Measurement of the Flux-Averaged Inclusive Charged-Current Electron Neutrino and Antineutrino Cross Section on Argon using the NuMI Beam and the MicroBooNE Detector
MicroBooNE collaboration, P. Abratenko, M. Alrashed, R. An, J. Anthony, J. Asaadi, A. Ashkenazi, S. Balasubramanian, B. Baller, C. Barnes, G. Barr, V. Basque, L. Bathe-Peters, O. Benevides Rodrigues, S. Berkman, A. Bhanderi, A. Bhat, M. Bishai, A. Blake, T. Bolton, L. Camilleri, D. Caratelli, I. Caro Terrazas, R. Castillo Fernandez, F. Cavanna, G. Cerati, Y. Chen, E. Church, D. Cianci, J.M. Conrad, M. Convery, L. Cooper-Troendle, J.I. Crespo-Anadon, M. Del Tutto, S.R. Dennis, D. Devitt, R. Diurba, L. Domine, R. Dorrill, K. Duffy, S. Dytman, B. Eberly, A. Ereditato, L. Escudero Sanchez, J.J. Evans, G.A. Fiorentini Aguirre, R.S. Fitzpatrick, B.T. Fleming, N. Foppiani, D. Franco, A.P. Furmanski, D. Garcia-Gamez, S. Gardiner, G. Ge, S. Gollapinni, O. Goodwin, E. Gramellini, P. Green, H. Greenlee, W. Gu, R. Guenette, P. Guzowski, L. Hagaman, E. Hall, P. Hamilton, O. Hen, C. Hill, G.A. Horton-Smith, A. Hourlier, R. Itay, C. James, J. Jan de Vries, X. Ji, L. Jiang, J.H. Jo, R.A. Johnson, Y.J. Jwa, N. Kamp, N. Kaneshige, G. Karagiorgi, W. Ketchum, B. Kirby, M. Kirby, T. Kobilarcik, I. Kreslo, R. LaZur, I. Lepetic, K. Li, Y. Li, B.R. Littlejohn, D. Lorca, W.C. Louis, X. Luo, A. Marchionni, C. Mariani, D. Marsden, J. Marshall, J. Martin-Albo, D.A. Martinez Caicedo, K. Mason, et al. (88 additional authors not shown)
MMeasurement of the Flux-Averaged Inclusive Charged-Current Electron Neutrino andAntineutrino Cross Section on Argon using the NuMI Beam and the MicroBooNEDetector
P. Abratenko, M. Alrashed, R. An, J. Anthony, J. Asaadi, A. Ashkenazi,
19, 33
S. Balasubramanian, B. Baller, C. Barnes, G. Barr, V. Basque, L. Bathe-Peters, O. Benevides Rodrigues, S. Berkman, A. Bhanderi, A. Bhat, M. Bishai, A. Blake, T. Bolton, L. Camilleri, D. Caratelli, I. Caro Terrazas, R. Castillo Fernandez, F. Cavanna, G. Cerati, Y. Chen, E. Church, D. Cianci, J. M. Conrad, M. Convery, L. Cooper-Troendle, J. I. Crespo-Anad´on,
10, 6
M. Del Tutto, S. R. Dennis, D. Devitt, R. Diurba, L. Domine, R. Dorrill, K. Duffy, S. Dytman, B. Eberly, A. Ereditato, L. Escudero Sanchez, J. J. Evans, G. A. Fiorentini Aguirre, R. S. Fitzpatrick, B. T. Fleming, N. Foppiani, D. Franco, A. P. Furmanski, D. Garcia-Gamez, S. Gardiner, G. Ge, S. Gollapinni,
34, 17
O. Goodwin, E. Gramellini, P. Green, H. Greenlee, W. Gu, R. Guenette, P. Guzowski, L. Hagaman, E. Hall, P. Hamilton, O. Hen, C. Hill, G. A. Horton-Smith, A. Hourlier, R. Itay, C. James, J. Jan de Vries, X. Ji, L. Jiang, J. H. Jo, R. A. Johnson, Y.-J. Jwa, N. Kamp, N. Kaneshige, G. Karagiorgi, W. Ketchum, B. Kirby, M. Kirby, T. Kobilarcik, I. Kreslo, R. LaZur, I. Lepetic, K. Li, Y. Li, B. R. Littlejohn, D. Lorca, W. C. Louis, X. Luo, A. Marchionni, C. Mariani, D. Marsden, J. Marshall, J. Martin-Albo, D. A. Martinez Caicedo, K. Mason, A. Mastbaum, N. McConkey, V. Meddage, T. Mettler, K. Miller, J. Mills, K. Mistry, A. Mogan, T. Mohayai, J. Moon, M. Mooney, A. F. Moor, C. D. Moore, L. Mora Lepin, J. Mousseau, M. Murphy, D. Naples, A. Navrer-Agasson, R. K. Neely, P. Nienaber, J. Nowak, O. Palamara, V. Paolone, A. Papadopoulou, V. Papavassiliou, S. F. Pate, A. Paudel, Z. Pavlovic, E. Piasetzky, I. D. Ponce-Pinto, D. Porzio, S. Prince, X. Qian, J. L. Raaf, V. Radeka, A. Rafique, M. Reggiani-Guzzo, L. Ren, L. Rochester, J. Rodriguez Rondon, H. E. Rogers, M. Rosenberg, M. Ross-Lonergan, B. Russell, G. Scanavini, D. W. Schmitz, A. Schukraft, W. Seligman, M. H. Shaevitz, R. Sharankova, J. Sinclair, A. Smith, E. L. Snider, M. Soderberg, S. S¨oldner-Rembold, S. R. Soleti,
24, 13
P. Spentzouris, J. Spitz, M. Stancari, J. St. John, T. Strauss, K. Sutton, S. Sword-Fehlberg, A. M. Szelc, N. Tagg, W. Tang, K. Terao, C. Thorpe, M. Toups, Y.-T. Tsai, M. A. Uchida, T. Usher, W. Van De Pontseele,
24, 13
B. Viren, M. Weber, H. Wei, Z. Williams, S. Wolbers, T. Wongjirad, M. Wospakrik, W. Wu, E. Yandel, T. Yang, G. Yarbrough, L. E. Yates, G. P. Zeller, J. Zennamo, and C. Zhang (The MicroBooNE Collaboration) ∗ Universit¨at Bern, Bern CH-3012, Switzerland Brookhaven National Laboratory (BNL), Upton, NY, 11973, USA University of California, Santa Barbara, CA, 93106, USA University of Cambridge, Cambridge CB3 0HE, United Kingdom St. Catherine University, Saint Paul, MN 55105, USA Centro de Investigaciones Energ´eticas, Medioambientales y Tecnol´ogicas (CIEMAT), Madrid E-28040, Spain University of Chicago, Chicago, IL, 60637, USA University of Cincinnati, Cincinnati, OH, 45221, USA Colorado State University, Fort Collins, CO, 80523, USA Columbia University, New York, NY, 10027, USA Fermi National Accelerator Laboratory (FNAL), Batavia, IL 60510, USA Universidad de Granada, Granada E-18071, Spain Harvard University, Cambridge, MA 02138, USA Illinois Institute of Technology (IIT), Chicago, IL 60616, USA Kansas State University (KSU), Manhattan, KS, 66506, USA Lancaster University, Lancaster LA1 4YW, United Kingdom Los Alamos National Laboratory (LANL), Los Alamos, NM, 87545, USA The University of Manchester, Manchester M13 9PL, United Kingdom Massachusetts Institute of Technology (MIT), Cambridge, MA, 02139, USA University of Michigan, Ann Arbor, MI, 48109, USA University of Minnesota, Minneapolis, MN, 55455, USA New Mexico State University (NMSU), Las Cruces, NM, 88003, USA Otterbein University, Westerville, OH, 43081, USA University of Oxford, Oxford OX1 3RH, United Kingdom Pacific Northwest National Laboratory (PNNL), Richland, WA, 99352, USA University of Pittsburgh, Pittsburgh, PA, 15260, USA a r X i v : . [ h e p - e x ] J a n Rutgers University, Piscataway, NJ, 08854, USA Saint Mary’s University of Minnesota, Winona, MN, 55987, USA SLAC National Accelerator Laboratory, Menlo Park, CA, 94025, USA South Dakota School of Mines and Technology (SDSMT), Rapid City, SD, 57701, USA University of Southern Maine, Portland, ME, 04104, USA Syracuse University, Syracuse, NY, 13244, USA Tel Aviv University, Tel Aviv, Israel, 69978 University of Tennessee, Knoxville, TN, 37996, USA University of Texas, Arlington, TX, 76019, USA Tufts University, Medford, MA, 02155, USA Center for Neutrino Physics, Virginia Tech, Blacksburg, VA, 24061, USA University of Warwick, Coventry CV4 7AL, United Kingdom Wright Laboratory, Department of Physics, Yale University, New Haven, CT, 06520, USA
We present a measurement of the combined ν e + ¯ ν e flux-averaged charged-current inclusive crosssection on argon using data from the MicroBooNE liquid argon time projection chamber (LArTPC)at Fermilab. Using the off-axis flux from the NuMI beam, MicroBooNE has reconstructed 214candidate ν e + ¯ ν e interactions with an estimated exposure of 2.4 × protons on target. Given theestimated purity of 38.6%, this implies the observation of 80 ν e + ¯ ν e events in argon, the largest suchsample to date. The analysis includes the first demonstration of a fully automated application of adE/dx-based particle discrimination technique of electron and photon induced showers in a LArTPCneutrino detector. We measure the ν e + ¯ ν e flux-averaged charged-current total cross section to be6 . ± .
