Band-edge positions evaluated via sumo-bandstats in collinear antiferromagnetic NiO

[FPrith]

Is it possible to change the method of searching for band-edge positions in spin-polarized systems?

My purpose is on the quick calculation of effective masses in various insulators including spin-polarized systems with non-zero spin magnetic moments.

As shown in the band-dispersion diagram attached here, in collinear antiferromagnetic NiO (AFM NiO), its VBM and CBM are in (0.500, 0.500, 0.500) (T point) and (-0.438, 0.312, 0.312) (in T-H2 line), respectively. This diagram was created via vise, which is developed by Y. Kumagai. The band-edge positions mentioned above are also visually confirmed in a vasprun.xml file and an OUTCAR file. band_NiO_ColinearAFM.pdf

However, the command sumo-bandstats seems to give a different result from the above. The CBM position is not consistent with the above result. Meanwhile, in spin-unpolarized systems, VBM and CBM positions seem to be successfully found. The part of the results of the command sumo-bandstats in AFM NiO is as follows:

Valence band maximum: Energy: 6.623 eV k-point: [0.50, 0.50, 0.50] k-point location: T k-point indices: 32 k-point degeneracy: 1 Band indices: 21, 22(Up), 21, 22(Down)

Conduction band minimum: Energy: 10.020 eV k-point: [0.61, 0.39, 0.50] k-point location: between GAMMA-H_2 k-point indices: 55 k-point degeneracy: 6 Band indices: 23(Up), 23(Down)

[ajjackson]

There is no direct Gamma-H2 segment in this band dispersion plot; it seems to be calculated on a different k-point path to the sumo-bandstats data? Do you have a sumo-bandplot output we can compare with the vise plot? Is it from the same calculation?

Can you also clarify which spacegroup we are looking at? By default sumo-kgen checks the symmetry from the atomic positions, but it can be instructed to use a different spacegroup if you know that would be more appropriate. That might be appropriate when dealing with a magnetically ordered system.

[FPrith]

Thank you very much for reply!

The high-symmetry k-point paths in the fully relaxed AFM-NiO cell were gererated via the SeeK-path program in vise. I knew that sumo had also the SeeK-path program yesterday, but in my calculation mentioned above, the SeeK-path program in vise was used.

A band dispersion diagram created via sumo-bandplot is here. Compared with the diagrams created via vise and sumo-bandplots, the same k-point positions of the band egdes (VBM and CBM) appeared; that is, the VBM position was at the T point, and the CBM position was at the point in the T-H2 line. band_NiO_CollinearAFM_sumo-bandplot.pdf

The space group in my AFM-NiO cell with magnetic moment in two Ni atoms was R-3m. R-3m appears only in the case of non-zero Ni magnetic moment. On the other hand, a conventional NiO cell without magnetic moment, as you may know, its space group is Fm-3m. Your opinion is that this difference in space group judging from atomic positions can be problematic in my analysis, isn't it? If there is a way to force the space group R-3m to be imposed, could you please let me know?

[ajjackson]

You can force sumo-kgen to use the seekpath path with --seekpath, and to use a given spacegroup with --spg. https://smtg-bham.github.io/sumo/sumo-kgen.html#usage You need the spacegroup number, which for R-3m is 166.

It looks like the paths are the same in your plots, anyway, so we can compare directly. Are you sure that sumo-bandstats gives the output in the initial post when run on the file plotted here? It is quite puzzling that it would find a point on the Gamma->H2 path if that path does not exist in the data.

Magnetic spacegroups can be a headache indeed and you will probably have to deal with them at some point: but as these plots are on the same path I don't think that is causing the discrepancy.