Measurement of the Ë b Lifetime in the Decay Mode Ë b ! J= Ë LEPP Journal Club January 22, 2004 Robyn Madrak Harvard University Decay Modes • First seen unambiguously at CDF in Л b ! J= Л – Only measured mass • Lifetimes measured at – CDF/LEP in Л b ! Л clч ц – LEP in baryon decays tagged by Лl 2 Why measure the Ë b Lifetime ? • Test theoretical models used in predictions of heavy quark quantities Not only interesting in themselves - need to extract weak interaction quantities from measureables • Currently, experiment and theory disagree • Besides this measurement, lifetime measured only in semileptonic decay modes • Initially expected to have huge amount of data by now very competitive measurement 3 Discrepancy: • Theory predictions work well for B à ; B s • But not Л b (though others claim 0.85 could be accommodated) 4 Measurements up to Now average b-baryon lifetimeЛ all semileptonic measurements b 5 Sources of Lifetime Differences In the simple spectator model, same lifetime for all hadrons with same heavy quark • The heavier the quark, the more valid the approximation È( b) = 9V2cbG 2FM 5b 192ù 3 • With more sophisticated theory, lifetimes are different, due to Pauli Interference Weak Annihilation (for mesons) Weak Exchange, or scattering (for baryons) 6 Sources of Lifetime Differences: 1.) Pauli Interference Same final state =>interference Different final states 7 Sources of Lifetime Differences: 2.) Weak Annihilation • Only in mesons • In B-but not B 0 8 Sources of Lifetime Differences: 3.) Weak Exchange: Main source of lifetime difference for B 0=( Ë b ) • Helicity suppressed in mesons, but not 9 baryons (no antiquarks) The Fermilab p pbar Accelerators • Upgraded for Run II • cm energy of 1.96 TeV (was 1.8) • 36p on 36pbar bunches (was 6X6) • 396 ns bunch crossing time: required many detector upgrades (was 3.5 us) 10 The CDFII Detector Central Calorimeter Time of Flight Muon Chambers CMU CMP CMX Plug Upgrade Calorimeter Beamline Silicon Vertex Detector (SVXII) Solenoid Drift Chamber (COT) 11 The CDFII Detector 12 The COT •End view: Fraction of the endplate •8 “superlayers” • alternating planes of sense wires (readout) and field sheets(ground) • alternating axial and stereo (2 °) superlayers Closeup of cell layout in 1 superlayer 13 The COT sense wires field sheet Segments (in 8 layers) Tracking in a Nutshell: •Form line segments in 4 axial layers •Do axial fit, connecting segments •Form segments in stereo layers •Add stereo segments to fit 14 •Final fit This measurement: Full Reconstruction Ë B ! J= Ë : baryon equiv. of golden mode J= K 0s • Ë ! pù J= ! ö + ö à • Reconstruct p + ; ù à ; ö + ö à Cons: • Fewer events - larger statistical error Pros: • Potentially smaller systematic error • Less background due to full reco • Have a signal invariant mass peak signal and background regions well defined • Good vertex for decay length (use multiple tracks) • Do not rely on MC as in semileptonic case (can’t 15 measure momentum of n boost(bg) is unknown) Data Sample • 65 pb^-1 with SVX fully functional • J/y sample: from dedicated J/y trigger Muon stubs (J/y->mm) MC event Level 1: Two online, opp. Q tracks, pT>1.5 GeV, good trackstub match (CMU or CMX) Level 2: Auto Level 3: Full tracking, cut on m(mm), track-stub matching • Reconstructing L->pp: compute invariant mass of all opp. Q tracks use p/p mass hypothesis for higher/lower momentum Proton track particle (Need maximal efficiency: No dE/dx or TOF) Pion track 16 Scheme for Measuring the Lifetime • Use control sample: B 0 ! J= K 0s • kinematically similar • well-known lifetime: ct = 2.7 cm 462 æ5öm ct = 7.9 cm same cuts except for a few cases (m(L) not m(Ks)) 4 track kinematic fit: vertex constrain J/y and L or Ks (V0’s) Mass constrain M(mm) to world average m(J/y) (better mass resolution) Constrain Ks to “point” back to J/y • Measure control sample lifetime agrees with world average? If so, measure Lb lifetime 17 Analysis Cuts •No dE/dx cut •No TOF use in future with more statistics Lb: Need max efficiency now •pt cut only for p’s from Ks, not efficient for L 0 •get rid of B •remove pairs of real J/y and L or Ks 0 that are not B, Lb Quantity Value Purpose Tracks N(COT axial) > 20 well -measured momenta N(COT stereo) > 16 Muons J/y > 2 rf SVX hits well -measured Lxy pT(m)>1.5 GeV reinforce trigger prob(c2) > 0.1% quality vertex 3.01 < m(mm) < 3.17 GeV V0 Ks prob(c2) > 0.1% quality vertex Lxy > 0.25 cm bg rejection 0.477 < m(pp) < 0.517 pT(p) > 0.4 GeV L B/Lb 1.104 < m(pp) < 1.128 ! 