CdWO4 scintillating bolometer for Double Beta Decay: Light and Heat anticorrelation, light yield and quenching factors
C. Arnaboldi, J.W. Beeman, O. Cremonesi, L. Gironi, M. Pavan, G. Pessina, S. Pirro, E. Previtali
CCdWO scintillating bolometer for Double Beta Decay: Light and Heat anticorrelation,light yield and quenching factors C. Arnaboldi a , J.W. Beeman c , O. Cremonesi a , L. Gironi a,b , M. Pavan a,b , G. Pessina a , S. Pirro a, ∗ , E. Previtali a a INFN - Sezione di Milano Bicocca I 20126 Milano - Italy b Dipartimento di Fisica - Universit`a di Milano Bicocca I 20126 Milano - Italy c Lawrence Berkeley National Laboratory , Berkeley, California 94720, USA
Abstract
We report the performances of a 0.51 kg CdWO scintillating bolometer to be used for future Double Beta Decay Experiments.The simultaneous read-out of the heat and the scintillation light allows to discriminate between di ff erent interacting particles aimingat the disentanglement and the reduction of background contribution, key issue for next generation experiments. We will describethe observed anticorrelation between the heat and the light signal and we will show how this feature can be used in order to increasethe energy resolution of the bolometer over the entire energy spectrum, improving up to a factor 2.6 on the 2615 keV line of Tl.The detector was tested in a 433 h background measurement that permitted to estimate extremely low internal trace contaminationsof
Th and
U. The light yield of γ/β , α ’s and neutrons is presented. Furthermore we developed a method in order to correctlyevaluate the absolute thermal quenching factor of α -particles in scintillating bolometers. Keywords:
Double Beta Decay, Bolometers, CdWO , Quenching Factor PACS:
1. Introduction
Double Beta Decay (DBD) searches became of critical im-portance after the discovery of the neutrino oscillations andplenty of experiments are now in the construction phase andmany others are in R&D phase [1, 2, 3, 4]. The main challengesfor all the di ff erent experimental techniques are the same [5]: i)increase the active mass, ii) decrease the background, and iii)increase the energy resolution.Thermal bolometers are ideal detectors for this survey: crys-tals can be grown with di ff erent interesting DBD-emitters and,fundamental for next generation experiments, they show an ex-cellent energy resolution.The CUORICINO Experiment [6], constituted by an array of62 TeO crystal bolometers, demonstrated not only the powerof this technique but also that the main source of backgroundfor these detectors arises from surface contaminations of ra-dioactive α -emitters. Moreover simulations show that this con-tribution will largely dominate the expected background of theCUORE Experiment [7, 8] in the region of interest, since thereis no possibility to separate this background from the two DBDelectrons. The natural way to discriminate this background is touse a scintillating bolometer [9]. In such a device the simulta-neous and independent read out of the heat and the scintillationlight permits to discriminate events due to γ / β , α and neutronsthanks to their di ff erent scintillation yield. Moreover if the crys-tal is based on a DBD emitter whose transition energy exceeds ∗ Corresponding author
Email address:
[email protected] (S. Pirro ) the 2615 keV γ -line of Tl then the environmental backgrounddue to natural γ ’s will decrease abruptly.CdWO is an ideal candidate for such kind of detector: • it is a well established scintillator • Cd (7.5 % i.a) has a DBD transition at 2805 keV • the light yield (LY) is rather large • the radiopurity of this compound is “naturally” highDue to these favourable features this crystal compound wasalready used to perform a DBD experiment [10] using “stan-dard” Photomultipliers. A large mass experiment, based onenriched CdWO crystals, readout by Photomultipliers, wasalso proposed [11]. But this technique, limited by the modestachievable energy resolution, key point of future experiments,is no more pursued.
2. Experimental details
The dimensions of the CdWO crystal tested are 4 cm diam-eter, 5 cm height. All the surfaces of the crystal are polished atoptical grade. The crystal was tested as a standard scintillator atroom temperature. It was wrapped with a reflecting sheet (3MRadiant Mirror film VM2000) and coupled with optical greaseto a Photomultiplier (Hamamatzu R6233). The energy resolu-tion evaluated on the Cs line is 12.3 % FWHM.
Preprint submitted to Atroparticle Physics October 24, 2018 a r X i v : . [ nu c l - e x ] M a y hermistor HeaterPTFEGe LD CdWO ReflectorPTFE
Figure 1: Setup of the detector. The two Cu columns, holding the two frames,are not visible from the chosen perspective.
