Scintillation in Low-Temperature Particle Detectors

Inorganic crystal scintillators play a crucial role in particle detection for various applications in fundamental physics and applied science. The use of such materials as scintillating bolometers, which operate at temperatures as low as 10 mK and detect both heat (phonon) and scintillation signals, significantly extends detectors performance compared to the conventional scintillation counters. In particular, such low-temperature devices offer a high energy resolution in a wide energy interval thanks to a phonon signal detection, while a simultaneous registration of scintillation emitted provides an efficient particle identification tool. This feature is of great importance for a background identification and rejection. Combined with a large variety of elements of interest, which can be embedded in crystal scintillators, scintillating bolometers represent powerful particle detectors for rare-event searches (e.g., rare alpha and beta decays, double-beta decay, dark matter particles, neutrino detection). Here, we review the features and results of low-temperature scintillation detection achieved over a 30-year history of developments of scintillating bolometers and their use in rare-event search experiments.


Bolometers as Viable Detectors of Rare-Event Processes
A unique class of particle detectors is represented by low-temperature calorimeters, often named bolometers, which operate typically at ∼10-100 mK and are instrumented by a temperature sensor to detect a particle interaction with a detector medium as a small temperature rise . An operation temperature near absolute zero is exploited to significantly reduce the detector heat capacity, allowing the thermal detection (e.g., ∼0.1 mK temperature rise per 1 MeV deposited energy). Consequently, common detector materials are crystalline insulators and semiconductors, for which the lattice-dominated heat capacity-being proportional to the cube of the ratio of the absolute temperature to a Debye temperature of the material (according to the Debye model) [28]-is very small at low temperatures. The particle energy release in a low-temperature detector (LTD) is converted to phonons (quantized modes of lattice vibrations), which are low-energy elementary excitations and thus their number fluctuations play a minor role on the detector energy resolution. Despite being largely affected by other sources of noise (read-out noise, thermal fluctuations of the detector, vibrational noise), bolometers offer a low detection threshold (in the range of tens eV to a few keV) and a high energy resolution (a few keV in a wide energy interval, from hundreds keV to a few MeV). Thanks to the absence of a dead layer, such instruments can be used for the detection of different types of radiation (α, β, γ, X-ray, neutrons, ...). However, a relatively slow response of thermal detectors (a typical signal duration of macrobolometers is within the ms-s range) imposes a strong constrain on the counting rate of such devices. Thus, taking into account excellent performances of bolometers and a variety of materials suitable for low-temperature applications, such detectors represent a great interest for low-counting experiments to search for rare-event processes (see Table 1). Table 1. Possible applications of particle detectors containing elements (isotopes) of interest for rare-event searches.

Application
Elements of Interest Most Promising Isotopes Ref.
A "price" to pay for bolometers-high performance detectors with a large choice of detector materials-is the complication in their operation, in particular the use of cryogenic and vacuum technologies. The required temperature conditions can not be fulfilled with cryogenic liquids as, for example, for semiconductor detectors (the coldest temperature of common cryogenic liquids is 4.2 K for liquid helium). In order to reach millikelvin temperatures, sophisticated and costly cryostats (dilution refrigerators), exploiting the thermodynamic properties of the 3 He/ 4 He mixture, are used [62][63][64][65][66][67]. Despite of the complexity, such instruments are commercially available and provide a stable ∼10 mK base temperature and a long duty cycle on a routine basis. There are two distinctive types of dilution refrigerators: the so-called "wet" and "dry" cryostats, that is respectively with and without a liquid helium bath to be refilled periodically with the costly cryoliquid (supplied by a company or recovered using a liquifier). The features of both types of cryostats can be mixed too [68]. Cryogen free dilution refrigerators become more attractive thanks to the use of commercial pulse tubes to pre-cool thermal stages down to 4 K. Modern pulse-tube cryostats also allow a fully automated cool-down process and require a minimal maintenance during the operation (a periodic refill and cleaning of a liquid nitrogen trap) [69]. Depending on the cryostat, the cooling down can last from a day (in the best case) to 1-5 weeks (for large facilities with massive internal shields), inducing a delay in the operation of bolometric detectors. It is important to stress that dilution refrigerators, especially pulse-tube-based ones, are sources of vibrations which can drastically affect the bolometric performance, thus, special interventions are needed to mitigate the noise issue (e.g., see [70][71][72][73][74][75][76][77]). The experimental volume in commercial cryostats, capable of hosting an array of bolometric detectors, is limited to ∼10-100 L. Possibilities to scale up to a 1 m 3 bolometer array and to collect one tonne-year data have been recently demonstrated in the CUORE experiment [77][78][79][80]. A cryostat of a similar size will be used in the AMoRE experiment [81]. A four time larger cryostat than that of CUORE is under discussion for the long-term prospects of the CUPID program [82]. Multiplication of cryogenic facilities can also be a (costly) solution to increase the number of individual detector modules. However, a large experimental space is mandatory only for some applications and this review illustrates that many physics results on rare-event searches can be achieved with either small-scale arrays or even with single bolometric modules.

Particle Identification with Low-Temperature Detectors
Rare-event search experiments face a common challenge in the suppression of different sources of background which can hide or mimic signals searched for. The background sources can be originated to environmental radioactivity, radioactive contaminants of detectors and experimental set-ups, cosmic rays, neutrons, and even neutrinos. (An interested reader can find details in a classical review on low-radioactivity techniques [83], as well as in the recent reviews and original papers on the related topic [24,51,[84][85][86][87][88][89][90][91][92][93][94][95][96][97][98][99].) Particle identification capability of a detector technology is thus beneficial for rare-event searches as a viable tool for background rejection. Bolometers based on some crystal scintillators show a particle-dependent response exploitable for pulse-shape discrimination (see details in the first original works [100][101][102] and some updates in the recent review [24]). However, the reproducibility of this method strongly depends on the experimental conditions (e.g., see [24,26] and references therein). It is important to stress that a lack or a poor performance of particle identification limits the application or sensitivity of a bolometric detector in rare-event searches.
The detection of scintillation light allows particle identification thanks to the dependence of the scintillation light yield on the energy loss mechanism, as described by a semi-empirical relation proposed by Birks [126][127][128]: where dL/dr is the specific scintillation yield per path length, S is the absolute scintillation efficiency, dE/dr is the particle stopping power in the material, B · dE/dr is the density of excitation centers along the track, and k is a quenching parameter. The product k · B is usually considered a single parameter expressed in g/cm 2 /MeV (the so-called Birks' factor) [127], which depends on the detector material, as well as conditions of measurements and data treatment can have a strong impact [128]. At small dE/dr (fast electrons) and at large dE/dr (e.g., α particles) Equation (1) approximates to dL/dr = S · dE/dr and dL/dr = S/(k · B), respectively. The particle-dependent difference in the scintillation light output is often presented as the ratio of a light yield of ions to that of electrons (quenching factor for ions, QF i ), which has the following approximate relation with the Birks' factor and the stopping power: The QF i value is usually smaller than one [128] and it can be explained by a saturation effect due to the high ionization density that characterizes the interaction of heavy particles in matter. In case of α particles, this parameter is often called the α/β ratio. Figure 1 illustrates the basic principle of particle identification with scintillating bolometers: particles with energy releases that are different from those of electrons, γ quanta, and muons populate band(s) with lower scintillation. The efficiency of the separation between different particle types increases with energy of the incident radiation (i.e., the amount of detected light). Figure 1. Schematic presentation of particle-dependent distributions of light signals measured by a photodetector in coincidences with the energy releases detected by a crystal scintillator-based bolometer. Circles represent individual events, which can form peaks shown as dark ovals on the β(γ, µ) and α bands. The light output induced by highly ionizing particles (alpha particles, nuclear recoils) impinging the scintillator is suppressed (quenched) with respect to the one induced by electrons (as well as γ quanta, cosmic-ray muons) of the same energy. Therefore, particles with different ionization properties become separated on such light vs. heat scatter-plot. The separation is more profound at higher energies, with increased amount of the detected scintillation light.

Importance of Scintillation Detection for Bolometric Rare-Event Searches
The necessity of scintillation detection in scintillating bolometers is briefly discussed in this section in view of possible applications of such devices in rare-event searches, as listed in Table 1 where charge-dependent coefficients a(Z) and b(Z) are determined by fitting experimental data for each isotopic series. Indeed, the Geiger-Nuttall plots in form of Equation (3) (e.g., see Figure 1 in [130]) show that a change in Q α of only 10% changes T 1/2 by about 10 3 . Consequently, the naturally occurring α-active nuclides (15 isotopes) have a large spread of half-lives (∼10 9 -10 21 yr) for a comparatively small range of Q α -values (∼2-5 MeV) [29,132].
There are also about 70 naturally occurring isotopes which are potentially α-unstable, but the α radioactivity of only 10% from the full list of candidates can possibly be detected (Q α ∼1.6-2.7 MeV and predicted T 1/2 ∼10 17 -10 22 yr) [29]. The list of isotopes of interest for investigations of rare α decays is given in Table 1.
The region of interest (ROI) in rare α decay searches is dominated by natural γ and β radioactivity, as well as by cosmic rays. The identification of α particles combined with the typically high energy resolution of bolometric detectors allow the realization of a nearly "background-free" experiment even at a surface laboratory, as for example, the first detection of the 209 Bi α decay [133], the rarest α decay observed (together with the 209 Bi α transition to the first excited level of 205 Tl [134]). Often, such experiments are by-products of scintillating bolometer development for dark matter and double-beta decay search programs.

Rare Beta Decay
Taking into account the total angular momentum change, ∆J, and the parity change, ∆π, between the initial and final states, β decays are classified as allowed (∆J = 0, 1 and ∆π = no) and nth-forbidden (∆J = n, n + 1 and ∆π = no, yes; β transitions with ∆J = 0 and with parity change also belong to first-forbidden branches) [135,136]. The partial half-lives of β emitters show a strong dependence on the selection rules of ∆J and ∆π, exhibited as significantly longer T 1/2 of β processes with higher forbiddenness [136]. The rarest β transitions ever observed, 4-fold-forbidden β decays of 50 V, 113 Cd, and 115 In (possible for only these three isotopes), are detected with the half-lives of 10 14 -10 16 yr [29].
Studies of highly-forbidden β decays are of particular interest to scrutinize the value of the axial-vector coupling constant, g A [30,31,[137][138][139]. A possible g A quenching with respect to the free nucleon value ( 1.27, see [140] and references therein) would affect the β spectral shape, notably at low energies. The g A value in this process is expected to be similar to the one involved in the neutrinoless double-beta decay (see Section 1.3.3) [31]. Therefore, a low-threshold bolometer containing isotope of interest for rare β decays (see Table 1) is a viable detector for the study of a β spectrum shape. In spite of the long half-lives of these processes, the induced counting rate and, subsequently, the probability of pile-ups in macrobolometers containing any of these nuclides is rather high (because of the high detection efficiency of the bolometric technique), strongly affecting the precision of the spectrum reconstruction. For that reason, a light-signal-based trigger comparable with a low-energy threshold of the main absorber could be useful to control pile-ups thanks to the much faster response of a photodetector.

Double-Beta Decay
Two-neutrino and neutrinoless double-beta decays (2νDBD and 0νDBD, respectively)spontaneous nuclear disintegrations with the emission of two electrons accompanied or not by two antineutrinos-are energetically allowed for only 35 isotopes [34]. Here, we consider only DBDs which increase the charge of nuclei by two units. The nuclear charge decreasing DBD processes-namely double-electron capture, electron capture with positron emission and double-positron decay [141][142][143][144][145][146][147]-are omitted because of suppressed decay probabilities and experimental issues (see, e.g., ref. [148] and references therein). It is worth underlining that the observation of a 0ν-mode of any of such charge decreasing DBD processes would be extremely important in view of the 0νDBD mechanism determination [141].
The 2νDBD is the rarest process ever observed and it has been detected for a dozen of isotopes with a typical half-life in the range of 10 18 -10 24 yr [37,145]. In contrast to 2νDBD, the 0νDBD violates the total lepton number and requires neutrinos to be massive Majorana particles (i.e., particle equal to its antiparticle) [35,36,142,143,[149][150][151]. Therefore, the 0νDBD is forbidden in the Standard Model (SM) of particle physics, but this process is allowed in many SM extensions [142,143,152]. Till today, the 0νDBD has not been detected and the most stringent half-life limits are in the range of 10 24 -10 26 yr [36,79,80,[153][154][155][156][157][158][159].
The interest in the 0νDBD detection is much stronger, because it would unambiguously imply the observation of new physics beyond the SM, the highly successful paradigm nowadays.
The energy spectrum of two electrons in the 2νDBD is a continuum up to the transition energy, Q ββ , peaked at ∼1/3 of Q ββ , while the 0νDBD signature is a peak centered at Q ββ and smeared by the detector energy resolution (e.g., see [33]). The DBD rate strongly depends on the transition energy (with a leading term proportional to Q 11 ββ and Q 5 ββ for the 2ν and 0ν modes, respectively), thus an isotope of interest with a high Q ββ is preferred. Only 11 (6) out of 35 potentially DBD-active isotopes have Q ββ values above 2.0 (2.7) MeV [34]; they are listed in Table 1. In case of a Q ββ higher than 2.7 MeV, the contribution of natural γ(β) radioactivity to the ROI becomes strongly suppressed and the major background contribution comes from surface originated α radioactivity [203]. Therefore, a radiopure detector with embedded high-Q ββ -isotope, high energy resolution, and particle identification capability (e.g., a scintillating bolometer) is well suited for DBD searches. A high energy resolution is of a particular importance to make the 0νDBD ROI narrow (thus reducing the background contribution), as well as to determine background components for the construction of a correct background model. It is worth noting that the energy resolution of a bolometer based on an efficient scintillator is strongly affected by the energy partition between heat, scintillation and energy trapping inside a crystal. Therefore, a correlation between heat and light signals has to be used to improve the detector energy resolution (e.g., see [204]).
Being intrinsically slow, bolometers can suffer from any counting rate higher than a few tens of Hz which can give rise to pile-up issues. Moreover, the 2νDBD event pile-ups can be an unavoidable source of background in the 0νDBD ROI [205,206]. The 2νDBD induced pile-ups are worrisome for 100 Mo-and 150 Nd-enriched LTDs due to considerably high 2νDBD rates of these isotopes (e.g., ∼20 mHz in 1 kg of 100 Mo), which are at least an order of magnitude faster than that of other 2νDBD-active nuclides [37]. This issue can largely be mitigated by a pulse-shape analysis of heat signals [205,206], while a further improvement in the pile-up-induced background reduction would be possible by the analysis of the faster scintillation light signals with enhanced signal-to-noise ratio [207].
Scintillating bolometers for 0νDBD searches have been extensively developed over the last decade, particularly in the framework of the following projects/experiments: BoLux R&D (research and development) experiment [208], AMoRE [209], ISOTTA [210], LUCIFER [211] and its follow-up CUPID-0 [212], LUMINEU [213] and its follow-up CUPID-Mo [214], CLYMENE [215], CROSS [216], and CANDLES [217]. Almost 30-year-long development on pure bolometric detectors based on tellurium dioxide, performed within the CUORE project and the predecessors (see [218] and references therein), was also crucial for a fast progress in R&D on scintillating bolometers. Moreover, the CUORE Collaboration has demonstrated the feasibility to realize a tonne-scale bolometric experiment, which is currently ongoing in the Gran Sasso underground laboratory (LNGS, Italy) [77,79,80]. Thus, all these activities play an important role towards the planned realization of CUPID [82,219] and AMoRE [81,209] next-generation large-scale bolometric experiments.
It is worth mentioning that low-temperature scintillation detection can be used not only for DBD detectors, but also for an active veto inside a cryogenic set-up. Such active shield is going to be developed within the BINGO [220] project using crystal scintillators and bolometric photodetectors. In case of the proof of concept, this technology is aiming at the implementation in the CUPID follow-up.

