A New Method for Measuring Coherent Elastic Neutrino Nucleus Scattering at an Off-Axis High-Energy Neutrino Beam Target
S.J. Brice, R.L. Cooper, F. DeJongh, A. Empl, L.M. Garrison, A. Hime, E. Hungerford, T. Kobilarcik, B. Loer, C. Mariani, M. Mocko, G. Muhrer, R. Pattie, Z. Pavlovic, E. Ramberg, K. Scholberg, R. Tayloe, R.T. Thornton, J. Yoo, A. Young
FFERMILAB-PUB-13-522-E
A New Method for Measuring Coherent Elastic Neutrino Nucleus Scatteringat an Off-Axis High-Energy Neutrino Beam Target
S.J. Brice, R.L. Cooper, F. DeJongh, A. Empl, L.M. Garrison, A. Hime, E. Hungerford, T. Kobilarcik, B. Loer, C. Mariani, M. Mocko, G. Muhrer, R. Pattie, Z. Pavlovic, E. Ramberg, K. Scholberg, R. Tayloe, R.T. Thornton, J. Yoo, and A. Young Fermi National Accelerator Laboratory, Batavia, IL, 60510, USA Indiana University, Bloomington, IN, 47405, USA University of Houston, Houston, TX, 77204, USA Los Alamos National Laboratory, Los Alamos, NM 87545, USA Virginia Tech, Blacksburg, VA 24061, USA North Carolina State University, NC 27695, USA Duke University, Durham, NC, 27708, USA
We present a new experimental method for measuring the process of Coherent Elastic NeutrinoNucleus Scattering (CENNS). This method uses a detector situated transverse to a high energyneutrino beam production target. This detector would be sensitive to the low energy neutrinosarising from pion decays-at-rest in the target. We discuss the physics motivation for making thismeasurement and outline the predicted backgrounds and sensitivities using this approach. We reporta measurement of neutron backgrounds as found in an off-axis surface location of the FermilabBooster Neutrino Beam (BNB) target. The results indicate that the Fermilab BNB target is afavorable location for a CENNS experiment.
PACS numbers: 29.25.-t, 13.15.+g, 13.40.Em, 23.40.Bw, 95.35.+d
I. INTRODUCTION
The Coherent Elastic Neutrino-Nucleus Scatteringprocess, or CENNS, has yet to be observed since itsfirst theoretical prediction in 1974 by D. Freedman [1].The condition of coherence requires sufficiently small mo-mentum transfer to a nucleon so that the waves of off-scattered nucleons in the nucleus are all in phase and addup coherently. Neutrinos with energies less than 50 MeVlargely fulfill this coherence condition in most target ma-terials. The elastic neutral current interaction leaves noobservable signature other than the low-energy recoilsof the nucleus with energies of up to tens of keV. Thetechnical difficulties of developing large-scale, low-energythreshold, and low-background detectors have hamperedthe experimental realization of the CENNS measurementfor more than four decades. However, recent innovationsin dark matter detector technology have made the unseenCENNS testable.Neutrinos and dark matter are similar in that theyboth exist ubiquitously in the Universe and interactvery weakly. All major dark matter direct detectionsearches rely on the postulation of coherent scatteringof these massive particles off of nuclei. Because of therelatively low momentum transfer, the total interactioncross-section scales as the atomic mass squared of the tar-get material. This is an analogy for low-energy neutrinosinteracting coherently with nuclei. In fact, the CENNSinteractions may prove to be an irreducible backgroundfor future direct detection dark matter searches.Besides its role as a fundamental background in darkmatter searches, measurement of the CENNS process im-pacts a significant number of physics and astrophysics topics, including supernova explosions, neutron form fac-tor, sterile neutrino, neutrino magnetic moments andother non-Standard Model physics.The method we outline uses low energy neutrinos aris-ing from pion decay-at-rest source in the existing highenergy neutrino beam [2]. This differs from other meth-ods for which detectors are proposed to be situated closeto the core of a nuclear reactor [3, 4] or spallation neutronsources [5, 6].In this paper, we present R&D for a measurementof CENNS. We start by discussing the physics moti-vation for the CENNS process in section II. The de-tails of the high-intensity and low-energy neutrino fluxfrom the Fermilab Booster Neutrino Beam (BNB) is ex-plained in section III. The beam-associated backgroundand cosmogenic background measurements at the BNBtarget building are described in section IV, a conceptualCENNS experiment is described in section V, and wesummarize this paper in section VI.
II. PHYSICS MOTIVATION
In the Standard Model, CENNS is mediated by Z vector boson exchange (see FIG. 1). In this process aneutrino of any flavor scatters off a nucleus with the samestrength; hence, the measurement will be insensitive toneutrino flavor and will be blind to neutrino oscillationsamong the active flavors. The dominant cross section fora spin-zero nucleus at an incident neutrino energy of E ν is given by σ νA (cid:39) π E ν [ Zw p + ( A − Z ) w n ] , (1) a r X i v : . [ phy s i c s . i n s - d e t ] N ov FIG. 1. Feynman diagram of the CENNS process. where the Z is an atomic number and A is an atomicmass. νA stands for neutrino-nuclei interaction. Thevector charge of Z to u -quark ( − sin θ w) and Z to d -quark ( − + sin θ w), where θ w is the Weinbergangle, causes the different coupling strength between w p and w n to the proton ( uud ) and the neutron ( udd ), re-spectively. The SM values are w p = G F (4 sin θ w − w n = G F . Since sin θ w (cid:39) . w p is suppressedand the νA cross section at a given neutrino energy iseffectively proportional to the square of the number ofneutrons, ( A − Z ) .Typical values of the total CENNS cross section formedium A nuclei are in the range of ∼ − cm whichis at least an order of magnitude larger than other neu-trino interactions in this energy range (see FIG. 2). Forexample, charged current inverse β decay on protons hasa total cross section of σ ¯ ν e p (cid:39) − cm and elasticneutrino-electron scattering has a total cross section of σ ν e e (cid:39) − cm . The maximum nuclear recoil energyfor a target nucleus of mass M is given by 2 E ν /M whichis in the sub-MeV range for E ν ∼
50 MeV and for typicaldetector materials.In the following sub-sections we briefly summarize theimportant physics cases where the CENNS interactionsplay a significant role.
A. CENNS in Particle Astrophysics
1. Dark Matter Physics
One of the most fascinating problems in Particle As-trophysics is the presence of dark matter. The StandardModel (SM) does not accommodate a suitable dark mat-ter particle candidate; therefore dark matter is crucialphenomenological evidence for physics Beyond the Stan-dard Model (BSM). The common theme of BSM scenar-ios is the introduction of new particles where at least oneis neutral and stable. These new particles in most sce-narios typically have non-gravitational interactions whichare sufficient to keep them in thermal equilibrium in theearly universe. In particular, particles with a mass of theelectroweak scale have a relic density in the right range
Neutrino Energy [MeV]
10 20 30 40 50 60 70 80 90 100 ] cm -38 Cross-section [10 -7 -6 -5 -4 -3 -2 -1
10 110 -e e n -e e n Ar - e n Ar - e n -e x n -e x n -A n coh FIG. 2. Neutrino cross sections on argon target in low energyregion. for a suitable candidate for dark matter.In the limit of vanishing momentum transfer the darkmatter to nuclei ( χA ) cross section becomes σ χA (cid:39) π µ χA [ Zf p + ( A − Z ) f n ] , (2)where µ χA is the reduced mass of the collision. A spin-independent χA interaction corresponds to a couplingto the nucleon density operators characterized by cou-pling constants f p and f n to protons and neutrons, re-spectively. In a wide range of BSM scenarios [7, 8], theHiggs-to-strange quark coupling is the dominant compo-nent of the χA interaction. Since the proton and neutronhave similar strange quark contents, it is usually assumedthat f p (cid:39) f n . The σ χA is, therefore, simplified to be pro-portional to A . This A –scaling of the cross section is avery strong driving force in the direct detection of darkmatter experiments and is analogous to the ( A − Z ) –scaling in CENNS. Moreover, it has been known thatthe CENNS of astrophysical and atmospheric neutrinosare irreducible backgrounds for future generation darkmatter detectors at spin-independent cross-sections. Arecent study showed background limits to future darkmatter searches coming from CENNS interactions of as-trophysical and atmospheric neutrinos [9]. There are afew possible ways to improve the limits by using direc-tional measurements of the neutrino interactions and/ormeasuring time variation of the interactions. However,this CENNS background limit is a robust lower boundwhich can not be substantially reduced. Measuring theCENNS cross section and performing subsequent testsof higher energy neutrino interactions on various targetmaterials will be extremely beneficial to future dark mat-ter experiments. The importance of the CENNS physicscases in dark matter searches is also pointed out in arecent Snowmass report [10].
