Tau Physics with BaBar

Tau Physics with BaBar Ohio State University & University of Paris VI,VII Outline of Talk Introduction & some facts Searches for Lepton Flavor Violation Tau lifetime Hadronic decays of the tau τ 7 hadrons τ 5 hadrons Summary & Conclusions

Fact #1 Since the cross section for e + e - Y(4S) BB is about the same as the cross section for e + e - τ + τ a B factory is also a τ factory.

Fact #2 BaBar is a great detector for τ physics! Detector of Internally Recflected Cherenkov Light (DIRC) 1.5 T Solenoid Electromagnetic Calorimeter (EMC) Drift Chamber (DCH) e - (9 GeV)e + (3.1 GeV) Instrumented Flux Return (IFR) Silicon Vertex Tracker (SVT) SVT, DCH: charged particle tracking vertex &mom. resolution EMC: electromagnetic calorimetry γ/π 0 /η DIRC, IFR, DCH: charged particle ID π/k/p

BaBar K/π ID D * + D 0 π + D 0 K + π - BaBar DIRC BaBar (and Belle) are the first e + e - experiments with: excellent charged particle tracking excellent EM calorimetry excellent Particle ID and LOTS of data.

Fact #3 PEP-II is a great* accelerator PEP-II delivered ~254 fb fb -1-1 BaBar recorded ~244 fb fb -1-1 Total # of of τ decays: ~220M Peak Luminosity 9 x 10 33 cm -2 s -1

Fact # 4 *Boosting CM is not optimal for τ physics τ + 9.1 GeV e - 3.0 GeV e + CM boost at Y(4S) βγ=0.56 τ Must boost from lab frame to CM frame for tau studies like to work with one tau per hemisphere harder to unravel tau decay products than at symmetric machines: for some analyses the penalty may be high! BaBar τ µγ <2 10-6 56 10 6 τ-pairs (ICHEP 2002) CLEO τ µγ <1 10-6 13 10 6 τ-pairs (PRD 2000) Detector is not symmetric impacts tracking systematic errors

Search for Lepton Flavor Violation (LFV) in tau decay In the neutrino sector there is large LFV! neutrino oscillations θ 23 45 0, θ 12 32 0 solar neutrinos: ν e ν e /3+ν µ /3+ν τ /3 But in the standard model charged LFV is heavily suppressed. 2 ν τ ν 2 µ ( 40 ) ( 52 ( ) mij τ µ B τ µγ O 10 O 10 ) W W γ However LFV OK in many extensions to standard model. m 2 W SUSY models can bring B(τ µγ) all the way to 10-7!

Theory Says.. color coding: now next few years never

LFV & b s! SUSY GUT models can relate CP asymmetry in B φk s to B(τ µγ)! 1 2 ( m ) ( m ) e ϕ ϕ 2 2 d% 23 l% 23 i( ) GUT relation: R L b s penguin processes and τ µγare related in SUSY GUTs. From J. Hisano TAU 2004 CP asymmetry in B d φk s m U ν M τ µ 3 N τ = 2 5 10 ev, = 1/ 2, = tan β = 30 14 5 10, GeV

Current published 90%CL limits on BR(τ µγ) 11 x 10-7 @90%CL CLEO, 13.8/fb, Phys. Rev. D 61, 071101(2000) 3.2 x 10-7 @90%CL BELLE, 86.3/fb, Phys. Rev. Lett. 92, 171802 (2004) Search strategy: Use the known beam energy as a constraint Fit for beam-energy constrained mass: m EC = (E beam2 p µγ2 ) τ µγ MC Search for τ µγ Ε = Ε µγ Ε beam 0 in absence of radiation τ µγ MC

τ µγ Simulation Blinding box: ±3σ of expected m τ ±3σ of expected E Signal box: ±2σ of expected m τ ±2σ of expected E ISR γ at edge of acceptance BLINDED

Preselection Strategy - Restrict sample to 1-prong Vs 1 or 3-prong events - m EC fit & require γ with E γ CM >200MeV - ID track as a µ using a NN selector optimal for this search - keep events with: 1.5 < m EC < 2.1 GeV -1.0 < E < 0.5 GeV preselection events skim µ ID Backgrounds surviving preselection 1 Vs N prong m EC fit, E γ > 200MeV m EC, E region e + e - µ + µ - γ e + e - τ + τ (γ) e + e - qq 827M 734M 380M 13M 1.6M Still lots of background τ µ(γ) νν (~82%) τ π(k)(γ) ν τ π(k)π 0 ν

Signal and Tags Signal has partner tau decaying to generic modes: tag-side To maintain high efficiency, many tag modes used. Different backgrounds targeted in each mode. τ signal µγ 9.1 GeV e - 3.0 GeV e + τ tag τ tag τ tag τ tag τ tag τ tag τ tag eνν eγνν µνν hν h 1π 0 ν 3h 0π 0 ν

Tag-mode dependent selection Global cuts reduce backgrounds: thrust, θ miss, pµ CM, ptag CM, tag-side mass signal γ not merge π 0 <200MeV of extra γs on signal side Different backgrounds contribute to the different tag-modes: optimize selection criteria separately for each tag-mode. Employ combination of cuts and Neural Net (NN) NN is trained using data in GSB outside blinded region Placement of cuts on NN output are optimized using MC (optimized for best expected Feldman&Cousins limit)

Neural Net Five well modeled variables before NN, all modes cosθ H (cuts at ±0.8) -ln(2 p misst /Ecm) (cuts at 4 for 3h 2 for rest) -ln(m miss /E CM ) P tag CM m ν 2 cuts at ±0.4 for h, 3h ±0.8 for hg E n =(E cm /2-E had ) M miss = (E cm2 -P miss2 ) p ν =p τ -p had

Neural Net After NN (all modes): Still Good agreement cosθ H -ln(2 p misst /E cm ) -ln(m miss /E CM ) m ν 2 P tag CM

Results on Data Unblind the signal region: 4 events observed Background expected from data using SB:6.2±0.5 events Signal Efficiency = (7.45±0.65)%

Results by Tag Mode Good agreement between data and MC for each mode before and after NN in GSB, in predicted background and in selected events. Data: find 4 events in signal region, predict 6.2 MC: 2.8±1.7 events (66% µ-pair; 30% τ-pair)

τ µ γ Result Total signal efficiency = (7.45±0.65)% N bkg = 6.2±0.5 events N obs = 4 events N τ+τ = 1.97 x 10 8 (full BaBar data set) Preliminary BR(τ µ γ) < 0.9 x 10-7 @ 90%CL Including systematic errors does not change limit to the level of significance quoted BELLE: 3.2 x 10-7 @90%CL 86.3/fb Phys. Rev. Lett. 92, 171802 (2004)

Search for LFV: τ lll & τ lhh SUSY Models: B(τ lll) & B(τ lhh) in range 10-17 to 10-6 µ - µ - τ - h 0 µ + Pre B-Factory measurements at CLEO: B(τ µµµ) < 1.9 10 6 90% C.L B(τ eee) < 2.9 10 6 90% C.L B(τ e + π - π - ) <1.9 10-6 90% C.L + many other modes.

Analysis Technique Search for 20 different decay modes 3 leptons: 3e, 3µ, µee (2), eµµ (2) ehh: ekk (2), ekπ (3), eππ (2) µhh: µkk (2), µkπ (3), µππ (2) different charge states e + µ µ and e - µ + µ Tag Side: One well identified tagging track with missing momentum Signal Side: 3 tracks, identified by PID No missing momentum allowed on signal side

Analysis Technique Use M = M ec -M τ and E = E ec(cm) -E CM /2 Signal has M, E 0 MC: τ µee Signal box in red M, E smeared by detector and radiative effects Signal box is optimised for each channel.

Background Estimate from Data For each mode fit the data in the ( M, E) sideband to estimate background in signal region QED hadronic ττ M E

Published Results For τ lllτ PRL 92, 121801 (2004) (91.5 fb -1 ) DATA

Results For τ lhh No signals, just upper limits DATA (221.5 fb -1 ) similar results for 6 other modes!

