Search for $t\bar{t}H/A \rightarrow t\bar{t}t\bar{t}$ production in proton-proton collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

TL;DR

The paper conducts a comprehensive search for heavy Higgs states $H$ or a pseudoscalar $A$ predicted by two-Higgs-doublet models, produced with a top-quark pair and decaying to a four-top final state. It leverages the ATLAS Run 2 dataset at $\sqrt{s}=13$ TeV with $139\mathrm{~fb}^{-1}$, employing 1L/2LOS channels and previously published multilepton results to set stringent 95% CL limits on $\sigma(pp\to t\bar t H/A)\times B(H/A\to t\bar t)$ across $m_{H/A}=400$–$1000$ GeV, translating into tan$\beta$ constraints in the alignment limit. The analysis introduces novel data-driven corrections for $t\bar t$+jets background using flavour rescaling and a multi-dimensional NN reweighting, plus an $m_{H/A}$-parameterised graph neural network to optimise signal discrimination. When combined with a prior ATLAS search, the results exclude tan$\beta$ values below 1.2–1.5 (depending on mass) for single-resonant scenarios and below about 1.9 (at 400 GeV) or 0.7 (at 1000 GeV) when both $H$ and $A$ contribute. Additionally, the study constrains a colour-octet scalar model (sgluon) via the same final state, excluding $m_{S_8}\leq1.3$ TeV. These findings advance the sensitivity to BSM Higgs sectors and provide complementary limits to gluon-initiated heavy Higgs searches in the same mass range.

Abstract

A search is presented for a heavy scalar ($H$) or pseudo-scalar ($A$) predicted by the two-Higgs-doublet models, where the $H/A$ is produced in association with a top-quark pair ($t\bar{t}H/A$), and with the $H/A$ decaying into a $t\bar{t}$ pair. Events are selected requiring exactly one or two opposite-charge electrons or muons. Data-driven corrections are applied to improve the modelling of the $t\bar{t}$+jets background in the regime with high jet and $b$-jet multiplicities. These include a novel multi-dimensional kinematic reweighting based on a neural network trained using data and simulations. An $H/A$-mass parameterised graph neural network is trained to optimise the signal-to-background discrimination. In combination with the previous search performed by the ATLAS Collaboration in the multilepton final state, the observed upper limits on the $t\bar{t}H/A \rightarrow t\bar{t}t\bar{t}$ production cross-section at 95% confidence level range between 14 fb and 5.0 fb for an $H/A$ with mass between 400 GeV and 1000 GeV, respectively. Assuming that both the $H$ and $A$ contribute to the $t\bar{t}t\bar{t}$ cross-section, $\tanβ$ values below 1.7 or 0.7 are excluded for a mass of 400 GeV or 1000 GeV, respectively. The results are also used to constrain a model predicting the pair production of a colour-octet scalar, with the scalar decaying into a $t\bar{t}$ pair.

Figures (12)

• Figure 1: Illustrative Feynman diagram showing the production of a heavy scalar or pseudo-scalar Higgs boson, $H/A$, produced in association with a pair of top quarks and the Higgs boson decaying into a pair of top quarks.
• Figure 2: Schematic view of the event categorisation for the (a) 1L and (b) 2LOS channels. The axes represent the jet multiplicities and $b$-tagging requirements defined in Table \ref{['tab:categorisation']}. Control, signal and validation regions are shaded with different colours. The regions used to derive the flavour rescaling and NN kinematic corrections are highlighted using rectangles.
• Figure 3: The relative contribution of the different background classes in the control, validation and signal regions in the (a) 1L and (b) 2LOS channels.
• Figure 4: The comparison of (a, b) $H_\mathrm{T}$, (c, d) $N_{\mathrm{jets}}$ and (e, f) $N_{\textrm{LR-jets}}$ distributions between data and background predictions before the likelihood fit to data (Pre-Fit), and before and after applying the data-driven corrections, separately for the 1L and 2LOS channels. The background prediction after applying the data-driven corrections is shown in coloured stacks, whilst the one before applying the corrections is shown with the red dashed line. The hashed area represents the total uncertainty in the background. The last bin in each distribution contains the overflow.
• Figure 5: The distributions of the GNN output evaluated at 400 $\text{Ge V}$ and 1000 $\text{Ge V}$ for the (a, b) 1L and (c, d) 2LOS channels with inclusive selections close to the signal region. The distributions are compared for the signal, SM $t\bar{t}t\bar{t}$ production and $t\bar{t}$+jets production.