[CP2K-user] [CP2K:20396] Re: PBE0 geometric optimization not converging

[Frederick Stein]

Dear Lory,
Regarding the convergence:
- you should replace your pseudopotential GTH-PBE-q4 with GTH-PBE0-q4 
(there is one in the file POTENTIAL_UZH).
- are you sure about the low ENERGY_GAP parameter? Such a low value may 
decrease the convergence. the default (0.02 Eh should work well enough). 
Simply comment this line out.
- please double-check the convergence of the parameters EPS_DEFAULT, 
EPS_SCHWARZ, CUTOFF. Especially EPS_SCHWARZ has to be chosen tighter 
(lower) if ADMM is requested. Also check within ENERGY_FORCE (not GEO_OPT) 
runs whether forces and stress tensors are well converged. You can reuse 
the thereby obtained wavefunctions as an initial guess for your actual 
geometry optimization

Performance-wise:
- EPS_PGF_ORB need not be chosen that tight the default value is usually 
good enough. You rarely need values lower than 1.0E-8 . Simply comment it 
out.
- MOLOPT basis sets are usually not a good choice for Hartree-Fock 
calculations. Try the correlation-consistent basis sets in BASIS_ccGRB_UZH 
or the MOLOPT-PBE0-GTH-q4 basis sets in BASIS_MOLOPT_UZH.

HTH,
Frederick

Lory Wang schrieb am Dienstag, 2. Juli 2024 um 08:56:04 UTC+2:

> Dear CP2K experts,
>
> I am currently experiencing difficulty in getting a geometric optimization 
> test with PBE0 to converge. I am starting with a PBE relaxed structure, and 
> the relaxed energy from PBE is approximately -212. From the output file 
> (cp2k_full_all.out), it seems like the very first SCF step is in the 
> correct energy range. However, the OT method then switches from DIIS to SD, 
> and the energy diverges significantly.
>
>      1 OT DIIS     0.15E+00  368.3     0.00789918      -212.0944035335 
> -2.12E+02
>      2 OT SD       0.15E+00   55.4    54.87357862     -3785.7999788923 
> -3.57E+03
>      3 OT SD       0.15E+00   55.1    78.83523027    -16117.9941684959 
> -1.23E+04
>      4 OT DIIS     0.15E+00   54.8    36.24089744     -3042.2730297100 
>  1.31E+04
>      5 OT SD       0.15E+00   55.1    25.41660926      -813.4054260626 
>  2.23E+03
>      6 OT SD       0.15E+00   55.1    35.73628016    -14638.4364940160 
> -1.38E+04
>
> I also tried different preconditioners and cutoff radii. I found that 
> FULL_SINGLE_INVERSE and a cutoff radius of 6 work better (see 
> cp2k_full_inverse.out), but I am still seeing energy spikes in the SCF 
> cycles.
>   ----------------------------------- OT 
> ---------------------------------------
>
>   Step     Update method      Time    Convergence         Total energy   
>  Change
>   
> ------------------------------------------------------------------------------
>      1 OT DIIS     0.80E-01  113.1     0.00645985      -213.0167602266 
> -2.97E-01
>      2 OT SD       0.80E-01   56.2     0.01323989      -213.9726833855 
> -9.56E-01
>      3 OT SD       0.80E-01   56.2     0.05110050      -219.8677669985 
> -5.90E+00
>      4 OT SD       0.80E-01   56.1     1.50838863      -526.6197195349 
> -3.07E+02
>      5 OT SD       0.80E-01   55.9     6.44077824     -3560.7235910833 
> -3.03E+03
>      6 OT DIIS     0.80E-01   55.9     6.93627366        -5.1310569943 
>  3.56E+03
>      7 OT SD       0.80E-01   56.3     0.00448314      -212.7274234258 
> -2.08E+02
>
> I have attached the full input and output files for reference. Could 
> someone help me identify what I might be missing?
>
> Thank you very much for your time and assistance.
>
> Best regards,
>
> Lory
>
>

-- 
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/8f25c603-de58-4f6a-859f-48534dedc1fdn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20240702/18aa2df5/attachment.htm>