Observation of charmless baryonic decays $B^0_{(s)} \to p \overline{p} h^+ h^{\prime-}$

Decays of $B^0$ and $B_{s}^0$ mesons to the charmless baryonic final states $p \overline{p} h^+ h^{\prime-}$, where $h$ and $h^\prime$ each denote a kaon or a pion, are searched for using the LHCb detector. The analysis is based on a sample of proton-proton collision data collected at center-of-mass energies of $7$ and $8\,$TeV, corresponding to an integrated luminosity of $3\,$fb$^{-1}$. Four-body charmless baryonic $B_{s}^0$ decays are observed for the first time. The decays $B^0_{s}\to p \overline{p} K^+ K^-$, $B^0_{s}\to p \overline{p} K^\pm \pi^\mp$, $B^0\to p \overline{p} K^\pm \pi^\mp$ and $B^0\to p \overline{p} \pi^+\pi^-$ are observed with a significance greater than $5$ standard deviations; evidence at $4.1$ standard deviations is found for the $B^0\to p \overline{p} K^+ K^-$ decay and an upper limit is set on the branching fraction for $B^0_{s}\to p \overline{p} \pi^+ \pi^-$. Branching fractions in the kinematic region $m(p \overline{p})<2850\,$MeV$/c^2$ are measured relative to the $B^0 \to J/\psi (\to p \overline{p}) K^*(892)^0$ channel.

secondary vertices consistent with the decay of a b hadron.
The final selection of B 0 (s) candidates, formed by combining four charged hadron candidates -a proton, an antiproton and an oppositely charged pair of light mesons -is carried out with a filtering stage followed by requirements on the response of a boosted decision tree (BDT) classifier [23,24] and on particle identification (PID). The filtering stage includes requirements on the quality, p, p T and χ 2 IP of the tracks, loose PID requirements and an upper limit on the pp invariant mass; the χ 2 IP is defined as the difference between the vertex-fit χ 2 of a PV reconstructed with and without the track in question. Each B 0 (s) candidate must have a good-quality vertex that is displaced from the associated PV (that with which it forms the smallest χ 2 IP ), must satisfy p and p T requirements, and must have a reconstructed invariant mass close to that of a B 0 (s) meson under the signal mass hypothesis. A requirement is also imposed on the angle ϑ dir between the candidate momentum vector and the line between the associated PV and the candidate decay vertex.
There are 15 input quantities to the BDT classifier: p T , η, χ 2 IP , ϑ dir and the flight distance of the B 0 (s) candidate; the quality of the B 0 (s) vertex fit; the p T and χ 2 IP of the tracks; and the largest distance of closest approach between any pair of tracks. The BDT is trained using simulated B 0 (s) → pphh signal candidates, generated with uniform distributions over phase space, and events in a high sideband of the ppKπ invariant mass in data (m(ppKπ) in the range 5450-5550 MeV/c 2 ) to represent the background. Tight PID requirements are applied to all final-state particles to reduce the combinatorial background, suppress the cross-feed backgrounds between the different pphh final states -background from other signal decays where one particle is misidentified -and ensure that the datasets for the three pphh final states are mutually exclusive. For each final state individually, the requirements on the PID and BDT response are optimized for the signal significance using simulation samples for the signal. After all selection requirements are applied, approximately 3% of events with at least one candidate also contain a second candidate; a candidate is then selected at random. The efficiency of the full reconstruction and selection, including the acceptance and the trigger selection, is approximately 0.1%.
To reject contributions from intermediate charm states, candidates with hh invariant mass consistent with a D 0 meson or phh invariant mass consistent with a Λ + c baryon are removed. The contribution from the charmonium region is removed by requiring the invariant mass of the pp pair to be less than 2850 MeV/c 2 , similar to the procedure in Refs. [5,25]. This last requirement is not applied to the normalization mode B 0 → J/ψ K * (892) 0 , where the vector mesons are reconstructed in the J/ψ → pp and K * (892) 0 → K + π − decay modes. All the other steps of the selection are in common for the signal and the normalization modes.
