Using LRHFCALC

[hong_tang1]

Dear VASP team,

I have a question regarding the long-range HF exact exchange (LRHFCALC=.TRUE.) in VASP 6.4.3. Please help.

The vasp wiki shows that " If LRHFCALC=.TRUE., then AEXX=1 is automatically set, but AEXX can be set to another value." So, if I set AEXX=0.25, then it will use only 25% of long-range HF exact exchange, and the other 75% in the long-range exchange will be made up by the underlying DFT. is this correct?

Also, if I use XC tag with metaGGA, then the underlying DFT will be the metaGGA, right? Which means I will have the short-range exchange is 100% from metaGGA, and for the long-range exchange, 25% from HF exact exchange, 75% from metaGGA. is this correct?

Thank you very much!

Best,
Hong

[manuel_engel1]

Dear Hong,

Thank you for reaching out. You are correct about your first question. In principle, you are also correct about the second question, however, in practice the short-range part for the hybrid calculation is only available for very simple density functionals (e.g., LDA, PBE). So it will not work with metaGGA DFT functionals. We currently mention this in this section on our wiki.

I have asked my colleague who is a bigger expert in these matters to see if we currently support any other functionals that would work with hybrids. If we find anything, I will report back with a more detailed answer.

[hong_tang1]

Dear Manuel,

Thanks so much for helping!
If I put the following lines in the INCAR file, what will be the short-range part of the exchange? LDA? GGA? but I inputted R2SCAN there.

I tested that the following calculation ran smoothly. What does the code really do?

Best,
Hong

LASPH=.TRUE.
LHFCALC = .TRUE.
XC = R2SCAN_X R2SCAN_C
XC_C = 1.0 1.0
AEXX = 0.25
NKRED = 1
HFSCREEN = 1.20
LRHFCALC = .TRUE.

[manuel_engel1]

Dear Hong,

The calculation you mentioned runs through without error but it does not produce the expected result. The Hartree-Fock part if screened but not the semilocal part when using r2SCAN. This is wrong on the theory level. We will add an error message that makes this clear in the future and that doesn't allow for such calculations if the functional is not supported.

I have checked with my colleagues and can confirm that the only functionals that can currently be used with HFSCREEN>0 are LDA, PBE and PBEsol. Anything else will not give expected results.