[Pw_forum] Pseudopotential Format

[Paolo Giannozzi]

On Thursday 27 October 2005 14:27, Lilia Boeri wrote:

> I have been trying to convert a Li pseudopotential downloaded from
> the ABINIT database (LDA TM pseudopotential generated with the 
> FHI98 code) to PWSCF UPF format; I thought I had succeeded in the
> conversion because the total energies and other properties (bulk 
> modulus, phonon frequencies etc) were very similar in the two 
> programs (the difference in total energies is of the order of 1mRy).
> The problem is that when I looked at the scf potential in real space 
> given by the two programs the shape was not the same... The 
> difference is really in the ionic part, which means that the 
> pseudopotential is not the same in the two calculations. 

I already heard complaints that the conversion to UPF of FHI98
pseudopotentials occasionally doesn't work, in particular when the
"nonlinear core correction" is present. In order to solve this problem, 
one needs to know what is inside the FHI98 files (what is inside the 
UPF format is documented in file upftools/UPF) and to modify the 
converter accordingly. Since the reward-to-effort ratio of such 
operation is unfavourable, I am afraid that this is not going 
to happen anytime soon.

Before claiming that the PP is not correctly converted, however, 
you should try to verify what you get in an atomic calculation:
the atomic code should work with UPF files, if the radial grid 
does not start from 0. If you can well reproduce one-electron levels
and total energy differences on different atomic configurations,
your PP does the job. The difference you observe may be due
to something else (for instance, you are not really looking at the 
same quantity)

Paolo
-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy