Kohn-Sham matrix in Hybrid.

Matt W MattWa... at gmail.com
Tue Jan 17 10:40:01 UTC 2017


Hi Albert,

maybe the ADMM method will not give correct eigenvalues straight up (see 
equations 33 and 34 in the original ADMM paper).

Try without using ADMM (if possible) or setting ADMM_PURIFICATION_METHOD to 
NONE, and see if that makes a difference?

Matt

On Tuesday, January 17, 2017 at 8:44:44 AM UTC, Albert de J. wrote:
>
> Dear Cp2k developers and users,
>
> In order to study the wave function in a material, I wish to extract the 
> orbital contributions of each orbital from the basis as well as the overlap 
> matrix. I originally wanted to use the PDOS print keywords, but I could not 
> find a way to obtain a decomposition based on the basis orbital. At best, I 
> can have the decomposition for each atomic orbital but those are the sum of 
> their constitutive basis orbital (unless the basis is a single zeta...) 
> Currently the best method I found is to output the Kohn-Sham matrix and the 
> overlap matrix in a file and then re-diagonals it to extract the 
> contributions.
>
> This usually works perfectly and I obtain the same contributions and 
> eigenvalues as those outputted by cp2k. I nevertheless have a problem when 
> I make a hybrid simulation (PBE0 with ADMM), in those situations, the 
> contributions and eigenvalues obtained after diagonalization are not right. 
> If I plot the DOS associated with these eigenvalues, it looks like a result 
> obtained in PBE instead of PBE0 (i.e. the gap is much smaller). I thus have 
> the impression that the KS matrix is outputted before the inclusion of the 
> HF part. Is that possible? Do I miss something ? Is there a way to get the 
> right KS matrix or, even better, directly obtain all the contributions ?
>
> Thanks !
> Albert.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170117/cc5f508d/attachment.htm>


More information about the CP2K-user mailing list