[Pw_forum] including vdwl interactions during phonon calculations

Stefano de Gironcoli degironc at sissa.it
Mon Jun 27 22:15:14 CEST 2011


dear Vikas Varshney,

I think you analyzed correctly the situation.

london (grimme) correction for vdW is included in the total energy,  
force and stress but not for dynamical matrix.

It should not be too difficult because it is an extra term in the  
energy depending explicitly on the atomic positions.

you could have a look on how this term is computed in  energy_london,  
force_london and stress_london (all contained in  
Modules/mm_dispersion.f90)
and work out a similar routine for dynmat_london...
that would be a great contribution to the community.

stefano
-
Stefano de Gironcoli - SISSA and DEMOCRITOS


Quoting Vikas Varshney <vv0210 at gmail.com>:

> Dear QE users,
> I tried looking for answer for this question (the one in the subject) but
> could not find, hence I am asking this.
>
> I wanted to calculate phonon dispersion curves for layered materials like
> MoS2. I have successfully relaxed the crystal structure of both of these
> systems using vc-relax and including vdwl interactions (using .london. flag)
> and got all the relaxed crystal dimensions within 1.5% of experimental
> values.
>
> Now I want to run ph.x to calculate normal modes for different wave vectors.
> However, I could not find, how I can include vdwl interactions in these
> calculations? It is also interesting to note that in ph.x input script,
> there is no need to specify pseudopotential file information. So, I am
> suspecting it is gathering information from certain directories which were
> created during scf run.
>
> In order to test whether vdwl interactions are taken into account in phonon
> calculations implicitly, I ran two phonon cases (in two different
> directories) after successful convergence of two SCF calculations (one
> including .london option and one without including that option). The
> coordinates and cell dimensions for both calculations were identical. The
> initial structure was based on vc-relaxed structure with .london=true was
> from previous vc-relax calculations.
>
> On looking the predicted frequencies, I find that both cases yeilded exactly
> same numbers for frequencies. This is pointing me to believe that while
> calculating dynamical matrix, vdwl interactions are not taken into account
> implicitly. Am I wrong? Is there a way to include such interactions?
>
> In addition, I got one of the 12 predicted frequencies to be -ve (~45 cm-1).
> I am not sure if it is because of non-inclusion of vdwl interactions. I also
> wanted to mention that higher optical modes are predicted very near to
> experimental values (within 10 cm-1) for all IR and Raman active modes.   I
> have read many such questions before on the forum on this issue (-ve
> frequency) and would be trying different suggestions but I wanted to address
> the issue of inclusion/exclusion of vdwl interactions in phonon calculations
> and how one can you about calculating their contribution towards dynamical
> matrix.
>
> Best Wishes,
>
> Regards,
> Vikas Varshney
> Computational Materials Scientist
> Wright Patterson Air Force Base, Dayton, OH
>
>
>
> Best Regards,
>
> Vikas Varshney
> Computational Research Scientist,
> Wright Patterson Air Force Base, Dayton OH
>



----------------------------------------------------------------
   SISSA Webmail https://webmail.sissa.it/
   Powered by Horde http://www.horde.org/





More information about the users mailing list