Low-Background gamma counting at the Kimballton Underground Research Facility
P. Finnerty, S. MacMullin, H.O. Back, R. Henning, A. Long, K.T. Macon, J. Strain, R.M. Lindstrom, R.B. Vogelaar
aa r X i v : . [ nu c l - e x ] F e b Low-background gamma counting at the Kimballton Underground Research Facility
P. Finnerty,
1, 2, ∗ S. MacMullin,
1, 2
H.O. Back,
3, 2
R. Henning,
1, 2
A. Long, K.T. Macon, J. Strain,
1, 2
R.M. Lindstrom, and R.B. Vogelaar Department of Physics and Astronomy, University of North Carolina, Chapel Hill, NC, USA Triangle Universities Nuclear Laboratory, Durham, NC, USA Department of Physics, North Carolina State University, Raleigh, NC, USA National Institute of Standards and Technology, Gaithersburg, MD, USA Department of Physics, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA (Dated: July 2, 2018)The next generation of low-background physics experiments will require the use of materials withunprecedented radio-purity. A gamma-counting facility at the Kimballton Underground ResearchFacility (KURF) has been commissioned to perform initial screening of materials for radioactivityprimarily from nuclides in the
U and
Th decay chains, K and cosmic-ray induced isotopes.The facility consists of two commercial low-background high purity germanium (HPGe) detectors.A continuum background reduction better than a factor of 10 was achieved by going underground.This paper describes the facility, detector systems, analysis techniques and selected assay results.
PACS numbers: 07.85.Fv, 29.40.-n, 07.85.Nc, 29.30.KvKeywords: Gamma spectroscopy; Low-background
I. INTRODUCTION
We have commissioned a low-background gamma-counting facility at the Kimballton Underground Re-search Facility (KURF). KURF is located at LhoistNorth America’s Kimballton mine in Ripplemead, Vir-ginia. The experimental hall is located on the mine’s 14 th level at a depth of 1450 m.w.e (meters of water equivalentshielding). The overburden consists of 520 m of dolomite,limestone and other sedimentary rock. Experiments arehoused in a 30 m ×
11 m laboratory building that wascompleted in October 2007 (Fig. 1). The laboratory’sgeneral infrastructure is maintained primarily by collab-orators at Virginia Polytechnic Institute and State Uni-versity. KURF has an office, air filtration, power, water,phone and ethernet. KURF also has the advantage ofdrive in access, making it simple to transport personneland equipment to the experimental hall. Liquid nitro-gen (LN ) used in the laboratory is stored in a 2.4 m portable dewar that can be transported to the surface tobe refilled as necessary. Radon levels in the laboratoryhave been found to vary from 37 Bq/m in the winter to122 Bq/m in the summer. The detectors used in thispaper are housed in sealed modified shipping containers(MSCs) within the laboratory building (Fig. 1). II. DETECTORS & SHIELDING
The counting facility consists of two high purity ger-manium (HPGe) detectors specifically designed for low-background assay work. The first detector, named “VT-1”, is a commercial ORTEC LLB (very low-background) ∗ Corresponding author, E-mail: [email protected]
Series coaxial detector [1]; the high voltage filter andpreamplifier are removed from the line of sight of thecrystal to reduce backgrounds from radioactive contami-nants that may exist in detector components. The samplecavity is cylindrical, 41 cm (height) ×
28 cm (diameter).Further specifications for VT-1 are in Table I. A VT-1background spectrum taken on the surface and a back-ground spectrum taken underground is shown in Fig. 2.The second detector, named “M elissa ,” is a CanberraLB (low-background) coaxial detector [2]. M elissa is ina vertical orientation with a dipstick style cryostat. Thepreamplifier is also removed from the cryostat, allowingfor shielding to be placed in between the preamplifier andcryostat. The sample cavity is 38 cm ×
38 cm ×
38 cm.Further specifications for
Melissa are in Table I.A
Melissa background spectrum is shown in Fig. 3.No active shielding is currently used for either detector.The sample cavities of both detectors are continuouslypurged with LN boil-off to flush out radon. A reductionof ∼
