[Pw_forum] Innards of ph.f90 and Matdyn.f90

[Paolo Giannozzi]

On Jun 6, 2008, at 6:16 , Paul M. Grant wrote:
> I always fear being scolded by Paolo and/or Axel for not looking at  
> past fora postings or within the intricacies of “the code.”  (No, I  
> missed seeing that covered in the posting of 28 June 1914…I was  
> otherwise engaged).
>
nobody who posts not-too-stupid questions is at risk of being scolded  
here.
> 3)      Question…why is it necessary to do two scf runs a la  
> Example07, one to set a DOS flag for separate use later, and  
> another to set the k-point grid for a subsequent el-ph run with ph.x?
>
it is sort of explained in the README that comes with example 07 (and  
with all
examples). The first scf calculates \epsilon_k on a dense grid,  
needed only in
the estimate of the electron-phonon coefficients, saves the  
\epsilon_k to a
specific file. The second scf calculates the \psi_k on a "normal"  
grid, used in
all further calculations, saves data to the data directory,  
overwriting all files
written by the previous run, except the special one containing the  
\epsilon_k
on a dense grid.
>   I know there might some resource issues for “serious”  
> calculations, but for Example07, one scf with la2F “on” and an MP  
> grid of order 12, “seems to” work just fine.  At least for a “non- 
> serious” calculation.
>
in that case, the dense and normal grid just coincide. The double- 
delta integral
needed for electron-phonon coefficients is calculated without  
interpolation over
the dense grid (or, more exactly, with interpolation, but the  
original and
interpolated grids are the same).
> 6)      The matdyn.x calculation to compute the frequency dependent  
> component of the Eliashberg kernel uses a “real space” sampling  
> grid…in the case of Example07 of size 19.
>
those are q-vectors in the BZ, not real-space points. You supply the  
points you
are interested in. Typicaly people draws phonons on high-symmetry  
lines in the
BZ and corresponding el-ph coefiicients
> 7)      Lastly, here’s where I really get confused and appreciate  
> Dr. Prasad’s posting.  Which “dos” calculation is being referred to…
>
phonon DOS. You need phonons calculated in a uniform q-point grid.
matdyn.x does this, starting from interatomic forces constants in  
real space
> What’s needed is a user’s manual for matdyn (sorry, Paolo)
>
if you write a manual for matdyn, I promise that I won't feel sorry  
at all.
Essential documentation is available in the header of file PH/dynmat.f90
> , or, better yet, three or four more examples like 07 on non-cubic  
> but otherwise simple symmetries and simple basis orbital sets with  
> more than one atom per unit cell…like MgB2!
>

the problem with el-ph is that the current algorithm is not  
especially satisfactory.
There are much better implementations around, but until now thay have  
failed to
materialize into the "official" release. Spending time and effort for  
the old algorithm
when a new one might be released soon(er or later) is not an  
appealing perspective.

Paolo
---
Paolo Giannozzi, Dept of Physics, University of Udine
via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222