Converged Phonon Frequency

[p.williams5]

Hi EPW users,

I am trying to calculate the electron phonon coupling matrix close to conduction band minimum in Si in order to calculate the acoustic deformation potential, however I have run into a problem. For a set polar angle between the electron and phonon vectors I am sweeping through a full circle of azimuthal angles and calculating the electron-phonon coupling matrix as well as the phonon frequencies. When calculating the phonon frequencies along this path using QE, they have the same symmetry as the BZ however this is not the case for the results calculated using EPW.

Attached is a figure showing a comparison of phonon frequencies calculated using QE and EPW for phonon wavevectors with a polar angle of 3π/10 from the electron wavevector as the azimuthal angle was varied from 0 to 2π. As you can see, the data calculated using EPW has a large peak in the curve before it comes down and is closer to the phonon frequencies given by QE. This example is indicative of all the sets of phonon frequencies calculated for Si regardless of the polar angle. This was calculated using the following epw input.

&inputepw
prefix = 'Si'
amass(1) = 28.0855
outdir = './'
dvscf_dir = '../../phonon/save'

elph = .true.
epwwrite = .true.
epwread = .false
lpolar = .false.
vme = 'dipole'

wannierize = .true.
nbndsub = 8
num_iter = 300
auto_projections = .true.
scdm_proj = .true.

prtgkk = .false.

band_plot = .true.
filqf = 'q_path.dat'
filkf = 'k_path.dat'

nk1 = 16
nk2 = 16
nk3 = 16
nq1 = 8
nq2 = 8
nq3 = 8
/


I have tried using different projections, including centred on the bonds and scdm to calculate them automatically but these make very little difference if any at all. Is there anything that can be done to improve the calculation other than increasing the number of q and k points?

Thank you,
Patrick.