Liquid Scintillator Response to Proton Recoils in the 10-100 keV Range
LLiquid Scintillator Response to Proton Recoils in the 10-100 keV Range
C. Awe, P.S. Barbeau, J.I. Collar, ∗ S. Hedges, and L. Li Department of Physics, and Triangle Universities Nuclear Laboratory, Duke University, Durham, NC 27708, USA. Enrico Fermi Institute, Kavli Institute for Cosmological Physics,and Department of Physics, University of Chicago, Chicago, IL 60637, USA (Dated: May 9, 2018)We study the response of EJ-301 liquid scintillator to monochromatic 244.6 ± PACS numbers: 29.40.Mc, 29.25.Dz, 28.20.Cz, 95.35.+d
I. MOTIVATION
Thirty years of direct searches for dark matter par-ticles, heavily focused on hypothetical Weakly Interact-ing Massive Particles (WIMPs) [1–4], have returned nounambiguous positive detection. As a reaction to this,a number of new initiatives seek to expand the reachof these efforts. In particular, recent phenomenologicalwork has concentrated on particle models involving low-mass candidates [5], incapable or limited in their abilityto produce signals above the energy threshold of existingWIMP detectors.For dark matter (DM) particle masses of order 1GeV/c , the match in mass of the projectile and a protontarget in hydrogenated organic scintillator results in anincrease in expected interaction rate, and in the maxi-mum proton recoil energy that can be imparted duringa DM-nucleus collision. This is advantageous in the con-text of searches for low-mass Strongly Interacting Mas-sive Particles (SIMPs), a DM possibility that has recentlyattracted renewed interest [6–9]. Furthermore, there areindications that a favorably large ( > ∼
5% around 100 keV.The change in this trend, towards the mentioned QF >
20% in the 0.1-1 keV proton energy range, has beenobserved by one group of experimenters only, duringneutron scattering calibrations employing NE-110 as thetarget [10, 11]. This organic plastic scintillator has ∗ Corresponding author: [email protected] been used for magnetic monopole searches [12, 13]. Inthese calibrations, the expected average scintillation sig-nal was well-below one single photoelectron (SPE). As aresult, less-than-straightforward statistical methods wereinvolved in the determination of the quenching factor.Due to the experimental difficulties that we de-scribe and bypass below, only a handful of measure-ments have been available for organic scintillators in the“turnaround” 1-100 keV energy region, where the QFwould be expected to display an abrupt monotonic in-crease with decreasing energy, in order to match the largevalues found using NE-110. Our dedicated measurementconfirms that this is the case, validating the findings in[10, 11], and the good prospects for use of hydrogenatedtargets in low-mass SIMP searches [9]. In combinationwith all previous data, our results also support an under-laying common physical basis for the production of lightin all aromatic organic scintillators.A second source of motivation for our work is the val-idation of QF models employed during neutron back-ground studies performed within the COHERENT col-laboration [14]. We have employed EJ-301 to charac-terize the flux and energy distribution of prompt andneutrino-induced neutrons at the Spallation NeutronSource (SNS) of the Oak Ridge National Laboratory.Unless controlled, these two sources of neutron back-grounds would be able to compete with Coherent ElasticNeutrino-Nucleus Scattering (CE ν NS), a process recentlymeasured for the first time at this site [15]. The mea-surements described below support the EJ-301 responsemodel used in [15] to demonstrate a negligible neutronbackground contamination of the CE ν NS signal.
II. NEUTRON SOURCE AND EXPERIMENTALSETUP
We follow a method previously implemented in a num-ber of QF studies of materials used in WIMP detectors.It consists of exposing a target detector under investiga-tion to a collimated beam of monochromatic neutrons,while registering scattered neutrons with a backing de- a r X i v : . [ phy s i c s . i n s - d e t ] M a y tector (Fig. 1). Knowledge of initial neutron energy andscattering angle θ is sufficient to define the nuclear recoil(NR) energy deposited in the target, whenever this tar-get is small enough to guarantee single scatters within.Delayed coincidences between target and backing detec-tor, separated by neutron time-of-flight (TOF), are usedto isolate these energy depositions. Gamma sources aretypically used to establish an energy scale for electron re-coils (ERs) in the target detector. In scintillators, a com-parison of the light yield for ERs and NRs of same energyleads to the determination of the QF. A wide range of NRenergies can be studied by varying neutron source energyand/or the angular position of the backing detector(s).There are several difficulties involved in the use of thismethod for proton recoils below 100 keV. First, the neu-tron energy necessary for this is modest, regardless ofvalue of θ . This precludes the use of organic scintillatorswith neutron/gamma pulse-shape discrimination (PSD)for the backing detector, an ideal choice able to reducethe severe gamma background contamination generatedby most neutron sources. For the low neutron energiesrequired, the majority of depositions in the backing de-tector would fall below the few tens of keV necessaryfor optimal PSD [16, 17]. In order to bypass this prob-lem, we employed a large 5 cm diameter x 1.5 cm length LiI[Eu] scintillator, isotopically enriched to > Li[18]. The Li(n, α ) neutron absorption reaction producesa well-defined signal at 3.1 MeV electron-equivalent en-ergy, distinguishable above the energy of most gamma backgrounds. We have successfully employed this alter-native backing detector before, during studies of sub-keVnuclear recoils in germanium [19], using a 24 keV neutronbeam [20]. Our present choice of neutron energy at 245keV was intentional, to profit from a broad resonancein the Li(n, α ) cross-section [21]. MCNP-PoliMi simula-tions [22] indicated that approximately 4% of neutronsentering the front of our backing detector would producethe characteristic signal sought, which was deemed suffi-cient. FIG. 1. Experimental arrangement, with all dimensions toscale, derived from the MCNP-PoliMi simulation geometry.The scattering angle θ and distance L between the LiI[Eu]backing detector and EJ-301 are indicated. A cross section ofthe small EJ-301 liquid scintillator cell is also shown. θ ( ◦ ) recoil energy(keV) L(cm) exposure time(m) integrated charge( × C) ± +3 . − . ± +0 . − . ± +6 . − . ± +0 . − . ± +5 . − . ± +0 . − . ± +8 . − . ± +0 . − . ± +9 . − . ± +0 . − . ± +10 . − . ± +0 . − . ± +14 . − . ± +0 . − . TABLE I. Parameters corresponding to the seven angular measurements performed (see text). The ”number of events”column lists the integrated number of proton recoils with <
25 PE in background-subtracted residual spectra (Fig. 4).
A second obstacle arises from the small light yield ex-pected from protons in this low-energy region. This con-cern was addressed through use of a Hamamatsu H11934-200 Ultra-Bialkali (UBA) photomultiplier (PMT) tomonitor the emissions from the EJ-301 target cell. Atthe 425 nm peak emission wavelength of EJ-301, the UBAphotocathode exhibits a 54% higher quantum efficiencythan a standard bialkali PMT [23]. The PMT was di-rectly coupled using optical RTV to a small ( ∼ − and D − , whichare then accelerated and converted into their positivelycharged counterparts inside of the tandem [25]. The ac-celerator is capable of energies up to 20 MeV, operatingin either direct-current (DC) or pulsed mode. A seriesof bending magnets constrain the energy of the beamand direct it to a suitable neutron-generating material,in our case 650 nm of lithium fluoride (LiF) evaporatedonto a 125 µ m-thick tantalum foil (Fig. 1), a substratechosen to minimize the production of gamma rays. Ourexperiment ran in DC mode and delivered approximatelyone microamp of protons-on-target (POT) over its dura-tion. The total charge delivered to the foil during eachrun was recorded at the accelerator console (Table I). Aproton energy of 2 MeV produces monochromatic ∼ Li(p,n)reaction [26, 27]. A precise neutron energy characteri-zation was determined by switching the proton beam topulsed mode, and measuring the difference in TOF be-tween gammas and neutrons, traveling from the lithiatedfoil to a plastic scintillator (Fig. 2).
FIG. 2. TOF for gammas (left peaks) and neutrons (rightpeaks) produced in the lithium foil, arriving to plastic scintil-lator placed in the forward ( θ = 0 ◦ ) direction. TOF is relativeto a logic signal in phase with POT pulses. Two distances d between foil and scintillator were tested. For each, neutron ve-locity v can be extracted from the expression d/v − d/c = ∆ t ,where c is the speed of light and ∆ t is the n- γ difference inTOF. Both measurements agree within 1%, yielding an aver-age neutron kinetic energy of 244.6 ± A high-density polyethylene (HDPE) collimator sur-rounding the neutron source was 7% enriched in naturallithium, to reduce the capture gamma background reach-ing the EJ-301. The target cell was placed in close prox-imity (19.5 cm) to the source, to maximize the neutronflux at its site. A double-conical tapering of the colli-mator (Fig. 1) blocks the line-of-sight between neutronsource and backing detector, while minimizing the fluxof neutrons moderated in the collimator that can reachthe EJ-301 cell. A borated cylindrical shield around thebacking detector served the purpose of reducing spuri-ous backgrounds from stray thermal neutron capture in LiI[Eu]. The gain of the PMT monitoring the backingdetector was stabilized against a θ dependence throughthe addition of µ -metal magnetic shielding. Laser tools were used for component alignment. Utilizing plumbbobs, projections on a two-dimensional grid drawn onthe ground were used to estimate the small uncertaintiesin θ and L (Table I). III. DATA ACQUISITION AND ANALYSIS
One of two identical signals from the base of the back-ing detector PMT was amplified, shaped, and routedthrough a single-channel analyzer set to generate an out-put for pulses in the 2.1-4.1 MeV energy interval, i.e.,centered around the LiI[Eu] (n, α ) signal. This outputis used as an external trigger to an Acqiris DP1400 fastdigitizer. It recorded the second LiI[Eu] output and EJ-301 PMT signals, at a sampling rate of 500 MS/s, oversufficiently long traces preceding the trigger. A -900Vbias was applied to the EJ-301 PMT, high enough toprovide SPE sensitivity. Traces were bundled and storedto disk. The triggering rate was well-below the maximumthroughput of this system, for all values of θ . FIG. 3. Signals in EJ-301, time-referenced to the onset ofsignals in the backing detector. The horizontal coordinateshows the onset of EJ-301 light emission, and the vertical itslight yield integrated over 80 ns. As in Fig. 2, a time differenceof ∼
90 ns between the indicated n and γ populations matchesthe expected 245 keV neutron TOF over the distance L = 60cm between detectors. The γ offset with respect to t = 0 isdue to a rise-time to reach an analysis threshold for LiI[Eu]signals. Variations by ∼
20 ns in the position of n signals wereobserved for different values of θ , as expected from neutronenergy losses in EJ-301 (Table I). An ER energy scale for EJ-301 was established usingthe 59.5 keV gamma emission from an
Am source,integrated over a 80 ns window. A light yield of 2.1PE/keV was obtained. A small non-linearity of order 5%in the response of EJ-301 to 10-100 keV ERs is neglectedin our analysis (EJ-301 is identical in formulation to theBC-501A studied in this respect in [28]).An offline analysis code was used to extract the am-plitude of EJ-301 pulses preceding the onset of LiI[Eu]trigger signals. Prompt coincidences from gamma scat-tering affecting both detectors were observed, as well as
FIG. 4. Examples of background-subtracted residual spectrafrom proton recoils in EJ-301. The electron-equivalent energyscale, defined via
Am gamma calibration, is 2.1 PE/keV.Solid histograms correspond to simulated spectra for best-fitQF values of 12.3%, 10.4%, 8.2%, top to bottom. The sensi-tivity of these fits is illustrated by dashed greyed histogramsfor QF = 11.3%, 11.4%, 7.2%, respectively. delayed coincidences from neutron interactions. The lat-ter resided in the [-160,-110] ns interval of Fig. 3, for allvalues of θ tested. A second interval [-660,-160] ns wasused to characterize a time-independent background ofspurious coincidences, consisting mainly of SPEs. Fol-lowing normalization to the same time span, the residualspectrum of signals falling in these two time intervalsprovides a background-free picture of energy depositionsfrom proton recoils in EJ-301 (Fig. 4).A detailed MCNP-PoliMi geometry of the setup anddetectors was constructed. Neutrons from a 244.6 ± ± ◦ forwardcone, uniformly sampled. This maximizes the efficiencyof the simulation, while correctly approximating the neu-tron angular distribution [27], and including the effectof the collimator. Post-processing of the output extractsproton recoil energies in EJ-301 for neutron histories pro-ducing (n, α ) reactions in the backing detector. For eachvalue of θ , a fine-grained sampling of QF values withinthe interval 2-20% were tested by the simulation, as fol-lows. Individual proton recoil energies were convertedinto an expected PE yield through a choice of QF, andthe electron-equivalent yield of 2.1 PE/keV. Poisson fluc-tuations around this expectation value generate a simu-lated number of PE for each NR. Multiple scatters andinfrequent carbon recoils (occurring for 16% of histories)were included. However, the smaller maximum recoil en-ergy of carbon nuclei and a modest ∼
1% quenching factor[29, 30] renders their contribution negligible.
FIG. 5. Comparison of experimental and simulated protonrecoil rates in EJ-301. Experimental rates are normalized tothe indicated reference proton current delivered to the lithi-ated foil, using charge, number of events, and exposure timesin Table I. Simulated rates are for the best-fit neutron flux atthe position of the EJ-301, stated in the label.
Simulated residual spectra were normalized to thesame number of events yielding <
25 PE as in the ex-perimental residuals (Fig. 4, Table I). A profile likelihoodestimator returned the QF providing the best fit betweenexperimental and simulated residuals, and its asymmet-ric one-sigma confidence interval (Table I, vertical errorsin Fig. 6). For each θ , the span of proton recoil ener-gies probed by the measurement was extracted from theasymmetric half-width at half-maximum of the simulateddistribution of recoil energies. This is listed as an energyuncertainty in Table I, and shown as horizontal error barsin Fig. 6. For this analysis, the QF was treated as ap-proximately constant over each of these energy spans.The simulation was also used to find the neutron fluxat the EJ-301 position that best matched the experimen-tally observed proton-recoil rates. Following the normal-ization of these rates to a reference current delivered tothe lithiated foil, an excellent agreement between simu-lation and measurement was noticed for all values of θ (Fig. 5), when a best-fitting neutron flux of 880 n/cm sis adopted. This flux matched estimates based on pre-vious operation of this neutron source. The agreementvisible in Fig. 5 provides an important cross-check onthe absence of systematics affecting our QF determina-tion at the lowest recoil energies probed. Neglecting thiscontrol comparison can lead to overestimated low-energyQF values [31]. IV. RESULTS AND CONCLUSIONS
Our new QF measurements are shown in Table I andFig. 6. The figure also displays all previously availabledata for low-energy proton recoils in aromatic organicscintillators. As mentioned in Sec. I, our present mea-surements confirm the EJ-301 response used to demon-strate a near-complete absence of neutron backgrounds
FIG. 6. Low-energy quenching factors for proton recoilsin organic scintillators [10, 11, 29, 30, 32], including presentresults. A dotted black line represents the modified Lindhardmodel proposed in [11], and adopted by the COHERENTcollaboration for neutron background studies in [15]. Protonrecoil light yields from [11] are converted here to a quenchingfactor via a 3 eV mean photon energy for NE-110 scintillation[10], and a 9.2 photon/keV scintillation light yield for ERs inthis material [33]. in a first CE ν NS measurement [15]. The ascending trendin QF found below 100 keV supports the measurementsbelow 1 keV described in [10, 11], and bodes well forplanned use of hydrogenated scintillators in low-massdark matter searches [9].The similar behavior noticeable in Fig. 6 for all aro-matic organic scintillator formulations suggests a com-mon physical basis in their production of light (namely,the excitation and transitions of π -electronic states inbenzenic rings [34]). Specifically, and similarly to whatwas recently reported in [32], our data are much bet-ter described by a modified Birks’ model of scintillationproduction containing a quadratic dependence on protontotal stopping power [35, 36], than by a standard Birks’model [37]. This preference for the modified model canbe quantified at the 5.4 sigma level, under a standardlikelihood ratio test. ACKNOWLEDGMENTS
We are indebted to Chris Maxwell and Chuck Hurlbutat Eljen Technology for their development of the minia-ture EJ-301 cell used in this study, and to John Dunhamfor his assistance setting up the neutron source. Thiswork was supported by NSF award PHY-1506357. [1] M.W. Goodman and E. Witten, Phys. Rev.
D31 (1985)3059.[2] I. Wasserman, Phys. Rev.
D33 (1986) 2071.[3] G. Jungman, M. Kamionkowski and K. Griest, Phys.Rept. (1996) 195.[4] G. Bertone, D. Hooper and J. Silk, Phys. Rept. (2005) 279.[5] C. Boehm and P. Fayet, Nucl. Phys.
B683 (2004) 219;C. Boehm, P. Fayet and J. Silk, Phys. Rev.
D69 (2004)101302; C. Bird, R. Kowalewski and M. Pospelov,
Mod.Phys. Lett.
A21 (2006) 457; J.L. Feng and J. Kumar,
Phys. Rev. Lett. (2008) 231301; D.G. Cerde˜no, C.Mu˜noz and O. Seto,
Phys. Rev.
D79 (2009) 023510;K.M. Zurek,
Phys. Rev.
D79 (2009) 115002; D.E. Ka-plan, M.A. Luty and K.M. Zurek,
Phys. Rev.
D79 (2009)115016; P. Draper et al. , Phys. Rev. Lett. (2011)121805; D. Hooper, N. Weiner and W. Xue,
Phys. Rev.
D86 (2012) 056009; K.Y. Choi and O. Seto,
Phys. Rev.
D88 (2013) 035005; R.C. Cotta et al. , Phys. Rev.
D90 (2014) 013020; K.M. Zurek,
Phys. Reports (2014)91; R. Foot, Int. J. Mod. Phys.
A29 (2014) 1430013.[6] Y. Hochberg et al. , Phys. Rev. Lett. (2014) 171301.[7] Y. Hochberg et al. , Phys. Rev. Lett. (2015) 021301.[8] Y. Hochberg, E. Kuflik and H. Murayama, JHEP (2016) 090.[9] J.I. Collar, arXiv:1805.02646 .[10] S.P. Ahlen et al. , Phys. Rev. Lett. (1985) 181.[11] D.J. Ficenec et al. , Phys. Rev. D36 (1987) R311.[12] S. Ahlen et al. , Phys. Rev. Lett. (1992) 1860.[13] M. Ambrosio et al. , Eur. Phys. J C25 (2002) 511.[14] COHERENT collaboration, arXiv:1803.09183 .[15] D. Akimov et al. , Science (2017) 1123. [16] X.L. Luo et al. , Nucl. Instr. Meth.
A767 (2014) 83.[17] E. Ronchi et al. , Nucl. Instr. Meth.
A610 (2009) 534.[18] Proteus Inc., Chagrin Falls, OH 44022, USA.[19] P.S. Barbeau, J.I. Collar, and O. Tench, JCAP (2007)009.[20] P.S. Barbeau, J.I. Collar, and P.M. Whaley, Nucl. Instr.Meth. A574 (2007) 385.[21] R.B. Murray, Nucl. Instr. (1958) 237.[22] M. Marseguerra et al. , Nucl. Instr. Meth. B213 (2004)289.[23] K. Nakamura et al. , Nucl. Instr. Meth.
A623 (1975) 57.[27] C.L. Lee and X.-L. Zhou, Nucl. Instr. Meth. B152 (1999)1.[28] A. Nassalski et al. , IEEE Trans. Nuc. Sci. (2008) 1069.[29] J. Hong et al. , Astropart. Phys. (2002) 333.[30] V.V. Verbinski et al. , Nucl. Instr. Meth. (1968) 8.[31] J.I. Collar, arXiv:1010.5187 .[32] S. Westerdale et al. , JINST (1959) 151.[35] R.L. Craun and D.L. Smith, Nucl. Instr. Meth. (1970)239.[36] C.N. Chou, Phys. Rev.87