2. Supernova Physics
The major unsolved problem of a supernova explosionis to understand how the burst of neutrinos transfers itsenergy to produce the shock wave that causes the star toexplode. CENNS plays a major role in an explosion ofa core-collapse supernova [11]. In the core of the dyingstar, neutrinos are scattered, absorbed, and reemittedby super-dense proton-neutron matter. Although yet tobe fully understood, modern numerical simulations showthat neutrino-driven convection eventually causes the gi-ant star to explode. A CENNS cross-section differentfrom the nominal SM prediction could have significantimpact on the understanding of supernova explosions.Moreover, CENNS is an important process for the de-tection of supernova neutrinos. Future large-scale low-energy threshold underground detectors, such as theCLEAN detector [12], will be sensitive to all active neu-trino species in a supernova burst, and will be flavorblind [13]. Hence, detecting supernova neutrinos in sucha detector may provide a total flux and spectrum of neu-trinos from supernova if the cross section of CENNS canbe independently and accurately measured. These re-sults combined with flavor-dependent interaction mea-surements [14, 15] can explain how neutrinos are ther-malized with matter in a supernova.
B. CENNS in Particle Physics
1. Neutrino Oscillations
Neutrino flavor oscillation is a well established physicsphenomenon studied over the last four decades. Neutrinodisappearance and appearance signatures are successfullyexplained by representing the neutrino flavor eigenstatesas a mixture of non-zero mass eigenstates. There hasbeen huge progress in measuring neutrino mixing anglesduring the last decades. Identifying mass hierarchies,measuring CP-phase(s) and determining whether neutri-nos are Dirac or Majorana particles are active topics inthe field. CENNS is a large and well-predicted cross-section in the Standard Model. If discovered at its pre-dicted rate, the CENNS process can become a powerfultool for future low energy neutrino physics, especially forneutrino oscillation experiments.A number of recent anomalous results suggest the exis-tence of a sterile neutrino [16, 17]. In these experiments,an excessive appearance of active-flavor neutrinos is seen.If confirmed, this excess requires a model which has rela-tively large mass differences (∆ m ∼ ) and requiresat least one more mass eigenstate ( m ) in the neutrinomass spectrum. Most of the previous experiments arebased on charged-current measurements, and hence areindirectly inferring the mixing matrix elements. How-ever, the sterile neutrino models can be clearly verified byCENNS interactions. The CENNS interaction is insensi-tive to the differences of active flavors of neutrinos, thus [keVnr] recoil E -2 -1
10 1 10 events/keV/(ton-year) B m -10 · = 2.0 n m B m -10 · = 1.5 n m B m -10 · = 1.0 n m B m -11 · = 5.0 n m FIG. 3. Differential yield as a function of nuclear recoil energyfor different values of neutrino magnetic moment ( µ ν ( ν µ )). ν µ flux of 2 . × ν /cm /s from pion decay-at-rest source isassumed. the measurement will be of total fluxes of active flavorneutrinos. Sterile neutrino oscillations manifest them-selves as a baseline- and energy-dependent disappearance of CENNS interactions. A short-baseline neutrino ex-periment measuring CENNS has the potential to probea wide range of oscillation hypotheses [18, 19].A sensitivity study of a future sterile neutrino searchusing CENNS has been carried out in reference [19].The study assumes neutrino fluxes of 2 . × (6 . × ) ν /cm /sec per flavor at 20 m (40 m) from the piondecay-at-rest neutrino source with one near (20 m) de-tector with 456 kg of liquid argon and four far (40 m)detectors. With this experimental scenario, one can testthe LSND best-fit mass splitting (∆ m = 1 . ) at the3.4 sigma.
2. Neutrino Magnetic Moment
As a consequence of non-zero masses, neutrinos canhave magnetic moments. In the minimally extended SM,Dirac neutrinos of mass m ν have a magnetic momentthrough one-loop radiative corrections [20]. The mag-netic moment is given by µ ν = 3 G F m e m ν π √ µ B (cid:39) . × − (cid:16) m ν (cid:17) µ B , (3)where G F is Fermi constant, m e is electron mass, and µ B (= e/ m e ) is Bohr magnetons. This predicted valuein an extended SM is exceedingly small to be measured.However, beyond the SM (BSM) models commonly pre-dict larger values of µ ν , and hence any measurementof excessive neutrino magnetic moment would be a sig-nature of BSM physics [21]. There are several con-sequences of the neutrinos having large magnetic mo-ments. The neutrino-electron scattering cross sectionwould be modified in low energies. Neutrinos would fliptheir spin in strong external magnetic fields which is, forexample, a natural configuration for the core region ofstars. Heavier-mass neutrinos would decay radiatively tolighter-mass neutrinos and emit photons.The best direct experimental limits result for ν − e scattering is from GEMMA experiment, µ ν (¯ ν e ) ≤ . × − µ B [22]. For muon neutrino scattering, the bestlimit is less stringent: µ ν ( ν µ ) ≤ . × − µ B [23].The most stringent limits are from astrophysical obser-vations with several assumptions. For example a model-dependent analysis of plasmon decay in red giant evolu-tion [24], and an analysis of neutrino spin-flip precessionin Supernova 1987A set limits of µ ν ≤ − µ B [25].A finite neutrino magnetic moment can be observed inthe recoil spectrum of CENNS. The magnetic scatteringcross section is given by [20], (cid:18) dσdE R (cid:19) m = πα µ ν Z m e (cid:18) − E R /E ν E ν + E R E ν (cid:19) , (4)where α is the fine structure constant, E R is the recoilenergy of nuclei. FIG. 3 shows the event rates as a func-tion of energy thresholds in a germanium detector withpion decay-at-rest ν µ flux of 2 . × ν /cm /s for vari-ous magnetic moment contributions. Future detectors forsub-keV recoil energy thresholds would begin to directlytest new regimes of neutrino magnetic moment.
3. Non-Standard Model Interactions
CENNS is a well-predicted cross-section in the Stan-dard Model. Therefore any deviation from the predictedvalue would be an indication of BSM physics. Any non-standard interactions (NSI) which are specific to the in-teractions of neutrinos and quarks can be parameterizedin a relatively model-independent way. An effective La-grangian of a neutrino with a hadron in the parametriza-tion of (cid:15) ij can be described as [26, 27]; L NSIνH = − G F √ (cid:88) q = u,dα,β = e,µ,τ [¯ ν α γ µ (1 − γ ) ν β ] (5) × ( ε qLαβ [¯ qγ µ (1 − γ ) q ] + ε qRαβ [¯ qγ µ (1 + γ ) q ]) , where the ε parameters represent either non-universal ( α = β ) or flavor-changing ( α (cid:54) = β ) interactions. Manyof these parameters are quite poorly constrained, andCENNS experiments can improve sensitivity by an orderof magnitude [26, 28, 29]. The cross section for CENNSof ν α off a spin-zero nucleus (A) is given by dVee ˛ -1 -0.5 0 0.5 1 u V ee ˛ -0.8-0.6-0.4-0.200.20.40.60.81 Ar 10% sysAr 5% sys Allowed by CHARM
FIG. 4. Allowed regions (red and yellow shaded areas) at90% C.L. assuming measurement of the SM-predicted CENNSrate, for (cid:15) µVee and (cid:15) dVee for 1 ton-year liquid argon detector at5 × ν/ cm /s per flavor of pion decay-at-rest neutrino flux,assuming 5% or 10% of systematic uncertainty in measure-ment. The energy threshold is assumed at 25 keV nr (nuclearrecoil). The shaded elliptical region corresponds to a sliceof the CHARM-experiment-allowed NSI parameter space, for (cid:15) qAee =0. (cid:18) dσdE (cid:19) ν α A = G F Mπ F (2 M E ) (cid:20) − M E k (cid:21) ×{ [ Z ( g pV + 2 ε uVαα + ε dVαα ) + N ( g nV + ε uVαα + 2 ε dVαα )] + (cid:88) α (cid:54) = β [ Z (2 ε uVαβ + ε dVαβ ) + N ( ε uVαβ + 2 ε dVαβ )] } , where g pV = ( − θ W ), g nV = − are the SM weakconstants. FIG. 4 shows allowed regions for (cid:15) µVee and (cid:15) dVee , for 1 ton-year of liquid argon detector data assum-ing high-intensity pion decay-at-rest neutrino flux. Theshaded elliptical region corresponds to constraints by theCHARM experiment [30]. Hence a CENNS experimentat an intense stopped-pion neutrino source would havesignificant sensitivity to currently-allowed NSI interac-tion parameters. C. CENNS in Nuclear Physics
Determination of the neutron distributions in nucleiis important not only for fundamental understanding ofnuclear physics, but also because of important implica-tions for astrophysics. For example, the primary physicsquantities of neutron stars such as masses, radii, andcomposition are determined using the equations of statesof neutron-rich nuclei. The equation of state is relatedto the nuclear symmetry energy, which is defined as, E ( n, δ ) (cid:39) E ( n ) + E sym δ and δ = ( n n − n p ) / ( n n + n p ),where n n and n p are the number densities of neutronsand protons. The symmetry energy is strongly corre-lated with the skin thickness of neutrons [31], and hencethe radii of neutrons. Therefore, the size of neutron starscan be predicted more precisely based on better measure-ments of the equation of state. Traditional methods ofmeasuring neutron radii through hadronic scattering re-port typical uncertainties of order 10% [32].The CENNS interaction is especially sensitive to neu-tron numbers in target nuclei, which provides a cleanway to measure the neutron part of nuclear form factors.At low momentum transfer, the form factor F ( Q ) ∼ Q values, small deviations from co-herence occur as higher-order terms of the nuclear formfactors come into play [33]. The CENNS cross section ofthe spin-zero nuclears is given by, (cid:18) dσdE (cid:19) νA = G F π Q w F ( Q ) M (cid:20) − M E R E ν (cid:21) , (6)where M is the nuclear mass, Q w is the weak charge and Q = √ M E R . The form factor F ( Q ) can be expandedas: F n ( Q ) ∼ N (cid:18) − Q (cid:104) R n (cid:105) + Q (cid:104) R n (cid:105) + · · · (cid:19) , (7)where (cid:104) R in (cid:105) are the even moments of the neutron density.Such deviations are observable as small distortions of theexpected recoil spectral shape and can be exploited tomeasure nucleon density distributions. With good con-trol of spectral shape uncertainties, multi-ton-scale ex-periments could make meaningful measurements of theneutron radius (cid:104) R n (cid:105) / and potentially higher-order mo-ments.According to reference [32], a exposure of 3.5 ton-yearwith a liquid argon detector with neutrino flux of 3 × ν/ cm /s per flavor is required to measure the 2ndand 4th moments of the form factor. The experimentalrequirements are challenging to reach in the near future;however, it is possible to determine the neutron radiusto a few percent by measuring neutron form factor withsufficient accuracy. The precise measurements of neutronradii then improve the predictive power of the equationof state of neutron matter, and thus the knowledge of thesize of neutron stars [31, 34]. D. Summary
In order to achieve the above physics goals, a phasedapproach is most appropriate, depending on the availableneutrino beam power and detector technology. 1. The first-generation CENNS experiment would bethe discovery of the CENNS interaction and mea-sure the cross-section with ∼
10% of accuracy, forexample at Fermilab. The experiment can be car-ried out with existing dark matter detector technol-ogy, existing beamline and target station at Fermi-lab. The result would be sensitive to the NSI rangesas well.2. The second-generation experiment would be theprecision measurement of the CENNS cross-section. The accurate measurement of the neu-trino flux, assuming cross section is exactly known,would be a powerful tool for neutrino oscillationstudy [35] and future low energy neutrino exper-iments. This would also allow an initial series ofmeasurements of supernova-related neutrino crosssections on a variety of targets [36], where they haverarely been measured. The precision measurementof the CENNS cross section will be a valuable inputto the next generation of dark matter experiments.3. The third-generation CENNS experiment woulduse a high-intensity neutrino beam and large-scaleneutrino detector with a lower energy threshold.The goal would be a search for the neutrino mag-netic moment, measurement of the neutron formfactor, and search for possible deviations of the SM.The major focus in this paper is the first-generationof CENNS experiment – the discovery of the CENNS.There are a few existing intensive pion decay-at-rest neu-trino sources, for example the Spallation Neutron Sourceat Oak Ridge National Laboratory [37]. In this paperwe present an alternative promising setup at an existingneutrino beam at Fermilab. This is a unique idea of us-ing the neutrinos at a location off-axis of existing highenergy neutrino beam [2].
