Gif Lectures on direct detection of Dark Matter

Eric Armengaud
arXiv:1003.2380v1 [hep-ex] 11 Mar 2010

Service de Physique des Particules, IRFU, CEA, F-91191Gif/Yvette Cedex, France

Abstract. These notes cover some of the topics associated with direct detection of
dark matter at an introductory level. The general principles of dark matter search are
summarized. The current status of some experiments is described, with an emphasis
on bolometric and noble liquid techniques. Plots and illustrations associated to these
notes may be found on transparencies presented during the lecture, on the web site of
Gif school 2009 (in French) : http://www-subatech.in2p3.fr/gif2009.html

1. Introduction

Current cosmological observations have lead to a concordance model of our Universe, in

which a major role is played by a still mysterious dark matter. Dark matter drives the
dynamics of galaxies and clusters, and generates the growth of large-scale structures,
from initial density perturbations of order 105 that are measured in the CMB
anisotropies, to inhomogeneities of amplitude 1 observed in the large scale structures
at low redshift. In the concordance model, dark matter cannot be of baryonic nature
since both primordial abundance measurements and CMB anisotropies point towards a
small global baryonic density b .
Dark matter, as well as dark energy, might be a manifestation of the break-down of
general relativity on large scales or in the regime of low accelerations. A large litterature
is now devoted to possible modifications of the laws of gravitation : in general, these
modifications are equivalent to the introduction of new fields in addition to the metric
tensor g . Another approch is to assume that dark matter is constituted of a gaz of
collisionless, stable and massive particles whose nature is still unknown. Indeed several
arguments strongly suggest that the Standard Model of particle physics is not complete,
at least for energies larger than the electroweak scale. A large number of extensions of
the Standard Model provide quite naturally a possible dark matter candidate in their
spectrum [1].
In these notes we present the experimental efforts for direct detection of possible
dark matter (DM) particles. All experiments for DM direct detection aim at testing the
hypothesis that our own galactic halo is filled with particle dark matter. For several
plausible DM hypothesis, like the gravitino in mSUGRA [2], the interaction rate of
[email protected]
Ecole de Gif, Batz sur mer, 21-25 September 2009
Gif Lectures on direct detection of Dark Matter 2

DM with ordinary matter is unfortunately far too small to be detected with current
technologies. Two main categories of DM particles can currently be probed by direct
detection experiments:

(i) Weakly Interacting Massive Particles (WIMPs) which have cross-sections with
ordinary matter driven by physics at the electroweak scale. Currently, the search
for WIMPs represents by far the strongest effort in direct detection experiments.
Several models of new physics at the electroweak scale provide such candidates,
like the neutralino in the minimal supersymetric extension of the standard model
(MSSM) or the lightest Kaluza-Klein particle (LKP) in universal extra dimension
(UED) models. The case for WIMPs is particularly strong due to the so-
called WIMP miracle: if dark matter particles are thermal relics, then from
thermodynamical considerations their current density M 0.3 implies a typical
annihilation cross-section during their freeze-out ha vi 3 1026 cm3 /s which is
characteristic of weak interactions.
(ii) Axions or axion-like particles are hypothetical pseudo-scalar particles. Axions were
initially introduced to solve the CP violation problem in the strong interaction
sector through the so-called Peccei-Quinn mechanism, but axion-like particles
constitute a generic prediction of some string models. They can constitute (non-
thermal) relics of the Big-Bang. The direct search for axion dark matter is possible
thanks to their coupling to photons, and will be briefly mentionned in Section 7.

We now concentrate mostly on the search for WIMPs, which started during the 80s after
the seminal works of, eg. [3, 4]. In the most studied models, typical WIMP masses may
range from 10 GeV to 10 TeV. For kinematic reasons, direct detection experiments
aim in general to observe nuclear recoils due to an elastic scattering of WIMPs on the
nuclei of a target. The WIMP-nucleon cross section is poorly constrained by current
measurements in frameworks like MSSM or UED . It may range typically from 106 to
1012 pb k. The last value corresponds to less than one interaction per ton of detector
per year. All the efforts of WIMP experiments are therefore devoted to the ability to
detect such a low rate of interactions.

2. Theory of WIMP direct detection

The formalism of WIMP direct detection is briefly sketched here. More details may be
found in [5, 6] .

2.1. Interaction rate

Let us consider a WIMP of mass m diffusing elastically on a nucleus with mass mN .
The nucleus recoils with an angle r with respect to the initial WIMP velocity. Since
the WIMP velocity v relative to the detector is of the order of the galactic rotation
k 1 pb = 1 picobarn = 1036 cm2
Gif Lectures on direct detection of Dark Matter 3

velocity 200 km/s, the kinematics is non-relativistic. The expression for the nuclear
recoil energy is easily found from energy and momentum conservation:
m  4 mN m

Er = v2 cos2 r
2 (mN + m )2
Typical recoil energies are in the range 1-100 keV : for WIMP searches, this requires
to use low threshold detectors, which are sensitive to individual energy deposits of this
order of magnitude. To compute the WIMP-nucleon interaction rate, one needs the
cross-section and the local phase-space density of WIMP:
For a given momentum transfert q we use the parametrization

d 0
2
= F 2 (q)
dq 4 m2r v 2
where mr is the reduced mass of the system, and F (q) is a dimensionless form
factor such that F (0) = 1. Since the maximum momentum transfert for a given
(v, mr ) is qmax = 2vmr , the parameter 0 corresponds to the total cross-section in
the case of F (q) = 1.
We note 0 the local WIMP mass density. The current observations contrain
0 0.3 GeV/cm3 . The distribution of WIMP velocities relative to the terrestrial
detector is noted f1 (v).
The interaction rate per unit mass of detector for WIMPs in the velocity range [v; v + dv]
is then given by:
 
0 d
dR = v 2 f1 (v) dv dq 2 (1)
m mN dq
After integration over the velocity distribution, this gives as a function of recoil energy
Er = q 2 /2mN :

dR 0 0 f1 (v)
Z
2
= 2
F (q) dv
dEr 2 m mr vmin v
q
mN Er
where vmin = 2 m2r
. We use a maxwellian velocity distribution for the galactic
WIMPs. Assuming that the detector is at rest with respect to the galactic halo, we
2 2 2
have f1 (v) vv3 ev /v0 . The integration is then straightforward and one finds
0
 
dR mN Er
exp
dEr 2 m2r v02
An approximately exponential recoil spectrum is therefore expected : as a consequence,
no really precise spectral signature such as a peak may be used, and in addition most
of the signal in a detector is expected at low recoil energies, which requires the energy
threshold of all WIMP detectors to be well understood experimentally.
In fact, the Earth velocity with respect to the WIMP halo must be written
as ve = v0 (1.05 + 0.07 cos t) where 1.05 v0 is the galactic velocity of the Sun and
Gif Lectures on direct detection of Dark Matter 4

= 2/1 year. The 7% modulation is due to the rotation of the Earth around the Sun.
In the former calculation, f1 (v) must be replaced by f1 (|~v v~e |) and the instantaneous
differential rate of WIMP interaction becomes
    
dR 0 0 2 vmin + ve vmin ve
= F (q) erf erf (2)
dEr 4 ve m m2r v0 v0
In the conventional model which is currently used in order to measure or constrain
cross-sections, and to compare experimental sensitivities, it is also assumed that the
WIMP maxwellian distribution is truncated for velocities larger than the galactic escape
velocity vesc = 650 km/s. This adds a small correction to the previous formula.

2.2. WIMP-nucleus cross-section for neutralinos

In this section we sketch qualitatively how the cross-section, ie. the parameters 0 and
F (q), is derived in a given particle physics model, and concentrate on the neutralino as
an important example. We refer to [5] for more details. Neutralinos interact with quarks
through the exchange of Z bosons, squarks and higgses. In the same way as for the Fermi
description of weak interactions, in the low-energy limit the neutralino-quark coupling
may be written as a coupling between currents that are of scalar, vector, pseudo-scalar,
axial or tensorial natures. Since the neutralino is a Majorana fermion, only axial and
scalar couplings exist.

2.2.1. Axial coupling : spin-dependent cross-section. The lagrangian for the axial
neutralino-quark coupling, due to Z and squark exchange, is Lq = dq ( 5 ) (
q 5 q)
where the parameter dq is calculated for a given SUSY model. At the scale of a nucleon
q 5 |ni is of the form 2 S q (n) where S
n - proton or neutron - the matrix element hn|
is the nucleon spin. The parameter q (n) depends on hadron physics and is estimated
from lepton-proton diffusion experiments. An effective neutralino-nucleon lagrangian is
therefore obtained by summation over the relevant quark flavors. We define
X
2 dq q (n) 2 2GF a(n)
q=u,d,s

With this convention, ap and an are dimensionless parameters. The neutralino-nucleon

coupling is then

Leff
n = 2 5 ) (
2 GF a(n) ( n 5 n)
The derivation of the neutralino-nucleus cross-section requires then some nuclear physics
input which we wont detail here. For a nucleus of spin J, with hSp i and hSn i being the
average spins carried by protons and neutrons respectively, the cross-section at zero
momentum transfert is

d 8 ap hSp i + an hSn i
2
(q = 0) = 2 G2F 2 J(J + 1) where =
dq v J
Gif Lectures on direct detection of Dark Matter 5

This is the so-called spin-dependent WIMP-nucleus cross section. There is an

additionnal form factor correction S(q) to take into account in the case of non-zero
momentum transfert. Quite often, it is a good approximation to say that the nucleus spin
is carried mostly by one kind of nucleon. In that case, depending on the target nucleus, a
measurement of d/dq 2 is almost a measurement of ap or an , or equivalently the proton
and neutron spin-dependent cross-sections pSD and nSD . Particularly appropriate nuclei
for spin-dependent measurements are 73 Ge (Sp = 0, Sn = 0.23) and 131 Xe (Sp = 0.04,
Sn = 0.24) for an measurements, and for example 19 F (Sp = 0.46, Sn = 0) for ap
measurements. The corresponding experiments are reviewed in section 6.1.

2.2.2. Scalar coupling : spin-independent cross-section. Higgses and squark exchanges

lead to a scalar neutralino-quark coupling Lq = fq (
q q)().
To express the neutralino-
(n)
nucleon coupling one needs the matrix element mq hn| q q|ni m(n) fT q , (n) being a
(n)
proton or neutron, and q is a quark (u, d or s). The fT q coefficients depend on hadron
physics and are obtained from pion-nucleon diffusion measurements. In addition to
coupling with quarks, there are also neutralino-gluon couplings which have an important
contribution to the scalar nucleon coupling through QCD effects, which we do not
describe here (see [5]). The resulting neutralino-nucleon coupling is of the form
(n)
X fT q
Leff
n = f(n) ()(
nn) where f(n) = fq +
mq
q=u,d,s

The additional gluon contributions to fp and fn are not explicited. It appears that
the QCD terms and the coupling to the sea quark s are often dominant in the f(n)
calculation. Therefore one has in general fp fn . For a given WIMP model, one can
therefore define a single spin-independent WIMP-nucleon coupling, independent on the
nature and spin of the nucleon. The summation over all nucleons of a nucleus is then
simpler than for the spin-dependent case. The classical nuclear density form factor F (q)
(the Fourier transform of spatial density of nucleons in the nucleus) appears, and one
gets, applying Fermis golden rule:
2
d 2 F (q)
= [Z fp + (A Z) fn ]
dq 2 v 2
We may then identify in this expression the 0 parameter defined above, and for
A2 m2r
fp = fn one finds 0 (A) = m 2 (p) 0 (nucleon). Therefore in the spin-independant
r
channel, all experiments measure approximately the same parameter 0 (nucleon) SI
independently of the target nucleus. Furthermore, contrarily to the spin-independant
case, as long as the nuclear form factor F (q) is of order unity (which is the case for low
nuclear recoil energies), there is a coherence effect which boosts the WIMP-nucleus cross
section by a factor A2 m2r (A). As a consequence, heavy nucleus targets are used in spin-
independent direct searches, and at least for the most popular SUSY neutralino models,
the spin-independent channel is currently the most sensitive to a potential WIMP signal.
Gif Lectures on direct detection of Dark Matter 6

We will therefore concentrate in the following sections on the experimental techniques

that are mostly dedicated to this channel.

2.3. Parameter uncertainties

In the framework of MSSM, predictions of the value of SI , ap and an may be achieved
by the use of exploratory scans in the cMSSM parameter space, which has 5 free
parameters (or even a wider parameter space). These scans are strongly sensitive on the
priors used for the parameter distribution (eg. flat or logarithmically distributed), as well
as on existing contraints from collider experiments and the cosmological measurement
of m . The same parameter space benchmark points as for collider searches may also
be used. For current direct detection experiments, a particularly interesting region
of parameter space is the so-called focus point where the neutralino is higgsino-like :
this favors Higgs exchanges in Feynman diagrams contributing to SI and generates a
relatively large WIMP-nucleon cross-section SI 108 pb.
While the uncertainty on the interaction rate for direct detection experiments is
dominated by the unknown WIMP physics which governs the value of cross-sections,
there are also other sources of errors due to more conventionnal physics used in the
interaction rate calculation.
(i) Uncertainties in hadronic and nuclear physics are still quite large, concerning for
example the fT s parameter or the spin content of individual nucleons in a given
nucleus. They generate uncertainties of order unity on the WIMP-nucleus cross-
section for a given WIMP model.
(ii) Uncertainties on the structure of our local dark matter halo. First, the truncated
maxwellian model for the velocity distribution is a simple approximation, with
two free parameters (its velocity dispersion related to the circular velocity at solar
radius, and the escape velocity). The presence of streams could change the function
f1 (v). Second, the local value of dark matter density is infered from measurements
of the Milky Way rotation curve and a modelization of its different components.
While the canonical value is 0 = 0.3 GeV/cm3 , there is certainly an uncertainty
of a factor 2 or 3 (see for example [7]). Furthermore other substructures may come
in addition to the smooth component of the halo. If the Solar System were in
a dark matter clump, the direct detection signal would be strongly boosted, but,
from current simulations, this is unlikely at the Solar System radius. Perhaps more
interestingly, some models of subhalo interactions within the Milky Way predict
the existence of a dark matter disk in the same plane as the Milky Way disk [8],
with a local density disk (0.25 1)halo . If such a disk exists, it could boost the
expected direct detection signals.
Gif Lectures on direct detection of Dark Matter 7

2.4. Comparison with dark matter indirect detection

In contrast to direct detection, indirect detection of WIMPs consists in observing fluxes
of secondary particles created by WIMP annihilation in various astrophysical objects.
The secondaries may be -rays, X-rays, neutrinos, or charged particles. Depending
on the sources, the fluxes may be diffuse or point-like. In the most common models,
there is in principle less undetermination in the cross-sections involved, which are
directly related to the annihilation cross-section governing the relic density m . On
the other hand, astrophysical uncertainties related to these indirect detection channels
are large : boost factors of several orders of magnitudes may arise due to the small-
scale structure of the halo, and astrophysical backgrounds to the dark-matter related
fluxes are difficult to control. As a consequence, the performances of direct and indirect
detection channels may be compared but this comparison is in general model-dependant,
and the systematics of both methods are different, making them largely complementary.
One indirect detection channel is particularly well-suited for comparison with the
spin-dependent searches for proton-WIMP coupling (pSD measurements). Models show
that the WIMP density clustered in the solar interior should be at equilibrium: the
solar WIMP density is driven by the capture rate due to diffusion on the solar protons,
which is itself proportional to the spin-dependant WIMP-proton cross-section pSD . The
annihilation of solar WIMPs generates, among others, a flux of high energy neutrinos
in an energy range where no other solar neutrino background exists. Upper limits on
the very high energy solar neutrino flux may therefore be translated into a limit on the
spin-dependant WIMP-proton cross-section. Currently, the limits on pSD from Super-
Kamiokande and Ice Cube [9] are more constraining than limits obtained by direct
detection.

3. General principles of WIMP detection, first experiments

To measure WIMP-induced nuclear recoils, detectors with a massive target and a low
detection threshold (typically a few keV) must be used. The expected signatures of
WIMP interactions are the following:
The interactions generate nuclear recoils, in contrast to electronic recoils from
radioactivity.
Due to the small interaction rate, only single interactions will be observed (no
multiple interactions).
For the same reason, WIMP interactions are uniformly distributed in the detector
volume, while interactions induced by external radioactivity, with penetration
length smaller than the detector size, do concentrate near the surface of the
detectors.
The recoil spectrum has an approximately exponential shape.
For the spin-independant channel, the interaction rate varies approximately as the
square of the target nucleus : the use of several nuclear targets can serve as a first
Gif Lectures on direct detection of Dark Matter 8

cross-check for a potential signal.

The movement of Earth in the galactic halo implies that 1) the distribution of recoil
directions is not isotropic ; 2) the recoil rate has an annual modulation. These two
signatures are studied in Sections 6.2 and 6.3.
The main challenge is the removal of environmental backgrounds due to
radioactivity or cosmic-ray induced signals. The following backgrounds are of
importance in the current WIMP searches [10]:

(i) Gamma-ray radioactivity, due to the intrinsic radioactivity of the surrounding

