MiniBooNE s First Neutrino Oscillation Result

Transcription

1 MiniBooNE s First Neutrino Oscillation Result Morgan Wascko Imperial College London HEP Seminar, 9 Nov 2007

2 Outline 1. Motivation and Introduction 2. Description of the Experiment 3. Analysis Overview 4. Two Independent Oscillation Searches 5. First Results 6. Updates Since First Result 2

3 Motivation if neutrinos have mass... a neutrino that is produced as a ν μ (e.g. π+ μ + ν μ ) might some time later be observed as a ν e (e.g. ν e n e- p) π + ν source μ + ν μ ν e e - X ν detector 3

4 Neutrino Oscillation ( νµ ν e ) = ( )( ) cosθ. sinθ ν1 sinθ.cosθ ν 2 Consider only two types of neutrinos ν µ ν 1 ν 2 If weak states differ from mass states ϴ ν e i.e. (ν µ νe) (ν1 ν2) Then weak states are mixtures of mass states ν µ (t) >= sinθ ν 1 > e ie 1t + cosθ ν 2 > e ie 2t P osc (ν µ ν e ) = < ν e ν µ (t) > 2 Probability to find ν e when you started with ν µ 4

5 Neutrino Oscillation In units that experimentalists like: ( ) P osc (ν µ ν e ) = sin 2 2θsin m 2 (ev 2 )L(km) Fundamental Parameters mass squared differences mixing angle Experimental Parameters L = distance from source to detector E = neutrino energy E ν (GeV) 5

6 LSND K2K Oscillation Signals hep-ex/ hep-ex/ Solar - Homestake,... SNO confirmed by reactors Atmospheric - Super-K,... confirmed by accelerators Accelerator - measured by LSND unconfirmed! KamLAND hep-ex/

7 The Problem Three different neutrino oscillation signals Three independent Δm 2 LSND νμ νe Reactor Limit νe νe M. Sorel Problem: We only need two! Explanation requires physics well beyond the standard model Is it true? 7

8 Verifying LSND P(ν µ ν e ) = sin 2 2θ 12 sin 2 (1.27 m 2 L 12 E ) PRD 64, LSND interpreted as 2 ν oscillation Verification requires same (L/E) and high statistics Different systematics MiniBooNE chose higher L and E Strategy: search for ν e excess in ν µ beam 8

9 MiniBooNE Collaboration Fermilab Visual Media Services TODAY: MiniBooNE s initial results on testing the LSND anomaly 1- Generic search for νe excess in ν µ beam 2- Analysis of data within 2 ν appearance only context 9

10 Outline 1. Motivation and Introduction 2. Description of the Experiment 3. Analysis Overview 4. Two Independent Oscillation Searches 5. First Results 6. Updates Since First Result 10

11 Overview Fermilab Visual Media Services MiniBooNE Overview JL Raaf 11

12 Target & Horn Main components of Booster Neutrino Beam (BNB) (96M and 146M+ pulses) MiniBooNE Overview JL Raaf 12

13 Meson Production External meson production data HARP data (CERN) Parametrisation of crosssections Sanford-Wang for pions Feynman scaling for kaons MiniBooNE Overview JL Raaf 13

14 ν Flux M. Wilking 99.5% pure muon flavour 0.5% intrinsic νe Constrain νe content with ν µ measurements MiniBooNE Overview JL Raaf 14

15 Detector MiniBooNE Overview JL Raaf 15

16 Neutrino Interactions ν µ,e µ -,e - n W + p ν µ CC / NC quasi-elastic scattering (QE) 42% / 16% ν µ p Z 0 p νµ µ - n W + π + n νµ CC / NC resonance production (1π) 25% / 7% νµ MiniBooNE is here p,n Z 0 p,n π 0 16

17 Mineral Oil Optics Production: Cherenkov and scintillation Secondary: Fluorescence and rescattering Wavefront Particle track θ C µ light Molecular energy levels of oil 17

18 PMT Hit Clusters PMT hits clusters in time form subevents ν µ events have 2 subevents µ, followed by e H.A. Tanaka ν e events have 1 subevent Simple cuts on subevents remove cosmic backgrounds pre-cuts 18

19 Track Reconstruction Charged particles produce Cherenkov and scintillation light in oil PMT PMTs collect photons, record t,q Reconstruct tracks by fitting time and angular distributions Find position, direction, energy 19

20 Track Images Muons full rings Electrons fuzzy rings Neutral pions double rings 20

21 Detector Stability 21

22 Outline 1. Motivation and Introduction 2. Description of the Experiment 3. Analysis Overview 1. Signal and Backgrounds 2. Strategy 4. Two Independent Oscillation Searches 5. First Results 6. Updates Since First Result 22

