Bug: atomic positions in charge density Gaussian cube file are incorrect

[kirk0830]

Describe the bug

Just run the example/lcao_Si2 with out_chg 1, open the SPIN1_CHG.cube, you can find:

This is definitely not an expected result.

[AsTonyshment]

@kirk0830 I suspect this bug arises from not properly handling the STRU file / cube file when the determinant of the LATTICE_VECTORS is negative, which is actually contrary to standard conventions. According to crystallographic conventions, the lattice vectors should span the lattice according to the right-hand rule.

The STRU file provided in examples/charge_density/lcao_nspin1_Si2 is actually left-handed (if you calculate the determinant, you will find it to be negative). I attempted to exchange two of the basis vectors and obtained the correct result:

P.S.: In VASP, if the POSCAR file is left-handed, the job is canceled and a warning is issued:

ERROR: the triple product of the basis vectors is negative exchange two basis vectors

Perhaps ABACUS could implement a similar check?

I hope my comment is helpful.