Search for top-squark pairs decaying into Higgs or Z bosons in pp collisions at √ s = 8 TeV

A search for supersymmetry through the direct pair production of top squarks, with Higgs (H) or Z bosons in the decay chain, is performed using a data sample of proton-proton collisions at √ s = 8 TeV collected in 2012 with the CMS detector at the LHC. The sample corresponds to an integrated luminosity of 19.5 fb − 1 . The search is performed using a selection of events containing leptons and bottom-quark jets. No evidence for a signiﬁcant excess of events over the standard model background prediction is observed. The results are interpreted in the context of simpliﬁed supersymmetric models with pair production of a heavier top-squark mass eigenstate (cid:101) t 2 decaying to a lighter top-squark mass eigenstate (cid:101) t 1 via either (cid:101) t 2 → H (cid:101) t 1 or (cid:101) t 2 → Z (cid:101) t 1 , followed in both cases by (cid:101) t 1 → t (cid:101) χ 01 , where (cid:101) χ 01 is an undetected, stable, lightest supersymmetric particle. The interpretation is performed in the region where the mass difference between the (cid:101) t 1 and (cid:101) χ 01 states is approximately equal to the top-quark mass ( m (cid:101) t 1 − m (cid:101) χ 01 (cid:39) m t ), which is not probed by searches for direct (cid:101) t 1 squark pair production. The analysis excludes top squarks with masses m (cid:101) t 2 < 575 GeV and m (cid:101) t 1 < 400 GeV at a 95% conﬁdence level.


Introduction
Supersymmetry (SUSY) with R-parity conservation [1] is an extension to the standard model (SM) that provides a candidate particle for dark matter and addresses the hierarchy problem [2][3][4][5][6][7]. The hierarchy problem originates in the spin-zero nature of the Higgs (H) boson, whose mass is subject to divergences from higher-order corrections. The leading divergent contribution from SM particles arises from the H boson coupling to the top quark. SUSY provides a possible means to stabilize the H boson mass calculation, through the addition of contributions from a scalar top quark (top-squark) with a mass not too different from that of the top quark [8][9][10][11][12]. Searches for direct top-squark production from the ATLAS [13][14][15][16][17][18] and Compact Muon Solenoid (CMS) [19] Collaborations at the Large Hadron Collider (LHC) at CERN have focused mainly on the simplest scenario, in which only the lighter top-squark mass eigenstate, t 1 , is accessible at current LHC collision energies. In these searches, the top-squark decay modes considered are those to a top quark and a neutralino, t 1 → t χ 0 1 → bW χ 0 1 , or to a bottom quark and a chargino, t 1 → b χ + 1 → bW χ 0 1 . These two decay modes are expected to have large branching fractions if kinematically allowed. The lightest neutralino, χ 0 1 , is the lightest SUSY particle (LSP) in the R-parity conserving models considered; the experimental signature of such a particle is missing transverse energy (E miss T ).
Searches for top-squark pair production are challenging because the cross section is approximately six times smaller than that for top-antitop quark pair (tt) production if m t 1 ∼ m t and decreases rapidly with increasing top-squark mass [20]. When the mass difference between the top-squark and the χ 0 1 is large, top-squark production can be distinguished from tt production, as the former is typically characterized by events with extreme kinematic features, especially large E miss T . This strategy is being pursued in existing searches and has sensitivity to top-squark masses up to about 650 GeV for low χ 0 1 masses [13][14][15][16][17][18][19]. The sensitivity of searches for direct top-squark pair production is, however, significantly reduced in the t 1 → t χ 0 1 decay mode for the region of SUSY parameter space in which m t 1 − m χ 0 1 m t . For example, in Ref. [19], the region |m t 1 − m χ 0 1 − m t | 20 GeV is unexplored. In this region, the momentum of the daughter neutralino in the rest frame of the decaying t 1 is small, and it is exactly zero in the limit m t 1 − m χ 0 1 = m t . As a result, the E miss T from the vector sum of the transverse momenta of the two neutralinos is typically also small in the laboratory frame. It then becomes difficult to distinguish kinematically between t 1 pair production and the dominant background, which arises from tt production. This region of phase space can be explored using events with topologies that are distinct from the tt background. An example is gluino pair production where each gluino decays to a top squark and a top quark, giving rise to a signature with four top quarks in the final state [21,22].  Figure 1: Diagrams for the production of the heavier top-squark ( t 2 ) pairs followed by the decays t 2 → H t 1 or t 2 → Z t 1 with t 1 → t χ 0 1 . The symbol * denotes charge conjugation.
states ( t 2 and t 1 ) due to the mixing of the SUSY partners t L and t R of the right-and left-handed top quarks. The t 2 decays to t 1 and a H or Z boson, and the t 1 is subsequently assumed to decay to t χ 0 1 , as shown in Fig. 1. Other decay modes such as t 1 → b χ + 1 → bW χ 0 1 are largely covered for m t 1 − m χ 0 1 m t by existing analyses [19]. The final states pursued in this search can arise in other scenarios, such as t 1 → t χ 0 2 , with χ 0 2 → H χ 0 1 or χ 0 2 → Z χ 0 1 . The analysis is also sensitive to a range of models in which the LSP is a gravitino [23,24]. The relative branching fractions for modes with the H and Z bosons are model dependent, so it is useful to search for both decay modes simultaneously. In the signal model considered, t 2 is assumed always to decay to t 1 in association with a H or Z boson, such that the sum of the two branching fractions is B( t 2 → H t 1 ) + B( t 2 → Z t 1 ) = 100%. Other possible decay modes are t 2 → t χ 0 1 and t 2 → b χ + 1 . These alternative decay modes are not considered here, since they give rise to final states that are covered by existing searches for direct top-squark pair production [13][14][15][16][17][18][19].
