[Pw_forum] A bug?
degironc
degironc at sissa.it
Fri Jul 6 15:26:50 CEST 2007
I agree the printout is different but this is because in the UPF format
the radial grid is given as a numerical array while in the original
pseudo it is given in terms of xmin, dx (and zmesh which is not
reported...) the two grids are (should) be identical and all results
should be exactly the same as, I understand, it is the case...
maybe we should change the printout when a numerical radial grid is used
(xmin and dx do not make sense anymore).
Stefano de Gironcoli
SISSA & DEMOCRITOS, Trieste (Italy)
Lazaro Calderin wrote:
> Hi,
>
> There is a possible hopefully not harmful bug in pw.x.
> If you use the same pseudo with the upf and old format there are a
> couple of line at the beginning that do not report the same info and in
> the case of upf it is wrong.
>
> For example for a pseudo for Al with the two formats you get the
> differences:
>
> 35c35
> < Exchange-correlation = SLA PZ NOGX NOGC (1100)
> ---
>
>> Exchange-correlation = LDA (1100)
>>
> 52,53c52,53
> < PSEUDO 1 is Al zval = 3.0 lmax= 2 lloc= 0
> < (in numerical form: 1135 grid points, xmin = 0.00, dx = 0.0000)
> ---
>
>> PSEUDO 1 is Al zval = 3.0 lmax= 2 lloc= 2
>> (in numerical form: 1135 grid points, xmin = -7.00, dx = 0.0125)
>>
>
> In the case of the upf lloc, xmin and dx are all zero while for the old
> format they are reported correctly.
>
> To reproduce the bug generate the pseudo potentials using, i.e.,
> espresso/atomic_doc/pseudo-gen/al.in and run example03 for the two
> pseudo potentials in the two formats.
>
> The results for the two pseudopotential are pretty much the same as
> compared with the reference. That is why I think this buggy is not
> harmful to the calculations.
>
> Regards,
> Lazaro
>
>
>
>
More information about the users
mailing list