Journal of Physics G: Nuclear

and Particle Physics

TOPICAL REVIEW You may also like

- A next-generation liquid xenon
Direct detection of WIMP dark matter: concepts observatory for dark matter and neutrino
physics
and status J Aalbers, S S AbdusSalam, K Abe et al.

- Light weakly interacting massive particles

Graciela B Gelmini
To cite this article: Marc Schumann 2019 J. Phys. G: Nucl. Part. Phys. 46 103003
- THE WIMP CAPTURE PROCESS FOR
DARK STARS IN THE EARLY UNIVERSE
Sofia Sivertsson and Paolo Gondolo

View the article online for updates and enhancements.

This content was downloaded from IP address 175.203.23.131 on 12/11/2024 at 00:36

 Journal of Physics G: Nuclear and Particle Physics
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 (36pp) https://doi.org/10.1088/1361-6471/ab2ea5

Topical Review

Direct detection of WIMP dark matter:

concepts and status
Marc Schumann
Physikalisches Institut, Universität Freiburg, D-79104, Freiburg, Germany

E-mail: [email protected]

Received 7 March 2019, revised 14 June 2019

Accepted for publication 2 July 2019
Published 20 August 2019

Abstract
The existence of dark matter as evidenced by numerous indirect observations
is one of the most important indications that there must be physics beyond the
Standard Model of particle physics. This article reviews the concepts of direct
detection of dark matter in the form of Weakly Interacting Massive Particles in
ultra-sensitive detectors located in underground laboratories, discusses the
expected signatures, detector concepts, and how the stringent low-background
requirements are achieved. Finally, it summarizes the current status of the field
and provides an outlook on the years to come.

Keywords: dark matter, direct detection, weakly interacting massive particle,

low background physics, rare event search

(Some figures may appear in colour only in the online journal)

1. Introduction: dark matter and WIMPs

More than 95% of the matter and energy content of the Universe is dark [1], i.e. it does not (or
only very weakly) interact with photons and ordinary baryonic matter. A total of 25.9% is made
of dark matter, a not-yet-identified form of matter which builds large scale structures in the
Universe. (The remaining 69.1% are due to dark energy, responsible for the accelerated
expansion of the Universe.) Since dark matter is five times more abundant than baryonic matter
and thus dominated the evolution of the Universe from the end of the radiation era (at about
70 000 years after the big bang) until redshifts of z0.5 (corresponding to 9.4×109 y), it
essentially shaped the Universe that can be observed today.
However, as of today there is only indirect evidence for the existence of dark matter,
based on the observation of gravitational effects. A plethora of experiments aim at the direct
detection of dark matter by searching for signals from dark matter particles scattering off
Earth-based detectors. This article reviews the concepts of these direct detection experiments,

0954-3899/19/103003+36$33.00 © 2019 IOP Publishing Ltd Printed in the UK 1

J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

summarizes important aspects for the interpretation of their results and provides an up-to-date
status of the field1.
In this first section, we will briefly discuss why most scientists are convinced that there is
plenty of dark matter in the Universe, how it is distributed in galaxies such as the Milky Way,
and will introduce the weakly interacting massive particle (WIMP), one of the most promising
candidates for the dark matter particle. The basics of direct detection experiments are pre-
sented in section 2; a discussion of the field’s current status follows in section 3.

1.1. Evidence for dark matter

Even though the dark matter particle has not yet been detected directly and its properties
remain largely unknown, there is plenty of indirect evidence from astronomy and cosmology
that a large fraction of the matter in the Universe is dark. Since dark matter neither interacts
electromagnetically nor strongly all evidence is based on effects of gravity. That it may
interact with (sub-)weak-scale cross sections with ordinary matter an assumption, see
section 1.3. By now, dark matter was indirectly seen galactic, galaxy-cluster and cosmologic
scales. The amount of dark matter in the solar system is too small to induce observable effects
in the motion of planets or spacecrafts [2].
The pioneering observations were done by Oort [3] and Zwicky [4]. Oort studied the
velocities of stars in the solar neighborhood and found that these are too high to be explained
by the luminous mass in the Galaxy. Zwicky applied the virial theorem, which relates the
average kinetic and potential energies of a gravitationally bound system, to the Coma galaxy
cluster. By estimating the time and mass-averaged galaxy velocity ⟨v 2⟩ in the cluster via
redshift measurements, he realized that the cluster mass derived via the virial theorem was
much larger than the luminous mass in the cluster. Zwicky attributed the discrepancy to a new
form of ‘dark matter’ (‘Dunkle Materie’ in the article written in German). This early work,
however, did not have a significant impact in the scientific community. (A comprehensive
review on the history of dark matter can be found in [5].)
Since the mid 1960s and starting with Andromeda Galaxy (M31), Vera Rubin studied the
rotation curves of spiral galaxies with a sensitive spectrograph built by Kent Ford, focusing
on H-II regions of ionized atomic hydrogen at different distances to the Galaxy centers. She
discovered that the observed rotation velocity vr did not follow the Keplerian vr µ r - 2
1

decrease expected from the distribution of stars. Instead, all galaxies show a flat (or even
slightly increasing) velocity profile after an initial rise attributed to the central bulge [6]. The
findings were later confirmed by radio-observations using the 21 cm line of atomic hydrogen
which could show that vr remains constant even in regions far outside of the visible disk of
the Galaxy [7]. It is straight forward to show that the flat profile can be explained by a density
distribution ρ(r) ∝ r−2, which is a much more modest decrease than the observed
rs (r ) µ r -3.5 of the number density in the stellar halo of the Milky Way [8]. One can
conclude that all galaxies appear to be embedded in a large halo of dark matter.
Further evidence comes from gravitational lensing, where invisible dark matter clumps in
the foreground distort the images of luminous objects in the background (for a review, see
[9]). Among the plethora of lensing systems, a particular iconic example is the Bullet-Cluster
[10]: it consists of two galaxy clusters which have traversed themselves. The collision led to a
1
Please note that I do not even try to provide a complete list of references in this review; in many cases I decided to
refer to recent reviews which provide an up-to-date view on a specific topic, rather than to the original work(s). In
other cases I quote particular examples from a generally much longer list of possible references. I also focus mainly
on the experiments which are most relevant at the time of writing. This does not imply that a project was not very
important at an earlier stage.

2
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 1. Temperature power spectrum of the cosmic microwave background (CMB)

for a dark matter density contribution WCDM varying between 0.11 and 0.43 (blue lines).
All other input parameters of the model are kept constant. The dashed black lines
shows the best fit to the Planck data from the 2018 release [13]. Small (large)
multipoles ℓcorrespond to lange (small) angular scales: the main acoustic peak at
ℓ∼200 is at ∼1°; ℓ=1800 corresponds to ∼0.1°.

separation of the baryonic matter, which is dominated by hot x-ray emitting plasma and
accumulates between the clusters, and the dark matter, which is observed by gravitational
lensing. The dissipation-less dark mass component traces the distribution of the cluster’s
galaxies which also do not undergo collisions. The separation cannot be explained by a
modification of the laws of gravity and is thus a clear observational evidence for the existence
of dark matter. In its entirety, measurements of gravitational lensing indicate that dark matter
is about five times more abundant than ordinary matter and not distributed in massive
objects of astronomical size. It moves at non-relativistic velocities (‘cold’ dark matter). All
interactions other than gravity are at most very small; this also holds for dark matter
self-interactions [9].
At largest scales, evidence for dark matter comes from the distribution of structure in the
Universe [11, 12] as well as from the precision analysis of the cosmic microwave background
(CMB) [1, 13]. This almost perfect black body spectrum is a relic of the big bang, created
when photons decoupled from the plasma of electrons and protons (at T≈3000 K) such that
the Universe became optically transparent. The expansion of the Universe has cooled down
the radiation to TCMB = 2.725 K . Temperature anisotropies at the 10−5-level can be observed
in the CMB; they originate from ‘acoustic’ temperature and density fluctuations in the early
Universe and carry plenty of cosmological information. The power spectrum of the fluc-
tuations shows their strength versustheir angular scale. It can be very well described by the
six-parameter ΛCDM model, which takes into account dark energy as a cosmological con-
stant Λ and cold dark matter (CDM). Especially the third acoustic peak in the power spectrum
is very sensitive to the total amount of dark matter in the Universe, see figure 1. The precise
contributions of the different components to the energy density of the Universe as derived
from the measurements of the Planck satellite are WL = 0.6911 (62), WCDM = 0.2589 (57) and
Wbaryon = 0.0486 (10) [1]. The position of the first acoustic peak shows that the Universe’s
geometry is flat.

3
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

1.2. Dark matter distribution in the galaxy

Direct detection experiments search for events induced by particles from the Milky Way’s
dark matter halo. Their distribution and kinematics governs the expected signal and is thus
essential for the interpretation of the experimental results. The exact distribution could not yet
be measured, however, it can be modeled using the knowledge derived from the gravity-based
observations discussed in section 1.1.
The most simple model assumes that dark matter is a collisionless gas with an isotropic
initial velocity distribution. Its equation-of-state relating pressurep and matter densityρ is
given by
p (r ) = r (r ) ⟨(v - v )2 ⟩ = r (r ) s 2, ( 1)
with the velocity dispersionσ2. In hydrostatic equilibrium, the pressure of the dark matter gas
balances the gravitational pull towards the center of the dark matter halo and the density
profile can be calculated to be
s2
r (r ) = . ( 2)
2pGr 2
The r−2-dependence of this ‘isothermal sphere’ correctly leads to the observed flat galactic
rotation profiles (see section 1.1).

1.2.1. Standard halo model. The Standard Halo Model commonly used for the interpretation
of direct detection experiments assumes that the dark matter particles are distributed in an
isotropic isothermal sphere with a Maxwellian velocity distribution
 ⎛ 3∣v ∣2 ⎞
f (v ) = N exp ⎜ - ⎟, (3)
⎝ 2s 2 ⎠
which follows from the solution of the Boltzmann equation for collisionless particles, with the
normalization N∝v2. In contrast to the stellar galactic disk, the halo does not rotate. The
velocity dispersionσ(R) is related to the average circular velocity vc(R) at which objects at a
distance R orbit around the galactic center: s (R) = 3 2 vc (R). The local circular velocity at
the solar distance R0;8.0 kpc is vc(R0)=220 km s−1. (Note that the Sun has a peculiar

motion with respect to vc.) Formally, the velocity distribution f (v ) is infinite, however,
particles with velocities above the escape velocity vesc = 2f are not bound by the

gravitational potentialf of the Galaxy and f (v ) has to be truncated. By analyzing samples of
stars with the highest measured velocities one obtains vesc  544 km s−1, with a 90%
confidence range from 498 to 608 km s−1 [14].
The question whether the assumed Maxwell–Boltzmann velocity distribution of the
Standard Halo Model is an over-simplification can be studied with large N-body simulations
(for a review, see [15]). Starting from initial conditions given by the CMB fluctuations, these
simulations model the evolution of parts of the Universe taking into account only dark matter
(=only gravitational interactions) or dark matter with baryonic feedback (from supernovae
and radiative feedback from massive stars) and provide information on Milky Way-like dark
matter halos. Recent studies indicate that velocity distributions taken directly from high-
resolution hydrodynamical simulations lead to direct detection results that are very similar to
the ones based on the Standard Halo Model [16]. Another important conclusion from
simulations is that the dark matter distribution at the solar distance R0 is expected to be rather
smooth, without relevant substructure. However, it is worth mentioning that some recent
observations, such as the distribution and motion of satellite dwarf galaxies, are in conflict

4
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

with simulations which predict them to be uniformly distributed and moving in random
directions [17].

1.2.2. Dark matter density. The simple isothermal sphere, equation (2), shows a divergence
at the galactic center. The precise shape of the density profile for r → 0 is unknown and
different possibilities were derived from N-body simulations. Some solutions exhibit a flat
core, others are more cuspy. Recent studies of dwarf galaxies indicate that the ‘cuspyness’ of
the density profile depends on the star formation rate which drives fluctuations in the
gravitational potential: galaxies with longer lasting star formation have more shallow dark
matter cores [18].
Nevertheless, at the solar radius R0, all models basically agree, such that the knowledge
of the density profile in the Milky Way center is irrelevant for direct detection experiments. (It
is, however, crucial for the interpretation of results from indirect detection searches looking
for particle excesses due to dark matter annihilation in the galactic center; for a review, see
[19].) The local dark matter density ρ0 can be derived from the measured rotation curve of the
Milky Way (assuming spherical symmetry) or from the vertical kinematics and position of
stars in the solar neighborhood. The canonical value adopted for the interpretation of direct
detection experiments is ρ0=0.3 GeV/c2/cm3. However, the number has a rather large
uncertainty of ∼50% as individual measurements show considerable variations [20]. A recent
+0.07
measurement (2017) using Sloan Digital Sky Survey data yields r0 = 0.46- 0.09 GeV/c /cm
2 3

[21]. In the absence of a signal, it is more important for the direct detection community to
adopt one common value to compare the different experiments rather than adopting an
updated result: a change of ρ0 will simply shift an exclusion limit/detection claim up or down
in cross section by the same amount.

1.3. WIMPs

Dark matter forms gravitationally interacting structures and it is generally assumed that it is of
particle nature, similar to all ‘normal’ matter. Dark matter in the form of Massive Compact
Halo Objects (MACHOs), e.g. brown dwarfs, lonely planets, black holes populating the
galactic halo, was not found in microlensing surveys in numbers sufficient to explain the
required amount of dark matter [22, 23]. Primordial black holes (PBHs) produced in the very
early phase of the Universe before big bang nucleosynthesis (and thus not being part of the
5% budget of baryonic matter in the Universe) came again into the focus of interest after the
first observations of gravitational waves from the inspiral event of black holes with unex-
pected high masses of 20–30Me [24]. However, the total number of detected events is too
small to explain the full amount of dark matter; PBHs in this mass range could only constitute
∼1% [25]. Other constraints come from microlensing searches of stars [26, 27], typeIa
supernovae [28], or cosmic ray data from VoyagerI [29] and basically exclude PBHs as
being all of the dark matter in any mass range. The recent observation of gravitational waves
in coincidence with electromagnetic radiation [30] also placed tight constraints on the pro-
pagation speed of gravitational waves and thus allows ruling out various popular models
which aim at explaining the indications for the existence of dark matter, most notably the flat
rotation curves (see section 1.1), by modifying the laws of gravity.
Dark matter made from baryons can be excluded by measurements of the primordial
abundance of light elements produced in the big bang nucleosynthesis (2H, 3He, 4He, 7Li, for
a recent review see [31]) and by precision studies of the CMB power spectrum, which
constrains the baryonic matter content of the Universe to Ωb≈5% [1]. As the Standard
Model of Particle Physics does not contain a single suitable dark matter candidate it is

5
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 2. The predicted mass range for the various prominent dark matter particle
candidates covers many orders of magnitude. This review focuses on the Weakly
Interacting Massive Particle (WIMP); the low-mass region discussed later is indicated
by the dashed bar.

assumed that dark matter must be made of one (or more?) new particle(s) which has not been
directly detected yet. (The free-streaming length of the massive but very light-weight neu-
trinos would wash out the observed large-scale structure of galaxies in the Universe.)
There are many suggestions for ‘new physics’ particles that would solve the dark matter
puzzle, see figure 2. One important class of potential dark matter candidates are WIMPs. They
arise naturally in various theories beyond the Standard Model, e.g. as the lightest super-
symmetric particle in supersymmetric theories (LSP; in many models this is the neutralino χ)
[32–34], or as the lightest Kaluza–Klein particle (LKP) in theories with extra spacetime
dimensions [35]. The lightest particle in Little Higgs models is also a possible WIMP can-
didate [36] if—as required for almost every theory—there is some new quantum number
(‘parity’) which is conserved and prevents the decay of the WIMP candidate into lighter
Standard Model particles. It is important to stress that none of these models was proposed to
explain the dark matter problem—the dark matter candidate comes for free.
WIMPs constitute a rather model-independent generic class of dark matter candidates,
with masses in the 1–105 GeV/c2 range and interaction cross sections from 10−41 to
10−51 cm2 [37]. The well-studied supersymmetric models usually predict WIMPs with masses
around 100 GeV/c2, however, other models favor different mass ranges. Asymmetric Dark
Matter [38], for example, relates the observed matter-antimatter asymmetry in the Universe
with dark matter and predicts

mc » 5 m proton » 5 GeV c 2. (4)

The massive and electrically neutral WIMPs are assumed to be thermally produced in the
early universe and are now moving with non-relativistic velocities, which makes them the
prime candidate for ‘cold’ dark matter. (A summary of dark matter production can be found,
e.g. in [39].) The interesting observation that the thermal production of WIMPs with weak-
scale cross sections naturally leads to the correct relic dark matter abundance of WCDM » 0.25
after their freeze-out from the thermal plasma is usually referred to as the ‘WIMP miracle’.
The result is almost independent of the WIMP mass and serves to further motivate the case for
WIMPs.
While many alternative dark matter candidates exist—prominent examples are heavy
right-handed neutrinos [40] and axions or axions-like particles [41], see figure 2—the WIMP
still is the prime target for experimental searches. The remaining part of this article will thus
concentrate on the WIMP, which, despite of many years of effort, was not yet discovered.
However, many detectors with an improved sensitivity will get online in the near future.

6
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

2. Direct detection basics

Information about the dark matter particle can be inferred indirectly from astrophysical and
cosmological observations. Examples are limits on its self-interaction strength from colliding
galaxy clusters [42], such as the Bullet cluster [10], or constraints from large scale structures
[43] and from the Universe’s ionization history [44].
Three more direct approaches are pursued in order to eventually measure the dark matter
particle physics properties, such as mass, coupling and interaction cross section, with bar-
yonic matter: these are the production of dark matter particles at hadron colliders such as the
LHC [45], the detection of the decay products following dark matter annihilation processes in
regions of high dark matter densities (e.g. the galactic center, the Sun, dwarf galaxies) [19],
and the direct detection of WIMP-nucleus scattering processes in ultra-sensitive low-back-
ground experiments. In this section, we will review the physics of direct detection and discuss
the characteristics of the used detectors.

2.1. Rates, spectra and interactions

The possibility to directly detect dark matter particles in the form of WIMPs was first
discussed by Goodmann and Witten [46]. Since the WIMP carries no electric charge, in most
scenarios it will not interact with the atomic electrons but will instead elastically scatter off the
atomic nucleus. The momentum transfer gives rise to a nuclear recoil (NR) which might be
detectable. The discussion below largely follows [47].

2.1.1. Rate and number of events. The expected rate of WIMPs scattering off a target
nucleus of mass mN is given by
dR r M vesc ds
dE nr
= 0
mN m c òv min
vf (v)
dE nr
dv. (5)

Enr is the NR energy, mχ denotes the WIMP mass, σ is the scattering cross section. M is the
target mass of the detector. The astrophysical parameters describe the WIMP distribution in
the Milky Way (see also section 1.2): f (v) is the normalized WIMP velocity distribution and
ρ0=0.3 GeV/c2/cm3 the local dark matter density. All velocities are defined in the
detector’s reference frame. The minimal velocity required for a WIMP to induce an NR of
energy Enr is

E nr mN (mN + mc )2 E nr mN 1
vmin = = . (6)
2 ( mN m c ) 2 2 m2
WIMPs with a velocity above the escape velocity vesc = 544 km s−1 [14] will not be bound to
the potential well of the Milky Way. μdenotes the reduced mass of the nucleon-WIMP
system.
The observed number of events in an experiment running for a time T is obtained by
integrating equation (5) from the threshold energy Elow to the upper boundary Ehigh :
E high dR
N=T òE low
dE nr  (E nr )
dE nr
, ( 7)

with the (typically energy-dependent) detector efficiencyò. The maximum recoil energy
Ehigh = 2m2vesc
2
mN-1 is given by kinematics and much less relevant than the energy threshold
Elow as the differential rate, equation (5), is eventually described by a simple steeply falling
exponential function [47] (see figure 3)

7
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 3. Nuclear recoil spectra induced by a mχ=100 GeV/c2 WIMP in various

common target materials, assuming a spin-independent WIMP-nucleon cross section
sSI = 1 ´ 10-47 cm2. The rate for spin-independent interactions increases with A2, i.e.
prefers heavy target materials. On the other hand, the rate for large nuclei (Xe, I, W)
decreases at higher energies due to the form factor suppression.

dR ⎛ E 4mc mN ⎞
µ exp ⎜ - nr ⎟, (8)
dE nr ⎝ E 0 ( m c + mN ) 2 ⎠
with E0 being the most probable kinetic energy of the incident WIMP and typical recoil
energies of  (10) keV nr only. The NR energy is given in keV nr (NR equivalent), which is
different from the electronic recoil scale (keV ee ) due to quenching effects caused by the
different energy-loss mechanisms (see section 2.2).

2.1.2. Annual modulation and directionality. The relevant velocity for the WIMP-nucleon
scattering in an Earth-based detector is not only the local circular velocity vc=220 km s−1
(as discussed in section 1.2), but it is the one of the Earth moving through the dark matter halo
with
vE = v + vÅ cos (q ) cos [w (t - t0)] , (9)
where ve=vc+12 km s−1 describes the motion of the Sun with respect to vc and v⊕=
30 km s−1 is the speed of the Earth orbiting around the Sun. θ≈60° is the inclination angle
measured between the Earth’s orbit and the galactic plane. The angular frequency ω=2π/T
is defined by T=1 y and the phase is fixed to t0=2 June, when ve and v⊕ add up
maximally. This detail leads to two observational consequences, which both are used in
experiments.
The different mean incident WIMP velocity in summer (ve and v⊕ parallel) and in winter
(ve and v⊕ anti-parallel) leads to a harder or softer WIMP spectrum, respectively, see figure 4.
Given a constant detector threshold, the number of observed dark matter signal events S(t)
will thus modulate over the course of the year [48]:
S (t ) = B + S0 + Sm cos [w (t - t0)]. (10)
The modulated part of the signalSm is rather small,  (vÅ vc ) ~ 5%, and most of the signal is
unmodulatedS0. The same is expected for detector-related backgrounds B, which are
typically much larger than S0. However, it is known that some possible background sources

8
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 4. Illustration of the Sun–Earth system moving around the galactic center and
through the dark matter halo in the direction of the constellation Cygnus. The varying
vector addition of the velocities over the course of a year is expected to induce a
modulating dark matter signature.

also show a seasonal modulation, e.g. the atmospheric neutrino flux depends on the
temperature-dependent density of the atmosphere [49]. Modifications to the simple dark
matter distribution model, e.g. dark matter streams etc would significantly alter the expected
signal and while astrophysical uncertainties of the local dark matter distribution have
generally a rather small impact for ‘standard’ direct detection experiments, they get more
significant for such annual modulation searches [50]. A detection of dark matter based solely
on the modulation signature is only possible in a statistical way, i.e.  (10 4) signal events are
required.
An even stronger darkmatter-specific signature that can be used to reduce background
 
events is to exploit the fact that vc » v points at a specific direction in the sky, towards the
constellation of Cygnus. Thus the ‘WIMP wind’ is expected to come from this region of the
sky. The differential recoil rate depends on the anglef between the direction of the WIMP
and the direction of the NR measured in the frame of the Galaxy,

d2R ⎡ 2 (v cos f - v )2 ⎤
µ exp ⎢  min
⎥, (11)
dE nr df ⎣ 3v2 ⎦

and is sharply peaked towards the Sun’s direction of motion. The Earth’s daily rotation thus
constantly changes the signal direction observed in a detector. vmin was defined in
equation (6). As most backgrounds are expected to be uniformly distributed, or come from the
Sun as, e.g. solar neutrinos, a measurement of the track direction (in addition to the recoil
energy) could be used to distinguish a dark matter signal from background events [51]. The
additional directional information (of a large statistics sample) would allow the study of many
particle and astrophysics dark matter properties. The experimental challenge is that the track
length of keV-scale NRs is very short, r<1 mm, and hard to reconstruct even in low-
pressure gas detectors [52] (see section 2.3).

2.1.3. Spin-independent and spin-dependent interactions.

Because of its large de Broglie
wavelength, the WIMP interacts coherently with all nucleons in the target nucleus. The

9
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

WIMP-nucleus scattering cross section in equation (5) is velocity and recoil-energy

dependent and given by
ds m
= 2N 2 (sSI FSI2 (E nr ) + sSD FSD
2
(E nr )). (12)
dE nr 2v m
Since the interaction of WIMPs with baryonic matter is a priori unknown, the cross section
consists of two terms for spin-independent (SI, which can be described by a scalar
S ~ ccqq or vector V ~ cg mc q g mq effective 4-fermion Lagrangian) and spin-dependent
(SD, axial-vector A ~ cg mg5 c q g mg5 q ) couplings. At small momentum transfers
q = 2m n Enr all partial waves of the nucleons add up and the WIMP scatters coherently
off the entire nucleus. For higherq the WIMP’s de Broglie wavelength λ=h/q is reduced
and only part of the nucleus participates in the interaction. This loss of coherence is accounted
for by the finite form factors FSI and FSD, which are only relevant for heavy WIMP targets
such as Xe, I or W (see figure 3). A usual choice for the (spin-independent) form factor is the
analytical approximation of Helm [47, 53] using the dimensionless parameter x=qRN/ÿ:
3 j1 (x ) ⎛ (xs)2 ⎞
F (x ) = exp ⎜ - 2 ⎟. (13)
x ⎝ 2RN ⎠
j1 (x ) = sin (x ) x 2 - cos (x ) x is the spherical Bessel function of the first kind. This
expression is based on the Woods–Saxon potential describing the nucleus of atomic number
A, with a radius RN » 1.2 A1 3 fm and a skin-thickness s≈0.5 fm. (In general, the
parameters have to be precisely obtained for every nucleus.)
The SI cross section is given by

m2 ( f p Z + fn (A - Z ))
2
m2 2
sSI = sn = sn A. (14)
mn2 f n2 mn2
μ is the WIMP-nucleus reduced mass as defined in equation (6), and mn is the reduced mass of
the WIMP-nucleon system. To allow for the comparison of different target nuclei employed
in different experiments, the WIMP-nucleus cross section σ is usually converted to a WIMP-
nucleon cross section σn. fp and fn describe the WIMP coupling strength to protons and
neutrons, respectively. The second expression assumes fp=fn which eventually leads to an
A2 dependence of the cross section: this implies that heavier target nuclei are expected
observe higher event rates, however, at the same time the expected recoil energy Enr is
smaller which poses bigger challenges on the detector’s thresholds.
The differential cross section for SD interactions, where the WIMP is assumed to be a
(Dirac or Majorana) fermion coupling to unpaired nuclear spinsJ, reads [54]

dsSD 8GF2 2 J + 1 S (∣q ∣) ,
 = [ a ⟨
p pS ⟩ + a ⟨
n nS ⟩] (15)
d∣q ∣2 pv 2 J S (0)

with the momentum transfer q . ⟨Sp⟩ and ⟨Sn⟩ are the expectation values of the total spin
operators for protons and neutrons in the the target nucleus which have to be calculated and
show some model dependence [55, 56]. There is no A2-scaling from coherence effects as in
equation (14), instead the cross section depends on the total nuclear spin J of the target

nucleus as well as its spin-structure function S (∣q ∣). While heavy nuclei are typically more
sensitive to SI-interactions, the situation is different for SD-scattering, as only the nucleus’
spin structure is relevant: nuclei without unpaired spins, e.g. argon targets which consist of
18 Ar, 18 Ar and 18 Ar, are ‘blind’ while the nuclear spin-structure of the rather light nucleus 9 F
40 38 36 19

makes it an excellent target to probe SD-interactions. Neutrons and protons typically

10
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Table 1. Relative strength of different isotopes used as targets in direct detection

experiments for spin-dependent WIMP searches. It is normalized the 19F, the target
with the highest sensitivity to spin-dependent coupling to protons. The expectation
values ⟨Sp, n⟩ to calculate the relative strength are taken from [57].

