Analysis of the theoretical bias in dark matter direct detection

TL;DR

This paper investigates how incorrect assumptions about dark matter interactions and velocity distributions can bias the results of direct detection experiments, highlighting the importance of accurate modeling.

Contribution

It provides a quantitative analysis of theoretical biases in dark matter detection, using simulated data and statistical methods within an effective theory framework.

Findings

Best fit dark matter mass can differ by up to 2 standard deviations from true values.

Dark matter-nucleon coupling constants can be biased by several standard deviations.

Common assumptions may lead to significant inaccuracies in parameter estimation.

Abstract

Fitting the model "A" to dark matter direct detection data, when the model that underlies the data is "B", introduces a theoretical bias in the fit. We perform a quantitative study of the theoretical bias in dark matter direct detection, with a focus on assumptions regarding the dark matter interactions, and velocity distribution. We address this problem within the effective theory of isoscalar dark matter-nucleon interactions mediated by a heavy spin-1 or spin-0 particle. We analyze 24 benchmark points in the parameter space of the theory, using frequentist and Bayesian statistical methods. First, we simulate the data of future direct detection experiments assuming a momentum/velocity dependent dark matter-nucleon interaction, and an anisotropic dark matter velocity distribution. Then, we fit a constant scattering cross section, and an isotropic Maxwell-Boltzmann velocity distribution to the simulated data, thereby introducing a bias in the analysis. The best fit values of the dark matter particle mass differ from their benchmark values up to 2 standard deviations. The best fit values of the dark matter-nucleon coupling constant differ from their benchmark values up to several standard deviations. We conclude that common assumptions in dark matter direct detection are a source of potentially significant bias.