220 likes | 362 Views
Update on the Inclusive Measurement of the b s Transition Rate Using a Lepton Tag Using Run I-V Data. Philip Bechtle (until 5/07) * , Rainer Bartoldus SLAC Colin Jessop, Kyle Knoepfel Notre Dame University Al Eisner, Bruce Schumm, Luke Winstrom UC Santa Cruz Minghui Lu
E N D
Update on the Inclusive Measurement of the b s Transition Rate Using a Lepton Tag Using Run I-V Data Philip Bechtle (until 5/07)*, Rainer Bartoldus SLAC Colin Jessop, Kyle Knoepfel Notre Dame University Al Eisner, Bruce Schumm, Luke Winstrom UC Santa Cruz Minghui Lu University of Oregon John Walsh University of Pisa Students * Now at DESY Bruce Schumm SCIPP 6/07 BaBar Coll. Meeting
Direct searches (LEP) b s is a leading constraint on new Electroweak scale physics… The SM transition is high order (two weak plus one EM vertex… So new physics can enter at leading order SUSY Extra Dimensions B s constraints MSSM Constraints
(J. Walsh) b motion 1.9 b s also provides universal constraints on hadronic effects Photon spectrum can be used to measure universal heavy quark parameters (largest uncertainty in |Vub| from inclusive measurement of b ul) In addition to partial BF, we measure 1st and 2nd moments of the photon distribution
Current Status of b s BF Measurements BaBar 2006 inclusive result (Run I-II only): B(B Xs ; 1.9 < EB < 2.7) = 3.67 0.29 0.34 0.29, where errors are statistical, experimental, and model uncertainty, and EB is the photon energy in the B rest frame. Phys.Rev.Lett.97:171803,2006 BaBar Sum of Exclusive Modes Run1-2 Babar Fully Inclusive To interpret the partial BF, one must extrapolate from EB = 1.9 GeV (experimental lower limit) to EB = 1.6 GeV (where theoretical calcul-ations are done). [We are not yet concerning ourselves with that step for Run 1-V analysis.]
qq + ττ BB XSγ Inclusive b s: little effect from long distance physics, but how do you eliminate backgrounds? • Continuum Bkgds: • Shape variables (was Fisher discriminant; now Neural Net) • Lepton tag indicates heavy flavor in “rest-of-the- event” decay • (4S) Bkgds: • Reconstruct (usually asym- metric) 0 and decays • Calorimeter cluster shapes sup- press merged 0s, hadrons Source: BAD 323, based on the 81 fb-1 Run I-II sample
Sig. Region B/Bbar background control region BB Cont. Signal After Selection Cuts Source: BAD 323, based on the 81 fb-1 Run I-II sample • And then… • Subtract off small remaining continuum using off-resonance (dominant statistical term) • Develop independent estimates B/Bbar backgrounds and subtract them (critical step) • Confirm B/Bbar estimates with control region • Theorists would love us to push below 1.9 GeV, but B/Bbar backgrounds intimidate… What are the sources of B/Bbar background?
B/Bbar background contribution “guess” (selection not yet finalized) 82% of B/Bbar background Nominal B/BBar Background Sources Electron categories x2 larger than that of prior simulation (was 3.7% combined). This raises questions, in-cluding the modeling of bremsstrahlung
Constraining the 0 - Background with a Measurement of Inclusive Production • Measure p0/h yields in on- and off-peak data and MC • Determine MC correction factors in bins of E(p0): Correction = [(On-peak data) – s*(off-peak data)]/[BB MC] • Use corrected MC to predict background contribution • Also need to know recon. efficiency. of background p0s gg invariant mass MC Correction Factors Fits done to both data and MC
How Do We Reconstruct 0s and ’s? From Run I-II Analysis; subject to further optimization for current Run I-V result • Begin with reconstructed high-energy (HE) with cms energy E* • Search GoodPhotonsLoose list for potential sibling with the following minimum lab energy (E2,lab) requirement (from Run 1-2; not yet optimized for current analysis): • Find potential sibling that, in combination with HE , has invariant mass M closest to the 0 () mass. • Reject event if 115 < M < 155 (508 < M < 588) MeV for the best 0 () combination.
Require 2nd photon to be above minimum energy cut Require 2nd photon to be in fiducial volume -.74 < coslab < .94 1 2 3 E* coslab Of remaining bkgd events, almost all make a good 0 candidate with the HE Require 2nd photon to have a truth match E* E* And with What Efficiency? If high-energy (HE) truth-matches to a 0 daughter, make succession of requirements on MC truth properties of other (low-energy) daughter • Observations: • Typically reconstruct only about ½ (depends on E*) of background 0s • 20% truth-matching inefficiency; only about 6% due to merged 0s. Could the rest be conversions? • must understand conversion effects to subtract background correctly (not appreciated before)
Material and the Inclusive Measurement of b s • Material enters into the measurement of b s in three substantial ways: • Conversions HE efficiency, • Conversions 0 reconstruction efficiency • Bremsstrahlung electron fake rate • There are complications associated with esti- • mating these effects. For example, a photon • converting in the DIRC may or may not be • reconstructed as the original photon, depending • on its energy, the depth in the DIRC, etc. • This must be understood, in addition to the distribution of material in the detector and the brem/conversion cross-sections. Additional control samples may need to be developed and applied (“radiative bhabha” to understand bremsstrahlung?).
