[Pw_forum] vc-relax analysis doubt

[Kevin May]

Hi Ricardo

If I understand correctly, you are saying that you no longer get scf
convergence after a few bfgs steps. This can be a frustrating problem, but
it has been mentioned on this forum before that after you have tried
changing things like smearing (yours is very small, why?) and mixing_beta
(sometimes going down to 0.2-0.3 helps), a useful thing to try is setting
scf_must_converge = .false. (see PW input file description).

Scf convergence can be very sensitive to changes in the atomic positions,
so if you just "power through" an intermediate geometry you might start
getting scf convergence again after another bfgs step. You want to keep an
eye on it to make sure it was just a one-time issue, and that you get scf
convergence in the bfgs steps that follow
​ (i.e. be careful).​


Good luck.

On 31 August 2017 at 15:28, Ricardo Afonso <ricardo at df.ufscar.br> wrote:

> Dear Lorenzo,
>
> I was not asking to someone to recompile my code and did what I have just
> done. By other reason, what I'm looking for is if there is any error in
> what I have just done, in terms of my input file. For example, changing
> upscale may bring some difficult to the code to keep the convergence.
> I did not mean to anyone testing my code or something like that. I'm sorry
> if that is what sounded. Anyway, I have gone through a good piece of this
> forum, but it does not seem that anyone came across the fact that after a
> few 'bfgs' cycles the limit of 'scf' cycles is reached. I'm open as wall
> for any reference, book, video or website of help, once this "error" keeps
> coming up for my structures.
>
> Otherwise, thank you for your answer.
>
> Regards.
>
> --
> Ricardo Afonso
> Student of Magnetism and Superconductivity Group
> Federal University of Sao Carlos
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>

--
Kevin May
PhD Candidate
Department of Mechanical Engineering, MIT
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170831/ffb97082/attachment.html>