Isotope Z Abundance Spin Unpaired nucleon Relative strength

19
F 9 100.0% 1/2 Proton 100
23
Na 11 100.0% 3/2 Proton 12
27
Al 13 100.0% 5/2 Proton 22
29
Si 14 4.7% 1/2 Neutron 11
73
Ge 32 7.7% 9/2 Neutron 34
127
I 53 100.0% 5/2 Proton 24
129
Xe 54 26.4% 1/2 Neutron 47
131
Xe 54 21.2% 3/2 Neutron 18

contribute differently to the total spin of the target such that the SD-results are commonly
quoted assuming that WIMPs couple either only to neutrons (coupling strength ap=0) or to
protons (an = 0): the above-mentioned 199 F is very sensitive to spin-dependent WIMP-proton
73
scattering while 32 Ge, 129 131
54 Xe and 54 Xe are sensitive to spin-dependent WIMP-neutron
interactions. The expression [ap ⟨Sp⟩ + a n ⟨Sn⟩]2 (J + 1) J in equation (15) allows defining an
isotope-specific sensitivity factor for SD interactions. Its relative strength for various isotopes
used as WIMP targets is given in table 1, assuming a coupling to the unpaired nucleon only.
Chiral two-body currents lead also to a sensitivity to SD-interactions with an even number of
nucleons, however at a reduced sensitivity [57].
The ‘traditional’ analyses in terms of SI and SD WIMP-nucleon (χ-N) interactions are a
subset of more general effective field theories (EFTs) [58–60] which construct a large number
of relativistic and non-relativistic operators to describe the various possible 4-point-
interactions, assuming a heavy mediator particle. EFTs provide a possibility to directly
compare results from direct detection and collider searches, however, the heavy mediator
assumption has important consequences for the interpretation of LHC searches as one needs
to ensure that the kinematics of the WIMP production process is described accurately [61].
This problem is overcome by more complex simplified models which replace the 4-point-
interaction by s- or t-channel exchange of the mediator. The SI and SD interactions
  are mainly
described by the non-relativistic EFT operators 1 = 1c 1N and 4 = Sc · SN , respectively,
however, the experimental result of a WIMP search can be interpreted in a plethora of
operators [62]; some of these would produce very different recoil spectra than expected for
the ‘standard’ SI case.
The shape of the expected recoil spectrum is determined by the interaction model and
kinematics (e.g. f (v), mN, mχ for SI scattering). If the local dark matter densityρ0 is fixed, the
expected total rate depends on the scattering cross section. SI rates of 3 events/ton/year
have already been excluded which corresponds to WIMP-nucleon cross sections
sn  4 ´ 10-47 cm2 for mχ∼30 GeV/c2 [63]. In order to be sensitive to such small SI
cross sections, the optimal WIMP detector should have a large total mass M, a high mass
number A, a low energy threshold Elow, an ultra-low background and the ability to distinguish
between (NR) signal and (electronic recoil) background events. To focus on a special part of
the parameter space or different interaction models, as well as practical reasons might require
compromises or departures from these general rules: e.g. very light targets are favorable for SI
interactions of low-mass WIMPs. Various detector designs are discussed in the next section.

11
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

2.2. Signal generation in WIMP detectors

The total energy loss of a recoil in a WIMP detector can be described by

⎛ dE ⎞ ⎛ dE ⎞ ⎛ dE ⎞
⎜ ⎟ =⎜ ⎟ +⎜ ⎟ . (16)
⎝ dx ⎠tot ⎝ dx ⎠elec ⎝ dx ⎠nucl

The majority of the NR energy induced by a WIMP-nucleus interaction is typically lost to

dE
heat, i.e. to nuclear energy losses dx ( )
nucl
leading to atomic motion. In solid materials this
gives rise to phonons. Smaller amounts of electronic energy losses are available to excite or
ionize the target atoms. In some materials, the atomic excitation leads to scintillation light of a
narrow emission spectrum, which can be detected by means of sensitive photosensors.
Because of the smallness of the signal, the number of physical quanta (phonons, electrons,
photons) available for detection is generally rather low.

2.2.1. Signal quenching. An experimental complication comes from the fact that the
expected NR signal of energy Enr is smaller than the same amount of energy deposited by an
electronic recoil signal Eee . This phenomenon is called signal quenching and is due to the
different energy-loss mechanisms of the two recoil types. In case of low-energy NRs,
significantly more energy is lost to atomic motion (heat), which is often not detected. This
leads to considerably smaller scintillation and ionization signals and makes their detection
even more challenging. Signal quenching is typically energy dependent and has to be
measured accurately in order to establish an energy scale for the detector. It is an intrinsic
feature of the detection material and independent from the actually used detector (if detector-
specific effects such as signal collection efficiency and thresholds are properly accounted for);
this allows for quenching factor measurements in detectors specifically designed for this
purpose. One usually distinguishes between absolute quenching factors Q (Enr ), which
directly compare the reconstructed energy scales for electronic (Eee , measured in electronic
recoil equivalent energy, keV ee ) and NRs (Enr , measured in NR equivalent energy, keV nr ) in
an energy-dependent fashion,
Eee [keVee] = Q (E nr ) ´ E nr [keVnr] , (17)
and relative quenching factors Q¢ (Enr ), see figure 5. The latter describe the observed (light or
charge) signal from an NR interaction relative to a conveniently chosen electronic recoil
calibration standard, e.g. the Eee = 122 keV γ-line from 57Co or the 31 keV conversion
electron line from 83mKr:
S (E nr )
Q¢ (E nr ) = . (18)
S (Eee = fixed)
An important example for a relative quenching factor is the relative scintillation efficiency
eff of liquid noble gases (Ar, Xe): it describes how a scintillation light signal from an NR
energy deposition Enr in the liquid is reduced compared to a calibration standard (e.g. the
122 keV line).
For some materials such as Ge or Sicrystals, signal quenching can be reliably calculated
using the Lindhard theory [64] which employs some basic assumptions on the motion of the
recoiling nucleus in the material (e.g. mainly elastic nuclear collisions at low E). For other
materials, e.g. Xe and Ar, the results are much worse and semi-analytical approaches are used
instead [65, 66], see figure 5.
Detectors which have the capability to actually measure the heat deposited in the target
(see cryogenic detectors below) do not suffer from this complication; the heat measurement

12
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 5. Examples for quenching factors. (Left) The quenching factor describes by
how much the ionization signal from an NR energy deposition in germanium at 77 K is
reduced compared to an ER of the same energy. Shown is a best-fit Lindhard model to
experimental data [158]. (Right) Light and charge yield (in number of photons or
electrons per keV nr ) in a liquid xenon (LXe) TPC operated with a drift field of
Ed=200 V cm−1. The light yield can also be expressed via a relative quenching factor
(right axis, often referred to as eff ), here normalized to the light yield of a 31.2 keV
ER signal at Ed=0 V cm−1. Data from [65]. In both cases, the data exhibit
uncertainties which are not shown: around 1 keV nr , the individual measurements have
relative uncertainties around ±1% and ±20%–30% for germanium and xenon,
respectively. The systematic differences between measurements can be even larger.

allows for a straightforward definition of the energy scale. The calibration can be done using
artificially induced signals from a heater.

2.3. Detectors for WIMP searches

Several detector designs exploiting various target materials exist to detect one or two of the
signal channels. (To date, no attempt to measure all three channels simultaneously in one
detector was successful.) They can be put into the following categories.

2.3.1. Anorganic crystal detectors. The very first experimental attempt to directly detect
WIMPs was realized shortly after the Goodmann and Witten suggestion [46]. The experiment
[67] used a 0.72 kg high purity Germanium (HPGe) crystal to search for dark matter-induced
charge signals. Only a very small amount of energy is required to create and electron–hole
pair in such semiconductor detectors (Ge: 2.9 eV, Si: 3.6 eV) which leads to many signal
carriers and thus a very good energy resolution. However, the signals are rather slow
(τ∼1 μs) and the large capacitance of Gediodes beyond the  (1)kg-scale, which leads to
high electronic noise levels, does not allow to build very massive detectors. Current state-of-
the-art experiments use kg-scale p-type point contact HPGe crystals, which are able to
achieve very low thresholds down to ∼160 eV ee [68]. Background events from the large n+
surface can be reduced based on their longer rise times compared to bulk events, which is
caused by the weaker electric fields next to the n+ region [69]. Some projects use Si crystals
(Z=14) in the form of CCD-chips to optimize their sensitivity to WIMPs at lower
masses [70, 71].
In order to increase the target mass a different type of experiments uses arrays of high-
purity scintillator crystals, mainly NaI(Tl) but also CsI(Tl). Their advantage is a rather simple
detector design which can be operated stably for long periods of time: an array of kg-scale
crystals making up a large target mass, read by low-background photomultipliers (see

13
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 6. (Left) The detector principle of scintillating crystals searching for an annually
modulating signals is rather simple. However, extremely low intrinsic background
levels are required. (Right) Cryogenic detectors are cooled down to mK-temperatures
and weakly coupled to a thermal bath. Measured observables are heat (in form of
phonons) and ionization (or scintillation light).

figure 6, left). The high mass numbers ofI (A=127) and Cs (A=133) lead to a high
sensitivity to spin-independent interactions. The shortcomings of this detector design are a
comparatively high intrinsic background level (the long-standing world-record for NaI-
crystals is ∼1 count/kg/d/keV [72]) which is enhanced by the inability to reduce
backgrounds effectively by fiducialization or NR/ER discrimination (see section 2.4).
Therefore, experiments based on this technology mainly search for a possible dark matter-
induced annually modulating low-energy signal above a constant contribution from
backgrounds and the much larger non-modulating WIMP signal fraction. An event-by-event
detection of dark matter candidates is not possible. Typical thresholds are 2 keV ee
(corresponding to 8 keV nr , 12 keV nr and 22 keV nr for Na, Cs and I, respectively) while
results with 1 keV ee have been published by DAMA/LIBRA recently [72].

