Direct Measurement of the NuMI Flux with Neutrino-Electron Scattering in MINERvA Jaewon Park University of Rochester On behalf of MINERvA Collaboration December 20, 2013 Fermilab Joint Experimental-Theoretical Seminar

2 Outline Neutrino experiments and their fluxes ν-e scattering: signal Event reconstruction Backgrounds and how to remove them Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

3 Oscillation Experiment Strategy In fact the flux doesn’t just decrease like 1/L 2 –Oscillations –Near detector sees line source, far detector sees point source Far detector sample is always very different from near detector sample Near Detector (ND) Far Detector (FD) or ? 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Isn’t this just 1/L 2 ? N: events in data B: Background  : Efficiency A: Acceptance  : Cross section

4 Needs of Precision Oscillation Experiments 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Precise measurement of oscillation parameters is the key to answer important questions like neutrino mass hierarchy and CP violation To achieve the highest precision, we need: –(High intensity beam) × (big detector) × (long operation) –Low uncertainties on flux prediction –Better understanding of neutrino interactions More accurate cross-section (May 10 JETP by MINERvA) Understanding nuclear effects (October 11 JETP by MINERvA) Detailed understanding of background interactions This talk is about a method to constrain or measure the neutrino flux using neutrino-electron scattering –This helps to reduce flux normalization uncertainties on MINERvA’s absolute cross-section measurements –This technique can be used in future high intensity beam experiments to measure the flux

5 Neutrinos from an Accelerator 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Near Detector Far Detector proton Horn 1 Horn 2 Target Decay pipeRock Neutrino beam is generated from a decay of secondary or tertiary particles  Hard to control beam itself, too hot to measure in situ Flux has large uncertainties due to poor knowledge of hadron production Non-perturbative QCD governs it  Difficult to calculate from basic principles ~15-30% normalization uncertainties on flux

6 Neutrinos from an Accelerator 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Near Detector Far Detector proton Horn 1 Horn 2 Target Decay pipeRock Kaon and muon decays are main source of electron neutrinos

7 Constraining flux with Hadron Production Data p π n target ν decay pipe 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Hadron production primarily function of x F =pion/proton momentum ratio and p transverse – Use NA49 measurements –Scale to 120 GeV using FLUKA (simulation) –Check by comparing to NA61 data at 31 GeV/c [Phys.Rev. C84 (2011)034604] Use MIPP (120GeV protons) for K/π ratio Particle production xFxF Reference NA49 GeV π±π± <0.5 Eur.Phys.J. C49 (2007) 897 K±K± <0.2 G. Tinti Ph.D. thesis p <0.9 Eur.Phys.J. C73 (2013) 2364 MIPP 120 GeV K/π ratio A. Lebedev Ph.D. thesis

8 NA49: pC → 158 GeV NA49 data vs. GEANT4 Uncertainties 7.5% systematic 2-10% statistical π + which make a ν μ in MINERvA focusing peak f(x F,p T ) = E d 3 σ/dp 3 = invariant production cross-section high energy tail 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

9 Need more than Hadron Production Measurements Hadron Production measurements don’t tell the whole story, only 70% –Some pion production is out of range of Hadron Production data –Tertiary production of neutrinos also important (n, , K L,S ) Beamline geometry and focusing elements contribute uncertainties 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

10 Special Runs to Understand Flux MINERvA integrated 10% of our total neutrino beam exposure in alternate focusing geometries: –Changed horn current –Changed Target Position Purpose is to disentangle focusing uncertainties from hadron production uncertainties –Different geometry focuses different parts of xF p T space, but same horn geometry and current MINERvA does this by using low hadron energy  charged current events, where energy dependence of cross section is very well understood 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Normal Running Target Moved upstream Pion Phase Space Neutrinos at MINERvA xFxF xFxF P t (GeV/c) Inclusive Event Spectra

11 Flux constraint using Near Detector Cross-section uncertainty goes into flux uncertainty MINERvA Flux uncertainty goes into cross-section uncertainty Neutrino Flux and Cross-section Measurement 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Flux and cross-section are anti-correlated with given Near Detector constraint σ (Cross Section) Φ (Flux) N: Events  : Efficiency A: Acceptance  : signal cross section Measurement uncertainty

12 Known Interaction (Standard Candle) ν-e scattering is well known interaction we can use to constrain the neutrino flux Flux constraint using ND Cross-section uncertainty goes into flux uncertainty ν-e Scattering 20 December 2013Jaewon Park, U. of Rochester FNAL JETP σ (Cross Section) Φ (Flux)

13 Outline Neutrino experiments and their fluxes ν-e scattering: signal Event reconstruction Event selection Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

14 ν-e Scattering History First unambiguous neutral current (Gargamelle) Solar ν oscillation measurement using ν-e scattering (SNO) Coherent forward scatterings Additional interaction for All flavor Matter effect is due to charged current ν e scattering on electrons, only for ν e Electroweak theory 20 December 2013 Jaewon Park, U. of Rochester FNAL JETP

15 Neutrino Scattering on Nucleon Let’s use well-known reaction to measure the flux Standard electroweak theory predicts it precisely –Point-like scattering Very small cross section (~1/2000 of ν-nucleon scattering) –Low center of mass energy due to light electron Very forward electron final state (Experimental signature) Good angular resolution is important to isolate the signal Very forward single electron final state νe→ νe candidate event Electron 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

16 ν-e Scattering E > 0.8 GeV –High background rate and tough reconstruction at low energy Predict 147 signal events for 3.43×10 20 Protons On Target (POT) –~100 events when you fold in (reconstruction + selection) efficiency of ~ 70% Not a large sample in low energy run but still useful to constrain absolute flux 20 December 2013Jaewon Park, U. of Rochester FNAL JETP FLUX e Scattering Events G F and θ W : well-known electroweak parameters

17 Signal Events 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Signal is mixture of in LE-FHC (neutrino beam) ~100 signal events for 3.43E20 POT Can’t distinguish neutrino type Still useful to constrain the flux –Total events: Constraint for integrated flux –Electron spectrum: Constraint for flux shape E>0.8 GeV E<0.8 GeV is not used Large background Tough reconstruction For remainder of talk, means and

18 Outline Neutrino experiments and their fluxes ν-e scattering: signal Event reconstruction Backgrounds and how to remove them Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

19 Data and Simulation Samples All Low Energy neutrino data is used for the analysis: more than previous analyses shown to date (3.43 × Protons on Target) Time-dependent effects (calibrations, accidental activity) included in the simulation MINERvA ran in three kinds of beam: Low Energy neutrino Low energy anti-neutrino “Special Runs”: higher energy runs to constrain flux model 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Thank you! for the excellent ν beam

20 MINERvA Detector 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Inner Detector 5m 3.5m 4m Outer Detector (steel + scintillator) Nuclear Targets (C, Pb, Fe, H 2 O) Tracker (Active target) Electromagnetic Calorimetry Hadronic Calorimetry

21 MINOS Near Detector (muon spectrometer) Inside the Detector −60° +60° x v x u Nuclear Target Tracker Ecal Hcal Number of channels: ~31k Number of scintillator plane: 128 Scintillator plane (X, U, V stereo angle) Pb Tracking Ecal (lead absorber + tracking plane) Fe Tracking Hcal (steel absorber + tracking plane)

22 Detector Technology 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Extruded plastic scintillator with wavelength shifting fiber readout 64 channel multi-anode PMT for photo-sensor Wavelength shifting fiber 8×8 pixels 64 channel multi-anode PMT Scintillator strip 17 mm 16 mm Position resolution: ~3mm 2.1m 127 strips into a plane 2.5 m

23 ν + e - → ν + e - candidate event X-ViewU-ViewV-View Data run: 2157/12/1270/2 Fiducial volume 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

24 Single Electron Reconstruction Nuclear Target Region (He,C/H2O/Pb/Fe) HCAL ECAL Track-like Shower-like Track-like part (beginning of electron shower) gives good direction 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

25 Single Electron Reconstruction Nuclear Target Region (He,C/H2O/Pb/Fe) HCAL ECAL Track-like part (beginning of electron shower) gives good direction 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Shower cone

26 Critical Variables for Signal Electron Identification –Must discriminate from photons Electron Energy Measurement Electron Angular Measurement 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

27 Electron Photon Discrimination using dE/dx 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Electromagnetic shower process is stochastic –Electron and photon showers look very similar Photon shower has twice energy loss per length (dE/dx) at the beginning of shower than electron shower –Photon shower starts with electron and positron Electron-induced electromagnetic shower Photon -induced electromagnetic shower MINERvA Preliminary

28 Energy and Angle Reconstruction 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Energy resolution ~ 5% Projected angle resolution ~ 0.3 degree (2 sigma truncated RMS) Precise angle reconstruction is critical to separate ν e elastic scattering from background –Lower energy angular resolution is worse due to multiple scattering Using simulated signal MINERvA Preliminary

29 Outline Neutrino experiments and their fluxes ν-e scattering: signal and backgrounds Event reconstruction Backgrounds and how to remove them Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

