Measurement of $CP$ asymmetries in $B^0\to \phi K^0_S$ decays with Belle II

We present a measurement of time-dependent rate asymmetries in $B^0\to \phi K^0_S$ decays to search for non-standard-model physics in $b\to q \overline{q}s$ transitions. The data sample is collected with the Belle II detector at the SuperKEKB asymmetric-energy $e^{+}e^{-}$ collider in 2019-2022 and contains $(387\pm 6)\times 10^6$ bottom-antibottom mesons from $\Upsilon(4S)$ resonance decays. We reconstruct $162\pm17$ signal events and extract the charge-parity ($CP$) violating parameters from a fit to the distribution of the proper-decay-time difference of the two $B$ mesons. The measured direct and mixing-induced $CP$ asymmetries are $A=0.31\pm0.20\pm0.05$ and $S=0.54\pm0.26^{+0.06}_{-0.08}$, respectively, where the first uncertainties are statistical and the second are systematic. The results are compatible with the $CP$ asymmetries observed in $b\to c\overline{c} s$ transitions.


I. INTRODUCTION
Measurements of CP asymmetries in loop-suppressed B meson decays are sensitive probes of physics beyond the standard model (SM).In particular, gluonicpenguin b → qqs modes, such as B 0 → ϕK 0 S , are sensitive to interfering non-SM amplitudes that carry additional weak-interaction phases.The SM reference is the mixing-induced CP asymmetry S ≡ sin 2ϕ 1 observed in tree-level b → ccs transitions, where ϕ 1 (or β) equals arg(−V cd V * cb /V td V * tb ) and V ij are Cabibbo-Kobayashi-Maskawa (CKM) quark-mixing matrix elements [1,2].The deviation from the value of S observed in b → ccs transitions, S = 0.699 ± 0.017 [3], is the key observable.For B 0 → ϕK 0 S decays, such a deviation is at most 0.02 ± 0.01 within the SM while the direct CP asymmetry C is expected to be zero [4].The current worldaverage values for B 0 → ϕK 0 S are S = 0.74 +0.11 −0.13 and C = 0.01 ± 0.14 [3].Therefore, experimental knowledge must be improved.We present a measurement of S and C in the sample of electron-positron collisions collected by the Belle II experiment in 2019-2022 [5].
At B-factories, BB events are produced from the decay of an Υ (4S) resonance, where B indicates a B + or B 0 meson.We denote pairs of neutral B mesons as B CP B tag , where B CP decays into a CP -eigenstate at time t CP , and B tag decays into a flavor-specific final state at time t tag .For quantum-correlated B-meson pairs, the flavor of B CP is opposite to that of B tag at the instant when the B tag decays.The probability to observe a B tag meson with flavor q (q = +1 for B 0 and q = −1 for B 0 ) and a proper-time difference ∆t ≡ t CP − t tag between the B CP and B tag decays is P(∆t, q) = e −|∆t|/τ B 0 4τ B 0 1 + q S sin(∆m d ∆t) where τ B 0 and ∆m d are the B 0 lifetime and B 0 − B 0 mixing frequency, respectively [6].
We reconstruct B 0 → ϕK 0 S decays in a sample of energy-asymmetric e + e − collisions at the Υ (4S) resonance provided by SuperKEKB and collected with the Belle II detector.The sample corresponds to (362 ± 2) fb −1 and contains (387 ± 6) × 10 6 BB events.We fully reconstruct B CP in the ϕK 0 S final state using the intermediate decays ϕ → K + K − and K 0 S → π + π − , while we only determine the position of the B tag decay.The flavor of the B tag meson is inferred from the properties of all charged particles in the event not belonging to B CP [7].In order to extract the CP asymmetries, we model the distributions of signal B CP and backgrounds in ∆t and other discriminating variables, and then perform a likelihood fit.The last measurements, by the Belle and BABAR experiments, used time-dependent Dalitz-plot analyses [8,9].This method models the interferences among the intermediate resonant and nonresonant amplitudes contributing to B 0 → K + K − K 0 S decays, thereby providing the best sensitivity on ϕ 1 .Due to the small dataset size, which may induce multiple solutions in the Dalitz-plot approach, we perform a quasi-two-body analysis by restricting the sample to candidates reconstructed in a narrow region around the ϕ mass.This strategy offers the advantage of a simpler analysis, albeit with a reduced statistical sensitivity.We use the knowledge from the previous Dalitz-plot analyses to estimate the effect of neglecting the interferences.We test our analysis on the CP -conserving B + → ϕK + decay, which has similar backgrounds and vertex resolution.Charge-conjugated modes are included throughout the paper.