Dark Matter (WIMP)
There are several evidences for the existence in the Universe of a large amount of invisible matter (about five times more than the ordinary matter), the so-called dark matter (DM), which could be constituted by unknown particles, which noticeably interact only gravitationally [40,221,222]. Direct detection of yet unknown DM particles [40,41,223,224] is therefore another hot topic in the searches for beyond-SM physics.
Weakly interactive massive particles (WIMP) are considered one of the most viable candidates for the DM composition [41,224]. The main detection principle of WIMP is based on a registration of a nuclear recoil (a few tens keV or less) induced by the WIMP scattering off a nuclei in a target material [40,41,224]. The WIMP-nucleus interactions are considered to be either spin-independent (with the whole nucleus) or spin-dependent (with nuclei exhibiting non-zero spin). Therefore, the experimental sensitivity to detect a WIMP signal depends not only on detector performance, exposure, and background level, but also on the chemical composition of the detector material (Table 1).
Scintillating bolometers are well placed among DM detection techniques thanks to a low threshold, particle identification capability, and a wide choice of detector (i.e., target) materials containing a variety of elements of interest (see Table 1). Indeed, the detection threshold plays a crucial role taking into account the exponential growth of the WIMPinduced event rate with the decrease of the nuclear recoil energy. The separation between electron and nuclear recoils provides the possibility of searching for a DM signal in an (almost) background-free region [225]. A scintillation-based particle identification at low energies is rather challenging due to the low amount of the expected light and thus it requires the use of efficient scintillators. The improvement in the particle identification at low energies allows us to achieve a lower detection threshold as well [225]. Moreover, taking into account that a spin-independent WIMP-nucleon elastic-scattering cross section is proportional to the square of the mass number of the nucleus, the difference in the light yield for ionizing particles is also crucial for distinguishing nuclear recoils originated from WIMP's scattering off light, mid-weight or heavy nuclei constituting a scintillation target [226]. If a scintillating bolometer has an anisotropic light output (which depends on the particle direction relative the main crystal axes), this feature can be exploited for the search for a diurnal modulation of a DM signal [227][228][229][230][231].
Scintillating bolometers for DM searches have been developed for almost 30 years by ROSEBUD [232] and CRESST [233], together with alternative developments of Ge and/or Si heat-ionization bolometers by CDMS [234] and EDELWEISS [235]. Extensive R&D activities (including those of ROSEBUD and CRESST) were performed to develop different scintillation materials for a potential use in a multi-target tonne-scale array of LTDs of the EURECA DM search project [236][237][238][239]. Due to a rapid progress of detector technologies based on liquid noble gases, achieved over the last decade in searches for high-mass WIMP [224], the realization of the EURECA project has become less relevant now. Therefore, the CRESST, CDMS, and EDELWEISS Collaborations are currently focused on the direct detection of low-mass DM particles [108,109,240]. In addition, an R&D on scintillating bolometers for the detection of the annual modulation of the DM signal is currently ongoing within the COSINUS project [241].

Solar Axions
The axion is a hypothetical particle postulated by the Peccei-Quinn theory to solve the "strong CP problem" in quantum chromodynamics [242,243]. The axions are predicted to be neutral, low-mass particles characterized by a low interaction cross section with ordinary matter and thus are considered to be possible DM candidates [222,223,244]. If axions exist, they can be largely produced in the solar core and subsequently detected in a laboratory experiment on Earth exploiting several mechanisms of axion interaction (see, e.g., [244]). The experimental signatures of axions are searched for typically at low energies (few-tens keV), with the exception of a resonant absorption of solar axions which are emitted with the energy of 478 keV (corresponding to the excited level of 7 Li) [43]. Moreover, the ROI for axion detection is dominated by irreducible γ(β) background. Therefore, the detection of scintillation light in a bolometric detector of axions would not provide an essential background rejection.

Solar and Supernova Neutrinos
Some of the most attractive isotopes for 0νDBD search (see Table 1) can also be interesting targets for the detection of solar and supernova neutrinos because of the high neutrino capture cross section expected and detector large mass foreseen in such experiments [3,[47][48][49][50]. Thus, a primary DBD search experiment with a large-scale array of scintillating bolometers can also be used for the detection in real time of solar and supernova neutrinos. Since the neutrino capture in a DBD detector would result in a continuum energy spectrum peaked at MeV energy (see, e.g., [50,88]), the scintillation-based particle identification can be exploited for the background suppression in the same way as DBD searches with scintillating bolometers.
The coherent elastic neutrino-nucleus scattering [245] (see Section 1.3.7), is another viable channel of neutrino detection [246]. This neutral-current process is characterized by several orders of magnitude higher cross section than that of neutrino-electron scattering. However, a detection signature is a low-energy nuclear recoil and, thus, low-threshold detectors are required. LTDs can therefore be suitable for such application [52,108,109]. At the same time, no scintillation detection can really be exploited as a particle identification tool due to a tiny light signal expected for low-energy release in even efficient scintillators.

Coherent Elastic Neutrino-Nucleus Scattering
The coherent elastic neutrino-nucleus scattering (CENNS)-a process predicted about 45 years ago [245] but detected only recently [247]-is a promising tool to search for effects beyond the SM [53,54]. The CENNS induced by solar neutrinos will be the irreducible background of near future direct DM search experiments because the CENNS signature is exactly the same as WIMP scattering off nuclei [51,52,108,109,224,248]. In turn, it is an opportunity to study this SM process. There is growing interest in precise CENNS investigation using artificial neutrino sources. An experimental approach based on a LTD placed near a nuclear reactor is now considered to be very attractive for such "tablescale" neutrino experiments (e.g., see [54,55] and references therein). Since the expected signal is a very low energy nuclear recoil, a scintillating bolometer approach would not be able to provide particle identification in the ROI (similarly to low-mass WIMP searches). However, this technique can be promising if a low-threshold detector contains a heavy target (as CENNS cross section is proportional to the square number of neutrons) and allows detection of unambiguously neutrons (see Section 1.3.8). R&D on such scintillating bolometers is now ongoing within the BASKET project [249].
It is worth mentioning a recently proposed method for nuclear recoils calibration at the ∼100 eV energy scale using a neutron capture reaction to induce recoils by de-excitation γ quanta (the CRAB project) [250]. The detection of escaped γ quanta in coincidence with nuclear recoils in the CENNS detector can significantly clean up the energy spectrum of recoils suppressing multi-γ-induced background contribution. The feasibility of a γ detector based on a high-Z scintillator with a bolometric light detection, in a similar way as an active shield for bolometric DBD search experiments (BINGO project [220]), is going to be investigated too.

Neutron Detection in Rare-Event Searches
A scintillating bolometer containing 6 Li or 10 B-isotopes actively used for neutron detection thanks to their high thermal neutron capture cross section (see, e.g., [57,61,[251][252][253])-can be exploited for the in-situ neutron flux monitoring in rare-event search experiments [59,60,74,254,255]. The detection principle is based on the 6 Li(n,t)α and 10 B(n,α) 7 Li reactions aiming at detection of the products. The neutron capture on 6 Li creates alpha (2052 keV) and triton (2731 keV) particles with a shared energy of 4783 keV. The neutron capture on 10 B can lead to the ground state of 7 Li (the branching ratio is 6%; the Q-value is 2792 keV), but the transition to the 478 keV excited state of 7 Li is dominant (the branching ratio is 94%). The isotopic abundance of 6 Li and 10 B is 7.5% and 19.8%, respectively, thus detectors with natural isotopic composition of lithium and/or boron can be used for neutron detection too. However, a detector enriched in 6 Li and/or 10 B would largely increase the detection efficiency. The cross sections of thermal neutron capture on 155,157 Gd (15% and 16% in natural Gd, respectively) are much higher than those of 6 Li and 10 B, however the products are γ quanta (see, e.g., [256]). Therefore, LTDs with 6 Li and/or 10 B content are preferred, while Gd-containing bolometers can have some specific application (e.g., if the suppression of the contribution induced by thermal neutrons is needed) [59].
Another viable way to detect neutrons with scintillating bolometers is related to the identification of nuclear recoils induced by the elastic neutron scattering off a nucleus in the detector material. Scintillation induced by nuclear recoils is even more quenched than that of α particles and it allows us to discriminate the dominant electron recoil background, as schematically shown in Figure 1. The presence of low atomic mass elements in the detector medium is preferred to get nuclear recoils with higher energies, improving particle identification. A simultaneous use of scintillating bolometers to detect both neutroninduced nuclear recoils and products of neutron captures can also be exploited for the neutron flux measurements [60,257,258].
Li-and/or B-containing scintillating bolometers for neutron detection were actively developed in 1993-2014, particularly for the ROSEBUD DM project. Over the last six years, a significant progress has been achieved in the development of other Li-containing bolometers for DBD (ISOTTA, LUMINEU, CLYMENE, CUPID-Mo, CROSS, CUPID, AMoRE), DM (CRESST) and CENNS (BASKET) oriented projects. The developments of scintillating bolometers for the efficient detection of fast neutron-induced nuclear recoils were and are mainly a subject of DM search programs, but such bolometric detectors developed for other applications can be used for this purpose too.

Key Ingredients and Performance of Scintillating Bolometers
Examples of construction elements of a scintillating bolometer and the assembled detector module are shown in Figure 2. The main ingredients of such a device and demands on its performance related to the scintillation detection are briefly discussed in this section.

Cryogenic Scintillator
At first glance, the term "scintillating bolometer" alludes to a detector material which possesses scintillation properties at low temperatures, but even non-scintillating (transparent) dielectric crystals, acting as Cherenkov radiators, can be suitable too. There is definitely an advantage in the use of scintillation materials in such devices. Moreover, different crystal scintillators at low temperatures have an increased light output compared to room temperature, as observed in tungstates, molybdates, and some other oxide compounds, as well as in selenides and alkali halides [260][261][262][263][264][265]. Such a feature opens the possibility to use also those materials that scintillate little at room temperature (e.g., lithium molybdate [266]). Some crystal scintillators exhibit the suppression of the scintillation at low temperatures (e.g., Tl-doped sodium iodide [263]), but the amount of the emitted light remains well detectable. It is also worth noting the temperature dependent scintillation kinetics of materials, which can result to a rather long fluorescence decay time at low temperatures (e.g., hundred(s) µs [261]), to be taken into account for a proper integration of photons by a light-sensitive device.
Some materials can remain be poorly-or non-scintillating even at low temperatures (e.g., tellurium dioxide [267][268][269]). In such occasion, the detection of a γ(β)-induced Cherenkov radiation can open a window to particle identification, because no emission is expected for α's due to a significantly higher energy threshold of the Cherenkov light production (e.g., 400 MeV for α's compared to 0.05 MeV for β's in tellurium dioxide) [270]. However, this approach is challenging due to a low radiation expected [270], requiring photodetectors with a low threshold (below 60 eV). For instance, the 2615 keV γ interaction with the full energy release in a tellurium dioxide crystal produces ∼300 Cherenkov photons [271].
In addition to the impact of temperature conditions, the light output depends on many other parameters such as the scintillation efficiency and optical properties of the detector material (chemical composition), the crystal chemical purity (the purity of starting materials) and quality (the crystal growth process), the thermal treatment (if needed), the sample shape and the roughness of the crystal surface (see, e.g., [272][273][274]). The choice of the detector material is crucial, but it depends on the physics goals and availability of crystal producers, starting materials (and their purification if required), and developed crystal growth process. The optimization of a proper crystal shape and/or surface treatment may drastically improve the light output [271,272,[274][275][276][277], if there are no restrictions (e.g., related to the crystal size and/or radiopurity). In its turn, the measured light signal additionally depends on the light collection efficiency and the photodetector sensitivity (as briefly discussed in Sections 2.2 and 2.4). Different scintillation materials used in scintillating bolometers and the results of scintillation detection are detailed in Section 3.

Reflector
By default, a reflector is present around a crystal to enhance the light collection efficiency; it is particularly crucial for scintillators with low light output. Widely used materials with a high reflection efficiency are reflective films from 3M (VM2000, VM2002, Vikiuti TM ) and TORAY (Lumirror ® ), an aluminum foil, a PTFE (Teflon ® ) tape, and an Agcoated detector housing [74,272,274,276,[278][279][280] (some other reflectors can be found, for example, in [279]). Another reason for the use of a reflector is related to possible scintillation properties of this material, which can be exploited for the identification of the following surface-induced backgrounds: (a) nuclear recoils in direct DM searches [281]; (b) α events degraded in energy in 0νDBD searches with poorly-or non-scintillating bolometers [282]. At the same time, the scintillating properties of a reflective film can spoil the particle identification of 0νDBD bolometric detectors with (reasonably) good scintillation [74]. The removal of the reflective film can also be driven by radiopurity considerations and/or improvement of coincidences in detector array [74,283].

Temperature Sensor
A particle interaction in a cryogenic scintillator is detected via a phonon signal collected by a special sensor. The sensor technology exploits the following working principles [17,284]: • temperature-dependent resistivity of highly doped semiconductors (neutron-transmutationdoped, NTD); • superconducting transition (transition-edge sensor, TES); • temperature-dependent magnetization of paramagnetic materials (metallic magnetic calorimeter, MMC); • kinetic inductance in superconducting materials (kinetic inductance detector, KID).
Up to now, sensor technologies based on NTD Ge (heavily doped germanium thermistor) and TES W (thin tungsten superconducting film) are the dominant choices for a phonon readout in scintillating bolometers. However, most of the scintillation materials were tested using NTD Ge thermistors and only some crystal compounds were studied with TES W phonon sensors (mainly within the CRESST and COSINUS programs). The use of MMC sensors in scintillating bolometers (mainly in the AMoRE project), began a decade ago [285] and is slowly increasing. A first operation of a cryogenic scintillator with a KID-based phonon sensor has been realized only recently [286].

Photodetector
In a scintillating bolometer, a photodetector working at millikelvin temperatures needs to be coupled to a cryogenic scintillator. The first proof of concept of scintillating bolometers was demonstrated with Eu-doped calcium fluoride scintillators and PIN silicon photodiodes [287,288] (further developments are reported in [289]). The use of an auxiliary bolometer as a photodetector in a scintillating bolometer was realized for the first time in [290,291], where a composite detector was constructed with the scintillator itself and a Bi-coated sapphire disk. A conception of a scintillation read-out using a lowtemperature photomultiplier tube, coupled to a scintillator-based (calcium tungstate or calcium molybdate) bolometer, has been recently demonstrated too [292].
The use of thin (∼0.05-1 mm) bolometric light detectors (LDs) in scintillating bolometers is dominant thanks to high radiopurity, high sensitivity to a wide range of photons emitted and more compact and simplified detector module structure. Moreover, being slow response photodetectors (a signal rise time is in the µs-ms range), LDs are also well suitable for slow scintillators. Thus, we focus here only on bolometric photodetectors. As for the energy absorber of such devices, the most frequently used materials are originally dark semiconductors (Ge or Si), but also transparent dielectric crystals (sapphire) coated typically with a Si thin layer.
In addition to the use of a reflective film, light collection in a scintillating bolometer can be enhanced by the optimization of the LD design, in particular following one or several actions:

•
The detection area of a photodetector and the crystal-side surface facing it are made comparable [24,74,272,274]. In an extreme case, an LD can cover a significant part of the cryogenic scintillator surface. For example, a beaker-shaped LD allows us to drastically improve (by a factor of 3) the detected light signal [293][294][295]. • A special LD coating is required to reduce the light reflection. The widely used coating materials are silicon dioxide and oxide, SiO 2 and SiO (e.g., see [75,272,[296][297][298][299]). For instance, the detection of ∼600 nm light signal by a Ge LD coated with a 70 nm SiO (SiO 2 ) layer is improved by approximately 30% (20%) [298]. Several other materials together with SiO 2 have been recently investigated aiming at the optimization of the antireflective coating [300]. • The distance between scintillator and photodetector is minimized, typically to a few millimeters. A method for putting an LD in direct contact with a crystal has been proposed recently [301].
The energy scale of scintillating bolometers is mostly determined with sources of γ quanta, but it does not allow calibration of LDs due to their small sizes. In principle, the calibration of LDs is not mandatory for particle identification. However, the knowledge of the LD energy scale provides a valuable information about the device performance, as well as the measurement of the detected scintillation light energy. The LD calibration can be realized in several ways [302,303], for example, using: • an X-ray source facing an LD (the most popular method; for example, 55 Fe with 5.9 and 6.5 keV doublet); • an external high-activity γ source to induce X-ray fluorescence near an LD (e.g., it can be useful for the calibration of LDs in low-background experiments, where the presence of an X-ray source near an LD is prohibited); • the energy distribution of cosmic-ray muons passing through an LD (not valid for deep underground measurements); • photon statistics (e.g., LED injected photons).
A low-threshold detection of scintillation light is of special importance for scintillating bolometers taking into account that only a small part of the particle energy release is converted into scintillation photons and detected by an LD (typically, ∼(0.1-1)% of the measured heat energy is detected in form of scintillation with the light collection efficiency of ∼(10-30)%). Certainly, the fluctuation of the LD noise determines the device threshold. Even if the threshold can be lowered by exploiting coincidences between phonon and scintillation signals (e.g., [304]), a primary goal of an LD technology is to achieve as low noise as possible. At the same time, the demand on the LD threshold is mainly determined by the scintillation efficiency of the chosen cryogenic scintillator and its application to rare-event searches (briefly discussed in Section 2.5). The state-of-the-art of bolometric LD technologies is summarized in Table 2. Table 2. Technologies of bolometric photodetectors and representative baseline noise resolution (RMS). The Neganov-Trofimov-Luke (NTL) gain of a signal-to-noise ratio is given only for technologies exploiting the signal amplification based on the NTL effect [305,306]. The transition-edge sensor (TES) technology-based QETs stands for quasiparticletrap-assisted electrothermal feedback transition-edge sensors (each QET consists of a W TES and an Al fin.). NTD, KID, and MMC stand for neutron-transmutation-doped semiconductors, kinetic inductance detector, and metallic magnetic calorimeter, respectively.

Demands on Particle Identification Efficiency
A commonly used particle identification parameter of scintillating bolometers is the ratio between a scintillation light signal measured by an LD (in keV) to an energy release in the cryogenic scintillator detected as a heat (in MeV), the so-called light-to-heat ratio, L/H. (This parameter is often called as "light yield", but such term can be confused with the absolute scintillation yield.) An illustration of the L/H parameter versus the particle energy extracted from the scintillating bolometer data [74] is shown in Figure 3 (left panel). As one can see, the γ(β) events are clearly separated from α's, exhibiting a factor of 5 difference in the scintillation light signal associated to these particles. Figure 3 (left panel) illustrates the application of scintillating bolometers to 0νDBD searches, showing a ROI at 3 MeV in the γ(β) band, which is free from α particles degraded in energy due to the decays at the surface of the detector materials (mimic using an α source). Some events present in Figure 3 (left panel), mostly at low heat energies, exhibit the negative values of light signals and consequently the negative L/H values. This situation is common for low efficient scintillators and is caused by the search for low-energy scintillation signals in the LD noise. In order to increase the signal-to-noise ratio, bolometric data are processed with a digital filter (e.g., the optimum filter [342,343]). Then, the light signal amplitudes (i.e., energies) are estimated at a certain time shift with respect to the heat events trigger positions (e.g., as proposed in [304]), taking into account the difference in the time response of the channels. It may happen that the filtered waveform contains a light signal completely hidden in the LD noise fluctuation, and the optimum filter can thus return a negative signal amplitude. If light signals are searched in a comparatively large time interval (i.e., tens of the sampled channels), positive values of the noise fluctuations are then represent amplitudes of low-energy scintillation signals (e.g., see Figure 4 in [74]).
The separation between α and γ(β) events is often expressed by a discrimination power parameter [101,344]: where µ (σ) denotes the mean value (width) of the corresponding L/H distributions of γ(β) and α events with energy E. The meaning of the DP α/γ(β) parameter is illustrated in Figure 3 (right panel), where the L/H distributions are shown for events detected around the 100 Mo 0νDBD ROI (3 MeV). As it is seen in Figure 3 (right panel), the DP α/γ(β) can be roughly interpreted as the α events rejection efficiency expressed in numbers of σ α . The widths of the L/H bands are determined (in an ideal case) by the fluctuations of the LD baseline noise and the detected photons (that should follow the Poisson distribution). Taking that into account, Figure 4 illustrates the expected discrimination power depending on the noise conditions and the L/H parameter of the detector. The L/H of γ(β)'s is varied from 0 keV/MeV (a non-scintillating material) to 1 keV/MeV (a scintillator with a low light output).  The smallest DP α/γ(β) value present in Figure 4 corresponds to the minimal α/γ(β) separation required for a scintillating bolometer technology to be used in 0νDBD searches. Indeed, with a DP α/γ(β) equal to 3.2, a rejection better than 99.9% of α events with a high acceptance of γ(β)'s (more than 90%) can be achieved. Consequently, the importance of highly performing LDs for scintillating bolometers with a low scintillation efficiency is evident in Figure 4.
The discrimination power parameter is typically used to characterize the particle identification capability of scintillating bolometers developed and/or used in searches for 0νDBD. Certainly, such a parameter can also be calculated for other particle types to be used instead of α's. An example of such particles is shown in Figure 5, where the detection of neutrons by a Li-containing scintillating bolometer and neutron-induced nuclear recoils in materials with different scintillation efficiency are illustrated. A significantly improved particle identification, especially at low energies, is evident for a scintillating bolometer based on an efficient scintillator ( Figure 5). Following the same approach as in Figure 4, we can illustrate the needs of LD performance and L/H for the separation between low energy nuclear recoils (DM search ROI) and γ(β) events, shown in Figure 6. As it was provisioned above (Section 1.3.4), an efficient particle identification for DM search scintillating bolometers requires both an ultra-low LD threshold and a scintillation material with a high light output.

Research and Development on Scintillating Bolometers
This section reports on development and applications of scintillating bolometers based on the following inorganic materials: The main properties of the above listed materials in terms of low-temperature scintillation detection (such as the wavelength of the maximum emission, the light-to-heat ratio for γ(β)'s and the quenching factor for α's) are collected in Table 3. It has to be noted that scintillator-based bolometers have typically a notable difference in the amplitude of phonon signals induced by γ(β)'s and α's of the same energy; this difference is exhibited as ∼(5-15)% higher than the nominal energy of α particles calibrated in the γ energy scale [74,204,347]. Such miscalibration of α's is often not corrected, resulting in a slightly lower QF α value. More details about the data of Table 3 are given below in sections corresponding to each material. Since phonon sensors based on NTD Ge technology were and are widely used in scintillating bolometers, we consider this technology as a default option, omitting to mention in the description of detectors. Table 3. Scintillation properties of inorganic materials used in low-temperature particle detectors with simultaneous phonon and scintillation readout. Crystal growth method(s), wavelength of the emission peak (λ max ) of the material at the quoted temperature, light-to-heat ratio for γ(β)'s, L/H γ(β) , scintillation light quenching for α particles, QF α , with respect to γ(β)s are listed. Typically used crystal growth processes are the following [348][349][350][351]: ordinary and low-temperature-gradient Czochralski (Cz and LTG Cz, respectively), Kyropoulos (Ky), Verneuil (Ve), and variations of Bridgman-Stockbarger (BS) techniques. The L/H γ(β) is evaluated in photons per MeV (ph/MeV) using the energy of photons corresponding to λ max .
First encouraging results for a possible DM search application of a CaWO 4 scintillating bolometer were reported two decades ago [392,393]. A TES-instrumented dual-readout LTD was made of a 6 g CaWO 4 crystal (5 × 10 × 20 mm) and a Si-on-sapphire wafer (SOS, 10 × 20 × 0.5 mm). The detector is characterized by a good scintillation signal of 8 keV/MeV for γ(β)'s, quenched to 29% and 13% for respectively α's and neutron-induced nuclear recoils (mainly scattering off the oxygen nuclei) [392,393]. (Later, the data have been reanalyzed with a more sophisticated pulse height determination and the quenching factors have been re-evaluated to be equal to 0.10-0.12 for nuclear recoils with energies 10-150 keV [378].) Thanks to good CaWO 4 scintillation properties and a good energy resolution of the LD, a particle identification down to 10 keV has been achieved [392,393].
Further prospects of CaWO 4 scintillating bolometers as DM detectors were then demonstrated in an underground operation of a device based on a commercial 54 g CaWO 4 crystal and a 25 mm Ge wafer, realized in the ROSEBUD experiment at the Canfranc laboratory (LSC, Spain). The detector showed a good L/H γ(β) of 6 keV/MeV and a scintillation quenching of 25% and 10% for α particle and nuclear recoils, respectively, thus allowing efficient particle identification with a heat-scintillation readout [267,384,394].
An extensive R&D on CaWO 4 scintillating bolometers (including the Cz-based growth of radiopure crystals [395,396]), as well as their use in the DM searches at LNGS have been realized by the CRESST Collaboration. Eighteen TES-W-instrumented detector modules made of large CaWO 4 crystals ( 40 × 40 mm, ∼300 g each) and either Si (30 × 30 × 0.4 mm) or SOS ( 40 × 0.5 mm) wafers, were operated in the CRESST-II experiment. All detectors of the CRESST-II Phase 1 were made according to a conventional design. Six out of eighteen detectors of the CRESST-II Phase 2 were constructed following new designs to improve an active background suppression: (a) all supporting elements of the crystal and the LD were made of materials with scintillating properties; (b) a complete active 4π-veto system was based on a beaker-shaped LD (see below) and a carrier crystal scintillator for the main energy absorber. The CRESST-III experiment is ongoing with the help of small CaWO 4 crystals (20 × 20 × 10 mm, 24 g) and SOS LDs (20 × 20 × 0.4 mm), aiming at a drastic improvement of the CaWO 4 detector threshold and the experimental sensitivity to low-mass WIMP. A typical light signal detected by the CRESST CaWO 4 scintillating bolometers is ∼(1.0-2.4)% of the total energy deposited by γ(β) radiation in crystals, that is ∼(10-24) keV/MeV, while the scintillation is quenched to 22% and 10% for α particles and nuclear recoils, respectively (e.g., see [397,398]). Thanks to a good scintillation efficiency of CaWO 4 crystals at low temperatures and high performance (i.e., low threshold) bolometric LDs, CRESST CaWO 4 -based detector modules show a powerful particle identification down to few-ten keV region [226,281,376,378,379,[381][382][383][384][385]396,397,399].
The study of light production and transport of CRESST-II detector modules [398] demonstrates the scintillation efficiency of CaWO 4 crystals is (7.4-9.2)%, while the detection efficiency is (18-28)%, or 23% on average. Additional measurements with a photomultiplier demonstrated a similar variation of the detected light, suggesting the dominant part of this variation is originated from the self absorption of the crystals (i.e., difference in the crystals quality) [398]. The detection efficiency was then improved by a factor 2 (to 34%) using two identical LDs facing a CaWO 4 crystal on top and bottom [398]. These results clearly show that the enhancement of the light detection with a single bolometric photodector is feasible via the detector module design optimization.
An impressive light collection optimization has been achieved with a large area, ≈60 cm 2 , beaker-shaped Si absorber (with a diameter and height of 40 mm, around 6 g in weight) allowing enhancement of the scintillation signal by a factor 2.5 compared to conventional CRESTT-II detector modules [293]. In particular, the scintillation light of two identical prototypes based on CaWO 4 crystals ( 35 × 38 mm, ≈220 g each) was detected by the beaker-design LDs at the level of 4.53% and 5.17% (e.g., compare with 1.95% of the average value for the conventional CRESST design [398]). It is also worth noting that the achieved baseline resolutions of such beaker-shaped devices show no degradation in performance compared to smaller CRESST-II LDs (see Table 2) [293,326].
The determination of quenching factors is essential for WIMP searches, in particular with a multi-element detector as CaWO 4 . With this in mind, a precise measurement of quenching factors for O, Ca, and W has been realized using a CaWO 4 LTD irradiated by a neutron beam and the CRESST neutron-calibration data [400]. The study found that the QFs in the 10-40 keV region are typically 0.11, 0.06, and 0.02 for O, Ca, and W, respectively [400].

