The ABINIT Users Mailing List ( CLOSED )

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

Hello Alaska,

I haven't seen your figures (probably more revealing than your inputs) but 
I suspect that the following is happening: the linewidths are interpolated 
along your chosen path in q space by a Fourier method, just like for the 
phonon bands. Empirically we have found that the linewidths are less 
amenable to this kind of interpolation, as they behave less sinusoidally. 
In particular this means you get a kind of aliasing and the interpolated 
lwd go negative, especially near gamma. We should try a more neutral 
interpolation like bilinear (do you have some spare time? :)

This can get worse with increasing kpoint sampling, since the 
non-sinusoidal behavior is resolved more and more precisely: the 
(eventually negative) oscillations should be confined to smaller and 
smaller intervals, since you have the exact values on the FT grid (the q 
grid), but they can get larger.

One option is to use the data labeled _QPT in the output and output files. 
This was added by Matteo Giantomassi, and calculates quantities without 
interpolation (just sums over the existing qpoints) so it its immune from 
interpolation problems.

Keep us posted, and eventually send me a figure of the lwd and the phonon 
bands if you want more diagnostic (no need to send it to forum)

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