II. EXPERIMENTAL SETUP
The Belle II detector [10] operates at the SuperKEKB accelerator at KEK, which collides 7 GeV electrons with 4 GeV positrons.The detector is designed to reconstruct the decays of heavy-flavor mesons and τ leptons.It consists of several subsystems arranged cylindrically around the interaction point (IP).The innermost part of the detector is equipped with a two-layer silicon-pixel detector (PXD), surrounded by a four-layer double-sided siliconstrip detector (SVD) [11].Together, they provide infor-mation about charged-particle trajectories (tracks) and decay-vertex positions.Of the outer PXD layer, only one-sixth is installed for the data used in this work.The momenta and electric charges of charged particles are determined with a 56-layer central drift-chamber (CDC).Charged-hadron identification (PID) is provided by a time-of-propagation counter and an aerogel ring-imaging Cherenkov counter, located in the central and forward regions outside the CDC, respectively.The CDC provides additional PID information through the measurement of specific ionization.Photons are identified and electrons are reconstructed by an electromagnetic calorimeter made of CsI(Tl) crystals, covering the region outside of the PID detectors.The tracking and PID subsystems, and the calorimeter, are surrounded by a superconducting solenoid, providing an axial magnetic field of 1.5 T. The central axis of the solenoid defines the z axis of the laboratory frame, pointing approximately in the direction of the electron beam.Outside of the magnet lies the muon and K 0 L identification system, which consists of iron plates interspersed with resistive-plate chambers and plastic scintillators.
We use simulated events to model signal and background distributions, study the detector response, and test the analysis.Quark-antiquark pairs from e + e − collisions, and hadron decays, are simulated using KKMC [12] with Pythia8 [13], and EvtGen [14], respectively.The detector response and K 0 S decays are simulated using Geant4 [15].Collision data and simulated samples are processed using the Belle II analysis software [16,17].

