[CP2K-user] SCF convergence using PBE+50%HFX on +ve dimer

[Travis]

Hi,

Do you mean to use the truncated Coulomb method? You have chosen a large 
screening parameter for omega of 7.0 for a screened Coulomb (HSE-like) 
method. That's an exceptionally small range for the HF potential. HSE 
functional variants reduce to PBE0 and PBE50 methods with omega = 0, I'm 
assuming this is closer to what you're after. Typically PBE0 and similar 
would be done with a truncated Coulomb method like that applied in the 
tutorial here, https://www.cp2k.org/exercises:2017_uzh_cp2k-tutorial:hybrid (see 
2nd task). You just need to change SCALE_X in &PBE and FRACTION in &HF to 
0.5 each for PBE50.

-T

On Tuesday, June 30, 2020 at 11:48:31 AM UTC-4 Jan Elsner wrote:

> I solved this by decreasing EPS_PGF_ORB to 1.0E-40 and removing 
> EPS_FILTER_MATRIX. 
>
> Jan
>
>
>
>
>
> On Friday, June 19, 2020 at 12:08:08 PM UTC+1, Jan Elsner wrote:
>>
>> Dear all, 
>>
>>  
>>
>> I'm running a single point calculation on a positively charged rubrene (C
>> 42H28) dimer (spin polarized calculation) using PBE+50%HFX, restarting 
>> from a converged PBE wavefunction. Unfortunately the SCF does not converge 
>> after 350 steps. The "convergence" value jumps to huge values at SCF step 2 
>> ("convergence" at SCF step1 = *3.3*, "convergence" at SCF step2 = *12281*) 
>> and at longer times (around SCF step 150 onwards), it shows oscillatory 
>> behaviour. The behaviour is the same if I use RESTART ATOMIC. 
>>
>>  
>>
>> I note:
>>
>>    - The SCF also doesn't converge for the neutral dimer
>>    - The SCF *does converges *for both the neutral and cationic *single 
>>    rubrene molecule*
>>    - For the dimer, I have tried a number of settings: different mixing 
>>    schemes (BROYDEN, DIRECT_P_MIXING), diagonalization and OT method, larger 
>>    CUTOFF, smaller EPS_SCHWARZ
>>    - I also can't converge the SCF using PBE0, where I observe similar 
>>    behaviour 
>>
>> I attach the relevant files. Any help would be much appreciated. 
>>
>>  
>>
>> Best,
>>
>>  
>>
>> Jan
>>
>>  
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200630/83ff1475/attachment.htm>