Internal error (bandgap_tools.F) during HSE calculation in VASP 6.4.3

[alicja_hertmanowska]

Hi,

I'm trying an HSE with AEXX=0.15 calculation for a LiMn0.17Ni0 oxide structure with VASP 6.4.3. However, I'm getting a bug with this calculation, and I'm wondering how to resolve this issue and whether there is anything wrong with the calculation.

The error is:

Code: Select all

internal error in: bandgap_tools.F at line: 316

Could not determine new bound in interval bisection. Most likely this means that the eigenvalues are not sorted.

The electronic minimisation doesn't appear to converge. I also performed this calculation with HSE directly, without the PBE calculation.

In a previous attempt, I also got this error:

Code: Select all

internal error in: rot.F at line: 805

EDWAV: internal error, the gradient is not orthogonal 1 1 4.372e-3

Any help with this would be much appreciated - thank you.

[henrique_miranda]

Hi, the bandgap_tools.F error is a known bug: the gap-finder assumed eigenvalues are always sorted, which isn't guaranteed (e.g. with ALGO = Fast). We fixed it in vasp 6.6.0 and I just added it to the KnownIssues#98.

The rot.F EDWAV error looks like a separate, unrelated convergence issue.
Does it persist when you use vasp 6.6.0?

[EDIT] I just noticed you did not specify initial magnetic moments, does the issue go away when you set them using the MAGMOM tag?