Low energy fast events from radon progenies at the surface of a CsI(Tl) scintillator
S. C. Kim, H. Bhang, J. H. Choi, W. G. Kang, H. J. Kim, K. W. Kim, S. K. Kim, Y. D. Kim, H. S. Lee, J. I. Lee, J. H. Lee, J. K. Lee, M. J. Lee, S. J. Lee, J. Li, J. Li, Y. J. Li, X. Li, S. S. Myung, S. L. Olsen, S. Ryu, I. S. Seong, J. H. So, Q. Yue
aa r X i v : . [ a s t r o - ph . C O ] A ug Low energy fast events from radon progeniesat the surface of a CsI(T ℓ ) scintillator S.C. Kim, a 1
H. Bhang, a J.H. Choi, a W.G. Kang, b H.J. Kim, c K.W. Kim, a S.K. Kim, a 2
Y.D. Kim, b H.S. Lee, a J.I. Lee, a J.H. Lee, a J.K. Lee, a M.J. Lee, a S.J. Lee, a J. Li, a J. Li, d Y.J. Li, d X. Li, a S.S. Myung, a S.L. Olsen, a S. Ryu, a I.S. Seong, a J.H. So, c Q. Yue d (KIMS Collaboration) a Department of Physics and Astronomy, Seoul National University, Seoul 151-742,Korea b Department of Physics, Sejong University, Seoul 143-747, Korea c Physics Department, Kyungpook National University, Daegu 702-701, Korea d Department of Engineering Physics, Tsinghua University, Beijing 100084, China
Abstract
In searches for rare phenomena such as elastic scattering of dark matter particlesor neutrinoless double beta decay, alpha decays of
Rn progenies attached tothe surfaces of the detection material have been identified as a serious source ofbackground. In measurements with CsI(T ℓ ) scintillator crystals, we demonstratethat alpha decays of surface contaminants produce fast signals with a characteristicmean-time distribution that is distinct from those of neutron- and gamma-inducedevents. Key words: alpha,
Rn, CsI(T ℓ ) crystal, dark matter, pulse shape discrimination Signals caused by radioactive decays of
Rn progenies that adhere to de-tector surfaces are known to be serious background for rare phenomena ex-periments such as searches for WIMPs (Weakly Interacting Massive Particles) [email protected] [email protected] Preprint submitted to Elsevier 6 November 2018 ark matter and neutrinoless double beta decay [1,2,3].
Rn is a noble gasthat permeates the air. It decays to
Po, which is a reactive metal that read-ily adheres to almost any surface [4]. Through successive decay chains, theprogenies of
Rn can accummulate on detector surfaces. When they decay,only a portion of the energy of the decay products is detected, thereby pro-ducing a troublesome low energy background. Among the progenies of
Rn,
Pb and
Po are the most dangerous because they have long half-lives,22.3 years and 138 days, respectively. If an experimental system is isolated,
Po becomes the principal alpha emitting contaminant. The energy spec-trum of these surface events appears as a continuum that ranges from the fullpeak energy of the alpha down to the very low energies. Anomalously fastevents have been observed in WIMP search experiments that use inorganicscintillators, and these have been attributed to the effects of surface alpha(SA) events since their rate is seen to be reduced when the detector surfacesare polished [1,5,6]. The KIMS experiment at the Yangyang UndergroundLaboratory in Korea looks for WIMP-induced nuclear recoils in an array ofCsI(T ℓ ) scintillators [7]. Figure 1 shows a scatter plot of energy deposits ver-sus LMT10 for events seen in the KIMS detector, where LMT10 denotes thenatural logarithm of the mean-time of each event calculated for a 10 µ s inter-val starting from the beginning of the event (MT10= P ti< µs A i × t i P ti< µs A i , A i is thearea of the i th cluster, which is usually equivalent to a single photo-electron,of an event). In the lower right-hand portion of the plot, strong alpha-decaypeaks can be seen with tails that extend down to zero energy and low LMT10values. Here one can see that LMT10 values for the low energy tail events aredistinctly smaller than those from gamma-ray induced signals.WIMPs are expected to scatter elastically from nuclei. The KIMS experimentis designed to detect the energy deposited by the recoiling nucleus and uses apulse shape discrimination (PSD) analysis to distinguish nuclear recoil eventsfrom gamma-induced backgrounds [7]. The presence of surface alpha events atenergies below 10 keV, the main region of interest for the dark matter search,complicates the PSD analysis. In order to investigate the characteristics of SA background events, we con-taminated the surface of a small CsI(T ℓ ) sample crystal with Rn progeniesby placing it for four days in a special chamber at the Korea Research Insti-tute of Standards and Science (KRISS) in which the
Rn concentration wasaround 4.33 MBq/m . The size of the contaminated crystal is 3 cm × × ig. 1. Deposited energy (horizontal) versus LMT10 (vertical) in data from theKIMS experiment. Here LMT10 denotes the natural logarithm of the mean-timeof each event computed over first 10 µ s interval. Here one can see that SA eventsare present as a band whose LMT10 is small compared to the background, whichindicates that they decay quickly.Fig. 2. A schematic depiction of the Radon progenies contaminated Double crystalDetector (RDD). Crystal A is the Rn-progenies contaminated CsI(T ℓ ) crystaland crystal B is a clean CsI(T ℓ ) crystal that is used to detect an alpha particle thatescapes from crystal A. The two crystals are separated by three 2 µ m-thick layersof aluminum foil. Rn nucleus decays to
Po with a half–life of 3.8 days. In aperiod of several tens of minutes, the
Po nucleus decays to
Pb via severaldecay steps.