0.48 < m(pp) < 0.513 Ks veto prob(c2) > 0.1% good 4-track vertex cos(q) > 0.9999 18 pointing J/y Candidates • ~600k candidates with 3 or more r-f SVX hits • S/B = 5.1 • width = 17.6 MeV 19 Ks and L Candidates After Cuts •Very clean Most B/Lb background comes from pairs of 0 real J/y and real Ks, L 20 0 0 Event Yields: B control sample 21 Event Yields: Lb 22 •J/y Ks background: expect 1 event at <= 5.5 GeV A Lb candidate event in data muon tracks proton pion 23 Proper Decay Time 500 mm • Need proper decay time ct for each event 10 cm primary vertex secondary vertex • Need pT(Lb) ct = L 1 бì í = L mB б pB = mB L xy б pB T 24 Unbinned Maximimum Likelihood Fit • Fit mass histogram to determine signal/background regions, and background fraction fb • Functional for for ct in signal region: F isi g( cьB ) = Exp( ct i ; cьB ) к G( ct i ; ыict ; sct ) G:detector resolution smearing • And in background region: F ibgr = (1 а f à а f + ) бG( ct i ; ыict ; sct ) Exp: real long lived distribution + { f+ ( à ct) i =õ + б e ; õ+ fà (ct) i =õ à б e ; õà • Fit signal and background simultaneously ( ct ) i > 0 ( ct ) i < 0 25 Control Sample Mystery • The control sample lifetime came out low! ct=363 27 mm PDG = 462 mm significance = 3.7s Why??? • Measured the correct 0 lifetime with 5k B events with full simulation • Checked for fitter bias/bugs using Toy MC (next) 26 Fitter Checks with Toy MC • Want to check fit for any possible bias, bugs Generate 5k “fake” experiments (Toy MC) ° same number of events as in data, same S/B these have: º distributions functional forms same as in data º errors on ct drawn from a histogram (of data errors) B toy MC generated ct = 414 mm ave fit ct = 4130.45 B toy MC - ct PULL fit mean = 4 2 mm width = 1.05 0.02 27 Answer: COT Tracking Algorithms • 1 Two algorithms: SL (from Run I) links full line segments (in each of 8 layers) 2 HL (new) for high pT efficiency - looks for hits along a line to the beamline after finding an outer 0 layer segment - easy to believe it would be biased - IS biased for Ks Segments (in 8 layers) Use ONLY HL: Use ONLY SL: ct = 33836mm ct = 41431mm •ALSO: •by dropping “supplementary” algorithm, we lose NO events For this analysis we 28 use SL only 0 Fit for B Control Sample F = (1 à f b)F s + f bF b F s = G ê Exp(cü) F b = (1 à f + à f а ) â G +{ (f + =õ + ) áe( ct=х + ) (f + =õ + ) áe(а ct=х + ) ct < 0 ct > 0 Fit Results G (prompt bg) l- (tails) within 1.5 s of PDG world ave of 462 mm l+ (tails + real long-lived bg) fit quantity value error ct 414um 31 sct 1.26 0.05 f- 0.029 0.009 l- 499um 139 f+ 0.13 0.02 l+ fb 307um 43 0.571 0.028 29 Fit for Lb Lifetime cь( Л B = 374 ж 78 цm) = 1:25 ж 0:26 ps 30 Cross Checks and Systematics •Cross Checks: •Fitting Method •Splitting of data into separate samples •Treatment of long-lived particles (V0’s) •Systematics •SVX Alignment •Fitting Model •Gaussian resolution function •“Careful Systematics” (from cross checks) 31 Alternative Fitting Method:2d Simultaneous Mass and Lifetime Fit • Fit mass and ct distributions simultaneously: F ilik = ( 1 а f B ) бF isig бM isig + f bgr бF ibgr бM ibgr • gaussian: M isig • linear: M ibgr • Fit parameters: f cь; sct ; f B ; f + ; f à ; х + х à ; m B ; ыm B ; C 0g 32 Cross Check #1: Results with 2d Simult. Fit B0 Lb • With this fit: 420 ж 42цm • With this fit: 351 ж 69цm • With separate fit: 414 ж 31цm • With separate fit: 374 ж 78цm 33 Cross Check #2: Luminosity Effect in Control Sample •Observed B mass width larger in later data than in earlier •Could