Spinors in vasp 6.2.1 with Wannier90 3.1.0

[yang-hao_chan1]

Hi VASP users,

I am wondering if anyone has succeed in writing out .mmn and .amn files with spinor calculations in vasp 6.2.1.

Here is my INCAR file:

Code: Select all

System = TiSe2
  
ISMEAR = 0
SIGMA  = 0.05
EDIFF  = 1.d-5
IBRION = -1
LREAL  = F
IALGO  = 38
ENCUT  = 320.d0

MAGMOM = 9*0

LSORBIT = .TRUE.
LMAXMIX = 4  
SAXIS =  0 0 1
GGA_COMPAT = .FALSE. 

NEDOS = 3001
LORBIT = 11
NBANDS = 48

NCORE = 1

LWANNIER90 = .TRUE.
WANNIER90_WIN = "
begin projections
Ti:l=2
Se:l=1
end projections
NUM_WANN = 22
spinors=.true.

After the scf steps, the code complains that " Calling wannier_setup of wannier90 in library mode
param_get_projections: too few projection functions defined
Error on node 0: examine the output/error files for details"

When checking the wannier90.win written by vasp, the content is the following,

Code: Select all

begin projections
Ti:l=2
Se:l=1
end projections

# This part was generated automatically by VASP
num_bands = 48
num_wann = 34
spinors = .true.
begin unit_cell_cart
     3.0639979    -1.7690000     0.0000000
     3.0639979     1.7690000     0.0000000
     0.0000000     0.0000000    17.8012579
end unit_cell_cart
begin atoms_cart
Ti       0.0000000     0.0000000     8.9006289
Se       2.0426653    -0.0000000     7.3483294
Se       4.0853305     0.0000000    10.4529285
end atoms_cart
mp_grid =     2     2     1
begin kpoints
      0.000000000000      0.000000000000      0.000000000000
      0.500000000000      0.000000000000      0.000000000000
     -0.000000000000      0.500000000000      0.000000000000
     -0.500000000000     -0.500000000000      0.000000000000
end kpoints

For some reason the num_wann is 34 which is different from the number set in the INCAR.
If I manually change the num_wann to 22 (10 d orbitals + 12 p orbitals) and remove the WANNIER90_WIN line in the INCAR, vasp can successfully write out .mmn file
but the .amn file again has the wrong number of bands. The header of .amn is
"File generated by VASP: TiSe2
48 4 34"
The log content after the scf runs is,

Code: Select all

 Computing MMN
 Projection [ 1/22] done
 Projection [ 2/22] done
 Projection [ 3/22] done
 Projection [ 4/22] done
 Projection [ 5/22] done
 Projection [ 6/22] done
 Projection [ 7/22] done
 Projection [ 8/22] done
 Projection [ 9/22] done
 Projection [10/22] done
 Projection [11/22] done
 Projection [12/22] done
 Projection [13/22] done
 Projection [14/22] done
 Projection [15/22] done
 Projection [16/22] done
 Projection [17/22] done
 Projection [18/22] done
 Projection [19/22] done
 Projection [20/22] done
 Projection [21/22] done
 Projection [22/22] done

Intel MKL ERROR: Parameter 4 was incorrect on entry to DLASCL.

Intel MKL ERROR: Parameter 4 was incorrect on entry to DLASCL.

If I go on with the wannier90.x calculation, the code complains that the info in .amn is not correct.

Please let me know if you have any insight on this problem.
Thank you and Happy Holidays!
Yang-hao

[henrique_miranda]

Hi Yang-hao,

Could you set NUM_WANN outside of the WANNIER90_WIN tag?
When NUM_WANN is not set vasp will try to guess the number of wannier states from the results of the wannier90 library call but this does not seem to work correctly.
The recommended approach is to always set NUM_WANN manually in the INCAR.

Let me know if this works.

[yang-hao_chan1]

Hi Henrique,

Yes, setting NUM_WANN outside of the block works.
VASP can now generate correct .amn file for wannier90 run.

Thank you very much for your help!