Perfromace of DFT functionals: meta-GGA vs. hybrid meta-GGA

[Mostafa Abedi]

Hi Everyone,

I'm trying to get a sense of the performance of different DFT functionals in bulk calculations (i.e.  under PBC), specifically meta-GGA vs. hybrid meta-GGA. As far as I know, hybrid meta-GGA ones must be more expensive. I ran some test single point + gradient calculations using the B97MV and wB97MV functionals and surprisingly I got the same calculation time for both. I guess I'm not doing the calculations with wB97MV correctly. The input files are attached. Any comments/suggestions are greatly appreciated.

Many thanks,

Mostafa  

[Frederick Stein]

Dear Mostafa,

You are missing the &HF section in the wB97MV functional. Check https://github.com/cp2k/cp2k/blob/master/tests/QS/regtest-libxc/H2O-hybrid-wb97mv-libxc.inp for reference. Please double-check the parameters! In case of PBC, you have to replace 'POTENTIAL_TYPE MIX_CL' with  'POTENTIAL_TYPE MIX_CL_TRUNC' and set the CUTOFF_RADIUS to a bit less than half of the nearest-neighbor distance.

In general, the HF section is not set automatically by CP2K but has to be provided by the user (at least, with LibXC).

HTH,

Frederick

[Mostafa Abedi]

Dear Frederick,

Many thanks for your reply. Yes, you're totally right. The &HF section is missing. Now the calculations make sense. Thank you.

Mostafa

[Mostafa Abedi]

Does anyone know any papers in the literature that compare/evaluate the performance of hybrid (meta-)GGA with meta-GGA and GGA functionals for PBC calculations? 

Many thanks,

Mostafa