Recent results from the Belle experiment
RRecent results from the Belle experiment
UCHEP-16-07
Z King and A J Schwartz(on behalf of the Belle Collaboration)
Physics Department, University of Cincinnati, P.O. Box 210011, Cincinnati, Ohio 45221 USAE-mail: [email protected], [email protected]
Abstract.
We review recent results from the Belle experiment, which took data at the KEKBasymmetric-energy e + e − collider in Japan. The experiment recorded about 1000 fb − of datarunning mainly at the Υ(4 S ) and Υ(5 S ) resonances. The results presented here are obtainedfrom the full data set.
1. Introduction
The Belle experiment successfully operated for more than a decade until 2010 at the asymmetric-energy e + e − collider KEKB [1]. The experiment took data at center-of-mass energiescorresponding to several Υ( n S) resonances; the total data sample recorded exceeds 1 ab − . Herewe present recent results based on the full data sample.The Belle detector is a large-solid-angle magnetic spectrometer that consists of a silicon vertexdetector (SVD), a 50-layer central drift chamber (CDC), an array of aerogel threshold Cherenkovcounters (ACC), a barrel-like arrangement of time-of-flight scintillation counters (TOF), andan electromagnetic calorimeter comprised of CsI(Tl) crystals (ECL) located inside a super-conducting solenoid coil that provides a 1.5 T magnetic field. An iron flux-return located outsideof the coil is instrumented to detect K L mesons and to identify muons (KLM). The detector isdescribed in detail elsewhere [2].
2. Search for the Decay B → φγ In the Standard Model (SM), the decay B → φγ proceeds through electroweak and gluonic b → d penguin annihilation processes. These amplitudes are proportional to the small Cabibbo-Kobayashi-Maskawa [3] matrix element V td and thus are highly suppressed. The branchingfraction has been estimated based on naive QCD factorization [4] and perturbative QCD [5] andfound to be in the range 10 − to 10 − . However, the internal loop can also be mediated bynon-SM particles such as a charged Higgs boson or supersymmetric squarks, and thus the decayis sensitive to new physics (NP). It is estimated that such NP could enhance the branchingfraction to the level of 10 − to 10 − [4]. Previously, no evidence for this decay has been found.Here we describe a recent Belle search for this decay with much higher sensitivity than previouslythat uses the full data set of 711 fb − recorded on the Υ(4S) resonance [6].Candidate φ mesons are reconstructed via φ → K + K − decays. The K + K − invariant massis required to be in the range [1 . , . c , which corresponds to 4 . σ in resolutionaround the φ mass. Candidate B mesons are identified using a modified beam-energy-constrainedmass M bc = (cid:113) E − | (cid:126)p B | c /c , and the energy difference ∆ E = E B − E beam , where E beam a r X i v : . [ h e p - e x ] D ec s the beam energy and (cid:126)p B and E B are the momentum and energy, respectively, of the B candidate. All quantities are evaluated in the center-of-mass frame. To improve the M bc resolution, the momentum (cid:126)p B is calculated as (cid:126)p φ + ( (cid:126)p γ / | p γ | ) (cid:113) ( E beam − E φ ) /c , where (cid:126)p γ is the photon momentum, and E φ and (cid:126)p φ are the energy and momentum, respectively, ofthe φ candidate. We require that events satisfy 5 .
