[Octopus-devel] Geometry optimization failure

[Micael Oliveira]

> So, should one then disregard the information of the total energy 
> and just look at the forces? But how should we do this? Or do you think
> there is a (small) bug getting the total energy?
I don't know. Maybe the problem is in the forces and not in the total 
energy. At least things are consistent at the global minimum.

  For now I would say that if you do not require a very high precision 
on the forces, the geometry optimization in octopus will give you 
reasonable results. I mean, for Na2 and H2O the geometries were not bad 
at all and I'm also getting decent results for the Xe clusters. The only 
problem is that sometimes you will not have a clean exit from the GSL 
routine.

  Anyway, this is something we should investigate more carefully when we 
have some time. Checking how the forces and energy minimums change with 
the spacing could be a good idea.

  By the way, previously I forgot to point out one important thing: when 
using the geometry minimization in octopus, one has to be very careful 
when using spheres around the atoms to build the box, as those atoms 
will move during the minimization but not the box boundaries...


  Cheers,

Micael