III. EVENT RECONSTRUCTION
Events containing a BB pair are selected online by a trigger based on the track multiplicity and total energy deposited in the calorimeter.We reconstruct B 0 → ϕK 0 S decays using ϕ → K + K − and K 0 S → π + π − decays, in which the four tracks are reconstructed using information from the PXD, SVD, and CDC [18].All tracks are required to have polar angle θ within the CDC acceptance (17 • < θ < 150 • ).Tracks used to form ϕ candidates are required to have a distance of closest approach to the IP less than 2.0 cm along the z axis and less than 0.5 cm in the transverse plane to reduce contamination of tracks not generated in the collision.
Kaon and pion mass hypotheses are assigned to tracks based on information provided by the PID subsystems.The ϕ candidates are formed by combining K + K − pairs consistent with originating from the IP and having invariant mass within [0.99, 1.09] GeV/c 2 , where the average ϕ mass resolution is approximately 3 MeV/c 2 .The K 0 S candidates are formed by combining two oppositely charged particles, assumed to be pions, and requiring their invariant mass to be within [0.480, 0.515] GeV/c 2 , where the average K 0 S mass resolution is approximately 2 MeV/c 2 .In order to suppress combinatorial background from mis-reconstructed K 0 S , we require K 0 S candidates to have a displacement of at least 0.05 cm from the ϕ decay vertex, where the average K 0 S flight distance is 10 cm.The beam-energy constrained mass M bc and energy difference ∆E are computed for each B 0 → ϕK 0 S candidate as , where E * beam is the beam energy, and E * B and p * B are the energy and momentum of the B CP candidate, respectively, all calculated in the center-of-mass (c.m.) frame.Signal B CP candidates peak at the known B 0 mass [6] and zero in M bc and ∆E, respectively, while continuum is distributed more uniformly.Only candidates satisfying M bc > 5.2 GeV/c 2 and |∆E| < 0.2 GeV are retained for further analysis.
The B 0 → ϕK 0 S decay vertex is determined using the TreeFitter algorithm [19,20].In addition, the B CP candidate is constrained to point back to the IP.The B tag decay vertex is reconstructed using the remaining tracks in the event.Each track is required to have at least one measurement point in the SVD and CDC subdetectors and correspond to a total momentum greater than 50 MeV/c.The B tag decay-vertex position is fitted using the Rave algorithm [21], which allows for weighting the contributions from tracks that are displaced from the B tag decay vertex, and thereby suppressing biases from secondary charm decays.The decay-vertex position is determined by constraining the B tag direction, as determined from its decay vertex and the IP, to be collinear with its momentum vector [22].
We estimate the proper-time difference using the longitudinal decay-vertex positions, ℓ CP and ℓ tag , of the B CP and B tag mesons, respectively, as where βγ = 0.28 is the Υ (4S) Lorentz boost and γ * = 1.002 is the Lorentz factor of the B mesons in the c.m. frame.The average distance between the B CP and B tag vertices is approximately 100 µm along the z axis.The Bdecay vertex resolution along the z axis is approximately 35 µm for simulated B 0 → ϕK 0 S decays.We apply loose χ 2 probability requirements to both the B CP and B tag vertices.Events having a ∆t uncertainty σ ∆t greater than 2.0 ps, where the average value is approximately 0.5 ps, are not included in the analysis, as they constitute less than 2% of the signal events and do not contribute to the determination of S.
The dominant sources of background come from continuum e + e − → qq events, where q indicates a u, d, c, or s quark.A boosted-decision-tree (BDT) classifier is trained on simulated samples to combine several topological variables that provide separation between continuum and signal events [23].The variables included in the BDT are the following, in order of decreasing discriminating power: the cosine of the angle between the thrust axes of B CP and B tag [24], the modified Fox-Wolfram moments introduced in Ref. [25], the thrust of B tag [26,27], the ratio of the zeroth to the first Fox-Wolfram moment [28], and the harmonic moments calculated with respect to the thrust axis.We impose a minimum requirement on the output of the BDT, O CS , that retains more than 95% of the signal, while rejecting more than 55% of the continuum events.The transformed output of the classifier, defined as where O min CS and O max CS are the minimum and maximum values of the selected events, is included in the fit.The signal and remaining background events are approximately Gaussian-distributed in this variable and are therefore simple to model.
An additional requirement |∆E| < 50 MeV further suppresses continuum and misreconstructed B → ϕK * decays.To reduce the contamination from nonresonant B 0 → K + K − K 0 S decays and other modes leading to the same final state, events are required to satisfy |m(K where m ϕ is the known ϕ meson mass [6].
The same event reconstruction is applied on B + → ϕK + decays, except for the K 0 S selection, which is replaced by a K + track with a stringent PID requirement.This is more than 90% efficient on the signal, while rejecting around 30% of misidentified charged particles.We achieve a total signal reconstruction efficiency of 33% for B 0 → ϕK 0 S and 40% for B + → ϕK + .Events with multiple candidates account for approximately 6% of the data.We keep the candidate with the highest B CP vertex χ 2 probability.The criterion retains the correct signal candidate 67% of the times using simulated events.We check that the candidate selection does not bias the ∆t distribution by comparing the results of lifetime fits to the B 0 and B + samples with known values [6].

