Page 1 of 1

Vasp 6.2.0 with Wannier 3.1.0: wrong num_wann

Posted: Wed Feb 17, 2021 4:45 pm
by mistguy
Hi,

I am using vasp 6.2.0 compiled with Wannier 3.1.0.

The INCAR is

Code: Select all

SYSTEM        =    FePS

ISTART        =    0
ICHARG        =    2
LWAVE         =    .FALSE.
LORBIT        =    11
LWANNIER90    =    .TRUE.
WANNIER90_WIN = "
begin projections
Fe:l=2
P:l=1
S:l=1
end projections
"

NCORE         =    1

ISYM          =    -1

LSORBIT       =    .TRUE.
MAGMOM        =    48*0 0 0 6 0 0 6 0 0 -6 0 0 -6
LMAXMIX       =    4

ISMEAR        =    0
SIGMA         =    0.04

NELM          =    240
NELMIN        =    5
EDIFF         =    1.0E-06

AMIX          =    0.2
AMIN          =    0.01
BMIX          =    0.0001
AMIX_MAG      =    0.8
BMIX_MAG      =    0.0001


GGA_COMPAT    =    .FALSE.
LREAL         =    .FALSE.
LASPH         =    .TRUE.
ALGO          =    Normal
NBANDS        =    240
ECUT          =    350
PREC          =    Accurate

Hubbard U
LDAU      =    .TRUE.
LDAUTYPE  =    2
LDAUL     =    -1 -1 2
LDAUU     =    0.00 0.00 3.00

IVDW          =    12
The POSCAR is

Code: Select all

PSFe
1.0
    5.94700000000000    0.00000000000000    0.00000000000000
    0.00000000000000   10.30000000000000    0.00000000000000
    0.99023394138609    5.15000000000000    6.42276729616994
    P    S   Fe
    4   12    4
cartesian
    6.44294882788179   12.87500000000000    4.30839230227080
    3.46778511350430    7.72500000000000    2.11437499389914
    3.46944882788179    7.72500000000000    4.30839230227080
    0.49428511350430    2.57500000000000    2.11437499389914
    4.46946454779196   12.87500000000000    4.80037627715741
    5.44126939359413    7.72500000000000    1.62239101901253
    1.49596454779196    7.72500000000000    4.80037627715741
    2.46776939359413    2.57500000000000    1.62239101901253
    1.47891601476266    4.28377000000000    4.80358766080550
    2.48481792662343    9.43377000000000    1.61917963536444
    2.48481792662343    6.01623000000000    1.61917963536444
    1.47891601476266   11.16623000000000    4.80358766080550
    4.45241601476266    9.43377000000000    4.80358766080550
    5.45831792662343    4.28377000000000    1.61917963536444
    5.45831792662343   11.16623000000000    1.61917963536444
    4.45241601476266    6.01623000000000    4.80358766080550
    0.49511697069304    6.00284000000000    3.21138364808497
    3.46861697069304    4.29716000000000    3.21138364808497
    0.49511697069304    9.44716000000000    3.21138364808497
    3.46861697069304   11.15284000000000    3.21138364808497
However, the wannier90 library complains too few projections defined. After examining the wannier90.win generated by vasp, I found that num_wann was set to 176, instead of the correct number of Wannier orbitals (136). I tried to defined num_wann in WANNIER90_WIN or change the projections, but num_wann in wannier90.win was always set to the magic number 176, which I don't know where it comes from.

Is this a bug of vasp?

Re: Vasp 6.2.0 with Wannier 3.1.0: wrong num_wann

Posted: Wed Feb 17, 2021 9:24 pm
by henrique_miranda
There is a new variable to set the number of Wannier projections when specifying the WANNIER90_WIN input inside INCAR:
wiki/index.php/NUM_WANN
I will soon update the wiki page with more detailed documentation about the new wannier interface in VASP 6.2.0.

Keep in mind that you can always run wannier90.x as a standalone using the wannier90.win file generated by vasp.

Re: Vasp 6.2.0 with Wannier 3.1.0: wrong num_wann

Posted: Wed Feb 17, 2021 10:56 pm
by mistguy
Thanks. However, the old behavior is that vasp will automatically append atom positions, k points, etc to wannier90.win. In vasp-6.2.0, these information is not appended if wannier90.win already exists.

Re: Vasp 6.2.0 with Wannier 3.1.0: wrong num_wann

Posted: Thu Feb 18, 2021 12:01 pm
by henrique_miranda
Yes, it is slightly different.
Note that in the previous versions, if the unit_cell_cart, atoms_cart, mp_grid, kpoints were already present in wannnier90.win they would not be replaced either. Instead, a warning message was displayed.

Using vasp.6.2.0 we created the WANNIER90_WIN input variable to have a single input file for the calculation.
This makes it easier to change the POSCAR or KPOINTS file without editing the wannier90.win file (even if editing, in this case, would mean just removing the old sections for unit_cell_cart, atoms_cart, mp_grid, kpoints from wannier90.win).

Another point in this change is that it makes it easier to ensure the consistency of the spinor handling in the VASP calculation and wannier90 as I mentioned here:
forum/viewtopic.php?f=4&t=18040#p20026