[CP2K-user] 自动回复: [CP2K:16272] Re: Improve extrapolation between relaxation steps?
Marcella Iannuzzi
marci.akira at gmail.com
Fri Nov 19 07:48:40 UTC 2021
Hi Nick
Actually EPS_SCHWARZ 1e-7. is not very tight. Maybe too many integrals are
screened out instead.
The ASPC extrapolation methods might help.
You could also try diagonalization+mixing+smearing with a high smearing
temperature, but this is in general slower than OT.
Best
Marcella
On Thursday, November 18, 2021 at 7:03:05 PM UTC+1 nwi... at berkeley.edu
wrote:
>
> There are not many energy fluctuations. Looking at printed force_eval
> (below) it shows that it pretty consistently decreases, with only one CG
> step moving away from minimum. I have not tried biasing the magnetization.
> Convergence does not seem to be unstable, it is very consistent, it's just
> much slower than I would expect for a proper wfn extrapolation considering
> the energy is changing by a small amount and the spin moment/charge on the
> site is not changing dramatically between steps.
>
> It is possible, and has been pointed out to me, that the issue is the
> combination of (a) longrange HF functional, (b) shallow state, and (c) a
> reasonably tight screening of EPS_SCHWARZ 1e-7. The very small integrals
> that this introduces may cause the condition number to go up, therefore
> making the calculation slower.
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.957685447298900
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.962799617469500
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.968174215491672
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.969982458291270
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.974875112599875
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.977152106254835
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.983613254400552
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.987573339468327
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.975501539662218
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.980535609151048
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.985348311781308
>
> ENERGY| Total *FORCE_EVAL* ( QS ) energy [a.u.]:
> -6495.988025124417618
> On Thursday, November 18, 2021 at 1:18:07 AM UTC-8 7510... at qq.com wrote:
>
>> 你好,我已收到邮件~谢谢
>
>
--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/48f93379-eb56-43e9-9ddf-3867ef47cd95n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20211118/4b25e648/attachment.htm>
More information about the CP2K-user
mailing list