Capri2010.dvi 1 B physics at CDF – the Beauty of hadron collisions D. Tonellia∗ aFermilab, P.O. Box 500, Batavia, IL, 60510, USA The CDF experiment at the Tevatron pp̄ collider established that extensive and detailed exploration of the b–quark dynamics is possible in hadron collisions, with results competitive and supplementary to those from e + e − colliders. This provides an unique, rich, and highly rewarding program that is currently reaching full maturity. I report a few recent world-leading results on rare decays, CP-violation in B 0 s mixing, and b → s penguin decays. 1. Introduction Precise results from the successful B–factories experiments disfavor large contributions from non standard model (SM) physics in tree-dominated bottom meson decays. Agreement with the SM within theory uncertainties is also manifest in higher-order processes, such as K0–K 0 or B0– B 0 flavor mixing. The emerging picture confirms the Cabibbo-Kobayashi-Maskawa (CKM) ansatz as the leading pattern of flavor dynamics. Non- SM contributions, if any, are small corrections, or appear well beyond the TeV scale (and the LHC reach), or have an unnatural, highly fine- tuned flavor structure that escaped all experi- mental tests to date. The last chances to avoid such a disappointing impasse include the physics of bottom-strange mesons, still fairly unexplored, along with a few rare B0 decays, not fully probed at the B-factories because of limited event statis- tics. The CDF experiment at the Tevatron is cur- rently leading the exploration of this physics, ow- ing to CP-symmetric initial states in √ s = 1.96 TeV pp̄ collisions, large event samples collected by a well-understood detector, and mature anal- ysis techniques. CDF is a multipurpose mag- netic spectrometer surrounded by 4π calorime- ters and muon detectors. Most relevant for B physics are the tracking, particle-identification (PID) and muon detectors, and the trigger sys- ∗ for the CDF Collaboration tem. Six layers of double-readout silicon mi- crostrip detectors between 2.5 and 22 cm from the beam, and a single-readout layer at 1.5 cm ra- dius, provide precise vertex reconstruction, with approximately 15 (70) µm resolution in the az- imuthal (longitudinal) direction. A drift cham- ber provides 96 samplings of three-dimensional charged-particles trajectories between 40 and 140 cm radii in |η| < 1, for a transverse momentum resolution of σpT /p 2 T = 0.1%/(GeV/c). Specific ionization measurements in the chamber allow 1.5σ separation between charged kaons and pi- ons, approximately constant at momenta larger than 2 GeV/c. A comparable identification is achieved at lower momenta by an array of scin- tillator bars at 140 cm radius, which measure the time-of flight. Muons with pT > 1.5(2.2) GeV/c are detected by planar drift chambers at |η| < 0.6 (0.6 < |η| < 1.0). Low-pT dimuon triggers select J/ψ, rare B, and bottomonia decays. They collected approxi- mately 40 millions J/ψ decays in 5 fb−1 of data, used to reconstruct roughly 6,000 B0s → J/ψφ, 20,000 B0 → J/ψK∗(892)0, and 52,000 B+ → J/ψK+ decays. A trigger on charged parti- cles displaced from the primary vertex collects hadronic heavy-flavor decays. It relies on dedi- cated custom electronics to reconstruct tracks in the silicon with offline-like (48 µm) impact pa- rameter resolution, within 20 µs of the trigger latency. This yielded approximately 50 millions D0 → K−π+, 13,000 B0s → D−s (π+π−)π+, and 12,000 B0 → K+π− decays in 5 fb−1 of data. FERMILAB-CONF-10-453-PPD 2 CDF has currently collected 8 fb−1 of physics- quality data. The sample size will reach 10 fb−1in October 2011. Additional 6 fb−1 will be col- lected if the proposed three-year extension will be funded. In the following I report some recent, world- leading results, selected among those more sensi- tive to the presence of non-SM particles or cou- plings. Branching fractions indicate CP-averages, K∗0 is shorthand for the K∗(892)0 meson, and charge-conjugate decays are implied everywhere. 