Our detector setup is schematized in Fig. 1. It is held bymeans of four L-shaped Teflon pieces fixed to the two cylindri-cal Cu frames. The frames are held together through two Cucolumns. The crystal is surrounded (without being in thermalcontact) by a 43 mm diameter reflecting sheet (3M VM2002).The Light Detector (LD) [12] is constituted by a 36 mm diam-eter, 1 mm thick pure Ge crystal absorber. The surface of theGe facing the crystal is further “darkened” through the depo-sition of a 600 Å layer of SiO in order to increase the lightabortion. Furthermore since the diameter of the LD is smallerwith respect to the one of the reflecting cavity, a “reflectingring” has been mounted in order to decrease light losses. Onthe opposite face of the crystal a reflecting sheet is mounted.The temperature sensor of the CdWO crystal is a 3x3x1 mm neutron transmutation doped Germanium, the same used in theCUORICINO experiment. The temperature sensor of the LDhas smaller volume (3x1.5x0.4 mm ) in order to decrease itsheat capacity, increasing therefore its thermal signal. A resis-tor of ∼
300 k Ω , realized with a heavily doped meander on a3.5 mm silicon chip, is attached to each absorber and acts asa heater to stabilize the gain of the bolometer [13, 14]. Thedetectors were operated deep underground in the Gran SassoNational Laboratories in the CUORE R&D test cryostat. Thedetails of the electronics and the cryogenic facility can be foundelsewhere [15, 16, 17].The heat and light pulses, produced by a particle interactingin the CdWO crystal and transduced in a voltage pulse by theNTD thermistors, are amplified and fed into a 16 bit NI 6225USB ADC unit. The entire waveform of each triggered voltagepulse is sampled and acquired. The time window has a width W a
600 1000 140010204050 1800 2200 260030
Energy (Heat) [keV] L i gh t [ k e V ] Figure 2: Scatter plot Light vs. Heat obtained in a 96 h calibration using anexternal
Th source. In the inset the highlight of the
Tl line. In the circletwo events due to the internal α -decay of W. The energy spectrum below 400keV is completely dominated by the Cd β -decay, with a rate close to 0.4 Hz. of 256 ms sampled with 512 points. The trigger of the CdWO is software generated while the LD is automatically acquired incoincidence with the former. The amplitude and the shape ofthe voltage pulse is then determined by the o ff line analysis thatmakes use of the Optimal Filter technique. The energy calibra-tion of the CdWO crystal is performed using γ sources placedoutside the cryostat. The Heat axis is calibrated attributing toeach identified γ peak the nominal energy of the line, as if allthe energy is converted into heat. Consequently this calibrationdoes not provide an absolute evaluation of the heat deposited inthe crystal. The dependency of amplitude from energy is pa-rameterized with a second order polynomial in log(V) where Vis the heat pulse amplitude. The three coe ffi cients of the poly-nomial are fitted on calibration data (the heat pulse amplitudecorresponding to the known γ lines visible in the spectrum).The choice of such a function was established by means of sim-ulation studies based on a thermal model of the detectors.The energy calibration of the LD is obtained thanks to a weak Fe source placed close to the Ge that illuminates homoge-neously the face opposed to the CdWO crystal. During theLD calibration its trigger is set independent from the one of theCdWO . The LD is calibrated using a simple linear function.The FWHM energy resolution of the LD, evaluated on the Xdoublet at 5.90 and 6.49 keV, is 480 eV FWHM. In the rangeof 0 ÷
50 keV, i.e. the energy interval of the light signals, theresponse of the LD can be definitely assumed to be linear.Three sets of data have been collected with this device: twocalibrations using
Th and K sources, a long backgroundmeasurement (433 h) and a neutron measurement (8 h) doneexposing the detector to an Am-Be source.
