Negative phonon linewidth !!

[appalakondaiah]

Dear all,

I am curious to know about the flag parameter degaussw!! Is the the same value as in electron-phonon calculations (i.e. degauss in phonon dispersion) or we need to cross check in EPW calculations ??

Also, I am quite surprised and puzzled with negative phonon linewidths in all acoustic modes as well as highest optical mode (occurring at the middle of Gamma to M (0.5,0.0) direction). My wannier spreads are between 1.8 to 1.9 Angstroms for all atoms and my wannier projected band structures in line with standard DFT. Indeed my electron phonon dispersion showing completely positive modes at acoustic level itself. I also checked with different eptemp, and found negative linewidths with different energies. Below I am attaching the part my epw.in for calculating the phonon line widths.

Kindly suggest me for further

elecselfen = .false.
phonselfen = .true.
a2f = .false.

parallel_k = .true.
parallel_q = .false.

fsthick = 2.0 ! eV
eptemp = 10 ! K (same as PRB 76, 165108)
degaussw = 0.25 ! eV

dvscf_dir = '../phonons/save'
filukk = './$prefix.ukk'
filqf = '/home/saint/Work/PWSCF//EPW/New/meshes/path.dat'
nkf1 = 16
nkf2 = 16
nkf3 = 1

nk1 = 8
nk2 = 8
nk3 = 1

nq1 = 8
nq2 = 8
nq3 = 1

with regards
S. Appalakondaiah

[carla.verdi]

Dear S. Appalakondaiah,

You should test your results with respect to degaussw, as it corresponds to the broadening of the delta functions in the calculation of the phonon linewidth.
Regarding the negative phonon linewidths you are getting, I can comment that the formula implemented is not positive definite so it is possible to find negative numbers. This cannot happen if you use the less accurate 'delta-delta approximation' (input delta_approx=.true.).
You could try incresing the fine k-mesh and decreasing degaussw in your calculations and check if the results change.

Best
Carla