materials as well as the detector itself. As an example, photomultiplier radioactivity
is a major source of background for scintillating detectors. Another example is the
presence of long-lived isotopes due to the cosmic activation of Argon (39 Ar with
a 269-year lifetime) and Germanium (65 Zn, 68 Ge with lifetimes of order a year),
which generate an intrinsic background for these nuclear targets.
The intensity of radioactive backgrounds may be attenuated by several orders of
magnitude by a careful selection of all materials in the experiment as well as by
the use of shieldings (typically lead). In addition, since -rays generate electronic
recoils, most detectors of current generation are designed to discriminate actively
nuclear recoils from electronic recoils. Finally, the external gamma-ray radioactivity
may be rejected using a position reconstruction which allows to define an inner,
fiducial volume within a detector.
(ii) Beta radioactivity due to materials in the immediate neighbourhood of the
detectors. The penetration depth in Germanium ranges from 350 nm at 10
keV to 700 m at 1 MeV, so that s generate surface interactions which require
special discrimination for heat-and-ionization detectors, see Section 4.1. A major
source of radioactivity is the 210 Pb, a daughter of radon present in the air with a
22-year lifetime.
(iii) Diffusion of fast neutrons. Fast neutrons are generated by interactions of cosmic-
ray induced muons in the rocks and materials surrounding the detectors. They also
originate from intrinseque radioactivity of the rock and surrounding materials, due
to (,n) and fission reactions generated by U/Th traces. A neutron with kinetic
energy of a few MeV generates a typical 10 keV nuclear recoil after elastic
diffusion. Therefore this recoil cannot be discriminated against WIMPs like -rays.
To reduce this background, the experiments are located deep underground where
the muon flux is attenuated. Polyethylen shields are also used to thermalize the
fast neutrons coming from external sources, and hence slow them down before they
reach the detectors. In addition, contrarily to WIMPs neutrons generate multiple
interactions : multiplicity measurements are a way to reject this background.
(iv) The coherent diffusion of solar neutrinos in the detectors would completely mimic
WIMP signals (single nuclear recoils). This background is expected to be relevant
only for extremely large exposures achieved with at least ton-scale detectors [11].
Gif Lectures on direct detection of Dark Matter 9

The observation of this background would moreover constitute an interesting by-

product of WIMP searches.

Direct detection experiments belong to the category of rare event search

experiments, like experiments looking for proton disintegration or neutrinoless double
beta searches. Since several decades, deep underground laboratories were built in order
to host these experiments. As an example, the largest of them is the Gran Sasso
laboratory [12] located along a road tunnel under the Gran Sasso mountain in Italy.
Some underground laboratories are located at the place of an ancient mine. Currently,
several construction or extension projects are under way, such as the Ulisse extension of
the Modane underground laboratory [13] in the Frejus tunnel between France and Italy,
or the DUSEL facility [14] in the United States.
The idea of dark matter direct detection was elaborated in the 80s. In the end
of 80s, first limits on the WIMP-nucleon cross-section as a function of WIMP mass
were obtained using ultra-pure, semiconducting germanium detectors with an ionization
measurement. As an example, the Oroville experiment [15], initially conceived for
double-beta search, was able to set limits excluding some WIMP models that were
envisaged at that time, in particular a possible 4th generation heavy Dirac neutrino
(LEP results on the number of neutrino flavors was not there yet). These ionization-only
experiments were severely limited by their unability to reject the abundant background
of -ray induced electronic recoils. Still, improved ionization-only high purity Ge
detectors remain competitive for the low-mass WIMP search (M 5 GeV), thanks to
their remarkably low energy threshold. As an example, the CoGeNT collaboration [16]
instrumented a specific HPGe detector with an intrinsic electronic noise of 70 eV,
obtaining a sensitivity for SI of 2 104 pb for M = 8 GeV.
In the beginning of the 90s, solid scintillators were developped for WIMP search:
the target materials are NaI or CsI cristals, and scintillation signals are read with
photomultipliers. The falltimes of the scintillation signals are different on average
depending on whether the initial recoil in the crystal is nuclear or electronic. As
a consequence, a statistical discrimination is possible between -induced events and
potential WIMP interactions, based on the observed pulse shape distribution. This
allowed an improvement in sensitivity with respect to Ge ionization detectors. Still,
the impossibility to perform an event-by-event discrimination along with the presence
of anomalous events eventually limited the spin-independent sensitivity of these
detectors to SI 104 pb. Currently, solid scintillators remain competitive, first
in spin-dependent searches (section 6.1), and second for annual modulation searches
(section 6.3), since it is relatively easy to operate large masses of such detectors during
long periods.
The best sensitivities in the spin-independent channel are currently obtained with
two technologies, which we describe in the following sections: low temperature heat-
and-ionization detectors and dual-phase noble liquids.
Gif Lectures on direct detection of Dark Matter 10

4. Bolometric dark matter detectors

The principle of these detectors is to measure, in addition to an ionization or scintillation

signal, a heat signal using a thermometer on the target to detect the temperature
elevation due to a WIMP interaction. For a given recoil energy, the ionization or
scintillation yield provides a discriminating variable between electronic and nuclear
recoils. This discrimination can be achieved on an event-by-event basis.
A cryogenic bolometer is made of an absorber : a crystal, with heat capacitance
C, in contact with a thermal bath with a thermal leakage G (expressed in J/K/sec).
A thermometer measures the temperature variations of the absorber. For dark
matter searches, bolometers are used in pulse mode: after the interaction of an
incident particle, a large fraction of its energy is converted into thermal energy.
This instantaneous energy deposit E0 in the absorber generates a temperature rise
T = E0 /C. The length of these temperature pulses is the time necessary for the
cristal to thermalize with the thermal bath, of order C/G : this is typically 100 ms,
therefore only small interaction rates may be detected (this is not a problem for WIMP
searches). The bolometer sensitivity is inversely proportional to C. Since the Debye law
C(T ) T 3 applies for semi-conductors, low energy depositions may be measured only
at ultra-low temperatures, typically T 10 100 mK. Furthermore, the thermometer,
consisting of a variable resistor R(T ) usually polarized at constant intensity, must have
a temperature coefficient = d ln R/d ln T as large as possible. Several technologies
were developped, in particular:
Semi-conducting materials which have a negative , such as germanium cristals
that are neutron transmutation doped (Ge-NTD).
Transition-Edge Sensors at the temperature of their supraconducting transition,
with a large positive .
To reach stable temperatures of 20 40 mK during continuous operation of
several months, dilution cryostats must be used. This is a particularly strong
technical constraint when large masses of detectors must be implemented, especially
for bolometers, which require a low microphonic noise environment.

