Question about CexO2x-1 f electron location

Message

dido · #1 Post by **dido** » Sun May 25, 2008 9:58 am

Hi, everything, i got a question about DFT(PBE)+U CexO2x with an oxygen vacancy. According to the literature, the leftover f electron were localized on two (out of 4) neighboring Ce atoms.
But in my PDOS result, the f electron are delocalized on all four 4 neighboring Ce (which happen to be the feature of non DFT+U result) instead of only 2. Here are my settings:
ISMEAR = 0
ENCUT = 550
LORBIT = 12

MD Stuff
NELMIN = 8
NELM = 200
EDIFF = 5E-3
EDIFFG = -5E-2
NSW = 140
IBRION = 2
ISIF = 2
POTIM = 0.5
ISPIN = 2

NPAR = 1
LDAU=.TRUE.
LDAUTYPE = 2
LDAUL = -1 3
LDAUU = 0.0 5.0
LDAUJ = 0.0 0.0

LMAXMIX = 6
ADDGRID = .TRUE.

AMIX = 0.2
BMIX = 0.0001
AMIX_MAG = 0.8
BMIX_MAG = 0.0001

Can any one help me out? Thanks a lot!

#2 Post by **admin** » Mon May 26, 2008 4:40 pm

please rather look at the charge density distribution of the f-state (first have a look at PROCAR to check which state corresponds to that f-state, then plot the PARCHG file which you obtain by setting LPARD=.T. (and all other tags related to the calculation of band decomposed charge densities).
the local DOS may suggest wrong contributions if the spheres overlap.
Of course first of all check if the calculation is fully converged.

dido · #3 Post by **dido** » Wed May 28, 2008 9:56 pm

Thank admin a lot!
I'm sure the columns in DOSCAR are s(2total) p(6 total) d(10 total) and f(14 total)
But i'm not quiet sure what the bands stand for in PROCAR.............

One more question about LPARD calculation--do i have to set up IBAND too? If yes, how would you decide what bands to calculation if i'm only interested in f electron?

#4 Post by **admin** » Thu May 29, 2008 11:52 am

in PROCAR, you will have the partial local charges resolved for each state (k-point and KS-eigenvalue), so you can check the predominant character of each single eigenstate.
So, if you want to pick out the state of the leftover f-electron: yes, you have to give IBAND as well, please set it to the band index you find from PROCAR (if the f-band crosses some other bands, you will have to separate the k-points as well)