aa r X i v : . [ h e p - e x ] O c t Charmless b -hadron decays at CDF Diego Tonelli (for the CDF Collaboration)
Fermilab, MS 223, P.O. Box 500 Batavia, IL 60510-500, USA, [email protected]
Measurements from the upgraded Collider Detector at the Fermilab Teva-tron (CDF II) are becoming increasingly competitive with B -factories resultson B decays into charged final states, and complementary to them in cor-responding B s and baryon modes [1]. In addition, the reached sensitivityto flavor-changing neutral current (FCNC) b -meson decays could reveal newphysics before the start-up of the Large Hadron Collider (LHC) [2].We present recent results on these topics, from samples corresponding totime-integrated luminosities of R L dt ≃ . − . C -conjugate modes areimplied throughout the text, branching fractions ( B ) indicate CP -averages,and the first (second) uncertainty associated to any number is statistical (sys-tematic). Details on the CDF II detector can be found elsewhere [3]. B s ) → h + h ′ − decay rates CDF is the only experiment, to date, that has simultaneous access to B and B s two-body decays into charged kaons and pions ( B s ) → h + h ′ − ). Joint studyof these modes, related by flavor symmetries, may allow (partial) cancellationof hadronic unknowns in the extraction of quark flavor-mixing parameters.We analyzed a R L dt ≃ − sample of pairs of oppositely-chargedparticles, used to form B s ) meson candidates, with p T > c and p T (1) + p T (2) > . c . The trigger also requires a 20 ◦ < ∆φ < ◦ transverse opening-angle between tracks to reject light-quark background. Inaddition, both charged particles must originate from a transversely-displacedvertex from the beam (100 µ m < d < B s ) meson candidatemust be produced in the primary p¯p interaction ( d ( B ) < µ m) and totravel a transverse distance L xy ( B ) > µ m. Diego Tonelli (for the CDF Collaboration) A B s ) → h + h ′ − signal ( B ≈ − ) of about 15,000 events and signal-to-noise ratio SNR ≃ . B or A CP ), as predicted from repeating the actual mea-surement on pseudo-experiments. We also exploit the discriminating power ofthe B s ) meson ‘isolation’ and of the information provided by the 3D-view ofCDF tracking, which both greatly improve signal purity. Isolation is defined as I ( B ) = p T ( B ) / [ p T ( B ) + P i p T ( i )], where the sum runs over every other trackin a cone of unit radius in η − φ around the B s ) meson flight-direction. The I ( B ) > . b -mesons withrespect to light-quark background. The 3D-view of tracking allows resolvingmultiple vertices along the beam direction. This halves the combinatoric back-ground, with little inefficiency on signal, by removing pairs of displaced tracksfrom distinct, uncorrelated, heavy-flavor decays.The resulting ππ -mass distribution (Fig. 1, right) shows a clean signal,estimated by a Gaussian plus an exponential (combinatoric background) andan Argus-shaped (partially reconstructed B decays) fit to contain about 7,000events, with standard deviation σ = 39 ± c and SNR ≃ . B s ) → h + h ′ − modes appear overlapping into an unresolvedmass peak. Indeed, the mass and PID resolutions are insufficient for separat-ing them on a per-event basis. We achieved a statistical separation with amultivariate, unbinned likelihood-fit that uses PID information, provided byspecific ionization energy loss ( dE / dx ) in the drift chamber, and kinematics.We exploit the kinematic differences among modes by using the correlationbetween masses and (signed ratios of) momenta (Fig. 1, left). Mass line-shapesare accurately described accounting for the effect of final state radiation ofsoft photons and non-Gaussian resolution tails. The dE / dx is calibrated overthe tracking volume and time using about 10 , 95% pure, D ∗ + → D ( → K − π + ) π + decays, where the identity of Cabibbo-favored D decay-productsis tagged by the strong D ∗ + decay [5]. A 1 . σ separation is obtained betweenkaons and pions with p > c . A 10% residual track-to-track correlationdue to uncorrected common-mode dE / dx fluctuations is included in the fit.Kinematic fit templates are extracted from simulation (signal) and from realmass-sidebands data (background); dE / dx templates (signal and background)are extracted from the D samples used in calibration.The fitted yields reveal the first observation of B s → K − π + (230 ± ± σ significance), Λ b → p π − (110 ± ±
16 events, 11 σ significance),and Λ b → p K − (156 ± ±
11 events, 6 σ significance) decays. After cor-recting for trigger, reconstruction, and selection efficiencies, we obtain the harmless b -hadron decays at CDF 3 q · ) /p = (1 - p a -1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1 ] c - m ass [ G e V / pp I n va r i a n t - p + p fi B - p + K fi B + p - K fi B - K + K fi B CDFII Monte Carlo ] -mass[GeV/c pp Invariant C a nd i d a t es p e r M e V / c ] -mass[GeV/c pp Invariant C a nd i d a t es p e r M e V / c -1 =1 fb int CDF Run II Preliminary L - p + K fi B + p - K fi B - K + K fi s0 B/ s0 B - p + p fi B/ B - p + K fi s0 B + + p - K fi s0 B + p p fi b0 L + - p p fi b0 L + Kp fi b0 L + - pK fi b0 L Combinatorial backg.Three-body B decays
Fig. 1.