E2,lab M E* More clever rejection of 0 backgrounds? ( analysis used likelihood based on mass and E2,lab) try NN rejection Run I-II analysis performance Signal Efficiency Using E* information Variables considered: M E* E2,lab coslab HE 2nd moment HE isolation HE Lat. Moment LE 2nd moment LE isolation LE Lat. Moment Signal Efficiency Ignoring E* information Most power in M, E2,lab (already in use) and E* (dangerous). Will not pursue. Background Efficiency
Continuum Suppression for Run I-V Analysis • Develop Neural Net to make most efficient use of shape variable information. Inputs include Fox-Wolfram moments, lepton tagging variables, energy-flow variables: • Two classes of NNs, separated by energy-flow approach: • Energy cones (three variants) • Two different cuts on NN output (standard and relaxed) • One without lepton momentum • Legendre moments plus momentum-tensor quantities (similar to sphericitiy tensor) • Prior (Run I-II) analysis used Fisher Discriminant composed only of shape variables • Note: At the end of the day, the continuum subtraction will be determined from the off-resonance data, not from a-priori understanding of the NN efficiency
% of total Error Statistitical Model Systematic Neural Net Selection: A Word About Run I-II Syst. Errors Run I-II Result (Phys.Rev.Lett.97:171803,2006 ) Br (BXsg) = (3.67 0.29 0.34 0.29) x 10-4 Different b s models (b mass, Fermi motion) E* [GeV] Selection efficiency vs. E* for Run I-II selection Important: Run I-V optimi-zation must consider both statistical and systematic (especially model) error! E* [GeV]
Event-Shape NN Selection • Consider both partial BF as well as moment calculations. All in all… • None of the candidate NNs is clearly preferable • Choose Legendre-moment-based NN in view of its modest dependence of signal efficiency on E* Eff vs. Eg* Econes I • better statistical precision • larger model error eff. slope = 3.2 Legendre Moments • more stats in p0/h control sample • reduced model error eff. slope = 1.5
Other Backgrounds: Antineutrons Was 7.7% of the B/Bbar background for RUN I-II Contribution can be constrained by looking at antiprotons. Must understand: Production Rate Two components: fragmentation and decay; have different isospin relations (p/n fraction) and different momentum spectra Working with hadronics group (D. Muller) to sort out. Signature in EMC Use -bar sample (high momentum) [Develop dE/dX-identified sample (low momentum) ?] Data MC ECAL Lateral Moment
Other Backgrounds: and ’ BAD 163 : nominally 2.1% of B/Bbar background; d/dp* measured; use to correct rates in MC (correction factor “”) BAD 179 + private updates /: nominally 0.8% of B/Bbar background; less well-constrained, but less of a contribution.
Other Backgrounds: B X Simulation estimates that HE backgrounds photons with B meson parents are twice as common than that of Run I-II simulation (1.4% vs. 0.7% of B/Bbar background) . These gammas seem to be coming predominantly from SL decay; how well do we understand this number? Why did it change in the MC simulation?
b s Outlook I An admirable goal would be Lepton/Photon – what kind of shape are we in? • The lepton-tagged inclusive analysis is gelling… • CM2 migration complete • Low-energy truth-matching work-around • Shape-variable selection (NN) finalized • 0 and production rates measured • 0 background rejection revisited • Several other selection cuts established (merged 0s …) • A number of “standard” things remain (treatment developed for Run 1-2) • Anti-neutron rejection criteria • Final optimization • “Control region” test of B/Bbar background contribution • Estimation of most sources of systematic errors
b s Outlook II • However, some new considerations have arisen • Brehmsstrahlung and conversions (material effects) • Non-DST level study of conversion, brehm properties • New control samples (radiative Bhabha?) • Understanding of direct B backgrounds. • Also, the loss of Philip Bechtle (to DESY) was a set-back, but students (Kyle, Luke) now coming up to speed on production code. • Initial preliminary results will include measurements of: • Partial branching fraction (1.9 < E* < 2.7) further tighten constraint on new physics • 1st and 2nd moments of photon energy distribution generic constraint on fermi motion of b quark • ACP Independent probe for new physics (current: -.110.115.017) • We have our work cut out for us…