No symmetries for Γ-point calculations

[abussy]

When doing Γ-point calculations, even with a perfectly symmetric crystal cell, the symmetries filed of the PlaneWaveBasis only contains a single element: One(SymOp). This is quite counterintuitive to me, and can be tracked down to:

DFTK.jl/src/symmetry.jl

Lines 167 to 171 in 781fa06

	function symmetries_preserving_kgrid(symmetries, kgrid::MonkhorstPack)
	if all(isone, kgrid.kgrid_size)
	# TODO Keeping this special casing from version of the code before refactor
	[one(SymOp)]
	else

This TODO was added in December 2023. Is it a bug or a feature, and should we add back symmetries for Γ-point calculations?

[antoine-levitt]

Yes that looks wrong

[mfherbst]

I kept this in the refactoring because of how it was before. Could be that this has been done for some optimisations or to avoid some edge cases, but I don't remember. Could be worth checking out if there is a better way to deal with this.

[mfherbst]

@Technici4n #1215 has fixed this, right ?

[Technici4n]

@Technici4n #1215 has fixed this, right ?

At least that branch is gone. But the basis has no symmetries in that Γ-point case, right?

[mfherbst]

Are you saying the basis should have no symmetries in the Γ-point case or are you saying DFTK master still gives a basis with no symmetries in the Γ-point case ?

[Technici4n]

I meant the former.

[Technici4n]

I guess what I said is wrong. There is no k-point reduction but the symmetries should all exist.

[mfherbst]

I agree with your last statement. in this case basis symmetries == model symmetries.

[mfherbst]

So we can close as done ?

[mfherbst]

Yes works as expected now: A gamma point calculation has all symmetries of the model.