Electronic structure of CeO2 greatly depends on initial guess of wave function in OT method

[stclair]

Dear CP2K Developers and Users,

I am trying to get the work function and orbital information around Fermi 
Level for CeO2 (111) surface (with one oxygen vacancy). 
In my case, DFT+U method is used. When I am testing the U effect, I realize 
that the electronic structure obtained greatly depends on the initial guess 
I used.
For example, if I set 'SCF_GUESS' as 'atomic', I got work function ~5.1 eV 
and defect states ~1.1 eV above valence band. But when restarting from the 
wave function I got from the calculation from higher U value, the work 
function ~4.74eV and defect states ~1.56 eV above valence band. Also, the 
total energy is ~0.89 eV lower than the previous one.
Does that indicate the OT method is not very suitable here?

Another problem is to use 'relax_multiplicity'. For CeO2 with one oxygen 
vacancy, the triplet is more stable. However, I don't want to fix the 
multiplicity in my calculation. So I am trying to use 'relax_multiplicity' 
to obtain correct state. However, despite of how large the value of 
'relax_multiplicity' I used, it still only converges to singlet state if I 
didn't specify it as triplet. Is there any suggestion? Or I have to switch 
to diagonalization method with smearing? Or there is other parameter I can 
use?


Thanks great in advance!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150410/865b611b/attachment.htm>