# Reliable estimation of the radius of convergence in finite density QCD

## TL;DR

This paper introduces new reliable estimators for the radius of convergence in finite density QCD, addressing limitations of previous methods and demonstrating their effectiveness on a toy model.

## Contribution

The authors propose three novel estimators for the radius of convergence that work reliably in finite volume and relate to Lee-Yang zeros, improving over the ratio estimator.

## Key findings

- New estimators accurately determine the radius of convergence.
- Estimators remain stable despite high-order Taylor coefficient errors.
- Ensemble contains sufficient information for convergence estimation despite large coefficient uncertainties.

## Abstract

We study different estimators of the radius of convergence of the Taylor series of the pressure in finite density QCD. We adopt the approach in which the radius of convergence is estimated first in a finite volume, and the infinite-volume limit is taken later. This requires an estimator for the radius of convergence that is reliable in a finite volume. Based on general arguments about the analytic structure of the partition function in a finite volume, we demonstrate that the ratio estimator cannot work in this approach, and propose three new estimators, capable of extracting reliably the radius of convergence, which coincides with the distance from the origin of the closest Lee-Yang zero. We also provide an estimator for the phase of the closest Lee-Yang zero, necessary to assess whether the leading singularity is a true critical point. We demonstrate the usage of these estimators on a toy model, namely 4 flavors of unimproved staggered fermions on a small $6^3 \times 4$ lattice, where both the radius of convergence and the Taylor coefficients to any order can be obtained by a direct determination of the Lee-Yang zeros. Interestingly, while the relative statistical error of the Taylor expansion coefficients steadily grows with order, that of our estimators stabilizes, allowing for an accurate estimate of the radius of convergence. In particular, we show that despite the more than 100\% error bars on high-order Taylor coefficients, the given ensemble contains enough information about the radius of convergence.