23 Blind Analysis Opened specific boxes with <1σ νe signal Initial Open Box Use all non-beam-trigger data calibration and MC tuning 0.25% random trigger unbiased data studies ν µ CCQE measure flux, E QE ν, oscillation fit ν µ NCpi0 measure rate for MC tuning ν µ CC1pi+ check rate for MC ν µ -e elastic check MC rate dirt measure MC rate all events with E ν > 1.4 GeV check MC rate One closed signal box Second Step explicitly sequester signal, 99% of data open 23

24 For robustness, MiniBooNE has performed two independent oscillation analyses. MC Tuning Calibration Data Quality Cuts DAQ Track Fitter Reconstruction Point light source Reconstruction Likelihood ν e Selection Boosting ν e Selection ν e /ν µ Ratio Oscillation Fit ν e +ν µ Combined Oscillation Fit one oscillation result cross check 24

25 Signal and Backgrounds stacked signal and backgrounds after ν e event selection RB Patterson Oscillation ν e Example oscillation signal Δm 2 = 1.2 ev 2 SIN 2 2θ = Fit for excess as a function of reconstructed ν e energy 25

26 Signal and Backgrounds stacked signal and backgrounds after ν e event selection RB Patterson ν e from K + and K 0 Use fit to kaon production data for shape Use high energy ν e and ν μ in-situ data for normalisation cross-check 26

27 Signal and Backgrounds stacked signal and backgrounds after ν e event selection RB Patterson ν e from μ + ν µ µ + p+be π + ν e ν µ e + Measured with in-situ ν μ CCQE sample Same ancestor π + kinematics Most important background - Constrained to a few % 27

28 Signal and Backgrounds stacked signal and backgrounds after ν e event selection RB Patterson MisID ν μ ~46% π 0 Determined by clean π 0 measurement ~16% Δ γ decay π 0 measurement constrains ~14% dirt Measure rate to normalise and use MC for shape ~24% other Use ν μ CCQE rate to normalise and MC for shape 28

29 Strategy Incorporate in-situ data whenever possible MC tuning with calibration data energy scale PMT response optical model MC tuning with neutrino data cross section nuclear model parameters π0 rate constraint Constraining systematic errors with neutrino data ratio method: ν e from µ decay combined fit to ν e and ν µ data Recurring theme: good data-mc agreement 29

30 MC Tuning data MC 30

31 Strategy Incorporate in-situ data whenever possible MC tuning with calibration data energy scale PMT response optical model MC tuning with neutrino data cross section nuclear model parameters π0 rate constraint Constraining systematic errors with neutrino data ratio method: ν e from µ decay combined fit to ν e and ν µ data Recurring theme: good data-mc agreement 31

32 ν µ CCQE events ν µ 12 C n p µ e E QE Used to measure flux and check E ν QE reconstruction ν = 1 2M p E µ m 2 µ 2M p E µ + (Eµ 2 m 2 µ)cosθ µ T. Katori Data Monte Carlo 2 subevents: e, µ Require e be located near end of µ track A.A. Aguilar Arevalo E ν QE resolution ~10% 32

33 Tuning CCQE MC Q 2 distribution fit to tune empirical parameters of nuclear model ( 12 C) Data χ 2 /ndf = 4.7 / 13 good data-mc agreement in variables not used in tuning! 33

34 π 0 Mis-ID Backgrounds ν µ γ π 0 12 C γ n(p) π0 s are reconstructed outside mass peak if: asymmetric decays fake 1-ring 1 of 2 photons exits high momentum π0 decays produce overlapping rings 34

35 π 0 Mis-ID Backgrounds The MC π 0 rate (flux xsec) is re-weighted to match the measurement in p π bins. J. Link J. Link good data-mc agreement in variables not used in tuning! 35

36 Strategy Incorporate in-situ data whenever possible MC tuning with calibration data energy scale PMT response optical model MC tuning with neutrino data cross section nuclear model parameters π0 rate constraint Constraining systematic errors with neutrino data ratio method: ν e from µ decay combined fit to ν e and ν µ data Recurring theme: good data-mc agreement 36

37 Analysis Strategy 1: Ratio Method MC predicts a range of ν µ fluxes π + ν µ µ + ν µ Use data/mc ratio of ν µ CCQE events to re-weight parent π + νe from µ decay unweighted νe from µ decay re-weighted e + ν e 37