not find specific change in detector configuration as cause Divide data into bins of instantaneous luminosity L < 1:17 cm а 2sа 1 1:17 < L < 1:7 cm а 2sа 1 Low lumi L > 1:7 cm а 2sа 1 mid lumi hi lumi 34 Cross Check #2: Luminosity Effect in B 0 Control Sample Standard (separate) fit: •low lum: 43839mm Standard (separate) fit: •hi lum: 38355mm OK (only 0.8 s difference) 35 Cross Check #3:Treatment of V0 ‘s •Ks and L are long-lived (ct=2.7 cm and 7.9 cm) •Many decay outside of SVX •We use only SVX hits consistent with being on tracks (based on COT-only info) •A reasonable analysis may have rejected all SVX hits for36V0’s V0 Tracking: Cross Check Results for B 0 •With COT-only tracks for Ks, lifetime is smaller: Compare to 414 31 mm • But in that case, the sample changes, and the difference is still < 1 s OK 37 d vs. f : 3 barrels: Systematic #1: SVX Alignment Getting the systematic: Barrel 1 Barrel 2 Barrel 3 Projection: RMS = 7 mm -Use higher statistics 80 mm B+ J/y K+ -Try a series of “inferior” alignment tables -Also try alignment where SVX wafers are “bowed” out systematically by 100mm -Quantify variation in lifetime: 5 mm 38 Systematic #2: Fitting Model •Current model is well motivated, but can imagine others: > Convolute background exponentials with gaussian resolution (instead of adding) B lifetime is 11 mm lower > Add an additional positive going tail in background function no change > Use mass errors and fit for scale factor on mass errors instead of B width B lifetime is 22 mm higher > Separately fit sidebands and signal region B lifetime is 2 mm higher •Take largest variation as sytematic: 22 mm 39 Systematic #3: Resolution Function 1 ) Fit for sct: scale factor for resolution 1.26 0.05 2) Other studies with inclusive J/y’s (not specifically B’s): 3) Fit those results to a function f 0:950 + 0:0308 г (pT(J= ) а 2:86) 1:075 1:036 pT( J= ) < 9 9 ф pT( J= ) ф 10 pT(J= ) > 10 4) Scale our errors by function and fix sct to 1.0 in lifetime fit 5) Remeasure B lifetime: 1mm smaller 40 Luminosity Effect in B Control Sample Revisited: Cross Check of a Cross Check 0 •Separate fit: •low lum: 43839mm •high lum: 38355mm •All OK (only 0.8 s difference) •BUT with 2-d simultaneous fit: •low lumi: 45540 mm, hi lumi: 32450 mm: 2s difference •STILL USE SEPARATE FIT FOR CENTRAL VALUE, THOUGH Assign systematic for Lb as N ss(t ot al) в ( cьpdg а cь(hi where ct(hi L) is from the separate fit N (hi L) L) ) = 25mm 41 V0 Tracking: Results for B 0 • Looked into treatment of Ks before… now we look a little deeper… •Divide into bins of Lxy(Ks): Cross check from before: this was OK •The worry here: Very small lifetime for COT-only in smallest bin of Lxy(Ks) 42 V0 Tracking: Another Effect • Given this and the previous effect, Assign systematic as: cь( B o)(our method) а cь( B 0)(COT а only K 0s) = 26цm Allows SVX hits when appropriate 43 Summary •Systematics: Source Value(mm) ct Resolution function SVX Alignment * 1 5 Occupancy 25 Fitting Model 22 V0 Tracking 26 Total 43 *From Bs lifetime control sample cü(Ë b) = 374 æ 78(stat:) æ 43(syst:)öm ü(Ë b) = 1:25 æ0:26( stat:) æ0:14( syst:)ps (world average = 1.229 0.080 ps ) 44 Implications with •ct(Lb) = 374 89 mm (this measurement) •ct(B) = 462 5 mm ü(Ë b) ü(B 0) = 0:81 æ0:19 45 More Data: Predictions < 10% error with 3X more data Already on tape! 46 Conclusions We’ve measured the Lb lifetime in LbJ/y L ü(Ë b) = 1:25 æ0:26( stat:) æ0:14( syst:)ps From this we extract a lifetime ratio which is consistent both with theory and the current world average (Though the two of those disagree) This result will be much more interesting with 3 times the data (The data are available now! Work in progress...) 47 Operator Product Expansion • Coefficients ci calculable within perturbation theory •Matrix elements contain long-distance physics •Terms 1 and 2 calculable both for mesons and baryons •Term 3 is for PI, WA, and WS, harder for baryons •lifetime differences between mesons begin in term 3, between baryons and mesons in term 2 48