The results are based on proton-proton collision data collected at √ s = 8 TeV by the CMS experiment at the LHC during 2012, corresponding to an integrated luminosity of 19.5 fb −1 . The analysis presented here searches for t 2 production in a sample of events with charged leptons, denoted by (electrons or muons), and jets identified as originating from bottom quarks (b jets). The four main search channels contain either exactly one lepton, two leptons with opposite-sign (OS) charge and no other leptons, two leptons with same-sign (SS) charge and no other leptons, or at least three leptons (3 ). The channels with one lepton or two OS leptons require at least three b jets, while the channels with two SS leptons or 3 require at least one b jet. These requirements suppress background contributions from tt pair production, which has two b quarks and either one lepton or two OS leptons from the tt → νqqbb or tt → ν νbb decay modes, where q denotes a quark jet. The sensitivity to the signal arises both from events with additional b quarks in the final state (mainly from H → bb), and from events with additional leptons from H, Z, or W decays. This letter is organized as follows: Section 2 briefly introduces the CMS detector, while Section 3 presents the event samples and the object selections used. Section 4 describes the signal regions, and Section 5 details the background estimation methods. The experimental results are presented in Section 6, and in Section 7 we discuss the interpretation of the results in the context of the signal model of the pair production of a heavier top-squark mass eigenstate t 2 decaying to a lighter top-squark mass eigenstate t 1 . measured from the z axis pointing in the direction of the counterclockwise LHC beam. The pseudorapidity is defined as η ≡ − ln[tan(θ/2)].
The silicon tracker, the muon system, and the electromagnetic calorimeter cover the regions |η| < 2.4, |η| < 2.4, and |η| < 2.5, respectively. The hadronic calorimeters extend up to |η| ≈ 5, improving momentum balance measurements in the plane transverse to the beam direction. The online trigger system that selects collision events of interest is based on two stages: a firstlevel hardware-based selection and a second set of requirements implemented in software.

Event samples, object selection, and event simulation
The data used for this search were collected with a high transverse-momentum (p T ) electron (e) or muon (µ) single-lepton trigger, which requires at least one electron with p T > 27 GeV or muon with p T > 24 GeV. The trigger efficiencies, as measured with a sample of Z → + − events, vary between 85% and 97% for electrons, and between 80% and 95% for muons, depending on the η and p T values of the leptons. Events were also collected with the ee, eµ, and µµ double-lepton triggers, which require at least one e or one µ with p T > 17 GeV and another with p T > 8 GeV. Events are also acquired with a double-lepton trigger targeting lower-p T leptons, requiring p T > 8 GeV, but with an additional online selection of H T ≡ Σ jet |p jet T | > 175 GeV, considering only jets with p T > 40 GeV in the sum. The efficiencies lie between 90% and 95% for the trigger targeting lower-p T leptons, and between 80% and 95% for the trigger targeting higher-p T leptons, depending on the η and p T values of the lower-p T lepton. For selections with more than two leptons, the triggers are fully efficient.
Events are reconstructed offline using the particle-flow (PF) algorithm [26,27]. Electron candidates are reconstructed by associating tracks with energy clusters in the electromagnetic calorimeter [28,29]. Muon candidates are reconstructed by combining information from the tracker and the muon detectors [30]. Signal leptons are produced in the decays of W and Z bosons. In order to distinguish these leptons from those produced in the decays of heavy-flavor hadrons, all lepton candidates are required to be consistent with originating from the primary interaction vertex, chosen as the vertex with the highest sum of the p 2 T of its constituent tracks. In particular they are required to have a transverse impact parameter with respect to this vertex smaller than 0.2 mm. A tighter requirement is used for the event category with two SS leptons (see Ref.
[31]). Furthermore, since misidentified lepton candidates arising from background sources, such as the decays of hadrons, are typically embedded in jets, all lepton candidates are required to be isolated from hadronic activity in the event. This is achieved by imposing a maximum allowed value on the quantity p sum T , defined as the scalar sum of the p T values of charged and neutral hadrons and photons within a cone of radius ∆R ≡ √ (∆η) 2 + (∆φ) 2 = 0.3 around the lepton candidate momentum direction at the origin. For the event category with at least three leptons, the isolation requirement is p sum T < 0.15p T . For the lower lepton-multiplicity selections, the isolation requirement is tighter (see Refs. [19] and [31] for details). The surrounding hadronic activity is corrected for the energy contribution from additional proton-proton interactions in the event (pileup), as described in Ref. [32]. Jets are reconstructed from particle-flow candidates using the anti-k T clustering algorithm [33] with a distance parameter of 0.5. Their energies are corrected for residual non-uniformity and non-linearity of the detector response using corrections derived from exclusive dijet and γ/Z+jet data [34]. The energy contribution from pileup is estimated using the jet area method for each event [35] and is subtracted from the jet p T . Only high-p T jets in the central calorimeter |η| < 2.4 are considered. Jets consistent with the decay of heavy-flavor hadrons are identified using the combined secondary vertex b-tagging algorithm at the medium or loose working points, defined such that they have tagging efficiencies of 70% or 80-85%, and misidentification rates for light-flavor jets less than 2% or 10%, respectively [36]. The E miss T is calculated as the magnitude of the vector sum of the transverse momenta of all PF candidates, incorporating jet energy corrections [37]. Quality requirements are applied to remove a small fraction of events in which detector effects such as electronic noise can affect the E miss T reconstruction. Events are required to have E miss T > 50 GeV to reduce background contributions from sources with a single W boson and from jet production via QCD processes.
Simulated event samples are used to study the characteristics of the signal and to calculate its acceptance, as well as for part of the SM background estimation. Pair production of t 2 squarks is described by the MADGRAPH 5.1.3.30 [38] program, including up to two additional partons at the matrix element level, which are matched to the parton showering from the PYTHIA 6.424 [39] program. The SUSY particle decays are simulated with PYTHIA with a uniform amplitude over phase space, so that all decays are isotropic [40]. The first two decay modes considered (see Fig. 1) are assumed to have a branching fraction of unity when setting limits on SUSY particle masses. The Higgs boson mass is set to 125 GeV [41], and its branching fractions are set according to the corresponding expectations from the SM [42]. For each decay mode, a grid of signal events is generated as a function of the two top-squark masses m t 2 and m t 1 . The t 1 squark is forced to decay to a top quark and a neutralino LSP assuming m t 1 − m χ 0 1 = 175 GeV. The topquark mass is set to 175 GeV. The signal event rates are normalized to cross sections calculated at next-to-leading order (NLO) in the strong coupling constant, including the resummation of soft-gluon emission at next-to-leading-logarithmic accuracy (NLO+NLL) [43][44][45][46][47][48].
The SM background processes considered are the production of tt; tt in association with a boson (H, W, Z, γ * ); W, Z, and γ * +jets; triboson; diboson; single-top quark in the s, t, and tW channels; and single-top quark in association with an additional quark and a Z boson. These processes are generated with the MADGRAPH, POWHEG-box 1.0 [49,50], or MC@NLO 2.0.0 beta3 [51,52] programs, using the CT10 [53] (POWHEG), CTEQ6M [54] (MC@NLO), and CTEQ6L1 [54] (MADGRAPH) parton distribution functions (PDFs). SM background event rates are normalized to cross sections [51,52,[55][56][57][58][59][60] calculated at next-to-next-to-leading order when available, otherwise at NLO. All the background samples are processed with the full simulation of the CMS detector based on GEANT4 [61], while the generated signal samples use a fast simulation [62]. The fast simulation is validated against the full simulation for the variables relevant for this search, and efficiency corrections based on data are applied [63]. The simulation is generated with inelastic collisions superimposed on the hard-scattering event. Events are weighted so that the distribution of the number of inelastic collisions per bunch crossing matches that in data.

Event categories and signal regions
The search is carried out through comparisons of the data and SM background yields in disjoint signal regions (SRs) targeting the SUSY processes shown in Fig. 1, while suppressing the contributions from SM backgrounds, predominantly tt production. The definitions of the SRs are summarized in Table 1, and are detailed in the following subsections. Events are classified according to the lepton multiplicity and charge requirements on the leptons. Four main event categories are considered. The first two include events with one lepton or two OS leptons. Since these lepton signatures also arise in the decays of top-antitop quark pairs, requirements of at least three b jets are used to suppress this background. The other two categories are events with exactly two SS leptons and events with three or more leptons, which do not typically arise in tt events. A requirement of at least one b jet is applied to further suppress the contribution from backgrounds from W and Z bosons. Lepton vetoes are used to ensure that the four main event categories do not overlap.

Event categories with a single lepton or two opposite-sign leptons
The event categories with one lepton or two OS leptons, accompanied in either case by at least three b jets, target signatures with H bosons, which have large branching fraction for H → bb. In the single-lepton channel, events are required to have exactly one electron with p T > 30 GeV and |η| < 1.44 or exactly one muon with p T > 25 GeV and |η| < 2.1. Events with an indication of an additional lepton, either an isolated track [31] or a hadronically decaying τ-lepton candidate τ h [64][65][66], are rejected in order to reduce the background from tt events in which both W bosons decay leptonically. In the double-lepton channel, events are required to contain exactly two charged leptons (ee, eµ, or µµ), each with p T > 20 GeV and |η| < 2.4. In this case, events with an additional e or µ with p T > 10 GeV are rejected. Any electron candidate in the region 1.44 < |η| < 1.57, a less well instrumented transition region between the barrel and endcap regions of the calorimeter, is excluded in the event selection since standard electron identification capabilities are not optimal. Jets are required to be separated from the candidate leptons by ∆R > 0.4.
In these event categories, a typical tt background event has two b jets in the final state, while signal events could have up to four additional b jets, two from each H decay. The requirement of more than two b jets greatly suppresses the tt background contribution. For events with exactly three b jets, the jet p T threshold applied is 40 GeV; for events with at least four b jets, the threshold is lowered to 30 GeV. In both cases, the medium working point of the b-jet tagger is used (see Section 3). To further reduce the tt background contribution in the sample with exactly three b jets, events are required to contain two additional jets with p T > 30 GeV, at least one of which must satisfy the loose but not the medium criteria of the b-jet tagger. Signal events can have large jet and b-jet multiplicities, while in the case of the tt background, additional jets are needed to satisfy this selection criterion. To reduce the contribution of jets from pileup in the event, a requirement is applied on a multivariate discriminating variable that incorporates the multiplicity of objects clustered in the jet, the jet shape, and the compatibility of the charged constituents of the jet with the primary interaction vertex [67].
Besides the requirements listed above, the analysis in the single-lepton channel selects events with large transverse mass of the ( , ν) system, defined as m T ≡ , where the p T of the selected lepton is used and the (x, y) components of the neutrino momen-tum are equated to the corresponding E miss T components. For events in which the E miss T arises from a single neutrino from a W boson decay, this variable has a kinematic endpoint m T ≈ m W , where m W is the W boson mass. The requirement of large m T (m T > 150 GeV for events with three b jets or m T > 120 GeV for events with at least four b jets) provides strong suppression of the semileptonic tt background.
The study of the OS dilepton channel uses information from pairs of b jets (ignoring their charge) to identify pairs consistent with H → bb decay: , with p z denoting the component of the momentum along the beam axis. Only b jets satisfying the medium working point of the tagger are used to form bb combinations. We denote the number of selected b-jet pairs as N bb and the invariant mass of a pair as m bb . Events are required to have either N bb = 1 and 100 ≤ m bb ≤ 150 GeV, or else N bb ≥ 2. For the signal models of interest, particularly the t 2 → H t 1 decay mode, the SRs with largest b-jet multiplicity (≥4 b jets) have the highest sensitivity.

Event category with two SS leptons
The event category with two SS leptons targets signatures with multiple sources of leptons. Standard model processes with two SS leptons are extremely rare. The analysis for this event category closely follows that described in Ref.
[31]. The only difference is the addition of a veto on events containing a third lepton, to remove the overlap with the 3 event category. These SRs also recover events with three leptons in which one of the three leptons falls outside the detector acceptance or fails the selection criteria. Multiple SRs are defined for the SS event category based on the jet and and b-jet multiplicities, E miss T , and H T , and on whether the leptons satisfy p T > 10 GeV (low-p T analysis) or p T > 20 GeV (high-p T analysis). The leptons must appear within |η| < 2.4. Also, a baseline selection is introduced, which requires at least two jets, moderate H T (>250 GeV in the low-p T analysis and >80 GeV in the high-p T analysis), and moderate E miss T (>30 GeV for events with H T < 500 GeV; otherwise, there is no E miss T requirement).

Event category with at least three leptons
The event category with at least three leptons and at least one b jet is sensitive to all of the processes shown in Fig. 1. These processes contain many sources of leptons, such as Z bosons from the top-squark decays, and τ leptons and W and Z bosons from the H boson decays. Even though signatures giving rise to three or more leptons have small production rates, this event category has good sensitivity because the backgrounds are strongly suppressed. The dataset is acquired using the double-lepton triggers. Events are selected offline by requiring at least three e or µ candidates with p T > 10 GeV, including at least one with p T > 20 GeV, and |η| < 2.4. Events with two leptons of opposite-sign charge with an invariant mass below 12 GeV are removed from the sample to reduce the contribution of leptons originating from low-mass bound states.
Events are required to have at least two jets with p T > 30 GeV and at least one b jet satisfying the medium working point of the tagger. Leptons within ∆R < 0.4 of a b-quark jet are not considered isolated and are merged with the b jet. This requirement imposes an additional isolation criterion for leptons and reduces the dominant background, tt production, by 25-40% depending on the SR, compared to the case where such an object is reconstructed as a lepton rather than a b jet. The efficiency for signal leptons is reduced by 1%. The remaining SM background in the ≥3 event category from WZ+jets production is highly suppressed by the b-jet requirement.
This three-lepton event sample is divided into several SRs by imposing requirements on the jet and b jet multiplicity, E miss T , and the hadronic activity in the event, as given by the kinematic variable H T , considering jets with p T > 30 GeV in the sum. Finally, events are classified as either "on-Z" if there is a pair of leptons with the same flavor and opposite charge that has an invariant mass within 15 GeV of the nominal Z boson mass, or "off-Z" if no such pair exists or if the invariant mass lies outside this range.
The separation of events into these SRs improves the sensitivity of the search. For the signal models of interest, the SRs with large b jet multiplicity (those designated N b jets = 2 and N b jets ≥3) and that with both high E miss T and high H T provide the greatest sensitivity. The on-Z regions are the most sensitive, when the decay to an on shell Z boson is kinematically allowed for the t 2 → Z t 1 decay mode. Conversely, the off-Z regions have more sensitivity when off-shell Z boson decays are kinematically allowed and for the t 2 → H t 1 decay mode.