Invariant ππ -mass for simulated B decays as a function of the signedratio of momenta, α = q min (1 − p min /p max ), where “min” and “max” refer to themagnitudes of momenta, and q is the sign of the charge. Similar dependencies holdfor B s and Λ b decays. Invariant ππ -mass after the offline selection with individualsignal components (cumulative) and backgrounds (overlapping) overlaid. following branching fractions: B ( B s → K − π + ) = (5 . ± . ± . × − , B ( B s → π + π − ) = (5 . ± . ± . × − , and B ( B → K + K − ) = (3 . ± . ± . × − . The extracted CP -violating asymmetries, A CP ( B → K + π − ) =( − . ± . ± . A CP ( B s → K − π + ) = (39 ± ± dE / dx shapes, isolationefficiency, combinatorial background shapes, and charge-asymmetries in back-ground. Further details on the analysis can be found in Ref. [7]. B meson decays In the standard model (SM), FCNC decays are strongly suppressed: O (10 − − − ) expected branching fractions for rare B s ) → µ + µ − decays are a factor O (100) beyond current experimental sensitivity. However, contributions fromnon-SM physics may significantly enhance these rates, making possible anobservation that would be unambiguous signature of new physics.We searched for B s ) → µ + µ − decays in R L dt ≃
780 pb − of data col-lected by the dimuon trigger. Offline, we require two oppositely-charged muoncandidates fit to a common decay-vertex. We cut on the dimuon transversemomentum to reject combinatoric background, on the 3D decay-length ( λ )and on its resolution to reject prompt background, and on the isolation; wealso require the candidate to point back to the primary vertex to further reducecombinatoric background and partially reconstructed b -hadron decays. Thisresults in about 23,000 candidates, mostly due to combinatoric background.Further purity is obtained by cutting on a the likelihood-ratio (LR) basedon three input observables: the isolation of the candidate, the decay-length Diego Tonelli (for the CDF Collaboration) ) R Likelihood Ratio (L0.9 0.95 1 / G e V / c mm M CDF II Preliminary ) -1 (780 pb d B s B CMU-CMUCMU-CMX ) ) (GeV/c fmm m( C a nd i d a t es p e r M e V / c Data f mm fi s BSignal regionSideband regionExtrapolated fit -1 CDF Run II Preliminary L~1fb
Fig. 2.