51 (stat.) ± .
33 (sys.) × − cm / nucleon, for neutrino energies above 250 MeV and anaverage neutrino flux energy of 905 MeV when this threshold is applied. The measurement issensitive to neutrino events where the final state electron momentum is above 48 MeV/c, includesthe entire angular phase space of the electron, and is in agreement with the theoretical predictionsfrom GENIE and
NuWro . This measurement is also the first demonstration of electron neutrinoreconstruction in a surface LArTPC in the presence of cosmic ray backgrounds, which will be acrucial task for surface experiments like those that comprise the Short-Baseline Neutrino (SBN)Program at Fermilab.
I. INTRODUCTION
The measurement of electron neutrinos ( ν e ) appearingin a muon-neutrino beam is the cornerstone of currentand future accelerator-based neutrino oscillation experi-ments. The appearance oscillation channel allows long-baseline experiments to determine the neutrino mass or-dering [1] and to search for CP violation in the neutrinosector [2, 3]. It further allows short-baseline experimentsto shed light on the possible existence of sterile neutri-nos [4]. The success of these experiments relies on a pre-cise understanding of electron-neutrino interactions withthe detector target. LArTPCs are being employed toperform all of the above-mentioned measurements. Mi-croBooNE [5] and ICARUS [4] are already running, whileSBND [4] and DUNE [6] are under construction. How-ever, only ArgoNeuT [7] has made a measurement ofelectron-neutrino interactions on argon which includeda sample of 13 selected events. In addition, only a hand-ful of measurements of electron-neutrino interactions onother nuclei in the hundred MeV to GeV range are avail-able [8–10].The lack of precise electron-neutrino cross section mea-surements has been mitigated in short-baseline oscillationmeasurements by using ν µ interactions to constrain theoscillated ν e flux and cross section models [11, 12]. In ∗ microboone [email protected] such an approach, any uncertainty on the ν e / ν µ crosssection ratio would reduce the strength of the constraintthat the ν µ can provide to a ν e measurement. These dif-ferences, predicted to be on the order of 10%, arise fromthe different final state lepton mass, radiative corrections,and modifications to the pseudo-scalar form factor [13].The last of these effects can be difficult to calculate. Sim-ilarly, recent theoretical calculations of the ν e charged-current (CC) to ν µ CC cross section ratios [14] predictdifferences of as much as 25% between the ν e and ν µ cross sections, particularly for forward-going leptons inthe sub-GeV range. The uncertainty on the electron neu-trino and antineutrino interaction model can be respon-sible for the majority of the uncertainty of the oscillationmeasurement [15]. Independent direct measurements ofelectron-neutrino cross sections are therefore crucial tofurther inform our understanding of different flavor neu-trino interactions. Performing these measurements withhigh precision requires suppression of backgrounds con-sisting of photon showers. This can be done using theamount of energy deposited per unit length at the originof the shower, usually referred to as dE/dx, combinedwith the distance of the shower from the interaction ver-tex. Both of these methods are strengths of LArTPCdetectors and their use is demonstrated in this work.In this paper, we present the first measurement of theelectron neutrino and antineutrino charged-current crosssection on argon using the MicroBooNE [5] detector atFermilab. We employ neutrinos from the Neutrinos fromthe Main Injector (NuMI) neutrino beam [16] runningin its “forward horn current” mode which selects neu-trinos over anti-neutrinos for the on-axis component ofthe flux. The measurement is performed for ν e + ¯ ν e energies above 250 MeV and an average neutrino fluxenergy at MicroBooNE of 905 MeV with this threshold.We select an inclusive sample of interactions defined inSection IV B 1 requiring at least one reconstructed elec-tromagnetic shower inside of the fiducial volume.While in principle, the L/E of the NuMI beam withMicroBooNE is similar to that of the Booster NeutrinoBeam (BNB), an analogous oscillation search using theNuMI beam is not practical given its significantly largerintrinsic electron neutrino content and flux uncertainties.This paper is structured as follows: first, we briefly de-scribe the MicroBooNE experiment (Section II) and themain features of the NuMI neutrino beam at the Micro-BooNE detector. We then describe the simulation andreconstruction chain (Section III), the event selection cri-teria, and their performance (Section IV). We report themeasured cross section (Section V) and conclude with adiscussion of systematic uncertainties (Section VI). e - e - e - e - e - e - LightReadoutCharge Readout from Wire Planes
ScintillationIonisation e - E D r i f t zyx θ φ x Cathode
FIG. 1. A diagram of a LArTPC with the coordinate sys-tem used in this analysis. The z coordinate points in thedirection along the Booster Neutrino Beam, MicroBooNE’sprimary neutrino beam (see text for details); y in the up-wards direction of the TPC; and x from the three wire planes(anode) to the cathode (colored in purple). The red crosssymbol marks the coordinate system origin. The angle θ isdefined as the angle off the z -axis and the angle φ is definedas the angle in the xy plane with φ = 0 ◦ pointing toward thecathode. II. THE MICROBOONE EXPERIMENT
The MicroBooNE experiment is a LArTPC with an85 tonne active mass, housed inside of a stainless-steelcryostat. The TPC has dimensions of 2.56 m (width, x ),2.30 m (height, y ), and 10.37 m (length, z ). Figure 1shows a diagram of the LArTPC together with the coor-dinate system used in this analysis. Here, we only focuson the elements of the detector crucial to this analysis.A more in-depth description of the MicroBooNE experi-ment is given in Ref. [5]. FIG. 2. A display of a selected electron neutrino candidaterecorded by the MicroBooNE detector using the NuMI beamalongside a number of cosmic ray tracks. The horizontal di-rection represents the wires on the collection plane and thevertical direction represents the electron drift time. Colorsrepresent the amount of charge deposited on the wires. Thegaps in some of the cosmic ray tracks and the electromagneticshower are due to unresponsive wires.
