Cross Section Uncertainties in the NOvA Oscillation Analyses Aaron - PowerPoint PPT Presentation

Cross Section Uncertainties in 
 the NOvA Oscillation Analyses Aaron Mislivec University of Minnesota 1

The NOvA Experiment NO ν A Off-axis, long-baseline neutrino oscillation experiment in the NuMI MINER ν A neutrino beam at Fermilab 2

ν ν μ 𝟑 𝜠𝒏 𝟒𝟑 ν 𝟑 θ δ 𝜠𝒏 𝟒𝟑 𝑁𝑏𝑡𝑡 𝐼𝑗𝑓𝑠𝑏𝑠𝑑ℎ𝑧 ν 𝟑 𝜠𝒏 𝟑𝟐 ν 𝟑 𝜠𝒏 𝟓𝟐 θ θ ν ν μ NOvA Physics Goals ν τ NC Coherent Pion Production Measurement c Long Baseline Neutrino Oscillation Measurements: – ν – • ν μ disappearance 
 Non-Oscillation Measurements: ν e appearance (±30% matter effect) • Cross sections (near detector) - θ 23 , Δ m 232 , δ CP , Mass Hierarchy • Supernova detection • NC disappearance • Exotic phenomena - Sensitive to Sterile Neutrinos - Magnetic monopoles - θ 24 , θ 34 , Δ m 241 - Neutrino magnetic moment 3

NOvA Detectors Functionally identical ND and FD • Same active materials and readout • No A-extrapolation between detectors • ND & FD correlations in cross sections, event selection, and reconstruction 4

NOvA Detectors Sampling Calorimeters (Near and Far) Far Detector PVC Extrusions filled with liquid scintillator - 
 • 14 kton, 344k channels • mineral oil + 5% pseudocumene 810 km from source • Near Detector WLS fiber collects and transports light to APD • 0.3 kton, 20k channels • Optimized for electron ID: Low-Z, 62% active • 1 km from source • 1 rad. length = 38cm (6 cell depths, 10 cell widths) • 5

NOvA Event Topologies p ν μ μ p 1 radiation length = 38cm ν e e (6 cell depths, 10 cell widths) π γ π 0 ν p γ 1m 1m 3 10 2 10 10 q (ADC) 6

Event Selection Learned varia+ons on the JINST 11 P09001 Input Image original image (2016) ν e ν μ ν τ NC Cosmic Events classified with Convolutional Visual Network (CVN) Events treated as images • Successive layers learn topological features • “Feed forward” neural network at end maps to event classes • ν μ analysis identifies μ track using a kNN • Inputs: track length, dE/dx, scattering, fraction of non-track planes 7

Energy Reconstruction NOvA Simulation ν e ν μ ν e FD MC A.U. (Area normalized) Arbitrary units CC ν µ Bkgd. Total − 1 − 0.5 0 0.5 (True - Reco)/True ν μ CC: 
 ν e CC: 
 E ν = E μ + E had E ν = f(E e , E had ) Δ E ν ~ 9% Δ E ν ~ 11% Calorimetric (not kinematic) E ν reconstruction 8

NOvA Neutrino Event Generator ND ν μ CC ND ν μ CC NOvA Preliminary NOvA Preliminary 20 2.85 10 P.O.T. 20 × 2.85 10 P.O.T. × 60000 60000 NOvA ND Data NOvA ND Data First Second MEC QE Analysis Analysis QE 40000 Events Events 40000 RES Non-RES 1 π RES DIS ✕ 0.5 DIS 20000 20000 Other Other 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Visible E (GeV) Visible E (GeV) had had NOvA Simulation 3 10 × 500 GENIE QE (+RPA) GENIE 2.12.2 with the following modifications: Empirical MEC 400 Addition of GENIE Empirical MEC scaled up 20% • Valencia MEC GENIE RES 300 Events Neutrino non-RES 1 π scaled down 50% per • deuterium data 200 CC QE RPA from Valencia & R. Gran 
 • 100 (Phys. Rev. D 88, 113007) 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 True q (GeV) 0 9

Cross Section Uncertainties Utilize GENIE’s standard systematics suite: Primary process ( e.g. , CC QE, RES M A ) • Hadronization • FSI • ND ν μ CC NO A Preliminary ν 6 10 × NO ν A specific uncertainties: 0.14 Simulated Selected Events Simulated Background 5% on CCQE M A per deuterium data • 0.12 Data Shape-only 1- syst. range σ CC QE RPA suppression & enhancement • 20 ND area norm., 8.09 10 POT 0.1 × Data mean: 0.31 GeV MC mean: 0.31 GeV Events (R. Gran, arXiv:1705.02932) 0.08 CC RES RPA f(Q 2 ) off → on (R. Gran) • 0.06 0.04 50% norm. uncertainty on DIS N π for • MEC 0.02 DIS 1.7 < W < 3.0 GeV MEC… 0 0.2 0.4 0.6 0.8 1 • Hadronic Energy Fraction 10

