inconsistent dielectric function matrix off-diagonal components

[jhpark]

Hello,

I used VASP 5.4.4 to calculate dielectric function of a material with magnetic moments and observed that identical structures, only differing the set of basis vectors (unit cell shape) in POSCAR results in inconsistent off-diagonal components of the dielectric function matrix: non-zero off-diagonal components in one set of basis vectors (basis1.zip) and zero values for the other set (basis2.zip).
Could you be able to help me figure out the reason for this behavior? Thank you!

[manuel_engel1]

Hi,

I would first try to minimize the differences between the two calculations. Make sure that both use the same INCAR tags, especially when it comes to accuracy (for example, same ENCUT). You should most likely also increase the number of bands. The [url=wiki/index.php/LOPTICS]LOPTICS[url] method usually requires a large number of conduction-band states to converge. The VASP wiki suggests a value two or three times as high as the VASP default. Once you are confident about convergence in both cases, I would try to compare the values again and see if the behavior persists.