[CP2K-user] Issues with tight binding methods

Pierre-André Cazade pierre.a... at gmail.com
Wed Feb 26 10:10:23 UTC 2020


Dear Maxime,

Thank you very much. That was so simple... I have tested the dispersions 
both d3 and uff. UFF works but leads to a contraction of the crystal on the 
c axis. D3 does not work, I obtain the same structure with it as when I 
disable the dispersion. This structure is particularly bad with the c axis 
doubling in value.

Regards,
Pierre

On Tuesday, February 25, 2020 at 1:31:15 PM UTC, Maxime Van den Bossche 
wrote:
>
> Dear Pierre,
>
> I don't know about the effect of dispersion, but this "3ob" parameter set 
> is definitely supposed to be used with charge self-consistency. To get the 
> SCC cycle to converge, usually only the DIRECT_P_MIXING method works here, 
> and with this setting I indeed get good convergence (see attached input 
> file, slightly modified from what you sent).
>
> Also the DFTB regtests only feature this mixing method, so I presume the 
> other methods like BROYDEN_MIXING don't work well with the current 
> implementation of DFTB in cp2k. I don't think this is explicitly documented 
> somewhere.
>
> best,
> Maxime
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200226/f0f88510/attachment.htm>


More information about the CP2K-user mailing list