Pb beta decays with a half–life of 22.3 years to
Bi, whichsubsequently beta decays to
Po. From this series of beta decays,
Po, themain source of surface alpha events, is continuously produced on the crystalsurface. It decays into
Pb with a half-life of 138 days, emitting a 5304 keValpha particle. The kinetic energy of the recoiling
Pb nucleus is 103 keV.
Fig. 3. MT25 versus the visible energy in crystal B, where MT25 is the mean-timeof each event that is determined over a 25 µ s event-time window. The events insidethe red solid lines are selected as alphas escaping from crystal A. Figure 2 shows a schematic diagram of the experimental setup, which wenamed the Radon progenies contaminated Double crystal Detector (RDD),that was used to study the characteristics of SA events. In order to tag escapingsurface alpha particles, we attached a clean CsI(T ℓ ) crystal (crystal B in Fig. 2)to a face of the contaminated one (crystal A in Fig. 2). We inserted three 2 µ m-thick aluminum foils between the two crystals to provide a barrier to preventcross-talk of the scintillation light between the crystals that the 5 MeV alphascan easily penetrate. According to the SRIM program [8], the energy loss of a5304 keV alpha in a 6 µ m-thick aluminum layer is 998 keV; thus alphas thatpenetrate the foils still have ample energy to provide a trigger signal. Theselayers also serve as a reflector for the collection of scintillation light. The sidesof the crystals are wrapped with teflon tape. A 3 inch photomultiplier tube(PMT), 9269QA from Electron Tubes, Ltd. is attached to each crystal todetect the scintillation photons. The PMT signals are amplified 100 times bya fast amplifier from Notice Co., Ltd. The amplified signals are digitized by a400 MHz Flash Analog-to-Digital Converter(FADC) of Notice Co., Ltd. thatis mounted in a Versa Module Eurocard (VME) crate that is read-out by alinux-operating PC via a VME–USB2 interface. The DAQ system is based onthe ROOT package [9]. When two or more photoelectrons (PEs) are detectedin each PMT within a 2 µ s time window, a trigger is generated. Events withpulse width longer than 300 ns are also triggered in order to include highenergy events in which many PEs are merged into a single big pulse. For each4vent, the PMT responses throughout a 40.96 µ s time window is recorded. Ofthese, pulses inside a 25 µ s window are used for analysis.To tag SA events in crystal A, we require a signal in crystal B that is con-sistent with that of an alpha particle. Because of the good PSD power ofCsI(T ℓ ) scintillators, as shown in Fig. 3, alpha signals are clearly separatedfrom other backgrounds. Here, MT25, the mean-time of each event determinedover a 25 µ s time window, is used. In Fig. 3, the visible energy is the measuredenergy which differs from the actual energy above around 100 keV because ofthe saturation effect resulting from the DAQ system optimized for collectinglow energy events. Furthermore, the energy of alpha events is underestimatedmore by the quenching factor in CsI(T ℓ ) , which is about 0.5. No correctionsfor these effects have been applied. Events that populate the region inside thered solid lines of Fig. 3 are selected as alphas that escape from crystal A.Figure 4 shows a scatter plot of energy in crystal A versus the visible energyin crystal B for events tagged as alphas. This figure shows the sum of theenergies at both crystals has an upper bound, as expected. The wide spreadin the sum of the energies reflects the energy loss in the Al foil layers andpossible dead layer on the surface of the detector ,which depends upon thealpha particle’s direction. To separate SA events from background, an upperbound of alpha energy at crystal B is set at 620 keV and a lower bound at70 keV. The experiment started two months after the radon contamination ofthe crystal was done and ran for about 3 months. The rate of events taggedas SA increased with time because of the continuous supply of Po from
Pb. The event rate became 947 events / day by the end of the measurementperiod. From this value, the amount of Pb contaminants implanted at thesurface of the crystal is estimated to be about 1 . × and the adhesion rateof radon progenies in the radon chamber is estimated to be about 13 / cm / s. Visible Energy, B (keV)0 200 400 600 800 1000 E n e r g y , A ( ke V ) Fig. 4. Energy of crystal A versus the visible energy of crystal B for events thatare tagged as an alpha. nergy (keV)2 4 6 8 10 12 14 16 18 20 L M T -1.5-1-0.500.511.5 gammaSA Fig. 5. LMT10 of SA events and gamma events in RDD
In the KIMS experiment, the quantity LMT10 is used as the PSD discrimi-nator. The distribution of LMT10 for SA events must be understood in orderto distinguish them from dark matter candidate events in KIMS. For this weanalyzed the response of crystal A in the SA-tagged events selected accordingto the description in the previous section.
