GGA + U and mag. mom.

Message

amoitra · #1 Post by **amoitra** » Sat Jun 28, 2008 6:08 am

I'm using paw-gga pp, with the following INCAR. I would like to do GGA + U calculation, what could be the possible problem that I am not getting the desirable result, and getting 1 miuB less than the expected, all the time??

ENCUT = 400
NSW = 50
IBRION = 2
ISIF = 3
LWAVE = .FALSE.
LCHARG = .FALSE.
ISPIN=2
VOSKOWN=1
LREAL = auto
MAGMOM = 2*0 2*4 2*4 4*-4 4*-4 12*4 38*0
LORBIT = 11
GGAU = .TRUE.
GGAUTYPE = 2
GGAUL = 2
GGAUU = 5.10
GGAUJ = 0.10
GGAUPRINT = 2
LMAXMIX = 4

appreciate your help.

#2 Post by **admin** » Wed Jul 02, 2008 1:15 pm

by defining the LDAU* parameters as GGAU*, they are ignored (please check your OUTCAR file): the input tags have to be set exactly by the same name as given in the handbook, whether the calculation is LDAU or GGAU is only specified by the XC-type of the POTCAR you use.

amoitra · #3 Post by **amoitra** » Wed Jul 02, 2008 8:07 pm

Thanks very much. I understood that using paw-GGA PP, and using notations like LDAU*, is the way of GGA+U calculation. Kindly let me know if I am wrong.

#4 Post by **admin** » Thu Jul 03, 2008 1:17 pm

the input parameters LDAU, LDAUL,... only are used by vasp to use the U corrections and to define the numerical values for U, the j and the l-splitting. They are not used to define which kind of XC (LDA,GGA,...) is used, this is exclusively done by the POTCAR (and the GGA-tag, if defined). When reading INCAR it is checked whether the sting "LDAU" (LDAUU,...) is defined in INCAR. Unless this string is matched precisely, the U-corrections are not taken into account at all.

magal · #5 Post by **magal** » Thu Nov 12, 2009 1:52 pm

I have a question about kind of pseudopotentials. Is it necessary to use PAW potentials to take into account the on site Coulomb interaction? Or ultra soft PP can also be used?

boris · #6 Post by **boris** » Thu Nov 12, 2009 3:31 pm

I think only PAW atomic data can be used to perform DFT+U calculations. The DFT+U calculations require a projection onto localized atomic orbitals used to calculate the occupancy matrices. The latter are also linked to the cutoff radius of the PAW spheres.

That's why you need PAW datasets.

magal · #7 Post by **magal** » Fri Nov 13, 2009 12:16 pm

Thank you for answer