Charged particles traversing the liquid argon ionizeand excite the argon atoms and generate free electronsand scintillation light along their path. This scintilla-tion light is detected by 32 photomultiplier tubes (PMTs)located behind the anode. A section of the full PMTsystem is depicted in Fig. 1. Each PMT gives a signalresponse within nanoseconds of the neutrino interactionwhich is significantly before the charge ionization signalis observed. To record a neutrino event, the MicroBooNEdetector NuMI online trigger requires a scintillation lightsignal above 9.5 photo-electrons (PE) to be in-time withthe accelerator beam spill window.An electric field produced by a 70 kV drop over the2.56 m drift distance attracts the free electrons towardsan anode consisting of three planes of sensing wires. Thefree electrons induce a signal on the inner two wire planes(induction planes, with wires oriented at ± ◦ from verti-cal) before being collected on the outer wire plane (collec-tion plane, with wires oriented vertically). The signal onthe wires provides position and calorimetric informationfor charged particles traversing the detector. Combiningthe information from the anode wire planes with timing LArTF A b s o r b e r T a r g e t H o r n s D e c a y P i p e N o t t o S c a l e ~ o ~ o MINOS,MINERvA,NOvA N u M I B ea m li n e S i d e V i e w B B M a i n I n j e c t o r G e V B e a m ~ o m ~ o N o t t o S c a l e BNB NuMIAbsorberNuMI Beamline Top View
FIG. 3. The position of the MicroBooNE detector relative to the NuMI neutrino beam target with views projected to the sideand above. The NuMI beamline is angled 3 ◦ downwards and the distance of the NuMI target to MicroBooNE is approximately679 m. The flux of neutrinos at MicroBooNE covers angles ranging from 8 ◦ to 120 ◦ relative to the NuMI beamline direction. obtained from scintillation light enables reconstruction ofthese interactions in 3D.A candidate electron-neutrino interaction in the Micro-BooNE detector recorded by the collection plane wirescan be seen in Fig. 2. The time needed for electronsto drift from the cathode to the anode is approximately2.2 ms. In this time frame, multiple cosmic ray trackscross the argon volume and can potentially contributeto the backgrounds of any neutrino analysis. Cosmicray tracks can be seen in Fig. 2 alongside the candidateelectron-neutrino interaction.MicroBooNE can detect neutrinos from the two neu-trino beams produced at Fermilab. The detector is ex-posed to an on-axis flux from the BNB [11], and an off-axis flux of neutrinos from the NuMI beam [16]. TheNuMI beam is created from collisions of protons accel-erated to an energy of 120 GeV with a graphite target.These collisions start a particle cascade resulting in parti-cles such as pions and kaons that can produce a neutrinofrom their decay. Particles of a particular electric chargefrom this cascade are focused by magnetic horns wherethe sign of the particles being focused depends on the di-rection of the electrical current applied to the horns. Thisanalysis uses data from the NuMI beam in forward horncurrent mode which uses a horn current of +200 kA. Thismode selects positively charged mesons and results in theon-axis flux being dominated by neutrinos. The major-ity of NuMI neutrinos interacting in MicroBooNE in theenergy range used in this analysis originate at the beamtarget and arrive at the detector at an angle close to 8 ◦ relative to the NuMI beamline direction. The position ofthe MicroBooNE detector relative to the NuMI target isshown in Fig. 3. Due to the energy of the protons gener-ating the beam and the position of MicroBooNE relativeto the NuMI beamline, the NuMI neutrino flux at Micro- BooNE has a composition of roughly 96% ν µ + ¯ ν µ and 4% ν e + ¯ ν e for energies above 250 MeV. The ν e componentis a factor of 10 larger than in the BNB making it anexcellent source of electron neutrinos. Neutrino Energy [GeV] P O T / . x ) / M e V / c m n ( F (56.6%) m n (39.4%) m n (2.5%) e n (1.5%) e n at MicroBooNEOff-axis NuMI FluxForward Horn Current Mode FIG. 4. The NuMI beam neutrino flux incident on the Micro-BooNE detector during NuMI forward horn current running.The percentages shown are calculated by applying a 250 MeVthreshold on the neutrino energy.
The NuMI beam neutrino flux at MicroBooNE for eachneutrino flavor is shown in Fig. 4. Each NuMI acceleratorbeam spill delivers ∼ protons on target (POT) overa duration of 9.6 µ s. Each accelerator spill consists of sixproton batches. To increase the neutrino intensity the ac-celerator complex can be run in slip-stacking mode, dou-bling the proton intensity in some batches [17]. BetweenOctober 2015 and July 2016, in the course of its first yearof data-taking, MicroBooNE collected 2 . × POT ofNuMI beam data while the NuMI beam operated in for-ward horn current mode. The majority of these data wascollected with the NuMI beam in a 4+6 slip-stacking con-figuration which means the first four out of the six protonbatches have double the usual intensity. A smaller frac-tion of the NuMI data taken during this period also con-tain 5+6 and 6+6 slip-stacked data. The NuMI simula-tion used by MicroBooNE consistently assumes 6 batches(i.e. no slip-stacking). Due to the low neutrino interac-tion rate at MicroBooNE, multiple interactions in onespill are rare. We thus scale the simulated events tomatch the integrated data exposure, neglecting the sub-percent effects of pile-up.
III. SIMULATION AND RECONSTRUCTION
The simulation and reconstruction of neutrino eventsfrom the NuMI beam in MicroBooNE is a complex set ofsteps that needs to account for both neutrinos and cosmicrays interacting within the detector. The proton inter-actions at the NuMI target, the meson re-interactions inthe NuMI beamline, and the resulting flux of neutrinosare simulated using a custom simulation package,
FLUGG ,developed by the MINOS collaboration. This packagecombines
FLUKA [18] to model the particle interactionsand
Geant4 [19] to model the beamline geometry [20].The flux prediction additionally uses the
PPFX software[21] which is also used by the NO ν A and MINER ν Aexperiments.
PPFX uses data from fixed-target experi-ments to constrain the hadron production in the NuMIbeamline. To constrain the NuMI flux at MicroBooNEwe use the
PPFX thin target (targets of few interac-tion lengths) constraints. More details can be found inRef. [21]. The neutrino flux is provided as input to the
GENIE [22] neutrino event generator [23].
GENIE simulatesthe neutrino-argon interactions inside the MicroBooNEcryostat volume. In parallel with the neutrino gener-ation, a spectrum of cosmic ray particles is simulatedusing the
CORSIKA [24] software package [25]. The result-ing Monte Carlo simulated (MC) events are processedusing the
LArSoft [26] software framework.
LArSoft isan event-based toolkit to perform simulation, analysisand reconstruction of LArTPC events. In the simula-tion chain the neutrino interaction products and cosmicrays are propagated through the detector using
Geant4 ,taking into account field inhomogeneities caused by spacecharge accumulation [27]. This is then fed into a detectorsimulation resulting in realistic waveforms on the anodesense-wires and PMTs. The version of the MicroBooNEsimulation software used in this analysis does not includethe effects of drift charge producing signals on neighbour-ing wires, known as dynamically induced charge [28, 29].This can impact the selection efficiency and backgrounds;its systematic effect is discussed in Section VI.The data acquired by the detector are reconstructedusing the same reconstruction chain as the MC generatedinteractions. The sense-wire waveform signals are de-convolved in one dimension with the electronics responsemeasured for each wire plane, resulting in waveforms with charge deposits having a uni-polar signature on all wireplanes. These charge deposits are then reconstructed as“hits” and fed into the
Pandora generic pattern-matchingreconstruction framework [30], which uses topologicaland calorimetric information to reconstruct and classifycharged particles as three-dimensional objects.
Pandora separates these objects into “tracks” (muon, proton andpion candidates) and “showers” (electron and photoncandidates), which are assembled into particle hierarchiesbased mainly on proximity and shared vertices.
Pandora also identifies the candidate neutrino interaction point asthe neutrino “vertex”. Calorimetric information is asso-ciated with the reconstructed 3D tracks and showers inthe form of dE/dx: the amount of energy deposited perunit length.The scintillation light signals acquired by the PMTsare also reconstructed. Light arriving at each PMT istranslated into photo-electrons and assembled into opti-cal hits. Optical hits from different PMTs are combinedinto “flashes”, which represent the total amount of lightrecorded from a single neutrino interaction or a cosmicray. Flashes are characterized by position, given by thePE-weighted positions of the included PMTs, and time.The scintillation light is used to determine the time ofeach interaction reconstructed in the TPC.The 2 . × POT dataset used in this analysis cor-responds to 6,361,077 NuMI accelerator beam triggerswith 734,221 of these passing the NuMI online trigger.We refer to these events as beam-on data. We comparethese to a sample of 728,500
GENIE generated ν µ , ¯ ν µ , ν e ,¯ ν e interactions inside the cryostat of MicroBooNE corre-sponding to 1 . × POT. Each of these events alsocontains
CORSIKA generated cosmic rays. Neutrinos thatinteract within and outside the cryostat walls of the Mi-croBooNE detector can produce daughter particles whichcan travel inside the cryostat and produce enough lightto pass the NuMI online trigger. These are known as out-of-cryostat interactions. We utilize a sample of 407,926
GENIE ν µ , ¯ ν µ , ν e , ¯ ν e interactions generated within andoutside the MicroBooNE cryostat walls (with daughterparticles that travel inside the cryostat) to estimate thisbeam-induced background. Each out-of-cryostat inter-action is combined with CORSIKA generated cosmic raysin the MicroBooNE cryostat. The out-of-cryostat samplecorresponds to 1 . × POT. All the MC samples gen-erated are normalized to the total POT of the beam-ondata sample.Not all accelerator spills result in a neutrino interactionin MicroBooNE. In many cases, the detector reads outexclusively cosmic rays in-time with the beam window.To characterize these readout triggers when no neutrinois present, a dedicated sample of 6,264,334 triggers wascollected explicitly when the beam was off. This sampleis normalized to the number of triggers for the beam-ondata.
IV. SELECTION OF INCLUSIVECHARGED-CURRENT ν e -LIKE INTERACTIONSA. Event Classification We define our signal as a CC ν e or ¯ ν e interaction insidea fiducial volume in the MicroBooNE detector above an(anti-)neutrino energy threshold of 250 MeV. This anal-ysis is optimised towards high energies and therefore weset this threshold to exclude a region where the efficiencybegins to rapidly decrease. Our signal events are iden-tified by the presence of an electron or positron showerin the final state, regardless of the presence of additionalparticles. Because MicroBooNE is not able to differen-tiate electrons from positrons and, therefore, ν e versus¯ ν e , the resulting selection contains both particles. Asa consequence, we calculate the final cross section for acombination of ν e and ¯ ν e .A pure selection containing ν e and ¯ ν e CC interactionsrequires the use of several variables to remove any cos-mic rays and other beam-induced backgrounds which aremis-reconstructed as showers. Due to the variety of in-teraction modes and detector effects, some interactionsmay be incorrectly classified, merged with other parti-cles, partially reconstructed, or entirely unreconstructed.In order to study the signal efficiency and various back-ground contributions, we classify events in the MC sim-ulation as follows: ν e CC : ν e or ¯ ν e interactions with an energy above250 MeV with the primary interaction vertex inside thefiducial volume. This is our signal classification. ν e CC Out-FV : ν e or ¯ ν e CC interactions whose pri-mary interaction vertex is reconstructed inside the fidu-cial volume, while the true simulated vertex is locatedoutside the fiducial volume but inside the MicroBooNEcryostat. As such, these are classified as background.
Cosmic : MC cosmic ray particles generated by
CORSIKA [24] which are selected as the neutrino candi-date. ν µ CC : MC generated particles originating from ν µ or¯ ν µ CC interactions. This background category includesall interaction topologies. NC : MC generated particles from a neutrino neutralcurrent (NC) interaction, including all topologies exceptthose including π in the final state. NC π : MC generated particles for a NC interactionwith one or multiple π in the final state. We classifythese separately as the photons originating from π de-cays can closely mimic electron showers. Out-of-Cryostat : This category contains neutrinocandidates originating from simulated neutrino interac-tions within and outside the cryostat walls.
Beam-Off Data : Any neutrino candidates originat-ing from the sample of data collected when the beamwas off fall under this background category. It containsexclusively cosmogenically produced activity.
B. NuMI ν e + ¯ ν e Selection
We combine information from the NuMI beam extrac-tion with the scintillation light recorded by the Micro-BooNE PMTs and with TPC pattern recognition tech-niques. The selection does not target a specific part ofthe electromagnetic (EM) shower phase space, neither inangle nor energy. In order to reject beam and cosmicray interactions that could mimic our signal, we apply aselection divided into six stages. These are listed in Ta-ble I, together with the number of signal and backgroundevents surviving each stage.The selection efficiency shown in Table I is defined asthe number of selected ν e + ¯ ν e CC interactions with anenergy above 250 MeV in the fiducial volume divided bythe number of simulated ν e + ¯ ν e CC interactions with thesame energy threshold in the fiducial volume before anyselection is applied. The selection purity is defined as thenumber of selected ν e + ¯ ν e CC interactions in the fiducialvolume with an energy above 250 MeV divided by thetotal number of selected neutrino candidates (signal andbackground).
1. Pre-selection
The goal of the first stage of the analysis is to identifyevents where a neutrino interaction happened inside avolume where they can be reliably reconstructed and hasa flash coincident with the beam window.
Pandora classifies each region of activity in the TPCas either a track or a shower. Our selection requires atleast one reconstructed shower associated to the
Pandora neutrino candidate. The leading shower in the event isdefined as the reconstructed shower object with the mostcharge deposition associated with it.The distribution of the reconstructed flash time isshown in Fig. 5 for data versus the stacked predictionof MC + NuMI beam-off data. The NuMI beam spillwindow occurs between 5.5 to 16.0 µ s. We reject flashesreconstructed outside of this window. The shoulder of theflash distribution on either side of the beam window isdue to the NuMI online trigger gate being slightly widerthan the NuMI beam spill window. This region is dom-inated by the beam-off data and is well-modeled. Theshape between 3 and 4.5 µ s is driven by flashes inducedby cosmic activity that happen before the NuMI onlinetrigger which have late scintillation light arriving insidethe NuMI online trigger window. Flashes between 17 and19 µ s are generated mainly by argon late-light scintilla-tion from interactions that happened during the beamwindow. The abrupt change in the number of flashes forbeam-on data around 12 to 15 µ s is a result of the 4+6slip-stacking configuration of the NuMI beam. The MCis generated uniformly across the beam window such thatthe integral is equivalent to the integral of the beam-ondata. This has no impact on normalization of the predic-tion to beam-on data because we normalize to the total TABLE I. A summary of the number of events in this analysis for data, simulated signal, beam background, cosmic MCbackground, and beam-off data background (scaled to the data POT/triggers). The final two columns show the efficiency andpurity at different stages of the selection.Selection stage Data Signal Beam Bgd. Cosmic MC Bgd. Beam-Off Bgd. Efficiency [%] Purity [%](1) Pre-selection 70691 632.1 7629.6 7736.4 52838.4 69.4 0.9(2) Flash matching 11135 417.5 2160.8 613.7 6642.8 45.1 4.2(3) Vertex reconstruction quality 7704 329.9 1462.4 457.3 4708.4 36.0 4.7(4) Shower hit threshold 1889 276.9 509.5 82.6 725.0 29.9 17.4(5) Electron-like shower 453 139.5 105.5 15.6 156.4 15.0 33.5(6) Final selection 214 83.8 41.5 9.3 82.3 9.1 38.6
POT delivered. s] m Flash Time [ E n t r i e s POT · MicroBooNE NuMI Data 2.4
Beam-On Data (Stat.) MCOut-of-Cryostat Beam-Off Data
FIG. 5. Beam-on data (points) compared to prediction (MC+ NuMI beam-off data) for the reconstructed flash time beforethe selection is applied. The initial flash time corresponds tothe start of the MicroBooNE detector readout window. TheNuMI beam spill occurs between 5.5 and 16.0 µ s, where thegreatest number of flashes are observed. A large fraction of cosmogenic activity is caused bylow energy neutrons and photons which deposit less en-ergy on average than neutrinos. We therefore reject neu-trino candidates where the total light signal observed bythe PMTs is less than 50 PE. This is a highly efficientmethod of removing a significant number ( ≈ . (and a fiducial mass of 57.6 tonnes).Any neutrino interaction candidate with a reconstructedvertex outside of this volume is removed. This reducesthe amount of selected out-of-cryostat and cosmic-ray in-teractions, and minimizes the impact of non-uniform elec-tric field on the reconstructed tracks and showers near the cathode due to space charge accumulation [27].The efficiency for the pre-selection is 69.4%. This largeinitial loss in efficiency is primarily driven by the showerreconstruction performance, where the electron from the ν e CC interaction is either mis-reconstructed or not re-constructed at all. At this stage, the purity is 0.9%, asthe selection is dominated by the cosmic ray background.
2. Flash Matching
A powerful method to reject cosmic ray backgroundsis to combine the TPC and light information. This isknown as “flash matching”. In this analysis we calculatethe distance from the position of the largest flash (i.e.the flash with the most PE) occurring inside of the beamwindow to the reconstructed
Pandora vertices in 2D ( yz plane). We assume that the largest flash in the beamwindow was produced by the neutrino interaction. Thedistance, ∆ YZ , is constructed as:∆ YZ = (cid:113) ( z flash − z tpc ) + ( y flash − y tpc ) , (1)where z flash and y flash are the reconstructed center of thelargest flash and z tpc and y tpc are the reconstructed neu-trino candidate vertex coordinates.Compared to neutrino interactions, cosmic-inducedbackgrounds (beam-off and CORSIKA
MC) tend to havelarger match distances when compared with the largestoptical flash registered during the beam window. Weselect events using the 2D match distance taking into ac-count the relative positions of the flash and the TPC in-teraction vertices in the z coordinate. Since the neutrinointeractions are usually forward going, the flash centershould be downstream of the neutrino interaction. Thismotivates a tighter selection if the reconstructed TPCvertex is downstream of the flash position: z tpc > z flash → ∆ YZ <
60 cm ,z tpc < z flash → ∆ YZ <
80 cm . (2)
3. Vertex Reconstruction Quality
To mitigate backgrounds resulting from mis-reconstruction or incorrect particle hierarchy asso-ciations, we examine the distance of showers and tracksfrom the vertex, which is a metric of reconstructionquality. This also removes background events where theleading shower originates from a photon. For example,in NC π events, the π decays to two photons whichpair convert after travelling some distance resulting inboth showers being displaced from the vertex.We remove any neutrino candidates where the lead-ing shower is reconstructed with a start point furtherthan 4 cm from the neutrino vertex. This requirementapplies only to the leading shower to ensure an inclu-sive selection of ν e CC topologies which produce show-ers distant from the neutrino interaction (e.g. ν e CC π production). If the neutrino candidate event includes re-constructed tracks, we additionally require that at leastone track must start within 4 cm of the reconstructedneutrino vertex; this additional selection removes eventswith associated cosmic activity, while allowing for somemis-recontruction effects due to unresponsive wires. Af-ter applying these vertex quality variables, the purityincreases to almost 5%, see Table I.
4. Shower Hit Threshold E n t r i e s Beam-On Data (Stat.)Beam-Off DataOut-of-Cryostat p NC NC CC m n Cosmic CC Out-FV e n CC e n Stat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4
Leading Shower Hits (All Planes) - - ( D a t a - M C ) / M C FIG. 6. Comparison of data (points) to the stacked prediction(MC + NuMI beam-off data) for the number of hits for theleading shower across all planes following the pre-selection,flash matching, and vertex reconstruction quality selectionstages. E n t r i e s True Electron Energy [GeV]
Lead i ng S ho w e r H i t s ( A ll P l ane s ) MicroBooNE Simulation
FIG. 7. Number of hits in the leading shower (all planes)against true electron and positron energy for all simulated ν e and ¯ ν e interactions. The more hits that are associated to a shower, theeasier it becomes to reconstruct its properties. Con-versely, showers with very small numbers of hits are dif-ficult to reconstruct precisely and are more likely to beaffected by spurious charge depositions. We integrate thenumber of hits for the leading shower across the threewire planes; the resulting distribution is shown in Fig. 6.The majority of the backgrounds cluster at values below200 total hits in the leading shower, mainly due to mis-reconstructed muon tracks or the reconstruction splittinglarger showers into sets of smaller showers. Requiringgreater than 200 hits for the leading shower is effectiveat removing these mis-reconstructed showers, however,it impacts low energy neutrinos as well as backgroundevents. Figure 7 shows the relationship between the sim-ulated electron energy and the leading shower hits. Whilethe total number of hits correlates with the energy of theshower, this correlation is non-trivial. We observe thatthis selection requirement removes events in a bin of elec-tron energy just above 0.2 GeV, but a large populationof such events are maintained.We also utilize the number of hits in the collectionplane, which is typically the best-performing plane forreconstruction due to its high signal-to-noise ratio andis the plane used to calculate dE/dx at the start of theshower. Showers with few hits in this plane can lead tosizable uncertainties in the dE/dx measurement. There-fore, we require there to be at least 80 hits on the collec-tion plane for the leading shower.
5. Electron-like shower
The presence of an electron shower is the identifyingcharacteristic for determining whether an event was in-duced by an electron neutrino interaction. To isolate suchevents, we employ the key feature of LArTPCs: the abil-ity to distinguish photon-induced showers from electron-induced showers using a combination of calorimetric andtopological information (i.e. the measurement of dE/dxand the distance between the shower and interaction ver-tex). The initial dE/dx for showers induced by an elec-tron correspond to the minimum ionizing particle (MIP)value. Showers induced from photons will instead regis-ter higher values of initial dE/dx corresponding to doubleMIP ionization. This is due to the electron-positron pairfrom photon pair production, which is the dominant in-teraction mode for photons at the energies of interest.Before the pair production occurs, a photon does notionize the argon. This can lead to an identifiable gapfrom the vertex which becomes another clear signatureof background EM-showers.Where the previous selection variables address mainlythe backgrounds from cosmic rays and shower qual-ity, requirements on the leading shower opening angleand the shower dE/dx further remove tracks that aremis-reconstructed as showers (mostly beam-off data and ν µ CC interactions) and limit the contamination fromphoton-induced showers originating from NC π and ν µ CC π interactions.The shower opening angle, α open , is calculated usinga principal component analysis (PCA) of the 3D hit po-sitions of the reconstructed shower and is given by theequation: α open = tan − (cid:32) (cid:112) PCA secondary (cid:112)
PCA principal (cid:33) , (3)where PCA principal and PCA secondary are the lengths ofthe principal and secondary eigenvectors. Figure 8 givesan intuitive view of this angle on a schematic of a recon-structed shower. The opening angle is a powerful discrim-inator for cases where tracks have been mis-reconstructedas showers. For example, cosmic rays with broken tracksnear the neutrino interaction can be mis-reconstructedas the leading shower resulting in a large opening an-gle. We select neutrino candidates where α open < ◦ .In order to remove neutrino candidates with a topologywhich is more track-like than shower-like, we require aminimum opening angle of α open > ◦ . Figure 9 showsthe data versus MC prediction for shower opening anglebefore this requirement is made.One of the most powerful features of LArTPC tech-nology is that its entire volume is an active calorimeter.This means that the energy loss of a particle can be cal-culated along its trajectory, enabling the use of dE/dxfor particle identification. Using the dE/dx in the firstfew centimeters of a shower is a powerful tool to distin-guish electron-induced from photon-induced showers, as FIG. 8. Schematic showing the shower opening angle. Theangle indicated by the dotted-red line shows the shower open-ing angle, α open . The median dE/dx calculation is performedusing the charge deposition from the shower which falls withinthe box shown in blue. E n t r i e s Beam-On Data (Stat.)Beam-Off DataOut-of-Cryostat p NC NC CC m n Cosmic CC Out-FV e n CC e n Stat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4
Leading Shower Opening Angle [deg] - - ( D a t a - M C ) / M C FIG. 9. Comparison of data (points) to the stacked prediction(MC + NuMI beam-off data) for the leading shower open-ing angle. Neutrino candidates pass the selection for leadingshowers between 3 ◦ and 15 ◦ . was demonstrated by ArgoNeuT [31]. Selecting the me-dian dE/dx value as a truer representation of the shower’sdeposition profile rather than the arithmetic mean mit-igates effects from single outlier hits which could resultfrom Landau fluctuations as well as mis-configured elec-tronics, detector effects, or mis-reconstruction.The calculation of the median dE/dx is performedby constructing a 1 × box starting at the recon-structed shower start point, shown in Fig. 8 and calcu-lating the dE/dx for the collection plane single chargedeposits along the start of the shower. The charge de-posits are converted to an energy using the Modified Boxmodel [32].0Figure 10 shows the distribution of the calculateddE/dx on the collection plane wires for the leading showerusing this method. The signal distribution peaks inthe 2 MeV/cm region and large fractions of backgroundlie to either side of the peak. Selecting those neutrinocandidates whose dE/dx lies between 1.4 MeV/cm and3 MeV/cm greatly increases the purity of the sample.As expected, the neutrino candidates containing photon-producing processes are centered around 4 MeV/cm. E n t r i e s Beam-On Data (Stat.)Beam-Off DataOut-of-Cryostat p NC NC CC m n Cosmic CC Out-FV e n CC e n Stat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4
Leading Shower dE/dx (Collection Plane) [MeV/cm] - - ( D a t a - M C ) / M C FIG. 10. Comparison of data (points) to the stacked pre-diction (MC + NuMI beam-off data) for the leading showerdE/dx (calculated as described in text) distribution after theshower opening angle selection requirement. Signal inter-actions are peaked at 2 MeV/cm and beam-induced back-grounds are mostly peaked around 4 MeV/cm. A large num-ber of background events are found between 0 and 2 MeV/cm.
A notable feature in Fig. 10 is the large populationof leading showers with a dE/dx of nearly 0 MeV/cm.This population is caused by tracks and showers thatare nearly perpendicular to the beam direction (60 ◦ <θ < ◦ ) where it is challenging to measure dE/dx. Infuture analyses, this effect can be mitigated with the useof all three wire planes to measure dE/dx enabled byusing methods such as 2D deconvolution as laid out inRefs. [28, 29].Figure 11 shows the stacked data versus MC predic-tion where θ is between 0 ◦ and 60 ◦ . This slice of θ isthe most populated region and has considerably higherpurity than the rest of the phase space. As the dE/dxdistribution at this angular slice includes showers run-ning roughly perpendicular to the collection plane wires,a very small fraction of showers have an unphysically lowdE/dx, which demonstrates the angular dependence ofthe dE/dx calculation in this analysis.For values of θ between 60 ◦ and 120 ◦ , where the dE/dx E n t r i e s Beam-On Data (Stat.)Beam-Off DataOut-of-Cryostat p NC NC CC m n Cosmic CC Out-FV e n CC e n Stat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4 o < 60 q < o Leading Shower dE/dx (Collection Plane) [MeV/cm] - - ( D a t a - M C ) / M C FIG. 11. Comparison of data (points) to the stacked pre-diction (MC + NuMI beam-off data) for the leading showerdE/dx for a slice of θ between 0 ◦ and 60 ◦ after the showeropening angle selection requirement. In this angular range,the leading showers all have a well reconstructed dE/dx andthe signal peak is well-defined. is not well reconstructed, both the neutrino interactionsand the considerable cosmic ray background are peakedat dE/dx values closer to 0 MeV/cm. In the range of θ between 120 ◦ and 180 ◦ , we expect relatively few electron-neutrinos and a high contamination of cosmic rays. How-ever, in this sample, the majority of the leading showershave well reconstructed dE/dx close to 2 MeV/cm.Given that the reconstructed leading shower directioncan affect the calculation of dE/dx, using it may intro-duce an angular bias to the selection. However, it is anextremely powerful tool for removing cosmic and beam-induced backgrounds which dominate at low dE/dx, aswell as photon backgrounds at higher dE/dx values. Fu-ture analyses can mitigate the angular dependence ofdE/dx by using three plane calorimetry.This selection stage is successful in removing alarge fraction of photon-induced shower backgrounds.Given the importance of removing these backgrounds inelectron-neutrino analyses we explore the performance ofphoton-rejection variables further in Section IV C.
6. Final Selection
Mis-reconstruction of the neutrino candidate can leadto associating physically uncorrelated showers with theinteraction. To remove these cases, we require the dis-tance between the sub-leading showers and the neutrinocandidate vertex to be less than 22 cm. The scope of this1selection requirement is to mitigate mis-reconstructedshowers while retaining a sizable fraction of ν e CC π in-teractions ( 44% relative to the previous selection stage). E n t r i e s Beam-On Data (Stat.)Beam-Off DataOut-of-Cryostat p NC NC CC m n Cosmic CC Out-FV e n CC e n Stat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4 ] -1 Leading Shower (Hits / Length) [cm - - ( D a t a - M C ) / M C FIG. 12. Comparison of data (points) to the stacked pre-diction (MC + NuMI beam-off data) for the leading showerhit density variable after the selection requirement on the sub-leading shower distance to the vertex. The threshold is placedat 3 hits per cm where all neutrino candidates that have aleading shower hit density less than this value are removed.
Further differentiating between shower and track ob-jects is necessary in order to remove the cosmic ray in-teractions which dominate the currently selected sam-ple. The object hit density is calculated by summingthe number of hits associated to the leading shower anddividing by its length. The hit density for the leadingshower is shown in Fig. 12, the cosmic rays largely pop-ulate the lower values of hit density. The effectiveness ofthis selection is particularly sensitive to the reconstruc-tion of the transverse component of the shower; poorlyreconstructed shower objects are removed by a selectionon this variable. Placing a higher threshold on the hitdensity improves the selection purity, however the selec-tion efficiency especially in the low energy signal regionis impacted. A conservative threshold is placed at a hitdensity of 3 hits per cm.The relationship between the length of the longesttrack and leading shower can be used to discriminatebetween ν µ and ν e interactions. For instance, a ν µ CCinteraction typically contains a rather long muon trackand any showers associated to the interaction are typ-ically much shorter in length. Contrast this to a ν e CC interaction, where the tracks produced are often ofcomparable length, or shorter than the leading showerlength even in the presence of charged pions. Select-ing on such a variable also removes cases where Michel electrons are produced by muon decays. The parametershown in Fig. 13 is defined as the length of the longesttrack in the neutrino candidate event divided by the lead-ing shower length. We select neutrino candidates whoseratio is below 1.0. E n t r i e s Beam-On Data (Stat.)Beam-Off DataOut-of-Cryostat p NC NC CC m n Cosmic CC Out-FV e n CC e n Stat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4
Longest Track / Leading Shower Length - - ( D a t a - M C ) / M C FIG. 13. Comparison of data (points) to the stacked predic-tion (MC + NuMI beam-off data) for the ratio between thelongest track and the leading shower length following the se-lection requirement on the shower hit density. The thresholdis placed at 1.0, where all neutrino candidates that have atrack longer than the leading shower are removed.
To increase the purity of shower + track topologies,we require that both the reconstructed start and end ofthe track are contained inside the fiducial volume. Thisrequirement is particularly effective at rejecting long cos-mic rays or muons which cross the fiducial volume and areassociated to a shower inside the fiducial volume. Thishas a small impact on the signal selection and resultsroughly in a 3% increase in purity.The final number of neutrino candidates remaining af-ter each selection stage is shown in Table I.
C. Electron-Photon Separation
A key requirement of any analysis searching for elec-tron neutrinos is the ability to differentiate electrons orig-inating from ν e CC interactions from photons originatingfrom any backgrounds. The two main features that sep-arate interactions containing electrons from those withphotons are the dE/dx at the start of the shower and thedistance between the shower and the interaction vertex.The latter is only well-defined when another charged par-ticle is present at the interaction vertex. Electron-photonseparation in a LArTPC has previously been demon-2strated using a semi-automated reconstruction chain [31]and only leveraging the dE/dx. In this measurement, wedemonstrate for the first time both of the electron-photonseparation techniques that the LArTPC technology offersusing a fully automated analysis chain.To examine the performance of the electron-photonseparation variables, we isolate the dE/dx and showervertex distance selection steps on the leading shower bymoving them to the end of the analysis chain; this ensuresthat the upstream part of the selection chain identifiesneutrino interactions with a well-defined leading shower.For this study, we additionally require the leading shower θ to be between 0 ◦ and 60 ◦ . This focuses on the topolo-gies unaffected by the absence of dynamically inducedcharge in our simulation chain and with dE/dx best re-constructed on the collection plane wires. The very goodagreement between the data and MC samples allows usto utilize the MC sample, which provides true informa-tion about the nature of the leading shower, to determinethe power of the two separation methods.After applying the ν e + ¯ ν e CC selection without thedE/dx and shower vertex distance selection steps, weobtain a sample of 1995 simulated neutrino events. Inthis sample, the true particle responsible for the leadingshower is an electron in 48% of cases, a photon in 39%of cases with 13% remaining for other particles. We thenexamine the individual and combined effect of applyingthe dE/dx and the shower to vertex distance selectionrequirements on these three groups. The value of dE/dxis required to be between 1.4 and 3 MeV/cm and the dis-tance between the shower and the vertex to be less than4 cm apart. The combination of these two requirementsselects 59% of electron neutrino events and rejects 81% ofphoton backgrounds and over 61% of other backgrounds.When applying the requirements individually, the dE/dxis the significantly more powerful method of rejectingevents with photons removing 73% of those backgroundsby itself compared to 28% for the shower distance to ver-tex. It is also responsible for the bigger drop in our ef-ficiency to select electrons: 35% compared to 11%. Wealso investigate the effect of the shower to vertex dis-tance selection requirement on a subset of events with atleast one candidate track present. For this sample, theselection requirement has an improved performance in re-jecting photon backgrounds with 47% rejected comparedto 28% for events where we do not require the presence ofa reconstructed track. The summary of the performancefor each selection requirement applied individually andcombined can be found in Table II.We find that the dE/dx variable is more effective inremoving photon-induced backgrounds. Figure 14 illus-trates its separation power in rejecting the photon-likeevents which dominate around the 4 MeV/cm peak inthe dE/dx distribution.
