PreprinttypesetinJINSTstyle-HYPERVERSION Gator: a low-background counting facility at the Gran Sasso Underground Laboratory 2 L. Baudisa∗, A.D. Ferellaa, A. Askina, J. Angleb, E. Aprilec, T. Brucha, A.Kisha, M. 1 0 Laubensteind, A. Manalaysaya, T. Marrodán Undagoitiaa and M. Schumanna 2 aPhysicsInstitute, n a UniversityofZürich,Winterthurerstrasse190,CH-8057,Zürich,Switzerland J bDepartmentofPhysics, 0 UniversityofFlorida,Gainesville,FL32611,USA 1 cDepartmentofPhysics, ] ColumbiaUniversity,NewYork,NY10027,USA M cGranSassoNationalLaboratory, I Assergi,L’Aquila,67010,Italy . h E-mail: [email protected] p - o tr ABSTRACT: Alow-backgroundgermaniumspectrometerhasbeeninstalledandisbeingoperated s a in an ultra-low background shield (the Gator facility) at the Gran Sasso underground laboratory [ in Italy (LNGS). With an integrated rate of ∼0.16 events/min in the energy range between 100- 2 2700keV, the background is comparable to those of the world’s most sensitive germanium detec- v tors. After a detailed description of the facility, its background sources as well as the calibration 5 2 and efficiency measurements are introduced. Two independent analysis methods are described 1 and compared using examples from selected sample measurements. The Gator facility is used to 2 . screenmaterialsforXENON,GERDA,andinthecontextofnext-generationastroparticlephysics 3 0 facilitiessuchasDARWIN. 1 1 : KEYWORDS: HPGespectrometers;lowlevelγ-rayspectrometry. v i X r a ∗Correspondingauthor. Contents 1. Introduction 1 2. TheGatorfacility 2 3. Calibrationmeasurementsandefficiencydetermination 5 4. Dataanalysismethods 6 4.1 Analysisofgamma-lines 7 4.2 Fitofthedatatoasimulatedspectrum 8 5. Results 9 5.1 BackgroundAnalysis 9 5.2 SampleAnalysis 10 6. Summary 12 7. Acknowledgements 14 1. Introduction Gamma-ray spectroscopy offers a standard method for material screening and selection for rare- eventsearches,suchasthedirectdetectionofWeaklyInteractingMassiveParticles(WIMPs)[1]or thesearchfortheneutrinolessdoublebetadecay[2]. Anultra-lowlevelgermaniumspectrometerin adedicatedlow-backgroundshield(theGatorfacility)hasbeenbuiltandinstalledattheLaboratori NazionalidelGranSasso(LNGS),Italy. Whilethefacilityisbeingoperatedmainlyinthecontext oftheXENONprogram[3][4][5][6],recentlyitalsohasbeenusedfortheGERDAexperiment[7] aswellasforR&Dpurposeswithinlow-backgroundparticleastrophysics. Thispaperisorganizedasfollows: inthenextsection,theGatorfacilityisdescribedindetail. In the third section, calibration measurements, efficiency determinations for the different sample geometries,andcross-checkswithstandard,calibratedsamplesaredescribed. Inthefourthsection, two different data analysis methods are introduced. The fifth section describes the determination ofthemainbackgroundsourcesofthefacilityaswellasselectedresultsforsamplemeasurements. Inthefinalsection,asummaryanddiscussionoffutureplansarepresented. –1– 2. TheGatorfacility Thedesignofthefacilityhasbeeninspiredbythelayoutoftheworld’smostsensitivegermanium spectrometers,operatedatLNGSinconnectionwiththeBorexinoandGERDAexperiments[8][9]. The core consists of a high-purity, p-type coaxial germanium (HPGe) detector with 2.2kg of sen- sitivemasswitharelativeefficiencyof100.5%1. Thedetectorconstructionhasbeenperformedin close cooperation with Canberra semiconductors[11], using only materials with ultra-low intrin- sicradioactivecontamination. Thecryostatismadeofultra-lowactivity,oxygen-freecopperwith