A Measurement of CP Asymmetry in b->sgamma using a Sum of Exclusive Final States

We perform a measurement of the CP asymmetry in b->sgamma decays using a sample of 383 BB events collected by the BABAR detector at the PEP-II asymmetric B factory. We reconstruct sixteen flavor-specific B decay modes containing a high-energy photon and a hadronic system (Xs) containing an s quark. We measure the \CP asymmetry to be -0.011+/- 0.030(stat)+/-0.014(syst) for a photon energy threshold at 1.6 GeV and the hadronic system mass between 0.6 and 2.8 GeV/c2.

The decay b → sγ is a flavor-changing neutral current process described by a radiative penguin diagram in the Standard Model (SM). It is sensitive to new physics which can appear in branching fraction or CP asymmetry measurements. Measurements of the branching fraction [1,2] are in good agreement with the SM [3] predictions.
A CP asymmetry between b → sγ andb →sγ decays is predicted by the SM to be ≤ 1% [4] but could be enhanced up to 15% [5,6,7] in models of physics beyond the SM. Existing measurements are consistent with zero CP asymmetry with a precision of 5% [8,9]. The increased precision obtained in this work allows us to better discriminate between various theoretical models [10].
We use a sample of 383 × 10 6 BB pairs collected at the Υ (4S) resonance by the BABAR detector [11] at the PEP-II e + e − B factory. In addition, we use 36.3 fb −1 collected 40 MeV below the Υ (4S) resonance to study backgrounds from non-B decays.
The hadronic system X s , formed from the kaons and pions, is required to have an invariant mass M Xs between 0.6 and 2.8 GeV/c 2 , corresponding to a photon energy threshold E γ > 1.9 GeV in the B meson rest frame.
Charged kaons are identified by combining information from the Cherenkov detector and the energy-loss measurements from the tracking system. The remaining tracks are assumed to be charged pions. The K 0 S candidates are reconstructed by combining two oppositely charged pions with an invariant mass within 9 MeV/c 2 of the nominal K 0 S mass [12] and a minimum flight distance of 2 mm from the primary event vertex. Both charged ‡ Now at Tel Aviv University, Tel Aviv, 69978, Israel § Also with Università di Perugia, Dipartimento di Fisica, Perugia, Italy ¶ Also with Università di Sassari, Sassari, Italy and neutral kaons are required to have laboratory momenta ≥ 0.8 GeV/c. Neutral pions and η candidates are reconstructed from pairs of photons with energies above 50 MeV in the laboratory frame and a lateral moment [13] less than 0.8. The lateral moment measures the spread of a shower in the calorimeter and provides good separation between electromagnetic and hadronic showers. The invariant mass of the pair of photons is required to be between 115 and 150 MeV/c 2 for π 0 candidates and between 470 and 620 MeV/c 2 for η candidates Monte Carlo (MC) samples based on EvtGen [14] and GEANT4 [15] are used to simulate the signal and background processes and the detector response. The b → sγ signal sample is generated with a photon spectrum derived from Ref. [4] assuming m b = 4.65 GeV/c 2 . The fragmentation of the X s system is modeled using JETSET [16] corrected to fit the BABAR data as described later.
The background to the B reconstruction is dominated by continuum processes (e + e − → qq, with q = u, d, s, c) that produce a high-energy photon either by initial-state radiation or from the decay of π 0 and η mesons. Continuum events tend to be less isotropic than B-decay events since they result from hadronic fragmentation of highmomentum quarks back-to-back in the CM frame. Highenergy photons in these events tend to be collinear with the thrust axis formed from the rest of the event (ROE), defined as those particles not used in reconstructing the signal B candidate. We reject such backgrounds by requiring that the cosine of the angle between the photon and the thrust axis of the ROE (in the CM frame) be less than 0.85. We further reject the continuum events by requiring the ratio of the second (L 2 ) and zeroth (L 0 ) Legendre moments for the ROE particles with respect to the B flight direction to be smaller than 0.46.
Continuum events with high-energy photons from π 0 and η decays are major backgrounds. To veto these events, we associate each high-energy photon candidate γ with another photon candidate γ ′ in the event. For multiple γ ′ candidate in an event, we choose the γγ ′ pairs whose invariant mass, determined from adding the four vectors, is closest to the nominal π 0 mass (or η mass in case of η veto). Events are rejected if the photon pairs are consistent with π 0 or η decays based on the output of a boosted decision tree (BDT) [17] constructed from the energy of the less energetic photon γ ′ and m γγ ′ .
We reject the remaining continuum events by constructing an additional BDT that combines information from a number of variables related to the event shape, the kinematic properties of the B meson, and the flavortagging [18] properties of the other B meson in the event. Examples of these variables are the Fox-Wolfram moments [19], and the cosine of the B flight direction computed in the CM frame with respect to the beam axis. Optimization of the selection criteria of the π 0 veto, η veto, and event selection BDTs is performed using an iterative method which maximizes the statistical signal significance. After the final event selection, we reject 97% of the continuum background while retaining 55% of the signal events.
Fully reconstructed b → sγ decays are characterized by two kinematic variables: the beam-energy substituted mass m ES = s/4 − p * B 2 , and the energy difference between the B candidate and the beam energy and p * B are the energy and momentum of the B candidate in the e + e − CM frame, and √ s is the total CM frame energy. Signal events are expected to have a ∆E distribution centered near zero and a m ES distribution centered at the mass of the B meson. For events with multiple B candidates, we select the one with the smallest |∆E|. We where the parameters are determined by an unbinned fit to the signal MC. In the above function, µ 0 is the peak position of the distribution, σ L,R are the widths on the left and right of the peak, and α L,R parameterize the tail on the left and right of the peak, respectively.
