tt̄H Coupling Measurement with the ATLAS Detector at the LHC

The Higgs boson was discovered on the 4th of July 2012 with a mass around 125 GeV by ATLAS and CMS experiments at LHC. Determining the Higgs properties (production and decay modes, couplings,...) is an important part of the high-energy physics programme in this decade. A search for the Higgs boson production in association with a top quark pair (tt̄H) at ATLAS [1] is summarized in this paper at an unexplored center-of-mass energy of 13 TeV, which could allow a first direct measurement of the top quark Yukawa coupling and could reveal new physics. The tt̄H analysis in ATLAS is divided into 3 channels according to the Higgs decay modes: H → Hadrons, H → Leptons and H → Photons. The best-fit value of the ratio of observed and Standard Model cross sections of tt̄H production process, using 2015-2016 data and combining all tt̄H final states, is 1.8±0.7, corresponds to 2.8σ (1.8σ) observed (expected) significance.


Introduction
The Higgs boson production in association with a top quark pair ttH ( Figure 1a) represents 1% of the total Higgs production making the observation of ttH very challenging with limited available statistics in Run 1. However, the expected SM ttH cross section increases by a factor of 3.9 from centre of mass energy of 8 TeV to 13 TeV. This factor, comparable to the one for the dominant backgrounds, allows to accumulate signal events at a faster rate at 13 TeV than during Run 1. Current efforts are dedicated to observe the ttH at the end of Run 2 with the full available statistics. The observation of the ttH production mode allows a direct measurement of the top Yukawa coupling (referred as g ttH ): the strongest coupling to the Higgs boson, giving that the top quark is the heaviest fundamental particle in the SM, with a mass equal to 173.21 GeV [2]. Its measurement is estimated using the ggF production ( Figure 1b) where the Higgs is indirectly coupled to the top quark via a loop but no new physics is assumed. Its value is found to be compatible with the SM expectation, using the top coupling modifier (κ t : defined as the ratio between the observed and SM top quark yukawa coupling) estimated as 0.87 ± 0.15, combining ATLAS and CMS data at Run 1 [3]. However, new physics could be hidden in the loops mediating the Higgs production via ggF. This issue could be solved by performing the measurements in ttH production mode involving a direct, tree-level, Higgs-top coupling. Therefore, a search for ttH is used to allow a first direct measurement of the top quark Yukawa coupling that could reveal new physics. It is worth to mention that all results, discussed in this paper, are obtained from . a e-mail: asma.hadef@cern.ch  Figure 1: Feynman diagram of (a) ttH and (b) ggF processes. The cross section of ttH is σ ttH = 507 fb at 13 TeV. The ttH process is tow order of magnitude lower than ggF: σ ttH /σ ggF ∼ 0.01.

Searches for ttH at ATLAS
A combination of searches for ttH production at 13 TeV has been performed by the ATLAS experiment in 3 Higgs decay channels: the bb, multilepton using 13.2 fb −1 of luminosity [4,7] and γγ using all available 2015/2016 data with luminosity of 36.1 fb −1 [5,8]. To increase the sensitivity, each channel is categorised into sub-channels according to the number of leptons and hadrons. Every channel is discussed individually starting with ttH (bb) representing the largest branching fraction.  3 ttH (bb) Analysis

Signal Region Selection
The signal region (SR) in this channel is divided into single-and di-lepton sub-channels. In singlelepton channel, exactly one lepton with at least 4 jets and at least 2 b-jets are required. In di-lepton channel, exactly 2 opposite sign (OS) leptons are required with at least 3 jets and 2 b-jets. The all hadronic channel is not treated yet at √ s = 13 TeV. Events are further categorised according to the number of jets and b-jets as illustrated in the scheme of Figure 3a where each row corresponds to a different jet-multiplicity and each column represents a different b-jet-multiplicity in the single-lepton channel. Control regions (CR), sheded in blue, are used to constraint the uncertainties.

Background Estimation
The background in ttH (bb) is dominating by tt +jets process ; divided according to the jet flavour into: tt +bjets, tt +c-jets and tt +light-jets. The pie-chart of Figure 3b shows the fractional contributions of the various backgrounds to the total background prediction in each region. The main background in the SR is tt +bb. Followed by small contribution of non-tt background: single top, diboson W/Z+jets and fake leptons. The tt Background is estimated from simulation and normalised to the prediction before the fit. The normalization of tt +b(c)-jets background are free parameters of the fit. Finally, fake lepton background in the single lepton channels is estimated using data-driven method.

Signal/Background Separation
In order to separate the signal from the background, two stage Multivariate techniques are used. First, reconstruction BDT (shown by blue in the left plot of Figure 4) is trained to match the reconstructed jets to the partons emitted from top and Higgs decays in ttH simulation (shown by red). For this 26 input variables are considered in single lepton channels, chosen to achieve optimal performance. Then, for each SR, information from the output of the reco BDTs is combined with other kinematic variables in a classification BDT (shown in the right plot of Figure 4) with 21 input variables for single lepton channels.