the cooling provided by a copper dipstick (’cold finger’) in thermal contact with a liquid nitrogen bath. The cryostat is of U-type, with the cooling rod shaped in a right angle below the cryostat to avoiddirectline-of-sighttotheoutside(seeFigure1). Whilethefield-effecttransistor(FET)used for the charge readout is cooled and close to the detector, the preamplifier is placed outside the low-backgroundshield,sinceitcontainsmoreradioactivecomponents. The shield of the detector has been designed to provide a large sample capacity, an ultra-low background and easy access to the germanium spectrometer itself. The sample chamber, with a dimension of 25×25×33cm3, is surrounded by 5cm (7cm for the base plate) of oxygen-free, radio-pure copper from Norddeutsche Affinerie[12]. Residual surface contaminations of the cop- per plates were removed by treating them with diluted sulfuric and citric acid solutions, followed by cleaning with deionized water. All steps were performed under clean-room conditions. The copper is surrounded by four layers of lead from Plombum[13], each 5cm thick. The inner 5cm lead layer has a nominal 210Pb activity of 3Bq/kg, while the outer 3 layers were built from lead with a higher 210Pb activity of 75Bq/kg. All lead bricks were cleaned with ethanol before being installedintheshield. Theirarrangementissuchthatnodirectline-of-sighttotheHPGedetector is possible. Large copper plates, which close the sample cavity and carry the upper lead bricks, were placed on sliding rails allowing for easy opening and closing of the shield: The full sample chambercanbeaccessedwhenthetopshieldisopen. Theleadissurroundedby5cmofpolyethy- leneasashieldagainstambientneutrons,andtheentireshieldisenclosedinanairtightaluminum boxinordertopreventradonenteringthesystem. Figure1showsaschematicviewofthedetectorandshieldconfiguration. Asamplehandling and glove box made of plexiglass, including an airlock system, is placed on top of the aluminum housing. The entire system is continuously flushed at slight over-pressure with boil-off nitrogen gas to suppress radon diffusion into the shield. A 5mm inner diameter PTFE tube allows sealed calibrationsourcestobebroughtclosetothegermaniumdetector. Thesamplestobescreenedare first cleaned in an ultra-sonic bath of ethanol (where applicable), then enclosed in a sealed plastic bag for transportation to the underground site, after which they are stored for a few days under nitrogenatmosphereabovetheclosedchamber. Thisallowstrappedradonandplate-out220Rnand 222Rnprogeniestodecaybeforetheactualcountingstarts. Toremotelycheckthestabilityofthedetectorwithtime,amonitorsystemhasbeeninstalled. Theliquidnitrogenlevel,theflowofnitrogengas,theleakagecurrentacrossthegermaniumdiode andthehigh-voltageappliedtothediodearereadevery2minutes,whiletheoveralltriggerrateis readevery6hours. Theliquidnitrogenlevelismeasuredwithalevel-meterconsistingoftwo40cm 1Thequotedefficiencyisdefinedrelativetoa7.62cm×7.62cmNaI(Tl)crystal,forthe1.33MeV60Cophoto-peak, atasource-detectordistanceof25cm[10]. –2– f e d a b c g Figure1. SchematicfigureoftheGatorfacilityatLNGS.TheHPGedetector(a)withitscoldfinger(b)and dewar(c),theopensamplechamber,thecopper(d)andlead(e)shieldwiththeslidingdoor,theglovebox (f)andpolyethyleneshield(g)canbeseen. long concentric metal tubes acting as a capacitor whose capacitance is read out with a universal transducer interface