Analysis of the decay B l νγ, l = e, µ Bad Liebenzell 1 Andreas Heller, Michael Feindt, Martin Heck, Anže Zupanc, Thomas Kuhr, Pablo Goldenzweig INSTITUT FÜR EXPERIMENTELLE KERNPHYSIK (EKP), KARLSRUHER INSTITUT FÜR TECHNOLOGIE (KIT) KIT Universität des Landes Baden-Württemberg und nationales Forschungszentrum in der Helmholtz-Gemeinschaft www.kit.edu
The Belle experiment The Belle experiment ran 1999-11 at KEKB accelerator at KEK in Tsukuba, Japan Asymmetric e e accelerator at Y (S) energy of 1.58 GeV B(Y (S) B- B) 96% (771.6 ± 1.6) 1 6 B- B pairs measured Andreas Heller Analysis of the decay B l νγ, l = e, µ /
Helicity suppression of B l ν Pure B l ν decay is helicity suppressed B meson has spin and decays into particle and antiparticle with opposite spin Both particles are almost exclusively right- or left-handed A heavier lepton has a lower momentum and thus a bigger coupling to the weak current SM prediction: B(B e (µ )ν) = 9. 1 1 (3.9 1 7 ) Measurement 1 : B(B τ ν) = (1.8 ±.5) 1 1 Belle: Phys. Rev. D 8 7111, BaBar: Phys.Rev. D88 (13) 311 Andreas Heller Analysis of the decay B l νγ, l = e, µ 3/
Theory of B l νγ B l νγ: Photon lifts helicity suppression but introduces additional electromagnetic coupling u γ W l SM expectation O(B(B l νγ)) 1 6 b νl Weak decay is precisely calculable Photon emission needs an approximation from heavy quark theory where E γ > 1GeV is required u b γ W l νl u l W γ b νl Beneke, Rohrwild: B meson distribution amplitude from B γl ν arxiv:111.38 (11) Andreas Heller Analysis of the decay B l νγ, l = e, µ /
Theory of B l νγ Branching fraction depends on λ B Describes the first moment of the quark distribution amplitude inside the B meson Theoretical calculation of the parameter unreliable Important parameter for several radiative B meson decays 3 Aubert et al.: Search for the radiative leptonic decay B γl ν, arxiv:7.178 (7) Beneke, Rohrwild: B meson distribution amplitude from B γl ν arxiv:111.38 (11) Andreas Heller Analysis of the decay B l νγ, l = e, µ 5/
Theory of B l νγ Branching fraction depends on λ B Describes the first moment of the quark distribution amplitude inside the B meson Theoretical calculation of the parameter unreliable Important parameter for several radiative B meson decays From other measurements λ B can be constrained O(B(B l νγ)) 1 6 3 Aubert et al.: Search for the radiative leptonic decay B γl ν, arxiv:7.178 (7) Beneke, Rohrwild: B meson distribution amplitude from B γl ν arxiv:111.38 (11) Andreas Heller Analysis of the decay B l νγ, l = e, µ 5/
Theory of B l νγ Branching fraction depends on λ B Describes the first moment of the quark distribution amplitude inside the B meson Theoretical calculation of the parameter unreliable Important parameter for several radiative B meson decays From other measurements λ B can be constrained O(B(B l νγ)) 1 6 BaBar measurement of B(B l νγ) gives upper limits of < 17 1 6 (electron channel), < 1 6 (muon channel) 3 3 Aubert et al.: Search for the radiative leptonic decay B γl ν, arxiv:7.178 (7) Beneke, Rohrwild: B meson distribution amplitude from B γl ν arxiv:111.38 (11) Andreas Heller Analysis of the decay B l νγ, l = e, µ 5/
Full reconstruction Hadronic full reconstruction covers 1% of the B branching fraction The efficiency is in the range of.5% Feindt et al.: A Hierarchical NeuroBayes-based Algorithm for Full Reconstruction of B Mesons at B Factories Andreas Heller Analysis of the decay B l νγ, l = e, µ 6/
B l νγ analysis Full reconstruction of one B meson Impuls des zweiten B-Mesons p Bsig = (E Beam /, p Btag ) m miss = (p Bsig p lepton p γ ) Normalized Data. B e ν γ. Generic MC.18 B X u l ν.16 Rare MC.1.1.1.8.6... m miss (GeV ) Andreas Heller Analysis of the decay B l νγ, l = e, µ 7/
