Hi Naveem,

If you set

RWIGS = [radii for each element]

in the INCAR file, you will get the spin on each atom within a sphere of these radii written in the OUTCAR file. That will make it easier to see what the magnetic state of your system is. Older verisons of VASP required you also set

NPAR = 1

if you want to get the momentns projecte on atoms, I don't know if that is still the case for VASP v6.3 that you are using.

Also, if you include OSZICAR files in your files archive, that makes it much easier for people to check how well your runs are converging.

Kind regards,

Peter

Naveen Sharma, you brought up a very relevant topic. After a few years of doing DFT using VASP, I just came to know the 'power' of the MAGMOM tag in the past few months. My understanding was that it just initializes the local magnetic moments, and the calculation takes care of finding the optimal values by relaxing the system to its lowest energy state. However, that's not the case, though! While working in an AFM-ordered system, I also noticed that different initial guesses end up in different magnetic configurations ~that was surprising to me, and like you said, I wasn't satisfied with the findings either. For FM-ordered systems, it appears to not be a problem.

This remains an open question for me. If knowing the precise spin configuration beforehand is mandatory, then are spin-polarized calculations for AFM-ordered systems truly first-principles? -No!

Peter Klaver, in the light of what I said, how can the 'RWIGS' tag help, or would you be able to spot where I am mistaken?

Gabriel Vinicius ,

If the precise spin configuration is not known, then one approach is to try different magntic configurations and see which one results in the lowest system energy. That takes out the non-first-principles aspect. See e.g. figure 1 in T. P. C. Klaver et al, PHYSICAL REVIEW B 85, 174111 (2012).

Knowing the moments through RWIGS can be helpful in that it lets you see what the moments are in the cxonverged state. That is not always the same as the initial state set in the MAGMOM line. Some systems can easily 'make up theirt own mind' as far as magnetic moments are concenred, which may be very diffewrent from what you set in the MAGMOM line.

Thanks Peter Klaver for clarifying and providing the reference. I will still go into detail on your suggestions, but after skimming through your paper, I have already found some useful things to think about.

Thanks for your help, though!

Thank you Peter Klaver and Gabriel Vinicius for your answers, I performed the calculations with some changes and have calculated the relative energy difference between all the configurations. Thank you.

You can also fix the total magnetic moment with NUPDOWN. You can use this to do a two-step calculation to nudge the system into some state: In the frist run, estimate the MAGMOMs and set them and NUPDOWN accordingly. After this finishes, restart the job with NUPDOWN set to allow the moments to relax. I used this in some tricky systems with interfaces between ferromagnetic and non-ferromagnetic parts which was difficult to converge to a reasonable state.

Martin Baeker, do you by any chance have an example INCAR where you might be able to show how you define an antiferromagnetic structure with NUPDOWN? The documentation on the VASP wiki has no information on this as best as I can tell.

I have been trying to do single layer and double layer antiferromagnets in Iron and am struggling to define a system where both atoms still have a "frozen" magnetic moment with NUPDOWN but the net moment of the structure is still 0. Thanks so much!

I don't think you can do that always.

NUPDOWN only fixes the total magnetic moment.

So if you have

NUPDOWN=0

MAGMOMS= +1 -1

you initialize the two atoms with opposite magmoms and nupdown=0 should ensure that the two moments stay opposite, but it might still converge to all moments being 0.

If you have more atoms, you have no handle in the magmoms of individual atoms (because this is not really a well-defined quantity in a plane-wave framework), you can only ensure that the total moment is zero.

In my experience (for bulk alloys), playing around with different initializations is necessary to get the magnetic state you want; I have seen cases where I could get alloying atoms in a Ni matrix being ferromagnetic wrt to the Ni atoms, antiferromagnetic or nonmagnetic depending on the initialization.