III. LOW ENERGY NEUTRINO SOURCE ATFERMILAB
Fermilab has two major neutrino beamlines (seeFIG. 5); the Neutrinos at the Main Injector (NuMI) andthe Booster Neutrino Beam (BNB). The energy range ofthese two neutrino sources on-axis is in the GeV range,which is too high to satisfy the condition for dominanceof coherent scattering. We found the far-off-axis ( > ∼
50 MeV. The BNB source has substan-tial advantages over the NuMI beam source owing to sup-pressed kaon production from the relatively low energy8 GeV proton beam on the target. Therefore, pion decayand subsequent muon decay processes are the dominantsources of neutrinos. At the far-off-axis area, the detec-tor can be placed close enough to the target to gain alarge increase in neutrino flux due to the larger solid an-gle acceptance. An initial study using the existing BNBMC has confirmed that this approach is promising.
FIG. 5. Fermilab neutrino beam lines: the Booster Neutrino Beam (BNB, red line) and Neutrinos at the Main Injector beam(NuMI, green line) [38]. The left insert figure shows the configuration around BNB target building (MI-12) area [39]. The redcross in the figure indicates the location of the target. No facility equipment occupies the area near the potential experimentsite.
The Fermilab Booster is a 474-meter-circumferencesynchrotron operating at 15 Hz. Protons from the Fer-milab LINAC are injected at 400 MeV and acceleratedto 8 GeV kinetic energy. The structure of the beam is aseries of 81 proton bunches each with a 2 ns width and19 ns apart. The maximum average repetition rate forproton delivery to the BNB is 5 Hz and 5 × protonsper pulse. The repetition limit is set by the horn designand its power supply. The target is made of berylliumdivided in seven cylindrical sections in a total of 71.1 cmin length and 0.51 cm in radius. In order to minimize up-stream proton interactions, the vacuum of the beam pipeextends to about 152 cm upstream of the target. Thehorn is an aluminum alloy toroidal electromagnet withoperating values of 174 kA and maximum field value of1.5 Tesla. A concrete collimator is located downstreamof the target and guides the beam into the decay region.The air-filled cylindrical decay region extends for 45 me-ters. The beam stop is made of steel and concrete. De-tails of the Fermilab BNB neutrino fluxes can be foundin [40].At very far-off-axis the pion decay region is no longer apoint source and the angle from on-axis is not a well de-fined quantity. Moreover, the geometry around the targetarea and shielding should be properly taken into accountin the neutrino flux calculation as the secondary hadronicprocesses in the shielding material also produce pions andhence neutrinos. In particular, the pion production fromthe 8 GeV Booster proton beam on the beryllium target,the multiplicity of pion production from the subsequent π - p interactions and defocused π − s from the horn all re-quire a well modeled MC study.In order to understand the neutrino flux at BNB far-off-axis, we adapted the Booster Neutrino Beam MonteCarlo (BNB MC). The BNB MC uses the Geant-4 frame-work for propagating particles, for electromagnetic pro-cesses, hadronic interactions in the beamline materialsand the decay of particles. The geometry of the targetarea and beamline is accurately modeled. The doubledifferential cross sections of pion and kaon production inthe simulation have been tuned to match external mea-surements. This is true for the hadronic cross sectionsfor nucleons and pions as well [40]. The original BNBMC, however, contains a hard-coded tracking thresholdcut to remove stopping pions (defined as below 1 MeV inkinetic energy). In fact, the stopping pions are the domi-nant neutrino source at far-off-axis. The cut does not af-fect any previous on-axis Booster Beam experiments suchas MiniBooNE and SciBooNE which focus on above-100-MeV neutrino interactions.The BNB MC simulation was carried out in neutrinomode with 173 kA horn current and 8 GeV proton mo-mentum. FIG. 6(a) shows the angular distribution ofthe neutrino flux 20 m away from a reference point ofthe upstream end of the decay pipe where the angle ismeasured from on-axis. The flux of the neutrinos, atthe 32 kW maximum Booster power (5 × protonson target (POT) per pulse), is estimated to be about10 ν /cm /pulse per flavor with 5 Hz frequency within apulse width of 1.6 µ s. Hence, the neutrino flux per unit q cos -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 POT] · /5 /cm n Flux [
10 (a) BNB neutrinos angular dist. e n m n m n [MeV] n E [Arbitrary Unit]
10 < 0.7 q BNB neutrinos at cos e n m n m n FIG. 6. Estimated neutrino flux from modified BNB MC in ν -mode, 173kA horn current and 8 GeV booster beam con-figuration. The neutrino flux is normalized per 5 × pro-tons on target. (a) The angular dependence of the neutrinofluxes for different flavors. The flux becomes uniform belowcos θ < .
7. See text for the definition of θ . (b) Energy spec-trum of neutrinos below cos θ < . time is about 5 × ν /cm /s. FIG. 6(b) shows the en-ergy spectrum of neutrinos at angles less than cos θ < . π + → µ + ν µ ) produces aprompt and monochromatic ν µ at 29.9 MeV. The µ + thendecays on a 2.2 µ s timescale to produce a ¯ ν µ and a ν e withenergies between 0 and m µ /
2. In FIG. 6(b), the ν µ , ν e and ¯ ν µ spectra follow the stopping π + decay kinematics.The small ν µ bump at ∼
100 MeV is due to the neutri-nos from µ − capture on nuclei. The peak at 235.3 MeVis from kaon decay at rest. These ν µ s above 55 MeVare potential background sources since the interaction ofneutrinos may scatter off neutrons from nuclei nearby orinside the detector. The existing radioactive shielding at the BNB targetarea is extensive and carefully thought out in order to sat-isfy the Fermilab radioactive safety regulations [41] (seeFIG. 7). The target itself is located ∼ × neutrons per 10 POT(year) are expected to be initially produced at the tar-get. These neutrons are produced in the forward beamdirection with a maximum kinetic energy of ∼ ∼ × neutrons/ m per 10 POT. Accordingto a simple scaling of neutron shielding, an additional ∼ >
200 MeV) from kaon decay. However,the distance from the NuMI target to the BNB far-off-axis site is more than 200 m away and the NuMI neutri-nos can be vetoed out using beam trigger information.Therefore the neutrinos from the NuMI beamline shouldbe significantly suppressed.Beam-uncorrelated backgrounds are mitigated by theBNB beam window; the timing allows a factor of5 × − rejection ( duty factor , here after) of steady-statebackgrounds assuming a 10 µ s detector time window.The total detector beam-on livetime per year is only ∼
26 min(=5 × − × year). Timing of individual eventsin the detector can be known to within ∼
10 ns usingdetectors with fast timing. Furthermore, these back-grounds can be subtracted using beam-off data. Cosmicray-related backgrounds will be significantly reduced bythe water shielding veto system.
FIG. 7. The top-down and elevation views of the BNB targetbuilding. The SciBath detector was operated at location (A)and the EJ-301 measurement carried out at location (B). Thedrawing is taken from [39] and modified. The red-filled circlein the top figure indicates the upstream end of the targetposition.
IV. NEUTRON BACKGROUNDSMEASUREMENT
A commercial EJ-301 liquid scintillator neutron de-tector and a newly-developed neutral particle detector,named
SciBath [42, 43] were used to measure the neu-tron backgrounds in the BNB target building.
A. EJ-301 liquid scintillator
To obtain a rough estimate of the neutron backgroundfrom the Booster beam, we attempted to measure theneutron flux with a commercial liquid scintillator detec-tor (Eljen 510-50x50-1/301 Liquid Scintillation DetectorAssembly, sealed system with 5” ETEL-9390KB PMTand EJ301 scintillator). The PMT signals were recordedfrom 3 µ s before to 20 µ s following the beam trigger usinga CAEN V1720 250 MS/s, 12-bit, 2 Vpp digitizer. Thescintillation response of the cell to gammas of various en-ergies was calibrated using the Compton edges of Ba,
FIG. 8. Calibration of the energy and pulse shape discrimi-nation (PSD) parameter F90 for the EJ301 scintillator detec-tor with neutron (
Cf) and gamma ( Na) sources. The511 keV annihilation gamma is easily visible in the Nadata. Neutrons are defined as events having F90 between0.76 and 0.91 and energy below the digitizer saturation point(around 2 MeV ee , electron equivalent energy, on this scale;saturating events are excluded from this plot) and above thepoint where the gamma and neutron F90 distributions merge(around 200 keV ee on this scale). Cs, and Na sources, from which the energy of pro-ton recoils can be obtained using Table 1 of reference [44].(The scintillation light output for 1 MeV proton recoils isquenched by a factor ∼ ∼
700 p.e. for neutron eventsdue to the slower scintillation pulses). We have measuredbeam-induced events with energies up to 8000 p.e., or > Cf and Na sources. Based on this calibration, neu-tron events will have F90 in the range 0.76-0.91, whilegammas have faster pulses with F90 > > s] m time [-4 -2 0 2 4 6 8 10 12 PSD Parameter
PSD Parameter vs Time neutronsgammas s] m time [-4 -2 0 2 4 6 8 10 12 energy [photoelectrons] pulses.npe Pulse energy vs time
FIG. 9. Top: Pulse shape discrimination parameter of scin-tillation events in the EJ301 detector vs time relative to thebeam trigger. Bottom: Energy of scintillation pulses mea-sured in the 5” EJ301 detector vs time. The color scale foreach point shows the PSD parameter, with darker colors beingmore neutron-like.
FIG. 9 top shows the F90 parameter vs detection timefor events in the 50–700 p.e. range. The tail of eventswith F90 < µ s beam spill is evident in the region from -0.6 to1.1 µ s on this time scale, and the events in this regionare overwhelmingly gamma-like; after the spill, the rateis dominated by neutron-like events. The rate of neutron-like events peaks partway through the beam spill, thendecays away with a characteristic time of a few µ s. FIG. 9(bottom) shows the event energy as a function of time,from which it is clear that the energy of the neutron-like events also decays with the same few µ s timescale.Both of these observations are roughly what one wouldexpect from neutrons gradually losing energy to elasticscattering in the shielding material and the building.Although this measurement lacked precise calibrationand required a small analysis window, with this first lookwe were able to determine the overall scale of the neu- tron background. Very few cuts were placed on the anal-ysis, but those that were applied should have had theeffect only of rejecting nuclear-recoil events while admit-ting a minimal amount of electron-recoil events. Themeasured rate of recoil-like events in the 0.3–1.6 MeVrange (assuming protons) in the liquid scintillator de-tector is ∼ > × − /cm /pulse (pulse (cid:39) × POT) neutronswith energy above 0.3 MeV at 19 m from the target. Amore in-depth characterization of the beam-induced neu-tron flux requires a detector with larger mass, more dy-namic range, and better particle discrimination, such asthe SciBath detector.
B. SciBath detector
The SciBath detector is a prototype for the proposedFINeSSE detector which is a 13 ton, fine-grained, liq-uid scintillator neutrino tracking detector [47]. Whilethe detection concept was originally optimized to be afine-grained neutrino tracker, it is also an excellent neu-tron detector. Below, we show results from a 2-monthmeasurement of the beam-correlated neutron flux (10-200 MeV) at the BNB target building. The SciBath de-tector will be described briefly here. More details aboutthe SciBath detector will appear in a future publication.
1. Detector Description
The SciBath detector is an 82 L, optically-open bathof mineral-oil-based liquid scintillator that serves as bothan active target and scintillator. Scintillation light isproduced by the recoiling charged particles from neutralparticle collisions with the mineral oil or by incomingcharged particles from outside the detector. This scintil-lation light is absorbed by a square 16 ×
16 array of wave-length shifting (WLS) fibers, oriented along each detectoraxis, with a spacing of 2.5 cm (i.e. 768 total fibers). Thelight entering each fiber is Stokes-shifted and re-emittedisotropically. Some of the wavelength-shifted light is thentransported by total internal reflection to a multi-anodePMT where it is read out and digitized by the DAQ. WLSfibers shift the ultraviolet bulk scintillation light to bluewhere it more effectively couples to the PMT quantumefficiency peak. A schematic of the detector is shown inFIG. 10.The liquid scintillator has a base of mineraloil combined with 15% pseudocumene (1,2,4-trimethylbenzene, C H ) by volume and 1.5 g / LPPO (2,5-diphenyloxazole, C H NO). The mixturewas created for this detector and was continuouslypurged with N . It is very similar in composition tocommercially available liquid scintillators EJ-321L [48]0 FIG. 10. A schematic drawing of the SciBath detector with its(45 cm) active volume indicated along with the other majorcomponents. and BC-517 H [49], but it lacks tertiary wavelengthshifters such as bis-MSB or POPOP. The scintillatoremission peaks at approximately 370 nm, and the atten-uation length for this light is over 1 m in the detectorand is adequate for the WLS fiber spacing. The 1.5 mmdiameter WLS fibers have an absorption peak at 345 nm,and reemission peaks at 435 nm which matches the peakquantum efficiency of the PMT. Approximately 8% ofthis reemitted light in the WLS fiber is collected at thePMT.The SciBath optical properties were calibrated withcosmic ray muons and an LED pulser system. A mini-mum ionizing muon will deposit approximately 65 MeVinto the SciBath detector and this yields approximately400 detected p.e.. The energy deposit to light output is6 p.e / MeV, and we found this calibration to be stableto within 5% over the entire 2 month run. Birks’ law isused to model quenching effects for large dE / dx parti-cles (e.g. protons). The Birks’ law coefficient kB used inthe Monte Carlo simulation is 0.013 g cm − MeV − whileKamLAND reports 0 . ± . − MeV − for thecommercially similar BC-517H [50]. A pulsed LED sys-tem was coupled to the opposite end, with respect to thePMT, of each WLS fiber. Low-light LED pulses wereused to measure the single p.e. response of the PMTsand calibrate the SciBath DAQ. These LED calibrationswere performed every three weeks, and the gains were stable to within 10% throughout the entire run. In fact,they were stable when compared to a previous deploy-ment six months prior.Each PMT is mounted to a custom, Indiana-designed,“Integrated Readout Module” (IRM) which serves asboth a digitizing readout and physical mounting for thePMT. They are built on a VME form factor, but theyare externally powered and connectivity is establishedthrough 1-Gigabit ethernet (in lieu of the VME powerand connectivity standard). The front-end electronicsof the IRM shapes and stretches the incoming pulses toenable simultaneous nanosecond timing resolution andspectroscopy with 20 MS / s, 12-bit flash ADCs. Addi-tional processing with onboard FPGAs and an ARM-9microcontroller digitize and transfer 64 PMT channelssimultaneously. For data collection, the DAQ was exter-nally triggered on the beam for 20 ms with a 1/3 p.e.threshold per channel and 100 µ s of pre-trigger data.The LED calibration runs were also externally triggered,but only recorded 1 µ s of data with no zero-suppressingthreshold.To exploit the tracking capabilities of 768 WLS fibersfor a large number of events, fast algorithms were devel-oped to determine the track-like properties of each event.The first four statistical moments of the WLS fiber lightoutput are calculated for each axis. A principal com-ponent analysis is then performed giving characteristiceigenvalues and eigenvectors of the fiber hit distribution.Point-and track-like objects can be discriminated by theircharacteristic eigenvalue spectra. Additionally, a pair oflikelihoods are created to further discriminate point- andtrack-like events. In this analysis, event topology is notused to construct the beam-correlated neutron spectrum,but it is used as a quality cut to select track-like eventsfor the direction spectrum.
