[Pw_forum] Phonon DOS in noncubic materials

[Paolo Giannozzi]

Stefano Baroni wrote:

> I think (others may want to confirm or deny) that this depends of the 
> non-analytic behavior of the dynamical matrix in the long-wavelength 
> limit. Because of this, phonon frequencies at q=0 are strictly speaking 
> not defined. Their limit as q->0 is indeed well defined only once the 
> direction of q and of the phonon polarization is specified. For cubic 
> materials, this non-analiticity gives rise to the well known LO-TO 
> splitting, whereas for crystals of lower symmetry the situation is more 
> complicated. If the DOS is calculated using a very fine mesh of q 
> points, I would expect that the result would be independent on what 
> happens around q=0. If, instead, too few points are used, I would not be 
> surprised if the above-mentioned non-analytic behavior would severly 
> undermine the accuracy of any sensible interpolation procedure needed 
> for any integration (such as the one used for DOS calculations). Hope 
> this is not too wrong, and of some help as well. SB

confirmed. I wrote some time ago the warning about phonon dos in
noncubic materials, because it wasn't clear to me what to do at
q=0 in that case. I don't expect a large error, since this is a
problem at just a single q-point, but it is safer to advertise
it.

Paolo
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy