Eur. Phys. J. C manuscript No. (will be inserted by the editor) Gerda The Performance of the Muon Veto of the Experiment K. Freund1,a, R. Falkenstein1, P. Grabmayr1,d, A. Hegai1, J. Jochum1, M. Knapp1,b, B. Lubsandorzhiev1,2, F. Ritter1,c, C. Schmitt1, A.-K. Schütz1, I. Jitnikov3, E. Shevchik3, M. Shirchenko3, D. Zinatulina3 1PhysikalischesInstitut,EberhardKarlsUniversität Tübingen,Tübingen,Germany 2InstituteforNuclearResearchoftheRussianAcademyofSciences,Moscow,Russia 3JointInstituteforNuclearResearch,Dubna,Russia 6 1 0 2 Received:date/Accepted: date n a J Abstract Low background experiments need a sup- tal signature of the 0νββ decay is a peak at Q , the ββ 2 pressionofcosmogenicallyinducedevents.TheGerda Q value of the decay. 2 experiment located at Lngs is searching for the 0νββ Gerdawasconstructedintheundergroundlabora- ] decay of 76Ge. It is equipped with an active muon veto tory of Laboratori Nazionali del Gran Sasso (Lngs) of et the main part of which is a water Cherenkov veto with InfninItaly,whichoffersanoverburdenof3500meter d 66 PMTs in the water tank surrounding the Gerda water equivalent (m.w.e.) of rock and hence a reduc- s- cryostat. With this system 806 live days have been tion of the muon flux by a factor of ∼106 to a rate of n recorded, 491 days were combined muon-germanium ∼3.4×10−4/(s·m2). This remaining muon flux however i data. A muon detection efficiency of ε = (99.935± . µd issufficienttocauseanon-negligiblebackgroundinthe s 0.015)% was found in a Monte Carlo simulation for c region of interest around Q =2039 keV when increas- ββ si the muons depositing energy in the germanium detec- ing the sensitivity beyond T10/ν2>1025 yr or when re- y tors.Byexaminingcoincidentmuon-germaniumevents questing a background index BI<10−2 cts/(keV·kg·yr). h a rejection efficiency of ε = (99.2+0.3)% was found. µr −0.4 However, also other analyses profit from the reduced p Withoutvetoconditionthemuonsbythemselveswould [ backgrounds (see e.g. Ref. [3,4]). 1 cctasu/s(ekeaVb·akcgk·ygrr)ouantdQind.exofBIµ =(3.16±0.85)×10−3 The origin of muons at Lngs for Phase I of Gerda ββ v wastwofold.Firstly,themajorityofthedetectablemu- 5 Keywords cosmic muons · water Cherenkov detec- ons are produced cosmogenically. Spectrum and angu- 3 tors · scintillation detectors · data analysis lar distribution of the muons are both altered by the 9 5 profile of the rock overburden and have been measured PACS 95.85.Ry · 29.40.Ka · 29.40.Mc · 29.85.Fj 0 for Lngs with high precision [5]. These muons have an 1. average energy of (cid:104)Eµ(cid:105) = 270 GeV. Secondly, a source 0 for muons was the Cngs neutrino beam from Cern [6] 6 1 Introduction which created muons via e.g. ν +d → µ− +u reac- µ 1 tions in the vicinity of the detector. This contribution : v Muonsmaycauseasubstantialbackgroundtorareevent amountedto2.2%ofthetotalmuonfluxinGerda.As Xi searches like Gerda (Germanium Detector Array) by theCngswasshutdownafter2012thefuturePhaseII generating counts in the region of interest (ROI) ei- of Gerda will be unaffected. In order to reduce muon r a ther through direct energy deposition in the detectors inducedbackground,amuonvetocomprisedofawater or through e.g. decay radiation of spallation products. Cherenkovvetoandascintillatorvetowasimplemented TheGerdaexperimentissearchingfortheneutrinoless totagmuonsandtouseitsresponseasavetosignalin doublebeta(0νββ)decayof76Ge[1,2].Theexperimen- the 0νββ analysis. This paper describes the hardware and setup of the apresent address: maxmentGmbH,Germany vetoandtheDAQsystem.Theperformanceoftheveto bpresent address: Areva,France cpresent address: BoschGmbH,Germany will be presented and compared to Monte Carlo simu- dCorrespondence,email:[email protected] lations. Detection and rejection efficiencies for the veto 2 plastic scintillator veto low-Z materials are used in order to reduce cosmogenic activation [1,12]. The Gerda muon veto consists of two independent twin lock parts which are read out by the same DAQ. The first and main part is a water Cherenkov veto that detects theCherenkovradiationofthetraversingmuons.Thus, glove box the water tank is instrumented with 66 encapsulated photomultipliers(PMTs).Thesecondpartofthemuon clean room veto is comprised of plastic scintillator panels. Since chimney shutter neck theCherenkovvetooffersreducedtaggingcapabilityat or around the neck of the cryostat, plastic panels were 590 m 3 > 0.17 M Ω m heat placed on top of the clean room in order to close this exchanger weak spot in the Cherenkov veto. water tank radon shroud 2.2 The Cherenkov veto Ge Below the water level there are 66 PMTs (8” size) of detector type 9354KB/9350KB by ETL [13] installed. In order array to protect the PMTs from the surrounding water, each copper PMT is housed in a stainless-steel