The ABINIT Users Mailing List ( CLOSED )

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

Dear Tyler,

I do not know what you intend by "low level implementation". The GW 
part of ABINIT originates precisely from Rex Godby's code. It has been 
released under GPL with his agreement. It is a complete implementation 
of the GWA. So, it should definitely work for bulk semiconductors as 
well as for their surfaces. In principle, it should also work for 
metals though this should be tested carefully. There is very little 
missing compared to other GW existing codes (e.g. non-symmorphic 
symmetries are not taken into account, limitation to LDA).

It is true that GW calculations are indeed very expensive. In 
particular, the GW part of ABINIT is not yet parallelized and lacks of 
optimization. This is maybe one of the major limitations.

I have not checked the manual to see what it states precisely. But, 
what is meant by "ionic states" is that in the GW method one computes 
the quasiparticles energies of states with N+1 or N-1 electrons (by 
contrast to the Bethe-Salpether method in which the number of electrons 
N is constant). So, within GW, there is no electron-hole interaction... 
That is intrisic to the method and it is not related to the 
implementation used in ABINIT.

Best regards,

Gian-Marco.