25 GeV /c < M bc < .
29 GeV /c and − .
30 GeV < ∆ E < .
15 GeV. The signal yield is calculated in a smaller “signal region”5 .
27 GeV /c < M bc < .
29 GeV /c and − .
20 GeV < ∆ E < .
10 GeV.Charmless hadronic decays suffer from large backgrounds arising from continuum e + e − → q ¯ q ( q = u, d, s, c ) production. To suppress this background, we use a multivariate analyzer basedon a neural network (NN) [7]. The NN generates an output variable C NN , which ranges from − C NN > . C NN to C (cid:48) NN , defined as C (cid:48) NN = ln [( C NN − C min ) / ( C max − C NN )] where C min = 0.3 and C max = 1.0. The variable C (cid:48) NN is well-described by a sum of Gaussian functions.After the above selections, 961 events remain. The remaining background consists of continuumevents and rare charmless b -decay processes.We obtain the signal yield using an unbinned extended maximum likelihood fit to theobservables M bc , ∆ E , C (cid:48) NN , and cos θ φ . The helicity angle θ φ is the angle between the K + momentum and the opposite of the B flight direction in the φ rest frame. The resulting branchingfraction is calculated as B (cid:16) B → φγ (cid:17) = N sig N BB · ε · B ( φ → K + K − ) (1)where N sig = 3 . +4 . − . is the signal yield in the signal region, N BB = (772 ± × is thenumber of BB events, ε = 0 . ± .
001 is the signal efficiency as calculated from Monte Carlo(MC) simulation, and B ( φ → K + K − ) = (48 . ± . B ( B → φγ ) < . × − at 90% confidencelevel (C.L.). This limit is almost an order of magnitude lower than the previous most stringentresult [9].
3. Angular Analysis of B → K ∗ (892) (cid:96) + (cid:96) − The rare decay B → K ∗ (892) (cid:96) + (cid:96) − involves the quark transition b → s(cid:96) + (cid:96) − , a flavor changingneutral current that is forbidden at tree level in the SM. Higher order SM processes such aspenguin or W + W − box diagrams allow for such transitions, leading to branching fractionsbelow 10 − . Various extensions to the SM predict NP amplitudes that can interfere with theSM amplitudes and lead to enhanced or suppressed branching fractions and modified angulardistributions.Belle performed an angular analysis using the decay modes B → K ∗ (892) e + e − and B → K ∗ (892) µ + µ − [10]. K ∗ candidates are reconstructed in the channel K ∗ → K + π − [11]. Largecombinatoric background is suppressed by applying requirements on kinematic variables. Largebackgrounds also arise from charmonium decays B → K ( ∗ ) J/ψ and B → K ( ∗ ) ψ (2 S ), in which ψ → (cid:96) + (cid:96) − . To maximize signal efficiency and purity, neural networks are applied sequentiallybeginning from the end of the decay chain. Signal and background yields are determined froman unbinned extended maximum likelihood fit to M bc in bins of q . The bin ranges used and theresulting fitted yields are listed in Table 1. In total, 117 . ± . B → K ∗ (892) µ + µ − , and 69 . ± . B → K ∗ (892) e + e − .We subsequently perform the angular analysis. The decay is kinematically described by threeangles ( θ (cid:96) , θ K , and φ ) and the invariant mass squared of the lepton pair ( q ). The definitionsof the angles and the full angular distribution are described in Ref. [12]. For this fit we require M bc > .
27 GeV /c , and the number of signal ( N sig ) and background events ( N bkg ) are fixed tothe values in Table 1. able 1. Fitted yields in bins of q for signal ( N sig ) and background ( N bkg ), for electron andmuon channels combined.Bin q range in GeV /c N sig N bkg . − .
00 49 . ± . . ± .
51 0 . − .
00 30 . ± . . ± .
12 4 . − .
00 49 . ± . . ± .
03 10 . − .
90 39 . ± . . ± .
44 14 . − .
00 56 . ± . . ± . Figure 1.
Results for P (cid:48) as compared to SM predictions and to previous measurements byLHCb [15][16].The angular observables P (cid:48) i =4 , , , introduced in Ref. [13] contain information about short-distance effects. These observables are considered to be largely free from form-factoruncertainties [14], and thus they are a promising place to search for NP. As the statistics inthis analysis are insufficient to perform a full eight-dimensional fit, a “folding technique” is usedas described in Ref. [15]. We determine P (cid:48) , , , by performing a three-dimensional unbinnedmaximum likelihood fit in four bins of q using the folded signal probability density functionand fixed background yields and shapes. We also fit for the longitudinal polarization F L andthe transverse polarization asymmetry A (2) T . The fit results are shown in Fig. 1. These resultsare consistent with SM predictions, although one value of P (cid:48) differs by 2 . σ from the SM value.It is notable that this deviation is for the same q bin and in the same direction as a similardeviation reported by LHCb [15][16].
4. Measurement of the branching fraction of B → D ∗ + τ − ν τ relative to that of B → D ∗ + (cid:96) − ν (cid:96) Semi-tauonic B decays of the type b → cτ ν τ are sensitive probes to search for NP. Charged Higgsbosons, which appear in supersymmetry and other models with at least two Higgs doublets,may induce measurable effects in the branching fraction due to the large mass of the τ lepton.Similarly, leptoquarks, which carry both baryon number and lepton number, can also contributeo this process. The ratio of branching fractions R ( D ( ∗ ) ) = B ( B → D ( ∗ ) τ − ν τ ) B ( B → D ( ∗ ) (cid:96) − ν (cid:96) ) ( (cid:96) = e, µ ) (2)is typically measured instead of the absolute branching fraction B ( B → D ( ∗ ) τ − ν τ ) to reducesystematic uncertainties arising from reconstruction efficiencies, form factors, and the CKMmatrix element | V cb | . Standard Model calculations predict R ( D ∗ ) = 0 . ± .
003 [17] and R ( D ) = 0 . ± .
017 [18][19]. Semi-tauonic B decays were first observed by Belle [20], withsubsequent studies reported by Belle [19][21], B A B AR [22], and LHCb [23]. The world averagevalues [24] are R ( D ∗ ) = 0 . ± . ± .
012 and R ( D ) = 0 . ± . ± . . σ and 1 . σ , respectively.Thus far, measurements of R ( D ( ∗ ) ) at Belle and B A B AR have been performed using either ahadronic [19][22] or an inclusive [20][21] tagging method. Here we report the first measurementof R ( D ∗ ) using a semileptonic tagging method. We reconstruct signal events in modes in whichone B decays as B → D ∗ + (cid:96) − ν (cid:96) and the the other B decays as B → D ∗ + τ − ν τ , τ − → (cid:96) − ν (cid:96) ν τ .To reconstruct normalization events corresponding to the denominator in Eq. (2), we requirethat both B mesons decay to D ∗ + (cid:96) − ν (cid:96) .Signal and normalization events are identified using a neural network. The dominantbackground contribution arises from events with falsely reconstructed D ( ∗ ) mesons. To separatesignal and normalization events from backgrounds, we use the energy variable E ECL , which isdefined as the sum of the energies of neutral clusters detected in the electromagnetic calorimeter(ECL) that are not associated with reconstructed particles.We extract the signal and normalization yields using a two-dimensional extended maximum-likelihood fit to the neural network output NN and E ECL . The resulting yields of signal andnormalization events are 231 ± ± R ( D ∗ ) is0 . ± . ± . . σ higher than the SM prediction.We investigate the compatibility of the data with a Type II two-Higgs doublet model(2HDM) [25] and with R -type and S -type leptoquark models [26]. We find that the dataallows for additional contributions from 2HDM scalar operators and also vector operators, butcontributions from a tensor operator with 0 . < C T < .
39, an R -type leptoquark model with0 . < C T < .
38, or an S -type leptoquark model with 0 . < C T < .
28 are disfavored.
5. Observation of the decay B s → K K The two-body decays B s → hh (cid:48) , where h ( (cid:48) ) is either a charged π ± or K ± , have all beenobserved [8]. However, the neutral modes π π , π K , and K K have not. The decay B s → K K is of particular interest because the branching fraction is predicted to be large:(16 − × − [27]. The presence of non-SM particles or couplings could measurably affectthis value [28]. The current upper limit, B ( B s → K K ) < . × − at 90% C.L., was setby Belle using only 23 . − of data recorded at the Υ(5 S ) [29]. Here we update that resultusing the full data set of 121 . − [30]. At the Υ(5 S ), B s B s pairs are produced in threedecay channels: B s B s , B ∗ s B s or B s B ∗ s , and B ∗ B ∗ s . The latter two channels dominate, withproduction fractions of f B ∗ s B s = (7 . ± . f B ∗ s B ∗ s = (87 . ± . K mesons are reconstructed via the decay K S → π + π − using a neural networktechnique. To suppress background arising from e + e − → q ¯ q ( q = u, d, s, c ) production, a secondNN is used. The output of the latter ( C NN ) ranges from − C NN > − . C NN to thevariable C (cid:48) NN as described in Section 2. igure 2. Projections of the fit for B s → K K : (a) M bc for − .
