Question about paramagnetic FeS2

[samar_fawzy]

Hi, I have a question regarding initializing iron pyrite FeS2 in a paramagnetic state.
How do I do that using the MAGMOM tag? would random initialization of the magnetic moment be correct?
for example in a super cell of 32 Fe and 64 S atoms, would something like MAGMOM = 17*2.5 15*-2.5 64*0 be correct?
And then maybe test for convergence?
Thanks a lot!

[fabien_tran1]

Hi,

The way you want to use MAGMOM seems technically correct. However, what is also important is to have some randomness in the position of the atoms with up and down spin moment, and since we don't know what they are, one can not say if "MAGMOM = 17*2.5 15*-2.5 64*0" is a reasonable choice.

The proper simulation of a paramagnetic state (random orientation of the magnetic moment) is in general not so easy, in particular with DFT. Considering several random configurations is maybe necessary. If you have not already done it, try to figure out the procedure that seems the most reasonable according the literature on the application of DFT on paramagnetic states.

[fabien_tran1]

It should be added that a state with a zero total magnetic moment would be more preferable. Thus, this should be "MAGMOM = 16*2.5 16*-2.5 64*0".

[mike_foster]

Modeling paramagnetic states where individual atoms have magnetic moments but the net spin is zero present challenges. There are many possible spin (up/down) configurations with 32 Fe atoms and they will all have different energies. The problem becomes worse considering non-colinear spin states. With that said, set NUPDOWN = 0 to maintain a net zero magnetic moment. I'm not sure what properties you are interested in but be careful comparing structures with different spin configurations. The energy difference between different spin configurations can easily be large than the property of interest.

[samar_fawzy]

Thank you so much for your responses.

Can you please elaborate more about "being careful comparing structures with different spin configurations"
I am actually trying to calculate the formation energy of native defects/interstitials, so setting NUPDOWN to 0 should make sense for the pristine structure, but it is expected for the different strucutres to have different magnetic moments/spins.
For instance, if NUPDOWN=0 is used for pristine, then for a strucutre with Fe vacancy in the -1 state, NELECT would increase by 1, and also NUPDOWN=1 should be used. Is my understanding correct?
But then yes, I have higher formation energies than those reported in literature, because of the large energy difference.

Can you please advise what is the best way to do this using MAGMOM and NUPDOWN tags?

[mike_foster]

Start by defining different up/down spin configurations and see what I'm talking about (e.g. MAGMOM= -2,2,-2,2... and 2 2 2 ... -2 -2 -2) and how much of an issue it may or may not be. You don't need to use NUPDOWN, it just ensures that you end up in the electronic state you want (assuming you know). With respect to defects, I would start by removing one Fe atom and maintain the same spin ordering as the case without a defect.