[Pw_forum] subtile error for non collinear calculations with a GGA pseudo

[Andrea Dal Corso]

On Mon, 2011-02-21 at 13:59 +0100, Cyrille Barreteau wrote:
> Dear all
> 
> I have noticed an usual behavior when doing non collinear calculation
> with a GGA pseudopotential.
> 
> During the iteration process there are some NaN appearing in the output
> file (see below). The strange thing is that the code continues and
> converges..
> 
> iteration #  1     ecut=    35.00 Ry     beta=0.30
>      Davidson diagonalization with overlap
>      ethr =  1.00E-02,  avg # of iterations =  2.0
>               NaN         NaN         NaN

This error has been already corrected in the cvs version a few months
ago. Note however that GGA + noncollinear is less stable than the
other parts of QE.

HTH, 

Andrea

> 
> This type of Pb is not present when using a LDA pseudo.
> 
> Moreover I have tried to do a calculation of a collinear configuration
> with or without the 'noncolin' option. In principle both should lead to
> almost the same solution. This is not the case. There are some
> differences which according to me are larger than just numerical
> errors..
> 
> I guess there is some kind of Pb when performing non collinear
> calculations with a GGA pseudo.
> 
> thanks to give me a hint to solve this problem
> 
>   cyrille
> 
-- 
Andrea Dal Corso                    Tel. 0039-040-3787428
SISSA, Via Bonomea 265              Fax. 0039-040-3787249
I-34136 Trieste (Italy)             e-mail: dalcorso at sissa.it