Does SOC band structure calculations require higher number of k-grids?

Question

In so many papers I see that they carry out soc band structure calculations with much higher kgrid than used for the geometry optimization. Whats the logic behind that?

Answer 1

Without knowing the specifics of the paper, I can give some general advice. It tends to be the other way around, geometry optimization tends to not require as fine of a kgrid and often it is reduced for computational savings.

Fine kgrids are generally speaking required for accurate electronic structure properties, while geometries tend to converge to the accuracy needed even with fairly coarse kgrids. It is also worth noting that with particularly large unit cells, you can get away with using just the gamma point for geometries sometimes but this would leave you with no band structure. Band structures often require a larger number of kpoints because you are tracing paths between high symmetry points in the Brillouin zone. This will be largely independent of cell size unless you use some sort of band unfolding procedure.

Maybe someone else can speak about your exact question, but I suspect this increase in kpoints is not directly related to SOC band structure calculations.

Answer 2

I can't understand what your logic is. The geometry optimization will use uniform k-mesh and the band calculation with SOC will use in general use the line-mode to plot the eigenvalue along the assigned k-path.

Due to the band calculation is a post-process, you of course can use a denser uniform k-mesh to plot smooth 3D band structure based on your previous converged self-consistent calculation with a suitable uniform k-mesh. Are you asking this?

Maybe you should link a paper to explain in detail your question.