Erratic magnetization output!

[pdalach]

I am calculating magnetic moments on Mn ions in a perovskite system. Whenever I calculate the system without cation vacancies on the A or B sites, my mag-moment rationally follows from my INCAR MAGMOM tag.


When I include cation vacancies, I get a mag-moment when I omit the MAGMOM tag (w/ ISPIN=2 and LORBIT=11) forcing the system into a ferromagnetic state. These mag-moments are extremely erratic.

When I include cation vacancies and set the MAGMOM tag for ferrimagnetic calculations, I always get a mag-moment=0 on every atom for every interaction.

I am expecting somewhat exotic magnetization behavior but this seems more like a code issue.

Thoughts?

Thanks!

[admin]

please note that MAGMOM is just an initial guess for the magnetic moments, the calculated moments correspond to the difference on the integrated spin up and down charge densities. The fluctuations might therefore appear beacuse the calculations are not properly converged.
I therefore suggest to first of all check whether the densities and the spin densities are converged properly:
please have a look at CONTCAR:
1) are the ionic steps converged electronically, or did the electronic scf loop break after 60 steps (default of NELM) without having reached convergence
2) the numbers in the last column should be about 10^-3 or less. If they are not, please decrese the mixing parameters (AMIX, BMIX) and the spin density mixing parameters (AMIX_MAG, BMIX_MAG)
3) you can try to enforce a high-spin solution by choosing large initial moments (eg by a factor of 2 larger than the expected final values)

[pdalach]

Thank you for your response!

Solutions converge very rapidly, often under 10 electronic steps.

My concern isn't that the final solution has mag=0.0000....the system can be AFM. Rather, in my OUTCAR, it seems that all magnetization for every atom, for every ionic step, has a mag-moment=0.

Your suggestion of starting with MAGMOM choices larger that expected sounds like a great idea...I'll give it a try.

[pdalach]

Choosing large initial moments fixed some of the issues. I had to chose a factor of about 3 and, in come cases, increase NBANDS. There are still some irregularities but I am more confident they can be resolved.

Thanks.