Results
The post-fit yields of signal and total background are compared to data where the signal is normalised to the best fit value and to the excluded value. A good agreement is seen between data and expectation in all bins within the uncertainties. The best fit value of the signal strength measurements (µ) of the signal lepton channel, obtained from a profile likelihood fit, is 1.6±1.1 where systematic uncertainties are dominating. The combination results have been removed because a problem was found in the dilepton channel.
Simulation Preliminary = 13 TeV s Single Lepton

Signal Region Selection
Events are categorised into six signal regions according to the number of light and hadronic tau leptons: 2 ss (ee, µµ and eµ), 2 ss+1τ had , 3 and 4 . Three main decays are contributing: WW * , ττ and ZZ * as shown in Table 2a. The Higgs decay into two W bosons is the dominant decay for all sub-categories except for 2 ss+1τ had where Higgs decay into two tau leptons is dominating. The ZZ * contribution is mostly νν and j j (since H → 4 is explicitly excluded in this analysis). The number of jets and b-jets differ from category to another and is summarized in Table 2b.

Background Estimation
Background is split into reducible and irreducible background as shown in Figure 5a. The reducible background is split into fake leptons (non-prompt leptons) and charge flip backgrounds. Fake leptons are coming from either mis-reconstructed objects as leptons, photon conversion or non prompt leptons decaying from heavy flavour jets. The latter is dominating in SRs. The charge flip background is mainly coming from the charge mis-reconstruction of an electron from bremsstrahlung. Reducible background is estimated using data-driven methods: the charge flip estimated from Z+jets OS events    control region and fake leptons estimated from tt SS/OS events CRs. Prompt lepton background is estimated from simulation validated using four regions. Fake leptons are the dominant background in 2 ss and 3 channels as shown in Figure 5b.

Results
The pre-fit yields of signal and total background for each category in ttH (ML) is presented in Figure  6a. The best fit value of the signal strength µ ttH for each category and for the ttH (ML) combination is presented in Figure 6b. Systematic uncertainties on combination are dominating and listed in Figure  7. Fake leptons are driving the error on the total background estimated in the 2 ss and 3 channels. Observed (expected) 95% C.L. upper limit on µ ttH is 4.9 (2.3).

Signal Region Selection
The ttH (γγ) channel has excellent di-photon mass resolution which is also sensitive to the tH channel (σ tH = 74 fb). Event are categorised according to the number of leptons and tH/ttH channels: three  leptonic channels, two of them are tH channels and six hadronic channels, 2 of them are tH channels. Cut&Count analysis is used in every category except the hadronic ttH channels where a BDT analysis is instead used. The number of jets and b-jets requirements for leptonic and hadronic channels are shown in Table 3. The signal is modeled using a double-sided Crystal Ball function where its parameters are fitted to the simulated signal samples.

Background Estimation
A data driven method is used to estimate the continuum background coming from γγ+jets, γ+jets and multi-jets. The shape of the background is build by reverting the photon identification or isolation criteria or removing b-tagging requirement. The shape is then normalised to data in the m γγ distribution outside the 120 < m γγ < 130 region. γγ and Vγ backgrounds are estimated from simulation. The total background with the fitted signal agree well with the expectation as shown in Figure 8.

Results
The best fit value of the signal strength µ ttH for ttH (γγ) analysis is obtained from a τ maximum likelihood fit for each event category and shown in the top of Figure 9. It corresponds to 1σ observed significance. Statistical uncertainties are dominating. Observed (expected) 95% C.L. upper limit on µ ttH are 1.7 (2.3).   6 ttH (ZZ * → 4 ) Analysis As mentioned in Section 4, the ttH (ZZ * → 4 ) channel is treated separately. A brief description of this analysis is presented in the following using 36.1 fb −1 of pp collision data at √ s =13 TeV. Despite the branching ratio of the Higgs decay to ZZ * → 4 is only 0.01%, this channel is privileged thanks to its clear signature and the high Signal/Background ratio. A quadruplet leptons originating from a Higgs boson decay are selected, shown by blue histogram in Figure 10a which is well separated from the resonant ZZ * → 4 background shown by the peak of the red histogram. The signal regions are classified according to the lepton flavour into 4 categories: 4e, 2e2µ, 2µ2e and 4µ. At least one b-jet is required with either at least four additional jets or one additional lepton with p T > 15 GeV and at least two jets. The main background in the analysis is the non resonant ZZ * , shown by the tail of the red histogram in Figure 10a. Other background could also contribute coming from triboson, tt +Z, tt, Z+jets and WZ. Due to the limited number of events in the control regions, the signal strength is just limited to be less than 7.5 (Figure 10b) where the fit parameters are constrained to positive values to avoid instabilities in the fit configuration. [GeV]

ttH Combination
There are two combination done at Run 2 so far: with 13.2 fb −1 data collision using ttH (bb), (ML) and (γγ) channels [7] and with 36.1 fb −1 data collision using both ttH and tH channels with only γγ and ZZ * → 4 Higgs decays [8]. The best fit value of the signal strength for 13.2 fb −1 data is 1.8 ± 0.7 which corresponds to 2.8σ observed significance. The observed (expected) 95% C.L upper limit on µ ttH is found to be 3.0 (1.2). The 13.2 fb −1 combination has smaller uncertainty than Run 1. Concerning the 36.1 fb −1 combination, 10% to 60% uncertainty reduction is achieved with respect to Run 1 ttH (γγ) analysis. The ration of the cross sections between the ttH and ggF is also computed and found to be 0.007 +0.010 −0.009 . The ratio between the top and gluon modifiers (called λ tg ) is also computed and found to be 0.74 +0.41 −0.63 . In total, all combined results are found to be consistent with the SM prediction.

Conclusion
Searches for the ttH production are performed in the ttH (bb), ttH (multileptons) and ttH (γγ) channels using 13.2 fb −1 of pp collision data at 13 TeV recorded by the ATLAS experiment. Results with