Cosmic Ray Induced Neutron Production in a Lead Target

TL;DR

This study models cosmic ray muon-induced neutron production in lead using Geant4 and compares it with experimental data from underground detectors, revealing good agreement in neutron multiplicity distributions and identifying systematic differences.

Contribution

It provides the first detailed comparison of Geant4 simulations with extensive underground neutron multiplicity data at multiple depths.

Findings

Neutron multiplicity distributions follow a power law with consistent slope parameters.

Simulations and data agree on the power law slope within uncertainties.

Discrepancies in amplitude are within systematic errors.

Abstract

Underground experiments searching for rare events, such as interactions from dark matter, need to exhibit background as low as possible. One source of background is from cosmic ray muons and muon-induced neutron production. Presently these background are not fully understood. In this study Geant4 is used to model cosmic ray muon induced neutron multiplicity production and compare the modeling with data collected using an $^3$He instrumented Pb-target detector system. The neutron event multiplicity production is taken from the 2002 NMDS-II data sets, consisting of 6504 hrs collected at 583 m.w.e. and 1440 hrs, with the identical detector system, collected at 1166 m.w.e.. The detector consists of a 30 cm cube Pb-target surrounded by 60 $^3$He tubes. The single particle detection efficiency is 23.2\%$\pm$1.2\% calibrated using a $^{252}$Cf neutron source. The highest neutron multiplicity event, observed at 583 m.w.e. was 54 observed neutrons corresponding to $\sim$ 233 produced neutrons. The neutron multiplicity, n, distributions fit well a 2-parameter power law fit, $k\times n^{-p}$. For the Monte Carlo simulations at both depths and the data collected at both depths, all are consistent with a single slope parameter p. For the simulation at 583 m.w.e., p=2.37$\pm0.01$ and for the data collected at 583 m.w.e, p=2.36$\pm0.10$. At 1166 m.w.e., p=2.31$\pm0.01$ for the simulation, and for the data with only 6 detected events above multiplicity 5, p=$2.50 \pm 0.35$ predicted using a Maximum Likelihood Estimation method. At both depths, the power law amplitudes of the Geant4 simulations differ by a factor of 2 larger than the data sets. However, the disagreement is within the estimated systematic error of the simulations.