[Pw_forum] problem in nmr calculation

[Davide Ceresoli]

Jones Tsz-Kai Wan wrote:
> Dear gipaw users,
> 
> I encountered the following error during nmr calculation:
> 
> ------------------------------------------------------------------------
>      Computing the magnetic susceptibility     isolve=0    ethr=0.1000E-09
>      k-point #    1 of    26      pool #  1
>      k-point #    2 of    26      pool #  1
>      k-point #    3 of    26      pool #  1
>      k-point #    4 of    26      pool #  1
>      k-point #    5 of    26      pool #  1
>      k-point #    6 of    26      pool #  1
>      k-point #    7 of    26      pool #  1
>      k-point #    8 of    26      pool #  1
>      k-point #    9 of    26      pool #  1
>      k-point #   10 of    26      pool #  1
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4785
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4793
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4793
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4793
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4793
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4793
>      ATTENTION: ik=  10  ibnd= 20  eigenvalues differ too much!
>         19.9838     19.4793
> ------------------------------------------------------------------------
> 
> What is the reason of this and how can I avoid this?

Dear Jones Tsz-Kai Wan,
     this is not really an error, it's more a warning. Most
of the times, it signals a problem with the diagonalization
at k+q. You can try to change the value of isolve (0 or 1)
in the gipaw input.
Sometimes, everything works out ok.
In any event, do the results look reasonable to you?

Davide