Neutron background in large-scale xenon detectors for dark matter searches

Abstract

Simulations of the neutron background for future large-scale particle dark
matter detectors are presented. Neutrons were generated in rock and detector
elements via spontaneous fission and (alpha,n) reactions, and by cosmic-ray
muons. The simulation techniques and results are discussed in the context of
the expected sensitivity of a generic liquid xenon dark matter detector.
Methods of neutron background suppression are investigated. A sensitivity of
$10^{-9}-10^{-10}$ pb to WIMP-nucleon interactions can be achieved by a
tonne-scale detector.

Full text

Page 1

arXiv:hep-ex/0404042v1 30 Apr 2004
Neutron background in large-scale xenon detectors
for dark matter searches
M. J. Carson, J. C. Davies, E. Daw, R. J. Hollingworth, V. A.
Kudryavtsev1, T. B. Lawson, P. K. Lightfoot, J. E. McMillan, B.
Morgan, S. M. Paling, M. Robinson, N. J. C. Spooner2, D. R. Tovey
Department of Physics and Astronomy, University of Sheffield, Sheffield S3 7RH, UK
Abstract
Simulations of the neutron background for future large-scale particle dark
matter detectors are presented. Neutrons were generated in rock and detector
elements via spontaneous fission and (α,n) reactions, and by cosmic-ray muons.
The simulation techniques and results are discussed in the context of the expected
sensitivity of a generic liquid xenon dark matter detector. Methods of neutron
background suppression are investigated. A sensitivity of 10−9− 10−10pb to
WIMP-nucleon interactions can be achieved by a tonne-scale detector.
Keywords: Dark matter, WIMPs, Neutron background, Neutron flux, Spontaneous fis-
sion, (α,n) reactions, Radioactivity, Cosmic-ray muons underground, Photomultipliers
PACS: 14.20.Dh, 14.80.Ly, 13.60.Rj, 13.75.-n, 13.85.-t, 28.20, 25.40, 98.70.Vc
Corresponding authors: V. A. Kudryavtsev, N. J. C. Spooner, Department of Physics
and Astronomy, University of Sheffield, Hicks Building, Hounsfield Road, Sheffield S3
7RH, UK
Tel: +44 (0)114 2224531;Fax: +44 (0)114 2728079;
E-mail: v.kudryavtsev@sheffield.ac.uk, n.spooner@sheffield.ac.uk
1Corresponding author, e-mail: v.kudryavtsev@sheffield.ac.uk
2Corresponding author, e-mail: n.spooner@sheffield.ac.uk
1

Page 2

1. Introduction
Future dark matter experiments planning to reach a sensitivity of 10−9− 10−10pb
to the WIMP-nucleon cross-section require a very low background environment, so-
phisticated techniques capable of discriminating WIMP-induced events from all kinds
of background and at least one tonne of target mass to achieve sufficient counting
rate. Hereafter, the quoted sensitivity refers to the minimum of the sensitivity curve,
which occurs for WIMP masses in the range 40-80 GeV (see, for instance, Figure
13 for examples of the sensitivity or exclusion curves). Some individual background
events, however, are indistinguishable from expected WIMP scattering events. WIMPs
are expected to interact with ordinary matter in detectors to produce nuclear recoils,
which can be detected through ionisation, scintillation or phonons. Identical events
can be induced by single elastic scattering of neutrons. Thus, only suppression of any
background neutron flux by passive or active shielding and proper choice of detector
materials will allow experiments to reach sufficiently high sensitivity to WIMPs. De-
signing detectors, their shielding and active veto systems requires simulation of neutron
fluxes from various sources.
Neutrons underground arise from two sources: i) local radioactivity, and ii) cosmic-
ray muons. Neutrons associated with local radioactivity are produced mainly via (α,n)
reactions initiated by α-particles from U/Th traces in the rock and detector elements.
Neutrons from spontaneous fission of238U also contribute to the flux. The neutron yield
from cosmic-ray muons depends strongly on the depth of the underground laboratory.
The suppression of the muon flux by a large rock overburden will also reduce the
neutron flux but by a smaller factor.
In the present study neutrons associated with radioactivity in rock and detector
elements are treated separately. Although these neutrons come from similar reactions,
the materials in which they are produced are certainly different, as are the methods of
their suppression. Neutrons from surrounding rock can be easily suppressed by passive
hydrocarbon shielding, whereas the internal neutron flux can be reduced by choosing
ultra-low-background materials and possibly by using an active veto to reject events
in the detector in coincidence with veto signals.
At deep underground sites (3 km w.e. or more), the neutron production rate from
muons is about 3 orders of magnitude lower than the rate for neutrons arising from
rock activity, depending strongly both on the depth and the U/Th contamination. The
muon-induced neutron flux can be important, however, for experiments intending to
reach high sensitivity to WIMPs. There are several reasons for this: 1) the energy
spectrum of muon-induced neutrons is hard, extending to GeV energies, and fast neu-
trons can travel far from the associated muon track, reaching a detector from large
distances; 2) fast neutrons transfer larger energies to nuclear recoils making them vis-
ible in dark matter detectors, while many recoils from α-induced neutrons fall below
detector energy thresholds; 3) a detector can be protected against neutrons from the
rock activity by hydrocarbon material, possibly with the addition of a thermal neutron
absorber; such material, however, will also be a target for cosmic-ray muons.
This work includes, for the first time, a detailed Monte Carlo simulation of pro-
duction, propagation and detection of neutrons from known sources, investigation of
techniques for neutron flux suppression and studies of systematics associated with the
neutron background in connection with the sensitivity of a future detector to WIMP-
2

