Spin orbit and noncollinear calculations

[bth20]

Hi there,

I am performing noncollinear calculations (with spin-orbit) on a 4-atom cobalt cluster. I know the result I should get as it has been produced by other authors using VASP. It should be highly noncollinear with the magnetization density basically spherically symmetric. However all my calculations result in a completely collinear magnetization density when I subsequently plot it.

I am following the VASP guidelines for performing the calculation. I begin with a collinear run with the following INCAR file:

SYSTEM = Cr cluster (collinear)
PREC = High
NELMIN = 4
NELM = 200
NBANDS = 40
EDIFF = 1.0e-4
ISPIN = 1
NSW = 100
IBRION = 1
ISIF = 2

I then use the resulting WAVECAR and CHGCAR files in a noncollinear calculation with the INCAR file:

SYSTEM = Cr cluster (noncollinear)
LNONCOLLINEAR = .TRUE.
LSORBIT = .TRUE.
ICHARG = 1
MAGMOM = -1 0 0 1 0 0 0 1 0 0 0 1
ISYM = 0
PREC = High
NELM = 200
NELMIN = 4
NBANDS = 80
EDIFF = 1.0e-6
ISPIN = 2
NSW = 100
IBRION = 1
ISIF = 2
RWIGS = 1.323

(I realize I am relaxing the system in both runs – this is not the problem as I have tried all possible combinations in this respect).

Is there something about what I’ve done here which is enforcing the collinearity? I am certain there are no problems with the KPOINTS, POSCAR and POTCAR files.

Any suggestions would be much appreciated.

Also, another (possibly related) question is this: the second method listed for specifying the initial magnetic moments is

MAGMOM = 0 0 total_magnetic_moment ! local magnetic moment parallel to SAXIS
SAXIS = x y z ! quantisation axis parallel to vector (x,y,z)

But how do you do this for cases in which there is more than one atom in the cell and you want them pointing in random directions? You can’t after all specify several SAXIS values.

Regards

Ben Hope

[admin]

From the technical point of view of the calculation, you have done everything
correctly.
-please mind that the magnetic structure is very closely related to the geometry,
there might be several different cluster configurations with different magnetic behaviour.
- have you also compared the local, direction-dependent magnetizations as written
in OUTCAR?
- concerning the MAGMOM tag in INCAR for non-collinear structures: The
tag specifies the magnetic moment at each site with respect to SAXIS,
SAXIS is the overall magnetization direction of the unit cell. Hence, if the
starting configuration of the moments should be random, define the MAGMOMs
as random vectors (they need not be given as integer numbers)