[gpaw-users] Re: Convergence error

[Jens Jørgen Mortensen]

On Mon, 2010-04-19 at 13:38 +0200, Vincent Davesne wrote:
> Hi!
> 
> Ask Hjorth Larsen wrote:
> > Hi Vincent
> >
> > On Fri, 16 Apr 2010, Vincent Davesne wrote:
> >
> >> Hi!
> >>
> >> I have a problem with the code I used.
> >>
> >> from ase import *
> >> from ase.lattice.surface import *
> >> from gpaw import GPAW
> >>
> >> slab = fcc111('Ag', size=(3,3,3), vacuum=5.0)
> >>
> >> calc = GPAW(mode='lcao',basis='dzp',h=0.2,nbands=-10,xc='RPBE', 
> >> txt='Ag111.out')
> >>
> >> slab.set_calculator(calc)
> >>
> >> constraint = FixAtoms(mask=[a.get_tag() != 1 for a in slab])
> >> slab.set_constraint(constraint)
> >>
> >> dyn = QuasiNewton(slab, trajectory='Ag111_f.traj')
> >> dyn.run(fmax=0.05)
> >>
> >> gives back:
> >> gpaw.KohnShamConvergenceError: Did not converge!
> >>
> >> It's just a relaxation of an Ag fcc 111 terrace surface. How can I 
> >> solve this issue?
> >> Thanks a lot,
> >> Vincent
> >
> > Increasing the electronic temperature probably works.  Also, you could 
> > try changing the mixer or adding some k-points.  Does the basis you 
> > are using correspond to RPBE?  Attaching the calculation log is 
> > generally a good idea.
> I tried to increase the electronic temperature, but
> calc = GPAW(mode='lcao',basis='dzp',h=0.2,nbands=-10,xc='RPBE', 
> txt='Ag111_2.out', occupations=FermiDirac(0.1))
> 
> gives back immediately
> NameError: name 'FermiDirac' is not defined
> 
> So I tried to increase the number of k points, by specifying
> calc = GPAW(mode='lcao',basis='dzp',h=0.2,nbands=-10,xc='RPBE', 
> txt='Ag111_2.out',kpts=(4,4,1))

Looks like more k-points works quite well.  I would also use more than
10 empty bands - this can also improve convergence.

> gives back after the run
> NotImplementedError

For version 0.6, forces are not implemented for non-orthogonal unit
cells.  You have to use the development version of GPAW to get forces.
Sorry :-(

JJ