Invariant µ + µ − -mass versus LR distribution (left). Both muons in the | η | < . . < | η | < . B s (blue box) and B (red box) signal regions are also shown. Invariant µ + µ − K + K − -mass for events satisfying the offline selection for the B s → µ + µ − φ search (right). probability ( e − ct/cτ ), and the ‘pointing’ to the primary vertex (i. e., the open-ing angle ∆α between the p T ( B )-vector and the vector of the displacementbetween the p¯p vertex and the candidate decay-vertex). We extract the signal(background) template from simulation (mass-sidebands in data).The B s ) → µ + µ − branching fractions are obtained by normalizing to thenumber of B + → J/ψ ( → µ + µ − ) K + decays collected in the same sample.The ratio of trigger acceptances between signal and normalization mode ( ≃ ≃ ≃
1) are extracted from unbiaseddata. The expected average background is obtained by extrapolating eventsfrom the mass-sidebands to the search regions. This estimate was checkedby comparing predicted and observed background yields in control samplessuch as like-sign dimuon candidates, and opposite-sign dimuon candidateswith negative decay-length or with one muon failing the quality requirements.Contributions of punch-through hadrons from B s ) → h + h ′ − decays are alsoincluded in the estimate of total background. The LR cut was optimized bysearching for the best a priori expected 90% confidence level (CL) upperlimit on B ( B s ) → µ + µ − ). The observed event yields in two, 120 MeV/ c -widesearch windows (to be compared with 25 MeV/ c mass-resolution) centered atthe world average B s ) meson masses (Fig. 2, left), are in agreement with theexpected background events. A Bayesian approach that assumes a flat prioris used to estimate the following upper limits for the branching fractions: B ( B s → µ + µ − ) < . × − at 90(95)% CL and B ( B → µ + µ − ) < . . × − at 90(95)% CL. These results improve by a factor of twoprevious limits and significantly reduce the allowed parameter space for abroad range of SUSY models [8]. harmless b -hadron decays at CDF 5 An analogous search is performed in 0.92 fb − for FCNC B → µ + µ − h decays, where B = B + , B , or B s and h = K + , K ∗ ( → K + π − ) , or φ ( → K + K − ), respectively [9]. While B + and B channels are already exploredat the B -factories, the B s mode is still unobserved. The strategy is similarto the one used for the B s ) → µ + µ − search: the selection is optimized bymaximizing S/ √ S + B , where S ( B ) are simulated signal (real background)events. Dimuon candidates consistent with J/ψ and ψ ′ decays are removed,as those consistent with B → Dπ decays in which hadrons are misidentifiedas muons. The observed signal yields are obtained by counting the eventsin a 2 σ -wide window centered at the relevant B meson mass after subtract-ing the background extrapolated from events in the higher-mass sideband.The yields are normalized to the reference B → J/ψh modes. The mea-sured branching ratio for the B s mode, (1 . ± . ± . × − (differentfrom zero at 2 . σ ), allows extraction of the most stringent limit to date: B ( B s → µ + µ − φ ) / B ( B s → J/ψφ ) < . . × − at 90(95)% CL. Theresults for other modes, B ( B + → µ + µ − K + ) = (0 . ± . ± . × − (4 . σ ) and B ( B → µ + µ − K ∗ ) = (0 . ± . ± . × − (2 . σ ), areconsistent and competitive with B -factories results. CDF continues to pursue an highly successful program in flavor physics: weobtained the first observation of B s → K − π + , Λ b → p π − , and Λ b → p K − decays, a competitive measurement of the CP -violating asymmetry in B → K + π − decays, and the first measurement of the corresponding asymmetry in B s → K − π + decays. In addition, we quote the most stringent upper limitson branching fractions of rare FCNC B s ) → µ + µ − and B s → µ + µ − φ decays,that contribute to exclude a broad portion of parameter space in several SUSYmodels and increase the sensitivity to the presence of new physics before theoperation of the LHC. References
1. A. Abulencia et al (CDF Collaboration): Phys. Rev. Lett. , 211802 (2006)2. A. Abulencia et al (CDF Collaboration): Phys. Rev. Lett. , 221805 (2005)3. R. Blair et al (CDF Collaboration): FERMILAB-PUB-96/390-E CDF (1996)4. W. Ashmanskas et al, Nucl. Instrum. Methods, A518 , 532 (2004)5. D. Tonelli, FERMILAB-THESIS-2006-23, (2006)6. H. J. Lipkin, Phys. Lett. B , 126 (2005)7. CDF Public Note 8579, and M. J. Morello, PhD thesis, in preparation8. see, for instance, R. Dermisek et al, J. High Energy Phys. , 037 (2003)and , 029 (2005), R. Arnowitt et al, Phys. Lett. B , 121 (2002)9. CDF Public Note 8543,, 121 (2002)9. CDF Public Note 8543,