Handling f electrons

[tjf]

I'm doing some calculations on a Ce-containing perovskite.

Formally this should be Ce III. Yet if I do the calulation with the standard Ce potential, in which the f electon is "valence" (allowing Ce IV), I get an energy that is 0.2 eV/atom lower than if I use the Ce_3, frozen f potential. Now I realise that the VASP total energy is only an approximation to the true cohesive energy, but this is quite a large energy difference. In addition, the calculated charge density suggests that if allowed, the Ce f electron delocalises. However, that's of course subject to the RWIGS, which I haven't explored very much.

My question is, what is the most appropriate way to treat this system? My normal reaction would be that adding more electrons to the valence space should not change the solution if they are supposed to stay localised, whereas artificially freezing them (using the Ce_3 potential) can cause spurious results.

A related issue is what is the reference state for these heavy elements for the potential-generation code? Are they forced to be spherically-symmetric? Can this symmetrisation of the reference f electron account for the observed 0.2 eV/atom stabalisation?

And on that note, is there a Ce_3 LDA potential available?

[Veronika]

My own experience is the following: you should NEVER use the Ce potential with the f electron in the core.

[admin]

1) if you have good hints that the Ce-f electrons delocalize, please use the standard Ce potential (not the Ce_3 one)
2) there were no PAW-LDA Ce_3 potentials generated
3) concerning the free atoms:
--) no, the symmetry which is imposed to the free atom calculation is not a spherically symmetric one, but the symmetry of the box.
--) please also be careful with the electronic configuration of the free atom: the PPs were generated for the electronic ground state of the respective element in the bulk, which may differ from the electronic ground state of the free atom in the gas phase. (this is not the case for Ce, however). In case of doubt, please check the PP's configuration in V_RHFIN (this file is in the same directory as the POTCAR file of the respective atom) and set the correct gas-phase occupancies in INCAR (FERWE, FERDO). Please have a look into the online-handbook, chapter examples, free atoms, for the appropriate INCAR file in such a case
--) please note that the GGA potentials are slightly non-symmetric, therefore the E of the free atom may be lowered significantly if the symmetry of the cell of the free-atom calculation is lowered: please have a look at the example given (for the free O atom) at the VASP workshop: pp. 3-17 of handsonI.pdf at
http://cms.mpi.univie.ac.at/vasp-workshop
(session talks and documentation)