Virtual-move Parallel Tempering

TL;DR

This paper introduces a new Monte Carlo method called Virtual-move Parallel Tempering that significantly improves sampling efficiency in parallel-tempering simulations by utilizing all trial moves, not just accepted ones, leading to better free-energy landscape estimation.

Contribution

The paper presents a novel Monte Carlo scheme that enhances parallel-tempering simulations by incorporating information from all trial moves, increasing sampling efficiency.

Findings

Sampling region increased by a factor of 20

Improved free-energy landscape estimation

Enhanced statistical averaging

Abstract

We report a novel Monte Carlo scheme that greatly enhances the power of parallel-tempering simulations. In this method, we boost the accumulation of statistical averages by including information about all potential parallel tempering trial moves, rather than just those trial moves that are accepted. As a test, we compute the free-energy landscape for conformational changes in simple model proteins. With the new technique, the sampled region of the configurational space in which the free-energy landscape could be reliably estimated, increases by a factor 20.