The ABINIT Users Mailing List ( CLOSED )

[bernard amadon <bernard.amadon@cea.fr>]

I agree that the effect of Hubbard correction on Ti 3d is the most 
important effect for the gap. However and as you mentionned,
for empty states, La 4f states play a role: In the paper mentionned by 
Joe Zwanziger, the Hubbard correction is applied to
La 4f states in order to change the band ordering of empty states (empty 
eg states with respect to empty La 4f).

Best,
Bernard Amadon


Vincenzo Fiorentini a écrit :
> PS - Of course neglecting La 4f is not a good idea in general as the 
> 4f and higher empty Ti d-like will be at similar energies, but still 
> should not affect too much the Mott gap.
> VF
>
> On 14 Feb 2008, at 8:43 AM, Vincenzo Fiorentini wrote:
>
> > I suppose that in LaTiO3 (as in other lanthanide oxides) one can get 
> > away without the unoccupied La 4f. Since the main playground of U is 
> > on Ti d (the Mott gap is due to Ti being 3+ in lantanide titanates), 
> > one may simply not apply it to La f's.
> >
> > Vincenzo Fiorentini
> >
> >
> > On 13 Feb 2008, at 5:32 PM, bernard amadon wrote:
> >
> > > Dear Joe,
> > >
> > > In fact, Abinit should give an error message instead of a crash. The 
> > > problem is only that LDA+U in PAW need projectors
> > > for the angular momentum on which the Hubbard correction is applied. 
> > > In other words, it it necessary physically to define the atomic
> > > wavefunction used to compute the occupations and the Hubbard 
> > > correction.
> > > I am thus surprised by the fact that the authors  are able to
> > > use LDA+U on 4f states of La without including 4f states in the 
> > > atomic data...
> > > You could construct new atomic data with 4f states in it to see if 
> > > you recover their results.
> > >
> > > Best
> > > Bernard
> > >
> > > jzwanzig@dal.ca a écrit :
> > > > Dear Colleagues,
> > > > I have been running calculations using PAW together with LDA+U with 
> > > > some success, on various transition metal oxides. Recently I've 
> > > > looked at LaTiO3, and in a paper by Okatov, Poteryaev, and 
> > > > Lichtenstein, Europhys. Lett 70, 499 (2005), this compound is 
> > > > treated using GGA+U on both the Ti 3d and La 4f states. What 
> > > > puzzles me is that their PAW data set as described in the paper, 
> > > > doesn't have 4f states in it. In abinit, when I use a La PAW set 
> > > > based on 5s and 6s, 5p and 6p, and 5d, similar to what they 
> > > > describe, abinit crashes if I turn on lpawu for the La 4f 
> > > > electrons. I assume that that happens because I don't have 4f 
> > > > states in the PAW set to start with. So either I am doing something 
> > > > completely wrong with abinit, or the code the other authors use 
> > > > works differently that abinit does. Can someone clarify this for me?
> > > >
> > > > thanks,
> > > > Joe