The ABINIT Users Mailing List ( CLOSED )

[TORRENT Marc <marc.torrent@cea.fr>]

Dear Willy,

From my opinon, your results are not suprising.

1- The differences on the lenght of the SCF cycle can be explained by 
the change of the default choice for the mixing scheme (iscf) between 
the 2 versions.
2- The differences on the absolute value of energy can be explained by 
several bug fixes, especially for spin-polarized systems. But, a 
difference on the absolute value of the energy does not means that that 
the physical results are wrong (energy differences may be correct).

Also note that a severe bug has been corrected in the computation of the 
on-site part of the XC energy for spin-polarized GGA calculations (your 
case here); it it correct from the 5.5 version...


Marc Torrent



Willy Offermans (WOffermans) a écrit :
> Dear Abinit friends,
>
> I have found some severe discrepancies in the results of two
> different abinit versions. At location A I use abinit 5.2.4 on FreeBSD
> 6.2-RELEASE-p7. Abinit has been installed via the ports. At location
> B I use the version 5.4.4, that is running on 64 bit Linux.
>
> The system under inspection is O2, that is calculated with PP and PAW. The
> results with PP are identical on both locations. Only the PAW
> results differ.
>
> I will include the input files and output files of the PAW calculations.
> The input is the same at both locations. The output differs severely.
>
> I hope someone can have a look and provide some feedback.
>
>