On Apr 24, 2006, at 7:31 AM, Aleksander Wittlin wrote:
> Dear Dr Troyer,
>
> I am sorry to bother you with perhaps trivial question.
> I am trying to run the spinmc code with
> dr's Todo fcc lattice definition. It
> works, however some diagnostics seems still cryptic to me.
> For example for the below enclosed parameters file
> I get for some temperatures ERROR "converged"="yes"
> and for somewhat higher, presumably closer to the
> phase transition "converged"="no". Should I tinker with
> parameters or is it OK?
converge=no means that you have to run for a larger number of sweeps.
This is natural close to the phase transition or at very low
temperatures since there the autocorrelation times increase. The best
procedure is to start from very small systems, and only increase the
system size once you see that you get good convergence of the errors.
As long as they are not converged, it is hard to trust the results
since the error bar is most likely much larger than estimated.
Best regards
Matthias Troyer