[gpaw-users] Phonon band structure

[Kristen Kaasbjerg]

My e-mail to Lauri missed the mailing list, so here it is:

Lauri Lehtovaara wrote:
> Dear all,
>
> How can I calculate phonon band structure with GPAW? Any tutorial or 
> example? I found one mail about phonopy. Is this the best way or are 
> there other ways to do it?
>
> Cheers,
>
>      - Lauri
>   

Mail #1:

Hi Lauri

I have just finished a brute force implementation for phonons based on
supercell calculations. It doesn't exploit space group symmetries to reduce
the number of force constants to be calculated. The possibility to enforce
the acoustic sum-rule to be fulfilled for the force constants will soon 
be included.

For the moment it is located in

gpaw/gpaw/dfpt/phonons.py

but will soon be transfered to ASE.

I attach here a simple script that calculates the phonon dispersion for 
Al using
effective medium theory. Note that this kind of supercell calculations 
must be
converged with respect to the size of the supercell.

Regards
Kristen

Lauri Lehtovaara wrote:


Great, thanks. We'll try it out. We want to check our ABINIT 
pseudopotential calculations with PAW. PAW & phonons does not work in 
abinit  :( 


Mail #2

Good to hear.

While we are at it, let me issue a little warning: the force constants, 
i.e. the
second derivative of the energy wrt atomic displacements, are calculated 
from
first order finite-difference derivatives of the atomic forces. For 
small displacements
of the atoms (keyword delta in the code), say < 0.05 Ang, these 
derivatives might be
dominated by the so-called egg-box effect. In order to decrease the 
sensitivity to this
highly unwanted effect, a larger displacement (~0.1 Ang) must be used. For
such large displacements one might not stay within the harmonic 
approximation. However,
this is the sacrifice one has to make to avoid these effects.

Another comment: we are presently working on an implementation of phonon 
calcuations
within the framework of density functional perturbation theory (as in 
ABINIT). So far,
it works only for the norm conserving HGH pseudo potentials. Also, here 
the above-mentioned
egg-box effect is harder to get rid of since no atoms are moved at any 
point.

That's it for now
Kristen