My Community

Hi everyone,

I need to do some calculations on Fe and FeCr dilute alloys in the ferro-magnetic, non-magnetic and anti-ferro magnetic states. The FM and NM states are not a problem, the AFM is. Just setting the initial spins antiFM for a system of mostly Fe doesn't work, VASP converges to a FM solution anyway most of the time. If it finds some other solution it is usually after very difficult convergence with a large difference between the input-output electron density at the final steps. I could try to enforce the moments with some tags in the INCAR file, but I wonder how accurate any results would be. Non-ground states are not too accurately described by DFT, how well would it do for an enforced AFM state for Fe? The difficult convergence and large input-output electron difference for anything other than the FM or NM states make me quite pessimistic about it. How accurate would a 'tortuously imposed' AFM state be? Does anyone know of any good reading on this subject?

greets,
Peter

If the relaxation inevitably tends towards a FM or NM ground state,
you can either
1) fix the total moment of the cell to 0 (NUPDOWN)
to avoid FM solutions or
2) set the directions of the magnetic moments: they can be enforced by constraining the moments:
the tags which have to be set in this case are
LNONCOLLINEAR = .True.
I_CONSTRAINED_M= 1
LAMBDA = r (real number the specify the weight of the penalty terms)
M_CONSTR= m_x m_y m_z (triplet for each ion)

please find additional hints in the VASP online manual,
chapter " Constraining the direction of magnetic moments "