MEC Uncertainties NOvA Simulation 3 Nieves et al. MEC (GENIE) Empirical MEC `Shape' ratio to Empirical MEC Martini et al. MEC (PRC 80, 065501) 2.5 Emp. MEC q QE q → 0 0 Megias et al. MEC (PRD 94, 093004) Arbitrary units Emp. MEC q RES q → 2 Uncertainty envelope 0 0 1.5 1 0.5 0 0 0.2 0.4 0.6 0.8 1 1.2 0 2 4 6 8 True q (GeV) 0 E (GeV) ν E ν shape from model comparisons • MEC q 0 shape → QE, RES q 0 shapes • Initial state np fraction from model comparisons: 
 • València via GENIE vs. SuSA-MEC via PRC94, 054610 np 0 . 7 ≤ np + nn ≤ 0 . 9 11

MEC q 0 NOvA Simulation NOvA Simulation 25000 Near Detector Far Detector True CC MEC only ν Empirical MEC µ 20000 P.O.T. All MEC uncertainties Emp. MEC q QE q → 0 0 15000 20 Arbitrary units Emp. MEC q RES q → 10 0 0 × Events / 9 10000 5000 0 5 0 0.2 0.4 0.6 0.8 1 Ratio 4 True q (GeV) 3 0 2 1 0 0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 True q (GeV) True q (GeV) 0 0 The MEC q0 shape is the largest cross section | 4 E | = π 2 L Δ m 2 systematic in the 2017 ν μ disappearance and 
 2 2 θ ν e appearance results: sin migrates events near ν μ oscillation dip • effect on selection efficiency larger for ν e than ν μ • 12

MEC q 0 NOvA Simulation NOvA Simulation NOvA Simulation 25 25000 Far Detector Far Detector True True CC MEC only CC MEC only ν ν Empirical MEC µ µ 20 20000 P.O.T. P.O.T. All MEC uncertainties All MEC uncertainties Emp. MEC q QE q → 0 0 15 15000 20 20 Arbitrary units Emp. MEC q RES q → 10 10 0 0 × × Events / 9 Events / 9 10 10000 5 5000 0 0 5 5 0 0 0.2 0.2 0.4 0.4 0.6 0.6 0.8 0.8 1 1 Ratio Ratio 4 4 True q True q (GeV) (GeV) 3 3 0 0 2 2 1 1 0 0 0 0.2 0.4 0.6 0.8 1 1.2 0 0 0.2 0.2 0.4 0.4 0.6 0.6 0.8 0.8 1 1 True q (GeV) True q True q (GeV) (GeV) 0 0 0 The MEC q0 shape is the largest cross section | 4 E | = π 2 L Δ m 2 systematic in the 2017 ν μ disappearance and 
 2 2 θ ν e appearance results: sin migrates events near ν μ oscillation dip • effect on selection efficiency larger for ν e than ν μ • 13

Near-to-Far Extrapolation 1 6 2 3 4 5 1. ND Data E ν Spectrum 4. P( ν x → ν y ) 2. ND Reco. → True E ν 5. FD True → Reco. E ν 3. FD / ND Event Ratio 6. FD Oscillated Prediction in True E ν Bins Systematic shifts affect 2-6 14

Near-to-Far Extrapolation N near ( E reco ) = Φ ( E true ) × � ( E true , A ) × R ( E true ) × ✏ ( ... ) ν ν ν ν N far ( E reco ) = P osc ( E true ) × Φ ( E true ) × � ( E true , A ) × R ( E true ) × ✏ ( ... ) ν ν ν ν ν ND data + extrapolation leverages ND ↔ FD correlations in constraining the FD prediction 15

Test Extrapolation: CC RES M A ν μ CC Selection ν μ CC Selection NOvA Simulation NOvA Simulation 1.2 10 ± σ extrap. ND shift in MaCCRES + σ shift in MaCCRES FD minus ND Residual difference (%) Ratio to nominal MC 1.1 5 FD shift MaCCRES ± σ - shift in MaCCRES FD minus ND σ 1 0 Residual 0.9 5 − 0.8 10 − 0 1 2 3 4 5 0 1 2 3 4 5 Reconstructed neutrino energy (GeV) Reconstructed neutrino energy (GeV) Replace ND data with ND MC under CC RES M A shift • Extrapolate and compare with FD MC under same shift • Shifted ND MC + extrapolation accounts for most of the • shift’s effect in the FD 16

Test Extrapolation: MEC q 0 Shape ν μ CC Selection ν μ CC Selection NOvA Simulation NOvA Simulation 1.3 10 1.2 Residual difference (%) Ratio to nominal MC 5 1.1 1 0 Residual 0.9 5 − + shift in MEC q0 shape FD minus ND σ ± σ extrap. ND shift in MEC q0 shape 0.8 - shift in MEC q0 shape FD minus ND σ FD shift MEC q0 shape ± σ 0.7 10 − 0 1 2 3 4 5 0 1 2 3 4 5 Reconstructed neutrino energy (GeV) Reconstructed neutrino energy (GeV) Replace ND data with ND MC under MEC q 0 Shape shift • Extrapolate and compare with FD MC under same shift • Shifted ND MC + extrapolation accounts for most of the • shift’s effect in the FD 17

Resolution Binning 3 10 × 1 300 0.8 Quantile 4 ν 0.6 200 / E had. Quantile 3 E 0.4 100 Quantile 2 0.2 Quantile 1 0 0 0 1 2 3 4 5 Reconstructed Neutrino Energy (GeV) • 2017 ν μ disappearance analysis extrapolates in bins of E had / E ν • Bins correspond to E ν resolution ( Δ E μ ~ 3%, Δ E had ~ 30%) • High-resolution bin helps resolve oscillation dip • Resolution binning further constrains FD prediction 18