[Pw_forum] U calculation in LDA+U

Burak Himmetoglu himm0013 at umn.edu
Tue Jul 10 17:11:45 CEST 2012


Hi Peng,

In my experience, the process of shifting the O atom to the origin is
recommended since it produces more stable results numerically, especially
in the second part of r.x where the program extrapolates to larger
supercells using the results from the starting supercell you have used.
However, if the crystal has already very  small number of symmetries, then
you should be able to avoid this shifting.

The first three entries in filepos are the lattice vectors in units of alat
(like the CELL_PARAMETERS when ibrav=0). The unit cell is expanded by a
factor of 10 in the tutorial. The length scale in filepos is not necessary.
That file is used only to compare the distances between pairs of atoms and
the scale does not really matter. It is generally better to scale up only
for numerical stability reasons.

Best regards,

Burak

On Tue, Jul 10, 2012 at 10:03 AM, Peng Chen <pchen at ion.chem.utk.edu> wrote:

> When O is perturbed in the case of NiO in tutorial, it
> is re-positioned and there are 2 different kinds of O (first with 1 atom,
> the second with 7 atoms).  Then the kind with only one O is perturbed. In
> the tutorial, it reads "This is done to preserve the largest possible
> number of symmetries".  Is this process necessary? Can we just run
> computation without re-position? In that case, there is only one kind of O
> with 8 atoms.
>
> In the position file "filepos", the scale of cell length is always 10,
> even if the scale is 7.9 in NiO and 5.42 in Fe. Shall we always use 10 in
> position file no matter what the real scale is?
>
> On Tue, Jul 10, 2012 at 9:37 AM, Burak Himmetoglu <himm0013 at umn.edu>wrote:
>
>> Hi Peng,
>>
>> r.x assigns the values of the response matrices only by considering the
>> distance between atoms and whether they have the same spin and type or not.
>> This information has to be given by the user in the positions file
>> "filepos" read by &input_mat. If B(1) and B(2) are different types
>> (e.g. crystallographically) you should perturb them separately, as you
>> outlined in the procedure, and make sure that ntyp=3 (type 1 is B(1), type
>> 2 is B(2) and type 3 is O) in the namelist &input_mat that goes into r.x
>>
>> Best regards,
>>
>> Burak
>>
>>
>> On Tue, Jul 10, 2012 at 1:45 AM, Stefano de Gironcoli <degironc at sissa.it>wrote:
>>
>>>  I don't know the details of r.x code but I think that, physically,
>>> what matters is the two atoms are crystallographically equivalent or not
>>> .therefore I think space group symmetry is the relevant one.
>>>
>>> stefano
>>>
>>>
>>> On 07/10/2012 05:54 AM, Peng Chen wrote:
>>>
>>> Dear All,
>>>
>>> In the calculation of U, how can we determine the equivalence of the ions,
>>> by site symmetry or space group symmetry?
>>> E.g  ABO3, B ions has 2 different site symmetries (B(1) B(2)). Shall we run
>>> calculation on perturbing B(1) and B(2) ions or  just a B ion?
>>>
>>>  In order to find U in B  ions, I tried to run calculation with perturbing
>>> just on B, and r.x gave some errors that it needs to read more data.
>>> So I guess it needs the perturbed data from other atoms. Shall I use the
>>> procedures listed below?
>>> 1. run scf calculation with Hubbard_U(i)= 1.d-20
>>> 2. run scf with perturbing on an A ion
>>> 3.   run scf with perturbing on a B(1) ion
>>> 4.   run scf with perturbing on a B(2) ion
>>> 5.   run scf with perturbing on an O ion
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Pw_forum mailing listPw_forum at pwscf.orghttp://www.democritos.it/mailman/listinfo/pw_forum
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Pw_forum mailing list
>>> Pw_forum at pwscf.org
>>> http://www.democritos.it/mailman/listinfo/pw_forum
>>>
>>>
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum
>>
>>
>
>
> --
>   Best Regards.
>         Peng
>
> _______________________________________________
> 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://www.democritos.it/pipermail/pw_forum/attachments/20120710/947ce6f7/attachment.htm 


More information about the Pw_forum mailing list