The ABINIT Users Mailing List ( CLOSED )

["Matteo Giantomassi" <Matteo.Giantomassi@uclouvain.be>]

> Dear Colleagues,
> I have two questions about GW approximation and I would be grateful for
> any
> help.
>
> 1. I would like to calculate GW energies of a SiH4 molecule in a large box
> with
> 2 additional electrons. To do that I try to put the variable charge=-2,
> but
> Abinit stops in Screening calculation with message:
>
> fermi : input and calculated number of electrons differ by  -2.00000000
>
> fermi : ERROR -
> Found too large difference in the number of electrons
>
> It seems that the program “fermi” does not accept additional
> electrons. Is
> there some way to put additional electrons in molecules or to push up the
> Fermi
> level in clusters?
>

 Dear Andrey

 The code stops in fermi.F90 simply because the total number of
 valence electrons inferred from the pseudopotentials is different
 from the number of electrons calculated from the occupation numbers.
 The latter quantity, indeed, contains the additional charge added during
 the generation of the KSS file.

 fermi.F90 should be able to deal with charged systems.
 You only have to skip the check on the number of electrons;
 for example you might comment the call to "leave_new" at line 116 of
src/15gw/fermi.F90.

>
> 2. I try also to impose a particular distribution of electrons in a spin
> polarized GW calculation of SiH4 molecule. For example instead of
> standard:
> 1 1 1 1 0 0 0 0
> 1 1 1 1 0 0 0 0
>
> I want to take:
> 1 1 1 0 1 0 0 0
> 1 1 1 0 1 0 0 0
>
> but I cannot impose this occupations in KKS stage of GW calculation,
> Abinit
> writes out electronic occupations in KKS dataset:
> 0 0 0 0 0 0 0 0
> 0 0 0 0 0 0 0 0

 You didn't attach the input file, but I guess you have
 generated the KSS file by performing a standard band structure
 calculation by reading a previous density file and using iscf -2

 In this particular case all the occupation numbers reported in the
 KSS file are set to zero, and the GW code recalculates the occupancies
 starting from the knowledge of the eigenvalues and taking into account
 if the system is a metal or a semiconductor.
 In your case, unfortunately, this procedure won't work as you
 want to enforce a particular electronic distribution by promoting some
electrons.

 To solve this problem you might generate a KSS file by performing
 a SCF calculation with occopt=2 and the conjugate gradient method
(kssform=3).
 In this case, indeed, the KSS file contains the occupation numbers
 defined by the input variable occ.
 Keep in mind, however, that, in case of kssform=3, all the bands you
 want to store in the KSS file have to be explicitely considered during
 the SCF calculation (i.e nband >= nbandkss)

 I have to say that I've never considered the possiblity of running GW
 calculations for systems with an arbitrary distribution of electrons.
 In particular the band parallelism (gwpara==2), as presently implemented,
 assumes that the electronic states are occupied according to their
 energy. Thus gwpara=2 wont't work properly in case of systems with
 promoted electrons.
 To be on the safe side I suggest using gwpara=1 (i.e parallelism
 over k-points).

 Hope this helps.
 Best regards,
 Matteo Giantomassi

>
> Is it possible to impose a particular electron distribution in GW
> calculation?
> Thank you,
> Andrey TITOV.
>
>
> Institut Matériaux Microélectronique et Nanosciences de Provence
> http://www.im2np.fr/doct/titov.html
>