2. Polarization in B0s → φφ decays The B0s → φφ decay proceeds through a penguin-dominated b → ss̄s transition and was first detected by CDF a few years back. We recently improved the measurement of branch- ing fraction using 295 ± 20 events reconstructed in 2.9 fb−1 of data collected by the displaced- track trigger. We obtain B(B0s → φφ) = [2.40 ± 0.21 (stat.) ± 0.27 (syst.) ± 0.82(B)] × 10−5 [1]. The last uncertainty is due to the uncertainty on B(B0s → J/ψφ), used as a reference. Three polarization amplitudes enrich the phe- nomenology of the B0s → φφ decay. They cor- respond to the allowed values of orbital angu- lar momentum (ℓ = 0, 1, or 2) in the decay of a pseudo-scalar into two vector particles (B → V V ). A determination of these amplitudes may contribute useful insight into the puzzling pic- ture of B → V V polarizations, where experi- mental data in B0(+) → φK∗(+) decays disfavor first-order predictions of small transverse compo- nent (fT = O(m2V /m2B) ≈ 4%, where m are the masses), possibly suggesting contributions from non-SM amplitudes [2]. We report the first measurement of B0s → φφ polarization, using the sample employed for the branching fraction measurement [3]. We fit the mass of the four kaons and their angular distri- butions (in helicity basis). The angular accep- tance, extracted from simulation, is validated by measuring the polarization of B0s → J/ψφ decays to be consistent with results from an indepen- dent sample [4]. The small bias due to decay- length requirements in the trigger is included in the systematic uncertainties. In analogy with measurements of similar b → s penguin decays, the results are at odds with naive theory predic- tions. We measure |A0|2 = 0.348±0.041 (stat.)± 0.021 (syst.), |A|||2 = 0.287 ± 0.043 (stat.) ± 0.011 (syst.), |A⊥|2 = 0.365 ± 0.044 (stat.) ± 0.027 (syst.) for the polarization amplitudes, and cos(δ||) = −0.91+0.15−0.13 (stat.) ± 0.09 (syst.) for the phase difference between the A|| and A0 am- plitudes. Further extensions of this analysis to larger samples will provide information on the decay-width difference ∆Γ in B0s mesons. 3. Rare B → µ+µ− and B → hµ+µ− decays Decays mediated by flavor changing neutral currents, such as B0 (s) → µ+µ− or B → hµ+µ− are highly suppressed in the SM because they occur only through higher order loop diagrams. Their phenomenology provide enhanced sensitiv- ity to a broad class of non-SM contributions. The B0 (s) → µ+µ− rate is proportional to the CKM matrix element |Vtd|2(|Vts|2), and is further suppressed by helicity factors. The SM expecta- tions for these branching fractions are O(10−9), ten times smaller than the current experimental sensitivity. An observation of these decays at the Tevatron would unambiguously indicate physics beyond the SM. Or, even improved exclusion- limits strongly constrain the available space of parameters of several SUSY models. The latest CDF search for B0 (s) → µ+µ− decays uses 3.7 fb−1 of data collected by the dimuon trigger with pT(µ) > 2 GeV/c [5]. A loose pres- election based on opposite-charge dimuon trans- verse momentum and muon identification criteria (quality of matching with track, energy deposit in the calorimeter, specific ionization energy loss), rejects combinatoric background and charmless B decays. Further rejection is achieved by se- lecting on the decay-length significance (λ/σλ) against prompt background and on the isolation of the B0 (s) meson candidate, to exploit the harder fragmentation of b–mesons with respect to light- quark background. In addition, we require the candidate to point back to the primary vertex to further reduce combinatoric background and partially reconstructed b–hadron decays. The re- sulting event sample contains about 55,000 can- 3 didates, mostly coming from combinatoric back- ground. To further enhance purity we use an ar- tificial neural network classifier (NN) that com- bines information from the above observables into a single scalar discriminating quantity. Signal dis- tributions are modeled from detailed simulation; backgrounds from mass-sidebands in data. The signal rate is normalized to 20,000 B+ → J/ψ(→ µ+µ−)K+ reconstructed in the same sample. The ratio of trigger acceptances between signal and normalization modes (≃ 25%) is derived from simulation, the relative trigger efficiencies (≃ 1) are extracted from unbiased data and the rel- ative offline-selection efficiency (≃ 80%) is de- termined from simulation and data. The ex- pected average number of background events in the search region is obtained by extrapolating events from the mass-sidebands. The validity of this extrapolation is checked by comparing pre- dicted and observed background yields in sev- eral independent control samples including like- sign dimuons, opposite-sign dimuons with neg- ative decay-length, and opposite-sign dimuons with one muon failing the muon-quality require- ments. Small contributions of punch-through hadrons from B0 (s) → h+h′− decays are included in the estimate of total background. We optimize the limit by combining three independent ranges for the NN discriminator (with efficiencies rang- ing from 44 to 12%) to obtain the a priori best ex- pected 90% C.L. upper limit on B(B0 (s) → µ+µ−) (see fig. 1). The resulting 90 (95)% CL upper- limits are B(B0s → µ+µ−) < 3.6(4.3) × 10−8, (1) B(B0 → µ+µ−) < 6.0(7.6) × 10−9. (2) These results are the most stringent currently available and reduce significantly the allowed pa- rameter space for a broad range of SUSY models. An update of this analysis with approximately doubled sample size will further improve them soon. We also updated the analysis of B0 → K∗0(→ K+π−)µ+µ−, B+ → K+µ+µ−, and B0s → φ(→ K+K−)µ+µ− decays to 4.4 fb−1of data [6]. These are suppressed in the SM (B ≈ 10−6), with amplitudes dominated by penguin and box Figure 1. Distribution of µ+µ− mass in three independent ranges of the NN classifier. b → s transitions. Despite the presence of final- state hadrons, accurate predictions greatly sensi- tive to non-SM contributions are possible for rel- ative quantities based on angular-distributions of final state particles. The B+ and B0 decays have been previously studied, while the B0s channel has not been observed yet. We use a large sample of high-purity dimuon candidates combined in a common vertex with one or two charged particles, whose masses are assigned as appropriate for each final state. In the B0 decay, the mass hypothe- sis yielding the Kπ mass closer to the known K∗0 mass is chosen, resulting in the proper assignment about 92% of the times. A NN that uses informa- tion from kinematics and particle-identification greatly improves signal-to-background discrimi- nation. It is trained on simulated signals and mass-sideband data. The simulation is tuned to reproduce accurately the data using the corre- sponding resonant channels (B → J/ψh). The NN is optimized for maximum expected statisti- cal resolution on the branching ratio and asym- metry measurements. Prominent signals of 120 ± 16 B+ → K+µ+µ− and 101 ± 12 B0 → K∗0µ+µ− events (fig. 2, left) are observed. The absolute branching fractions, measured using the resonant decays as a refer- ence, are [0.38 ± 0.05 (stat.) ± 0.03 (syst.)] × 10−6 and [1.06 ± 0.14 (stat.) ± 0.09 (syst.)] × 10−6, re- spectively, consistent