Observation of Higgs boson production in association with a top quark pair at the LHC with the ATLAS detector

TL;DR

This work reports the first or a highly significant observation of Higgs boson production in association with a top-quark pair (ttH) using $pp$ collision data from the LHC with the ATLAS detector. A multi-channel strategy combines $H\rightarrow\gamma\gamma$ and $H\rightarrow ZZ^*\rightarrow4\ell$ at 13 TeV with earlier $H\rightarrow b\bar{b}$ and multilepton analyses from 7 and 8 TeV, employing profile-likelihood fits and channel-specific systematics to extract the ttH cross section. The 13 TeV dataset yields a measured $\sigma_{t\bar{t}H}=670\pm90\mathrm{~(stat)}^{+110}_{-100}\mathrm{~(syst)}$ fb, consistent with the SM prediction, and the combined 7/8/13 TeV result reaches $6.3\sigma$ significance, demonstrating a direct test of the Higgs–top Yukawa coupling. The analysis reinforces the SM description of Higgs interactions and provides a crucial benchmark for top-quark–Higgs dynamics at the LHC.

Abstract

The observation of Higgs boson production in association with a top quark pair ($t\bar{t}H$), based on the analysis of proton-proton collision data at a centre-of-mass energy of 13 TeV recorded with the ATLAS detector at the Large Hadron Collider, is presented. Using data corresponding to integrated luminosities of up to 79.8 fb$^{-1}$, and considering Higgs boson decays into $b\bar{b}$, $WW^*$, $ττ$, $γγ$, and $ZZ^*$, the observed significance is 5.8 standard deviations, compared to an expectation of 4.9 standard deviations. Combined with the $t\bar{t}H$ searches using a dataset corresponding to integrated luminosities of 4.5 fb$^{-1}$ at 7 TeV and 20.3 fb$^{-1}$ at 8 TeV, the observed (expected) significance is 6.3 (5.1) standard deviations. Assuming Standard Model branching fractions, the total $t\bar{t}H$ production cross section at 13 TeV is measured to be 670 $\pm$ 90 (stat.) $^{+110}_{-100}$ (syst.) fb, in agreement with the Standard Model prediction.

Figures (6)

• Figure 1: Distribution of the BDT output in the (a) Had and (b) Lep region in the $H \rightarrow \gamma \gamma$ analysis. The distribution of the simulated $t\bar{t}H$ signal is compared with that of the other Higgs boson production modes, as well as to the continuum background from data in the diphoton invariant-mass sidebands of $105~\gev < m_{\gamma \gamma} < 120~\gev$ and $130~\gev < m_{\gamma \gamma} < 160~\gev$. Events to the left of the vertical line are rejected. The distributions are normalized to unity.
• Figure 2: Weighted diphoton invariant mass spectrum in the $t\bar{t}H$-sensitive BDT bins observed in 79.8 of 13 data. Events are weighted by $\ln(1 + S_{90}/B_{90})$, where $S_{90}$ ($B_{90}$) for each BDT bin is the expected $t\bar{t}H$ signal (background) in the smallest $m_{\gamma\gamma}$ window containing 90% of the expected signal. The error bars represent 68% confidence intervals of the weighted sums. The solid red curve shows the fitted signal-plus-background model with the Higgs boson mass constrained to 125.09 $\pm$ 0.24 . The non-resonant and total background components of the fit are shown with the dotted blue curve and dashed green curve. Both the signal-plus-background and background-only curves shown here are obtained from the weighted sum of the individual curves in each BDT bin.
• Figure 3: Number of data events in the different BDT bins of the $H \rightarrow \gamma \gamma$ analysis, in the smallest diphoton mass window that contains 90% of the $t\bar{t}H$ signal. The expected background and $t\bar{t}H$ signal (for a signal strength $\mu = \sigma / \sigma_\mathrm{SM}$ of 1.4) are shown as well. The expected continuum background is extracted from the diphoton mass fits. The lower panel shows the residuals between the data and the background. The red line shows the expected signal. The BDT bins are shown in ascending order of signal purity.
• Figure 4: Observed event yields in all analysis categories in up to 79.8 of 13 data. The background yields correspond to the observed fit results, and the signal yields are shown for both the observed results ($\mu = 1.32$) and the SM prediction ($\mu = 1$). The discriminant bins in all categories are ranked by $\log_{10}(S/B)$, where $S$ is the signal yield and $B$ the background yield extracted from the fit with freely floating signal, and combined such that $\log_{10}(S+B)$ decreases approximately linearly. For the $H \rightarrow \gamma \gamma$ analysis, only events in the smallest $m_{\gamma\gamma}$ window containing 90% of the expected signal are considered. The lower panel shows the ratio of the data to the background estimated from the fit with freely floating signal, compared to the expected distribution including the signal assuming $\mu = 1.32$ (full red) and $\mu = 1$ (dashed orange). The error bars on the data are statistical.
• Figure 5: Combined $t\bar{t}H$ production cross section, as well as cross sections measured in the individual analyses, divided by the SM prediction. The $\gamma \gamma$ and $ZZ^* \rightarrow 4\ell$ analyses use 13 data corresponding to an integrated luminosity of 79.8 , and the multilepton and $b\bar{b}$ analyses use data corresponding to an integrated luminosity of 36.1 . The black lines show the total uncertainties, and the bands indicate the statistical and systematic uncertainties. The red vertical line indicates the SM cross-section prediction, and the grey band represents the PDF+$\alphas$ uncertainties and the uncertainties due to missing higher-order corrections.