The ABINIT Users Mailing List ( CLOSED )

[Anurag Chaudhry <anuragchaudhry@yahoo.co.uk>]

Hello Everyone,

I am have been running ABINIT to calculate the density of states for few Cerium doped insulators for some time now.  In all calculations there is a single Ce ion sitting in a host supercell matrix (as a substitutional impurity).  And all systems have an odd number of total electrons.

I have run calculations with no spin polarization for these systems. A typical total DOS has Ce 4f electron state inside the host bandgap, i.e., there is a gap between 4f and the valence band maximum (VBM). Recently I repeated a calculation (same goemetry and everything else same) except for spin polarization.  Now the spin polarized DOS shows a different 4f - VBM gap. 

The 4f for Ce has only 1 electron in the system the spin up and spin down differ by only 1 electron.  I do not understand how spin polarization vs. no spin calculations should give different results for the filled 4f - VBM gap ? 

Any suggestions/comments are welcome.

All calculations are with TM pseudopotentials.  LDA and LSDA were used for no spin and spin polarized calculations.

I did the calculations for the same compound with VASP (PAW, PBE) and found that it gives the same 4f-VB gap for spin polarized as well as no spin calculations.

best regards,
Anurag