and competitive with pre- 4 )2) (GeV/c * KµµM( 5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 )2 E v e n ts / ( 2 0 M e V /c 0 10 20 30 40 50 60 -1CDF Run II Preliminary L=4.4fb -µ+µ*0 K→0B 12 (102 expected)±Yield:101 2 3 MeV/c±Mass:5284 Data Total Fit Signal Background )2) (GeV/cφµµM( 5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 )2 E v e n ts / ( 2 0 M e V /c 0 2 4 6 8 10 12 14 16 18 20 22 -1CDF Run II Preliminary L=4.4fb -µ+µφ →0sB 6 (31 expected)±Yield:27 2 5 MeV/c±Mass:5365 Data Total Fit Signal Background Figure 2. Distribution of K+π−µ+µ− mass (left) and K+K−µ+µ− (right). vious determinations [2]. In addition, 27 ± 6 B0s → φµ+µ− events are reconstructed, corre- sponding to the first observation of this decay, with a significance in excess of 6σ. The branch- ing ratio, (1.44±0.33 (stat.)±0.46 (syst.))×10−5, is consistent with theory predictions, and corre- sponds to the rarest B0s decay ever observed to date. We use the B+ and B0 signals for the first measurement in hadron collisions of branching ra- tios, muon forward-backward asymmetry (AFB ), and K∗0 longitudinal polarization, as a function of the dimuon mass. The asymmetry is greatly sensitive to non-SM particles and is determined from a fit to the cos(θµ) distributions, θµ being the helicity angle between the µ+ (µ−) and the opposite of the B (B̄) direction in the dimuon rest frame. Angular acceptances are determined from simulation. Figure 3 shows the asymme- try as a function of dimuon mass. Integrated in the 1–6 GeV2/c4 range, where theory predic- tions are most reliable, it equals AFB (1 < q 2 < 6) = 0.43+0.36−0.37 (stat.) ± 0.06 (syst.), consistent with Belle and Babar determinations [2]. CDF plans to achieve world-leading results by summer 2011 with the anticipated 2–3 factors increase in statistics, due to additional data, triggers, and reconstructed final states. )2/c2 (GeV2q 0 2 4 6 8 10 12 14 16 18 F B A -0.5 0 0.5 1 1.5 2 Data SM SM 7=-C7C )-µ+µ*0 K→ 0(BFBA -1CDF Run II Preliminary L=4.4fb Figure 3. Forward-backward asymmetry in B0 → K∗0µ+µ− as a function of dimuon mass. 4. Measurement of the B0s mixing phase Non-SM contributions have not yet been ex- cluded in B0s-B 0 s mixing. Their magnitude is constrained to be small by the precise determi- nation of the frequency [7]. However, knowledge of only the frequency leaves possible non-SM con- tributions to the (CP-violating) mixing phase un- constrained. The time evolution of flavor-tagged B0s → J/ψφ decays allow a determination of this phase largely free from theory uncertain- ties. These decays probe the phase-difference be- tween the mixing and the b̄ → c̄cs̄ quark-level transition, β J/ψφ s = β SM s + β NP s , which equals βSMs = arg(−VtsV ∗tb/VcsV ∗cb) ≈ 0.02 in the SM and is extremely sensitive to non-SM physics in the mixing. A non-SM contribution (βNPs ) would also enter φ J/ψφ s = φ SM s − 2βNPs , which is the phase difference between mixing and decay into final states common to B0s and B 0 s, and is also tiny in the SM: φSMs = arg(−M12/Γ12) ≈ 0.004. Because the SM values for β J/ψφ s and φ J/ψφ s cannot be resolved with the precision of cur- rent experiments, the following approximation is used: φ J/ψφ s ≈ −2βNPs ≈ −2β J/ψφ s , which holds in case of sizable non-SM contributions. Note that the phase φ J/ψφ s also modifies the decay- width difference between light and heavy states, ∆Γ = ΓL − ΓH = 2|Γ12| cos(φJ/ψφs ), which en- ters in the B0s → J/ψφ amplitude and equals 5 ∆ΓSM ≈ 2|Γ12| = 0.086 ± 0.025 ps−1 in the SM [8] . We updated the measurement of the time- evolution of flavor-tagged B0s → J/ψ(→ µ+µ−)φ(→ K+K−) decays to a sample of 5.2 fb−1 collected by the dimuon trigger [4]. Im- provements over the previous version of the anal- ysis include (a) a