Open JaydevSR opened 3 months ago
Thanks @JaydevSR.
I think parse_system
is probably an important one to fix, that gets used a lot. Otherwise I think a helpful message in irrfbz_path
for the case that dim > 1
and no atoms are given should cover most cases, I believe. Also I suppose adding a unit test to ensure we don't again trigger such a bug in the future would be useful.
Oh and having looked at the code for 1s: I think if you actually manually supply the kpath
it should work in 3D, but I might overlook something.
For a non-atomic system, i.e.,
atoms
andpositions
vectors are empty for the model, the band computation usingcompute_bands
gives the following error:The main problem is that there are multiple places where reduction over atoms is performed without passing an initial value. Particularly, even if the line
src\external\atomsbase.jl:48
is changed to provide a value to theinit
kwarg, there are places inSpglib.jl
andBrillouin.jl
where the same problem occurs.After discussing with @mfherbst over slack, other than making changes wherever reduction happens, one possible fix would be to short circuit for the case of no atoms where there's no need for creation of an spglib cell.
Also, a quick workaround is to define dummy atoms without adding them to the Hamiltonian, but it's very hacky.