3. Light-Heat scatter plot
In this field the usual way to present the results is to drawthe Light vs. Heat scatter plot. Here each event is identified2y a point with abscissa equal to the heat signal (recorded bythe CdWO bolometer), and ordinate equal to the light signal(contemporary recorded by the LD). In the scatter plot, γ/β , α and neutrons give rise to separate bands, in virtue of their char-acteristic Light to Heat ratio (see Figs. 5 and 6). This featureis the result of the di ff erent LY’s characterizing these particles: γ/β events belong to a distribution (see Fig. 3) characterized bythe θ βγ angle, while the α events are characterized by θ α .The scintillation Quenching Factor (QF) is defined as the ra-tio of the scintillating yield of an interacting particle ( α , neu-tron, nucleus) with respect to the LY of a γ / β event at the sameenergy. From Fig. 3 we have QF α = tan( θ α ) / tan( θ βγ ). But thisholds only in first approximation, as will be exposed in Sec. 5and Sec. 6.Within each band, monochromatic events (i.e. those corre-sponding to the complete absorption of a monochromatic parti-cle in the CdWO crystal) appear as sections of lines with neg-ative slope, showing a strong anticorrelation between Heat andLight. This is evident in the Light vs. Heat scatter plot ob-tained with a source of Th (Fig. 2) where di ff erent γ -linesare clearly visible. Despite the much lower statistics, the samefeature is perceptible also for the α lines that appear in the back-ground spectrum (Fig. 7).A simple model accounts for the observed pattern. Theenergy E of a monochromatic particle, fully absorbed in theCdWO crystal, is divided into two channels: a fraction ( L ) isspent in the production of Light (photons) and a fraction ( H )is spent in the production of Heat (phonons). In absence of“blind” channels (i.e. channels in which the energy depositedinto the crystal is stored in some system that do not takes partin signal formation) the energy conservation requires: E = E Heat + E Light = (1 − k ) E + kE = H + L (1)being k the fraction of the total energy that escapes the crystal inform of light. The last equation states not only the energy con-servation, but also a subtle distinction between the Heat (whatis measured) and the energy definition.Monochromatic events should produce a spot in the scatterplot, with a size determined by the intrinsic energy resolutionof the heat and light detectors. The observed spread along anegative slope line, however, can be accounted for assumingthe existence of fluctuations in the H and L signal amplitudes.Since the energy has to be conserved the amplitude of the fluc-tuations ( δ L and δ H) must compensate each other: δ L = - δ H. Inother words the fluctuations are anti-correlated between eachother and produce the observed pattern. They are visible be-cause their amplitude is much larger than the intrinsic resolu-tion of the detectors.We can devise two mechanisms that introduce a fluctuationin the L / H ratio: i) the statistical Poissonian fluctuation in theemitted light: the more the energy spent in light production, theless the energy spent in heat; ii) any variation of the LY as thosedue to position dependent e ff ects (inhomogeneities and defectsof the crystals that modify the light emission or self absorptionof scintillation photons). Regardless the mechanism, we willdemonstrate, in the next section, how this anticorrelation can H ea t θ θ βγ θ α γ γ α γ α α γ γ α α L=0 α1 θ fit θ L i gh t Heat
Iso−energy Line γ1 L=0 =E( )E( )H( )=H( )=E( ) E( ) =E( )H H
Figure 3: Light vs. Heat scheme for the interpretation of the energy correctionmethod. The blue (dark) spots represent the γ/β monochromatic events, whilethe red (light) represents an α . be used in order to increase (correct) the energy resolution ofthe detector over the entire energy spectrum.Fig. 3 tries to summarize the model here discussed. Thedark spots represent monochromatic energy deposition in theCdWO crystal, as those observed during a γ calibration. Foreach of them a negative slope (iso-energy) line is tracked. Theintercept on the Heat axis (H Light = ) is the Heat that would cor-respond to a full heat conversion of the deposited energy (inabsence of light emission or in the case in which all the light isabsorbed by the crystal itself). In this simple model γ/β ’s and α ’s releasing the same energy within the CdWO crystal lie onthe same iso-energy line. But due to the larger LY, γ/β eventswill convert less heat within the crystal with respect to an α particle releasing the same energy: part of the energy escapesthe crystal in form of photons. This feature can be easily de-duced from Fig. 3: the α particle release the same total energyof the γ particle (they belong to the same iso-energy line) butthey show a di ff erent position on the Heat axis. In fact the α shows the same heat of the γ particle that releases a larger en-ergy within (but not “into”) the crystal. The “usual” picture (inwhich γ/β and α of the same energy show the same heat signal)corresponds to the limit in which the light emitted is negligi-ble (or completely re-absorbed by the crystal): in this case thedistinction between H and E becomes meaningless.It is important to remark that up to now it was assumed tobe able to measure the absolute values of energy converted intoheat and light.Actually, the measured experimental heat and light signalscan be written as H = α (1 − k ) EL = β(cid:15) kE (2)being α and β the absolute calibration factors for the Heat andLight axis. The factor (cid:15) , instead, takes into account the overalllight collection e ffi ciency.3nder the condition that α and β are not depending on theenergy, and considering that the conservation of energy requiresthat δ [ kE ] = - δ [(1 − k ) E ], Eq. 2 implies that δ L δ H = − β(cid:15)α = − tan ( θ f it ) (3)The last equation states that the e ff ect of miss-calibrations ofthe Heat and Light axis simply implies a variation of the slopeof monochromatic lines, which is exactly what we observe ex-perimentally. Also the light collection e ffi ciency plays a funda-mental role in the evaluation of the slope, as discussed in thenext section.It is important to note that, unlike the Heat channel, the LDenergy calibration is more delicate. The energy calibration onthis detector is performed using ionizing X-Rays from the Fesource.