Page 3

nucleus interactions. The work is part of a programme of neutron background studies
for the dark matter experiments at Boulby mine (North Yorkshire, UK) (see Ref. [1]
for a review of dark matter searches at Boulby). Similar studies have been initiated for
dark matter experiments at Gran Sasso and Modane [2, 3]. The present simulations
were carried out for a large-scale xenon detector and are relevant to several programmes
around the world, including other potential large-scale dark matter detectors. A tonne-
scale xenon dark matter detector is planned for the Boulby Underground Laboratory
[4]. A similar detector has been proposed for a new underground laboratory in the
USA [5]. Another liquid xenon based detector is XMASS II [6] to be built in Japan for
solar neutrino, double-beta decay and dark matter searches. The double-beta decay
experiment EXO [7] will also be based on liquid xenon. The simulations presented here
are important for many detectors designed for rare event studies.
The paper is organised as follows. Generation of neutron spectra from (α,n) re-
actions is described in Section 2. Neutrons from rock and the required shielding are
discussed in Section 3. Simulations of muon-induced neutrons are presented in Section
4. Neutron background from detector elements is investigated in Section 5. Systematic
effects caused by neutron background in connection with the sensitivity of a large-scale
xenon detector to WIMP-nucleon cross-section are studied in Section 6. The summary
and conclusions are given in Section 7.
2. Neutron production by radioactive isotopes
Neutron production by radioactive isotopes in the decay chains of uranium and
thorium was calculated using the SOURCES code [8]. The main features of the code
are as follows. Spontaneous fission (of238U mainly) was simulated using a Watt spec-
trum [9]. Neutron fluxes and spectra from (α,n) reactions were obtained taking into
account the lifetimes of isotopes, energy spectra of alphas, cross-sections of reactions
as functions of alpha energy, branching ratios for transitions to different excited states,
stopping power of alphas in various media, and assuming isotropic emission of neutrons
in the centre-of-mass system.
SOURCES provides a treatment of (α,n) reactions only up to 6.5 MeV α-energies.
This is likely to restrict significantly the reliability of the results because the cross-
sections of (α,n) reactions rise with energy and the average neutron energy also in-
creases with the parent alpha energy. Hence, the 6.5 MeV cut reduces the total neu-
tron yield from (α,n) reactions and artificially shifts the neutron spectrum to lower
energies. The effect of the neutron spectrum shift can be significant. We tested this by
generating neutron spectra in NaCl with the original SOURCES code and by taking
a different spectrum from Ref. [10], calculated for the Modane rock. We then prop-
agated neutrons with both of these spectra through lead and various thicknesses of
hydrocarbon material (CH2) usually used to shield detectors from rock neutrons (see
Section 3 for details of propagation procedure), and compared the two results. Even if
both spectra are normalised to the same neutron production rate, there remains about
a two order of magnitude difference in the predicted neutron flux above 10-100 keV
after 30 cm of lead and 35 g/cm2of hydrocarbon, the spectrum from SOURCES giving
a lower rate because of the smaller neutron energies. The effect is mainly due to the
decrease in neutron-proton elastic scattering cross-section with energy. It became ob-
vious that the neutron production code had to be modified to provide a more realistic
3