board. The nitrogen gas flow into the shield is monitored and regulated with an electronic flowmeter and the leakage current is measured as the voltage drop across the feed- backresistorandreadoutwithananalog-to-digitalconverter. Therelevantparametersareplotted versustimeandcanbemonitoredonawebinterfacewhichisrefreshedevery10minutes. Incase pre-defined thresholds for these parameters are not met, email and SMS alarms are being issued. Thedataacquisitionconsistsofahigh-voltageunit,aspectroscopyamplifier,andaaselftriggering, 16kchannelmultichannelanalyzer. Thespectra, acquiredandclearedevery6hours, aresavedin asciifilesandanalyzedoffline. Prior to its installation at LNGS, the germanium spectrometer was operated in the Soudan undergroundlaboratoryinnorthernMinnesota,withintheSOLOfacility[14]. AttheSoudanlabo- ratory,thedetectorwasusedforthescreeningofXENON10[3]materials,andseveralbackground runswereacquired[15]. Duringitswater-andground-basedtransportationtoLNGS,thedetector was exposed for several months to the cosmic ray flux at the Earth’s surface, leading to cosmic activation of the crystal and the surrounding copper of the cryostat. The shield was improved in October2008,anditwascleanedonceagaininFebruary2009. Table1showstheintegralbackgroundinthe100-2700keVregionintheSoudanconfiguration, andforthreemeasurementstakenatLNGS.TheintegralcountingrateatLNGShasbeenconstantly decreasing due the decay of cosmogenic radio-nuclides such as 54Mn, 57Co, 58Co and 65Zn, with typical half-lives around one year on the one hand, and due to an improved shield and overall –3– sealing of the system on the other hand. This is reflected in the decrease of prominent lines from 214Biand214Pb,asshowninTable2. Itgivestheintegralcountingratesinthe±3σ-regionsforthe mainprimordialγ-lines,aswellasforthe137Cs,60Coand40Klines,alongwithacomparisonwith theGeMPIdetector,whichisoneoftheworld’smostsensitivelow-backgroundspectrometers[8]. Table1. IntegralbackgroundcountingratesforGatorasmeasuredatSoudanandatLNGSinthreedifferent runs. Theintegralisevaluatedintheenergyrange[100,2700]keV. Run Lifetime[days] Rate[events/min] GatoratSoudan 22.96 0.842±0.005 GatoratLNGS(09-2007) 14.90 0.258±0.003 GatoratLNGS(10-2008) 22.59 0.186±0.003 GatoratLNGS(04-2010) 51.43 0.157±0.001 Table 2. Background counting rates (in events/day) in the ±3σ-regions for the main primordial and the gammalinesof137Cs,60Coand40K. Energy Chain/nuclide Peakintegralbackgroundrate[counts/day] [keV] Gator Gator Gator Gator GeMPI[8] (Soudan) (LNGS, (LNGS, (LNGS, 09-2007) 10-2008) 04-2010) 239 232Th/212Pb 1.1±0.7 0.7±0.1 0.13±0.08 <0.5 NA 911 232Th/228Ac 0.9±0.3 0.4±0.2 0.4±0.1 <0.5 <0.2 352 238U/214Pb 4.9±0.7 4.3±0.7 1.1±0.2 0.7±0.3 <0.5 609 238U/214Bi 4.5±0.5 4.0±0.5 1.1±0.2 0.6±0.2 0.50±0.45 1120 238U/214Bi 1.6±0.3 2.7±0.4 1.3±0.2 0.3±0.1 NA 1765 238U/214Bi 1.3±0.2 1.5±0.3 0.2±0.1 0.08±0.06 NA 662 137Cs 2.9±0.4 0.5±0.3 <0.5 0.3±0.1 NA 1173 60Co 0.5±0.1 0.5±0.2 0.5±0.2 0.5±0.1 0.6±0.4 1332 60Co 0.6±0.1 0.6±0.2 0.5±0.1 0.5±0.1 0.4±0.3 1461 40K 5.8±0.4 0.5±0.2 0.4±0.1 0.5±0.1 0.6±0.4 2615 208Tl 0.7±0.1 0.2±0.1 0.2±0.1 0.2±0.1 NA Figure 2 (top) shows the comparison between the latest background spectrum acquired at LNGS(2010),aspectrumtakenintheSOLOfacilityatSoudan(2007)andthebackgroundofthe GeMPIdetector [8]. Italsoshows(bottom)theGatorspectrumundergroundatLNGSpriortoits installation in the shield, inside the shield, and inside the shield with the radon protection system on. The background decrease is more than four orders of magnitude, and shows that a careful shieldingisneededevenwhenthedetectorisoperateddeepunderground. –4– ]100 1 ! V e k !1 10!1 d 1 ! g k e [10!2 at R 10!3 500 1000 1500 2000 2500 Energy [keV] 105 104 ] 1 !V103 e 1 k102 ! 1 d101 ! g100 k [ e 10!1 at R10!2 10!3 200 400 600 800 1000 1200 1400 1600 Energy [keV] Figure 2. (Top) Background spectra of Gator at Soudan (red), at LNGS (black) and the spectrum of theGeMPIdetector [8](blue). (Bottom)GatorbackgroundspectrumatLNGS:outsidetheshield(black), insidetheshield(blue)andinsidetheshieldwiththeradonprotectionsystemon(red),clearlyshowingthe suppressionofthemaingammalinesassociatedwithradondecays. 3. Calibrationmeasurementsandefficiencydetermination The HPGe detector is calibrated regularly with radioactive sources such as 109Cd, 133Ba, 137Cs, 60Co,57Co,22Na,54Mnand228Th. InFigure3(left)thecomparisonofthespectrumobtainedfrom a 60Co calibration with the one from a Monte Carlo simulation of the source-detector geometry is shown. The FWHM of the two 60Co lines at 1173keV and 1332keV are 2.5keV and 3.0keV, respectively. The energy resolution of the detector, defined here as the ratio of the σ to the mean energyofthegammaline,isshowninFigure3(right). Thecalculationofthedetectionefficienciesofthevariousgammalinesusedintheanalysisof theexperimentaldata(seeSection4)isbasedonMonteCarlosimulationsusingGeant4[16]. For each measured sample, a detailed geometry is included into a Geant4 model of the facility. The efficiency ε of a specific γ-line is defined as the ratio between the number of events detected in the line to the number of gammas of that energy emitted by the source. In order to simulate each decay chain, the G4RadioactiveDecay class, which takes into account the branching ratios for the differentgammalinesinonedecay,isused. –5– 0.01 103 0.008 gy)0.006 er n nts102 / E!0.004 Cou on ( 101 oluti0.002 s e R 0.001 100 0 500 1000 1500 2000 2500 102 103 Energy [keV] Energy [keV] Figure 3. (Left) Comparison of the Gator spectrum from a 60Co calibration (black, data points) with the onefromaMonteCarlosimulationofthesource-detectorgeometry(red,solid). (Right)Energyresolution (defined here as the ratio of σ/energy) as a function of energy. The solid curves represent a fit using the functionσ2(E)=E2(2.35×10−7)+E(7.70×10−4)+(4.43×10−1). To cross-check the efficiency determination, a measurement of two extended sources and a comparison with the certified values for their activities has been performed. The sources used for this measurements, which had similar dimensions and weights, are CANMET-STSD2 (from the Canada Centre for Mineral and Energy Technology) and IAEA-Soil6 (from the International Atomic Energy Agency). Both sources are soils from different places on Earth, which have been thermally treated and sieved several times in order to destroy any remaining organic matter. The homogeneityofthematerialiscertifiedbytheprovider. Tables3showstheresultsGator’sscreen- ing of these two sources and a comparison with the certified values. Within the uncertainties of the measurements, we find a good agreement. We also show in Figure 3 the comparison of the efficiencies as determined by the Monte Carlo method, with those from the data, as a function of energy,aswellastherelativedifferenceamongthese. These results indicate that the measurements performed with our spectrometer provide a re- liable value for the activity of a given sample. Further cross-check were done by using well- calibrated,commerciallyavailablepointsources. 4. Dataanalysismethods Two methods are used to determine the concentrations of radioactive nuclides in a given sample. In the first method, the most prominent γ-lines are analyzed, using efficiencies as determined by a full Monte Carlo simulation. In the case of 238U, the γ-lines from the daughters of 226Ra (214Pb and214Bi)andincaseof232Ththeγ-linesfrom228Acandfrom212Pb,212Biand208Tlareused. In thesecondmethod,theoveralldataspectrumiscomparedtotheoneobtainedfromaMonteCarlo simulation,aftersubtractingthemeasuredbackgroundspectrum,andtheactivitiesaredetermined fromthebestfit. Theresultsfromthetwomethodsagreewithinthestatisticalerrors, asshownin Section5. Bothmethodsareexplainedinmoredetailbelow. –6– Table3. ResultsfromscreeningtheCANMET-STSD2(493g)andIAEASoil6(530g)sourcesandcom- parisonwithcertifiedvaluesprovidedbytwoagencies(fordetails,seetext). STSD2Activity[Bq/kg] Nuclide Gatorresults Certifiedvalues 228Th 75±4 70±5 226Ra 230±30 230±10 40K 590±10 540±20 Soil6Activity[Bq/kg] Nuclide Gatorresults Certifiedvalues 226Ra 88±5 80±7 137Cs 57±2 54±2 0.07 STSD2 Soil6 Data 0.06 ! Monte Carlo y, c n0.05 e ci Effi0.04 n o cti0.03 e et D 0.02 0.01 D ! ) / D 0.2 ! ! 0 MC!0.2 ! ( 0 500 1000 1500 2000 2500 Energy [keV] Figure4. Upperpanel:Efficienciesasafunctionofenergyforthetwocertifiedsources,asdeterminedfrom data(filledsymbols)andMonteCarlo(opensymbols). Lowerpanel: relativedifferencebetweensimulated (MC)andmeasured(D)efficiencies. 4.1 Analysisofgamma-lines The first method is based on counting events at the location of the most prominent lines, after subtracting the background spectrum closest in time. The Compton background, estimated from the regions left and right of a peak, is subtracted as well. The decision on whether the signal exceedsthebackgroundisbasedoncomparingthenetsignalnumberofcounts S =S−B·t /t −B (4.1) net S B C withtheso-calleddetectionlimitL (thelevelofatruenetsignalthatcanbedetectedwithagiven d –7– probability)asdefinedin[18]fora∼95%C.L.: (cid:114) t S L =2.86+4.78 B +B· +1.36. (4.2) d C t B Sisthenumberofcountsinthe±3σ-regionaroundapeak,BandB arethenumberofbackground C and Compton-background counts in the same region, andt ,t are the measuring times for signal S B andbackground,respectively. Foreachpeak,threecasesareconsidered[18]: 1. S <0: theupperlimitissettoL (nonetcontributionfromasignal) net d 2. 0<S <L : theupperlimitissettoS +L (thereisanindicationofasignal,butitcan net d net d notbeconfirmedfortheexistingbackgroundlevelandsampleexposure) 3. S >L : thedetectionlimitisexceeded(clearindicationforasignalat95%C.L.) net d Forthethirdcase,thespecificactivityandits1σ erroriscalculatedas S ∆A ∆S net net A[Bq/kg]= with = . (4.3) r·ε·m·t A S net withthepeakefficienciesε asdeterminedbyMonteCarlosimulations,thebranchingratiorforthe specificline,themassmofthesample(inkg),andthemeasuringtimet (inseconds). Forthecase in which an upper limit is reported, S is replaced by L or by (S +L ) in equation (4.3). As net d net d concreteexamples,weshowtheabovequantitiesinTable4,togetherwiththedeterminedspecific activitiesorupperlimitsforacopperandastainlesssteelsample,usingdifferentgammalinesfrom theirmeasuredspectrum. Table4. Examplesofupperlimitorspecificactivitiescalculation. Detailsaregiveninthetext. Sample Usedline[keV] S B·t /t B L Condition Activity[mBq/kg] net S B c d Copper 239 0 0 93 49 S <L <0.33 net d 352 -5 19 71 48 S <L <0.36 net d 1173 42 6 19 27 S >L 0.24±0.06 net d Stainlesssteel 352 66 7 58 42 S >L 4.3±0.9 net d 1173 236 2 19 25 S >L 7.2±0.9 net d 1461 -3 3 10 21 S <L <5.7 net d 4.2 Fitofthedatatoasimulatedspectrum The individual activities are also determined by a global fit based on a χ2-minimization of the simulatedspectrumtotheexperimentaldata. Theaimistomodel(cid:126)y,themeasurednumberofcounts ineachenergybin,asafunctionof(cid:126)x,theenergybinvaluefromtheMonteCarlosimulation,with afunctionaldependenceoftheform M (cid:126)y= ∑a · f ((cid:126)x), (4.4) k k k=0 –8– where M is the number of simulated radioactive isotopes. a ≥0 are the scaling factors for each k isotope,whicharekeptasfreeparameters,and f ((cid:126)x)areM MonteCarlospectrawhicharealready k smearedwiththeenergyresolutionofthedetector. Foruncorrelatedstatisticalerrorsineachenergy bini,theχ2 valueisdefinedas χ2= ∑N (cid:2)yi−∑Mk=0ak· fk(xi)(cid:3)2, (4.5) σ2 i=0 i √ whereσ = y isthevarianceintheobservednumberofcountsineachbinandN isthenumber i i of energy bins over which the fit is performed. The statistical uncertainty in the Monte Carlo component is negligible. Before the analysis, the measured background spectrum is subtracted from the sample spectrum. The outcome of this procedure are the scaling factors a ±∆a for k k every decay chain/isotope k. In the subsequent analysis, these factors are assumed to describe a normalizedGaussianprobabilitydistributionfunction(PDF)withmeana =a andσ =∆a . 0 k k The method applied to decide between a real detection and an upper limit is based on [17]. If the lower of the two symmetric limits providing 95% statistical coverage is positive, a peak detectionisclaimedandtheactivityiscalculatedfroma ,takingintoaccountthemeasuringtime k t inseconds,thesamplemassminkg,andthenumberofsimulatedeventsn . Therelative±1σ sim errorisgivenby∆a /a : k k n a ∆A ∆a sim k k k A [Bq/kg]= with = . (4.6) k mt A a k k Ifthelowerlimitisequalorlessthanzero,nodetectioncanbeclaimedat95%C.L.andanupper limitisgiven. Itsvaluea isdeterminedbythe95%quantileofthepositivepartoftheGaussian up PDFdefinedbya =a andσ =∆a . Thelimitontheactivity(95%C.L.)iscalculatedusinga 0 k k up inequation(4.6). Asanexample,Figure4.2showsameasurementofastainlesssteelsample: thedataspectrum iscomparedwiththeMonteCarlosumspectrum,andtheindividualcontributionsfrom238U,232Th, 40K, and 60Co, as given by the best-fit, are shown. The derived activities and upper limits for this samplearegiveninTable6. 5. Results In this section, we first outline results obtained from a detailed study of the background of the facility, using the analysis method described in Section 4.2. We then present screening results for a few selected samples, the goal being to compare the outcome of the two analysis methods introduced in the previous section. Results from a much larger selection of screened samples are givenanddiscussedindetailin[20]. 5.1 BackgroundAnalysis TomodeltheresidualbackgroundoftheGatorfacility, thedetailedgeometryofthecrystal, cryo- stat system and shields has been simulated with Geant4. The following potential contributions to the background have been simulated: the natural decay chains of 238U, 232Th and 40K decays in the copper of the shield and of the cryostat, the decays of the cosmogenic radio-nuclides 54Mn, –9–