A CsI low-temperature detector for dark matter search
G. Angloher a , I. Dafinei b , A. Gektin
c , L. Gironi d , e , C. Gotti d , A. Gütlein
f , g , D. Hauff a , M. Maino
d , S.S. Nagorny
h , S. Nisi i , L. Pagnanini h , L. Pattavina
i , G. Pessina
d , F. Petricca
a , S. Pirro
i , F. Pröbst a , F. Reindl a , ∗, K. Schäffner h , ∗, J. Schieck
f , g , W. Seidel a , S. Vasyukov
c
a Max-Planck-Institut für Physik, D-80805 München, Germany b INFN - Sezione di Roma I, I-00185 Roma, Italy c Institute for Scintillation Materials, U-61001 Kharkov, Ukraine d INFN - Sezione di Milano Bicocca, Milano I-20126, Italy e Dipartimento di Fisica, Università di Milano-Bicocca, Milano I-20126, Italy f Institut für Hochenergiephysik der Österreichischen Akademie der Wissenschaften, A-1050 Wien, Austria g Atominstitut, Vienna University of Technology, A-1020 Wien, Austria h Gran Sasso Science Institute, I-67100 L’Aquila, Italy i INFN - Laboratori Nazionali del Gran Sasso, Assergi (AQ) I-67010, Italy
a r t i c l e i n f o
Article history:
Received 1 March 2016
Revised 16 August 2016
Accepted 19 August 2016
Available online 20 August 2016
Keywords:
Dark matter
Cryogenic bolometer
Alkali halides
Transition edge sensor
Scintillation
a b s t r a c t
Cryogenic detectors have a long history of success in the field of rare event searches. In particular scin-
tillating calorimeters are very suitable detectors for this task since two signals are induced by a particle
interaction in a scintillating crystal. The thermal signal provides a precise measurement of the deposited
energy while the simultaneously measured scintillation light signal yields particle discrimination as the
amount of produced scintillation light depends on the nature of the interacting particle. We investi-
gate the calorimetric properties and background rejection capabilities of two large CsI (undoped) crystals
( ∼122 g each) operated as scintillating calorimeters at milli-Kelvin temperatures. Furthermore, we dis-
cuss the feasibility of this detection approach towards a future background-free dark matter experiment
based on alkali halide crystals, with active particle discrimination via the two-channel detection.
G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77 71
Table 1
Properties of undoped CsI and CaWO 4 including the dom-
inant emission wavelength λmax at low temperatures and
the Debye temperature �D [15–22] .
Properties CsI CaWO 4
Density [g/cm
3 ] 4.5 6.12
Melting point [ °C] 894 1650
Structure Cubic Scheelite
λmax [nm] ∼340 at 10 K 420 at 8 K
Hygroscopic slightly no
�D [K] 125 335
Hardness [Mohs] 2 4.4-5
Mean energy
of emitted photon [eV] 3.9 3.1
(
t
m
b
l
t
b
m
2
d
c
b
a
t
t
s
l
l
c
e
c
c
(
T
s
b
t
fi
P
O
T
i
o
s
a
t
I
4
i
t
l
Fig. 1. Schematics of the detector module consisting of an undoped CsI target crys-
tal and a light detector. Both detectors are read out by transition edge sensors (TES)
and are surrounded by a reflective foil.
i
a
t
w
P
m
C
t
o
s
o
2
r
f
l
g
t
c
h
s
m
t
p
3
3
F
f
t
f
c
i
e
t
u
t
r
scintillation light only), thus not exploiting an active particle iden-
ification technique via an additional and independent channel.
We believe a scintillating calorimeter based on alkali halide
aterial which provides a particle identification on an event-
y-event basis by simultaneously detecting the thermal and the
ight signal produced by an interacting particle, is a powerful tool
o study material-dependent interactions and allows to suppress
ackground. These are two distinct requirements of future dark
atter detectors utilizing alkali halides as target.
. Experimental set-up
In this work we present the results from measurements of two
ifferent commercial undoped CsI crystals operated as scintillating
alorimeters at milli-Kelvin temperatures: CsI-Hilger and CsI-ISMA
oth of cubic shape with a side length of 30mm and a mass of
bout 122 g each. All flat surfaces of both crystals were polished
o optical quality. 1 The crystal named CsI-Hilger was acquired from
he Hilger crystal company (UK), the CsI-ISMA was produced at In-
titute for Scintillating Materials in Kharkov, Ukraine. In Table 1 we
ist important properties of undoped CsI and of the crystal scintil-
ator calcium tungstate (CaWO 4 , used in CRESST-II) for comparison.
CsI is a very “soft” crystal. Any mechanical contact with the
rystal may induce stress to the crystal lattice, creating relaxation
vents that can mix up with particle events in the CsI operated as
alorimeter. Thus, we designed a holding structure where the CsI
rystal is solely in contact to a so-called carrier crystal , a thin disc
40 mm in diameter, 1.5 mm in thickness) made from CdWO 4 .
he contact between CsI and carrier crystal is realized by means of
ilicon oil. The mass ratio of CsI to carrier crystal is 8.1.
The carrier crystal is kept in its copper structure via three
ronze clamps and is equipped with a highly sensitive thermome-
er: a transition edge sensor (TES) consisting of a thin tungsten
lm (200 nm, W-TES) directly evaporated onto the carrier disc.
article interactions in the CsI create tiny temperature excursions
( μK) which are measured by the change in the resistance of the
ES. TESs are operated between the normal and the superconduct-
ng phase. A dedicated heater is used to stabilize the temperature
f the W-TES and crystal at the desired operating point in the tran-
ition. The heater consists of a gold stitch bond (gold wire with di-
meter of 25 μm), which is attached to a small gold structure on
he TES. The gold also serves as thermal link.
The CsI crystal is facing a cryogenic light detector of CRESST-
I type [4] , which consists of a thin sapphire disc (thickness of
60 μm and diameter of 40 mm). A 1-μm-thick layer of silicon
s epitaxially grown onto the sapphire disc to absorb the blue scin-
illation light of CsI. We refer to light detectors of this type as SOS
ight detectors (Silicon on Sapphire).
1 A more precise discussion on the surface treatment is given later in Section 4.2 .
t
t
Like the CsI, the light absorber is read out by a W-TES, which
s optimized in shape and dimension for the light detector. Crystal
nd light detector are enclosed in a reflective housing (Lumirror ®)
o maximize the light collection efficiency. The dedicated W-TESs
ere produced in collaboration with the Max-Planck-Institute for
hysics in Munich, Germany. A detailed scheme of the detector
odule , the ensemble of cryogenic light detector and scintillating
sI crystal, is depicted in Fig. 1 .
The whole series of measurements was carried out in the
est facility of the Max-Planck-Institute for Physics at the Lab-
ratori Nazionali del Gran Sasso (LNGS), a deep underground
ite (3600 m.w.e. [23] ) in central Italy. The facility is composed
f a dilution refrigerator that is laterally surrounded by about
0 cm of low-background lead to reduce the environmental β/ γ -
adioactivity. In order to ensure a low noise condition, a plat-
orm attached by a spring to the mixing chamber of the di-
ution unit mechanically decouples the detector from the cryo-
enic facility. The set-up is a double stage pendulum, similar to
he one described in [24] . The TESs are operated with commer-
ial dc-SQUID electronics (Applied Physics Systems company). The
ardware-triggered signals are sampled in a 164 ms window at a
ampling rate of 50 kHz. Both detectors are always read out si-
ultaneously, no matter of which one triggered.
For a detailed descriptions of the DAQ, the control of detec-
or stability and the pulse height evaluation and energy calibration
rocedures the reader is referred to [25,26] .
. Detector performance
.1. Data processing
Both crystals are mounted according to the set-up shown in
ig. 1 . In all measurements the crystals are permanently irradiated
rom the lateral side with an
55 Fe X-ray source. Additionally, an ex-
ernal and removable 241 Am source emitting 59-keV- γ -rays is used
or a γ -calibration.
In one particular measurement dedicated to study particle dis-
rimination, a Ra-source is additionally placed close to the crystal,
n position identical to the 55 Fe source.
Heater pulses are injected every 4 s to control stability, to lin-
arize the detector response 2 and to calibrate the detectors down
o threshold energy. The overall count rate of these detector mod-
les is about 1 Hz, dominated by natural radioactivity arising from
he non-radiopure materials of the cryostat.
Three basic cuts are applied to our data: right-left baseline pa-
ameter, pulse shape differences with respect to the template pulse
2 For a TES the resistance change is to very good approximation proportional to
he energy deposit. Using heater pulses allows to measure the small deviations from
his linearity and, hence, to maximize the energy resolution.
72 G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77
Fig. 2. Background data of an undoped CsI crystal (0.458 kg-days of exposure;
CsI-Hilger) in the light yield-energy plane. In between the solid lines 80% of the
events of the corresponding event class are expected; blue for electron recoils, red
for nuclear recoils off Cs or I, green for potential events not producing scintillation
light. The dashed lines mark the mean values of the central 80% probability bands.
A trigger threshold as low as 4.7 keV is achieved within the present detector set-
up, see text for detailed information. (For interpretation of the references to colour
in this figure legend, the reader is referred to the web version of this article.)
e
b
a
e
o
f
g
e
[
r
c
w
t
i
c
p
a
d
1
b
m
d
o
i
K
i
e
via the RMS of the template fit, and rise time of the pulse via the
peak position-onset parameter. The right-left baseline parameter
is defined as the difference between the average level of the last
and of the first 50 samples of a record. It allows to remove a large
part of so-called decaying baselines, which follow pulses induced
by large energy depositions that do not relax back to equilibrium
before the reactivation of the trigger. Additionally, this parameter
discards artifacts from the readout electronics causing a change of
the baseline level within the record. A cut on the RMS of the tem-
plate fit of the registered particle pulse removes any pulse devi-
ating from the nominal pulse shape (e.g. events with a strongly
tilted baseline) and/or an enhanced noise level. Moreover, the class
of events that certainly has to be identified and removed from the
data are events originating from interactions taking place in the
carrier, so-called carrier events . The fraction of carrier events that
survives the RMS cut can be alternatively spotted by the rise time
of the pulse, via the peak position-onset parameter [27] .
3.2. Calibration
The CsI crystal is calibrated with γ s from an externally applied241 Am source. In the case of the light channel an equivalent energy
of 59.5 keV ee is assigned to the scintillation peak in the energy
spectrum of the light detector induced by the 241 Am γ s in the CsI.
To denote that the scale is only valid for electron recoils producing
scintillation light in the CsI the subscript ee (electron-equivalent)
is added to the energy scale. 3
To additionally allow for a direct energy calibration of the light
detector, a weak 55 Fe X-ray source (electron capture with subse-
quent de-excitation via X-ray emission) was faced directly to the
light absorber. The energy resolution achieved for the 55 Mn K α-
and K β -lines is σ = 90.7 eV. The RMS of the baseline of the light
detector is about σ = 20eV. 4 The same light detector is utilized for
all data presented within this work.
3.3. Results of the CsI-Hilger crystal
Fig. 2 depicts about 0.458 kg-days of background data in the
light yield versus energy plane. Light yield we define as the ra-
tio of energy detected in the light detector expressed in keV ee
(electron-equivalent) and the energy deposited in the CsI crystal
in keV. 5 Thus, electron recoils get assigned a light yield of one. For
other particle types, as mentioned later in the text, the respective
quenching factors, which quantify the reduction in light output for
different particles, have to be considered.
The regions in the light yield-energy plane where different
event types are expected are described by bands. A band is defined
by two functions, both depending on energy: the mean light yield
depending on the type of particle (quenching) and the width set by
the finite resolution of the detectors. Since the e −/ γ -band is highly
populated, mean and width can be determined precisely by a fit to
the data. With the parameters found for the e −/ γ -band and the
quenching factor (QF) for the various event types the correspond-
ing quenched bands are calculated. The QF is defined as the ratio
of scintillation light produced by an interaction of a particle of a
certain type to the scintillation light produced by a γ of the same
3 For other particle types the respective QFs have to be considered, see
Section 3.3 . 4 Light detectors used in CRESST-II experiment have demonstrated an RMS of
baseline as good as σ ∼ 5eV. Thus, it is possible to further improve the perfor-
mance of the light detector by improving the electronic noise condition at the test
facility. 5 Since the energy measured in the crystal is for this measurement practically
independent of the particle type the index ee can be dropped for the phonon de-
tector.
w
e
c
t
c
y
nergy. We restrict ourselves to this qualitative description of the
ands since a more detailed explanation and the validity of this
pproach was already proven, e.g. in [4,28] .
In Fig. 2 we show the bands where possible event classes are
xpected for the CsI-Hilger crystal. In between the solid lines 80%
f all events of the corresponding event class are expected; in blue
or electron recoils, in red for nuclear recoils off Cs or I and in
reen for potential events not producing scintillation light. 6 The
nergy-dependent QFs for Cs- and I-recoils are taken from Tretyak
29] .
At low energies the scintillation light output is no longer di-
ectly proportional to the deposited energy. For non-alkali halide
rystals (like e.g. CaWO 4 ) the relative light output decreases to-
ards lower energies. Alkali halide crystals, on the other hand,
ypically show an increase down to ∼10 keV before decreasing go-
ng to even smaller energies [31] . A comparison of the two CsI
rystals measured in this work with CaWO 4 and CdWO 4 is de-
icted in Fig. 3 showing the mean light yield for e −/ γ -events as
function of energy. For both CsI-crystals (Hilger and ISMA) we
etermine the light yield at 10 keV to be O(10%) higher than at
00keV, in particular visible as the tilt of the mean line in Fig. 3 ,
ut also obvious in Fig. 2 and 7 . This observation is in good agree-
ent to other measurements of CsI (undoped) [32] .
In Fig. 4 we show an energy spectrum of the scintillation light
etected by the SOS light detector. The peak at around 6 keV ee
riginates from scintillation light emitted by the crystal due to its
rradiation with X-rays from an
55 Fe source (K α = 5.9 keV and
β = 6.5 keV). The physical position of the source is indicated
n Fig. 1 . The light energy resolution at this peak is (727.3 ± 6.3)
V ee . 7
The presence of an additional 55 Fe X-ray source, oriented to-
ards the light absorber, gives us the possibility to gain a direct
nergy calibration of the light detector. With this information we
an calculate the total amount of energy detected in the light de-
ector of our set-up. We find that for CsI-Hilger about 8.1% of the
6 Since the QF for Cs and I is very similar, only one band is shown for reasons of
larity. 7 The distance between the two X-ray lines is � 1 σ , thus fitting only one Gaussian
ields a conservative estimate on the energy resolution.
G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77 73
Fig. 3. Mean light yield of e −/ γ -events for different crystals as a function of energy
deposited in the crystal. The two magenta lines correspond to the two CsI crystals
analyzed in this work (dashed for CsI-ISMA and dotted for CsI-Hilger). In addition,
the mean light yields for CdWO 4 (solid blue) and CaWO 4 (dashed dotted red) are
drawn (data from [30] ). By convention, all lines are normalized to a light yield of
one at 122 keV - the energy of γ s from a 57 Co source, typically used for calibration.
(For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
Fig. 4. Energy spectrum of the SOS light detector for the scintillation light produced
due to particle interactions in the CsI-Hilger crystal. The peak at around 6 keV ee is
due to the scintillation light emitted from the CsI-crystal due to the absorption of
X-rays originating from a dedicated 55 Fe X-ray source. The data correspond to an
exposure of about 0.986 kg-days (sum of background and calibration data).
e
t
o
t
r
h
t
t
m
t
a
l
f
t
d
T
t
h
Fig. 5. Coincident pulse as recorded with the CsI detector module. The phonon and
light signal are colored in red and blue respectively. This event is due to an X-ray
from an 55 Fe source (5.9 keV) and being absorbed in the CsI crystal. (For interpre-
tation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
p
i
w
n
p
b
o
o
c
c
h
i
fi
s
v
[
m
s
m
m
i
i
a
l
c
d
I
a
X
t
p
b
n
γ
t
e
nergy deposited in the CsI crystal is detected in form of scintilla-
ion light.
The physical properties of alkali halide crystals per se are not
ptimal for the application as cryogenic calorimeters: the produc-
ion, propagation and thermalization of phonons, is best for mate-
ials with high Debye temperature, hence materials with generally
igh melting point and hardness.
In cryogenic scintillating calorimeters using a crystal scintilla-
or with a relative high Debye temperature, thus suitable phonon
ransportation properties, the energy threshold of the detector
odule is typically driven by the performance of the phonon de-
ector. 8 In other words, particle discrimination in such calorimeters
t some point breaks down due to the limited sensitivity of the
ight detector or rather the amount of produced scintillation light
or the given particle type at given energy deposit. For the CsI de-
ector module the situation is somehow reversed. The amount of
8 In a CRESST-II CaWO 4 detector with similar layout typically ∼ 2% of the energy
eposited in the crystal is detected in form of scintillation light for β/ γ -events.
hus, in comparison to the phonon detector the sensitivity of light detector needs
o be about a factor 100 higher for β/ γ interactions; for other particle types the QF
as to be considered in addition.
t
f
fi
c
roduced scintillation light is increased by a factor of about two
n comparison to e.g. a CaWO 4 crystal. However, at the same time
e observe generally small pulse amplitudes in the phonon chan-
el. The small signal amplitudes may find an explanation in the
hysical properties of CsI (also see Table 1 ). Typically a low De-
ye temperature is not suitable for a calorimeter, a detector relying
n good phonon propagation properties. Instead excellent results
n signal amplitude and energy resolution are observed e.g. using
alorimeters made of sapphire crystals having a �D = 1041 K [33] .
Furthermore, since we apply the carrier - absorber detector
oncept, phonons cannot directly propagate to the W-TES but first
ave to be transmitted between the two carefully bonded dissim-
lar crystalline media (CsI → silicon oil → carrier crystal → metal
lm (W-TES)). In this the phonons will experience a boundary re-
istance. On such interfaces phonons can get reflected and con-
erted, as described by the acoustic impedance mismatch model
34] .
As mentioned before, the Debye temperature �D is a good first
easure on the quality of the phonon transmission between solid-
olid interfaces (see Table 1 ). The more the �D differ for two bond
aterials the lower the transmission probability. A qualitative esti-
ate of the transmission probability from material 1 to material 2
s given by the ratio of speeds of sound (v m 1 / v m 2 ) 3 [35] . Estimat-
ng the transmission probability for the material combination CsI
nd CaWO 4 by using the simple ratio and the velocity values from
iterature [36] results in a transmission of only about 50%. 9
The pulses depicted in Fig. 5 are recorded simultaneously and
orrespond to the heat (red) and scintillation light (blue) signal
ue to an energy deposition of an
55 Fe X-ray in the CsI crystal.
t is intuitively clear from Fig. 5 that the light detector has still
definite signal whereas the heat signal induced by the 5.9 keV
-ray in the crystal is already close to baseline noise. To estimate
he trigger threshold we use pulses which are created by superim-
osing a signal template, scaled to the desired energy, and empty
aselines. The signal template is acquired by averaging a large
umber of pulses of the same deposited energy (usually from a
-calibration peak) and, thus, provides a noise-free description of
he detector response to an energy deposition. Determining the en-
rgy resolution from these simulated pulses is a precise measure of
he impact of the baseline noise, which also is the decisive factor
or the achievable trigger threshold. For the CsI-Hilger crystal we
nd a resolution of σ = (946 eV ± 5 eV(stat.)). This method was
9 Since values for CdWO 4 are absent in literature, the values for the very similar
rystal CaWO 4 are used for the calculation instead.
74 G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77
Fig. 6. Energy spectrum of the deposited energy recorded in the CsI-ISMA crystal
in about 0.254 kg-days of exposure. The line from the external 241 Am γ -source is
detected at 59.89 keV ( σ= 3.1 keV). Supported by a simulation, we identify the
small peak at about 30 keV to originate from X-ray emission of iodine.
C
a
t
l
4
o
b
S
a
d
i
C
p
S
r
fl
4
c
t
[
i
f
-
t
2
h
T
h
b
d
a
α
w
t
e
u
s
4
f
s
t
o
a
d
H
t
11 In comparison to VM2002 (3M company) the Lumirror ® material does not scin-
tillate, but only works as a reflector.
validated by the CRESST experiment [6] proving long-term stability
for a threshold setting of five times the baseline resolution, which
in our measurement corresponds to 4.7 keV.
Alternatively, the baseline resolution of a cryogenic calorime-
ter can also be extracted from the energy resolution of injected
heater pulses. This approach is more conservative, for the measure-
ment discussed here we find an energy resolution of σ = 979 eV
for the smallest injected test pulse with an equivalent energy
of 7.9 keV. The estimated threshold value of 4.7 keV is indicated
as a vertical dashed line in Fig. 2 . The events observed at ener-
gies below are mostly attributed to triggers of the light detector
which simultaneously toggles the read-out of the phonon detector.
However, it should be emphasized that for events with only the
light detector triggering, nuclear recoil events are strongly disfa-
vored with respect to e −/ γ -events, due to the quenched light sig-
nal of nuclear recoils.
3.4. Results of CsI-ISMA crystal
The performance of the CsI-ISMA detector module is compara-
ble to the CsI-Hilger detector discussed before. We interpret this
as a positive sign, since the two crystals were produced in differ-
ent growth facilities, starting from different raw materials. Only the
light detector was affected by some additional microphonic distur-
bances induced by the cryostat, resulting in a slightly worse energy
resolution. The energy spectrum of the CsI-ISMA crystal acquired
in about 0.254 kg-days of exposure and in presence of an
241 Am
γ -source is shown in Fig. 6 . A peak is observed with a fitted mean
of (59.89 ± 0.10) keV ( σ = (3.11 ± 0.14) keV). This peak corre-
sponds to the 241 Am γ -line. 10 The smaller peak at about 30 keV
is attributed to the emission of X-rays from iodine.
We find the energy resolution of the smallest injected heater
pulse, with an equivalent energy of 7.0 keV, to be σ = 702 eV.
This value is slightly better than the resolution determined for
the CsI-Hilger crystal. Consequently, the CsI-ISMA crystal exhibits
a slightly lower estimated trigger threshold of about 3.5 keV. For
CsI-ISMA the fraction of energy detected in form of scintillation
light is about 6.5%, thus ∼20% less than the CsI-Hilger. Chemical
impurities in the crystal are amongst others known to suppress
scintillation in crystals [38] . In Section 5 we discuss the presence
of impurities in the CsI crystals on basis of the results from an
ICP-MS (Inductively Coupled Plasma Mass Spectrometry) analysis
of samples taken from these crystals.
10 Literature value for gammas from
241 Am is 59.54 keV [37] .
e
p
l
To summarize, the two cryogenic measurements carried out on
sI-Hilger and CsI-ISMA prove that CsI can be successfully oper-
ted as a scintillating cryogenic calorimeter with a threshold, in
he case of e.g. CsI-Hilger as low as 4.7 keV and for CsI-ISMA as
ow as about 3.5 keV, in the present set-up.
. Particle discrimination
In order to study the response of the CsI to particles recoiling
ff the Cs and I target nuclei, the most suitable solution would
e to irradiate it with neutrons from an external neutron source.
ince, at the time of the measurement, such a source was not
vailable, we decided for a first measurement with a source pro-
ucing α-particles and accordingly recoiling nuclei, the latter be-
ng of interest for the demonstration of particle discrimination in
sI. The crystal used for this measurement is CsI-Hilger. The ex-
erimental set-up is identical to the measurement described in
ection 3.3 , only a small aluminum foil (2 cm
2 , 30 μm thick) car-
ying the source was inserted and attached onto the Lumirror ® re-
ector, as indicated in Fig. 1 . 11
.1. Source
The source was made by exposing an aluminum foil in vacuum
onditions to a surface-implanted
228 Th-source, thus having a cer-
ain chance for implanting 224 Ra nuclei onto the aluminum foil
39,40] . The penetration depth of a 224 Ra nucleus of e.g. 90 keV
n aluminum is O(100 nm) [41] . 224 Ra, with a half life of 3.66 d
urther α-decays to 220 Rn followed by 216 Po, 212 Pb, 212 Bi and
212 Po
the fifth and last α-decay in the chain to stable 208 Pb.
Given the sequence of α-decays, there is a realistic probability
o have nuclear recoils escaping from the aluminum foil. Especially24 Ra atoms which were implanted at a shallow depth have the
ighest probability to provide nuclei that can escape the source. 12
hese slow and heavy recoils interacting in the CsI mimic the be-
avior of a recoiling nucleus of the lattice itself. The energy distri-
ution of such recoil events is expected to reach from ∼100 keV
own to threshold energy of the CsI detector. For a more precise
nswer a simulation would have to be performed on the various
-decays and their density profile in the aluminum foil. However,
ithout having information on the surface condition of the foil and
he CsI crystal, thus not having in hand the correct starting param-
ters, we do not have a reliable basis to start such a complex sim-
lation which completely relies on surface properties, due to the
hort range of nuclei in such material.
.2. Experimental results
Combining both a theoretical approach to the detector response
rom nuclei as e.g. Rn and Po and experience of QFs measured for
uch heavy nuclei in other scintillating materials at low tempera-
ures [4] , we would expect the recoil-events to accumulate in form
f a narrow band at a low light yield value. Such a band for Cs
nd I recoils was already shown in Fig. 2 , whereby the energy-
ependent QFs for Cs and I were adapted from Tretyak [29] .
In Fig. 7 data of an exposure of about 0.163 kg-days of the CsI-
ilger module is shown in the light yield-energy plane. Apart from
he 224 Ra recoil source an external 241 Am γ -source was present
12 At the same time the source produces α-particles with a higher rate and a good
nergy resolution of the α-lines, thanks to the shallow implantation profile of the
rimary 224 Ra nuclei (max. ∼ 550 nm) being almost negligible in sense of energy
oss considering few MeV αs with an energy loss of O(100 keV/ μm).
G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77 75
Fig. 7. Data of CsI-Hilger in the light yield-energy plane of an exposure of about
0.163 kg-days. During the measurement time the CsI was constantly exposed by γ s
of an external 241 Am source as well as by recoils induced by a 224 Ra source, directly
faced to the crystal (see Fig. 1 ). The bands are colored in analogy to Fig. 2 , blue for
e −/ γ -events, red for recoils off Cs and I and green for potential phonon-only events
not producing scintillation light. The upper and lower cyan dashed lines mark the
5 σ -boundaries of the e −/ γ -band and the phonon-only band. The events in between
the cyan lines (marked with a red circle) are therefore statistically incompatible
with leakage from the respective bands, thus providing reasonable evidence to be
induced by heavy nuclei (Rn/Po from
224 Ra source) interacting in the very near-
surface layers of the CsI. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
d
t
l
t
e
o
d
a
b
r
p
e
o
i
t
o
s
p
F
a
a
t
m
w
p
T
o
e
p
c
c
m
1
b
g
f
p
d
e
i
c
o
t
o
o
a
n
l
f
s
c
t
a
t
p
e
m
a
b
c
t
p
o
c
s
e
o
g
t
m
s
o
l
s
H
a
k
o
r
t
i
i
s
[
u
n
uring the whole data taking for the energy calibration of the de-
ectors. 13
Looking at Fig. 7 the highly populated event-distribution at a
ight yield (LY) around 1 can be assigned to e −/ γ -interactions in
he CsI. The color code for the band description, blue for e −/ γ -
vents, red for recoils off Cs and I and green for potential phonon-
nly events not producing light is chosen identical to Fig. 2 . In ad-
ition, we display cyan-colored dashed lines which mark the lower
nd upper 5 σ -boundary of the e −/ γ -band and the phonon-only
and, respectively. Thus, the distribution of events (marked with a
ed circle) in between the cyan lines are statistically neither com-
atible with leakage from the e −/ γ -band, nor to the phonon-only
vents band. 14 Two characteristic features of these events may be
bserved. Firstly, the broad light yield distribution (0.2 � LY � 0.6)
s not in accordance with a band-like interpretation and, secondly,
he energies of most events are between threshold and ∼90 keV
nly. Solely two events are present at slightly higher energies but
ignificantly higher light yield.
Since this class of events (marked with a red circle) is not
resent in the previous measurements without recoil-source (see
ig. 2 exposure of a factor 2.8 larger), detector specific artifacts and
strong leakage-component from e −/ γ -events can be excluded in
dvance, moreover as the same crystal was used. To evaluate if
hese events can be due to 220 Rn recoiling off Cs or I nuclei, we
ay investigate the α-particles produced by this source. In total
e find 2106 α-events in the complete data set from the four ex-
ected lines at 5685 keV, 6288 keV, 6778 keV and 8784 keV.
he events in the cloud-like distribution count 19, thus a fraction
f solely 0.9%, in comparison to the total number of recorded α-
vents. Recoiling nuclei can travel only O(100 nm) in aluminum
13 Due to the short half life of 224 Ra the whole measurement, including detector
reparation, cool-down of the cryostat and the optimization of the detectors in-
luding energy calibration had to be carried out in a most efficient and least time
onsuming way. Furthermore, the operating points of the detectors were not opti-
ized at the level of the previous measurement ( Fig. 2 ) due to lack of time. 14 In total there are 5003 events in our data set with energies lower than
30 keV.
i
s
y
n
C
o
2
efore they are stopped, however the range of α-particles with the
iven energies is few O(10 μm). Thus, given the range differences
or nuclear recoils and α-particles, we expect only such a small
ercentage.
That heavy nuclei interacting in the near-surface layers in un-
oped CsI can in fact reveal a peculiar LY is further supported by
xperimental observations discussed in [42–44] : they observe an
ntensive blue luminescence in CsI due to the formation of vacan-
ies as a consequence of e.g. plastic deformation of the crystal. The
ptically polished surface of our CsI sample implies a mechanical
reatment of the surface, known to damage the near-surface layers
f CsI. Such near-surface layers exceed - typically by many orders
f magnitude - the equilibrium concentration of vacancies, which
re known to change the correlation between the various compo-
ents of the luminescence, resulting in an additional and intense
uminescence contribution in the range of 440–550 nm [43] . De-
ect surface layers are difficult to remove, especially in case of ab-
ence of a cleavage plane, as in CsI. In [44] a surface-removing pro-
edure is discussed which relies on two important steps: first, af-
er mechanical surface treatment of the CsI it has to be stored in
mbient air (relative humidity above 30%) to allow for recombina-
ion/relaxation of vacancies ( O(days)). Second, still remaining im-
urities and defects may be removed by chemical polishing using
.g. Methanol.
The CsI-Hilger underwent a partially similar procedure: after
echanical polishing of the surface it was chemically treated with
mixture of alcohols in order to remove the defect layer induced
y the machining. But, the relaxation time for near-surface va-
ancies in ambient air was not respected, hence, there is a cer-
ain chance to still have defect layers, even after the chemical
olishing.
Thus, there is reasonable evidence that the events induced by
ur 224 Ra recoil-source can create events in the CsI which are ac-
ompanied by a larger scintillation light signal, in particular since
uch nuclei are stopped in CsI within the first O(100) atomic lay-
rs, depending on the initial energy of the nucleus. The short range
f such nuclei is also an excellent measure for the overall homo-
eneity of the defect surface, present in our sample.
From the measurement we can also conduct an upper limit on
he thickness of the defect layer present in the CsI samples. In all
easurements we irradiate the CsI with X-rays from
55 Fe. The ab-
orption length of 5.9 keV X-rays in CsI is ∼5 μm. As we do not
bserve a distorted scintillation peak from
55 Mn K α and K β in the
ight detector (see Fig. 4 ) we can acknowledge that the defect near-
urface layer does not exceed a thickness of about 5 μm in CsI-
ilger and CsI-ISMA.
