Background modeling for dark matter search with 1.7 years of COSINE-100 data
G. Adhikari, P. Adhikari, E. Barbosa de Souza, N. Carlin, J.J. Choi, S. Choi, M. Djamal, A.C. Ezeribe, L.E. Franca, C. Ha, I.S. Hahn, E.J. Jeon, J.H. Jo, W.G. Kang, M. Kauer, G.S. Kim, H. Kim, H.J. Kim, K.W. Kim, N.Y. Kim, S.K. Kim, Y.D. Kim, Y.H. Kim, Y.J. Ko, V.A. Kudryavtsev, E.K. Lee, H.S. Lee, J. Lee, J.Y. Lee, M.H. Lee, S.H. Lee, D.S. Leonard, W.A. Lynch, B.B. Manzato, R.H. Maruyama, R.J. Neal, S.L. Olsen, H.K. Park, H.S. Park, K.S. Park, R.L.C. Pitta, H. Prihtiadi, S.J. Ra, C. Rott, K.A. Shin, A. Scarff, N.J.C. Spooner, W.G. Thompson, L. Yang, G.H. Yu
EEur. Phys. J. C manuscript No. (will be inserted by the editor)
Background modeling for dark matter search with 1.7 years ofCOSINE-100 data
G. Adhikari , P. Adhikari , E. Barbosa de Souza , N. Carlin ,J.J. Choi , S. Choi , M. Djamal , A.C. Ezeribe , L.E. Fran¸ca , C. Ha ,I.S. Hahn , E.J. Jeon a,7 , J.H. Jo , W.G. Kang , M. Kauer , G.S. Kim ,H. Kim , H.J. Kim , K.W. Kim , N.Y. Kim , S.K. Kim , Y.D. Kim ,Y.H. Kim , Y.J. Ko b,7 , V.A. Kudryavtsev , E.K. Lee , H.S. Lee ,J. Lee , J.Y. Lee , M.H. Lee , S.H. Lee , D.S. Leonard , W.A. Lynch ,B.B. Manzato , R.H. Maruyama , R.J. Neal , S.L. Olsen , H.K. Park ,H.S. Park , K.S. Park , R.L.C. Pitta , H. Prihtiadi , S.J. Ra , C. Rott ,K.A. Shin , A. Scarff , N.J.C. Spooner , W.G. Thompson , L. Yang ,G.H. Yu Department of Physics and Astronomy, Sejong University, Seoul 05006, Republic of Korea Wright Laboratory, Department of Physics, Yale University, New Haven, CT 06520, USA Physics Institute, University of S˜ao Paulo, 05508-090, S˜ao Paulo, Brazil Department of Physics and Astronomy, Seoul National University, Seoul 08826, Republic of Korea Department of Physics, Bandung Institute of Technology, Bandung 40132, Indonesia Department of Physics and Astronomy, University of Sheffield, Sheffield S3 7RH, United Kingdom Center for Underground Physics, Institute for Basic Science (IBS), Daejeon 34126, Republic of Korea Department of Science Education, Ewha Womans University, Seoul 03760, Republic of Korea Department of Physics and Wisconsin IceCube Particle Astrophysics Center, University of Wisconsin-Madison, Madison, WI53706, USA Department of Physics, Kyungpook National University, Daegu 41566, Republic of Korea IBS School, University of Science and Technology (UST), Daejeon 34113, Republic of Korea Korea Research Institute of Standards and Science, Daejeon 34113, Republic of Korea Department of Accelerator Science, Graduate School, Korea University, Sejong 30019, Republic of Korea Department of Physics, Sungkyunkwan University, Seoul 16419, Republic of Korea Department of Physics, University of California, San Diego, La Jolla, CA 92093, USAReceived: date / Accepted: date
Abstract
We present a background model for dark mat-ter searches using an array of NaI(Tl) crystals in theCOSINE-100 experiment that is located in the Yangyangunderground laboratory. The model includes backgroundcontributions from both internal and external sources,including cosmogenic radionuclides and surface
Pbcontamination. To improve the model in the low energyregion, with the threshold lowered to 1 keV, we used adepth profile of
Pb contamination in the surface of theNaI(Tl) crystals determined in a comparison betweenmeasured and simulated spectra. We also consideredthe effect of the energy scale errors propagated fromthe statistical uncertainties and the nonlinear detectorresponse at low energies. The 1.7 years COSINE-100data taken between October 21, 2016 and July 18, 2018were used for this analysis. The Geant4 toolkit version10.4.2 was utilized throughout the Monte Carlo simula- a e-mail: [email protected] b e-mail: [email protected] tions for the possible internal and external origins. Inparticular, the version provides a non-Gaussian peakaround 50 keV originating from beta decays of Pb ina good agreement with the measured background. Thisimproved model estimates that the activities of
