[CP2K-user] [CP2K:16901] Slow force calculation step timing using hybrid functionals

[Бранислав Миловановић]

Dear Frederick, Augustin,

Thank you for your answers. Both answers improved performance of the 
calculation, especially Augustin's solutions.
Solution 1) is straightforward and easy to implement.
I also tried 2) and 3) and which are helpful as well.

Thanks once again!

Best
Branislav

On Wednesday, April 27, 2022 at 3:02:40 PM UTC+2 Augustin Bussy wrote:

> Hi Branislav,
>
> your input file is in principle correct. The main issue is that the 
> DZVP-MOLOPT-GTH basis set is really not optimized for HFX type 
> calculations. Therefore, a lot of time is spent calculating 4-center 
> integrals, and even more time is spent on 4-center derivatives.
>
> I see 3 ways that you can improve your timings:
>
> 1) Use a basis set that is better suited. For example, Pople style 
> all-electron basis sets such as 6-31G* (requires a switch to GAPW)
>
> 2) Use a RI approximation for the HFX: add  &RI &END to the &HF section 
> (this requires CP2K 9.1 or more recent)
>
> 3) Use the ADMM method (
> https://manual.cp2k.org/cp2k-9_1-branch/CP2K_INPUT/FORCE_EVAL/DFT/AUXILIARY_DENSITY_MATRIX_METHOD.html
> )
>
> I think solution 1) is the easiest.  I also suggest that you use the 
> FULL_ALL preconditioner for OT. It performs better than FULL_S_INVERSE for 
> smaller system (in that case, you also need to remove the ENERGY_GAP 
> keyword).
>
> Hope that helps. Best,
>
> Augustin
> On 4/26/22 16:10, Бранислав Миловановић wrote:
>
> Dear CP2K users, 
>
> I'm experiencing problems when using hybrid functionals such as B3LYP or 
> M062X and while optimizing simple test system (hydrogen bonded acetic acid 
> dimer).
>
> SCF convergence appears to be fast and normal but the Force calculation 
> step is extremely slow. Maybe I made some mistake in the input file?
>
> I attached my input files and output file (B3LYP).
> Also, I don't really understand the timings at the end output file.
>
> Thanks in advance!
> Branislav
>
> -- 
> 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+uns... at googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/e568c56a-bfea-411c-9274-119d735cc83cn%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/e568c56a-bfea-411c-9274-119d735cc83cn%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
> -- 
> Augustin Bussy
> Postdoctoral researcher
> Hutter Group
> University of Zurich
>
>

-- 
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 on the web visit https://groups.google.com/d/msgid/cp2k/51b5bc7a-de20-4c3e-8326-fe9b1a47979fn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20220427/2c2d1e87/attachment-0001.htm>