Some AFM states converge to wrong magnetic states.

[SangLee]

Dear Sir:

The relative energies of FM and AFM spin states have been tried to be calculated for a ABX3 crystal where A is an alkali earth metal, B is a 3d transition metal and X is a nonmetallic atom. Spin-polarized calculations for the crystal with the unit cell are ok but those for the crystal with the 2x1x1 supercell show some problems.

1. Some AFM states converge to an AFM state having a different spin arrangement from that given by MAGMOM keyword. For example, the initial AFM state with “MAGMOM=2*1, 2*-1, 2*1 2*-1� has converged to the final AFM state with “MAGMOM= 0.6, 2*-0.6, 2*0.6 2*-0.6, 0.6�.

2. Some AFM states converge to the nonmagnetic state, that is, the magnetic state having very small magnetic moments on B sites.

The calculations have been performed with the PAW-PBE potential and PBE functional. The KPOINTS partition has been chosen such that the E0 energy shows the convergence of 1 meV/FU for the nonmagnetic state. The ENCUT has been chosen such that the relative energy between nonmagnetic and FM states converges to 1 meV/FU. The initial CHGCAR for the spin-polarized calculations is the same as the CHGCAR obtained for the nonmagnetic state.

How do I make those AFM states converge to right magnetic states ?

Please, help me.

Thank you.

[admin]

1) the values chosen for MAGMOM are starting guesses , the system will relax to its self-consistent mgnetic solution.
2) it seems that this up/down boundary in your cell is more favorable than the one in your input,. Please check whether these 2 solutions are symmetry-equivalent. The question of the absolute values of local charges and magnetic moments has been discussed several times in the forum, please check.
3) have a look if the total energy of the non-magnetic states is lower than the one for the magnetic state to see whether it is the ground state. Please note that sometimes it is necessary to start with high MAGMOMs in order to relax into the high spin solution of the system.

[SangLee]

Dear Head-admin and users:

Thank you, Head-admin, for your kind reply.

I have figured out the solutions to these strange behaviors.

1) The sympton 1 mentioned in the original post was originated from the insufficient significant number for the cartesian coordinates of the ions within the supercell provided in POSCAR. When the significant number has been increased, the symption disappears.

2) The sympton 2 has been disappeared when ISYM=0 option has been included in the INCAR file.

I hope this reply help to other users.

Thank you.