A priori the thermal signal that arises by the absorptionof 1 keV of photons at the boundaries of the Ge crystal (more-over covered by the SiO layer) could give a di ff erent thermalsignal with respect to a 1 keV ionizing energy. In any case thise ff ect still preserves proportionality, so that it can be consideredwithin the factor β .
4. Light-Heat anticorrelation: energy resolution
The energy anticorrelation between the two signals ofa double readout system was already observed in ioniza-tion / scintillation detectors [18] as well as in heat / scintillationbolometers [19] with a device very similar to the one presentedhere. In the former case this anticorrelation was demonstratedto improve the energy resolution by a factor close to ∼
25 %(evaluated @570 keV). In paper [19] two di ff erent γ -lines werestudied at rather low energy. An evident anticorrelation wasfound but, however, the correlation factor of the two lines wasfound to strongly depend on the energy.The energy correction procedure is graphically explained inFig. 3. Here blue (dark) and red (light) spots represent γ/β and α produced by monochromatic events. In order to obtain aspectrum with improved energy resolution we have to combinethe heat and light values of each single event in an appropriateway. This can be obtained in a natural way performing a simpleSU(1) rotation in the scatter plane an projecting the points onthe Heat θ axis:Heat θ = Heat cos θ + Light sin θ Ligh θ = − Heat sin θ + Light cos θ (4)The value of θ can be evaluated in two di ff erent ways: i) itcan be “optimized” in order to obtain the best energy resolu-tion on the Heat θ axis (energy minimization); ii) by fitting themonochromatic spots in the scatter plot with a negative slopeline, evaluating the θ f it angle for each single distribution (forconstruction we have θ = π/ − θ f it )It is clear from this scheme that some assumptions should besatisfied in order for this procedure to work properly: • the slope of each monochromatic spot (i.e. the θ f it angle)or the θ angle minimizing the FWHM has to be the samewithin all the lines; F W H M E n e r gy R e s o l u ti on [ k e V ] Rotation angle q [deg]
583 keV 4858 keV 88 keV
Figure 4: FWHM energy resolutions evaluated for di ff erent θ angles. For eachangle a spectrum is produced and the corresponding peaks are fitted. The curvesthrough the data points are only to guide eyes. For two peaks the correspondingerror on the FWHM is not plotted for better clarity. • the LY of the di ff erent class of interacting particles ( α ’sand γ/β ) has to be independent from energy.The first condition ensures that the rotation minimizes theFWHM on the whole energy spectrum. It has to be noted that ifthe first condition is not verified, this technique still provides auseful tool for resolution optimization. In case of di ff erent θ f it values the rotation angle will be chosen in order to reach thebest performances in the region of more interest for the physics(e.g. the region where the DBD peak should appear).The second condition ensures that the projected spectrumkeeps linearity and energy calibration. This point is less triv-ial and will be treated in more detail in Sec. 5.We applied both the methods in order to evaluate the rota-tion angle. Using the energy minimization method, the value ofthe rotation θ angle is allowed to vary over a wide range (0-80degrees) and for each angle a spectrum is produced projectingthe data on the Heat θ axis. Finally the FWHM energy resolu-tion of the more intense peaks is evaluated by a fit (we use anasymmetric Gaussian reporting then the average FWHM).We applied this method not only to the calibration spectrumbut also to the background measurement. In this way we wereable to study also the the 88 keV γ -line of Cd (accidentallycontamination of our CdWO sample) and the α peaks due tointernal trace contaminations: in this latter case we observea line at the full energy of the decay (Q = α + nuclear recoil).The internal α -lines observed are U- α (Q = W- α (Q = U- α (Q = θ and projecting on Heat θ . The energy resolutions of the peaksare evaluated for each single rotated spectrum. The minimumof the energy resolution occurs within θ = (67.5 ± Cd.In fact the minimum of the energy resolution for this line,1.03 ± θ =
60 deg while at 67.5deg it becomes 1.11 ± ff erent behaviour with respect to all theother lines. We do not have an explanation for this enhancementin the anticorrelation at low energies, but the same e ff ect wasalready observed in [19].Each monochromatic events distribution in the scatter plotwas also fitted with a linear function, as previously discussed.The θ f it angles as well as the most relevant parameters of thisanalysis are summarized in Tab. 1. The values of the θ f it anglesfor the α ’s and γ / β ’s lines reported here are compatible withinthe experimental errors within the two groups (with the excep-tion of the 88 keV line). However, with respect to the γ / β ’s,the angular coe ffi cient calculated by fitting linearly the singlelines gives a result that is systematically slightly larger with re-spect to the value (67.5 ± ff erence could arise fromthe fact that while the bidimensional fit is model independent,the energy minimization depends on the fitting procedure of thepeaks. Moreover it should be pointed out that the quoted erroron θ f it don’t include systematic e ff ects (introduced, for exam-ple, by