Search for heavy Higgs bosons A / H decaying to a top quark pair in pp collisions at √ s = 8 TeV with the ATLAS detector

A search for heavy pseudoscalar ( A ) and scalar ( H ) Higgs bosons decaying into a top quark pair ( t ¯ t ) has been performed with 20.3 fb − 1 of proton–proton collision data collected by the ATLAS experiment at the Large Hadron Collider at a center-of-mass energy of √ s = 8 TeV. Interference e ﬀ ects between the signal process and Standard Model t ¯ t production, which are expected to distort the signal shape from a single peak to a peak–dip structure, are taken into account. No signiﬁcant deviation from the Standard Model prediction is observed in the t ¯ t invariant mass spectrum in ﬁnal states with an electron or muon, large missing transverse momentum, and at least four jets. The results are interpreted within the context of a type-II two-Higgs-doublet model. Exclusion limits on the signal strength are derived as a function of the mass m A / H and the ratio of the vacuum expectation values of the two Higgs ﬁelds, tan β , for m A / H > 500 GeV.

Search for heavy Higgs bosons A/H decaying to a top quark pair in pp collisions at √ s = 8 TeV with the ATLAS detector

ATLAS Collaboration
A search for heavy pseudoscalar (A) and scalar (H) Higgs bosons decaying into a top quark pair (tt) has been performed with 20.3 fb −1 of proton-proton collision data collected by the ATLAS experiment at the Large Hadron Collider at a center-of-mass energy of √ s = 8 TeV. Interference effects between the signal process and Standard Model tt production, which are expected to distort the signal shape from a single peak to a peak-dip structure, are taken into account. No significant deviation from the Standard Model prediction is observed in the tt invariant mass spectrum in final states with an electron or muon, large missing transverse momentum, and at least four jets. The results are interpreted within the context of a type-II two-Higgs-doublet model. Exclusion limits on the signal strength are derived as a function of the mass m A/H and the ratio of the vacuum expectation values of the two Higgs fields, tan β, for m A/H > 500 GeV. c 2017 CERN for the benefit of the ATLAS Collaboration. Reproduction of this article or parts of it is allowed as specified in the CC-BY-4.0 license. The signal process gg → A/H → tt, including the decays of the top quarks and resulting W bosons, was simulated using MadGraph5_aMC@NLO [30] v2. 3.3 with the model of Ref. [31], which implements the production of (pseudo)scalar particles through loop-induced gluon-gluon fusion with loop contributions from top and bottom quarks at leading order (LO) in QCD. The CT10 set [32] of parton distribution functions (PDFs) was used and the renormalization and factorization scales were set to decay products (p 2 T + m 2 ). For the statistical interpretation, the tt invariant mass distributions in the signal regions in data were compared to a combination of the expected distributions from all background processes B, the pure signal process S , and the signal-plus-interference component S + I for a given signal hypothesis, as illustrated in Eq. (1) below. The most reliable description of the tt background [33] is obtained at next-to-leading order (NLO) with Powheg-Box [34][35][36][37] + Pythia6 [38]. Therefore, the S + I contribution was modeled separately from this background process by modifying the MadGraph5_aMC@NLO software to remove the pure SM tt process to yield only the S + I contribution on an event-by-event basis. The S + I events obtained with the modified software can have positive or negative weights. Figure 1 shows the tt invariant mass distributions for the S and S + I components in a model with tan β = 0.68 and a pseudoscalar of mass m A = 500 GeV (m A m H ). The S + I component exhibits a peak-dip structure with the minimum around m A/H for all signal hypotheses studied in this search. The width of both the S and S +I distribution decreases with increasing tan β.
The S + I distributions from the modified MadGraph5_aMC@NLO software were validated against those from the unmodified program. The latter were obtained by generating a large inclusive sample S + I + B tt for a given parameter point and a LO SM tt background B tt sample with the same generator settings. The event selection and reconstruction algorithms were applied separately to each sample. The difference between the resulting two m tt distributions corresponds to the S + I component, which agrees with that obtained with the modified software within 0.4% across the whole spectrum. The difference is taken as a systematic uncertainty in S + I.
Pythia6 with the Perugia 2011c set of tuned parameters [39] was used to model the parton shower and hadronization for all S and S + I samples and the stable particles obtained after hadronization were passed through the ATLAS fast detector simulation [40]. The effects of additional collisions within the same or nearby bunch crossings were simulated by overlaying additional pp collisions, simulated with Pythia v8.1 [41], on each event. Correction factors were applied to adjust the trigger and selection efficiencies in simulated events to those measured in data. The S and S + I samples with this setup were generated separately for pseudoscalar and scalar Higgs bosons.
Event samples for both the S and S + I components for different values of (m A/H , tan β) were obtained from signal samples S after the detector simulation by applying an event-by-event reweighting. This reweighting substantially reduces the computing time required. The weight is the ratio of the Mad-Graph5_aMC@NLO matrix elements, calculated from the four-momenta of the incoming gluons and outgoing top quarks of the generated event with the new and the old values of (m A/H , tan β), respectively. All S + I and a small number of S samples were obtained through reweighting. Signal hypotheses with m A/H < 500 GeV were not considered as they require an accurate modeling of the Higgs boson decay into virtual top quarks and the implementation of higher-order corrections that are not available in the MadGraph5_aMC@NLO model. The requirement tan β ≥ 0.4 was imposed to ensure that all amplitudes involving scalars preserve perturbative unitarity [25,42].
Correction factors K S were applied to normalize the generated signal (S ) cross-section to the value calculated at partial next-to-next-to-leading-order (NNLO) precision in QCD [43][44][45]. The correction factor for the interference component I is K I = √ K S × K B , as suggested in Ref. [46], where K B = 1.87 is the correction factor to normalize the total cross-section of the SM tt background generated at LO with MadGraph to the cross-section calculated at NNLO accuracy in the strong coupling constant α S , including resummation of next-to-next-to-leading-logarithmic soft gluon terms. The values of K S range between two and three for the tested signal hypotheses.
The event selection criteria for the signal regions provide a high selection efficiency for tt events in the +jets channel. Only events with a resolved topology, in which the three jets from the hadronically decaying top quark are well separated in the detector, are selected. This is the most efficient selection strategy for signal hypotheses with m A/H < 800 GeV. Events with a merged topology, in which the hadronically decaying top quark is reconstructed as a single large radius jet, are not considered. The event reconstruction and selection criteria are identical to those for the resolved topology in Ref.
[17] except that events that would satisfy the criteria for both topologies are classified as "resolved" instead of "merged".
Events are required to contain exactly one isolated electron or muon that is geometrically matched to the corresponding trigger-level signature. applied to jets with p T < 50 GeV and |η| < 2.4. At least one of the jets must be identified as originating from the decay of a b-hadron (b-jet) using a multivariate tagging algorithm with a 70% efficiency for b-jets [54].
Jets are assigned to the top quarks using a χ 2 algorithm that relies on kinematic constraints and the expected values of the top quark and W boson masses [17]. The invariant mass m reco tt of the candidate tt pair is reconstructed from the four selected jets, the lepton, and the E miss T vector. The experimental resolution for the tt invariant mass is 8% for a resonance mass of 500 GeV. Events in the e+jets and µ+jets channels are further classified into three orthogonal categories, based on whether a b-tagged jet was assigned to either the hadronically or the semileptonically decaying top quark, or to both of them. Each category defines a signal region, hence six independent signal regions are used in the statistical analysis.
The systematic uncertainties can be divided into experimental and modeling uncertainties. The impact on both the normalization and the shape of the m reco tt distributions is taken into account for all systematic uncertainties. The average impact of the dominant uncertainties on the event yields is summarized in Table 1. Table 1: Average impact of the dominant uncertainties on the estimated yields for the total background and for a pseudoscalar A with m A = 500 GeV and tan β = 0.68 in percent of the nominal value. The spectra of the e+jets and µ+jets channels are added. Only uncertainties with a yield impact > 0.5% are shown. A bar (−) indicates that an uncertainty is not applicable to a sample.