Background estimation
The main background arises from SM tt events, which usually have two b jets and at most two leptons from W boson decays. Thus, tt events can only satisfy the selection criteria if accompanied by sources of additional b jets or leptons. Such backgrounds are estimated using control samples in data, as described below. This method greatly reduces the dependence of the background prediction on the accurate modeling in simulation, the knowledge of the inclusive tt production cross-section, the measurement of the integrated luminosity, and the accuracy of the object-selection efficiency determination.
Additional backgrounds arise from processes involving one or more W and Z bosons, although these contributions are suppressed by the b-jet requirements. Finally, all event categories have backgrounds from rare SM processes, such as ttZ and ttW production, whose cross sections have not been precisely measured [68]. The prediction for these contributions is derived from simulation, and a systematic uncertainty of 50% is assigned to account for the uncertainty in the NLO calculations of their differential cross sections. The remainder of this section describes the background predictions for each of the specific event categories.

Backgrounds in event categories with a single lepton or two OS leptons
For the single-lepton or two-OS-lepton event categories, the dominant background is from tt events (85-95% of the total). These events can have three or more b jets if the tt pair is accompanied by additional jets that may be mistagged in the case of light-parton jets or that may contain genuine b jets from gluon decays to a bb pairs. In the case of semileptonic tt events, there are small additional contributions from W → cs decays, with a charm-quark jet misidentified as a b jet, and from the rare W → cb decay mode. In the case of dileptonic tt events, τ leptons from the second W boson decay that are misidentified as b jets also contribute. Scale factors, defined as the ratio of the yield in data to the yield in simulation, are used to normalize the background predictions from simulation. For each SR, the corresponding scale factor is derived from a control region enhanced in background tt events. These control regions are defined by 50 ≤ m T ≤ 100 GeV for the single-lepton selections and by either N bb = 0 or N bb = 1, with either m bb ≤ 100 GeV or m bb ≥ 150 GeV, for the OS-dilepton case. The contribution from non-tt events is evaluated from simulation and subtracted from the data before deriving the normalization. To reduce the contribution from a possible signal in these control regions, the samples are restricted to events with low jet multiplicity: for the three-b-jet category, only events with exactly five jets are used, and for the category with four b jets, only events with exactly four jets are used. The dominant source of uncertainty for the background prediction arises from the 8 5 Background estimation limited number of events in the control sample (15-35% on the total background). The tt background prediction also depends on the ratio of events in the signal and control regions, which is evaluated from simulation and validated using tt-dominated control samples obtained by selecting events with fewer than three b jets.
In the single-lepton channel, the modeling of the high-m T tail is critical for the background estimation. Genuine semileptonic tt events have an endpoint at m T ≈ m W , with E miss T resolution effects primarily responsible for populating the m T > m W tail. The effect of E miss T resolution on the m T tails is investigated by selecting events with one or two b jets and by varying the number of additional jets. The comparison of simulation with data in the m T tail region is used to extract scale factors and uncertainties for the semileptonic tt prediction. The scale factors are in the range 1.1-1.2, depending on the m T requirement, with corresponding uncertainties of 5-10%. The semileptonic background contributes 50-60% of the total background in the single-lepton SRs. Events from genuine dileptonic tt events can also satisfy the single-lepton event selection if the second lepton is not identified or is not isolated and can give rise to large values of E miss T and m T due to the presence of two neutrinos. This tt → + jets contribution constitutes ∼30-40% of the total background and is derived from simulation, with scale factors consistent with unity, as determined from comparison of data with simulation in the dilepton control regions.
In the channels with two OS leptons, the most important issues for the background prediction are related to the construction of b-jet pairs (see Section 4.1 for the full list of requirements). Modeling of the emission of additional radiation leading to jets and gluon splitting to bb pairs, and of effects such as τ-lepton mistagging, c-quark-jet mistagging, and b-jet identification efficiency, can affect the m bb variable. The modeling of these effects is validated using the statistically precise single-lepton control sample with 50 ≤ m T ≤ 100 GeV, in which the m bb distributions in data and simulation are compared as a function of the b-jet multiplicity. These studies of the ratio of events in the signal and control regions are used to derive scale factors, which are found to be consistent with unity, and uncertainties, corresponding to 20-30% of the total background uncertainty.

Backgrounds in the event category with two SS leptons
For the SRs with two SS leptons, the background estimates and uncertainties are derived following the procedures described in Ref. [31]. There are three main categories of backgrounds. Non-prompt leptons are produced from heavy-flavor decays, misidentified hadrons, muons from the decay-in-flight of light mesons, and electrons from unidentified photon conversions. Charge misidentification arises mainly from electrons that undergo severe bremsstrahlung in the tracker material, leading to a misreconstruction of the charge sign. Finally, rare SM processes yielding two genuine SS leptons (typically a tt pair in association with a H, W, or Z boson) can contribute significantly, especially in SRs with tight selection requirements. Backgrounds from non-prompt leptons and rare SM processes dominate, each contributing 20-80% of the total, while charge misidentification contributes 1-5%.
The background from non-prompt leptons is evaluated with a sample in data where at least one lepton fails the full identification criteria but satisfies a combination of relaxed isolation and lepton identification requirements. The number of events in this sample is scaled by the probability of a loosely identified lepton to satisfy the full set of selection requirements, which is determined from an independent sample of events. The charge misidentification background is obtained using a sample of OS ee and eµ events that satisfy the full kinematic selection weighted by the p T -and η-dependent probability of electron charge misassignment. The sys-tematic uncertainty of the total background prediction is dominated by the uncertainties from rare SM processes and from events with a jet misidentified as a prompt lepton (30-50% of the total background).

Backgrounds in the event category with at least three leptons
For SRs with at least three leptons, there are two main types of backgrounds. In the off-Z SRs, the background with two prompt leptons and an additional object misidentified as a prompt lepton dominates, comprising 50-90% of the total. In the on-Z SRs, the dominant background is typically from SM processes with at least three genuine prompt leptons, corresponding to 60-100% of the total.
The background sources with two prompt leptons from W or Z boson decay and a third object misidentified as a prompt lepton are predominantly from tt production, although the Z + jets and WW + jets processes also contribute. The procedure to estimate this background contribution follows closely that used for the analysis of events with two SS leptons [31]. The probability for a lepton from the decay of a bottom hadron to satisfy the analysis selection criteria is measured in a data sample enriched in QCD dijet events. This probability is applied to a sample of ≥ 3 events, in which the isolation requirement on one of the leptons is removed, providing an estimate of the background contribution from non-prompt leptons. A systematic uncertainty of 30% is derived for this background based on studies of the method in simulation. This uncertainty accounts for the difference in the p T spectrum of b jets in the control sample, where the probability is measured, compared to the spectrum in the signal sample, where it is applied. This systematic uncertainty dominates the uncertainty in the background prediction in the SRs with looser kinematic requirements. SRs with tight kinematic requirements also have a significant statistical uncertainty due to the size of the sample used to derive this background estimate. These are the dominant sources of uncertainty in the backgrounds in the off-Z signal regions, corresponding to 20-90% uncertainty on the total background. The background contribution from events with two vector bosons that produce three genuine prompt isolated leptons, mainly WZ + jets and ZZ + jets events, is estimated from simulation and is validated by comparing data and simulation in control samples in which the full selection is applied and the b-jet requirement is inverted. A control sample enhanced in the WZ background is obtained by selecting events with three high-p T leptons. One pair of leptons is required to form a Z → + − candidate. The third lepton is combined with the E miss T vector, and this system is required to form a W boson candidate (50 < E miss T < 100 GeV and 50 < m T < 120 GeV). A second control sample, enhanced in the ZZ background, is obtained by selecting events with four leptons and E miss T < 50 GeV. Two leptons are required to form a Z candidate. A scale factor of 0.9 is derived for the ZZ background prediction based on the comparison of data and simulation in this sample. The systematic uncertainty for the diboson background is derived based on these comparisons, which are limited by the statistical precision of the control samples. A 50% uncertainty is assigned to account for possible mismodeling of additional partons required to satisfy the b-jet requirement.

Results
The results of the search are shown in Tables 2-4 Table 2.
as a control sample to derive the normalization for this background contribution. As shown in Fig. 2 (top), the backgrounds in the SR are mainly semileptonic and dileptonic tt events.
For the SRs with two OS leptons, Fig. 2 (bottom) shows a comparison of the m bb distribution in data and simulation. The sample in the region outside the m bb signal window is used to derive the normalization for the tt → + jets background prediction for events with three b jets. In the case of events with at least four b jets, multiple bb pairs are possible. The control region is not indicated in Fig. 2 (bottom right) since the m bb requirement is not applied when N bb ≥ 2.
The dominant background in the SRs is from tt → + jets events. The results for the SRs with one lepton or two OS leptons are summarized in Table 2. The predicted and observed yields agree within 1.4 standard deviations of local significance [69], given the statistical uncertainty in the predicted yields.  For the baseline event selection with two SS leptons, Fig. 3 shows a comparison of data and the predicted backgrounds for three selection observables: the jet multiplicity, the b-jet multiplicity, and E miss T . The amount of background varies strongly among the signal regions; some of them including tens of background events while others have essentially none. The relative contribution from rare SM processes increases as the requirements are tightened. As shown in Table 3, the SM background predictions and observations are in agreement for both the high-p T and low-p T selections.
Finally, for the event sample with at least three leptons, Fig. 4 shows a comparison of data and the predicted backgrounds for the jet and b-jet multiplicities and for the E miss T distribution. The dominant background is from processes with two prompt leptons and additional nonprompt leptons, mainly due to tt events, although in the case of the on-Z selection, background sources with Z bosons also contribute significantly. The results of the search, summarized in Table 3: SS dilepton event category: predicted total background and observed data yields as a function of the jet multiplicity, b-jet multiplicity, E miss T , and H T requirements, for the low-p T and high-p T regions. The uncertainties in the total background predictions contain the statistical and systematic components.    Table 4, demonstrate agreement between background predictions and observations for all the SRs considered.
