The ABINIT Users Mailing List ( CLOSED )

[Matteo Giantomassi <gmatteo@pcpm.ucl.ac.be>]

Dear Hamann, 

I think that the different results for the dielectric constant are
mainly due to an inconsistent treatment of the small imaginary part 
 i\delta used to avoid the poles in the time ordered \Chi_o (\omega). 

In case of a sequential run with abinis or abinip (on 1 proc), the
imaginary shift \delta is not used if \omega is purely imaginary or
zero, while \delta is always included in the denominator in case of
parallelization over bands.

This should explain why the results of the automatic test are so
different. 

This inconsistency has been fixed in my branch 5.4, and I will commit 
the modification in version 5.3.4

These are the results I obtained runnig test tN just 2 minutes ago
(log0 for abinis, log1 and log2 for abinip with 1 or 2 processors):
 
grep diele ../TN/log*

../TN/log0: dielectric constant =  27.4114

../TN/log1: dielectric constant =  27.4114

../TN/log2: dielectric constant =  27.4054

Now the value of the dielectric constant reported in log2 is close to
the other two values, although a (smaller) difference still remains.

Indeed I looked at the matrix elements of \chi_0 reported in log2 and
log1, and I found that they are not perfectly equivalent. 

I have to study better the implementation, but these differences might
be related to a different expression for the frequency dependent part of
\Chi_o which is presently used in case of gwpara==2.

Best regards
Matteo  

On Fri, 2007-03-30 at 11:42 -0400, D. R. Hamann wrote:
> Perhaps whoever can answer this can also address the question of why in 
> 5.3.3/tests/paral/Refs,
> 
> grep 'dielectric constant =' tN*.out gives:
> 
> tN0.out: dielectric constant =  27.4114
> tN10.out: dielectric constant =  27.3806
> tN1.out: dielectric constant =  27.4114
> tN2.out: dielectric constant =  27.3804
> tN4.out: dielectric constant =  27.3805
> 
> is considered OK?  These are band-parallelism GW tests, and it looks to 
> me as if serial (tN0 and tN1) and parallel tests give unacceptably 
> different answers.
> 
> Don Hamann
> 
> Philippe Blaise wrote:
> > Please, someone could explain more in details what is written in the 
> > release notes for the 5.3 ?
> >
> > B.6 *Parallelism over bands is available for GW*. See the input 
> > variable gwpara.
> >    This parallelism allows for memory savings, and is quite efficient.
> >    At present, either the band parallelism or the k point parallelism 
> > is available
> >    in GW, they cannot be used simultaneously.
> >    This has been implemented by R. Shaltaf.
> >    (Note, however, that parallelism over bands in GW is not compatible 
> > with nsppol=2 as of v5.3.0)
> >    Automatic test for band parallelism : paral#N
> >
> > and the corresponding action of the parameter gwpara=2 : 
> > parallelisation on bands.
> >
> > I'm not sure to understand : "bands" stand for matrix elements ? 
> > wavefunctions ?
> >
> > Thanks a lot,
> >
> > Philippe
> >
> >
>