PHYSICAL REVIEW D 99, 092004 (2019) Search for heavy particles decaying into a top-quark pffiffi pair in the fully hadronic final state in pp collisions at s = 13 TeV with the ATLAS detector M. Aaboud et al.* (ATLAS Collaboration) (Received 27 February 2019; published 14 May 2019) A search for new particles decaying into a pair of top quarks is performed using proton-proton collision data recorded with the ATLAS detector at the Large Hadron Collider at a center-of-mass pffiffiffi energy of s ¼ 13 TeV corresponding to an integrated luminosity of 36.1 fb−1 . Events consistent with top-quark pair production and the fully hadronic decay mode of the top quarks are selected by requiring multiple high transverse momentum jets including those containing b-hadrons. Two analysis techniques, exploiting dedicated top-quark pair reconstruction in different kinematic regimes, are used to optimize the search sensitivity to new hypothetical particles over a wide mass range. The invariant mass distribution of the two reconstructed top-quark candidates is examined for resonant production of new particles with various spins and decay widths. No significant deviation from the Standard Model prediction is observed and limits are set on the production cross-section times branching fraction for new hypothetical Z0 bosons, dark-matter mediators, Kaluza-Klein gravitons and Kaluza- Klein gluons. By comparing with the predicted production cross sections, the Z0 boson in the topcolor- assisted-technicolor model is excluded for masses up to 3.1–3.6 TeV, the dark-matter mediators in a simplified framework are excluded in the mass ranges from 0.8 to 0.9 TeV and from 2.0 to 2.2 TeV, and the Kaluza-Klein gluon is excluded for masses up to 3.4 TeV, depending on the decay widths of the particles. DOI: 10.1103/PhysRevD.99.092004 I. INTRODUCTION events containing tt̄ pairs, where both the top and antitop The Large Hadron Collider quarks decay hadronically (tt̄ → W þ bW − b̄ with W → qq̄0 ), pffiffiffi(LHC), currently operating at is presented. The analysis is based on 36.1 fb−1 of proton- a center-of-mass energy of s ¼ 13 TeV, has the potential pffiffiffi to discover phenomena beyond the Standard Model (SM) at proton collision data at a center-of-mass energy of s ¼ the TeV scale. The heaviest elementary particle known in 13 TeV recorded with the ATLAS detector at the LHC in the SM, the top quark, is produced abundantly at the LHC. 2015 and 2016. It is often predicted to be a probe for new physics The fully hadronic final state is characterized by the phenomena at the TeV scale, in models such as the two- presence of multiple hadronic jets, two of which contain Higgs-doublet model (2HDM) [1], topcolor-assisted-tech- b-hadrons, and the absence of reconstructed leptons. This nicolor [2–4] and Randall-Sundrum (RS) models of warped all-jets topology benefits from the largest top-quark decay extra dimensions [5,6]. Resonant production of a pair of top branching fraction (45.7% of tt̄ decays), but suffers from and antitop quarks (tt̄) is particularly interesting as it large backgrounds due to QCD multijet production. provides a clear signature indicating the existence of Dedicated top-quark reconstruction and identification tech- new heavy particles decaying into tt̄. Such new particles niques are used to enhance selection of tt̄ over multijet could manifest themselves as a localized deviation from the events to maximize the sensitivity to the benchmark signals SM prediction in the high invariant mass distribution of the considered. Two different search strategies are employed, tt̄ system (mtt̄ ). In this paper, a search for new particles in each targeting a different mass range of the hypothetical resonance. In the mass range below approximately 1.2 TeV, * Full author list given at the end of the article. where the decay products of the top quarks can be resolved as separate small-radius jets, the “buckets of tops” algo- Published by the American Physical Society under the terms of rithm [7] is used to optimize the reconstruction of top- the Creative Commons Attribution 4.0 International license. quark-pair candidates. At higher masses, top-quark decay Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, products often merge into a single large-radius jet due to the and DOI. Funded by SCOAP3. high transverse momentum (pT ) of the top quarks, hence a 2470-0010=2019=99(9)=092004(38) 092004-1 © 2019 CERN, for the ATLAS Collaboration M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) second strategy with a jet-substructure-based top-quark Γ ¼ 1% to 30% to cover resonances narrower or wider than identification technique [8,9] is exploited. In the inter- the typical detector resolution of about 10%. mediate mass range of about 1.1 to 1.6 TeV, signals are As the first benchmark, a topcolor-assisted-technicolor searched for using both strategies separately. The two (TC2) model [2,3] is considered, which predicts a spin-1 results are compared at each mass point and the one with color-singlet boson. This leptophobic Z0 boson (denoted the better expected sensitivity is selected. by Z0TC2 ), referred to as Model IV in Ref. [4], couples The ATLAS and CMS collaborations performed only to first- and third-generation quarks and is mainly searches for heavy particles produced by qq̄ annihilation. The model parameters are pffiffidecaying ffi into tt̄ using pp collision data recorded at s ¼ 7 TeV [10–14], 8 TeV chosen to maximize the branching fraction for the [15–18] and 13 TeV [19–21] and set lower limits on the Z0TC2 → tt̄ decay, which reaches 33%, and the width is masses for several benchmark signal models. The ATLAS set to Γ ¼ 1% or 3%. search at 13 TeV [19], using data equivalent to 36.1 fb−1 , A framework of simplified models for dark matter (DM) exploits the lepton-plus-jets topology, where a high-pT interactions is considered as the second benchmark. An electron or muon and large missing transverse momentum axial-vector mediator Z0med;ax and a vector mediator Z0med;vec are required, and excludes masses below 3.0 (3.8) TeV for are used, following the recommendation of the LHC Dark the new Z0TC2 boson with an intrinsic decay width1 of Γ ¼ Matter Working Group in Ref. [25]. In the simplified model 1% (3%) in the topcolor-assisted-technicolor model [2,3] there are five parameters relevant for pp → Z0med → tt̄ (described in Sec. II). The CMS search with lepton-plus- processes (Z0med is either Z0med;ax or Z0med;vec ): the mediator jets, all-jets, and dilepton topologies at 13 TeV [21] mass mmed , the dark-matter mass mDM , and the mediator excludes the Z0TC2 boson with Γ ¼ 1% up to 3.8 TeV using couplings to quarks gq , to leptons gl , and to dark matter 35.9 fb−1 . The Kaluza-Klein (KK) excitation of the grav- gDM . This search considers the coupling parameters defined iton GKK predicted in the specific “bulk” RS model [22,23] in the A1 (V1) scenario of Ref. [25] for the axial-vector decaying into tt̄ (see details in Sec. II) was also searched for (vector) mediator. The branching fraction of the mediators by the ATLAS Collaboration and the mass range from 0.45 into tt̄ is 8.8% and the width is approximately constant at to 0.65 TeV is excluded assuming k=M̄Pl ¼ 1, where k is Γ ¼ 5.6% over the search range considered. The DM mass the curvature pffiffiffiffiffiffi of the warped extra dimension and M̄Pl ¼ mDM is fixed to 10 GeV. MPl = 8π is the reduced Planck mass. The KK excitation of An RS model with the SM fields propagating in the bulk the gluon, gKK , predicted in an RS model with a single of a single warped extra dimension [6] is used as the third warped extra dimension [6] with Γ ¼ 15% (30%) is benchmark, which predicts a spin-1 color-octet boson, the excluded by the ATLAS search up to 3.8 (3.7) TeV. The first KK excitation of the gluon, gKK . The gKK is primarily CMS search [21] considered a slightly different model produced in qq̄ annihilation and decays predominantly into [24], including a KK gluon with Γ ¼ 20% and larger tt̄ with a branching fraction of approximately 92.5% as production cross section, and set a lower limit of predicted in Ref. [6]. In this analysis, the coupling of the 4.55 TeV on the mass. KK gluon to quarks is set to gq ¼ −0.2gs , where gs is the The paper is organized as follows. The signal models considered are discussed in Sec. II. After a brief description strong coupling constant in the SM. The left-handed of the ATLAS detector in Sec. III, the data and simulation coupling to the top quark is fixed to gs while the right- samples are summarized in Sec. IV. The analysis strategy handed coupling is varied to change the intrinsic width. including event selection, reconstruction and categorization The “bulk” RS model [22,23] with the SM fields is presented in Sec. V. The background estimation is propagating in the bulk, inherited from the original RS described in Sec. VI and the systematic uncertainties in model, is used as the fourth benchmark to predict a spin-2 the background and signal predictions in Sec. VII. After color-singlet boson. The first KK excitation of the graviton, describing the signal search and the statistical procedure in GKK , in this model is mainly produced in gluon-gluon Sec. VIII, the results are presented in Sec. IX with the fusion, and the production rate and width are controlled by conclusions given in Sec. X. a dimensionless coupling constant k=M̄Pl . In this analysis k=M̄Pl is chosen to be 1, resulting in the GKK width varying II. SIGNAL MODELS from Γ ¼ 3% to 6% in the mass range between 0.5 and 3 TeV. The branching fraction of the GKK into tt̄ increases Several benchmark signal models are considered in this from 18% to 50% between 400 and 600 GeV and stays analysis, in which new spin-1 or spin-2 color-singlet and approximately constant at 68% for masses larger than color-octet bosons with masses ranging from 0.5 to 5 TeV 1 TeV. In addition, the GKK can decay into a pair of W, are introduced. The width of these bosons can vary from Z or Higgs bosons and, with negligible branching fraction, into light fermions or photons. 