wB97 functionals

[giaan_kler-young]

Hello,

I'm looking for some advice on the implementation of range-separated hybrid functionals in VASP.

An older forum post (forum/viewtopic.php?p=22644) mentions that wB97 functionals can't intrinsically be run without manipulating source code. Then there is this paper from 2024: https://doi.org/10.1038/s41467-024-48609-2 which states they used wB97X-D3(BJ), and the author states this is what they specified in their INCAR:
GGA = LIBXC
LIBXC1 = HYB_GGA_XC_WB97X_V
LHFCALC = .TRUE.
AEXX = 0.167000
HFSCREEN = 0.566918

I cannot make sense of how these settings would be equivalent to specifying both short- and long-range Fock exchange separately, or where the HFSCREEN parameter comes from, or how this would introduce long-range Fock exchange without the LRHFCALC keyword. Any wisdom on the correctness/incorrectness of this implementation would be much appreciated!

Many thanks,
Giaan Kler-Young

[michael_wolloch]

Dear Giaan Kler-Young,

I am sorry for the somewhat late reply, and I apologize for not having a clear answer for now. I am not familiar with that part of the code, and the colleague who is the expert in these matters (Fabien Tran) is on holiday until after Easter. I chatted with some other devs, but the consensus was to wait until Fabien returns.

I, or Fabien, will get back to you next week and hopefully have some insight.

Thanks for your patience,
Michael

[fabien_tran1]

Hi,

VASP calculations with the wB97X functional are currently not possible. The reason is that it is not possible to combine the short- and long-range exact Hartree-Fock exchanges as specified by Eq. (2) in C3CP54374A. Therefore, the wB97X calculations presented in s41467-024-48609-2 should be wrong, unless they modified the source code, which I doubt. Furthermore, they used the old VASP.5.4.4 that does not support Libxc. I suggest to contact the authors of s41467-024-48609-2.

P.S. Hopefully with the next release of VASP, calculations with wB97X will be possible.