[DFTB-Plus-User] Fwd: SCC convergence problem for cluster calculation

[Ben Hourahine]

Hello Anjali,

On extra things to try, is to increase the electron temperature.
10 Kelvin is quite small, several hundred is more common (but
the choice depends on your band structure).

Regards

Ben

On 17/10/12 07:01, Bálint Aradi wrote:
> Dear Anjali,
>
>    Apart of some special cases, DFTB+ shows usually a nice SCC
> convergence behaviour. Are you sure, that your input is physically
> meaningful? Especially, that atoms are not unphysically close to each other?
>
>   One thing I note in your input, that your Broyden-mixing parameter is
> extremely low. Usually, 0.1-0.2 is the right range. Also, you could as
> suggested by Rodrigo already, set SCCTolerance less tight as the default
> value (e.g. set it to 1e-4 instead of the default 1e-5).
>
>    Alternatively, if you geometry is rather unphysical, you could set
> ConvergentForcesOnly = No in the ConjugateGradient section. This would
> allow for relaxation along non-convergent forces as long as your system
> does not become more physical. Similarly, you could just set SCC=No for
> a while, relax the sytem, and restart then with SCC = Yes.
>
>   I hope, some of the ideas will help you.
>
>    Best regards,
>
>    Bálint
>
>


-- 
      Dr. B. Hourahine, SUPA, Department of Physics,
    University of Strathclyde, John Anderson Building,
           107 Rottenrow, Glasgow G4 0NG, UK.
     +44 141 548 2325, benjamin.hourahine at strath.ac.uk
    The University of Strathclyde is a charitable body,
registered in Scotland, with registration number SC015263