Analysis of tt̄H and tt̄W production in multilepton final states with the ATLAS detector

Content	Preview
Analysis of tt̄H and tt̄W production in multilepton final states with the ATLAS detector ATLAS-CONF-2019-045 16 October 2019
Main document (CDS record) - internal	pdf from CDS
Figures Tables	-
Abstract
A search for the associated production of a top-quark pair with the Higgs boson ($t\bar{t}H$) in multilepton final states is presented. The search is based on a dataset of proton--proton collisions at $\sqrt{s}=13 $TeV recorded with the ATLAS detector at the CERN Large Hadron Collider and corresponding to an integrated luminosity of 80 fb$^{-1}$. Six final states, defined by the number and flavour of charged-lepton candidates, and 25 event categories are defined to simultaneously search for the \ttH signal and constrain several leading backgrounds. The $t\bar{t}W$ background normalisation is left unconstrained in the statistical analysis and the resulting $t\bar{t}W$ normalisation is found to be higher than the theoretical prediction. An excess of events consistent with $t\bar{t}H$ production, over the expected background from Standard Model processes, is found with an observed significance of 1.8 standard deviations, compared to an expectation of 3.1 standard deviations. Assuming Standard Model branching fractions, the best-fit value of the $t\bar{t}H$ production cross section is $\sigma_{t\bar{t}H} = 294^{+182}_{-162}$ fb, which is consistent with the Standard Model prediction. The impact on the $t\bar{t}H$ cross section measurement of the assumptions made on the $t\bar{t}W$ background modelling is discussed.
Figures
Figure 01a: Comparison between data and signal-plus-background prediction for the distribution of jet multiplicity in (a) the 2ℓSS channel and (b) the 3ℓ channel after event selection and before further event categorisation (see Section 5). The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total signal-plus-background prediction before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the signal-plus-background prediction is indicated by the blue hatched band. The ratios of the data to the total pre- and post-fit predictions are shown in the lower panel. The last bin in each figure contains the overflow. png (279kB) eps (2MB) pdf (51kB)
Figure 01b: Comparison between data and signal-plus-background prediction for the distribution of jet multiplicity in (a) the 2ℓSS channel and (b) the 3ℓ channel after event selection and before further event categorisation (see Section 5). The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total signal-plus-background prediction before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the signal-plus-background prediction is indicated by the blue hatched band. The ratios of the data to the total pre- and post-fit predictions are shown in the lower panel. The last bin in each figure contains the overflow. png (267kB) eps (1MB) pdf (74kB)
Figure 02a: Comparison between data and signal-plus-background prediction for the event yields in (a) the 2ℓSS channel and (b) the 3ℓ channel after event selection and before further event categorisation (see Section 5), split in four separate categories depending on the total charge and b-jet multiplicity. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total signal-plus-background prediction before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the signal-plus-background prediction is indicated by the blue hatched band. The ratios of the data to the total pre- and post-fit predictions are shown in the lower panel. png (148kB) eps (124kB) pdf (38kB)
Figure 02b: Comparison between data and signal-plus-background prediction for the event yields in (a) the 2ℓSS channel and (b) the 3ℓ channel after event selection and before further event categorisation (see Section 5), split in four separate categories depending on the total charge and b-jet multiplicity. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total signal-plus-background prediction before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the signal-plus-background prediction is indicated by the blue hatched band. The ratios of the data to the total pre- and post-fit predictions are shown in the lower panel. png (152kB) eps (120kB) pdf (36kB)
Figure 03a: (a) Comparison between data and prediction for the event yields in the 3ℓ validation region enriched in Z → μ⁺μ^- γ^* (→ e⁺e^-) candidate events, divided intro three categories depending on the requirements satisfied by the electron: internal conversion, material conversion, or very tight. The comparisons are made after correcting the simulation by the normalisation factors for internal conversions and material conversions resulting from the likelihood fit to data in all categories ("Post-Fit"). (b) Comparison between data and prediction for the distribution of the scalar sum of the lepton p_T (H_{T, lep}) in the 2ℓSS tt̄ control region at low jet multiplicity with a sub-leading muon, after performing the likelihood fit to data in all categories ("Post-Fit"). The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. In both figures the size of the combined statistical and systematic uncertainty in the total prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. The last bin in each figure contains the overflow. png (120kB) eps (673kB) pdf (13kB)
