[CP2K-user] [CP2K:22220] Re: Converge EPS_SCHWARZ and CUTOFF_RADIUS for hybrid functionals

[Frederick Stein]

Dear Max,
a larger cell should not be necessary. I have just seen that you set 
SCF_GUESS ATOMIC instead of RESTART. Thus, you start from an atomic guess 
instead of your restart file. I also would not use XC_GRID and I would not 
touch the *_CUTOFF parameters in the XC section.
Best,
Frederick

Max H schrieb am Dienstag, 5. Mai 2026 um 12:22:19 UTC+2:

> Dear Frederick,
>
> thank you for your kind explanation.
> I have been doing a few tests with CUTOFF_RADIUS 8.2 (slightly smaller 
> than half the smallest cell dimension) and various values of EPS_SCHWARZ 
> (from 1E-6 to 1E-8).
> Unfortunately, the SCF does not converge at all. The first iteration 
> yields a negative energy, but the energies become positive after the second.
> I am restarting the SCF from a previously converged calculation at the PBE 
> level of theory (convergence with fewer than 50 iterations) and employ the 
> "MIN_PAIR_LIST_RADIUS -1" setting in the QS block of the input (attached).
>
> Is it possible to conclude from the SCF behaviour that calculations on 
> multiple unit cells with a yet larger CUTOFF_RADIUS are required?
>
> Kind regards,
> Max
>
>
> On Friday, April 24, 2026 at 6:24:33 PM UTC+2 Frederick Stein wrote:
>
>> Dear Max,
>> Both parameters are determined differently.
>> CUTOFF_RADIUS needs to be less than half of the distance of an atom to 
>> its nearest image. For sufficiently large cells, you may even reduce it to 
>> values at the order of 5-7 A (please check convergence of your results in 
>> that case). If this parameter is not yet converged, you employ the 
>> MULTIPLE_UNIT_CELL as you suggested. Convergence is expected if 
>> CUTOFF_RADIUS can be set to values of at least 5 A.
>> EPS_SCHWARZ is determined by checking convergence of total energies. A 
>> good starting point is 1E-6 but larger values may be necessary, especially 
>> if you set "SCREEN_ON_INITIAL_P T" or with ADMM (ADMM is recommended).
>> Best,
>> Frederick
>>
>> Max H schrieb am Freitag, 24. April 2026 um 17:05:35 UTC+2:
>>
>>> Dear all,
>>>
>>> I plan on doing some single point calculations with the PBE0 functional 
>>> and a ccGRB basis and would appreciate advice on the selection of 
>>> appropriate parameters, because I have not found information of sufficient 
>>> detail.
>>>
>>> I read in the recent review (JPhysChemB, "The CP2K Program Package Made 
>>> Simple") and in various other sources that the EPS_SCHWARZ and 
>>> CUTOFF_RADIUS parameters are crucial for accuracy and computational 
>>> efficiency.
>>>
>>> What is the typical approach to converge these values? Are single-point 
>>> calculations with one SCF cycle sufficient as soon as energy convergence is 
>>> observed (comparable to the convergence protocols for CUTOFF and 
>>> REL_CUTOFF)?
>>>
>>> When studying metallic systems, which values of CUTOFF_RADIUS can be 
>>> expected?
>>> Supposing that the CUTOFF_RADIUS did not converge at half the smallest 
>>> cell dimension, is the best approach then to set MULTIPLE_UNIT_CELL 2 2 1 
>>> (xy periodic system) in the &CELL block or do special restrictions arise?
>>>
>>> I appreciate your kind advice.
>>>
>>> Best,
>>> Max
>>>
>>>
>>>
>>>

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/c53420f7-6d42-4b0c-b014-e6dfd02866b1n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260505/fb4cf996/attachment.htm>