2.3.2. Cryogenic detectors. Crystalline detectors allow for the detection of the heat signal in
form of phonons by measuring the temperature increase following a particle interaction. The
sensitivity σ to record the temperature increase from an energy deposition E is given by
s 2 = c1 kT ´ (T C + c2 E ) , (19)
where T is the operation temperature, C the detector’s (T-dependent) heat capacity and
c1,2 »  (1) are detector-specific constants describing the thermal coupling of the detector to
the heat bath, readout noise, etc. To optimize the sensitivity, i.e. to reduce σ2 it is thus
mandatory to operate the detector at cryogenic temperatures, typically „50 mK, and to reduce
the heat capacity C. Dielectric crystals, such as Ge or Si, are particularly well-suited for
cryogenic operation as their heat capacity is given by C∝M×T3 below their Debye
temperature which is well above room temperature for Ge and Si. Mdenotes the mass of the
detector. One possibility to detect the tiny temperature rise ΔT are transition edge sensors:
they consist of thin wires (e.g. made from W) which are operated at the transition temperature
between their super-conducting and the normal-conducting state. A small temperature
increase will have a big impact on the resistivity of the wire and thus the current running
through it. An alternative detector-type are neutron transmutation doped germanium
thermistors whose resistivity strongly depends on the temperature [73].
It was demonstrated already in the90s that the simultaneous measurement of the
ionization signal allows for signal/background discrimination in the WIMP search as the
partition of the signal into the two channels depends on the recoil type [74]. All modern

14
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

cryogenic experiments follow this two-channel approach, where the second channel can be
ionization or scintillation, see figure 6 (right).
The main advantages of cryogenic detectors are their precise energy measurement with
almost no quenching in the heat channel, their excellent energy resolution and the possibility
to exploit two detection channels for excellent background rejection. However, the rejection
breaks down at low energies around 5 keV nr , where the distributions start to overlap. The
operation at mK-temperatures, typically using dilution refrigerators, is challenging and
expensive and the mass of a single detector is limited to the kg-scale due to the requirement of
a small heat capacity. This is overcome by using arrays of quasi-identical crystals, however,
the surface-to-volume ratio of such experiment is not optimal and surface contamination has
to be rejected. This is achieved by, e.g. optimizing the shape of the electric field across the
crystals to distribute the ionization signal of surface events on the segmented readout
electrodes differently to the signals from events happening in the bulk. This allows for the
definition of an inner fiducial target [75]. Rejection of surface events based on the rise-time of
the phonon signal is employed as well: acoustic phonon pulses from events close to the
detector surface rise faster than pulses from the bulk since impurities, defects etcon the
surface enhance the conversion of the initially created optical phonons to the detected low-
frequency acoustic phonons [76].
In the recent years, the sensitivity of cryogenic HPGe detectors to low-mass WIMP
(5 GeV/c2) has been extended by operating the crystals at a relatively high bias voltage
Vbias. This leads to a conversion of the charge signal into heat by the production of Neganov–
Luke phonons [77] during the transport of the ne ionization electrons (and the corresponding
holes). These phonons are detected on top of the primary heat signal Eheat from the primary
radiation,
Edep = E heat + n e e Vbias, (20)
what allows reaching very low thresholds down to ∼50 eVee. Since the initial ionization
signal is not detected the ionization yield cannot be used for background rejection. However,
the pulse shape carries some information on the position of the interaction in the detector.

2.3.3. Noble liquid detectors. The noble gases argon and xenon are excellent scintillators and
can be ionized easily. Krypton has similar properties but is not used for dark matter searches
because of its high intrinsic background from long-lived isotopes. Neon has a low mass
number and was so for only used in gaseous state to search for low-mass WIMPs [78]. Both,
argon and xenon can be liquefied to build dense and compact dark matter targets, with boiling
points of 87.2 K (−186.0 °C) and 165.2 K (−108.0 °C), respectively. Both boiling points lie
conveniently above the temperature of liquid nitrogen.
A particle interaction in the liquid noble gas target produces heat (which is not detected)
as well as excited X* and ionized atoms X+. The X* combines with neutral atomsX forming
excimer states X*2 , which subsequently decay under the emission of ultraviolet light:
+X
X* ⟶ X*2  2 X + hn. (21)
The photons have a wavelength of 128 and 178 nm for argon and xenon, respectively. While
sensitive photocathodes exist for the xenon scintillation light, this is not the case for argon.
Before being detected by photosensors, the light thus needs to be shifted to optical
wavelengths using a wavelength shifter such as tetraphenyl butadiene (TPB). This fluorescent
material emits blue light at 425 nm.
Depending on the relative orientation of the two atomic spins in the excimer X2*, the
scintillation light is emitted with different decay constants. With 4 and 22 ns, the singlet and

15
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 7. Detectors using the liquid noble gases argon or xenon as WIMP target either
(left) measure only the primary scintillation signal (single phase detectors) or (right)
detect the primary scintillation light as well as the ionization signal in a dual-phase time
projection chamber (TPC). To achieve the lowest possible threshold, the ionization-
induced charges are usually converted into a light signal via proportional scintillation
after extracting the electrons into the gas phase above the liquid target.

triplet state lifetimes are very similar for xenon, however, they differ by almost three orders of
magnitude for argon (7 ns and 1.6 μs). Since the relative population of singlet or triplet states
depends on the ionization density and hence the interaction type, this leads to very different
scintillation pulse shapes in argon and allows for the very efficient rejection of electronic recoil
background to the 10−8 level (see section 2.4). As natural argon gas contains the radioactive
β-decaying isotope 39Ar (T1/2=269 y) at a level of ∼1 Bq kg−1, the high rejection is
mandatory for argon-based WIMP searches. The use of 39Ar-depleted argon from underground
sources reduces this problem by more than three orders of magnitude [79], however, it
significantly increases the price of the gas. Since the Earth’s atmosphere contains only a fraction
of 9×10−8 of xenon it is also rather costly, however, it does not contain any long-lived
radioactive isotope besides the double-beta decaying 136Xe (T1/2=2.12×1021 y).
The ions X+ form singly ionized molecules X+ 2 with neutral atoms. In a time projection
chamber (TPC) a charge measurement is possible by removing  the ionization electrons from
the interaction site by means of a strong electric field Ed . Otherwise the electrons will
recombine forming excited xenon atoms, which eventually will again decay via the emission
of scintillation light:
+2X + e-
X+ + e- ⟶ X +
2 + X ⟶ 2 X + X**  2 X + X* + heat ⟶ (21) 4 X + hn . (22)

The recombination fraction depends on the strength of the electric field Ed . Scintillation and
the ionization signal are anti-correlated: a simultaneous measurement improves the energy
resolution of a detector [80].
Two types of detectors filled with a liquid noble gas target are employed and ton-scale
experiments of both types are being operated: single-phase detectors measure solely the
scintillation signal. Their spherical noble-liquid target is surrounded by photomultipliers in a
4π geometry to achieve a very high light yield, see figure 7 (left). While single-phase argon
detectors employ pulse-shape discrimination background reduction in xenon detectors is
limited to target fiducialization, i.e. defining a clean inner part of the active volume. The
interaction position is defined based on photon timing and the signal distribution in the
photosensors with a typical position resolution of a fewcm.

16
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Dual-phase TPCs also record the primary scintillation signal (usually called S1) by
means of two arrays of photomultiplier tubes (PMTs) installed above and below a cylindrical
noble-liquid dark matter target, see figure 7 (right). The  upper PMT array is placed in the
gaseous phase above the liquid. The electric drift field (Ed =0.1–1.0 kV cm−1) created across
the liquid target by wire or mesh electrodes removes the ionization electrons from the
interaction site and drifts them towards the gas phase on top of the detector. A second field
(‘extraction field’, ∼10 kV cm−1) across the liquid-gas interface pulls the electrons into the
gas phase where they produce a secondary scintillation signal (S2) by collisions with the gas
atoms. The S2signal is proportional to the number of electrons and measured with the PMTs
as well. The simultaneous measurement of both signals allows for identification of multiple
scatters and the reconstruction of the location of the interaction with mm-precision. Thanks to
the high target densities (r Xe = 2.94 g cm−3, rAr = 1.40 g cm−3 at their boiling points)
backgrounds are efficiently reduced by target fiducialization. The partition into excitation and
ionization depends on the ionization density and the ratio S2/S1 can thus be used to
distinguish electronic from NRs. Compared to other methods, the rejection levels are only fair
reaching ∼1 × 10−3 at 30% NR acceptance. Liquid argon-based detectors are thus mainly
relying on background rejection based on the pulse-shape of the scintillation signalS1.
The detector threshold of dual-phase TPCs is determined by the scintillation light signal
and can be greatly reduced in a charge-only (‘S2-only’) dark matter search [81]. Here, no
detection of an S1signal is required and threshold around 1 keV nr can be achieved. However,
backgrounds are significantly higher since one loses the possibility to reject electronic recoils
and the z-position of the event which reduces the power of fiducialization. The same data can
also be analyzed in terms of sub-GeV dark matter scattering off electrons [82], see section 3.3.

2.3.4. Bubble chambers. Superheated liquids, usually refrigerants such as CF3I, C3F8,
C4F10, C2ClF5 or C3ClF8, are used as WIMP target in bubble chambers. The liquids are kept
at a temperature just below their boiling point. A sufficient energy deposition into a certain
micro-volume will lead to a local phase transition of the superheated liquid and start the
formation of a bubble. The probability for bubble formation depends on the specific energy
loss dE/dx of the recoiling particle and can be tuned such that only NR events from
α-particles, neutrons or WIMPs can create bubbles. The detector is then almost immune to the
usually dominant electronic recoil background sources (γ, β) which can be rejected by a
factor <10−9. α-particles can be rejected based on the detailed acoustic characteristics of the
bubble ‘explosion’ [83].
After each event, the bubble chamber has to be compressed to remove the bubble and
decompressed to bring the liquid back into the superheated state. This involves mechanical
motion and induces a rather long detector deadtime and complicated detector calibration. As
in the early ages of particle physics, the (typically cylindrical) bubble chambers are read out
by means of cameras, which are usually arranged in a stereoscopic way around the target, see
figure 8 (left). The image allows the 3D reconstruction of the event position with mm-
resolution as well as measuring the multiplicity of the event.
The direct reconstruction of the recoil energy is not possible in bubble chambers as every
energy deposition above the nucleation threshold will lead to a bubble. They are thus operated
as threshold detectors measuring the integral rate above a certain minimal energy, which can
be adjusted by controlling the pressure and temperature of the used fluid. This is calculated
using the Seitz-model [84]; thresholds down to 3.3 keV were already achieved [85].
Superheated droplet detectors rely on the same detection principle. Their deadtime is reduced
compared to bubble chambers as the bubbles are trapped in a water-based polymer matrix

17
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 8. (Left) The expected signal from a WIMP interaction in a bubble chamber is a
single bubble in the superheated fluid which is recorded photographically. (Right)
Directional detectors, such as a TPC filled with a low-pressure gas and a high-
granularity readout at the anode as illustrated here, aim at detecting the direction of the
nuclear recoil signal.

[86]; however, bubble chambers were realized with much larger target masses at the 50 kg
scale [87].
The possibility to operate the bubble chamber with various target fluids of different
composition makes them very versatile instruments for the WIMP search. Almost all used
target fluids contain the isotope 19F, which has the highest sensitivity to spin-dependent
WIMP-proton couplings (see table 1 in section 2.1) and makes bubble chambers the leading
technology for this channel. The possibility to add the heavy element iodine (A≈127) offers
a good sensitivity to spin-independent interactions.

2.3.5. Directional detectors. The Sun is moving around the Galactic center such that the
average direction of the ‘WIMP wind’ through the solar system originates from the
constellation of Cygnus, see section 2.1. If the direction (or at least an head-tail asymmetry) of
NR events would be detected, a statistical discrimination of a WIMP signal from the
uniformly distributed background would be possible if ∼30 WIMP events are observed [88].
Since the Earth rotates once a day with respect to the expected direction of the WIMP flux, the
detectors either observe a daily modulation of the WIMP direction, or they need to be fixed on
a parallactic mount to compensate for the rotation. If detectors have track reconstruction
capabilities but integrate over long periods of time without following the Earth’s rotation, the
preferred WIMP direction is preserved, however, more than double the number of signal
events is required to detect dark matter [89].
As the NR’s track length depends on the target density, and a longer track facilitates the
reconstruction of the track direction, most directional detectors feature low-pressure gas
targets (∼40–100 mbar) with either photographic or a fine-granularity track readout in a TPC
geometry, see figure 8 (right). The large number of channels required to reconstruct the short
tracks makes the detector readout rather costly. The most common target gas used is CF4,
which provides sensitivity to spin-dependent WIMP-nucleon interactions; sometimes it is
used in mixtures with other gases (e.g. CS2, CHF3). The typical threshold of existing detectors
is around 20 keVee, while a competitive low threshold of 2 keVee has been achieved in the
MIMAC prototype [90]. MIMAC also demonstrated an angular resolution of 14°for
6.3 keVnr recoils [91]; the typical numbers achieved by other detectors are considerably
higher.

18
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

The most sensitive directional detector is the 1 m3 scale DRIFT-II TPC [92]. Its target
comprises 0.140 kg when filled with 55 mbar of a CS2 + CF4 + O2 mixture. Due to the
electronegativity of the CS2 gas, the electrons generated by a particle interaction create
negative ions which are drifted towards a multi-wire proportional chamber readout plane.
This significantly reduces diffusion and allows for fiducialization of the target. Electronic
recoil background can be rejected to high levels (2×10−7) based on their longer range and
lower ionization density. The achieved limit of 2.8×10−37 cm2 at mχ=100 GeV/c2 for
spin-dependent WIMP-proton couplings from a 7.66 kg × d run of DRIFT-II has to be
compared with the current best limit in this channel from PICO-60, a bubble chamber filled
with 52 kg of C3F8. An efficiency corrected exposure of 2571 kg × d led to a 3.5 orders of
magnitude more sensitive upper limit at the same mass [93]. Ignoring thresholds,
backgrounds and the different fraction of 19F in the two targets, and assuming that the
WIMP would have a cross-section at the current best upper limit, the directional detector
would need to be ∼104 times larger (or measure ∼104 times longer) to detect an excess of
∼30 signal events. The prospects for directional detectors are more promising at low WIMP
masses mχ5 GeV/c2, where the neutrino floor (see section 3.4) becomes relevant at cross
sections four orders of magnitude higher than for larger mχ: measuring the direction of an NR
provides the possibility to distinguish the neutrino background from the Sun from a WIMP
signal [94].