The background surviving the final selection can be attributed to one of three sources: continuum events, BB events other than b → sγ decays (referred to as generic BB), and "cross-feed events", defined as events containing a b → sγ decay, but in which the true decay was not correctly reconstructed. The shape of the crossfeed and BB background is described by a binned PDF, determined from MC with 1 MeV/c 2 binning.
The continuum background is described by an ARGUS function [20] determined from a fit to the off-resonance data. In this fit, the m ES distribution is shifted to have the same end-point as that of the on-resonance data.
In the maximum-likelihood fit, all parameters are fixed with the exception of the normalizations of the various components as well as µ 0 , which is determined from fitting the data, since the peak position is not well modeled in the MC simulation. The signal, BB and cross-feed shapes are constrained by the MC, while the continuum background shape is fixed to that of off-resonance data. The shapes of the distributions are assumed to be the same for B and B candidates, with the exception of the BB and cross-feed background, which are allowed to vary between b and b in order to eliminate the possibility of a false CP asymmetry. In Figure 1 we present the final fits to the m ES distributions for b → sγ andb →sγ events for the four M Xs sub-regions. As expected, the signal to background ratio decreases from lower to higher M Xs regions. In Figure 2, we present the final fits to the m ES distribution for b → sγ andb →sγ events for the entire M Xs region. The direct CP asymmetry is calculated as where N b and N b are the yields of the b → sγ and b →sγ signals respectively. A det , described in details below, is the flavor bias caused by the detector responses to positively and negatively charged particles. Table I presents the fitted values for ( , the flavor-bias of the detector: A det , the systematic error arising from the BB and cross-feed modeling and the systematic error arising from the continuum background modeling. The last column shows the final results for the CP asymmetries. fraction between b and b decays, and D = 1 − (ω + ω) is the dilution factor from the average wrong-flavor fraction. The small wrong-flavor fractionω (ω), defined to be the fraction of b (b) reconstructed as the opposite flavor, is due to charged pions misidentified as charged kaons. Using the particle misidentification rate measured in control samples in data we calculate ∆D = (5 ± 4) × 10 −5 and 1 − D = (5.4 ± 0.1) × 10 −3 . The flavor bias of the detector A det is due to asymmetric K + , K − interaction cross-sections in the detector at low momenta. Such an asymmetry could produce a false CP asymmetry in the signal events. We perform a measurement of A det in data using two independent methods. The first approach determines this asymmetry from events in the m ES sideband, 5.22 < m ES < 5.27 GeV/c 2 , which is dominated by continuum background with no expected CP asymmetry. The second approach uses a control sample where we replace the high-energy photon from the B decay with a high-energy π 0 with p CM ≥ 1.6 GeV/c. The same selection criteria used in the signal selection are applied, except for π 0 and η veto requirements, and the CP asymmetry in the m ES sideband is measured. In both control samples we apply appropriate weights to the events to ensure that the fraction of each reconstructed final state is identical to that in the signal sample. We find the CP asymmetry measured using both of these approaches to be nearly identical, and average the two measurements to obtain A det = −0.007 ± 0.005. The mean value is used to shift the ( mean value, while the error contributes to the systematics. The values of A det computed in each X s mass region are reported in Table I.
The shape of the BB and cross-feed background, determined from MC, is also a potential source of flavor bias in the fit to the data. This background peaks broadly in the signal region, and a small shape difference as a function of flavor could create a false CP asymmetry in the signal. We measure the size of this effect by correcting the BB and cross-feed shapes separately. The high-energy π 0 control sample is used to study the uncertainty of the BB background shape. We use the differences found between the data and MC m ES shapes in this control sample to correct the nominal BB background shape built from the MC. The biggest uncertainty in the cross-feed shape is due to the fact that JETSET does not reproduce the observed fragmentation structure of data. We thus correct the simulation shape using the fragmentation previously determined from BABAR data [21]. We then construct new b and b binned PDFs using these corrected crossfeed and BB events and fit the data a second time with them. The difference between the nominal A CP and A CP from this fit, shown in Table I, is used as the systematic error from shape modeling of the B background.
The systematic error arising from the continuum background modeling is determined by varying the ARGUS shape parameters within the experimental errors, and is found to be 0.006 for the combined M Xs region. Systematic errors due to possible differences in the signal shape between b and b events, CP content of the peaking background, and possible contaminations from b → dγ decays are all found to be negligible. Contributions from D , ∆D and signal modeling are neglected due to their small impact on A CP . The dominant systematic errors are therefore due to the uncertainties in the flavor bias of the detector and the background shapes as described above.
The total systematic errors are calculated as the sum in quadrature of errors on A det , systematic errors arising from the continuum, BB and cross-feed shape modeling. The results are shown in Table I.
In summary, we measure the direct CP asymmetry in b → sγ to be A CP = −0.011 ± 0.030 ± 0.014 in the region 0.6 < M Xs < 2.8 GeV/c 2 . This result represents the most accurate measurement of this quantity to date. The mea-surement is consistent with zero CP asymmetry and with the SM prediction. The CP asymmetry in each M Xs region considered in our study is also consistent with zero.
We are grateful for the excellent luminosity and machine conditions provided by our PEP-II colleagues, and for the substantial dedicated effort from the computing organizations that support BABAR. The collaborating institutions wish to thank SLAC for its support and kind hospitality. This work is supported by DOE and NSF (USA), NSERC (Canada), CEA and CNRS-IN2P3 (France), BMBF and DFG (Germany), INFN (Italy), FOM (The Netherlands), NFR (Norway), MES (Russia), MEC (Spain), and STFC (United Kingdom). Individuals have received support from the Marie Curie EIF (European Union) and the A. P. Sloan Foundation.