Minimal Autocorrelation in Hybrid Monte Carlo simulations using Exact Fourier Acceleration

Abstract

The hybrid Monte Carlo (HMC) algorithm is a ubiquitous method in computational physics with applications ranging from condensed matter to lattice QCD and beyond. However, HMC simulations often suffer from long autocorrelation times, severely reducing their efficiency. In this work two of the main sources of autocorrelations are identified and eliminated. The first source is the sampling of the canonical momenta from a sub-optimal normal distribution, the second is a badly chosen trajectory length. Analytic solutions to both problems are presented and implemented in the exact Fourier acceleration (EFA) method. It completely removes autocorrelations for near-harmonic potentials and consistently yields (close-to-) optimal results for numerical simulations of the Su-Schrieffer-Heeger and the Ising models as well as in lattice gauge theory, in some cases reducing the autocorrelation by multiple orders of magnitude. EFA is advantageous for and easily applicable to any HMC simulation of an action that includes a quadratic part.

15 + 8 pages, 4 + 1 figures, 3 algorithms; v2: removed gauge fixing; v3: minor revision, published

References in corpus (10)

• Ab initio calculation of the neutron-proton mass difference
• Critical slowing down and error analysis in lattice QCD simulations
• Optimised Trotter Decompositions for Classical and Quantum Computing
• Exploring the HMC trajectory-length dependence of autocorrelation times in lattice QCD
• SmoQyDQMC.jl: A flexible implementation of determinant quantum Monte Carlo for Hubbard and electron-phonon interactions (version 2.0 release)
• First-principle quantum Monte-Carlo study of charge carrier mobility in organic molecular semiconductors
• Gauge-Fixed Fourier Acceleration
• Overcoming Ergodicity Problems of the Hybrid Monte Carlo Method using Radial Updates
• Fourier Acceleration in a Linear Sigma Model with Spontaneous Symmetry Breaking
• Hybrid Monte Carlo Simulation with Fourier Acceleration of the $N=2$ Principal Chiral Model in two Dimensions

Cited by in corpus (1)

• Fully ergodic simulations using radial updates