[Pw_forum] Too large susceptibility in NMR/ g-tensor calculations using GIPAW

[Davide Ceresoli]

On 10/17/2010 11:46 AM, Aurab wrote:
> Dear QE users and developers,
> I am facing a strange problem calculating g tensor of Al doped ZnO
> systems. Some elements of the susceptibility tensor (chi_bare) is
> printed as ****** . I assume this is probably due to occurrence of a
> very large number for that element. Could you please tell me why the
> susceptibility is blowing up and also what is chi_bare pGv and chi_bare
> vGv. The block of the output file is given below.
>
>
>
> End of magnetic susceptibility calculation
>
>       f-sum rule:
>        -285.026632      1.736420      0.008960
>           1.735646   -277.370374      0.014146
>           0.008277      0.011673   -280.653668
>
>       f-sum rule (symmetrized):
>        -281.198503      0.000000      0.000000
>           0.000000   -281.198503      0.000000
>           0.000000      0.000000   -280.653668

Dear Aurab,
     it would be useful if you could post your input files.
In the meanwhile I can only guess: is your cell hexagonal?
GIPAW is incompatible with symmetries that do not map cartesian
axis on themselves. Try nosym = .true.

How many electrons do you have? the f-sum rule is a good indicator
of the convergence of GIPAW calculations. In my experience,
the % error on chemical shift is similar to the % error on the
f-sum rule, but the error on susceptibilities is much larger.
Is your system spin-polarized?

To converge the f-sum rule, you must increase the number of
k-points.

HTH.

Cheers,
     Davide