2. Results
The SciBath detector was placed about 20 m away fromthe BNB target at nearly 180 ◦ with respect to the beamdirection, and the detector position is shown schemati-cally in FIG. 7. SciBath recorded ten-minute beam-ondata runs starting on February 29, 2012 and ending onMay 3, 2012 with a 95% total livetime. After the BNBshut down on April 23, 2012, various calibrations wereperformed. During the entire run, 4 . × protons ontarget (POT) were delivered to the BNB target. Approx-imately 5.5 weeks of production-quality data are used inthe analysis below, and this data set contains 3 . × POT. The remainder of the time was used for LED cali-brations and other systematic checks. A total of 2.5 TBof data was collected with the majority of events hav-ing low fiber multiplicity ( <
5) these were unused in theanalysis.FIG. 11 shows the distribution of events in time aroundthe beam window for various p.e. subgroups. Theblack trace with the highest count rate is all events with1 s) m Time from Accelerator Trigger (110 115 120 125 130 135 140
Counts per 100 ns p.e. > 2060 < p.e. < 200p.e. > 200 FIG. 11. The time distribution of events around the beamwindow for the given selection of p.e.. The dominant, blacktrace shows all events with p.e. >
20, the red trace selects60 < p.e. < > p.e. >
20. The red colored trace is the group of eventswith 60 < p.e. <
200 that has an excess of events abovebackground for a few µ s after the beam pulse. Afterwhich, the count rate returns to pre-beam, backgroundlevels. This is consistent with high-energy neutrons los-ing energy in the shielding, slowing down, and arrivingat delayed times. On the other hand, the blue coloredtrace selecting events with p.e. >
200 does not show anappreciable excess, and its count rate returns to back-ground levels quickly after the beam pulse. The rateimmediately after the beam for the p.e. >
20 data re-mains significantly elevated above background levels fora longer timescale ( ∼ µ s). This is consistent with the2.2 MeV, neutron-capture gamma rays from the hydrogenin mineral oil. For low event rates, neutron capture tag-ging can be used to discriminate primary neutrons fromgamma rays, but this is not possible here because of thehigh event rate per beam spill. Correlating a specificneutron-capture candidate to a specific neutron primaryscatter is impossible.A minimal set of cuts is used to select events for ana-lyzing the neutron energy spectrum and the high-energyneutron direction spectrum. A 3 µ s window surroundingthe beam from 120 µ s to 123 µ s after the accelerator trig-ger is used to select in-beam events (see FIG. 11). Also,background events are selected in a 10 ms window from9 ms to 19 ms after the beam trigger and scaled for sub-traction. For the neutron energy spectrum, events withp.e. >
60 are selected to minimize the gamma-ray con-tamination, and for the direction spectrum events withp.e. >
700 are selected to choose track-like events. (+z) q cos -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 + y ) fi ( + x f FIG. 12. Direction spectrum for high energy proton recoilswith track-like fiber hit distribution. In our right-handed co-ordinate system, + z is pointed towards the beam target and+ y is vertical. Back-projecting the peak of this distributionpoints in line with the beam, but about 10 m upstream of thetarget.
3. Direction Spectrum
The direction-spectrum for high-energy proton recoilswith track-like detector response is measured. In addi-tion to the p.e. and timing cuts described above, eventsare required to be reconstructed within the inner 20%fiducial volume and a modest set of track-like qualitycuts are made. FIG. 12 shows the proton recoil directionspectrum for energetic proton recoils after cuts. Back-projecting the peak of the direction spectrum locates apossible neutron source that is approximately 10 m up-stream of the BNB target. The spatial distribution ofpoint-like events within the SciBath detector corrobo-rates this result. The tracking capabilities were vali-dated against the cosmic ray muon background and muonflux from the NuMI beam during a previous deployment.When validated against the cosmic ray spectrum, our re-sults agree with the results of Mei and Hime [51] andMiyake [52] to within 10%.
4. Neutron Energy Spectrum
To analyze the neutron energy spectrum, the in-beamp.e. spectrum is background-subtracted for the entiredata set. As shown in FIG. 13, the in-beam rate clearlydominates the background rate when scaled for the to-tal beam exposure time of 23 s. The background sub-tracted data shows a cutoff at 1600 p.e. and this is con-sistent with the maximum SciBath response to a single,200 MeV proton recoil. Higher p.e. events are occasion-ally observed, but their origin is consistent with hadroniccascades and multiple, energetic scattering events.The neutron energy spectrum is then unfolded from the2 -2 C o un t s BeamBG -1 C o un t s Data Fit
FIG. 13. The p.e. spectrum used to unfold the neutron energyspectrum. Variable width bins are used with 10 p.e. binsbelow 1000 p.e. and 100 p.e. bins above. The insert figureshows a comparison between in-beam measurements (blackline) and background measurements (dashed line: 9 to 19msec off from the beam trigger). p.e. spectrum by using the SciBath detector response ascalculated with a Monte Carlo simulation (MC). Fromthe results of the direction spectrum, we simulated a di-verging beam of neutrons with a large cross-sectional areaimpinging on the SciBath detector. The simulation showsthat SciBath has a 0.19 m effective cross-sectional areafor neutron acceptance. Neutrons were uniformly gener-ated up to 200 MeV in 20 MeV bins, and the simulationthen tallied the total p.e. response for each 20 MeV neu-tron energy bin. The p.e. response was binned in thesame way as the data, 10 p.e. bins for p.e. < ≥ with the systematic uncertainties added inquadrature with the fit uncertainty. The total energy res-olution is approximately 30% near the 60 p.e. threshold,and this gives an effective neutron energy threshold of ap-proximately 10 MeV. From the unfolded neutron energyspectrum, we find the total number of neutrons above10 MeV per pulse per m is 6 . ± .
7. Shielding the lowenergy neutron flux should not be challenging, but shield-ing will moderate high energy neutrons to potentiallydangerous energies in the CENNS detector. With thisin mind, the neutron flux above 40 MeV is particularlydangerous as a background, and we measure 2 . ± . above 40 MeV. Above 200 MeV,the SciBath detector loses sensitivity because recoiling N e u t r o n s a t m f r o m T a r g e t ( m - p u l s e - ) ( p u l s e ≈ . × P O T ) FIG. 14. The measured neutron energy spectrum by Sci-Bath 20 m behind the proton target is shown. We measure3 . ± .
38 neutrons per m per beam pulse above 40 MeV,and the low energy bin is strongly influenced by the detectorthreshold. The SciBath sensitivity above 200 MeV is signifi-cantly reduced, and these energy bins have large uncertainties. protons at these energies are no longer fully containedby the detector. Fits above 200 MeV show very littlesignificance, and the correlation matrix for the fit showsthat we are unable to differentiate higher energy neutronsfrom 200 MeV neutrons.
5. Systematic Uncertainties
In the analysis, we identified four classes of uncertain-ties to the neutron energy spectrum: energy scale calibra-tion, fiducial cut, fit uncertainty, and the threshold. Thedominant uncertainty above 60 MeV is due to extrapolat-ing the energy scale calibration defined at approximately400 p.e. (6 p.e. / MeV) from cosmic ray muons to higherenergies. We found that this conversion factor varied by5% for a number of reasons: uncertainty of the muon pathlengths, detector energy resolution, p.e. counting statis-tics, light collection efficiency as a function of position,muon input into the MC, and analysis cuts. Above the10 MeV neutron energy threshold, the variation of theBirks law coefficient kB had a negligible impact whencompared to the other systematic uncertainties.At low neutron energy, the choice of fiducial cut, un-certainty of the p.e. threshold, and the fit contributeroughly equally to the total uncertainty. The extractionof the neutron energy spectrum with the unfolding pro-cedure should be independent of the choice of the centraldetector fiducial if the MC is correct. The neutron en-ergy spectrum in FIG. 14 uses the entire detector, butwe found very little variation even down to 10% of the3total volume. Its effect on low neutron versus high neu-tron energies can be understood, because attenuation atthe detector edges is stronger for low energy neutrons,whereas high energy neutrons are more penetrating pro-duce longer track proton recoils which with average posi-tions closer to the center of the detector. Because we donot have neutron-gamma discrimination at low energies,we set the p.e. threshold to 60 to remove gamma raysbelow 10 MeV. Due to gain shifts during the run and theextrapolation of the energy calibration to low energy, wefound that a 10% variation in threshold was reasonable,and we used the MC to examine this variation on theunfolded neutron spectrum. As expected, the thresholdwill vary the first bin (10-20 MeV) very strongly, but hasno effect above 40 MeV.
6. Cosmogenic Neutron Flux C o un t s FIG. 15. Raw cosmogenic p.e. spectrum (black) with doubleexponential fit to neutron data (red). The unfolding pro-cess fits to the double exponential (blue markers), and theexpected Gordon p.e. spectrum is also overlaid (green).
