Identification of photons in double beta-decay experiments using segmented germanium detectors - studies with a GERDA Phase II prototype detector ∗ 7 I. Abt, A. Caldwell, K. Kr¨oninger , J. Liu, X. Liu, 0 0 B. Majorovits 2 n Max-Planck-Institut fu¨r Physik, Mu¨nchen, Germany a J 4 Abstract 1 v 5 The sensitivity of experiments searching for neutrinoless double beta-decay of ger- 0 manium was so far limited by the background induced by external γ-radiation. 0 Segmented germanium detectors can be used to identify photons and thus reduce 1 0 this background component. 7 The GERmanium Detector Array, GERDA, will use highly segmented germa- 0 nium detectors in its second phase. The identification of photonic events is inves- / x tigated using a prototype detector. The results are compared with Monte Carlo e - data. l c u Key words: double beta-decay, germanium detectors, segmentation n PACS: 23.40.-s, 14.60Pq, 29.40.-n : v i X r a 1 Introduction Neutrinoless double beta-decay (0νββ) is expected to occur, if the neutrino is a massive Majorana particle. The observation of the 0νββ-process would not only reveal the nature of the neutrino as a Majorana particle but could also provide information about the absolute neutrino mass scale (see, e.g. [1]). The germanium isotope 76Ge is a prominent candidate for the observation of the 0νββ-process. Experiments searching for neutrinoless double beta-decay ∗ Max-Planck-Institut fu¨r Physik, Mu¨nchen, Germany, Tel. +49-(0)89-32354-337 Email address: [email protected] (K. Kro¨ninger). Preprint submitted to 8 February 2008 of 76Ge use high purity germanium detectors as source and detector simulta- neously. Their sensitivity is limited by unidentified background events which in previous experiments were mostly induced by external γ-radiation. The Heidelberg-Moscow and IGEX experiments set 90% C.L. lower limits on the half-lifeoftheprocess ofT > 1.9·1025 years [2]andT > 1.6·1025 years [3], 1/2 1/2 respectively. An evidence for the observation of the 0νββ-process was claimed bypartsoftheHeidelberg-MoscowcollaborationwithT = 1.2·1025 years[4]. 1/2 The GERmanium Detector Array, GERDA [5], is a new germanium dou- ble beta-decay experiment being installed in Hall A of the INFN Gran Sasso National Laboratory (LNGS), Italy. Its main design feature is to operate ger- manium detectors directly in liquid argon which serves as cooling medium and as a shield against external γ-radiation. A detailed description of the experi- ment can be found in [5,6]. The detectors for the second phase of the experiment (Phase II) will be en- riched in 76Ge to a level of about 86% and will have a mass of approximately 2 kg each. For the first time, highly segmented germanium detectors will be usedinadoublebeta-decayexperiment. Thesegmentation scheme ischosen to minimize the background level in the energy region around Q = 2039 keV. ββ The current detector design foresees a 6-fold segmentation in the azimuthal angle φ and a 3-fold segmentation in the height z. All segments and the core are read out separately to allow a better identification of photons. The esti- mated gain in background reduction for the GERDA experiment is discussed in [6]. In this paper the results of a study with a GERDA Phase II prototype de- tector are presented. The identification of events with photons in the final state is investigated. Section 2 summarizes the photon identification using coincidences between segments. The underlying physics processes and their signatures are described as are the event selection and the analysis strategy. Section 3 describes the experimental setup of the prototype detector and the data sets collected. The Monte Carlo simulation is introduced in section 4. The results of the study and comparisons with Monte Carlo data are given in section 5. Section 6 concludes and discusses the significance for the GERDA experiment. 