Cadmium Tungstate
Cadmium tungstate (CdWO 4 )-a classic material for scintillation detectors [374,375]has been actively used in rare-event search experiments with scintillating counters for about three decades (e.g., see [168,387,[401][402][403][404][405][406][407][408]). CdWO 4 -based LTDs with heat-scintillation readout [296,409] can drastically improve the detector performance and particle identification capability compared with scintillation detectors, thus enhancing the sensitivity to rare processes. A particular interest in CdWO 4 scintillating bolometers is the search for DBD of 116 Cd, one of the most promising isotopes for 0νDBD searches from both theoretical and experimental points of view (e.g., see [168,346,406] and references therein).
The first realization of a CdWO 4 scintillating bolometer has been reported 15 years ago [296]; the detector was constructed from a 140 g CdWO 4 crystal and a SiO 2 -coated Ge slab (see Table 4), and tested at LNGS. The scintillation signal ∼6 keV/MeV has been detected for γ and β particles, while the light yield for α particles, well separated from γ(β)'s, is quenched to 18%. It should be emphasized that the LD calibration is done using photon statistics for the observed γ peaks and the noise resolution of this large-area device was estimated as 16 photons, that is ≈48 eV (assuming ∼3 eV/photon) [71,296], never reported later for this LD. Thus, we can speculate that such "low" L/H γ(β) value of the CdWO 4 sample is underestimated. (The results for very old 433 g CdWO 4 crystal, characterized by L/H γ(β) ∼ 14 keV/MeV [410], could also support this assumption.) It is interesting to note that the bolometric LD measured the scintillation peak at 2615 keV with an energy resolution of 3% FWHM, which is better than the resolutions achieved with room temperature CdWO 4 scintillation detectors (3.4%-5.0% [406,411,412]) developed for rare-event search experiments.
A few years later, the technical feasibility of the CdWO 4 scintillating bolometer approach through an array of detectors has been demonstrated using four 213 g crystals and a single 426 g sample viewed by LDs with a Ge wafer diameter of 66 and 35 mm, respectively (Table 4) [413]. In particular, a full suppression of an α-induced background has been illustrated for the 426 g CdWO 4 scintillating bolometer (the quenching factor for α's of 210 Po is around 0.18) [413]. It was also observed that some α events, ascribed to decays at the crystal surface, have a loss of the scintillation signal with respect to bulk α's.
Subsequently, an extensive investigation of a 510 g CdWO 4 scintillating bolometer (Table 4) has been realized [204]. Similar to previous prototypes, the detector exhibits a high L/H γ(β) (17.6 keV/MeV) and QF α = 0.19. The high light yield of the CdWO 4 scintillator worsen the energy resolution of the heat channel because of the heat-light anti-correlation [204,414], however, it can be improved by considering the energy partition between heat and scintillation channels. It was also found that the energy resolution of 6.8% at the 2615 keV scintillation peak of the LD cannot be explained by the fluctuation of the Poisson statistics of the absorbed photons and it is probably dominated by a light yield variation due to position dependent effects. Table 4. Results of low-temperature scintillation detection with scintillating bolometers based on natural/ 116 Cd-enriched CdWO 4 crystals and thin bolometric light detectors (LDs). The crystal growth method, the mass (two significant digits) and the size of the sample, the size and the material (including coating) of the coupled LD, as well as the L/H γ(β) and QF α values are listed. Recently, a 433 g CdWO 4 -based scintillating bolometer (cylindrical sample, Table 4) has been investigated at LSC, aiming at precise investigation of the β-spectrum shape of 113 Cd [410]. The crystal was produced about 25 years ago [404] and for most of this period it has been stored underground, where it was also used in a low-background experiment to investigate precisely the rare β decay of 113 Cd [404]. The choice of this crystal for the experiment is mainly driven by the precise measurement of the Cd isotopic composition in the sample and the high crystal radiopurity [404]. The L/H γ(β) for 2.6 MeV γ quanta of a 232 Th source, measured with the help of a SiO-coated Ge LD, amounts to 14 keV/MeV. However, the light yield exhibits a strong energy dependence, dropping down to L/H γ(β) = 10 keV/MeV for the γ(β) energy deposition in the crystal below 0.1 MeV. The LD exploits the NTL signal amplification and that allowed us to reach a low-threshold of around 60 eV (equal to 5 sigma of the noise fluctuation). CdWO 4 has been also examined for potential use in DM searches exploiting the CRESST technology [415]. In particular, the low-temperature scintillation of a small and of a CRESSTsize crystals (8 g and 400 g, respectively) was measured with Si and SOS LDs, respectively ( Table 4). The small sample was cut from the ingot grown using the LTG Cz method [350,418]. Among two CdWO 4 samples, only a large crystal was equipped with a phonon sensor (TES on a small CdWO 4 carrier). The L/H γ(β) values of the small and large CdWO 4 samples were measured to be 27 and 14 keV/MeV respectively, which is comparable to CaWO 4 . Thus, CdWO 4 is a promising alternative target to CaWO 4 -based DM searches.
Cadmium tungstate produced from cadmium enriched in 116 Cd ( 116 CdWO 4 ) is preferred for both DBD and DM search experiments, because a number of 116 Cd nuclei per unit volume can be increased by one order of magnitude consequently reducing the amount of β-active 113 Cd (12% in natural Cd, 0.56 Bq/kg activity in CdWO 4 ) [350,406]. The first test of a 116 CdWO 4 scintillating bolometer has been recently performed using a 35 g sample and a SiO-coated Ge LD [346]. The sample was cut from a 1.9 kg boule developed from highly purified enriched cadmium (82% enrichment in 116 Cd) [406]. Thanks to a high crystal quality, the measured L/H γ(β) of 31 keV/MeV is the highest ever achieved with CdWO 4 scintillating bolometers. The quenching factors for α particles and nuclear recoils were calculated as 0.18 and 0.08, respectively. The detector exhibits a small light-heat anticorrelation, with a minor impact on the detector energy resolution. Similar results have been recently obtained with two large 116 CdWO 4 scintillating bolometers constructed from twin crystals of around 580 g each cut from the same ingot (the 35 g sample was produced from the same boule too) and operated one at LSC and the other at the Modane underground laboratory (LSM, France). The L/H γ(β) values of 27 and 25 keV/MeV were measured, thus providing an efficient separation of α-induced background [417]. The investigations of 116 CdWO 4 scintillating bolometers [346,417] reinforce results of early studies with such devices based on natural crystals [204,296,413,415], confirming the good prospects of this material for a large-scale bolometric experiment to search for the 116 Cd 0νDBD, in particular using the advantage of multi-target approach [419].

Lithium Tungstate with Mo Content
Due to issues of direct crystallization of the lithium tungstate (LiWO 4 ) stoichiometric melt, such material has been produced by a solid solution crystal growth with molybdenum admixture resulting in Li 2 Mo 1−x W x O 4 , where x is the molybdenum mole ratio [420]. High quality crystals have been produced from a stoichiometric mixture with x = 0.05 using Cz [420] and LTG Cz [352,421] crystal growth methods.
In 2017, this material was selected for the BASKET project [249] aiming at the development of a cryogenic detector suited to the study of CENNS in above-ground conditions. The Li ( 6 Li) content ensures the neutron detection capability, as demonstrated with other Li-containing scintillating bolometers, while the presence of a heavy target (W) increases the CENNS events detection rate [55]. Therefore, a low-threshold LiWO 4 (Mo) scintillating bolometer is particularly interesting to be used as CENNS detector, neutron flux monitor, and/or active veto [422].
A first scintillating bolometric test of the material was done with a Cz-grown Li 2 Mo 0.08 W 0.92 O 4 crystal ( 18 × 7 mm, 8 g) coupled to a 44 mm Ge LD [422]. A rather low scintillation signal (L/H γ(β) = 0.17 keV/MeV) was detected, but the result is affected by non-optimal light collection (the absence of a reflective film around the lateral side of the crystal and uncoated Ge wafer). However, a good α/γ separation has been achieved using the NTL amplification of the LD signals (QF α was computed as 0.28).
A poor scintillation efficiency of the compound has been also observed in lowtemperature tests of two LTG-Cz-grown samples of Li 2 Mo 0.05 W 0.95 O 4 (10 × 10 × 10 mm, 4.4 g and 25 × 25 mm, 52 g) [423]. In particular, the 4.4-g-based detector measured L/H γ(β) = 0.15 keV/MeV in conditions similar to the above mentioned test of the 8 g sample. The 52 g Li 2 Mo 0.05 W 0.95 O 4 scintillating bolometer exhibited L/H γ(β) = 0.4 keV/MeV (quenched to 26% for α+t events) thanks to improved light collection conditions (a SiO-coated Ge LD, a reflective film at bottom and around the lateral side of the sample). An α/γ(β) separation has been achieved thanks to the NTL mode operation of the LD.

Sodium Tungstate
A Na-and/or I-containing scintillator without hygroscopic properties, in contrast to sodium iodide (and cesium iodide), can be interesting for DM searches, in particular to scrutinize the nature of the signal modulation observed in the DAMA/NaI [424] and DAMA/LIBRA [425][426][427] DM search experiments. Taking into account such a possible application, a sodium tungstate (Na 2 W 2 O 7 ) represents a great interest as a cryogenic scintillator [428]. The first bolometric operation of the material has been recently realized at LNGS using a 5.6 g Na 2 W 2 O 7 scintillation element (10 × 10 × 10 mm, LTG Cz growth) coupled to a large-area Ge LD ( 50.8 × 0.2 mm) [429]. An efficient scintillation (L/H γ(β) = 12.8 keV/MeV) together with an excellent particle identification capability (QF α = 0.20) have been observed. These results show good prospects of a Na 2 W 2 O 7 -based scintillating bolometer for searches for DM particles.

Lead Tungstate
Thanks to the presence of the heaviest stable element (Pb) in lead tungstate (PbWO 4 ) and increased scintillation at low temperatures (e.g., [430]), this well-known scintillator is attractive for bolometric DM searches [431,432]. In addition, there are four Pb isotopes that can potentially undergo α decay, but the theoretical predictions [29] are rather pessimistic from the experimental point of view. At the same time, the PbWO 4 production using modern lead is a drawback of the material for rare-event searches, in particular using Pbcontaining bolometers [409]. Indeed, a considerably high contamination (tens-thousands Bq/kg) of modern lead by β-active 210 Pb (Q β = 63.5 keV, T 1/2 = 22.3 yr) drastically affects the background and the operation of a thermal detector. The 210 Pb issue can be solved by using lead produced hundreds of years ago, the so-called Roman, or ancient, or archeological lead (see [409,433,434] and references therein). The production of PbWO 4 crystals (Cz method) from ancient lead is reported in [407,408,435]. A first test of the material as a scintillating bolometer has been realized at LNGS with a device fabricated from a large PbWO 4 sample (30 × 30 × 61 mm, 454 g) and a Ge wafer ( 36 × 1 mm) [435]. The measured PbWO 4 scintillation signal for γ(β) interactions is reasonably good (L/H γ(β) = 1.78 keV/MeV), being five times lower for α particles (QF α = 0.2) [435].

Zinc Tungstate
Zinc tungstate (ZnWO 4 ) has a rather long-standing interest as a detector material for the searches for DBD ( 64,70 Zn and 180,186 W isotopes) and DM (spin-independent and spindependent interactions, diurnal modulation) [96,[436][437][438][439][440]. Such an interest is driven by a reasonable scintillation efficiency (comparable with CaWO 4 ), one of the highest radiopurity among crystal scintillators, the presence of isotopes with non-zero spin ( 67 Zn, 183 W), and anisotropic properties of the material.
The first test of this compound as a scintillating bolometer was done using a CRESSTlike detector module with a 8 g ZnWO 4 crystal (20 × 10 × 5 mm, LTG Cz growth) and a SiO 2 -coated silicon wafer (30 × 30 × 0.5 mm) both instrumented with W superconducting thermometers [441]. A pulse-shape difference between the ZnWO 4 luminescence and particles directly impinging on the LD was observed. A high scintillation signal, resulted to L/H γ(β) ∼ 14 keV/MeV, was measured. A non-linearity of the ZnWO 4 light output for electron recoils was evident below ∼150 keV heat energy deposition (a curious reader can find details about a non-proportional scintillation response, for example, in [374,442,443]). The light signal for nuclear recoils is quenched to ∼10% and, thanks to that, such events can be clearly separated from γ(β)'s for phonon signals above 20 keV. This test shows that the light output of ZnWO 4 crystals is among the highest for oxide scintillators. Moreover, the ZnWO 4 light output can be further increased by annealing. For example, a ∼(10-30)% improvement of a light output at room temperature has been achieved for LTG-Cz-grown ZnWO 4 crystals annealed at 800 • C [444].
Recently, the light collection efficiency of the CRESST conventional detector design has been investigated using a ZnWO 4 sample ( 40 × 40 mm) and several CaWO 4 crystals [398]. The ZnWO 4 scintillation efficiency is estimated to be 8%, while only one fourth is detected as scintillation light (1.9%) due to ∼24% of the light collection efficiency [398]. The results obtained for the ZnWO 4 detector are comparable with those of the CaWO 4 -based scintillating bolometers [398].
An anisotropy in the ZnWO 4 light output (at room temperature) for α particles [437,445] and nuclear recoils [445]-a key feature for the ZnWO 4 -based detection of the diurnal asymmetry of WIMP direction [231,437]-has been also reported recently in studies of a ZnWO 4 scintillating bolometer (1 cm 3 ) with MMC readout of the phonon and scintillation channels [446].
Because of the W content, a ZnWO 4 -based detector is also suitable for the investigation and searches for rare α decays of W isotopes [447]. Moreover, an interesting approach has been proposed in [448], where a scintillating bolometer made of a 22 g ZnWO 4 crystal doped with Sm isotopically enriched in 148 Sm to 95.54% (0.16% of 148 Sm 2 O 3 powder in the initial charge; Cz growth) was used for the precise measurement of the 148 Sm α decay half-life.
Furthermore, an internal active shield based on ZnWO 4 scintillators with a bolometric light readout is considered, for the first time in an array of macrobolometers, by the BINGO project [220] aiming at the development of an advanced background rejection for a possible follow-up of the CUPID tonne-scale DBD search experiment. The BINGO concept is going to be demonstrated by the MINI-BINGO small-scale experiment using lithium molybdate and tellurium dioxide crystals of natural isotopic composition with a ZnWO 4 -based 4π active shield. As a very first step, a scintillating bolometer constructed from an optically polished small ZnWO 4 crystal (10 × 10 × 10 mm, 8 g, LTG Cz growth) coupled to a SiO-coated Ge LD ( 44 × 0.2 mm) has been recently investigated at IJCLab (Orsay, France) with excellent results in terms of the detected light (∼15 keV/MeV for γ(β)'s) [449]. A subsequent bolometric test of low-temperature scintillation detection from a 60-mm-long ZnWO 4 bar (30 mm in diameter) shows a similar light yield of the material (the measured L/H γ(β) is ∼14 keV/MeV) [450].