In the frame of introducing cryogenic calorimeters based on an
lkali halide target for the search of dark matter particles such
ind of surface effect is not of concern. First, as discussed previ-
usly, Kudin et al. [44] started to develop a recipe to successfully
emove defect surface layers. Second, α-induced background has
o be prevented in any case, as nuclear recoils can mimic WIMP
nteractions. This can be achieved by using a completely scintillat-
ng housing [45] of the detector or an active shielding to tag the
imultaneously emitted α-particle in order to reject such events
27] . Third, dark matter interactions are expected to take place
niformly in the volume, thus practically are not affected by very
ear-surface properties. From another point of view the observed
ncreased surface luminescence may be used in order to reject any
urface-related background, a powerful tool in rare event searches.
To summarize, we observe a distribution of events at low light
ield values which in number reasonably agree with the expected
umber of recoils induced by the 224 Ra recoil-source faced to the
sI. This is further supported by the observation of such events
nly up to an energy of about 100 keV, the maximum energy a24 Ra recoil can acquire from the decay. The broad distribution in
76 G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77
Table 2
Impurities in CsI-Hilger and CsI-ISMA from re-
sults of ICP-MS analyses. Uncertainties of given
values are at 30%. Limits are computed at 68%
C.L.
Element Unit CsI-Hilger CsI-ISMA
Tl [ppb] 2 .9 1 .9
Na [ppm] 0 .16 0 .92
Ca [ppm] 13 .3 3 .3
K [ppm] 8 .08 3 .88
Cr [ppb] < 19 < 17
Fe [ppb] < 375 < 167
Co [ppb] < 2 < 2
Ni [ppb] < 28 < 25
As [ppb] < 8 < 7
Mn [ppb] < 19 < 5
l
f
i
i
i
4
s
i
c
t
s
l
f
t
t
T
o
t
i
6
m
g
a
n
i
d
t
m
b
e
t
s
N
D
n
T
w
n
p
(
t
e
d
t
i
d
s
[
A
d
light yield can have its explanation in the presence of defect layers
which, as discussed in literature, are known to show an increased
scintillation light production due to the high amount of available
vacancies. Thus, there is room to interpret the observed event dis-
tribution being due to nuclear recoils, hence proving the capability
of particle identification in CsI (undoped) by making use of the si-
multaneous detection of the thermal and the light signal.
However, a measurement with a neutron source is needed to
quantify the discrimination power that can be achieved in this ma-
terial. Such a measurement is indispensable in order to judge the
potential of alkali halide crystal for rare event searches in detail.
At last we want to refer back to the background measurement
of CsI-Hilger discussed in 3.3 and displayed in Fig. 2 . In these data
we find one event without an associated light signal at around
14 keV, a so-called phonon-only event . Since we expect an in-
creased light yield from such surface nuclear recoil events induced
by surface α-decays (see 4.2 above), we expect this event to result
from lattice relaxation due to the not perfect surface quality of the
CsI-Hilger crystal.
There are few events detected slightly above the upper 80% line
of the almost empty nuclear-recoil band (red) as well as three
events around a light yield of 0.5. In fact, since no passive neu-
tron shielding is available, we expect to see neutron-induced recoil
events in our data set. 15 The overall high population of the e −/ γ -
band has its explanation in the shielding situation of the test fa-
cility. The cryostat itself is not made from radiopure materials and
the external 20 cm lead shielding is not enclosing the set-up com-
pletely, but leaving the top part uncovered.
5. Chemical purity of CsI crystals
In order to investigate the amount of chemical contaminations
in the CsI-Hilger and CsI-ISMA material we performed material
screenings using ICP-MS carried out on-site at LNGS.
Especially, we are interested in understanding to which extent
the discrepancy in the observed amount of scintillation light for
CsI-Hilger and CsI-ISMA ( ∼ 20%) originates from chemical impuri-
ties, known to harm the scintillation performance (mainly involv-
ing contributions from V, Cr, Mn, Fe, Co and Ni [38] ). Last but not
least also the bolometric performance of the CsI can be affected
by adding an additional contribution to its heat capacity (involving
ferromagnetic and paramagnetic elements [46] ).
Table 2 lists the measured concentrations of chemical impuri-
ties present in the two samples. We find very low limits, at ppb
15 From a sequence of measurements carried out in the test cryostat the neutron
count rate is estimated to 2–3 neutrons/(kg day).
e
t
evel, for Cr, Fe, Co, Ni, As and Mn. The amount of potassium is
ew ppm, as expected due to the chemical affinity of K and Cs. 16
In comparison to other elements, the content of Ca is different
n the samples, with about 13 ppm in CsI-Hilger and about 3 ppm
n CsI-ISMA. Doping of CsI with divalent ions (Me 2+ ), as e.g. Ca,
s known to introduce an additional blue emission band around
15 nm [47] , thus an enhancement in scintillation emission.
As for Tl and Na, typically used as dopants in CsI, we find only
purious traces of Tl in both samples. Thus, special care was taken
n avoiding such contaminations by using for example a dedicated
rucible. While doping CsI with Tl and Na is known to increase
he light yield at room temperature, it will suppress the intrinsic
cintillation, the mechanism responsible for the light production at
ow temperatures.
The measured amounts of Na and Ca in CsI-ISMA are about a
actor of ∼ 5 higher and ∼ 4 lower, respectively, than in CsI-Hilger,
hough it should be mentioned that standard doping levels are in
he range of 10 0 0 ppm, thus three orders of magnitude higher.
he presence of 1 ppm of Na and about 3 ppm of Ca can, from
ur understanding, have only a marginal effect on the overall scin-
illation light production, possibly insufficient to explain the miss-
ng ∼ 20% in CsI-ISMA in comparison to CsI-Hilger.
. Perspective
In this manuscript we report the successful detector develop-
ent and first results of a scintillating calorimeter using CsI as tar-
et, a crystal belonging to the family of alkali halides. CsI might be
n interesting target for dark matter search by itself, even if we do
ot follow this path right now. In particular, if some other exper-
ment apart from DAMA/LIBRA finds hints for dark matter many
ifferent tar get materials will be needed to confirm this poten-
ial finding and to characterize the interaction properties of dark
atter particles. Then CsI, with only rather heavy nuclei, might
e of high interest. Our measurements show that CsI can be op-
rated as cryogenic detector with particle discrimination and we,
herefore, think that CsI is of relevance for direct dark matter
earch.
