The ABINIT Users Mailing List ( CLOSED )

[Valerio Olevano <valerio.olevano@grenoble.cnrs.fr>]

Dear Manuel,

I think you don't mind if I answer
the whole ABINIT forum since it is a question
potentially of interest for other people.

On Wed, 2006-02-15 at 15:38, mperez@mpi-halle.mpg.de wrote:
> Dear Prof Olevano
> 
> this is an Abinit question (sorry, i had technical problems with the
> Abinit forum submissions), i hope you do not mind my asking directly.
> 
> In your GW implementation in ABINIT, the routine screening.F90 uses for
> the RPA dielectric matrix (DM), a symmetric DM, defined in some notes that
> you provide in Abinit. According to the literature (Aulbur et al, von der
> Linden and Horsch) the symmetric DM is defined as a Hermitian DM, which is
> required in order to make DM causal.
> 
> My question is this:
> 
> consistently with the GW implementation, there is an output file in
> screening.F90 which provides -(Im(epsilon^-1)), which apparently can
> be used as a check up, since this is EELS, and can be compared to
> experiment. EELS needs causal response!
> 
> BUT, CAN WE USE EPSILON (NOT EPSILON^-1) AS CALCULATED IN SCREENING.F90
> FOR COMPARISON TO PHOTOABSORPTION (Im(epsilon))?
> I UNDERSTAND PHOTOABSORPTION REQUIRES THE RETARDED DENSITY-DENSITY
> RESPONSE FUNCTION (AND EPSILON), NOT THE CAUSAL ONE, AND screening.F90
> ONLY PROVIDES THE CAUSAL RESPONSE (AND DM)
> 
I am sorry. But aren't the retarded correlation functions the same
as the causal ones? I knew that the term causal was opposed
to time-ordered/ antitime-ordered correlation functions to indicate
both the retarded and advanced correlation functions. 
I think it is a question of terminology but normally
causal is used as synonymous of the physical retarded corr. funct.
Or am I wrong?
hence you can use it to compare with photoabsorption.
though that GW code is not the ideal to do it since it is
interested to the whole microscopic dielectric matrix 
entering into the Self-Energy,
at omega=0 or few others, imaginary even.
and not to the macroscopic omega-dependent dielectric function.
There is another code, the TDDFT DP code 
http://theory.polytechnique.fr/codes/dp/
to do exactly that
job, and we plan to distribute it open source, when we recover
our forces....
But to calculate optical properties RPA or TDLDA is not enough
at all, since excitonic effects, important in photoabsorption,
are completely neglected. you need more elaborate xc kernels
or use a MBPT method like the solution of the Bethe-Salpeter
equation.
And there we can give you a good news:
we succeeded to create all the stuff for an open
source project for the Bethe-Salpeter EXC code that now
is available at
http://www.bethe-salpeter.org/
it is fully and preferencially interfaced with ABINIT,
though we hope in the future that also people using other codes 
can use it, to not create favoritisms.

I hope it helps

regards

Valerio Olevano

> thanks in advance
> 
> Manuel Perez Jigato, MPI fuer Mikrostrukturphysik, Halle, Germany
>