Search for the lightest scalar top quark in events with two leptons in p-pbar collisions at sqrt(s) = 1.96 TeV

Data collected by the D0 detector at a p-pbar center-of-mass energy of 1.96 TeV at the Fermilab Tevatron Collider have been used to search for pair production of the lightest supersymmetric partner of the top quark decaying into $b \ell \tilde{\nu}$. The search is performed in the $\ell\ell' = e\mu$ and $\mu \mu$ final states. No evidence for this process has been found in data samples of approximately 400 pb^-1. The domain in the [$M(\tilde{t}_1),M(\tilde{\nu})$] plane excluded at the 95% C.L. is substantially extended by this search.

Data collected by the D0 detector at a pp center-of-mass energy of 1.96 TeV at the Fermilab Tevatron Collider have been used to search for pair production of the lightest supersymmetric partner of the top quark decaying into bℓν. The search is performed in the ℓℓ ′ = eµ and µµ final states. No evidence for this process has been found in data samples of approximately 400 pb −1 . The domain in the [M (t1), M (ν)] plane excluded at the 95% C.L. is substantially extended by this search.
PACS numbers: 14.80.Ly; 12.60.Jv Supersymmetric theories [1] predict the existence of a scalar partner for each standard model fermion. Because of the large mass of the standard model top quark, the mixing between its chiral supersymmetric partners is the largest among all squarks; therefore the lightest supersymmetric partner of the top quark,t 1 (stop), might be the lightest squark. If thet 1 → bℓν decay channel is kinematically accessible, it will be dominant [2] as long as thet 1 → bχ ± 1 andt 1 → tχ 0 1 channels are kinematically closed, whereχ ± 1 andχ 0 1 are the lightest chargino and neutralino, respectively. In this letter we present a search for stop pair production in pp collisions at 1.96 TeV with the D0 detector, where a virtual charginoχ ± decays into a lepton and a sneutrino, and where the sneutrinoν, considered to be the next lightest supersymmetric particle, decays into a neutrino and the lightest neutralinõ χ 0 1 ; in pp collisions, stop pairs are dominantly produced via the strong interaction in quark antiquark annihilation and gluon fusion. We use the Minimal Supersymmetric Standard Model (MSSM) as the phenomenological framework for this search. We assume the branching ratio Br(χ ± 1 → ℓν) = 1 with equal sharing among all lepton flavors, and we consider only cases where ℓ = e, µ. For stop pair production, we consider bb ℓℓ ′ ννχ 0 1χ 0 1 final states with ℓℓ ′ = e ± µ ∓ and ℓℓ ′ = µ + µ − (eµ and µµ channels); the signal topology consists of two isolated leptons, missing transverse energy ( / E T ), and jets. D0 has also searched for scalar top in the charm jet final state [3].
The D0 detector [4] comprises a central tracking system surrounded by a liquid-argon sampling calorimeter and a system of muon detectors. Charged particles are reconstructed using a multi-layer silicon detector and eight double layers of scintillating fibers in a 2 T magnetic field produced by a superconducting solenoid. The calorimeter provides hermetic coverage up to pseudo-rapidities |η| ≃ 4 (where η=-log(tan(θ/2), and where θ is the polar angle with respect to the proton beam direction) in a semi-projective tower geometry with longitudinal segmentation. After passing through the calorimeter, muons are detected in the muon detector comprising three layers of tracking detectors and scintillation counters located inside and outside of 1.8 T iron toroids. Events containing electrons or muons are selected for off-line analysis by a trigger system. A set of dilepton triggers is used to tag the presence of electrons and muons based on their energy deposit in the calorimeter, hits in the muon de-tectors, and tracks in the tracking system.
Three-body decays of thet 1 are simulated using comphep [5] and pythia [6] for generation and hadronization respectively. Standard model background processes are simulated using the pythia and alpgen [7] Monte Carlo (MC) generators. These MC samples are generated using the cteq5l [8] parton distribution functions (PDF); they are normalized using next-to-leading order cross sections [9]. All generated events are passed through the full simulation of the detector geometry and response based on geant [10]. MC events are then reconstructed and analyzed with the same programs as used for the data.
