On the Statistics of Baryon Correlation Functions in Lattice QCD

TL;DR

This paper analyzes the noise structure in baryon correlation functions in lattice QCD, revealing a sign problem and proposing a new method to improve energy measurement accuracy at large times.

Contribution

It introduces a novel analysis of noise in baryon correlators and proposes a method to maintain a constant signal-to-noise ratio, enabling more accurate large-time energy measurements.

Findings

Noise is linked to a sign problem and characterized by normal and wrapped normal distributions.

Large-time noise region causes unreliable standard energy measurements.

A new analysis method maintains constant signal-to-noise ratio, improving nucleon mass extraction.

Abstract

A systematic analysis of the structure of single-baryon correlation functions calculated with lattice QCD is performed, with a particular focus on characterizing the structure of the noise associated with quantum fluctuations. The signal-to-noise problem in these correlation functions is shown, as long suspected, to result from a sign problem. The log-magnitude and complex phase are found to be approximately described by normal and wrapped normal distributions respectively. Properties of circular statistics are used to understand the emergence of a large time noise region where standard energy measurements are unreliable. Power-law tails in the distribution of baryon correlation functions, associated with stable distributions and "Levy flights", are found to play a central role in their time evolution. A new method of analyzing correlation functions is considered for which the signal-to-noise ratio of energy measurements is constant, rather than exponentially degrading, with increasing source-sink separation time. This new method includes an additional systematic uncertainty that can be removed by performing an extrapolation, and the signal-to-noise problem re-emerges in the statistics of this extrapolation. It is demonstrated that this new method allows accurate results for the nucleon mass to be extracted from the large-time noise region inaccessible to standard methods. The observations presented here are expected to apply to quantum Monte Carlo calculations more generally. Similar methods to those introduced here may lead to practical improvements in analysis of noisier systems.