[Wien] Properties of corresponding atoms in afm calculations

[Ulrich Keßler]

Thank you for the information on the forces. It seems that there are 
problems with the FCOxxx (see below).

Peter Blaha wrote:

>The forces are calculated from 4 contributions.
>
>:FHFxxx (coulomb forces) are calculated in lapw0 for each atom
>individually. I guess here is the source of your (very small) differences.
>There is NO symmetry constraint on these numbers and my guess is, that you 
>did not start with an exactly antiferromagnetic startup ? (Setting all 
>moments to ZERO in case.inst, except for the AFM atoms)?
>How did FHFxxx look during scf-cycles ?
>
>  
>
:FHFxxx are the same for corresponding atoms in the first execution of 
lapw0 and stay the same for corresponding atoms in the course of the 
scf-cycles.

>:FCOxxx  Core forces: Also these numbers are calculated without
>symmetry constraint and it might be that the potential inside the
>atoms is not 100% identical.
>
>  
>
:FCOxxx are exactly the same in the first scf-cycle except for one pair 
of atoms (17.875 vs. 17.877). I have no idea why these values differ (I 
checked the struct-file again). In the course of the scf-cycles the 
:FCOxxx diverge for all corresponding atoms, e. g. 71.021 vs. 71.779. It 
is remarkable that the different values change between the corresponding 
atoms from one cycle to the next!
By the way: How could the potential be different when using the symmetry 
of the supergroup in constructing the case.struct etc.?

>:FVAxxx and :FSUxxx : these are the IBS corrections calculated in lapw2
>and these numbers should have the "symmetry constraint". Usually they are
>added only in the last scf-iteration.
>
>Please check already for the first scf iteration (starting with dstart)
>if the force contributions show differences (and if they become smaller 
>during scf, because you mix in a "true" AFM density). 
>  
>
I couldn't check this because :FVAxxx and :FSUxxx are not written to the 
scf-file, and scf2-files are not saved by save_lapw. But I will run a 
test and check it.

Hope you can make something of this information.

Ulrich Keßler

>
>  
>
>>>What does "different forces" mean ? Could that be "numerical noise"
>>>(like forces of +20 and +21), or is it more like +20 and +10 ??
>>>
>>> 
>>>
>>>      
>>>
>>You may call it "numerical noise" (it's like 10.276 vs. 10.400 etc.). 
>>But in my opinion this must not happen in an afm calculation since the 
>>down electrons are identical to the up electrons (by clmcopy). Thus 
>>everything that happens on atom A must be _exactly_ the same on atom A' 
>>(maybe with different sign, but same absolute value).
>>The different forces are quite inconvenient when doing a minimization as 
>>pointed out before.
>>
>>    
>>
>>>In the "noise" case, your procedure should be fine (eventually you may need
>>>better scf-convergence ?), otherwise something more
>>>serious may happen.
>>>
>>> 
>>>
>>>      
>>>
>>I think the scf-convergence is okay (-ec 0.00001 -cc 0.00001). By the 
>>way this has no influence on the afm argument that properties of 
>>corresponding atoms should be the same.
>>    
>>
>
>See my arguments above.
>
>                                      P.Blaha
>--------------------------------------------------------------------------
>Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
>Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
>Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
>--------------------------------------------------------------------------
>_______________________________________________
>Wien mailing list
>Wien at zeus.theochem.tuwien.ac.at
>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
>  
>