[Pw_forum] Convergence of Gamma phonon

[Stefano Baroni]

Dear Subhadip:

On Feb 4, 2005, at 7:56 AM, Subhradip Ghosh wrote:

> Dear All,
>
> I have a question regarding convergence of Gamma phonon for metallic
> cases.
>
> I understand that to check the convergence of Gamma phonon frequencies,
> one should test convergence with respect to PW cut-off, k-point mesh,
> smearing parameter sequentially.
>
> My question is: what is the criteria for acceptable convergence?
>
> Is it only if the difference in gamma phonon frequencies are ~1% witrh
> respect
> to all of the above parameters?
>
> Is a higher tolerance(say 3%) for gamma phonons permitted?

only you can say. if you want to compare with an experimental datum  
which is know to, say, 10%, I would say that 3% is quite acceptable. if  
instead, you want to calculate a derived quantity which depends  
sensitively on the phonon frequency, then a higher accuracy may be  
required. you have also to be aware of the overall inaccuracy coming  
from systematic approximations (mainly, in your case, the lda/gga, and  
the use of pseudopotentials). it may be silly to pursue very high  
numerical accuracy when you know from the beginning that the best  
calculations you can possibly do will always be affected by such  
systematic errors. the estimate of such systematic errors, in turn, is  
often more the matter of an art than of a science ... only experience  
can guide you in this unexplored land ...


>
> I am asking this for a specific reason.
>
> I am doing fcc Copper. For the Gamma phonon, increasing the PW cut-off
> upto 60 Ryd hasn't given me convergence ~1%. I am using a k-mesh of
> 12x12x12 without offset.
>
> It seems strange to me that such a high cut-off is required.

first-row elements and first-row transition metals are known to have  
the "hardest" pseudopotentials. first-row transition metals more than   
first-row elements. the heavier, the worst. overall, copper is one of  
the worst cases. I do not find anything strange in the lack of  
convergence at 60 Ry. of course, it depends on the specific  
pseudopotential (choice of the core radii, etc.). this is THE case  
where the use of ultra-soft pseudopotentials would be kind of mandatory  
(that should allow to decrese the cutoff to below 30, Ry, I would say)

> Am I going the wrong way? Should I first look for the convergence with
> respect to k-mesh?

the convergence with respect to the cutoff depends on the properties of  
the atomic core, that with respect to kpoints on the properties of the  
crystal. they are independent and they should be checked independently.

hope this clarifies a bit
Stefano B.

>
> Subhradip
>
> --  
> *********************************************************************** 
> *
> 			Dr. Subhradip Ghosh
> 			Assistant Professor
> 			Department of Physics
> 		Indian Institute of Technology, Guwahati
> 			North Guwahati, Assam-781039
> 				India
> 			E-mail: subhra at iitg.ernet.in
> 			Phone: +91 361 2582717(O)
>                                +91 361 2584717(R)
> *********************************************************************** 
> *
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
---
Stefano Baroni    ---  SISSA  &  DEMOCRITOS National Simulation Center
via Beirut 2-4 34014 Trieste Grignano / [+39] 040 3787 406 (tel) -528  
(fax)

Please, if possible, don't  send me MS Word or PowerPoint attachments
Why? See:  http://www.gnu.org/philosophy/no-word-attachments.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 3527 bytes
Desc: not available
Url : /pipermail/attachments/20050204/f64e4379/attachment.bin