TABLE II. Survival rate of a sample of 1995 neutrino eventswhere the leading shower is classified as originating from anelectron, photon, or other based on MC information. TheEM shower selection row refers to the ν e + ¯ ν e CC selectionwithout the dE/dx and shower to vertex distance selectionrequirements. The subsequent rows show the effect of thedE/dx and shower distance to vertex selection requirementsapplied individually and combined for this sample of events.The final row shows the performance of the shower vertexdistance selection requirement applied to events with at leastone candidate track present.Selection stage Electrons Photons OtherEM Shower Selection 951 771 273dE/dx (only) 65% 27% 52%Shower-Vertex Dist. (only) 89% 72% 73%Combined 59% 19% 39%Shower-Vertex Dist. 89% 53% 64%(only, ≥ E n t r i e s Beam-On Data (Stat.)Out-of-CryostatBeam-Off DataNeutronMuonKaonPionPhotonProtonElectronStat. UncertaintyMC + Beam-Off
POT · MicroBooNE NuMI Data 2.4 o < 60 q < o Leading Shower dE/dx (Collection Plane) [MeV/cm] - - ( D a t a - M C ) / M C FIG. 14. dE/dx of leading showers for neutrino candidatesbroken down by particle type. This plot is made for lead-ing shower θ between 0 ◦ and 60 ◦ where the reconstructionof showers is good. Electrons are gathered in the MIP peak,while most photons are around 4 MeV/cm. D. Selection Performance
Many of the selection requirements focus on removingcosmic ray interactions reconstructed as showers. Theoverall decrease in the cosmic ray contamination is a fac-tor of 10 compared to the initial Pandora reconstructionstage which can have many reconstructed cosmic raysin a readout window. This ultimately brings the cos-mic ray contamination to roughly the same size as thenumber of selected electron neutrino and antineutrino in-teractions. The remaining non-cosmic backgrounds con-3tribute to approximately 19% of the selected neutrinocandidates. This demonstrates the selection’s ability toreliably remove beam-induced backgrounds such as ν µ CC and π interactions. We find that the selection issensitive to neutrino events where the final state electronmomentum is higher than 48 MeV/c, and includes theentire angular phase space of the electron. N o S e l e c t i on ( ) P r e - s e l e c t i on ( ) F l a s h M a t c h i ng ( ) V e r t e x R e c o . Q ua li t y ( ) S ho w e r H i t T h r e s ho l d ( ) E l e c t r on - li k e S ho w e r ( ) F i na l T un i ng ( ) MicroBooNE Simulation
EfficiencyPurityPurity (Beam Only)EfficiencyPurityPurity (Beam Only)
FIG. 15. Summary of the selection performance, includingefficiency, purity, and purity (beam only). The beam only pu-rity corresponds to the selection purity if cosmic backgroundscould be completely removed.
A summary of the purity and efficiency at differentstages of the selection can be seen in Fig. 15. The steep-est decrease in the efficiency, once the basic pre-selectionis applied, occurs with the application of the showeropening angle and dE/dx selection requirements whichare both included in the electron-like shower category.For the latter the decrease occurs because of the diffi-culty in calculating the dE/dx for the shower directionof certain signal events. This is also the step with thelargest increase in purity. With improvements to calcu-lating dE/dx for all shower directions, this variable isexpected to become even more powerful. The beam-only purity (in green) is shown in Fig. 15 and has afinal value above 60%. This is calculated by consider-ing only beam-induced backgrounds which would be theideal case if all cosmic ray backgrounds in this analysiscould be completely removed. In future analyses withimproved cosmic rejection tools such as using the cosmicray tagger system installed around MicroBooNE, con-tamination from cosmic ray backgrounds should signif-icantly decrease which will enable much higher puritiesand improved selection performance. The final selectionefficiency is 9 . ± .
3% for the ν e + ¯ ν e sample, whichcan be divided into 8 . ± .
4% and 12 . ± .
0% for ν e and ¯ ν e respectively. - · / nu c l eon ] CC C r o ss S e c t i on [ c m e n + e n Data (stat. + sys.)GENIE v2.12.2GENIE v3.0.6NuWro v19.02.1
POT · Data 2.4MicroBooNE NuMI
FIG. 16. The extracted flux-averaged inclusive electron neu-trino and antineutrino charged-current total cross section onargon compared to the predictions made by
GENIE and
NuWro .This measurement is for energies above 250 MeV and theaverage ν e + ¯ ν e flux at MicroBooNE above this threshold is905 MeV. The measurement and predictions are in agreementwithin the statistical uncertainty. V. FLUX-AVERAGED INCLUSIVE ν e + ¯ ν e CCTOTAL CROSS SECTION
We employ the standard formula to extract the crosssection (cid:104) σ (cid:105) : (cid:104) σ (cid:105) = N − B(cid:15) × N target × Φ ν e +¯ ν e (4)where N is the total number of selected neutrino candi-dates, B the number of selected background events, (cid:15) theselection efficiency, N target the number of target nucleons,and Φ ν e +¯ ν e the integrated ν e + ¯ ν e POT-scaled flux. Thenumber of target nucleons in the fiducial volume definedin this analysis is 3 . × . The number of selectedsignal neutrinos, ( N − B ), in data is calculated to be80 . ± . N = 214 and B = 133 . ν e and ¯ ν e fluxes is 905 MeV, which iscalculated by integrating the flux from 250 MeV.The cross section, 6 . ± .
51 (stat.) × − cm , isin agreement with the GENIE and
NuWro [33] predictedvalues as seen in Fig. 16. The N − B term is theleading contribution to the 22% statistical uncertaintyon the extracted cross section. We find similar agree-ment with GENIE v2.12.2 as ArgoNeuT does with
GENIE v2.12.10c [7]. The theory predictions for the flux-averaged cross section between these versions of
GENIE are equivalent.
VI. SYSTEMATIC UNCERTAINTIES
The systematic uncertainties considered in the analysisarise from the simulation of neutrino interactions, propa-4gation of secondary particles, detector response, and neu-trino flux. The simulation of interactions on an argonnucleus is complex due to both the nature of the largenuclear target and the interplay of the different interac-tion modes in the 1 GeV energy region.
