[CP2K-user] [CP2K:22054] Re: How to perform hybrid functional calculations for large systems using the pob-TZVP-rev2 basis set?

[Augustin Bussy]

Hi Yisichi,
Your calculation is most likely not stuck, but proceeding. Hybrid DFT in 
the condensed phase can be very expensive, even when using ADMM and 
non-diffuse basis sets like pob-DZVP-rev2. Given an optimized input file, 
the only solution is to used more resources.

That being said, your input in not perfect. In particular, you are using 
and *extremely tight* value for EPS_PGF_ORB. The default value if 1E-5, and 
particularly accurate calculations can used 1E-6. I recommend you to change 
it. The MIN_PAIR_LIST_RADIUS option that you commented out is a much more 
efficient solution to the "Kohn Sham matrix occupancy warning": use this 
instead. For performance, make sure to also optimize the MAX_MEMORY keyword 
to the specifications of your computer. See the "Hartree−Fock and Hybrid 
Density Functional Theory" section of our latest paper 
(https://pubs.acs.org/doi/full/10.1021/acs.jpcb.5c05851) to understand the 
meaning of these keywords.

Best,
Augustin

On Wednesday, 21 January 2026 at 17:41:23 UTC+1 yis... at 163.com wrote:

> To the CP2K Development Team,
>
> Hello,
>
> I am a beginner in DFT calculations and am currently learning to use CP2K 
> for calculations with the PBE0 hybrid functional. My research involves a 
> Mo-doped NiFe-LDH system. I have constructed a 6×8×3 supercell containing a 
> total of 720 atoms. Given the significant presence of Ni, Fe, and O in the 
> system, when using the GTH pseudopotential with the DZVP-MOLOPT-SR basis 
> set, convergence was only achieved with very high values of CUTOFF and 
> REL_CUTOFF. Therefore, I intend to use the GAPW method to reduce 
> computational cost and improve accuracy.
>
> Currently, I am employing the pob-TZVP-rev2 basis set for all-electron 
> calculations, along with pob-DZVP-rev2 as the ADMM basis set to accelerate 
> the hybrid functional calculation. During the calculation, I first 
> performed a pure functional calculation using the pob-TZVP-rev2 basis set. 
> Then, I used the wavefunction generated from this calculation as the 
> initial guess for the hybrid functional calculation. Unfortunately, the 
> hybrid functional calculation does not seem to proceed—after running 
> continuously for 4 hours, the calculation remains stuck at the interface 
> shown in the attached image and does not appear to enter the 
> self-consistent iteration process.
>
> The supercomputer I am using is equipped with Intel Xeon Platinum 
> Processors (3rd Gen, 64 cores) and 256 GB of memory. I would like to ask 
> the developers for their insight into the following:
>
>    1. 
>    
>    What might be causing this issue?
>    2. 
>    
>    How should hybrid functional calculations be conducted for such a 
>    system?
>    3. 
>    
>    Could you recommend a more suitable basis set for this system?
>    
> For your reference, I have attached the input and output files from my 
> CP2K calculation at the end of this email.
>
> Thank you for your attention and assistance.
>
> Best regards,
>
> yisichi
>
> [image: 微信图片_20260122004005_251_55.png]
>

-- 
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/5c913ae1-0307-4862-8651-c397f5446498n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260122/0520ad35/attachment.htm>