1: \begin{abstract}

2: In a recent article (J. Comput. Chem. 2013, 34, 132-140), convective

3: Replica-Exchange (convective-RE) has been presented as an alternative

4: to the standard even-odd transition scheme. Computations on systems of

5: various complexity have shown that convective-RE may increase the

6: number of replica round-trips in temperature space with respect to the

7: standard exchange scheme, leading to a more effective sampling of

8: energy basins. Moreover, it has been shown that the method may prevent

9: the formation of bottlenecks in the diffusive walk of replicas through

10: the space of temperature states. By using an ideal temperature-RE

11: model and a classical harmonic-oscillator RE scheme, we study the

12: performances of convective-RE when ergodicity is not broken and

13: convergence of acceptance probabilities is attained. In this dynamic

14: regime, the round-trip ratio between convective and standard-RE is at

15: maximum $\sim 1.5$, a value much smaller than that observed in

16: non-ergodic simulations. For large acceptance probabilities, the

17: standard-RE outperforms convective-RE. Our observations suggest that

18: convective-RE can safely be used in either ergodic or non ergodic

19: regimes; however, convective-RE is advantageous only when bottlenecks

20: occur in the state-space diffusion of replicas, or when acceptance

21: probabilities are globally low. We also show that decoupling of the

22: state-space dynamics of the stick replica from the dynamics of the

23: remaining replicas improves the efficiency of convective-RE at low

24: acceptance probability regimes.

25: \end{abstract}

26: