Observation of Upsilon(4S) decays to pi(+)pi(-)Upsilon(1S) and pi(+)pi(-)Upsilon(2S).

Observation of Upsilon(4S) decays to pi(+)pi(-)C and pi(+)pi(-)Upsilon(2S)We present the first measurement of Upsilon(4S) decays to pi(+)pi(-)Upsilon(1S) based on a sample of 230 x 106(4S) mesons collected with the BABAR detector. We measure the product branching fractions Beta(Upsilon(4S) --> pi(+)pi(-)Upsilon(1S)) x BetaUpsilon(1S) --> mu(+)mu(-) = (2.23 +/- 0.25(stat) +/- 0.27(syst))x 10(-6) and Beta(Upsilon(4S) --> pi(+)pi(-)Upsilon(2S) x Beta(Upsilon(2S) --> mu(+)mu(-))=(1.69 +/-0.26(stat) +/- 0.20(syst)) x 10(-)6, from which we derive the partial widths Gamma(Upsilon(4S) --> pi(+)pi(-)Upsilon(1S))=(1.8 +/-0.4) keV and Gamma(Upsilon(4S) --> pi(+)pi(-)Upsilon(2S))=(2.7 +/- 0.8) keV.

We present the first measurement of 4S decays to ÿ 1S and ÿ 2S based on a sample of 230 10 6 4S mesons collected with the BABAR detector. We measure the product branching fractions B4S ! ÿ 1S B1S ! ÿ 2:23 0:25 stat 0:27 syst 10 ÿ6 and B4S ! ÿ 2S B2S ! ÿ 1:69 0:26 stat 0:20 syst 10 ÿ6 , from which we derive the partial widths ÿ4S ! ÿ 1S 1:8 0:4 keV and ÿ4S ! ÿ 2S 2:7 0:8 keV. DOI The 4S meson is known to decay predominantly to B B, with small, but as of yet unobserved, decays to other bottomonium states or to light hadrons. Partial widths for hadronic transitions in heavy quarkonia have been extensively studied both experimentally and theoretically over the past decades [1]. In particular, the values of the partial widths for dipion transitions between vector states 2S ! ÿ J= and mS ! ÿ nS, where the principal quantum number m > n, can be related to the radial wave function within the framework of the QCD multipole expansion [2]. This picture may be significantly altered by mixing and coupled channel effects [3] when states are close to the threshold for open charm or bottom production. Hence these states are the ideal laboratory to investigate these effects. Exclusive non-D D decays of the 3770 (believed to be predominantly 3 D 1 ) have recently been observed [4 -6], but only upper limits have been published for exclusive non-B B decays of the 4S [7]. We search for the decays 4S ! ÿ nS, where n 1; 2 [the 4S ! 3S transition is kinematically not allowed], using a sample of 230 10 6 4S events corresponding to an integrated luminosity of 211 fb ÿ1 acquired near the peak of the 4S resonance with the PEP-II asymmetric-energy e e ÿ storage rings at SLAC. An additional 22 fb ÿ1 sample collected approximately 40 MeV below the resonance is used as a control sample.
The BABAR detector is described in detail elsewhere [8]; here we summarize only the features relevant to this analysis: charged-particle momenta are measured in a tracking system consisting of a five-layer double-sided silicon vertex tracker (SVT) and a 40-layer central drift chamber (DCH), both situated in a 1.5-T axial magnetic field. Charged-particle identification is based on the dE=dx measured in the SVT and DCH, and on a measurement of the photons produced in the synthetic fused-silica bars of the ring-imaging Cherenkov detector (DIRC). A CsI(Tl) electromagnetic calorimeter (EMC) is used to detect and identify photons and electrons, while muons are identified in the instrumented flux return of the magnet (IFR).
An mS ! ÿ nS transition, denoted by mS ! nS, is detected by reconstructing the nS meson via its leptonic decay to ÿ . The sensitivity to 4S ! nS transitions is much smaller in the ÿ e e ÿ final state due to the presence of larger backgrounds, and to a trigger-level inefficiency introduced by the prescaling of Bhabha scattering events. Data collected at a nominal center-of-mass energy s p near 10.58 GeV include 3S ! nS (n 1; 2) and 2S ! 1S events from initial state radiation (ISR) production that are used as control samples. The signature for mS ! nS transition events, where the nS decays to muons, is a ÿ invariant mass, M , that is compatible with the known mass [9] of the nS resonance, MnS, and an invariant mass difference M M ÿ M that is compatible with MmS ÿ MnS.
The rms values of the reconstructed M and M distributions are, respectively, 7 MeV=c 2 and 75 MeV=c 2 . The center-of-mass momentum p cand should be compatible with 0 for 4S ! nS candidates, or with s ÿ M 2 mS= 2 s p for mS ! nS candidates from ISR.
Simulated Monte Carlo (MC) events are generated using the EVTGEN package [10]. The angular distribution of generated dilepton decays incorporates the nS polarization, while dipion transitions are generated according to phase space. These events are passed through a detector simulation based on GEANT4 [11], and analyzed in the same manner as data. The events in the data sample whose values of M and M are within 60 MeV=c 2 and 300 MeV=c 2 , respectively, of the values expected for any known mS ! nS transition were not examined until the event selection criteria were finalized. Events outside these regions were used to understand the background.
