The ABINIT Users Mailing List ( CLOSED )

["张�s" <zhangting1980323@gmail.com>]

 Dear 
Matthieu:

      Thanks to your kindly reply. My calculation shows that the phonon frequency at q=0 drops a lot compared  with other q values. It's almost negative. I'll use |gkk|^2 to perform further calculations. Thanks again for your advice.

Regards

                                                       Zhang Ting
                                                       Peking Univ.
                                                       June 20th, 2008

2008/6/30 <mjv500@york.ac.uk>:

Hello again,

ok here we go: what you read from the gkk file is what abinit calculates raw, which is "gkk" matrix elements, but not for responses not to phonon modes but to single perturbations of 1 atom in 1 direction. The hand waving argument is that if you start from an equilibrium structure, the deformation potential you are sandwiching to get gkk should be somehow positive as you should be increasing the energy. I don't have an article for you - that would contradict the meaning of hand waving. However,

(1) In your case I'm still not sure that the structure is mechanically stable (not just wrt to phonons but wrt strains as well)

(2) what you are examining in the gkk file are full matrices btw all bands in the simulation. Even if my hand waving convinces you that the operator should be positive, that has no bearing on it's matrix elements in a random basis.

(3) I'm still not convinced of my hand waving because in the definition of the gkk there is a scalar product with the phonon displacement vector d. In many systems (and at least in the harmonic approximation) d and -d are both valid displacement vectors, and this makes the sign of d's scalar product with the raw gkk part have an arbitrary sign (the atomic perturbations I described above are a specific non-diagonal choice of basis vectors for the displacements). Even worse, if you have degenerate modes then you can make any rotation of the phonon eigenvectors corresponding to those modes and still get the same ep coupling.

anyhow, I wouldn't worry about the sign and instead look at the derived physical quantities you are interested in.

good luck

Matthieu

On Jun 29 2008, [GB2312] ÕÅœs wrote: