.amn files have projectability larger than 1

[hugodd]

Hi all,

I have encountered a problem when trying to perform projectability disentanglement in graphene.

I run the SCF calculation with the following INCAR:

Code: Select all

System = graphene
PREC = Accurate               

ISTART = 0                    

ICHARG = 2                   

LORBIT=11
NBANDS=24

KPAR=1
NCORE=1
ENCUT = 500
ISMEAR = 0                    

SIGMA = 0.1

ISYM = -1
LWAVE = .TRUE.           

LCHARG = .TRUE.            


LSORBIT = .TRUE.
MAGMOM=6*0.0
LNONCOLLINEAR = .TRUE.

LWANNIER90=.TRUE.
NUM_WANN=4
LWRITE_MMN_AMN=.TRUE.
WANNIER90_WIN = "

begin_projections
C:pz
end_projections

write_hr = .true.
spinors = .true.

num_iter          =   0
num_print_cycles  =   50
dis_win_min = -4.0
dis_win_max =  1.0
dis_froz_min = -2.75
dis_froz_max = -1.0
dis_num_iter  = 100
"

Which runs without issues. Then when I run wannier90 using the .mmn, .amn, .eig files with the following wannier90.win input:

Code: Select all

begin_projections
C:pz
end_projections

write_hr = .true.

num_iter          =   0
num_print_cycles  =   50
dis_froz_proj = .true.

# This part was generated automatically by VASP
num_bands = 24
num_wann = 4
spinors = .true.
begin unit_cell_cart
     2.4580000     0.0000000     0.0000000
    -1.2290000     2.1286904     0.0000000
    -0.0000000    -0.0000000    25.2060000
end unit_cell_cart
begin atoms_cart
C        0.0000000     1.4191270     0.0000000
C        1.2290000     0.7095635     0.0000000
end atoms_cart
mp_grid =     5     5     1
begin kpoints
      0.000000000000      0.000000000000      0.000000000000
      0.200000000000      0.000000000000      0.000000000000
      0.400000000000      0.000000000000      0.000000000000
     -0.400000000000      0.000000000000      0.000000000000
     -0.200000000000      0.000000000000      0.000000000000
      0.000000000000      0.200000000000      0.000000000000
      0.200000000000      0.200000000000      0.000000000000
      0.400000000000      0.200000000000      0.000000000000
     -0.400000000000      0.200000000000      0.000000000000
     -0.200000000000      0.200000000000      0.000000000000
      0.000000000000      0.400000000000      0.000000000000
      0.200000000000      0.400000000000      0.000000000000
      0.400000000000      0.400000000000      0.000000000000
     -0.400000000000      0.400000000000      0.000000000000
     -0.200000000000      0.400000000000      0.000000000000
      0.000000000000     -0.400000000000      0.000000000000
      0.200000000000     -0.400000000000      0.000000000000
      0.400000000000     -0.400000000000      0.000000000000
     -0.400000000000     -0.400000000000      0.000000000000
     -0.200000000000     -0.400000000000      0.000000000000
      0.000000000000     -0.200000000000      0.000000000000
      0.200000000000     -0.200000000000      0.000000000000
      0.400000000000     -0.200000000000      0.000000000000
     -0.400000000000     -0.200000000000      0.000000000000
     -0.200000000000     -0.200000000000      0.000000000000
end kpoints

I get the following error message from wannier90:

Code: Select all

......
Projectability:    0.00000    0.00000    1.60130    0.00000    0.00000    0.00000    0.00000    0.00000    0.00000    0.00000
   0.00000    0.00000    0.00000    0.00000    0.00000    0.00000    0.00000    0.00000    0.00000    0.00000
   0.00000    0.00000    0.00000    0.00000
 Exiting.......
 dis_windows_proj: projectability < 0.0 or > 1.0
 

This seems to suggest that the .amn files are not being normalised correctly? I checked this with vasp 6.5.0 and vasp 6.5.1, which give the same issue. I attach the input/output files, let me know if I have missed something.

Many thanks,
Hugo

[manuel_engel1]

Hello Hugo, and welcome to the forum!

Thank you for posting about this issue. You are using a rather recent feature of the wannier90 code (projectability disentanglement) that we haven't tested on our side yet. Therefore, I cannot recommend a standard course of action here. However, I can point you to a few things you can still try.

Your current INCAR file has a WANNIER90_WIN input section as well as a NUM_WANN tag. I would like you to remove both of those from the INCAR and instead use the LOCPROJ tag to specify the projections, like this:

Code: Select all

LOCPROJ = 1 2 : pz : Hy

The "1 2" simply means atoms 1 and 2, the "pz" is the orbital character and the "Hy" means hydrogen-like (this will also generate a LOCPROJ file which contains some information about the projections you made). You can try to change the "Hy" to "Ps" or "Pr" which pick different projection sites from the pseudopotential information (this is something the Wannier90 team recommend in their documentation about projectability disentanglement). It's possible that this will give you the correct normalization, but I wouldn't bet on it. Importantly, in this workflow, you do not need a projections block in the wannier90 input file and also you shouldn't manually specify the number of wannier orbitals via NUM_WANN.

Please let me know how it goes. In the meantime, I will add this to my to-do list for future improvements of the VASP-wannier90 interface.

[hugodd]

Hi Manuel,

Many thanks for your quick answer!

I have tried with the LOCPROJ tag as well and ran into the same issue, but I will test the exact parameters that you recommend. I will update you with the results.

I can also say that I faced the same issue when not using spin-orbit coupling, in case this helps.

Hugo

[hugodd]

Hi Manuel,

I have tested it again with the following INCAR:

System = graphene
PREC = Accurate
ISTART = 0
ICHARG = 2
LORBIT=11
NBANDS=24

KPAR=1
NCORE=1
ENCUT = 500
ISMEAR = 0
SIGMA = 0.1

ISYM = -1
LWAVE = .TRUE.
LCHARG = .TRUE.

LSORBIT = .TRUE.
MAGMOM=6*0.0
LNONCOLLINEAR = .TRUE.

LWANNIER90=.TRUE.
LWRITE_MMN_AMN=.TRUE.
LOCPROJ = 1 2 : pz : Hy

and got a very similar error, although with different value for the projectability:

Projectability: 0.00000 0.00000 1.30326 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
0.00000 0.00000 0.00000 0.00000
Exiting.......
dis_windows_proj: projectability < 0.0 or > 1.0

Thanks

[manuel_engel1]

Hi Hugo,

Thanks for reporting back with the new data. It's unfortunate but this confirms that the feature will not work until we implement a proper support for it. It might work by chance with the PS and PR projectors, since those are normalized differently compared to the Hydrogen-like orbitals, but this is something I cannot guarantee in general as of now. It's an interesting feature, so I will see how we can support this in the future.

[hugodd]

I see, many thanks Manuel!