4.1. Heat-and-ionization detectors

Heat-and-ionization bolometers are used by the EDELWEISS and CDMS collaborations.
The absorber is a Germanium (or Silicium) cristal of typical mass 400 g nowadays, cooled
down to 20 mK or 40 mK. Electrodes cover the cristal, and a voltage of order a few
volts is applied in order to collect the ionization signals. No more than a few volts
may be applied due to the low temperature. Also, with large voltages, the heat channel
would only measure the Joule effect associated to charge drift, and would not add any
information complementary to the ionization channel.
The ionization yield measurement results in a remarkable discrimination of -ray
induced interactions within the absorber. An issue with these detectors comes with
Gif Lectures on direct detection of Dark Matter 11

surface interaction discrimination: when an interaction takes place close to the electrode
surface, the charge collection mechanism cannot be fully completed, resulting in a loss
of ionization signal. A surface electronic recoil may therefore mimic a bulk nuclear
recoil. A source of such surface events is the radioactivity due to 210 Pb, which is a
daughter element of radon. The ubiquity of radon in the air, together with the 22 year
lifetime of 210 Pb, make it difficult to reduce the presence of this background. Typical
surface rates of a few events/kg/day are observed. These events of radioactive origin
have limited EDELWEISS sensitivity to WIMP-nucleon cross-sections of 106 pb in
2003 [17].
The rejection of surface events by CDMS-II is currently based on their specific
phonon measurement. The CDMS thermometer covers the surface of the absorber
and is able to measure the athermal component of phonons generated during an
interaction, which consists in a fast signal with respect to the thermal component.
Surface interactions generate large athermal signals. A timing parameter using both the
relative heat-ionization time and the heat pulse shape is estimated for each event and is
used as a second discriminating variable, enabling the rejection of surface events at the
price of a relative loss of efficiency. In 2008, with 15 detectors in use, an effective exposure
of 121 kg.days was unblinded, with a post-cut surface event background estimated to
be 0.5 events. No WIMP candidate was observed above a recoil energy of 10 keV [18].
In the end of 2009, the unblinding of 194 more kg.days lead to the observation of
two low-energy WIMP interaction candidate events, for 0.8 0.3 background events
expected [19]. Even with these two background events, an upper limit of 3.8 108 pb
could be set on SI for m = 70 GeV, which is the best sensitivity achieved so far for
all direct detection searches.
The rejection of surface events for EDELWEISS-II is based on the recently
developped InterDigit electrode design [20]. Plane electrodes are replaced by a set
of concentric rings alternatively polarized for example at 4V and 1.5V on one side,
and the opposite on the other side. This modifies the electric field near the detector
surface. An interaction taking place in the bulk of the target generates charges only on
the collecting (4V) electrodes, while a near-surface event also creates a charge signal on
one of the field shaping (1.5V) electrodes. A rejection factor approaching 105 was
demontrated using a 210 Pb source. With 10 germanium detectors, a first limit of 107
pb could be set after 6 months of data taking in 2009 with an effective exposure of 144
kg.d and a recoil energy threshold of 20 keV [21]. One WIMP candidate was observed
at 21 keV, with a preliminary background estimation of less than 0.23 events expected
from neutron, beta and gamma interactions.

4.2. Heat-and-scintillation detectors

In heat-and-scintillation bolometers, the absorber is a scintillating material instru-
mented with a thermometer. A second much lighter calorimeter faces the main absorber
in order to detect the photon signal generated by an interaction in the target detector.
Gif Lectures on direct detection of Dark Matter 12

The scintillation yield of each event is a discriminating variable between nuclear and
electronic recoils. An advantage of this technique is the large choice of materials avail-
able as scintillating absorbers.
The CRESST-II experiment used 10 bolometers with a CaWO4 absorber to set a
limit on SI of 5 107 pb at 50 GeV. The sensitivity is limited by the presence of
a background apparently due to the low scintillation yield for nuclear recoils in these
detectors : WIMPs must be searched as heat-only interactions with no light signal.
Such no-light events are also due to other backgrounds related directly to the detector
properties that are under study [22].
More ambitious project of bolometric dark matter detectors are planned both in
Europe and in the USA. The CDMS program consists in a future installation of the
Super-CDMS setup at the new deep SNOLab laboratory, with 100 kg of detectors.
A further ton-scale experiment could be installed at DUSEL. The european project
EURECA consists in a multi-target array of cryogenic detectors installed in an extension
of the LSM laboratory. These programs would require an unprecedented control of
background noises as well as a heavy cryogenic infrastructure.

5. Noble liquid detectors

At temperatures below 165 K and 88 K, Xenon and Argon behave as dense liquids,
with good scintillation yields of 40 103 photons/MeV, as well as reasonably good
electron mobilities. An interaction in these liquids ionizes the medium and produces
excited states of the Xenon or Argon atoms. The excited states in turn generate an UV
luminescence signal. There are two excited states (singlet and triplet); for Argon, the
triplet lifetime is 1.6 s, large enough to be measured and discriminated from the singlet
state using a pulse shape analysis.
The principle of a dual-phase noble gaz TPC is the following : the Xenon (or
Argon) is kept in a vessel in thermodynamical conditions where it is in equilibrium
between liquid and gaz, the gaz phase being a small volume above the liquid phase.
The vessel is instrumented with an array of detectors, typically photomultipliers. A
strong, vertical electric field of order 1 kV/cm is applied within the whole volume. An
interaction within the liquid phase generates a first direct scintillation signal, called S1.
Depending on the ionizing power of the incoming particle, free electrons are created and
drift vertically until they reach the gaz phase. When crossing the boundary between the
two phases, a second light S2 is emitted due to the difference of amplitude of the electric
fields in both phases. From the different signals of the PMT array, and their relative
timings, the position of the primary interaction may be reconstructed. Furthermore,
since the ionization yield is smaller for nuclear recoils, the S2/S1 ratio is used as a
discriminating parameter between electronic and nuclear recoils.
Gif Lectures on direct detection of Dark Matter 13