For 10 ms after each beam trigger, we collected back-ground events with a total exposure of 8 . × s. Theraw p.e. spectrum is shown in FIG. 15, and the peak cen-tered at 400 p.e. contains the minimum-ionizing, cosmicray muons. To extract the neutron p.e. spectrum, the to-tal p.e. plot is fit to a double exponential plus the muonresponse functions as calculated by the Monte Carlo. Thedouble exponential is then fit with the same least-squaresfitting procedure that was used for the in-beam data set.Gordon et al. [53] give a parameterization of the expectedbackground neutron flux. For comparison, we applied ourMC response function to the Gordon spectrum to gener-ate the expected p.e. spectrum we would measure in our T o t a l N e u t r o n s p e r D a y p e r M e V FIG. 16. The unfolded neutron spectrum (red) overlaid withthe Gordon neutron spectrum with 10 MeV neutron energythreshold (black trace) and 5 MeV threshold (black) markers. detector. The p.e. spectrum from Gordon was scaledby the effective area for neutron acceptance and by thetotal exposure time. To match the measured data, andadditional scale factor 2.4 was required. The overlaid p.e.spectra are shown in FIG. 15. FIG. 16 shows the unfoldedneutron energy spectrum and the expected neutron spec-trum from Gordon. Again, the Gordon spectrum requiresa factor of 2.4 to match the neutron spectrum unfoldedfrom our data. Aside from the factor of 2.4, our raw p.e.and unfolded neutron energy spectra shapes agree wellwith the parameterizations from Gordon above 20 MeV.Our disagreement in the lowest energy bin seems to beindicative of threshold effects. The uncertainties shownare from the fit only, but the systematic uncertainty issimilar to those in the in-beam data.
V. CENNS EXPERIMENTA. CENNS Detector
Liquid Argon (LAr) has several advantages as a detec-tion medium. As in all of the noble liquids, LAr is nat-urally transparent to its own scintillation light and canbe made very pure, leading to long attenuation lengthsfor the UV photons. Most critically, the time profile ofthe scintillation light created by the nuclear recoil signalis dramatically different than that for electron-like back-grounds. Radiation interacting with a noble liquid leadsto the formation of dimers, in the form of trapped exci-ton states [54]. Both singlet and triplet states are formedand create ultraviolet scintillation light at 128 nm whenthey decay. The lifetimes of these states are very differ-4
FIG. 17. Conceptual sketch of a ton-scale low energy thresh-old liquid argon detector. The active volume of the innerliquid argon detector is of a ton-scale and viewed by ∼ π coverage. The inner de-tector is enclosed in a vacuum insulation chamber. The outerwater tank is designed for muon veto and neutron shielding. ent in LAr; 6 ns for the singlet and 1.6 µ s for the triplet.Moreover, the relative amplitudes of these states dependon the type of ionizing radiation [55–57]. Boulay andHime [58] recognized that this Pulse-Shape Discrimina-tion (PSD) allows for unprecedented rejection of Arbeta decay background intrinsic to the argon target, aconcept that has since been demonstrated in small pro-totype detectors [59, 60] and has led to major efforts forthe direct detection of dark matter.Of particular utility to a CENNS measurement is theso-called “single-phase” approach to dark matter whereinonly the primary scintillation light is recorded [58]. Thisapproach allows one to design a detector with the highphoto-coverage necessary to achieve the desired lightyield and low-energy threshold. The PMTs are the onlyactive component in the detector, affording simplicity indesign. Moreover, the speed for recording digital pulsesis governed by the triplet lifetime of the argon scintilla-tion light, thus avoiding difficulties with pulse pileup anddead time associated with a time projection chamber.The basic conceptual design of a single-phase detectoris shown in FIG. 17 which is similar to the CLEAR de-tector concept [5]. Key to measuring CENNS is a detec-tor with a sufficiently large target mass and low-energythreshold to reveal a clean nuclear recoil signal that is freeof background. The detector requirements for a CENNSmeasurement are similar to those for dark matter detec-tion, however, with key differences: dark matter detec-
FIG. 18. Leakage probability of the internal Ar back-ground as a function of energy threshold. The dotted red lineindicates the statistical leakage rate of Ar events into thesignal region for 1 ton-year detector livetime. The solid redline indicates the leakage rate tolerable after duty factor cor-rection (5 × − ). Solid blue curves show the impact of PSDcuts in the leakage probability for two different light yieldassumptions. tors need to be operated deep underground and free ofcosmic ray induced background while a CENNS detectorwould be placed on the surface in a neutrino beam withits associated beam-related backgrounds. A great advan-tage of exploiting the BNB at Fermilab comes from itsshort-pulse time structure which provide a 5 × − re-duction factor against steady state backgrounds.By far the largest activity in the detector arises from Ar in the LAr target. Ar is a beta emitter ( Ar → K+ e − + ¯ ν e , Q = 535 keV, τ / = 269 year). In nat-ural argon it is present at ∼ , yielding thedecay rate of ∼ Ar background is a verystrong function of the light yield, which in turn dictateswhat can be achieved as an analysis energy threshold.The pulsed structure of the BNB provides significant re-duction in this background. If, for example, we assume 6p.e./keV ee for light yield, then one can expect to achievean energy threshold of 10 keV ee (40 keV nr ) with leakageof only one Ar event in an exposure of 1 ton-year. Theassumption of 6 p.e./keV ee is based on that measured inmicroCLEAN and projected for MiniCLEAN using theHamamatsu R5912-02MOD PMTs submerged and oper-ating cold in LAr [59, 61].Significant improvements are foreseen with PMT tech-nologies that increase the efficiency of 19% for the R5912-02MOD to ∼ Energy Threshold Signal Background(keV ee /keV nr ) 6 p.e./keV ee ee
10 p.e./keV ee
12 p.e./keV ee Ar background (events/year) for a 1 ton detector assuming 50% acceptance in rejecting electronand gamma background. The background rate is determined for the energy window between energy threshold and 100 keV nr (25 keV ee ). with light yield as high as 12 p.e./keV ee . As can be seenin Table I, this would yield a detector with an energythreshold as low as 6 keV ee (24 keV nr ) that is essentiallyfree of steady state and detector-related background.In addition to Ar in the sensitive volume, there areexternal backgrounds arising from the detector construc-tion materials themselves. Table II contains a projec-tion of the non- Ar backgrounds after scaling the Mini-CLEAN backgrounds to a 1-tonne detector target andappropriate surface area [61]. Unlike a dark matter de-tector, the CENNS detector can employ the full targetmass without fiducialization since the duty factor of theBNB is such as to make the steady backgrounds fromneutron backgrounds and surface radon progeny negli-gible. Therefore, CENNS experiment does not requirethis extreme level of radon background control. Hence,we assume 100 /m /day or lower of modest level radondaughter decay rate in the energy region of interest whichis reasonably achievable [63].FIG. 19 shows the event rate of CENNS in a one tonliquid argon neutrino detector given a neutrino flux of5 × ν /cm /s when the detector is located 20 m awayfrom the target at a far-off-axis site. Assuming flat ∼
50% detection efficiency, which is mostly from the PSDcut efficiency [59, 60], we expect about ∼
250 CENNSevents/ton/year at 25 keV nr energy threshold after back-ground subtraction (at 32 kW beam power). The beam-induced neutron backgrounds and systematic uncertain-ties are discussed in the following sections. B. Neutron Shielding
The measured beam-induced neutrons (see FIG. 14)can be significantly reduced with proper shieldings. Thefast neutron component, above 100 MeV, requires specialattention in shielding design. These neutrons may slowdown in the shielding material itself and then become amore difficult background component with slower neu-trons of less than a few MeV energy. We carried outMCNP and Geant-4 based Monte Carlo simulations inorder to evaluate the overall level of neutron shieldingthat is needed for a CENNS experiment. We used themeasured beam-induced neutron fluxes as input to thesimulation. We found these neutron fluxes can be sub-stantially suppressed by more than 7 orders of magnitude [keV]
Recoil E [Events/keV/year/ton] -2 -1
10 110 n Total m n Prompt m n + e n Delayed Ar gammasradiogenicneutrons (cosmic)neutrons (beam) n Total m n Prompt m n + e n Delayed Ar gammasradiogenicneutrons (cosmic)neutrons (beam) n Total m n Prompt m n + e n Delayed Ar gammasradiogenicneutrons (cosmic)neutrons (beam) FIG. 19. Number of expected CENNS events with far-off-axisBNB (32 kW) neutrino flux. The liquid argon detector is as-sumed to be located at 20 m away from the target. The beam-induced (cosmogenic) neutron background estimated basedon SciBath measurements and assuming 7 m (4 m) of concreteshielding but without water shielding (see FIG. 20). Flat 50%detection efficiencies are applied for nuclear recoil events. after 7 m of concrete shielding. FIG. 20 shows results ofthe MC from a Geant-4 based simulation. MCNP resultsare consistent with the Geant-4 results. We also foundthat measured cosmogenic neutrons can be significantlysuppressed with 4 m of concrete shielding. Given theselevels of concrete shielding, the total number of neutronsthat enter the detector’s water shielding within the de-tector livetime can be less than 20 neutrons/m per yearof operation time.The neutrons entering the water shielding (10 m indiameter) are then passed to the liquid argon detectorin Geant-4 MC. In order to boost statistics of neutronswe simulated one million neutrons, then scaled to theexpected input neutron fluxes. The resulting neutron-nucleus event rate in the liquid argon detector with watershielding is negligible (less than 10 − events/ton/year).Therefore in FIG. 19, we show MC results of neutron-nuclear recoil events without water shielding but with6 Source Production Rate Detection Rate E <
25 keV ee < E <
25 keV ee (/ton/year) (events/ton/year)PMT( α ,n) 66,700 11,340 1,520 710Steel( α ,n) 3,680 495 65 30Total( α ,n) 70,380 11,835 1,585 740Total( α ,n) × duty factor 3.5 0.6 0.08 0.04Radon 15,880 7,147 (25 < E <
100 keV nr )Radon × duty factor 0.8 0.36TABLE II. Backgrounds in the 1 ton CENNS detector arising from ( α ,n) neutrons from the PMTs and steel. The radonbackground is from TPB and acrylic. En [MeV]0 20 40 60 80 100 120 140 160 180 200 /20MeV/day neutrons/m -5 -4 -3 -2 -1
10 110
10 Beam-induced Neutrons (SciBath)Beam-induced Neutrons (7m Concrete)Cosmogenic Neutrons (SciBath)Cosmogenic Neutrons (4m Concrete)
FIG. 20. Neutron flux reduction with concrete shielding. Thethick black solid line is beam-induced neutrons and thick redsolid line is cosmogenic neutrons measured by SciBath detec-tor. The detector livetime corrections are made for both inputneutron fluxes. The thin black line is beam-induced neutronflux after passing 7 m of concrete shielding. The dotted redline is cosmogenic neutron flux after passing 4 m of concreteshielding. nr to 100 keV nr ). This low-background configuration sug-gests that the CENNS detector can be placed as closeas 14 m away from the target where we expect twice theneutrino flux than at the 20 m location. However it isalso true that predicting the neutron flux over a mas-sive shielding without accurate understanding of shield-ing configurations is quite challenging. Therefore, beamtests of various neutron shielding configuration would beneeded. One important check is to see if the neutrons are from “sky shine”, directly from the target or fromthe beamline. C. Systematics and Discovery Potential
There are two major sources of systematic uncertain-ties in a CENNS experiment: (1) Uncertainties of stop-ping pion production at the BNB target, and hence therelated systematic uncertainties of absolute flux of neu-trinos at the far-off-axis. (2) Uncertainties of scintillationyield (
Leff ) in liquid argon detector for the measurementof low-energy nuclear recoil events. The other sources ofsystematics such as beam-induced neutron backgrounds,cosmogenic neutrons, gamma backgrounds, ambient ra-dioactive decays and uncertainties from high energy neu-trino interactions near or in the detector, depend onthe specific experimental design or are minor backgroundcontributions.