Calcium Molybdate
R&D on calcium molybdate (CaMoO 4 ) scintillation detectors, in particular scintillating bolometers, for DBD search (mainly 100 Mo) has been going on for nearly two decades [296,451,452]. This material has also been studied for the bolometric detection of DM particles [444,453].
A good prospect for the application of CaMoO 4 scintillating bolometers in this field was first demonstrated 15 years ago, achieving a full α/γ(β) separation with a few-grams detector coupled to a Ge bolometer [296]. Later, a powerful particle identification of such detectors was shown with a massive CaMoO 4 crystal (158 g, 40 × 35 mm) [101]. The L/H γ(β) was measured as 1.87 keV/MeV and the QF α parameter was found to be ∼0.15 [101]. The investigation of the scintillation of this material, obtained with a 20 × 10 × 5 mm crystal and a 30 × 30 × 0.5 mm SiO 2 -coated Si LD, instrumented by a TES W sensor, reports a L/H γ(β) value of 4.77 keV/MeV induced by 60 keV γ quanta of an 241 Am source [453].
The extensive developments of CaMoO 4 scintillating bolometers have been realised over the last decade for the AMoRE project [81,454]. In particular, the fabrication of high quality CaMoO 4 crystals from calcium depleted in 48 Ca and molybdenum enriched in 100 Mo ( 48depl Ca 100 MoO 4 ) has been developed [455]. The 2νDBD-activity of 48 Ca (in spite of 0.2% content in natural calcium [456]) can represent a major background contribution to the 100 Mo 0νDBD ROI [296,451,452], thus, the 48 Ca content in calcium used for the AMoRE crystals production is reduced below 0.001% [457]. The purification of the starting materials has been adopted to improve the crystal quality and radiopurity [81,455,458,459]. The growth of the 48depl Ca 100 MoO 4 crystals is done using the Cz method; the double crystallization is applied to further improve the radiopurity of the ingots [81].
The main milestones achieved in the AMoRE R&D on MMC-instrumented CaMoO 4 scintillating bolometers can be summarized as follows: • first single-readout CaMoO 4 prototypes based on small (10 × 10 × 6 mm, 2.7 g) [285] and large-volume ( [464][465][466][467][468], currently in progress at Y2L. A highly efficient identification of α events has been demonstrated with all these dual-readout bolometers [465]. Unfortunately, the measurements lack the LD calibrations, thus, the information about the L/H γ(β) values is not available. The only particle identification parameter which can be extracted from the published data is QF α , estimated as ∼0.2 [461,462]. In spite of a huge progress in the developments of CaMoO 4 scintillating bolometers, the AMoRE Collaboration is performing R&D on other Mo-containing crystal scintillators [468,469] for the AMoRE-II stage of the experiment (with a 200 kg scale detector array) to avoid the needs of the 48 Ca-depleted starting material and to mitigate the remaining issue with the purification of Ca-based compounds.

Cadmium Molybdate
Cadmium molybdate (CdMoO 4 ) was considered a promising scintillation material for cryogenic detectors about 15 years ago [260], in particular the luminescence of this material at 9 K was reported to be 80% of CaWO 4 . In short, an excellent α particle identification was demonstrated with a first scintillating bolometer based on a small (10 × 10 × 5 mm) CdMoO 4 crystal [266,470]. In spite of the first encouraging results, the interest in this material has been reactivated only recently, in particular by proposing to use CdMoO 4 enriched in 116 Cd and 100 Mo for a bi-isotope search for 0νDBD [471]. As a follow-up, a first test of a large scintillating bolometer based on this compound has been recently realized with a colorless 134 g CdMoO 4 crystal ( 25 × 45 mm, BS growth) and 44 mm Ge LD without anti-reflective coating [472]. These measurements demonstrate good prospects for CdMoO 4 cryogenic scintillators; in particular, a L/H γ(β) value of ∼2.6 keV/MeV and a quenching to 16% for α particles ensures an excellent α/γ(β) separation.

Lithium Molybdate
Lithium molybdate (Li 2 MoO 4 ) is the best example of recent developments of scintillation materials with a strong impact on the strategy of bolometric DBD search experiments and related activities. Li 2 MoO 4 was first considered a prospective material for 100 Mo DBD search with LTDs about a decade ago [473]. An outstanding progress in the development of Li 2 MoO 4 crystal scintillators and their application in rare-event searches with scintillating bolometers has been achieved over the last six years. Consequently, Li 2 MoO 4 is now considered the most viable detector material for tonne-scale bolometric 0νDBD searches to be realized in the near future.
Going back to the first operation of a Li 2 MoO 4 crystal (1.3 g; Cz growth) as a scintillating bolometer [266], it was hard to imagine such a success of the material because the results of the test were far from being appealing. In particular, the light yield was estimated to be ≈20% of the CaMoO 4 one [266] (i.e., L/H γ(β) ∼0.4 keV/MeV using CaMoO 4 data from [101]). Despite the poor detector performance, a hint on the α/γ(β) separation was demonstrated [266]. Few years later, a similar L/H γ(β) (0.43 keV/MeV) combined with good performance were reported for a 33 g Li 2 MoO 4 scintillating bolometer [474] (from the same crystal producer of [266]). The encouraging results of this study were then reinforced by the characterization of a 151 g Li 2 MoO 4 (LTG Cz growth) scintillating bolometer [311], showing excellent performance, almost doubled L/H γ(β) (∼0.7 keV/MeV) and a highly efficient α/γ(β) separation.
Moreover, 100 Mo-enriched crystals (Li 2 100 MoO 4 ; 97% of enrichment in 100 Mo) have been already developed and used in the LUMINEU (4-detector array) [74,173,345] and its follow-up CUPID-Mo (20-detector array) [159,259,[497][498][499] bolometric DBD search experiments at LSM. Despite of the modest exposure of these small-scale CUPID demonstrators, they have provided valuable physics results. Notably, the most precise measurement of the 100 Mo 2νDBD half-life (the second highest precision among all 2νDBD-active nuclides) [173] and the most stringent half-life limit on the 0ν-mode [159] have been achieved. Furthermore, Li 2 100 MoO 4 bolometers will be used in the CROSS DBD search experiment (32-52 detectors) [69,118,500] and possibly in the BINGO demonstrator [220]. In addition to the enriched detectors, several crystals have been produced from molybdenum depleted in 100 Mo (Li 2 100depl MoO 4 ) [494] and are going to be used together with enriched ones for the investigation of the 2νDBD spectral shape. Preliminary results of the first scintillating bolometer test of the Li 2 100depl MoO 4 sample ( Table 5) are encouraging [500]. Despite the absence of the reflective film inside the Cu holder, that caused to register a lower light [345], an excellent α/γ(β) separation has been achieved thanks to the good performance of the LD. Table 5. Results of low-temperature scintillation detection with scintillating bolometers based on natural/ 100 Mo-enriched Li 2 MoO 4 crystals and thin Ge LDs. The crystal growth method, the mass (two significant digits) and the size of the sample, the size and the coating of the coupled LD, as well as the L/H γ(β) and QF α values are listed.  Figure 7 [474] is ∼0.7 keV/MeV, in contradiction to ; this discrepancy is probably the origin of the reported QF α = 0.43 [474]. b Estimated for α+t events (4.8 MeV sum energy), products of neutron capture by 6 Li. c Affected by smaller area of an LD compared to one of the faced crystal side. d No reflective foil, but a Cu housing. e Such spread is mainly due to the light collection difference imposed by the detector design (see text). f Combination of two identical LDs. g No reflective cavity, that is fully open detector structure.
Finally, Li 2 100 MoO 4 scintillators have been selected from the list of 82 Se-, 100 Mo-, 116 Cd-and 130 Te-containing crystals for the realization of the CUPID tonne-scale bolometric experiment [82,283]. Last but not least, Li 2 MoO 4 detector material is also a part of the AMoRE DBD project [81,[483][484][485]. Several Li 2 100 MoO 4 scintillating bolometers (few out of 18 detectors) are operating in the AMoRE-I DBD experiment [465], aiming at investigating the possibility to use Li 2 MoO 4 (instead of CaMoO 4 , see Section 3.2.1) in the large-scale AMoRE-II detector array [81]. In addition to the great interest for DBD searches, Li 2 MoO 4 low-threshold scintillating bolometers are promising detectors for low-mass DM searches with a high sensitivity to spin-dependent interactions with 7 Li, as demonstrated for the first time in [322]. A large content of 7 Li can also be exploited as a target for a resonant absorption of solar axions [43,266,474,502,503]. Thanks to the 8% content of 6 L in natural lithium, Li 2 MoO 4 scintillating bolometers can be used for neutron detection in a ROI populated only by α events [74,311,345,474,478,479,485,498]. A high radiopurity of the material [69,74,345,479,497,498] allows the suppression of the contribution of bulk/surface radioactivity down to the ROI for neutron spectroscopy.
The measurements of scintillation light with natural and 100 Mo-enriched Li 2 MoO 4 scintillating bolometers are listed in Table 5 and can be summarized as follows: • The Cz-grown Li 2 MoO 4 crystals (developed by CLYMENE) [478,479] exhibit similar light yield to the LTG Cz produced scintillators. The amount of the detected light for a standard detector design envisaging the use of a reflective film is compatible with highly efficient particle identification in the 100 Mo 0νDBD ROI. • Crystals produced by the LTG Cz growth from the purified starting materials show reproducible value of the light yield within a minor variation for the same detector structure. In particular, the largest L/H γ(β) (0.90 keV/MeV median value) for the 20 similar size Li 2 100 MoO 4 detectors of CUPID-Mo has been measured for crystals viewed by a single LD (placed at bottom) [259,498]. The use of two LDs reduces the amount of the light detected by each of them to 0.64 and 0.74 keV/MeV (median values) for the bottom and top photodetectors, respectively [259,498]. A small difference in the light signals seen by the top and bottom LDs is explained by a slightly reduced entrance window for the bottom one (required to place the crystal). The combination of two LDs allows us to double the measured scintillation light signal (median L/H γ(β) is 1.33 keV/MeV) and, subsequently, to enhance the particle identification efficiency [159,283,498,504]. • A notably smaller photodetector area than the crystal side facing it (e.g., a factor of 3 difference [74,275]) can decrease the light collection drastically. An order of tens % difference [69,283,501] can be tolerated, because an efficient particle identification capability would be still possible without the needs of a high-performing LD. • The absence of a reflective film around a crystal inserted inside a fully closed Cu housing, decreases the light collection by almost a factor 2 [345]. A similar reduction factor is observed for bare crystals compared to ones surrounded by the reflective film in an opened detector structure [283]. This result combined with Monte Carlo simulations of the scintillation light production, propagation, and absorption show that the surface roughness does not play an important role for Li 2 MoO 4 crystals [283] (for instance, similar observations are reported for tungstates [276,277], while a stronger impact of the surface roughness on the light collection is expected, for example, for zinc molybdate [275] and tellurium dioxide [271,505]). • Despite a large variation in the measured L/H γ(β) , imposed by light collection efficiency, the QF α value remains rather similar, ∼0.2, showing a small variation. The quenching factor for α+triton events (QF α+t ), detected in neutron calibrations of Li 2 MoO 4 scintillating bolometers, is about 10% larger than that of α's of similar energy. The difference in the light yield induced by α and α+t interactions in a scintillator illustrates Birck's formula (see Section 1.2): more than a half of the energy release in the 6 Li(n,t)α reaction is taken away by a lighter nucleus, triton, which induces a higher light output than α of the same energy loss. Table 5), other phonon sensor technologies have been recently used. Among the features of these technologies, it is important to emphasize a faster detector response and the possibility of channel multiplexing. Fast timing of heat and light signals is of special importance for 100 Mo-enriched bolometers to suppress the background in the 100 Mo 0νDBD ROI induced by random coincidences (mainly due to 2νDBD events) [205][206][207]. The AMoRE prototypes of Li 2 MoO 4 scintillating bolometers are instrumented with MMC sensors [484,485]. Only partial scintillation-based particle separation (improved to 99.9% by the heat pulse-shape analysis) has been achieved in the measurements with a small crystal coupled to a Ge LD (15 × 15 × 0.5 mm) [484]. Conversely, an efficient particle identification around the 100 Mo 0νDBD ROI has been demonstrated with a much larger ( 50 × 48 mm) crystal paired with a Ge disk ( 50 × 0.5 mm) [485]. It was also observed that a weak hygroscopicity of the material can drastically impact the phonon signal amplitude if no precaution is taken to avoid exposition to a humid environment [485]. Another technology, employing KIDs developed within the CALDER project [286,338,506], has been used to test a Li 2 MoO 4 scintillating bolometer based on a 24 g crystal (20 × 20 × 20 mm; LUMINEU sample) and a Si wafer (20 × 20 × 0.3 mm). The LD signals are characterized by a rise time of ∼0.2 ms [286], a factor 3-5 faster than ever reported for NTD-Ge-instrumented LDs. A study of the KID LD response shows that the Li 2 MoO 4 scintillation decay time, measured as 85 (5) µs, is constant in the 10-190 mK range [286].

Lithium Magnesium Molybdate
Lithium magnesium molybdate (Li 2 Mg 2 (MoO 4 ) 3 ) is another Mo-containing material recently developed and tested at low temperatures for rare-event searches [356]. The Li 2 Mg 2 (MoO 4 ) 3 compound contains one of the largest number of Mo atoms per crystal volume (e.g., comparable to CaMoO 4 , and 20% more than Li 2 MoO 4 ). The absence of hygroscopic properties of the material (in contrast to weak hygroscopicity of Li 2 MoO 4 ) is an advantage too. An optically-clear quality Li 2 Mg 2 (MoO 4 ) 3 crystal growth was realized successfully using a 5N grade MoO 3 powder and the LTG Cz technique [356]. An element with a size of 19 × 14 × 10 mm and a mass of 10.2 g was produced for a bolometric test. Scintillation detection was done with the help of a SiO-coated Ge LD ( 44 × 0.2 mm) assisted with the NTL signal amplification. The measured L/H γ(β) of 1.3 keV/MeV is comparable to the results of ZnMoO 4 (Section 3.2.10) and slightly exceeds the values reported for Li 2 MoO 4 (Section 3.2.3) crystal scintillators. The scintillation induced by α particles is quenched to 22%. The detector provides an efficient particle identification satisfying the requirements of a bolometric 0νDBD search experiment and allowing the use of the material for a bolometer-based neutron detection [356]. The variety of elements with different atomic masses present in Li 2 Mg 2 (MoO 4 ) 3 is of particular interest for DM search applications [356].

