Simulated-tempering approach to spin-glass simulations

TL;DR

This paper applies simulated tempering to 2D Ising spin glass simulations, demonstrating reduced slowing down and advantages like vectorization and direct access to the canonical ensemble.

Contribution

It introduces an iteration procedure for parameter determination in simulated tempering and successfully applies it to spin-glass simulations.

Findings

Reduction of slowing down comparable to multicanonical algorithm

Allows full vectorization of programs

Provides direct access to the canonical ensemble

Abstract

After developing an appropriate iteration procedure for the determination of the parameters, the method of simulated tempering has been successfully applied to the 2D Ising spin glass. The reduction of the slowing down is comparable to that of the multicanonical algorithm. Simulated tempering has, however, the advantages to allow full vectorization of the programs and to provide the canonical ensemble directly.