A Study of J/psi-->gamma gamma V(rho,phi) Decays with the BESII Detector

Using a sample of $58\times 10^6$ $J/\psi$ events collected with the BESII detector, radiative decays $J/\psi\to\gamma\gamma V$, where $V=\rho$ or $\phi$, are studied. A resonance around 1420 MeV/c$^2$ (X(1424)) is observed in the $\gamma\rho$ mass spectrum. Its mass and width are measured to be $1424\pm 10(stat)\pm 11(sys)$ MeV/c$^2$ and $ 101.0\pm 8.8 \pm 8.8$ MeV/c$^2$, respectively, and its branching ratio $B(J/\psi\to \gamma X(1424)\to \gamma \gamma \rho)$ is determined to be $(1.07\pm0.17 \pm 0.11)\times 10^{-4}$. A search for $X(1424)\to \gamma\phi$ yields a 95% C.L. upper limit $B(J/\psi\to \gamma X(1424)\to \gamma\gamma \phi)<0.82 \times 10^{-4}$.


Introduction
Experimentally the structure of the η(1440) remains unresolved. The existence of two overlapping pseudo-scalar states has been suggested: one around 1410 MeV/c 2 decays into both KKπ and ηππ, and the other around 1470 MeV/c 2 decays only to KKπ [1,2]. It is therefore conceivable that the higher mass state is the ss member of the 2 1 S 0 nonet [3], while the lower mass state may contain a large gluonic content [4].
Standard perturbative theory predicts [5] that if the η(1440) is a qq state which decays in a flavor independent way, the partial width relationship between its γρ, and γφ final states should be Γ γρ : Γ γφ = 9 : 2. A simultaneous search for a resonance near 1440 MeV/c 2 in the γρ and γφ mass spectra and a determination of the branching ratios of the resonance may shed light on the internal structure of the η(1440).
Radiative decays of a resonance near 1440 MeV/c 2 to γV (V = ρ, and φ) have been studied previously in J/ψ → γγV events by Crystal-Ball [6], MarkIII [7] and DM2 [8]. The situation here is further complicated by the proximity of the f 1 (1420) to the η(1440). MarkIII finds that the 0 − is only slightly favored over the 1 + in a fit to their angular distributions in J/ψ → γγρ [7].
In this letter, we report a study of decays J/ψ → γγρ and J/ψ → γγφ selected from a sample of 58 × 10 6 J/ψ events collected by the Beijing Spectrometer (BESII) detector.

Event selection
We want to study where X is a resonance, and V denotes vector mesons ρ or φ, which are reconstructed via their decays ρ → π + π − and φ → K + K − .
The BESII detector has been described in detail elsewhere [9]. In this study two oppositely charged particles must be detected in the main drift chamber. Photons are detected by the barrel shower counter (BSC) which covers 80% of the 4π solid angle with an energy resolution δE/E = 21%/ √ E. In order to remove electronic noise, the energy deposited in the BSC by each neutral particle is required to have a minimum of 70 MeV. A photon is required to be isolated from charged tracks (cos θ γπ(K) < 0.98, where θ γπ(K) is the angle between the photon and a charged particle) to reject any photons radiated by a charged particle in the event, and to be consistent with originating from the event interaction point. Photon candidates satisfying these criteria are used for this analysis. The highest energy photon in an event is taken as the radiative photon directly produced in J/ψ → γX events.

Selection of
Monte Carlo (MC) simulations have been carried out for both the signal and background processes. The backgrounds considered here are radiative J/ψ decays into two charged tracks, namely (mγ)π + π − (m=1,2,3,4) and (nγ)K + K − (n=1,2,3,4) for which known branching ratios compiled by the Particle Data Group (PDG) [1] are used to form the correct mixture of these processes in the Monte Carlo background simulation. The most important backgrounds for this channel are J/ψ → γη c , J/ψ → γηππ, J/ψ → γf 1 (1510) → γηππ, J/ψ → ωπ 0 , J/ψ → ωη, J/ψ → a 2 (1320)ρ, and J/ψ → b 0 1 (1235)π 0 . The generated Monte Carlo samples of signal and background are analyzed, and selection variables are varied until an optimized ratio of signal to background is reached. As a result, the following criteria are chosen for the J/ψ → γγρ → γγπ + π − analysis: (1) The sum of momenta of the charged tracks in the event ( P miss ) is less than 1.14 GeV/c, (2) at least one of the two charged tracks in the event must have a higher particle identification confidence level for the pion hypothesis than for the kaon hypothesis by combining the information from TOF and dE/dx, (3) the χ 2 of a four constraint kinematic fit of the event to a γγπ + π − topology is less than 10.0, (4) the total energy of any photons not used in the kinematic fit in criterion (3) is less than 250 MeV, (5) the invariant mass of the two selected photons must be greater than 0.66 GeV/c 2 , and (6) the helicity angle θ of the dipion in the γπ + π − system must satisfy | cos θ| < 0.86. Fig. 1 shows the π + π − invariant mass spectrum of the selected γγπ + π − events, where a clear ρ signal is visible. To select ρ candidates, π + π − pairs must satisfy |M π + π − − M ρ | ≤ 0.28 GeV/c 2 , as indicated in Fig. 1. Combining the ρ candidate with the lower energy photon in the γγπ + π − event, the γρ mass distribution, shown in Fig. 2, is obtained.

Selection of
The selection criteria for J/ψ → γX → γγφ (φ → K + K − ) have been chosen in a similar way to those for the γγπ + π − . According to the Monte Carlo simulation, we find that the major sources of backgrounds are from J/ψ → φf 0 (980), J/ψ → γη(1440) → γKKπ, and J/ψ → γf 1 (1420) → γKKπ. The final requirements are: P miss ≤ 1.31 GeV/c, χ 2 ≤ 25 for the four constraint kinematic fit to the J/ψ → γγK + K − hypothesis, and at least one identified charged kaon must be present in the event. The other criteria remain the same for this mode as for the γγπ + π − analysis.   Fig. 3 are the side-band regions, as well as the φ signal region. The γφ mass distribution is shown in Fig. 4. The lower energy γ is also combined with the φ side-band events forming the dashed distribution, also shown in Fig. 4.

Analysis and Results
We have fitted the mass distributions in Figs. 1 and 3, and estimate that they contain 38249±490 ρ and 764±64 φ events, respectively. The insert in Fig. 2 shows the full γρ mass range, where a strong J/ψ → γη ′ (958) → γγπ + π − signal is observed, as expected. To verify that the mass scale is correct, we have fitted the η(958) signal and obtain 957.5±0.2 MeV/c 2 for the mass and 0.20±0.04 MeV/c 2 for the width, which are in excellent agreement with the world average values of 957.78±0.14 MeV/c 2 and 0.202±0.016 MeV/c 2 , respectively.
Two enhancements above 1.2 GeV/c 2 in the γρ mass spectrum are evident in Fig. 2. We have examined the γπ + π − mass distribution for π + π − pairs with masses just above the upper edge of the ρ mass band. The distribution does not exhibit any distinct structures. We conclude that the peaks in Fig. 2 are associated with the ρ.
The identical selection criteria have been applied to a sample of 30 million Monte Carlo inclusive J/ψ events which do not contain the decay J/ψ → γX(1420) → γγρ → γγπ + π − . The resulting Monte Carlo γπ + π − mass distribution does not show the enhancement at 1420 MeV/c 2 but does show the f 1 (1285), as expected.
In order to extract the resonance parameters in Fig. 2, we perform an unbinned maximum-likelihood fit to the data. The fit function consists of two Breit-Wigner functions, each convoluted with a Gaussian with a mass resolution of 12 MeV/c 2 , for the signals (f 1 (1285) and X(1424)), and a polynomial function for the background. The χ 2 /dof of the fit is 68.3/48. In order to check whether the background shape in our fit is correct, we compared it with the background from our J/ψ inclusive MC sample and find that the backgrounds are consistent. The results of the fit are shown in Fig. 2 and summarized in Table 1, where the first errors are statistical errors obtained from the fit and the second are systematic.
The systematic errors on the mass and the width for the first resonance (1276) are determined from the variations when different background functions are used in the fit, about 0.07% and 19.5%, respectively, and from the uncertainty of the Monte Carlo simulation, about 0.6% and 12.5%, respectively. The systematic errors on the mass and the width for the second resonance (1424) include the background function variations, about 0.01% and 2.2%, respectively, and the uncertainty of the Monte Carlo simulation, about 0.8% and 8.4%, respectively.
The detection efficiencies for J/ψ → γX → γγρ → γγπ + π − are determined from a Monte Carlo simulation to be (9.3 ± 0.1)% at 1.285 GeV/c 2 and (8.81 ±0.09)% at 1.420 GeV/c 2 . The systematic errors on the branching ratios are determined by combining the Monte Carlo uncertainty on the efficiencies (8.4%), the error on the number of J/ψ events (5.0%), and the variation in the number of signal events due to the different background shapes used in the fit (13.2% and 6.5% for the first and second resonances, respectively).
To determine whether the X(1424) is more likely to be the f 1 (1420) or the Comparing this limit to our measurement of B(J/ψ → γX(1424) → γγρ) = (1.07 ± 0.17 ± 0.11) × 10 −4 , we conclude that of X(1424) in J/ψ → γγρ 0 channel should be predominantly η(1440). For the resonance around 1276 MeV/c 2 , MarkIII [7] finds that the 1 + hypothesis is preferred over the 0 − by about 4 σ, leading to the conclusion that it is f 1 (1285). The BESII results for the mass, width, and branching fraction of the lower mass peak are consistent with those of MarkIII, as shown in Table 2. From MarkIII analysis, it is not distinguishable between 0 − and 1 + hypothesis for the X(1432) state, but the 0 − slightly better than 1 + by about 2σ .
By combining the information from TOF and dE/dx one can clearly distinguish K from π for charged kaon momenta between 200 MeV/c and 800 MeV/c. Therefore the contamination from π misidentified as K can be neglected.
According to the Monte Carlo simulation, we find that the main backgrounds are as mentioned in Section 2.2. They arise from the decays J/ψ → φf 0 (980), and J/ψ → γη(1440) → γKKπ and J/ψ → γf 1 (1420)) → γKKπ. These background events are very difficult to reject in our event selection. For these processes a comparison of the two γK + K − invariant mass spectra, derived from the K + K − mass around the φ signal and from K + K − from the φ sidebands, shows comparable contributions in the γK + K − mass region around 1400 MeV/c 2 . Therefore, the φ side-bands can be used to estimate the background in the γφ spectrum. In Fig. 5 the side-bands subtracted γφ mass spectrum is shown, and no significant peak around 1420 MeV/c 2 is observed.  Table 2 shows a comparison of results from BESII (this work) and other experiments.

Acknowledgments
We would like to thank Profs. F.A. Harris, X.C. Lou, D.V. Bugg, H. Yu and Dr. J.D. Richman for valuable discussions.