Summary of 3 Body LFV Decays No signals found Set limits O(10-7 ) for 20 LFV modes (6 lll and 14 lhh) Limits have met up with upper end of theoretical predictions eg: SUSY with Higgs Triplet: B(τ lll) is 10-7 Can probe 10-8 (SUSY) region with higher statistics: backgrounds are under control: For τ lll analysis expect 3.41 events find 3. For τ lhh analysis expect 11.1 events find 10.

Tau Lifetime Measurement Motivation: precision check of lepton universality

Lifetime Measurement Technique Reconstruct decay length (λ t ) in transverse plane only Convert transverse decay length to 3D using: λ=λ t /sinθ approximate direction (θ) of tau using 3-prong momentum vector Do not use decay length errors as weights equalize acceptance in φ using 60 bins Average the λ measurements to get <λ> do not use maximum likelihood method calculate statistical error in <λ> using variance of 100 subsamples Calculate the average lifetime using: Mτ τ τ =< λ τ > < P > <P τ > calculated using MC Koralb (1st order) KK2f (2nd order, exponentiation) Subtract measurement bias using MC τ

Tau Lifetime Analysis Philosophy: sacrifice statistics for purity Select events where the tag tau decays via τ tag eνν use electron ID Signal tau decays into 3-prong no electrons or γ-conversions allowed no particle ID for the 3 hadrons Reduce backgrounds due to bhabhas, 2-photon, hadrons using: thrust visible charged energy (E ch ) missing energy ( s-e ch -E γ s ) min and max 1-prong momentum sinα=p T /( s-e ch ) High quality charged particle tracking 6 SVT hits on all 3-prong tracks 3-prong vertex fix P(χ 2 )>1% cut on vertex error BLINDED ANALYSIS Analysis uses 80fb -1 Total efficiency: 0.44% ~ 310k events

Data Vs MC Very detailed MC-data studies. Excellent agreement between MC and data

Tau Decay Length Measurement λ τ =λ t /sinθ <λ τ >=238.145±0.752 µm Detailed MC study to measure the bias in this procedure: event selection: -0.029±0.127% tau momentum reconstruction: +0.016±0.010% decay length reconstruction: +0.350±0.179% Total: +0.336±0.220%

Tau Backgrounds bhabha and 2-photon backgrounds: use data to estimate hadronic backgrounds: use combination of data and MC Note: large systematic uncertainty for continuum events (u,d,s,c). But negligible effect since contamination is very small.

Everything looks great BUT: : Data MC Measured decay length Vs φ is not flat! (but decay length Vs θ is!) Believe azimuthal variations in data are related imperfect SVT alignment Detailed studies using different SVT alignment sets show that the variations in φ DONOT lead to a decay length bias at 0.1% level.

Biases and Systematic Errors total bias total systematic

Tau Lifetime Results τ τ = 289.4 ± 0.91 ± 0.90 fs preliminary Lifetime asymmetry preliminary no systematic error yet

Lepton Universality

Search for 7-prong 7 τ decays Very rare no observation to date. Previous Experimental Result: BR(τ 7π(π 0 )ν τ ) < 2.4 10-6 (CLEO, 1997, PRD 56, 5297) Theory: Low BR a bit of a strong interaction puzzle. BR(1-prong): ~8.5 10-1 BR(3-prong): ~1.4 10-1 BR(5-prong): ~1 10-3 BR(τ 5hπ 0 ν τ ): ~2 10-4 BR(7-prong): <3 10-6 Perhaps can be explained (S. Nussinov, M. Purohit, 2002, PRD 65) BR(τ 7π(π 0 )ν τ ) < 6 10-11 (assumes no substructure) BaBar s Motivation: 25 CLEO s statistics possible observation Large enough sample: could help put bound on tau neutrino mass interesting channel to study resonant substructure

7-Prong Search Strategy Look for events satisfying 1-7 topology! Tag the 1-prong as: electron +0 or 1 γ muon +0 or 1 γ π/k, 0 γ ρ, 0 γ Develop criteria for the 7-prong that reject backgrounds: B mesons hadronic events (u, d, s, c) cross feed from tau events τ (5π) π 0 γγ e+ e - τ tag τ rec e - e + MC 1-7 event bhabhas, 2-photon, µ pairs negligible backgrounds Invariant mass of 7-prong side is the key! 1-prong side 7-prong side

7-Prong Search Strategy Invariant mass of 7-prong side must satisfy: m 7 <m τ (m τ ~1.78 GeV) BABAR preliminary Signal region Mass (GeV/c 2 ) Analysis is blinded : events below 2 GeV/c 2 are removed from the data.

Pseudomass! Pseudomass was introduced by ARGUS in 1992 to measure the τ lepton mass. Assume neutrino is massless τ direction is approximated by 7 ch. tracks m* τ2 =2(E beam E 7π )(E 7π P 7π )+m 7π 2 Events / 0.005 GeV/c 2 MC 7-prong Invariant and Pseudomass BABAR preliminary MC Advantage of pseudomass: Mass (GeV/c 2 ) Sharp cut-off at the τ mass (1.777 GeV/c 2 ). significant improvement of signal-background separation Events / 0.01 GeV/c 2 BABAR preliminary MC Events / 0.01 GeV/c 2 MC BR=2.4 10-6 BABAR preliminary Invariant Mass (GeV/c 2 ) Pseudomass (GeV/c 2 )

Background Estimate Fit pseudomass distribution from 2 to 2.5 GeV/c 2 after thrust cut with a Gaussian function Use these fit parameters on the pseudo-mass spectrum after all cuts. Extrapolate the fit below 2 GeV/c 2 Integrate from 1.3 to 1.8 GeV/c 2 (signal region) Check procedure using data and MC Events / 0.025 GeV/c 2 DATA After thrust cut fit extrapolate integrate Events / 0.025 GeV/c 2 DATA After all cuts µ, σ BABAR preliminary Pseudomass (GeV/c 2 ) Pseudomass (GeV/c 2 )

Background Estimate Validation Use 1-8 prong DATA 1-8 Data after thrust cut 1-8 Data after all cuts Events / 0.025 GeV/c 2 BABAR preliminary Events / 0.025 GeV/c 2 BABAR preliminary Pseudo-Mass (GeV/c 2 ) Pseudo-Mass (GeV/c 2 ) Cuts Expected Observed Thrust mag. 41 ± 10 57 7-prong π ID 29 ± 7 32 p T 19 ± 5 22 DOCA XY /p T 7.7 ± 2.3 8 1-prong tag 2.0 ± 0.6 1 Also do MC study expect: 1.8 ± 0.7 observe: 1

Preliminary Results After thrust cut After all cuts Events / 0.025 GeV/c 2 BABAR preliminary extrapolation of fit Events / 0.025 GeV/c 2 BABAR preliminary signal region Pseudo-Mass (GeV/c 2 ) Events in signal Region: observed events: 7 expected background: 11.9 ± 2.2 Pseudo-Mass (GeV/c 2 ) No evidence for 7-prong τ decay

Preliminary 7-prong 7 BR Upper Limit B A B A R p r e l i m i n a r y Ν τ+τ 1.1 10 8 τ + τ background 0.6 ± 0.4 qq background 11.3 ± 2.2 Total expected background 11.9 ± 2.2 τ 4π 3π + ν τ efficiency (8.05 ± 0.55) % τ 4π 3π + π 0 ν τ efficiency (8.04 ± 0.55) % systematic error BR (τ 4π 3π + (π 0 ) ν τ ) @ 90% CL < 2.7 10-7 Experiment CLEO (1997) BaBar Luminosity (fb -1 ) 4.6 124.3 Observed (predicted) events 0 (2.8) 7 (11.9) BR (τ 4π 3π + (π 0 ) ν τ ) @ 90% CL < 2.4 10-6 < 2.7 10-7

τ 5(hadron)ν final States Tau semi-leptonic decays are an ideal place to study strong interaction effects sum rules, spectral functions, Wess-Zumino anomaly The large sample of tau data from BABAR allows us to re-examine many of the well-known hadronic decays and to look in detail at the decays that were previously limited by statistics (Previous 5-prong studies done with ~300 events) This study focuses the 5-hadron final states: τ 2π π + 2π 0 ν τ (η, ω, φ π π + π 0 ) τ 3h 2h + ν τ (no particle ID for hadrons)

Analysis Strategy Reduce non-tau backgrounds using standard techniques: event variables: thrust, transverse momentum, missing energy split event into signal and tag hemispheres (1 Vs 3 or 1 Vs 5) require 1-prong tag to be either electron or muon τ 2π π + 2π 0 ν τ Tracks required to be π s 2 π 0 candidates M(5pion) < 1.8 GeV 9900 candidates Background 15% Mainly tau decay background Efficiency 0.65% τ 3h 2h + ν τ π or kaon tracks allowed Veto electrons, γ conversions and π 0 s M(5hadron) < 1.8 GeV 15869 candidates* Background 21% tau decays with π 0 s or K 0 s Efficiency 4.0% *PDG result based on ~ 300 events

Comparison with other measurements systematic errors: τ 2π π + 2π 0 ν τ : 12.3% Will decrease soon! dominated by the background (9.3%) and π 0 finding (6.5%) τ 3h 2h + ν τ : 4.7% tracking (3.1%), luminosity σ (2.3%), π 0 s (2%) B h h τ + 4 ( τ 3 2 ν ) (10 ) BaBar 8.5 (0.4) OPAL 9.1 (1.5) ALEPH 8.7 (1.7) Preliminary B( τ 2π π 2 π ν ) (10 + 0 3 τ BABAR 4.51± 0.07 ± 0.56 CLEO 5.2 ± 0.5 ALEPH 5.0 ± 0. 7 ± 0.7 ) CLEO 7.7 (1.0) PDG 8.2 (0.6) Good Agreement! 6 8 10 12

Entries / 2 MeV/c 2 140 120 100 80 60 40 20 η π π + π 0 BABAR P R E L I M I N A R Y Structure in τ 2π π + 2π 0 ν τ Β(τ ηπ π 0 ν τ ) Β(τ φπ π 0 ν =(1.71±0.12±0.16) 10-3 τ )< 4.5 10-4 90% CL 3 (1.74 ± 0.24) 10 PDG proceeds through Wess-Zumino anomaly Entries / 5 MeV/c 2 250 200 150 BABAR P R E L I M I N A R Y φ π π + π 0 0 0.5 0.52 0.54 0.56 0.58 0.6 M(3π) / GeV/c 2 100 Entries / 5 MeV/c 2 1200 1000 800 600 ω π π + π 0 BABAR P R E L I M I N A R Y Β(τ ωπ π 0 ν τ ) =(3.66±0.10±0.31) 10-3 3 (4.3± 0.5) 10 PDG 50 0 first upper limit! 0.9 0.95 1 1.05 1.1 1.15 1.2 M(3π) / GeV/c 2 400 200 0 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0.9 M(3π) / GeV/c 2 η s and ω s but No φ s!

The fun here is just beginning Unraveling all the substructure is a challenge many possible final states. states with charged K s many resonances: e.g. have hint of τ f 1 (1285)π ν τ (Β(f 1 (1285) 2π 2π + )=11% ) understanding of 5 prong final states improvement in ν τ mass limit Also a challenge to QCD and model builders! τ 3h 2h + ν τ Present MCs (tauola/kk2f) use phase space which does not represent data! m(5h)

Summary & Conclusions BaBar is now turning out tau results! new results in many areas: searches lifetime BRs resonant substructure Many more results to come: more LFV searches, e.g. τ eγ second class currents (e.g. τ ηπν, wrong G parity for vector current) final states with kaons (BRs, spectral functions, V us, m s ) improved limit on mass of tau neutrino (~10MeV) searches for CPV??????