Figure 03b: (a) Comparison between data and prediction for the event yields in the 3ℓ validation region enriched in Z → μ⁺μ^- γ^* (→ e⁺e^-) candidate events, divided intro three categories depending on the requirements satisfied by the electron: internal conversion, material conversion, or very tight. The comparisons are made after correcting the simulation by the normalisation factors for internal conversions and material conversions resulting from the likelihood fit to data in all categories ("Post-Fit"). (b) Comparison between data and prediction for the distribution of the scalar sum of the lepton p_T (H_{T, lep}) in the 2ℓSS tt̄ control region at low jet multiplicity with a sub-leading muon, after performing the likelihood fit to data in all categories ("Post-Fit"). The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. In both figures the size of the combined statistical and systematic uncertainty in the total prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. The last bin in each figure contains the overflow. png (154kB) eps (166kB) pdf (43kB)
Figure 04a: Comparison between data and signal-plus-background prediction in the 2ℓSS channel after event selection for (a) the event yield, split in four separate categories depending on the flavour of the sub-leading lepton and the b-jet multiplicity, and (b) the score of the BDT trained to discriminate tt̄H signal from tt̄ background. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (129kB) eps (129kB) pdf (40kB)
Figure 04b: Comparison between data and signal-plus-background prediction in the 2ℓSS channel after event selection for (a) the event yield, split in four separate categories depending on the flavour of the sub-leading lepton and the b-jet multiplicity, and (b) the score of the BDT trained to discriminate tt̄H signal from tt̄ background. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (155kB) eps (220kB) pdf (63kB)
Figure 05a: Comparison between data and prediction for the event yields in (a) the eight tt̄H categories and (b) the 17 control-region categories. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (136kB) eps (2MB) pdf (113kB)
Figure 05b: Comparison between data and prediction for the event yields in (a) the eight tt̄H categories and (b) the 17 control-region categories. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (120kB) eps (2MB) pdf (100kB)
Figure 06a: Comparison between data and prediction for the BDT discriminants used in different tt̄H categories: (a) 2ℓttH+, (b) 2ℓttH-, (c) 3ℓttH, and (d) 1ℓ2τ_had. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit is shown as a dashed blue histogram. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (138kB) eps (135kB) pdf (36kB)
Figure 06b: Comparison between data and prediction for the BDT discriminants used in different tt̄H categories: (a) 2ℓttH+, (b) 2ℓttH-, (c) 3ℓttH, and (d) 1ℓ2τ_had. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit is shown as a dashed blue histogram. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (133kB) eps (135kB) pdf (36kB)
Figure 06c: Comparison between data and prediction for the BDT discriminants used in different tt̄H categories: (a) 2ℓttH+, (b) 2ℓttH-, (c) 3ℓttH, and (d) 1ℓ2τ_had. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit is shown as a dashed blue histogram. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (145kB) eps (154kB) pdf (40kB)
Figure 06d: Comparison between data and prediction for the BDT discriminants used in different tt̄H categories: (a) 2ℓttH+, (b) 2ℓttH-, (c) 3ℓttH, and (d) 1ℓ2τ_had. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit is shown as a dashed blue histogram. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (140kB) eps (148kB) pdf (37kB)
Figure 07: Event yields as a function of log₁₀(S/B) for data, background and a Higgs boson signal with m_H = 125 GeV. The discriminant bins in all categories are combined into bins of log₁₀(S/B), where S is the expected tt̄H signal yield and B the background yield from the unconditional fit. The background yields are shown as the fitted values, while the tt̄H signal yields are shown for the fitted value (μ=0.58) and the SM prediction (μ=1). The total background before the fit is shown as a dashed blue histogram. The size of the combined statistical and systematic uncertainty in the background prediction is indicated by the blue hatched band. The pull (residual divided by its uncertainty) of the data relative to the background-only prediction is shown in the lower panel, where the full red line (dashed orange line) indicates the pull of the prediction for signal with μ=0.58 (μ=1) and background relative to the background-only prediction. png (150kB) pdf (48kB)
