The ABINIT Users Mailing List ( CLOSED )

[Sébastien Hamel <sebastien.hamel@UMontreal.CA>]

Selon Xavier Gonze <gonze@pcpm.ucl.ac.be>:

> Dear Sebastien,
> 
> sebastien.hamel@umontreal.ca wrote:
> >  I was wondering if anaddb could handle the calculation of the phonon
> spectrum
> >  of a molecule with a partly optimized/partly contrained geometry.
> > 
> >  How should it react to an uncomplete DDB?
> > 
> >  From what I can see,aznaddb returns 0 for every values if the DDB is used
> as is.
> 
> Yes, this is the present way of treating incomplete information.
> The rationale was the following : an incomplete DDB is usually
> useless, except if the provided information just spans one or more
> representations
> of the symmetry group for phonons (i.e., if the phonons spanned by the
> treated perturbations decouple from the other phonons). This often happens
> with rather symmetrical solids. If the missing information is set to zero,
> then, the only non-zero frequencies corresponds to the "treated" phonons.
> (I hope this is not too diffocult to understand).
> 
> Do you think you can make something else with an incomplete DDB ??

Well, abinit does give phonon energies for a constrained geometry but not the
eigenvectors. in short we wanted anaddb to give us the eigenvectors so we 
could
compare with the fully optimized calculation to see how much the phonon
energies changed. possible solutions, tell 1) anaddb to put a big value 
(force)
for the missing numbers or 2) use the SVD to isolate the eigenvalues and
eigenvectors of interest. what do you think?

Thanks,
Sebastien

> Good continuation,
> Xavier
>