The yields of the signal decays are obtained from a simultaneous unbinned extended maximum likelihood fit to the B 0 (s) candidate invariant mass distributions in the three pphh final states in the range 5165-5525 MeV/c 2 . This approach accounts for potential cross-feed from one channel to another due to particle misidentification. Each signal component is modeled with a double-sided Crystal Ball (DSCB) function [26]. For each signal the tail parameters of the DSCB functions are determined from simulation. The peak position of the B 0 signals is common to the three final states, while the difference between the peak positions of the B 0 and B 0 s signals is constrained to its known value [11]. The width of the B 0 signal is a free parameter in the ppKπ final state and it is related to the width in the other two final states by scale factors determined from simulation. The same applies to the width of the B 0 s signals, which is a free parameter only in the ppKK final state.
For each final state the dominant B 0 (s) → pphh cross-feed background is included: the B 0 → ppKπ mode in the ppKK and ppππ invariant mass distributions, and the B 0 → ppππ mode in the ppKπ spectrum. Each cross-feed background is modeled with a DSCB function with all the shape parameters fixed according to simulation; the yield is fixed relative to the yield in the correctly reconstructed final state taking into account the (mis)identification probabilities calibrated using data, as described below. In addition, a combinatorial background component modeled by an exponential function, with both parameters free to vary, is present for each final state.
The yield of the normalization decay is determined from a separate simultaneous fit to the ppKπ, pp and Kπ invariant mass distributions in the ranges 5180-5380 MeV/c 2 , 3047-3147 MeV/c 2 and 642-1092 MeV/c 2 , respectively. The B 0 → J/ψ K * (892) 0 component is parameterized in the Kπ invariant mass distribution by a relativistic spin-1 Breit-Wigner function and in the ppKπ and pp invariant mass distributions by DSCB functions with the tail parameters fixed from simulation. The Kπ S-wave component is modeled in the Kπ invariant mass distribution by the LASS parametrization [27,28] that describes nonresonant and K * 0 (1430) 0 S-wave contributions; this component is modeled in the ppKπ and pp invariant mass distributions with the same shape as the B 0 → J/ψ K * (892) 0 component. A combinatorial background component modeled by a freely varying exponential function is also present in each spectrum.
The pphh invariant mass distributions with the results of the fit overlaid are shown in Fig. 1 while the signal yields and the significances are collected in Table 1. The significance of each of the signal modes is determined from the change in likelihood when the corresponding yield is fixed to zero, with systematic uncertainties taken into account [29]. The B 0 s → ppKπ, B 0 → ppKK and B 0 s → ppππ modes are found to have significances of 6.5 standard deviations (σ), 4.1 σ and 2.6 σ, respectively, while the other signal modes have significances greater than 25 σ.
The branching fractions of the B 0 (s) → pphh decays are determined relative to the visible branching fraction of the B 0 → J/ψ K * (892) 0 decay using where f s /f d = 0.259 ± 0.015 (included only for the B 0 s ) is the ratio of b hadronization probabilities, f q , to the hadron B q [30], and N corr denote efficiency-corrected fitted signal yields. The yields are obtained from the mass fits, while simulation is used to evaluate the contribution to the efficiency from each stage of the selection except for the effect of the PID criteria. The latter is determined from calibration data samples of kinematically identified pions, kaons and protons originating from the decays D * + → D 0 (→ K − π + )π + , Λ → pπ − and Λ + c → pK − π + and weighted according to the kinematics of the signal particles [31,32]. For each final state the efficiencies are determined as a function of the position in phase space, and efficiency corrections for each candidate are applied using the method of Ref. [33] to take the variation over the phase space into account. Explicitly, N corr = i W i /ε i , where the sum runs over the candidates in the fit, W i is the sWeight for candidate i determined with the sPlot method [34] and ε i is the efficiency for the candidate i which depends only on its position in the five-dimensional phase space. The visible branching fraction of the normalization mode, defined as B(B 0 → J/ψ K * (892) 0 ) ×   B(J/ψ → pp) × B(K * (892) 0 → K + π − ), is B vis (B 0 → J/ψ K * (892) 0 ) = (1.68 ± 0.12) × 10 −6 , where the B 0 → J/ψ K * (892) 0 branching fraction is taken from Ref. [35] and the others from Ref. [11].
The branching fraction of each signal mode is reported in Table 1. The significance for the B 0 s → ppππ mode is less than 3 σ; an upper limit on its branching fraction is found to be B(B 0 s → ppππ) < 6.6 × 10 −7 at 90% confidence level, by integrating the likelihood after multiplying by a prior probability distribution that is uniform in the region of positive branching fraction. The values of the ratios of branching fractions between different B 0 (s) → pphh decay modes are reported in Table 2. The signal distributions in m(hh ) and m(pp) are obtained by subtracting the background using the sPlot technique [34], with the B 0 (s) candidate invariant mass as the discriminating variable. Per-candidate weights are applied to correct for the variation of the selection efficiency over the phase space. Figure 2 shows the hh invariant mass distributions of the B 0 → ppKπ, B 0 s → ppKK and B 0 → ppππ decay modes. A peak from a vector meson is identifiable in each mass spectrum, corresponding to a K * (892) 0 , a -- Table 2: Ratios of branching fractions among different B 0 (s) → pphh modes. The first uncertainty is statistical, the second systematic and the third, where present, comes from the uncertainty on φ(1020) and a ρ(770) 0 meson, respectively. The pp invariant mass distributions are also shown for the same decay modes. An enhancement near threshold, typical in baryonic B decays [3,4], is clearly visible in each case. Detailed amplitude analyses of the B 0 (s) → pphh decays will be of interest with larger samples.
The sources of systematic uncertainty on the absolute branching fractions and on the ratios of branching fractions arise from the fit model, the knowledge of the efficiencies and, where appropriate, from the uncertainties on the branching fraction of the normalization mode and on the ratio of b-quark hadronization probabilities. Pseudoexperiments are used to estimate the effect of using alternative shapes for the fit components, or of including additional components in the fit. In particular, the effect of adding other cross-feed backgrounds, partially reconstructed backgrounds and components coming from Λ 0 b decays have been investigated. These are the dominant sources of systematic uncertainty for the B 0 → ppKK and B 0 s → ppππ modes. The effect of fixing the yields of the cross-feed backgrounds based on the (mis)identification probabilities is also assessed by varying these probabilities within their uncertainties. Intrinsic biases in the fitted yields are investigated with pseudoexperiments and are found to be negligible. Uncertainties on the efficiencies arise due to the limited size of the simulation samples, the uncertainty on their evaluated distributions across the phase space of the decays and from possible residual differences between data and simulation. The unknown decay kinematics are the principal source of systematic uncertainty for the B 0 s → ppKπ mode, while for the B 0 s → ppKK, B 0 → ppKπ and B 0 → ppππ modes the dominant source of systematic uncertainty comes from the uncertainty on the efficiency of the hardware stage of the trigger. As the efficiencies depend on the signal decay-time distribution, the effect coming from the different lifetimes of the B 0 s mass eigenstates has been evaluated. The systematic uncertainties due to the vetoes of charm hadrons are also included.
In summary, a search for the four-body charmless baryonic decays B 0 (s) → pphh has been carried out by the LHCb collaboration with a sample of proton-proton collision data corresponding to an integrated luminosity of 3 fb −1 . First observations are obtained for the decays B 0 → ppππ, nonresonant B 0 → ppKπ, B 0 s → ppKK and B 0 s → ppKπ, while first evidence is reported for the B 0 → ppKK mode and an upper limit is set on the B 0 s → ppππ branching fraction. In particular, four-body baryonic B 0 s decays are observed for the first time and a threshold enhancement in the baryon-antibaryon mass spectra is confirmed for baryonic B 0 s decays [2]. The LHCb collaboration has recently published studies of CP violation with four-body Λ 0 b → ph − h + h − decays studying triple-product correlations, and presented first evidence for CP violation in baryons [36]. The decays of B 0 and B 0 s mesons to pphh final states reported in this paper may be used in the future for similar studies of CP violation in baryonic B decays.