Unreasonably large magnitude eigenvalues for spin-polarized calculations with VASP 5.3
Posted: Thu Aug 22, 2013 10:45 pm
I have a user doing a spin-polarized (ISPIN=2) of NiO using the PAW_PBE potentials
for Ni and O_s. Whereas this calculation is reasonable with VASP 5.2.x, runs with
VASP 5.3.2 and 5.3.3 have very large magnitude eigenvalues (exceed the range
of the output formatting) for the spin-component-2 terms. Has anyone else seen
this behavior? I need to get the user's permission to submit his files within a
formal bug report. Only the band energies and the average electrostatic potentials
of the Ni ions are large enough to exceed the output formatting and deviate
considerably between the VASP 5.2.x and VASP 5.3.x runs.
Thanks,
Scott Kajihara
for Ni and O_s. Whereas this calculation is reasonable with VASP 5.2.x, runs with
VASP 5.3.2 and 5.3.3 have very large magnitude eigenvalues (exceed the range
of the output formatting) for the spin-component-2 terms. Has anyone else seen
this behavior? I need to get the user's permission to submit his files within a
formal bug report. Only the band energies and the average electrostatic potentials
of the Ni ions are large enough to exceed the output formatting and deviate
considerably between the VASP 5.2.x and VASP 5.3.x runs.
Thanks,
Scott Kajihara