ausblenden:
Schlagwörter:
-
Zusammenfassung:
The problem of critical slowing down is the appearance of a very long (diverging) correlation time associated with a corresponding apperance of a diverging correlation lenght. This is elucidated by a trivially solvable model, the one-dimensional Gaussian model. The general strategy to fight this effect, in order to sample phase space more efficiently, is to update large lenght scales with artificially high rates.This approach is exemplified by the methods of Fourier acceleration, and multigrid simulations. For polymer chains, the long correlations arise directly from the molecular connectivity. Some Monte Carlo algorithms which attack this problem again by non-local moves (dimerization, pivot, slithering snake, connectivity-altering methods) are presented.