Page 4

treatment of (α,n) reactions. Note that the neutron production was simulated in Ref.
[10] assuming a transition of the nucleus to the ground state only (this overestimates
the neutron energy) and an emission of a neutron at 90◦(which means that neutron
energy was directly calculated from the alpha energy and the neutron spectrum was a
delta-function).
The following modifications were made to SOURCES to overcome the 6.5 MeV
limit. Existing cross-sections were extended to 10 MeV, taking into account available
experimental data. For some materials, new cross-sections were added to the code.
For example, we added the cross-section for23Na measured up to 10 MeV [11] as an
alternative to those already present in the code library. The cross-section for35Cl was
not present initially in the code library and was added from Ref. [12]. The cross-
sections on Na and Cl were needed for calculation of neutron production in the salt
rock. If the cross-section for a material was measured or calculated for low energies
only, then it was extrapolated from low energies up to 10 MeV. The Nuclear Data
Services of the Nuclear Data Centre at the International Atomic Energy Agency [13]
were extensively used to obtain cross-sections.
The branching ratios for transitions to the ground and excited states above 6.5 MeV
were chosen to be the same as at 6.5 MeV. This resulted in a small overestimate of
neutron energies for alphas above 6.5 MeV, since the increased probability of transition
to the higher states was neglected, but the total neutron flux was not affected. The
uncertainties associated with the calculations of such probabilities were not negligible,
however. The differences as large as (20-30)% exist between different calculations of
the transition probabilities in the SOURCES library. If the excited levels were not in
the code library, as was the case for elements for which the cross-sections were absent
too, then in adding the cross-section we assumed that the transition was occuring to
the ground state only.
3. Neutrons from rock
Here, our main objective was to find the thickness of hydrocarbon shielding needed
to suppress the neutron flux from rock down to a level allowing the required sensitivity
to WIMP-nucleus interactions. We started with neutron production in rock, then we
propagated neutrons through the rock to the rock/cavern boundary and further on
through lead and hydrocarbon shielding to the detector. Finally we generated nuclear
recoils from neutrons in the xenon target within the detector.
Simulation of neutron production via spontaneous fission and (α,n) reactions in rock
was carried out with the modified SOURCES code (see Section 2). Rock was assumed
to be halite (NaCl), which is the case for the Boulby Underground Laboratory (UK)
and the Waste Isolation Pilot Plant at Carlsbad (USA), both being the proposed sites
for underground experiments. The contamination levels of radioactive elements in
rock vary from site to site and can vary also from hall to hall within an underground
laboratory. In these simulations they were taken as 60 ppb of U and 300 ppb of Th in
secular equilibrium. The energy spectrum of neutrons at production from SOURCES is
shown in Figure 1. The total neutron production rate was found to be about 1.05×10−7
cm−3s−1. The neutron energy spectra in NaCl obtained with SOURCES are similar
in shape for U and Th initiated neutrons (see Figure 1). This means that for other
contamination levels of U and Th the spectrum of neutrons (as well as the spectrum of
4

Page 5

nuclear recoils in a detector) can be scaled from that reported here taking into account
the difference in contamination levels. For equal concentrations in NaCl, uranium
gives roughly twice as many neutrons as thorium. For 100 ppb of uranium the neutron
production rate is equal to 5.2×10−8cm−3s−1, whereas for a 100 ppb of thorium it is
2.5×10−8cm−3s−1. Decreasing the Th level down to 150 ppb (50% of the basic value
used in the present work) results in a decrease of the neutron yield down to about 65%
of the initial value with only a tiny decrease of the mean neutron energy from 1.81
MeV down to 1.75 MeV.
Neutron propagation through the rock was simulated using the GEANT4 package
[14]. Neutrons from the rock wall were produced in a slab of rock 1 × 1 m2with 3 m
depth into the rock (simulations with varying rock thicknesses showed that only those
neutrons within 3 m of the rock surface are capable of reaching it). Neutrons from this
region were allowed to propagate isotropically into a much larger region (100×100 m2
also with 3 m depth). This avoids neutron losses due to rock edge effects. The total
spectrum observed from this region was then re-scaled to the original 1×1 m2surface
element. Parameters of neutrons reaching the rock/cavern boundary were stored and
neutrons were propagated later through lead and hydrocarbon shielding.
Neutrons were also simulated in a realistic cavern with a size of 30 × 6.5 × 4.5 m3
in the rock (see Figure 2). The neutron flux at the rock/cavern boundary for this
configuration was found to be 4.36 × 10−6cm−2s−1above 100 keV and 2.20 × 10−6
cm−2s−1above 1 MeV. In practice these values are affected by the back-scattering
of neutrons from other walls of the cavern: a neutron can enter the cavern, reach the
opposite wall and be scattered back into the cavern increasing the total flux through the
boundary. To check the effect of back-scattering, neutrons were propagated through
the cavern and counted each time they entered the cavern. In this case the calculated
neutron flux was 1.19×10−5cm−2s−1above 100 keV and 4.10×10−6cm−2s−1above
1 MeV. For a real detector the back-scattering of neutrons can occur also on detector
elements, shielding etc.
Lead of low radioactivity is widely used to shield dark matter detectors from gam-
mas produced in the rock and the laboratory walls. Some detectors, however, can be
made insensitive to these gammas [15]. So the use of lead as shielding and its thickness
is decided for each particular experiment. Normally the thickness is such that gammas
from rock contribute only a minor part to the total gamma flux at the detector, while
a major contribution comes from the detector itself. Neutrons produced via sponta-
neous fission in low activity lead do not contribute significantly to the neutron flux
coming from the rock. Simulations were carried out with and without lead shielding
to investigate the effect of lead on the neutron flux. Neutrons coming from the rock in
a simple geometry (neutron production volume – 1 × 1 × 3 m3, neutron propagation
volume – 100 × 100 × 3 m3as described above) were propagated through 30 cm of
lead and those emerging on the opposite side were stored. Note that neutrons can be
scattered back from the lead into the rock and then to the lead again. In order not to
lose these neutrons the rock was present at this stage of the simulations, although the
neutrons were generated only on its boundary. Then hydrocarbon shielding of varying
thicknesses was added to the set-up after lead. Similar simulations were carried out
without lead (with hydrocarbon only). Figure 3 shows neutron spectra after 30 cm of
lead and slabs of hydrocarbon of various thicknesses (Figure 3a), and similar spectra
obtained without lead (Figure 3b). Due to the high cross-section of inelastic neutron
5