2 2 Identification of photon events using segment coincidences The volume over which energy in a single event is deposited inside a detector depends on the incident particles. Segmented detectors can be used to identify events with photons in the final state by requiring coincidences between the segments of a detector. This technique is well established in nuclear experi- ments such as AGATA [7] and GRETA [8], and provides a basis for γ-ray tracking [9]. The potential of segmented detectors for double beta-decay experiments has also been investigated by the Majorana collaboration [10] using a clover de- tector, consisting of four detectors with two longitudinal segments each, and Monte Carlo simulations [11]. 2.1 Signatures and physics processes The signatures of events encountered in double beta-decay experiments canbe classified according to the particles in the final state. A detailed classification for these events is given in [6]. For the identification of photon events only two such classes are considered here: • ClassL:Localenergy deposit. Three different types of events arepart ofthis class: (a)Events with onlyelectrons inthefinalstate. Electrons intheMeV- energy region have a range of the order of a millimeter in germanium [6,12]. Energy is therefore deposited locally. Double beta-decay events which have two electrons in the final state are of this type (these events correspond to Class I events in reference [6]). (b) Events with photons in the final state in which a photon Compton-scatters only once inside the fiducial volume of the detector. Energy is thus deposited locally. (c) ’Double escape’ events: If a photon produces an electron-positron pair and both photons from the subsequent annihilation escape, energy is deposited locally. • Class M: Multiple energy deposits. Photons emitted in radioactive decays have energies in the MeV-energy region and interact dominantly through Compton scattering in germanium. The range of these photons is of the order of centimeters. The different interactions are separated by distances large compared to the scale of Class L events. This class is composed of Classes II-IV in reference [6]. It should be noted that with the technique presented in this paper the three event types in Class L cannot be separated but only be distinguished from Class M events. 3 2.2 Event selection and identification of photon events DuetothewellseparatedmultipleenergydepositsClassMeventsareexpected to deposit energy predominantly in more than one segment. Class L events will predominantly deposit energy in only one segment. Events in which more than one segment measures deposited energy can thus be identified as Class M events. 3 Experimental setup and data sets 3.1 Experimental setup The GERDAPhase II prototype detector under study is a high purity n-type germanium crystal with a true coaxial geometry. It is 70 mm high and has an outer diameter of 75 mm. The inner diameter is 10 mm. The detector is 6-fold segmented in the azimuthal angle φ and 3-fold segmented in the height z. It is placed inside a two-walled aluminum cryostat with a combined thickness of 6 mm. The operation voltage of the detector is (+)3000 V. A schematic diagram of the detector and the experimental setup is given in Figure 1. The core and each segment are read out using charge sensitive PSC 823 pre-amplifiers. The pre-amplified signals are digitized using a data acquisition system basedon514-bitADCPIXIE-4 modulesatasampling rate of 75 MHz. In this configuration the energy resolution of the core is approxi- mately 2.6 keV (FWHM at 1333 keV), the energy resolution of the segments varies between 2.4 keV and 4.7 keV with an average segment energy resolu- tion of 3.3 keV. The threshold of the core and the segments was set to 20 keV. Cross-talk between the core and the segment pre-amplifiers and a constraint in the DAQ system, which resulted in the inability to handle late arriving signals, caused a fraction of less than 10% of individual segment signals to not be recorded. A detailed description of the setup and the prototype properties will be pub- lished [13]. 4 Fig. 1. Schematic diagram of the detector and the experimental setup. 