HAPPY 50TH ANNIVERSARY..!
SEARCH FOR STERILE NEUTRINOS AT THE NOVANEAR DETECTOR
SIVA PRASAD K FOR THE NOVACOLLABORATION UNIVERSITYOF HYDERABAD NEW PERSPECTIVES | FERMILAB JUNE 05, 2017 LSND/MINIBOONEEXCESS
2 2 The excess can be interpreted as "Sterile Neutrinos" with ∆m ≥1 eV . In the 2-flavor neutrino oscillation model, the excess can be expressed as 1.27∆m2L P(ν → ν ) = sin2 2θ sin2 µ e E
A. Aguilar et al., Phys. Rev. D 64, 112007 (2001).
Siva Prasad K Short-Baseline Oscillations June 5, 2017 2/32 (3+1) MODEL
For our analysis, we consider (3+1) model. The PMNS matrix is given by νe Ue1 Ue2 Ue3 Ue4 ν1 νµ Uµ1 Uµ2 Uµ3 Uµ4 ν2 = ντ Uτ1 Uτ2 Uτ3 Uτ4 ν3 νs Us1 Us2 Us3 Us4 ν4
νe Appearance Probability
2 2 SBL,3+1 2 2 2 ∆m41L 2 2 ∆m41L P(−) (−) = 4|Uµ4| |Ue4| sin = sin 2θµe sin νµ → νe 4E 4E
νµ Disappearance Probability
2 2 SBL,3+1 2 2 2 ∆m41L 2 2 ∆m41L P(−) (−) = 1−4|Uµ4| (1−|Uµ4| ) sin = 1−sin 2θµµ sin νµ →νµ 4E 4E
2 2 2 2 2 2 sin 2θµe = sin 2θ14 sin θ24 and sin 2θµµ = cos θ14 sin θ24
Siva Prasad K Short-Baseline Oscillations June 5, 2017 3/32 NOVAEXPERIMENT
NOvA (NuMI Off-Axis Electron Neutrino Appearance) is an accelerator based neutrino oscillation experiment. Near Detector 1 km from source, 100 meter deep underground 0.3 kton in mass. Far Detector 810 km from Fermilab on surface. 14 kton in mass. This analysis is done with the near detector alone.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 4/32 NOVADETECTORS
NOvA detectors are made of plastic PVC cells which are glued together in vertical and horizontal planes.
Each cell is filled with mineral oil based scintillator with the liquid scintillant.
Each cell has a Wavelength Shifting Fiber inside to carry light photons to the Avalanche Photodiode read out.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 5/32 NUMIBEAMIN NOVA
Both detectors use the same NuMI beam which peaks at 2 GeV, at 14.6 mrad off-axis.
NuMI beam composition: ≈ 98% νµ’s and ≈ 2% νe’s. NOvA is operating at it’s designed NuMI power, 700 kW. Siva Prasad K Short-Baseline Oscillations June 5, 2017 6/32 WHAT NOVACAN DO?
NOνA Simulation
120
100 νµ total
80 νµ from Kaon
60 νµ from Pion
40 Events / 0.02 (km/GeV) 1E21 POT
7 20 10
0 LSND excess with L/E 0 0.2 0.4 0.6 0.8 L/E (km/GeV)
NOvA Near Detector is well placed to probe the Short-Baseline Neutrino Oscillations with ∆m2 ≥ 1eV2.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 7/32 ANALYSIS STRATEGY
Siva Prasad K Short-Baseline Oscillations June 5, 2017 8/32 JOINT FIT METHOD SENSITIVITIES
NOvA Simulation 102
Comparison of joint fit sensitivities with the independent νµ and νe fits. 10 ) 2
Joint Fit allows for the partial (eV
2 41 m
cancellation of systematic ∆ uncertainties, significantly 1 Sensitivity - 90% C.L. 1 Year improving sensitivity of the analysis. NOvA with single fit NOvA with joint fit NOvA with single fit (Beam+Cross-section) NOvA with joint fit (Beam+Cross-section)
10−1 10−2 10−1 1 2 θ sin 2 14 2 2 2 sin 2θµe = sin 2θ14 sin θ24.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 9/32 EVENT SELECTION
Fiducial - removes neutrino-induced events outside the detector Containment - to avoid the events going out of the detector Particle Identification - distinguish signal from background
Siva Prasad K Short-Baseline Oscillations June 5, 2017 10/32 EVENT SELECTION - νe
NOνA Simulation NOνA Simulation
105 15 Total MC (No Oscillations) ν CC Total MC µ ν e CC ν CC µ 3.72×1020 POT NC 20 4 ν Total MC (3+1 osc.)
