[CP2K:10213] Re: Cut-off problrm

Ari Paavo Seitsonen ari.p.s... at gmail.com
Tue Apr 24 13:23:42 CEST 2018


Hello,

  Just one entry in the input; I have not used PBE a lot with CP2K, but at
least with BLYP I would go stricter than 1e-10 in &QS / EPS_DEFAULT
1.0E-10; and I usually also convergence the electronic structure up to
EPS_SCF 1e-8. But this might be different with PBE, in general being
smoother than BLYP.

    Greetings from Sunny Zurich,

       apsi


2018-04-24 13:16 GMT+02:00 simin pahlavi <simin.... at gmail.com>:

> Hello Matteu,
> I would like to thank you very much for your Helpful comments again.
> Have a nice day
> Simin
>
> On Tuesday, April 24, 2018 at 12:42:37 PM UTC+2, Matt W wrote:
>>
>> Personal opinion is that converging total energy is not very important in
>> this case - it is normally just a rigid shift based on inaccuracy of the
>> core part of the wavefunctions. I am still (again personally) not really a
>> big fan of the smoothing, I'd rather have a slightly bigger cutoff but a
>> not messed with functional.
>>
>> I would look at forces on atoms and see how they converge and see how
>> errors in forces propagate to energy conservation.
>>
>> Matt
>>
>> On Tuesday, April 24, 2018 at 10:24:34 AM UTC+1, simin pahlavi wrote:
>>>
>>> Hello Matteu
>>> Thank you very much fir your helpful comment.
>>> As I am supposed to use this DFT calculations for every step of an ab
>>> initio molecular dynamics, I wonder if I should
>>> go for the heavy smoothing and get reasonable forces or I should be
>>> worry about convergence and cutoff dependence of minimization?
>>> I appreciate any reply in advance
>>> Simin
>>>
>>> On Monday, April 23, 2018 at 10:41:00 PM UTC+2, Matt W wrote:
>>>>
>>>> Sorry, I posted too quickly without really looking at your setup.
>>>>
>>>> The big problem for you is the XC_GRID settings.
>>>>
>>>>    &XC_GRID
>>>>     XC_SMOOTH_RHO NN50
>>>>     XC_DERIV NN50_SMOOTH
>>>>    &END XC_GRID
>>>>
>>>> You really do not want to be using heavy smoothing, especially for this
>>>> purpose. These settings (see original paper) help get reasonable forces
>>>> with a rather low cutoff, but as you increase the cutoff lead to you
>>>> continuously redefining the XC functional you use (you are smoothing over
>>>> different areas of real space), so they do not converge.
>>>>
>>>> Try changing to the defaults
>>>>
>>>>    &XC_GRID
>>>>     XC_SMOOTH_RHO NONE
>>>>     XC_DERIV PW
>>>>    &END XC_GRID
>>>>
>>>> you should see at least a significant improvement.
>>>>
>>>> Matt
>>>>
>>>>>
>>>>>>>> --
> 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+uns... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at https://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.
>



-- 
-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
  Ari Paavo Seitsonen / Ari.P.S... at iki.fi / http://www.iki.fi/~apsi/
    Ecole Normale Supérieure (ENS), Département de Chimie, Paris
    Mobile (F) : +33 789 37 24 25    (CH) : +41 79 71 90 935
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180424/b02388ee/attachment.html>


More information about the CP2K-user mailing list