2.4. Backgrounds and background mitigation strategies

The search for WIMP dark matter, with its featureless exponentially falling recoil spectrum as
expected for spin-(in)dependent WIMP-nucleus scattering2, requires an extremely low
background in the energy region of interest (ROI). Usually, the design of the experiments
aims for ‘zero background’ (1 events in the ROI for the design exposure) such that the
observation of a few signal events would already have a high statistical significance. In this
section we will review the most common background sources and background mitigation
strategies.

2.4.1. Background sources. There are various possibilities to classify the sources of
background in dark matter experiments. We chose to organize them around the signature they
leave in the detector.

2.4.2. Electronic recoil background. The overall background of most current WIMP search
experiments is dominated by γ-backgrounds from the environment or the detector setup itself
and by β-particles at the surfaces or in the bulk of the detector. The particles generate
electronic recoils (ER) by electromagnetic interactions with the atomic electrons.
The backgrounds originate from long-lived natural radioisotopes such as the elements of
the primordial 238U and 232Th chains and 40K, but also from anthropogenic isotopes in the
environment such as 85Kr, 137Cs or 110mAg which stem from nuclear fuel element recycling
plants and accidents. 60Co is usually present in steel components; grade-1 titanium has often a
lower radioactivity level at only little weaker material strength [95]. Oxygen-free high-
conductivity copper is considered the best construction material in terms of background,
however, it is rather soft and has a limited mechanical strength.
2
Other interaction models predict different signatures, but no model expects a ‘smoking gun’-signature like a
monoenergetic line.

19
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Another source of β- and γ-backgrounds are isotopes in the material of the shielding or
the detector itself which were produced by cosmic ray activation in proton or neutron-capture
reactions. A prominent example with a very long half-life of T1/2=269 y is 39Ar which is
produced in the atmosphere by cosmic ray neutrons in the reaction 40Ar(n,2n)39Ar or
underground by muon (39K(μ−,νμ)39Ar) or neutron capture (39K(n,p)39Ar) on 39K [96]. In
other cases, the resulting isotopes are short-lived (T1/2<1 y) and underground-storage of the
materials prior to the dark matter search reduces their impact on the experiment. Nevertheless,
to build a dark matter experiment it is important to procure freshly produced materials (e.g.
Cu, Ar, Xe, Ge-crystals) with a low activation whenever possible and to optimize the
transportation of the materials to the laboratory. Air transport is often avoided as exposure at
high elevations is significantly worse than at sea level [97]. At the current scale and
background level, no experiment has to be worried about underground activation induced by
cosmogenic muons (which themselves are never a problem since they depot huge signals),
however, this potential background needs to be studied for the ultimate detectors (see
section 3.4).
A special background isotope is 222Rn3, a member of the 238U chain with the long-lived
mother 226Ra. As a noble gas, it is constantly emanated from all surfaces and thus present in
most experiments, especially if they use gas or liquid targets. Its α-decay deposits a huge
energy and is usually not contributing to the dark matter background directly, however,
β-decays of its daughter 214Pb to the ground state of 214Bi, which are not accompanied by a
prompt γ-emission, can lead to single scatter ER events in the WIMP search energy window.
In addition, radon plate-out on surfaces, i.e. the collection of α-decaying daughters, can lead
to background if most of the α-energy of the subsequent decays are lost in the surface. With a
specific activity of only 0.18 μBq per kg of target material the lowest 222Rn concentration in a
dark matter experiment so far was achieved by DEAP-3600 [98].
Even though ER backgrounds are usually dominating the overall background budget, all
high-sensitivity, low-background detectors have means to reject ER signals at a very high
efficiency (see discussion below in section 2.4.5). This usually renders ER backgrounds sub-
dominant for the actual WIMP search. The ultimate ER background will be elastic collisions
of low-energy solar pp- and 7Be neutrinos with atomic electrons [99] as this background
source can neither be avoided nor shielded but only rejected in the analysis based on their
ionization density.

2.4.3. α-background. α-contamination in the dark matter target or the detector materials
facing the target is usually uncritical due to the large energy deposited in the detector. It only
becomes relevant if the major part of the α-energy is lost in insensitive detector regions (e.g.
detector wall), producing artifacts which can leak into the signal region (for an example see
[100]). A notable exception are bubble chambers and some directional detector concepts as
described in section 2.3.4, where α-particles are the main source of background since the
detectors are insensitive to ER signals.

2.4.4. Nuclear recoil background. The most critical background for direct detection
experiments are single-scatter neutron-induced NRs from (α, n) and spontaneous fission
reactions (radiogenic background), or NRs induced by cosmic ray muons (cosmogenic
background), as such events cannot be distinguished from a WIMP signal. For this reason,
NRs induced by the coherent scattering of neutrinos off the target nuclei will eventually
3
The argumentation is similar for 220Rn from the 232Th chain, however, its background contribution is lower thanks
to its shorter half-life and reduced abundance in detectors.

20
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

constitute the ultimate background for direct WIMP searches (see section 3.4) as there is no
possibility to reject this background on an event-by-event basis. Compared to γ-rays of the
same energy, neutrons have a significantly longer mean free path (e.g.  (10) cm in liquid
xenon) and are thus much harder to shield. However, in larger detectors they have a rather
high probability to scatter elastically off the target nuclei several times creating a non WIMP-
like multi-scatter signature.
Radiogenic neutrons can originate from (α, n) reactions where an α-particle emitted by a
radioactive decay, usually from the primordial decay chains, is absorbed by a nucleus and
knocks off a neutron. A well-known example for this process is the 241AmBe neutron source,
which is employed for calibration purposes in many dark matter detectors. Due to the short
path length of the α-particle, the process is only relevant if the α-emitter is mixed into a
material with a high (α, n) cross section. The reaction probability depends on its Q-value, the
difference of the (α, n) threshold energy and the α-energy as well as on the Coulomb barrier
which effectively suppresses the (α, n) process for heavy nuclei. Therefore, the neutron
production via (α, n) reactions is highest for low-Z materials, typical examples being
polyethylene ((C 2 H 4)n ) or PTFE (Teflon, (C 2 F4)n ). The latter is often used in dark matter
detectors and contains 17F with a particularly high (α, n) cross section. Another isotope to be
avoided due to its high cross section is 9Be which typically occurs in springs made from CuBe
bronze etc.
Since their high Coulomb-barrier prevents (α, n) reactions, very heavy nuclei (> 232Th)
can essentially only generate neutrons via spontaneous fission. The highest contributions
usually come from 238U and 235U present in materials. The spontaneous fission neutron yield
normalized to the activity of the 238U decay chain is of the order of 6×10−7 n/decay; the
neutrons have MeV-energies.
Cosmogenic neutrons are created by fast muons with mean energies above 40 GeV that
make it into the underground laboratories (depth > 100 m.w.e.) where they interact with the
detector and the surrounding environment. The produced neutrons have energies up to the
GeV-range and thus have a great penetration power: even neutrons generated inside the rock a
few meters away from an underground cavern can enter the experiment. The neutrons are
produced in hadronic (had) or electromagnetic showers (em); in the latter case, the main
process is photoneutron production by real photons in giant dipole resonances. A smaller
fraction of the neutrons is produced by virtual photons (v). Muon-capture reactions are
irrelevant at the underground depths considered here. The neutron yields Y of the different
processes are ordered in the following way [101]
Yhad > Yem  Yv.
The main production processes in the dominant hadronic showers are deep-inelastic πN and
π−N capture reactions.

2.4.5. Background mitigation. In order to be sensitive to the rare signal from WIMP
interactions inside a detector, the contribution from the various backgrounds has to be
minimized. There are a plethora of different reduction strategies which are commonly dubbed
‘low-background techniques’ (for a review see [102]). All sensitive experiments make use of
most of them in one or another way.

2.4.6. Shielding. Massive shields installed around the dark matter detector are used to
suppress backgrounds originating from the experiment’s surroundings. High-Z materials such
as lead and copper or large amounts of water are efficient against external γ-rays. Dense
materials with a high hydrogen content (e.g. polyethylene or paraffin) or water efficiently

21
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 9. Muon flux versusdepth of various underground laboratories (measured in

‘meters water equivalent’ (m.w.e.). While the depth increases from 620 m.w.e.to
6720 m.w.e, the muon flux decreases by more than four orders of magnitude.

moderate and eventually absorb neutrons. In order to reduce muon-induced neutrons, dark
matter detectors are installed in deep-underground laboratories. To allow for the direct
comparison between different laboratories, their effective depth is expressed in ‘meters water
equivalent’ (m.w.e.), see figure 9. The hadronic cosmic ray component (p, n) is already
blocked at depths 10 m.w.e. Above a few hundred m.w.e., the remaining muon intensity is
reduced to below the level of radiogenic neutrons [102]. At the typical rock overburden of
2K6×103 m.w.e.of the laboratories hosting dark matter detectors, the muon flux is
suppressed by 5–7 orders of magnitude compared to the sea level flux.
As many background sources predominantly lead to events close to the detector surfaces,
their contribution can be further reduced by exploiting the self-shielding capability of the
target material. This fiducialization is especially effective for high-Z detector materials and
relies on the definition of an inner detector volume with reduced backgrounds. It is either
based on the measurement of the interaction position of every single event with some
precision, or by designing the detector in a way such that only events away from the walls
will create a measurable signal. The latter approach is successfully employed by Ge-based
experiments [75]. In some cases, the timing or pulse shape of the detector signal contains
information on its place of origin inside the detector [69, 76].
Muon-induced neutrons are further reduced by active muon vetoes, typically made of
large sheets of plastic scintillator or water-based Cherenkov detectors. These systems detect
the minimum ionizing muon with a high efficiency which allows for the rejection of every
coincident signal in the detector as this might stem from a muon-induced neutron. The veto
systems usually also have some sensitivity to detect hadronic or electromagnetic showers; this
is important for cases where the shower is created outside of the veto and no muon can be
detected.

2.4.7. Material selection. Thanks to the effective shielding, the ‘external’ background from
the detector surroundings (with the exception of cosmogenic neutrons) are irrelevant for
almost every experiment. The main ER and NR backgrounds thus originate in the (active or
passive) parts of the detector and the shield itself. They are reduced by using only materials

22
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 10. γ-spectrum of a sample of 10low-background PMTs (Hamamatsu R11410-

21, used in LXe TPCs) to measure their intrinsic radioactive contamination, together
with the background spectrum of the HPGe detector [159].

and components with a low intrinsic radioactive contamination for detector construction.
Several methods are available to identify such ‘clean’ materials:
(i) γ-spectrometry: A non destructive method to measure the γ-activity of a material or
detector component in a low-background HPGe detector (see figure 10). Instruments with
large sample-cavities of  (25) liters are available. The method is limited by the long
measuring times required even for massive samples and the different sensitivity to
different parts of the decay chains which complicates the detection of possible breaks of
the radioactive equilibrium.
(ii) Mass spectroscopy: Materials (mostly metals, but also some non-metals) are evaporated
for glow discharge (GD-MS) or inductively coupled plasma mass spectrometry (ICP-
MS). The methods can measure the concentration of elements to ppt levels, however,
the radioactive isotopes themselves remain undetected and one has to make assumptions
on their natural abundance. In case of the U, Th-chains, the early part of the chain is
measured (which is harder to assess with γ-spectrometry).
(iii) Neutron activation analysis: Small samples are irradiated by neutrons and the
characteristic γ-lines of activated isotopes are measured in γ-spectrometers. The method
gives access to isotopes which are not accessible by the other methods.
(iv) Rn-emanation: The emanation of radon isotopes from materials is measured by collecting
the daughters (e.g. 218Po, 214Po) on a PIN-diode where they undergo α-decay. The
sample is often placed in a large-volume chamber and the charged daughters are collected
on the small diode via an electric field [103]. Sensitivities of  (100)μBq/sample can be
achieved. If the sample chamber is sufficiently large, the method is non-destructive.
(v) α-spectroscopy: The effect of Rn-plate out can be measured without destroying the
sample with surface α-counters.

2.4.8. Target purification. In many cases the dark matter target has to be purified from
radioactive impurities. In case of solid targets, this is done prior to the start of the experiment,
e.g. during the process of growing crystals (e.g. NaI, CsI, CaWO4, Al2O3, Ge). While pulling
the crystal from the melt, the lattice will attract certain impurities and repel others. In the
zone-refining procedure, part of a crystal is molten and slowly moved across the ingot,
pushing the impurities to its end [104]. This allows the production of crystals that are
significantly cleaner than the pre-selected raw material. Cryogenic distillation exploits the
different boiling points of target and background gases [105] and is used to clean the noble

23
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

gases argon and xenon from nat Kr (which contains radioactive 85Kr at the 10−11-level). An
alternative approach relies on gas chromatography in a charcoal column [106].
It is more complicated to remove the radioactive 39Ar which is present in an nat Ar target
at a concentration of 1 Bq kg−1 as isotopic separation is required. The most efficient way to
produce low-background argon gas is to extract it directly from underground wells where the
39
Ar-level is reduced by more than three orders of magnitude [79]. Gaseous or liquid targets
can in principle be purified even after detector assembly, even on-line during data taking,
however, for practical reasons and efficiency it is usually done before the start of an
experiment. A notable exception is the active removal of Rn, which is constantly regenerated
inside the detector by emanation from the walls. In this case, on-line removal is mandatory
and very promising first results have been achieved by running a cryogenic distillation
column for 85Kr removal from Xe in ‘reverse mode’ [107]. An alternative approach to reduce
the background from Rn-emanation is the coating of surfaces to avoid that the isotope enters
the active target.