5.1. Xenon detectors

The XENON10 experiment [23, 24] instrumented 22 kg of Xenon with 89 PMTs on
the top and bottom sides of a TPC. The Xenon was kept at 180 K and 2.2 atm with a
specifically developped pulse tube cryocooler. Calibrations show that the populations of
electronic and nuclear recoil are not completely separated by the discriminating variable
based on S2/S1: they overlap partly, therefore to reject most of the background
one must also reject 50% of nuclear recoils, resulting in a 50% WIMP search
efficiency loss. A 5.4 kg fiducial volume was defined in the bulk of the TPC using the
event position reconstruction. Since most of the radioactivity comes from surfaces (in
particular PMTs), this allows an efficient self-shielding against the background. In
2007 a WIMP search was accomplished using 136 kg.days of effective exposure in the
energy range 4.5 < Er < 27 keV. Some background remained present especially for
Er > 15 keV. These anomalous events are probably due to double interactions taking
place in the detector, among which one takes place in a dead region of the liquid Xenon
and generates S1 but no S2 signal. A limit on SI of 4.4 108 pb for M = 30 GeV
was obtained, which is still the best published sensitivity for WIMPs with masses below
44 GeV. An important systematic uncertainty on this sensitivity is the exact value of
liquid Xenon scintillation yield for low energy nuclear recoil [25]. Recent measurements
show that the XENON-10 sensitivity is more probably 5.6 108 pb for M = 30 GeV.
Similar results were obtained by the ZEPLIN-III experiment in 2008 [26]. More
massive Xenon detectors are now under construction or in operation. In particular the
XENON100 experiment [27] at the Gran Sasso laboratory is starting taking WIMP
search data. It consists of a 170 kg Xenon TPC with a fiducial volume ranging from 30
to 50 kg depending on the position cuts. The overall radiopurity of the experiment was
improved with respect to XENON10, and the side and bottom shields were equipped
with PMTs to veto multiple scatterings. A more ambitious 1 ton phase is scheduled. A
similar program is being developped in the US with the LUX experiment [28] consisting
of a 350 kg Xe TPC. A particularity of LUX will be the use of a water shielding against
and neutron backgrounds.
A single-phase liquid Xenon program is also taking place in Japan at the Kamioka
mine : the XMASS experiment consists of a 900 kg Xe sphere completely surrounded
by low-radioactivity PMTs. Since only liquid Xenon is present and no electric field
is applied, only the scintillation (S1) signal will be measured. The lack of -ray
discrimination parameters should be compensated by the large Xe mass and the full
PMT coverage with which a strict fiducial cut will be applied.

5.2. Argon detectors

Argon detectors work similarly to Xenon ones, with two main differences:
The scintillation from triplet state may be discriminated from the singlet state
thanks to a slow decay time. Since the singlet/triplet ratio depends on the ionizing
Gif Lectures on direct detection of Dark Matter 14

power of the incident particle, the pulse shape analysis of S1 signals provides a
second, independent discriminating variable to reject electron recoils.
On the other hand Argon contains naturally the 39 Ar radioactive isotope generating
a large electron recoil background in the fiducial volume of a detector. Argon
reservoirs with a low 39 Ar concentration have been found in underground cavities
and may be used in the future. Methods of depletion in 39 Ar are also investigated.
The WArP experiment [29] at the Gran Sasso laboratory has demonstrated the
efficiency of the combination of S2/S1 + pulse-shape discriminations to reject electron
recoils with a quoted rejection factor of 108. A first WIMP search with a 2.6 kg
prototype lead to a sensitivity of 106 pb. This sensitivity was limited by a large
background and a small number of instrumented PMTs. A 140 kg dual-phase Argon
TPC is now being installed at Gran Sasso, with a large coverage of low-radioactivity
PMTs and an active veto. We only mention here the existence of other large mass
projects for either dual-phase (ArDM) or single-phase (DEAP/CLEAN) argon detectors,
that are also being planned or commissionned.

5.3. Comparison with bolometers

At present, both noble liquids and germanium bolometers are the most competitive
techniques to probe the spin-independent channel in direct detection of WIMP. Both
technologies require heavy cryogenic set-ups, the cryogeny for Ge bolometers being even
more imposing for an equivalent target mass due to the necessity to cool-down massive
targets at 20-40 mK. The definition of a fiducial volume may now be done for both
noble liquids and bolometers. For noble liquids, the self-shielding effect will be more
and more efficient as the instrumented mass grows. For bolometers, since the individual
bolometer mass will be limited for heat capacitance reasons, the detector segmentation
will be important. The electronic recoil vs nuclear recoil discrimination is much less
efficient for Xenon detectors than for bolometers. This is less the case for Argon TPCs,
but current Argon detectors must face a stronger internal -ray background. While it
seems that going to larger-mass detectors will be easier with noble liquid detectors, there
is a real complementarity between these techniques which would allow for an efficient
cross-check in case of detection of a WIMP signal.

6. Complementary detection channels

Until now we focussed on the search for a spin-independent WIMP-nucleon coupling by

measuring a nuclear recoil spectrum. There are other direct detection channels, which
are complementary in the sense that a first spin-independent direct detection signal
could be confirmed by one of these channels. For the most standard WIMP models,
and with current technologies, these channels are less sensitive than the former one, but
they have advantages of their own.
Gif Lectures on direct detection of Dark Matter 15

6.1. Spin-dependent WIMP-nucleon cross-section

As described above, this component of the WIMP-nucleon interaction cross-section
depends strongly on the spin content of the target nucleus. Measurements or limits
on the WIMP interaction rate on a specific nuclear target result in a specific constraint
on the WIMP coupling coefficients ap and an , or equivalently pSD and nSD . We provide
here only a short selection of measurements, which are currently the most competitive.

The WIMP-neutron spin-dependent cross-section is currently best constrained by

the XENON10 experiment, using the same data as described in Section 5.1. The
constraint on an makes use of the 20% natural abundance of the 129 Xe and 131 Xe
isotopes. A limit of 5 103 pb was set on nSD for M = 30 GeV [30].
Solid scintillators like NaI and CsI provide competitive constraints on ap . As an
example, the KIMS experiment [31] in Korea using CsI detectors has obtained
in 2007 the best sensitivity on ap for high-mass WIMPs. As explained above,
these detectors do not provide an event-by-event discrimination of the gamma
background, but both the high exposure achieved (3400 kg.days) and the use of a
statistical pulse-shape discrimination using the falltime of signals enabled to reach
a sensitivity of 0.2 pb for pSD .
The superheated droplet technique, used by the PICASSO [32] and COUPP [33]
experiments currently provides the best sensitivity to ap at low WIMP mass. The
19
F is used as a target in, for example, liquid CF3 I for COUPP. In appropriate
temperature and pressure conditions, the interaction of a particle in the detector
creates a bubble in the target volume, only for nuclear recoils. This efficiently
suppresses the background. Temperature or pressure scans enable to constrain the
energy distribution of the observed nuclear recoils in a WIMP search. A sensitivity
of 0.16 pb on pSD for a WIMP mass of 24 GeV was recently reached by PICASSO.