Muons are reconstructed by finding tracks pointing to hit patterns in the muon system. Non-isolated muons are rejected by requiring the sum of the transverse momenta (p T ) of tracks inside a cone with ∆R = (∆φ) 2 + (∆η) 2 = 0.5 (where φ is the azimuthal angle) around the muon direction, and the calorimeter energy in an annulus of size 0.1 < ∆R < 0.4 around the muon to be less than 4 GeV/c and 4 GeV. Isolated electrons are selected based on their characteristic energy deposition in the calorimeter, their fraction of deposited energy in the electromagnetic portion of the calorimeter and their transverse shower profile inside a cone of radius ∆R = 0.4 around the direction of the electron; furthermore, it is required that a track points to the energy deposition in the calorimeter and that its momentum and the calorimeter energy are consistent with the same electron energy; an "electron-likelihood" is defined as a variable combining information from the energy deposition in the calorimeter and the associated track. Backgrounds from jets and photon conversions are further suppressed by requiring the tracks associated with the muons and electrons to each have at least one hit in the silicon detector. Jets are reconstructed from the energy deposition in calorimeter towers using the Run II cone algorithm [11] with radius ∆R=0.5, and corrected for the jet energy scale (JES) [12]; in this search, jets are considered with p T > 15 GeV/c. The / E T is defined as the energy imbalance of all calorimeter cells in the plane transverse to the beam direction, and is corrected for the JES, the electromagnetic energy scale, and reconstructed muons. All efficiencies are measured with data [13].
In both eµ and µµ channels, the signal points [M (t 1 ), M (ν)] = (110,80) GeV/c 2 and (145,50) GeV/c 2 , respectively referred as "soft" (point A) and "hard" (point B) signals, have been used to optimize the selection of signals of different kinematics because of different ∆m = M (t 1 ) − M (ν). The choice of these points was also motivated by the sensitivity of the D0 search during Run I [14]. The main background processes imitating the signal topology are Z/γ * , W W , tt production, and multijet background. All but the latter are estimated with MC simulation. The multijet background is estimated from data. In the eµ channel, two samples each dominated by a different multijet background are obtained by inverting the muon isolation requirements, and by inverting the cut on the electron-likelihood; in the µµ channel, such a sample is obtained by selecting same-sign muon events. Factors normalizing each sample to the selection sample are also obtained from data, and applied to the background samples to obtain the multijet background estimation, this, at an early stage of the selection.
For the eµ channel, the integrated luminosity [15] of the data sample is (428 ± 28) pb −1 . The preselection is concluded by requiring the transverse momenta of the electron and muon (see Fig. 1 (a) and (b)) to be greater than 10 and 8 GeV/c, respectively. In this final state, the data are dominated by the multijet and Z/γ * → τ τ backgrounds. In these processes, poorly reconstructed leptons are correlated with / E T , giving rise to higher event populations at high and low values of the azimuthal angular difference between the leptons and the / E T , a low value of the angular difference for one lepton being correlated with a high value of the other. Taking  We require / E T to be greater than 15 GeV to reduce contribution of both the multijet and Z/γ * → τ τ backgrounds. To reject multijet events in which leptons are associated with a jet, we require the two leptons to be at a ∆R distance greater than 0.5 from any reconstructed jet. To further reduce the multijet contribution, we require the z component of the origin of the highest transverse momentum muon track to be within four standard deviations σ from the z component of the primary vertex: To reduce the Z/γ * → τ τ background, we cut on low values of the transverse mass of the muon and / E T (M T (µ, / E T ), see Fig. 1(c)). To further reduce this background, we make use of the correlation between the angular differences ∆φ(µ, / E T ) and ∆φ(e, / E T ), and require their sum (see Fig. 1 (d)) to be greater than 2.9: The contributions of different backgrounds, and the expected numbers of signal and observed data events in the eµ final state at different selection levels are summarized in Table I. After all selections, the W W (dominating the diboson contribution) and tt contributions are the dominant backgrounds. To separate soft signals such as point A from these backgrounds, we consider the variable S T defined as the scalar sum of the transverse momentum of the muon, the electron, and the / E T (see Fig. 1(e)). To separate hard signals such as point B from background contributions, we consider the variable H T defined as the scalar sum of the transverse momentum of all jets (see Fig. 1(f)). Rather than cutting on these two variables, the H T and S T spectra predicted for signal and background are compared with the observed spectra in twelve [S T , H T ] bins (see Table II) when extracting limits on the signal cross section, thus allowing a separation of signals of different kinematics from the W W and tt backgrounds.
