Constraining the direction of magnetic moments

[caokun]

Hi all,

I am performing noncollinear calculations with a supercell containing 16 Cu atoms. I am following the instruction of VASP manual and trying to constain the magnetic moment of Cu ions.
The set of INCAR file:

PREC = high
ISTART = 0
ICHARG = 11
ISYM = 0

ISPIN = 2
LNONCOLLINEAR = .TRUE.
LSORBIT = .TRUE.
RWIGS = 1.17

NSW = 0
EDIFF = 1E-08

I_CONSTRAINED_M =1
M_CONSTR =0 -1 0 0 -1 0 0 0 1 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 -1 0 0 1 0 0 -1 0 0 -1 0 0 1 0 0 -1 0

Only RWIGS and M_CONSTR of Cu are shown above.
In this calcuation, initial charge density produced by previous collinear calculation is read.

According to the value of M_CONSTR, most of the moments should be constained to be parallel to y direction with one exception along z direction.
But the local magnetic moments of Cu in OSZICAR
are all collinear and parallel to z direction.

The OSZICAR:
ion MW_int M_int
1 0.000 0.000 -0.534 0.000 0.000 -0.680
2 0.000 0.000 -0.534 0.000 0.000 -0.680
3 0.000 0.000 0.534 0.000 0.000 0.680
4 0.000 0.000 -0.534 0.000 0.000 -0.680
5 0.000 0.000 0.534 0.000 0.000 0.680
6 0.000 0.000 0.534 0.000 0.000 0.680
7 0.000 0.000 -0.534 0.000 0.000 -0.680
8 0.000 0.000 0.534 0.000 0.000 0.680
9 0.000 0.000 0.534 0.000 0.000 0.680
10 0.000 0.000 0.534 0.000 0.000 0.680
11 0.000 0.000 -0.534 0.000 0.000 -0.680
12 0.000 0.000 0.534 0.000 0.000 0.680
13 0.000 0.000 -0.534 0.000 0.000 -0.680
14 0.000 0.000 -0.534 0.000 0.000 -0.680
15 0.000 0.000 0.534 0.000 0.000 0.680
16 0.000 0.000 -0.534 0.000 0.000 -0.680

I am sure that POSCAR, POTCAR and KPOINTS are correct.
Is there anything wrong in my settings?

Any suggestions would be appreciated.











<span class='smallblacktext'>[ Edited ]</span>

[forsdan]

From your settings above you seem to not have set the weight of the penalty terms (LAMBDA). The default is most likely zero, so there will be no penalty contribution. Try to set LAMBDA and subsequently increase it, to see if it changes the results.

I also assume that you are using a binary compiled without NGZhalf in order to be able to run non-collinear calculations.

Cheers,
/Dan

[caokun]

Thanks for your reply.
Actually, I set LAMBDA=10. I am sorry that it was missed in the first thread. I am indeed using a binary compiled without NGZhalf.
In addition, I have tried to subsequently increase LAMBDA, but the results didn't change.

Any other suggestions?