[Pw_forum] problems with integrated charge

[Liping YU]

Paolo Giannozzi wrote:
>
> On Oct 25, 2006, at 24:22 , Vivek Ranjan wrote:
>
>> I am trying to relax a structure. Within 3 scf steps the program aborts
>> with the following :
>>
>> WARNING: integrated charge= 207.40294106, expected= 208.00000000
>>
>> from electrons : error # 1
>> charge is wrong
>> [...]
>> What could be the reason ?
>
> incorrect data. The only reason related to the code I am aware of, is
> a failure of the parallel algorithm for subspace diagonalization. You
> can try to disable it, following the instructions I posted a few days
> ago.
I often get this error message when parallel algorithm for subspace
diagonalization
is used for large supercell ( especially when celldm(3) is big for
tetragonal cell )
calculations. For the same system with smaller super cell, there is no
such problem
when a serial algorithm is used. So I think this is due to parallel
algorithm in the
code not the bad data.

To let program continue running without this error message, I changed the
line in PW/electron.90 " ABS( charge - nelec ) / charge > *1.0D-3* " to
" ABS( charge - nelec ) / charge > *1.d0* ". When scf calculation is
close to
convergence, such WARNING message will disappear. I don't know if this is
a good way to do. If yes, then criterion for relative charge difference
is too
strict and we need to change it. If not, as Paolo suggested, maybe we need
to disable the parallel algorithm for subspace diagonalization.

Liping

> If you have evidence that your data is good and the code still doesn't
> work, please provide a test
>
> Paolo
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum