The ABINIT Users Mailing List ( CLOSED )

[Ganesh Panchapakesan <gpanchap@gmail.com>]

Dear Takeshi and Bernard,

Thank you for the suggestions and references.

The fatbands seem to look like what I expected.  I wonder why the fermi level is zero when printed from this routine (even for the test NiO case)? Perhaps so as not to shift the bands by Efermi. The computed bands seem to be otherwise consistent with the L decomposed DOS.

I am computing (L,M) decomposed partial DOS using 'prtdosm 1'  after commenting out the last few lines of 'partial_dos_fractions_paw.F90' which convert the real DOS to Imaginary. The DOS for m and -m is not exactly identical as was the case before.  But I still don;t see the integrated partial DOS to be related to the occupancy from in LDA+U method.  I think I am doing the conversion correctly, but it has to be same for all M values.  Any suggestions would be helpful.

Thanks.

Ganesh

2010/1/14 Takeshi NISHIMATSU <t_nissie@yahoo.co.jp>

Ganesh,

Please see
https://listes-2.sipr.ucl.ac.be/abinit.org/arc/forum/2007-11/msg00037.html
and the source code.

Good luck,
-- Takeshi

--- Ganesh Panchapakesan <gpanchap@gmail.com> wrote:

> Dear Bernard,
>
> Thank you for the reply.  Indeed, I see that using imag. Ylm's,
> DOS(L,M)==DOS(L,-M).
>
> It is also clear that if I knew the projected 'dos' in terms of the real.
> sph. harms., then DOS(L,M)=(dos(L,M)+dos(L,-M))/2.0 , where 'dos' is for
> projections on real sph. harms.  But going the other way around, from DOS to
> dos, doesn't seem to be possible due to the identity in the first line.  It
> seems I have to explicitly compute 'dos' .  Maybe there is a relation
> between dos(L,M) and dos(L,-M) that I am missing, but I don't see it in my
> derivation.
>
> Thanks.
>
> Ganesh
>
>
>
> On Wed, Jan 13, 2010 at 9:11 AM, Bernard Amadon <bernard.amadon@cea.fr>wrote:
>
> > Dear Ganesh
> >
> > Prtdosm computes M-resolved partial dos in the complex spherical harmonics,
> > it explains why you have
> > DOS(L,M) == DOS(L,-M) (without spin-orbit). In the contrary, LDA+U
> > occupation matrix is in the real spherical harmonics basis.
> > You could use prtfatbands which computes band structure in the real
> > spherical harmonics basis, but
> > to have the partial dos, you can use the conversion from complex to real
> > spherical harmonics,
> > apply it to the dos, and you will have the partial dos in the real
> > spherical harmonics which is more
> > convenient for the interpretation.
> >
> > Best regards
> > Bernard
> >
> >
> >
> >
> >

> > Ganesh Panchapakesan a 将アcrit :

> >
> >  Dear All,
> >>
> >> I am trying to compute the partial DOS (projected on to (atom,L,M) ) of an
> >> anti-ferro. system using PAW - GGA with v5.7.4.  The system is metallic.  I
> >> use the following options: prtdos 3, prtdosm 1, ngkpt 10 10 10, shiftk 0 0
> >> 0, natsph 1, iatsph 1, ratsph 2.31, iscf -3 to compute the DOS using the
> >> tetrahedron method and project onto the (L,M) centered on atom 'i' (ratsph =
> >>  r_c of PAW dataset)
> >>
> >> Although the partials obey the condition: Sum(m)DOS(L,M) == DOS(L), the
> >> result of the partials always follows this identity: DOS(L,M) == DOS(L,-M),
> >> even when I expect the different d-orbitals (i.e. L=2) to have a different
> >> occupations (I see that from the occupation matrices in a +U calculation).
> >> As a result, the integral of the  DOS(L,M) up to the Fermi level does not
> >> correspond to the occupation number from the +U calculation.  I have tried
> >> with both the pure GGA density and GGA+U density as inputs to the DOS
> >> calculation.
> >>
> >> Any suggestions/clarifications will be very much appreciated. If anyone
> >> knows that this is a problem but it is fixed in v5.8.4 that would also help.
> >>
> >> Thanks.
> >>
> >> Ganesh
> >>
> >
> > ---------------------------------------------------------------------
> >
> >
> >
>