Letter of Intent to Construct a nuPRISM Detector in the J ... · Letter of Intent to Construct a nuPRISM Detector in the J-PARC Neutrino Beamline S.Bhadra,24 A.Blondel,3 S.Bordoni,5

Letter of Intent to Construct a nuPRISM Detector in the J-PARC Neutrino Beamline

S. Bhadra,24 A. Blondel,3 S. Bordoni,5 A. Bravar,3 C. Bronner,9 J. Caravaca Rodrıguez,5 M. Dziewiecki,23

T. Feusels,1 G.A. Fiorentini Aguirre,24 M. Friend,4, ∗ L. Haegel,3 M. Hartz,8, 22 R. Henderson,22 T. Ishida,4, ∗

M. Ishitsuka,20 C.K. Jung,11, † A.C. Kaboth,6 H. Kakuno,25 H. Kamano,13 A. Konaka,22 Y. Kudenko,7, ‡

M. Kuze,20 T. Lindner,22 K. Mahn,10 J.F. Martin,21 J. Marzec,23 K.S. McFarland,15 S. Nakayama,18, †

T. Nakaya,9, 8 S. Nakamura,12 Y. Nishimura,19 A. Rychter,23 F. Sanchez,5 T. Sato,12 M. Scott,22 T. Sekiguchi,4, ∗

M. Shiozawa,18, 8 T. Sumiyoshi,25 R. Tacik,14, 22 H.K. Tanaka,18, † H.A. Tanaka,1, § S. Tobayama,1 M. Vagins,8, 2

J. Vo,5 D. Wark,16 M.O. Wascko,6 M.J. Wilking,11 S. Yen,22 M. Yokoyama,17, † and M. Ziembicki23

(The nuPRISM Collaboration)1University of British Columbia, Department of Physics and Astronomy, Vancouver, British Columbia, Canada

2University of California, Irvine, Department of Physics and Astronomy, Irvine, California, U.S.A.3University of Geneva, Section de Physique, DPNC, Geneva, Switzerland

4High Energy Accelerator Research Organization (KEK), Tsukuba, Ibaraki, Japan5Institut de Fisica d’Altes Energies (IFAE), Bellaterra (Barcelona), Spain

6Imperial College London, Department of Physics, London, United Kingdom7Institute for Nuclear Research of the Russian Academy of Sciences, Moscow, Russia

8Kavli Institute for the Physics and Mathematics of the Universe (WPI), TodaiInstitutes for Advanced Study, University of Tokyo, Kashiwa, Chiba, Japan

9Kyoto University, Department of Physics, Kyoto, Japan10Michigan State University, Department of Physics and Astronomy, East Lansing, Michigan, U.S.A.

11State University of New York at Stony Brook, Department of Physics and Astronomy, Stony Brook, New York, U.S.A.12Osaka University, Department of Physics, Osaka, Toyonaka, Japan

13Osaka University, Research Center for Nuclear Physics(RCNP), Ibaraki, Osaka, Japan14University of Regina, Department of Physics, Regina, Saskatchewan, Canada

15University of Rochester, Department of Physics and Astronomy, Rochester, New York, U.S.A.16STFC, Rutherford Appleton Laboratory, Harwell Oxford, and Daresbury Laboratory, Warrington, United Kingdom

17University of Tokyo, Department of Physics, Tokyo, Japan18University of Tokyo, Institute for Cosmic Ray Research, Kamioka Observatory, Kamioka, Japan

19University of Tokyo, Institute for Cosmic Ray Research, Research Center for Cosmic Neutrinos, Kashiwa, Japan20Tokyo Institute of Technology, Department of Physics, Tokyo, Japan

21University of Toronto, Department of Physics, Toronto, Ontario, Canada22TRIUMF, Vancouver, British Columbia, Canada

23Warsaw University of Technology, Institute of Radioelectronics, Warsaw, Poland24York University, Department of Physics and Astronomy, Toronto, Ontario, Canada

25Tokyo Metropolitan University, Department of Physics, Tokyo, Japan(Dated: December 16, 2014)

As long-baseline neutrino experiments enter the precision era, the difficulties associated with understanding neutrino inter-action cross sections on atomic nuclei are expected to limit experimental sensitivities to neutrino oscillation parameters. Inparticular, the ability to relate experimental observables to the incident neutrino energy in all previous experiments has reliedsolely on theoretical models of neutrino-nucleus interactions, which currently suffer from very large theoretical uncertainties.

By observing charged current νµ interactions over a continuous range of off-axis angles from 1◦ to 4◦, the nuPRISM waterCherenkov detector can provide a direct measurement of the far detector lepton kinematics for any given set of oscillationparameters, which largely removes neutrino interaction modeling uncertainties from T2K oscillation measurements. Thisnaturally provides a direct constraint on the relationship between lepton kinematics and neutrino energy. In addition, nuPRISMis a sensitive probe of sterile neutrino oscillations with multiple energy spectra, which provides unique constraints on possiblebackground-related explanations of the MiniBooNE anomaly. Finally, high-precision measurements of neutrino cross sectionson water are possible, including electron neutrino measurements and the first ever measurements of neutral current interactionsas a function of neutrino energy.

The nuPRISM detector also provides significant benefits to the proposed Hyper-Kamiokande project. A demonstration

that neutrino interaction uncertainties can be controlled will be important to understanding the physics reach of Hyper-K. In

addition, nuPRISM will provide an easily accessible prototype detector for many of the new hardware components currently

under consideration for Hyper-K. The following document presents the configuration, physics impact, and preliminary cost

estimates for a nuPRISM detector in the J-PARC neutrino beamline.

∗ also at J-PARC, Tokai, Japan† affiliated member at Kavli IPMU (WPI), the University of

Tokyo, Japan

‡ also at Moscow Institute of Physics and Technology and NationalResearch Nuclear University ”MEPhI”, Moscow, Russia§ also at Institute of Particle Physics, Canada

arX

iv:1

412.

3086

v2 [

phys

ics.

ins-

det]

13

Dec

201

4

2

CONTENTS

I. Introduction 3A. Uncertainties in Neutrino Energy

Determination 4B. ND280 Capabilities and Limitations 4C. Detector Overview 6

II. Physics Capabilities 6A. Off-Axis Fluxes 6B. Monochromatic Beams 6C. Simulation Inputs 7D. Event Pileup 9E. Event Selection for Sensitivity

Studies 12F. T2K νµ Disappearance Sensitivities 12G. nuPRISM 1-Ring e-like Ring

Measurements 191. Beam νe and νe cross section

study 192. Predicting oscillated νe for the

appearance measurement 203. Backgrounds from νµ’s 204. Sterile Neutrino Sensitivity 21

H. νµ Measurements 21I. Cross Section Measurements 22

1. CC Inclusive 242. CC0π 243. CC1π+ and CC1π0 254. NC1π+ and NC1π0 25

III. Detector Design and Hardware 26A. Site Selection 26B. Civil Construction 26C. Liner and Tank 27D. Detector Frame and Lifting

Mechanism 271. Detector Shape, Support and

Positioning 272. Water Flow and Optical

Isolations 283. Walls of Inner Detector (ID) 294. Detector in the shaft 295. Detector Surveying 30

E. Scintillator panels 31

1. Scintillator counters withWLS/avalanche photodiodereadout 31

2. Veto counters for nuPRISM 32F. Photomultiplier Tubes 32G. Electronics 32

1. FADC Digitization 322. Signal Conditioning And PMT

HV 333. Digitization

Performance/Optimization 33H. Water System 33

1. Gd option 35

IV. Detector Calibration 37A. Overview of Super-K Calibration

Systems 371. Detector hardware calibrations 372. Calibrations for physics analyses 38

V. Conclusion 39

References 39

3

I. INTRODUCTION

With the publications of the first ever observation of νeappearance, and the world’s most precise measurementof θ23, T2K has achieved its initial experimental goalswith only 8.5% of the approved protons on target (POT).The next phase of the experiment will make even moreprecise measurements of νe appearance and νµ disappear-ance using both neutrinos and anti-neutrinos in order toprobe the value of δCP , the θ23 octant, and

∣∣∆m232

∣∣. Inconjunction with measurements from NOνA, these mea-surements may also provide a constraint on the neutrinomass hierarchy.

In order to achieve these goals, a more precise under-standing of neutrino interaction cross sections is required.Currently, T2K is forced to rely on neutrino interac-tion generators to translate experimental observables intoconstraints on the neutrino energy spectrum, which de-pends on the value of the oscillation parameters. Mea-surements of very forward-going muons on the carbontarget employed by the existing near detector, ND280,are translated into constraints on the 4π muon angulardistribution on a water target seen at the far detector.The interactions of final-state hadrons both within thenucleus and within each detector medium can have a sig-nificantly different impacts on the near and far detec-tor analyses. Some of the backgrounds at far detector,Super-Kamiokande (Super-K), are poorly constrained atND280. This is particularly true of NCπ+ events, whichare problematic both because the cross section is not wellmeasured, and because π+ reconstruction at Super-K isnot well understood. This results in a contamination ofboth the νµ and νe samples that produces large system-atic uncertainties.

It is also necessary to measure events with single,electron-like rings in order to constrain any differencesin the νe and νµ cross sections. These events can becaused by a variety of sources, such as beam νe, single γproduction, π0 production, external γ background, sterileneutrino oscillations and radiative muon production. Anexcess of such events has been observed by MiniBooNE.It is important to confirm whether a similar excess existson a water target, ideally with a water Cherenkov detec-tor, and if found, the cause must be understood in orderto perform precision νe appearance measurements.

The least constrained component of these neutrino in-teraction models, however, is the relationship betweenthe experimentally observable lepton kinematics and theenergy of the incident neutrino. At present, there is anexperimentally-unconstrained and potentially large biasin the ability to translate lepton kinematics to neutrinoenergy. Current estimates, based solely on new, recentlydeveloped models, suggest that this bias may be one ofT2K’s largest systematic uncertainties, and no existingdataset can provide a constraint on this uncertainty in amanner that does not rely on neutrino interaction mod-els. Had neutrino interaction models been trusted toprovide this relationship just 5 years ago, current mod-

els suggest that 20 to 30% of events where only the finalstate lepton was observed would have been reconstructedwith an incorrect neutrino energy in a way that wouldnot have been constrained or even detectable. Evena high-performance near detector, capable of preciselymeasuring all charged particles in the final state, wouldbe forced to rely on models that relate lepton kinemat-ics to hadronic final states, and no modern theoreticalmodels offer a prediction for such a relationship within anuclear environment.

The nuPRISM water-Cherenkov detector takes advan-tage of the energy dependence of the neutrino flux withoff-axis angle by spanning a continuous range of 1 to 4degrees in off-axis angle. This technique has the poten-tial to significantly reduce uncertainties from neutrinointeraction modeling in T2K oscillation analyses, as isdemonstrated for the muon neutrino disappearance mea-surement described in Section II. In particular, thesemeasurements will provide the first direct experimentalconstraint on the relationship between lepton kinematicsand neutrino energy using measurements of final statemuons at many different off-axis angles. In order to con-struct a more cost-effective detector that can reasonablybe built on a timescale that is applicable to T2K, thisdocument proposes to instrument a subset of the full wa-ter volume on a frame that moves vertically within thewater tank, which sequentially samples the full off-axisrange of the shaft in 5-6 separate running periods.

The construction of a nuPRISM detector in the next3-5 years can also provide significant benefits to Hyper-Kamiokande (Hyper-K). The problems with understand-ing neutrino interactions can have a larger impact onHyper-K, since Hyper-K will have much smaller statis-tical errors, and a demonstration that these uncertain-ties can be managed with a nuPRISM near detectorwill significantly enhance the physics case for Hyper-K. In addition, nuPRISM is an easily accessible waterCherenkov detector that provides an ideal environmentto test Hyper-K technology. Hyper-K proposes to usenew, in-water electronics, new solid state hybrid-PMTs(HPDs), and a new tank and liner construction to preventleaks, all of which require extensive testing in a prototypedetector. Finally, nuPRISM will provide an intermediatephysics program that bridges the gap from T2K phaseI to Hyper-K, which can provide continuity within theJapanese physics community while Hyper-K is being de-signed and constructed.

The remainder of this document provides an overviewof the detector components and physics potential ofnuPRISM. The results for a full T2K νµ disappear-ance analysis are provided, and a variety of additionalnuPRISM neutrino energy spectrum fits are presentedto demonstrate how the nuPRISM technique can con-strain νe cross sections, which will be important for mea-surements of νe appearance and δCP , as well as severaldifferent oscillation backgrounds. Cost estimates havebeen obtained for the items that are expected to dom-inate the cost of the project, in particular photomulti-

4

plier tubes (PMTs) and civil construction. For the ad-ditional less expensive items, cost estimates from a verysimilar project proposed in 2005, the T2K 2 km waterCherenkov detector, are used to guide expectations forthe full nuPRISM project cost.

A. Uncertainties in Neutrino EnergyDetermination

Prior to 2009, neutrino interaction models assumedthat neutrinos, when encountering a nuclear target, in-teract a single nucleon. The initial state of the nucleonwas characterized by a binding energy and Fermi momen-tum, which were drawn from either a Fermi gas [2, 3] or amore specialized spectral function treatment [4]. In thisparadigm, all the remaining dynamics of charge-currentquasi-elastic (CCQE) interactions, in which the targetneutron is converted into an outgoing proton, are encap-sulated in a set of three vector and three axial-vectorform factors. Most of these form factors are tightly con-strained from external electron and pion scattering ex-periments (for a detailed discussion, see Ref. [10]). Thelargest remaining uncertainty is on the axial vector formfactor, which is assumed to take a dipole form,

FA(Q2) =FA(0)

(1 + Q2

M2A

)2. (1)

The parameter FA(0) is precisely known from nuclearbeta decay, which leaves MA as the remaining uncertainparameter. Modifying MA simultaneously alters boththe overall CCQE cross section and the shape of the Q2

distribution.In 2009, the first comparison of MiniBooNE CCQE-

like data at neutrino energies around 1 GeV and NO-MAD data at higher energies was released. A reproduc-tion of that comparison is shown in Figure I A. The Mini-BooNE data are consistent with an MA value of 1.35 GeV(an additional empirical parameter, κ is consistent withno modification at 1 σ), while the NOMAD data preferan MA of 1.03 GeV. This discrepancy is currently unex-plained by neutrino-nucleus interaction models and is anoutstanding experimental question that can be addressedby nuPRISM (see Section II I 2).

Later in 2009, the Marteau [5] formalism for the treat-ment of neutrino scattering on nucleon pairs in nuclei wasresurrected by Martini et al. [6–8] to explain the higherevent rate and muon kinematic distributions observed byMiniBooNE. If this explanation of the MiniBooNE eventexcess were correct, it would imply that neutrino energyreconstruction for all previous neutrino experiments onnuclei at the GeV scale could have significant biases for20-30% of CCQE-like events. In the past few years, themodels of Martini et al. and Nieves et al. [9] have be-gun to incorporate these effects, but such calculations arevery difficult and the predictions of just these two mod-els produce significantly different results when applied toT2K oscillation analyses [1].

There exists circumstantial experimental evidence formultinucleon interaction mechanisms in both neutrinoand electron scattering, but nothing that allows us toconclusively solve the problem or even to down-selectamong the various calculations. In electron scattering,the reaction mechanism is different due to the absence ofan axial-vector current component. In neutrino scatter-ing experiments with broadband beams, the evidence isonly circumstantial, since we must rely on the predictionsof the models themselves to extract the neutrino energyfor any given event. Other approaches, such as mak-ing precise measurements of the hadronic final state, arelimited by a lack of theoretical understanding of the ex-pected hadron kinematics for multinucleon events. Eventhe final state hadron spectra for CCQE events are modi-fied by nuclear effects which are also not well understood.

Figure 2 illustrates the challenge associated with us-ing near detector data to constrain the interaction modelthat predicts far detector event rates. The detectors mea-sure the convolution of the neutrino spectrum with theinteraction model. Since the near and far detector spec-tra are different due to neutrino oscillations, the mea-surement of this convolution in the near detector doesnot directly constrain the event rate in the far detector,and neutrino energies that represent a small fraction ofthe event rate in the near detector can be a significantlyimpact the measurement of oscillation parameters in thefar detector.

In addition to multinucleon effects, other effects suchas long range correlations and final state interactionswithin the target nucleus can also produce distortionsto the neutrino energy spectrum that can be difficult tomodel. In order to perform precision oscillation measure-ments with uncertainties at the level of the few percentstatistical errors expected for 7.8× 1021 POT, it will benecessary to provide a data-driven constraint on theseneutrino interaction model uncertainties.

B. ND280 Capabilities and Limitations

T2K oscillation analyses rely on precise constraints offlux and cross section model parameters from ND280.While a 3.2% uncertainty on the predicted number ofelectron neutrinos at the far detector has been achievedfor the combination of flux and cross section parame-ters that are currently constrained by the near detec-tor, there remains a 4.7% uncertainty on the far detectorevent sample due to additional cross section parametersthat remain unconstrained. This unconstrained uncer-tainty is dominated by uncertainties in the modeling ofthe target oxygen nucleus, and largely depends on thetheoretical model used to extrapolate measurements oncarbon to oxygen.

The full capabilities of the T2K near detector have notyet been exploited. For example, the near detector analy-ses have thus far used interactions in the most-upstreamFine-Grained Detector (FGD1), which is composed en-

5

FIG. 1. The CCQE cross section measurements are shown for MiniBooNE and NOMAD. The data show significant differencesbetween measurements made at low and high energies.

FIG. 2. A cartoon of the effect of energy reconstruction biasesis shown for both the T2K near detector (top) and the fardetector (bottom). At the far detector, these biases directlyimpact the measurement of the oscillation dip, but the biasesare largely unconstrained at the near detector due to the largeunoscillated sample of unbiased CCQE events.

tirely of alternating layers of horizontally- and vertically-oriented scintillator bars. Since the FGD scintillator lay-ers are predominantly composed of carbon and hydrogen,FGD1 measurements cannot directly probe interactionson oxygen. An additional FGD (FGD2) contains layersof water interspersed within its scintillator layers. A si-multaneous fit of the interactions in both FGDs can pro-

vide a constraint on nuclear uncertainties in oxygen, andmay potentially reduce the corresponding nuclear modeluncertainties.

Another expected improvement to ND280 is the exten-sion of the measured phase space of the outgoing leptonkinematics from a charged-current neutrino interaction.In the currently available analyses, muons are requiredto be produced in an FGD and traverse a minimum dis-tance through the downstream TPC in order to make ameasurement of both muon momentum and particle iden-tification. This limits the muon acceptance to forwardangles. Improvements to detector timing calibration andto track matching to the Electromagnetic Calorimeters(ECALs) and Side Muon Range Detectors (SMRDs) sur-rounding the FGDs and TPCs will allow for the recon-struction of charged-current events with backward-goingand sideways-going muons. These additional samples willadd less than 20% to the total even sample with a de-graded energy resolution relative to events that enter aTPC, however they may be able to improve constraintson the cross section modeling in previously inaccessibleand potentially important new regions of phase space.

An additional sample of events that has not yet beenincorporated into the oscillation analysis are charged-current interactions in the pi-zero detector (P0D). TheP0D is capable of operating with and without water tar-gets dispersed throughout its active volume, and by mea-suring the event rates separately in these two configura-tions, it is possible to extract constraints on interactionsin water. The requirement to match a track in the TPClimits the angular acceptance for muons produced in theP0D, however the larger fiducial volume of the P0D pro-duces a higher event sample.

In order for any of these new samples to reduce sys-tematic uncertainties, it is necessary to choose a neu-trino interaction model that can characterize all possiblevariations of the neutrino cross sections as a function ofboth neutrino energy and the final state particle kine-matics. In other words, model dependent choices willhave to be made that will directly impact the strength

6

of the constraint that can be extracted. Given the dif-ficulties in understanding neutrino-nucleus interactions,it may not be possible to justify reductions in cross sec-tion systematic uncertainties beyond their current levelwithout a direct experimental constraint. In addition,the aforementioned uncertainties due to multinucleon ef-fects have not yet been incorporated into T2K oscillationanalyses. Preliminary studies within T2K indicate thatthese effects would be difficult to constrain using onlylepton kinematics from ND280 at the level required forthe full-statistics T2K sensitivity, and may be as largeas the current dominant systematic uncertainties. Theuse of additional hadronic information is being explored,but any such constraint would be subject to even furthermodel dependence.

C. Detector Overview

The nuPRISM detector uses the same water Cherenkovdetection technology as Super-K with a cylindrical watervolume that is taller than Super-K (50-100 m vs 41 m)but with a much smaller diameter (10-12 m vs 39 m).The key requirements are that the detector span thenecessary off-axis range (1◦-4◦) and that the diameteris large enough to contain the maximum required muonmomentum. The baseline design considers a detector lo-cation that is 1 km downstream of the neutrino interac-tion target with a maximum contained muon momentumof 1 GeV/c. This corresponds to a 50 m tall tank witha 6 m diameter inner detector (ID) and a 10 m diame-ter outer detector (OD). A larger, 8 m ID is also beingconsidered at the expense of some OD volume at thedownstream end of the tank. As the nuPRISM analysisstudies mature, the exact detector dimensions will be re-fined to ensure sufficient muon momentum, νe statisticsand purity, etc.

The instrumented portion of the tank is a subset ofthe full height of the water volume, currently assumedto be 10 m for the ID and 14 m for the OD. The novelfeature of this detector is the ability to raise and lowerthe instrumented section of the tank in order to span thefull off-axis range in 6 steps. The inner detector will beinstrumented with either 5-inch or 8-inch PMTs to ensuresufficient measurement granularity for the shorter lightpropagation distances relative to Super-K. Also underconsideration is to replace the OD reflectors with largescintillator panels, such as those used in the T2K SideMuon Range Detector (SMRD), although this has notyet been integrated into the overall detector design. Moredetails regarding the detector hardware can be found inSection III

II. PHYSICS CAPABILITIES

The physics goals of nuPRISM include reducing sys-tematic uncertainties on the T2K oscillation analyses,

using electron-like events to search for sterile neutrino os-cillations and constrain electron neutrino cross sections,and making the first ever energy dependent neutral cur-rent (NC) and charged current (CC) cross section mea-surements that do not rely on neutrino generators to pro-vide the incident neutrino energy.

A. Off-Axis Fluxes

The nuPRISM detector concept exploits the fact thatas a neutrino detector is moved to larger off-axis anglesrelative to the beam direction, the peak energy of theneutrino energy spectrum is lowered and the size of thehigh-energy tail is reduced. This effect can be seen inFigure 3, which shows the neutrino energy spectra at sev-eral different off-axis angles in the T2K beam line. Sincethe off-axis angle for a single neutrino interaction can bedetermined from the reconstructed vertex position, thisextra dimension of incident neutrino energy dependencecan be used to constrain the interaction rates and finalstate particles in a largely model independent way.

A typical nuPRISM detector for the T2K beam linewould span a continuous range of off-axis angles from 1◦

to 4◦. For T2K, the best choice of technology is a wa-ter Cherenkov detector in order to use the same nucleartarget as Super-K, and to best reproduce the Super-Kdetector efficiencies.

B. Monochromatic Beams

The detector can be logically divided into slices ofoff-axis angle based on the reconstructed vertex of eachevent. In each slice, the muon momentum and angle rel-ative to the mean neutrino direction can be measured.By taking linear combinations of the measurements ineach slice, it is possible to produce an effective muon mo-mentum and angle distribution for a Gaussian-like beamat energies between 0.4 and 1.2 GeV. Qualitatively, anydesired peak energy can be chosen by selecting the ap-propriate off-axis angle, as shown in Figure 4, and thenthe further on-axis measurements are used to subtractthe high energy tail, while the further off-axis measure-ments are used to subtract the low energy tail. Figure 4shows three such “pseudo-monochromatic” neutrino en-ergy spectra constructed in this manner. These spectraare for selected 1-ring muon candidates and systematicerrors from the flux model are applied using the T2K fluxsystematic error model. The statistical errors for an ex-posure of 4.5× 1020 protons on target are also shown. Inall cases the high energy and low energy tails are mostlycanceled over the full energy range and the monochro-matic nature of the spectrum is stable under the fluxsystematic and statistical variations.

Figure 5 shows the reconstructed energy distributionsfor 1-ring muon candidates observed with the pseudo-monochromatic beams shown in Figure 4. The candidate

7

φ

10

20

30

40

1510×

Fluxµν Flux x 100eν

° = 1.0OAθ

(GeV)νE

0 1 2 3 4 5

eνmax

φ/ µνφ

-110

1

10

210

φ

20

40

1510×


° = 2.5OAθ

(GeV)νE

0 1 2 3 4 5

eνmax

φ/ µνφ

-110

1

10

210

φ

10

20

30

40

1510×


° = 4.0OAθ

(GeV)νE

0 1 2 3 4 5

eνmax

φ/ µνφ

-110

1

10

210

FIG. 3. The neutrino energy spectra for νµ and νe fluxes inthe T2K beam operating in neutrino mode are shown for off-axis angles of 1◦, 2.5◦, and 4◦. The νµ flux normalized by themaximum νe flux is shown at the bottom of each plot, demon-strating that feed-down from high energy NC backgrounds toνe candidates can be reduced by going further off-axis.

events are divided into quasi-elastic scatters and non-quasi-elastic scatters, which include contributions fromprocesses related to nuclear effects such as multinucleoninteractions or pion absorption in final state interactions.With these pseudo-monochromatic beams, one sees astrong separation between the quasi-elastic scatters andthe non-quasi-elastic scatters with significant energy re-construction bias, especially in the 0.8 to 1.2 GeV neu-trino energy range. These measurements can be used todirectly predict the effect of non-quasi-elastic scatters inoscillation measurements and can also provide a uniqueconstraint on nuclear models of these processes.

The nuPRISM technique can be expanded beyondthese pseudo-monochromatic beams. This linear com-bination method can be used to reproduce a wide varietyof flux shapes between 0.4 and 1.0 GeV. In particular,as described later in this section, it is possible to repro-duce all possible oscillated Super-K spectra with a linearcombination of nuPRISM measurements, which signifi-cantly reduces many of the uncertainties associated withneutrino/nucleus interaction modeling.

C. Simulation Inputs

To perform nuPRISM sensitivity analyses, the officialT2K flux production and associated flux uncertaintieshave been extended to cover a continuous range of off-axis angles, and the standard T2K package used to gen-erate vertices in ND280 has also been modified to handleflux vectors with varying energy spectra across the de-tector. However, for the analysis presented in this note,full detector simulation and reconstruction of events werenot available. Instead, selection efficiencies and recon-struction resolutions for vertex, direction, and visible en-ergy were tabulated using the results of fiTQun run onSuper-K events. The efficiency for electrons (muons) wasdefined as events passing the following cuts: OD veto, 1-ring, e-like (µ-like), 0 (1) decay electrons, and the T2KfiTQun π0 rejection (no π0 cut). The efficiency tabu-lation was performed in bins of the true neutrino en-ergy, the visible energy and distance along the track di-rection to the wall of the most energetic ring, and sepa-rate tables were produced for charged current events withvarious pion final states (CC0π, CC1π±0π0, CC0π±1π0,CCNπ±0π0, and CCother) for both νe and νµ events, aswell as a set of neutral current final states, also charac-terized by pion content (NC0π, NC1π±0π0, NC0π±1π0,NCNπ±0π0, and NCother). To determine the smearingof true quantities due to event reconstruction, vertex, di-rection, and visible energy resolution functions were alsoproduced for the 1-ring e-like and µ-like samples in binsof visible energy and distance along the track directionto the wall of the most energetic ring.

The neutrinos in nuPRISM are simulated with theT2K flux simulation tool called JNUBEAM. The versionof JNUBEAM used is consistent with what is currentlyused by T2K and it includes the modeling of hadronic

8

(GeV)νE0.5 1 1.5 2

Eve

nts/

50 M

eV

0

5000

10000 Event Spectrumµ1 Ring

Absolute Flux Error

Shape Flux Error

Statistical Error

Gaussian Fit

Fit Mean: 0.60 GeVFit RMS: 0.08 GeV

Linear Combination, 0.6 GeV Mean

(GeV)νE1 2 3

Eve

nts/

50 M

eV

0

5000


Absolute Flux Error

Shape Flux Error

Statistical Error

Gaussian Fit



(GeV)νE1 2 3 4

Eve

nts/

50 M

eV

0

500

1000

1500


Absolute Flux Error

Shape Flux Error

Statistical Error

Gaussian Fit



FIG. 4. Three “pseudo-monochromatic” spectra centered at0.6 (top), 0.9 (middle) and 1.2 (bottom) GeV. The aqua errorbars show the 1 σ uncertainty for flux systematic variations,while the black error bars show the flux systematic variationafter the overall normalization uncertainty is removed. Thetan error bars show the statistical uncertainty for samplescorresponding to 4.5× 1020 protons on target.

(GeV)recE0.5 1 1.5 2

Eve

nts/

50 M

eV

0

2000

4000


Absolute Flux Error

Shape Flux Error

Statistical Error

NEUT QE

NEUT Non-QE


(GeV)recE0 1 2 3

Eve

nts/

50 M

eV

0

2000

4000


Absolute Flux Error

Shape Flux Error

Statistical Error

NEUT QE

NEUT Non-QE


(GeV)recE1 2 3

Eve

nts/

50 M

eV

0

500

1000


Absolute Flux Error

Shape Flux Error

Statistical Error

NEUT QE

NEUT Non-QE


FIG. 5. The reconstructed energy distributions for 1-ring muon candidate events produced using “pseudo-monochromatic” spectra centered at 0.6 (top), 0.9 (middle)and 1.2 (bottom) GeV. The aqua error bars show the 1 σuncertainty for flux systematic variations, while the black er-ror bars show the flux systematic variation after the overallnormalization uncertainty is removed. The tan error barsshow the statistical uncertainty for samples correspondingto 4.5 × 1020 protons on target. The red and blue his-tograms show the contributions from non-quasi-elastic andquasi-elastic scatters respectively.

9

interactions based on data from the NA61/SHINE ex-periment. We define the off-axis angle for a particularneutrino as the angle between the beam axis and thevector from the average neutrino production point alongthe beam axis to the point at which the neutrino crossesthe flux plane, as illustrated in Fig. 6. The off-axis angleis defined in terms of the average neutrino productionpoint so that an off-axis angle observable can be con-structed based on the location of the interaction vertexin nuPRISM. The off-axis angle and energy dependencefor each neutrino flavor is shown in Fig. 7. The neutrinoflux files are produced for both neutrino mode (focussingpositively charged hadrons) and antineutrino mode (fo-cussing negatively charged hadrons), although only theneutrino mode flux is used for the analysis presented inthis note.

νPRISM Flux Planes

Beam direction

Average neutrino production point Point crossing

the flux plane

θOA

FIG. 6. The definition of the off-axis angle for individualneutrinos.

The positions of the neutrino interaction vertices in thenuPRISM water volume are shown in Fig. 8. The rateof simulated interactions has been cross checked againstthe observed INGRID ratesand found to be consistent.

D. Event Pileup

The baseline design of nuPRISM is an outer detector(OD) volume with radius of 5 m and height of 14 m,and an inner detector (ID) volume with a radius of 3m and height of 10 m, located 1 km from the T2K tar-get. We have carried out a simulation of events in thenuPRISM ID and OD volumes, as well as the surround-ing earth to study the event pile-up in nuPRISM. Thesimulation is carried out for the earth+nuPRISM geom-etry shown in Fig. 9. The flux at the upstream end of thevolume is simulated using the JNUBEAM package withhorn currents set to 320 kA. Interactions in the earthand detector volumes are generated using the same toolsfrom the NEUT package used for ND280 neutrino vectorgeneration. The earth volume is filled with SiO2 with adensity of 1.5 g/cm3. The water volume has three de-tector sub-volumes: the ID detector, the OD detector

and an intermediate volume. The vertical position of thedetector volumes in the water column can be adjustedto study the event pile-up at different off-axis angles. AGEANT4 simulation of the particles from the neutrinovectors is carried out and all particles with visible energygreater than 10 MeV are recorded if they originate inany of the detector volumes or cross any of the detectorvolume boundaries.

We break up the visible events into five categories forthe pile-up studies:

1. Events originating outside of the ID and enteringthe ID.

2. Events originating inside the ID with visible par-ticles escaping the ID. These are called partiallycontained (PC) ID events.

3. Events originating inside the ID with no visible par-ticles escaping the ID. These are called fully con-tained (FC) ID events.

4. Events originating in the OD with no visible parti-cles entering the ID.

5. Events originating outside the OD with visible par-ticles entering the OD, but not the ID.

The first three categories represent the event rate in theID, while all but the second category represent the eventrate in the OD. Table I shows the simulated event ratesper 2.5 × 1013 protons on target, the assumed protonsper bunch for full 750 kW operation. Rates are shownfor the nuPRISM configurations where the ID covers off-axis angle ranges of 0.0-0.6, 1.0-1.6, 2.0-2.6 or 3.0-3.6degrees. While the current design does not include apit that extends to on-axis, the 0.0-0.6 degree position isused to make comparisons to the INGRID event rates.

For the off-axis angle 1.0-1.6 degree position, the to-tal rate of ID+OD visible events in a spill (8 bunches) is6.12. If a bunch contains an event, the probability thatthe next bunch contains at least one visible event is 53%.This suggests that nuPRISM should employ deadtime-less electronics that can record events in neighboringbunches and that the after-pulsing of PMTs should becarefully considered. The rate of ID events per bunch is0.230 and the probability of two or more visible ID eventsin a single bunch with at least one visible event is 20%.Hence, most bunches will not require the reconstructionof multiple interactions in the ID volume. However, theprobability of 2 or more ID events per spill is 84%, so thereconstruction of out of time events such as decay elec-trons needs to be carefully studied. Decay electrons in aspill may potentially be matched to their parent interac-tions using both spatial and timing information. For in-teractions inside the ID, a spatial likelihood matching thedecay electron to the primary vertex may be constructedbased on the reconstructed decay electron vertex positionand the reconstructed primary vertex or reconstructedstopping point of the candidate muons or charged pions

10

(GeV)νE0 1 2 3 4 5

(D

egre

es)

OA

θ

1

1.5

2

2.5

3

3.5

4

2000

4000

6000

8000

10000

1210× FluxµνPRISM ν

(GeV)νE0 1 2 3 4 5

(D

egre

es)

OA

θ

1

1.5

2

2.5

3

3.5

4

50

100

150

200

250

300

350

4001210× FluxµνPRISM ν

(GeV)νE0 1 2 3 4 5

(D

egre

es)

OA

θ

1

1.5

2

2.5

3

3.5

4

10

20

30

40

50

1210× FluxeνPRISM ν

(GeV)νE0 1 2 3 4 5

(D

egre

es)

OA

θ

1

1.5

2

2.5

3

3.5

4

0

1000

2000

3000

4000

910× FluxeνPRISM ν

FIG. 7. The neutrino flux (arbitrary normalization) as a function of off-axis angle and energy for each neutrino flavor with thehorn in neutrino-mode operation.

TABLE I. The event rates per 2e13 POT for nuPRISM with horn currents at 320 kA.

Off-axis Angle (◦) Entering ID PC ID FC ID OD Contained Entering OD0.0-0.6 0.4179 0.2446 0.3075 1.2904 0.70761.0-1.6 0.1005 0.0550 0.0741 0.3410 0.19392.0-2.6 0.0350 0.0198 0.0230 0.1234 0.06353.0-3.6 0.0146 0.0092 0.0156 0.0564 0.0291

in the event. For decay electrons originating from muonsproduced outside of the ID, a similar spatial likelihoodmay be constructed using OD light, ID light, and hitsfrom scintillator panels (if they are installed between theOD and ID) from the entering particle. Since the muonmean lifetime (2.2 µs) is shorter than the spill length( 5 µs), there will also be statistical power to match de-cay electrons to their primary vertex based on the timeseparation of the decay electron vertex and primary ver-tex. On the other hand, the muon lifetime may providea cross-check for the spatial matching of primary and de-cay electron vertices since significant mismatching wouldtend to smear the time separation distribution beyondthe muon lifetime. Studying the matching of decay elec-

trons to primary interactions is a high priority and workis underway to address this issue with a full simulationof nuPRISM and the surrounding rock.

The rate of events producing light in the OD is 0.690per bunch. Hence, the probability that an FC ID eventwill have OD activity in the same bunch is 50%. Ne-glecting out of time events, the rejection rate of FC IDevents would be 50% if a veto on any OD activity in thebunch is applied. This rejection rate falls to 21% and10% in the 2.0-2.6 and 3.0-3.6 degree off-axis positionsrespectively. Of the OD events, about 30% are enter-ing from the surrounding earth, and most of those aremuons. The scintillator panels may be used to relax theveto on these types of pile-up events by providing ad-

11

Z (m)995 1000 1005

X (

m)

-6

-4

-2

0

2

4

6

900

950

1000

1050

1100

Vertex: X vs. Z

Z (m)995 1000 1005

Y (

m)

-50

-40

-30

-20

-10

0

10

20

1

10

210

310

Vertex: Y vs. Z

X (m)-6 -4 -2 0 2 4 6

Y (

m)

-50

-40

-30

-20

-10

0

10

20

1

10

210

310

Vertex: Y vs. X

FIG. 8. The distribution of simulated vertices shown inprojections to the x-z (top), y-z (middle) and x-y (bottom)planes. Here x is defined as the horizontal axis perpendicularto the beam, z is the horizontal axis in the beam directionand y is the vertical axis.

Beam

Sand Volume

Water VolumeR=5 m

53 m22 m

78 m

ID, OD and intermediate volumes

FIG. 9. The GEANT4 geometry used in the pile-up simula-tion.

ditional spatial and timing separation between the ODand ID activity in the same bunch. If the veto can beremoved for all events entering the OD from the earth,then rejection rates due to OD pile-up drop to 39%, 16%and 8% for the 1.0-1.6, 2.0-2.6 and 3.0-3.6 degree off-axisangle positions respectively.

We can cross-check the estimated nuPRISM eventrates by extrapolating from the event rates observed byINGRID. We assume that the rate of interactions insidethe detector will scale with the detector mass, and therate of entering events from the earth will scale with thecross-sectional area of the detector. The rates shouldalso scale with 1/d2, were d is the distance from the av-erage neutrino production point to the detector, about240 m for INGRID and 960 m for nuPRISM. INGRIDobserves 1.74 neutrino events per 1 × 1014 POT in 14INGRID modules with a total mass of 5.7× 104 kg. Foran OD mass of 8.2× 105 kg, we extrapolate the INGRIDrate, assuming 60% detection efficiency in INGRID, toobtain 0.66 interactions in the OD for each 2.5 × 1013

12

POT bunch. The simulated rate of visible OD interac-tions in nuPRISM is 1.50 and 0.39 for the 0.0-0.6 and1.0-1.6 degree positions respectively. Since INGRID cov-ers an angular range of about ±1 degree, it is reasonablethat the extrapolated value from INGRID falls betweenthe simulated nuPRISM values at these two positions.

INGRID also observes a event rate from earth interac-tions of 4.53 events per 1×1014 POT in 14 modules witha cross-sectional area of 21.5 m2. These earth interac-tion candidates are INGRID events failing the upstreamveto and fiducial volume cuts. The selection of enter-ing earth-interaction events is > 99% efficient and 85.6%pure. Scaling to the OD cross-sectional area and distancewhile correcting for the efficiency and purity gives a rateof 0.31 events entering the OD per bunch. The rate fromthe nuPRISM pile-up simulation is 0.903 or 0.239 for the0.0-0.6 and 1.0-1.6 degree positions respectively. Onceagain, the extrapolated INGRID rate falls between thesimulated rates for these two nuPRISM positions.

In summary, the event pile-up rates for nuPRISM ap-pear manageable. Even for the most on-axis positionand high power beam, most bunches with interactionswill only have a single interaction with visible light inthe ID. The OD veto rate from pile-up can be as largeas 50%, hence careful studies of the OD veto are needed.The OD veto rate may be reduced and better understoodwith the inclusion of scintillator panels at the outer edgeof the OD or at the OD/ID boundary. The electronicsfor nuPRISM should be deadtime-less to handle multipleevents per spill.

Further studies of the event rates will be carried out.These will include the study of entering neutral particlesto be used in the optimization of the OD and fiducialvolume sizes, more realistic studies of how the scintillatorpanels may be used to optimize the OD veto cut, andupdates to the earth density to better reflect the surveyeddensity of the rock strata at potential nuPRISM sites.

E. Event Selection for Sensitivity Studies

We select samples of single ring muon and electron can-didates for the long and short baseline sensitivity studiesdescribed in the following sections. As described in Sec-tion II C, the efficiencies for single ring electron or muonselections are applied using tables calculated from the SKMC. The efficiency tables are calculated with the follow-ing requirements for muon and electron candidates:

• Muon candidate requirements: fully contained, asingle muon-like ring, 1 or fewer decay electrons

• Electron candidate requirements: fully contained, asingle electron-like ring, no decay electrons, passesthe fiTQun π0 cut

Additional cuts are applied on the smeared νPRISM MC.For the muon candidates the cuts are similar to the SKselection for the T2K disappearance analysis:

• Muon candidate cuts: dWall > 100 cm, toWall >200 cm, Evis > 30 MeV, pµ > 200 MeV/c

where dWall is the distance from the event vertex to thenearest wall, and toWall is the distance from the vertexto the wall along the direction of the particle.

For the single ring electron candidates, the cuts ontoWall and Evis were reoptimized since the separationbetween electrons and muons or electrons and π0s de-grades closer to the wall. The cut on dWall is set to 200cm to avoid entering backgrounds. The cuts are:

• Electron candidate cuts: dWall > 200 cm,toWall > 320 cm, Evis > 200 MeV

The tight fiducial cuts for the electrons candidates areneeded to produce a relatively pure sample, but there isa significant impact to the electron candidate statistics.A simulation with finer PMT granularity may allow forthe toWall cut to be relaxed, increasing the statisticswithout degrading the purity.

F. T2K νµ Disappearance Sensitivities

The most straightforward application of the nuPRISMconcept to T2K is in the νµ disappearance measurement.A full νµ analysis has been performed in which nuPRISMcompletely replaces ND280. In the future, it will be use-ful to incorporate ND280 into nuPRISM analyses, par-ticularly the sterile neutrino searches, but for simplicitythis has not yet been done.

The main goal of this νµ disappearance analysis is todemonstrate that nuPRISM measurements will removemost of the neutrino cross section systematic uncertain-ties from measurements of the oscillation parameters.This is achieved by directly measuring the muon momen-tum vs angle distribution that will be seen at Super-Kfor any choice of θ23 and ∆m2

32.To clearly compare the nuPRISM νµ analysis with

the standard T2K approach, the full T2K analysis isreproduced using nuPRISM in place of ND280. Thisis done by generating fake data samples produced fromthrows of the flux and cross section systematic parame-ters and fitting these samples using the standard oscilla-tion analysis framework. In each flux, cross section andstatistical throw, three fake data samples using differ-ent cross section models were produced at both ND280and Super-K: default NEUT with pionless delta decay,NEUT with the Nieves multinucleon model replacing pi-onless delta decays, and NEUT with an ad-hoc mult-inucleon model that uses the final state kinematics ofthe Nieves model and the cross section from Martini etal. For each throw, all three fake data samples were fitto derive estimates of the oscillation parameters. Thedifferences between the fitted values of sin2 θ23 for theNEUT nominal and NEUT+Nieves or NEUT+Martinifake data fits are shown in Figure 10. The systematicuncertainty associated with assuming the default NEUT

13

model rather than the model of Martini or Nieves is givenby the quadrature sum of the RMS and mean (i.e. bias)of these distributions. For the ND280 analysis, there is a3.6% uncertainty when comparing with the Nieves model,and a 4.3% uncertainty in the measured value of sin2 θ23

when comparing with the Martini model. These uncer-tainties would be among the largest for the current T2Kνµ disappearance analysis, and yet they are based solelyon model comparisons with no data-driven constraint.

FIG. 10. The results of fitting fake data with and withoutmultinucleon effects are shown. The measured differences insin2 θ23 when comparing the Nieves model (blue) to defaultneut (black) and the Martini model (red) to default neut giveRMS values of 3.6% and 3.2%, respectively, and biases of 0.3%and -2.9%, respectively.

As was discussed in Section I A the limitation of usingND280 data to predict observed particle distributions at

Super-K is that the neutrino flux at these two detectorsis different due to oscillations. Therefore, any extrapo-lation has significant and difficult to characterize crosssection model dependent uncertainties. In the nuPRISMbased analysis, this limitation is resolved by deriving lin-ear combinations of the fluxes at different off-axis an-gles to produce a flux that closely matches the predictedoscillated flux at Super-K. The observed particle distri-butions measured by nuPRISM are then combined withthe same linear weights to predict the particle distribu-tion at Super-K. In this way, the analysis relies on theflux model to determine the weights that reproduce theoscillated flux while minimizing cross section model de-pendence in the extrapolation.

The first stage of the nuPRISM νµ analysis is to sepa-rate the 1-4 degree off-axis range of the detector into 300.1 degree or 60 0.05 degree bins in off-axis angle. Theneutrino energy spectrum in each off-axis bin is predictedby the T2K neutrino flux simulation. For each hypoth-esis of oscillation parameter values that will be testedin the final oscillation fit, the oscillated Super-K energyspectrum is also predicted by the T2K neutrino flux simu-lation. A linear combination of the 30 (60) off-axis fluxesis then taken to reproduce each of the Super-K oscillatedspectra,

ΦSK(Eν ; θ23,∆m

232

)Eν =

30∑i=1

ci(θ23,∆m

232

)EνΦνPi (Eν),

(2)where ci

(θ23,∆m

232

)is the weight of each off-axis bin, i.

The extra factors of Eν are inserted to approximate theeffect of cross section weighting. The ci

(θ23,∆m

232

)are

determined by a fitting routine that seeks agreement be-tween the Super-K flux and the linear combination overa specified range of energy. An example linear combina-tion of nuPRISM off-axis fluxes that reproduces the SKflux is shown in Figure 11. These fits can successfullyreproduce Super-K oscillated spectra, except at neutrinoenergies below ∼ 400 MeV. The maximum off-axis angleis 4◦, which peaks at 380 MeV, so at lower energies it isdifficult to reproduce an arbitrary flux shape. This couldbe improved by extending the detector further off-axis.

The determination of the ci(θ23,∆m

232

)weights to re-

produce the oscillated flux is subject to some optimiza-tion. Figure 12 shows two sets of weights for a particu-lar oscillation hypothesis. In the first case a smoothnessconstrain was applied to the weights so that they varysmoothly between neighboring off-axis angle bins. In thesecond case the weights are allowed to vary more freelyrelative to their neighbors. Figure 13 shows the compar-isons of the nuPRISM flux linear combinations with theSuper-K oscillated flux for a few oscillation hypothesesin the smoothed and free weight scenarios. The oscil-lated flux in the maximum oscillation region is nearlyperfectly reproduced when the weights are allowed tovary more freely. When they are constrained to varysmoothly, the agreement is less perfect, although stillsignificantly better than the agreement between ND280

14

and Super-K fluxes. An analysis using the free weightsis less dependent on the cross section model assumptionsin the extrapolation to Super-K since the Super-K fluxis more closely matched. On the other hand, the analy-sis with the smoothed weights is less sensitive to uncer-tainties on the flux model and nuPRISM detector modelthat have an off-axis angle dependence since neighboringbins have similar weight values. The statistical errors arealso smaller for the smoothed weight case since the sumin quadrature of the weights in a given neutrino energybin is smaller when there are less fluctuations in weightvalues. In the analysis presented here, the smoothedweights are used, although the optimization of the levelof smoothness is an area where the analysis will be im-proved in the future.

FIG. 11. A sample fit of the flux in 30 nuPRISM fluxes toan oscillated Super-K flux is shown. Good agreement canbe achieved, except at low energies due to the 4◦ maximumoff-axis angle seen by nuPRISM.

The nuPRISM candidate events are events with a sin-gle observed muon ring and no-other observed particles,matching the selection applied at Super-K. After theci(θ23,∆m

232

)coefficients are derived, they are used to

make linear combination of observed candidate event dis-tributions from each nuPRISM off-axis bin. In this casethe observables are the momentum and polar angle ofthe scattered muon candidate, and hence the expectedSuper-K distribution of these observables is predicted bythe linear combination of observed nuPRISM events.

In order to use these nuPRISM measurements to makean accurate prediction of Super-K muon kinematics, a se-ries of corrections are required. First, non-signal eventsfrom either neutral current events or charged currentevents with another final state particle above Cherenkovthreshold, must be subtracted from each near detectorslice. This is particularly important for neutral currentevents, which depend on the total flux rather than theoscillated flux at Super-K, but depend on the oscillatedflux in the nuPRISM linear combination. This back-ground subtraction is model dependent, and is a source

FIG. 12. The weights for each off-axis bin produced in thenuPRISM flux fits are shown after requiring that neighboringbins have similar values (top; as in Figure 13 left column) andwith neighboring bins allowed to vary more freely relative toeach other (bottom; as in Figure 13 right column).

of systematic uncertainty, although neutral current inter-actions can be well constrained by in situ measurementsat nuPRISM. The differences in detector efficiency andresolution must also be corrected. The efficiency differ-ences are due to differences in detector geometry and arelargely independent of cross section modeling. Detec-tor resolutions must be well determined from calibrationdata, but this effect is somewhat mitigated due to thefact that the near and far detector share the same de-tector technology. Finally, for the present analysis, thetwo dimensional muon momentum vs angle distributionis collapsed into a one dimensional Erec distribution us-ing a transfer matrix, Mi,p,θ (Erec). This is an arbitrarychoice that does not introduce model dependence into thefinal result, and has only been used for consistency withexisting T2K νµ disappearance results. Future analysescan be conducted entirely in muon momentum and angle

15

FIG. 13. Fits of the nuPRISM flux bins to oscillated Super-K fluxes are shown for three different sets of(θ23,∆m

232

): top -

(0.61, 2.56 ∗ 10−3), middle - (0.48, 2.41 ∗ 10−3), and bottom - (0.41, 2.26 ∗ 10−3). In the left column, the weights for the off-axisbins are forced to vary smoothly with off-axis angle, while in the right column they are allowed to vary more freely.

variables.The final expression for the nuPRISM prediction for

the Super-K event rate is then

NSK(Erec; θ23,∆m

232

)= δ (Erec) +BSK

(Erec; θ23,∆m

232

)+

30∑i=1

∑p,θ

ci(θ23,∆m

232

) (NνPi,p,θ −BνPi,p,θ

)×εSKp,θενPi,p,θ

Mi,p,θ (Erec) ,

(3)

where NSK (Erec) and NνPi,p,θ are the number of expected

events in Super-K Erec bins and nuPRISM off-axis an-gle, muon momentum, and muon angle bins, respectively,BSK (Erec) and BνPi,p,θ are the corresponding number of

background events in these samples, and εSKp,θ and ενPi,p,θ

are the efficiencies in each detector. The final correctionfactor, δ (Erec), accounts for any residual differences be-tween the nuPRISM prediction and the Super-K eventrate predicted by the Monte Carlo simulation. These aremostly due to the previously described imperfect fluxfitting, and the fact that nuPRISM is not sensitive toneutrino energies above ∼ 1.5 GeV since most muonsat that energy are not contained within the inner de-tector. Comparisons of the Super-K event rate and thenuPRISM prediction for Super-K prior to applying theδ (Erec) correction factor are given in Figure 14.

The nuPRISM technique effectively shifts uncertaintiesin neutrino cross section modeling into flux predictionsystematic uncertainties. This is quite helpful in oscilla-tion experiments since many flux systematic uncertain-ties cancel, and the important physical processes in theflux prediction, the hadronic scattering, can be directly

16

FIG. 14. The Super-K Erec distributions and nuPRISM Erecpredictions corresponding to the flux fits in Figure 13 (leftcolumn) are shown prior to applying the δ (Erec) correctionfactor.

measured by dedicated experiments using well charac-terized proton and pion beams. Figure 15 shows the ef-fect of a few selected flux uncertainties on the Super-Kenergy spectrum and the nuPRISM linear combination.The largest flux uncertainty is due to pion production inproton-carbon interactions, but this uncertainty mostlycancels when applied at both the near and far detector.The more problematic uncertainties are those that affectthe off-axis angle, such as horn current and proton beampositioning, since these effects will impact Super-K andthe nuPRISM linear combinations differently. Figure 16

shows four examples of how the Super-K Erec distribu-tion and the corresponding nuPRISM predicted distri-bution vary for different throws of all the flux and crosssection systematic uncertainties. The predicted spectrafrom the nuPRISM linear combination closely tracks thetrue spectrum at SK, indicating a correlated effect frommost systematic parameters on the nuPRISM linear com-bination and SK event rates.

The final covariance matrices are shown in Fig-ure 17. The largest errors are at high energies where nonuPRISM events are present due to the smaller diameterof the detector relative to Super-K. In this region, theSuper-K prediction is subject to the full flux and crosssection uncertainties with no cancelation at the near de-tector. Similarly, at energies below 400 MeV the errorsget larger since the current 4◦ upper bound in off-axisangle prohibits the nuPRISM flux fit from matching theSuper-K spectrum at low energies.

Using the nuPRISM covariance matrices shown in Fig-ure 17 in place of those produced by ND280, the standardT2K νµ disappearance oscillation analysis is repeated.The results are shown in Figure 18. As expected, thenuPRISM analysis is largely insensitive to cross sectionmodeling. Replacing the default new model with theNieves multinucleon model now produces a 1.0% uncer-tainty in sin2 θ23, and the corresponding Martini uncer-tainty is 1.2%. More importantly, this uncertainty is nowconstrained by data rather than a pure model compari-son. These uncertainties are expected to be further re-duced as the flux fits are improved, and nuPRISM con-straints on NC backgrounds and information from ND280are incorporated into the analysis.

17

(GeV)νE0.5 1 1.5 2 2.5 3

SK P

redi

ctio

n R

atio

0.85

0.9

0.95

1

1.05

1.1

1.15

SK MC (Random Throw)/Nominal

PRISM Linear Comb. (Random Throw)/Nominalν

Pion Multiplicity Throw

(GeV)νE0.5 1 1.5 2 2.5 3

SK P

redi

ctio

n R

atio

0.85

0.9

0.95

1

1.05

1.1

1.15

Change)/NominalσSK MC (1

Change)/NominalσPRISM Linear Comb. (1ν

Horn Current +5 kA Change

(GeV)νE0.5 1 1.5 2 2.5 3

SK P

redi

ctio

n R

atio

0.85

0.9

0.95

1

1.05

1.1

1.15

Change)/NominalσSK MC (1

Change)/NominalσPRISM Linear Comb. (1ν

Proton Beam -1 mm Y Shift

FIG. 15. Systematic uncertainties on the neutrino flux pre-diction due to pion production (top), horn current (middle),and proton beam y-position (bottom) are shown.

FIG. 16. Variations in the Super-K Erec spectrum and thecorresponding nuPRISM prediction are shown for 4 throws ofall the flux and cross section parameters. Significant correla-tions exist between the the near and far detector, which helpto reduce the systematic uncertainty.

18

FIG. 17. Covariance matrices are shown (from top to bot-tom) for the total, statistical, systematic, and flux only un-certainties. The bin definitions (in GeV) are 0: (0.0,0.4), 1:(0.4,0.5), 2: (0.5,0.6), 3: (0.6,0.7), 4: (0.7,0.8), 5: (0.8,1.0),6: (1.0,1.25), 7: (1.25,1.5), 8: (1.5,3.5), 9: (3.5,6.0), 10:(6.0,10.0), 11: (10.0,30.0)

FIG. 18. The variation in the measured sin2 θ23 due to mult-inucleon effects in the nuPRISM νµ analysis are shown. Forthe Nieves and Martini fake datasets, the RMS produces 1.0%and 1.2% uncertainties, respectively, with no measurable bias.This is a large improvement over the standard T2K resultsshown in Figure 10

19

G. nuPRISM 1-Ring e-like Ring Measurements

Single ring e-like events in nuPRISM at an off-axis an-gle of 2.5◦ in principle provide a reliable estimate of theνe appearance background at SK, since the near-to-farextrapolation correction is small. This includes beam νe,NCπ0, and NC single γ (NCγ) backgrounds with produc-tion cross section and detection efficiency in water foldedin. For a νe background study with better than ∼10%precision, more careful studies are required: for example,the γ background from outside the detector scales differ-ently between the near and far detectors due to the differ-ences in surface to volume ratio. Contributions from CCbackgrounds, e.g. CCπ0 events created outside the detec-tor, would also be different between near and far detectordue to oscillation. Careful identification of each type ofsingle ring e-like event is required. As described below,the nuPRISM capability of covering wide off-axis rangesmakes such a study possible. It also enables relative crosssection measurements between νe and νµ, which are likelyto be limiting systematic uncertainties for measuring CPviolation.

The nuPRISM detector will also provide a unique andsensitive search for sterile neutrinos in the νµ → νe chan-nel, and eventually the νµ → νµ channel, particularlywhen ND280 is incorporated into the analysis. The 1kmlocation of nuPRISM for the off-axis peak energies of 0.5-1.0GeV matches the oscillation maximum for the sterileneutrinos hinted by LSND and MiniBooNE. The pres-ence or absence of an excess of νe events as a functionof off-axis angle will provide a unique constraint to ruleout many currently proposed explanations of the Mini-BooNE excess, such as feed-down in neutrino energy dueto nuclear effects. The off-axis information also allowsfor a detailed understanding of the backgrounds, sincethey have a different dependence on off-axis angle thanthe oscillated signal events.

FIG. 19. Reconstructed neutrino energy distribution for theνe appearance analysis of MiniBooNE [20].

Figure 19 shows the single ring e-like events observedby MiniBooNE. There are several sources of events:

• Beam νe from muons and kaons

• NCπ0 with one of the photons missed

• NCγ (∆→ Nγ)

• ”Dirt” events: background γ coming from outside

• Others, such as CC events with µ misidentified aselectron

• Possible sterile neutrino contribution causing νµ →νe oscillation

There is a significant discrepancy between data and theMonte Carlo prediction. For precision νe appearancestudies, such as CP violation, it is essential to under-stand the origin of this discrepancy.

1. Beam νe and νe cross section study

The beam νe represents only 1% of the total neu-trino flux and about 0.5% at the off-axis peak energy atEν=600MeV. Thanks to the excellent µ/e particle identi-fication and π0 suppression in water Cherenkov detectorswhen using fiTQun, the νµ background is expected to besuppressed, similar to the suppression seen at Super-K.Since the beam νe’s originate from three body decays ofmuons and kaons, their off-axis dependence is more mildthan the dependence seen in the νµ flux. By taking ad-vantage of the steep off-axis angle dependence of the νµflux, it is possible to study background contamination indetail. For example, the νµ backgrounds are largely sup-pressed compared to beam νe at an off-axis angle largerthan 3 degrees. The beam νe events at nuPRISM providean opportunity to precisely study νe cross sections, forwhich there is currently very little data available. Thecross section difference between νe and νµ, which doesnot cancel in the near to far detector extrapolation inνµ → νe appearance, is considered to be an eventual lim-itation of the CP violation sensitivity [21]. The differ-ences in the νe and νµ cross sections come from kinemat-ical phase space differences due to the difference in massbetween electron and muons, radiative corrections, pos-sible second class currents, which also depend on leptonmass, and nuclear effects [10].

nuPRISM provides a unique method for canceling theflux differences between νe and νµ. Using a techniquesimilar to that used in the nuPRISM νµ disappearanceanalysis, it is possible to use linear combinations of νµmeasurements at different off-axis angles to reproducethe shape of the intrinsic νe flux in the large off-axisangle section of nuPRISM:

Φνe(Eν) = ΣciΦiνµ(Eν), (4)

where Φνe(Eν) is the nuPRISM νe flux of interest,Φiνµ(Eν) is the νµ flux at the ith off-axis position and

ci is the weight factor for the ith off-axis position. Usingthis combination, the ratio of the νe and νµ double dif-ferential cross sections in momentum and angle can bedirectly measured, averaged over the νe flux spectrum.

20

Fig. 20 shows that the nuPRISM 2.5◦−4.0◦ off-axis νeflux can be reproduced by the linear combination of νµfluxes for the 0.3-1.5 GeV energy range. Above 1.5 GeVthe νe flux cannot be produced since the fall-off of theνµ fluxes is steeper. However, this region will have littleimpact for the ratio measurement for a couple of reasons.First, Fig. 20 shows the flux multiplied by the energy toapproximate the effect of the cross section, but the crosssection for CC interactions producing no detectable pi-ons is growing more slowly than this linear dependenceand the rate from the high energy flux will be lower thanit appears in the figure. Second, the analysis will beapplied in the limited lepton kinematic range where thenuPRISM muon acceptance is non-zero, cutting out for-ward produced high momentum leptons. This will alsosuppress the contribution from the high energy part ofthe flux.

(GeV)νE0 0.5 1 1.5 2 2.5 3

ν)*

Eν

(EΦ

0

100

200

300

400

610×

sk_nom_numode_nue

Entries 2.27779e+07

Mean 1.521

RMS 0.7749

)° (2.5-4.0eνPRISM ν

Linear Combo.µνPRISM ν

(GeV)νE0 0.5 1 1.5 2 2.5

ν)*

Eν

(EΦ

0

5000

10000sk_nom_numode_nue

Entries 2.27747e+07

Mean 0.9408

RMS 0.6043

eνSK Beam+Osc.


FIG. 20. Fits of the off-axis nuPRISM νµ fluxes tothe nuPRISM 2.5◦ − 4.0◦ off-axis νe flux (top) andthe oscillated+intrisic beam νe at SK (bottom) assum-ing sin22θ13=0.094, δcp=0, ∆m2

32 = 2.4 × 10−3eV2 andsin2θ23=0.5.

2. Predicting oscillated νe for the appearance measurement

As discussed in the previous section, the cross sectionratio of σνe/σνµ can be measured using beam νe and νµinteraction candidates in nuPRISM. The measured crosssection ratio can be used to apply the nuPRISM extrapo-lation method to predict the νe candidates at SK for theappearance measurement. Following the procedure usedfor the disappearance analysis, the oscillated+intrinsicbeam νe flux is described by a linear combination of thenuPRISM off-axis νµ fluxes:

ΦSKνµ (Eν)Pνµ→νe(Eν |θ13, δcp, ...) + ΦSKνe (Eν)

=∑

ci(θ13, δcp, ...)Φiνµ(Eν).

(5)

ΦSKνµ (Eν) and ΦSKνe (Eν) are the predicted νµ and νefluxes at SK in the absence of oscillations. Pνµ→νe is

the νµ to νe oscillation probability. Φiνµ(Eν) is the ith

off-axis νµ flux in nuPRISM and the ci are the derivedcoefficients that depend on the oscillation hypothesis be-ing tested. Fig. 20 shows the level of agreement that canbe achieved between the linear combination of nuPRISMfluxes and the predicted SK νe flux for a particular os-cillation hypothesis. The agreement is excellent between0.4 and 2.0 GeV. Below 0.4 GeV, the second oscillationmaximum is not reproduced, but the rate from this partof the flux is small.

Using the derived ci coefficients, the measured muonp, θ distributions from nuPRISM are used to predict theSK p, θ distribution for the νe flux. An additional correc-tion must be applied to correct from the predicted muondistribution for νµ interactions to the predicted electrondistribution for νe interactions. This correction is derivedfrom the cross section models which are constrained bythe ratio measurement described in the previous section.

3. Backgrounds from νµ’s

The backgrounds from νµ comes from NCπ0 eventswith one γ missed, NCγ events (∆ → Nγ), CC eventswith e/µ mis-ID, γ’s coming from ν (mainly νµ) interac-tion outside the detector (dirt or sand events). Becausethe νµ energy spectrum changes dramatically as a func-tion of vertex positions (= off-axis angles) in nuPRISM,these background processes can be studied and verifiedby comparing their vertex distributions.

The NCπ0 rate can be measured by detecting two γ’sin nuPRISM. By using the hybrid π0 technique used inT2K-SK analysis, the π0 backgrounds with a missing γcan be estimated using the beam νe and Michel elec-trons as electron samples combined with a Monte Carloγ event. The NCπ0 rate can also be used to estimate theNCγ rate. As mentioned above, dirt/sand background issuppressed by having fully active outer veto detector andthe fiducial volume cut. The vertex distribution of the νeevents as a function of the distance from the (upstream)

21

wall provides an excellent confirmation of the suppressionof the background, as is done in the T2K-SK analysis.

4. Sterile Neutrino Sensitivity

The position of nuPRISM, at 1 km from the neutrinosource, as well as its huge fiducial mass makes this detec-tor an excellent candidate for the studies of non-standardshort-baseline neutrino oscillations. This section presentsan initial, conservative sensitivity of nuPRISM to the so-called LSND anomaly. The LSND and MiniBooNE ex-periments detect an undetermined excess in their νe andνe channels, which may be explained by sterile neutrinomixing with a sin2(2θµe) ∼ 10−3 and ∆m2

41 ∼ 2eV 2 inthe 3+1 model [20].

Here we present the sensitivity studies for a two dif-ferent layouts of the nuPRISM detector: 3 m radius and4 m radius. We performed our νe selection analysis con-sidering an exposure of 4.6 × 1020 p.o.t. with a hornconfiguration enhancing neutrinos and defocusing anti-neutrinos. The possible νe disappearance due to sterilemixing is neglected as in the case of the LSND and Mini-BooNE analyses. This is justified by the fact that we haveonly 1% of νe in the beam and the νµ → νe channel willbe dominant. In the case where both νe disappearanceand appearance are considered, our current results canbe seen as lower limits for the mixing angle sin2(2θµe).

We test the simplest sterile neutrino model by addingto the standard three-neutrino parametrization one ad-ditional mass state, mainly sterile, with a mass differencerelative to the other states of ∆m2

41. Since the mixingwith the sterile neutrino is dominant at short baselines,such as the nuPRISM baseline, the new mass state isexpected to be much larger (∼ eV 2) than the two stan-dard neutrino mass splittings. In such conditions thetwo-neutrino approximation is valid and provides the fol-lowing νe appearance probability,

Pνµνe = P (νµ → νe)

= sin2(2θµe) sin2

(1.27∆m2

41[eV2]L[km]

E[GeV]

),

(6)

where L is the neutrino flight path fixed at 1km and Ethe energy of the neutrinos. sin2(2θµe) = 4|Ue4|2|Uµ4|2where U are the new elements in the extended PMNSmatrix. We consider an analysis on the reconstructedenergy (Erec) and off-axis angle (OAA) shape informa-tions, so both rate and shape are taken into account bybuilding bidimensional binned templates. The expectednumber of background and signal events entering in theνe selection are shown in Table II for different oscillationhypothesis and both detector radius cases.

The systematic errors due to the flux and the cross-section uncertainties are included through a covariancematrix that is calculated using toy Monte Carlo throws.A χ2 test for a binned template of 10 ERec bins and 10OAA bins is performed between 0.2 MeV and 4 MeV, in

TABLE II. Expected number of events in the νe selection foreach oscillation hypothesis, and for the two detector innerdiameters being considered.

(sin2(2θµe),∆m241) 3 m radius 4 m radius

νµ → νe Signal (0.001, 1 eV 2) 87.6 484.3(0.005, 1eV 2) 437.8 2421.7(0.01, 10eV 2) 635.2 3521.0(0.001, 10eV 2) 63.5 352.1

Background νe 1076.2 6695.5νµ 983.8 4700.7

order to obtain the expected sensitivity in the bidimen-sional oscillation parameter space (sin2(2θµe),∆m

241).

For each oscillation hypothesis the χ2 value is given by

χ2 = ~ns(sin2(2θµe),∆m

241

)T × V −1

×~ns(sin2(2θµe),∆m

241

) (7)

where ~ns is the n-tuple of number of expected signalevents due to νµ → νe in ERec and OAA bins, and V isa 100×100 covariance matrix that includes the statisticsand systematic errors. The χ2 is computed for each pointof a bidimensional grid and the constant ∆χ2 methodis applied to determine the contours for the regions ex-cluded at the 90% C.L. The final sensitivity is shown inin Fig. 21 for the 90% C.L. along with a comparison withthe MiniBooNE antineutrino results.

We observe that the final sensitivity, taking into ac-count statistical uncertainties as well as flux and crosssection systematic errors, contains, for the 4m inner de-tector radius case, the full MiniBooNE allowed region at90% C.L. Regarding the 3m case, the detector is ableto explore the whole low ∆m2

41 region allowed by Mini-BooNE and it covers most of the high ∆m2

41 part. Thesensitivity has been computed without using any con-straints from ND280. In the nuPRISM analysis sce-nario, ND280 has the role of reducing model uncertain-ties in flux and cross sections, so the final errors for a fullnuPRISM +ND280 analysis are expected to be signifi-cantly reduced, but have not yet been computed. More-over, a νe appearance analysis allows for the use of thenuPRISM νµ analysis to further constrain the flux andcross section systematics, which should further improveupon the sensitivity predicted in this study.

H. νµ Measurements

In principle, the nuPRISM technique of using multi-ple off axis angles to measure the oscillated pµ and θµfor each oscillated flux will work for anti-neutrinos aswell. However, when running the T2K beam in anti-neutrino mode, there is a significant wrong-sign back-ground from neutrino interactions. To disentangle these

22

)θ(22sin

310 210 110

2 m

∆

210

110

1

10

No Systs

Flux Systs

All Systs

MiniBooNE

)θ(22sin

310 210 110

2 m

∆

210

110

1

10

No Systs

Flux Systs

All Systs

MiniBooNE

FIG. 21. 90% C.L. expected sensitivities for an exposure of4.6 × 1020 p.o.t. for three scenarios: statistical uncertaintyonly, both statistical uncertainties and flux systematic un-certainties, and statistical uncertainties with flux and cross-section systematic uncertainties. The sensitivity curves areshown for the two detector configuration considered: 3m (top)and 4m (bottom) inner detector radius. For comparison, theMiniBooNE allowed region at 90% C.L. in antineutrino modeis shown in red.

neutrino and anti-neutrino interactions, linear combina-tions of the neutrino-mode data can be used to constructthe wrong-sign flux in anti-neutrino mode, analogous tothe procedure used in Section II F to construct the Super-K oscillated spectra and in Section II G 1 to construct theelectron neutrino spectrum. Hence, the neutrino flux inthe anti-neutrino mode is described with the linear com-bination of neutrino mode fluxes:

Φνmodeνµ (Eν , θoa) =∑

ci(θoa)Φi,νmodeνµ (Eν). (8)

Φνmodeνµ (Eν , θoa) is the anti-neutrino mode νµ (wrong-

sign) flux for a given off-axis angle θoa. Φi,νmodeνµ (Eν)

is the neutrino mode νµ (right-sign) flux for the ith off-

axis bin and ci is the weight for the ith off-axis bin thatdepends on the off-axis angle for which the anti-neutrinomode wrong sign flux is being modeled.

Linear combinations to reproduce the wrong-sign 1.0−2.0◦, 2.0− 3.0◦ and 3.0− 4.0◦ anti-neutrino mode fluxesare shown in Figure 22. As with the combinations to pro-duce the νe flux, the agreement is good up to about 1.5GeV in neutrino energy. As discussed in Section II G 1,it is less important to reproduce the high energy part ofthe flux since high energy interactions are suppressed bythe event topology selected and the muon acceptance ofnuPRISM.

As shown Figure 23, there is significant correlation be-tween the wrong-sign neutrino flux in anti-neutrino modeand the neutrino-mode flux, so the flux uncertainties willgive some cancelation using this method. After subtract-ing the neutrino background, the remaining νµ eventscan then be combined as in the neutrino case to produceoscillated spectra at Super-K.

I. Cross Section Measurements

A unique feature of nuPRISM is the ability to measurethe true neutrino energy dependence of both CC and NCinteractions using nearly monoenergetic beams. Thesemeasurements are expected to significantly enhance thereach of oscillation experiments, since the energy depen-dence of signal and background processes must be un-derstood in order to place strong constraints on oscil-lation parameters. As explained in Section II F, addi-tional multinucleon processes, with a different energy de-pendences than the currently modeled CCQE and CC1πcross sections can affect the T2K oscillation analysis. Inthe current disappearance analysis, there are also sub-stantial uncertainties on NC1π+ and NC1π0 processes(for disappearance and appearance respectively). As aresult, future proposed experiments which use water asa target (e.g. Hyper-Kamiokande and CHIPS) will di-rectly benefit from the nuPRISM cross section program;other programs benefit less directly through a criticalvalidation of our assumptions of the energy dependenceof the cross section on oxygen. It is also not just longbaseline oscillation programs which benefit, as cross sec-tion processes at T2K’s flux peak are also relevant forproton decay searches and atmospheric neutrino oscilla-tion analyses. Finally, should T2K run an antineutrinobeam during nuPRISM operation, all arguments madeabove equally apply for antineutrino cross section mea-surements at nuPRISM.

One should also consider the study of neutrino inter-actions interesting in its own right as a particle/nucleartheory problem. As an example, MiniBooNE’s cross sec-tion measurements have received much attention fromthe nuclear theory community who predominantly studyelectron scattering data.

Some of the difficulties in improving our understandingof neutrino cross sections stems from the fact that we do

23

(GeV)νE0 0.5 1 1.5 2 2.5 3

ν)*

Eν

(EΦ

0

50

100

910×

sk_nom_numode_numu

Entries 8639860

Mean 1.762

RMS 0.6999

Fluxµν Mode νPRISM Anti-ν


°-2.0°=1.0OAθ

(GeV)νE0 0.5 1 1.5 2 2.5 3

ν)*

Eν

(EΦ

0

20

40

60

80

910×

sk_nom_numode_numu

Entries 8639860

Mean 1.63

RMS 0.7181



°-3.0°=2.0OAθ

(GeV)νE0 0.5 1 1.5 2 2.5 3

ν)*

Eν

(EΦ

0

10

20

30

40

50

910×sk_nom_numode_numu

Entries 8639860

Mean 1.454

RMS 0.7483



°-4.0°=3.0OAθ

FIG. 22. The nuPRISM anti-neutrino mode wrong-sign νµfluxes for 1.0− 2.0◦ (top), 2.0− 3.0◦ (middle) and 3.0− 4.0◦

(bottom), and the nuPRISM linear combinations of neutrinomode νµ fluxes.

Cor

rela

tion

-1

-0.5

0

0.5

1

, 0-5 GeVµν Mode, ν , 0-5 GeVµν Mode, ν

, 0-5

GeV

µν M

ode,

ν

, 0-5

GeV

µν M

ode,

ν

Right Sign/Wrong Sign Flux Correlations

FIG. 23. The correlations between the flux normalizationparameters for energy bins from 0 to 5 GeV for the neutrinomode and anti-neutrino mode νµ fluxes.

not know, for a given interaction, the incident neutrinoenergy. Any given measurement is always averaged overthe entire flux. The observed rate N in a given observablebin k depends on the convolution of the cross section, σ,and the flux, Φ:

Nk = εk

∫σ(Eν)Φ(Eν)dEν (9)

where ε is the efficiency. Therefore, our understand-ing of the energy dependence of neutrino interaction fora particular experiment is limited by the flux width andshape. One then attempts to use different neutrino fluxes(with different peak energies) to try to understand thecross section energy dependence. As discussed later inthis section, for CC interactions we have many examplesof disagreements between experiments, and for NC, wehave a limited number of measurements made, and thelack of information and conflicting information leaves un-resolved questions about the true energy dependence ofthe cross section.

In addition to providing new measurements on oxygen,there are two main advantages of nuPRISM over the cur-rent paradigm. First, we can directly infer the energy de-pendence of the cross section by combining measurementsat different off-axis angles into a single measurement, asif we would have had a Gaussian neutrino flux source.Second, and equally important, we can fully understandthe correlations between energy bins, in a way not possi-ble previously when comparing across experiments withentirely different flux setups.

In CC interactions, previous experiments use the muonand hadronic system to try to infer the neutrino energydependence. nuPRISM has the capability to directly testif the neutrino energy dependence inferred from the lep-

24

TABLE III. Expected number of events in the fiducial volumeof nuPRISM for 4.5×1020 POT, separated by true interactionmode in NEUT.

Int. mode 1-2◦ 2-3◦ 3-4◦

CC inclusive 1105454 490035 210408CCQE 505275 271299 128198CC1π+ 312997 111410 39942CC1π0 66344 23399 8495CC Coh 29258 12027 4857NC 1π0 86741 32958 12304NC 1π+ 31796 11938 4588NC Coh 18500 8353 3523

ton information is consistent with the energy informationdetermined from the off-axis angle. nuPRISM will alsofor the first time probe the energy dependence of NCcross sections within a single experiment.

Furthermore, there is no data for the kinematic in-formation of pions out of NCπ+ interactions. However,NCπ+ is one of the backgrounds in the current T2K1Rµ-like selection used for the disappearance analysis.A direct measurement of NCπ+, and a measurement ofthe pion momentum and angular distributions would re-duce the substantial uncertainties on this process (in bothcross section and detector efficiency) in the analysis.

Oxygen is an interesting target material for studyingcross sections because few measurements exist and it is amedium sized nucleus where the cross section is calcula-ble. nuPRISM will provide differential measurements inmuon and final state pion kinematic bins. While thesekinds of measurements will be done with the ND280 P0Dand FGD2 detectors in the near term, nuPRISM will havemore angular acceptance than those measurements andso enhances the T2K physics program.

Possible cross section measurements, based on observ-able final state topologies, at nuPRISM include:

• CC inclusive

• CC0π

• CC1π+, π0 (resonant and coherent)

• NC1π+, π0 (resonant and coherent)

• NC1γ

The above list is based on expected water Cherenkovdetector capabilities from experience with MiniBooNE,K2K 1 kton and Super-Kamiokande (SK) analyses. AllCC measurements can be done for νµ and νe flavors dueto the excellent e-µ separation at nuPRISM. Antineu-trino cross section measurements are also possible withsimilar selections. A brief summary of each measurementfollows. Table III shows the number of events in the FVof nuPRISM, broken down by interaction mode.

1. CC Inclusive

Inclusive measurements are valuable because they arethe most readily comparable to electron scattering mea-surements and theory, as there is minimal dependance onthe hadronic final state. Also, external CC inclusive neu-trino data was used in the estimation of the T2K neutrinooscillation analyses to help determine the CCDIS and CCmulti-π uncertainties.

The CC νµ cross section has been measured on car-bon by the T2K [22] and SciBooNE [23] experiments.MINERvA has produced ratios of the CC inclusive crosssection on different targets (C,Fe,Pb) to scintillator [24].In addition, the SciBooNE results include the energy de-pendence of the CC inclusive cross section from the muonkinematic information. The CC νe cross section on car-bon is in preparation by T2K.

nuPRISM should be able to select CC νµ and νe eventswith high efficiency and produce a CC inclusive measure-ment vs. true neutrino energy on water. Using the latestT2K simulation tools, we estimate a CC inclusive νµ (νe)selection to be 93.7% (50.4%) efficient relative to FCFVand 95.9% (39.5%) pure based on observable final state.The low purity of the νe selection is predominantly dueto the small νe flux relative to νµ.

2. CC0π

The CCQE νµ cross section has been measured oncarbon by MiniBooNE [25] and is consistent with alarger cross section than expected which could corre-spond to an increased value of an effective axial mass(MA) over expectation; SciBooNE’s analysis was pre-sented at NuInt2011 [26] but not published and is con-sistent with MiniBooNE. In addition, a measurement byNOMAD [27] was done at higher neutrino energies whichis not in agreement with MiniBooNE and SciBooNE.This is shown in Figure 24, along with the recent T2KND280 Tracker analysis results. An indirect measure-ment of the cross section was done with the K2K neardetectors, where a higher than expected value of the QEaxial mass, MA, was also reported [29]. There are alsorecent results from MINERvA [28].

MiniBooNE’s selection was CC0π, that is 1 muon andno pions in the final state, and was 77.0% pure and 26.6%efficient; the 1Rµ-like selection at SK is 91.7% pure and93.2% efficient, based on observable final state. It is pos-tulated that the MiniBooNE selection, but not the NO-MAD one, is sensitive to multinucleon processes, wherea neutrino interacts on a correlated pair of nucleons andthat this resulted in the higher cross section reportedby MiniBooNE. However, the two experiments have verydifferent flux, selection and background predictions andsystematics.

By measuring the CC0π cross section at different ver-tex points in nuPRISM, we should be able to infer thedifferent energy dependence and constrain multinucleon

25

FIG. 24. The CCQE cross section as predicted by NEUT(pink dashed) vs. true neutrino energy. Also overlaid areresults from MiniBooNE, NOMAD and T2K.

and CC1π+ pionless ∆ decay (PDD) processes. Thiscan be seen in Figure 5, which shows the momentum ofCCQE and MEC (Nieves’ npnh) events for a particularangular range (0.85 <cos(θ)< 0.90) generated accord-ing to the T2K flux, and for a 1 GeV nuPRISM flux.MiniBooNE and T2K have difficulty separating the MECcomponent of the CCQE cross section due to the shape oftheir neutrino energy spectra, but the nuPRISM detec-tor would give us additional information to separate outthat component and characterize it, as demonstrated inFigure 5. Even though nuPRISM is not a measurementon carbon, oxygen is of a similar density to carbon andso will be helpful in understanding the difference betweenthe MiniBooNE and NOMAD results if it is indeed dueto MEC.

3. CC1π+ and CC1π0

The CC1π+ and CC1π0 cross sections have been mea-sured on carbon by MiniBooNE [30],[31]; K2K also pro-duced measurements CC1π+ [32] and CC1π0 [33] withthe SciBar detector. One may infer the coherent contri-bution to the CC1π cross section from the angular dis-tribution of the pion; this was done by K2K [34] and Sci-BooNE. Improvements to the SK reconstruction couldyield a similar efficiency and purity to the the Mini-BooNE selections for CC1π+ (12.7%, 90.0%) and CC1π0

(6.4%, 57.0%) based on observable final state.

The CC1π resonant cross section for the T2K flux isdominated by contributions from the ∆ resonance [35],so nuPRISM would provide clear information about theN∆ coupling and form factors. We can also compare thepion momentum produced out of CC1π+ interactions fordifferent neutrino energies in order to better understandhow final state interactions affect pion kinematics.

4. NC1π+ and NC1π0

The NC1π0 cross section has been measured on car-bon by MiniBooNE [36] (36% efficient, 73% pure) andSciBooNE. A measurement of the ratio of NC1π0 to theCCQE cross section has been done water by the K2K1kton near detector [37]. The efficiency and purity ofthe K2K selection is 47% and 71% respectively. A mea-surement of NCπ+ exists [39] on a complicated targetmaterial (C3H8CF3Br) but has no differential kinematicinformation. Figure 25 shows this measurement with aprediction from the NUANCE neutrino event generator.

(GeV)E1 10 210

/ nu

cleo

n)2

cm

-38

) (10

+n

µ

p µ

(0

0.05

0.1

0.15

0.2

0.25

0.3Br3 CF8H

3GGM, NP B135, 45 (1978), C

=1.1 GeV)A

NUANCE (M

FIG. 25. The NCπ+ cross section as predicted by NUANCEvs. true neutrino energy overlaid with the only measurement(on C3H8CF3Br). Figure from Ref. [38]

A measurement of NCπ+ will be challenging but pos-sible at nuPRISM. T2K already has developed an “NC”enhanced selection for Super-K that is 24% NCπ+, 14%NC1proton, and 55% CCνµ, by interaction mode. Recentdevelopments in event reconstruction at Super-K includea dedicated pion ring finder, which should make possiblea more pure selection of NCπ+ from which the pion mo-mentum and angular distribution can also be measured.Since nuPRISM will allow for a first measurement of theenergy dependence of the NC channels and like the CCchannels, it will be particularly interesting to measure theoutgoing pion spectra of these events in order to probenuclear final state interactions.

To summarize, nuPRISM’s measurement of true neu-trino energy dependence of the cross section is a uniqueand potentially critical input to our overall understand-ing of cross section processes around 1 GeV neutrino en-ergy. In particular, nuPRISM will help us understandfor CC0π events, if the shape and size of the PDD andmulit-nucleon components are modeled correctly. Fur-thermore, nuPRISM can provide new information on thepion kinematics out of NC interactions relevant to theoscillation analysis and the energy dependence of thosecross sections.

26

III. DETECTOR DESIGN AND HARDWARE

The nuPRISM detector uses the same water Cherenkovdetection technology as Super-K with a cylindrical watervolume that is taller than Super-K (50-100m vs 41m) butwith a much smaller diameter (6-10m vs 39m). The keyrequirements are that the detector span the necessary off-axis range (1◦-4◦) and that the diameter is large enoughto contain the maximum required muon momentum. Thebaseline design considers a detector location that is 1 kmdownstream of the neutrino interaction target with amaximum contained muon momentum of 1 GeV/c. Thiscorresponds to a 50 m tall tank with a 6 m diameterinner detector (ID) and a 10 m diameter outer detector(OD), as shown in Figure 26. A larger, 8 m ID is alsobeing considered at the expense of some OD volume inthe downstream portion of the tank. As the nuPRISManalysis studies mature, the exact detector dimensionswill be refined to ensure sufficient muon momentum, νestatistics and purity, etc.

FIG. 26. The planned configuration of the nuPRISM detectorwithin the water tank is shown. The instrumented portion ofthe tank moves vertically to sample different off-axis angleregions.

The instrumented portion of the tank is a subset ofthe full height of the water volume, currently assumedto be 10 m for the ID and 14 m for the OD. The novelfeature of this detector is the ability to raise and lowerthe instrumented section of the tank in order to span thefull off-axis range in 6 steps. The inner detector will be

instrumented with either 5-inch or 8-inch PMTs to en-sure sufficient measurement granularity for the shorterlight propagation distances relative to Super-K. Also un-der consideration is to replace the OD reflectors withlarge SMRD-style scintillator panels, as discussed in Sec-tion III E, although this has not yet been integrated intothe overall detector design.

The remainder of this section describes the elementsneeded for nuPRISM and corresponding cost estimates,where available. The cost drivers for the experiment arethe civil construction and the cost of the PMTs, and, cor-respondingly, more detailed cost information is presentedin those sections.

A. Site Selection

The nuPRISM detector location is determined by sev-eral factors, such as signal statistics, accidental pile-uprates, cost of digging the pit, and potential sites available.At 2.5o off-axis position at 1 km with a fiducial volumesize of 4 m diameter and 8 m high cylinder, the neutrinoevent rate at nuPRISM is more than 300 times that ofSK. At 2km, the number of events drops by a factor of 4,which yields 75 times more events than SK, for the samesize of the detector. The impact of the number of eventscollected on the physics sensitivities is described in Sec-tion II. The event pile-up is dominated by sand muons,but at 1 km, the pile-up rate appears to be acceptable,which is explained in more detail in Section II, The de-tector depth and diameter scales with the distance to thenuPRISM detector. In order to cover from 1-4◦ off-axisangles, the depth of the detector is 50 m at 1 km and100 m at 2 km. There are standard Caisson-based exca-vation procedures available for pit depths of up to 65 mand diameters of up to 12 m. For deeper depth or largerdiameter, more specialized construction may be required,and could increase the cost per cubic meter of excavationdramatically, as discussed in the next section.

Potential sites for nuPRISM have been identified alongthe path from the neutrino beam to both the Mozumimine, where Super-K is located, and to the Tochiboramine, which is a candidate site for Hyper-Kamiokande,and is positioned at the same off-axis angle as Mozumi.No specific sites are discussed in this public version of thedocument. Land use will require consensus from the localcommunity and involvement from one or more Japanesehost institutions. There are existing facilities that areoperated just outside J-PARC, such as the KEK-Tokaidormitory, KEK Tokai #1 building at IQBRC, and thedormitory of the Material Science Institute of Tokyo uni-versity.

B. Civil Construction

Based on the current baseline design of the nuPRISMdetector described previous sections, we have communi-

27

cated with companies for the preliminary cost estimationof nuPRISM civil construction; the water tank construc-tion and detector construction. The nuPRISM detec-tor is also considered as a prototype detector of Hyper-Kamiokande (Hyper-K) for testing new photo-sensors,readout electronics, and the water containment systemdesign.

Two groups have been contacted to provide prelimi-nary cost estimates for the civil construction associatedwith fabricating a 50 m deep cylindrical volume with a10 m diameter. The first group consists of a generalconstruction company and a heavy industrial companycurrently providing cost estimates for Hyper-K. The sec-ond group is a single general construction company thatwas associated with the cost estimates from the originalT2K 2 km detector proposal [41].

There are several techniques to construct the 10 mφand 50 m long vertical “tunnel”; Pneumatic Caisson (PC)method, Soil Mixing Wall (SMW) method, New AustrianTunneling (NAT) method, Urban Ring (UR) method.Each of the construction methods have pros and cons,and some of the methods are not applicable dependingon the actual geological condition.

C. Liner and Tank

The nuPRISM detector can be used for proof-testingvarious designs and components which will be adoptedin the Hyper-K detector. The nuPRISM water tank willhave the same liner structure as that designed for Hyper-K.

The structure of the nuPRISM tank liner is shown inFigure 27. The innermost layer contacting with the tankwater must be a water-proofing component to seal thewater within the tank. We use High-Density Polyethy-lene (HDPE) sheets, which are commonly used as awater-proofing tank liner material. The sheets have ex-tremely low water permeability and also are resistant tolong-term damages from the ultra pure water. The ad-joining sheets are heat-welded, and the welded part alsokeeps the water-proof functionality.

We select the HDPE sheet with a number of studsprotruding from one side. These studs work for anchoringthe sheet firmly on the backside concrete layer. To buildthis ”HDPE on concrete” liner, a HDPE sheet is fastenedto the inside of a concrete form beforehand, then theconcrete is poured into the form for making the backfillconcrete layer. While the thickness of the HDPE liner is5-10mm, the thickness of the backfill concrete layer is yetto be determined.

Though we aim to construct the HDPE sheet liner suchthat the tank water can not leak, an additional water-proof layer is made between the backfill concrete layerand the shotcrete. This layer works as a catcher and aguide for the water by the unexpected leakage throughthe HDPE liner (and also the sump water through theshotcrete). This leaked water is drained via pits placed

!"#$%&"'()

1

*+,-%./(('.

012340"#5

61#7)('(%

&"'()89)11:%./(('.

0/1'7)('(

+)"2#%92'

FIG. 27. A schematic view of the nuPRISM tank liner.

under the water tank.

D. Detector Frame and Lifting Mechanism

This section describes a proposed design for the framethat supports the nuPRISM PMTs and defines both theinner and outer detector. We will also describe the sys-tem by which this frame can be moved up and down in or-der to be able to make the nuPRISM measurements. At-tention will be paid to the question of providing adequatewater flow through the nuPRISM frame while maintain-ing optical separation.

1. Detector Shape, Support and Positioning

Figure 28 shows a simple cylindrical design, the wallsof the Inner Detector (ID) being 0.5 meters thick. Thehalf circles represent the 20” PMTs (0.5m) facing out-ward for the veto region (OD). The smaller half circlesrepresent 8” PMTs (0.2m) facing inwards to the ID re-gion. The 0.5m thickness of the detector wall is to containthe bodies of the PMTs (and PMT electronis) and, withinternal stiffening braces, be stiff enough to accuratelyposition the PMTs and not deform significantly underthe weights and buoyancies.

Figure 28 also shows a conceptual support and posi-tioning system. The detector is positioned on four ver-tical rails fixed to the shaft walls, and supported on topand bottom rings. Struts connect the detector to thesetwo rings. The struts are positioned at the corners of thedetector where the structure is strongest, and angled sothat the distance from the detector to the start of thereflector is 1.7m top and bottom, and 1.5m on the sides.The reflector encloses the OD region and is required tobe optically isolated from the ID volume, and from theshaft water volumes above and below. We discuss thereflector in more detail below.

28

FIG. 28. The Detector is positioned on four rails inside shaft,and supported on top and bottom rings. Struts connect thedetector to its rings. Four vertical cables support the assem-bly. Ballast can be added to the rings, if required. Distancesare in meters.

2. Water Flow and Optical Isolations

Figure 29 shows views of the top-left corner of the de-tector and reflector. Two section views indicate the con-ceptual features and functions involved. The volume ofthe ID and OD are ≈264m3 and ≈790m3, respectively,with a combined volume of ≈1,190m3 (including all wallvolumes). If the apparatus is to traverse the shaft lim-its in ≈24 hours, the speed would be ≈1.5 meters/hour.Since the reflector side walls are close to the shaft, thedisplaced water needs to flow through the reflectors. Thisspeed corresponds to a water flow of ≈118 m3/hour = 2.0m3/min. Even if the water could flow past the sides ofthe reflector enclosure, 1,190 tons of water would alsobe in motion, which would be difficult to accommodate.With no water flowing through the sides of the reflec-tor enclosure, the sides can be simple metal panels witha white inner surface to enhance the OD light collec-tion. As indicated in Figure 29, these vertical reflectorwalls need to notch around the four rails and the asso-ciated couplings on the rings, and would be screwed tothe top/bottom rings. With a height of 13.8m and cir-cumference of 33.5m, it will need to be segmented withoverlapping joints (or added joint strips). When the de-tector is out of the water, it would be useful to be able toeasily remove the side reflector segments. Minimal seg-mentation would be four, with joints at the center of the

‘notches’. This would allow the segments to slid out pastthe rails and the support towers. The top and bottomreflectors are also bolted to the top/bottom rings, butthey have to be thicker to allow them to be strong andstiff due to the quantity of water flowing through them.The stiffness is achieved by making the top/bottom re-flectors 0.2m thick and them having an internal bracingstructure. The top/bottom reflectors need an optical sealto the rest of the shaft, yet allow ≈2.0 tons/min of waterto flow through. Figure 29, shows two possible solutions:

FIG. 29. This shows views of the top-left corner of the detec-tor and reflector. Two sections views indicate the conceptualfeatures and functions involved. A system of offset black pipes(or flaps) would allow water to flow through.

1. The first is a system of offset black pipes, so thatwater can flow through, but any light would needat least two reflections off black surfaces. The in-ner surface of the top/bottom reflectors would bewhite to enhance light collection. The tubes at theinner surface would have white ‘tube covers’. Thisis easily done by having the tube extend, with thetube cover fixed to it, but the tube having fourside large slots, leaving webs of material to holdthe cover. The cover and outer surface of tube ex-tension would be white. To prevent water beingtrapped in the reflector wall when the detector islifted out of the water, there would be ‘Drain holes’in the tubes, just inside the inner wall. Alternately,there could be some small drain tubes+covers ex-tending slightly into OD volume. This scheme isa little complicated, but has the advantage of nomoving parts. If the flow is fully distributed overthe 55.0 m2 area, the movement water flow wouldthen be ≈36 liters/minute/m2.

2. Another way solve this problem would to have asystem of flaps that open only when the detectoris moved, and close automatically when it stops.Half the flaps open when the detector moves down

29

(water moves up), these flaps close under their ownweight when movement stops. The other flaps openwhen the detector moves up (water down), these‘down’ flaps would need to be spring loaded orcounterweighted to close when movement stops. InFigure 29, we show both these flaps in the openposition. The inner surfaces of the flaps would bewhite. In this scheme most of the water would drainthrough the spring loaded ‘down flaps’, but wouldalso require a system to small holes or pipes to drainout the last of the water. This system has manymoving parts that cannot be lubricated, so bindingand galling would be concerns, but it can probablybe made to work. It would need to be made veryreliable, a few flaps stuck closed wouldn’t be a con-cern, but some stuck open could be a problem. Thissystem has the disadvantage that it prevents lowerlevels of circulating water during data taking. Thisrecirculating loop will probably be required for; thepurification and temperature control of the water,cooling of electronics etc. For these reasons, weprefer the offset tubes option.

When the detector is out of the water, the bottomreflector would need to be segmented to be removed be-tween the support towers. With four towers (see Fig-ure 30), the four bottom cover segments would be 4.6x4.6meters. Higher segmentation (multiples of 4) would alsobe possible. We imagine a scissor cart rolled under thedetector, lifted to contact a segment. It could then beunbolted, lowered and rolled away. The segments wouldneed to overlap on the inner surface for light seal, and onthe outer surface for joining (or have extra joint strips).The top reflector would be craned out, in one piece or insegments.

3. Walls of Inner Detector (ID)

The top/bottom walls of the ID would also need to al-low water flow, otherwise one would have to allow for theinertia of 400 tons of trapped water. The movement flowswould be 42 tons/hour = 0.7 tons/minute. Distributed,this is 27 liters/minute/m2. This is somewhat less thanthe 36 liters/minute/m2 of the reflector, but this wall hasall the PMTs as well. In Figure 29, I show the tubes andflaps options for this wall, similar to that for the top andbottom reflectors.

4. Detector in the shaft

The detector is guided within the shaft by a set of rails.The current proposal has four rails and support cablesbut it could be three, five, etc. if dictated by other de-sign considerations. It is important to understand thatthe ring connections to the rails do not need to be highprecision rail bearings. Because the positioning accuracy

required is only ≈1cm, they could be simple guides (seeFigure 30). Similarly, the rails do not need to be com-plex. The loose tolerance makes it far less likely that thedetector will jam on the rails. When the detector hasbeen moved, there may be a system to lock two of thefour guide locations to eliminate small position changesduring data taking. Another reason for a looser coupling(before locking), is that then the rails do not need tobe so precisely positioned on the shaft walls, i.e. severalmillimeters versus 0.1mm.

Figure 30 shows the detector in the shaft, the shaftcovers and the external towers. Four vertical cables sup-port the assembly. Ballast can be added to the rings,if required. Above ground, there would be four towersextending upwards 17.6 meters. Four motors, acting to-gether, lift or lower the detector in the shaft, or even liftit completely out of the water. The load will increaseas it leaves the water (loss of buoyancy), if the load istoo much, the top ballast can be removed by crane as itclears the water. Or, a lifting frame could be attachedwhen the top ring clears the water, allowing the crane toraise it further, then it can be locked in the out position,freeing the crane.

In this concept, the signal and power cables for thedetector would travel up out of the water beside the foursupport cables. They would nominally go up and over thetowers, then down to the ground racks. With this schemethere would be no extra length in the water, wherever thedetector was positioned in the shaft. When the detectoris slowly lowered further down the shaft, the cables etc.should be cleaned before entering the water.

Once the detector is entirely out of the water, the shaftcovers can be craned back into position (see Figure 31).Adding counterweights will make sure the Center-of-Gravity (COG) of the covers are beyond the detectorshadow when the covers are pushed in. The covers wouldbe bolted to the ground. Lightweight seals cover thejoints, the central region, and the four small areas wherethe support cables, signal and power cables exit the wa-ter.

Figure 32 shows the detector out of the water and cov-ers reinstalled. It is important that the covers and sealsare safe for people and light equipment, so that the bot-tom of the detector can be worked on. Scaffolding canbe erected to work on all parts of the detector. The fig-ure also shows the detector moved to a stand. To movethe detector, the lifting frame would be installed, the de-tector supported, then the eight ring guides removed andtwo of the towers removed (or laid down), opening a pathfor the detector move.

Whether above the shaft or on a separate stand, itwould probably be useful to be able to remove the reflec-tor sections and get access to parts of the ID. If the IDwere bolted together sections, it might be possible to par-tially disassemble to make repairs and/or replacements.

30

FIG. 30. This shows the detector in the shaft, the shaft coversand the external towers. Four vertical cables support theassembly. Above ground, there would be four towers upwardsextending 17.6 meters. Four motors, acting together, lift orlower the detector in the shaft, or even lift it completely outof the water.

FIG. 31. The four covers can be craned in and out. Addedcounterweights make sure the center of gravity is beyondthe detector shadow when in. The covers are bolted to theground. Light weight seals cover the joints and the centralregion.

5. Detector Surveying

As mentioned earlier, after the detector has beenmoved, there may be a system to lock two of the fourguide locations to eliminate small position changes dur-ing data taking. A laser surveying system could be inplace to look down through the water to periodicallycheck the detector position at the four rail locations. ThePMTs may have to be turned off during these times. Thepositioning of PMTs within the detector would be sur-veyed during its assembly (out of water) and then shouldonly be subject to thermal expansion/contraction shiftsin the water, plus deflections due to loads (primarily thetop/bottom PMTs.

The thermal expansion/contractions of the detectorwill depend on its material. For a 10 meter Aluminum

31

FIG. 32. The four towers allow the detector to be raised outof the water (units in meters). The covers can be reinstalledunder the detector, allowing people to work underneath it. Alifting frame can be craned over the detector, attached, andthe towers removed. The detector can then be craned to astand.

piece, the expansion would be 2.2mm for a 10 OCchange. 306 stainless steel would be 1.6mm. The shaft,being reinforced concrete, should expand 1.3mm for 10degrees. Stainless steel is a better thermal match, thedifferential expansion being 0.3mm for 10 degrees, com-pared to 0.9mm for an Aluminum detector frame, butthe difference is not likely to be significant.

E. Scintillator panels

The veto system of the nuPRISM detector can be com-posed of plastic scintillator detectors which completelysurround the the water Cherenkov detector. The mainpurpose of the veto system is to identify backgroundsfrom beam neutrino interactions in the surrounding pitwalls and to provide a cosmic trigger signal for calibra-tion purposes. The technology developed for the ND280SMRD detector can be applied for this veto system.

1. Scintillator counters with WLS/avalanche photodiodereadout

Scintillator counters with wavelength-shifting (WLS)fibers and opto-electronic readout are an establishedtechnology for neutrino detectors in long-baseline neu-trino oscillation experiments. ND280 consists of severalsubdetectors which use extruded plastic scintillators ofvarious shape and dimensions [11]. Each of these subde-tectors is comprised of plastic slabs and bars, wavelengthshifting fibers and compact photosensors - multi-pixelavalanche photodiodes. The Kuraray double-clad Y11WLS fibers are used in all ND280 scintillator detectorsfor transportation of the reemitted light to photosensors.

SMRD counter. The SMRD detector was made ofthe polystyrene-based scintillator slabs, each with anembedded wave-length shifting fiber. The slabs wereproduced at the Uniplast Factory (Vladimir, Russia).The scintillator composition is a polystyrene doped with1.5% of paraterphenyl (PTP) and 0.01% of POPOP. Theslabs were covered by a chemical reflector by etchingthe scintillator surface in a chemical agent that resultsin the formation of a white micropore deposit over apolystyrene[18]. The chemical coating is an excellent re-flector, besides it dissolves rough surface acquired duringthe cutting process. The WLS fiber was read out onboth ends to increase light yield, improve uniformity andposition accuracy, and provide redundancy.

A key feature of these counters is the usage of the oneserpentine-shaped WLS fiber for readout of scintillatingsignal. The serpentine geometry of a groove consists of 15half-circles, each with a diameter of 58 mm and straightsections connecting the semi-circles. A 1 mm diameterY11 (150) Kuraray WLS fibers of flexible S-type and withdouble-cladding was used for the SMRD counters. Fibersare bent into a serpentine-shape and glued into grooveswith BC600 Bicron glue. The mean light yield for sum ofboth ends was about 40 p.e./MIP after subtraction of theMPPC cross-talk and after pulses. The high light yieldallowed us to obtain the efficiency of more than 99.9%for detection of minimum ionizing particles.

The light yield of about 14 p.e. per a minimum ioniz-ing particle (∼ 7 p.e./MeV for 1 cm thick bar) providesthe efficiency for detection of minimum ionizing particlesof more than 99% in an individual scintillator bar for adetection threshold of 1.5 p.e. Time resolution dependson the light yield as ∼ 1/

√Np.e. where Np.e.− is the

number of photoelectrons. For the l.y. of 20 p.e. thetypical resolution is obtained to be σ 1 ns. Detectorswith shorter WLS fibers were also tested. Light yield ofthe detector with a 5 m long WLS Y11 fiber is shown inFig. 33

In this case, the minimum light yield of more that 40p.e./MIP (sum of both ends) is obtained.

32

Light yield (p.e./MIP)

0

20

40

60

80

100

Distance (m)

0 1 2 3 4 5

5 m long Y11 fiber, bar 0.7x2x90 cm, T=20.5-21.5oC

left MPPCrigt MPPC left+right

FIG. 33. Light yield of a scintillator counter with 5 m longWLS fiber vs position along the fiber. The T2K 667 pixelMPPC’s were used in this measurement.

2. Veto counters for nuPRISM

i

The excellent performance of the SMRD counters withone serpentine WLS fiber per counter gives a possibil-ity to make a veto system using similar approach. Oneoption is to construct the nuPRISM veto system fromscintillator counters, each of 0.2 m2. One WLS Y11 S-type fiber is embedded in the extruded plastic slab of2000× 200× 7mm3. Half-circles have the radius of 3 cmthat allows to keep the performace of the fiber withoutloosing the transmission of the reemitted light along thefiber. A 6 m long Y11 fiber is readout on both ends byMPPC’s. Taking into account the improved parametersof new MPPC’s, for exmple, higher PDE, we can expectto obtain miminum light yield of 20-30 p.e./MIP andtime resolution of about 1 ns for these detectors. Moreaccurate information can be obtained after tests of theconter prototypes.

F. Photomultiplier Tubes

The original T2K 2 km detector proposal used 8”PMTs to better match the granularity of the 20” PMTsused in the much-larger Super-K detector. The baselinedesign for the nuPRISM detector is only 6 m in diame-ter and 10 m tall, which corresponds to 3,120 PMTs for40% photocathode coverage. This is significantly smallerthan the 11,129 PMTs used at Super-K, so to improvethe granularity of the detector, 5” PMTs are also be-ing investigated, of which 7,385 PMTs would be requiredfor 40% coverage. Additional options such as avalanchephotodiodes and high quantum efficiency coating are alsobeing explored.

G. Electronics

Part of the goal of the nuPRISM is to serve as a pro-totype for the Hyper-K. We therefore want nuPRISM touse a set of electronics that is as close as possible to theelectronics being proposed for Hyper-K. Some of the keyfeatures of the Hyper-K electronics are the following:

• Front-end electronics will be placed in the water,as close as possible to the PMTs.

• Front-end electronics are expected to find all hitsabove 0.25 PE and send all information about hitsup to back-end electronics. In back-end computerstrigger decisions will be made using software. Noglobal triggers will be propagated to the front-endelectronics.

• PMT digitization should provide 0.05 PE chargeresolution, 0.25 ns timing resolution (for 1PE hits)and 0.1-1250 PE dynamic range.

We shall note various aspects of the nuPRISM elec-tronics where we may differ from the default HK elec-tronics plan. In particular, one clearly different aspectof nuPRISM will be the much higher rate of ‘pile-up’events during beam spills. The rate of sand muon eventsentering the ID may be as high as 0.19 per bunch. Atminimum we therefore need electronics that can cleanlydistinguish between PMT hits in different bunches; ie,hits with separation of order ≈ 600ns. We may also wantto have some capacity to distinguish between hits withina single bunch; ie hits that differ by 10s of ns. This wouldbe a more challenging requirement.

1. FADC Digitization

Given this requirement for inter-bunch and intra-bunch hit resolution we propose using FADC digitiza-tion with basic digital signal processing in the front-endelectronics. The basic scheme is as follows:

1. The stretched/shaped PMT signal is fed into theFADC. Use a standard commercial FADC, withsampling frequency between 80-500 MHz and 12-16 bit resolution.

2. The digital output of FADC is fed into an FPGA(on the front-end electronic card), where we do ba-sic digital pulse processing (on the fly, at same rateas original digitization). Digital pulse processingwould involve the following:

• Finding PMT hits (for instance, by using sim-ple threshold comparison).

• Calculating the pulse time and charge.

3. The digital pulse information is then transferred tothe back-end electronics. We send different typesof data depending on the pulse charge.

33

It is worth emphasizing that the expected timing res-olution using FADCs is not intrinsically limited by thesampling. For instance, if you appropriately stretch andshape a PMT pulse you can easily achieve 0.25 ns timingresolution using a 100 MHz digitizer (ie a sample each 10ns), as long as you have high signal to noise ratio and areasonable number of ADC samples on the leading andfalling edges. We will explore the trade-offs involved inoptimizing the performance of such a system in SectionIII G 3.

2. Signal Conditioning And PMT HV

We propose to use differential transmission in orderto deliver signals from the PMT bases to the digitiza-tion board. An advantage of such a solution is that, inprinciple, it would allow us to use a standard unshieldedtwisted pair cable, while still maintaining fairly good im-munity to pickup of electromagnetic interference. Thebase of the PMT would contain shaping circuitry, whichwould stretch PMT signals, limiting their bandwidth tomatch FADC requirements and converting them into asymmetric form, suitable for transmission via a twistedpair cable. Preliminary studies show that signal shap-ing using a 5-th order Bessel-type low pass filter shouldprovide satisfactory results.

One of the design goals for the nuPRISM is minimiza-tion of the amount of necessary cables. As such, it wouldbe advisable to use a single cable to provide both highvoltage to the PMT and to transmit the signal fromthe PMT base to the digitization board. Therefore, thepreferable solution would be to synthesize the high volt-age directly on the PMT base, from a 48-200 V DC sup-ply, using either a commercial high voltage module or acustom designed voltage multiplier structure. This way,power to the PMT base could be delivered via an addi-tional twisted pair of the same cable that would be usedto transmit the shaped PMT signal. The slow controllink necessary to tune the high voltage for specific PMTcould be realized via a DC power line, thus avoiding theneed to use additional cables. In any case, it should beemphasized that the details of the PMT HV implemen-tation will depend strongly on the exact PMTs that arechosen.

3. Digitization Performance/Optimization

There is a strong inter-dependance of the digitizationperformance on the signal conditioning, type and param-eter of the chosen analog-to-digital converter (ADC) andthe applied signal processing algorithms. The key pa-rameters here are the speed and accuracy of the ADC aswell as signal to noise ratio (SNR) of the whole system.Cost-wise, it would be best to use as slow and as leastaccurate ADC as possible while still meeting the per-formance requirements. Therefore, a Monte-Carlo study

has been performed in order to estimate impact of theelectronics chain on the overall system performance.

Simulation setup is presented in Fig. 34. The photo-multiplier has been simulated as a current source (iPMT ),connected in parallel with a base resistor RB and a capac-itor CB , which together form the first pole of the shapinglow-pass filter. Both the RB and the CB were chosen tofulfil the dynamic range requirement while maintainingthe best possible signal to noise ratio, i.e. to provide thehighest possible PMT signal for maximum pulse charge(2000 p.e.) without saturating the amplifiers. The PMTcurrent pulse waveform was approximated using a trape-zoid pulse, with timing parameters (trise, tFWHM ) corre-sponding to manufacturer specification given in the PMTdatasheet. The rise and fall times were assumed to beequal. Given the time constants of the shaper, the PMTpulse can be treated as a delta function.

The output of the shaper’s response simulated in theSPICE program was then sampled, quantized and subse-quently analyzed using a digital Constant Fraction Dis-crimator (CFD), modeled in MATLAB. Using the resultof the CFD, the difference between the calculated timeand the real time was calculated, as well as the differenceof calculated pulse charge and the real pulse charge. Asummary of the results is presented in Figures 36 and35. As can be seen, there is some difficulty in achiev-ing the desired timing and charge resolution for the 1p.e. pulses, which is due to poor signal to noise ratio.In particular, even with a 16-bit, 250MHz sampling wecan only achieve approx. 0.8 ns timing resolution for thesingle p.e. (compared to the desired 0.25ns resolution).

As such, further studies are ongoing in order to find aworking solution. The considered options include split-ting the signal from the PMTs to separate high and lowgain branches which would then be digitized by their ownADCs. Other possibilities include dropping the linearityrequirement for the PMT response to large number ofphotons and running it at higher gain. A significant effortis also foreseen to optimize signal processing algorithmsfor poor SNR conditions, in particular an adoption ofmatched filtering approach is planned.

H. Water System

Starting with the very first large-scale WaterCherenkov detector – the Irvine Michigan Brookhaven[IMB] proton decay experiment, which began taking datain the early 1980’s – exceptional water clarity has been ofkey importance for massive devices of this kind. Thereis little benefit in making a very large detector unlessthe target mass contained within the detector can beefficiently observed. Good water quality has two mainadvantages: the light generated by physics interactionsin the water can propagate long distances with minimalattenuation until it is collected by photomultiplier tubesor other technologies, aiding accurate energy reconstruc-tion, and the light can traverse these distances (10’s of

34

FIG. 34. Simulation setup for the study of the FADC performance.

FIG. 35. Estimated timing resolution for FADC digitization,as a function of sampling frequency and ADC precision. Thetop plot is for 1PE pulses; bottom is for 10PE pulses.

FIG. 36. Estimated Charge resolution for FADC digitization,as a function of sampling frequency and ADC precision. Thetop plot is for 1PE pulses; bottom is for 10PE pulses.

35

meters) with minimal scattering, which aids in the pre-cise reconstruction of event vertices.

The strategy employed to create kilotons of extremelyclear water has been to remove all suspended solids, dis-solved gases, ions, and biologics from solution via a se-ries of filtration elements. These include microfiltrationfilters, degasifiers (vacuum and/or membrane type), re-verse osmosis membranes [RO], de-ionization resins [DI],and exposure to intense ultraviolet light [UV].

These water systems typically run in one of two modes:fill or recirculation. During the fill mode, water suppliedby the local municipality or ground water in the vicinityof the experiment is first brought up to ultrapure lev-els and then injected into the detector. The capacity ofthe water system, along with availability of water, defineshow long it will take to fill the detector. During recircula-tion mode, already high-quality water from the detectoris continuously passed through the filtration system andreturned to the detector after being cleaned even further.This is necessary as transparency-impairing materials aresteadily leaching into the chemically active ultrapure wa-ter. In addition, during the process of filtration the wateris typically chilled to further impede biological growth,with the added benefit of simultaneously reducing PMTdark noise which is typically strongly temperature de-pendent.

In the current baseline design, nuPRISM will have in-terior dimensions ten times smaller than Super-K. It istherefore possible that a commensurately less powerfulwater filtration system would be able to provide suffi-cient water transparency. Nevertheless, for now we willbase our initial system design and flow rates on water sys-tems known to have worked and produced useful physicsin the past.

Following this approach, a baseline design and costestimate for the nuPRISM water system has been pre-pared. The primary components described above arerepresented graphically in Figure 37. This system will becapable of filling the detector at a rate of 6.3 tons/hour,such that a complete fill can be completed in one monthof operations. It will be capable of recirculating the waterat a rate of 6.3 tons/hour through the entire system plusan additional 22.8 tons/hour through what is known asa secondary ”fast recirculation” path which trades somefiltration components for faster overall flow. The com-bination of complete cleaning and fast recirculation hasbeen shown at previous experiments (including the K2Kone kiloton near detector) to be the most cost-effectiveway of achieving the desired water transparencies. A pre-liminary cost estimate for this baseline water system fromSouth Coast Water in the is $350,000, including shipping,duties, and installation at the detector site.

1. Gd option

If it is decided to add 0.2% gadolinium sulfate by massto Super-Kamiokande in order to provide efficient tag-

Industrial water input during fill, water from

nuPRISM tank during recirculation

Pre-treatment and RO

To nuPRISM

@ 6.3 tons/hr

Uranium Removal,

DI and

UV

Chiller and Degas

FIG. 37. A preliminary baseline design of the nuPRISM watersystem.

ging of neutrons in water, it will likely be useful for anear detector at Tokai to also be Gd-loaded such thatthe responses of both detectors are as similar as possible.As a large water Cherenkov detector, nuPRISM is a nat-ural candidate for eventual Gd-loading. Therefore, theimplications this has on the water system design must betaken into account.

Over the past decade there have been focused R&Dprograms both in the US and Japan aimed at devisinga method capable of maintaining the exceptional watertransparency discussed above, while at the same timemaintaining the desired level of dissolved gadolinium insolution. In other words, somehow the water must becontinuously recirculated and cleaned of everything ex-cept gadolinium sulfate.

Starting in 2007 with a 0.2 ton/hour prototype at theUniversity of California, Irvine, since 2009 the Kamioka-based EGADS (Evaluating Gadolinium’s Action on De-tector Systems) project has shown that such a selec-tive water filtration technology – known as a ”molec-ular band-pass filter” and schematically shown in Fig-ure 38 – is feasible at 3 tons/hour. As the EGADSdesign is modular and uses off-the-shelf and readilyavailable equipment, albeit in novel ways, scaling itup from the current 3 tons/hour to 60 tons/hour forSuper-Kamiokande, is straightforward, while scaling tonuPRISM’s 6.3 tons/hours would be trivial.

36

Molecular Band-Pass Filter

Ultrafilter Nanofilter

Reverse

Osmosis

Larger and smaller

impurities to drain

(UF Flush + RO Reject)

Pure water

(RO product)

plus Gd2(SO4)3

Pure water

plus Gd2(SO4)3

Gd2(SO4)3

(NF Reject) Gd2(SO4)3

plus smaller impurities

(UF Product)

Impurities smaller than Gd2(SO4)3

(NF Product)

Impurities larger

than Gd2(SO4)3

(UF Reject

flushed

periodically )

FIG. 38. A schematic illustration of the principle of the”molecular band-pass filter”. Successively fine filter elementsisolate the dissolved gadolinium sulfate ions and return themto the main tank, bypassing water system elements whichwould be fouled if they were to trap gadolinium.

37

IV. DETECTOR CALIBRATION

The calibration systems for nuPRISM will largely bor-row from the existing Super-K calibration systems. How-ever, nuPRISM will also face some unique challenges:

• The PMT frame will move within the water volume.

• Accessing the inner detector is more difficult whenthe position of the top of the detector is not fixed.

To address these issues, nuPRISM will consist of calibra-tion sources that are fixed within the ID (e.g. laser balls,LEDs, and scintillation cubes), as well as sources thatcan be lowered though remote-controlled access portals(e.g. radioactive sources). It is expected that each timethe detector is moved, all of the PMTs will need to berecalibrated. This can be accomplished using the fixedlight sources within the ID, and additional calibrationruns with radioactive sources will be taken for each newdetector position.

In addition to the detector response, it will also be nec-essary to precisely determine both the relative position ofthe PMTs within the ID, as well as the absolute positionof the PMT frame within the water volume. This will beaccomplished with a laser calibration system. An R&Dprogram is planned to demonstrate the effectiveness ofsuch a system when operated in water.

As nuPRISM will essentially reuse many of the estab-lished Super-K calibration techniques, the remainder ofthis section will provide a brief description of Super-Kcalibration systems. Further details can be found else-where [15, 16].

A. Overview of Super-K Calibration Systems

This section overviews Super-K detector calibrations.For further details, reader can also refer to [15, 16].

The Super-K detector calibration can be divided intotwo steps; the detector hardware calibrations and the cal-ibrations for physics analyses. The first step is commonover all physics analyses, but the second step is designedfor each physics analysis goal.

1. Detector hardware calibrations

The detector hardware calibrations (measurements)consist of several parts:

• Geometrical surveys: tank geometry, PMT posi-tions

• Geomagnetic field

• PMT calibration: gain, photo-detection efficiency

• Readout channel (PMT and electronics) calibra-tions: linearity, timing, timing resolution

• Optical properties: water, PMT glass, black sheet,etc (for detector MC tuning)

• Water temperature

All of these calibrations and measurements are indispens-able to understand the detector and to model the detec-tor in the simulation. This section focuses on the PMTcalibrations and readout channel calibrations, which willbe most relevant to nuPRISM.

The PMT calibration procedure can be divided intothree large steps; 1) pre-calibration, 2) post-installationcalibration, 3) detector monitoring. At the stage of ‘pre-calibration’, a fraction of all Super-K PMTs have beencalibrated prior to the installation, e.g. a tuning of PMTgain. The pre-calibrated PMT, called standard PMTs,were used to calibrate all other PMTs in-situ after in-stalled, at the stage of post-installation calibration. Onceall PMT are calibrated, the stability of the PMTs is mon-itored continuously for the lifetime of the experiment.The following sections discuss our ideas for each of thePMT calibration steps.a. Pre-calibration SK has 420 standard PMTs,

which corresponds to about 4% of all SK PMTs. The SKstandard PMTs were calibrated prior to the installationby adjusting HV values to have identical charge (∼ 30p.e.) over the standard PMTs. For the pre-calibration,SK employed a xenon lamp and scintillator ball. Fig-ure 39 shows a schematic diagram of the pre-calibrationset-up.

less than 0.011. This small temperature difference clearly indicatesfull convection of water in the SK tank and thus indicates the bestuniformity in optical properties of water in the ID. This watercondition period was used for measurement of relative differencesof quantum efficiency in each PMT as described in Section 3.1.5.

3. Inner detector calibration

3.1. PMT and electronics calibrations

3.1.1. IntroductionTo provide background for this section, a brief description of PMT

calibration is presented here. The 20-in. diameter PMTs developed byHamamatsu Photonics K.K. (R3600-05(A)) [10] are used in the innerdetector. These PMTs have a photo-cathode made of bialkali (Sb-K-Cs), and has its maximal photon conversion probability in thewavelength range of Cherenkov light. The PMT dynodes are of aVenetian-blind type, and their base circuit is an optimized 11-stagevoltage divider. The high-voltage system for the PMTs was manu-factured by CAEN Co. and consists of distributors (A933K), controllers(SY527), and interface modules (V288).

Since the timing behavior of PMTs depends on the charge of themeasured pulse, we begin discussing ID-PMT calibrations withcharge-related issues. In the definition for the PMT charge calibra-tion, “gain” is a conversion factor from the number of photoelec-trons to charge (in units of pC), and “QE” is the product of thequantum efficiency and collection efficiency of photoelectronsonto the first dynode of the PMT. Low-energy physics events likesolar neutrinos largely consist of single-photoelectron (single-pe)

hits and rely heavily on the QE calibration for their interpretation,whereas high-energy events like those involving TeV-scale muonsdepend more on proper gain calibration. Knowledge of both gainand QE is important and must be available on a PMT-by-PMT basis.

Unfortunately, the old ATMs used in SK-I, II, and III did notallow us to record meaningful single-pe distributions on a PMT-by-PMT basis, however, a cumulative distribution for all PMTscould be obtained after the relative gains had been properlycalibrated.

This situation forces us to set up PMT calibration in the followingway. First, we need to determine a suitable high-voltage value to beapplied to each ID-PMT. This determination is described in Section3.1.2. Next we need to understand the differences in gain betweenindividual ID-PMTs. Section 3.1.3 details this effort and its results. Oncewe are able to obtain meaningful cumulative single-pe distribution forall ID-PMTs, Section 3.1.4 describes how to use this cumulative single-pe distribution to calibrate the average gain over all ID-PMTs.Referencing, in turn, the gain variation for an individual PMT to theaverage gain gives the individual gain of each ID-PMT. In Section 3.1.5we use Monte Carlo simulations to extract a calibration of the QE foreach individual PMT. This new procedure which determines the gainand QE of an ID-PMT's independently is a major improvement overthe procedure used previously. Section 3.1.6 describes the validation ofboth the gain and QE calibrations, including verifications of theirconsistency. Discussion of charge-related calibration issues is con-cluded in Section 3.1.7, which describes measurements for assessingthe linearity of charge determinations. Section 3.1.8 addresses the ID-PMT timing calibration.

These calibrations, except for the establishment of 420 refer-ence PMTs, were performed in the beginning of SK-I, II and III. Inaddition, a real-time calibration system monitors crucial para-meters throughout normal operations of the experiment to allowus to consider variability as well as ensure stability during data-taking. For this purpose, light sources are permanently deployednear the center of the ID. During SK data-taking, the lights flash inturn at approximately 1-s intervals. As detailed in Sections 3.1.2,3.1.8, and 3.2.1, they monitor ID-PMT gains and timing as well asoptical parameters of ID water.

3.1.2. Determination of the high-voltage setting for each PMTTo establish the high-voltage (HV) setting for each PMT, we

require that all PMTs give the same output charge for the sameincident light intensity. For this purpose, an isotropic light source isplaced at the center of the SK tank. Since the SK tank is a largecylinder about 40 m in both diameter and height, we expect theamount of light reaching each PMT from that source to be about afactor of two different between the closest and farthest PMT.Correcting for only this geometrical difference is insufficient, because

z (m)-15 -10 -5 0 5 10 15

Tem

pera

ture

(deg

rees

)

13.1

13.15

13.2

13.25

13.3

Fig. 4. The vertical dependence of the water temperature in the ID.

Fig. 5. Schematic view of the setup for the pre-calibration. A Xe flash lamp, placed inside a box, emitted light that was guided by optical fibers through a fiber bundle to twoavalanche photodiodes and a scintillator ball located in another μ-metal shielded dark box, where a 20 in. PMT was exposed to the light from the scintillator ball. Two 2-in.PMTs monitor the light output of the scintillator ball.

K. Abe et al. / Nuclear Instruments and Methods in Physics Research A 737 (2014) 253–272 257

FIG. 39. SK pre-calibration set-up. (Figure quoted from [16])

The SK standard PMTs were installed in the tank in ageometrically symmetric configuration. Figure 40 showsthe location of the standard PMTs in SK inner detector.

b. Post-installation calibrations In the post-installation calibration, all PMTs other than thestandard PMTs were calibrated in-situ after installed.At this stage, all PMT parameters were determined andmeasured. We will discuss the following items in thissection,

• HV (gain) tuningTune HV for all PMTs, referencing to the stan-dard PMTs by using the Xe lamp and deploying ascintillator ball in the tank (the same light sourceused in the pre-calibration). Move the scintilla-tor ball along Z-axis (height direction), and tune

38

The second measurement uses low-intensity flashes in which onlya few PMTs are hit in each event, therefore, we can be reasonablysure that each of these is a single-pe hit. We count the number oftimes Nobs(i) that PMT i records a charge that is greater thanthe threshold value. Since the location of the light source is notchanged between the two measurements, the complicating factorsin estimating those two intensities Qobs(i) and Nobs(i) are almostidentical:

Qobs!i"p Is # a!i" # ɛqe!i" # G!i" !1"

Nobs!i"p Iw # a!i" # ɛqe!i" !2"

where Is and Iw are the average intensities of high and lowintensity flashes, respectively, a(i) is the acceptance of ID-PMT i,ɛqe denotes its QE, and G(i) its gain. The threshold is sufficientlylow that the relative changes in gain, which we want to track, havelittle effect on Nobs(i), for example, 10% gain change makesthe Nobs(i) just 1.5% change. The low threshold enables us toignore, in the above calculations, differences in probability forhaving a charge below the discriminator threshold among PMTs.The gain of each PMT can then be derived by taking the ratio of

Eqs. (1) and (2), except for a factor common to all PMTs:

G!i"pQobs!i"Nobs!i"

: !3"

Then the relative gain of each ID-PMT can be obtained by normal-ization with the average gain over all PMTs.18

To perform this calibration we need a means to change theintensity of the flashes of the light source. The light source isnitrogen-laser-driven dye laser (Section 3.1.8). To manipulate theoverall intensity of the light delivered into the ID, we used a filterwheel with neutral density filters between the dye laser, and theoptical fiber that feeds light into the diffuser ball.

Fig. 10 shows the ratio (3) for each PMT, the RMS of thedistribution was found to be 5.9%. Since the HV value for eachPMT was determined to make Qobs be the same, we infer that thisdeviation is due to differences in QE among PMTs. The observedratio in Eq. (3) for each PMT, normalized by the average over allPMTs, contributed to a table of relative gain differences amongPMTs. These factors for relative gain differences of each PMT are

Fig. 8. The location of “standard PMTs” inside the SK inner detector (left). The red points indicate the locations of the standard PMTs. These PMTs served as references forother PMTs belonging to the same group with similar geometrical relationship to the light source (right). (For interpretation of the references to color in this figure caption,the reader is referred to the web version of this paper.)

differences (%)-6 -4 -2 0 2 4 6

0

200

400

600

800

1000mean = 0.00

rms = 1.27

Fig. 9. The observed percent charge differences for all ID-PMTs from theirrespective reference value.

relative gain0.6 0.8 1 1.2 1.40

500

1000 Mean 1.000RMS 0.059

Fig. 10. Distribution of relative gain of PMTs.

18 The common factor Is=Iw is also eliminated by this normalization. In theactual measurement, Nobs was corrected by occupancy.

K. Abe et al. / Nuclear Instruments and Methods in Physics Research A 737 (2014) 253–272 259

FIG. 40. Layout of SK standard PMTs. (Figure quoted from[16])

HV group-by-group, where the group is defined byFig. 40.

• Charge to photo-electron conversionConversion factor of charge (pC) to photo-electron(p.e.) were obtained by measuring 1 p.e. distri-bution. SK deployed “nickel source” in the tank,that generate 1 p.e. level of light, where the nickelsource is nickel-californium source; Ni(n,γ)Ni,Eγ ∼9 MeV. Figure 41 shows the SK nickel source.

Cf

γ "n

Ni

(prompt)

γ "(Ni captured) ~9MeV

FIG. 41. SK “Nickel source” (Figure quoted from [16])

• Photo-detection efficiencyThe photo-detection efficiency, ε, is defined byQuantum Efficiency times Collection Efficiency(CE). Hit rate (Nhit) for 1 p.e. level of lightis proportional to the photo-detection efficiency;Nhit ∝ Nphoton · ε. For this measurement, SK usedthe Nickel source to evaluate the hit rate, and com-pare with MC to evaluate relative efficiency over allPMTs.

• Timing calibrationCalibration for time response of readout channel(PMT and electronics), e.g. time-walk effect. SK

employed N2-dye laser and deployed diffuser-ball intank, that light source can generate 0.1∼1000 p.e.level light and covers the entire dynamic range ofelectronics. Evaluate TQ-maps for every singlePMTs, and evaluate detector timing resolution (forMC input).

2. Calibrations for physics analyses

The calibrations for physics analyses need to be de-signed for physics goal basis. This section describes thecalibrations used for SK atmospheric neutrino and T2Kanalyses, that relevant to nuPRISM physics goals.a. Photon yield and charge scale Although several

detailed detector calibrations have been carried out, thereare uncertainties on the photon propagation and pho-ton detection of the detector, that need to be tuned inthe detector simulation using a well known control sam-ples. For that, SK uses cosmic-ray muons, called “verticalthrough-going muons”. Figure 42 shows a schematic ofvertical through-going muon event of SK. The absolute

photon travel length

FIG. 42. Schematic of SK vertical through-going muonevents.

photon yield and charge scale in the detector simulationhave been tuned to data using the vertical through-goingmuon events that provide known muon track length andCherenkov photon travel distance.b. Momentum and energy scale SK event recon-

struction algorithm uses a conversion table that translatethe observed total charge in the Cherenkov ring to theparticle (muons and electrons) momentum. The conver-sion table is called “momentum table” have been evalu-ated using the detector simulation by generating parti-cles in momentum range of 10’s MeV/c to GeV/c. Basedon all detector calibrations and the simulation tuning,the detector and simulation are ready to use for physicsanalyses. Absolute energy scale is checked using natu-ral sources; decay electron, π0 mass, sub-GeV stoppingmuons, and multi-GeV stopping muons, these sources

39

cover the energy range of 10’s MeV to 10 GeV. SK de-tector simulation reproduces data within ∼ 2% and thathave been continuously monitored. SK defines the en-ergy scale uncertainty as the data-MC difference. If thesimulation does not reproduce the data reasonably well,the detector calibrations and simulation tuning need tobe revised.

V. CONCLUSION

The proposed nuPRISM detector has the potential toaddress the remaining systematic uncertainties that arenot well constrained by ND280. In particular, this detec-tor can constrain the relationship between measured lep-ton kinematics and incident neutrino energy without re-lying solely on rapidly-evolving neutrino interaction mod-els. Since nuPRISM is a water Cherenkov detector, theneutral current backgrounds with large systematic uncer-tainties at Super-K, particularly NCπ+ and NCπ0, canbe measured directly with a nearly identical neutrino en-ergy spectrum. The ability to produce nearly monoener-getic neutrino beams also provides the first ever ability

to measure neutral current cross sections as a functionof neutrino energy. Finally, nuPRISM provides a mecha-nism to separate the many single-ring e-like event typesto simultaneously constrain νe cross sections, neutral cur-rent background, and sterile neutrino oscillations.

The main long-baseline oscillation analysis presentedin this note was a νµ disappearance measurement, sincethe effects of various cross section models on this mea-surement had already been well studied, which provided auseful basis for comparison. However, it is also expectedthat nuPRISM will provide a significant improvementto the ultimate T2K constraint on δCP by constrainingneutral current backgrounds and electron-neutrino crosssections. Initial studies have also been presented thatdemonstrate the impact nuPRISM can have on both νeappearance measurements and anti-neutrino oscillationmeasurements. Other planned improvements to the anal-ysis include a realistic detector simulation and event re-construction. Thanks to the work done on event simula-tion and reconstruction in Hyper-K, these tools alreadyexist and can be quickly incorporated into the currentanalysis to perform more detailed studies of event pileupand detector performance for various detector configura-tions and PMT sizes and coverage.

[1] T. Dealtry et al. T2K-TN-172 (2014)[2] C. H. Llewellyn Smith, Phys. Rept. 3, 261 (1972).[3] R. A. Smith and E. J. Moniz, Nucl. Phys. B 43, 605

(1972) [Erratum-ibid. B 101, 547 (1975)].[4] O. Benhar, A. Fabrocini and S. Fantoni, Nucl. Phys. A

505, 267 (1989).[5] J. Marteau, Eur. Phys. J. A 5, 183 (1999) [hep-

ph/9902210].[6] M. Martini, M. Ericson, G. Chanfray and J. Marteau,

Phys. Rev. C 80, 065501 (2009) [arXiv:0910.2622 [nucl-th]].

[7] M. Martini, M. Ericson, G. Chanfray and J. Marteau,Phys. Rev. C 81, 045502 (2010) [arXiv:1002.4538 [hep-ph]].

[8] M. Martini, M. Ericson and G. Chanfray, Phys. Rev. C84, 055502 (2011) [arXiv:1110.0221 [nucl-th]].

[9] J. Nieves, I. Ruiz Simo and M. J. Vicente Vacas, Phys.Rev. C 83, 045501 (2011) [arXiv:1102.2777 [hep-ph]].

[10] Melanie Day, Kevin S. McFarland, Differences in Quasi-Elastic Cross-Sections of Muon and Electron NeutrinosPhys.Rev. D86 (2012) 053003

[11] Yu. Kudenko (for the T2K Collaboration). Nucl. Instr.and Meth. A598 (2009) 289-295; K. Abe et al., (The T2KCollaboration) Nucl. Instr. and Meth. A 659 (2011) 106.

[12] D. Renker, and E. Lorenz, JINST 4 (2009) P04004.[13] M. Yokoyama et al., Nucl. Instr. and Meth. A 610 (2009)

128; M.Yokoyama et al., Nucl. Instr. and Meth. A 622(2010) 567; A. Vacheret et al., Nucl. Instr. and Meth. A656 (2011) 69.

[14] Yu.Musienko, New developments in solid state photomul-tiplier, talk at the International Conference ”Instrumen-tation for Colliding Beam Physics” (INSTR14), 24 Febru-ary - 1 March 2014, Novosibirsk, Russia.

[15] Y. Fukuda et al. [Super-Kamiokande Collaboration],Nucl. Instrum. Meth. A 501, 418 (2003).

[16] K. Abe, Y. Hayato, T. Iida, K. Iyogi, J. Kameda,Y. Kishimoto, Y. Koshio and L. .Marti et al., Nucl.Instrum. Meth. A 737, 253 (2014) [arXiv:1307.0162[physics.ins-det]].

[17] O.Mineev et al., JINST 6 (2011) P12004;arXiv:1110.2651 [physics.ins-det].

[18] Y. Kudenko et al., Nucl. Instr. Meth. A469 (2001) 340.[19] S. Nakayama et al. [Hyper-Kamiokande Working Group],

”Hyper-K Tank”, Talk at the 4th Open Meeting for theHyper-Kamiokande Project, http://indico.ipmu.jp/

indico/getFile.py/access?contribId=29&sessionId=

4&resId=0&materialId=slides&confId=29

[20] MiniBooNE collaboration,Phys.Rev.Lett.102(2009)101802

[21] nuSTORM - Neutrinos from STORed Muons: Proposalto the Fermilab PAC, arXiv:1308.6822

[22] K. Abe et al. [T2K Collaboration], Phys. Rev. D 87,092003 (2013) [arXiv:1302.4908 [hep-ex]].

[23] Y. Nakajima et al. [SciBooNE Collaboration], Phys. Rev.D 83, 012005 (2011) [arXiv:1011.2131 [hep-ex]].

[24] B. G. Tice et al. [MINERvA Collaboration],[arXiv:1403.2103 [hep-ex]].

[25] A. A. Aguilar-Arevalo et al. [MiniBooNE Collaboration],Phys. Rev. D 81, 092005 (2010) [arXiv:1002.2680 [hep-ex]].

[26] Y. Nakajima, http://nuint11.in/final_nuint/

cc%20quasi%20and%20nc%20elastic%20scattering/

nakajima_nuint11.pdf

[27] VLyubushkin et al. [NOMAD Collaboration], Eur. Phys.J. C 63, 355 (2009) [arXiv:0812.4543 [hep-ex]].

[28] G. A. Fiorentini et al. [MINERvA Collaboration], Phys.

http://arxiv.org/abs/hep-ph/9902210

http://arxiv.org/abs/hep-ph/9902210

http://arxiv.org/abs/0910.2622






http://indico.ipmu.jp/indico/getFile.py/access?contribId=29&sessionId=4&resId=0&materialId=slides&confId=29








http://nuint11.in/final_nuint/cc%20quasi%20and%20nc%20elastic%20scattering/nakajima_nuint11.pdf




40

Rev. Lett. 111, no. 2, 022502 (2013) [arXiv:1305.2243[hep-ex]].

[29] R. Gran et al. [K2K Collaboration], Phys. Rev. D 74,052002 (2006) [hep-ex/0603034].



[32] A. Rodriguez et al. [K2K Collaboration], Phys. Rev. D78, 032003 (2008) [arXiv:0805.0186 [hep-ex]].

[33] C. Mariani et al. [K2K Collaboration], Phys. Rev. D 83,054023 (2011) [arXiv:1012.1794 [hep-ex]].

[34] M. Hasegawa et al. [K2K Collaboration], Phys. Rev. Lett.95, 252301 (2005) [hep-ex/0506008].

[35] O. Lalakulich and U. Mosel, Phys. Rev. C 88, no. 1,017601 (2013) [arXiv:1305.3861 [nucl-th]].


[37] S. Nakayama et al. [K2K Collaboration], Phys. Lett. B619, 255 (2005) [hep-ex/0408134].

[38] J. A. Formaggio and G. P. Zeller, Rev. Mod. Phys. 84,1307 (2012) [arXiv:1305.7513 [hep-ex]].

[39] E. Hawker, Proceedings of the Second InternationalWorkshop on Neutrino-Nucleus Interactions in the Few-GeV Region, Irvine, CA, unpublished

[40] T. Kikawa et al. T2K-TN-132 (2013)[41] E. Kearns et al. http://www.phy.duke.edu/∼cwalter/nusag-

members/2km-proposal-05-05-30.pdf (2005)


http://arxiv.org/abs/hep-ex/0603034










http://www.phy.duke.edu/~cwalter/nusag-members/2km-proposal-05-05-30.pdf

http://www.phy.duke.edu/~cwalter/nusag-members/2km-proposal-05-05-30.pdf

Documents

Letter of Intent to Construct a nuPRISM Detector in the J ... · Letter of Intent to Construct a nuPRISM Detector in the J-PARC Neutrino Beamline S.Bhadra,24 A.Blondel,3 S.Bordoni,5