[Wannier] Spin polarized calculations with f-electrons

[Jonathan Yates]

On 29 Mar 2012, at 10:09, Mario Burbano wrote:
> 
> The results I have obtained so far for yttrium-doped CeO2 and scandium-doped CeO2 have been consistent with each other using either DFT program. By this I mean that the forces and the dipole moments obtained from the DFT codes show excellent agreement between them. However, it is when I carry out simulations with elements that have f-electrons (Ce3+, Gd3+, Sm3+) that I start having some serious problems. In the case of Gd-doped ceria, when I visualize the wannier90_centres.xyz file from the Wannier90 analysis of the data, I can see that the f-electrons that should have been assigned to Gd, have instead been localized near the nucleus of the Ce atoms, in addition to the expected 8 valence electrons. I have encountered similar problems with systems that include reduced cerium (Ce3+). In this case the single f-electrons from each Ce3+ are assigned to a single Ce atom (which according to the VASP output should not be reduced). These wannier centres are conspicuous among the others given that their spreads are typically higher (>4 Bohr^2) than those of all the other WCs in the system (2.5 – 3 Bohr^2). In addition, the up and down spin components from the wannier90 calculation are the same, which, I think, is unexpected. Finally, a comparison with the results from CPMD is not available here because CPMD does not have the DFT+U method implemented. 

Mario,

 I'm going to make some generic comments first, rather than think about the chemistry of your system. Although I don't think there should be particular issues in dealing with f-electron systems.

You should investigate if the WF you have found correspond a true minimum in the spread functional. 

A good indication is to check how real the WF are.
You are starting from random phases. What happens if you use a starting guess?

You are using the gamma-point. Is the vasp interface set up to take advantage of the fact that the bloch states are real at gamma? (the PWSCF interface is) If so, you can use the gamma-point routines within W90. The special gamma point routines are quicker, and have been found to be more robust ie less reliant on the start guess.



 Jonathan


-- 
Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK
tel: +44 (0)1865 612797                http://users.ox.ac.uk/~oums0549/