I'm trying to simulate LiFePO4 with PAW under the GGA approximation.
Before modelling the final compound I'd like to check the single PAWs for the atoms.
I've taken the following input files for atompaw:
- Fe from the abinit table (semicore version)
- Li, O, P from the atompaw table
Li (bcc), Li2O (fcc) and Li3PO3 are well described with no use of compensation densities (usexcnhat=0 or noxcnhat in atompaw). I've checked mainly the lattice constant in comparison with LAPW calculations.
For iron the option usexcnhat in atompaw is required to get a good description of ferromagnetic Fe (bcc) and FeO (fcc).
In particular noxcnhat in atompaw causes a warning "mkdenpos : WARNING -Density went < 0" and the cell parameter are not correct.
If I switch on usexcnhat in Li, O and P I get wrong results in Li, Li2O, Li3PO4, so I'd like to generate a PAW for iron with no use of compensation charge density.
Do you have any suggestion? Or is it not wise?
Thanks a lot in advance and best regards,
MAuro.