Relating centrality to impact parameter in nucleus-nucleus collisions

In ultrarelativistic heavy-ion experiments, one estimates the centrality of a collision by using a single observable, say $n$, typically given by the transverse energy or the number of tracks observed in a dedicated detector. The correlation between $n$ and the impact parameter, $b$, of the collision is then inferred by fitting a specific model of the collision dynamics, such as the Glauber model, to experimental data. The goal of this paper is to assess precisely which information about $b$ can be extracted from data without any specific model of the collision. Under the sole assumption that the probability distribution of $n$ for a fixed $b$ is Gaussian, we show that the probability distribution of the impact parameter in a narrow centrality bin can be accurately reconstructed up to $5\%$ centrality. We apply our methodology to data from the Relativistic Heavy Ion Collider and the Large Hadron Collider. We propose a simple measure of the precision of the centrality determination, which can be used to compare different experiments.
Comment: 9 pages, 8 figures; v2, published version: added analysis of CMS data; added fit of STAR data in Fig.6; ancillary file (fit.py) contains a python script which performs the fit of Trento data (trento.dat) on distribution of entropy