30 Initial Background Rejection -e scattering is very rare, even for interactions: Simple cuts can eliminate most background events while keeping high fraction of signal events –Obvious muon-like event rejection –Upstream energy rejection Removes neutrino interactions upstream of detector that make  20 December 2013Jaewon Park, U. of Rochester FNAL JETP e Quasi-elastic (CCQE) Coherent  0 Most Events (  Charged or neutral Current) Rare but hard to reject:

31 Background Events 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Use Eθ 2 to select very forward signal Electron neutrino fraction in flux is small ~ 1%. electron proton z x MeV MC If recoil nucleon is not observed, it looks similar to signal Angles of electron have wide spread while signal is very forward NC-coherent π 0 NC-resonant π 0 Neutral current single π 0 1. Small opening angle between two gammas π 0 (1.1 GeV) γ (67 MeV ) π 0 (7.5 GeV) 2. One of gammas is not observed in the detector Simulated event Also, photon has wide spread of angle In addition, use dE/dx to reject

32 Example: Neighborhood Energy Neighborhood energy = energy around shower cone Small neighborhood energy means isolated shower 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Signal × 200 Not Full Sample MINERvA Preliminary 5 cm Neighborhood Shower cone

33 Event Selection 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Shower cone Reconstruction Electron Energy>0.8GeV Fiducial cut Other reconstruction quality cuts Signal sample Eθ 2 dE/dx Kinematic constraint on e scattering, using Mandelstam variables: in CM frame in lab frame

34 dE/dx Cut All cuts made on this sample except for the dE/dx cut Neutrino interaction doesn’t always produce only single electron or single photon (from π 0 ) Non-single particle activity affects dE/dx tuned dE/dx<4.5MeV/1.7cm 20 December 2013Jaewon Park, U. of Rochester FNAL JETP MINERvA Preliminary

35 Eθ 2 Cut All cuts but E  2 cut Kinematic limit for signal –Eθ 2 < 2m e Clean separation of signal tuned 20 December 2013Jaewon Park, U. of Rochester FNAL JETP MINERvA Preliminary

36 Electron Spectrum after all cuts tuned 20 December 2013Jaewon Park, U. of Rochester FNAL JETP True electron energy (signal only) Reconstructed electron energy MINERvA Preliminary

37 Outline Neutrino experiments and their fluxes ν-e scattering: signal and backgrounds Event reconstruction Backgrounds and how to remove them Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

38 Backgrounds after all Cuts 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Background prediction is affected by the flux and physics model Cross-section of various neutrino reactions are uncertain –That’s what MINERvA is trying to measure Data-driven background prediction tuning is used to handle the uncertainty of predicted background Sideband Signal Need to know energy spectrum of background MINERvA Preliminary

39 4 Background Processes, 4 Sidebands 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Sideband = Outside of major Eθ 2 and dE/dx cuts (b) region is not used because there are not many events for tuning Further, cut is slightly loosened on sideband so it gets some ν μ CC for tuning purpose dE/dx (MeV/1.7cm) (a) Sideband signal 20 (b) Unused Eθ 2 (GeV∙rad 2 ) 3 Min dE/dx Energy Sideband 1 Sideband 3 Sideband 2 Sideband 4 Sideband 1, 2, 3 (not sideband 4) (Coherent π 0 rich region) No side-exiting muon Narrow shower at beginning Eθ 2 <0.1

40 Sideband Populations 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Coherent  0 Most Events (  Charged or Neutral Current ) Rare but hard to reject: e Charged Current

41 Sideband Tuning Scale three MC components to match to data Minimize χ 2 across 7 histograms 3 parameters tuned in this step Minimize χ 2 across 2 histograms 1 parameter tuned in this step After tuning Before tuning Events with tracks in downstream Hadron Calorimeter are mostly ν μ CC 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Track Length in HCAL (modules) ParameterTuned value 0.83 ± ± ± ± 0.08 ν e ν μ NC ν μ CC COH π 0 MINERvA Preliminary

42 dE/dx and   in Sidebands after tuning 20 December 2013Jaewon Park, U. of Rochester FNAL JETP 4.5 dE/dx (MeV/1.7cm) Sideband Signal Eθ 2 (GeV∙rad 2 ) (a) (b) (c) Unused Both dE/dx and Eθ 2 are well simulated in the sideband region after fitting dE/dx (MeV/1.7cm) Eθ 2 (GeV∙radians 2 ) # Events (Eθ 2 < 0.2) MINERvA Preliminary

43 Outline Neutrino experiments and their fluxes ν-e scattering: signal and backgrounds Event reconstruction Backgrounds and how to remove them Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

