[Octopus-users] Geometry optimisation

[Robertson Burgess]

Hi David,

Sorry, perhaps I didn't explain my problem well enough.
My SCF cycles are converging. It just continues to converge SCF cycle after SCF cycle. The problem I have is it never reaches a minimisation iteration. i.e. the output file never comes up with a line like
+++++++++++++++++++++ MINIMIZATION ITER #:    1 ++++++++++++++++++++++

I got two such lines in my first geometry optimization run, then it seems to have gotten stuck after two. Now when I'm running it another time I still haven't gotten one, and it has converged 21 SCF cycles in a row.

Bob


>>> David Strubbe <dstrubbe at berkeley.edu> 21/09/10 4:53 PM >>>
Hi Bob,

If you can't get convergence, then you need to play with the Mixing and/or
Smearing. The lack of SCF convergence probably has nothing to do with
geometry optimization; if you start over for a gs calculation with the same
geometry you should have the same results (or lack thereof). It is useless
to specify so many ExtraStates unless they are all occupied to some
non-negligible degree. Also you may not need to go to ConvAbsEv = 1e-4. What
if you just use the default tolerance?

David

On Mon, Sep 20, 2010 at 8:42 PM, Robertson Burgess <
Robertson.Burgess at newcastle.edu.au> wrote:

> Dear octopus users.
>
> I'm performing a geometry optimization on a cluster containing 109 atoms. I
> ran the program once and it performed two minimization iterations before it
> just seemed to converge SCF cycle after SCF cycle without doing a
> minimization iteration.
>
> I then stopped that run, took the structure that it had already minimized
> to and put that as the new input structure. I then tried to run another
> geometry optimization run. So far during this run it has completed 17 SCF
> cycles, but still has not done one minimization iteration. Is there
> something I'm doing wrong, or something I can change to help the geometry
> optimization? Here is the contents of my inp file:
>
>
> calculationmode=go
>
> units=2
> %species
> 'Au'|196.9666|spec_ps_psf|79|3|0
> 'P'|30.974|spec_ps_psf|15|3|0
> 'H'|1.0079|spec_ps_psf|1|3|0
> 'Cl'|35.453|spec_ps_psf|17|3|0
> %
>
> extrastates=25
>
> XYZCoordinates='struct.txt'
>
> ConvAbsEv=1e-4
> EigensolverFinalTolerance=1.e-13
> EigensolverInitTolerance=1.e-8
> BoxShape=3
>
> Radius=3.5
>
>
>
> SpinComponents=1
> Spacing=0.20
> TypeOfMixing=2
> What2Mix=1
>
>
> periodicdimensions=0
>
> LRConvAbsDens = 1e-3
> LinearSolverTol = 1e-5
> Mixing=0.2
> ParallelizationStrategy=1
>
> OutputHow= xcrysden
>
> NumberUnoccStates=5
>
> SmearingFunction = fermi_dirac
> Smearing = 0.01
>
> MaximumIter=50
>
> OutputBandsGnuplotMode=yes
> DOSEnergyMin=-100
> DOSEnergyMax=0
> DOSEnergyPoints=1000
> DOSGamma=0.3
>
> GOMethod=cg_bfgs2
>
> TDDeltaStrength=0.01
> TDPolarizationDirection=2
> TDPolarizationEquivAxes=1
>
> TDExponentialMethod=lanczos
> TDExpOrder=100
>
> tmax=30
> TDEvolutionMethod= aetrs
> TDTimeStep =0.004
> #TDMaximumIter=10
> TDMaximumIter=tmax/TDTimeStep
>
>
> Thanks,
> Bob Burgess
> University of Newcastle
>
> _______________________________________________
> Octopus-users mailing list
> Octopus-users at tddft.org
> http://www.tddft.org/mailman/listinfo/octopus-users
>