[Pw_forum] nscf calculation failure~

[lan haiping]

Hi,All.

I came to a  CRASH with this error information in nscf calculation :
"%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
    task #         1
    from cgramg1_k : error #         1
    negative or zero norm in S
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
"

I browsed the mail list, it seems this error is related to ultrasoft
pseudopotential's bad behavior, as pointing out by paolo
"S is the overlap matrix in the Ultrasoft pseudopotential formalism. It has
to be positive definite. Likely one (or more than one) of the
pseudopotentials is bad "
But no CRASH is occured in 'scf'  or 'relax' calculation, it's too strange
to belevie the problem of pseudopotential,
I also try to use  'cg' method for diagonalization instead of  'david '
method, and no such CRASH would be occured.

Regards,
hai-ping
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/attachments/20061111/aa7e1a11/attachment.htm