1 In the rest of this paper, the decay width of a resonance Representative leading-order (LO) Feynman diagrams of divided by the resonance mass is referred to as the width. the benchmark signals are presented in Fig. 1. 092004-2 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) (a) (b) (c) FIG. 1. Representative Feynman diagrams for leading-order production in the selected signal models: (a) Z0 , (b) gKK and (c) GKK . The details of each signal model are described in the text. III. ATLAS DETECTOR trigger implemented in custom electronics, which uses a subset of the detector information to reduce the event rate to The ATLAS detector at the LHC is a multipurpose, 100 kHz. This is followed by a software-based trigger that forward-backward symmetric detector2 with nearly full reduces the accepted event rate to 1 kHz on average by solid angle coverage, as described in Refs. [26–28]. It refining the first-level trigger selection. consists of an inner tracking detector (ID) surrounded by a thin superconducting solenoid, a calorimeter system com- IV. DATA AND SIMULATION posed of electromagnetic (EM) and hadronic calorimeters, and a muon spectrometer. This analysis is based on 36.1 fb−1 of pp collisions The ID consists of a silicon pixel detector, a silicon recorded by the ATLAS experiment at the LHC at a center- microstrip tracker and a transition radiation tracker, all of-mass energy of 13 TeV in 2015 and 2016. A number of immersed in a 2 T axial magnetic field, and provides quality criteria were imposed to ensure that the data were charged-particle tracking in the range jηj < 2.5. The EM collected during stable beam conditions with the relevant calorimeter is a lead/liquid-argon (LAr) sampling calorim- detectors operational. Simulated signal and background eter with accordion geometry. It is divided into a barrel events are used to optimize the event selection, to estimate section covering jηj < 1.475 and two endcap sections the background contribution and to perform the hypothesis covering 1.375 < jηj < 3.2. For jηj < 2.5 it is divided into test of the benchmark signal models considered. three layers in depth, which are finely segmented in η and The main backgrounds after applying criteria to enhance ϕ. In the region jηj < 1.8, an additional thin LAr presam- potential signals originate from SM tt̄ and multijet pro- pler layer is used to correct for energy losses in the material duction. The tt̄ contribution and the related modeling upstream of the calorimeters. The hadronic calorimeter is a uncertainties are evaluated using Monte Carlo (MC) sampling calorimeter composed of steel/scintillator tiles in simulated events, while the multijet contribution is esti- the central region (jηj < 1.7), while copper/LAr modules mated directly from data. However, simulated events of are used in the endcap (1.5 < jηj < 3.2) regions. The multijet processes are used to optimize selection criteria forward region (3.1 < jηj < 4.9) is instrumented with and derive residual corrections to the multijet distributions. copper/LAr and tungsten/LAr calorimeter modules opti- For the generation of SM tt̄ events, the next-to-leading- mized for electromagnetic and hadronic measurements, order (NLO) generator POWHEG-BOX v2 [30–32] was used respectively. Surrounding the calorimeters is a muon with the CT10 [33,34] parton distribution function (PDF) spectrometer that includes three air-core superconducting set in the matrix element calculations. The tt̄ production pffiffiffi toroidal magnets and multiple types of tracking chambers, cross section in pp collisions at s ¼ 13 TeV is σ tt̄ ¼ providing precision tracking for muons with jηj < 2.7 and 832þ46 −52 pb for a top-quark mass of 172.5 GeV. It was trigger capability in the range jηj < 2.4. calculated at next-to-next-to leading order (NNLO) in QCD A two-level trigger system is used to select events for including resummation of next-to-next-to-leading logarith- offline analysis [29]. Events are first selected by the level-1 mic soft gluon terms with Top++2.0 [35–41]. Parton showering, hadronization and the underlying event were 2 ATLAS uses a right-handed coordinate system with its simulated using PYTHIA v6.428 [42] with the CTEQ6L1 origin at the nominal interaction point (IP) in the center of the [43] PDF set and the corresponding Perugia 2012 set of detector and the z axis along the beam pipe. The x axis points tuned parameters [44]. The hdamp parameter, which controls from the IP to the center of the LHC ring, and the y axis points the transverse momentum of the first additional parton upwards. Cylindrical coordinates ðr; ϕÞ are used in the transverse plane, ϕ being the azimuthal angle around the z axis. The emission beyond the Born configuration, was set equal to pseudorapidity is defined in terms of the polar angle θ as the top-quark mass. The top-quark kinematics in tt̄ events η ¼ − ln ptanðθ=2Þ. Angular ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi distance is measured in units of were corrected to account for electroweak higher-order ΔR ≡ ðΔηÞ2 þ ðΔϕÞ2 . effects [45]. The generated events were weighted by this 092004-3 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) correction factor as a function of the flavor and center-of- corrections are applied so that the object identification mass energy of the initial partons, and of the decay angle of efficiencies, energy scales and energy resolutions match the top quarks in the center-of-mass frame of the initial those determined from control samples of data. partons. The value of the correction factor decreases with increasing mtt̄ from 0.98 at mtt̄ ¼ 0.4 TeV to 0.87 at V. EVENT RECONSTRUCTION, SELECTION mtt̄ ¼ 3.5 TeV. Multijet processes were simulated with AND CATEGORIZATION the PYTHIA v8.186 [46] generator using the LO NNPDF2.3 [47] PDF set. The production of a pair of hadronically decaying top Simulated signal samples of spin-1 color-singlet Z0TC2 quarks is characterized by the presence of multiple had- bosons decaying into tt̄ were generated with PYTHIA v8.165 ronic jets. When the top quarks have moderate transverse [46] with the LO NNPDF2.3 PDF set and the A14 set [48] momentum, pT , of less than approximately 500 GeV, the of tuned parameters. To account for higher-order contri- decay products can be reconstructed as separate jets, which butions, the LO calculation of the cross section was is referred to as the “resolved” event topology. At higher multiplied by a factor 1.3 obtained at NLO in QCD [49] transverse momentum, the decay products of each of the using the PDF4LHC2015 PDF set [50]. For the spin-1 two top or antitop quarks are merged into a single large- mediators Z0med in the DM simplified model, the same radius jet, referred to as the “boosted” event topology. For samples are used after being reweighted to have the both topologies the identification and reconstruction of the approximate mediator width and cross section as jets originating from the top quarks is crucial for recon- simulated by MADGRAPH5_aMC@NLO [51]. The structing the top-quark pair, resulting in a better separation production cross sections were calculated at LO accuracy of signal from background. The resolved and boosted event using the LO NNPDF2.3 PDF set. The production of a analyses are employed in parallel in the analysis. spin-2 bulk RS graviton GKK was performed using MADGRAPH5_aMC@NLO with the LO NNPDF2.3 PDF A. Object reconstruction and event preselection set, interfaced to PYTHIA v8.165 with the A14 set of tuned Events are required to have at least one pp interaction parameters for parton shower and hadronization. Simulated vertex associated with two or more tracks with samples of spin-1 color-octet KK gluons gKK with Γ ¼ 30% pT > 400 MeV. If more thanP one vertex is found in an were generated with PYTHIA v8.165 with the same PDF and event, the one with the largest p2T of associated tracks is tuned parameters as those used for the Z0TC2 samples. chosen as the primary interaction vertex. Depending on the Samples of gKK with different widths (from 10% to 40%) kinematic regime of the top