Long-range hybrid functionals

[heisen]

Dear All,
Can general range-separated hybrid functionals be used in VASP? For example, the SRSH functional ( S. Refaely-Abramsonet al., Gap renormalization of molecular crystals from density-functional theory.Phys. Rev. B88, 081204 (2013) ) with a certain fraction of short-range Fock exchange and a different fraction of long-range fock exchange. I saw that the functionals of libxc can be used in VASP 6.3 Does that include long-range separated hybrid functionals? Does it include functionals such as SRSH ? Does it include wb97x-v?
I've been using a long-range separated hybrid in VASP 5.4.4 with:
LHFCALC = .TRUE.
HFSCREEN = 0.048
LRHFCALC = .TRUE.
AEXX = 1.0
ALDAC = 1.0
Which works when the fraction of short-range fock exchange fraction is zero and long-range fock exchange fraction is 1. I saw that this had been implemented in xclib.F by Iann Gerber.
How does one set-up the input parameters for a general SRSH, or can one only use the functionals specified explicitly in libxc? The manual was not so clear on this : https://www.vasp.at/wiki/index.php/LIBXC1
Thank you,
Dr. Helen Eisenberg,
Fritz Haber Center for Molecular Dynamics,
Hebrew University of Jerusalem

[fabien_tran1]

Hi,

Regarding range-separated functionals, the main limitation in VASP is that it is not possible to use two types of HF exchange at the same time. One can use full HF (HFSCREEN=0), SR-HF (HFSCREEN>0) or LR-HF (LRHFCALC=.TRUE., HFSCREEN>0), but not a combination of them. That means that the SRSH and w-B97X-V functionals can not be used with VASP.

Concerning Libxc, it provides only the semilocal part of a hybrid functional, and all codes using Libxc need their own implementation of the (SR/LR)-HF exchange.

Making general range-separated functionals available in VASP is something that we will keep in mind.

[heisen]

Thank you for your reply, it was very helpful.

I've been looking at the code for xclib.F and fock.F and I think with the current code I could use a range separated functional where the range separation is 1/r = AEXX*erf(HFSCREEN*r)/r +(1-AEXX*erf(HFSCREEN*r))/r. i.e SRSH with a zero fraction of short-range Hartree Fock.
I think implementing the full SRSH with a fraction (BEXX) of exact-exchange also in the short-range would require me to add to the fock.F routine in a similar way to the L_MODEL_HF:

POTFAK(NI)=SCALE/GSQU*(BEXX-AEXX*EXP(-GSQU*(TPI*TPI/(4*HFSCREEN*HFSCREEN))))
and elsewhere that L_MODEL_HF is called in fock.F

Does this look correct and sufficient?

Thank you,
Dr. Helen Eisenberg,
Fritz Haber Center for Molecular Dynamics
Hebrew University of Jerusalem

[fabien_tran1]

I'm not familiar with the HF implementation in VASP, but what you proposed seems plausible, except that it concerns only the plane-wave part. For the PAW spheres, you should have a look at pawfock.F. I can try to get some more info from colleagues if necessary.

[heisen]

Thank you, I'd really appreciate that

[fabien_tran1]

Hi,

I can also just mention that in pawfock.F the parameters have different names: RMU, BEXX and BRSEXX (see subroutine RSPARAMETERS in fock.F).

[mturiansky]

Hello!

I was wondering if there are any updates on the status of SRSH or general hybrid functionals in VASP?