[Pw_forum] Li pseudopotential

[Lorenzo Paulatto]

Replying to message "Re: [Pw_forum] Li pseudopotential" from Miguel  
Martínez (17/11/09):
> Maybe this doesn't apply to US or PAW pseudopotentials, since I've
> "succesfully" created a Cs pseudo with 9 electrons using TM and both
> the Li-1s and the Na-2s2p inputs in the PAW library use tm.

Indeed, in the ultrasoft and PAW formalism pseudo-wavefunctions do not  
have to be orthogonal.

> As a side note, I'm getting various errors when building the PAW
> library pseudopotentials with espresso 4.1.1:
>  -Cl, Ge, P, S and Si produce a "negative rho" error
>  -Li-1s produces a "q_i coefficients" error

Yes, it is a recent change in the code which I haven't had time to check  
yet, you can get rid of it by increasing r_match_augfun, or even removing  
it from the input.

> A diff between reference and
> available output shows noticeable differences only in the bmat matrix
> before the crashes.

I think the bmat matrix depends on the normalization of wavefunctions,  
which has changed too, so this should be unrelated and harmless.  
Apparently when the check fails the PAW dataset would have contained a  
substantial amount of negative charge, which can cause some problem during  
the plane-wave calculation. It is better to work around it by changing the  
parameters a bit than just removing the check.

regards

-- 
Lorenzo Paulatto
SISSA  &  DEMOCRITOS (Trieste)
phone: +39 040 3787 511
skype: paulatz
www:   http://people.sissa.it/~paulatto/

     *** save italian brains ***
  http://saveitalianbrains.wordpress.com/