The ABINIT Users Mailing List ( CLOSED )

[Gian-Marco Rignanese <rignanese@pcpm.ucl.ac.be>]

Dear all,

In answer to the message of Vadim, I would like to make a few comments 
about the GW part of ABINIT starting with a few words about history.

CHI and SIG appeared for the first time in version 3.2 (see 
Release_notes point B1). The release was far from complete but it was 
definitely a good start. Since then, quite a lot of work has been put 
into complying with ABINIT rules of coding, completing the distribution, 
and improving the userfriendliness.

There have been significant changes between versions 3.3 and 3.4. First, 
Valerio Olevano has worked a lot on the '_KSS' file, which now includes 
informations on the non-local part of pseudopotential. These are 
required for a correct treatment of the case q=0 in the CHI calculation. 
This feature was not included in version 3.3. The '_KSS' file has now 
become the default, it replaces the '_STA' file of version 3.3 (this 
format still exists for those who need it). Second, CHI and SIG have 
been fully incorporated in ABINIT. The input files are now ABINIT-like 
(you can even use "ndtset"): CHI calculations are done by setting 
"optdriver 3", while SIG calculations correspond to "optdriver 4". In 
version 3.4.2, a preliminary manual has also been included, it can be 
found in the 'Infos' directory (gwmanual.txt). Examples can be found in 
Test_v3: for version 3.3, see t30.in, t31.in, and t32.in; for version 
3.4, see t30.in and t31.in. Kiril Tsemekhman also told me that he was 
putting together all the notes he had taken while learning how to use 
the codes. I hope that we will soon have a tutorial as well.

There are several important limitations that I would like to point out:
1) the GW code is not parallel (James Raynolds has proposed to work on 
this issue)
2) the GW code only works (for the moment) with pseudopotentials with 
one projector per angular momentum (Valerio Olevano will work on this 
point)
3) the GW code only works with the Perdew-Zunger exchange-correlation 
functional (I hope to fix this as soon as possible)
4) the GW code only works with "nsppol 1"
5) the GW code included in version 3.4 has not been fully tested 
(because of hollydays) and I am presently working on a possible bug...

As far as results are concerned, the original version of the GW codes 
(prior to its incorporation into ABINIT) has lead to quite lot of 
publications (V. Olevano, G. Onida, L. Reining, M. Torrent). But I am 
not aware of any work done with the version included in ABINIT.

Best regards,

Gian-Marco.

[Valerio, please correct me if I forgot to say something]