The XENON100 exclusion limit without considering Leff as a nuisance parameter

TL;DR

This paper reevaluates the XENON100 dark matter exclusion limits by directly fitting the scintillation efficiency Leff from data, avoiding nuisance parameter profiling, to better understand the impact of Leff uncertainties on light dark matter constraints.

Contribution

It introduces a maximum likelihood approach that treats Leff as a data-driven parameter rather than a nuisance, providing more robust exclusion limits for light dark matter.

Findings

Exclusion limits are sensitive to assumptions about Leff at low energies.

Direct fitting of Leff yields different constraints compared to nuisance parameter profiling.

Results highlight the importance of accurate Leff modeling for light dark matter searches.

Abstract

In 2011, the XENON100 experiment has set unprecedented constraints on dark matter-nucleon interactions, excluding dark matter candidates with masses down to 6 GeV if the corresponding cross section is larger than 10^{-39} cm^2. The dependence of the exclusion limit in terms of the scintillation efficiency (Leff) has been debated at length. To overcome possible criticisms XENON100 performed an analysis in which Leff was considered as a nuisance parameter and its uncertainties were profiled out by using a Gaussian likelihood in which the mean value corresponds to the best fit Leff value smoothly extrapolated to zero below 3 keVnr. Although such a method seems fairly robust, it does not account for more extreme types of extrapolation nor does it enable to anticipate on how much the exclusion limit would vary if new data were to support a flat behaviour for Leff below 3 keVnr, for example. Yet, such a question is crucial for light dark matter models which are close to the published XENON100 limit. To answer this issue, we use a maximum Likelihood ratio analysis, as done by the XENON100 collaboration, but do not consider Leff as a nuisance parameter. Instead, Leff is obtained directly from the fits to the data. This enables us to define frequentist confidence intervals by marginalising over Leff.