IV. TIME-DEPENDENT CP -ASYMMETRY FIT
The distributions of signal and backgrounds are described in a likelihood fit to extract the CP asymmetries.We consider the following contributions to the sample composition: signal B 0 → ϕK 0 S events, nonresonant B 0 → K + K − K 0 S background, and continuum background.Additional BB background events are treated as a source of systematic uncertainty, as they are estimated to be at most 2% of the signal yield, according to simulation.Low-multiplicity events contribute at less than the level of the BB backgrounds in the simulation, and are distributed like continuum in the variables used in the fit, so they are treated as part of the continuum background.We model the distributions of signal and background events in the M bc , O ′ CS , cos θ H , and ∆t variables.The M bc and O ′ CS variables provide discrimination between signal and continuum background.The helicity angle θ H , defined as the angle between the momentum of the B 0 and that of the positively charged kaon in the ϕ rest frame, is used to distinguish between signal and nonresonant components.The ∆t variable and tag-flavor q provide access to the time-dependent CP asymmetries.In addition, we use σ ∆t as a conditional observable to model the per-event resolution.
We extract the CP asymmetries using an extended maximum-likelihood fit to the unbinned distributions of the discriminating variables.The total probability density function (PDF) is given by the product of the four one-dimensional PDFs, since the dependences among the fit observables are negligible.We model the M bc distribution using an ARGUS function [29] for continuum and a Gaussian function with shared parameters for the B 0 → ϕK 0 S and B 0 → K + K − K 0 S components.The continuum shape is fixed from a fit to the |∆E| > 0.1 GeV sideband, while the signal-shape parameters are determined by the fit.We check that the continuum shapes are not biased by B 0 → ϕK * 0 , B + → ϕK * + , and other B 0 and B + decay modes, contributing in total to less than 1% of the events in the ∆E sideband.The O ′ CS distribution is modeled using the sum of two Gaussian functions with a common mean and constrained proportions for continuum, and a Gaussian function with asymmetric widths and shared parameters for the B 0 → ϕK 0 S and B 0 → K + K − K 0 S components.The O ′ CS shapeparameters are determined from events in the ∆E sideband for continuum, and using simulated events for signal.The cos θ H distribution of continuum is modeled with a second-order polynomial determined from ∆E sideband events.We verify using simulated samples that the B 0 → ϕK 0 S and B 0 → K + K − K 0 S components follow a cos 2 θ H and a uniform distribution, respectively, as expected from angular momentum conservation, and the detector acceptance does not affect their shapes.
The B tag flavor is identified using a category-based Bflavor tagging algorithm from the particles in the event that are not associated with the B CP candidate [7].The tagging algorithm provides for each B tag candidate a flavor (q) and the tag-quality r = 1 − 2w.The latter is a function of the wrong-tag probability w and ranges from r = 0 for no discrimination power to r = 1 for unambiguous flavor assignment.Taking into account the effect of imperfect flavor assignment, Eq. (1) becomes  where ∆w is the wrong-tag probability difference between events tagged as B 0 and B 0 , and a tag ϵ is the taggingefficiency-asymmetry between B 0 and B 0 .
The effect of finite ∆t resolution is taken into account by modifying Eq. (3) as follows: where R is the resolution function, conditional on the per-event ∆t uncertainty σ ∆t .Its parametrization, as determined in B 0 → D ( * )− π + decays [30], consists of the sum of three components, where δt is the difference between the observed and the true ∆t.The first component is described by a Gaussian function with mean m G and width s G scaled by σ ∆t , which accounts for the core of the distribution.The second component R t is the sum of a Gaussian function and the convolution of a Gaussian with two oppositely sided exponential functions, where exp > (kx) = exp(kx) if x > 0 or zero otherwise, and similarly for exp < (kx).The exponential tails arise from intermediate displaced charm-hadron vertices from the B tag decay.The fraction f t is zero at low values of σ ∆t and steeply reaches a plateau of 0.2 at σ ∆t = 0.25 ps.The third component, which accounts for outlier events contributing with a fraction of less than 1%, is modeled with a Gaussian function having a large width σ 0 of 200 ps.The effect on the resolution function of the small momentum of the B 0 in the Υ (4S) frame is taken into account as a systematic uncertainty.We divide our sample into seven intervals (bins) of the tag-quality variable r, with boundaries (0.0, 0.1, 0.25, 0.45, 0.6, 0.725, 0.875, 1.0), to gain statistical sensitivity from events with different wrong-tag fractions.The response of the tagging algorithm and detector ∆t resolution is calibrated from a simultaneous fit of w, ∆w, a tag ϵ , and resolution-function parameters in the seven r-bins, using flavor-specific B 0 → D ( * )− π + decays [31].The effective flavor tagging efficiency, defined , where ε i is the fraction of events associated with a tag decision and w i is the wrong-tag probability in the ith r bin, is (31.69 ± 0.35)%, where the uncertainty is statistical.We verify in simulation the compatibility of the flavor tagging and resolution function between the calibration and signal decay modes.We use the flavor-tagging parameters obtained from B + → D 0 π + decays to calibrate the flavor tagger and resolution function in the B + → ϕK + control channel.
The ∆t distribution of the continuum background is modeled using events from the ∆E sideband and allowing for an asymmetry in the yields of oppositely tagged events.A double Gaussian parametrization, with means and widths scaled by σ ∆t , describes the data accurately.The ∆t distribution of the B 0 → K + K − K 0 S background is parametrized using the same detector response as for signal.Its CP asymmetries are fixed to the known values [3].
The nominal fits to the control and signal samples determine the continuum yields and the sum of the resonant and nonresonant yields in the seven r-bins.We also determine the fraction of the resonant yields with respect to the sum of the resonant and nonresonant yields directly in the data.In addition, the mean and width of the Gaussian function describing the resonant and nonresonant components in M bc and the asymmetry in the normalization of oppositely tagged continuum-background events are determined by the fit.Finally, the fit determines the CP asymmetries, for a total of 20 free parameters.The fit results are reported in Table I.In the control sample, we find 581 ± 33 signal B + → ϕK + , 70 ± 23 nonresonant, and 5730 ± 77 continuum events.The relevant data distributions are displayed in Fig. 1, with fit projections overlaid, under selections in the analysis variables that enhance the signal component.The control-sample CP asymmetries are C = −0.12± 0.10 and S = −0.09± 0.12, where the uncertainties are statistical only, with correlation coefficient ρ = 0.06.The results are compatible with the null asymmetries we expect.In the fit to the signal B 0 → ϕK 0 S sample, displayed under the same signal-enhancing selections in Fig. 2  S and (right) B + → ϕK + candidates subtracted of the continuum background.The fit PDFs corresponding to q = −1 and q = +1 tagged distributions are shown as dashed and solid curves, respectively.The yield asymmetries, defined as (N (q = +1) − N (q = −1))/(N (q = +1) + N (q = −1)), are displayed in the bottom subpanels.
uncertainties are statistical only, with correlation coefficient ρ = 0.01.The observed continuum background asymmetry is compatible with zero.The ∆t distributions for tagged signal decays, after subtracting the continuum background [32], are displayed in Fig. 3, along with the resulting CP -violating asymmetries.

