[Pw_forum] about U value determination in DFT+U calculation

[Gabriele Sclauzero]

Dear Jia Chen,

  On 07/16/2010 06:34 PM, jia chen wrote:
> Dear All,
>
> I am doing a calculation to determine U value for DFT+U calculation
> following method of Prof Cococcioni by pw 4.2. I did the similar work
> several months before and I always could get making sense values, so I
> think I can handle this method. But this time, results I obtained are
> very wired. Because I am using semi-core pseudopotential which
> contains both 3d and 4d atomic orbitals, I wonder if this can cause a
> problem. When projecting KS states to atomic orbitals in order to
> calculate occupation number, if it is possible that pw code could mess
> up 3d and 4d orbitals? In my case, 3d is occupied and 4d should be
> empty.
If the 4d orbital has the occupation set to zero in the pseudopotential 
(look inside the header of the PP file), then the code should select the 
3d orbitals to build the local occupations needed for DFT+U. Please have 
a look at subroutine PW/offset_atom_wfc.f90 and eventually add some 
debugging printout instruction to be sure that the correct wavefunctions 
are selected (compare with the header of the PP file, the atomic 
wavefunctions should be processed in the same order as listed therein).
If you find still problems, please add more information to your next 
post, otherwise it will be hard to help you further. For instance, if 
you compute the isolated atom with pw.x, how much are the occupations 
written on output when lda_plus_U=.true.? Do they coincide with the 
occupation of the 3d orbital as written in the PP?

Regards,

GS


> If pw code can pick up 3d orbitals without any problem, I made
> some mistakes and apologize here. Any suggestion is greatly
> appreciate!
>
> Best Wishes
>    


-- 

Gabriele Sclauzero, EPFL SB ITP CSEA
PH H2 462, Station 3, CH-1015 Lausanne