In summary, the data yields are found to be consistent with the background predictions across all event categories and SRs. Of the 96 SRs, the largest discrepancy corresponds to a 1.6 standard deviation excess of local significance (30 events compared to 16 ± 5 expected, see Table 4), computed following the recommendations of Ref. [69]. Thus, no indication of top-squark pair production is observed.

Interpretation
The results are used to set upper limits on the cross section times branching fraction for pair production of t 2 squarks for the decay modes shown in Fig. 1. The upper limits are calculated at a 95% confidence level (CL) using the LHC-style CL S method [70][71][72]. The exclusion curves on particle masses at 95% CL are evaluated from a comparison of the cross section upper limits and the theoretical signal cross section predictions. As explained below, the results from the various SRs are combined in the limit-setting procedure in order to improve the sensitivity of the search.
The limit calculation on the cross section times branching fraction depends on the signal selection efficiency and the background estimates. The SRs with at least three leptons have the highest expected sensitivity because of the small level of SM background. For SRs with at least three leptons, the off-Z SRs with H T > 200 GeV are used for the ttHH interpretation, while both the off-Z and on-Z SRs with H T > 200 GeV are used for the ttZZ interpretation. The total signal acceptance for all SRs with at least three leptons varies from around 0.4-0.5% for the ttHH signal, to 1.2-1.5% for the ttZZ signal. The acceptance for the most sensitive SR alone is around ∼0.1% for ttHH and approximately three times larger for ttZZ. This difference in acceptance is due to the larger leptonic branching fraction for Z boson decays compared to H boson decays. The SRs with lower lepton multiplicities also have sensitivity to the ttHH signal. All SRs of the high-p T SS dilepton analysis are used in the limit setting. While only the high-p T results are used in the interpretation presented in this letter, the low-p T experimental results are included in Table 3 for potential use in future interpretations. In SRs with two SS leptons, the overall acceptance for ttHH events is 0.3-0.5%, where the most sensitive signal regions contribute ∼0.15%. In the case of SRs with one lepton or two OS leptons, the acceptance for ttHH events is approximately 0.2-0.4%. The acceptances for the single-lepton and dilepton final states are slightly lower for the ttZZ signal. Because of the large branching fraction for the H → bb decay mode, SRs with higher b-jet multiplicity requirements dominate the expected sensitivity for scenarios with H bosons. SRs with low b-jet multiplicities are most sensitive for scenarios with Z bosons.
The systematic uncertainties, listed in Table 5, are evaluated for the signal selection efficiency in every SR and for every signal point separately. The total uncertainty in the signal selection efficiency is in the 9-30% range. The dominant source of uncertainty depends on the SR and decay mode considered. An important source of uncertainty arises from the estimation of the trigger and lepton identification efficiencies, which are derived using Z → + − samples and contribute 6-13%. The uncertainty due the knowledge of the energy scale of hadronic jets increases with tighter kinematic requirements and corresponds to an uncertainty of 1-15%. The uncertainty due the knowledge of the b jet identification performance depends on the event properties, such as the jet flavor and p T value, and gives rise to an uncertainty of 2-20%. For smaller differences between the m t 2 and m t 1 mass values, uncertainties in the modeling of initial-state radiation (ISR) become important. The uncertainty related to the PDFs on the acceptance is determined using the PDF4LHC recommendations [73] and contributes 2-5%. The corresponding uncertainty in the signal selection efficiency is of 3-15%, increasing for smaller m t 2 -m t 1 mass differences. The systematic uncertainties, including their correlations, are treated consistently in the different analyses. The correlations between the different analyses have a small impact on the combined result.  Figure 5 (left) shows the 95% CL upper limits on the cross section times branching fraction in the m t 1 versus m t 2 plane for the (a) t 2 → H t 1 and (b) t 2 → Z t 1 decay modes. The contour bounds the excluded region in the plane assuming the NLO+NLL cross section calculation in the decoupling limit for all the SUSY sparticles not included in the model. The results are presented assuming a branching fraction of 100% to each decay mode. The 95% CL expected (thick dashed red) and observed (solid black) limits are obtained including all uncertainties with the exception of the theoretical uncertainty in the signal production cross section. The expected limit is defined as the median of the upper-limit distribution obtained using pseudoexperiments and the likelihood model considered. The bands around the expected limit correspond to the impact of experimental uncertainties, and the bands around the observed limit indicate the change for a ±1 standard deviation (σ) variation in the theoretical cross section (mainly due to uncertainties in the renormalization/factorization scales and in the knowledge of the PDFs). In the t 2 → H t 1 decay mode, taking a −1σ theory lower bound on signal cross sections, a t 2 squark with m t 2 525 GeV is excluded at a 95% CL for t 1 squarks with m t 1 300 GeV. Similarly, in the t 2 → Z t 1 decay mode, a t 2 squark with m t 2 575 GeV is excluded at 95% CL for t 1 squark with m t 1 400 GeV.