CC 10 10 e × NC
20 LSND best-fit signal ∆ 2 2 10 10 m41 = 1.2 eV × θ 14 = 0.16 rad θ 24 = 0.17 rad 103 Events / 3.72 Events / 0.5 GeV 3.72
2 5 102 10
0 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 CVN Reconstructed Neutrino Energy (GeV)
Selection efficiency of νe’s with CVN > 0.95 is 49% and purity is 86%.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 11/32 EVENT SELECTION - νµ
NOνA Simulation NOνA Simulation 5
40 Total MC Total MC (No Oscillations) ν CC ν CC µ µ ν 4 e CC ν CC e 3.72×1020 POT NC 20 NC Total MC (3+1 osc.) 10
× 30 20 LSND best-fit signal 10
× ∆ 2 2 3 m41 = 1.2 eV θ 14 = 0.16 rad θ 24 = 0.17 rad 20 2 Events / 3.72 5 10 Events / 0.5 GeV 3.72 4 10 10 1
0 0 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 CVN Reconstructed Neutrino Energy (GeV)
Selection efficiency of νµ’s with CVN > 0.5 is 58% and purity is 96%.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 12/32 PREDICTION WITH SYSTEMATICS
ν ν e prediction NOνA Simulation µ prediction NOνA Simulation
8 8
6×1020 POT 6×1020 POT 6 Beam and Cross-Section 6 Beam and Cross-Section uncertainties (±1σ) uncertainties (±1σ)
∆ 2 2 ∆ 2 2 m41 = 3.0 eV m41 = 3.0 eV θ θ 14 = 0.10 rad 14 = 0.10 rad Events Events
3 θ 5 θ 4 24 = 0.15 rad 4 24 = 0.15 rad 10 10
2 2
0 0 0 2 4 6 0 2 4 6 Reconstructed Neutrino Energy (GeV) Reconstructed Neutrino Energy (GeV)
Predictions for both νe (left) and νµ (right) after including the flux and cross-section uncertainties added in quadrature at ±1σ. Any correlated systematics will be cancelled in the joint fit, e.g cross-section systematics.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 13/32 SENSITIVITY - θµe
NOvA Simulation 102 Fitting simultaneously both νe and νµ prediction with νe and νµ
selected MC spectrum with no 10 (3+1) oscillations. ) Fit for θµe, profiling θ14, θ24 and 2
(eV 1 θ34, and keeping all other 2 41 m parameters fixed. ∆ LSND 90% C.L LSND 99% C.L MiniBooNE 90% C.L Included flux and cross-section −1 10 MiniBooNE 99% C.L uncertainty into the fit. KARMEN 90% C.L NOvA with flux and cross-section syst (1yr) NOvA with flux and cross-section syst (3yr)
−2 10 − − 10 5 10−4 10 3 10−2 10−1 1 2 θ sin 2 µe
2 2 2 sin 2θµe = sin 2θ14 sin θ24.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 14/32 SUMMARY
NOvA is well positioned to search for the sterile neutrinos with ∆m2 ≥ 1 eV2 at its baseline with the NuMI beam.
Joint fit provides significant improvement in sensitivities over the individual fits.
Current NOvA sensitivities are competitive with the on-going and future experiments like Short-Baseline Near Detector program at Fermilab etc.
Stay tuned for the exciting results.
THANK YOU
Siva Prasad K Short-Baseline Oscillations June 5, 2017 15/32 Siva Prasad K Short-Baseline Oscillations June 5, 2017 16/32 BACKUP
Siva Prasad K Short-Baseline Oscillations June 5, 2017 17/32 NUMIBEAM COMPOSITION -FHC
Siva Prasad K Short-Baseline Oscillations June 5, 2017 18/32 PREDICTION WITH FLUX SYSTEMATICS
ν ν e prediction NOνA Simulation µ prediction NOνA Simulation
8 8
6×1020 POT 6×1020 POT ± σ ± σ 6 Beam Uncertainty ( 1 ) 6 Beam Uncertainty ( 1 )
∆ 2 2 ∆ 2 2 m41 = 3.0 eV m41 = 3.0 eV θ θ 14 = 0.10 rad 14 = 0.10 rad Events Events
3 θ 5 θ 4 24 = 0.15 rad 4 24 = 0.15 rad 10 10
2 2
0 0 0 2 4 6 0 2 4 6 Reconstructed Neutrino Energy (GeV) Reconstructed Neutrino Energy (GeV)
Siva Prasad K Short-Baseline Oscillations June 5, 2017 19/32 PREDICTION WITH CROSS-SECTION SYSTEMATICS
ν ν e prediction NOνA Simulation µ prediction NOνA Simulation
8 8
6×1020 POT 6×1020 POT 6 Cross-Section 6 Cross-Section Uncertainty (±1σ) Uncertainty (±1σ)