Pband H are the dominant sources of the backgroundswith an average level of 2.73 ± · years. COSINE-100 is an NaI-based experiment for the di-rect detection of dark matter particles [1,2], with anarray of 106 kg NaI(Tl) crystals. It has been operatingat the Yangyang underground laboratory (Y2L) sinceSeptember 2016 [3,4,5]. One of the COSINE-100 goalsis to test DAMA/LIBRA’s assertion of an observationof annual modulation signal with a statistical signifi- a r X i v : . [ a s t r o - ph . I M ] J a n cance that is now more than 12.9 σ in the energy region(2–6) keV [6,7,8]. There are several groups, such asDM-Ice [9,10], ANAIS [11,12], KamLAND-PICO [13],SABRE [14], and COSINUS [15], developing ultra-low-background NaI(Tl) crystals with the goal of reproduc-ing the DAMA/LIBRA results. The DAMA/LIBRAcollaboration claimed that the improved configurationlowers the energy threshold and reinforces the annualmodulation signature at 9.5 σ C.L. in the energy regionof 1–6 keV [8].To verify the DAMA/LIBRA modulation signal, acomplete understanding of the background energy spec-trum is required. We have developed a backgroundmodel by performing Monte Carlo simulations usingthe Geant4 toolkit (V.10.4.2) [16]. We used 1.7 years ofCOSINE-100 data taken from October 21, 2016 to July18, 2018 with a 106 kg array of low background NaI(Tl)crystals. We used the measured spectrum obtained witha newly developed threshold of 1 keV electron equivalentenergy [17]. To build a complete background model withthe energy threshold as low as 1 keV, we precisely in-vestigated the low-energy contribution from the surface
Pb contamination in the NaI(Tl) crystals [18], in ad-dition to background simulations of internal radioactivecontaminants, such as natural radioisotopes and cosmo-genically activated isotopes inside NaI(Tl) and externalbackground sources from the exterior of crystals. More-over, an adjustment coefficient was applied to matchthe data and Monte Carlo (MC) simulations, taking theeffect of the energy scale errors into consideration forthe simulated spectrum at low energy.
The main detector of COSINE-100 is a 106 kg arrayof eight ultra-pure NaI(Tl) crystals (named as C1-C8)stacked in two layers. Each crystal is equipped with two3–inch Hamamatsu R12669SEL photomultiplier tubes(PMTs). The NaI array is immersed in a 2200-liter liquidscintillator (LS) that serves both as an active veto anda passive shield. Four shielding layers exist comprisingplastic scintillator panels, a lead-brick castle, a copperbox, and a tank of liquid scintillator. The experimentalsetup is described in detail in Ref. [3]. All the usedmaterials are properly assigned while constructing thegeometry used for the simulations, as shown in Fig. 1.We present here the background modeling to repre-sent the initial 1.7 years of COSINE-100 data named as“SET2 data”. Three crystals C1, C5, and C8 showed highbackground rates, twice as high as the other crystals.Also, C1 has a high noise rate and C5 and C8 havelow light yields. All three crystals are excluded fromthis analysis. The total exposure used in this analysis is 97.7 kg · years with a total mass of 61.3 kg from theother five crystals.To determine the light characteristics of the crystalsand the liquid scintillator (LS), including light yields,energy scales, and energy resolutions, an energy calibra-tion has been performed by tracking internal β – and γ -ray peaks from radioactive contaminants in the crys-tals, as well as γ -ray sources. Internal background peaksat 49 keV from Pb, 238 keV from
Pb, 295 keVand 352 keV from
Pb, 1462 keV from K, 2614 keVfrom
Tl, 609, 1764, and 2204 keV from
Bi andthe external background peak at 1173 keV from Coare used for the high-energy calibration. A nonlineardetector response of NaI(Tl) crystals in the low energyregion, as reported in Ref. [19], is studied and an em-pirical function is modeled to describe this nonlinearity.The charge-to-energy ratios of the calibration points forfive NaI(Tl) crystals describe the nonlinear detector re-sponse that is fitted by an empirical function, as shownin Fig. 2; the peaks at 0.9 keV from Na, 3.2 keV from K, 25.5 keV from
Cd, 30.5 keV from
Te, 49 keVfrom
Pb, and 67.8 keV from
I are used for thelow-energy calibration.The event selection criteria for the 1 keV energythreshold [17] are applied in this analysis. Events classi-fied as single-hit are for the ones that show more than4 photoelectrons in only one of the crystals and none inany of the other crystals or the LS. Multiple-hit eventsare those with the LS signal and/or more than 4 photo-electrons in other crystals. The LS veto threshold wasset at 20 keV for both single-hit and multiple-hit events.The PMTs are configured to generate two readouts.Both signals from the high-gain anode and the low-gain5th-stage dynode are stored as independent channels forall the events. The two readouts have different energyresolutions. The anode signals are used for low energyevents with less than a boundary energy of 70 keV, whilethe 5th-stage dynode signals are used for those largerthan the boundary.
The COSINE-100 simulation framework developed forthe background spectra of the first 59.5-day data fromthe COSINE-100 detector (“SET1 data”) [20] is usedfor the background simulations in this analysis. It is aGeant4-based simulation framework with a later version(10.4.2) than the Geant4 version 9.6.2 that was used inthe previous analysis. The newer version better describesthe non-Gaussian peak around 50 keV from the betadecay of
Pb to
Bi; only a small part, 4.3%, is dueto the 46.5 keV gamma-ray line and most of the eventsin the peak are from the conversion electrons, Auger (a) (b)
Fig. 1
Detector geometry (front view) used in the Geant4 simulation. (a) Two white-colored cylindrical shapes inside thecenter box represent the NaI(Tl) detectors supported by the acrylic frame (red) inside the liquid scintillator. (b) Eighteen5-inch PMTs are attached to two sides of the copper box to detect LS-produced photons.
Energy [keV] R e l a t i v e R a t i o Fig. 2
Detector response to the calibration data points isdescribed by an empirical function (red line) for five NaI(Tl)crystals at low energy. electrons, and X-rays, followed by beta electrons fromthe decay to
Bi, which results in a non-Gaussian peak.It was not well reproduced by the simulations using theGeant4 version 9.6.2.Each simulated event records all energies depositedin the crystals within an event window of 10 µ s fromthe time when a decay is generated, to account for theconditions in the data acquisition system (DAQ) of theexperimental setup [7]. Consecutive decays occurring ina short time, such as Bi–
Po decays with
Po’shalf-life of 300 ns may appear together in a 10 µ s timewindow, resulting in pileup events. They are treated as asingle event in the simulation. Based on this framework,we carried out Monte Carlo simulations for all the possi-ble background sources to build a complete model of thebackground measurements with 1 keV energy threshold. The simulated spectrum was convolved with anenergy-dependent energy resolution function developedduring the calibration.3.1 Internal and external backgroundsWith the the first 59.5 days of SET1 data, the detectorbackground was investigated with simulated backgroundspectra from the internal radioactive sources, such as fulldecay chains of U, Th, K, and
Pb inside theeight NaI(Tl) crystals assuming a chain equilibrium [3].However, the background level may vary over the timeif the internal activities are not in a chain equilibrium.We indeed found an evident increase in the
Th back-ground level during the 1.7 years of SET2 data whencompared with the SET1 result. To take broken chaineffects of the background sources into account for thebackground simulation, the
U and
Th decay chainsare treated as broken at the long-lived parts of the chain.The
U chain was broken into five distinct groups andthe
Th chain was broken into three groups. The ac-tivities of the
U and
Th decay chains in the eightgroups are treated as fitting parameters to quantify theunknown fractions of the background composition.We simulated external background sources in theCOSINE-100 experiment configuration: PMTs, greases,copper cases, bolts, cables, acrylic supports, liquid scin-tillator, copper box, and a steel structure that supportsthe lead block housing. Overall, the background spec-trum is well matched to the data in the high-energy re-gion, except for the peaks around 185 keV and 2.6 MeVin single-hit events. The background contribution dueto the
U chain from the PMTs was treated in twogroups as broken at the long-lived part of the chain and the 185.7 keV γ –ray from the decay of U in the firstgroup was helpful to improve the modeling between 100and 200 keV. This observation is similar to the one fromother group [21] that reported the existence of
U inthe same PMTs. The background peak around 2.6 MeVis well reproduced in multiple-hit events by the simula-tions while not in single-hit events for the same energyregion, which indicates that there could be an unknownexternal component located a short distance away fromthe NaI(Tl) crystals. The inclusion of γ –rays from thedecay of Tl from the exterior of the crystals improvedthe background model around 2.6 MeV in the single-hitevents; it is possible that there still exist backgroundcontributions from materials close to the detector.3.2 Cosmogenic radioisotopesTable 1 lists all the cosmogenic radioactive isotopesproduced in the NaI(Tl) crystals in COSINE-100, asreported in Ref. [22], with their half-lives and decaymodes; short-lived isotopes, for which half-lives are lessthan a year, are I, Te, m Te, m Te, m Te, m Te, and
Sn and long-lived isotopes are
Cd, Na, H, and
I. Since we use the 1.7-year data, theshort-lived (T / < H, Na, and
Cd, which have low energy depositsand are, therefore, potentially troublesome. The beta-decay spectrum of tritium has an endpoint energy of18 keV. The electron capture decay of Na produces0.87 keV emissions, and the electron capture decay of
Cd contributes peaks at 25.5 keV and around 3.5 keV
Table 1
Cosmogenic radionuclides in the NaI(Tl) crystalsidentified in other studies and considered here [22]. We listthe contributing cosmogenic isotopes with their half lives anddecay modes: β + , β − , electron capture (EC), and isomertransition (IT).Cosmogenic Half-life [23,24,25] Decay typeisotopes (days) & emissions energy I 59.4 EC, 35.5+31.7=67.2 keV
Te 19.17 EC, 4.1–4.7 and 30.5 keV m Te 164.2 EC, 4.1–4.7 and 30.5 keV m Te 119.3 IT, 247 keV m Te 57.4 IT, 145 keV m Te 106.1 IT, 88 keV
Sn 115.1 EC, 3.7–4.2 and 28 keV
Cd 462 EC, 25.5 and 88 keV Na 950 β + , 511 and 1274.6 keV H 4494 β − I 1.57 × yr β − which are at the binding energies of the Ag K-shell andL-shell electrons. It is thus essential to understand theirbackground contributions to the low energy spectra re-gions, especially in the (1–6) keV dark matter signalregion of interest (ROI). We, therefore, simulated back-grounds from cosmogenic radioactive isotopes, listed inTable 1. The simulated background spectra are used inthe data fitting, by floating their unknown fractions andthe fitted results are compared with the measurementsreported in Ref. [22]. The details of these comparisonsare discussed in Sect. 4.As the presence of cosmogenic I was introducedby DAMA/LIBRA with the estimated concentration of I/ nat I = (1.7 ± × − [26], we included it in ourbackground fitting model. Figure 3 shows the fitted simu- Fig. 3
Background modeling for low energy single hit events for Crystal 6. Background models not including
I (left) andincluding
I cosmogonic component (right) are shown. The solid green line corresponds to the
I isotope. The dashed blueline, the black dots, and the thick red line represent internal
Pb isotope, data, and the total MC, respectively.
Energy (keV) A r b i t r a r y U n i t Total Bi fi Pb Po fi Bi Fig. 4
Low-energy spectra due to the beta decays of
Pbthat are distributed within the surface thickness of 1 µ m inthe crystal. Further details are given in Sect. 3.3. lation spectra (a) not including of I and (b) includingof
I (green solid line). The simulated spectral shapeof the
I in Crystal 6, as shown in Fig 3(b), improvesthe background modeling around 30 to 70 keV. This isbecause the beta decay of
I to Xe ∗ is followed by Xe ∗ transitioning to the stable Xe isotope via theemission of a 39.6 keV γ –ray.3.3 Surface PbThe Q value of the beta decay of
Pb to
Bi is63.5 keV. The decay associates with low-energy emis-sions of electrons and γ /X–rays typically less than60 keV. The light outputs of the low-energy eventsdepend on the depth of Pb distribution on the crys-tal surface. It has also been suggested that sources areattributed to the
Rn contamination that occurredanytime during the powder- and/or crystal-processingstages. To understand the energy spectra from the betadecays of
Pb, we simulated them by generating
Pbat random locations within the surface thickness of 1 µ min the crystal. The simulated spectra are depicted inFig. 4, where each color represents the beta decays of Pb (dotted red line) and
Bi (dashed blue line),respectively. The peaks at approximately 10 and 50 keVare attributed to the X-rays and 46.5 keV emissions to-gether with about 4 keV mean energy of beta electronsfrom the decay to the
Bi metastable state. In addi-tion, the conversion electrons contribute to the peaks atapproximately 20 and 35 keV. These spectral featurescan be affected by the depth distribution of
Pb on
Fig. 5
Simulated energy spectra for the beta decays of
Pbwithin the surface of Crystal 7; we simulated the energy spectrafor beta decays of
Pb that are exponentially distributed inthe surface of Crystal 7 by following two exponential functionswith mean depths of 1.39 µ m (black solid line) and 0.107 µ m(red dashed line). the crystal surface and, therefore, the depth profile ofthe surface Pb contamination should be taken intoaccount for the detector background in the low-energyregion in particular for a low-background experimentusing NaI(Tl) crystals.We have studied the surface
Pb contaminationwith a test setup at Y2L using a NaI(Tl) crystal fromthe same ingot as C6 and C7. We measured its depthprofile by using the measured spectra from both betadecay of
Pb and alpha decay of
Po at the decaysequence of the surface
Pb contamination that is ob-tained using a
Rn-contaminated crystal, as reportedin Ref. [18]. Using this study, it was found that thelow-energy spectrum of the surface
Pb contamina-tion is primarily attributed to depth profiles of
Pbexponentially distributed within a shallow surface witha mean depth of (0.107 ± µ m, as well as a deepsurface with a mean depth of (1.39 ± µ m. We thussimulated energy spectra from beta decays of Pb thatare exponentially distributed in the surface by follow-ing two exponential functions with the mean values of1.39 µ m and 0.107 µ m for deep and shallow depths,respectively and the simulated spectra are used in thedata fitting. The ratio of the amplitudes of the expo-nential distributions is treated as a floating parameterbecause it could be affected by the Rn exposure. Fig-ure 5 shows the simulated energy spectra of the surface
Pb contamination that is weighted by an exponentialcurve as a function of the surface depth. The black solidline represents the background spectrum of
Pb withthe mean depth of 1.39 µ m and the red dashed linerepresents the background spectrum of Pb with themean depth of 0.107 µ m. Fig. 6
Top panel shows Crystal 6 energy spectra of single-hit events in the low energy region. Black dots are data and thered (blue) line shows the total MC without (with) applicationof the adjustment coefficient in the background modeling fit.In the bottom panel, the red (blue) dots are the ratio of data toMC without (with) application of the adjustment coefficient.
Pb contamination and long-lived cosmogenicisotopes. However, there is still a little mis-matchingbetween data and MC spectra at low energies, whichis because presumably the energy scales that are setseparately for the anode readout and the dynode read-out, based on linear fits of calibration data points, haveerrors propagated from the statistical uncertainty, aswell as the nonlinear detector response, as describedin Sect. 2. We, thus, consider an adjustment coefficientin the MC spectrum for the energy scale errors, usingthe method described in Sect. 3.4.1. Figure 6 shows theresults considering the adjustment coefficient (blue line)and without considering the adjustment coefficient (redline) in the background modeling fit. It is shown thatthe background modeling has been improved with theadjustment coefficient.
The energy scale set for the dynode readout is basedon the linear fit of calibration data points and, thus,the energy E in the MC spectrum corresponding to thescaled energy of the dynode readout is adjusted as E → E (1 + (cid:15) ) , (1)where (cid:15) is a coefficient that represents a change in energy.The i th bin content of the MC spectrum, B i , can be approximated as B i → B i + (cid:15) · ∂B i ∂(cid:15) (cid:12)(cid:12)(cid:12)(cid:12) (cid:15) =0 , (2)where we use a numerical approach to obtain the deriva-tive as ∂B i ∂(cid:15) (cid:12)(cid:12)(cid:12)(cid:12) (cid:15) =0 ≈ B ( E i (1 + δ(cid:15) )) − B ( E i (1 − δ(cid:15) ))2 δ(cid:15) , (3)where δ(cid:15) represents a very small change in (cid:15) and E i denotes the central value of the i th energy bin. A linearinterpolation of the MC spectrum is used for the smallvariation of δ(cid:15) .Since there is the nonlinear detector response mod-eled by the empirical function obtained in Sect. 2, atlow energies, adjusting the energy in the MC spectrumcorresponding to the scaled energy of the anode readoutfollows a different procedure from that of the dynodereadout, and is expressed by E → E [1 + (cid:15) · { f ( E ) − C } ] , (4) B i → B i + (cid:15) · { f ( E ) − C } · ∂B i ∂(cid:15) (cid:12)(cid:12)(cid:12)(cid:12) (cid:15) =0 , (5)where f ( E ) is the empirical function shown in Fig. 2and C is a coefficient for a linear component. In this analysis we used data collected between October21, 2016 and July 18, 2018 (“SET2 data”) with an energythreshold lowered to 1 keV. To model the measuredenergy spectrum ranged from 1 keV quantitatively, wehave performed Geant4 Monte Carlo simulations for thebackground spectra, as described in Sect. 3, which arefitted to the measured data to quantify their backgroundfractions.We use a binned likelihood method with the followingformula [27], − λ ( (cid:126)α ) = 2 N bins (cid:88) i =1 N components (cid:88) j =1 α j S ij − D i + D i ln D i (cid:80) N components j =1 α j S ij (cid:35) + N components (cid:88) j =1 (cid:18) α j − m j σ j (cid:19) , (6)where λ ( (cid:126)α ) is the likelihood ratio in terms of the frac-tions of the MC components (cid:126)α = ( α , α , · · · , α N components ), D i is the number of events in the i th energy bin of thedata histogram and S ij is the number of events in the i th bin of the j th simulation component. The last term Fig. 7
The energy spectra of single-hit (top) and multiple-hit (bottom) events in Crystal 7. The MC was carried out to fit themeasured data. The shaded area in the energy spectra of single-hit events is excluded from the data fitting. denotes a prior for the fraction α j of the j th componentand is only active if there is a pre-measurement of thiscomponent; m j and σ j are the measured value and theerror, respectively.As mentioned in Sect. 3.4, low- and high-energydata are taken through anode and dynode channels,respectively, and they have different energy resolutions.Thus, we perform a four-channel simultaneous fit: single-hit low-energy, single-hit high-energy, multiple-hit low-energy, and multiple-hit high-energy spectra. In thefour-channel fitting the energy ranges are set to be[6, 3000] and [1, 3000] keV for single-hit and multiple-hit events, respectively. The lower bound of the energyfor multiple-hit events is extended to 1 keV based onthe study of lowering the energy threshold, reported inRef. [17]. The lower bound for the single-hit events isset to 6 keV as not to bias the WIMP signal in the ROI.Figure 7 shows the measured and simulated back-ground spectra of Crystal 7 in both low and high energy regions. The spectra for single-hits and multiple-hits areshown in the top and the bottom, respectively. One cansee that the SET2 data is well reproduced overall exceptfor the energy region higher than ∼ Pb components, we used depth profilesof
Pb distributed within a shallow surface with themean depth of 0.107 µ m, as well as a deep surfacewith the mean depth of 1.39 µ m; their fractions areallowed to float. As a result of the fit, the surface Pbcontamination of three Crystals (2, 3, and 4) is primarily (a) (b)(c) (d)(e) (f)
Fig. 8
Comparison of the measured [3][20][22] and the fitted activity levels in five NaI(Tl) crystals. attributed to the deep depth profile and for Crystal 7 isprimarily attributed to the shallow depth profile. Thedepth distribution of
Pb for Crystal 6 consists of bothshallow and deep depth profiles. As studied in Ref. [18],surface
Pb contamination could be affected by
Rnexposure.eIn Fig. 8(a) and (b), we compared the fitted activi-ties of internal K and
Pb to their measured levels for the five crystals with an agreement at the ∼ / < H, Na, and
Cdare expected to contribute more. The fitted activities oflong-lived cosmogenic isotopes: H, Na, and
Cd forthe five crystals are compared with the measured ones,
Table 2
Background contributions in the energy range of 1 to 6 keV. There are only statistical uncertainties for the data row,and uncertainties for other rows are from the modeling.[Unit: Counts/keV/kg/day] Crystal 2 Crystal 3 Crystal 4 Crystal 6 Crystal 7Data 2 . ± .
010 2 . ± .
010 2 . ± .
007 2 . ± .
008 2 . ± . . ± .
189 2 . ± .
369 2 . ± .
325 2 . ± .
324 2 . ± . Pb 1 . ± .
006 0 . ± .
010 0 . ± .
012 1 . ± .
017 0 . ± . K 0 . ± .
004 0 . ± .
002 0 . ± .
001 0 . ± .
004 0 . ± . . ± . . ± . . ± . . ± . . ± . . ± .
101 0 . ± .
194 0 . ± .
198 0 . ± .
297 0 . ± . Pb Teflon 0 . ± .
005 0 . ± .
008 0 . ± .
007 0 . ± .
004 0 . ± . H 1 . ± .
160 2 . ± .
313 1 . ± .
258 0 . ± .
128 0 . ± . Sn 0 . ± .
003 0 . ± .
002 0 . ± .
004 0 . ± .
002 0 . ± . Cd 0 . ± .
003 0 . ± .
004 0 . ± .
004 0 . ± .
004 0 . ± . . ± .
003 0 . ± .
001 0 . ± .
001 0 . ± .
002 0 . ± . × − ) 5 . ± .
103 3 . ± .
039 2 . ± .
037 3 . ± .
089 3 . ± . Fig. 9
The low-energy spectra of single-hit events in Crystal 7.The measured energy spectrum is compared with the total ofthe simulations. The range of 1 to 6 keV in the MC spectrumis extrapolated from the modeling. reported in Ref [22], as shown in Fig. 8(c),(d),(e); thesevalues are in reasonable agreement with the measure-ments that averaged measured activities in the 1.7 yearsof data.
Cd contributes peaks at 25.5 and ∼ Sn that produces 28 keV and 3.7 ∼ Sn inthe modeling fit. However, because
Sn has a shorthalf-life of 115.1 days, the 1.7 year period is long enoughto reduce its activity to a negligible level while
Cdhas 562 days of half-life and enough activity levels whenproduced. As a result of the fit, they are compared withthe measurements in Fig. 8(f).Based on the background model, we found the back-ground levels for the five NaI(Tl) detectors in unit of dru (counts/day/keV/kg) in 1–6 keV as listed in Ta-ble 2. The dominant background contributions are from
Pb and H. Figure 9 shows the low-energy spectraof single-hit events of Crystal 7 in the 1–20 keV energyregion. The range of 1 to 6 keV in the MC spectrumis extrapolated from the modeling. The measured andsimulated results are in a good agreement.
COSINE-100 has been taking data at Y2L from Octo-ber 21, 2016. We present the background model for theWIMP search during the first 1.7 years of COSINE-100data. Our previous analysis with 59.5-day data showedthat
Pb and H produce the dominant contributionsin the energy region of 2–6 keV. As we lowered thethreshold to 1 keV, the background modeling was car-ried out accordingly. The model includes backgroundcontributions from both internal and external sources,including cosmogenic radionuclides and surface
Pbcontamination. To improve the background model withthe energy threshold as low as 1 keV, we used a depthprofile of the surface
Pb contamination that is pro-vided by the measurement with a test setup at Y2L.We also considered the effect of the energy scale errorspropagated from the statistical uncertainty and the non-linear detector response for the simulated spectrum atlow energy. With this improved background model, theoverall energy spectrum summed over all the simulationresults is well matched to the measured data not only forsingle-hit events but also for multiple-hit events. The im-proved model estimates the averaged background levelof 2.73 ± Pb and H for the five crystalsof COSINE-100 during the 1.7 years period of SET2data with a total exposure of 97.7 kg · years. Acknowledgments
We thank the Korea Hydro and Nuclear Power (KHNP)Company for providing underground laboratory spaceat Yangyang. This work is supported by: the In-stitute for Basic Science (IBS) under project codeIBS-R016-A1, NRF-2016R1A2B3008343, and NRF-2018R1D1A1B07048941, Republic of Korea; UIUC cam-pus research board, the Alfred P. Sloan Foundation Fel-lowship, NSF Grants Nos. PHY-1151795, PHY-1457995,DGE-1122492, WIPAC, the Wisconsin Alumni ResearchFoundation, United States; STFC Grant ST/ N000277/1and ST/K001337/1, United Kingdom; and Grant No.2017/02952-0 FAPESP, CAPES Finance Code 001, andCNPq 131152/2020-3, Brazil.
References
1. R. Gaitskell, Annu. Rev. Nucl. Part. Sci. 54, 315(2004).2. L. Baudis, Phys. DarkUniv. 1, 94 (2012).3. G. Adhikari et al. , Eur. Phys. J. C (2018) 107.4. G. Adhikari et al. , Nature (7734) (2018) 83-86.5. G. Adhikari et al. , Phys. Rev. Lett. , 031302 (2019).6. R. Bernabei et al. , Int. J. Mod. Phys. D (2004) 2127.7. R. Bernabei et al. , Eur. Phys. J. C (2013) 2648.8. R. Bernabei et al. (DAMA/LIBRA), Nucl. Phys. At. En-ergy (2018) 307. arXiv:1805.10486.9. J. Cherwinka et al. , (DM-Ice Collaboration), Phys. Rev. D (2014) 092005.10. E. Barbosa de Souza et al. (DM-Ice Collaboration), Phys.Rev. D (2017) 032006.11. J. Amare et al. , Nucl. Instrum. Meth. Phys. Res., Sect. A (2014) 187.12. J. Amare et al. , Phys. Rev. Lett. (2019) 031301.13. K. Fushimi et al. , Physics Procedia (2015) 67.14. C. Tomei et al. (SABRE Collaboration), Nucl. Instrum.Meth. A (2017) 418.15. G. Angloher et al. , Eur. Phys. J. C (2016) 441.16. S. Agostinelli et al. , Nucl. Instrum. Meth. Phys. Res.,Sect. A (2003) 250.17. G. Adhikari et al. (COSINE-100 Collaboration),arXiv:2005.13784.18. G.H, Yu. et al. , Astropart. Phys. (2021) 102518.19. L. Swiderski et al., Nucl. Instruments Methods Phys. Res.Sect. A Accel. Spectrometers, Detect. Assoc. Equip. et al. , Eur. Phys. J. C (2018) 490.21. J. Amare et al. , Eur. Phys. J. C (2016) 429.22. E. Barbosa de Souza et al. (COSINE-100 Collaboration),Astropart. Phys. (2020) 102390.23. Decay Data Evaluation Project,
24. S. Ohya, Nucl. Data Sheets 111, 1619 (2010).25. WWW Table of Radioactive Isotopes, http://nucleardata.nuclear.lu.se/toi/
26. R. Bernabei at al. , The DAMA/LIBRA apparatus, Nucl.Instrum. Meth. A (2008) 297.27. P.A. Zyla et al. (Particle Data Group), Prog. Theor. Exp.Phys.2020