[Pw_forum] (no subject)

[Guido Roma]

Raghani Pushpa wrote:
> 
> Dear users,
> I am calculating the phonon spectrum for a bulk system. For that I
> calculate the dynamical matrices at q points in IBZ and then
> calculate the force constants in real space. I have the force
> constants for 6x6x6 and 12x12x12 mesh. Then I use the matdyn.f
> program to calculate the phonon spectrum. With this program for 6x6x6 mesh
> the phonon spectrum is qualitatively correct but when I use the 12x12x12
> mesh, frequencies at many intermediate pts are incorrect. Does anybody
> have any idea as to why does it happen? With a larger q point mesh, the
> phonon spectrum should improve, isn't it?

Hello Pushpa,

I think it would not necessarily improve: using larger q meshes could
introduce spurious high frequency (i.e. long range, for force constants)
components to your dispersion.
By going from dynamical matrices (reciprocal space) to real space (force
constants) then back to q space, you do an FFT interpolation on your
original 12x12x12 mesh; if you don't have the same level of convergence
(coming from k-points, cutoff etc.) on each of the mesh points, some
spurious components will show up as noisy oscillations in your phonon
dispersions.
I think that people knowing better than me the way matdyn is programmed,
could confirm this.

Guido
 
 


-- 
Guido Roma <roma at cea.fr> -- CEA-Saclay - DEN/DMN/SRMP Bat.520/130 
Phone: [+33]-1-69085738 -- Fax: ...6867 -- Mobile: [+33]-6-20069085