∆ 2 2 ∆ 2 2 m41 = 3.0 eV m41 = 3.0 eV θ θ 14 = 0.10 rad 14 = 0.10 rad Events Events
3 θ 5 θ 4 24 = 0.15 rad 4 24 = 0.15 rad 10 10
2 2
0 0 0 2 4 6 0 2 4 6 Reconstructed Neutrino Energy (GeV) Reconstructed Neutrino Energy (GeV)
Siva Prasad K Short-Baseline Oscillations June 5, 2017 20/32 TRUE LENGTH
NOνA Simulation NOνA Simulation
2 100
80 1.5 ν ν e total µ total
ν from Kaon νµ from Kaon e 60 ν ν 1 e from Muon µ from Pion
40
0.5 Events / 0.02 km 1E21 POT Events / 0.02 km 1E21 POT 7 7 20 10 10
0 0 0.2 0.4 0.6 0.8 1 1.2 0.2 0.4 0.6 0.8 1 1.2 True Length (km) True Length (km)
Siva Prasad K Short-Baseline Oscillations June 5, 2017 21/32 TRUE ENERGY
NOνA Simulation NOνA Simulation 100
0.4 80
νµ total
0.3 ν 60 ν e total µ from Kaon
ν ν e from Kaon µ from Pion 0.2 40 ν e from Muon Events / 0.5 GeV 1E21 POT Events / 0.5 GeV 1E21 POT 7 0.1 7 20 10 10
0 0 0 2 4 6 8 10 0 2 4 6 8 10 True Energy (GeV) True Energy(GeV)
Distribution of true energy of neutrinos break down by their parents in the near detector.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 22/32 TRUE LENGTH OVER TRUE ENERGY
NOνA Simulation NOνA Simulation
2 120
ν 100 ν 1.5 e total µ total
ν 80 ν e from Kaon µ from Kaon
1 ν ν e from Muon 60 µ from Pion
40 0.5 Events / 0.02 (km/GeV) 1E21 POT Events / 0.02 (km/GeV) 1E21 POT
7 7 20 10 10
0 0 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 L/E (km/GeV) L/E (km/GeV)
Distribution of true length over true energy of neutrinos break down by their parents in the near detector.
Siva Prasad K Short-Baseline Oscillations June 5, 2017 23/32 SIGNAL PREDICTION
NOνA Simulation NOνA Simulation 15
4
POT ν POT ν Unoscillated e Unoscillated µ 20 20
10 (ν → ν ) 10 (ν → ν ) µ × µ µ × e 10 3
∆ 2 2 ∆ 2 2 m41 = 3.0 eV m41 = 3.0 eV θ 2 θ 14 = 0.1 rad 14 = 0.1 rad θ θ 24 = 0.15 rad 24 = 0.15 rad 5 y+z/2 x = #frac{y+z/2}{y^{2}+1} : x = y2+1 1 Events / 0.5 GeV 3.72 Events / 0.5 GeV 3.72 2 5 10 10
1.20 1.20 0 1 2 3 4 5 0 1 2 3 4 5 RecoE (GeV) RecoE (GeV) 1.1 1.1
1 1 Osc Osc Unosc Unosc
0.9 0.9
0.8 0.8 0 1 2 3 4 5 0 1 2 3 4 5 Reconstructed Neutrino Energy (GeV) Reconstructed Neutrino Energy (GeV)
Predictions, Left: νe prediction and Right: νµ prediction, after applying the νe 2 2 and νµ selection respectively, for the oscillation parameters ∆m41 = 3.0 eV ; θ14 = 0.1(rad) ; θ24 = 0.15(rad). Siva Prasad K Short-Baseline Oscillations June 5, 2017 24/32 SELECTION
νe selection νµ selection DQ: flasher cut (hits per plane <= 8) kNumuQuality Reco: each slice has a vertex and 1 3D kNumuContainND FuzzyK prong and one shower kNumuNCRej (ReMID > 0.75) Vertex containment: abs(vX)<140, abs(vY)<140, 100 Energy: Slice calorimetric energy < 5 GeV Loose containment and nplanestofront > 6 Hits: 20 < nhits < 200 Length: 140 < longest fuzzyk prong < 500 cm Gap: Shower start to vertex gap < 100 cm LID > 0.95 Siva Prasad K Short-Baseline Oscillations June 5, 2017 25/32 2 2 SIN 2θ14 VS ∆M41 NOvA Simulation Single Fit: Performed fit for only 102 νe appearance. Joint Fit: Performed fit for both 10 νe appearance and νµ disappearance. ) 2 (eV 1 Fitting νe/νµ prediction with 2 41 m νe/νµ selected MC spectrum ∆ with no (3+1) oscillations. −1 10 Sensitivity - 90% C.L. 1 Year Fit for θ , profiling θ , and NOvA with single fit 14 24 NOvA with joint fit keeping all parameters fixed. −2 10 − − − No systematics included. 