[Pw_forum] ld1.x problem (not really)
calderin at physics.queensu.ca
Fri Jan 25 14:30:57 CET 2008
> You should only put the valence states in test configuration, e.g.
> configts(1) = '6s2 6p3'
Thanks in deep Andrea and Lorenzo.
I completely missed that ...
> Oh, and you have a ghost in the s component at ~ -4 rydberg and one in the
> p channel at ~ 0.8 Ry. It means that your rcloc is probably far too small.
Thanks Lorenzo, I am aware of the ghosts. That's the very first try, so
there is a lot of run for improvement.
> Even if the test works in ld1 this pseudo will never be usable in pw
Humm .. I wonder why.
Those ghosts are out of the interval of -2.5 to 0 Ry which one would
expect to be the important one for the bonding in this case. Besides the
log derivative is reproduced very well in that interval. So, all the
ground state properties calculated with that psp should be fine except,
perhaps, the band gap in systems with one. It should also work for metals
without any problems.
More information about the users