doubled event sample along with a newly optimized selection (b) a fully data-driven recalibration of the flavor-tagging al- gorithms (c) inclusion of possible non-φ scalar K+K− contributions as B0s → J/ψf0(980). ]2) [GeV/cφ ψMass(J/ 5.28 5.32 5.36 5.4 5.44 2 C a n d id a te s p e r 2 M e V /c 0 100 200 300 400 500 600 700 800 900 -1CDF Run II preliminary L = 5.2 fb -1Mixing Frequency in ps 10 20 30 A m p lit u d e -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Amplitude A -1Sensitivity: 37.0 ps -1CDF Run 2 Preliminary, L = 5.2 fb -1Mixing Frequency in ps 10 20 30 A m p lit u d e -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Figure 4. Distribution of J/ψK+K− mass after the optimized selection. A kinematic fit to a common space-point is ap- plied to the candidate J/ψ and two tracks con- sistent with being kaons originated from a φ me- son decay. A NN trained on simulated data (to identify signal) and B0s mass sidebands (for back- ground) exploits kinematic and PID information for an unbiased optimization of the selection: we minimize the average expected statistical uncer- tainty on the mixing phase extracted from en- sembles of statistical trials generated with dif- ferent input values. Discriminating observables used include kaon-likelihood, from the combina- tion of dE/dx and TOF information; transverse momenta of the B0s and φ mesons; the K +K− mass; and the quality of the vertex fit. The fi- nal sample contains approximately 6500 signal events over a comparable background (fig. 4, left). The angular distributions of final state particles provide a statistical determination of the CP- composition of the signal thus improving sensi- tivity to the phase. Possible contributions from scalar J/ψK+K− or J/ψf0(980) final states are included in the angular distributions. The pro- duction flavor is inferred using two classes of al- gorithms. Opposite-side tags (OST) exploit fla- vor conservation in strong interaction. Being b– quarks predominantly produced in pairs with b̄– quarks, the production flavor is inferred from the charge of decay products (e, µ, or jet) of the b– hadron emitted in the opposite hemisphere of the signal B0s. Same-side tags (SST) exploit flavor conservation in the fragmentation process that leads to the signal B0s meson. Because a ss̄ pair will be required to form a meson from the bottom quark, the charge of a kaon kinematically close to the bottom-strange meson is correlated to its fla- vor. The performance of flavor taggers is modeled as function of many event properties using simu- lations; however overall scale factors between the simulation and reality are allowed and extracted from data. The uncertainty in their determina- tion contributes to the final systematic uncer- tainty of the measurement. The tagging power, ǫD2, is the product of an efficiency ǫ, the fraction of candidates with a flavor tag, and the square of the dilution D = 1 − 2w, where w is the mis- tag probability. We calibrate independently b and b̄ OST tags using 52,000 B+ → J/ψK+ decays. Predicted and measured dilutions agree with scale factors close to unity. The observed tagging effi- ciency is (94.2 ± 0.4)% and the average predicted dilution on signal is √ 〈D2〉 = 0.110 ± 0.002, with 1.03 ± 0.06 scale factor. The SST calibration is obtained by repeating the full mixing analysis on 13,000 B0s → D−s π+(π+π−) [7]. The result- ing frequency, ∆ms = 17.79 ± 0.07 (stat.) ps−1, is fully consistent with published values (fig. 4, right). The scale factor, A = 0.94 ± 0.15 (stat.) ± 0.13 (syst.) is given by the size of the amplitude at the mixing frequency. This corresponds to a tagging efficiency of (52.2 ± 0.7)% and an aver- age dilution on signal of √ 〈D2〉 = 0.275 ± 0.003. Multiple tags, if any, are combined