The ABINIT Users Mailing List ( CLOSED )

[m.foulkes@imperial.ac.uk]

As a test, one of my research students took a self-consistent Si potential 
from Abinit and fed it into a program of his own, which carries out a single 
non-selfconsistent solution of the Kohn-Sham Hamiltonian to find the electron 
density. When he disabled the non-locality of the pseudopotential (in Abinit 
and his own program) his density agreed exactly with the density calculated 
by Abinit; but when he used the full non-local pseudopotential in both 
programs there were significant differences.

We cannot explain this discrepancy, especially since the non-local 
pseudopotential form factors we use were written out from inside Abinit 
itself. Our best guess is that Abinit's way of implementing non-local 
(Kleinman-Bylander) pseudopotentials is not as we had been assuming. We have 
already looked at (parts of) the source code and scanned the literature, but 
remain unsure.

Are there any papers or other documentation in which the details of Abinit's 
implementation of KB pseudopotentials are explained? If not, does Abinit do 
anything unusual that might have caught us out?