5 3.2 Measurements and data sets Several measurements were performed with different radioactive sources posi- tioned 10 cm above the center of the detector. Energy and time information in all segments and the core were recorded on an event-by-event basis. An event was recorded, if the energy measured in the core exceeded the thresh- old. Measurements were performed with three different sources: (1) a 60 kBq 60Co source, (2) a 100 kBq 228Th source and (3) a 75 kBq 152Eu source. The corresponding data sets are referred to as “source data sets” in the following and contain approximately 4 · 106 events each. An additional measurement without any source was performed in order to estimate the background in the laboratory. This “background data set” contains approximately 106 events. Figure 2 shows the raw energy spectra obtained with the core electrode for the three source data sets and the background data set. E10-1 E10-1 d d N/ N/ d10-2 d10-2 N N 1/ 1/ 10-3 10-3 10-4 10-4 10-5 10-5 10-6 10-6 10-7 10-7 0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500 E [keV] E [keV] E10-1 E10-1 d d N/ N/ d10-2 d10-2 N N 1/ 1/ 10-3 10-3 10-4 10-4 10-5 10-5 10-6 10-6 10-7 10-7 0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500 E [keV] E [keV] Fig. 2. Raw energy spectra obtained with the core electrode for the 60Co (top, left), 228Th (top, right) and the 152Eu (bottom, left) source data sets. The sources were placed 10 cm above the detector. The energy spectrum for the background data set is also shown (bottom, right). The binning is 1 keV and the spectra are normalized to the area. 6 4 Monte Carlo simulation A Monte Carlo simulation of the prototype setup was performed using the GEANT4 [14] based MaGe framework [15]. The energy deposited in each segment is recorded and the core energy is calculated by adding all segment energies. The drift anisotropy of charge carriers inside the germanium diode [16] can cause electrons and holes to deviate from their drift path. It is therefore possi- ble to measure energy in one segment even if the energy was deposited in the neighboring segment. Hence, a correction is applied to the segment energies. An effective model is used which assigns a segment to each energy deposit de- pending on its position with respect to the axes of the crystal and the segment ◦ borders. The maximum angular shift is 3.5 . The directions of the crystal axes were measured and used as input for the Monte Carlo.This includes an overall variation of this effect by 40% with respect to the two hemispheres. Each segment is assigned a relative efficiency with respect to the core on the order of 90%. This effectively models the DAQ-inefficiency. The segment and core energies are individually smeared according to the energy resolution of the prototype detector measured in each channel. 5 Results The results of the measurements are presented in the following and compared to Monte Carlo data. In order to account for background from radioactive isotopes in the laboratory the fraction of background events in each data set is estimated. 5.1 Background estimate The number of background events in a given source data set is estimated using characteristic photon lines in the spectrum. These lines are associated with thedecays of214Pb(352keV), 214Bi(609keV, 1120keV, 1765keV, 2204keV) and 40K (1461 keV). The photon lines are fitted with a Gaussian plus linear function and the number of events, n , under each peak is calculated. i 7 For the background data set, the fraction of events under the ith peak is denoted f = N /N, where N is the number of events under the ith peak i i i and N is the total number of events in the spectrum. For each source data set the total number of background events, n , is estimated by minimizing bkg a χ2-function defined as (n ·f −n )2 χ2 = χ2(nbkg) = X bkσg2 +i n i , (1) i i i where σ is the Poissonian uncertainty on the expression n ·f . i bkg i The fraction of background events in the source data sets are estimated as 14.0% (χ2/d.o.f. = 1.0) for 60Co, 8.3% (χ2/d.o.f. = 0.7) for 228Th and 15.8% (χ2/d.o.f. = 8.4)for152Eu.Anuncertainty onthebackground fractionof0.1% is estimated. 