as indepen- dent for a total tagging power ǫD2 ≈ 4.5%. The proper time of the decay and its resolution are known on a per-candidate basis with an average resolution of approximately 90 fs−1. Information on B0s candidate mass and its uncertainty, angles 6 between final state particles’ trajectories (to ex- tract the CP-composition), production flavor, and decay length and its resolution are used as observ- ables in a multivariate unbinned likelihood fit of the time evolution that accounts for direct de- cay amplitude, mixing followed by the decay, and their interference. Direct CP-violation is expected small and neglected. The fit determines the phase β J/ψφ s , the decay-width difference ∆Γ, and many other “nuisance” parameters including the mean B0s lifetime (2/(ΓL + ΓH)), the magnitudes of lin- ear polarization amplitudes, the CP-conserving phases (δ‖ = arg(A‖A ∗ 0), δ⊥ = arg(A⊥A ∗ 0)), and others. The acceptance of the detector is calcu- lated from simulation and found to be consistent with angular distributions of random combina- tions of four tracks in data; the fit model was validated by measuring lifetime and polarization amplitudes in B0 → J/ψK∗0 decays to be consis- tent with B–factories measurements [9]. For enhanced precision, all parameters but the mixing phase are determined in a fit with phase fixed to zero: cτ(B0s) = 458.6 ± 7.6 (stat.) ± 3.6 (syst.) µm, ∆Γ = 0.075 ± 0.035 (stat.) ± 0.010 (syst.) ps−1, |A|||2 = 0.231±0.014 (stat.) ± 0.015 (syst.), |A0|2 = 0.524 ± 0.013 (stat.) ± 0.015 (syst.),and δ⊥ = 2.95 ± 0.64 (stat.) ± 0.07 (syst.) rad. These results represent the cur- rent best measurements of these quantities from a single experiment. With phase floating, fits on simulated samples show biased, non-Gaussian distributions of esti- mates and multiple maxima, because of known likelihood symmetries. We use a frequentist con- fidence region in the (β J/ψφ s , ∆Γ) plane using a profile-likelihood ratio ordering [10], which is close to optimal for limiting the impact of sys- tematic uncertainties. These are included by ran- domly sampling a limited number of points in the space of all nuisance parameters. A specific value (β J/ψφ s , ∆Γ) is excluded only if it can be excluded for any assumed value of the nuisance parameters within 5σ of their estimate on data. The resulting allowed region is greatly reduced with respect to the previous measurement and is fairly consistent (0.8σ) with the SM: the range [0.02, 0.52]∪[1.08, 1.55] contains the βJ/ψφs phase (rad) sβ -1 0 1 ) -1 ( p s Γ ∆ -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 -1 CDF Run II Preliminary L = 5.2 fb 95% CL 68% CL SM prediction Figure 5. Confidence region in the (β J/ψφ s , ∆Γ) plane. at the 68% CL and the range [-π/2, -1.44] ∪[-0.13, 0.68]∪[0.89, π/2] at the 95% CL. With the full Run II data sample, we expect to observe a non-SM phase or exclude it for any value of β J/ψφ s larger than approximately 0.4 rad. REFERENCES 1. CDF Collaboration, Note 10064, and D. Acosta et al. (CDF Collaboration), Phys. Rev. Lett. 95, 031801 (2005). 2. Heavy Flavor Averaging Group, arXiv:[1010.1589(hep-ex)]. 3. CDF Collaboration, Note 10120. 4. CDF Collaboration, Note 10206, and T. Aal- tonen et al. (CDF Collaboration), Phys. Rev. Lett. 100, 161802 (2008). 5. CDF Collaboration, Note 9892, and T. Aal- tonen et al. (CDF Collaboration), Phys. Rev. Lett. 100, 101802 (2008). 6. CDF Collaboration, Note 10047, and T. Aal- tonen et al. (CDF Collaboration), Phys. Rev. D79, 011104(R) (2009). 7. CDF Collaboration, Note 10108, and A. Abu- lencia et al. (CDF Collaboration), Phys. Rev. Lett. 97, 242003 (2006). 8. U. Nierste, talk at CKM Workshop 2010. 9. CDF Collaboration, Note 8950. 10. D. Tonelli (for the CDF Collaboration), arXiv:[0810.3229(hep-ex)].