BLYP functional

[Neutrino]

Is BLYP functional supported in VASP 5? How can I use it and with which Pseudopotential ?

Thanks.

[admin]

The extensive comparison of semilocal and hybrid functionals is published in
New Journal of Physics 10(2008)063020.
Details on the use of BLYP and B3LYP functionals can be found in:
http://cms.mpi.univie.ac.at/vasp-forum/ ... 3lyp#p8621

[Neutrino]

Thank you for your reply, admin. If I understand correctly the BLYP is composed of the Becke88 exchange and the Lee-Yang-Parr correlation. As such one needs to set in INCAR:

GGA=B3 ; AGGAX=1 ; AGGAC=0 ; ALDAC=1

However, I think for that to work one needs to start from a LYP LDA pseudopotential. Clarification is very appreciated in terms of which pseudopotential one need to start with and whether the tags I mentioned above are the right way to set BLYP.

Thank you also for the paper. The "surface argument" presented in this paper to explain the performance of semilocal functionals is interesting. It would be nice to see how this argument can be employed to explain the better performance of RPBE and BLYP in describing liquid water.

[Neutrino]

Dear admin, some clarification related to setting BLYP in VASP would be very appreciated.

Thank you,
M. Y.

[Neutrino]

Dear users and admin,

Some clear guidance on how to start BLYP calculations using VASP in terms of pseudopotential and INCAR tags is very appreciated. Thanks! M.Y.

[Neutrino]

Dear VASP admin and users,

I'm still confused about running BLYP calculations in VASP. On one hand I'm convinced now that starting from PBE PAW file should be fine as long as one is interested in energy differences. But The setup in INCAR is not clear. I think what I wrote above

Code: Select all

GGA=B3 ; AGGAX=1 ; AGGAC=0 ; ALDAC=1 

should be the correct way to employ the BLYP functional. But I tried to do one benchmark case on FCC Ag by calculating its lattice constant and comparing it with the results of the paper the admin posted (New Journal of Physics 10(2008)063020.) I get 4.34 A while in the paper it was reported to be 4.27.
My INCAR is:

Code: Select all

Metal
ISTART=0 ; ICHARG = 2
ISMEAR =1 ; SIGMA =0.1
ENCUT= 400
IBRION=2 ;  ISIF=3; NSW=1000
ADDGRID=.TRUE.
EDIFF = 1E-06
EDIFFG = -0.001
POTIM = 0.1
LWAVE =.FALSE.
PREC=Accurate
NPAR=1
ISPIN=2
GGA=B3
AGGAX=1
AGGAC=0
ALDAC=1

I also tried to set GGA=B3 but without any change in the defaults of AGGAX, AGGAC, ALDAC, and I get a very different value of 4.15 A.
I really appreciate any help in setting up accurate BYLP calculations in VASP.


Thank you,
Mostafa

[sungeng]

I am confused by using BLYP in vasp too,
It seems that PBE functional in vasp contains the exchange part of Becke03, so we can use the PBE in GGA tag. But I don't know how to set the correlation part of LYP functional.
If I understand correctly, BLYP is not hybrid functional but GGA functional. so BLYP must be set combined with correct POTCAR file.
(But I am not sure with this, Hopping for the admin to give clear instruction)

[sungeng]

Ok,

I am doing some dig in the vasp code,

It seems that, AGGAC should be set as 1.0. which stands for using the correlation part of B3LYP(from Lee, Yang and Parr, same as what is used in BLYP).

In B3LYP, the AGGAC =0.81, but in BLYP AGGAC is 1.0

[admin]

The comparison of semilocal and hybrid functionals,
published in New Journal of Physics 10(2008)063020,
uses the PBE functional as starting point
and for BLYP the following set of parameters:
GGA = B5
ALDAX = 1.00
AGGAX = 1.00
AGGAC = 1.00
ALDAC = 0.00

[Neutrino]

Thank you very much, admin! I could reproduce some of the lattice constants in the above mentioned paper using the setup you explained.