1. Uncertainty of neutrino flux
The uncertainty in neutrino production from stoppedpions and muons is dominated by the uncertainty of thepion production in the BNB target and surrounding ma-terials. The HARP experiment at CERN measured pionproduction from both thin beryllium targets and a replicaBNB target at the 8 GeV proton energy that the BNBuses. The uncertainty of the pion production measuredby HARP was 7% [64, 65]. In addition to the uncer-tainty in direct pion production there are uncertaintiesthat arise from the secondary production of pions anduncertainties in the fraction of pions and muons that getto decay rather than interact. These additional uncer-tainties are estimated to be at the 5% level [40]. Thisgives a total of 9% neutrino flux uncertainty.
2. Uncertainties from Leff of liquid argon detector
The scintillation efficiency for nuclear recoils relative tothe electron-equivalent efficiency, referred to as
Leff , has7 [keVnr]
Recoil
E0 50 100 150 200 250
Scintillation Yield
LindhardMei et al.McKinseyRegenfusWArP
FIG. 21. The scintillation efficiency for nuclear recoils rela-tive to the electron-equivalent measured in microCLEAN [59],Regenfus et al. [66] and the single, averaged value fromWArP [67]. The model of Mei et al. [68] combines the Lind-hard theory with Birks saturation providing the phenomeno-logical description indicated.
Energy Threshold [keVnr]0 10 20 30 40 50 60 -Ar) [%] n ( s Uncertainty in
FIG. 22. Extracted cross section uncertainty as a functionof energy threshold due to the intrinsic uncertainty in
Leff of6.5%. The dashed curve indicates the uncertainty calculatedusing an analytical approximation to the shape of the differen-tial neutrino-nucleus scattering spectrum and the solid curveuses the true spectrum as simulated for the BNB. The un-certainty in
Leff effectively induces an uncertainty in knowl-edge of the energy threshold and thus the integrated eventrate above or below that threshold. The bump structure near ∼
47 keV nr comes from the similar structure in the event rateat the same energy (see FIG. 19 black-curve) as the Leff ischanging monotonically in these energies. been measured for LAr in microCLEAN [59] and inde-pendently by Regenfus et al. [66] as 0.25 ± ± ± ∼ nr . The combined measurements provide Leff =0.262 ± Leff at the lowestenergies measured is interesting and worth further explo-ration. It is very likely due to an intrinsic energy depen-dence in the scintillation yield for gamma rays. Measure-ments are typically made of the nuclear recoil scintillationyield relative to a calibrated energy scale for gammas andit is usually assumed that the scintillation yield for gam-mas is independent of energy. FIG. 22 shows expectedcross section uncertainty as a function of energy thresh-old due to the
Leff . At the energy threshold of 25 keV nr the measurement uncertainty of cross section by Leff is7.5%.
3. Uncertainties from high energy neutrino interactions
The high energy neutrinos ( >
55 MeV, see FIG. 6-(b))are produced by muon-capture and kaon decay at rest.These neutrinos represent only 3% of the total neutrinofluxes. However these high energy neutrinos may pro-duce two types of background events; (1) direct neutrinointeractions in the liquid argon volume, and (2) neutrinointeractions in the water shield which result in secondaryneutrons reaching the sensitive detector volume and leavenuclear recoils in the signal region.We carried out a detector simulation for high en-ergy neutrino interactions using FLUKA [69–71]. Theneutrino interactions were weighted by neutrino-nucleuscross sections obtained with the GENIE(2.8.0) [72] neu-trino simulation package. Table III shows the above twobackground cases.
Liquid argon Water shieldAll events w/neutrons All events w/neutrons ν e ν µ ν µ Sum
TABLE III. Expected background events by E ν >
55 MeVwhich deposit energy of 25 keV to 100 keV per ton liquid argondetector per year. The numbers of events with secondariesproduced in (or reaching) the sensitive volume are presentedin the ‘all events’ columns. More critical events containingone or more neutrons are given by the ‘w/neutrons’ columns.
The CENNS signal is identified by single nuclear re-coils in the energy range 25 keV to 100 keV, and themost serious background is expected from nuclear recoilscaused by undetected neutron scattering. An upper limitof 0.42 events (=0.30+0.12 events or 0.21 events afterapplying 50% detection efficiency) per ton-year is foundfor the neutrino-induced background. As the number ofexpected background events is small, statistical uncer-tainties in the simulation are not expected to be rele-vant. The largest systematic uncertainty of this study8
UncertaintyNeutrino flux 9%
Leff of LAr 7.5%High energy neutrinos < < < Ar and gammas < < Total uncertainty 12%
TABLE IV. Systematic uncertainties of the event rate ofCENNS experiment. The detector energy threshold is as-sumed to be E th ≥
25 keV nr . arises from the neutrino-argon cross-section uncertain-ties in the GENIE model in the relevant neutrino energyrange (55 MeV to 250 MeV), which has never been mea-sured. However, even if we assume an order of magnitudeof uncertainty in the GENIE cross section model in thisenergy region, the backgrounds by the high energy neu-trinos are expected to be about 1% of the total numberof CENNS signal events.
4. Uncertainties from beam-induced neutrons
The neutron flux measurement by SciBath and re-sults from a neutron shielding MC study indicate thatthe beam-induced neutrons can be substantially reducedwith proper shielding design and could have a negligi-ble impact on the CENNS event rates. However, due tothe potential unknowns of these fast neutron shielding ef-fects, and our current uncertainty in neutron sources anddirections we assign a systematic uncertainty of beam-induced neutrons on the CENNS event rate at the 1%level.