V. SYSTEMATIC UNCERTAINTIES
Contributions from all considered sources of systematic uncertainty are listed in Table II.We consider uncertainties associated with the calibration of the flavor tagging and resolution function, fit model, and determination of ∆t.
The leading contribution to the total systematic uncertainty on C arises by neglecting a possible timeintegrated CP asymmetry from BB backgrounds.The main systematic uncertainty on S comes from the fit bias, due to the modest statistical precision to which the fraction of B 0 → K + K − K 0 S backgrounds can be determined with the current sample size.We assess the uncertainty associated with the resolution function and flavor tagging parameters using simplified simulated samples.We generate ensembles assuming for each an alternative value for the above pa-rameters sampled from the statistical covariance matrix determined in the B 0 → D ( * )− π + control sample.Each ensemble is fitted using the nominal values of the calibration parameters and the standard deviation of the observed biases is used as a systematic uncertainty.
A similar procedure is used to assess a systematic uncertainty due to the systematic uncertainties on the calibration parameters, in which the ensembles are generated by varying each parameter independently within their systematic uncertainty.
We estimate the impact of differences in the resolution function and tagging performance between the signal and calibration samples.We apply the resolution function and flavor-tagging calibration obtained from a simulated B 0 → D ( * )− π + sample and repeat the measurement of C and S over an ensemble of simulated B 0 → ϕK 0 S events.The average deviation of the CP asymmetries from their generated values is assigned as a systematic uncertainty.