Lithium Zinc Molybdate
Lithium zinc molybdate (Li 2 Zn 2 (MoO 4 ) 3 ) was developed a decade ago as a potential LTD of DBD processes in Zn and Mo isotopes (with the main interest in 100 Mo) [357]. A small sample (20 × 10 × 2 mm; LTG Cz growth), cut from a crystal boule grown from purified materials, was successfully operated as a single readout LTD [357]. Despite a rather low scintillation of the material detected at 223 K (∼(3-4)% of CaMoO 4 ), an increase of the Li 2 Zn 2 (MoO 4 ) 3 light output at low temperatures has been observed [357], opening a possibility to use Li 2 Zn 2 (MoO 4 ) 3 as a cryogenic scintillator.

Magnesium Molybdate
A possible application of magnesium molybdate (MgMoO 4 ) to rare-event searches with scintillating LTDs was investigated for the first time 15 years ago [358]. The scintillation of the material increases steeply below ∼30 K [260,507]. A bolometric test of this scintillator (89 g sample, 32 × 31 × 24 mm) has been realized with a single readout only (the assembly structure did not allow for the mounting of an LD) [101]. Moreover, the detector performances were strongly affected (possibly by a crystal crack under the glued thermistor). Despite of that issue, a clear pulse-shape difference between γ(β) and α events has been demonstrated [101]. These results definitely indicate the possibility of particle identification with a MgMoO 4 -based scintillating bolometer.

Sodium Molybdate
In addition to DM search applications (as for sodium tungstate, Section 3.1.4), sodium molybdate might be interesting for Mo-based DBD search experiments, in particular for the AMoRE project [508]. A good progress in the development of a large volume Na 2 Mo 2 O 7 crystals using the Cz growth has been achieved within the AMoRE R&D [508]. The light output of the Na 2 Mo 2 O 7 scintillator at 10 K was found to be 55% of the CaMoO 4 light yield [508], which is rather promising for scintillating bolometer applications.
A growth and low-temperature characterization of Na 2 Mo 2 O 7 and Na 2 Mo 4 O 13 crystals have been reported in [265]. A 1.6 g sample of Na 2 Mo 4 O 13 crystal grown by the vertical Bridgman method was of low quality because of the difficulties in the growing process of this compound. Despite this problem, a scintillation-assisted particle identification has been demonstrated with the Na 2 Mo 4 O 13 scintillating bolometer. A Cz-grown Na 2 Mo 2 O 7 sample was larger (11 g) and of better quality single crystal. An aboveground characterization of the sample at low temperatures also demonstrates a clear separation between γ/β/µ-induced events and α particles [265].
A highly purified MoO 3 powder and a commercial Na 2 CO 3 (4N purity grade) were used as starting materials for the LTG-Cz-based production of the Na 2 Mo 2 O 7 scintillator [429]. In a bolometric operation of a small sample (10 × 10 × 10 mm, 3.6 g), realized together with the same-size Na 2 W 2 O 7 crystal (see Section 3.1.4), the detected scintillation is characterized by L/H γ(β) = 1.61 keV/MeV and QF α = 0.16 [429]. In spite of relatively modest performance of the LD (0.3 keV RMS noise), a particle identification was clearly demonstrated. Another LTG-Cz-grown Na 2 Mo 2 O 7 crystal (10 × 10 × 10 mm) together with a Ge slab (15 × 15 × 0.05 mm) were used in an MMC-instrumented scintillating bolometer [484]. Using the dual readout, the detector achieved only a partial separation of α particles from γ and β events, while the α rejection efficiency was improved to higher than 99.9% with a pulse-shape analysis of the heat channel [484].

Lead Molybdate
A large increase of the lead molybdate (PbMoO 4 ) fluorescence at cryogenic temperatures has been reported decades ago [430,509], emphasizing the good prospects of the material for scintillating bolometer searches for the 100 Mo 0νDBD [509,510]. A powerful α/γ(β) separation with a small, a few grams, PbMoO 4 scintillating bolometer was demonstrated for the first time 15 years ago [296]. As in the case of PbWO 4 (see Section 3.1.5), the 210 Pb contamination is a limiting factor of this material to be used for bolometric DBD searches [296]. However, PbMoO 4 remains attractive for rare-event searches [431,432], especially if 210 Pb-free lead can be used for the crystal production.
Recently, highly purified ancient lead samples were used for the PbMoO 4 growth by Cz [511] and LTG Cz [512] methods and 57 g ( 20 × 30 mm) and 570 g ( 44 × 55 mm) samples were cut respectively from the grown ingots for bolometric characterizations. Two scintillating bolometers based on these crystals were constructed in the same way and using similar size Ge LDs ( 44 × 0.3 mm and 44 × 0.2 mm). Low-temperature tests were realized in the same underground facility at LNGS. The material exhibits the highest scintillation yield ever reported for molybdates at low temperatures. However, the measured L/H γ(β) is significantly different between the modules: 5.2 and 12.0 keV/MeV for the Cz-and LTG-Cz-grown PbMoO 4 crystals, respectively [511,512]. This difference is probably due to higher quality (and/or purity) of the latter sample. The α particle induced scintillation is quenched to a ∼20% level for both PbMoO 4 samples (the QF α parameters are 0.23 and 0.18 for the Cz-and LTG-Cz-grown crystals, respectively [511,512]).

Strontium Molybdate
Strontium molybdate (SrMoO 4 ) is a potential scintillation material for the searches for DBD of Sr and Mo isotopes. The feasibility of SrMoO 4 as a phonon-scintillation LTD, in particular a clear α/γ(β) separation, as well as a problem with the crystal radiopurity were demonstrated about 15 years ago [296]. In contrast to the γ and β events, the detector response to α radiation exhibits a non-linearity of the scintillation light output with the increase of particle energy. Therefore, the QF α value can only be roughly estimated as ∼0. 26.
The interest in the material has been renewed recently in view of a progress in the crystal growth (Cz method). The measurements of a SrMoO 4 scintillation at 11 K, carried out using the multi-photon counting technique [513], has determined its light yield to be 15(5)% of ZnWO 4 [361], that is L/H γ(β) ∼ (1-3) keV/MeV taking into account the results for ZnWO 4 crystals (Section 3.1.6).

Zinc Molybdate
Zinc molybdate (ZnMoO 4 ) is another crystal scintillator extensively developed over the last decade, being considered one the most promising Mo-containing scintillation materials for bolometric experiments to search for the 100 Mo 0νDBD [74,344,[514][515][516][517]. Early bolometric tests of the material [101,275,307,344,514,518,519] have been realized mainly within Bolux R&D Experiment, LUCIFER, and ISOTTA projects, while a technology of radiopure, natural and 100 Mo-enriched ZnMoO 4 scintillating bolometers has been developed as a part of the LUMINEU project [74,318,362,475,[515][516][517][520][521][522][523][524]. The developed ZnMoO 4 scintillating bolometers and the results of scintillation detection with these devices are listed in Table 6. Table 6. Results of low-temperature scintillation detection with scintillating bolometers based on natural/ 100 Mo-enriched ZnMoO 4 crystals and thin Ge LDs. The crystal growth method, the mass (two significant digits) and the size of the sample, the size and the coating of the coupled LD, as well as the L/H γ(β) and QF α values are listed. The low-temperature investigations of ZnMoO 4 found that this material is characterized by a reasonably good scintillation (see Table 6), which allows for a highly efficient particle identification in the ROI of 100 Mo 0νDBD. The only abnormal peculiarity, observed in scintillation-vs.-heat data of ZnMoO 4 -based LTDs, is a tail (with a negative slope) of α lines of bulk contaminants to lower light signal values [74,101,313,518]. Since the QF α value is lower than 1, this anomaly results in an increased separation of such α's, but worsen the α energy resolution. This feature is probably due to inclusions seen in the samples. In spite of a great progress in the development of large-mass crystal boules (about 1 kg mass) using highly purified natural/ 100 Mo-enriched molybdenum compounds [74,522], the difficulties with the ZnMoO 4 solidification process affect the crystal quality. It was found that doping with a small amount of W helps to mitigate the ZnMoO 4 growing issues and to get better quality crystals with similar low-temperature scintillation efficiency [523]. However, the growth of a large-volume crystal boule needs further R&D to improve the presently achievable crystal quality [74]. Therefore, another Mo-containing material (lithium molybdate, Section 3.2.3) has been developed as a viable alternative to zinc molybdate.

Lithium Europium Borate
In view of the poor scintillation of lithium fluoride, largely investigated for neutron detection (Section 3.6.2), lithium europium borate (Li 6 Eu(BO 3 ) 3 ) was studied as a possible scintillating bolometer for neutron spectroscopy exploiting (n, α) reactions on 6 Li and 10 B [59]. In particular, a Li 6 Eu(BO 3 ) 3 -based LTD would have an efficiency for neutron detection similar to lithium fluoride [59]. The feasibility of this approach has been investigated with a 5 mm side cube Li 6 Eu(BO 3 ) 3 scintillating bolometer irradiated by a neutron source. A good capability to distinguish three expected neutron capture reactions- 6 Li(n, t)α, 10 B(n, α) 7 Li, and 10 B(n, α) 7 Li+γ-has been demonstrated [59]. Li 6 Eu(BO 3 ) 3 was also considered a viable Eu-containing material to search for α decays of naturally occurring europium [525] (e.g., as an alternative to Eu-doped calcium fluoride [526], Section 3.6.1), in particular, using a Li 6 Eu(BO 3 ) 3 -based LTD. Such idea has been lately realized using a scintillating bolometer made of a 6.2 g Li 6 Eu(BO 3 ) 3 crystal (Cz-grown) and a Ge LD ( 44 × 0.3 mm) to detect the rare α decay of 151 Eu [527,528]. The material has a good scintillation at low temperatures, as exhibited by the measured L/H γ(β) of 6.55 keV/MeV. The α particles induced scintillation was found to be quenched to 8%, providing excellent identification capability.

Lithium Gadolinium Borate
Lithium gadolinium borate (Li 6 Gd(BO 3 ) 3 ), developed two decades ago [529], is an attractive scintillation material for neutron detection thanks to the content of 6 Li, 10 B, and 155,157 Gd. It is worth noting that the large thermal neutron cross sections for 155,157 Gd(n, γ) reactions result in relatively low detection efficiency of a Li 6 Gd(BO 3 ) 3 -based detector to thermal neutrons as compared to, for example, Li 6 Eu(BO 3 ) 3 or lithium fluoride [59]. On the contrary, such a feature of Li 6 Gd(BO 3 ) 3 can be exploited to get a cleaner fast neutron spectrum near the thermal peak [59].
A small-size (5 × 5 × 5 mm) Li 6 Gd(BO 3 ) 3 enriched at 95% in both isotopes 6 Li and 10 B has been investigated as a scintillating bolometer [59,371]. A poor L/H γ(β) (0.26 keV/MeV) has been measured. The scintillation for α particles is found to be quenched to 23% of the γ(β)-induced one. It is interesting to note that the scintillation yield registered for α+triton events (4.8 MeV) is 35% higher than that of 241 Am α's (5.5 MeV). This evidence might suggest a good scintillation of the 6 Li/ 10 B-enriched Li 6 Gd(BO 3 ) 3 bolometer, possibly affected by non-optimal light collection conditions (e.g., caused by a low reflectivity of the cavity for the scintillation emission spectrum occurring in the ultraviolet range [530]). Indeed, such a large difference between QF α and QF α+t has been recently observed with an efficient Li-containing scintillator (lithium indium diselenide, Section 3.5.1).

Aluminium Oxide
Aluminium oxide (Al 2 O 3 ) is a very promising target for low-mass WIMP spinindependent searches allowing also investigation of the spin-dependent interactions ( 27 Al, 100% isotopic abundance) [531]. (Al 2 O 3 is called sapphire for all synthetic crystals, except red-colored ones (rubies) which are with the admixture of Cr.) In particular, this material has been used in the single readout bolometric DM search detectors of the EDEL-WEISS/MANOLIA [532], ROSEBUD [533][534][535], and the CRESST (instrumented with TES W) [536,537] experiments. Scintillating bolometers based on this target (as well as bismuth germanate, Section 3.4.2, and lithium fluoride, Section 3.6.2) have been extensively developed and studied as part of the ROSEBUD DM search program [257,267,[538][539][540][541]. In addition to being used as DM detector, an Al 2 O 3 scintillating bolometer, operated alone or together with a 6 Li-containing LTD, can be exploited for the monitoring of fast neutrons (which can mimic a DM signal) [60,257,258].
A scintillating bolometer based on a Ky-grown nominally undoped sapphire (Ti content is lower than 10 ppm in weight) exhibits a good scintillation, in particular the L/H γ(β) values of 50 g prototypes coupled to 25 mm Ge LDs were measured as 12.7-13.5 keV/MeV [267,530,538]. A very similar scintillation signal of a Ti-doped Al 2 O 3 crystal (4 g, Ky growth), that is 13 keV/MeV, was detected by a Ge LD [267]. The scintillation induced by α particles and nuclear recoils is quenched to around 10% and 6%, respectively [267,538], also demonstrating the dependence of quenching factors on energy [541]. Such performance of a sapphire-based scintillating bolometer provides a powerful particle identification, down to 10 keV [539].
In the study of two scintillating bolometers made of Ve-grown undoped Al 2 O 3 crystals with mass of 25 and 200 g coupled respectively to 25 and 40 mm Ge LDs, five times lower L/H γ(β) values (2.5-2.8 keV/MeV) have been measured [530]. Such a large difference in the detected scintillation of Ve-and Ky-grown Al 2 O 3 crystals cannot be completely explained by the difference in the light collection efficiency (a relative factor of 1.7, originated to the difference in the reflectivity between uncoated and Ag-coated Cu cavities) [530]. The evaluated quenching factors for α's and nuclear recoils (QF α = 0.30-0.36 and QF recoil = 0.12 [530]) are 2-4 times larger than those obtained for Ky-grown crystals [267,538]. Due to a lower light output (combined with a lower scintillation quenching for nuclear recoils), a clear separation between γ(β) and nuclear recoils has been achieved only down to around 60 keV [530].
Further investigations of sapphire-based scintillating bolometers have been realized with four 4-50 g crystals from several suppliers and with different concentration of Ti and Cr, as part of the SciCryo project [542]. Either Ge or Si LDs were coupled to the studied scintillators. The L/H γ(β) values (8-14 keV/MeV) measured for three out of four samples [542,543] are similar to previously reported results for Ky-grown crystals [267,538], while a considerably lower scintillation (2.5 keV/MeV) was detected from the fourth one [542,543]. A 50 g Al 2 O 3 scintillating bolometer was also operated in the EDELWEISS set-up at LSM, demonstrating the integration of a scintillation-phonon device into the ionization-phonon detector array of the cryogenic DM search experiment [542,543]. This test, relevant for EURECA, shows the feasibility of the combination of different DM detector technologies and the use of sapphire as scintillating bolometer to act as a neutron flux monitor [542,543].
An anti-correlation between light and heat pulse amplitudes of a scintillating bolometer was reported for the first time for a sapphire-based device [538]. This observation was exploited for the analysis of the energy partition in an Al 2 O 3 scintillating bolometer, fabricated from a 50 g crystal [540]. It was found that the absolute light yield is 11%, while the same amount is trapped by the crystal and the rest of the energy release goes to heat [540]. However, a significantly lower L/H γ(β) value is measured due to a low light collection efficiency (around 12%) [540].