quarks, resolved or boosted, were derived by reweighting the shapes of corresponding different jet reconstruction techniques are applied. Events samples with Γ ¼ 30% and adjusting their normalization containing leptons (electrons or muons) are included in the according to the appropriate prediction. The Z0TC2 and gKK complementary search targeting the lepton-plus-jets top- samples were generated for the mass range between 0.5 and ology [19] but are rejected in the analyses presented here. 5 TeV. Signal masses were sampled at intervals of 100– Small-R jets are built from three-dimensional topological 150 GeV below 1 TeV, 250 GeV between 1 and 3 TeV and clusters of energy deposits in the calorimeter [57], cali- 500 GeV above 3 TeV for the Z0TC2 . The gKK samples were brated at the electromagnetic energy scale, using the anti-kt produced at fixed intervals of 500 GeV in all mass ranges. algorithm [58] with a radius parameter R ¼ 0.4. These jets The GKK samples were generated between 0.5 and 3 TeV in are calibrated to the hadronic energy scale by applying pT - steps of 250 GeV (1 TeV) below (above) 1 TeV. The and η-dependent corrections derived from MC simulations simulated samples are also used to evaluate the acceptance and in situ measurements pffiffiffi obtained from Z=γ þ jets and and selection efficiencies for the signals considered in the multijet events at s ¼ 13 TeV [59]. Jets from pileup search. interactions are suppressed by applying the jet vertex tagger The EVTGEN v1.2.0 program [52] was used in all [60], which uses information from tracks associated simulated samples to model the properties of heavy-flavor with the hard-scatter and pileup vertices, to jets with pT < hadron decays. All simulated samples include the effects of 60 GeV and jηj < 2.4. Events containing jets from calo- multiple pp interactions in the same and neighboring rimeter noise or noncollision backgrounds are removed by bunch crossings (pileup) and are processed through the discarding events containing at least one jet failing to ATLAS detector simulation [53] based on GEANT4 [54]. satisfy the loose quality criteria defined in Ref. [61]. Jets Pileup effects were emulated by overlaying simulated that satisfy all the selection requirements and have pT > minimum-bias events generated with PYTHIA v8.186, using 25 GeV and jηj < 2.5 are considered in the resolved the MSTW2008LO PDF set [55] and the A2 set of tuned analysis. Small-R jets containing b-hadrons are identified parameters [56]. The number of overlaid minimum- using an algorithm [62] based on multivariate techniques to bias events was adjusted to match the luminosity profile combine information from the impact parameters of dis- of the recorded data. Simulated events were processed placed tracks as well as topological properties of secondary through the same reconstruction software as the data, and and tertiary decay vertices reconstructed within the jet. 092004-4 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) Two working points with 70% (tight) and 85% (loose) and 23.8% (40.5%) for light-flavor quarks and c-quarks, efficiencies for b-quark-induced jets are chosen, where the respectively. efficiencies are averaged values derived from simulated Electrons are reconstructed from clusters of EM calo- SM tt̄ events. The corresponding misidentification rates rimeter energy deposits matched to an ID track with of the tight (loose) working point are 0.26% (3%) and jηj < 2.47, excluding the barrel and endcap transition 8% (32%) for jets containing hadrons composed of light- region of 1.37 < jηj < 1.52. The electron candidates are flavor quarks and c-quarks, respectively. Efficiencies to tag required to have ET > 25 GeV and to satisfy the “tight” jets from b- and c-quarks in the simulation are corrected identification criteria defined in Ref. [70]. To suppress to match the efficiencies in data using pT -dependent contamination from misidentified hadrons, the electron factors, whereas the light-jet efficiency is scaled by pT - and candidates are further required to be isolated from other η-dependent factors [62]. hadronic activity in the event. This is achieved by requiring Large-R jets are built from three-dimensional topo- the scalar sum of track pT within a cone around the electron logical clusters of energy deposits in the calorimeter direction, excluding the track associated with the electron, calibrated with the local cluster weighting (LCW) pro- to be less than 6% of the electron transverse momentum cedure [57] using the anti-kt algorithm with a radius peT . The cone size is given by the minimum of ΔR ¼ parameter R ¼ 1.0. The noncompensating response of the 10 GeV=peT and ΔR ¼ 0.2. calorimeter and the energy loss in dead material and due Muons are reconstructed by combining tracks separately to out-of-cluster leakage from charged and neutral par- reconstructed in the ID and the muon spectrometer. The ticles are corrected in the LCW procedure before jet muon candidates are required to have pT > 25 GeV and reconstruction. The reconstructed jets are “trimmed” [63] jηj < 2.5, and satisfy the “medium” quality requirements to mitigate contributions from pileup and soft radiation. defined in Ref. [71]. The muons are also required to be In the trimming procedure, the jet constituents are isolated by using the same track-based isolation conditions reclustered into subjets using the kt algorithm [64–66] as for electrons, except that the value of ΔR ¼ 0.2 is with a radius parameter R ¼ 0.2 and subjets with pT less replaced with ΔR ¼ 0.3. Electron and muon candidate tracks are required to be than 5% of the pT of the parent jet are removed [67]. associated with the primary vertex using criteria based on Finally, the large-R jets are formed from the momentum the longitudinal and transverse impact parameters. To avoid vectors of the remaining subjets and selected by requiring the misidentification of jets as electrons and electrons pT > 200 GeV and jηj < 2.0 in the boosted analysis. For from heavy-flavor decays, ffi the closest small-R jet within pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi highly boosted top quarks, the mass resolution of a large- 2 2 R jet containing the top-quark decay products deteriorates ΔRy ¼ ðΔyÞ þ ðΔϕÞ ¼ 0.2 around a reconstructed with increasing top pT due to the limited angular electron is removed.3 If an electron is then found within granularity of the calorimeter. To overcome this the mass ΔRy ¼ 0.4 of a jet, the electron is removed. If a muon is of the large-R jet, mJ , is calculated by combining found within ΔRy ¼ 0.04 þ 10 GeV=pμT of a jet (where pμT the calorimeter energy measurement with the track is the muon transverse momentum), the muon is removed if information from the ID, as described in Ref. [68]. the jet contains at least three tracks, otherwise the jet is The two jets with the highest pT in the event are required removed. to have 50 GeV < mJ < 350 GeV. In the resolved analysis, the event selection is based on Track-jets are built from charged-particle tracks using multijet triggers requiring the presence of at least five the anti-kt algorithm with a radius parameter R ¼ 0.2. small-R jets with pT > 60–65 GeV depending on the data- Tracks used in the reconstruction are selected by requiring taking periods. Events are further required to have at least that they are associated with the primary vertex, and have six jets with pT > 25 GeV and jηj < 2.5, out of which the pT > 400 MeV and jηj < 2.5. Track-jets composed of at five highest-pT jets must have pT > 75 GeV and jηj < 2.4. least two constituent tracks and having pT > 10 GeV and Among those six jets at least two of them are required to be jηj < 2.5 are used to identify jets containing b-hadrons in b-tagged with jηj < 1.6 using the loose efficiency working the boosted analysis. In the dense environment character- point. The trigger efficiency for the events satisfying the istic of the boosted topology, the b-tagging is more efficient offline selection criteria is estimated using a lower-thresh- if performed on track-jets than on calorimeter jets [69]. The old multijet trigger. The trigger efficiency is above 99% and same b-tagging algorithm as used for small-R jets with 77% consistent between data and the simulated events. (tight) and 85% (loose) efficiency working points from b- In the boosted analysis, events are selected using quark-induced jets is employed. The training of the triggers that require at least one large-R jet with pT > multivariate algorithm and the evaluation of systematic 360–420 GeV depending on the data-taking periods. uncertainties associated with the track-jet b-tagging effi- ciency are performed separately from those for the small-R The rapidity is defined as y ¼ 12 ln Eþp 3 z E−pz where E is the energy calorimeter jets. The corresponding misidentification and pz is the longitudinal component of the momentum along the rates at the tight (loose) working point are 1.7% (5.3%) beam direction. 