[Pw_forum] LDA or LSDA?

[Stefano de Gironcoli]

dear Maxim Ivanov,

    the LDA+U calculation is an LDSA+U calculation with the added 
constraint of having equal up- and down-spin levels.
  As such, I would say that the LSDA+U, providing additional variational 
freedom, should result in lower energy and be a better approximation to 
the real system. However L(S)DA+U may be tricky and your calculation 
might remain trapped in
local minima more easily than in ordinary LDA calculation. a few checks 
may be useful.
  is the energy of your LSDA+U calc lower than the one of the LDA+U one, 
as it should ?
  are you using the same U parameter in the two cases ? have you tried 
different guesses for the starting occupations.
  on which levels are you applyaing U ? the "defect w U and spin 
polarized" DOS is remarkably similar to the "perfect" case...
  are you sure you are really applying a +U correction in that case ?
  what would be the result of a plain LSDA (no +U) calculation  ? this, 
rather than LDA, should be your reference defective calculation.
 best,
   stefano

Максим Иванов wrote:
> Dear all,
>
> I have made
> TDOS calculations for rutile-TiO2 with oxygen vacancy in the center of the
> supercell with 96 atoms. I have used LSDA+U and LDA+U approaches. Non spin-polarized
> calculations showed enlargement of the energy gap with slight shift of defect
> level. Spin-polarized calculations didn’t give changes in the value of Eg,
> however defect level appreciably went down. Experimental value of the gap is 3
> eV, defect level is 2 eV above valence band top.
>
> So, I have
> a question: which of these two approaches give me true result?
>
>  
>
> Thanks in
> advance,
>
> Maxim
>
>
>       ________________________________________________________
> Вы уже с Yahoo!? 
> Испытайте обновленную и улучшенную. Yahoo! Почту! http://ru.mail.yahoo.com
>   
>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>