For the µµ channel, the integrated luminosity [15] of the data sample is (395 ± 26) pb −1 . The selection of the signal in this final state is more challenging because of the strongly dominating Z/γ * → µµ background. The preselection is concluded by requiring the transverse momenta of the two highest transverse momenta oppositesign muons to be greater than 8 and 6 GeV/c. While the signal is characterized by the presence of jets originating from the hadronization of b quarks, the Z/γ * → µµ background owes the presence of jets to initial state radiation gluons which hadronize into softer jets, resulting in a lower multiplicity of jets; the latter is also valid for soft signals such as point A. To keep sensitivity to soft signals while rejecting substantial background, we require at least one jet: To further remove Z/γ * → µµ background events, where poorly reconstructed muons correlate with the / E T , we require the / E T to be greater than the contour shown on Fig. 2(a), using a cut parametrized by the following equation: where µ 1 is the highest transverse momentum muon. To augment the search sensitivity in this channel, we take advantage of the presence of jets originating from the fragmentation of long-lived b quarks in the signal. An algorithm based on the lifetime of hadrons calculates the probability P for the tracks of a jet to originate from the primary interaction point [16]. This b jet tagging probability is constructed such that its distribution is uniform for light-flavor jets while peaking at zero for heavy-flavor jets which have a vertex significantly displaced from the primary vertex ( Fig. 2(b)). Considering the highest transverse energy jet, we require P(jet) < 1%. (Dimu 3) A cut on the dimuon invariant mass (Fig. 2(c)) in the vicinity of the Z boson resonance only at low / E T (Fig.  2(d)) further suppresses the Z/γ * → µµ background while preserving the signal:  [17].
After applying all selection cuts for eµ and µµ data sets, no evidence fort 1 production is observed. We combine the number of expected signal and background events and their corresponding uncertainty, and the number of observed events in data from the twelve bins of the eµ selection (Table II) and the five bins of the µµ selection (Table IV) to calculate upper-limit cross sections for signal production at the 95% C.L. for various signal points using the modified frequentist approach [18]. In this calculation, correlated uncertainties are taken into account; no overlap is expected nor observed between the two samples. Regions for which the calculated cross section upper limit is smaller than the theoretical one are excluded at 95% C.L. Figure 3 shows the excluded region as a function of the scalar top quark and sneutrino masses, for nominal (solid line) and for both minimal and maximal (band surrounding the line) values of thet 1t1 production cross section; the latter variation corresponds to the PDF uncertainty for the signal cross section, quadratically added to the 2µ r and µ r /2 renormalization scale variations of thet 1t1 cross section. Although the numbers of expected and observed events are similar (Tables  I and III), their distribution across the bins (Tables II  and IV) causes the expected cross section limit to be lower than the observed one. For minimal values of the production cross section, the search in the eµ final state individually excludes a stop mass of 176 GeV/c 2 for a sneutrino mass of 60 GeV/c 2 , and a sneutrino mass of 97 GeV/c 2 for a stop mass of 130 GeV/c 2 ; the search in the µµ final state, once combined with the eµ final state, extends the final sensitivity by approximately 10 GeV/c 2 for small and large mass differences.
In summary, we have searched for the lightest scalar top quark decaying into bℓν; events with an electron and a muon, and two muons have been considered for this search. No evidence for the lightest stop is observed in these decays, leading to a 95% C.L. exclusion in the [M (t 1 ), M (ν)] plane. The largest stop mass excluded is 186 GeV/c 2 for a sneutrino mass of 71 GeV/c 2 , and the largest sneutrino mass excluded is 107 GeV/c 2 for a stop mass of 145 GeV/c 2 ; these mass limits are obtained with the most conservative theoretical production cross section, taking into account the PDF uncertainty and the variation of the renormalization scale. This is the most sensitive search for stop decaying into bℓν to date.