Bismuth Germanate
Bismuth germanate (Bi 4 Ge 3 O 12 , commonly abbreviated as BGO) is another classic scintillation material [374,375] with a long-standing interest for direct DM searches with scintillating bolometers. In particular, this material was considered a possible scintillating target for the CRESST [392] and used in the ROSEBUD [267,394,539,[544][545][546] experiments. The content of the heaviest "stable" element, 209 Bi (100% abundance in natural bismuth), with non-zero spin of the ground state provides a viable target for high-mass WIMP spin-independent and spin-dependent searches.
Excellent particle identification capability of BGO scintillating bolometers was reported for the first time almost two decades ago in studies with 46 and 91 g crystals [133,267,394]. In particular, L/H γ(β) = 7.5 keV/MeV, QF α = 0.17, and QF recoil ∼ 0.07 have been measured with a 46 g BGO bolometer ( 20 × 20 mm) coupled to a Ge LD ( 25 × 0.1 mm), providing an efficient separation down to ∼20 keV recoil energies [133,267,539]. A detailed study of QF α and QF recoil as a function of the deposited energy in that 46 g BGO scintillating bolometer is presented in [541].
As in the case of sapphire (Section 3.4.1), the light-heat anti-correlation of the 46 g BGO scintillating bolometer [544] allowed us to determine that the absolute light yield is 5.8%, while the rest of the deposited particle energy is released in heat or trapped with a similar fraction (46% vs. 48%) [540]. At the same time, due to a low light collection efficiency (estimated as 12%), the measured L/H γ(β) remains much lower (7.0 keV/MeV).
The considerable contamination of BGO crystals by 207 Bi [96,547], which might be mitigated by a careful selection of Bi-containing starting materials, is a limiting factor for DM search applications. In any case, a BGO scintillating bolometer (especially, with a reasonably low 207 Bi content) represents an interest for γ and/or α spectroscopy [544,548] and related nuclear and particle physics applications. In particular, the 207 Bi contamination of the 46 g BGO scintillating bolometer has been exploited for the first measurement of the L/K electron capture ratio of the 207 Bi decay to the 1633 keV level of 207 Pb [549]. Moreover, excellent performance and α/γ(β) separation of BGO scintillating bolometers allowed us to detect, for the first time, the α decay of 209 Bi [133], the rarest ever observed α decay. This discovery has been made with the above-mentioned 46 and 91 g BGO scintillating bolometers. Furthermore, about twice higher L/H γ(β) , 16.6 keV/MeV, measured with a large-volume (50 × 50 × 50 mm, 891 g) BGO scintillating bolometer working in coincidences with a SiO 2 -coated Ge LD ( 36 × 0.3 mm) was exploited to detect the 209 Bi α decay transition to the first excited state of 205 Tl [134]. It is worth noting that the LD energy resolution was measured as 3.5% FWHM at 2615 keV [134,548]. Last but not least, the operation of a 4-crystal array of 0.89-kg BGO bolometers coupled to a single Ge LD ( 50 × 0.3 mm) at LNGS has enabled competitive searches for axioelectric absorption of solar axions [550].
Taking into account good scintillation properties of BGO and well-established growth of large-volume crystals (particularly, using the LTG Cz method [551]), this material is considered in the BINGO project as a viable alternative to ZnWO 4 for the development of an active shield of a bolometric DBD search experiment [449]. A first low-temperature test of the BGO-based BINGO prototype has been recently realized at IJClab using a LTG-Cz-grown sample ( 30 × 60 mm) and two identical Ge LDs ( 44 × 0.2 mm; only one LD, coated with SiO, was operational). The data analysis shows an efficient pulse-shape discrimination between signals induced by BGO scintillation and muons passing through the LD [449]. The L/H γ(β) is measured as ∼28 keV/MeV [449], which is the highest value ever reported for BGO and is twice higher than that of the same-size ZnWO 4 crystal investigated in the same experimental conditions (see Section 3.1.6).

Lithium Aluminate
A lithium aluminate (LiAlO 2 ) crystal scintillator is another Li-containing material investigated recently using a scintillating bolometer approach and applied for the searches for spin-dependent DM interactions [323,552]. One detector module based on a small LiAlO 2 crystal (20 × 10 × 5 mm, 2.8 g, Cz growth) with an NTD readout was tested aboveground with a CRESST-III LD instrumented with a TES sensor. (A twin sample with a TES directly deposited on the crystal was operated too, but without LD.) Another module was investigated at LNGS. It was constructed from a large sample ( 50 × 70 mm, 373 g), cut from the same crystal boule, and equipped with both NTD and TES-on-CaWO 4 -basedcarrier sensors. A CRESST-II LD was also facing the top surface of the LiAlO 2 crystal. Both phonon-scintillation LTDs demonstrate good particle identification capability thanks to high-performance LDs and a reasonable light yield. In particular, the small LiAlO 2 scintillating bolometer is characterized with L/H γ(β) = 1.18 keV/MeV, QF α+t = 0.6 and QF recoil = 0.24. The quenching factor for α particles is 16% lower than that of α+t events, as found in the study of the large LiAlO 2 scintillating bolometer.

Tellurium Dioxide
Tellurium dioxide (TeO 2 ) is the most extensively developed and used material for DBD search experiments with pure thermal detectors. In particular, thanks to almost 30years-long development on TeO 2 LTDs [218], 988 large-volume detectors (50 × 50 × 50 mm, 750 g) are now operating at LNGS in the first tonne-scale bolometric experiment, CUORE [77][78][79][80]. A major drawback of this material is the negligibly low scintillation of pure or doped crystals either at room or low temperatures [267][268][269]553]. Despite that issue, particle identification with a TeO 2 crystal operated as a scintillating bolometer, first reported in [267,554], can be realized by the detection of the Cherenkov radiation induced by γ(β)'s, but not α's [270] (see Section 2.1). However, as one can see in Table 7, this task is extremely challenging because of the poor amount of the emitted light (L/H γ(β) ∼ 0.04 keV/MeV), which requires low-threshold LDs. Indeed, in order to reach about 99.9% rejection of the α background (i.e., DP α/γ(β) = 3.2) for a TeO 2 bolometer, the baseline noise resolution of an LD is required to be ∼20 eV RMS [555], for example, comparable with the noise resolution of the best performance NTD-instrumented LDs (the same phonon sensor technology of the CUORE experiment). A feasible optimization of the light collection could relax this constraint to only ∼30 eV RMS [271]. The state-of-the-art of the Cherenkovdominated signal detection and particle identification with TeO 2 crystals, operated using a scintillating bolometer approach, is reported in Table 7. Table 7. Results of low-temperature scintillation and Cherenkov radiation detection with scintillating bolometers based on natural/ 130 Te-enriched TeO 2 crystals and thin LDs. The mass (two significant digits) and the size of the sample, the size, the material (including coating), the sensor technology, and the baseline noise RMS of the coupled LD, as well as the L/H γ(β) value and the achieved DP α/γ(β) at the 130 Te 0νDBD ROI are listed.

Yttrium Orthovanadate
Yttrium orthovanadate (YVO 4 ), given its large mass fraction of vanadium (about 18%), is an attractive compound for the investigation of the 4-fold-forbidden β decay of 50 V, in particular using a bolometric technology [557]. The first characterization of the YVO 4 material using a scintillating bolometer technique has been realized with a Cz-grown undoped crystal ( 18 × 20 mm, 22 g) viewed by a Ge LD [557]. The YVO 4 scintillating bolometer shows an efficient scintillation at low temperatures: the measured L/H γ(β) is 59.4 keV/MeV. (An order of magnitude lower light signal is seen in Figure 5 [557] compared to Figure 4 and results presented ibid., seems to be due to the LD miscalibration. Indeed, YVO 4 has a rather high light output at room temperature (6300-11,000 photons/MeV [558,559] and a significantly increased scintillation efficiency with the temperature decrease (a factor 5 gained light yield at ∼140 K compared to one at room temperature [559]).) The scintillation light for α particles is quenched to 20% [557], thus providing an excellent particle identification capability. Thanks to the encouraging results of this bolometric test, an innovative approach based on triple coincidence between YVO 4 , TeO 2 and Ge LTDs has been proposed [557] to detect the rare β transition of 50 V to the first excited state of 50 Cr, improving the present knowledge on the 50 V decay scheme [560].

Zirconium Dioxide
Zirconium dioxide (ZrO 2 ) is potentially interesting to search for DBD of Zr isotopes [34,296,561,562], especially 96 Zr, an isotope with one of the largest Q ββ -value. An operation of a small ZrO 2 sample (0.14 g) as a scintillating bolometer has been realized about a decade ago [208,470]. The ZrO 2 scintillation has been detected with the help of a Ge LD ( 35 × 0.3 mm), allowing for a separation between γ(β) and α events. No ZrO 2 L/H γ(β) is reported, but a poor scintillation light output of the material together with Li 2 MoO 4 is concluded from the low-temperature tests [208,470]. Using a rough comparison of a light signal amplitude induced by similar energy deposition in the ZrO 2 detector and a neighbour scintillating bolometer (cesium iodide) operated in the same assembly [470], the ZrO 2 light yield is estimated as ∼2% of cesium iodide (or L/H γ(β) ∼ 2 keV/MeV, using results presented in Section 3.7.1) and the QF α parameter is evaluated as ∼0.2.