encapsulation which shield cryostat is further developed from the design of the Borexino man 3 64 m LAr hole capsules[14].Thecapsuleoflowradioactivitystainless- 66 PMT Cherenkov steel [15] is closed with a custom made PET cap and is veto filledwithIRspectroscopyoil[16].Theoilkeepstheop- 2m tical transition between PET cap and PMT as smooth 5m as possible. Thus, efficiency losses by total internal re- Fig. 1 Asketchofthe Gerdaexperiment. flections are minimized. The bottom of the base of the PMT is encased in polyurethane [17] and sealed with silicon gel [18]. The lower part of the PMT, i.e. the will be discussed. Parts of this work have been pub- dynode structure, is protected from magnetic fields by lished during the respective PhD periods [7–11]. a cone of µ-metal [19]. Each capsule is equipped with an optical fibre for optical calibration pulses from out- side the water tank. To keep the number of cables and 2 Instrumentation connectorswithinthecleanwaterataminimumanun- derwater high-voltage cable [20] connects the PMT to Here, a technical description of the apparatus is given. a signal splitter outside of the water tank. This band- Purposeandfunctionofbothpartsoftheveto,Cheren- pass separates the signal from the high voltage (HV). kov and scintillator panels, is introduced. The trigger The former is digitized and stored on disk upon a trig- logic, calibration and data acquisition are summarized. ger (sec. 2.4). The HV is supplied by a HV multichan- nelsystembyCAEN[21]equippedwithsix12-channel boards [22]. A sketch of the capsule is shown in Fig. 2 2.1 Gerda and images of the components of the Cherenkov veto are shown in Fig. 3. In Gerda an array of bare germanium detectors is op- The water in the Gerda water tank was provided eratedinacryostatthatcontains64m3 ofliquidargon initiallly by the Borexino water plant. The purifica- (LAr) as seen in Fig. 1. The cryostat is surrounded by tion system which consists of filters, de-ionizers and an a water tank with a diameter of 10 m and a height of osmosisunitisrunningwith2.4m3/hourandkeepsthe 9.4mthatisfilledwith590tultra-purewater;itswalls water at the ultra-pure level of of >0.17 MΩ m [23]. are covered with a reflective foil. The cryostat has a The PMTs are arranged in seven rings in the water connection (neck) through the water tank to the clean tank, their distribution is shown in Fig. 4. One ring of room above from which the germanium detectors are six PMTs (101-106) is pointing inside the separate vol- loweredintotheLAr.Bothwatertankandcryostatare ume under the cryostat (commonly referred to as “pill- partoftheinnovativeshieldingdesignof Gerdawhere box”), two rings ofeightPMTs (201-208)and 12 PMTs 3 Fig. 3 Images of the components of the Cherenkov veto during the installation. Top, left: a capsule mounted on the floor; top, right: a capsule on the wall; bottom, left: a diffuser ball; bottom, right: PMTs and Daylighting Film in “pillbox” below thecryostat PET-window flanges chimney μ-metal 6.5m 710 709 708 707 706 705 704 703 702 701 steel capsule 610 609 608 607 606 605 604 603 602 601 5m 3.5m 510 509 508 507 506 505 504 503 502 501 oil filling tubes fibre holder 2m 410 409 408 407 406 405 404 403 402 401 PMT base cable feedthrough 307 305 306 308 204 304 205 203 309 104103 303 105102 206 106101 202 Fig. 2 Parts ofaPMTcapsule 310207 201302 208 311 301 312 Fig. 4 AsketchofthePMTdistributioninsidetheGerda (301-312) respectively are placed on the bottom of the watertank.Thevioletareaonthebottomplatesignifiesthe water tank looking upwards and four rings of 10 PMTs circumferenceofthecryostatandthegapsshowtheposition are placed on the wall (401-710), pointing horizontally of manholes into the pillbox. The other three violet areas showthelocationofthemanholeintothecryostat.Allcables towards the cryostat. The number of PMTs and their andfibresareleadoutofthewatertankthroughthree50cm placement was chosen after an extensive Monte Carlo flangesmountedina“chimney” abovethewaterlevel(Fig.1) study [7]. The PMTs closest to the cryostat, i.e those 4 ] 30 ] % . u [ . a cy 3 [ n 20 ns e o ci t ffi 2 o h e p m 10 # u 1 t n a u q 100 300 500 700[nm] Fig. 5 Efficiency (red curve) of a 9354KB type PMT adapted from Ref. [13] and simulated photon spectrum with wave-lengthshiftingeffect(bluehistogram) fromthepillboxandtheinnerringonthebottom,were Fig. 6 SchematicdrawingofadiffuserballafterRef.