38 For each Eν bin i, i = N DATA i Analysis Strategy 2: Combined Fit N MC i Raster-scan in Δm2 and sin 2 2θ µe to calculate χ 2 over νe and ν µ bins ( ) νe ν µ νe νeν µ ν µ Correlations between E ν QE bins from the optical model: χ 2 = N bins i=1 N bins j=1 i Mi 1 j j Systematic error matrix includes uncertainties for νe and ν µ 38

39 Error Matrix M i j = 1 N α N α α=1 (N α i N MC i )(N α j N MC j ) Use MC variations to study systematic uncertainties Vary underlying parameters and compare to central value MC Example of Eν distributions for several MC variations Central Value (MC) 1 st variation (α=1) 2 nd varaiation (α=2)... Total error matrix is sum of individual matrices i 1 i 2 i 3 i 4 i 5 i 6 i 7 i 8 E ν QE (GeV) 39

40 Systematic Errors Neutrino flux predictions meson production cross sections meson secondary interactions focussing horn current target and horn system alignment Neutrino interaction cross sections nuclear model rates and kinematics for relevant processes resonance width and branching fractions Detector modelling optical model of light propagation PMT charge and time response electronics & DAQ model neutrino interactions in dirt surrounding detector constraint? 40

41 Outline 1. Motivation and Introduction 2. Description of the Experiment 3. Analysis Overview 4. Two Independent Oscillation Searches 1. Reconstruction and Event Selection 2. Systematic Uncertainties 5. First Results 6. Updates Since First Result 41

42 2 Independent Searches Method 1: Track Based Analysis Careful Reconstruction of particle tracks Identify particle type by likelihood ratio Use ratio method to constrain backgrounds Strengths: Relatively insensitive to optical model Simple cuts on likelihood ratios Primary Analysis Method 2: Boosted Decision Trees Classify events using boosted decision trees Cut on output variables to improve event separation Use combined fit to constrain backgrounds Strengths: Combination of weak variables to form strong classifier Better constraints on backgrounds Cross-check Analysis 42

43 Particle Identification e-like e-like π 0- like μ-like Monte Carlo Monte Carlo Reconstruct under 3 hypotheses: µ-like, e-like and π0 -like ν e particle ID cuts on likelihood ratios chosen to maximise ν µ νe oscillation sensitivity 43

44 e/µ Likelihood ν µ CCQE data (with muon decay electron) compared to ν µ data with no decay electrons ( All but signal ) Removes most muon events 1 decay electron 0 decay electrons 44

45 e/π 0 Likelihood All but signal data and MC Signal Region Cut on likelihood ratio and reconstructed mass Opened sidebands before unblinding full data sample 45

46 Signal and background Predicted νe efficiency Analysis region defined to be MeV Signal efficiency higher at low energy Backgrounds higher there too... 46

47 Signal and background Predicted νe energy distribution Analysis region defined to be MeV Signal efficiency higher at low energy Backgrounds higher there too... 47

48 Signal and background Predicted νe energy distribution MeV νe(µ decay) νe(k decay) Radiative Δ NCπ 0 Dirt Other Total 358 Signal

49 Uncertainties source uncertainty (%) Flux from π + /μ + decay 6.2 Flux from K + decay 3.3 Flux from K 0 decay 1.5 Target and beam models 2.8 ν-cross section 12.3 NC π 0 yield 1.8 External interactions 0.8 Optical model 6.1 Electronics & DAQ model 7.5 constrained total 9.6 Note: total is not the quadrature sum-- errors are further reduced by constraints from ν μ data 49

50 Sensitivity Sensitivity to oscillations Primary analysis chosen on the basis of this plot Chosen before opening the box! 50

51 Outline 1. Motivation and Introduction 2. Description of the Experiment 3. Analysis Overview 4. Two Independent Oscillation Searches 5. First Results 6. Updates Since First Result 51

52 After applying all analysis cuts Step 1: Fit sequestered data to oscillation hypothesis Don t return fit parameters Opening The Box Apply unreported parameters to MC, check diagnostic variables Return χ2 for diagnostic variables Step 2: Open plots from Step 1 Plots chosen to be useful but not revealing Step 3: Report only the (unsigned) χ2 from fit No fit parameters returned Step 4: Compare EnuQE for data and MC Blindness broken Step 5: Present results within two weeks 52

53 Training for a blind search MOW c (blinded) On March 26, 2007 we opened the box...

54 Opened box! Counting Experiment ( MeV) Expect 358 ± 19(stat) ± 35(sys) Observe 380 Significance 0.55 σ 54

55 Exclusion Curve No evidence for ν µ νe 2ν appearance only oscillations MiniBooNE First Result Independent second analysis finds similar result Incompatible with LSND at 98% CL cf. KARMEN2 compatible at 64% 55

56 What Does It Mean? With the blind analysis, we have asked the question: Do νμs oscillate directly to νes with Δm 2 ~ 1eV 2, ala LSND? We have a clear answer: NO More work yet to do... 56

57 At lower energy... Lowering the energy threshold reveals νe excess Excess not consistent with LSND signal Currently under investigation 57

58 Outline 1. Motivation and Introduction 2. Description of the Experiment 3. Analysis Overview 4. Two Independent Oscillation Searches 5. First Results 6. Updates Since First Result 58

59 Low E checklist Data integrity checks New backgrounds? Double check background calculations (i.e. not considered in original analysis) N.B. If this is a background it may be relevant for other experiments searching for νμ νe New physics? Looking at new/more data 59

60 Integrity checks event/pot vs day, 475<Enu<1500 MeV Detector anomalies: none found Example: time distribution of νe events is flat Hand scanned all events: nothing pathological found event/pot vs day, 300<Enu<475 MeV event display of typical νe 60

61 Muon Internal Brem Data-MC excess, but note the scale! Statistical uncertainties only! Apply recon and PID to clean muon CCQE events Directly measure rate of final state muon ν e backgrounds Paper on this work: arxiv: [hep-ex] Submitted to PRD 61

62 Dirt Backgrounds before box-opening, fit yielded meas/pred = 1.00±0.15 fit in different (open) sample yields meas/pred = 1.08±0.12 ν µ dirt γ shower results from dirt-enhanced fits visible energy (GeV) dist to tank wall along track (cm) 62

63 Lower energy threshold More data should help Extended threshold to lower energy required extension of systematics Excess persists below 300 MeV New bin is even more dominated by mis-id ν µ 63

64 Lower energy threshold More data should help Extended threshold to lower energy required extension of systematics Excess persists below 300 MeV New bin is even more dominated by mis-id ν µ New bin 64

65 Background Breakdown reconstructed ν energy bin (MeV) total BG 284±25 274±21 358±35 νe intrinsic νμ induced NC π NC Δ Nγ Dirt other DATA 375±19 369±19 380±19 65

66 Visible Energy & Angles cos θ, 200<Eν<3000 MeV E l, 200<Eν<3000 MeV cos θ Recall: two-body kinematics allow ν energy reconstruction from Elepton and θlepton no anomalies in these distributions 66

67 El & θl in Eν bins 200< Eν<300 MeV 300 <Eν<475 MeV 200< Eν<300 MeV 300< Eν<475 MeV cos θ cos θ Excess distributed among El, cosθl bins 475 <Eν<3000 MeV 475< Eν<3000 MeV cos θ At higher energy, data are welldescribed by predicted background 67 41

68 New BG? Physics? Difficulty distinguishing single photons from electrons Photo-nuclear absorption Can produce low energy ν e events No effect on E ν>475 MeV Anomaly-mediated photon production arxiv: [hep-ex] Both under active investigation 68

69 More data should help! Double check everything in MiniBooNE ν Same detector with different beam NuMI ν Same beam with different detector SciBooNE ν 69

70 Same Det. Diff. Beam MiniBooNE can see neutrinos from the NuMI beam Off-axis beam 110 mrad Decay Pipe MiniBooNE ν Beam Absorber MINOS near Enriched ν e sample Very different energy for ν µ components Results coming soon! 70

71 Same Beam Diff. Det. New experiment at Fermilab Near Detector in BNB Better at distinguishing photons from electrons Check MiniBooNE s background estimates νe event in SciBar detector Spokespeople: T. Nakaya, Kyoto University, Imperial College 71

72 Three subdetectors: SciBar, EC, MRD Data run started June 2006 Now taking data (as I speak!) 72

73 Data Progress Expect 2.0e20 POT total 2007 ν run 1.0e20 neutrino 1.0e20 antineutrino Collected 0.54e20 POT antineutrinos already Now running in neutrino mode ν run Only 1 dead channel in 14,

74 What s Next? MiniBooNE is publishing more papers: Neutrino cross section measurements Joint analysis of MiniBooNE, LSND and KARMEN data More exotic oscillation analyses νe disappearance 2 or 3 sterile neutrinos with CP violation MiniBooNE analysis coming soon Results of NuMI-MB analysis very soon Fermilab Wine & Cheese Seminar Dec 14 MiniBooNE is pursuing ν e appearance search now 74

75 Summary MiniBooNE observes no evidence for ν µ νe 2ν oscillations MiniBooNE First Result Incompatible with LSND ν µ νe oscillation signal at 98% CL Low energy excess under investigation More data coming soon Phys.Rev.Lett. 98, (2007) arxiv: [hep-ex] 75