the continuum background) whose values could domi-nate the total error.The values of the θ f it angles reported in Tab. 1 seems to showa slightly di ff erence between α and γ/β . However it has tobe pointed out that the quoted error on the θ f it for the α par-ticles has a (further) systematic error that cannot be easily eval-uated. The point is that the α events were collected during thelong background measurement. During this period the CdWO crystal was slowly cooling down (the LD, on the contrary, wasrather stable during the same period). This drift, as explainedin Sec. 2, is usually corrected through the Heater pulse. If thetemperature drift is rather small ( (cid:46) µ K) then the correctionis independent from the energy. On the contrary, if the driftis larger (in this case the drift was ≈ α peakwas “self-stabilized”. This can introduce an uncontrolled sys-tematics since we are somehow forcing a distribution to obtaina minimum. On the other hand, a check was made on the weak2615 keV γ -line present in the background spectrum. As in thecase of the α ’s, we had to self-correct the drift. The obtained θ f it value is 71.5 ± α background.For the following we will assume θ γ/β f it = θ α f it . The Th and K calibrations, as well as the 88 keV analysis, on the otherhand, were made at the end of the measurement in a rather sta-ble working temperature.As a final remark it has to be noticed that the projection ofthe events on the Heat θ axis introduces an Heat (energy) nor-malization factor K that will be di ff erent for di ff erent classes of particles, namely K γβ,α = cos θ (1 + tan θ tan θ γβ,α ). This feature israther evident in Fig. 3: in the Heat axis we have H( α = H( γ θ axis we have H( α = H( γ α and γ/β is di ff erent within the Heat spectrum andthe Heat θ spectrum. The second is that the same considerationholds within the same class of events: if the LY of α particles(for example) depends slightly on energy, then the linearity ofthe projected spectrum will change.
5. Light Yield and Scintillation Quenching Factor
In Tab. 1 we also present the values of the LY, determinedon the di ff erent monochromatic lines. We define the LY as theenergy released in the LD (in keV) for a nominal energy depo-sition of 1 MeV in the scintillating crystal. Table 1: Table with the main parameters of the detector. The theoretical FWHMenergy resolution of the CdWO crystal, evaluated through Optimal Filtering,is 0.8 keV. The LY of the α -particles (last three lines) is evaluated for the overall Q-value of the decay. The correct value (assuming negligible scintillation forthe nucleus recoil) is ≈
2% larger. † See text.
Line FWHM θ = FWHM θ = . LY π/ − θ f it [keV] [keV] [keV] [keV / MeV] [deg]88 1.4 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± † ± † ± ± † ± † ± † ± ± † ± † ± † ± ± † The table shows that the LY of the γ/β events is, within theerror, the same for all the observed lines (including the 88 keVline that shows a di ff erent resolution curve in Fig. 4). On theother hand a small, but rather clear, energy dependence is evi-dent for α particles. This e ff ect can also be observed in Fig. 6 inwhich it is evident that the α particles belong to a curve ratherthan a straight line. Moreover, as already stated, the observed α -lines are internal so that the e ff ect cannot be ascribed to sur-face e ff ects. A theoretical explanation for this behaviour anda detailed discussion of the energy dependence of LY and QFin di ff erent scintillating crystals can be found in [20]. Essen-tially, the observed behaviour reflects the fact that an α particlescintillates less with respect to an electron because of its largerdE / dx which can induce saturation e ff ects in the scintillator (theBirks law [21]). The energy dependence of the LY and QF is aconsequence of the energy dependence of the stopping power.Electrons, due to their low stopping power, do not su ff er of sat-uration e ff ects and their LY is, consequently, energy indepen-dent. For this reason the definition of QF α = tan( θ α ) / tan( θ βγ ),given in Sec. 3, represents only an approximation.Using the values in the table we can evaluate the scintil-lation Quenching Factors of α particles (QF = LY( α ) / LY( β )):QF W = ± .006, QF U = ± .006, QF U = ± .006.5ithin this framework, furthermore, it is possible to evaluatethe total amount of light k that escapes the crystal.Using Eq. 3 and considering that, for definition,LY ≡ L / E = β(cid:15) k , we obtain: LY = tan ( θ f it ) α k (5)Now we have to note that the energy calibration we adoptedfor the Heat axis (as discussed in Sec. 2) implies H ≡ E. Thismeans (see Eq. 2) that α = / (1-k γ ), being k γ the absolute LYfor γ/β events. Using this last relation in the last equation wefinally get for γ/β k γ = LY γ tan ( θ f it ) + LY γ (6)Using the values of Tab. 1 we get k γ = is of the order of 2.615 · =
46 keV. Assuming ≈ / photon we get (assuming a Fano fac-tor =
1) the variation in the light channel to be of the order of1 / √ / = ff erent CdWO crystal samples we observed the same anti-correlation (in terms of θ f it ) but with a much smaller spread.Thus the magnitude of this spread is probably dominated byfluctuation in the overall light collection e ffi ciency. In partic-ular we believe that the grade of the surfaces of the crystal (interms of di ff usion of the scintillation light) plays an importantrole in this mechanism.