[11] selectedaccordingtotheirperformanceandradioactiv- ity (type 9354KB is a low activity module). However, due to the distance to the germanium detectors and thelowoverallmasstheradioactivityofthemuonveto areworkingverywell,thecalibrationfibresattachedto is negligible compared to the one of the stainless steel every PMT are currently not in use. cryostat [7]. The floor and the walls of the water tank and of the cryostat are covered with the reflective Daylighting 2.3 Scintillator Veto Film DF2000MA (commonly known as “VM2000”) [24] which offers a reflectivity of >99% in order to increase Muons passing through the neck of the cryostat may the light yield of each event. This comes at the cost either traverse a too short distance in the “pillbox” or of a reduction in tracking capabilities of each individ- in the water tank to be detected. In order to keep the ual muon. The reflective foil does not only increase the muon rejection efficiency as high as possible, a veto of collectedlightyieldofeachmuon,butitactsasawave- plastic scintillator panels was conceived and installed. lengthshifteraswell.Itshiftsthepredominantlyultra- Each scintillator panel contains a 200×50×3 cm3 violet Cherenkov photons to around 400 nm where the sheet of plastic scintillator based on polystyrol with an PMTs are most efficient. The efficiency of a 9354KB addition of PTP (2%) and POPOP (0.03%) [26]. In type PMT and the wave-length shifting effect imple- addition the panels contain optical fibres [27] on the mentedintheGeant4simulationsareshowninFig.5. narrow sides as light-guides, an electronics board with ThePMTsarecalibratedregularlywithasetoffive a trigger and shaper and a PMT. Half of the panels custom made diffuser balls shown in Fig. 6 which are areequippedwithPMTsbyHamamatsuPhotonics[28] constructed to provide a light source as isotropic as and the other half with PMT-085 by Kvadrotech. The possible.Theseareglassballswithadiameterof4.5cm PMT-085 are powered by the same HV supply as the thatarefilledwithamixtureofsilicongel[18]andglass muon veto PMTs however connecting 3 PMTs to one pearls[25]withadiameterof∼50µm.Anopticalfibre HV output. The front-end electronics, the Hamamatsu is glued into a small vial inside the ball with a higher PMTs and all other components are powered by a cus- content of glass pearls. The cladding at the end of the tom made power source with +12 V and ±6 V. fibre is removed and the fibre is roughened in order to The panels are arranged in three layers covering an obtainadiffuselightemission.Ultra-fastLEDsoutside area of 4×3 m2 centered over the neck of the cryostat. of the water tank can be pulsed to illuminate the five It was aimed to keep both trigger rate and thus data diffuser balls [11]. Four of these balls are distributed volume of the veto as low as possible. Therefore, the in the main water tank and a fifth is placed inside the triggerisatriplecoincidenceofallthreelayerssincethis “pillbox”. With an appropriate setting of the LEDs it option offers a high discrimination against non-muonic is possible to illuminate all PMTs simultaneously with background events in the panels such as γ rays from single photons and thus record single-photon responses environmental radioactivity. Thus, the scintillator veto of every PMT at the same time. As the diffuser balls recordsanalmostpuremuonsample.Individually,each 5 120 scintillator panels ax panel 2 [mV] 100 events 1110002310 20 40 60 8G0ERD1A0 105/04 670000 ligHhPVt M pduTal sisseiyg nchaalin muslitgipnleaxler pulser m 80 max [mV] 500 difbfuasller LED driver splitter PMT signal RC 400 FADCs 60 muon PMT & controller HV supply Ge-array 300 40 cryostat CAEN HV 200 water tank remote PC, 20 PMT data storage 100 Fig. 8 Schematic drawing of the data acquisition (DAQ) setup 0 0 0 20 40 60 80 100 120 max panel 1 [mV] Fig.7 Scatterplotofthepulseheightmaximaoftwopanels inatriplestackwithaproductcut(redline)asgiveninEq.1. events in which PMTs next to each other have fired. Theinsetshowsthepulseheightdistributionofonepaneland The final trigger condition is set such that five FADCs aLandau-fit(red) must trigger on at least 0.5 photo-electrons (p.e.) each within 60 ns. That signal is realized by a FPGA and it starts the readout of all traces covering a period of panelshowsapulse-heightdistributionwhichtakesthe 4 µs. These are stored together with the time stamp, form of a Landau peak. Despite the triple coincidence, lately from a GPS clock. A schematic drawing of the small γ contamination at low pulse heights remains. entire muon veto and its data flow is shown in Fig. 8. These events can be discarded with a product cut of Thetriggersignalisfurthermoresenttothegermanium the form DAQ recording it as a redundant but immediate veto (x −ρ )(x −ρ )<c (1) information for the germanium data stream. 1 1 2 2 The scintillator panels are arranged in three layers for the pulse heights x of any pair of panels in the 1,2 of12panelseach.ThreeadditionalFADCsofthesame stack and constants ρ and c. The 2-D pulse height 1,2 type digitize the signal of the scintillator veto. The sig- distributionoftwopanelsinastack,theappliedcutand nals of two non-neighboring panels within a layer are a spectrum of a single panel with a fit to the Landau multiplexedontooneFADCchannelusingcustommade peak is shown in Fig. 7. reflection-free modules with an amplification of -6 dB. Thusthe36panelsoccupyonly18FADCchannelssuch that each layer is read out by one FADC module. The 2.4 Data acquisition panel stack which was hit can be determined by the uniquecombinationoffiredchannels.Thesametrigger The entire apparatus is read out and operated by a logicasfortheCherenkovvetoisappliedforthepanels, VME data acquisition system (DAQ) which is almost albeitthetriggerwindowislargerinordertoaccommo- identicaltotheoneofthegermaniumDAQ(seeRef.[1] datethemuchlongeroutputsignalsofthepanelPMTs for details). Ten Flash-ADCs [29] with 8 channels each because they are shaped with a larger time constant. digitize the input signals of the Cherenkov veto with For a panel event, all three FADCs (i.e. all three pan- 100 MHz. The signals in each channel are processed els in a stack) need to have fired. Both types of trigger by a trapezoidal filter and if the height exceeds the signals are accepted during data taking. threshold set to 0.5 photo-electrons (p.e.) an internal triggerisgenerated.EachFADCmodulehasonetrigger For a calibration run of the PMTs, the standard output which is the logic OR of the internal triggers of datatakingisstopped.Theultra-fastcalibrationLEDs itseightchannels.Thus,thePMTsignalsareconnected are activated with a pulser and the LED luminosity is to the input channels in such a way over the FADCs, controlled by a current source in form of a digital-to- thatneighboringPMTsarealwaysreadoutbydifferent analog converter [30]. A separate FADC reads out the FADCs. This allows the proper detection of clustered pulsersignalandtriggerstheentirevetoforeachpulse. 6 nts [cts] 103 high iSlluPmPi n=a=t io1n p:.eG.ERDA 15/04 T response [a.u.]1105////// PPPMMMTTT 576000111 GERDA 15/04//////1105 ve SPP @ 1.2 p.e. PM // PMT 402 // e102 DPP @ 2.2 p.e. 5// PMT 301 //5 // PMT 201 // PMT 101 0 0 Apr/11 Jul/11 Oct/11 Jan/12 Apr/12 Jul/12 Oct/12 Dec/12 Apr/13 10 Fig. 11 Summary of the stability of the light output of selectedPMTs;notethearbitraryoffsets.Thehatchedareas indicatemaintenanceperiods 1 0 100 200 300 400 500 FADC channel July 2013, 805.6 live days have been recorded and 491 Fig.9 CherenkovPMTresponsewithdifferentforwardvolt- days of combined muon-germanium data. The duty cy- ages of the calibration LEDs, i.e. luminosities. With a low cle is shown in Fig. 10 together with the accumulated luminosity, only the single photon peak (SPP, broken green livetime(redline).DuringPhaseIthemuonDAQwas line) is visible, if the luminosity is too high, the double pho- tonpeak(DPP,blue)emergesaswell.Thepedestalisshown onlystoppedduringbreaksofthegermaniumdatatak- inred ing in order to perform short maintenance work and to calibrate the PMTs by adjusting the HV of each PMT so that each module shows the same response to sin- TwocalibrationspectracanbeseeninFig.9.Oneis gle photons. The PMTs were very stable and since the showingtheconventionallyrecordedsingle-photonpeak beginning only few PMTs needed to be readjusted. As (SPP) set to channel 100, the other comes from a too example the daily light output per day is shown for brightLEDsetting.Thiscausesacontaminationofthe selected seven PMTs in Fig. 11. The offsets are for dis- SPP by the double photon peak (DPP) and a shift of play only. For e.g. PMT101 there was no readjustment theamplitudesofbothpeakstohighervalues.TheSPP necessary since May 2011, while PMT301 needed sev- emergesverywellinmostPMTsandpeak-to-valleyra- eral tunings of the HV. However, between the breaks tios between 1.2 and 3.0 are observed. thelightoutputremainsstable.Duringthegermanium datatakingthemuonvetowasalwaysfullyoperational. During Phase I two PMTs were lost due to implo- 3 Veto performance sion of the tube (PMTs 401 & 604). The implosion was mostly contained by the encapsulation and no other The Gerda muon veto was installed in 2009 and its PMTsinthevicinitywereharmed.Theimplosionshap- operationstartedinNovember2010.Duringthegerma- pened in February and April 2012 and the PMTs were nium commissioning runs the panel veto was installed over 10 m apart. Hence a direct influence can be ruled so that the complete veto was operational at the start out.AthirdPMTwaslostrightafterinstallationdueto of the physics runs of Gerda in November 2011. Until apuncturedcable(PMT305)andafourthPMTceased working during the installation and could be immedi- ately exchanged (PMT 203). In July 2013, the Gerda e 1.0 GERDA 15/04 800s water tank was drained, the veto inspected and two of o duty cycl0.8 670000live day t4h0e1s)e.APlMloTtshewrePreMsTucscsetsisllfuwlloyrkexacshianntegnedde(dPaMnTdssh3o0w5e&d vet0.6 500 little to no signs of deterioration. 400 0.4 300 3.1 Simulation studies 0.2 un 25 un 28 un 31 un 34 un 37 un 40 un 43 un 46 120000 r r r r r r r r The performance of the muon veto was simulated with 0.0 0 Jul/11 Jan/12 Jul/12 Dec/12 Jul/13 theGeant4-based[31]frameworkforGerdaandMa- Fig. 10 Duty cycle of the muon veto. The veto uptime jorana (MaGe) [32]. First, the simulations were used (black) and accumulated live days (red) are marked as well to find initial placements and efficiencies [7] and re- as the Gerda physics runs for Phase I (filled light red and peatedoncetheexactgeometriesoftheapparatuswere green) finalized[9].ThemuonspectraprovidedbytheMacro 7 experimentwereusedasinputforthesimulationofcos- 102 103 104 mogenic muons [5]. The simulations were mainly used a.u.] 105 cosmGicER DmA 15/04 105 to determine the efficiency of the apparatus in case the s [ pillbox muon caused any energy deposition in the germanium vent104 simulation 104 e crystals. 103 103 For the efficiency of the Cherenkov veto the sim- 102 102 ulation was undertaken in two parts. Firstly, muons were simulated with the Cherenkov effect switched off. a.u.] 103 102 103 deetd. eligph. t m[P.E.1] 04103 Theprimaryverticesofthosemuonsthatcausedenergy nts [102 pillbox 102 e simulation deposition in the germanium detectors were extracted ev 10 10 from these events. Secondly, these selected muons were 1 1 used in a second simulation with Cherenkov effect en- 10-1 10-1 abled. This two-step procedure was applied in order to 102 103 det. light [p.e.] 1 04 accelerate the simulation as only a minute fraction of Fig. 12 Photo-electron spectra for all cosmogenic muons the muons interact with the germanium detectors and (top)andthosewithenergydepositioninthegermaniumde- becausethesimulationofopticalphotonsisaresource- tectors (bottom). In each panel the total recorded p.e. spec- trum (red) is compared to the spectrum which is recorded demanding procedure. A detection efficiency for the just in the pillbox (green). Spectra derived from simulations veto was derived by determining the fraction of energy (blue)arenormalizedtothe sameexposure depositing muons which caused a trigger signal in the muon veto. The trigger condition were the same as for themuonvetoDAQdescribedinSec.2.4.Fortheentire 3.2 Multiplicity and photon spectra veto a detection efficiency of The light yield of a single muon is determined by ob- εMC =(99.935±0.015)% (2) µd serving the total number of recorded p.e. in all PMTs. For comparison, the pillbox is treated as an individual for muons with energy deposition in the germanium volume. In Fig. 12 histograms of the recorded and sim- detectors was found in the simulated data. ulatedeventsareshownfortheperiodofPhaseI.Inthe ByremovingcertainPMTsfromtheefficiencycalcu- spectra for the cosmogenic muons, apart from the very lations, a veto degradation (e.g. possibly broken PMTs lowlightyieldtherearemaximaatabout167p.e.inthe or malfunctioning FADCs) was simulated. Even with pillbox data and about 605 p.e. for the total spectrum. the first two FADCs removed (14 PMTs in total, two These broad peaks correspond to the mean traversed in each of the seven rings shown in Fig. 4), the effi- distance of 1.8 m for the pillbox and of 9 m for the wa- ciency is still (99.525+0.025)%. However when only four ter tank for muons with a mean incident angle of 60◦. −0.035 PMTs in the pillbox are removed the value drops to Light from muons in the water tank is subject to at- (97.855±0.065)%.ThepillboxPMTscanhencebecon- tenuation, the attenuation length of photons in water sidered the most critical ones in case of a break-down. being ∼10 m. Assuming a mean distance of 5 m of the The light produced inside the pillbox can illumi- PMTsfromthemuontrackinthewatertankand2min nate the main water tank through two small manholes. the pillbox, muons of any track deposit approximately However, the light coming from these two holes is not the same amount of light. Thus, each muon generates sufficient in order to generate a trigger. The insensitiv- (115±39)p.e./m.Thisisreproducedbythesimulation. ity against a loss of a few PMTs gives high reliability A peak structure is visible in the data for muons of the efficiency of the veto against small variations of which deposit energy in the germanium detectors as trigger conditions or changes in the amplitude of the well. The peak for the pillbox is at slightly lower p.e. PMTs. valuesbecausetheaverageincidentzenithangleislower In an earlier work the efficiency was estimated as andthusthetracklengthshorter.Thep.e.spectrumfor εMC = (99.56±0.42)% and thus some what lower de- all PMTs shows a double peak feature. This is due to µd spite lower trigger conditions [7]. The previous simula- the fact that the muon has to cross the cryostat in or- tion neglected the optical connection between the “pill- dertodepositenergyinthegermanium.Themuoncan box” volume and the main water tank completely. Sev- deposit light in the water tank before and after inter- eral other simulation studies have been performed like acting with the germanium detectors. The higher peak the veto response to regular cosmogenic muons which correspondstomuonswhichpassthetanktwiceandthe are shown and compared to experimental data in the smaller corresponds to muons that pass the water tank next sections. just once (e.g. shallow angles close to the neck of the 8 u.] 106 m events, 18 P.E. cut GERDA 15/04 3.3 Coincident muon-germanium events s [a.105 panel trigger The muon veto and the germanium systems have been ent sim. events, FADC cut operational in common during Phase I of Gerda over v e104 a time of 491 d. During this period an exposure of E = (21.6 + 6.2 ) kg·yr of germanium data was 103 enr nat recorded. During Phase I the muon veto was only shut 102 down during pauses in the germanium data taking and hence there is no additional loss of exposure due to the 10 veto. 0 10 20 30 40 50 60 PMT multiplicity M The two data streams were correlated by using the Fig. 13 PMTmultiplicityspectrumoftheCherenkovveto. timestamps of the events. Prior to Phase I both DAQ Themultiplicityofallevents(green)iscomparedwithevents systems were operated with their own internal clock where the panels have fired as well (red) and with simulated whichpermittedundesiredjumpsinthetimeoffsetbe- data(blue).Histogramsaftercutsaremarkedbydashedlines tween the two systems. For Phase I both DAQ systems were equipped with the same GPS timing system so that events can be correlated via the timestamp with high precision. The length of the germanium trace of cryostat). Again, the simulations agree with this even 160 µs is taken as a coincidence window. Most interac- though the double peak structure is less pronounced. tionsbetweenmuonsandthegermaniumarrayhappen within ±10 µs, however delayed interactions cannot be Another characteristic of a muon event is the num- excluded. The germanium DAQ is described in detail ber of fired PMTs. This multiplicity M is shown in in Ref. [1]. Fig. 13 for several classes of muon events. The spec- Bothsystemswerephysicallyandelectronicallyvery trum of all measured events (green line) shows a peak stableovertime.AfterthedeploymentofthenewBEGe at M>60 and another one at M<10. The peak at high detectorsinthesecondhalfofPhaseItheset-upofop- M is the regular response of the veto to muons. This erational modules was unchanged for the rest of the is verified by either the simulated data (blue line) and data taking with four BEGes, six enrGe and one natGe a subset of all events in which the panel veto has to coaxialdetector.Inthisperiodthemeanratewasr = µ havefiredaswell(redline).Theshapeofthespectrais (4.01±0.04)×10−2/s for the veto (no cuts applied), slightly different which is due to different incident an- r = (2.87±0.06)×10−2/s for all germanium detec- Ge gles (panel trigger) or the lack of random coincidences torsandr =(9.5±0.6)×10−5/stherateofphysical coin at low multiplicities in the simulated data. In addition, coincidences. Due to the low rate of veto and germa- four PMTs were lost in the Cherenkov veto which en- nium random coincidences are negligible compared to hances the peak at M>60. This is due to events that the true coincident rate. would trigger all PMTs are now recorded as having an M=(66−x),wherexistheamountoflostPMTs.Thus, the counts at or just below 66 are shifted to lower mul- 3.4 Muon rejection efficiency tiplicities. A muon rejection efficiency (MRE) can be obtained by Only the measured spectrum from the Cherenkov definingacutforclearlyidentifiedmuonhitsintheger- trigger shows the low multiplicity enhancement, which manium detectors and testing for coincident veto sig- ischaracterizedbynotonlyalownumberoffiredPMTs nals. The rejection efficiency ε is given as the ratio µr butalsoanunusuallylowamountofrecordedp.e.With of these events which are vetoed in comparison to the acutof18p.e.thisenhancementcanbealmostentirely entire set. The following cuts were applied to the ger- suppressed. This is the standard cut condition for the manium events of the Gerda Phase I data to identify Cherenkovvetodata.Thesourceofthisenhancementis muons:eitherasinglehitshowedanenergydepostionof discussed in Sec. 3.7. A cut which emulates the trigger morethan8.5Mevorthesummedenergyofamulti-hit conditionimplementedintheDAQdescribedinSec.2.4 exceeded 4 MeV. This cut excluded energy depositions (“FADC cut”) was applied to the simulated data. The from the U and Th decay chains. In addition, the ger- resulting spectrum (dashed blue line) indicates the be- manium test-pulse and quality cuts were applied, but havioroftheCherenkovvetowithoutunphysicalevents no muon veto cuts. Out of the 848 candidate muon atlowM likerandomcoincidencesorthelowmultiplic- eventsidentifiedaccordingtoenergyreleaseandmulti- ity enhancement. plicityinthe Germaniumdetectors,841areactuallyin 9 coincidence with a valid muon veto signal. This leads 3.6 Panel detection efficiency to a MRE of: Inordertodeterminetheefficiencyofthepanels,adata ε =(99.2+0.3)% (3) µr −0.4 sample of clearly identified muons was selected. A cut This is a conservative number since the studied events on the Cherenkov events of M ≥20 was chosen which are not standard events either in germanium multiplic- discards unphysical events at low M. The pre-selection ity or in energy range where saturation might set in. of muon events by the Cherenkov veto is necessary be- The derived MRE is slightly lower in comparison causethestandardtriggeroftheplasticvetocannotbe to the efficiency derived from the simulation given in used. For the panels a simple cut on the pulse heights Eq. 2. Assuming that the given MRE can be projected is insufficient due to the γ coincidences at low pulse to standard events, i.e. with multiplicity m = 1 and heights (see Fig. 7). In addition to the pulse height cut withenergiesatoraroundtheROI,therejectionpower definedbythetriggerthreshold,acutontheproductof of the veto is reliably high and close to unity. twopulseheightsintheformofEq.1wasapplied.The detection efficiency of one panel can now be given as the ratio of events, in which the top and bottom panel 3.5 Muonic background index of a stack have fired in comparison to the events where all three panels in a stack have fired. Inordertoestimatetheimprovementofthebackground index given in cts/(keV·kg·yr) due to the muon veto, Thedatasetcontainedeventssincethebeginningof a ±100 keV window was chosen around Q . Analog the muon data taking with the panels in August 2011. ββ to the germanium background a blinding window of Inthisset,30044eventswerefoundwhichtriggeredthe ±20 keV around Q was omitted from the analysis topandbottompanel.Oftheseevents29951triggered ββ hence the ROI of this study is 160 keV wide. Out of the third panel as well. This leads to an average muon a total exposure of E=27.8 kg·yr of germanium data, detection efficiency per panel of: 14 vetoed events were found in this ROI with a ger- εP =(99.70+0.03)% (7) manium multiplicity of one. Were these 14 events not µd −0.05 vetoed,theywouldhaveledtoacontributiontotheBI Since this value is an average over different panels, it of: can be seen as an approximation for a general panel BI =(3.16±0.85)×10−3 cts/(keV·kg·yr) . (4) efficiency. The efficiency of a triple stack of panels is µ hence (εP )3 =(99.10+0.09)%. Due to insensitive areas A simulated value for the muonic background in the µd −0.15 at the panel borders (scintillator edges, encasing) the germaniumarraysurvivinganti-coincidencecutsis[33] effective area of the panels is reduced by ∼5 mm per border or less than <0.25% of the area. BI (MC)=(1.6±0.1)×10−3 cts/(keV·kg·yr) . (5) µ As this simulation was undertaken before construction of the experiment was finalized the geometry differs to 3.7 Low multiplicity enhancement what was realized. Due to the different geometry, the low statistics and the subsequent large errors, both re- The enhancement at low multiplicities is characterized sults can be considered sufficiently in agreement. by a very low number of recorded photo-electrons (in WithBI andthepreviouslyderivedMRE,anesti- most cases one or two p.e. per PMT) without any ob- µ mationoftheunvetoedbackgroundcontributioncanbe servableclusteringeffectsandamountstoabout8.7%of given. It is assumed that the MRE is constant over the the overall Cherenkov activity (i.e. 3.1×10−3/s). This entire energy range of the germanium detectors. The behavior suggests a very faint source of light inside the given vetoed BI is equivalent to the amount of success- watertankthatisnotdirectlycausedbymuons.Events fully vetoed muons, i.e. 99.2%. An amount of unvetoed which trigger the panel veto and can hence be consid- muons is found: ered true muon events do not show this anomaly. It was already suggested that this anomaly is caused by BI =(2.87±0.77)×10−5cts/(keV·kg·yr) . (6) µ,unvet. scintillation of the reflective foil under irradiation by α The design goal of Phase II of Gerda aims to reach a sourcesinthestainlesssteelofthewatertank[1,8].The totalBIof10−3 cts/(keV·kg·yr).Thus,withthecurrent foilhasahighlyreflectivefrontsideandiscoveredwith settings of the muon veto unchanged, unvetoed muons an adhesive on the back side. According to the manu- wouldcontribute1/40oftheBI“allowance”.ForPhaseI facturer the foil itself has a thickness of 66 µm and the of Gerdathisisequivalentto0.16eventsina200keV