RRecent Cross Section Work from NOvA
J. Wolcott, for the NOvA Collaboration
Department of Physics and Astronomy, Tufts University574 Boston Ave., Medford, MA 02155 USA
The NOvA experiment is an off-axis long-baseline neutrino oscillation experiment seekingto measure ν µ disappearance and ν e appearance in a ν µ beam originating at Fermilab. Inaddition to measuring the unoscillated neutrino spectra for the purposes of predicting theoscillated neutrino spectrum in the far detector, the 293-ton near detector also enables high-statistics investigation into neutrino scattering in numerous reaction channels. We discuss thevarious near detector analyses currently in progress, including inclusive measurements of bothelectron and muon neutrino charged-current interactions and efforts to constrain the off-axisNuMI flux using the elastic scattering of neutrinos from atomic electrons. Over the course of the last three decades, neutrino oscilliation experiments have sought to usethe quantum-mechanical properties of the neutrino as a probe of the fundamental nature of thelepton family. Since the weak-force coupling of neutrinos to other particles is extremely small,terrestrial neutrino oscillation experiments, such as NOvA, typically construct large detectorsfrom materials composed of heavy nuclei in an effort to maximize the neutrino interaction rate.But the intractibility of calculating the dynamics of nucleons within the nucleus in the low-energy limit of the strong force introduces significant uncertainties into the reaction predictionsused in measurements made with these detectors. Even in the two-detector paradigm usedby NOvA and other experiments, in which a detector close to the neutrino source (the neardetector, ND) is used to constrain the product of interaction cross section models and the fluxprediction (which is then extrapolated to the far detector, FD, where oscillations are observed),direct measurements of neutrino interaction cross sections on the target materials are extremelyvaluable for constraining and choosing between models.The 293-ton NOvA near detector is an ideal instrument to use for this sort of cross sectionmeasurement for several reasons. First, its location 14.6 mrad off-axis in the Fermilab NuMIneutrino beam it samples yields a narrow neutrino energy spectrum centered on 2 GeV, producingan event sample rich in interaction types (including copious examples of quasielastic scattering,baryon resonance production, and deep inelastic scattering) and exhibiting multiple kinds ofnuclear effects (including coherent meson production, multi-nucleon scattering, and final-statehadron rescattering). Second, the detector itself is a mostly-active, fine-grained, segmentedtracking calorimeter constructed of PVC cells filled with liquid scintillator with excellent spatialand energy resolution. We present status reports on a number of measurements currently inprogress using the NOvA ND. a r X i v : . [ h e p - e x ] N ov uon ID E v en t s Simulated selected eventsSimulated backgroundData syst. range s Full 1- POT · ND POT norm., 3.72
NOvA Preliminary
Figure 1: Predicted distribution of muon particle identification classifier described in the textfor tracks ( ν µ CC signal, red line; other predicted reactions, blue line) compared to ND data(black points). Events with Muon ID > . ν µ CC events. ν µ charged-current inclusive scattering During its lifetime the NOvA ND is expected to record an immense sample of charged-current(CC) interactions of muon neutrinos on the liquid scintillator ( ν µ CH → µ − X ) ultimately num-bering in the millions. The statistical power of this sample offers an unprecedented opportunityboth to verify the basic nucleon-level models for CC reactions in detail and to examine therelevant nuclear effects near E ν = 2 GeV; this energy range has previously been explored mostlyin light bubble chamber experiments in measurements reporting only total cross sections. The comparatively long lifetime and clean ionization profile of muons make the lepton kinematicsin CC reactions particularly amenable to precise measurement. NOvA reconstructs muons astracks and separates them from the hadronic background using a k -nearest neighbors (kNN)algorithm trained with four variables: the track length, the longitudinal energy profile ( dE/dx ),the scattering along the track, and the fraction of energy in the neutrino event associated withthe track. The distribution of the resulting classifier is shown in figure 1; the observed datadistribution is well-described by the prediction. Events which have Muon ID > . ν µ CC events. The predicted resolutions in both muon energy andangle for this sample are very good (averaged over the sample, 50 MeV → .