5. Uncertainties from cosmogenic neutrons, gammas,radons and Ar The non-beam-related backgrounds can be signifi-cantly suppressed by the duty factor. Therefore the back-ground requirement of the CENNS experiment is far lessstringent than that of typical dark matter or other lowbackground experiments. Cosmic-ray backgrounds canbe further reduced by an active veto system in the watershielding, or it can also be significantly suppressed by 4 mof passive concrete shielding (see FIG. 20). The expectedsystematic uncertainty of the cosmogenic neutrons eventsin the signal rate is less than 1%. The gamma back-grounds are produced mostly by the decay chain of U, Th, and K in the PMT glass windows. These gammabackgrounds can also be suppressed by the duty factor,PSD and fiducial volume cuts. As shown in the FIG. 19 [keV] threshold E
20 25 30 35 40 45 50 [Events/year/ton] n Total n Delayed n Prompt [keV] threshold E
20 25 30 35 40 45 50 ] s Discovery potential[
FIG. 23. CENNS discovery potential. The integrated sig-nal event rates per ton detector after one-year operation as afunction of detector energy threshold (top plot) and the dis-covery potential in σ (bottom plot). A flat detection efficiencyof 50% over the energy range is assumed. The error bands onthe top plot are 1 sigma quadratic-sum errors of statisticaland systematic errors. the contribution of the gamma backgrounds in the sig-nal region is negligible. The backgrounds from the radondaughters, especially Po can produce nuclear recoils inthe signal region. The radon daughter backgrounds in thesignal region is expected to be negligible after the pulsetiming cut. Moreover, the steady-state backgrounds canbe separately measured by the beam-off data in the en-ergy region of interest and can be subtracted from thesignal shape. Therefore, the systematic uncertainty dueto radiogactive backgrounds is conservatively assumed tobe less than 1%.Table IV summarizes the systematic uncertainties.The total systematic uncertainty in event rate is expectedto be 12%. FIG. 23 shows the discovery potential ofthe CENNS interaction as a function of detector energythreshold with 1 ton-year exposure at 20 m from the BNBtarget. A 7.5 sigma discovery of the CENNS is expectedat the detector energy threshold of 25 keV nr . VI. SUMMARY
We presented a new experimental method for mea-suring the Coherent Elastic Neutrino Nucleus Scattering(CENNS), utilizing low energy neutrinos emitted at thefar-off-axis of a high energy neutrino beam. To determinethe feasibility of this approach, we have made neutron9background measurements at the Fermilab Booster Neu-trino Beam (BNB). Our results indicate that this methodcan result in a successful experiment. With the BNBneutrino source, non-beam related backgrounds such ascosmic rays, internal and external radioactivity are sub-stantially suppressed by the beam duty factor. The mea-sured beam-induced neutron backgrounds can be safelyreduced with proper shielding. We show that a one-tonfiducial mass single-phase liquid argon detector can makea 7.5 sigma discovery of CENNS at the detector energythreshold of 25 keV nr . Further development of a low en-ergy neutrino source at Fermilab as part of programs likeProject-X [73] and nuSTORM [74] will provide excellentresources for the future low energy neutrino physics ex- periments. ACKNOWLEDGEMENTS
The authors are grateful to R. Davis, W. Jaskierny,W. Miner, K. Taheri and J. Volk for providing technicaland engineering help of neutron flux measurement at theFermilab Booster target building. We would also like toacknowledge H. O. Meyer for crucial work on the Sci-Bath concept and detector construction. This work issupported by the Department of Energy, Fermilab Uni-versity Research Association Visiting Scholar program,NSF-1068712, NSF-1306942 and NSF-1005233. [1] D. Z. Freedman, Phys. Rev. D9 , 1389 (1974).[2] J. Yoo, Short-Baseline Neutrino Workshop (2011) ,https://indico.fnal.gov/conferenceDisplay.py?confId=4157.[3] H. T. Wong, H.-B. Li, J. Li, Q. Yue, and Z.-Y. Zhou,J.Phys.Conf.Ser. , 266 (2006), hep-ex/0511001.[4] P. Barbeau, J. Collar, and O. Tench, JCAP , 009(2007), nucl-ex/0701012.[5] K. Scholberg et al. , (2009), 0910.1989.[6] CSI Collaboration, D. Akimov et al. , (2013), 1310.0125.[7] G. Jungman, M. Kamionkowski, and K. Griest, Phys.Rept. , 195 (1996), hep-ph/9506380.[8] S. Arrenberg, L. Baudis, K. Kong, K. T. Matchev, andJ. Yoo, Phys. Rev. D78 , 056002 (2008), 0805.4210.[9] J. Billard, L. Strigari, and E. Figueroa-Feliciano, (2013),1307.5458.[10] P. Cushman et al. , (2013), 1310.8327.[11] D. Z. Freedman, D. N. Schramm, and D. L. Tubbs,Ann.Rev.Nucl.Part.Sci. , 167 (1977).[12] J. A. Nikkel, R. Hasty, W. H. Lippincott, and D. N.McKinsey, Astropart. Phys. , 161 (2008), astro-ph/0612108.[13] C. J. Horowitz, K. Coakley, and D. McKinsey, Phys.Rev. D , 023005 (2003).[14] K. Scholberg, Ann.Rev.Nucl.Part.Sci. , 81 (2012),1205.6003.[15] S. Chakraborty, P. Bhattacharjee, and K. Kar, (2013),1309.4492.[16] LSND Collaboration, A. Aguilar-Arevalo et al. ,Phys.Rev. D64 , 112007 (2001), hep-ex/0104049.[17] MiniBooNE Collaboration, A. Aguilar-Arevalo et al. ,Phys.Rev.Lett. , 161801 (2013), 1207.4809.[18] G. Garvey et al. , Phys.Rev.
D72 , 092001 (2005), hep-ph/0501013.[19] A. Anderson et al. , Phys.Rev.
D86 , 013004 (2012),1201.3805.[20] P. Vogel and J. Engel, Phys. Rev.
D39 , 3378 (1989).[21] N. F. Bell, M. Gorchtein, M. J. Ramsey-Musolf, P. Vo-gel, and P. Wang, Phys.Lett.
B642 , 377 (2006), hep-ph/0606248.[22] A. Beda et al. , Phys.Part.Nucl.Lett. , 406 (2010),0906.1926.[23] LSND Collaboration, L. B. Auerbach et al. , Phys. Rev.D , 112001 (2001).[24] G. G. Raffelt, Phys. Rev. Lett. , 2856 (1990). [25] A. Ayala, J. C. D’Olivo, and M. Torres, Phys. Rev. D , 111901 (1999).[26] J. Barranco, O. G. Miranda, and T. I. Rashba, JHEP , 021 (2005), hep-ph/0508299.[27] S. Davidson, C. Pena-Garay, N. Rius, and A. Santamaria,JHEP , 011 (2003), hep-ph/0302093.[28] K. Scholberg, Phys. Rev. D73 , 033005 (2006), hep-ex/0511042.[29] J. Barranco, O. G. Miranda, and T. I. Rashba, Phys.Rev.
D76 , 073008 (2007), hep-ph/0702175.[30] CHARM Collaboration, J. Dorenbosch et al. , Phys.Lett.
B180 , 303 (1986).[31] A. Steiner, M. Prakash, J. Lattimer, and P. Ellis, PhysicsReports , 325 (2005).[32] K. Patton, J. Engel, G. C. McLaughlin, and N. Schunck,Phys.Rev.
C86 , 024612 (2012), 1207.0693.[33] P. Amanik and G. McLaughlin, J.Phys.G
G36 , 015105(2009).[34] C. Horowitz and J. Piekarewicz, Phys.Rev.Lett. , 5647(2001), astro-ph/0010227.[35] OscSNS Collaboration, M. Elnimr et al. , (2013),1307.7097.[36] Y. Efremenko and W. Hix, J.Phys.Conf.Ser. , 012006(2009), 0807.2801.[37] A. Bolozdynya et al. , (2012), 1211.5199.[38] Figure is taken from a MiniBooNE site plan, DrawingNo. 6-7-52 CDR-2 (1999).[39] R. Stefanski et al. , Fermilab internal (Project 6-7-55)(2000).[40] MiniBooNE Collaboration, A. A. Aguilar-Arevalo et al. ,Phys. Rev. D79 , 072002 (2009), 0806.1449.[41] I. Stancu, Fermilab internal (2001).[42] R. Tayloe et al. , Nucl.Instrum.Meth.
A562 , 198 (2006).[43] R. Cooper et al. , (2011), 1110.4432.[44] V. V. Verbinski et al. , Nucl. Instr. Meth. , 8 (1968).[45] J. M. Adams and G. White, Nucl. Instr. Meth. , 495(1978).[46] R. A. Cecil, B. D. Anderson, and R. Madey, Nucl. Instr.Meth. , 439 (1979).[47] FINeSSE Collaboration, L. Bugel et al. A623 , 1046 (2010).[51] D. Mei and A. Hime, Phys.Rev.
D73 , 053004 (2006),astro-ph/0512125.[52] S. Miyake, International Cosmic Ray Conference , 3638(1973), Denver 1973.[53] M. Gordon et al. , IEEE.Trans.Nucl.Sci , 3427 (2004).[54] S. Kubota et al. , Phys. Rev. B , 1649 (1976).[55] A. Hitachi et al. , Phys. Rev. B , 5279 (1983).[56] T. Doke et al. , Nucl. Instrum. Meth. A269 , 291 (1988).[57] T. Doke et al. , Jpn. J. Appl. Phys. , 1538 (2002).[58] M. G. Boulay and A. Hime, Astropart. Phys. , 179(2006).[59] W. H. Lippincott et al. , Phys. Rev. C78 , 035801 (2008),0801.1531.[60] P. Benetti et al. , Astropart.Phys. , 495 (2008), astro-ph/0701286.[61] A. Hime, Proceedings of the DPF-2011 Conference, Prov-idence, RI. (2011). [62] T. Alexander et al. , Astroparticle Physics , 44 (2013).[63] M. Boulay et al. , (2009), 0904.2930.[64] HARP Collaboration, M. Catanesi et al. , Eur.Phys.J. C52 , 29 (2007), hep-ex/0702024.[65] D. W. Schmitz, PhD Thesis, Columbia University (2008).[66] C. Regenfus, IDM2006 Proceedings, World Scientific ,325 (2006).[67] WArP Collaboration, R. Brunetti et al. , New Astron.Rev. , 265 (2005), astro-ph/0405342.[68] D.-M. Mei, Z.-B. Yin, L. Stonehill, and A. Hime, As-troparticle Physics , 12 (2008).[69] G. Battistoni et al. , AIP Conference Proceeding , 31(2007).[70] A. Fass`o, A. Ferrari, J. Ranft, and P. Sala, (2005).[71] G. Battistoni, A. Ferrari, M. Lantz, P. R. Sala, and G. I.Smirnov, (2009).[72] C. Andreopoulos et al. , Nucl. Instrum. Meth. A614 , 87(2010), 0905.2517.[73] A. S. Kronfeld et al. , (2013), 1306.5009.[74] nuSTORM Collaboration, D. Adey et al.et al.