My Community

I have done a band structure calculation for elemental Ce, PAW, GGA as follows:
1. self-consistent step with ISMEAR=1, and,
2. non self-consistent step with ISMEAR=1.
The two respective INCAR files are shown below. My problem is that the band structure I get with these settings is wrong compared to other published band structures eg. my valence band minimum is located at the X point instead of being at the Gamma point (which happens to be 1.5 eV higher in my case). Does anyone know why this happens? I also tried to use ISMEAR = -5 with IBRION=-1 and what I got was not recognizable in any way as a Ce band structure!

INCAR 1
ISMEAR=1
SIGMA=0.1

ENCUT=400
PREC=High
NELMIN=8
ISPIN=2

LWAVE=T
LCHARG=T

IBRION=2
NSW=0
EDIFF=1E-5
EDIFFG=-0.001

INCAR 2
ISMEAR=1
SIGMA=0.1

ENCUT=400
PREC=High
NELMIN=8
ISPIN=2

LWAVE=T
LCHARG=T

IBRION=2
NSW=0
EDIFF=1E-5
EDIFFG=-0.001

ICHARG=11

Are you using the right Potential (Ce_h)?
ENCUT 400 is not enough in this case, also EDIFF and EDIFFG are set too low. Cerium is an extremely difficult system to treat (1f electron) ...

i think you need to add U.

elelel does obviously not perform an LDA+U calculation.

you should check that you have input the special kpoints in the correct format (cartesian or fractional) for that crystal structure type!

Hi guys,
Thank you all for your suggestions. My special kpoints are all specified correctly. Yes, my accuracy settings were too stringent and I have adjusted them appropriately but I still dont get the correct bandstructure. LDA+U appears to be the way to go I think but getting the settings right is a bit of a challenge. What settings have others successfully used for the LDAU+U part of the INCAR file?
<span class='smallblacktext'>[ Edited Wed Sep 13 2006, 08:01AM ]</span>

Hi elelel

#LDA plus U switched on
LDAU = .TRUE.
LDAUTYPE = 2
LDAUL = 3 -1
LDAUU = 5.0 0.0
LDAUJ = 0.0 0.0
LMAXMIX=6

should do it (similar to settings in published papers) assuming you have your atoms ordered Ce then O in POS(T)CAR...whether you really want to use LDA+U is a whole other ball game;)

i solve the ce2o3 problem, but when i cal ceo2 with two vacancy , one week, the job can not converge.
the incar is as follow:
general:
System = CeO2
ISTART=0
ISPIN=2
GGA=91
ENCUT = 500 eV
PREC=high
ISIF=2
IBRION=1
NFREE=10
POTIM=0.2
ALGO=Fast
EDIFFG=-0.01
ISMEAR =0; SIGMA = 0.20 #-5 for accurate total energy and DOS 0 for large cell 1 for metal
NSW = 100
LREAL=A
AMIX = 0.2
BMIX = 0.0001
LMAXMIX = 6
LWAVE =.True.
VOSKOWN =1 # 0 for PBE 1 for PW91
NELM=200
NSIM=4
LDAU=.TRUE.
LDAUTYPE=2
LDAUL= 3 -1
LDAUU= 5 0
LDAUJ=0.5 0
any suggestion for me. thanks

Hi Christine,

Have you tried ISIF=3 or 4? Will there be anything wrong with that? Thats what I have used in my calculations but I havent yet had a chance to test the settings Matt suggested for LDA+U - I can already see something I wasnt doing correctly though.

thanks, i dont think ISIF should be 3 or 4. the job as i mentioned has been converged now.
but i have a question. if i only consider energy of the reaction, such as O vacancy formation, are the results with U and without U similar? i know DFT cant describ the band gap correctly. but for me, band gap isnot important. the energy is important. any suggestion?

Dead right! Its CeO2 you are doing. I was thinking Ce2O3. No idea about other question, sorry.

The issue on the energy is probably not easy to solve (at least with VASP). See J. Phys. Chem. B 109 (2005) 22860. Fig. 1 in it, and all the discussion related to it, is interesting.

thanks. i have seen this paper before. but the code this group used is not vasp.

Check the literature on Ceria in PRB, there were just recently some interesting papers about it, also in J. of Chem. Phys.

In VASP, is this possible to specify Hubbard U for both the f and d orbitals of the same atom simultaneously ?