6. Heat absolute scale and Heat Quenching Factor
From the previous sections is evident that for scintillatingbolometers the Heat / Energy scale is di ff erent for di ff erent kindsof interacting particles (or, better, for particles characterized bydi ff erent LY’s). In other words, with our convention for theHeat axis calibration, H measures the total energy of a γ/β particle while it doesn’t for an α : the energy calibration of α -particles has to be dealt in a separate way.This “displacement” for the α lines is evident in our exper-imental data. In particular the W- α line (as can be seen inFig. 2), whose nominal Q-value is 2516 keV [22], appears (inthe γ -calibrated spectra) at 2627 ± γ -line used for the calibra-tion of the Heat signal is indeed the Tl line at 2615 keV).The observed shift in keV, ∆ E = E α − E γ/β , between theexperimentally reconstructed energy and its nominal value isgiven - for the three mentioned α ’s - by: ∆ E W = ± ∆ E U = ± . ∆ E U = ± α γ E ( α < E ( γ ff erent iso-energy lines up to reach the case in whichH( α = H( γ ∆ E = H L = γ − H L = α . By simple geometric considerationwe get E Light = γ/β = E + Light γ/β tan ( θ f it ) = E (cid:32) + LY γ/β tan ( θ f it ) (cid:33) (8)So, finally, we obtain ∆ E = E [ MeV ] · (cid:32) LY γ/β − LY α tan ( θ f it ) (cid:33) [ keV ] (9)Inserting the experimental values we get: ∆ E W = ± ∆ E U = ± ∆ E U = ± ff erences in the response of adetector with respect to di ff erent type of interacting particles isto introduce relative quenching factors. This is exactly whatis done for the scintillation light. In the same way, followingwhat is often done in literature [23, 24, 25, 26], it is possi-ble to introduce the HQF as the ratio between the Heat sig-nal of an α particle and that of a γ/β of the same energy, fullyabsorbed in the detector. With this definition, for our detec-tor we can quote the HQF measured for the W α decay as:HQF W = / = = + LY γ/β − LY α tan( θ fit ) = + LY γ/β tan( θ fit ) (1 − QF α ).Even if the definition of the (scintillation) QF and the HQFis identical, due to the di ff erent physical mechanism that gener-ates them, they show a substantial di ff erence. While the formerdepends, substantially, only on the “nature” of the scintillator,the HQF, instead, depends on the amount of light (energy) thatescapes the crystal: it depends on the size and / or on the qualityof the crystal. It is also obvious that since some (minor) amountof light that escapes the crystal can be reflected by the reflectingfoil and can be absorbed by the crystal itself, then the HQF willdepend also by the setup itself.With the 3x3x2 cm CdWO crystal measured in [9], wefound the position of the W line at 2670 keV, giving thereforeHQF = / Light ratiosin our framework). Then we can investigate if this fraction isdi ff erent between a γ/β and an α . In other words we could askourselves if the sum of the heat and light signals is the same forparticle converting the same total energy in the scintillator. We6 L i gh t [ k e V ] Energy (Heat) [keV]
Figure 5: Scatter plot Heat vs. Light obtained in a 8 hour measurement withan Am-Be neutron source. In the inset the low energy region. The energyspectrum extends up to 9 MeV. The 4.44 MeV “line” due to the C ∗ is visible.The intrinsic energy spread of this peak is rather large ( ≈ C ∗ in the source. provide therefore an alternative definition of HQF as the ratioof E Light = α and E Light = γ/β .Provided that our model that compensates the energy lossesdue to scintillation is correct, using the values of Eq. 10 we canevaluate the (corrected) HQF:HQF( W) = ± .002 HQF( U) = ± .002 andHQF( U) = ± .002.These HQF’s are compatible with 1 meaning that there are nointrinsic di ff erences in the heat signal generated by an α withrespect to a signal generated by a γ/β . Alternatively we canstate that any eventually present blind channel have the samebehaviour for the two.