2.4.9. Active rejection. While the methods discussed so far aim at avoiding backgrounds in
the first place, there are always background events remaining that are recorded by the dark
matter detector. These can be reduced by making use of the expected signature of a WIMP: a
single-scatter, low-energy NR4. It is thus important to detect the event multiplicity (i.e. the
number of interactions inside the detector) as WIMPs will only scatter once in the detector
due to their tiny cross section. In particular neutrons with their longer mean-free path will
often generate multiple-scatter signatures. Depending on the used technology, multi-scatter
events can be identified either directly by measuring several interaction vertices or by pulse-
shape studies. A less powerful but still viable alternative is to segment the detector in an array
of smaller units and to reject events which are seen in several segments. All sensitive WIMP
detectors can therefore identify and reject multi-scatter events.
The different ionization density of ERs compared to the WIMP-induced NRs is often
used to reject ER signals. This effect has already been discussed in context of signal
quenching in section 2.2 and is exploited if two of the experimental observables charge, light
and heat/phonons are available on an event-by-event basis as their ratio differs for ERs and
NRs. For cryogenic experiments measuring phonons and another observable, this method
very effectively rejects ERs (<10−6 at an NR acceptance around 100% [108, 109]), however,
the ER and NR populations overlap at low energies, see figure 11. The method is less
powerful for liquid noble gases, with ER rejection levels at the 99.9% level at 30% NR
acceptance, but at least in xenon the ER discrimination works effectively down to lowest
energies and does thus not impact the threshold (figure 11). In general, the achievable ER
rejection factor is directly coupled to the NR acceptance: the first can be usually increased by
lowering the latter. All experiments have to find ways to reduce the impact of events
happening at or close to the detector surface where partial signal loss can lead to leakage of
ERs into the NR region.
As mentioned above, the liquid noble gas excimer states X2* de-excite with two different
time constants, depending on whether the spins of the two atoms are in a singlet () or a triplet
configuration (). The relative population of the two states depends again on the ionization
density ρI and thus the recoil type, with less triplet states being produced for high ρI. This is
effectively exploited in liquid argon, where the two time constants differ by three orders of
magnitude: t = 6 ns, t = 1.5 μs [110]. By comparing the fast part of the scintillation signal
4
If one searches for non-standard WIMP-nucleon interactions predicted by models beyond spin-(in)dependent
scattering, one, two or even all of these expectations might be subject to change.

24
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 11. Examples for active rejection of electronic recoil backgrounds by measuring
two independent signals in a detector. Both plots show the mean (thick) and the 10%
and 90% quantiles of the distributions. The vertical dashed lines indicates the energy
threshold used for the WIMP search. (Left) Cryogenic experiments using germanium
targets measure ionization and the phonon signal, which is used to set the energy scale
(in keV nr ). The ionization yield for nuclear recoil (NR) signals is reduced compared to
electronic recoils (ER) and the distributions are well separated above ∼10 keV nr . Data
from EDELWEISS-III [108]. (Right) Liquid xenon time projection chambers measure
scintillation light (S1) and ionization charge (S2) from an interaction. The ratioS2/S1
is reduced for NR signals. The distributions are rather close together but ER rejection at
reduced efficiency works down to the detector threshold. The energy scale is
reconstructed from both signals (green lines). Data from LUX [160].

Fprompt to the total pulse area ERs can be rejected to a very high level (< 2×10−8 at 90% NR
acceptance [111]). However, the need for a large number of detected photons for pulse shape
discrimination increases the analysis threshold to ∼30 keV nr . For xenon, the two time constants
are too similar to be exploited for ER rejection in a meaningful way [112].
Bubble chambers operated at the right temperature regime are almost immune (at the
10−9 level) to interactions from particles producing ERs with a low ionization density as these
do not create detectable bubbles. The remaining α-background can be reduced by recording
the acoustic signal from the bubble nucleation [83]: bubbles from α-interactions sound
different (louder) than the ones from NRs from neutrons (or WIMPs) due to the formation of
multiple bubbles along the α-track. Typical rejection levels are ∼99% at ∼95% signal
acceptance.

3. Direct detection: status and future

Since the first attempt to experimentally detect cold WIMP dark matter using a standard
0.72 kg high purity germanium crystal in 1987 [67], numerous experiments utilizing the
techniques discussed in section 2.3 have published their results. Apart from a few ‘anomalies’
(see, e.g. [113]), which usually faded away within a few months after triggering lots of
theoretical activities to explain the result, no experiment observed a statistically significant
excess above its background expectation. As of today, the WIMP still remains elusive.
Here we focus on the current status of the WIMP direct detection landscape. The leading
experiments are summarized in table 2.

25
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Table 2. Alphabetical list of some of the leading direct detection experiments that
published results on WIMP interactions; some of them are not operational anymore.
The quoted mass refers to the active target; some detectors have position sensitivity and
can select the (cleaner) innermost part of the detector for the analysis. Please note that
the mass-range covers 7orders of magnitude. The references list the latest or most
relevant publication from the experiment; projects marked with an asterisk (*) indicate
prototype phases, i.e. the quoted mass does not represent the final stage.
Experiment Type Target Mass (kg) Laboratory References
ANAIS-112 Crystal NaI 112 Canfranc [121]
CDEX-10 Crystal Ge 10 CJPL [68]
CDMSLite Cryogenic Ge 1.4 Soudan [164]
COSINE-100 Crystal NaI 106 YangYang [119]
CRESST-II Cryogenic CaWO4 5 LNGS [165]
CRESST-III Cryogenic CaWO4 0.024 LNGS [123]
DAMA/LIBRA-II Crystal NaI 250 LNGS [72]
DarkSide-50 TPC Ar 46 LNGS [122]
DEAP-3600 Single phase Ar 3300 SNOLAB [166]
DRIFT-II Directional CF4 0.14 Boulby [92]
EDELWEISS Cryogenic Ge 20 LSM [167]
LUX TPC Xe 250 SURF [168]
NEWS-G Gas Counter Ne 0.283 SNOLAB [78]
PandaX-II TPC Xe 580 CJPL [169]
PICASSO Superheated Droplet C4F10 3.0 SNOLAB [129]
PICO-60 Bubble Chamber C3F8 52 SNOLAB [93]
SENSEI* CCD Si 9.5×10−5 FNAL [137]
SuperCDMS* Cryogenic Si 9.3×10−4 above ground [138]
XENON100 TPC Xe 62 LNGS [170]
XENON1T TPC Xe 1995 LNGS [63]
XMASS Single phase Xe 832 Kamioka [171]

3.1. Annual modulation signature

The only ‘anomaly’ surviving since many years is the signal observed by DAMA/LIBRA.
This collaboration operates a massive array of low-background NaI(Tl) crystals in the Italian
Gran Sasso underground laboratory and searches for a dark matter-induced annually mod-
ulating signal. Such modulation, with period and phase consistent with the expectation from
the standard halo model, has been observed in various incarnations of the project. The long-
awaited results from the DAMA/LIBRA-phase2 were published recently [72]. As expected
from dark matter, the modulating signal is only present in single scatter events at low energies
<6 keV ee . The modulation has been observed over a total of 20 annual cycles, with a total
exposure of 2.46 t × y. By now, the signal reached a statistical significance of 12.9 σ and
could not be related to other, non-WIMP explanations.
However, there are serious concerns about the dark-matter nature of the DAMA/Libra
observation: if the signal is interpreted in the standard scenarios which expect NRs signals, it
is excluded by more sensitive experiments by many orders of magnitude [63]. It is worth
mentioning that the phase2 results of DAMA, obtained with a threshold lowered from 2 to
1 keV, are also inconsistent with the results from the first phase of the same experiment when
interpreted as a dark matter WIMP signal [114]. In order to avoid an interpretation bias due to
a specific model, liquid xenon experiments searched for modulation signals in their data as
well. These experiments are sensitive to the energy interval in which DAMA observes its

26
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 12. The current experimental parameter space for spin-independent WIMP-
nucleon cross sections. Not all published results are shown. The space above the lines
is excluded at a 90%confidence level. The two contours for DAMA interpret the
observed annual modulation in terms of scattering of iodine (I) and sodium (Na),
respectively [161]. The dashed line limiting the parameter space from below represents
the ‘neutrino floor’ [143] from the irreducible background from coherent neutrino-
nucleus scattering (CNNS), see section 3.4.

signal, have a significantly lower background and were operated stably for several years. The
electronic structures of xenon and iodine are very similar and the non-observation of a
modulation signal thus allowed XENON100, XMASS and LUX to exclude the DAMA signal
at 5.7 σ, ∼2 σ and 9.2 σ, respectively [115–117].
Attempts to solve the long-standing discrepancy of the experimental results with so-
called ‘isospin-violating’ dark matter models, which favor NaI over Xe targets [118], have
recently been challenged by the first results of COSINE-100 [119]. This experiment uses a
total of 106 kg of low-background NaI(Tl) crystals which are installed inside an active veto-
shield made of liquid scintillator. This allows for the reduction of the background from a
3 keV x-ray line from 40K right in the ROI by detecting the coincident 1.4 MeV γ-ray in the
veto. COSINE-100 did not observe an excess of signal-like events above background in a first
59.5 d run and thus challenges the DAMA/LIBRA signal as being caused by spin-inde-
pendent WIMP-nucleon interactions using the same NaI target. The first searches for an
annually modulating signal with COSINE-100 [120] and the similarly-sized ANAIS-112
using 112 kg of NaI(Tl) crystals [121] are both consistent with the absence of a signal,
however, both need more data to draw definite conclusions on the DAMA claim.

3.2. Nuclear recoils from WIMP-nucleon scattering

3.2.1. Spin-independent interactions.

The current status of the search for spin-independent
WIMP-nucleon scattering is shown in figure 12. (Note that only a selection of the most
relevant results are shown.) Above WIMP masses of ∼5 GeV/c2, the strongest limits are
placed by the the liquid xenon (LXe) TPCs XENON1T, LUX, and PandaX-II. The

27
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

experiment with the best sensitivity to WIMPs in this mass-range is XENON1T, a LXe TPC
with a 2.0 t target. It achieved a background of only 85 events/(t × y × keV ee ) in the WIMP
search region (before ER rejection) and excluded spin-independent WIMP-nucleon
interactions with cross sections above 4.1×1047 cm2 at 30 GeV/c2 and 90% confidence
level in a run with a 1.0 t × y exposure [63]. The results from the liquid argon (LAr)
experiments DarkSide-50 (TPC, 46 kg target) and DEAP-3600 (single-phase, 3.6 t target) are
weaker due to their higher threshold and lower exposure.
In a mass range from 1.8 GeV/c2mχ 5 GeV/c2, the most stringent exclusion limit
was placed by DarkSide-50 using a LAr target depleted in 39Ar [122]. The result from a
0.019 t × y run is a based on using the ionization signal only, which allowed reducing the
analysis threshold to 0.1 keV ee . The observed background of 1.5 events/(kg × d × keV ee ),
corresponding to 5.5×105 events/(t × y × keV ee ), can be attributed to known background
sources above ∼1.4 keV nr (corresponding to 8 e−).
Due to their much smaller total target mass and higher backgrounds, the cryogenic
experiments using Ge-crystals with ionization and phonon readout (EDELWEISS, (Super)
CDMS) or scintillating CaWO4-crystals with light and phonon readout (CRESST) cannot
compete in the search for medium to high-mass WIMPs. However, due to their ability to
reach extremely low thresholds well below 1 keV nr , they are very sensitive to low-mass
WIMPs with masses 5 GeV/c2. The Germanium-based detectors SuperCDMS and
EDELWEISS could improve their low-mass sensitivity by operating the detectors with a
high bias voltage, converting the ionization signals into Neganov–Luke phonons. CRESST-
III is currently placing the most stringent constraints below mχ=1.8 GeV/c2 [123]
extending the mass range down to 0.16 GeV/c2. The result was achieved using a
24 gCaWO4 crystal with a threshold of 31 eV. The cryogenic crystal was operated
underground and acquired an exposure of 3.64 kg × d. In this sub-GeV window competitive
limits were also placed by NEWS-G, a spherical proportional counter with 60 cm diameter
and filled with a Ne+CH4 (0.7%) gas-mixture at 3.1 bar (corresponding to 283 g) [78]. With
its low threshold of 36.5 eV ee and the use of the low-A gas neon the instrument was optimized
to search for low-mass WIMPs.
It has recently been proposed that the reach of WIMP detectors could possibly be
extended further into the sub-GeV region by exploiting the Migdal effect [124, 125]: if the
WIMP-nucleus interaction does not lead to a signal above detector threshold but the recoling
atom gets excited and ionized by the process, this might lead to an additional signal which
might be detected. Several results taking into account this effect were already published
[126–128], however, at the moment it is not clear how the detector response to this effect can
be calibrated and whether the effect is actually present at all.