In the very near future we aim to switch from CsI to
aI, the target material used by the DAMA/LIBRA collaboration.
AMA/LIBRA observes a statistically robust annual modulation sig-
al which can be interpreted as a signature for dark matter [8] .
he development of a first prototype NaI scintillating calorimeter
ill be carried out within COSINUS (Cryogenic Observatory for SIg-
atures seen in Next-generation Underground Searches), an R&D
roject funded by CSN5 of Istituto Nazionale di Fisica Nucleare
INFN) and located at LNGS in Italy.
In case the performance of such NaI calorimeters can be proven
o be comparable to already existing scintillating calorimeters (en-
rgy threshold of 1 keV and lower), as e.g. used in the CRESST
ark matter search, such cryogenic NaI detectors have the potential
o give an answer on the particle interaction channel participat-
ng in the DAMA/LIBRA modulation signal with higher sensitivity
ue to the significantly lower energy threshold for nuclear recoil
ignals and within a very moderate exposure of few 10 kg-days
48] .
cknowledgments
This work was supported by the Italian Ministry of Research un-
er the PRIN 2010ZXAZK9 2010–2011 grant. We grateful acknowl-
dge the generous support of LNGS of this activity.
16 Both crystals are commercial crystals without any specification on low radioac-
ivity.
G. Angloher et al. / Astroparticle Physics 84 (2016) 70–77 77
t
u
M
f
R
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
In particular, we want to thank the LNGS mechanical workshop
eam E. Tatananni, A. Rotilio, A. Corsi, and B. Romualdi for contin-
ous and constructive help in the overall set-up construction and
. Guetti for his constant technical support in the underground
acility.
eferences
[1] M.W. Goodman, E. Witten, Detectability of certain dark-matter candidates,Phys. Rev. D 31 (12) (1985) 3059–3063, doi: 10.1103/PhysRevD.31.3059 .
[2] G. Bertone , et al. , Particle Dark Matter, Cambridge University Press, 2010 . [3] F.E. Simon , Application of low temperature calorimetry to radioactive measure-
ments, Nature 135 (1935) 763 . [4] G. Angloher, et al. CRESST Collaboration, Results from 730 kg-days of the
CRESST-II Dark Matter Search, Eur. Phys. J. C 72 (4) (2012) 1971, doi: 10.1140/epjc/s10052- 012- 1971- 8 .
[5] G. Angloher , et al. CRESST Collaboration, Results on low mass WIMPs using an
upgraded CRESST-II detector, Eur. Phys. J. C 74 (2014) 3184 . [6] G. Angloher, et al. CRESST Collaboration, Results on light dark matter particles
with a low-threshold CRESST-II detector, Eur. Phys. J. C 76 (1) (2016) 25, doi: 10.1140/epjc/s10052- 016- 3877- 3 .
[7] K. Schäffner, et al., Alternative Scintillating Materials for the CRESST DarkMatter Search, J. Low Temp. Phys. 167 (5–6) (2012) 1075–1080, doi: 10.1007/
s10909- 012- 0488- 8 .
[8] R. Bernabei, P. Belli, F. Cappella, V. Caracciolo, S. Castellano, R. Cerulli, et al.DAMA Collaboration, Final model independent result of DAMA/LIBRA–phase1,
Eur. Phys. J. C 73 (12) (2013) 1–11, doi: 10.1140/epjc/s10052- 013- 2648- 7 . [9] J. Amaré, S. Cebrián, C. Cuesta, E. García, C. Ginestra, M. Martínez, et al., Back-
ground analysis and status of the ANAIS dark matter project, AIP Conf. Proc.1672 (2015). http://dx.doi.org/10.1063/1.4928001 .
[10] J. Cherwinka, D. Grant, F. Halzen, K.M. Heeger, L. Hsu, A.J.F. Hubbard, et al. DM-
Ice Collaboration, First data from DM-Ice17, Phys. Rev. D 90 (2014) 092005,doi: 10.1103/PhysRevD.90.092005 .
[11] K. Kim, W. Kang, S. Oh, P. Adhikari, J. So, N. Kim, et al., Tests on NaI(Tl) crystalsfor WIMP search at the Yangyang Underground Laboratory, Astropart. Phys. 62
(2015) 249–257. http://dx.doi.org/10.1016/j.astropartphys.2014.10.004 . [12] S.C. Kim, H. Bhang, J.H. Choi, W.G. Kang, B.H. Kim, H.J. Kim, et al. KIMS Collab-
oration, New limits on interactions between weakly interacting massive parti-
cles and nucleons obtained with CsI(Tl) crystal detectors, Phys. Rev. Lett. 108(2012) 181301, doi: 10.1103/PhysRevLett.108.181301 .
[13] E. Shields, J. Xu, F. Calaprice, SABRE: a new NaI(T1) dark matter direct de-tection experiment, Phys. Procedia 61 (2015) 169–178 . 13th International Con-
ference on Topics in Astroparticle and Underground Physics, TAUP 2013 http://dx.doi.org/10.1016/j.phpro.2014.12.028 .
[14] K. Fushimi, S. Nakayama, R. Orito, R. Sugawara, Y. Awatani, H. Ejiri, et al., PICO-
LON dark matter Search, J. Phys. 469 (1) (2013) 012011 . URL: http://stacks.iop.org/1742-6596/469/i=1/a=012011
[15] C. Woody , et al. , Readout techniques and radiation damage of undoped cesiumiodide, IEEE Trans. Nucl. Sci. 37 (1990) 4 92–4 99 .
[16] A. Boyle , G. Perlow , Debye-Waller factor for the cesium ion in the CesiumHalides by measurements of the Mössbauer effect in Cs133, Phys. Rev. 151
(1966) 211–214 .
[17] P. Nadeau, M. Clark, P.D. Stefano, J.-C. Lanfranchi, S. Roth, M. von Sivers, et al.,Sensitivity of alkali halide scintillating calorimeters with particle identifica-
tion to investigate the DAMA dark matter detection claim, Astropart. Phys. 67(2015) 62–69. http://dx.doi.org/10.1016/j.astropartphys.2015.02.001 .
[18] P. Schotanus , Scintillation characteristics of pure and Tl-doped CsI crystals,Nucl. Sci. IEEE Trans. 37 (2) (1990) 177–182 .
