Variation in Magnetic Moments depending on Supercell Size

[mdh1992]

Dear all,

I am looking at running some bulk calculations for some antiperovskite materials (A3BX). I have only just started with some very preliminary calculations, so I am running some tests to determine suitable parameters. I got a rather odd result - starting from the crystallographic unit cell (i.e. .cif file), the wavefunctions were converged, but oddly I got different local magnetic moments for each A atom in my unit cell, despite the fact that they ought to be symmetry equivalent (the experimentally determined unit cell is undistorted), although they were all ferromagnetically aligned. The difference between the largest and smallest local moment was 0.75 BM. I haven't attempted any optimisation yet, so the atoms should be precisely fixed on their crystallographic lattice sites.

I repeated the calculation but multiplied the unit cell to a 2x2x2 supercell, and adjusted the MP k-point mesh accordingly. This time, the A atom local magnetic moments were almost the same (identical to 1 decimal point, but varying with a range of 0.02 BM). The B and X atom local moments were unchanged from the 1x1x1 cell.

I am puzzled as to why this is the case - it would make more sense if there were a complex antiferromagnetic ordering that could not be accommodated within the symmetry imposed by the smaller cell. However, in both cases, the calculation converged on a ferromagnetic solution. Moreover, the total magnetic moment per formula unit was more or less unchanged, i.e. the average of the local magnetic moments for the A atoms in the 1x1x1 roughly equal that of each atom in the 2x2x2 cell.

I am probably missing something obvious here. Any comments or suggestions would be most appreciated.

[mdh1992]

Just an update - I repeated the calculation for the 1x1x1 cell with symmetry switched off (ISYM=0), which appears to have resolved the issue (of course, the calculation was much slower). Again, perhaps I am missing something, but it seems counterintuitive to me that switching the symmetry off would in fact appear to result in symmetric local magnetic moments. Again, comments and suggestions are welcome.

[marie-therese.huebsch]

Hi,
very nice work to carefully take into account the symmetry of your system. It is difficult to make any comments without seeing the details of your calculation, though.

As a first step, I would compare the symmetry operations you expect to be present based on the MAGMOM tag and POSCAR file with the symmetry reported in the OUTCAR file.

For further discussion, please consider sharing the input files, OUTCAR and stdout of your calculation.
Best regards,
Marie-Therese