[Pw_forum] CuO, LDA+U, Pseudopots, etc.

[Matteo Cococcioni]

Dear Paul,

I have no experience with CuO and can't give you a very useful  
advise. I suspect that even if you change occ_loc = 9 for Cu the
code will converge to the same ground state. This is however quick to  
try and you may find better results (sometimes the code gets trapped in
local minima with a finite U).

On Dec 13, 2007, at 6:26 AM, Henning Glawe wrote:

> On Wed, Dec 12, 2007 at 11:40:56PM -0800, Paul M. Grant wrote:
>> Simplistically, the initial d-state configuration of Cu2+ in CuO  
>> by Hund’s
>> Rule should be 3d9 (one empty minority spin state state…a hole).    
>> But, with
>> nspin=2, lda_plus_u set true, and U=5 eV (arbitrary), the initial  
>> minority
>> spin occupation gets set to unity, not 4/5 on average (3/5 for Ni  
>> and 1/5
>> for Fe).  Is this due to occ_loc being set to 10 in tabd.f90 and  
>> not 9 (it’s
>> 6 for Fe and 8 for Ni)?  The occupation distribution converges to  
>> around
>
> as the d shell can be filled with 10 electrons, getting a initial  
> hole into the
> d shell while occ_loc=10 is IMHO impossible.
>
>
>> functionals have equivalent occupations and give similar  
>> results).  When I
>> try Cu.pz-d-rrkjus.UPF, with 3d10, 4s1 and 4p-1(!!!), the first scf
>> iteration yields individual spin occupations of nearly 5(!!!), and  
>> the
>> situation gets really ugly thereafter.
>
> I also encountered this problem with this pseudo potential; the  
> problem lies
> in the normalization of the 3d wave functions saved in the pp file.  
> if you
> search in the archives of the pw forum, you will find a patch for  
> pwscf
> allowing to recalculate the norm of the atomic states after the pp  
> file has
> been read.
>


the patch has been implemented in the cvs version of the code. if you  
download it and read INPUT_PW
you find the following lines (you are interested in norm-atomic):


U_projection_type   CHARACTER   (default='atomic')
                Only active when lda_plus_U is .true., specifies the  
type
                of projector on localized orbital to be used in the  
LDA+U
                scheme.
                Currently available choices:
                'atomic': use atomic wfc's (as they are) to build the  
projector
                'ortho-atomic': use Lowdin orthogonalized atomic wfc's
                'norm-atomic': Lowdin normalization of atomic wfc.  
Keep in mind:
                               atomic wfc are not orthogonalized in  
this case.
                               This is a "quick and dirty" trick to  
be used when
                               atomic wfc from the pseudopotential  
are not
                               normalized (and thus produce  
occupation whose
                               value exceeds unity). If  
orthogonalized wfc are
                               not needed always try 'atomic' first.
                'file': use the information from file "prefix".atwfc  
that must
                        have been generated previously, for instance  
by pmw.x
                        (see PP/poormanwannier.f90 for details)
                NB: forces and stress currently implemented only for the
                    'atomic' choice.


Hope this helps.

Matteo


> -- 
> c u
> henning
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20071213/f36ec999/attachment.html>