[Pw_forum] phonon errors (vtmtrinh at caltech.edu)

[Eyvaz Isaev]

Hi,

--- On Thu, 10/8/09, vtmtrinh at caltech.edu <vtmtrinh at caltech.edu> wrote:

> My main purpose is to calculate the phonon dispersion curve from Gamma 
> to X  for Si. I'd like to reproduce the Si phonon dispersion curve from > Gamma  to X.  

In order to do this one correctly first you should calculate dynamical matrices for a given q-mesh, then perform FFT by q2r.x, then apply matdyn.x to calculate phonons along  given directions (specifying a list of q-points). In this way you apply ASR for both q2r.x and matdyn.x calculations.

If you have only \Gamma-point phonons you can use dynmat.x to apply ASR. 

> I thought that after the scf calculation for 10x10x10, I can
> run phonon calculation directly for a series of q points from Gamma to
> X (let say 5 points).  Then from there, I use q2r.x and then
> dynmat.x to calculate  intermediate points in order to fix the non-zero > frequency  at Gamma point  (applying ASR) and to get a smooth dispersion > curve  curve.  

This is a wrong way to calculate phonons. Of course, you can calculate phonons for any q-point you like, but they are not oblidged to form a proper mesh for Fast-Fourier-Transformation. In this way you always  will meet troubles with q2r.x (FFT).

> However, the phonon calculation (the first crash) crashed all the >times.  Now I am  testing by running phonon calculation for a mesh of >4x4x4  to see if it  still crashes. If it works, this means that direct
> calculation of phonon  by specifying directly q points along G-X does not
> work.  However, this  will be expensive calculation if to get phonon 
> dispersion a long one  direction, I have to do the 3D-messh phonon 
> calculation  first.
> 
> If I did only a single q-point calculation one at a time,
> which would be  much faster than does the approach above, 

Sorry, I  am not sure I exactly understand what you mean, but calculations performed in the right way (see above) give you Interatomic Force Constants (IFC) matrix which allows you to evaluate many thermodynamical properties, besides  phonons along high symmetry directions.

>how could I  specify q-grid in  q2r.x?

You should not specify q-grid in q2r.x's input file. It (q2r.x) reads Sysname.dyn0 file generated by ph.x which contains info about the q-grid (divisions along each basis vectors, total number of q-points and q-points itself.

Of course, you could calculate dynamical matrices in the old manner, specifying a list of q-points (but they have to be a result of KxLxM mesh, and calculated before you start phonon calculations). In this case you can 
follow your example given below.

> 
> Also, I have another question, if I would like to apply ASR
> to fix the phonon freq. at Gamma point, is it correct to just specify
> the q-grid in  the input files for q2r.x and matdyn.x as follows:
> 
> &input
>     fildyn='si.dyn', zasr='crystal',
> flfrc='si.fc'
>   /
>  1 1 1
>  1
>  si.dyn1    !this the *.dyn file at gamma

I do not think so, as q2r.x calculates IFC Matrix, not phonon modes as matdyn.x does this job. 
How to apply ASR for Gamma-point phonons  see above. 

Bests,
Eyvaz.

-------------------------------------------------------------------
Prof. Eyvaz Isaev, 
Theoretical Physics Department, Moscow State Institute of Steel & Alloys, Russia, 
Department of Physics, Chemistry, and Biology (IFM), Linkoping University, Sweden 
Condensed Matter Theory Group, Uppsala University, Sweden 
Eyvaz.Isaev at fysik.uu.se, isaev at ifm.liu.se, eyvaz_isaev at yahoo.com

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com