Figure 08: The observed best-fit values of the tt̄H signal strength μ and their uncertainties by analysis channel and combined. The individual μ values for the channels are obtained from a simultaneous fit with the signal-strength parameter for each channel floating independently. The SM prediction corresponds to μ=1. png (114kB) pdf (36kB)
Figure 09: Pre-fit S/B (black line) and S/√ B (red dashed line) ratios for each analysis category. The background prediction methods are described in Section 6. png (47kB) eps (2MB) pdf (122kB)
Figure 10: The fractional contributions of the various backgrounds to the total predicted background in each of the 25 event categories. The background estimation methods are described in Sect. 6: "Non-prompt (e/μ)", "Mat Conv", and "QMisID" refer to the data-driven background estimates (largely tt̄ but also include other electroweak processes), and rare processes (tZ, tW, WtZ, tt̄WW, VVV, tt̄t, tt̄tt̄, and tH) are labelled as "Other". png (234kB) pdf (92kB)
Figure 11a: The distributions of the kinematic variables used in the four binned control regions: (a) Δ R(ℓ,ℓ) in the 2ℓLJ(e1) control region, (b) and (c) scalar sum of the lepton p_T (H_{T, lep}) in the 2ℓLJ(e2) control region (see Figure 3(b) for the corresponding distribution in the 2ℓLJ(μ) control regions), and (d) BDT score for the tt̄ category in the 3ℓtt control region. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H (tt̄W) signal, scaled according to the results of the fit, is shown as a filled red (yellow) histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. The last bin in each figure contains the overflow. png (156kB) eps (160kB) pdf (41kB)
Figure 11b: The distributions of the kinematic variables used in the four binned control regions: (a) Δ R(ℓ,ℓ) in the 2ℓLJ(e1) control region, (b) and (c) scalar sum of the lepton p_T (H_{T, lep}) in the 2ℓLJ(e2) control region (see Figure 3(b) for the corresponding distribution in the 2ℓLJ(μ) control regions), and (d) BDT score for the tt̄ category in the 3ℓtt control region. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H (tt̄W) signal, scaled according to the results of the fit, is shown as a filled red (yellow) histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. The last bin in each figure contains the overflow. png (149kB) eps (170kB) pdf (44kB)
Figure 11c: The distributions of the kinematic variables used in the four binned control regions: (a) Δ R(ℓ,ℓ) in the 2ℓLJ(e1) control region, (b) and (c) scalar sum of the lepton p_T (H_{T, lep}) in the 2ℓLJ(e2) control region (see Figure 3(b) for the corresponding distribution in the 2ℓLJ(μ) control regions), and (d) BDT score for the tt̄ category in the 3ℓtt control region. The background contributions after the likelihood fit ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H (tt̄W) signal, scaled according to the results of the fit, is shown as a filled red (yellow) histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. The last bin in each figure contains the overflow. png (148kB) eps (158kB) pdf (41kB)
Figure 12: Observed correlations between the signal strength μ and the normalisation factors for the tt̄W background in the profile likelihood fit to the data. png (144kB) pdf (54kB)
Figure 13: Ranking of the parameters included in the fit according to their impact on the signal strength μ. Only the 15 most highly ranked parameters are shown. The empty blue rectangles correspond to the pre-fit impact on μ and the filled blue ones to the post-fit impact on μ, both referring to the upper x-axis scale. The impact of each nuisance parameter (NP), Δμ, is computed by comparing the nominal best-fit value of μ with the result of the fit when fixing the considered NP to its best-fit value, hatθ, shifted by its pre-fit (post-fit) uncertainties ±Δθ (±Δhatθ). The black points show the pulls of the NPs relative to their nominal values, θ₀. The nominal value for all NPs is θ₀=0, with the exception of the NP associated with the limited sample size in the estimation of the fake tauhad background in the 1ℓ2tau_had channel, for which the nominal value is θ₀=1. These pulls and their relative post-fit errors, Δhatθ/Δθ, refer to the lower x-axis scale. The tt̄W normalisation factors (red points) refer to the floating normalisations of the tt̄W background, for which the pre-fit impact on μ is not defined. The nominal value of the tt̄W normalisation factors is 1, corresponding to the tt̄W prediction as discussed in Section 3. For experimental uncertainties that are decomposed into several independent sources, "NP I" corresponds to the first nuisance parameter, ordered by its impact on μ. png (802kB) pdf (924kB)
Figure 14: Comparison between data and prediction for the event yields in the categories in the 2ℓSS and 3ℓ channels from the cut-and-count cross-check analysis. The background contributions after the likelihood fit in the cut-and-count analysis ("Post-Fit") are shown as filled histograms. The total background before the fit ("Pre-Fit") is shown as a dashed blue histogram. The tt̄H signal, scaled according to the results of the fit, is shown as a filled red histogram added to the post-fit background. The size of the combined statistical and systematic uncertainty in the total signal-plus-background prediction is indicated by the blue hatched band. The ratio of the data to the total post-fit prediction is shown in the lower panel. png (190kB) eps (65kB) pdf (26kB)
