Characteristic time scales and overlap distributions in replica exchange Monte Carlo simulations of spin glasses*

We present a large-scale numerical study using replica exchange Monte Carlo (parallel tempering) of time scales of the three- dimensional Ising spin glass. We measure the integrated and exponential autocorrelation times for several observables, as well as the round-trip times for different disorder realizations in order to investigate the relationship between the characteristic time scales of a disorder realization and its overlap distribution.