Systematic uncertainties [%]
Total bkg S S + I Luminosity [55] 1.7 1.9 1.9 PDF 2.5 2. A detailed breakdown of the observed and expected event yields in the e+jets and µ+jets channels and their associated total uncertainties is shown in Table 2. Good agreement is found between the observed number of events in data and the expected total number of background events.
The fitted variable is √ µ and the case µ = 1 corresponds to the type-II 2HDM in the alignment limit, while the case µ = 0 corresponds to the background-only hypothesis. This approach relies on the assumption that, for a given signal hypothesis, the shape of the tt invariant mass distributions for S and S + I in Eq. (1) does not change if the signal strength is varied. The terms S and S + I on the right-hand side of Eq. (1) correspond to the m reco tt distributions obtained from the S and S + I samples, respectively, while B stands for the expected m reco tt distribution of the total SM background.
The level of agreement between the observed and expected mass spectra is quantified in a fit under the background-only hypothesis with µ = 0 in which only the nuisance parameters are allowed to vary. The observed m reco tt spectra are compatible with the expected spectra after the background-only fit within the (constrained) uncertainty bands, as shown in Figure 2.
The upper limits on µ at 95% confidence level (CL) are obtained with the CL s method [60] for a number of (m A/H , tan β) values. The upper limits at intermediate points are obtained from a linear interpolation among the three closest points. In Figure 3, the observed and expected exclusion regions for the type-II 2HDM (µ = 1) are shown for the three mass hierarchies discussed in the introduction. The excluded values of tan β for the different mass hypotheses are listed in Table 3.
In conclusion, the search for massive pseudoscalar and scalar resonances decaying into a top quark pair in 20.3 fb −1 of pp collisions at 8 TeV recorded by the ATLAS experiment at the LHC, yields no statistically signifiant deviations from the SM prediction. The results are interpreted in the type-II 2HDM model in the alignment limit, and upper limits are set on the signal strength µ at 95% CL in the tan β versus resonance mass plane. Unlike previous searches for massive resonances in the same final state, this analysis takes into account interference effects between the signal process and the background from SM tt production. It tightens significantly the previously published constraints on the 2HDM parameter space in the low tan β and high mass (m A/H > 500 GeV) region.
We thank CERN for the very successful operation of the LHC, as well as the support staff from our institutions without whom ATLAS could not be operated efficiently. [ [47] The ATLAS experiment uses a right-handed coordinate system with its origin at the nominal interaction point (IP) in the center of the detector and the z-axis along the beam pipe. The x-axis points from the IP to the center of the LHC ring, and the y-axis points upward. Cylindrical coordinates (r, φ) are used in the transverse plane, φ being the azimuthal angle around the z-axis.
The pseudorapidity is defined in terms of the polar angle θ as η = − ln[tan(θ/2)]. Transverse momenta are computed from the three-momenta, p, as p T = |p| sin θ.