3.2.2. Spin-dependent interactions. As discussed in section 2.1, bubble chambers filled with
targets containing the isotope 19F have the highest sensitivity to spin-dependent WIMP-
proton couplings. The best limit to date is from PICO-60, operated with 52 kg of C3F8
(octafluoropropane), see figure 13 (top). No excess of WIMP candidates was observed above
the background expectation in a combined exposure of (1167+1404)kg × d (with
thresholds of 3.3 keV nr and 2.45 keV nr , respectively), excluding spin-dependent cross
sections above 2.5×10−41 cm2 for 25 GeV/c2 WIMPs [93]. At low WIMP masses between
2 and 4 GeV/c2, the best limits on spin-dependent WIMP-proton interactions are from
PICASSO which operated 32 superheated droplet detectors with a total mass of 3.0 kg C4F10
[129]. The experiments are probing the same parameter space as the neutrino telescopes
IceCube, ANTARES and Super-Kamiokande, which constrain spin-dependent WIMP-proton
scattering via dark matter capture in the Sun (and subsequent annihilation into τ-leptons or

28
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 13. Current status of the searches for spin-dependent couplings. (Top) WIMP-
proton interactions. The search is dominated by bubble chambers and superheated
droplet detectors which contain the isotope 19F. The results from the much larger LXe
detectors are an order of magnitude weaker. Also shown are limits from indirect
searches [162, 163]. (Bottom) WIMP-neutron interactions. The best results are from
LXe TPCs.

b-quarks which decay into detectable neutrinos). At lower WIMP masses, the parameter space
is also explored by ATLAS and CMS via ‘mono-X’ searches.
Due to a ∼50% natural abundance of xenon isotopes with unpaired neutrons, the
strongest constraints on spin-dependent WIMP-neutron scattering are set again by the
massive liquid xenon TPCs (LUX, PandaX, XENON) which dominate the spin-independent
searches at medium to high masses, see figure 13 (bottom). The most sensitive result to-date is

29
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

from XENON1T [130], re-interpreting the data already used to constrain spin-independent
interactions. Argon has no stable isotopes with unpaired spins such that LAr detectors cannot
contribute to spin-dependent WIMP searches. There are published results from the cryogenic
Germanium experiments CDMS-II [131] and SuperCDMS (calculated by [132] based on
[133]), however, due to the significantly lower detector mass, these are not competitive.
SuperCDMS operating one Ge-crystal (600 g) in Neganov–Luke mode (‘CDMSLite’)
put constraints on spin-dependent WIMP-proton and WIMP-neutron interactions extending to
mχ=1.5 GeV/c2 [134], however, the results are more than six (five) orders of magnitude
weaker than the best high-mass results for proton (neutron)-couplings (and thus do not fit to
the axis chosen for figures 13). Recent results from a 2.66 g Li2MoO4 cryogenic scintillating
crystal calorimeter prototype operated by the CRESST collaboration place limits down to
0.8 GeV/c2 [135], however, are much weaker than the one from SuperCDMS.

3.3. Electronic recoils from WIMP-electron scattering

Very light WIMPs with masses in the MeV/c2-range will not transfer sufficient momentum to
the target nucleus to generate a detectable NR signal. The search for such particles thus
concentrates on WIMP-electron scattering.
Thanks to their high sensitivity to single-electron signals and the low ionization energy of
1.2 eV, experiments using Si-based sensors are ideally suited to search for such signals. The
most stringent constraints in the 0.5–5 MeV/c2 mass range come from a prototype run of
the SENSEI experiment operated in the shallow MINOS cavern at Fermilab. SENSEI uses the
Skipper-CCD technique featuring an ultra-low electronic readout noise of only 0.07 electrons
per CCD pixel [136]. A repeated operation of the 0.094 7g detector for 120 s before readout
led to a total exposure of 0.069 g × d and the best for WIMP-electron scattering below
5 MeV/c2 [137], see figure 14. An above-ground prototype run of a SuperCDMS detector
with a resolution of 0.1electron-hole pairs and eV-resolution [138] gave a similar result. The
tiny detector masses and exposures imply that the results are expected to improve sig-
nificantly once the experiments leave the prototyping stage. Above 5 MeV/c2, the data from
dual-phase TPCs (XENON10, XENON100, DarkSide-50) analyzed in ionization-only mode
provide the most stringent constraints as these detectors provide more exposure but a higher
threshold than the Si detectors [82, 139].
Many more experimental efforts are currently being discussed to search for very light
(eV-MeV) dark matter as an alternative to the WIMP paradigm (see, e.g. [140]). These are,
however, beyond the scope of this review.

3.4. Outlook: towards the neutrino floor

The signature for WIMP dark matter in direct detection experiments are single scatter NRs.
The same signal is produced (in any detector) by coherent neutrino-nucleus scattering
(CNNS) which leads to an irreducible background for the WIMP search. The cross section for
neutrinos of energyEν scattering coherently off a nucleus of massmN (Z protons and A − Z
neutrons), resulting in an NR of energyEnr , is given by [141]

ds G 2 mN ⎛ m E ⎞
= F [A - Z - (1 - 4 sin2 qW ) Z ] ⎜1 - N 2nr ⎟ F 2 (Q 2) ; (23)
dE nr 4p ⎝ 2En ⎠

GF is Fermi’s constant and θW the Weinberg angle. As in the description of WIMP-nucleon

interactions the form factor F(Q2) depends on the momentum transfer and can be described by
the standard Helm form factor. The neutrinos are high-energetic 8B neutrinos from the Sun

30
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Figure 14. Parameter space for MeV-scale dark matter interacting with electrons
mediated by a heavy dark photon (dark matter form factor FDM = 1 independent of
momentum transfer). The space is probed by experiments with sensitivities to single
electrons based on solid state Si detectors (SENSEI, SuperCDMS) or dual-phase TPCs
filled with LXe (XENON10) or LAr (DarkSide-50) analyzed in charge-only mode.

Figure 15. Differential recoil spectra in xenon from WIMPs of various masses mχ and
spin-independent cross sections and from neutrinos scattering coherently off the xenon
nuclei (CNNS). At low recoil energies, the CNNS sum spectrum is dominated by solar
8
B neutrinos, at high energies by atmospheric neutrinos. An NR acceptance of 50% is
assumed and the detector signal is converted to the electronic recoil scale (keVee) using
the scintillation signal only.

(dominating at low WIMP masses mχ) and atmospheric neutrinos (dominating above
mχ∼10 GeV/c2). Neutrinos from the hep-process and from the diffuse supernova
background (DSNB) have subdominant contributions [142], see figure 15. This neutrino

31
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

background provides the ultimate border for ‘traditional’ direct WIMP searches [143–145],
whereas directional detectors could go beyond the border provided that a significant number
of events are detected [94] (see also discussion on directional detectors in section 2.3.5).
The ‘neutrino floor’ plotted in figure 12 was introduced as ‘WIMP discovery limit’ in
[143]. It is defined by the detection of a WIMP signal at 3σ on top of a background of
500CNNS events above a threshold of 4 keV nr in a LXe detector, assuming infinite energy
resolution. Extremely large (and rather unrealistic) exposures around 5300 t × y are required
to detect 500CNNS events from atmospheric neutrinos. The cross-section where one CNNS
event is expected in the dataset is about one order of magnitude higher than the neutrino floor
defined in this way.
While the ton-scale LXe experiment XENON1T has already almost reached the neutrino
floor at mχ∼8 GeV/c2 [63], see figure 12, 2–3orders of magnitude of cross-section are still
to be explored at higher and lower WIMP masses. At low WIMP masses, where the neutrino
floor is almost a factor105 higher than at higher masses, a low detector threshold (at a
moderate exposure) is the key experimental requirement. The cryogenic experiments
SuperCDMS [146] (Ge, Si-targets with standard ionization-phonon readout or with Neganov–
Luke amplification) and CRESST [147] (CaWO4, Al2O3) have presented strategies how to
reach the neutrino floor in a staged approach with increasing detector mass and a significant
background reduction compared to the state of the art. CRESST plans to reach the neutrino
floor around 5 GeV/c2 with an exposure of 1000 kg × d of CaWO4. The CYGNUS colla-
boration works towards 10–1000 m3 scale directional detectors to probe this region [148].
Probing the mass range mχ  10 GeV/c2 down to the neutrino floor requires large
‘background-free’ exposures and thus massive detectors with multi-ton targets and very low
background levels. Due to the exponentially falling recoil spectra, low thresholds are bene-
ficial but not absolutely required for masses above ∼40 GeV/c2. This range is the realm of the
massive dual-phase TPCs filled with the liquid noble gases xenon and argon. (Single-phase
experiments will soon be terminated with XMASS (LXe) joining the XENON project and
DEAP (LAr) joining DarkSide.) A total of four dual-phase TPCs are currently being prepared
to start taking data in the next years: the LXe experiments PandaX-4T (4.0 t active target)
[149], XENONnT (5.9 t) [150] and LZ (7.0 t) [151] and DarkSide-20k (23t) operated with
depleted LAr [152]. After a few years of data taking they will be able to probe spin-
independent cross sections of  (10-48) cm2 at the sensitivity maximum, which is still an order
of magnitude above the neutrino floor. The design sensitivities of the smaller PandaX-4T and
the LAr-based DarkSide-20k experiments are a bit weaker than for XENONnT and LZ,
however, the achieved sensitivity of any experiment will eventually depend on the actual
background level. For spin-dependent WIMP-proton couplings, the upcoming bubble
chamber PICO-500 [153] with a 500 l C3F8 target will reach similar sensitivities as
XENONnT and LZ for spin-dependent WIMP-neutron scattering.
To eventually probe heavy-WIMP cross sections at the neutrino floor, very large
exposures are required. The Global Argon Dark Matter Collaboration anticipates a LAr TPC
with a 300 t fiducial mass to acquire an exposure of 3000 t × y [154]. Using a LXe detector,
exposures of about 200 t × y and background-levels dominated by neutrinos are required
[155]. The DARWIN collaboration proposes to build a 40 t LXe TPC with lowest back-
grounds to achieve this goal. This instrument will not only be sensitive to spin-independent,
spin-dependent and various inelastic WIMP-nucleon scattering interactions, but will also have
a rich neutrino and astroparticle physics program [99, 156, 157].

32
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

Acknowledgments

I would like to thank the editors for the invitation to write this review and their patience while
waiting for it. This work was supported by the German Ministry for Education and Research
(BMBF) through the project CRESST-XENON and the European Research Council through
the ERC consolidator grant ULTIMATE.

ORCID iDs

Marc Schumann https://orcid.org/0000-0002-5036-1256

References

[1] Ade P A R et al (Planck) 2016 Astron. Astrophys. 594 A13

[2] Pitjev N P and Pitjeva E V 2013 Astron. Lett. 39 141
[3] Oort J H 1932 Bull. Astr. Inst. Netherlands 6 249
[4] Zwicky F 1933 Helv. Phys. Acta 6 110
[5] Bertone G and Hooper D 2018 Rev. Mod. Phys. 90 45002
[6] Rubin V C, Ford W K and Thonnard N 1978 Astrophys. J. 225 L107
[7] Bosma A 1981 Astron. J. 86 1825
[8] Slater C T et al 2016 ApJ 832 206
[9] Massey R, Kitching T and Richard J 2010 Rep. Prog. Phys. 73 086901
[10] Clowe D et al 2006 Astrophys. J. 648 L109
[11] Percival W J et al 2001 MNRAS 327 1297
[12] Eisenstein D J et al (SDSS) 2005 Astrophys. J. 633 560
[13] Aghanim N et al (Planck) 2018 arXiv:1807.06209
[14] Smith M C et al 2007 Mon. Not. R. Astron. Soc. 379 755
[15] Kuhlen M, Vogelsberger M and Angulo R 2012 Phys. Dark Univ. 1 50
[16] Bozorgnia N et al 2016 J. Cosmol. Astropart. Phys. JCAP05(2016)024
[17] Müller O et al 2018 Science 359 534
[18] Read J I, Walker M G and Steger P 2019 Mon. Not. R. Astron. Soc. 484 1401
[19] Gaskins J M 2016 Contemp. Phys. 57 496
[20] Read J I 2014 J. Phys. G: Nucl. Part. Phys. 41 063101
[21] Sivertson S et al 2018 Mon. Not. R. Astron. Soc. 478 1677
[22] Tisserand P et al 2007 Astron. Astrophys. 469 387
[23] Brandt T D 2016 Astrophys J. 824 L31
[24] Abbott B P et al (LIGO) 2016 Phys. Rev. Lett. 116 061102
[25] Ali-Haïmoud Y, Kovetz E D and Kamionkowski M 2017 Phys. Rev. D 96 123523
[26] Alcock C et al (MACHO) 2001 Astrophys. J. Lett. 550 L169
[27] Wyrzykowski L et al (OGLE) 2011 Mon. Not. R. Astron. Soc. 416 2949
[28] Zumalacárregui M and Seljak U 2018 Phys. Rev. Lett. 121 141101
[29] Boudaud M and Cirelli M 2019 Phys. Rev. Lett. 122 041104
[30] Abbott B et al (LIGO/VIRGO) 2017 Phys. Rev. Lett. 119 161101
[31] Fields B D 2011 Ann. Rev. Nucl. Part. Sci. 61 47
[32] Nilles H P 1984 Phys. Rep. 110 1
[33] Lahanas A B 2007 Lect. Notes Phys. 720 35
[34] Peskin E 2015 Proc. Natl Acad. Sci. 112 12256
[35] Hooper D and Profumo S 2007 Phys. Rep. 453 29
[36] Chen C-R et al 2014 J. High Energy Phys. JHEP06(2014)074
[37] Roszkowski L 2004 Pramana 62 389
[38] Zurek K M 2014 Phys. Rep. 537 9
[39] Baer H et al 2015 Phys. Rep. 555 1
[40] Shaposhnikov M and Tkachev I 2006 Phys. Lett. B 639 414
[41] Ringwald A 2012 Phys. Dark Univ. 1 116
[42] Harvey D et al 2015 Science 347 1462