11 GeV < ∆ E < .
02 GeVand C (cid:48) NN > .
5; (b) ∆ E for 5 .
405 GeV /c < M bc < .
427 GeV /c and C (cid:48) NN > .
5; and (c) C (cid:48) NN for 5 .
405 GeV /c < M bc < .
427 GeV /c and − .
11 GeV < ∆ E < .
02 GeV. The points witherror bars are data, the (green) dashed curves show the signal, (magenta) dotted curves showthe continuum background, and (blue) solid curves show the total.We measure the signal yield by performing an unbinned extended maximum likelihood fit tovariables M bc , ∆ E , and C (cid:48) NN . The results are 29 . +8 . − . signal events and 1095 . +33 . − . backgroundevents. Projections of the fit result are shown in Fig. 2. The branching fraction is calculated as B ( B s → K K ) = N sig N B s B s (0 . B K ε (3)where N sig is the fitted signal yield; N B s B s = (6 . ± . × is the number of B s B s events [32]; B K = (69 . ± . K S → π + π − [8]; and ε = (46 . ± . K K → K S K S (since K K is CP even). Inserting these values gives B ( B s → K K ) = (19 . +5 . − . ± . ± . × − , where the first uncertainty is statistical, thesecond is systematic, and the third reflects the uncertainty due to the total number of B s B s pairs. This result is in good agreement with the SM prediction [27].The signal significance is calculated as (cid:112) − L / L max ), where L is the likelihood valuewhen the signal yield is fixed to zero, and L max is the likelihood value of the nominal fit.Systematic uncertainties are included in the significance by convolving the likelihood functionwith a Gaussian function whose width is equal to the systematic uncertainty associated withthe signal yield. We obtain a signal significance of 5 . σ , and thus our measurement constitutesthe first observation of this decay.
6. Measurement of the branching fraction and CP asymmetry for D → V γCP asymmetries in the charm sector are a promising area in which to search for new physics.The CP asymmetry ( A CP ) of many two-body D decays has been measured [24], but all resultsare consistent with zero. Here we present a first search for a CP asymmetry in radiative D → V γ decays, where V is a neutral vector meson φ , K ∗ , or ρ . Within the SM, such asymmetriesare predicted to be small: A CP ∼ − . However, theoretical studies [33][34] indicate thatextensions to the SM with chromomagnetic dipole operators can increase A CP up to a fewpercent. Several D → V γ decay modes have been measured by Belle and B A B AR [35][36].The current world average branching fractions are B ( D → φγ ) = (2 . ± . × − and B ( D → K ∗ γ ) = (32 . ± . × − [8]. The D → ρ γ mode has not yet been observed, andthe current upper limit on the branching fraction is B ( D → ρ γ ) < × − at 90% C.L. [8].Here we present new measurements of all three modes using the full Belle data set. Our resultsinclude the first measurement of the CP asymmetries. For our analysis, the D is requiredto originate from D ∗ + → D π + decays; this tags the D flavor and suppresses combinatoricackground. The vector mesons in the final state are reconstructed via φ → K + K − , K ∗ → K − π + , and ρ → π + π − .All branching fractions and A CP are determined by normalizing to decay channels D → K + K − for D → φγ , D → K − π + for D → K ∗ γ , and D → π + π − for D → ρ γ . The branchingfractions are calculated as B sig = B norm × N sig N norm × ε norm ε sig (4)where B and N are the branching fraction and fitted yield, respectively, of signal or normalizationmodes, and ε is the reconstruction efficiency. For B norm , the world-average value [8] is used.The extracted raw asymmetry A raw = N ( D ) − N ( D ) N ( D ) + N ( D ) (5)has several contributions: A raw = A CP + A FB + A ± ε , where A FB is the forward-backwardproduction