092004-5 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019)   Events are required to have at least two large-R jets with  mkl mW    pT > 400 GeV to ensure that the jets can fully contain the  m − m  < 0.15; Bi top top-quark decay products. The large-R jets with the highest and the second-highest pT in the event are referred to as the where mkl is the invariant mass of the (k, l) jet combination leading and sub-leading jets, respectively. The leading jet inside Bi , and mW ¼ 80.4 GeV is the W-boson mass. To has to satisfy pT > 500 GeV to ensure a nearly full trigger retain tt̄ events where one of the jets originating from the top- efficiency. The trigger efficiency is measured using a quark decay, presumably the softer quark from W → qq̄0 , control sample in data and found to be approximately falls outside the top buckets, two-jet top buckets are formed. 100% in this pT range. The invariant mass mJJ of the two The metric used to form the bucket is adjusted to be leading large-R jets is required to be mJJ > 1 TeV to avoid a kinematic bias caused by the jet pT requirements. The two Δbj B ¼ jmB − 145 GeVj leading jets are required to have an azimuthal angle difference larger than 1.6. In addition, each jet is required if the bucket mass mB is smaller than 155 GeV, otherwise to have at least one track-jet within ΔR ¼ 1.0 satisfying the Δbj B is set to an arbitrary large number. The mass criteria are loose b-tagging efficiency working point. The fraction of based on the top-decay kinematics in which only the b-quark events with more than two b-tagged jets is negligibly small, and the harder quark from W → qq̄0 fall inside the bucket. and those events are rejected to simplify the data-driven When the two top buckets are classified as (tW , tW ) the event multijet background estimation. is kept. If the buckets are classified as (tW , t− ) or (t− , tW ) with the notation that the first bucket in the parentheses is always B. Top-quark pair reconstruction chosen to be B1 , the t− bucket is recalculated using the new In the resolved analysis, the top-quark pair reconstruc- metric Δbj B from all jets excluding those belonging to any tW tion is achieved by exploiting the “buckets of tops” bucket in the event. Hereafter these two categories are algorithm [7] using small-R jets. In this algorithm, all jets collectively referred to as (tW , t− ). If the two top buckets are in the event are assumed to originate from tt̄ events, (t− , t− ), the new buckets are formed from all jets in the event including those from initial- or final-state radiation, and by minimizing the sum of a new metric Δbj bj B1 þ ΔB2 . The new are assigned to one of three groups, referred to as “buckets.” two-jet bucket is finally required to satisfy the mass window The first two buckets correspond to reconstructed candidates requirement of of the two top quarks in tt̄ events and the third bucket contains all jets from extra radiation. The assignment of 75 GeV < mbj Bi < 155 GeV: small-R jets to buckets is performed by taking all jet combinations and minimizing a metric based on the differ- If an event has no buckets satisfying the mass window ence between the invariant mass of jets falling into one of the requirements, the event is classified as (t0 , t0 ). Finally, the first two buckets and the top-quark mass. In this analysis the top-quark candidate, reconstructed as the sum of the metric Δ2 is defined as momentum vectors of the jets in the tW , t− or t0 bucket, Δ2 ¼ ωΔ2B1 þ Δ2B2 ; ΔB1ð2Þ ¼ jmB1ð2Þ − mtop j; ω ¼ 100; is required to have pT > 200 GeV to suppress multijet backgrounds. The performance of the resolved tt̄ where mB1ð2Þ is the invariant mass of the jets falling into reconstruction is summarized in Table I. The resolution of the reconstructed tt̄ mass for the resolved analysis is bucket 1(2), denoted by B1ð2Þ, and mtop ¼ 173.5 GeV is the typically 6%. top-quark mass. The difference from mtop used in the For the boosted analysis, a top-quark pair is recon- simulation (172.5 GeV) does not affect the performance structed using the top-quark tagging requirements based on of the tt̄ reconstruction. The ω factor is introduced to ensure the jet mass and a jet substructure variable called n- that B1 has a mass closer to mtop than B2 , i.e., ΔB1 < ΔB2 , as subjettiness, τn [8,9]. For each large-R jet, τn is calculated described in Ref. [7]. No restriction is imposed on the by reconstructing exactly n subjets with the “winner-take- multiplicity of jets falling into the buckets except that B1 and all” recombination scheme [72] from the large-R jet B2 are required to contain exactly one b-tagged jet each. constituents using the kt algorithm [64–66] with a radius Furthermore, the mass window requirements of parameter of R ¼ 0.2: 155 GeV < mB1;2 < 200 GeV 1X i τn ¼ p × minðΔR1;i ; ΔR2;i ; …; ΔRn;i Þ; are applied to increase the fraction of tt̄ events. The preferred d0 i T two “top buckets” B1;2 are further classified according to the hadronic W-boson decay. If the following condition is where piT is the transverse momentum of the ith large-R jet satisfied for at least one combination of two non-b-tagged constituent and ΔRj;i is the y-ϕ distance between the jets (k, l), the bucket is considered to contain a W-boson and the ith constituent. The τn variable is scaled by subjet j P candidate and labeled tW , otherwise it is labeled t− : d0 ¼ ð i pT i × RÞ−1 with R ¼ 1.0, the radius parameter −1 092004-6 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) TABLE I. Performance of the resolved tt̄ reconstruction with the “buckets of tops” algorithm estimated using simulated SM tt̄ and Z0TC2 (850 GeV) events in the fully hadronic final state. The fraction of events in each of the five possible top bucket categories is shown for all events satisfying the selection criteria described in Sec. VA. For each event category the relative fraction of events that have correctly matched top-quark pairs is presented. The measure of accuracy is based on a geometrical matching in the η-ϕ plane. Specifically the matched top buckets are required to be within ΔR ¼ 0.3 of a simulated top quark. The momenta of the simulated top quarks are evaluated immediately before the decay. The errors indicate the statistical uncertainty only. Fraction of events [%] Matched top-quark pairs [%] Top buckets category SM tt̄ Z0TC2 (850 GeV) SM tt̄ Z0TC2 (850 GeV) ðt0 ; t0 Þ 16.5  0.3 12.6  0.7 57.1  1.0 63.6  2.7 ðt− ; t− Þ 17.5  0.3 15.0  0.9 66.7  0.9 74.2  2.6 ðt− ; tW Þ 7.8  0.2 7.9  0.8 72.2  1.3 80.0  3.9 ðtW ; t− Þ 30.2  0.4 30.9  1.2 78.9  0.6 82.6  1.5 ðtW ; tW Þ 28.0  0.4 33.6  1.3 88.7  0.5 90.7  1.1 TABLE II. Performance of the boosted tt̄ reconstruction in the boosted analysis estimated using simulated SM tt̄ and Z0TC2 (3 TeV) events in the fully hadronic final state. The fraction of events in each of the eight possible boosted signal regions is shown for all events satisfying the selection criteria described in Sec. VA, together with the relative fraction of events that have correctly matched top-quark pairs. The measure of accuracy is based on a geometrical matching in the η-ϕ plane. Specifically the matched large-R jets are required to be within ΔR ¼ 0.4 of a simulated top quark. The notation used to define each signal region is described in Sec. V C. The momenta of the simulated top quarks are evaluated immediately before the decay. The errors indicate the statistical uncertainties only. Fraction of events [%] Matched top-quark pairs [%] Signal region category SM tt̄ Z0TC2 (3 TeV) SM tt̄ Z0TC2 (3 TeV) Medium R1 1b 1.80  0.07 2.41  0.08 89.8  4.4 86.7  4.1 Medium R1 2b 5.24  0.11 4.39  0.10 94.0  2.7 84.3  2.8 Tight R1 1b 2.55  0.08 2.07  0.10 93.8  4.0 83.5  4.2 Tight R1 2b 7.75  0.14 4.18  0.10 97.2  2.3 83.5  2.8 Medium R2 1b 1.20  0.06 1.99  0.07 83.8  5.3 86.4  4.4 Medium R2 2b 3.13  0.09 3.08  0.08 91.4  3.3 86.3  3.3 Tight R2 1b 0.89  0.05 1.54  0.06 90.0  6.6 89.8  5.2 Tight R2 2b 2.25  0.07 2.59  0.07 93.9  4.1 86.5  3.6 of the large-R jet. To distinguish fully contained top quarks In the resolved analysis, events satisfying the preselection with a three-prong structure from other backgrounds domi- criteria in Sec. VA are classified according to the recon- nated by a single-prong or two-prong structure, the τ32 variable structed top buckets and number of b-tagged jets in the defined as τ32 ¼ τ3 =τ2 is used as a discriminant. Since there events. The combination of four possible pairs of top buckets, are two top quarks in signal events, the τ32 variables from the (tW , tW ), (tW , t− ), (t− , t− ) and (t0 , t0 ), and the two b-tagging two leading large-R jets are used to construct a single criteria, i.e., (1) satisfying the tight or (2) satisfying the loose likelihood ratio Lτ32 , which is then used to suppress the but failing to satisfy the tight efficiency working points for multijet background. The likelihood ratio is computed as both b-tagged jets, are used to classify events into eight Lτ32 ¼ Ps =ðPs þ Pb Þ where Ps and Pb are the probability different regions A–D, A0 , A− , C0 and C− defined in density functions for the signal and background, respectively, Table III. By construction those regions have no