Lithium Indium Diselenide
Lithium indium diselenide (LiInSe 2 ) has been under development as a neutron detector material for several years [265] and it can potentially be interesting for the searches for 0νDBD of Se isotopes and/or solar neutrino detection. However, the presence of the β-active 115 In (T 1/2 = 4.4 × 10 14 yr) with almost 96% abundance in natural indium limits a suitable crystal mass of a bolometric detector to ∼10 g, in which the 115 In activity of around 1 Bq is the maximum acceptable in order to avoid serious pile-up issues. On the contrary, such cryogenic detector can be used to investigate the 4-fold-forbidden β decay of 115 In, particularly its spectral shape would help to disentangle a possible quenching of the axial-vector coupling constant and this information is important for theoretical predictions on 0νDBD rate [30]. With this aim, a small LiInSe 2 crystal (19 × 15 × 8 mm, 10.2 g, BS growth) was operated as a scintillating bolometer [265,563,564]. A Ge LD ( 44 × 0.2 mm) with the NTL signal amplification was used to detect the LiInSe 2 scintillation. The measured notable L/H γ(β) (14 keV/MeV [265]) does not require a high-performance LD, but it is important for the control of pile-ups and for the reconstruction of the spectrum at near threshold energy. The quenching of scintillation light was found to be 55% and 79%, respectively, for α particles of 210 Po and α+t events from thermal neutron captures by 6 Li [265]. Thus, LiInSe 2 is extremely promising as a scintillating bolometer for neutron spectroscopy because the products of the neutron capture can be detected in the region free of α/β/γ/µ-induced radiation.
The first investigation of ZnSe scintillating bolometers [347] reports comprehensively the features of the material at low temperatures, which were then confirmed and extended by further studies of natural and 82 Se-enriched ZnSe crystals. The findings related to the ZnSe scintillation detection can be summarized as follows: • A good L/H γ(β) of several keV/MeV and a large spread (up to a factor 10) among samples are observed (see Table 8). A large variation in the light yield was also reported in studies of ZnSe scintillation at low temperatures with a bolometric photodetector [262].
• Light and heat signals exhibit correlation [347,565,570,573] (in contrast to other efficient scintillation materials showing anti-correlation). Taking into account that this feature (and anti-correlation as well) deteriorates the energy resolution of a bolometer, a simultaneous detection of heat and scintillation signals is thus needed to improve the energy resolution of ZnSe-based scintillating bolometers [347,565,570,573]. • A scintillation signal induced by an α particle is few times higher than the one of γ(β)'s of the same energy. Therefore, the QF α parameter is larger than 1, and amounts to ∼3-5 (see Table 8), in contrast to other scintillation materials tested with a scintillating bolometer approach. • The QF α dependence on an α source position (e.g., about 20% difference for a source irradiating different crystal sides) is reported [347]. A small difference (around 12%) in the QF α is found between surface and bulk α events [565,573]. • The reduction of scintillation light is observed for some α events (similar to surface α events in a CdWO 4 scintillating bolometer, see Section 3.1.2), which are then leaking into the band of γ(β) events spoiling particle identification [347,565,569,573]. Moreover, the number of events in the tail of the α population depends on the crystal surface quality (optical or rough) facing the α source [347]. • Scintillation light signals induced by α's are faster than those of γ(β)'s, thus a highly efficient particle identification can be done using a pulse-shape analysis of signals acquired by a bolometric LD [312,347,565,568,569,571,573,574]. The particle-dependent difference in the pulse-shape also affects the QF α estimate. Indeed, if one uses the pulse area (instead of the signal maximum amplitude obtained with the optimal filtering technique) as an energy estimate, the QF α value is reduced by a factor 1.5, but still higher than 1 [347]. • A thermal treatment of ZnSe crystals (under an argon atmosphere) improves their homogeneity and reduces microcracks [575,576]. However, the pulse-shape, the signal amplitude, and the light yield of scintillating bolometers based on the annealed ZnSe crystals are found to be deteriorated [575]. It is interesting to note that the QF α parameter is found to be less than 1 [575]. These observations suggest changes of the annealed crystal defect structure (affecting the bolometric performance of the material) and reductions of the luminescence centers or changes their nature (affecting the light yield and the kinetics of luminescence) [575]. A subsequent characterization shows that the donor-acceptor pairs consisting of Zn vacancies and Al donors, present in as-grown material with concentrations of the order of ppm and responsible for the good scintillation properties, are lost during the ZnSe thermal treatment [576]. Therefore, more studies are needed (and some are proposed in [576]) to understand the influence of the defects present in ZnSe aiming at the optimization of ZnSe-based scintillating bolometers.
Despite the complexity of the ZnSe growth process (which affects the crystal quality and radiopurity) and the features of ZnSe scintillating bolometers (which affect the detector performance), the results of the LUCIFER project and the CUPID-0 experiment proof good prospects of this material for high-sensitivity searches for the 82 Se 0νDBD. In particular, thanks to the scintillation-based rejection of α events, CUPID-0 has achieved the lowest background in the 0νDBD ROI ever reported for a bolometric experiment [571] and has set the most stringent limit on the 82 Se 0νDBD half-life [157,568]. It is worth noting that the 82 Se 2νDBD has been measured with the best precision ever reported for such process [169]. Therefore, ZnSe scintillators remain promising for a large-scale bolometric DBD search experiment, particularly exploiting a multi-target array of scintillating bolometers [419]. n/a n/a ibid.
a The same sample used in [565,573], but after the thermal treatment. b A median value of 12 ZnSe scintillating bolometers including four crystals studied in [565,566].
3.6. Alkali Metal Fluorides 3.6.1. Calcium Fluoride Calcium fluoride (CaF 2 ), particularly Eu-doped, is a well-known scintillation material [374,375] with a long history of applications to the searches for DBD of calcium isotopes (see, e.g., [577][578][579][580][581][582][583] and references therein). This material is especially interesting for the search for 0νDBD of 48 Ca (the nuclide with the largest transition energy Q ββ ), particularly with CaF 2 scintillating bolometers [287]. There is also remarkable interest in CaF 2 as a target for DM searches (in particular spin-dependent interactions with 19 F) [291,579,584,585]. A Eu-doped CaF 2 can also be used for the investigation of α decay of natural europium [526] (the first observation of the α decay of 151 Eu was done using a CaF 2 room-temperature scintillation detector [526]).
As mentioned in Section 2.4, CaF 2 (Eu) crystal scintillators were used in first composite bolometers, developed in the 1990's, with a simultaneous detection of phonons (by a thermistor) and photons (with either a photodiode [287][288][289] or a bolometric LD [290,291]). These studies show that the Eu (paramagnetic element) doping at the level of 0.01% to 0.07% does not affect the bolometric performance of a CaF 2 (Eu) thermal detector [287] and that a 0.03% doping concentration maximizes the scintillation yield [289]. The operation of gram-scale CaF 2 (Eu) scintillating bolometers (based on 2 g [287-289] and 0.3 g [290,291]) crystals) demonstrates the discrimination of α particles thanks to the heat-scintillation dual readout. The QF α of 0.14(1) for ∼5-6-MeV α's was reported in [291]. A similar value of the quenching (0.18) was obtained in bolometric measurements of the scintillation emitted by a 20 g CaF 2 (Eu) crystal under α and γ irradiation [267]; the L/H γ(β) value was measured to be ∼14 keV/MeV. In spite of such good scintillation properties at low temperatures, no R&D on CaF 2 scintillating bolometers have been realized until recently. It can be explained by the fact that more appealing materials (e.g., CaWO 4 ) were investigated and used for DM search applications. The very low isotopic abundance of 48 Ca (0.2% [456]) combined with the not-yet-possible industrial enrichment in this isotope [35] is a limiting factor for CaF 2 -based DBD search experiments.
The interest in developments of pure / Eu-doped CaF 2 scintillating bolometers has been renewed recently [586][587][588], in particular within the CANDLES project [217]. Two scintillating bolometers based on CaF 2 and CaF 2 (0.17% Eu) crystals (312 g, 50 × 50 mm each) and 2 in. Ge wafers, both with MMC phonon readout as in the AMoRE experiment, have demonstrated an efficient particle identification [586,587]. A significantly higher α/γ(β) separation has been achieved for the Eu-doped detector [587], thanks to the enhanced light output of the doped sample. A QF α = 0.17 is found for ∼5-MeV α events, in agreement with early investigations. The R&D is ongoing to improve both the energy resolution of CaF 2 (Eu)-based LTDs (possibly affected by the spin-lattice interaction of paramagnetic Eu ions) and to understand the large distribution of the α-induced scintillation measured with the undoped CaF 2 scintillating bolometer [587,588].

Lithium Fluoride
Lithium fluoride (LiF) is among the pioneering materials investigated extensively as thermal detectors for DM searches (e.g., see [589][590][591]) and remains attractive for such applications [531]. Indeed, the content of only light nuclei in LiF provides sensitivity to low-mass WIMP. 7 Li and 19 F are viable targets for the investigation of spin-dependent interactions. Moreover, a large content in 7 Li makes LiF suitable for the searches for resonant absorption of solar axions [502,503]. Furthermore, the presence of 6 Li in a Licontaining bolometer can be exploited for neutron spectroscopy, as demonstrated for the first time with LiF LTDs [589,591].
A first operation of a LiF-based scintillating bolometer (16 g crystal coupled to 25-mm Ge LD) was realized about 15 year ago [267,554] and this study reports a low scintillation of the material (L/H γ(β) = 0.38 keV/MeV). The scintillation light for α particles and nuclear recoils was found to be quenched to 29% and 15%, respectively. Events induced by the thermal and fast neutron captures by 6 Li were clearly separated form γ/β/µ events, but not from α's (despite of ∼20% more light detected for α+t events of the same energy release as α's). Similar results were achieved with a twice massive LiF scintillating bolometer operated aboveground [592] and underground [545]. Since such low light yield did not provide particle discrimination at low thresholds, this material was considered only for neutron detection in further studies with natural and 6 Li-enriched LiF scintillating bolometers [59,60,254,255,257,258,371]. In particular, a scintillating bolometer prototype based on a 32 g 6 LiF crystal ( 25 × 25 mm, 95% enrichment in 6 Li) has been operated successfully [59,371]. The measured L/H γ(β) is twice lower (0.21 keV/MeV [371]) than that of the 16 g LiF detector reported in [267], but α particle identification above 2 MeV was clearly demonstrated with the 6 LiF scintillating bolometer.

Strontium Fluoride
Strontium fluoride (SrF 2 ) can potentially be used in searches for DBD of Sr isotopes and DM particles [371]. A 54 g SrF 2 scintillation crystal ( 25 × 25 mm) and a Ge wafer with the same diameter were coupled together in a scintillating bolometer module tested aboveground [371,593]. An excellent α/γ(β) separation has been achieved, in particular thanks to a good scintillation (L/H γ(β) = 2.9 keV/MeV) and quenching to 26% and 10% for α particles and nuclear recoils, respectively [371]. The crystal was found to be contaminated by 226 Ra (0.5 Bq/kg), 210 Po (0.07 Bq/kg) and 228 Th (0.02 Bq/kg) [371]; it was used to study the light output in fast subsequent α decays. In particular, a positive correlation between the light yield of 224 Ra and 220 Rn α emitters indicates a light output position dependence in SrF 2 [371,593]. It is also worth noting the particle identification capability of SrF 2 by using either heat or scintillation signals [593].

Cesium Iodide
Cesium iodide (CsI) is among the oldest discovered and widely used inorganic scintillators [374,375]. This material, doped with Tl to enhance light output, is actively used in scintillation detectors for DM searches [594][595][596][597][598][599][600][601][602][603]. Undoped CsI is very attractive for bolometric searches for DM signals [415,604,605], particularly for the detection of the annual modulation of the DM rate [606]. Indeed, a low light yield of CsI at room temperature (a factor ∼20 lower than the light yield of Tl-doped cesium and sodium iodides [607]) can be about an order of magnitude higher at 3.4-10 K [264,608]. At the same time, a weak hygroscopicity and a low hardness of CsI has to be taken into account for bolometric operation. Moreover, a low Debye temperature of the material and thus a high specific heat expected at low temperatures is the main issue which would affect the bolometric performance (in particular, it would result in a low signal amplitude affecting the detector energy resolution and threshold).
A large scintillation signal together with an efficient particle identification of a CsI scintillating bolometer, made of a small cubic crystal (10 × 10 × 10 mm) and a Ge LD ( 35 × 0.3 mm), has been reported about a decade ago [470]. More detailed study of the material as scintillating LTD has been then realized using the CRESST detection technology [415]. The L/H γ(β) of a small (20 × 10 × 5 mm) and a CRESST-size ( 40 × 40 mm) CsI crystals measured with SOS LDs ( 40 × 0.4 mm each) was found to be enormous: 71 and 49 keV/MeV, respectively. On the contrary, a small phonon signal, measured by a TES on a small sapphire carrier glued onto the large CsI, was the reason of an order of magnitude worse energy resolution than that of the same-size CaWO 4 bolometer. The large difference in the light yield might be explained by the difference in the samples' shape and size (i.e., difference in the internal trapping of scintillation light).
These observations were also confirmed by further low-temperature tests of two CsI crystals of the same size (30 × 30 × 30 mm, 122 g), but from different producers [604]. The phonon readout was done using a small CdWO 4 -based carrier with a W TES evaporated on it. The scintillation was detected with the help of a SOS LD ( 40 × 0.46 mm) with the same TES-based readout. The L/H γ(β) of the detector modules was measured as 81 and 65 keV/MeV. It was also observed that the light yield of both CsI crystals at 10 keV is around 10% higher than at 120 keV. The results of the detection of nuclear recoils (from an α source) in the DM search ROI were found to be unclear, requiring a dedicated study with a neutron source, while no α region is shown in [604]. Concerning particle identification, we can quote QF α ∼ 0.5, which was measured at 3.4-10 K with an undoped CsI crystal viewed by two photomultipliers and using an optical cryostat with an experimental space encapsulated in a glove box [608].

Sodium Iodide
Sodium iodide, doped with Tl to improve light output, is a "standard" in the world of inorganic scintillators and the most extensively used material in scintillation detectors [374,375]. Both pure and Tl-doped NaI exhibit a huge light yield at low temperatures (at ∼2 K), which amounts to about 65% of the light output of NaI(Tl) at room temperature [263]. Therefore, a NaI-based LTD with particle identification is of a great interest for DM search experiments [585,605,606,609], in particular aiming at the investigation of the nature of the galactic DM signature detected in the DAMA/LIBRA and DAMA/NaI experiments with room-temperature NaI(Tl) scintillation detectors (e.g., see [424][425][426][427] and references therein). However, in addition to a low Debye temperature (as in the case of CsI, Section 3.7.1), a high hygroscopicity of NaI is a challenge for bolometric applications of this material.
In order to mitigate the hygroscopicity issue and therefore to simplify the mounting and operation of a NaI-based LTD, a vapor-deposited parylene (poly-p-xylylene) coating of NaI(Tl) and NaI crystals has been demonstrated to be a viable solution [610].
The development of another technology of NaI scintillating bolometers is now ongoing withing the COSINUS project [241,611]. The first prototype [612] was fabricated using a 66 g undoped NaI (30 × 30 × 20 mm) attached to a CdWO 4 -based carrier ( 39 × 1.6 mm), equipped with a TES W sensor. The LD was made of a SOS wafer ( 40 × 0.5 mm) with a W TES deposited on it. A reflective film without scintillating properties (LuMirror TM ) was placed around the NaI crystal. The detector module was housed in an air-tight copper container, which was then evacuated using a dedicated cryogenic valve opening during the cool-down. The L/H γ(β) of this prototype was measured to be 37 keV/MeV [612], close to the final COSINUS design goal of 40 keV/MeV [611].
The second prototype [294] was made of the same-size undoped NaI, while a Si beaker ( 40 × 38 mm, the wall thickness is about 0.6 mm), foreseen for the final detector design, was used as LD and active shield. Thanks to significantly improved light collection provided by such a photodetector (see also Section 3.1.1), a huge L/H γ(β) of 131 keV/MeV, never reported for scintillating bolometers, has been measured [294,295].

Conclusions
Active developments of scintillating bolometers have been ongoing over the past three decades aiming at the implementation of low-temperature particle detectors for rareevent searches (as rare alpha and beta decays, double-beta decay, dark matter, neutrino detection) and/or for a background control in these experiments (e.g., neutron detection). An important feature of such devices is the detection of the scintillation light emitted by a cryogenic scintillator that allows an event-by-event particle identification, a crucial tool for background rejection. In most cases, the scintillation of low-temperature devices is detected using an auxiliary (thin) bolometer.
This review reports on scintillation of low-temperature detectors based on different scintillation materials (used as particle energy absorbers). Among them we find tungstates (CaWO 4 3 ), and some other oxide scintillators (Al 2 O 3 , Bi 4 Ge 3 O 12 , LiAlO 2 , TeO 2 , YVO 4 , ZrO 2 ) together with selenides (LiInSe 2 , ZnSe) and alkali halides (CaF 2 , LiF, SrF 2 , CsI, NaI). Some of the considered scintillators have been also produced from starting materials enriched or depleted in isotopes of interest for rare-event searches.
A particular interest of this review is devoted to the measurements of particledependent scintillation of materials, represented by the fraction of the energy detected by a photodetector to the energy measured as a heat of a cryogenic scintillator (lightto-heat ratio). The amount of scintillation light measured by a photodetector plays an important role for particle identification capability of a scintillating bolometer and defines the performance requirements of the light detector. Taking into account that most of the scintillating bolometer tests follow "universal" detector design (i.e., a scintillator inside a reflective cavity is viewed by a bolometric photodetector), the scintillation efficiency of the material largely contributes to the observed difference in the light-to-heat values, ∼0.1-100 keV/MeV, of the investigated compounds. At the same time, an optimization of the light output and light collection efficiency can also be a viable way to enhance the scintillation signal measured by the light detector.
Thanks to the extensive R&D on scintillating bolometers realized over the last 30 years, some of the reported materials have already been used in frontier searches for rare-event processes. Moreover, new experiments, in particular with a tonne-scale array of scintillating bolometers, are under development. Furthermore, known scintillation materials of advanced quality and new ones are still appealing for a wide field of applications of cryogenic scintillators, as well as for a possible improvement and/or extension of the physics goals achievable with the presently developed scintillating bolometers. Therefore, investigations of low-temperature detectors based on new or advanced scintillation materials are greatly encouraged.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Acknowledgments:
The author is grateful to Andrea Giuliani, Pierre de Marcillac and anonymous Reviewers for their careful reading of this manuscript and their many insightful comments and suggestions.

Conflicts of Interest:
The author declares no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript: