[CP2K-user] Orbital transformations for system with shallow defect states

Matt W mattwa... at gmail.com
Thu Mar 4 12:45:29 UTC 2021


Couple of tricks I regulalry used are:
(i) include distortion around the defect likely to cause localisation
(ii) include a core correction (excess charge) on atoms to break symmetry / 
get localisation. Then restart from the wavefunction without the core 
correction (it is in the KIND section).

On Tuesday, February 16, 2021 at 8:53:34 AM UTC Marcella Iannuzzi wrote:

> Dear Nick,
>
> It depends on the system. 
> If you expect localised defect states, it might be that DFT is not able to 
> find the right state.
> You can try to bias at the initial guess the localisation and hope that 
> the SCF falls in the right minimum. 
> Using OT you can help the SCF convergence by tuning the preconditioner.
> If the system is really metallic, you better give up OT.
>
> Regards
> Marcella
>
>
> On Monday, February 15, 2021 at 11:42:01 PM UTC+1 n... at berkeley.edu 
> wrote:
>
>> Hello all,
>>
>> I am running some defect calculations for vacancies using CP2K, and some 
>> of my calculations induce states very close to the band edge. 
>>
>> I believe this is causing the OT method to be very slow in its 
>> convergence, since OT requires (to my knowledge) a clear delineation 
>> between occupied/unoccupied states. I would like to continue using OT 
>> because of its good scaling behavior, so I'm wondering if there are any 
>> tricks to help converge systems with shallow states, or if I simply need to 
>> let the SCF proceed for a very long time.
>>
>> Any suggestions would be appreciated, thanks.
>>
>> -Nick
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210304/9007ca81/attachment.htm>


More information about the CP2K-user mailing list