[CP2K-user] [CP2K:22172] Re: Issues with HSE06 calculations

[Frederick Stein]

Dear Lorenzo,
Regarding the oscillations, you may need to decrease EPS_DEFAULT and 
EPS_SCHWARZ. With OT, you may also try other preconditioners and try to 
lower ENERGY_GAP in the OT section. I am also not sure whether 
PBE-pseudopotentials are suitable with HSE06 and should be replaced with 
pseudopotentials optimized for hybrid functionals (PBE0).
HSE06 is generally more expensive than PBE but you may try to 
set SCREEN_ON_INITIAL_P T in the HF section (at the expense of a lower 
EPS_SCHWARZ) to decrease the number of necessary integrals. The runtime of 
the second step should be significantly lower than the first SCF-step. To 
achieve that, you should set the MAX_MEMORY keyword and increase its value 
to fit as many integrals into RAM as possible. If there are still too many 
integrals to be recalculated, you can reduce the number of MPI ranks while 
increasing the number of OpenMP threads per MPI rank (and MAX_MEMORY).
It may also help to increase the print level and as soon as the accuracy 
looks fine, to post a timing report of a few SCF steps for further advice 
by either using a larger value of EPS_SCF or adding  
IGNORE_CONVERGENCE_FAILURE 
<https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/DFT/SCF.html#CP2K_INPUT.FORCE_EVAL.DFT.SCF.IGNORE_CONVERGENCE_FAILURE> to 
the SCF section.
Best,
Frederick

Lorenzo Lagasco schrieb am Mittwoch, 8. April 2026 um 11:11:34 UTC+2:

> Good morning,
> I'm writing here because I'm facing some troubles with a 
> HSE06 calculation, in order to get info about band gap and atomic pdos from 
> my system (delafossite CuAlO2 with cumarin343 anchored above the surface). 
> In particular, following some hints from manual/seminars, I launched a 
> preliminar energy calculation GGA only (PBE). Then, I used the restart wfn 
> file for the following hybrid calculation.
> However, the calculation is extremely slow (probably because I'm using 
> only 1 node with 16 CPU), but also the energy fluctuates strangely among 
> the SCF iterations.
> Here attached the input files of both PBE and HSE06 calculations, the 
> output for the former calculation (still running) and xyz system (395 
> atoms).
> Thanks in advance for the help 
>

-- 
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/a312531a-a26f-498a-9704-c48dc5214ed4n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260408/bf2d04e0/attachment.htm>