The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

Hi,

Problem of SCF loop convergency are frequent in DFT, although quite unusal for Si.
Could you please post the input file that leads to your problem so that we can check what is happening.
Also wich version of Abinit are you using: older version uses CG by default (iscf 5) while newer uses pulay mixing (iscf 7) ? Are you sure you use CG ? Have you tryed Pulay mixing ?  And if yes have you tryed to increase "npulayit" ?

Usually self consistency problems for silicon appears when your supercell includes dangling bonds. In this case it is advised to use a preconditioner. For instance "diemac 30" or "iprcel 45" in your input file will help.

regards

PMA

On 7/12/07, yunhee.chang@kriss.re.kr < yunhee.chang@kriss.re.kr> wrote:

Dear Abinit users,

I'm calculating Si surfaces.

In the CG atomic relaxation step, there is a problem.
In every second ionic step, electronic SCF loop requires more than 100 steps.
And it seems difficult to converge in optimization of ionic step.

Is it conventional problem in abinit code using CG method?
How relaxed the ionic position using CG method in abinit?
Is there any special principle in abinit code?

Would you give me helpful comments?

With best regards,

Yunhee Chang

-- 
Pierre-Matthieu Anglade