B. Fit model
To validate how accurately the fit determines the underlying physics parameters in the presence of backgrounds, we generate ensemble datasets that contain all the fit components.For each ensemble, we sample alternative values of C and S within the physical boundaries, and the fraction of the resonant events over the sum of resonant and nonresonant decays between 0.7 and 1.0, to account for the statistical precision on the observed value f ϕK = 0.89 ± 0.07.Due to the limited sample size, we assign a conservative systematic uncertainty for the fit bias by taking the largest deviations of the fitted values of C and S from their generated values.We also check that the relative magnitude of this systematic uncertainty with respect to the statistical uncertainty remains constant for larger sample sizes.
We study the effect of neglecting interference between the signal and nonresonant backgrounds using simulated samples, where the B 0 → ϕK 0 S and B 0 → K + K − K 0 S components are generated coherently using a complete Dalitz-plot description of the decay [8].We apply the nominal fit to these samples, where the nonresonant yields are determined by the fit and the CP -asymmetries of the backgrounds, C K + K − K 0 S and S K + K − K 0 S , are fixed to their generated values, neglecting interference with the signal.The difference between the generated and fitted values of the CP -asymmetries of the signal is assigned as a systematic uncertainty.
The effect of fixing the PDF shapes of the M bc , O ′ CS , cos θ H , and ∆t distributions in continuum, and O ′ CS distribution in signal and nonresonant background, is estimated from ensemble datasets.We generate simulated datasets by varying the shape parameters, in order to cover for the empirical parametrization and statistical uncertainty, and fix them to their nominal values in the fit.The resulting standard deviation on the distributions of C and S is used to estimate the corresponding systematic uncertainty.
The same procedure is applied to estimate the systematic uncertainty associated with the external inputs used for the lifetime τ B 0 = (1.519± 0.004) ps, mixing frequency ∆m d = (0.507 ± 0.002) ps −1 , and CP asymmetries C = 0.06 ± 0.08 and S = −0.68+0.09 −0.10 of the nonresonant background.
Simulation shows that the residual BB backgrounds is at most 2% of the signal yield.We generate ensemble datasets containing an additional BB background component with PDF shapes modeled after the B 0 → ϕK 0 S or B 0 → K + K − K 0 S distributions and by conservatively varying the BB background CP asymmetries between +1 and −1.The BB backgrounds are neglected in the fit to these datasets.The corresponding systematic uncertainty is obtained by taking the largest deviations of C and S from their generated values.
The time evolution given in Eq. ( 1) assumes that the B tag decays in a flavor-specific final state.We study the impact of the tag-side interference, i.e., neglecting the effect of CKM-suppressed b → ucd decays in the B tag in the model for ∆t [33].The observed asymmetries can be corrected for this effect by using the knowledge from previous measurements [3].We conservatively assume all events to be tagged by hadronic B decays, for which the effect is largest, and take the difference with respect to the observed asymmetries as a systematic uncertainty.
The effect of multiple candidates is evaluated by repeating the analysis with all the candidates and taking the difference with respect to the nominal candidate se-

C. ∆t measurement
The impact of the detector misalignment is tested on simulated samples reconstructed with various misalignment configurations.
The uncertainty on the momentum scale of charged particles due to the imperfect modeling of the magnetic field has a small impact on the CP asymmetries [31].
Similarly, the uncertainty on the coordinates of the e + e − interaction region (beam spot) has a subleading effect [31].
We do not account for the angular distribution of the B meson pairs in the c.m. frame when calculating ∆t using Eq. ( 2).Therefore, we estimate the effect of the ∆t approximation on simulated samples, where the generated and reconstructed time differences can be compared.

VI. SUMMARY
A measurement of CP violation in B 0 → ϕK 0 S decays is presented using data from the Belle II experiment.We find 162 ± 17 signal candidates in a sample containing (387 ± 6) × 10 6 BB events.The values of the CP asymmetries are C = −0.31± 0.20 ± 0.05 and S = 0.54 ± 0.26 +0.06 −0.08 , where the first uncertainty is statistical, and the second is systematic.The results are compatible with previous determinations from Belle and BABAR [8,9] and have a similar uncertainty on C, despite using a data sample 2.0 and 1.2 times smaller, respectively.When compared to measurements using a similar quasi-two-body approach [34,35], there is a 10% to 20% improvement on the statistical uncertainty on S for the same number of signal events.No significant discrepancy in the CP asymmetries between b → qqs and b → ccs transitions is observed.

Figure 1 :
Figure 1: Distributions of (top left) M bc , (top center) O ′ CS , (top right) cos θH , (bottom left) ∆t for B + -tagged and (bottom right) ∆t for B − -tagged B + → ϕK + candidates (data points) with fits overlaid (curves and stacked shaded areas).The M bc distribution is displayed for candidates with O ′ CS > −1 and the O ′ CS distribution is displayed for candidates with M bc > 5.27 GeV/c 2 .The cos θH and ∆t distributions are displayed for candidates with O ′ CS > −1 and M bc > 5.27 GeV/c 2 .

Figure 2 :
Figure 2: Distributions of (top left) M bc , (top center) O ′ CS , (top right) cos θH , (bottom left) ∆t for B 0 -tagged and (bottom right) ∆t for B 0 -tagged B 0 → ϕK 0 S candidates (data points) with fits overlaid (curves and stacked shaded areas).The M bc distribution is displayed for candidates with O ′ CS > −1 and the O ′ CS distribution is displayed for candidates with M bc > 5.27 GeV/c 2 .The cos θH and ∆t distributions are displayed for candidates with O ′ CS > −1 and M bc > 5.27 GeV/c 2 .

Table I :
Results of the fit to the signal and control samples.