overlapping obtained from MC simulations (see Sec. IV). The performance events. Region D, which contains events with (tW , tW ) of the tt̄ reconstruction in the boosted analysis is summa- buckets and tight b-tagged jets, is the most sensitive to rized in Table II, where signal regions as defined in Sec. V C the benchmark signals and hence chosen to be the main are used for illustration. The resolution of the reconstructed tt̄ signal region (SR) for the resolved analysis. Regions A–C are mass for the boosted analysis is typically 10%. used in a joint likelihood fit with the SR to extract the multijet background in the SR as detailed in Sec. VI. The regions with C. Event categorization the (t− , t− ) and (t0 , t0 ) buckets (A0 , A− , C0 and C− ) are used For both the resolved and boosted analyses, the recon- to estimate systematic uncertainties associated with the structed events are categorized into several sub multijet background modeling (see Sec. VII). samples used for the signal search and background In the boosted analysis, preselected events are first estimation. categorized by the number of tight b-tagged track-jets 092004-7 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) TABLE III. Event categorization in the resolved analysis. The multijet-enriched regions A–C and the main signal region D, as well as the additional validation regions A0 , A− , C0 , C− selected with looser requirements on the top-quark pair candidates are shown. The events are also classified according to the two b-tagging criteria, i.e., satisfying the tight or satisfying the loose but failing to satisfy the tight efficiency working points for both b-tagged jets. The expected fraction of tt̄ events to the total background events in each region, as estimated from the simulation, is given in parentheses. The error indicates the statistical uncertainty only. Top buckets category ðt0 ; t0 Þ ðt− ; t− Þ ðtW ; t− Þ ðtW ; tW Þ Loose b-tag A0 ð2.1  0.0Þ% A− ð4.2  0.1Þ% A ð12.3  0.2Þ% B ð38.9  0.9Þ% Tight b-tag C0 ð8.0  0.1Þ% C− ð16.9  0.2Þ% C ð44.9  0.5Þ% D ð79.6  1.3Þ% (nb ) and the τ32 -likelihood ratio (Lτ32 ) as shown in Fig. 2(a). optimizing the signal sensitivity while the lower boundary Most signal events have nb ¼ 1 or 2, which define the 1b of the Loose region is used to ensure that events have and 2b regions. The events with nb ¼ 0 (0b region) are kinematic properties similar to those in the Tight and used to model the multijet background. Medium regions. The Loose region is used for validation For the Lτ32 variable, the three criteria 0.35 ≤ Lτ32 < 0.6, of the background estimation across the Lτ32 regions (see 0.6 ≤ Lτ32 < 0.8 and 0.8 ≤ Lτ32 ≤ 1.0 define Loose, Sec. VI for details). The possible contamination from Z0TC2 Medium and Tight regions, respectively, while 0.35 ≤ signal events in the Loose region is a few percent as Lτ32 < 1.0 is referred to as Inclusive. The lower boundaries estimated for a signal with a cross section that has already of the Tight and Medium regions are determined by been excluded by previous analyses. It is hence negligible Selected events in the boosted analysis nb: =0 =1 =2 L 32 : [0.35, 0.6) [0.6, 0.8) [0.8, 1.0] [0.35, 0.6) [0.6, 0.8) [0.8, 1.0] [0.35, 0.6) [0.6, 0.8) [0.8, 1.0] (Loose, (Medium, (Tight, (Loose, (Medium, (Tight, (Loose, (Medium, (Tight, Category: 0b) 0b) 0b) 1b) 1b) 1b) 2b) 2b) 2b) (a) 340 Sub-leading jet mass [GeV] CR1 CR2 190 R1 140 CR3 R2 CR4 50 50 140 190 340 Leading jet mass [GeV] (b) FIG. 2. Schematic diagram of the event categorization in the boosted analysis. (a) Events selected in the boosted analysis are classified into nine categories based on the number of tight b-tagged jets (nb ) and Lτ32 , i.e., Loose, Medium and Tight regions for nb ¼ 0, 1 and 2. At least two loose b-tagged jets are already required in the preselection. The region 0.35 ≤ Lτ32 < 1.0 is referred to as Inclusive. (b) In each category, events are further classified into three regions, R1, R2 and CR1–4, according to the leading and sub-leading large-R jet masses. 092004-8 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) ATLAS Data ATLAS Data 1600 1600 s = 13 TeV, 36.1 fb-1 tt s = 13 TeV, 36.1 fb-1 tt Boosted MultiJet Boosted MultiJet 1400 1400 Tight 1b Total Bkg unc. Tight 2b Total Bkg unc. 1200 Post-Fit Post-Fit 1200 Events / 10 GeV Events / 10 GeV 1000 1000 800 800 600 600 400 400 200 200 1.50 1.50 Data / Bkg Data / Bkg 1.25 1.25 1 1 0.75 0.75 0.5 0.5 50 100 150 200 250 300 350 50 100 150 200 250 300 350 Leading Large-R Jet mass [GeV] Leading Large-R Jet mass [GeV] (a) (b) 1200 ATLAS Data ATLAS Data 1400 s = 13 TeV, 36.1 fb-1 tt s = 13 TeV, 36.1 fb-1 tt Boosted MultiJet Boosted MultiJet 1000 1200 Tight 1b Total Bkg unc. Tight 2b Total Bkg unc. Post-Fit Post-Fit 800 1000 Events / 10 GeV Events / 10 GeV 800 600 600 400 400 200 200 1.50 1.50 Data / Bkg Data / Bkg 1.25 1.25 1 1 0.75 0.75 0.5 0.5 50 100 150 200 250 300 350 50 100 150 200 250 300 350 Sub-leading Large-R Jet mass [GeV] Sub-leading Large-R Jet mass [GeV] (c) (d) FIG. 3. Comparison between data and predicted background after the fit (“Post-Fit”) in events satisfying the criteria for the Tight Lτ32 requirement and one (a),(c) or two (b),(d) b-tagged jets in the boosted analysis. Shown are (a),(b) the mass of the leading reconstructed top-quark candidate, and (c),(d) the mass of the sub-leading reconstructed top-quark candidate. The background components are shown as stacked histograms and the shaded areas around the histograms indicate the total systematic uncertainties after the fit. The lower panel of the distribution shows the ratio of data to the background prediction. The multijet contribution also contains all other small non-tt̄ backgrounds. for the signals with higher masses, which have lower Representative distributions of the jet masses are shown in predicted cross sections, and also for other benchmark Fig. 3 for events satisfying the ðTight; 1bÞ or ðTight; 2bÞ signals with kinematic properties similar to the Z0TC2 . In requirements. The jet mass distributions are shown for the each category, events are further classified into different data and background predictions obtained after the fit to regions using the masses mJ1 and mJ2 of large-R jets with data (“Post-Fit”), as detailed in Sec. VIII. Signal regions are the leading and sub-leading pT as shown in Fig. 2(b). defined in the ranges 140 75 GeV. The small-R JER uncertainty (typ- The uncertainty in the combined 2015 þ 2016 integrated ically 2%–3% at pT ¼ 50 GeV) obtained from an in situ luminosity is 2.1%. It is derived, following a methodology measurement of jet response using dijet events [59] is similar to that detailed in Ref. [74], and using the LUCID-2 also included. The uncertainty in the efficiency of the jet detector for the baseline luminosity measurements [75], vertex tagger (Sec. VA) is also considered following from calibration of the luminosity scale using x-y beam- Ref. [60]. The impact on the total background yield (for separation scans. The pileup modeling uncertainty is a 850 GeV Z0TC2 signal) in the resolved analysis is about considered by varying the average number of pp collisions 9% (11%) for the JES uncertainty and 3% (11%) for the in simulated events. JER uncertainty. In the resolved analysis the trigger efficiency is The large-R JES uncertainties are estimated with the corrected around the jet pT threshold at the trigger Rtrk method using dijet data control samples [68,73]. level. The uncertainty in the correction factor, estimated The method assumes that the track-related uncertainties to be below 1%, is dominated by the statistical uncer- are uncorrelated with the calorimeter cluster-related tainty of the lower-threshold trigger data. In the boosted uncertainties. The procedure works by measuring the analysis the uncertainty in the trigger efficiency is found ratio rtrk of an observable (which can be the pT , mJ or to be negligible. τ32 variables) using calorimeter jets to that using track- jets reconstructed within the same detector region. The B. Background modeling uncertainties deviation of the average data-to-simulation ratio hRtrk i ¼ In this section, uncertainties associated with the data- hrdata trk i=hrtrk i from unity is taken as the uncertainty, MC driven estimates of multijet background and theory uncer- together with the uncertainties associated with the track tainties in the SM tt̄ prediction are discussed. measurement, charged particle multiplicity modeling in As discussed in Sec. VI, in both the resolved and boosted simulation and the statistical uncertainty of the dijet analyses the multijet background in the SRs is estimated by sample. The impact on the total background yield (for a extrapolating the mreco tt̄ shape obtained from the regions 3 TeV Z0TC2 signal) in the boosted analysis is about where the b-tagging criterion is loosened compared with 3% (4%) for the large-R JES uncertainty and 3% (2%) that in the SRs. Uncertainties in the mreco tt̄ shape and the for the large-R JER uncertainty. yield of the multijet background are estimated separately as Correction