Is the broad histogram random walk dynamics correct?

Abstract:

We show explicitly that the broad histogram single-spin-flip random walk dynamics does not give correct microcanonical average even in one dimension. The dynamics violates the detailed balance condition by an amount proportional to the inverse system size. As a result, in distribution different configurations with the same energy can have different probabilities. We propose a modified dynamics which ensures detailed balance and the histogram obtained from this dynamics is exactly flat. The broad histogram equation relating the average number of potential moves to density of states is generally valid.