[gpaw-users] Severe convergence problems

[Michael Busch]

Hi all,

I am currently studying the activity of different doped MnO2 slabs. All 
systems are based on a MnO2 structure  and contain Ta + a transition 
metal as dopants. The structures where taken from previous calculations 
so the starting guesses should be reasonable. While most systems 
converge without significant problems I am having severe convergence 
problems for several of the Ni doped oxides. All calculations are spin 
polarised with the spin being allowed to change during the calculations. 
The obtained spins seem to be reasonable for the system at hand. A 
summary of the used parameters can be found below.

Until now I tried the following to achieve convergence:

- different eigensolvers: cg and Davidson
- gradually increasing smearing from 0.1 eV to values as high as 1.5 eV (!)
- modifying the mixing parameters between 0.001 and 0.1


Is there any possibility to obtain convergence that I did not consider 
until now?

All the best

Michael



calc = GPAW(
             xc='RPBE',
              h=0.15,
         charge=0,
        spinpol=True,
      occupations=FermiDirac(
                 width=1.0,
             fixmagmom=False
                          ),
           kpts=(2, 1, 1),
      convergence={'energy': 0.0005},
        maxiter=1024,
      eigensolver='cg',
        usesymm=True,
  mixer=Mixer(0.05, 5, weight=100.0),
            )

-- 
Dr. Michael Busch
Postdoc
Center for Atomic-Scale Materials Design
Department of Physics
Technical University of Denmark
DK-2800 Kgs. Lyngby
Denmark

Phone : +45 45 25 32 35
E-mail: mbus at fysik.dtu.dk