factors to the simulated event samples are follows. applied, separately for small-R jets and track-jets, to The different b-tagging criteria between the signal and compensate for differences observed between data and control regions could produce a bias in the predicted mreco tt̄ simulation in the b-tagging efficiency of b-, c- and light- distributions. In the resolved analysis this effect is esti- quark and gluon-induced jets [62]. The correction factor mated by comparing the mreco tt̄ distributions in the vali- for b-jets is derived from tt̄ events with final states dation regions A0 and C0 (see Table III) and the difference containing two leptons, and is consistent with unity observed is assigned as a systematic uncertainty in the within uncertainties at the level of a few percent over multijet background shape. The assumption that the most of the jet pT range. Uncertainties in the correction potential bias is caused by the b-tagging instead of top- factors for the b-tagging identification efficiency result in quark tagging is verified by repeating the same procedure a variation of the total background yield of about 5% (4%) using the validation regions A− and C− , which gives a for the resolved (boosted) analysis. Uncertainties due to result comparable to the one from the validation regions A0 possible correlations between the correction factors in and C0 . For the boosted analysis, the variations of the the signal and control regions are checked to have a correction factor applied to the psum T distribution (see negligible impact on the final results. An additional term Sec. VI) are considered as an uncertainty in the multijet is included to extrapolate the measured uncertainties to background shape. These include the statistical uncertainty the high-pT region of interest. This term is calculated of the multijet simulation samples and a small residual from simulated events by considering variations of the difference observed in the mreco tt̄ distributions after the quantities affecting the b-tagging performance such as shape correction. A possible bias arising from using the the impact parameter resolution, percentage of poorly inclusive Lτ32 range [0.35, 1.0] for the multijet template measured tracks, description of the detector material, and extraction from TRiðInclusive; 0bÞ is also taken into track multiplicity per jet. The impact on the 3 TeV Z0TC2 account as a source of systematic uncertainty. The multijet signal yield due to such high-pT extrapolation uncer- mreco tt̄ distribution obtained from TRiðInclusive; 0bÞ is tainty is about 3%. compared with those obtained from the individual Lτ32 092004-13 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) regions [TRiðj; 0bÞ; j ¼ Medium and Tight] and the HERWIG 7 [82,83]. The uncertainty arising from the maximum difference in shape is considered. choice of PDF set is estimated by taking into account the The impact on the multijet yield due to correlation variations from the PDF4LHC15 PDF set, which between the top- and b-quark tagging variables in the includes 30 separate uncertainty eigenvectors [50], and resolved analysis is evaluated by using the (t0 , t0 ) or the difference between the nominal PDF4LHC15 and (t− , t− ) categories instead of the (tW , t− ) category. As a CT10 PDF sets. For the boosted analysis, an additional result, an uncertainty of 20% is added to the normali- uncertainty is considered in the mreco tt̄ shape due to the zation of the multijet background, resulting in a 3% extrapolation procedure using an exponential function at uncertainty in the total background yield. In the boosted high mreco tt̄ above 1.5 TeV (Sec. VI). This includes the analysis, the uncertainty in the multijet background statistical uncertainty in the exponential fit and the normalization is estimated by taking the maximum stability of the fit results estimated by varying the fit deviation of the expected yields in the four CRs from range. The overall impact on the SM tt̄ event yields from the average. This leads to a 3% uncertainty in the overall these uncertainties is estimated to be 29% in the resolved background yield. analysis and 24% in the boosted analysis. There are several sources of theoretical uncertainties affecting the modeling of SM tt̄ background processes in VIII. STATISTICAL ANALYSIS all regions including signal, control and validation regions. The cross-section uncertainty given in Sec. IV A binned maximum-likelihood fit to the mreco tt̄ distribu- accounts for the choice of PDF and strong coupling tions is performed to estimate the signal and background constant calculated using the PDF4LHC prescription yields, separately in the resolved and boosted analyses. The [76] with the MSTW2008 68% C.L. NNLO [55,77], likelihood is defined as a product of the Poisson proba- CT10 NNLO [33,34] and NNPDF2.3 5f FFN [47] PDF bilities to observe ni events when λi events are expected in sets, as well as the renormalization and factorization bin i. The λi is expressed as λi ¼ μsi ðθÞ þ bi ðθÞ where μ is scale uncertainties. In addition to this pure normalization the signal strength, defined as a signal cross section in units uncertainty, the following modeling uncertainties affect- of the theoretical prediction, to be determined by the fit, and ing both the acceptance and shape of the tt̄ kinematic si ðθÞ and bi ðθÞ are the expected numbers of signal and distributions are considered. The impact from the mod- background events, respectively. The fit includes two eling of extra QCD radiation is evaluated using POWHEG background components; tt̄ and multijet processes, which +PYTHIA samples in which the renormalization and are estimated by the simulated samples and the data-driven factorization scales and the hdamp parameter are varied methods, respectively, as described in Sec. VI. The sys- within the ranges consistent with the measurements of tt̄ tematic uncertainties are taken into account as nuisance production in association with jets [78–80]. Additionally, parameters, θ, constrained by Gaussian or log-normal the uncertainty in the tt̄ event kinematics due to higher- penalty terms in the likelihood. Nuisance parameters are order QCD effects is considered by adding an uncer- also determined by the fit, varying the normalization and tainty covering the difference between NLO and NNLO shape of the mreco tt̄ distribution for each component of the QCD calculations of tt̄ production. The recent QCD signal and background. calculations in Ref. [81] are used to derive the differ- In the resolved analysis, the likelihood fit is performed ence, which is applied as a function of top-quark pT and simultaneously in the three multijet-enriched regions A–C the transverse momentum of the tt̄ system at the particle and the main signal region D. In each region, the mreco tt̄ level taking into account the final-state radiation, to distribution is divided into 19 bins spanning the range 0 to estimate this uncertainty. The variation of the event yield 2 TeV. The shape of the multijet background is determined at the reconstruction level is less than 4% at mreco tt̄ below by bin-by-bin unconstrained normalization factors. 500 GeV, but approaches 11% at mreco tt̄ of 1.2 TeV in the Assuming that the mreco tt̄ shape does not depend on the resolved analysis and 20% above 3 TeV in the boosted b-tagging requirement, the bin-by-bin multijet normaliza- analysis. The electroweak corrections to top-quark kin- tion factors for regions A and C as well as for regions B ematics in tt̄ events have an associated uncertainty of and D are treated as fully correlated. In order to consider the about 10%, which varies as a function of mreco tt̄ [45]. normalization component not depending on the top-tagging The uncertainty associated with the choice of requirement but depending on the b-tagging requirement, a event generator is evaluated by taking the difference common free-floating normalization factor is additionally between the predictions from the tt̄ samples generated applied to regions C and D. Thus, the correlation between with POWHEG-BOX and aMC@NLO both interfaced to the (tW , t− ) and (tW , tW ) categories is introduced in the HERWIG++ v2.7.1 [82]. The uncertainty in the parton background parameterization. shower modeling is evaluated by comparing the tt̄ events The SRs in the boosted analysis cover the mreco tt̄ range simulated with the default POWHEG+PYTHIA with those between 1 and 6 TeV, which is divided into 19 bins. The fit with the same version of POWHEG-BOX but interfaced to is performed simultaneously in the eight SRs defined in 092004-14 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) Sec. V C. The mreco tt̄ shape and normalization of the multijet A test statistic based on the profile likelihood ratio [84] is background are constrained by the variations due to used to extract information about μ from a likelihood fit to systematic uncertainties estimated in Sec. VII by using data under the signal-plus-background hypothesis, sepa- them as nuisance parameters in the fit. rately for each model considered. The