We select events having at least 4 charged tracks with a polar angle within the fiducial volume of the tracking system (0:41 < < 2:54 rad). Each muon candidate is required to have a center-of-mass momentum greater than 4 GeV=c, and to be compatible with the muon hypothesis based on the energy deposited in the EMC and the hit pattern in the IFR along the track trajectory. A dipion candidate is formed from a pair of oppositely charged tracks. The two pion candidates are each required to have a transverse momentum greater than 100 MeV=c. The dimuon and the dipion are constrained to a common vertex, and the vertex fit is required to have a 2 probability larger than 10 ÿ3 .
A large fraction of the background is due to ÿ events where a photon converts in the detector material. To reduce this background we apply an ''electron veto,'' rejecting events where any of the following is true: either of the two pion candidates is positively identified as an electron; the e e ÿ invariant mass of the two charged tracks associated with the pion candidates satisfies M ee < 100 MeV=c 2 ; or the dipion opening angle satisfies We also observe signals for 2S ! 1S, 3S ! 2S, and 3S ! 1S from ISR at M; M 0:563; 9:460GeV=c 2 , 0:332; 10:023 GeV=c 2 , and 0:895; 9:460 GeV=c 2 , respectively. The diagonal band is predominantly due to events, while the cluster at M; M 0:332; 9:460 GeV=c 2 is due to 3S ! ÿ 2S decays, where 2S ! 1SX.
The number of signal events N sig is extracted by an unbinned extended maximum likelihood fit to the M distribution for events with p cand < 200 MeV=c and jM ÿ M1Sj < 200 MeV=c 2 for the 4S ! 1S mode or jM ÿ M2Sj < 150 MeV=c 2 for the 4S ! 2S mode (Fig. 2). In each case, the background is parametrized as a linear function, and the signal as the convolution of a Gaussian with standard deviation and a Cauchy function with width ÿ, which is found to adequately describe the non-Gaussian tails of the M distribution. The values for and ÿ are, for each mode, fixed to the values determined from a fit to a MC signal sample subjected to the detector simulation and reconstruction algorithms. We verify that the experimental M resolution is well described by the MC simulation for 2S ! 1S and 3S ! nS and M4S ÿ M2S 0:5567 0:0035 GeV=c 2 [9]. These values can-not be interpreted as a new measurement of the 4S mass: the data were collected at s p equal to the world average value of M4S. Since the 4S width is larger than the spread in the s p of the e e ÿ collisions, a scan of the 4S line shape would be needed to measure the mass. The cuts described above are also applied to ÿ e e ÿ candidates, with the additional requirement on the polar angle of the electron, e ÿ > 0:75 radians, to reject Bhabha events. The fits to the electron samples are also shown in Fig. 2, and give yields and M values consistent with expectations based on the fits to the muon samples.
The significance, estimated from the likelihood ratio n ' 2 logLN sig =L0 q between a fit that includes a signal function and a fit with only a background hypothesis, is 10:0 for 4S ! 1S and 7:3 for 4S ! 2S in the ÿ ÿ final states. The significance of the signals in the ÿ e e ÿ final states is 3:6 and 2:5 for 4S ! 1S and 4S ! 2S, respectively.
The event selection efficiency sel is determined using the MC samples. The largest source of systematic uncertainty (10%) is due to the unknown distribution of the dipion invariant mass in the 4S ! ÿ nS transition, and is estimated by comparing the acceptance for a phase space distribution to that obtained using the QCD multipole model [2]. The second largest source of systematic uncertainty is due to uncertainty in the track reconstruction efficiency, which is 1.3% per track, resulting in a 5.2% uncertainty in sel . The systematic uncertainties as-  sociated with the event selection (4.3%) and muon identification (1.4%) criteria are estimated by comparing the efficiency of each selection criterion determined from MC samples to the corresponding efficiency measured with the ISR control samples. We have also considered the systematic uncertainties due to the choice of signal and background parametrizations by using different functions or different parameters, and the systematic uncertainties due to the choice of the fit range. The contributions from these sources are negligible in comparison to the previously mentioned sources.
The product branching fraction (Table I) is determined from the ÿ ÿ sample using: The dipion invariant mass distribution, M ÿ (Fig. 3), is determined by fitting the M distribution in equal intervals of M ÿ , and dividing the number of signal events in each interval by the corresponding selection efficiency. The measured distribution for the 4S ! 1S transition has a shape similar to the prediction of the Kuang-Yan model [2]. This model provides a good description of the observed distributions for 2S ! 1S, 3S ! 2S, and also 2S ! ÿ J= , but fails to describe the 3S ! 1S distribution. Our measured distribution for the 4S ! 2S transition has a marked enhancement at low M ÿ that is incompatible with this model.
In conclusion, we measure The dipion invariant mass distribution is measured for 4S ! ÿ 1S and 4S ! ÿ 2S transitions; the latter is found to be incompatible with predictions from QCD multipole expansions. 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.