The ABINIT Users Mailing List ( CLOSED )

[TORRENT Marc <marc.torrent@cea.fr>]

Dear Matthieu,
You should have a warning for these high values of rhoij, producing 
instabilities in the famous "nonlop" routine.
Indeed it's connected to "too high projectors", but not in absolute value.
What produces high rhoij values are projectors with values too far from 
their corresponding  partial waves... in orther words, you should have 
partial waves and projectors with the same order of magnitude.
(as rhoij contains a product of partial waves and projectors, a simple 
sclaing of projectors is inneficient).

The solution is:
1- identify then angular momentum l concerned with (s, p...) by having a 
look at the rhoij array (that's why it's output).
2- change the values of the reference energies for this l (in order to 
decrease projectors, it's better to move reference energies away from 
each other; but not too much as you can degrade your log derivatives).
3- if 2- is unefficient, try to change your local potential by changing 
rcloc value or changing the pseudization scheme (iploctype).
4- if 3- isd innefficient, try to change the value of rc for this l...

good luck,

Marc

Matthieu Verstraete a écrit :
> More info!
>
> The problem appears to be linked to large values of the RHOIJ, and in 
> principle of the projectors. I have plotted out the TPROJECTOR parts 
> of the paw datafiles, and they look quite similar in Ag and Pd. 
> Maximum value ~ 200. Does anything else enter in the RHOIJ which could 
> be mis-represented?
>
> Matthieu