Multicanonical Algorithm, Simulated Tempering, Replica-Exchange Method, and All That

TL;DR

This paper extends multicanonical algorithms, simulated tempering, and replica-exchange methods to multiple dimensions by incorporating additional physical quantities as energy terms, enabling more flexible and comprehensive sampling in complex systems.

Contribution

It introduces a multi-dimensional generalization of these algorithms, allowing simultaneous sampling over multiple parameters and energies, which was not previously available.

Findings

Multi-dimensional algorithms enable efficient sampling in extended parameter spaces.

Replica-exchange simulations help determine weight factors for multi-dimensional methods.

The approach facilitates comprehensive exploration of complex physical systems.

Abstract

We discuss multi-dimensional generalizations of multicanonical algorithm, simulated tempering, and replica-exchange method. We generalize the original potential energy function $E_0$ by adding any physical quantity $V$ of interest as a new energy term with a coupling constant $\lambda$. We then perform a multi-dimensional multicanonical simulation where a random walk in $E_0$ and $V$ space is realized. We can alternately perform a multi-dimensional simulated tempering simulation where a random walk in temperature $T$ and parameter $\lambda$ is realized. The results of the multi-dimensional replica-exchange simulations can be used to determine the weight factors for these multi-dimensional multicanonical and simulated tempering simulations.