5.2 Photon identification and reduction The segment multiplicity, N , is defined as the number of segments with mea- S sured energies larger than the threshold of 20 keV. A measure for the identification of photonic events is the number suppression factor, SF , defined as the ratio of the number of events within a 10 keV N region around a certain energy and the number of events which, in addition, have a segment multiplicity of N = 1. In order to quantify the identification S of photons which deposit their full energy within the detector the line sup- pression factor, SF , is defined similarly to SF , but the number of events is L N replaced by the number of events under the photon peak under study. For the calculation of the suppression factors the number of events in the source data sets are corrected for the background by subtracting the back- ground contribution. ThenumbersuppressionfactoriscalculatedfortheQ regionof76Ge(2039keV) ββ whereas the line suppression factors are calculated for the photon lines of 60Co (1173 keV, 1333 keV and the summation peak at 2506 keV), 208Tl (511 keV, 583 keV, 861 keV, 2615 keV and the corresponding single and double escape peaks at 2104 keV and 1593 keV), 212Bi (1620 keV) and 152Eu (122 keV, 245 keV, 344 keV, 779 keV, 964 keV, 1086 keV, 1112 keV and 1408 keV). 8 The results are displayed in Table 1 for data and Monte Carlo data. For a discussion of the agreement between data and Monte Carlo see section 5.5. Table 1 Suppressionfactors for different sources and energies calculated for data and Monte Carlo. The background has been subtracted from the data. For a discussion on the agreement between data and Monte Carlo see section 5.5. The uncertainties are statistical uncertainties only. Source Energy SF (data) SF (data) SF (MC) SF (MC) N L N L [keV] 60Co 1173 - 2.56 ± 0.01 - 2.56 ± 0.01 1333 - 2.63 ± 0.01 - 2.63 ± 0.01 2506 - 34.6 ± 5.7 - 43.0 ± 11.0 2039 14.2 ± 2.1 - 12.5 ± 2.1 - 228Th 511 - 1.92 ± 0.01 - 1.91 ± 0.02 583 - 2.04 ± 0.01 - 2.01 ± 0.01 861 - 2.35 ± 0.03 - 2.37 ± 0.05 1593 - 1.09 ± 0.02 - 1.09 ± 0.04 1620 - 2.85 ± 0.01 - 2.84 ± 0.13 2104 - 3.13 ± 0.01 - 3.20 ± 0.11 2615 - 3.04 ± 0.02 - 3.23 ± 0.04 2039 1.68 ± 0.02 - 1.66 ± 0.05 - 152Eu 122 - 1.01 ± 0.002 - 1.01 ± 0.003 245 - 1.26 ± 0.01 - 1.22 ± 0.01 344 - 1.54 ± 0.01 - 1.55 ± 0.01 779 - 2.29 ± 0.01 - 2.26 ± 0.02 964 - 2.46 ± 0.02 - 2.41 ± 0.02 1086 - 2.54 ± 0.02 - 2.50 ± 0.03 1112 - 2.52 ± 0.02 - 2.54 ± 0.04 1408 - 2.64 ± 0.02 - 2.72 ± 0.02 The line suppression factors increases from 1.01±0.002 at 122 keV to 3.04± 0.01at 2615keV, where thesuppression increases with increasing energy.This is expected as the average number of Compton-scattering processes increases. Figure 3 shows the line suppression factors as a function of the core energy for data and Monte Carlo data. 9 The double escape peak of the 2615 keV photon from the de-excitation of 208Tl at 1593 keV is basically not suppressed. These Class L events have a very localized energy deposition. In comparison, the 1620 keV line from the 212Bi is suppressed by a factor of 2.85±0.01. These events are predominantly encompassed in Class M. Figure 4 shows the energy spectrum for the 228Th source with and without a segment multiplicity requirement of N = 1. The S left figure shows the energy region up to 3 MeV, the right figure shows a close-up of the region around 1.6 MeV. Note that background has not been subtracted from the data spectra. The number suppression factor for the 60Co source is SF = 14.2 ± 2.1. It N is large compared to the suppression factor for the 208Th source of SF = N 1.68± 0.002. For the latter, the Q region lies within the Compton contin- ββ uum of the 208Tl photon. A single scattering process can cause a local energy deposit.Incontrast,forthe60Co sourceanenergydeposit inthisenergy region is only possible if both photons (1173 keV and 1333 keV) deposit energy in the same segment. In contrast, the number suppression factor for 0νββ-decay events is expected to be close to unity as the electrons in the final state mostly deposit energy in only one segment. Figure 5 shows the energy spectrum for the 60Co source with and without a segment multiplicity requirement of N = 1. The left figure shows the energy S region up to 3 MeV, the right figure shows a close-up of the region around the Q -value of 76Ge. Note that background has not been subtracted from the ββ spectra. 10