6.2. Directionnality of nuclear recoils

Due to the average relative velocity of the Solar System with respect to the dark matter
halo, a strong anisotropy is expected in the WIMP-induced nuclear recoil distribution in
the direction opposite to the Solar System velocity in the halo [34]. Using the notations
of section 2, and being the angle of the nuclear recoil with respect to the direction
(, b) = (90 , 0) in galactic coordinates (which is approximately located in the Cygnus
region, hence the expression Cygnus wind), the recoil angular distribution is:

(ve cos vmin )2

 
dR
exp
dEr d cos v02
Since the Solar System velocity is of the same order of magnitude than v0 , the
modulation amplitude is large, of order 80% at appropriate recoil energies. Observing
such a modulation in a population of nuclear recoils would provide a strong evidence
Gif Lectures on direct detection of Dark Matter 16

that these recoils are indeed induced by WIMPs. To measure the direction of low-
energy nuclear recoils in a WIMP-search experiment, a favored technique is the use
of gazeous detectors. The corresponding challenge is the necessity to build very large
volumes of detectors, with a large number of associated readout channels. Several R&D
developments were made in the past years to accomplish this [35]. Let us mention two
recent achievements as examples of the current activity:
The DM-TPC collaboration, using a CF4 gazeous TPC instrumented with a CCD
camera, has observed the so-called head-tail effect for energies down to 100 keV.
The asymetry in observed traces allowed the reconstruction of the vector direction
of the recoils.
The MIMAC project has recently demonstrated the ability to reconstruct low
energy traces at 6 keV using micromegas detectors in a gazeous 3 He TPC.

6.3. Annual modulation of the recoil rate

The Earth rotation around the Sun at 30 km/s generates an annual modulation of
the Earth-WIMP gaz velocity ve , and therefore an annual variation of the shape and
amplitude of the WIMP-induced nuclear recoil spectrum, according to Eq. 2. At a
given recoil energy, the interaction rate is sinusoidally modulated. The amplitude of
the modulation depends on the recoil energy and is typically 7%. The observation of
such a modulation in a nuclear recoil spectrum therefore constitutes one more test of the
WIMP hypothesis. Since the modulation is weak, a high statistics of WIMP-induced
nuclear recoils as well as a good control of long-term detector stability are required in
order to perform such a test.
Since 1996, the DAMA experiment uses NaI(Tl) solid scintillators in order to
search for such a modulation. With these cristals, an exceptionnally high exposure
of 0.82 ton.year could be collected, using first 100 kg during 7 years, and then 250
kg during 4 years in the DAMA/LIBRA setup [36, 37]. As explained above, the -
ray background cannot be rejected on an event-by-event basis in these scintillators,
and actually the DAMA collaboration does not perform any pulse shape rejection of
electronic recoils in its annual modulation search. Therefore an integral spectrum,
including radioactivity backgrounds as well as possible WIMP interactions, is observed,
and an annual modulation signal is searched for in this spectrum for energies E > 2
keV. The combination of all DAMA/NaI and DAMA/LIBRA data shows a statistically
compelling (8 ) modulation signal at low energies, for 2 . E . 4 keV. This modulation
signal is compatible with a WIMP interpretation:
Making use of the segmentation in different cristals, the modulation is only observed
for single events.
The dependance of the modulation with energy as well as the observed phase of
this modulation, with a maximum in the beginning of June, are compatible with
WIMP predictions.
Gif Lectures on direct detection of Dark Matter 17

Still, this result remains controversial from an experimental point of view. In particular,
the modulated signal is near 3 keV, an energy where a peak at 3.2 keV associated to
40
K dominates the radioactivity background, and where the detector efficiency drops
sharply, opening the way to several possible systematic effects. The KIMS experiment
(see section 6.1), now also searching for an annual modulation in a different underground
laboratory and with different cristals (CsI), could provide a possible direct cross-check
of the modulation signal.
From a phenomenological point of view, the DAMA signal is merely compatible
with other searches, both in the spin-independent and spin-dependent channels. An
interpretation of DAMA signal as due to a spin-dependent WIMP-proton coupling is
now severely constrained by KIMS, COUPP and PICASSO. All currently competitive
bolometric and noble liquid searches exclude the DAMA signal in the hypothesis of a
purely spin-independent coupling, except for very-low mass WIMPs M < 10 GeV. In
this low-mass range, the CoGeNT experiment (see section 3) could put constraints on
the DAMA signal, but a window remains clearly open, mostly due to the uncertainties
on detector properties (such as a hypothetical channeling effect in NaI cristals) and on
astrophysical parameters (especially the local WIMP velocity distribution).
An abundant litterature therefore exists in order to interpret the potential DAMA
signal in light of other null-result searches. As an example we focus here on a plausible
phenomenological explanation which has been the subject of recent interest, named
inelastic dark matter [38] : when scattering on nuclei, WIMPs could be able to
jump to an excited state with m = m + , 100 keV. This induces a change
of kinematics with respect to the elastic scattering, suppressing the recoil spectrum
for low energies and increasing the relative annual modulation signal. The new free
parameter opens more possibilities but still the analysis or re-analysis of null-search
results, especially of XENON10 and CDMS, is able to put severe constraints on this
scenario [39, 19].