Tables
Table 01: The configurations used for the event generation of signal and background processes. The samples used to estimate the systematic uncertainties are indicated in parentheses. V refers to production of an electroweak boson (W or Z/γ^*). The matrix element (ME) order refers to the order in the strong coupling constant of the perturbative calculation. If only one parton distribution function (PDF) is shown, the same one is used for both the ME and parton shower generators; if two are shown, the first is used for the ME calculation and the second for the parton shower. Tune refers to the underlying-event tune of the parton shower generator. MG5_aMC refers to MadGraph5_aMC@NLO 2.2.X or 2.3.X [38]; Pythia 6 refers to version 6.427 [39]; Pythia 8 refers to version 8.2 [26]; Herwig++ refers to version 2.7 [40]; Herwig7 refers to version 7.0.4 [36]; MePs@Nlo refers to the method used in Sherpa [41,42,43,44,45] to match the matrix element to the parton shower. Samples using Pythia 6 or Pythia 8 have heavy flavour hadron decays modelled by EvtGen 1.2.0 [46]. All samples include leading-logarithm photon emission, either modelled by the parton shower generator or by PHOTOS [47]. The masses of the top quark and SM Higgs boson were set to 172.5 GeV and 125 GeV. png (56kB) pdf (64kB)
Table 02: The requirements applied to select loose (L), loose and minimally-isolated (L), tight (T) and very tight (T) light leptons. png (21kB) pdf (55kB)
Table 03: Offline selection criteria applied to the channels. The common selection criteria for all channels are listed in the first line under the title "Common". Same-charge (opposite-charge) lepton pairs are also referred to as same-sign (opposite-sign) with abbreviation SS (OS). Same-flavour (SF), OS lepton pairs are referred to as SFOS pairs. In the categories for conversions the selection requirements on one of the leptons are loosened as discussed in Section 4. png (79kB) pdf (73kB)
Table 04: Sources of systematic uncertainty considered in the analysis. "N" means that the uncertainty is taken as normalisation-only for all processes and channels affected. Some of the systematic uncertainties are split into several components, as indicated by the number in the rightmost column. The PDF uncertainties are correlated between the tt̄H signal and the tt̄W background. png (328kB) pdf (63kB)
Table 05: Breakdown of the contributions to the uncertainties in hatμ. The contribution of the different sources of uncertainty is evaluated after the fit described in Section 7. The total statistical uncertainty is evaluated, as described in the text, by fixing all the nuisance parameters in the fit except for the free-floating background normalisation factors. The contribution from the uncertainty in those normalisation factors is then included in the quoted total statistical uncertainty rather than in the systematic uncertainty component. The statistical uncertainty evaluated after also fixing the background normalisation factors is then indicated as "intrinsic statistical uncertainty". Statistical uncertainties from data-driven background estimates are included within the experimental uncertainties. The other quoted numbers are obtained by repeating the fit after having fixed a certain set of nuisance parameters corresponding to a group of systematic uncertainty sources, and subtracting in quadrature the resulting total uncertainty of μ from the uncertainty from the full fit. The same procedure is followed for quoting the individual effects of background normalisation factors. Due to rounding effects and small correlations between the different sources of uncertainty, the total systematic uncertainty is different from the sum in quadrature of the individual sources. png (44kB) pdf (53kB)
Table 06: Offline selection criteria applied to the 2ℓSS and 3ℓ channels in the cut-and-count analysis, together with the event categories defined. The common selection criteria for all channels are listed in the first line under the title "Common". Same-charge (opposite-charge) lepton pairs are also referred to as same-sign (opposite-sign) with abbreviation SS (OS). In the 2ℓSS channel the leading and trailing (in p_T) SS leptons are denoted ℓ₀ and ℓ₁ respectively. In the 3ℓ channel, the OS lepton (w.r.t the SS pair) is denoted ℓ₀, but is not necessarily the one with highest p_T; the remaining SS leptons are denoted ℓ₁ (closest in Δ R to ℓ₀) and ℓ₂ (the remaining one). Same-flavour (SF), OS lepton pairs are referred to as SFOS pairs. png (114kB) pdf (215kB)

© 2019 CERN for the benefit of the ATLAS Collaboration. Reproduction of the article, figures and tables on this page is allowed as specified in the CC-BY-4.0 license.
2024-05-19 01:28:55