10 3 10 2 10 1 1 2 θ sin 2 14 Siva Prasad K Short-Baseline Oscillations June 5, 2017 26/32 CONVOLUTIONAL VISUAL NETWORK This analysis is based on using the new algorithm for particle identification, Convolutional Visual Network. Based on machine learning and deep learning methods. No other information is needed except initial hit clustering and energy deposited in hits. All interactions with at least 15 hits are used for training. CVN achieves the figure of merit A. Aurisano et al., arXiv:1604.01444. equivalent to collecting 30% more exposure Siva Prasad K Short-Baseline Oscillations June 5, 2017 27/32 CUT FLOW TABLE Selection Total MC νµ νe NC No Cut 2.749e+08 2.279e+08 4.062e+06 4.289e+07 Data Quality 3.812e+07 3.366e+07 4.339e+05 4.029e+06 Fiducial 2.703e+06 1.998e+06 5.393e+04 6.516e+05 Containment 7.801e+05 4.525e+05 2.012e+04 3.075e+05 FrontPlanes 7.764e+05 4.505e+05 2.001e+04 3.059e+05 # Hits 7.242e+05 4.354e+05 1.654e+04 2.723e+05 Energy 7.213e+05 4.353e+05 1.469e+04 2.714e+05 ProngLength 5.96e+05 3.862e+05 1.406e+04 1.958e+05 Ptp 5.163e+05 3.412e+05 1.337e+04 1.617e+05 CVN > 0.95 7242 944.1 5171 1127 For νe for 6e+20 POT. Selection Total MC νµ νe NC No Cut 2.749e+08 2.279e+08 4.062e+06 4.289e+07 Data Quality 6.81e+07 6.212e+07 5.978e+05 5.382e+06 Containment 3.565e+06 2.484e+06 5.777e+04 1.022e+06 CVN > 0.5 1.908e+06 1.889e+06 365.4 1.894e+04 For νe for 6e+20 POT. Siva Prasad K Short-Baseline Oscillations June 5, 2017 28/32 JOINT FIT METHOD The formalism to combine several measurements by performing a joint maximum-likelihood fit of different datasets. This analysis is done using log-likelihoods. LJoint = LA + LB where L is Log-likelihood. A and B are datasets. Combine νµ disappearance and νe appearance performing a joint fit for the datasets after applying νµ and νe selection. Siva Prasad K Short-Baseline Oscillations June 5, 2017 29/32 SENSITIVITY - θ14 NOvA Simulation 102 Fitting νe/νµ prediction with νe/νµ selected MC spectrum with no (3+1) oscillations. 10 Fit for θ14, profiling θ24, and ) keeping all parameters fixed. 2 (eV 1 2 41 Included flux and cross-section m ∆ uncertainty. Sensitivity - 90% C.L. 1 Year With the improved selection 10−1 Bugey NOvA truth-based selection NOvA after event selection and understanding of NOvA with beam uncertainties NOvA with cross-section uncertainties systematics, improves the NOvA with both uncertainties 10−2 sensitivity in a better way. −3 −2 −1 10 10 10 1 2 θ sin 2 14 Siva Prasad K Short-Baseline Oscillations June 5, 2017 30/32 SENSITIVITY - θ24 NOvA Simulation 102 Fitting νe/νµ prediction with νe/νµ selected MC spectrum with no (3+1) oscillations. 10 Fit for θ14, profiling θ24, and ) keeping all parameters fixed. 2 (eV 1 2 41 Included flux and cross-section m ∆ uncertainty. Sensitivity - 90% C.L. 6e+20 POT CDHS −1 10 CCFR With the improved selection SciBooNE + MiniBooNE NOvA truth-based selection and understanding of NOvA after event selection with stat. error only systematics, improves the NOvA with flux and cross-section uncertainties −2 10 − − − − sensitivity in a better way. 10 4 10 3 10 2 10 1 2θ sin 24 Siva Prasad K Short-Baseline Oscillations June 5, 2017 31/32 SENSITIVITY - θµe NOvA Simulation 102 Fitting simultaneously both νe and νµ prediction with νe and νµ selected MC spectrum with no 10 (3+1) oscillations. ) Fit for θµe, profiling θ14, θ24 and 2 (eV 1 θ34, and keeping all other 2 41 LSND 90% C.L m parameters fixed. ∆ LSND 99% C.L MiniBooNE 90% C.L MiniBooNE 99% C.L Included flux and cross-section −1 10 KARMEN 90% C.L uncertainty into the fit. SBND 90% C.L NOvA with flux and cross-section syst (1yr) NOvA with flux and cross-section syst (3yr) −2 10 − − − 10 6 10 5 10−4 10 3 10−2 10−1 1 2 θ sin 2 µe 2 2 2 sin 2θµe = sin 2θ14 sin θ24. Siva Prasad K Short-Baseline Oscillations June 5, 2017 32/32