Discrepancy in using Range-Separated Hybrid Functional with D4 Correction

[Lim Jia Yang]

Dear CP2K community,

CP2K version 2025.1, ORCA version 6.1.0
I am doing benchmarking on vertical detachment energy. Both functionals used are from Libxc for calculations in ORCA and CP2K.

First method: LC-ωPBE+D3BJ(2016)
Second method: LC-ωPBE08+D4

The first method gives similar values with a ~0.0002 eV difference between ORCA and CP2K. The second method shows differences of ~0.024 eV, with CP2K being lower.

Any advice is really appreciated. I hope I didn't make some mistakes in the input.

Attached are the CP2K input files for both methods. The output file for the second method shows "POWELL| Error in trust region". The same error appears when D4 correction is removed.

ORCA input line for the first method:
! D3BJ aug-cc-pVTZ TightSCF DEFGRID3
%method
        method dft
        functional hyb_gga_xc_lc_wpbe
        D3S6 1.00
        D3A1 0.563761
        D3S8 0.906564
        D3A2 3.593680
end
ORCA input line for the second method:
! D4 aug-cc-pVTZ TightSCF DEFGRID3
%method
        method dft
        functional hyb_gga_xc_lc_wpbe08_whs
        D4S6 1.00
        D4A1 0.378
        D4S8 1.170
        D4A2 4.816
end

[Jürg Hutter]

Hi

I have to guess, but here I my two top options:

1) Poisson solver: use WAVELET, it is more accurate than MT
2) D4: check that CP2K and ORCA have the same D4 contributions

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Lim Jia Yang <limjiay...@gmail.com>
Sent: Tuesday, July 29, 2025 2:53 PM
To: cp2k
Subject: [CP2K:21699] Discrepancy in using Range-Separated Hybrid Functional with D4 Correction

--
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...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/6860d009-2149-4781-84ee-227085e3a60bn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/6860d009-2149-4781-84ee-227085e3a60bn%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Johann Pototschnig]

Hi

You could try comparing to the alternative implementation:

D4_REFERENCE_CODE F

Did you also check the other parameters?

FACTOR_S9_TERM

D4_CUTOFF

D4_CN_CUTOFF

R_CUTOFF

best wishes