[Octopus-users] geometry optimisation

[Robertson Burgess]

Hi David,

No, what I mean is that after a certain point, the SCF cycles converge, and then it just starts another cycle. It just runs SCF cycle after SCF cycles, always converging them, but never actually appearing to get anywhere except to run endless SCF cycles. It never comes up with the line saying "MINIMIZATION ITER #: 1". Overnight my calculation converged 601 SCF calculations without making a minimization iteration.

I'll try and see what making a smaller step size does, maybe that's my problem. It's not the mixing, as the SCF cycles readily converge.

Bob

>>> David Strubbe <dstrubbe at berkeley.edu> 24/02/2010 6:01 pm >>>
Hi Bob,

Do you mean that after a certain point, the SCF cycles don't converge? The
number of SCF iterations per minimization step is highly dependent on your
system, the size of the steps you are taking, and also the mixing, so it's
hard to say in general, but the number of SCF iterations should be less for
subsequent steps than the first, since the new charge density will be close
to the previous charge density. To make it converge, you probably should
either fiddle with the mixing, or make the step size smaller.

David

On Tue, Feb 23, 2010 at 7:50 PM, Robertson Burgess <
Robertson.Burgess at newcastle.edu.au> wrote:

> Dear Octopus Users,
>
> I'm trying to relax some gold clusters using the geometry optimisation run
> in octopus. I've tried playing around with the various settings and I can't
> find anything that seems to work significantly better than the defaults.
> Usually when I run octopus it will perform a couple of MINIMIZATION ITER
> and then it seems to get stuck. It will just run SCF cycle after SCF cycle
> without ever seeming to work towards another minimization iteration.
>
> As a guide, about how many SCF cycles are expected per Minimization
> Iteration. What would cause the geometry optimization to get stuck, and what
> variables can I change to prevent it happening?
>
> Thanks,
> Robertson Burgess
> University of Newcastle
>
> _______________________________________________
> Octopus-users mailing list
> Octopus-users at tddft.org 
> http://www.tddft.org/mailman/listinfo/octopus-users 
>