Energy (keV)10 20 30 40 50 60 70 N o . o f eve n t s Fig. 6. The energy spectrum in crystal A for SA-tagged events.
Figure 5 shows the LMT10 distribution for SA- and gamma ray-induced eventsfrom RDD in the low energy range that is used in the dark matter search. The6A events range down to 3 keV, the energy threshold for this study. Figure 6shows the energy spectrum of SA events. The energy deposited in crystal A isthe sum of the recoil energy of the recoiling
Pb nucleus and some partialenergy of the alpha particle. The
Pb recoil energy is 103 keV, and it isexpected to show up as 7–8 keV due to the quenching factor of CsI(T ℓ ) ifwe assume the quenching factor for Pb is the same as that for Cs and Inuclei [10,11]. The FWHM energy resolution at this energy is 4 keV. Thereis no clear
Pb signal such as that which was seen in another study thatused a phonon sensor [3]. The scintillator may have an inactive surface layerwhere the scintillation efficiency is very small that it can cause smearing of
Pb signal.
LMT10-1.5 -1 -0.5 0 0.5 1 1.5 N o . o f eve n t s Fig. 7. LMT10 distibutions for various reference data at 3 keV. (a) LMT10 forSA-induced events in crystal 1, (b) gamma-induced events in crystal 1, (c) neu-tron-induced events in crystal 2 and (d) gamma-induced events in crystal 2.
The LMT10 distributions for 3–4 keV energy bin are shown in Fig. 7. Thecrystal used for the SA study with
Rn contamination is labeled crystal 1 inthe figure; the crystal used for the neutron response study is a different crystal,crystal 2, which was exposed to an Am-Be neutron source. Here, we alsopresent the results of gamma calibration for both test crystals for comparison.The gamma calibration was done by irradiating the crystals with
Cs source.The temperature for SA study setup was maintained as (25 . ± . ◦ C, andfor neutron study setup, (25 . ± . ◦ C. The results in the figure show thatSA-induced signals are, on average, faster than neutron-induced signal events,and their LMT10 distributions are distinct. Figure 8 shows the value of thepeak position of LMT10 distribution as a function of the energy for eachtype of event, which is not necessarily same with the mean value since thedistribution is an asymmetric Gaussian. The SA LMT10 distribution obtainedfrom this study is used to distinguish possible WIMP-induced nuclear recoilsfrom surface alpha-induced events in the KIMS data.7 nergy (keV)3 4 5 6 7 8 9 10 11 L M T -0.200.20.40.60.81 SA, crystal 1gamma, crystal 1neutron, crystal 2gamma, crystal 2
Fig. 8. The value of the peak position of LMT10 distributions for SA-, neutron-and gamma-induced events. Crystal 1 refers to the crystal used for SA study andcrystal 2 is the one used for the neutron response measurement.
We have studied surface alpha background from
Rn progenies with aCsI(T ℓ ) scintillator contaminated with Rn progenies, mainly
Pb andits daughters. The main alpha emitter is
Po. When
Po alpha decaysat the surface, it doesn’t deposit its full energy into the detector. Its energyspectrum ranges from the full peak energy to the very low energy. We directlyshow that SA events decay faster than neutron and gamma events. We alsoobtained the distribution of LMT10, the PSD parameter, for SA events. Thiscan be used to analyse the dark matter search data of KIMS experiment.
Acknowledgments
We thank Dr. Jongman Lee for providing us access to the KRISS radon cham-ber. We are very grateful to the Korea Midland Power Co. and their staff forproviding the underground laboratory space at Yangyang. This research wassupported by the WCU program (R32-10155) and Basic Science Research Pro-gram (2010-0005332) through National Research Foundation of Korea fundedby the Ministry of Education, Science and Technology.8 eferences [1] V. A. Kudryavtsev et al. , Astropart. Phys. , 401 (2002).[2] M. Pavan et al. , Eur. Phys. J. A , 159 (2008).[3] R. F. Lang et al. , Astropart. Phys. , 60 (2010).[4] S. Cooper et al. , Phys. Lett. B , 6 (2000).[5] V. A. Kudryavtsev et al. , Phys. Lett. B , 167 (1999).[6] N. J. T. Smith et al. , Phys. Lett. B , 9 (2000).[7] H. S. Lee et al. , Phys. Rev. Lett. , 091301 (2007).[8] J. F. Ziegler, J. P. Biersack and U. Littmark, The Stopping and Range of Ionsin Matter, SRIM Co., Chester, Maryland (2008).[9] R. Brun and F. Rademakers, ROOT - An Object Oriented Data AnalysisFramework, Proceedings of AIHENP ’96 Workshop, Lausanne, Sep. 1996; Nucl.Instrum. Methods A 389 , 81 (1997), see also http://root.cern.ch.[10] H. Park et al. , Nucl. Instrum. Methods
A 491 , 460 (2002).[11] V. I. Tretyak, Astropart. Phys. , 40 (2010)., 40 (2010).