80% in the activity of radon daughters was observedafter introducing the LN boil-off, however activity dueto radon remains as a dominant source of background.The shield design of the detectors has made it difficult todesign a hermetically sealed radon exclusion system.Integral count rates for these background spectra(Figs. 2 and 3) are shown in Table II. While direct com-parisons with other underground gamma-counting facili-ties are difficult, integral count rates for other selected fa-cilities can be found in [3–6]. Although the integral countrate in the 40-2700 keV region is higher than for otherdetectors at a similar depth, most of our background is atlow energy. An overview of low-radioactivity backgroundtechniques and a comparison of low-level counting meth-ods can be found in [7, 8] and references therein.FIG. 1: KURF before and after construction of the laboratory building. Modified shipping containers (MSCs) areshown in the right panel (Color online).TABLE I: Detector specifications.M elissa VT-1Manufacturer Canberra ORTECRelative Efficiency 50 % 35%Performance FWHM at 1.33 MeV (keV) 1.70 1.80Threshold (keV) 20 20Shield Lead thickness (cm) 15.2 10.1Oxygen-free high conductivity(OFHC) copper thickness (cm) 2.54 0.3Crystal (coaxial) Mass (kg) 1.1 0.956Length (mm) 64.5 75.7Diameter (mm) 65 55.8Hole diameter (mm) 7.5 9.1Hole depth (mm) 50 63.2Outer electrode thickness (mm) 1.06 0.7Inner electrode thickness ( µ m) 0.3 0.3Cryostat End cap diameter (mm) 82 70End cap thickness (mm) 4.2 1.3End cap to crystal (mm) 5 4End cap material high purity Al MgIR window material Al Mylar TM , Kapton Al Mylar TM TABLE II: Integral count rates for M elissa and VT-1in [10 counts/day].Melissa VT-1 VT-1 (surface)40-2700 keV 7.8 7.6 8440-1000 keV 7.5 6.5 681000-2700 keV 0.27 0.71 16 III. SAMPLE PREPARATION
To achieve the required assay sensitivity for next gen-eration low-background experiments, unwanted radionu-clides must be removed from sample surfaces to preparefor gamma-counting. Samples are prepared for assay ina cleanroom environment using ultra-pure reagents andclean plastics that have been screened for radioactivity.Depending on the sample material and required detectionlimit, a variety of methods can be used to treat samplesurfaces. These methods include acid leaching for plas-tics, acid etching for metals or cleaning with ultra-puresolvents. After cleaning, samples are bagged in nylon to
Energy [keV] ] - d - k g - C oun t R a t e [ ke V -1 Pb K Bi Bi Tl
511 keV
FIG. 2: VT-1 surface (black/top) and underground(red/bottom) comparison. Both spectra were takenwith the lead shield in place (Color online). Theunderground spectrum was taken with a LN boil-offpurge in place Energy [keV] ] - d - k g - C oun t R a t e [ ke V -1 Pb Tl Bi Bi K Bi Tl FIG. 3: Background spectrum for the M elissa detector. The spectrum was taken with lead shieldingand a LN boil-off purge in place (Color online).prevent any recontamination of surfaces. IV. ANALYSIS TECHNIQUESA. Monte Carlo simulations and efficiencycalculations
The full energy peak (FEP) detection efficiency, de-fined as the ratio of the number of events detected in thegamma peak to the number of events emitted from thesource for a particular radioactive isotope in a sample,depends on many factors, including the crystal, cryo-stat, shielding and source geometries. There are cur-rently several methods to determine the FEP detectionefficiency. One is to use analytical calculations [9], how-ever this technique is limited to simple geometries andrequires complex calculations. In some cases, a physicalmodel of the sample can be created using known stan- dards [10]. This process is complicated and time con-suming and is of limited accuracy for complex geometries.Another method is to use a point source calibration atrepresentative points determined by Monte Carlo simu-lation. The efficiency curve generated is then correctedfor absorption by a sample matrix and sample container[11].At KURF, a detailed Monte Carlo simulation for eachsample is done using
MaGe [12].
MaGe is a
Geant4 -based [13, 14] simulation package maintained and devel-oped by a joint group of the M ajorana [15–17] andG erda [18] collaborations. Once a detailed sample ge-ometry has been coded into the simulation, it is uniformlydoped with isotopes of interest from the U, Th, and K decay chains, along with any other isotopes that maybe present in the sample, e.g. Co. The primary γ -raysfrom these decays are tracked from the emission at thesource to absorption in the detector active region. Byusing a pure Monte Carlo simulation to determine theFEP detection efficiency, self-attenuation in the sampleis accounted for and there are no limitations on sourceor detector configurations.The simulated spectrum is then convolved with the fi-nite energy resolution of the detector. The energy resolu-tion for both detectors has been measured as a function of γ -ray energy from 303–1836 keV using radioactive pointsources.The peak area in the simulated spectrum is deter-mined by fitting the peak of interest with a Gaussianand subtracting a linear background. The FEP detec-tion efficiency, including branching ratios, can then bedetermined from ǫ γ = peak areanumber of events simulated (1) B. Monte Carlo validation
It is our goal to know the efficiency of any volumesample to <
10% and attribute <
10% systematic un-certainty to these efficiencies. In order to validate theMonte Carlo FEP detection efficiency calculations, pointsources of well-known activity were used to baseline thesimulation using a method similar to [19]. Once the de-tector geometries are well understood, the Monte Carlocan be used to simulate all relevant physical processesand accurately determine peak efficiencies.Experimental data were taken for each detector usingpoint sources emitting γ -rays with energies ranging from303-1836 keV. The activity of each source was known towithin 3%. To understand the effects of source place-ment relative to the crystal, the sources were placed inthe locations shown in Fig. 4 for VT-1 (the same methodwas used for M elissa ). Enough experimental data weretaken for each source placement to minimize uncertain-ties from counting statistics to < elissa , the ratio of the experimental to sim-ulated efficiencies is 0.96. A constant correction factor toaccount for this difference is applied when determiningthe activity of a sample. C. Activity Calculations
With the FEP detection efficiency well understood, wecan precisely determine or set an upper limit on the ac-tivity of a sample. First, a sample spectrum is comparedto the background. The background spectrum is takenclose to the time of the assay to limit the effects of tem-poral variation in background, such as seasonal changesin radon activity. A Gaussian plus a linear backgroundfunction is fit to each peak used for analysis. The fit isused to obtain the centroid of the peak and define thelimits of the peak. The peak area is then extracted bysumming the counts in a ± σ width region for both the Energy [keV] D a t a / S i m u l a t i on Above crystalBelow crystal
FIG. 5: Experimental and simulated efficiency in theVT-1 detector. Multiple data points at each γ –rayenergy correspond to a different placement of the sourcein the detector. Energy [keV] D a t a / S i m u l a t i on Above crystalBelow crystal
FIG. 6: Experimental and simulated efficiency in theM elissa detector. Multiple data points at each γ –rayenergy correspond to a different placement of the sourcein the detector.sample and background spectra. Subtraction of the con-tinuum is done using one of two methods: (1) Integratethe linear part of the fit extrapolated under the peakand subtract, (2) Define two regions, one 5 σ to the leftand the other 5 σ to the right of the peak, average thecounts in the two regions, and subtract it from the to-tal integrated peak area. If another peak is within 5 σ to the left or right, the background region with no peakpresent is used rather than the average (See Fig. 7 forillustration). We prefer method (2) since this does notrely on the goodness-of-fit. The net peak area is foundby subtracting the background peak area from the cor-responding sample peak.The activity of a specific peak is then given by: A γ = net peak area [counts/second] ǫ γ m (2)where ǫ γ is the peak efficiency, and m is the mass of thesample. Energy [arb] C oun t R a t e [ a r b ] σ σ σ σ FIG. 7: Generic Gaussian peak as an illustration ofintegration limits. The peak is fit with a linear plus aGaussian function. The limit of the Gaussian is ± σ (dashed) and the linear background (solid) isextrapolated under the peak.If no statistically significant peak is present, an upperlimit can be placed on the activity of a sample. Theactivity at a given energy is required to be less thanor equal to 1.64 ×√ background counts (90% confidencelevel (C.L.)).Once activities for individual peaks are calculated,they must be related to the overall U and
Th con-tent. In many cases secular equilibrium cannot be as-sumed, so activities or upper limits are considered sep-arately for
U (from
Th, m Pa) and
Ra (from
Ra,
Pb,
Bi) in the
U decay chain and
Ra(from
Ac) and
Th (from
Pb,
Tl) in the
Thdecay chain. The presence of radon gas, which decays to
Bi and
Pb, makes the measurement of
Ra diffi-cult, so measures are taken to eliminate as much radonas possible inside the sample cavities through the use ofLN boil-off. D. Estimating detection limits A MaGe / Geant4 simulation was performed to esti-mate the detection limits, defined as the net signal levelthat may be expected to lead to a detection [20], for ageneric sample placed in a Marinelli beaker, modeled by aTeflon R (cid:13) ring of dimensions 14.1 cm (OD) × × U and
Th decay chains and K were calculated forM elissa and VT-1 using the best-achieved backgroundspectrum for each detector. We define a detectable signalby Eq. (3), where the background counts are determinedfrom experimental background spectra and signal countsare the excess expected from the simulations of the sam-ple. Counts within ± σ width of the signal-peak energyare included. signal counts ≥ . × p background counts (3) The MaGe / Geant4 simulation determines the effi-ciency for detecting a decay, including geometric effectsand branching ratios, so that the number of signal countsin a detector can be related to the source activity: signal counts = ( source activity rate ) × ( counting live time ) × ǫ γ (4)The sensitivity is determined by combining Eqs. (3)and (4), and solving for the detectable activity rate. Thisgives the sensitivity, S : S = 1 . √ background counts ( counting live time ) × ǫ γ (5)The sensitivity results are shown in Table III. V. ASSAY RESULTS
Table IV shows the sample assay results to date sinceMay 2008. Samples are listed in the order that they wereassayed. Activity limits for later assays were improvedas a result of progress made with background reductionand detector performance. Disequilibrium in natural un-closed systems, such as plants, soil and rock, is commonas observed in the Table Mountain rock sample. For thesample of aluminum stock flange coupling, disequilibriumwas observed in the
U and
Th decay chains. Thedecrease in activity of the signature
Ra daughters havebeen observed in previous analyses of both pure and alu-minum alloys. This can be explained through differencesin the chemistries of U, Th, Ra and Pb, as could occurthrough the steps required to recover aluminum from itsore [21]. For lead samples,
Pb ( T / = 22 . · y ), activ-ity was measured separately since it is not expected to bein equilibrium with the rest of the U chain. The leadsample from Sullivan Metals had less than 2.5 Bq/kg of
Pb activity. Lead bricks of unknown origin acquiredfrom the University of Washington showed no measur-able
Pb signatures, but an upper limit of 10 Bq/kgwas placed on the
Pb activity.
VI. CONCLUSIONS & REMARKS
A gamma-counting facility has been commissioned atKURF. The background signal rates for the M elissa andVT-1 detectors have been pushed to low levels. This wasaccomplished by building the facility in an undergroundlocation and using passive shielding, radio-pure detectorcomponents, and radon mitigation techniques. We havesuccessfully demonstrated the analysis procedures andassay sensitivities required for screening materials for thenext generation of low-background experiment.
VII. ACKNOWLEDGEMENTS
This worked was primarily support by NSF grant , 167 (2004).[4] D. Budj´aˇs et al., AIP Conf. Proc. , 26 (2007).[5] D. Budj´aˇs et al., in The Proceedings of the 14th Inter-national Baksan School “Particles and Cosmology-2007” (Institute for Nuclear Research of the Russian Academyof Sciences, Moscow 117312, Russia, 2008), pp. 228–232,ISBN 978-5-94274-055-9, arXiv:0812.0723v1 [physics.ins-det].[6] D. Budj´aˇs et al., in
The Proceedings of the 14th Inter-national Baksan School “Particles and Cosmology-2007” (Institute for Nuclear Research of the Russian Academyof Sciences, Moscow 117312, Russia, 2008), pp. 233–238,ISBN 978-5-94274-055-9, arXiv:0812.0768v1 [physics.ins-det]. [7] G. Heusser, Annu. Rev. Nucl. Part. Sci. , 543 (1995).[8] M. Hult, Metrologia , S87 (2007).[9] T.-K. Wang et al., Appl. Radiat. Isot. , 933 (1995).[10] A. Smith, private communication (2009).[11] J. Saegusa et al., Appl. Radiat. Isot. , 1383 (2004).[12] M. Bauer et al., J. Phys. Conf. Ser. , 362 (2006).[13] S. Agostinelli et al. (GEANT4), Nucl. Instr. and Meth.A , 250 (2003), http://geant4.web.cern.ch/geant4/.[14] J. Allison et al., IEEE Trans. Nucl. Sci. , 270 (2006).[15] C. E. Aalseth et al. ( Majorana ), AIP Conf. Proc. ,88 (2009), arXiv:0907.1581 [nucl-ex].[16] S. R. Elliott et al., J. Phys. Conf. Ser. , 012007 (2009).[17] V. Guiseppe et al., in
Nuclear Science Symposium Con-ference Record, 2008. NSS ’08. IEEE (2008), pp. 1793–1798.[18] I. Abt et al. (2004), arXiv:0404039 [hep-ex].[19] D. Karamanis et al., Nucl. Instr. and Meth. A , 477(2002).[20] L. Currie, Anal. Chem. , 586 (1968).[21] A. Smith, private communication (2008). TABLE III: Detector sensitivities (90% C.L.) for M elissa and VT-1 for a Teflon R (cid:13) ring.Melissa VT-1Energy [keV] Isotope (Chain) counts/day Detection Limit counts/day Detection Limit[mBq/kg] [mBq/kg]63 Th (
U) 81 ± ± Th (
U) 96 ± ± m Pa (
U) 2.7 ± ± Ra (
U) 145 ± ± Pb (
U) 117 ± ± Pb (
U) 114 ± ± Bi (
U) 59 ± ± Bi (
U) 13 ± ± Bi (
U) 9.2 ± ± Ac (
Th) 65 ± ± Ac (
Th) 5.7 ± ± Pb (
Th) 133 ± ± Tl (
Th) 18 ± ± Tl (
Th) 2.9 ± ± K ( K) 19 ± ± Th - activity measured in the
U decay series from
Th and m Pa.
Ra -activity measured in the
U decay series from
Ra,
Pb and
Bi.
Ra - activity measured in the
Thdecay series from
Ac.
Th - activity measured in the
Th decay series from
Pb and
Tl. Activities from
Tl are divided by the branching ratio (35.94%). Not measured = NM.Sample Detector
Th [Bq/kg]
Ra [Bq/kg]
Ra [Bq/kg]
Th [Bq/kg] K [Bq/kg] Co [Bq/kg]Table Mountainrock (latite) M elissa
NM 100 ±
40 100 ±
40 270 ±
120 790 ±
320 NMTable Mountainrock (latite) VT-1 NM 100 ±
40 100 ±
40 300 ±
120 730 ±
290 NMSuperinsulationpanels M elissa
NM 3.0 ± ± ± elissa ± ± < ± < elissa < ± ± ± ± < ± ± ± ± elissa ± ± ± ± ± TM foaminsulation M elissa NM < < < R (cid:13) VT-1 < < ± ± ± < elissa < < < < < elissa < < < < < < < < < < <<