7. Direct detection of axion dark matter

Axions or axion-like particles are coupled to the photon field through the interaction
term La ga F F a = ga (E ~ B)
~ a. The coupling constant ga is not strongly
constrained for generic models, but in the case of the standard Peccei-Quinn axion
models, ga ma .
The mass of dark matter axions is constrained by cosmological and astrophysical
observations. As dark matter axions must be non-thermal relics of the Big Bang,
predictions are more model-dependant than in the case of thermal relics like WIMPs.
In general, ma 106 eV would lead to a relic density 1. For ma 1 eV, relic
axions would constitute hot dark matter.
Therefore, in the specific framework of Peccei-Quinn axions with a relic density
M = a 0.3, a region of particular interest is 106 ma . 103 eV, and the
corresponding coupling ga 1017 1013 GeV1 .
Gif Lectures on direct detection of Dark Matter 18

Current and proposed experiments using resonant cavities are exploring this region
of parameter space. Dark matter axions are searched by looking for the conversion of a
magnetic field into an electric field in a resonant microwave cavity.
As an example, in the case of the ADMX experiment [40], the resonant cavity is
filled with an 8T magnetic field, and the electric field is measured with an appropriate
low-noise readout, now using a SQUID amplifier. The electric field power is measured for
different values of the resonance frequency f of the cavity which is varied mechanically by
changing its effective dimensions. The signal from an axion with mass ma would appear
as a peak in the electric field intensity vs f for f = ma . At a given mass, the sensitivity
to ga is related to the measured electric field noise, which is limited by the readout
noise as well as the thermal noise of the cavity. Current ADMX results already explore
the Peccei-Quinn axion models in a narrow mass range 1.9 106 < ma < 3.5 106
eV. Upcoming upgrades as well as other projects such as CARRACK aim at extending
the range of mass scan as well as the sensitivity to ga .

Acknowledgments

I thank the Ecole de Gif organizers for having provided such a nice atmosphere during
the school.

References

[1] F. D. Steffen, Eur. Phys. J. C 59 (2009) 557 [arXiv:0811.3347].

[2] J. R. Ellis, K. A. Olive, Y. Santoso and V. C. Spanos, Phys. Lett. B 588 (2004) 7
[arXiv:hep-ph/0312262].
[3] M. W. Goodman and E. Witten, Phys. Rev. D 31 (1985) 3059.
[4] J. R. Primack, D. Seckel and B. Sadoulet, Ann. Rev. Nucl. Part. Sci. 38 (1988) 751.
[5] G. Jungman, M. Kamionkowski and K. Griest, Phys. Rept. 267 (1996) 195 [arXiv:hep-ph/9506380].
[6] J. D. Lewin and P. F. Smith, Astropart. Phys. 6 (1996) 87.
[7] R. Catena and P. Ullio, arXiv:0907.0018.
[8] J. I. Read, G. Lake, O. Agertz and V. P. Debattista, arXiv:0803.2714.
[9] R. Abbasi et al. [ICECUBE Collaboration], Phys. Rev. Lett. 102 (2009) 201302 [arXiv:0902.2460].
[10] J. A. Formaggio and C. J. Martoff, Ann. Rev. Nucl. Part. Sci. 54 (2004) 361.
[11] J. Monroe and P. Fisher, Phys. Rev. D 76, 033007 (2007) [arXiv:0706.3019 [astro-ph]].
[12] http://www.lngs.infn.it/
[13] http://www-lsm.in2p3.fr/
[14] http://www.lbl.gov/nsd/homestake/
[15] D. O. Caldwell, R. M. Eisberg, D. M. Grumm, M. S. Witherell, B. Sadoulet, F. S. Goulding and
A. R. Smith, Phys. Rev. Lett. 61 (1988) 510.
[16] C. E. Aalseth et al. [CoGeNT Collaboration], Phys. Rev. Lett. 101 (2008) 251301 [Erratum-ibid.
102 (2009) 109903] [arXiv:0807.0879].
[17] S. Fiorucci et al. [EDELWEISS Collaboration], Astropart. Phys. 28 (2007) 143
[arXiv:astro-ph/0610821].
[18] Z. Ahmed et al. [CDMS Collaboration], Phys. Rev. Lett. 102 (2009) 011301 [arXiv:0802.3530].
[19] Z. Ahmed et al. [The CDMS-II Collaboration], arXiv:0912.3592.
[20] A. Broniatowski et al., Phys. Lett. B 681 (2009) 305 [arXiv:0905.0753].
[21] E. Armengaud et al. [EDELWEISS Collaboration], arXiv:0912.0805.
Gif Lectures on direct detection of Dark Matter 19

[22] J. Schmaler et al., arXiv:0912.3689.

[23] E. Aprile et al. [XENON Collaboration], arXiv:1001.2834.
[24] J. Angle et al. [XENON Collaboration], Phys. Rev. Lett. 100 (2008) 021303 [arXiv:0706.0039].
[25] A. Manzur, A. Curioni, L. Kastens, D. N. McKinsey, K. Ni and T. Wongjirad, arXiv:0909.1063.
[26] V. N. Lebedenko et al. [ZEPLIN Collaboration], Phys. Rev. D 80 (2009) 052010 [arXiv:0812.1150].
[27] E. Aprile, L. Baudis and f. t. X. Collaboration, arXiv:0902.4253.
[28] S. Fiorucci et al., arXiv:0912.0482.
[29] P. Benetti et al., Astropart. Phys. 28 (2008) 495 [arXiv:astro-ph/0701286].
[30] J. Angle et al., Phys. Rev. Lett. 101 (2008) 091301 [arXiv:0805.2939].
[31] H. S. Lee. et al. [KIMS Collaboration], Phys. Rev. Lett. 99, 091301 (2007) [arXiv:0704.0423].
[32] S. Archambault et al., Phys. Lett. B 682 (2009) 185 [arXiv:0907.0307].
[33] E. Behnke et al. [COUPP Collaboration], Science 319, 933 (2008) [arXiv:0804.2886].
[34] D. N. Spergel, Phys. Rev. D 37 (1988) 1353.
[35] S. Ahlen et al., arXiv:0911.0323.
[36] R. Bernabei et al. [DAMA Collaboration], Nucl. Instrum. Meth. A 592 (2008) 297
[arXiv:0804.2738].
[37] R. Bernabei et al. [DAMA Collaboration], Eur. Phys. J. C 56 (2008) 333 [arXiv:0804.2741].
[38] D. Tucker-Smith and N. Weiner, Phys. Rev. D 64 (2001) 043502 [arXiv:hep-ph/0101138].
[39] J. Angle et al. [XENON10 Collaboration], Phys. Rev. D 80 (2009) 115005 [arXiv:0910.3698].
[40] L. D. Duffy et al., Phys. Rev. D 74 (2006) 012006 [arXiv:astro-ph/0603108].