[CP2K-user] SCF run not converged-wB97X-V

Stanislav Šimko stanisl... at gmail.com
Mon Oct 28 20:02:11 UTC 2019

Dear Farzaneh,

I had only a very brief look at your input, however, according to "latest" 
literature, you're not using "recommended" setup. I suggest to use 
"PARAMETERS 6.2 0.01" for the rVV10 section, as can be found 
at 10.1021/acs.jpclett.6b02527

I know it's quite a bit late but hopefully you read it. Besides this, my 
"latest" setup is rather similar to yours, however, I do not have got 
"benchmark" for the cp2k results yet, so I am still "unsure" whether we 
don't miss something important. Please, let me know if you're interested in 
helping with the "benchmark".
Best regards,

On Thursday, August 8, 2019 at 4:31:09 PM UTC+2, farzaneh sarrami wrote:
> Each step will take about 10min.
> I will try with MINIMIZER CG and let you know.
> Thanks for your advices.
> Farzaneh
> On Thursday, August 8, 2019 at 4:14:01 PM UTC+2, Frederick Stein wrote:
>> The input file looks correct to me. When you still have convergence 
>> issues, you might try the conjugate gradient minimizer (MINIMIZER CG). How 
>> many MPI ranks do you use? How much time does one step need?
>> Concerning the costs you can try to use a higher EPS_SCHWARZ (if it 
>> really is Hartree-Fock and the results do not get worse) or a different 
>> preconditioner like FULL_KINETIC or FULL_SINGLE.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20191028/28b7a2d7/attachment.htm>

More information about the CP2K-user mailing list