[19] Y. Zdesenko , et al. , Scintillation properties and radioactive contamination ofCaWO 4 crystal scintillators, Nucl. Instr. Meth. Phys. Res. A 538 (2005) 657–667 .
20] A. Senyshyn, H. Kraus, V.B. Mikhailik, V. Yakovyna, Lattice dynamics and
thermal properties of CaWO 4 , Phys. Rev. B 70 (2004) 214306, doi: 10.1103/PhysRevB.70.214306 .
[21] V.B. Mikhailik, H. Kraus, Performance of scintillation materials at cryogenictemperatures, Phys. Status Solidi (b) 247 (7) (2010) 1583–1599, doi: 10.1002/
pssb.20 094550 0 . 22] V.B. Mikhailik, V. Kapustyanyk, V. Tsybulskyi, V. Rudyk, H. Kraus, Luminescence
and scintillation properties of CsI: a potential cryogenic scintillator, Phys. Sta-
tus Solidi (b) 252 (4) (2015) 804–810, doi: 10.1002/pssb.201451464 . 23] M. Ambrosio, R. Antolini, G. Auriemma, R. Baker, A. Baldini, G.C. Barbarino,
et al., Vertical muon intensity measured with MACRO at the Gran Sasso labo-ratory, Phys. Rev. D 52 (1995) 3793–3802, doi: 10.1103/PhysRevD.52.3793 .
[24] S. Pirro, Further developments in mechanical decoupling of large thermal de-tectors, Nucl. Instr. Meth. Phys. Res. A 559 (2) (2006) 672–674. http://dx.doi.
org/10.1016/j.nima.2005.12.197 . 25] G. Angloher, et al. CRESST Collaboration, Commissioning run of the CRESST-
II dark matter search, Astropart. Phys. 31 (4) (2009) 270–276, doi: 10.1016/j.astropartphys.20 09.02.0 07 .
26] G. Angloher, et al. CRESST Collaboration, Limits on WIMP dark matter usingscintillating CaWO 4 cryogenic detectors with active background suppression,
[27] F. Reindl , et al. , Status update on the CRESST dark matter search, Astropart.Part. Space Phys. Detect. Phys. Appl. Proc. 14th ICATPP Conf. 8 (2014) 290–296 .
28] R. Strauss, et al., Energy-dependent light quenching in CaWO 4 crystalsat mk temperatures, Eur. Phys. J. C 74 (7) (2014) 2957, doi: 10.1140/epjc/
s10052- 014- 2957- 5 . 29] V. Tretyak , Semi-empirical calculations of quenching factors for ions in scintil-
lators, Astropart. Phys. 33 (2010) 40–53 .
30] K. Schäffner , Study of Backgrounds in the CRESST Dark Matter Search, Technis-che Universität München, München, 2014 PhD Thesis .
[31] M. Moszy ́nski, A. Syntfeld-Ka ̇zuch, L. Swiderski, M. Grodzicka, J. Iwanowska,P. Sibczy ́nski, et al., Energy resolution of scintillation detectors, Nucl. Instrum.
Methods Phys. Res. Sec. A 805 (2016a) 25–35, doi: 10.1016/j.nima.2015.07.059 . 32] M. Moszy ́nski, A. Syntfeld-Ka ̇zuch, L. Swiderski, P. Sibczy ́nski, M. Grodzicka,
T. Szcz ̨e ́sniak, et al., Energy resolution and slow components in undoped CsI
[33] F. Pröbst , et al. , Model for cryogenic particle detectors with superconductingphase transition thermometers, J. Low. Temp. Phys. 100 (1995) 69 .
34] A.C. Anderson , Nonequilibrium Superconductivity, Phonons, and KapitzaBoundaries, Plenum, New York, NY, 1981 .
[35] K. Schäffner, G. Angloher, F. Bellini, N. Casali, F. Ferroni, D. Hauff, S.S. Nagorny,
L. Pattavina, F. Petricca, S. Pirro, F. Pröbst, F. Reindl, W. Seidel, R. Strauss, Parti-cle discrimination in TeO 2 bolometers using light detectors read out by transi-
36] M. Weber , Handbook of Optical Materials, CRC Press, 2002 . [37] M.B. Chadwick, M. Herman, P. Obložinský, M.E. Dunn, Y. Danon, A.C. Kahler,
et al., ENDF/B-VII.1 nuclear data for science and technology: cross sections, co-
variances, fission product yields and decay data, Nucl. Data Sheets 112 (12)(2011) 2887–2996, doi: 10.1016/j.nds.2011.11.002 .
38] I. Dafinei, The LUCIFER project and production issues for crystals needed inrare events physics experiments, J. Crystal Growth 393 (2014) 13–17. http://dx.
doi.org/10.1016/j.jcrysgro.2013.10.017 . 39] A. Alessandrello, C. Brofferio, D. Camin, P. Caspani, P. Colling, O. Cremonesi,
et al., The thermal detection efficiency for recoils induced by low energy nu-
clear reactions, neutrinos or weakly interacting massive particle, Phys. Lett. B408 (14) (1997) 465–468. http://dx.doi.org/10.1016/S0370- 2693(97)00765- X .
40] C. Arnaboldi, S. Capelli, O. Cremonesi, L. Gironi, M. Pavan, G. Pessina, et al.,Characterization of ZnSe scintillating bolometers for Double Beta Decay, As-
45] R. Strauss, G. Angloher, A. Bento, C. Bucci, L. Canonica, A. Erb, et al., A de-tector module with highly efficient surface-alpha event rejection operated
in CRESST-II Phase 2, Eur. Phys. J. C 75 (8) (2015) 352, doi: 10.1140/epjc/s10052-015-3572-9 .
46] T.O. Niinikoski , A. Rijllart , A. Alessandrello , E. Fiorini , A. Giuliani , Heat capacityof a Silicon calorimeter at low temperatures measured by alpha particles, EPL
(Europhys. Lett.) 1 (10) (1986) 499 .
[47] S. Myagkota , A. Pushak , G. Stryganyuk , S. Novosad , I. Pashuk , Spectral and ki-netic characteristics of luminescence in CsI-Ca crystals, Funct. Mater. 15 (2)
(2008) 187–191 . 48] G. Angloher, P. Carniti, L. Cassina, L. Gironi, C. Gotti, A. Gütlein, et al., The COS-
INUS project: perspectives of a NaI scintillating calorimeter for dark mattersearch, Eur. Phys. J. C 76 (8) (2016) 441, doi: 10.1140/epjc/s10052- 016- 4278- 3 .
