A New Method for Analysing Dark Matter Direct Detection Data

TL;DR

This paper introduces a new Bayesian analytical method for Dark Matter direct detection data that is robust and flexible, capable of handling complex data and providing stronger exclusion limits, with implications for interpreting experimental results.

Contribution

The paper presents a novel Bayesian method applicable to all direct detection experiments, improving data analysis robustness and sensitivity compared to existing approaches.

Findings

Agreement with XENON100 limits for 225 days data

Stronger exclusion limits for 100 days data

Indications of potential low-mass Dark Matter or unknown background in 225 days data

Abstract

The experimental situation of Dark Matter Direct Detection has reached an exciting cross-roads, with potential hints of a discovery of Dark Matter (DM) from the CDMS, CoGeNT, CRESST-II and DAMA experiments in tension with null-results from xenon-based experiments such as XENON100 and LUX. Given the present controversial experimental status, it is important that the analytical method used to search for DM in Direct Detection experiments is both robust and flexible enough to deal with data for which the distinction between signal and background points is difficult, and hence where the choice between setting a limit or defining a discovery region is debatable. In this article we propose a novel (Bayesian) analytical method, which can be applied to all Direct Detection experiments and which extracts the maximum amount of information from the data. We apply our method to the XENON100 experiment data as a worked example, and show that firstly our exclusion limit at 90% confidence is in agreement with their own for the 225 Live Days data, but is several times stronger for the 100 Live Days data. Secondly we find that, due to the two points at low values of S1 and S2 in the 225 days data-set, our analysis points to either weak consistency with low-mass Dark Matter or the possible presence of an unknown background. Given the null-result from LUX, the latter scenario seems the more plausible.