8% and 4 ◦ → . Because NOvA is a tracking calorimeter, it offers detailed reconstruction of the hadronic part of ν µ CC interactions as well. Here the effect of the nucleus on neutrino interactions takes centerstage; we observe clear evidence for an extra reaction type beyond those predicted by defaultGENIE 2.10.4 lying in between the quasielastic (QE) and baryon resonance (RES) channels inmomentum transfer variables (where E µ and E had are the reconstructed muon and non-muonenergies in the system): q = E had E ν = E µ + E had Q = 2 E ν (cid:0) E µ − p µ cos( θ µ ) − M µ (cid:1) | (cid:126)q | = Q + q (1)This is illustrated in figure 3. Inspired by recent work in neutrino scattering , we interpretthis absence as the lack of a model for a two-particle, two-hole (2p2h) process, where the neutrinoscatters from a nucleus and ejects two of the nucleons (which were previously in some kind ofcorrelated state) together.GENIE 2.10.4 does ship with an “optional” (not enabled by default), mostly empiricalmodel for 2p2h reactions , “Empirical MEC a ” (previously called “Dytman MEC,” after itsauthor). Because it is unclear whether the kinematic assumptions built in to this model thatwere constructed largely from observations at lower E ν should extrapolate correctly to NOvA’sneutrino energy range, we further modify this model as follows:1. We reverse the linear turn-off of the cross-section between 1 and 5 GeV (so that the Empir-ical MEC cross section becomes a constant fraction of the QE one) since there are recentindications that 2p2h exists with similar strength at energies above 5 GeV.
2. We reverse the fraction of scattering from neutron-neutron and neutron-proton pairs inthe model to 20% and 80%, respectively, based on indications from electron scattering and expectations from theoretical expectations in neutrino scattering . b a Meson Exchange Currents (MEC) are one predicted class of 2p2h which have generated intense theoreticalinterest in recent years. Good summaries of the various strategies can be found elsewhere. b The typo that led to the need for this correction has been corrected in GENIE 2.12. eco “q ” (=E had,vis ) E ve n t s q |/GeV < 0.9 0.9 < | q |/GeV < 10.7 < | q |/GeV < 0.80.4 < | q |/GeV < 0.5 0.5 < | q |/GeV < 0.6 0.6 < | q |/GeV < 0.70.2 < | q |/GeV < 0.3 0.3 < | q |/GeV < 0.40.1 < | q |/GeV < 0.2 NOvA ND Data
NOvA Preliminary
Figure 3: Simulated reactions from GENIE 2.10.4 broken down by true reaction type (coloredareas) compared to NOvA ND data (black points) in reconstructed energy transfer q , dividedinto slices of momentum transfer | (cid:126)q | (both described in the text).3. We apply a momentum-transfer-dependent weight derived from our ND data as describedin the next paragraph.To construct weights that constrain the Empirical MEC to better fit our observed data, we firstexamine the data excess in | (cid:126)q | (effectively the difference of the integrals of data and simulation c in each panel of figure 3). We reweight the Empirical MEC such that it agrees with the dataexcess in this variable. To set the fourth component of the four-momentum transfer, q , we fix itto the shape of the predicted q distribution in each bin of | (cid:126)q | taken from the GENIE quasielasticchannel. This somewhat underestimates the E had in the observed distribution, as illustrated infigure 4b, but the overall agreement relative to the untuned version (figure 4a) is substantiallyimproved. The GENIE 2.10.4 prediction with tuned Empirical MEC is the base prediction forcurrent oscillation analysis efforts, including those discussed elsewhere in this volume. ν e charged-current inclusive scattering Electron neutrinos are expected to undergo the same types of reactions and their interactions areexpected to experience the same types of nuclear effects as ν µ , up to the influence of the differencein the charged lepton masses. Understanding whether this is actually the case is very importantfor oscillation experiments like NOvA, for which the interactions of ν e appearing via oscillationfrom a ν µ beam comprise a critical signal channel. However, at energies around several GeV,until recently it has been challenging to accumulate enough ν e interactions to make statisticallysignificant measurements. The very intense NuMI beam used by NOvA, on the other hand, hasabout a 1% admixture of ν e , opening the door for high-statistics investigation.For a cross section analysis, NOvA begins selecting ν e interactions using a likelihood classi-fier constructed from the longitudinal energy profiles of various particle templates; the perfor-mance of this classifier (after a baseline selection requiring containment and rejecting especiallyminimum-ionizing tracks to reject ν µ CC), and the selection cut made on it, is illustrated in c After applying the correction to non-resonant 1 π production from neutrons suggested by Rodrigues et al. (GeV) had Visible E E v en t s NOvA ND DataQERESDISOther P.O.T. · NOvA Preliminary (a) No MEC (GeV) had
Visible E E v en t s NOvA ND DataMECQERESDISOther
P.O.T. · NOvA Preliminary (b) Tuned Empirical MEC
Figure 4: Visible hadronic energy distributions in ND selected ν µ CC events before (left) andafter (right) the addition of GENIE 2.10.4 “Empirical MEC” constrained as discussed in thetext.