33
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

[43] Frenk C S and White S D M 2012 Ann. Phys. 524 507

[44] Lopez-Honorez L et al 2017 Phys. Rev. D 96 103539
[45] Kahlhoefer F 2017 Int. J. Mod. Phys. A 32 1730006
[46] Goodman M W and Witten E 1985 Phys. Rev. D 31 3059
[47] Lewin J D and Smith P F 1996 Astropart. Phys. 6 87
[48] Drukier A K, Freese K and Spergel D N 1986 Phys. Rev. D 33 3495
[49] Bellini G et al 2012 J. Cosmol. Astropart. Phys. JCAP05(2012)015
[50] Green A M 2017 J. Phys. G: Nucl. Part. Phys. 44 084001
[51] Spergel D N 1988 Phys. Rev. D 37 1353
[52] Mayet F et al 2016 Phys. Rep. 627 1
[53] Helm R H 1956 Phys. Rev. 104 1466
[54] Engel J, Pittel S and Vogel P 1992 Int. J. Mod. Phys. E 1 1
[55] Toivanen P et al 2009 Phys. Rev. C 79 044302
[56] Menendez J et al 2012 Phys. Rev. D 86 103511
[57] Klos P et al 2013 Phys. Rev. D 88 083516
Klos P et al 2014 Phys. Rev. D 89 029901 (erratum)
[58] Fan J, Reece M and Wang L 2010 J. Cosmol. Astropart. Phys. JCAP11(2010)042
[59] Fitzpatrick A et al 2013 J. Cosmol. Astropart. Phys. JCAP13(2013)004
[60] Hoferichter M et al 2016 Phys. Rev. D 94 063505
[61] Buchmueller O, Dolan M J and McCabe C 2014 J. High Energy Phys. JHEP01(2014)025
[62] Aprile E et al (XENON) 2017 Phys. Rev. D 96 042004
[63] Aprile E et al (XENON1T) 2018 Phys. Rev. Lett. 121 111302
[64] Lindhard J et al 1963 Selskab. Mat. Fys. Medd. 33 1
[65] Szydagis M and (NEST) (https://doi.org/10.5281/zenodo.1314669)
[66] Szydagis M et al 2011 JINST 6 P10002
[67] Ahlen S P et al 1987 Phys. Lett. B 195 603
[68] Jiang H et al (CDEX-10) 2018 Phys. Rev. Lett. 120 241301
[69] Aalseth C E et al (CoGeNT) 2011 Phys. Rev. Lett. 106 131301
[70] Aguilar-Arevalo A et al (DAMIC) 2016 Phys. Rev. D 94 082006
[71] Crisler M et al (SENSEI) 2018 Phys. Rev. Lett. 121 061803
[72] Bernabei R et al (DAMA/LIBRA) 2018 Universe 4 116
[73] Haller E E et al 1994 SPIE 2198, Instrumentation in Astronomy VIII p 630
[74] Shutt T et al 1992 Phys. Rev. Lett. 69 3425
[75] Broniatowski A et al (EDELWEISS) 2008 J. Low. Temp. Phys. 151 830
[76] Akerib D S et al (CDMS) 2007 Nucl. Phys. Proc. Suppl. 173 137
[77] Luke P N et al 1990 Nucl. Instrum. Meth. A 289 406
[78] Arnaud Q et al (NEWS-G) 2018 Astropart. Phys. 97 54
[79] Agnes P et al (DarkSide-50) 2016 Phys. Rev. D 93 081101(R)
[80] Aprile E et al 2006 Phys. Rev. Lett. 97 081302
[81] Angle J et al (XENON10) 2011 Phys. Rev. Lett. 107 051301
Angle J et al (XENON10) 2013 Phys. Rev. Lett. 110 249901 (erratum)
[82] Essig R et al 2012 Phys. Rev. Lett. 109 021301
[83] Aubin F et al (PICASSO) 2008 New J. Phys. 10 103017
[84] Seitz F 1958 Phys. Fluids 1 2
[85] Amole C et al (PICO-2L) 2016 Phys. Rev. D 93 061101
[86] Barnabe-Heider M et al (PICASSO) 2005 Nucl. Instrum. Meth. A 555 184
[87] Amole C et al (PICO-60) 2017 Phys. Rev. Lett. 118 251301
[88] Copi C J, Heo J and Krauss L M 1999 Phys. Lett. B 461 43
[89] O’Hare C A J, Kavanagh B J and Green A M 2017 Phys. Rev. D 96 083011
[90] Santos D et al (MIMAC) 2013 J. Phys.: Conf. Ser. 469 012002
[91] Tao Y et al (MIMAC) 2019 arXiv:1903.02159
[92] Battat J B R et al (DRIFT) 2017 Astropart. Phys. 91 65
[93] Amole C et al (PICO-60) 2019 arXiv:1902.04031
[94] Grothaus P, Fairbairn M and Monroe J 2014 Phys. Rev. D 90 055018
[95] Akerib D S et al (LZ) 2017 Astropart. Phys. 96 1
[96] Šrámek O et al 2017 Geochim. Cosmochim. Acta 196 370
[97] Baudis L et al 2015 Eur. Phys. J. C 75 485
[98] Amaudruz P A et al (DEAP-3600) 2018 Phys. Rev. Lett. 121 071801

34
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

[99] Baudis L et al 2014 J. Cosmol. Astropart. Phys. JCAP01(2014)044

[100] Strauss R et al (CRESST-II) 2015 J. Cosmol. Astropart. Phys. JCAP06(2015)030
[101] Malgin A S 2017 Phys. Rev. C 96 014605
[102] Heusser G 1995 Ann. Rev. Nucl. Part. Sci. 45 543
[103] Kiko J 2001 Nucl. Instrum. Meth. A 460 272
[104] Pfann W G 1952 JOM 4 747
[105] Abe K et al (XMASS) 2009 Astropart. Phys. 31 290
[106] Bolozdynya A I et al 2007 Nucl. Instrum. Meth. A 579 50
[107] Aprile E et al (XENON100) 2017 Eur. Phys. J. C 77 358
[108] Armengaud E et al (EDELWEISS-III) 2017 JINST 12 P08010
[109] Agnese R et al (SuperCDMS) 2013 Appl. Phys. Lett. 103 164105
[110] Hitachi A et al 1983 Phys. Rev. B 27 5279
[111] Amaudruz P-A et al (DEAP) 2016 Astropart. Phys. 85 1
[112] Hogenbirk E et al 2018 JINST 13 P05016
[113] Gresham M I and Zurek K M 2014 Phys. Rev. D 89 016017
[114] Baum S, Freese K and Kelso C 2019 Phys. Lett. B 789 262
[115] Aprile E et al (XENON100) 2017 Phys. Rev. Lett. 118 101101
[116] Abe K et al (XMASS) 2016 Phys. Lett. B 759 272
[117] Akerib D S et al (LUX) 2018 Phys. Rev. D 98 062005
[118] Feng J L et al 2011 Phys. Lett. B 703 124
[119] Adhikari G et al (COSINE-100) 2018 Nature 564 83
[120] Adhikari G et al (COSINE-100) 2019 arXiv:1903.10098
[121] Amaré J et al (ANAIS-112) 2019 arXiv:1903.03973
[122] Agnes P et al (DarkSide-50) 2018 Phys. Rev. Lett. 121 081307
[123] Abdelhameed A H et al (CRESST-III) 2019 arXiv:1904.00498
[124] Ibe M et al 2018 J. High Energy Phys. JHEP03(2018)194
[125] Dolan M J, Kahlhoefer F and McCabe C 2018 Phys. Rev. Lett. 121 101801
[126] Akerib D et al (LUX) 2019 Phys. Rev. Lett. 122 131301
[127] Armengaud E et al (EDELWEISS) 2019 Phys. Rev. D 99 082003
[128] Liu Z Z et al (CDEX-1B) 2019 arXiv:1905.00354
[129] Behnke E et al (PICASSO) 2017 Astropart. Phys. 90 85
[130] Aprile E et al (XENON1T) 2019 Phys. Rev. Lett. 122 141301
[131] Ahmed Z et al (CDMS-II) 2009 Phys. Rev. Lett. 102 011301
[132] Marcos C, Peiro M and Robles S 2016 J. Cosmol. Astropart. Phys. JCAP03(2016)019
[133] Agnese R et al (SuperCDMS) 2014 Phys. Rev. Lett. 112 241302
[134] Agnese R et al (SuperCDMS) 2018 Phys. Rev. D 97 022002
[135] Abdelhameed A H et al (CRESST) 2019 arXiv:1902.07587
[136] Tiffenberg J et al 2017 Phys. Rev. Lett. 119 131802
[137] Abramoff O et al (SENSEI) 2019 Phys. Rev. Lett. 122 161801
[138] Agnese R et al (SuperCDMS) 2018 Phys. Rev. Lett. 121 051301
Agnese R et al (SuperCDMS) 2019 Phys. Rev. Lett. Erratum 122 069901
[139] Essig R, Volansky T and Yu T 2017 Phys. Rev. D 96 043017
[140] Battaglieri M et al 2017 arXiv:1707.04591
[141] Freedman D Z, Schramm D N and Tubbs D L 1977 Ann. Rev. Nucl. Part. Sci. 27 167
[142] Strigari L E 2009 New J. Phys. 11 105011
[143] Billard J, Strigari L and Figueroa-Feliciano E 2014 Phys. Rev. D 89 023524
[144] O’Hare C A J 2016 Phys. Rev. D 94 063527
[145] Boehm C et al 2019 J. Cosmol. Astropart. Phys. JCAP01(2019)043
[146] Agnese R et al (SuperCDMS) 2017 Phys. Rev. D 95 082002
[147] Angloher G et al (CRESST) 2015 arXiv:1503.08065
[148] Spooner N 2018 CYGNUS—a multi-latitude directional WIMP experiment, talk at IDM
[149] Zhang H et al (PandaX) 2019 Sci. China Phys. Mech. Astron. 62 31011
[150] Aprile E et al (XENON) 2016 J. Cosmol. Astropart. Phys. JCAP04(2016)027
[151] Akerib D S et al (LZ) 2018 arXiv:1802.06039
[152] Aalseth C E et al (DarkSide-20k) 2018 Eur. Phys. J. Plus 133 131
[153] Vázquez-Jáuregui E 2017 PICO-500L: Simulations for a 500L Bubble Chamber for Dark Matter
Search, talk at TAUP

35
J. Phys. G: Nucl. Part. Phys. 46 (2019) 103003 Topical Review

[154] The Global Argon Dark Matter Collaboration (GDAMC) Future Dark Matter Searches with Low-
Radioactivity Argon, input to the European Particle Physics Strategy Update 2018-2020
[155] Schumann M et al 2015 J. Cosmol. Astropart. Phys. JCAP10(2015)016
[156] Aalbers J et al (DARWIN) 2016 J. Cosmol. Astropart. Phys. JCAP11(2016)017
[157] Lang R F et al 2016 Phys. Rev. D 94 103009
[158] Scholz B J et al 2016 Phys. Rev. D 94 122003
[159] Sivers M V et al 2016 JINST 11 P12017
[160] Akerib D S et al (LUX) 2014 Phys. Rev. Lett. 112 091303
[161] Savage C et al 2009 J. Cosmol. Astropart. Phys. JCAP04(2009)010
[162] Aartsen M G et al (IceCube) 2017 Eur. Phys. J. C 77 146
[163] Choi K et al (Super-Kamiokande) 2015 Phys. Rev. Lett. 114 141301
[164] Agnese R et al (SuperCDMS) 2016 Phys. Rev. Lett. 116 071301
[165] Angloher G et al (CRESST) 2016 Eur. Phys. J. C 76 25
[166] Ajaj R et al (DEAP-3600) 2019 arXiv:1902.04048
[167] Hehn L et al (EDELWEISS) 2016 Eur. Phys. J. C 76 548
[168] Akerib D S et al (LUX) 2017 Phys. Rev. Lett. 118 021303
[169] Cui X et al (PandaX) 2017 Phys. Rev. Lett. 119 181302
[170] Aprile E et al (XENON100) 2016 Phys. Rev. D 94 122001
[171] Abe K et al (XMASS) 2019 Phys. Lett. B 789 45

36