distributions of the ATLAS Data ATLAS Data 5000 tt 1000 tt s = 13 TeV, 36.1 fb-1 Multijet s = 13 TeV, 36.1 fb-1 Multijet Resolved Total Bkg unc. Resolved Total Bkg unc. Region A Pre-Fit Bkg Region B Pre-Fit Bkg 4000 800 Post-Fit Post-Fit Events / 50 GeV Events / 50 GeV 3000 600 2000 400 1000 200 1.10 1.20 Data / Bkg Data / Bkg 1.05 1.1 1 1 0.95 0.9 0.9 0.8 200 400 600 800 1000 1200 1400 1600 1800 2000 200 400 600 800 1000 1200 1400 1600 1800 2000 mreco tt [GeV] mreco tt [GeV] (a) (b) ATLAS Data 1800 ATLAS Data 3500 tt tt s = 13 TeV, 36.1 fb-1 Multijet s = 13 TeV, 36.1 fb-1 Multijet 1600 Resolved Total Bkg unc. Resolved Total Bkg unc. 3000 Region C Pre-Fit Bkg Region D Pre-Fit Bkg 1400 Post-Fit Post-Fit 2500 Events / 50 GeV Events / 50 GeV 1200 2000 1000 800 1500 600 1000 400 500 200 1.250 1.20 Data / Bkg Data / Bkg 1.15 1.1 1.05 1 0.95 0.9 0.85 0.8 200 400 600 800 1000 1200 1400 1600 1800 2000 200 400 600 800 1000 1200 1400 1600 1800 2000 mreco tt [GeV] mreco tt [GeV] (c) (d) FIG. 6. Observed mreco tt̄ distributions in the multijet-enriched regions (a) A, (b) B, (c) C and (d) the main signal region D after the fit (“Post-Fit”) under the background-only hypothesis for the resolved analysis. The shaded areas around the histograms indicate the total uncertainties in the background. The lower panel of the distribution shows the ratio of data to the fitted background prediction. The distributions before the fit are shown by the dashed lines and the background components are shown as stacked histograms. The multijet contribution also contains all other small non-tt̄ backgrounds. 092004-15 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) test statistic under the signal-plus-background and the computing the local p0 value, defined as the probability background-only hypotheses are obtained from pseudo to observe an excess at least as large as the one observed in experiments. The probability that the observed data is data, under the background-only hypothesis. The global p0 compatible with the SM prediction is estimated by value is computed by considering the look-elsewhere effect 108 108 ATLAS Data ATLAS Data 107 tt 107 tt s = 13 TeV, 36.1 fb-1 Multijet s = 13 TeV, 36.1 fb-1 Multijet 106 Boosted Total Bkg unc. 106 Boosted Total Bkg unc. 105 SR1(Medium,1b) Pre-Fit Bkg 105 SR1(Medium,2b) Pre-Fit Bkg 4 Post-Fit Post-Fit 104 Events / 100 GeV Events / 100 GeV 10 103 103 2 10 102 10 10 1 1 −1 −1 10 10 −2 10 10− 2 −3 10 10− 3 −4 1.5 −4 1.5 10 10 Data / Bkg Data / Bkg 1.25 1.25 1 1 0.75 0.75 0.5 0.5 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 mreco tt [GeV] mreco tt [GeV] (a) (b) 108 108 ATLAS Data ATLAS Data 107 tt 107 tt s = 13 TeV, 36.1 fb-1 Multijet s = 13 TeV, 36.1 fb-1 Multijet 106 Boosted Total Bkg unc. 106 Boosted Total Bkg unc. 10 5 SR1(Tight,1b) Pre-Fit Bkg 105 SR1(Tight,2b) Pre-Fit Bkg Post-Fit Post-Fit 104 104 Events / 100 GeV Events / 100 GeV 3 10 103 102 102 10 10 1 1 −1 −1 10 10 10− 2 10− 2 10− 3 10− 3 −4 1.5 −4 1.5 10 10 Data / Bkg Data / Bkg 1.25 1.25 1 1 0.75 0.75 0.5 0.5 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 mreco tt [GeV] mreco tt [GeV] (c) (d) FIG. 7. Observed mreco tt̄ distributions in (a) Medium R1 1b (b) Medium R1 2b (c) Tight R1 1b and (d) Tight R1 2b after the fit (“Post- Fit”) under the background-only hypothesis for the boosted analysis. The shaded areas around the histograms indicate the total uncertainties in the background. The lower panel of the distribution shows the ratio of data to the fitted background prediction. The open triangles indicate that the ratio values are outside the plotted range. The distributions before the fit are shown by the dashed lines and the background components are shown as stacked histograms. The multijet contribution also contains all other small non-tt̄ backgrounds. 092004-16 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) [85,86] associated with the multiple testing to scan the the CLs prescription [87]. The results of the resolved and signal mass points. If no significant excess is observed over boosted analyses are compared in the mtt̄ region covered by the background, expected and observed upper limits on the both analyses and the one providing the better expected signal strength are set at 95% confidence level (C.L.) using limit is selected. The upper limits on μ are converted into 108 108 ATLAS Data ATLAS Data 107 tt 107 tt s = 13 TeV, 36.1 fb-1 Multijet s = 13 TeV, 36.1 fb-1 Multijet 106 Boosted Total Bkg unc. 106 Boosted Total Bkg unc. 105 SR2(Medium,1b) Pre-Fit Bkg 105 SR2(Medium,2b) Pre-Fit Bkg Post-Fit Post-Fit 104 104 Events / 100 GeV Events / 100 GeV 3 10 103 102 102 10 10 1 1 10− 1 10− 1 10− 2 10− 2 −3 10 10− 3 −4 1.5 −4 1.5 10 10 Data / Bkg Data / Bkg 1.25 1.25 1 1 0.75 0.75 0.5 0.5 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 mreco tt [GeV] mreco tt [GeV] (a) (b) 108 108 7 ATLAS Data ATLAS Data 10 tt 107 tt s = 13 TeV, 36.1 fb-1 Multijet s = 13 TeV, 36.1 fb-1 Multijet 106 Boosted Total Bkg unc. 106 Boosted Total Bkg unc. 10 5 SR2(Tight,1b) Pre-Fit Bkg 105 SR2(Tight,2b) Pre-Fit Bkg Post-Fit Post-Fit Events / 100 GeV Events / 100 GeV 104 104 103 103 2 10 102 10 10 1 1 −1 −1 10 10 10− 2 10− 2 10− 3 10− 3 −4 1.5 −4 1.5 10 10 Data / Bkg Data / Bkg 1.25 1.25 1 1 0.75 0.75 0.5 0.5 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 mreco tt [GeV] mreco tt [GeV] (c) (d) FIG. 8. Observed mreco tt̄ distributions in (a) Medium R2 1b (b) Medium R2 2b (c) Tight R2 1b and (d) Tight R2 2b after the fit (“Post- Fit”) under the background-only hypothesis for the boosted analysis. The shaded areas around the histograms indicate the total uncertainties in the background. The lower panel of the distribution shows the ratio of data to the fitted background prediction. The open triangles indicate that the ratio values are outside the plotted range. The distributions before the fit are shown by the dashed lines and the background components are shown as stacked histograms. The multijet contribution also contains all other small non-tt̄ backgrounds. 092004-17 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) TABLE V. Expected and observed yields in the main signal region D and multijet-enriched regions A–C for the resolved analysis. The yields and their uncertainties are evaluated after the fit to data under the background-only hypothesis. The expected Z0TC2 signal yields with masses of 0.75 and 1 TeV are calculated using the μ ¼ 1 hypothesis. The multijet contribution also contains all other small non-tt̄ backgrounds. Type Region A Region B Region C Region D tt̄ 4300  280 2740  190 9820  460 8990  250 Multijet (template) 31420  770 4440  360 12840  530 1820  250 Total background 35720  770 7180  370 22660  350 10800  190 Data 35722 7186 22665 10 821 Z0TC2 ð0.75 TeVÞ 470  68 367  91 1200  140 1200  180 Z0TC2 ð1 TeVÞ 460  65 296  37 1020  130 1010  150 TABLE VI. Expected and observed yields in the signal regions for the boosted analysis. The yields and their uncertainties are evaluated after the background-only fit to the data. The expected Z0TC2 signal yields with masses of 1.5 and 3 TeV are calculated using the μ ¼ 1 hypothesis. The multijet contribution also contains all other small non-tt̄ backgrounds. Type SR1ðMedium; 1bÞ SR1ðMedium; 2bÞ SR1ðTight; 1bÞ SR1ðTight; 2bÞ tt̄ 320  50 930  50 440  70 1350  70 Multijet (template) 1360  60 810  50 510  70 330  50 Total background 1680  40 1740  40 950  30 1680  50 Data 1689 1730 952 1676 Z0TC2 ð1.5 TeVÞ 100  20 280  20 150  20 460  30 Z0TC2 ð3 TeVÞ 41 81 41 81 Type SR2ðMedium; 1bÞ SR2ðMedium; 2bÞ SR2ðTight; 1bÞ SR2ðTight; 2bÞ tt̄ 250  40 690  60 190  30 510  40 Multijet (template) 2760  60 1640  70 820  50 510  50 Total background 3010  50 2330  50 1010  30 1020  30 Data 3006 2322 989 1021 Z0TC2 ð1.5 TeVÞ 80  10 210  20 70  10 190  20 Z0TC2 ð3 TeVÞ 41 61 31 51 TABLE VII. The relative impact of the post-fit uncertainties on the signal strength parameter μ using the Z0TC2 benchmark model with m ¼ 0.75 ð3Þ TeV in the resolved (boosted) analysis. The eight systematic uncertainties with the highest impact on the signal strength parameter in the resolved and boosted analyses, respectively, are shown. The uncertainty on the extrapolation using an exponential function at high mreco tt̄ above 1.5 TeV applies to the boosted analysis only. To estimate the impact from a given source of systematic uncertainty, the fit is performed with the nuisance parameter for the test fixed to the 1σ value after the nominal fit and the other nuisance parameters floated. The differences between the best-fit μ values in the tests and the nominal fit are divided by total post-fit uncertainty in μ are shown in this table. The total systematic uncertainty is different from the sum in quadrature of the different components due to correlations between nuisance parameters built by the fit. The statistical uncertainty in the data is evaluated by fixing all the nuisance parameters in the fit to the best-fit values except for the free-floating normalization factors. Resolved (Z0TC2 m ¼ 0.75 TeV) Boosted (Z0TC2 m ¼ 3 TeV) Source of uncertainty