[Pw_forum] Should I scrap the nonanalytic part of DM in electroded thin films?

[Serge Nakhmanson]

Stefano, Nicola,

Thank you for your comments -- this is some good food for thought!
Making electrodes an explicit part of the calculation will
surely be necessary in the case of thin-ish films, but it
opens its own big can of worms (band alignment between components
with DFT band gaps, etc; I bet you know the issues very well)
I do not want to deal with in this particular project. For
the case of thick-ish films with assumed perfect screening
I could probably guesstimate things around Gamma by taking an
"honest" bulk case as a reference and then playing with phonon
postprocessing routines (easy, since no new time-consuming DM
calculations are required).

But, I guess, the thought of how to add screening to this
process in a more rigorous fashion is worth some additional
thinking.

THX,

Serge

Stefano Baroni wrote:
> Dear Serge:
> 
>>
>> I understand that in bulk calculations the dynamical
>> matrices at small q are split into analytic and
>> nonanalytic parts, with the latter accounting for the
>> effects of macroscopic electric field. However, if
>> we want to study phonons in a model representing an
>> electroded thin film instead of a bulk crystal (i.e.,
>> the electrodes are not explicitly included in such a
>> calculation), we should assume that such an electric
>> field will be screened out by the free charges and
>> its influence onto ionic vibrations should not be
>> considered.
>>
>> I am under an impression that to achieve this in pwscf
>> I need to block the calls to nonanal(...) subroutine
>> in dynmat and matdyn, while leaving everything else
>> intact. Would this work out or could it possibly be not
>> as trivial as I imagine it?
> 
> I am afraid it is indeed not as trivial as you imagine it ... I am no 
> longer familiar with the guts of the code, but the only thing the code 
> can possibly do is to make the phonons calculated at q=0 equal to the 
> values that the phonon branches would have in the long-wavelength limit 
> (and those depend in a non trivial way on the combination of 
> wavevector/polarization, because of the long range character of the 
> Coulomb interaction). I think that, in the presence of free carriers, 
> the results of the code are in principle correct whenever the phonon 
> wavelength is smaller that the screening length. At q=0 just setting the 
> effective charges equal to zero (which is what you would presumably get 
> by commenting out some appropriate routine calls) should do the job, but 
> you would then get frequencies that are inconsistent with those 
> calculated at finite q. To get fully consistent results, I think that 
> screening from free carriers should be explicitly accounted for in the 
> calculation of the dynamical matrix. Possible, but "not as trivial as 
> one could imagine".
> 
> Cheers - stefano
> 


-- 
*********************************************************
  Serge M. Nakhmanson               phone: (630) 252-5205
  Assistant Scientist                 fax: (630) 252-4798
  MSD-212, Rm. C-224
  Argonne National Laboratory
  9700 S. Cass Ave.
  Argonne, IL 60439
*********************************************************