For the pure t 2 → H t 1 decay (Fig. 5 upper right), the SRs with at least three leptons, no Z → + − candidates, and large b-jet multiplicities are the most sensitive. Nevertheless, the SRs with lower lepton multiplicities (one lepton or two leptons) have significant expected sensitivity in the t 2 → H t 1 decay mode. Including the final states with lower lepton multiplicities in the combination lowers the cross section upper limit results by 15-20% compared to the threelepton results alone. Therefore, all lepton multiplicity categories are used in the interpretation of the ttHH signal.
In the case of the signals with Z bosons (Fig. 5 lower right), the SRs with at least three leptons completely dominate the expected sensitivity. The different SRs with at least three leptons provide sensitivity to different types of signals. In particular, off-Z SRs are sensitive to the region of parameter space in which the Z bosons are off-shell, m t 2 − m t 1 < m Z , while the on-Z regions provide sensitivity to signals with larger mass differences. Only the SRs with at least three leptons are used in the interpretation of the ttZZ signal.
Mixed-decay scenarios, with non-zero branching fractions for the Z and H decay modes, are also considered, assuming these to be the only decay modes possible. Figure 6 shows the corresponding limits as a function of the relative branching fraction of the Z and H decay modes. The scenario with the least expected sensitivity is where the H boson decay mode dominates, while the best expected sensitivity is achieved when the Z boson decay mode dominates.
The cross section upper limits are obtained neglecting the contribution of direct t 1 squark pair production, which can satisfy the selection criteria for the single-lepton or OS-lepton SRs if a light-parton jet is misidentified as a b jet or if there is additional radiation leading to genuine b jets. Including direct t 1 squark pair production in the single-lepton or two OS lepton SRs typically lowers the cross section limit by a few percent, with the most pronounced differences occurring at larger t 2 mass. The contribution in the case of events with two SS leptons or at least three leptons is small due to the low probability of misidentifying non-prompt leptons. Since the signature with three leptons has the best sensitivity overall, the impact on the combined limit is much smaller than the uncertainty in the production cross section.

Summary
This letter presents results of a search for the pair production of the heavier top-squark mass eigenstate t 2 decaying to the lighter eigenstate t 1 , producing a signature of a top-antitop quark pair in association with Higgs or Z bosons. The analysis explores final states with exactly one lepton and at least three identified bottom-quark jets (b jets), with exactly two leptons of opposite charge and at least three b jets, with exactly two same-sign leptons and at least one b jet, and with three or more leptons and at least one b jet, where by "lepton" we mean an electron or muon. No significant excess event yield above standard model expectations is observed. The results are used to exclude a range of t 2 masses below approximately 575 GeV for t 1 masses below approximately 400 GeV. The interpretation assumes that the t 1 squark always decays to t χ 0 1 and that m t 1 − m χ 0 1 m t , where the χ 0 1 particle represents a stable, weakly interacting lightest supersymmetric particle neutralino whose signature in the detector is missing transverse energy. This region of phase space is not probed by searches for direct t 1 squark pair = 175 GeV. The shaded maps (plots on the left) show the upper limit (95% CL) on the cross section times branching fraction at each point in the m t 1 vs. m t 2 plane for the process pp → t 2 t * 2 , with t 2 → H t 1 , t 1 → t χ 0 1 (upper plots) and t 2 → Z t 1 , t 1 → t χ 0 1 (lower plots). In these plots, the results from all channels are combined. The excluded region in the m t 1 vs. m t 2 parameter space is obtained by comparing the cross section times branching fraction upper limit at each model point with the corresponding NLO+NLL cross section for the process, assuming that (a) B( t 2 → H t 1 ) = 100% or (b) that B( t 2 → Z t 1 ) = 100%. The solid (dashed) curves define the boundary of the observed (expected) excluded region. The ±1 standard deviation (σ) bands are indicated by the finer contours. The figures on the right show the observed (expected) exclusion contours, which are indicated by the solid (dashed) curves for the contributing channels. As indicated in the legends of the righthand figures, the thinner curves show the results from each of the contributing channels, while the thicker curve shows their combination. The four event categories for the t 2 → H t 1 study are shown in the upper plots, while the on-Z and off-Z categories for events with at least three leptons are shown in the lower plots.  Figure 6: Upper limits on the cross section for t 2 pair production for different branching fractions of t 2 → H t 1 and t 2 → Z t 1 , assuming that B( t 2 → H t 1 ) + B( t 2 → Z t 1 ) = 100%. The t 1 squark is assumed to always decay to a top quark and a neutralino χ 0 1 with m t 1 − m χ 0 1 = m t .
The decay t 2 → H t 1 is only considered when the H boson production is kinematically allowed, m t 2 − m t 1 > m H . production.
fice; the Fonds pour la Formationà la Recherche dans l'Industrie et dans l'Agriculture (FRIA-Belgium); the Agentschap voor Innovatie door Wetenschap en Technologie (IWT-Belgium); the Ministry of Education, Youth and Sports (MEYS) of the Czech Republic; the Council of Science and Industrial Research, India; the Compagnia di San Paolo (Torino); the HOMING PLUS programme of Foundation for Polish Science, cofinanced by EU, Regional Development Fund; and the Thalis and Aristeia programmes cofinanced by EU-ESF and the Greek NSRF.