforced treshold for detection of the imaginary part of real space force constant

[gjoalland13]

An error is raised when extracting the force tensor out of the attached dynamical matrix, when using the following code with the default img_trsh=1.0e-6:

import cellconstructor as CC, cellconstructor.Phonons
import cellconstructor.ForceTensor
PATH = "GZ"
N_POINTS = 10
HARM_DYN = "dynamical-matrix-"
SPECIAL_POINTS = {"G": [0 ,0 ,0], "Z": [0, 0.5, 0]}
NQIRR = 8

harmonic_dyn = CC.Phonons.Phonons(HARM_DYN, NQIRR)
qpath, data = CC.Methods.get_bandpath(harmonic_dyn.structure.unit_cell, PATH, SPECIAL_POINTS, N_POINTS)
harmonic_dispersion = CC.ForceTensor.get_phonons_in_qpath(harmonic_dyn, qpath)

The following issue is raised:

Traceback (most recent call last):
  File "/home/gjoalland/work/WO3/jupyter/phonon/plot_phonons.py", line 63, in <module>
    harmonic_dispersion = CC.ForceTensor.get_phonons_in_qpath(harmonic_dyn, qpath)
  File "/home/gjoalland/work/.venv-flare/lib/python3.9/site-packages/cellconstructor/ForceTensor.py", line 3072, in get_phonons_in_qpath
    t2.SetupFromPhonons(dynamical_matrix)
  File "/home/gjoalland/work/.venv-flare/lib/python3.9/site-packages/cellconstructor/ForceTensor.py", line 196, in SetupFromPhonons
    super_dyn = current_dyn.GenerateSupercellDyn(phonons.GetSupercell(), img_thr  =1e-6)
  File "/home/gjoalland/work/.venv-flare/lib/python3.9/site-packages/cellconstructor/Phonons.py", line 1921, in GenerateSupercellDyn
    dyn_supercell.dynmats[0] = self.GetRealSpaceFC(supercell_size, img_thr = img_thr)
  File "/home/gjoalland/work/.venv-flare/lib/python3.9/site-packages/cellconstructor/Phonons.py", line 2782, in GetRealSpaceFC
    fc = GetSupercellFCFromDyn(dynmat, np.array(self.q_tot), self.structure, super_structure, img_thr = img_thr)
  File "/home/gjoalland/work/.venv-flare/lib/python3.9/site-packages/cellconstructor/Phonons.py", line 4083, in GetSupercellFCFromDyn
    assert imag < img_thr, ASSERT_ERROR.format(imag)
AssertionError: 
    Error, the imaginary part of the real space force constant
    is not zero. IMAG=1.2811627450721814e-06

I fixed this error by manually changing img_thr=2.0e-6 in the ForceTensor.py file.
dynamical-matrix-0.txt
dynamical-matrix-1.txt
dynamical-matrix-2.txt
dynamical-matrix-3.txt
dynamical-matrix-4.txt
dynamical-matrix-5.txt
dynamical-matrix-6.txt
dynamical-matrix-7.txt
dynamical-matrix-8.txt

[mesonepigreco]

Yes, this is quite tricky, as usually this error indicates that there is some numerical noise in the way q points are defined in Cartesian coordinates. Increasing the img_thr may solve one specific test, but not all cases.

The solution would be to add a function that cleans the float truncation arising from finite precision in the printing of the q points of the Quantum Espresso files (which have single precision) to satisfy the correct q points computed from the primitive cell.
There should be a function to do that in the code; we need to check it carefully and call it automatically from the function that reads the dynamical matrix from a file.