[Paper Review] Remarks on the construction of worm algorithms for lattice field theories in worldline representation

Mario Giuliani, Christof Gattringer|arXiv (Cornell University)|Feb 15, 2017

Scientific Research and Discoveries13 references3 citations

TL;DR

This paper presents a generalized worm algorithm for lattice field theories in worldline representation, where site-dependent weight factors depend on all fluxes at a site. It introduces an amplitude parameter A to control worm starting and terminating probabilities, enabling optimization of worm length and simulation efficiency. The method ensures detailed balance and is validated numerically on the 4D relativistic Bose gas, showing A significantly influences average worm length and flux change per step.

ABSTRACT

We introduce a generalized worldline model where the partition function is a sum over configurations of a conserved flux on a d-dimensional lattice. The weights for the configurations of the corresponding worldlines have factors living on the links of the lattice, as well as terms which live on the sites x and depend on all fluxes attached to x. The model represents a general class of worldline systems, among them the dual representation of the relativistic Bose gas at finite density. We construct a suitable worm algorithm and show how to correctly distribute the site weights in the various Metropolis probabilities that determine the worm. We analyze the algorithm in detail and give a proof of detailed balance. Our algorithm admits the introduction of an amplitude parameter A that can be chosen freely. Using a numerical simulation of the relativistic Bose gas we demonstrate that A allows one to influence the starting and terminating probabilities and thus the average length and the efficiency of the worm.

Motivation & Objective

• To address the lack of theoretical foundation for worm algorithms in lattice field theories with site-dependent weight factors.
• To construct a worm algorithm that correctly incorporates site weights depending on all fluxes at a site into Metropolis acceptance probabilities.
• To introduce an amplitude parameter A that controls worm starting and terminating probabilities, enabling optimization of worm length and simulation efficiency.
• To prove detailed balance for the proposed algorithm in a general worldline model with both link and site weights.
• To demonstrate the algorithm's performance and tunability via numerical simulation of the 4D relativistic Bose gas.

Proposed method

• The model generalizes worldline systems with conserved flux on a d-dimensional lattice, including link weights and site weights that depend on all fluxes at a site.
• A worm algorithm is constructed where the worm moves through the lattice, changing fluxes on links and updating weights via Metropolis steps.
• Site weights are distributed in Metropolis probabilities such that they appear in the denominator during worm starting and in the numerator during termination, ensuring correct statistical weights.
• An amplitude parameter A is introduced in the starting and terminating probabilities to compensate for small or large site weights and to tune worm length.
• Detailed balance is proven by decomposing transition probabilities into individual worm steps and matching each step with its inverse, ensuring correct equilibrium distribution.
• Numerical simulations of the 4D relativistic Bose gas are performed to study the effect of A on worm length, flux changes, and step efficiency.

Experimental results

Research questions

• RQ1 How can site-dependent weight factors, which depend on all fluxes at a site, be correctly incorporated into the Metropolis probabilities of a worm algorithm?
• RQ2 What is the role of the amplitude parameter A in controlling the starting and terminating probabilities of the worm, and how does it affect worm length?
• RQ3 Can the proposed worm algorithm be proven to satisfy detailed balance in the presence of complex site weights?
• RQ4 How does the amplitude parameter A influence the average flux change per worm step and the overall simulation efficiency?
• RQ5 To what extent can the amplitude parameter A be used to optimize performance across different coupling regimes in lattice field theories?

Key findings

• The amplitude parameter A allows independent control over the starting and terminating probabilities of the worm, enabling tuning of average worm length.
• As A increases, the normalized flux change per worm step (Δflux / N_steps) decreases, indicating longer worms do not necessarily yield higher flux changes per step.
• The normalized flux change per link (Δflux / 4V) increases with A, confirming that longer worms are generated as A increases.
• Finite-size effects in the number of accepted steps and total flux changes are attributed to worms winding around periodic boundaries.
• The proposed algorithm maintains detailed balance for general worldline models with both link and site weights, providing a theoretical foundation for its use.
• The method is generalizable to other systems and worm algorithm variants, offering a systematic way to optimize performance via A.