GENIE is usedto simulate the neutrino interactions with argon, usingcross section models that depend on a number of param-eters. Estimates of the uncertainties on these parametersare provided by
GENIE . These parameters in
GENIE aresimultaneously sampled 1000 times within their adopteduncertainties and are used to modify the simulated eventrates. This primarily modifies the background rates,though it does have a small impact on the signal effi-ciency. For each of the 1000 variations, the measuredcross section is re-calculated and the uncertainty is cal-culated as the standard deviation of the 1000 modifiedcross sections, leading to an uncertainty of 5%.The model parameter uncertainties provided by
GENIE v2.12.2 are supplemented by considering alternativemodels within
GENIE for charged-current quasi-elastic(CCQE) and meson exchange current (MEC) interac-tions - the dominant reaction mechanisms at Micro-BooNE. For low 4-momentum transfers (low- Q ), thecollective behavior of nucleons in the nucleus can lead toa suppression of the CCQE cross section. This physicseffect is not included in our default GENIE model set.MEC interactions are simulated using an empirical modelwhich does not include any associated uncertainties. Analternative CCQE model which includes suppression atlow- Q (calculated using the random phase approxima-tion, or RPA) and a theory-driven MEC model for CCinteractions [34–36] are used to modify the default simu-lation. Again, the cross section is recalculated with thismodified simulation. This alternative model set intro-duces a 9% change in the calculated cross section whichis mostly driven by the effect of the alternative CCQEmodel on the efficiency (6%).Uncertainties for proton and charged pion re-interactions in argon are estimated by recalculating thesurvival probability as a function of momentum aftermodifying their cross sections within their uncertainty(conservatively estimated to be 30%). The re-interactioncross sections for protons and charged pions are sampledsimultaneously 250 times resulting in a new cross sectionin each case. Taking the standard deviation of these 250modified cross sections results in an uncertainty of 2%.Systematic uncertainties originating from the detectormodeling are evaluated using independent modificationsto the detector simulation. Individual parameters in theunderlying detector model are varied and the events re-simulated. The measured cross section is recalculatedusing these modified simulations and the difference withrespect to the central value is taken to be the uncertainty.The uncertainties from the various detector parametersare added in quadrature, resulting in a 23% uncertainty.The largest contribution to this uncertainty is the dy-namically induced charge variation (16%) where we usea simulation sample with this effect included. This vari- ation affects the reconstruction of showers and greatlyimproves the data to MC agreement in variables such asthe shower multiplicity and momentum. Future itera-tions of this analysis will include this effect in the defaultsimulation. TABLE III. Summary of systematic uncertainties on the crosssection measurement in this analysis. The interaction un-certainty quoted here includes the
GENIE (5%), alternativeCCQE and MEC models (9%), and re-interaction (2%) un-certainties. The beam flux uncertainty includes the hadronproduction (21%) and beamline modeling (6%) uncertainties.Systematic Source Relative Uncertainty [%]Interaction 10Detector Response 23Beam Flux 22POT Counting 2Cosmic Simulation 4Out-of-Cryostat Simulation 6Total 34
The uncertainty in the flux prediction arises primarilyfrom the modeling of the particle cascade following theproton-target collision. An alternate beamline simulationcompatible with
PPFX was run and reweighted in neutrinoenergy and angle to the NuMI beamline to match thenominal flux prediction from
FLUGG . This reweighted fluxwas then modified by
PPFX according to the hadron pro-duction uncertainties from data. Each time the flux wasmodified, the cross section was recalculated, includingbackground subtraction and efficiency correction, givinga total uncertainty on the measurement from the hadronproduction of 21%.Additionally, we evaluate the uncertainties due to themodeling of the NuMI beamline by re-running the
PPFX -compatible beamline simulation with parameters such asthe target location, horn current, and beam spot sizechanged by their estimated uncertainties. The combina-tion of beamline uncertainties added in quadrature leadto a 6% uncertainty on the measured cross section. Anadditional uncertainty of 2% is added to account for po-tential inaccuracies in the counting of POT from beam-line monitors [21, 37].To estimate the uncertainty of the simulation of cosmicrays, we calculate the difference in selection rate of cos-mic rays between a sample of simulated neutrino interac-tions overlaid on beam-off data and
CORSIKA simulation.This difference is used to scale the selected
CORSIKA cos-mic rays in this analysis which varies the calculated crosssection by 4%.Simulation of out-of-cryostat interactions are depen-dent on many factors such as the geometry of the build-ing around MicroBooNE, the density of various materialsaround the building and
GENIE modeling of the neutrinointeractions in these materials. Due to this large set ofunknowns, the number of selected out-of-cryostat inter-actions is varied by 100% and the cross section is recal-culated. This gives an uncertainty on the cross section5of 6%.Table III shows a summary of all the systematic un-certainties considered. We obtain a total uncertainty of34% with the flux and detector modeling being the mostdominant. The final value of the electron neutrino andantineutrino CC total cross section on argon is therefore, (cid:104) σ (cid:105) = 6 . ± .
51 (stat.) ± .
33 (sys.) × − cm nucleon . This result is consistent with the
GENIE predictionwithin statistical and systematic uncertainties as shownin Fig. 16.
VII. CONCLUSIONS
We presented the measurement of the flux-averagedinclusive electron neutrino and antineutrino charged-current total cross section on argon using the Mi-croBooNE detector and the NuMI beam at Fermilab.For ν e + ¯ ν e energies above 250 MeV and an aver-age neutrino flux energy of 905 MeV calculated byapplying this threshold, we find the cross section tobe 6 . ± . ± . × − cm / nucleon,which is in agreement with the predictions from GENIE and
NuWro . This is the first such measurement per-formed in a large-scale LArTPC and the first one from a LArTPC placed on the surface. It is also the first mea-surement from an off-axis beam at MicroBooNE, withneutrinos arriving with a minimum angle of 8 ◦ relativeto the NuMI neutrino beamline direction. Using thelargest sample of electron-neutrino interactions on ar-gon acquired to date, consisting of 214 selected ν e and¯ ν e CC events with a purity of 38.6%, we demonstratethe electron-photon dE/dx separation power of LArT-PCs using a fully-automated analysis chain. The mea-surement techniques presented here will be of immediateuse for electron-neutrino appearance experiments such asthe SBN program and DUNE.
VIII. ACKNOWLEDGMENTS
This document was prepared by the MicroBooNE col-laboration using the resources of the Fermi National Ac-celerator Laboratory (Fermilab), a U.S. Department ofEnergy, Office of Science, HEP User Facility. Fermilab ismanaged by Fermi Research Alliance, LLC (FRA), act-ing under Contract No. DE-AC02-07CH11359. Micro-BooNE is supported by the following: the U.S. Depart-ment of Energy, Office of Science, Offices of High EnergyPhysics and Nuclear Physics; the U.S. National ScienceFoundation; the Swiss National Science Foundation; theScience and Technology Facilities Council (STFC), partof the United Kingdom Research and Innovation; andThe Royal Society (United Kingdom). Additional sup-port for the laser calibration system and cosmic ray tag-ger was provided by the Albert Einstein Center for Fun-damental Physics, Bern, Switzerland. [1] M.A. Acero et al. (NOvA), “New constraints on oscil-lation parameters from ν e appearance and ν µ disappear-ance in the NOvA experiment,” Phys. Rev. D , 032012(2018), arXiv:1806.00096 [hep-ex].[2] B. Abi et al. , “Long-baseline Neutrino OscillationPhysics Potential of the DUNE Experiment,” (2020),arXiv:2006.16043 [hep-ex].[3] K. Abe et al. (Hyper-Kamiokande), “Hyper-KamiokandeDesign Report,” (2018), arXiv:1805.04163 [physics.ins-det].[4] M. Antonello et al. (MicroBooNE, LAr1-ND, ICARUS-WA104), “A Proposal for a Three Detector Short-Baseline Neutrino Oscillation Program in the Fermi-lab Booster Neutrino Beam,” (2015), arXiv:1503.01520[physics.ins-det].[5] R. Acciarri et al. , “Design and construction of the Mi-croBooNE detector,” Journal of Instrumentation ,P02017 (2017).[6] B. Abi et al. , “Deep Underground Neutrino Experiment(DUNE), Far Detector Technical Design Report, VolumeII: DUNE Physics,” (2020), arXiv:2002.03005 [hep-ex].[7] R. Acciarri et al. (ArgoNeuT), “First measurement ofelectron neutrino scattering cross section on argon,”Phys. Rev. D , 011101 (2020), arXiv:2004.01956 [hep- ex].[8] J. Blietschau et al. , “Total cross sections for ν e and ¯ ν e interactions and search for neutrino oscillations and de-cay,” Nuclear Physics B , 205 – 219 (1978).[9] K. Abe et al. (T2K Collaboration), “Measurement of theInclusive Electron Neutrino Charged Current Cross Sec-tion on Carbon with the T2K Near Detector,” Phys. Rev.Lett. , 241803 (2014).[10] J. Wolcott et al. (MINERvA Collaboration), “Measure-ment of electron neutrino quasielastic and quasielasticlikescattering on hydrocarbon at
Neutrino Flux Prediction for the NuMIBeamline , Ph.D. thesis, William-Mary Coll. (2016).[22] C. Andreopoulos, C. Barry, S. Dytman, H. Gallagher,T. Golan, R. Hatcher, G. Perdue, and J. Yarba, “TheGENIE Neutrino Monte Carlo Generator: Physics andUser Manual,” arXiv:1510.05494 [hep-ph].[23] GENIE v2.12.2 was used in this analysis.[24] D. Heck, J. Knapp, J.N. Capdevielle, G. Schatz, andT. Thouw, “CORSIKA: A Monte Carlo Code to Sim-ulate Extensive Air Showers,” Forschungszentrum Karl-sruhe Report