[gpaw-users] PoissonSolver convergence error

[askhl at fysik.dtu.dk]

Dear Dominik

> Dear Gpaw users,
>
>  I was wondering if it is possible to increase the maximum number of
> iterations of the Poisson solver above the standard 1000. The Poisson
> Solver does not have a "maxiter" parameter, but maybe there is a way
> around it.
>
> My calculator setup is as follows:
>
> calc = GPAW(xc='RPBE',
>             mode = 'lcao',
>             gpts = (16, 16, 672),
>             kpts = (8,8,1),
>             txt=name_system+'_calc_state.txt',
>             poissonsolver=PoissonSolver(relax='GS', eps=1e-12),
>             mixer=Mixer(0.01, 20, weight=100.0),
>             occupations = FermiDirac(0.1, maxiter=2000),
>             maxiter = 500)
>
> I have followed the tips from the page on convergence issues, so the
> number of points of the real grid is divisible by 8 and I am using the
> Gauss-Seidel method. The SCF loop does not converge for eps=1e-10 so I am
> trying with eps=1e-12 (another tip from the page on convergence issues).
>
>
> Best Regards
> Dominik
>
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>

If it is necessary to increase even the iteration count of the
*occupations* object, then the occupations (and thus density) are
oscillating so wildly as to cause considerable suspicion.  In other words,
the density mixing is probably the real cause of the error.  Try with even
lower 'alpha' than 0.01, and maybe higher weight also (since the system
has such an unusual shape).

I'm sure you're more than welcome to implement a maxiter keyword for the
Poisson solver also. :)

Regards
Ask