Electron Likelihood ID E v en t s Data e n CC m n / m n CC NCROCK e n CC A Preliminary n NO POT · (a) - - E v en t s A Preliminary n NO Data e n CC ROCK e n CC NC m n CC POT · BDT output - - D a t a / M C (b) Figure 5: Performance of variables used to select ν e CC interactions, as described in the text:likehood classifer preselector (left); final boosted decision tree output (right). The inset in theright plot is a zoom showing only the distribution above the cut.figure 5a. Once this electromagnetic cascade-enhanced sample is obtained, further purification isaccomplished using a boosted decision tree using shower shape variables (both longitudinal andtransverse); its performance is shown in fig. 5b. Studies of sideband regions in these variablesare underway in order to better constrain the predicted backgrounds and understand what thedominant uncertainties will be for a cross section. ν − e elastic scattering The neutrino flux prediction is an essential ingredient to any cross section measurement be-cause it represents the normalization coefficient as a function of neutrino energy; traditionallyflux uncertainties comprise the largest source of error for extracted cross sections. This owesprimarily to the fact that ab initio calculations of horn-focused neutrino beams like NuMI de-pend on predictions of the strong-force dynamics of protons colliding (and re-interacting) withcomplex molecular targets like graphite, which are difficult. However, it is in principle possibleto constrain the flux prediction using an in situ measurement of a neutrino scattering process rad · (GeV q E P O T · E v en t s / . n MC e n MC Beam CC + NC m n MC NOvA Simulation (a)
Electron Energy(GeV) P O T · E v en t s / . n MC e n MC Beam CC + NC m n MC NOvA Simulation (b)
Figure 6: Electron Eθ (left), where events with Eθ < .
005 GeV × rad are considered signalcandidates, and resulting electron energy distribution (right) in the ν + e → ν + e scatteringanalysis.with a well-understood cross section. Because of the complexities of neutrino interactions withnuclei, however, purely leptonic processes like ν + e → ν + e scattering (neutrinos with atomicelectrons) are the the reactions most amenable to use in this fashion. Unfortunately, the crosssection of ν + e → ν + e scattering is suppressed relative to nucleon scattering by the ratio ofthe electron to nucleon masses and other kinematic factors, resulting in σ ν − e /σ ν − N ∼ − .Therefore statistics are typically low in this channel.As in the ν e CC case, NOvA uses two PID classifiers to identify candidate electron showersfor this analysis: one that distinguishes between electromagnetic showers and other backgrounds,and one that specifically distinguishes between electron-induced and photon- or neutral pion-induced showers. After selections on these variables, we employ a cut at 0 .
005 GeV × rad on thekinematic variable E e θ e , which is limited to very small values by the kinematics of the interactionitself, to further enrich the signal; this is illustrated in figure 6a. The resulting electron energyspectrum, which will be used to constrain the flux, is shown in figure 6b. Currently effortsare being devoted to quantifying the size of uncertainty in the signal efficiency and backgroundcross section and flux predictions. It is expected that this technique will constrain the fluxnormalization to around 10% uncertainty. Acknowledgments
NOvA is supported by the US Department of Energy; the US National Science Foundation; theDepartment of Science and Technology, India; the European Research Council; the MSMT CR,Czech Republic; the RAS, RMES, and RFBR, Russia; CNPq and FAPEG, Brazil; and the Stateand University of Minnesota. We are grateful for the contributions of the staffs of the Universityof Minnesota module assembly facility and NOvA FD Laboratory, Argonne National Laboratory,and Fermilab. Fermilab is operated by Fermi Research Alliance, LLC under Contract No. De-AC02-07CH11359 with the US DOE.1. P. Rodrigues et al. , Eur. Phys. J.
C76 , 474 (2016).2. P. Rodrigues et al. (MINERvA Collaboration),
Phys. Rev. Lett. , 071802 (2016).3. T. Katori,
AIP Conf. Proc. , 030001 (2015).4. A. Ankowski et al. , Phys. Rev.
D93 , 113004 (2016).5. R. Subedi et al. , Science , 1476 (2008).6. I. Ruiz-Simo et al. , Phys. Lett.