asymmetry and A ± ε is the detection asymmetry between positively and negativelycharged particles. Both of these asymmetries are eliminated by measuring A CP of a signal moderelative to that of its normalization mode, which has the same final-state charged particles. The CP asymmetry of a signal mode is then A sig CP = A sigraw − A normraw + A norm CP , where A norm CP is the PDGvalue of the CP asymmetry for the normalization mode [8].To extract the signal yields and CP asymmetries, we perform a simultaneous two-dimensionalunbinned extended maximum likelihood fit of the D and D samples. The fit variables are theinvariant mass of the D and the cosine of the helicity angle θ H , defined as the angle betweenthe D and the positively or negatively charged hadron in the rest frame of the V meson. For D candidates we use the K + / K − / π + for φ / K ∗ / ρ decays, and we use the oppositely chargedparticles for D candidates. The fitted signal yields are N φγ = 524 ± N K ∗ γ = 9104 ± N ρ γ = 500 ±
85. The resulting branching fractions are B ( D → φγ ) = (2 . ± . ± . × − B ( D → K ∗ γ ) = (4 . ± . ± . × − B ( D → ρ γ ) = (1 . ± . ± . × − , where the first uncertainty is statistical and the second is systematic. The result for φγ isimproved with respect to the previous result [35]. The result for K ∗ γ is 3.3 σ higher than the B A B AR result [36]. The result for ρ γ is close to that for φγ , which is consistent with theoreticalexpectations. The significance is calculated as (cid:112) − L / L max ) as described in Section 5.Systematic uncertainties are included in the significance by convolving the likelihood functionwith a Gaussian whose width is equal to the systematic uncertainty associated with the signalyield. The significance for ρ γ is 5 . σ , and thus our measurement constitutes the first observationof this decay.The results for A CP are A CP ( D → φγ ) = − . ± . ± . A CP ( D → K ∗ γ ) = − . ± . ± . A CP ( D → ρ γ ) = 0 . ± . ± . . No CP asymmetry is seen in any of these radiative decays. Acknowledgements
The authors thank the BEACH 2016 organizers for a well-run workshop and excellent hospitality.This research is supported by the U.S. Department of Energy. eferences [1] Kurokawa S and Kikutani E 2003
Nucl. Instrum. Methods Phys. Res., Sect. A et al Prog. Theor. Exp. Phys. et al (Belle Collaboration) 2002
Nucl. Instr. and Meth. A Phys. Rev. Lett. Prog. Theor. Phys. Eur. Phys. J. C Eur. Phys. J. C Chin. Phys. Lett. et al (Belle Collaboration) 2016 Phys. Rev. D Nucl. Instrum. Methods Phys. Res. Sect. A et al (Particle Data Group) 2014 Chin. Phys. C et al ( B A B AR Collaboration) 2005
Phys. Rev. D et al (Belle Collaboration) 2016 arXiv:1604.04042[11] Charge-conjugate modes are implicitly included unless noted otherwise[12] Aaij R et al (LHCb Collaboration) 2013 JHEP JHEP JHEP et al (LHCb Collaboration) 2013
Phys. Rev. Lett. et al (LHCb Collaboration) 2016
JHEP Phys. Rev. D Phys. Rev. D et al (Belle Collaboration) 2015 Phys. Rev. D et al (Belle Collaboration) 2007 Phys. Rev. Lett. et al (Belle Collaboration) 2010 Phys. Rev. D et al ( B A B AR Collaboration) 2012
Phys. Rev. Lett. et al ( B A B AR Collaboration) 2013
Phys. Rev. D et al (LHCb Collaboration) 2015 Phys. Rev. Lett. et al
Zeit. Phys. C Phys. Rev. D Phys. Lett. B Phys. Rev. D Phys. Rev. D Phys. Rev. D Phys. Rev. D Phys. Rev. D Phys. Lett. B Phys. Rev. D J. Phys. G et al (Belle Collaboration) 2010 Phys. Rev. D et al (Belle Collaboration) 2016 Phys. Rev. Lett. et al (Belle Collaboration) 2013
Phys. Rev. D et al (Belle Collaboration) 2015 Phys. Rev. D Phys. Rev. Lett. et al (Belle Collaboration) 2014
Phys. Rev. Lett. et al ( B A B AR Collaboration) 2008
Phys. Rev. D78