B l νγ analysis Full reconstruction of one B meson Impuls des zweiten B-Mesons p Bsig = (E Beam /, p Btag ) m miss = (p Bsig p lepton p γ ) Normalized Data. B e ν γ. Generic MC.18 B X u l ν.16 Rare MC.1.1.1.8.6.. Analysis is performed on MC (blind analysis). m miss (GeV ) Signal branching fraction is assumed to be 5 1 6 Background simulation consists of: generic MC with b c decays and semi-leptonic b ulν MC with B π /ηl γ processes Andreas Heller Analysis of the decay B l νγ, l = e, µ 7/
Pre-selection After full reconstruction only signal side particles are expected to be left in the detector Signal signature consists of a charged track and a high energetic photon Efficient signal selection Highest energetic photon in the event with an energy above 1 GeV No charged tracks left after selection Lepton selection based on compounded likelihood variable Little remaining energy deposition in the calorimeter Cut on the mass of the fully reconstructed B meson Andreas Heller Analysis of the decay B l νγ, l = e, µ 8/
Signal like background Dominant background after pre-selection: B π l ν and B ηl ν B(π γγ) 99%, B(η γγ) % Decays are similar to signal if only one photon is found Second photon originating from the meson is searched in the detector with specific variables Number of Expected Events 5 3 1 B µ ν γ Generic MC B X u l B X u l ν fixed decays ν free decays m miss (GeV ) Andreas Heller Analysis of the decay B l νγ, l = e, µ 9/
Meson vetoes Meson masses of π, η are reconstructed from signal photon candidate an the remaining photons in the calorimeter Meson candidate with mass closest to its nominal mass is kept Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Meson vetoes Meson masses of π, η are reconstructed from signal photon candidate an the remaining photons in the calorimeter Meson candidate with mass closest to its nominal mass is kept High combinatorics can lead to the construction artificial peaks at the nominal mass Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Meson vetoes Meson masses of π, η are reconstructed from signal photon candidate an the remaining photons in the calorimeter Meson candidate with mass closest to its nominal mass is kept High combinatorics can lead to the construction artificial peaks at the nominal mass Mass spectra are computed with different energy cuts on the background photons Leads to different mass distributions with complementary information Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Meson vetoes Meson masses of π, η are reconstructed from signal photon candidate an the remaining photons in the calorimeter Meson candidate with mass closest to its nominal mass is kept High combinatorics can lead to the construction artificial peaks at the nominal mass Mass spectra are computed with different energy cuts on the background photons Leads to different mass distributions with complementary information Meson mass spectra are gathered in neural network which is used for selection Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Meson vetoes Normalized Data.5..15.1.5 B µ ν γ B π µ ν B η µ ν Remaining B X u l ν Generic MC...5.1 m(π.15..5.3 ) (GeV) without cut on γ.35. bkg Normalized Data. B µ ν γ. B π µ ν B η µ ν.18 Remaining B X u l ν.16 Generic MC.1.1.1.8.6..... m(π..6.8 ) (GeV) with γ > 6 MeV 1. bkg Andreas Heller Analysis of the decay B l νγ, l = e, µ 11/
Network training Neural net is trained with NeuroBayes 5 Training: Signal-MC against X u lν MC and especially B π l γ Additional variables in the network are the remaining energy in the calorimeter and angles among the signal candidate particles Signal is measured with a fit of the missing mass in 6 bins of the network output Normalized Data.1.8.6... B µ ν γ Generic MC B X u l ν...6.8 Network output 5 Feindt: A Neural Bayesian Estimator for Conditional Probability Densities, arxiv:physics/93 () Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Network output data-mc comparison 35 3 B µ ν γ Exclusive X u l ν MC Background MC Data sample 5 35 B e ν γ Exclusive X u l ν MC Background MC Data sample 5 3 5 15 15 1 1 5 5....6.8 1. Network output Muon channel....6.8 1. Network output Electron channel Andreas Heller Analysis of the decay B l νγ, l = e, µ 13/
Network bin optimization Expected Significance B µ νγ.8.6... e νγ Expected Significance B.8.6... 1.8 1.8 1 3 5 6 7 8 Fit in no. of network bins Muon channel 1 3 5 6 7 8 Fit in no. of network bins Electron channel Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Fit shapes, muon channel Number of Expected Events.7.6.5..3..1 1..8.6.. 1. 1..8.6... Network Bin 1. Network Bin. Network Bin 3. 3. 9 Number of Expected Events. 1.8 1.6 1. 1. 1..8.6...5. 1.5 1..5 8 7 6 5 3 1. Network Bin. Network Bin 5 Network Bin 6 Andreas Heller Analysis of the decay B l νγ, l = e, µ 15/
Fit shapes, electron channel.8 1. 1. Number of Expected Events.7.6.5..3..1.8.6.. 1. 1..8.6... Network Bin 1. Network Bin. Network Bin 3 1.8 1.6.5 7 Number of Expected Events 1. 1. 1..8.6... 1.5 1..5 6 5 3 1. Network Bin. Network Bin 5 Network Bin 6 Andreas Heller Analysis of the decay B l νγ, l = e, µ 16/
Signal extraction Yield of the dominant B X u l ν backgrounds is fixed to MC prediction Two free fit parameters are the signal and one background yield Andreas Heller Analysis of the decay B l νγ, l = e, µ 17/
Signal extraction Yield of the dominant B X u l ν backgrounds is fixed to MC prediction Two free fit parameters are the signal and one background yield Electron and muon channel are fitted separately and simultaneously Andreas Heller Analysis of the decay B l νγ, l = e, µ 17/
Signal extraction Yield of the dominant B X u l ν backgrounds is fixed to MC prediction Two free fit parameters are the signal and one background yield Electron and muon channel are fitted separately and simultaneously Several tests are performed with toy MC studies: linearity test, confidence interval coverage, bootstrapped toy MC No bias is found for significant measurements Confidence interval coverage is too large Andreas Heller Analysis of the decay B l νγ, l = e, µ 17/
Systematic error Sample muon channel electron channel Meson veto network -.58 -.66.75.6 Fit shapes 1.3 1.6 Fixing of B X u l ν ±.18 ±. B l νγ model -.1 -.5 Tag side efficiency ±.35 ±.3 Continuum suppression -.13 -. Tracking efficiency -.1 -.1 Lepton ID ±. ±.18 N B B ±.11 ±.11 Sum Sum 1.1 1.59 Simultaneous fit 1.81.9 1.1 1.39 Andreas Heller Analysis of the decay B l νγ, l = e, µ 18/
Measurement on data, muon channel Number of Events 3..5. 1.5 1..5.5. 3.5 3..5. 1.5 1..5 7 6 5 3 1. Network Bin 1. Network Bin Network Bin 3 Number of Events 1 8 6 1 1 8 6 18 16 1 1 1 8 6 Network Bin Network Bin 5 Network Bin 6 Andreas Heller Analysis of the decay B l νγ, l = e, µ 19/
Measurement on data, electron channel.5..5. 6 5 Number of Events 3.5 3..5. 1.5 3.5 3..5. 1.5 3 1..5. Network Bin 1 1..5. Network Bin 1 Network Bin 3 8 1 Number of Events 7 6 5 3 8 6 18 16 1 1 1 8 1 Network Bin Network Bin 5 6 Network Bin 6 Andreas Heller Analysis of the decay B l νγ, l = e, µ /
Fit results on data one bin B µ ν γ Exclusive X u l ν MC 35 Background MC 3 Data sample 5 3 5 15 B e ν γ Exclusive X u l ν MC Background MC Data sample 15 1 5 1 5 m miss (GeV ) Muon channel m miss (GeV ) Electron channel Andreas Heller Analysis of the decay B l νγ, l = e, µ 1/
Measurement on data Default analysis E(γ sig ) > 1 GeV Sample Yield significance in σ BR limit 1 6 (9% CL) B e νγ 6.1.91.1 3.9 1. 1. 13. B νγ.9 3.61.1.6 1.6. 7.1 B l νγ 6.6 5.71.8.7.3 1. 7.3 Secondary analysis with E(γ sig ) > MeV Sample Yield significance in σ BR limit 1 6 (9% CL) B e νγ 11.9 7.1.9 6... 1. B µ νγ.1 5.1.9.1.3-6.3 B l νγ 11.3 8.3.1 7. 3.6 1.5 7.6 For the yields the first error is statistical, the second error is systematic. Significances and upper limits contain systematic errors. Andreas Heller Analysis of the decay B l νγ, l = e, µ /
Backup Slides Andreas Heller Analysis of the decay B l νγ, l = e, µ 3/
Measurement on data, muon channel, secondary analysis 6 7 6 5 6 5 Number of Events 3 5 3 3 1 1 1 Network Bin 1 Network Bin Network Bin 3 1 5 Number of Events 1 8 6 15 1 5 3 1 Network Bin Network Bin 5 Network Bin 6 Muon channel with E(γ sig ) > MeV. Andreas Heller Analysis of the decay B l νγ, l = e, µ /
Measurement on data, electron channel, secondary analysis 7.5 6 6. 5 Number of Events 5 3 1 3.5 3..5. 1.5 1..5 3 1 Network Bin 1. Network Bin Network Bin 3 16 Number of Events 8 7 6 5 3 1 1 1 8 6 5 3 1 1 Network Bin Network Bin 5 Network Bin 6 Electron channel with E(γ sig ) > MeV. Andreas Heller Analysis of the decay B l νγ, l = e, µ 5/
Hierarchische Ansatz der vollständigen Rekonstruktion Andreas Heller Analysis of the decay B l νγ, l = e, µ 6/
Netwerkvariablen Extraenergie im ECL Winkel zwischen Signal γ und ν m(π ) mit E(γ rem ) > MeV m(π ) ohne Schnitt auf E(γ rem ) m(η) mit E(γ rem ) > 3 MeV Winkel zwischen Signal γ und Lepton m(η) mit E(γ rem ) > 1 MeV m(π ) mit E(γ rem ) > 6 MeV m(π ) mit ECL-Schnitten skaliert um.6 Andreas Heller Analysis of the decay B l νγ, l = e, µ 7/
Fitformen: Myonkanal für Sekundäranalyse E(γ sig ) > MeV Number of Expected Events.6. 1...5 1.8 1. 1.6. 1..8 1..3.6 1..8...6.1.... Network Bin 1. Network Bin. Network Bin 3 Number of Expected Events 3..5. 1.5 1..5. Network Bin 7 6 5 3 1 Network Bin 5 18 16 1 1 1 8 6 Network Bin 6 Andreas Heller Analysis of the decay B l νγ, l = e, µ 8/
Fitformen: Elektronkanal für Sekundäranalyse E(γ sig ) > MeV Number of Expected Events.8 1. 1.8 1.6.7 1. 1..6.8 1..5 1...6.8.3..6....1.. Network Bin 1. Network Bin. Network Bin 3 Number of Expected Events.5. 1.5 1..5. Network Bin 5 3 1 Network Bin 5 18 16 1 1 1 8 6 Network Bin 6 Andreas Heller Analysis of the decay B l νγ, l = e, µ 9/
Übergroßes Konfidenzintervall für oberes Limit Relative number of values above the 9% boundary.1.8.6... 1 3 5 6 7-6 BR(B l ν γ) 1 Relative number of values above the 9% boundary.8.6... 1 3 5 6 7-6 BR(B l ν γ) 1 Muon channel Electron channel Andreas Heller Analysis of the decay B l νγ, l = e, µ 3/
Übergroßes Konfidenzintervall für oberes Limit Relative number of values above the 9% boundary.1.8.6... 1 3 5 6 7-6 BR(B l ν γ) 1 Simultaneous fit Andreas Heller Analysis of the decay B l νγ, l = e, µ 31/
Systematische Fehler für die Sekunäranalyse mit E(γ sig ) > MeV Datensatz Myonkanal Elektronkanal Mesonveto Netzwerk -.91-1.8 1.18 1.7 Fitformen 1.69 1.9 B X u l ν Normierung ±.31 ±.1 B l νγ Modell.8. Tagseiteneffizienz ±.5 ±.51 Kontinuumsunterdrückung.19 -.8 Trackingeffizienz -. -. Lepton ID ±.6 ±.7 N B B ±.17 ±.17 Summe Summe 1.9.7 Simultanfit 3.5 3.58 1.9.39 Andreas Heller Analysis of the decay B l νγ, l = e, µ 3/
Faltung eines signifkanten Likelihoods.35.3 Likelihood Arbitrary Scale.5..15.1 Convoluted Likelihood.5. 5 1 15 5 3 35 Signal Yield Electron channel Likelihood für den Elektronkanal Andreas Heller Analysis of the decay B l νγ, l = e, µ 33/
Faltung eines signifkanten Likelihoods.5 Likelihood. Convoluted Likelihood Arbitrary Scale.3..1. 5 1 15 Signal Yield Electron channel Likelihood für den Elektronkanal Andreas Heller Analysis of the decay B l νγ, l = e, µ 3/
Erwartete Signifikanz Primäranalyse mit E(γ sig ) > 1 GeV Datensatz Erwartete Messung Signifikanz in σ Erwartetes Limit 1 6 (9% CL) Elektron 8. ±.5 1.1 1.. (.1) 7. (7.6) Myon 8.7 ±.6 1.1 1.6.5 (.) 6.5 (6.93) Simultanfit 16.5 ± 6.5 1.8.3 3.6 (.9).35 (.76) Sekundäranalyse mit E(γ sig ) > MeV Datensatz Erwartete Messung Signifikanz in σ Erwartetes Limit 1 6 (9% CL) Elektron 1. ± 6. 1.9.. (.1) 6.5 (6.78) Myon 11.9 ± 6. 1.9.3.5 (.) 5.98 (6.3) Simultanfit.9 ± 8.7 3.1 3.6 3. (.9).8 (.3) Erwartete Messung und Signifikanz für Signal mit BR = 5 1 6 und erwartetes oberes Limit ohne Signal. Werte in Klammern beinhalten systematische Fehler. Signifikanzen werden durch eine Likelihoodverhältnis bestimmt Obergrenzen werden durch Integration eines Likelihoods bestimmt, wobei nur der positive Teil berücksichtigt wird Andreas Heller Analysis of the decay B l νγ, l = e, µ 35/