Quantifying the evidence for Dark Matter in CoGeNT data

TL;DR

This paper re-analyzes CoGeNT data to assess evidence for dark matter, focusing on background separation and statistical methods, ultimately finding weak support for light dark matter signals.

Contribution

It introduces a refined background separation method using rise-time distributions and applies comprehensive statistical analyses to evaluate dark matter evidence.

Findings

Weak evidence for light dark matter at less than 1 sigma

Improved background modeling with rise-time distribution fitting

Robustness checks show results are sensitive to analysis choices

Abstract

We perform an independent analysis of data from the CoGeNT direct detection experiment to quantify the evidence for dark matter recoils. We critically re-examine the assumptions that enter the analysis, focusing specifically on the separation of bulk and surface events, the latter of which constitute a large background. This separation is performed using the event rise-time, with the surface events being slower on average. We fit the rise-time distributions for the bulk and surface events with a log-normal and Pareto distribution (which gives a better fit to the tail in the bulk population at high rise-times) and account for the energy-dependence of the bulk fraction using a cubic spline. Using Bayesian and frequentist techniques and additionally investigating the effect of varying the rise-time cut, the bulk background spectrum and bin-sizes, we conclude that the CoGeNT data show a preference for light dark matter recoils at less than 1 sigma.