adhesive a thickness of 38 µm [24]. If the adhesive can analysis window. be assumed to be an organic compound a mean free 10 path of ∼30 µm for a 5 MeV α particle is expected. the solid angle (8.5% of 4π) and efficiency (0.25) of the Hence it is unlikely that α particles coming from the PMT used for this test, the efficiency (0.3) and sur- backsideareabletodepositenergyinthefoilandpro- face coverage (0.005) of the PMTs in the Gerda water ducescintillationphotonswhichcanexitthefoilonthe tank the effect of this scintillation can be calculated. It frontside.Thiswastestedbyilluminatingthefoilwith isfoundthatthebulkoftheseeventswoulddeposit2-6 an 241Am α source and by recording the scintillation p.e. in the PMTs per event. These events will on aver- light on the front side with a 9235Q PMT. When the age not fulfill the trigger conditions given in Sec. 2.4 It foil was illuminated from the back side, i.e. through is still likely that about 1% of these events could still the adhesive, almost no additional light was recorded. triggertheveto.Thiswouldputtheexpectedandmea- Whentheadhesivewasremovedfromthebacksidethe sured rate as well as the expected photon yield of one samemeasurementyieldedasmalleffect.Ifilluminated totwop.e.perPMTintothesameorderofmagnitude. onthefrontsidesufficientphotonswererecordedwhich Thus, the scintillation caused by β particles has to be could explain the enhancement if applied to the condi- considered the most likely source of this enhancement. tionsinthewatertank.However,thissuggeststhatthe α source is either solved in the ultra-pure water or ad- heringtothefrontsideofthefoil.Theactivityofwater 4 Summary fromthisplantwasmeasuredtohaveanoverallαactiv- ity in the range of 10−6–10−7 Bq/kg [23] and measure- In this work, the muon veto deployed during Phase I ments of Gerda water samples agree with these val- oftheGerdaexperimentwasintroduced,itshardware was presented and its performance was shown. In ad- ues[34].Thisactivityistoolowtoexplaintheseexcess dition, the cosmogenic components of the background events. If the front side of the foil had a radon contam- in Gerda was systematically identified, analyzed and ination, a higher rate would have been expected after compared to Monte Carlo simulations. theopeningofthewatertank.Ahigherratewasindeed Thehardwarethresholdswerechoseninaway,that measured but this was in accordance to a higher dark by design a very pure muon sample is recorded with rateafterprolongedexposuretolightoftheCherenkov only a few percent contamination by random coinci- PMTs. dencesorothersourcesofbackground.Withthispower- Another explanation for the origin of this anomaly fulmuonvetoovertwoyearsofdatahavebeenrecorded are β sources since electrons have much higher specific including the 491 days coincident with the germanium ranges in comparison to α particles. The stainless steel of the Gerda water tank exhibits a low level of ra- detectors. The Monte Carlo simulations of earlier works were dioactivityandoneofthestrongestcontributiontothe radionuclides in the steel is the β emitter 60Co with extendedtoaccommodateforamorerealisticset-upof an activity of ∼20 mBq/kg [35] which leads to a 60Co themuonveto.PMTmultiplicityandp.e.spectrawere found in good agreement with the data and the light activity on the surface of the water tank of ∼5 Bq. To testtheeffectthefoilwasilluminatedbya60Co-source deposition of cosmogenic and energy-depositing muons could be related to their different track-lengths in the (Q =0.35 MeV) and the results are shown in Fig. 14. β water tank. With the updated geometry a detection The foil was illuminated from the front and from the back side (adhesive not removed). The back illumina- tion shows only a slightly smaller scintillation effect in s GERDA 15/04 comparison to the front illumination and in both spec- ent104 back illumination v tra a low-energy β spectrum emerges. In order to de- e front illumination termine an efficiency for this process, the scintillation 103 single photon rate of a 5 mm thick sheet of plastic scintillator was recorded that is assumed to have an efficiency of unity. 102 By comparing the rate of the foil with the rate of the scintillator, the efficiency of the foil towards 0.35 MeV 10 β particles can be calculated: 1 εfoil =(12.0+1.1)%. (8) 0 10 20 30 40 50 60 70 80 β −1.0 detected light [p.e.] With this efficiency, the activity expected from 60Co Fig. 14 The effect of irradiation of the VM2000 reflective foil by a β source (60Co) recorded by a PMT. The single from the steel is reduced to ∼0.6/s which is still too photonresponse(red)iscomparedtoanirradiationfromthe high in comparison to the measured rate. Using the front(blue)orbackside (green)whichcarriestheadhesive meanlightrecordedbytheilluminationmeasurements,