kgen: primitive and conventional cells

[ajjackson]

Summary of discussion with @utf and @scanlond based on user feedback.

• Users can get confusing/incorrect results when using kgen with a non-primitive POSCAR
• kgen warns that results may be incorrect and writes a _POSCARprim file along with a KPOINTS which corresponds to that primitive/standard cell setting.
• If the user ignores this warning and uses their current POSCAR with this KPOINTS file, their results will usually be wrong.
• A solution to this would be to refuse to output KPOINTS until kgen is satisfied it will match POSCAR
• A scenario in which this could be a problem is magnetically-ordered structures in which the true electronic unit cell resembles a supercell of an underlying lattice.
  • A simple solution is to provide an "I promise this is the primitive cell" flag which causes kgen to work with the cell given.
  • For magnetic materials it could be possible to provide information about magnetic moments, in which case this stuff should be inferred correctly (and a primitive cell consistent with the magnetic ordering might still be found).
  • If the user is determined to treat a magnetic material as though it has the periodicity of an underlying structural unit, they really need to look into band-unfolding instead.
• A key point is that we should not endorse the calculation of "band structure diagrams" with non-primitive cells as the results will contain band-folding.

Proposal from @utf

• By default refuse to generate the k-points if the cell convention doesn't match the one expected
• Have an option to override this behaviour
• Implement a magmom option to set the magnetic moments on each atom (the functionality for this is available in spglib, seekpath and pymatgen)
• Make kgen support line mode band structures to make band unfolding easier

[ajjackson]

I've had a bit of a play with the VASP line-mode. As far as I can determine:

• The same number of points will always be used along each line segment. This is the first number in the second line of the file, unless you set that to less than 2 in which case it is set to 2 (i.e. just the end points).
• Weighting information is ignored and all points will be evenly weighted.

At this point I can't really recommend actually using line-mode for anything other than a single segment. Our motivation for including this mode was for use with BandUP, which it appears actually uses a non-standard version of the format. I suggest that we instead provide a --bandup flag which generates a valid _KPOINTS_primcell.in file.

[ajjackson]

Adding "good first issue tag" to this; that mostly applies to the first two parts (i.e. refuse by default). Adding magmom shouldn't be many lines of code but will need a bit more familiarity with the ecosystem.