The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <xavier.gonze@uclouvain.be>]

Dear Soraya and Pierre-Matthieu

Well, for non-magnetic insulators, strictly speaking,
nband= ne/2
will work already. And there are many examples of this in the
input files provided with ABINIT.
The goal of adding a few unoccupied bands is only to monitor whether
the system is actually metallic or insulating ! And sometimes,
by adding some unoccupied bands you can get a faster convergence,
but this is not always the case.

For metallic systems, nband= ne/2 will definitely not be enough.

Soraya : in general, there is no need to define nband by hand,
there are default values in ABINIT that usually work quite well,
except perhaps for some response function calculations (phonons).

Xavier


On 06 Mar 2009, at 18:06, Anglade Pierre-Matthieu wrote:

> On Fri, Mar 6, 2009 at 5:49 PM, Souraya Goumri-Said
> <Souraya.Goumri-Said@fundp.ac.be> wrote:
> > Dear PMA,
> >
> > Thank's very much for you detailed explanation
> >
> > I still have some confusion <:-| :
> >
> > Suppose we have a system of 8 atoms (A, B and C) , and we have 4 A  
> > atoms, 3
> > B atoms and 1 C atom
> >
> > following your explanation: the valence electron : ne = 4 x ZION(A) 
> > + 3 x
> > ZION(B)+ ZION(C)  ( ZION extracted from PSPNC files) ( is it  
> > correct ?)
> >
> > which expression is correct : nband = 1/2 [  ne+2 ] or  nband= ne/2  
> > +2
>
> nband= ne/2 +2 (at least)
>
>
> > occopt=1 is it the best choice or can be equal =< 3 for SC
>
> have a look at :
> http://www.abinit.org/Infos_v5.5/input_variables/varbas.html#occopt
> You will see that using occopt = {1, 3, 4, 5, 6, 7}
> is easy while other options are used in very special cases.
> For a semiconductor it is natural to use occopt 1 yet any of the
> choice above will work as well.
>
>
>
> > Best regards,
> >
> > Souraya,
> >
> >
> >
> >
> >
> > At 09:18 6/03/2009, you wrote:
> >
> > For basic ground state calculation in semiconductors, nband is to be
> > chosen has, at least, half the number of electrons (ne)  within the
> > cell + 2. The number of electrons within the cell is the sum of all
> > the valence electron of the atoms in the cell. The number of valence
> > electrons of an atom is written in the pseudopotential file you use
> > for this species. If you have a look at Abinit pseudopotentials, they
> > contain a variable zion. This is the number of valence electron of  
> > the
> > atom.
> >
> > A DFT calculation consist in writting an hamiltonian (made from your
> > atomic positions, pseudopotential, cell size, cutoff energy, exchange
> > correlation energy) and getting the eigenvalue (energy) associated
> > with each band (at convergence the bands are eigenvectors of your
> > hamiltonian). When solving the problem one seeks only the lowest
> > eigenvalues and associated eigenvectors since electrons are going to
> > occupate only the lowest energy levels. For semiconductors, the
> > occupied bands (vlaence) are separated from the unoccupied one
> > (conduction) by a gap. Therefore, all the valence bands are totally
> > occupied and all the conduction bands are totally empty ; then the
> > laters make no contribution to the total energy and it is not usueful
> > from a physical point of view to calculate them. Nevertheless
> > including a few conduction bands helps solving the hamiltonian and
> > that's why I recommended you to use at least nband = ne/2 +2 for a
> > semiconductor. That's a minimum. For safety, if you don't understand
> > very well, you'd better use nband = ne/1.5.
> >
> > In metals things are a little more complicated. Since conduction  
> > bands
> > and valence bands are very close, temperature can spread the  
> > electrons
> > over the fermi level. Moreover getting an accurate evaluation of the
> > total energy is more difficult and can be made easier by using a high
> > temperature for the electrons. Depending on the way you want to deal
> > with those two facts you can choose with occopt a distribution for
> > your population of electrons whose parameter is the electronic
> > temperature (tsmear). Yet, if you actually deal with semiconductors
> > you can safely choose occopt 1.
> >
> > regards
> >
> > PMA
> >
> >
> >
> > On Thu, Mar 5, 2009 at 7:42 PM, lagoun brahim  
> > <lag17_brahim@yahoo.fr> wrote:
> > > HI
> > > i am not a devloper then if i am wrong someone  should correct me
> > > we need a convergence study for ecution in the case of metals, but  
> > > for
> > > insulator was not  need that,
> > > OCCUP x 2 =the number of valence electrons in your pseudopotentiel  
> > > (in the
> > > cell)
> > >
> > >
> > > --- En date de : Jeu 5.3.09, Souraya Goumri-Said
> > > <Souraya.Goumri-Said@fundp.ac.be> a écrit :
> > >
> > > De: Souraya Goumri-Said <Souraya.Goumri-Said@fundp.ac.be>
> > > Objet: [abinit-forum] nband value and occopt
> > > À: forum@abinit.org
> > > Date: Jeudi 5 Mars 2009, 17h56
> > >
> > > Dear Abinitors,
> > >
> > > I have some problem to understand the "choice" of the nband and  
> > > occopt
> > > value
> > > ? By reading the input variables for nband and occopt, I don't  
> > > know how
> > > can
> > > I attribute the values of these input essentially for an  
> > > insulator  (
> > > semiconductors) ?
> > >
> > > Can someone explain to me how to know in adavance  these values ?
> > >
> > > Regards,
> > >
> > >
> > >
> > > Cordialement,
> > > Sincerely yours,
> > >
> > > ====================================
> > > Dr. Souraya Goumri-Said
> > > Centre de recherche en Physique de la Matière et du Rayonnement  
> > > (PMR)
> > > Laboratoire de Physique du Solide
> > > Facultés Universitaires Notre-Dame de la Paix
> > > Rue de Bruxelles 61 - B - 5000 Namur
> > > Tel : +32(0)81 72 47 01
> > > Fax : +32(0)81 72 47 10
> > > http://sourayagoumri-said.weebly.com/
> > > email : Souraya.Goumri-Said@fundp.ac.be
> >
> >
> >
> > --
> > Pierre-Matthieu Anglade
> >
> > Cordialement,
> > Sincerely yours,
> >
> > ====================================
> > Dr. Souraya Goumri-Said
> > Centre de recherche en Physique de la Matière et du Rayonnement (PMR)
> > Laboratoire de Physique du Solide
> > Facultés Universitaires Notre-Dame de la Paix
> > Rue de Bruxelles 61 - B - 5000 Namur
> > Tel : +32(0)81 72 47 01
> > Fax : +32(0)81 72 47 10
> > http://sourayagoumri-said.weebly.com/
> > email : Souraya.Goumri-Said@fundp.ac.be
>
>
>
> --
> Pierre-Matthieu Anglade