44 Systematic Uncertainties Error in background contribution –Flux uncertainties –Cross Section Uncertainties Error in efficiency and Acceptance 20 December 2013Jaewon Park, U. of Rochester FNAL JETP N: events in data B: Background  : Efficiency A: Acceptance  : signal cross section

45 Uncertainty in e  CCQE extrapolation from sideband Previous MINERvA results on  Quasi- elastic process shows that momentum transfer squared (Q 2 QE ) distribution is not what GENIE predicts Phys. Rev. Lett. 111, (2013), Phys. Rev. Lett. 111, (2013). Q 2 QE and Eθ 2 are highly correlated Compare e  background prediction E  2 extrapolation with two different models: one is GENIE, the other is one inspired by MINERvA  data: systematic uncertainty: 3.3% 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

46 Flux and Cross Section Systematic Uncertainties on MC Background Sideband tuning reduced systematic uncertainty on predicted background –Predicted background (before tuning): 38.9 ± 6.2 (stat) ± 10.3 (sys) –Predicted background (after tuning): 32.9 ± 5.3 (stat) ± 5.7 (sys) The tuning didn’t eliminate systematic uncertainty but it gives confidence on background prediction Uncertainty Sources MC background uncertainty [events] Before tuningAfter tuning MC background events MC bkg statistical Total systematic Flux_BeamFocus Flux_NA Flux_Tertiary GENIE CCQE Shape Total December 2013Jaewon Park, U. of Rochester FNAL JETP

47 Reconstruction Systematic Uncertainties Electromagnetic Energy Scale: look at electrons from stopped  decays (Michel): see agreement at 4.2% level, add as systematic uncertainty 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Angular Alignment: look at data- simulation differences in  angles for  CC events with low hadron energy –3 (1) mrad correction in y (x) –uncertainty is ±1mrad MINERvA Preliminary  Charged Current Events with hadron energy<100MeV

48 Reconstruction Uncertainties Source Uncertainty on Source Systematic Uncertainty Beam angle uncertainty  x and  y : ± 1 mrad 1.1% and 1.3% Energy scale4.2%1.9% EM calorimeter energy smearing Additional energy smearing0.0% Absolute Electron Reconstruction Efficiency 2% based on muon studies2.8% All Reconstruction Uncertainties 5.4% Simulation statistics (Bckgd) 6.0% Flux (Bkgd) Beam focusing, Beam tuning 1.3% Cross Section (Bkgd) GENIE, CCQE Shape 6.3% 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

49 Outline Neutrino experiments and their fluxes ν-e scattering: signal and backgrounds Event reconstruction Backgrounds and how to remove them Background Prediction Systematic uncertainty Result and Conclusions 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

50 Result Found: 121 events before background subtraction -e scattering events after background subtraction and efficiency correction: ± 17.0 (stat) ± 9.1 (sys) total uncertainty: 15% Prediction from Simulation: ± 22.9 (flux) –Flux uncertainty: 15.5% 20 December 2013Jaewon Park, U. of Rochester FNAL JETP Observed ν-e scattering events give a constraint on flux with similar uncertainty as current flux uncertainty, consistent with prediction

51 Future Flux Measurement at NuMI Assuming similar signal/background ratio as in Low Energy Run: –Can expect statistical uncertainty = ~2% –Total systematic uncertainty on this measurement: 7% –Total uncertainty = ~7.3% Medium Energy Flux: Now being produced in the NuMI Beamline, as of September 2013 ~20 times the low energy signal sample MINERvA Preliminary

52 Conclusions -e scattering provides an independent flux measurement for ν-nucleon cross-section normalization Uncertainty on ν-e based flux measurement in Low Energy beam is 15% –It is similar size as current flux prediction uncertainty –Will be used as constraint in future MINERvA cross section measurements In Medium Energy run, estimate a 7% uncertainty on total flux –Currently dominant uncertainty is statistical uncertainty This technique could be used in future higher intensity experiments like NOvA and LBNE to provide a precise flux measurement 20 December 2013Jaewon Park, U. of Rochester FNAL JETP

53 (Backup)

54 Neutrino Beam Spread Diameter of decay pipe: ~ 2m 0.5 m transverse position at 500 m distance  1 mrad angle MINERvA Preliminary

55 Data Energy Stability (Michel Electron) Some odd behavior is found between minerva1 and rest of playlists –1 or 2% energy variation Use 4.2% as the systematic uncertainty on energy scale MINERvA Preliminary

56 Sideband 4 (After Tuning) MINERvA Preliminary

57 Background Subtraction MINERvA Preliminary

58 Comparing Observed and Predicted Spectra MINERvA Preliminary