Response of undoped cryogenic CsI to low-energy nuclear recoils
RResponse of undoped cryogenic CsI to low-energy nuclear recoils
C.M. Lewis ∗ and J.I. Collar Enrico Fermi Institute, Kavli Institute for Cosmological Physics,and Department of Physics, University of Chicago, Chicago, Illinois 60637, USA (Dated: January 12, 2021)The bright scintillation of pure CsI operated at liquid-nitrogen temperature makes of this materiala promising dark matter and neutrino detector. We present the first measurement of its quenchingfactor for nuclear recoils. Our findings indicate it is indistinguishable from that for sodium-dopedCsI at room temperature. Additional properties such as light yield, afterglow, scintillation decayproperties for electron and nuclear recoils, and energy proportionality are studied over the 108-165 Ktemperature range, confirming the vast potential of this medium for rare-event searches.
I. INTRODUCTION
An ongoing interest in characterizing the responseof radiation detectors to low-energy nuclear recoils, in-duced by the elastic scattering of neutral particles, istraceable to the first direct search for Weakly Interact-ing Massive Particles (WIMPs) [1], popular dark mattercandidates. Neutrinos with energies below a few tens ofMeV can scatter coherently from nuclei via the weakneutral current [2], also producing few-keV nuclear re-coils as the single outcome from this process. The recentobservation of this so-called Coherent Elastic Neutrino-Nucleus Scattering (CE ν NS) [3, 4] has added thrust toa quest for new materials well-adapted to the detectionof these subtle low-energy interactions.Scintillating sodium-doped cesium iodide (CsI[Na]),operated at room-temperature, was chosen as the fa-vored detector material for the first CE ν NS measure-ment. A long list of virtues leading to its selection isdescribed in [5, 6]. Among those is a large and essen-tially identical CE ν NS cross-section for both Cs andI, a high light-yield, reduced afterglow, and a quench-ing factor (QF) of order 10 % in the few-keV nuclearrecoil (NR) energy region of interest. This QF is theratio between the light yield for NRs and that for elec-tron recoils (ERs) of the same energy. A precise un-derstanding of the energy dependence of the QF is ofcrucial importance in the interpretation of WIMP andCE ν NS searches [7].Undoped CsI exhibits a large increase in light yieldat liquid-nitrogen temperature, reaching a theoreticallimit in light-conversion efficiency that exceeds 100 scin-tillation photons per keV of ER energy deposition [8–16]. This is close to three times the room-temperatureyield of CsI[Na]. When monitored with silicon lightsensors combined with state-of-the-art waveshifters ableto maximize their quantum efficiency, the potential todetect NRs as low in energy as 1 keV appears to be ∗ [email protected] Figure 1. Detector cross-section, derived from the MCNPsimulation: 1) voltage divider, 2) Hamamatsu R8520-506cryogenic PMT, 3) copper holder, 4) position of thermocou-ple within reach [17]. This is a NR energy regime unprece-dented for scintillators; one where the new physics be-yond the Standard Model that is reachable via CE ν NSconcentrates [17]. However, the assumption on whichthis promise is based is that the QF of this cryogenicmaterial, unknown for NRs until now, is at least as fa-vorable as for doped CsI at room-temperature.This work describes a first measurement of this QFin the temperature range 108-165 K, using the customdetector assembly in Fig. 1, exposed to monochromatic2.25 MeV neutrons from a D-D generator. Neutronsscattering from the CsI crystal are detected by a Bi-cron 501A liquid scintillator cell with neutron/gammadiscrimination capability. This cell is placed at a user-defined angle from the initial neutron trajectory, allow-ing to select the energy deposited by NRs in CsI. This a r X i v : . [ phy s i c s . i n s - d e t ] J a n experimental setup, data acquisition system (DAQ),and analysis method have been previously employed byus for NaI[Tl] and CsI[Na] room-temperature QF mea-surements. Details of these technical aspects are pro-vided in [5, 7, 18]. In this new implementation, a PIDalgorithm was used to monitor the temperature at bothends of the CsI crystal and control the power injectedinto a heating element (manganin wire, Fig. 1) resultingin a temperature stability of ∼ < II. ISOLATION OF LOW ENERGY NUCLEARRECOILS AND QF MEASUREMENT
The use of a small 7.24 cm CsI scintillator [19]ensures that single-scatters dominate neutron interac-tions in the crystal. Multiple scatters make up for just17-27% of the total, depending on the selected scat-tering angle, and are accounted for in simulations. AHamamatsu R8520-506 cryogenic bialkali photomulti-plier (PMT) was directly coupled to the sample us-ing optical RTV. While operation of this PMT downto 87 K is possible, the lowest temperature of 108 Kachieved in this study was limited by the cooling powerof the horizontal-arm cryostat employed.Among the lessons derived from our latest CsI[Na]experimentation is the impact on QF measurements ofPMT saturation at high bias. As a preliminary precau-tion, we compared the PMT charge output for
Am59.5 keV gammas and for single photo-electrons (SPE)following the test procedure developed in [7]. The nor-malized ratio of these outputs provides a light yield inunits of PE/keV. As in [7], the energy reference used inthe definition of the QF was given by this
Am emis-sion, assuming direct proportionality for lower ER en-ergies. Tests over a range of PMT voltage biases weremade at 108 K, a temperature corresponding to themaximum light yield observed in this work. As can beseen in Fig. 2 the chosen 820 V bias is well within thelinear response of this PMT and away from saturationeffects at the light levels involved in this study.Incremental improvements to the MCNP-PoliMi sim-ulations [20] used in [7] were made by accounting forsubdominant inelastic neutron scattering through de-exitation gamma escape from CsI. Similarly to [7],charge was integrated over the 3 µ s following the onsetof scintillation signals. However, due to electrical safety Figure 2. Tests of R8520-506 PMT saturation under 59.5keV gamma irradiation of the CsI crystal. Light yield isnormalized to the average of all measurements. Error barscombine the uncertainties from fits to SPE and 59.5 keVcharge distributions [7]. A bias of 820 V was adopted.Figure 3. Implementation of an offline baseline correctionusing an inverse high-pass algorithm on an example co-added ensemble of 1,000
Am gamma signals. The finalintegrated charge over the first 3 µ s is corrected by ∼ % with respect to the original trace. concerns for metallic-envelope PMTs like the R8520-506, a positively-biased voltage divider was used. Theresulting capacitive coupling to the DAQ produces awell-documented overshoot of the PMT signal [21]. Un-corrected, this leads to an underestimation of the inte-grated charge carried by a scintillation signal. Remedialanalysis techniques have been put forward in a numberof affected experiments [22–24]. The impact of this ef-fect and its correction on our charge measurements can Figure 4. Scatter plot of CsI events passing the Bicron 501AIRT cut for the 65 ◦ neutron scattering angle. Prompt coinci-dences between the backing detector and CsI crystal appearat ∼
225 ns along the horizontal scale in this DAQ [18]. Ascintillation decay time of ∼
600 ns at 108 K [8] results in amodest spillage of the onset of few-PE signals to later times. be assessed from Fig. 3. The corrective procedure usesan inverse high-pass algorithm (an offline pole-zero can-cellation) to allow for accurate charge integration belowthe median waveform baseline. Special attention waspaid to ensure that this average charge correction alsoapplied to signals at lower energy. As expected, sincethe response of the output capacitor causing the over-shoot depends on signal frequency, and not on ampli-tude, the magnitude of the integrated charge correctionwas found to be consistent at 18 % for all energies.Also following [7], we utilize an integrated rise-time (IRT) analysis [25, 26] to separate neutron- fromgamma-induced events in the Bicron 501A backing de-tector. The resulting data quality is illustrated in Fig.4 following rejection of initially-dominant gamma con-taminations. The modest background of random coinci-dences between CsI and backing detector is removed bysubtraction of the energy spectrum of events within the105-225 ns time range of Fig. 4 from that for true co-incidences concentrated within the 225-345 ns interval.The residual spectra of NR signals from elastic neutronscattering at 108 K are shown in Fig. 5, for each of thesix scattering angles explored.The extraction of a best-fit QF is accomplished identi-cally to [7]: simulated energy depositions are translatedto a corresponding number of photoelectrons (PE), ac-counting for Poisson smearing of PE statistics and theeffect of the assumed QF. The obtained simulated distri-butions are then compared with the experimental resid-uals of Fig. 5, with a log-likelihood analysis selectingthe most adequate QF. Fig. 6 illustrates this procedure Figure 5. Energy deposition by NRs from neutron scat-tering on CsI at 108 K. Datapoints are experimental data,histograms correspond to simulated distributions at best-fitQF. Scattering angle, backing detector distance to the CsIcrystal, simulated mean NR energy, and best-fit QF are in-dicated. The decrease in event rate with increase in angle ischaracteristic of forward-peaked elastic scattering. for the lowest-energy NRs measured. The uncertaintiesin the QF values, manifested as vertical error bars inFig. 7, combine the 1-sigma log-likelihood error and thesmall dispersion in the
Am light yield. This energyreference was measured repeatedly during these runs.Horizontal error bars in Fig. 7 are akin to those in [7],i.e., derived from the simulated spread in NR energiesprobed.The totality of our QF measurements for pure CsI at108 K are reported in Fig. 7. An excellent match to themodified Birks model developed for room-temperatureCsI[Na] in [7] is noticeable. At least from the point ofview of the adiabatic factor included in that model thisagreement is not surprising: the band gap on which thisadiabatic factor depends is not expected to change sig-nificantly from room-temperature to 108 K [27, 28], anargument supported by observations in other cryogenicscintillators [29]. However, this apparent constancy ofthe QF for NRs over the 108-295 K temperature rangeis in contrast with a reported factor of ∼ Figure 6. Comparison between light yield observed for ∼ ± σ uncertainty in the best-fit QF.Figure 7. Quenching factor for low-energy nuclear recoils inundoped cryogenic CsI. The recoil energies probed span theCE ν NS range of interest for CsI at a spallation source [3, 17].A dashed line shows the modified Birks model developed in[7] for 295 K CsI[Na], a grayed band its ± σ uncertainty. decreasing temperature in the QF for alpha particles,over the same range, for this material [11, 12].To confirm the observed independence of the QF onoperating temperature, measurements at the 56 ◦ scat-tering angle (13.9 keV NR energy) were repeated forfour additional temperatures, up to 165 K. The maxi-mum temperature that could be explored was limitedby the rapidly decreasing light yield. Fig. 8 showsthe result of these measurements. As expected, no Figure 8. CsI quenching factor measurements as a func-tion of temperature, normalized to their average, for the 56 ◦ scattering angle (13.9 keV NRs). No significant dependenceof QF on temperature is observed. The Am light yieldshown follows a trend of rapid change as in [8, 14–16] (errorbars are encumbered by datapoints). Extrapolated to 87 K,the lowest operating temperature of modern bialkali PMTs,the observed light yield triples that for room-temperatureCsI[Na] during the first CE ν NS observation [3, 4]. statistically-significant variation in the QF is visible,over a temperature range for which the overall lightyield nevertheless more than tripled.
III. ANCILLARY MEASUREMENTS
Small differences in the scintillation decay propertiesof ERs and NRs have been exploited in past experi-ments. Even when these are too subtle for event-by-event ER-NR discrimination they can still be appliedto a large enough ensemble of events, statistically im-proving the sensitivity of a search for rare NR events[30]. To explore this possibility, a dedicated ER dataset was collected containing Compton scatters from acollimated beam of
Ba gammas impinging on the CsIcrystal. Low-energy events were favored by triggeringthe DAQ on coincidences with a backing detector placedat a small angle with respect to the incoming beam [4].Five hundred events were selected from this data set,and the same number from available NR data, with thecriterion that both groups should have similar distribu-tions in the number of PE registered per event (Fig. 9inset, [5]). This PE range selection corresponds to a NRenergy of ∼ Figure 9. Decomposition of the scintillation decay time ofpure CsI at 108 K into fast and slow components, for co-added ensembles of low-energy NRs and ERs (see text). Onein ten waveform points is displayed, for clarity. For an unbi-ased ER-NR comparison, the PE distributions (inset) werechosen for similarity between both data sets. Best-fit slow(s) and fast (f) scintillation decay constants, and the ratioof PMT current in each decay component are shown. ysis [5]. The average ER and NR traces thus obtained,shown in Fig. 9, were fitted allowing for fast and slowscintillation decay components [8, 14]. The PMT over-shoot corrections for each data subset had identical de-cay components, to avoid introducing an artifact in thefits. This direct comparison between few-keV ER andNR events in CsI at 108 K shows only subtle differences,probably too difficult to exploit even for statistical ER-NR discrimination.Separately, exposures to a variety of gamma-emittingradiosotopes were obtained in order to define the lightyield proportionality of pure CsI at 108 K. A lowest-energy datapoint at 5.9 keV was acquired by placing anevaporated Fe source adjacent to the CsI crystal, incontact with its PTFE reflector. These results are dis-played in Fig. 10, along with all other available similardata for this material [9, 31, 32]. Attempts have beenmade to understand the considerable dispersion in theseresults as a function of operating temperature and ofCsI sample origin [31]. Our measurements using Am-crys/Proteus stock [19] show a characteristic absence ofreduction in light yield below ∼
30 keV ER energy, seenin other datasets. The deviation from the assumptionmade in the definition of the QF that direct propor-tionality exists below 59.5 keV ER energy, is modest forour data. As emphasized in [7], this assumption is inany case immaterial as long as the NR energy scale itdefines is applied consistently to both QF calibrationsand in the interpretation of physics runs.
Figure 10. All known light yield proportionality data forERs in undoped CsI at various temperatures, made relativeto 662 keV [9, 31, 32], including the present measurement.Figure 11. Afterglow in cryogenic pure CsI. The procedureto obtain these data follows our previous CsI[Na] and CsI[Tl]measurements, also shown [5]. Each data point combines500 measurements. This was 100 for doped material, leadingto smaller present error bars, shown one-sided for clarity.
The light yield previously demonstrated for pure CsIat liquid nitrogen temperature ( ∼
80 K) is in the rangeof 80-125 scintillation photons per keV at a referenceER energy of 662 keV [8, 10, 12, 13], displaying a depen-dence on CsI stock [9, 33]. The presently measured yieldis 26.13 ± % quantum efficiency of R8520-506 PMTs at the ∼
340 nm emission characteristic ofcryogenic pure CsI [8, 14], and a non-proportionalityof 8.5 % between 59.5 keV and 662 keV (Fig. 10), gives96.3 ± ∼
10% increaseis to be expected at the minimum 87 K operating tem-perature of present-day cryogenic bialkali PMTs. Asemphasized in [12, 13, 15, 17], this uncommonly-highyield is ideal for low-energy NR detection.A final study was performed to quantify the afterglow(phosphorescence) of cryogenic pure CsI, thus far alsoan unknown quantity. These long-delayed few-PE emis-sions following a primary energy deposition can leadto a continuum of low-energy pulses that impede theidentification of NRs, raising the effective threshold ofthe detector. The abatement of afterglow is of particu-lar importance for CE ν NS searches, performed withoutthe benefit of a significant overburden, and thereforesubject to frequent, energetic primaries from cosmic-raytraversal. CsI[Na] was preferred over CsI[Tl] during thefirst CE ν NS measurement for this reason [5]: however,the removal of residual afterglow events still resulted insignificant signal acceptance losses [3, 4]. Fig. 11 illus-trates our results, following the same procedure as in[5] (integration of afterglow over 1 µ s periods followinga ∼ IV. CONCLUSIONS
We have presented a first measurement of the quench-ing factor for low-energy nuclear recoils in undoped CsIat cryogenic temperature. Our results indicate that it is indistinguishable from that for room-temperatureCsI[Na], further validating a physical response modeldeveloped for that scintillator in [7], and the modi-fied interpretation of the first CE ν NS measurement de-scribed in that same publication. The combination ofa sizeable and by now well-understood NR quenchingfactor, negligible afterglow, and light yield in excess of100 scintillation photons per keV defines an exception-ally promising material for WIMP and CE ν NS detec-tion. Specifically, when combined with high quantum-efficiency light sensors, cryogenic CsI can provide a sen-sitivity to ∼ ν NS studies [17].Additional work using a Y/Be photoneutron sourceand the QF measurement technique laid out in [34–36]is planned. This will probe nuclear recoils below 4.6keV in CsI. A search for low-energy deviations from themodel depicted in Fig. 7, stemming from Migdal-likeprocesses [37–39], should be possible using this uniquematerial.
V. ACKNOWLEDGEMENTS
We are indebted to P. Parkhurst at Proteus Inc. formany useful consultations. This work was supported byNSF awards PHY-1806722, PHY-1812702, and by theKavli Institute for Cosmological Physics at the Univer-sity of Chicago through an endowment from the KavliFoundation and its founder Fred Kavli. [1] S. Ahlen, F. Avignone, R. Brodzinski, A. Drukier,G. Gelmini, and D. Spergel, Phys. Lett. B , 603(1987).[2] D. Z. Freedman, Phys. Rev. D , 1389 (1974).[3] D. Akimov et al. , Science , 1123 (2017).[4] B. Scholz, Ph.D. thesis, University of Chicago (2017),arXiv:1904.01155.[5] J. I. Collar, N. E. Fields, M. Hai, T. W. Hossbach, J. L.Orrell, C. T. Overman, G. Perumpilly, and B. Scholz,Nucl. Instr. Meth. A , 56 (2015).[6] N. Fields, Ph.D. thesis, University of Chicago (2014).[7] J. I. Collar, A. R. L. Kavner, and C. M. Lewis, Phys.Rev. D , 033003 (2019).[8] C. Amsler, D. Grogler, W. Joffrain, D. Lindelof,M. Marchesotti, P. Niederberger, H. Pruys, C. Regen-fus, P. Riedler, and A. Rotondi, Nucl. Instr. Meth. A , 494 (2002). [9] M. Moszynski, M. Balcerzyk, W. Czarnacki, M. Ka-pusta, W. Klamra, P. Schotanus, A. Syntfeld, M. Szaw-cowski, and V. Kozlov, Nucl. Instr. Meth. A , 357(2005).[10] M. Moszynski, W. Czarnacki, W. Klamra, M. Sza-wlowski, P. Schotanus, and M. Kapusta, Nucl. Instr.Meth. A , 307 (2003).[11] P. Nadeau, Ph.D. thesis, Queen’s University (2015).[12] M. Clark, P. Nadeau, S. Hills, C. Dujardin, and P. D.Stefano, Nucl. Instr. Meth. A , 6 (2018).[13] J. Liu, M. Yamashita, and A. Soma, J. Instrum. ,P10003 (2016).[14] C. L. Woody, P. W. Levy, J. A. Kierstead, T. Skwar-nicki, Z. Sobolewski, M. Goldberg, N. Horwitz,P. Souder, and D. F. Anderson, IEEE Trans. Nucl.Sci. , 492 (1990).[15] X. Zhang, X. Sun, J. Lu, and P. Lu, Radiat. Detect.Technol. Methods , 15 (2018). [16] V. B. Mikhailik, V. Kapustyanyk, V. Tsybulskyi,V. Rudyk, and H. Kraus, Phys. Status Solidi (b) ,804 (2015).[17] D. Baxter, J. I. Collar, P. Coloma, C. E. Dahl, I. Es-teban, P. Ferrario, J. J. Gomez-Cadenas, M. Gonzalez-Garcia, A. R. L. Kavner, C. M. Lewis, F. Monrabal,J. Muñoz Vidal, P. Privitera, K. Ramanathan, andJ. Renner, JHEP , 1 (2020).[18] J. I. Collar, Phys. Rev. C , 035806 (2013).[19] Amcrys stock, Czochralski grown in dedicated furnacesfor trace-level dopant concentration. Procured fromProteus Inc., Chagrin Falls, Ohio 44022, USA.[20] S. A. Pozzi, E. Padovani, and M. Marseguerra, Nucl.Instr. Meth. A , 550 (2003).[21] A. Wright and T. Wright, The Photomultiplier Hand-book (Oxford University Press, 2017).[22] R. Acciarri et al. , J. Instrum. , P08003 (2017).[23] Y. Abe et al. (Double Chooz Collaboration), Phys. Rev.D , 052008 (2012).[24] H. Zhang, Z. Wang, Y. Zhang, Y. Huang, F. Luo,P. Zhang, C. Zhang, M. Xu, J. Liu, Y. Heng, C. Yang,X. Jiang, F. Li, M. Ye, and H. Chen, J. Instrum. ,T08002 (2019).[25] X. L. Luo et al. , Nucl. Instr. Meth. A , 83 (2014).[26] E. Ronchi, P.-A. Soderstrom, J. Nyberg, E. A. Sunden,S. Conroy, G. Ericsson, C. Hellesen, M. G. Johnson,and M. Weiszflog, Nucl. Instr. Meth. A , 534 (2009).[27] Y. Varshni, Physica , 149 (1967). [28] A. M. Karo and J. R. Hardy, J. Chem. Phys. , 3173(1968).[29] D. Spassky, V. Nagirnyi, A. Savon, I. Kamenskikh,O. Barinova, S. Kirsanova, V. Grigorieva, N. Ivan-nikova, V. Shlegel, E. Aleksanyan, and et al., J. Lumin. , 195–202 (2015).[30] P. Smith, G. Arnison, G. Homer, J. Lewin, G. Al-ner, N. Spooner, J. Quenby, T. Sumner, A. Bewick,J. Li, D. Shaul, T. Ali, W. Jones, N. Smith, G. Davies,C. Lally, M. van den Putte, J. Barton, and P. Blake,Phys. Lett. B , 299 (1996).[31] X. Lu, Q. Li, G. A. Bizarri, K. Yang, M. R. Mayhugh,P. R. Menge, and R. T. Williams, Phys. Rev. B ,115207 (2015).[32] G. Salakhutdinov and D. Efanov, , 345 (2015).[33] R. Donghia, Nuovo Cimento C , 276 (2016).[34] B. Scholz, A. Chavarria, J. Collar, P. Privitera, andA. Robinson, Phys. Rev. D (2016).[35] J. I. Collar, Phys. Rev. Lett.
110 21 , 211101 (2013).[36] A. E. Chavarria et al. , Phys. Rev. D , 082007 (2016).[37] M. Ibe, W. Nakano, Y. Shoji, and K. Suzuki, JHEP2018