IImproving the light collection using a new NaI(Tl)crystal encapsulation.
J.J. Choi a , B.J. Park b,d , C. Ha c , K.W. Kim d , S.K. Kim a , Y.D. Kim d,b ,Y.J. Ko d , H.S. Lee d,b , S.H. Lee b,d , S.L. Olsen d a Department of Physics and Astronomy, Seoul National University,Seoul 08826, Republic of Korea b IBS School, University of Science and Technology (UST),Daejeon 34113, Republic of Korea c Department of Physics, Chung-Ang University,Seoul 06973, Republic of Korea d Center for Underground Physics, Institute for Basic Science (IBS),Daejeon 34126, Republic of Korea
Abstract
NaI(Tl) crystals are used as particle detectors in a variety of rare-event searchexperiments because of their superb light-emission quality. The crystal lightyield is generally high, above 10 photoelectrons per keV, and its emission spec-trum is peaked around 400 nm, which matches well to the sensitive region ofbialkali photocathode photomultiplier tubes. However, since NaI(Tl) crystalsare hygroscopic, a sophisticated method of encapsulation has to be applied thatprevents moisture from chemically attacking the crystal and thereby degradingthe emission. In addition, operation with low energy thresholds, which is essen-tial for a number of new phenomenon searches, is usually limited by the crystallight yield; in these cases higher light yields can translate into lower thresholdsthat improve the experimental sensitivity. Here we describe the developmentof an encapsulation technique that simplifies the overall design by attachingthe photo sensors directly to the crystal so that light losses are minimized. Thelight yield of a NaI(Tl) crystal encapsulated with this technique was improved bymore than 30%, and as many as 22 photoelectrons per keV have been measured.Consequently, the energy threshold can be lowered and the energy resolution ∗ corresponding author Email address: [email protected] (C. Ha)
Preprint submitted to Journal of L A TEX Templates August 13, 2020 a r X i v : . [ phy s i c s . i n s - d e t ] A ug mproved. Detectors with this higher light yield are sensitive to events withsub-keV energies and well suited for low-mass dark matter particle searches andmeasurements of neutrino-nucleus coherent scattering. Keywords:
NaI(Tl), light yield, inorganic scintillator, dark matter,encapsulation
1. Introduction
It is well established that dark matter exists in our universe and correspondsto 26.4% of its total energy content [1]. Weakly Interacting Massive Particles(WIMPs) are dark matter candidate particles that are frequently consideredbecause of the balance among their inferred relic abundance, the measured darkmatter density and the strength of the weak interactions [2, 3]. The scatteringof relic WIMPs in the galactic dark matter halo from ordinary nuclei is beingsearched for in a number of experiments with a variety of target nuclei.The experimental signature for these searches is the detection of the recoilnucleus. For WIMPs with masses in the range between a few hundred MeVand a few hundred GeV, the recoil nuclei have kinetic energies of a few keV andthe current state of the art searches have set upper limits on the WIMP-nucleiinteraction cross-sections of 10 − cm to 10 − cm [4]. Recoil energies of low-mass WIMP nuclei scattering would typically be in the sub-keV energy range,and ultra-low-threshold experiments are needed to probe these masses. Interest-ingly, low-energy neutrino-nuclei scattering has the same experimental signatureas WIMP-nuclei interactions, where the neutrinos might originate from man-made accelerators, nuclear reactors, or the cosmos. An incoming neutrino withan energy of a few hundred MeV or less can interact coherently with the entiretarget nucleus with a cross section that can be as high as 10 − cm . The firstunambiguous detection of the coherent neutrino-nuclei scattering was reportedin 2017 [5], some 40 years after it was first predicted [6]. The count rates ofthese rare signals decrease exponentially with increasing nuclear recoil energy.Thus, experiments with low-thresholds and high-light yields can, in general, im-2rove the sensitivity of both low-mass WIMP searches and coherent neutrinoscattering measurements.Thallium-doped sodium iodide crystals (NaI(Tl)) [7] are suitable for low-energy-threshold rare-event search experiments. In the currently operatingCOSINE-100 WIMP-search experiment, NaI(Tl) crystals with light yields of15 PEs/keV are operated with an energy threshold of 1 keV [8, 9, 10]. At thatthreshold, the main limitation is the low number of detected photo-electrons inthe detected signal pulses, which makes them difficult to distinguish from thecopious photomultiplier tube (PMT) noise-induced pulses. Therefore, to accessevents with energies below 1 keV, improvements in the light yield are needed.It may be possible to improve the intrinsic light output of the crystal materialitself by changes in the crystal production process [11]. This is a long-term pro-gram that addresses a number of non-trivial technical issues that are currentlybeing pursued in parallel with the detector encapsulation studies. The encap-sulation R&D uses existing NaI(Tl) crystals and is being reported here. Espe-cially, we focus on improvements in the efficiency for collecting of the radiation-generated scintillation photons by means of a simplified light coupling scheme.Also, because of the hygroscopic properties of NaI(Tl) crystals, a small amountof moisture contaminations during the detector assembly procedures can alsoaffect the light yield. We, therefore, have given special attention to the devel-opment of techniques that limit the crystal’s exposure to humidity during theassembly procedure.
2. Method
The new encapsulation design principle that we follow is the minimization oflight losses during the transit of scintillation photons from their generation siteto the PMT photocathode. For comparison, we use the existing COSINE-100crystals which are cylindrical with dimensions of 4.75-inches in diameter and PE stands for photoelectrons.
Figure 1: The COSINE-100 detector design (top) is 4.75-inches in diameter and is indepen-dently encased by copper with an optical pad and a quartz window inside of each end of thecylindrical module. The detector module is, in turn, coupled at each end to a 3-inch PMT viaan optical gel. The new design (bottom) encapsulates a 3-inch diameter crystal and a 3-inchPMT that is directly coupled to the NaI(Tl) end-surface via a single optical pad. There is aPTFE ring spacer with 1 cm thickness (grey) that pushes against the PMT to give pressureon the optical pad.
In the COSINE-100 detector modules, shown in the top panel of Fig. 1,generated photons that are incident on the portion of the outer edge of the 4-inch quartz window in 4.75-inch crystal that is not covered by the 3-inch PMTphotocathode have a low probability of being detected. By matching the sizeof the crystal end face to that of the PMT photocathode, all of the generatedphotons that are incident on the endface of the crystal are collected with highefficiency. In addition, we carefully polish the entire crystal surface and use only4 single optical pad between the PMT window and the NaI(Tl) end surface. Thisreduces light losses due to reflections at each optical surface. The design of thisnew encapsulation configuration is shown in the bottom panel of Fig. 1.The COSINE-100 detectors are a detector-sensor separated design while thenew design is a detector-sensor combined assembly where the PMTs are inte-gral components of the airtight crystal encapsulation system. This removes thequartz window and the optical gel but at the cost of a more difficult encap-sulation procedure. The NaI(Tl) crystal’s vulnerability to moisture requires atight seal that is secure from any air leakage, while applying a limited amountof pressure onto the relatively fragile PMT structure. To accomplish this, weplaced a 1 cm-thick PTFE ring spacers shaped to fit the neck part of PMT glassbetween the endcap of the copper cylinder and the back of the PMT glass enve-lope that applied just enough pressure to couple the optical pad to the crystalend face and maintain the airtight integrity of the encapsulated structure. Theforce is applied via four screws to the back of the PTFE using a preset torquewrench. Since any leaked air would quickly degrade the crystal’s surface qualityand, thereby, reduce the light output, we use the measured light yield as theprimary monitor of the long-term stability and airtightness of the assembly.We have tested three crystals with these encapsulations that are labeled asNEO-1, NEO-2, and NEO-3.
In this development, we use 3-inch low-background PMTs (R12669SEL) thathave a high (40%) quantum efficiency for 400 nm photons. The crystal compo-nents were cut to match the 3-inch photocathode. For the first crystal, NEO-1,the original size of the crystal ingot was 4-inches in diameter; its diameter wasreduced to 3-inches using a lathe, as shown in Fig. 2. On the other hand, the4-inch length of the crystal is chosen to keep as much of the original ingot lengthas possible. In COSINE-100, we have seen no noticeable effect in light yield dueto the length of crystal when two crystals with twice different lengths are com-pared [8]. Since the NaI(Tl) crystal is brittle, we had to pay special precautions5 igure 2: NEO-1 crystal machining. Rough machining was used to reduce the diameter of a4-inch crystal to 3-inches in a normal atmosphere with a lathe. While machining, we had topay special attention because the brittleness of the crystal and weakness for its limited abilityto support stress. The crystal was turned with a very sharp tool bit at a slow revolutionspeed. during the machining, but some cracks were inevitable in our first attempt. Af-ter seeing the results of NEO-1, for the next two crystals, NEO-2 and NEO-3,we designed and ordered 3-inch diameter cylinder detectors and measured theoriginal light yields prior to re-encapsulation.
The next step was to polish all of the surface areas of the 3-inch crystals. Thelateral areas are included in this procedure because we found that the roughenedsurfaces of the COSINE-100 crystals had some radioactive surface contaminationthat originated from either the polishing film or the environment [11]. Wedid the polishing in a low-humidity glovebox that was continuously flushedwith nitrogen gas with the humidity maintained below 100 ppm v of H O bymeans of a molecular sieve trap. The polishing was done in several stages withlapping papers of different grits using a small lathe that was located inside6he glovebox (see Fig. 3). As soon as the crystal polishing was finished, theencapsulation procedure was started in order to minimize additional radioactivecontaminants on the surfaces of the crystal. Also, we used carefully cleaned anddried encapsulation components that had been baked at high temperature andkept in the glovebox for a long enough time so that the level of H O emanationfrom them was too low to affect the crystal’s surface quality.
Figure 3: Fine polishing inside the low-humidity glovebox (left) and the final product (right).For the polishing, a small lathe with lapping papers was used. All surfaces were polished untilthey were of optical quality.
As shown in the top diagram of Fig. 1, one end of the COSINE-100 crystalis coupled through an optical pad, a quartz window, and an optical gel in se-ries to the PMT photocathode. The optical pad has 90% light transmissionfor 400 nm photons and a refractive index of 1.43; the optical gel is nearly EJ-560 from Eljen Technology EJ-550
Figure 4: The NEO-1 (top) and NEO-3 (bottom) detectors. In both detectors, the PMTs aresealed along with the polished crystal ingots. The crystals are in the middle section betweenthe copper flanges that, together with the PMT glass envelope body, form the airtight seal. . Measurements To measure the crystal light yields and resolutions, we used a simple testsetup in a surface-level laboratory that included a 4 π , 20 cm-thick lead and5 cm-thick copper shield against environmental background radiation. A Amsource located at the middle of the crystal scintillator provided a 59.5 keVgamma line that is produced during its alpha transition to
Np.Additionally, we tested the detectors in a facility at the Yangyang Under-ground Laboratory (Y2L), where the cosmic-ray muon rate is strongly sup-pressed by the 700 m rock overburden, and shielding comprised of lead, copper,and polyethylene attenuated the environmental radiation. The Y2L setup has12 low-background CsI(Tl) crystals that surround the test volume that are usedto tag accompanying radiation, which facilitates the evaluation of internal back-grounds in the NaI(Tl) crystal that is being studied.Figure 5 shows the surface laboratory setup and the Y2L setup that are usedin these tests.
CsICsICsI
CsI
CsICsICsI
CsICsI CsI
CsI CsINaI(Tl)
Figure 5: A photo of the surface laboratory setup (left) and the schematic of the 700 m un-derground setup at Y2L (right). The Y2L setup contains rectangular-shaped CsI(Tl) crystalsthat facilitate the identification of background contaminants in the NaI(Tl) crystal that isbeing tested. The green horizontal bar made of PMMA on the right schematic supports thetop row CsI(Tl) crystal detectors. . Results The shape of the waveform produced by single photoelectrons was charac-terized using isolated signals in the tail part of NaI(Tl) pulses associated with59.5 keV gamma rays for which the full energy is deposited in the crystal. Fromthis, the light yield is determined from the ratio of the total deposited chargeto the single photoelectron’s (SPE) mean charge, scaled to 59.5 keV.NEO-1 was made from a 4-inch diameter crystal with a one-window en-capsulation. So, for this test crystal, the disentangling of the effects of crystalresizing and the modified light coupling was difficult. On the other hand, NEO-2and NEO-3 started out with 3-inch diameter ingots with a one-window encap-sulation made by the same vendor. The measured light yields of the originaldetector configurations were 10.7, 16.9, and 17.7 PEs/keV for NEO-1, NEO-2,and NEO-3, respectively. The lower yield for the original NEO-1 measurementwas likely due to the size mismatch between the crystal end face and the PMTphotocathode.The light yields for these crystals after the re-encapsulation are measuredto be 20.5, 19.3, and 21.8 PEs/keV. For NEO-2 and NEO-3, the new designimproves the light yields by 14% and 23%, respectively. This improvementlikely comes from the clear-polishing of the crystal combined with the simpli-fied optical coupling. It is likely that the marginal improvement for NEO-2compared to NEO-3 is due to a few cracks developed near the endface whenthe re-encapsulation was performed. In case of the NEO-3 measurements, wehave additionally verified the yields with an SPE charge spectrum that wasdetermined with a LED source. The light yields are summarized in Table 1.
Since the light yield directly affects the energy resolution of a crystal de-tector, we compare the light yield and resolution of the peak with those fromprevious COSINE-100 measurements. Figure 6 shows the light yield compar-ison between a COSINE crystal and the newly designed detectors. Figure 710 able 1: The light yield measurements before and after the encapsulation change. The lastcolumn shows the light yield for one of the COSINE-100 crystals measured in the same way.NEO-1 shows a higher light yield after the resizing of the crystal. However, the originallight yield of the 4-inch crystal was not accurately measured due to a mismatch between thecrystal base size and PMT photocathode size. The units of the light yield measurements arePEs/keV.
NEO-1 NEO-2 NEO-3 COSINE-100after(before) after(before) after(before) C620.5 ± ± ± ± ± ± ± Am gamma peak, with anenergy resolution of 3 . ± . stat. ) keV from Gaussian fits to the peak. TheNEO-2 and NEO-3 resolutions are determined in the same way in Fig. 7 to be2 . ± . stat. ) and 2 . ± . stat. ) keV, respectively. We have measured the NEO-2 energy spectrum at Y2L for a four weekcontinuous period as a check on its stability. For this we used crystal’s internalpeaks from
Pb (46.5 keV gamma plus X-rays) and cosmogenic
I (67.2 keV)and m Te (30.5 keV) [12] to monitor the low energy spectrum. Figure 8 showsthat the peak position did not change between first 100 hour data period andthe succeeding 100 hour data period, which indicates that the encapsulationdoes not have an air leak.
5. Conclusion
We have developed a method for NaI(Tl) crystal encapsulations that in-cludes a well matched crystal-PMT window size with a simplified light coupling11
PE800 900 1000 1100 1200 1300 1400 1500 A r b i t r a r y un i t NEO-3NEO-2COSINE-100
Figure 6: The distributions of the number of photoelectrons (NPEs) associated the
Amgamma peaks in the COSINE-100 (black dot-dashes), NEO-2 (red dashes) and NEO-3 (solidblue line) detectors.
Energy (keV)40 45 50 55 60 65 70 75 80 A r b i t r a r y un i t COSINE-100
Energy (keV)40 45 50 55 60 65 70 75 80 A r b i t r a r y un i t NEO-2
Energy (keV)40 45 50 55 60 65 70 75 80 A r b i t r a r y un i t NEO-3
Figure 7: The energy resolutions for 59.5 keV gamma rays of the three detectors. The blackdata points (left) are from COSINE measurements while the red (middle) from NEO-2 and theblue(right) from NEO-3, respectively. The resolutions are: 3.4 keV for the COSINE detector,2.5 keV for NEO-2 and 2.8 keV for NEO-3 obtained by Gaussian fits (red lines). design. The results show 22–38% light yield improvements and as much as 30%improvement in energy resolution. The absolute 22 PE/keV value in NEO-3 isby far the highest ever reported for a large-size NaI(Tl) crystal. In addition,12 nergy (keV) 0 10 20 30 40 50 60 70 80 90 100 c oun t s / da y / k g / k e V I-125Pb-210I-125Te/I ECNEO-2 Crystal
Figure 8: The low energy background spectra between 0 keV and 100 keV for two differenttime periods separated by two weeks. First 100 hours of data(blue) and the next 100 hoursof data (magenta) are compared in cases of beta/gamma events. Several radioisotopes aredecaying away as the cosmic activation is terminated in the underground laboratory. Weapply the same energy calibration for the two data periods. Below 5 keV, there remaineda residual noise contamination that obscured the beta/gamma spectrum. A modest eventselection was applied to reject noise events and multiple-site events are removed using thesurrounding CsI(Tl) veto detectors.
6. Acknowledgments
This research was funded by the Institute for Basic Science (Korea) underproject code IBS-R016-A1; This research was supported by the Chung-AngUniversity Research Grants in 2020.
ReferencesReferences [1] M. Tanabashi, et al., Review of particle physics, Phys. Rev. D 98 (2018)030001. doi:10.1103/PhysRevD.98.030001 .[2] B. W. Lee, S. Weinberg, Cosmological lower bound on heavy-neutrinomasses, Phys. Rev. Lett. 39 (1977) 165. doi:10.1103/PhysRevLett.39.165 .[3] G. Jungman, M. Kamionkowski, K. Griest, Supersymmetric dark matter,Phys. Rept. 267 (1996) 195–373. doi:10.1016/0370-1573(95)00058-5 .[4] M. Schumann, Direct detection of WIMP dark matter: concepts and sta-tus, J. Phys. G46 (10) (2019) 103003. arXiv:1903.03026 , doi:10.1088/1361-6471/ab2ea5 .[5] D. Akimov, et al., Observation of coherent elastic neutrino-nucleusscattering, Science 357 (6356) (2017) 1123–1126. arXiv:https://science.sciencemag.org/content/357/6356/1123.full.pdf , doi:10.1126/science.aao0990 . 146] D. Z. Freedman, Coherent effects of a weak neutral current, Phys. Rev. D9 (1974) 1389–1392. doi:10.1103/PhysRevD.9.1389 .[7] R. Hofstadter, Alkali halide scintillation counters, Phys. Rev. 74 (1948)100–101. doi:10.1103/PhysRev.74.100 .[8] G. Adhikari, et al., Initial performance of the COSINE-100 experiment,Eur. Phys. J. C 78 (2) (2018) 107. arXiv:1710.05299 , doi:10.1140/epjc/s10052-018-5590-x .[9] G. Adhikari, et al., An experiment to search for dark-matter interactionsusing sodium iodide detectors, Nature 564 (7734) (2018) 83–86. arXiv:1906.01791 , doi:10.1038/s41586-018-0739-1 .[10] G. Adhikari, et al., Lowering the energy threshold in COSINE-100 darkmatter searches arXiv:2005.13784 .[11] B. Park, et al., Development of ultra-pure NaI(Tl) detector for COSINE-200 experiment arXiv:2004.06287 .[12] E. Barbosa de Souza, et al., Study of cosmogenic radionuclides in theCOSINE-100 NaI(Tl) detectors, Astropart. Phys. 115 (2020) 102390. arXiv:1905.12861 , doi:10.1016/j.astropartphys.2019.102390doi:10.1016/j.astropartphys.2019.102390