Fitting a sum of exponentials to lattice correlation functions using a non-uniform prior

TL;DR

This paper introduces a Bayesian approach with a non-uniform prior to fit lattice correlation functions, effectively extracting excited states and comparing favorably with traditional methods like effective mass tables.

Contribution

The paper presents a novel Bayesian methodology for fitting sums of exponentials in lattice correlation functions, improving excited state extraction.

Findings

Bayesian method performs well compared to variational analysis.

Effective in analyzing torelon and glueball operators.

Provides a new approach for orthogonalization of correlation functions.

Abstract

Excited states are extracted from lattice correlation functions using a non-uniform prior on the model parameters. Models for both a single exponential and a sum of exponentials are considered, as well as an alternate model for the orthogonalization of the correlation functions. Results from an analysis of torelon and glueball operators indicate the Bayesian methodology compares well with the usual interpretation of effective mass tables produced by a variational procedure. Applications of the methodology are discussed.