7. Background Rejection: α ’s and neutrons The detector was also exposed to a neutron source. It con-sists of a 185 kBq Am-Be source with a neutron productionrate of ≈
10 n / s. The neutron spectrum has its maximum at ∼ γ sources are inserted. The scatter plot is shown inFig. 5: neutron direct interactions are clearly visible, especiallyin the inset. The γ / β events extends well above 2615 keV dueto (n, γ ) reaction on the surrounding materials but, mainly, bythe source itself: for each neutron produced there is 60 % prob-ability to produce an excited state of C that emits a γ of 4.44MeV.Moreover it has to be remarked (see also Sec. 8) that Cdhas a huge neutron capture cross section, with a Q-value largerthan 9 MeV, so that “mixed events” are possible (a neutron We note here that although we always discuss of α ’s it is true that we arealways considering the system ( α + nuclear recoil), where however the nuclearrecoil transports only a small fraction of the total energy. L i gh t [ k e V ] Energy (Heat) [keV]
Figure 6: Scatter plot Light vs. Heat obtained in a 93 hours
Th calibrationmeasurement with a smeared α source facing the CdWO . The liquid standardalso contains a small amount of U. The alpha curve is completely separatedfrom the γ / β line. Moreover it can be noted that the α events are not belongingto a straight line, showing a decrease of the QF with the energy, as discussed inSec. 5. scatters on the Oxygen of the crystal and then is absorbed bythe Cd with subsequent de-excitation of the nucleus). FromFig. 5 we evaluate the neutron scintillation QF with respect to γ / β to be (0.14 ± ÷
300 keV, where most of the events are recorded.In a subsequent test, an α source was mounted close to thedetector in order to evaluate its rejection capability. The customsource was build up using 2 µ L of a 0.1 %
U liquid Standard.The U “droplets” were dried on a 1 cm Al tape and then cov-ered with a 6 µ m Aluminized Mylar foil in order to “smear”the alpha energy down to the region of interest. The source wasthen faced to the crystal on the face opposed to the LD. In Fig. 6we present the scatter plot obtained with the above mentionedsource during a 93 hours Th calibration measurement. Fix-ing at 4 σ the acceptance on the light signal of the 2615 keV γ -line, we evaluate a rejection factor for α particles > σ at thesame energy. As far as the fast neutron interaction is concerned, the rejection factor will be obviously larger, but the most dan-gerous contribution will arise from thermal neutron absorptionthrough Cd. But, at least in principle, thermal neutron can bee ff ectively shielded while fast neutron, via µ -spallation withinthe shielding or close to it, cannot.
8. Background measurement: internal contaminations
In Fig. 7 we present the Light vs. Heat scatter plot obtainedin the 433 h live time background measurement. It can be eas-ily recognized that we observe only two transitions belongingto the Uranium chain, namely
U and
U, corresponding toan internal contamination of 3.1 · − g / g and 5.7 · − g / g, re-spectively. The Uranium chain is broken at U, while a con-tamination of
Po is clearly seen. As it often happens, it israther di ffi cult to determine if this Po contamination arise from7 U U
234 210
PoTl K L i gh t [ k e V ] Energy (Heat) [keV]
Figure 7: Scatter plot Light vs. Heat obtained in a 433 hours live time back-ground measurement. Part of the
Po “bump” is characterized by smallerlight emission. This could arise from a worse light collection e ffi ciency, espe-cially from the lateral part of the crystal. The energy scale of the α is correctedaccording with what exposed in Sec. 6. its parent, Pb. Moreover the absence of a clear peak indi-cates that the contamination is partially penetrating (few µ m)the surface of the crystal (or its surroundings). No events canbe ascribed to the Thorium chain, giving a limit of 9 · − g / g(95% CL) for Th, one of the most “dangerous” contaminantfor DBD searches. But the main (unexpected) feature result-ing from Fig. 7 is the γ / β background above 2615 keV. These11 events cannot arise from external γ ’s due to Tl or
Bisince the observed background lines (583 keV, 2615 keV and1764 keV) are too weak to allow such contribution at high en-ergy. The same holds for internal contaminations belonging toTh and U chain. Unrecognized pile-up of the
Cd- β decay(Q =
320 keV) with a 2615 keV γ cannot contribute at this level(we expect ∼ / year). Internal contaminations due to rarehigh-energy β emitters like Rh (Q = ffi cult to evaluate. But, in any case, we observe γ / β eventsup to 4.5 MeV, extremely large to be associated with standard“known” β emitters. On the other hand we previously tested dif-ferent CdWO scintillating crystals [28], obtaining, with largerstatistics, no events above 2615 keV. The main di ff erence withrespect to [28] is that in the present work the environmentalneutron shielding (consisting of 7 cm of polyethylene and 1 cmof CB ) was removed from the top of the cryostat, leaving ∼ opening, corresponding to ≈ Cd is the most probable mechanism in order to explainthe anomalous background above 2615 keV.
