The ABINIT Users Mailing List ( CLOSED )

[Matthieu Verstraete <mjv500@york.ac.uk>]

Things are becoming clearer:

the alpha2F functions are calculated for frequencies in a certain range, 
which is actually determined in the calculation of the linewidths 
(dependent on your path in qspace). The bounds for the frequency are 
given in the log

> elph_linwid : omega_min, omega_max =   -1.2425615382345503E-003
vs
> elph_linwid : omega_min, omega_max =   -1.0928984516948099E-003

which as you see is a bit different for the 2 qpaths you chose. As a 
result the frequency integration of a2F will have slightly different 
bounds. When you go to denser kpoints, omega_min=0 and everything is fine.

In principle the frequency range should be independent of the qpath, I 
agree, and I might change this. In practice the qpath is a good place to 
look for the omega_min omega_max because the extrema of bands are often at 
zone edges. Plus, I add 10% to the omega_min, omega_max, to have a buffer. 
Not quite enough in your case, maybe because of a soft mode which is off 
the qpath, or because of a too large broadening in the a2F calculation.

We'll see what the best strategy is for the next release. Anyone else 
using the elphon part of the code, please beware. This will be most acute 
for negative phonon modes, and the use of the elphon code is quite 
debatable in this case anyway (not necessarily wrong, but the 
interpretation is not straighforward).

Matthieu

--
================================================================
Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
Dept. of Physics, University of York,     tel: +44 1904 43 22 08
Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14