Relative impact on μ Relative impact on μ Luminosity <0.01 þ0.03=− 0.03 b-tagging efficiency þ0.05= − 0.04 þ0.07=− 0.07 Small- and large-R JES and JER þ0.20= − 0.24 þ0.21=− 0.09 tt̄ modeling þ0.34= − 0.33 þ0.10=− 0.09 Multijet estimation þ0.25=− 0.27 þ0.16=− 0.13 Extrapolation  þ0.34=− 0.33 PDF þ0.07=− 0.08 þ0.10=− 0.10 Pileup reweighting þ0.07=− 0.05 <0.01 Simulation statistical uncertainty 0.41  Total systematic uncertainty 0.92 0.67 Data statistical uncertainty 0.39 0.74 092004-18 SEARCH FOR HEAVY PARTICLES DECAYING INTO A … PHYS. REV. D 99, 092004 (2019) limits on the cross-section times branching fraction of new respectively. The systematic uncertainties with the largest particles decaying into tt̄. post-fit impact on the signal strength parameter μ in the resolved and boosted analyses are presented in Table VII. The observed data agree well with the estimated SM IX. RESULTS background and no significant excess is observed. The observed mreco distributions in the regions A–D for tt̄ Assuming a narrow-width resonance modeled by the the resolved analysis and in the signal regions SR1 and SR2 Z0TC2 signal, the minimum local p0 value is observed in for the boosted analysis after the fit (Post-Fit) with the the boosted analysis to be 0.02 (2.1σ) at m ¼ 1.75 TeV. The background-only hypothesis are shown in Figs. 6–8, respec- observed excess corresponds to a global significance tively. The expected signal and background yields as well as of less than 1σ. While the excess is mostly driven by the observed number of data events are summarized in SR1ðTight; 2bÞ region, it is worth noting that the other Tables V and VI for the resolved and boosted analyses, regions contribute significantly to the overall sensitivity, e.g., adding the SR2 regions can improve the sensitivity by up to 20% (for a 3 TeV signal) and adding the 1b regions to 105 ATLAS s = 13 TeV, 36.1 fb-1 the 2b ones adds about 10% more sensitivity. The data and 104 Observed 95% CL upper limit expected background spectra are also compared using Expected 95% CL upper limit 103 Expected 95% CL upper limit ± 1 σ BumpHunter [88], which performs a hypothesis test to look Expected 95% CL upper limit ± 2 σ 102 LO Z’TC2 Γ =1.2% cross−section× 1.3 for local excesses or deficits in data relative to the back- σ × B [pb] NLO Z’TC2 Γ =3% cross−section ground, taking the look-elsewhere effect into account as 10 NLO Z’TC2 Γ =1% cross−section well. No significant deviation from the background is found. 1 In the absence of a significant excess above the back- −1 10 ground prediction, 95% C.L. upper limits on the cross- Resolved section times branching fraction of new particles decaying Boosted 10−2 into tt̄ are calculated at each mass value for the different 10−3 benchmark signal models considered. The expected and 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 mZ’ [TeV] observed upper limits on the cross-section times branching fraction of Z0TC2 → tt̄ are presented in Fig. 9. Due to the FIG. 9. Observed and expected upper limits on the cross- strength of the expected limits, results from the resolved section times branching fraction of Z0TC2 decaying into tt̄ as a analysis are shown at mZ0TC2 below 1.2 TeV, whereas the function of the Z0TC2 mass. The theory predictions of the cross sections for the Z0TC2 with Γ ¼ 1% and 3% are shown by the results of the boosted analysis are shown above that value. dotted and dashed lines at NLO and by the solid line with Γ ¼ The NLO theory cross-section predictions for the Z0TC2 with 1.2% at LO, respectively. The results from the resolved and Γ ¼ 1% and 3%, as well as those at LO with Γ ¼ 1.2% are boosted analyses are shown to the left and right of the vertical overlaid. The observed (expected) 95% C.L. exclusion range dashed line, respectively. is set for the Z0TC2 masses between 0.58 and 3.1 TeV 105 105 ATLAS s = 13 TeV, 36.1 fb-1 ATLAS s = 13 TeV, 36.1 fb-1 104 Observed 95% CL upper limit 104 Observed 95% CL upper limit Expected 95% CL upper limit Expected 95% CL upper limit 103 Expected 95% CL upper limit ± 1 σ 103 Expected 95% CL upper limit ± 1 σ Expected 95% CL upper limit ± 2 σ Expected 95% CL upper limit ± 2 σ 102 Dark Matter, Axial-vector mediator 102 Dark Matter, Vector mediator σ × B [pb] σ × B [pb] 10 10 1 1 10−1 10−1 Resolved Resolved Boosted Boosted 10−2 10−2 10−3 10−3 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Mediator mass [TeV] Mediator mass [TeV] (a) (b) FIG. 10. Observed and expected 95% C.L. upper limits on the cross-section times branching fraction of Z0med decaying into tt̄ as a function of the Z0med mass. The theoretical predictions of the cross sections for the Z0med in the (a) A1 axial-vector mediator and (b) V1 vector mediator scenarios of the benchmark DM models are shown by the solid lines. The resolved and boosted analyses are shown to the left and right of the vertical dashed line, respectively. 092004-19 M. AABOUD et al. PHYS. REV. D 99, 092004 (2019) 105 and 2.0 TeV < mZ0med;vec < 2.1 TeV (1.99 TeV < mZ0med;ax < 104 ATLAS s = 13 TeV, 36.1 fb-1 Observed 95% CL upper limit 2.04 TeV). The upper limit on the cross-section times 103 Expected 95% CL upper limit Expected 95% CL upper limit ± 1 σ branching fraction of the GKK in the bulk RS model 102 Expected 95% CL upper limit ± 2 σ LO GKK cross−section is shown in Fig. 11. The cross-section times branching σ × B [pb] 10 fraction for GKK production with the model parameters 1 described in Sec. II is too low to be excluded with the 10−1 sensitivity of this measurement, hence the limit is presented only up to 3 TeV. Figure 12 shows the upper limit on Resolved 10−2 Boosted the cross-section times branching fraction of the gKK with 10−3 Γ ¼ 30% in the RS model with a single warped 10−4 extra dimension. The observed and expected lower limits 0.5 1 1.5 2 2.5 3 mG [TeV] KK on the gKK mass are 3.4 and 3.3 TeV, respectively. The exclusion limit is also extracted for the gKK as a FIG. 11. Observed and expected 95% C.L. upper limits function of the width at representative mass values. on the cross-section times branching fraction of GKK decaying Figures 13(a)–13(e) show the results for mgKK ¼ 0.5, 1.0, into tt̄ as a function of the GKK mass. The theoretical prediction 1.5, 2.0 and 5.0 TeV, respectively. For mgKK > 0.5 TeV, the of the cross section for the GKK in the bulk RS model with limits on the cross-section times branching fraction deterio- k=M̄ Pl ¼ 1.0 is shown by the solid line. The resolved and rate with increasing gKK width as the signal peak of the boosted analyses are shown to the left and right of the vertical dashed line, respectively. reconstructed mreco tt̄ distribution becomes broad. The limit at mgKK ¼ 0.5 TeV does not depend on the signal width since the events with reconstructed mrecott̄ < 0.5 TeV are covered (0.57 and 2.8 TeV) and 0.53 and 3.6 TeV (0.51 and 3.6 TeV) by one bin, as shown in Fig. 6. for Γ ¼ 1% and 3%, respectively. Limits are also set on the The extracted lower limits on the masses for various cross-section times branching fraction of the vector and signal hypotheses where the sensitivity of the analysis axial-vector mediators Z0med in the simplified DM model, as allows for it are summarized in Table VIII. shown in Fig. 10. The vector (axial-vector) mediator Z0med is excluded in the mass ranges of 0.74 TeV < X. CONCLUSION mZ0med;vec < 0.97 TeV and 2.0 TeV < mZ0med;vec < 2.2 TeV A search for resonant production of tt̄ decaying into the −1 (0.80 TeV < mZ0med;ax < 0.92 TeV and 2.0 TeV < mZ0med;ax < pffiffiffi using 36.1 fb of fully hadronic final state is performed 2.2 TeV) at 95% C.L. by the data with the corresponding pp collision data recorded at s ¼ 13 TeV with the expected mass ranges of 0.75 TeV < mZ0med;vec < 1.07 TeV ATLAS detector at the LHC. Depending on the mass of new hypothetical particles, the search exploits two analysis techniques optimized for the reconstruction of a top-quark pair and background suppression. No significant deviation 105 from the Standard Model expectation is observed over the ATLAS s = 13 TeV, 36.1 fb-1 104 Observed 95% CL upper limit search range considered. Upper limits are set on the 103 Expected 95% CL upper limit production cross-section times branching fraction for sev- Expected 95% CL upper limit ± 1 σ Expected 95% CL upper limit ± 2 σ eral benchmark signals, such as Z0TC2 boson predicted in the 102 LO g cross−section topcolor-assisted-technicolor model, vector and axial-vec- σ × B [pb] KK 10 tor mediators Z0med in the dark-matter simplified model, and 1 the Kaluza-Klein excitations of the graviton GKK and gluon −1 gKK in the specific models based on the Randall-Sundrum 10 scenario of warped extra dimensions. The Z0TC2 boson is Resolved Boosted 10−2 excluded in the mass range of 0.58 and 3.1 TeV (0.53 and 10−3 3.6 TeV) for the decay width of 1% (3%). The vector 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 mg [TeV] (axial-vector) mediator Z0med is excluded in the mass KK ranges of 0.74 TeV < mZ0med;vec < 0.97 TeV and 2.0TeV< FIG. 12. Observed and expected 95% C.L. upper limits on the mZ0med;vec <2.2TeV (0.80TeV