9. Conclusions
For the first time a “large” scintillating bolometer to be usedfor future DBD searches was fully characterized in details interms of energy resolution, internal contaminations and parti-cle identification capabilities. It was shown how the use of the anticorrelation between light and heat improves the energy res-olution by a factor 1.26 @ 88 keV and up to a factor 2.6 @2615 keV. We developed, for the first time, a procedure thatevaluates the Heat Quenching Factor corrected for the loss ofthe scintillation light.This CdWO , grown without any precaution in terms of ra-diopurity, shows extremely low trace contaminations in U andTh. Nonetheless a future experiment based on this compoundshould use isotopic enrichment, for two reasons: i) increase themass of Cd; ii) decrease the fraction of
Cd in order toavoid β decay and, more important, decrease (n, γ ) reaction dueto Cd.As a final remark we point out that such a crystal, depletedin
Cd, would be an extremely interesting compound not onlyfor DBD experiments but also for Dark Matter searches.
10. Acknowledgements
The results reported here have been obtained in the frame-work of the Bolux R&D Experiment funded by INFN, aim-ing at the optimization of a cryogenic DBD Experiment for anext generation experiment. Thanks are due to E. Tatananni,A. Rotilio, A. Corsi and B. Romualdi for continuous and con-structive help in the overall setup construction. Finally, we areespecially grateful to Maurizio Perego for his invaluable helpin the development and improvement of the Data Acquisitionsoftware.
References [1] V. I. Tretyak and Yu. G. Zdesenko, Atomic Data and Nuclear Data Tables, (2002): 83;[2] S. Elliott and P. Vogel, Ann. Rev. Nucl. Part. Sci. (2002):115;[3] A. Morales and J. Morales, Nucl. Phys. B (Proc. Suppl.) (2003):141;[4] F. T. Avignone III, S. R. Elliott and J. Engel (2006), Rev. Mod. Phys., (2008):481.[5] S. Pirro, Eur. Phys. J. A 27 , S1 (2006):25[6] C. Arnaboldi, et al. , Phys. Rev.
C 78 (2008):035502[7] C. Arnaboldi, et al. , Nucl. Instr. and Meth.
A 518 (2004):775[8] M. Pavan , et al. , Eur. Phys. J
A 36 (2008):159[9] S. Pirro, et al. , Physics of Atomic Nuclei No.12 (2006):2109[10] F.A. Danivich, et al. , Phys. Rev.
C 68 (2003):035501[11] G. Bellini, et al. , Phys. Lett.
B 493 (2000):216[12] S. Pirro, et al. , Nucl. Instr. and Meth.
A 559 (2006):361[13] A. Alessandrello, et al. , Nucl. Instr. and Meth.
A 412 (1998):454[14] C. Arnaboldi, G. Pessina, E. Previtali, IEEE Tran. on Nucl. Sci. (2003):979[15] S. Pirro, Nucl. Instr. and Meth. A 559 (2006):672[16] C. Arnaboldi, G. Pessina, S. Pirro, Nucl. Instr. and Meth.
A 559 (2006):826[17] C. Arnaboldi, et al. , Nucl. Instr. and Meth.
A 520 (2004):578[18] E. Conti, et al. , Phys Rev.
B 68 (2003):054201[19] J. Amare, et al. , Appl. Phys. Lett. (2005):264102[20] V.I. Tretyak, Astropart. Phys. (2010):40[21] J.B. Birks, Proc. Phys. Soc. A 64 (1951):874[22] C. Cozzini, et al. , Phys Rev.
C 70 (2004):064606[23] A. Alessandrello, et al. ,Phys Lett
B 408 (1997):465[24] N. Coron, et al. , Phys Lett