Search for Higgs boson pair production in association with top-quark pairs using 196 fb$^{-1}$ of proton-proton collision data at $\sqrt{s}=$ 13 and 13.6 TeV with the ATLAS detector

Abstract

This paper presents the first search for non-resonant Higgs boson pair production in association with a top-quark pair ($t\bar{t}HH$) using proton-proton collision data collected with the ATLAS detector at the Large Hadron Collider. The data sample corresponds to an integrated luminosity of 196 fb$^{-1}$, comprising 140 fb$^{-1}$ at a centre-of-mass energy of $\sqrt{s}=13$ TeV and 56 fb$^{-1}$ at 13.6 TeV. The search targets three distinct final states expected from $t\bar{t}HH$ decays: (i) one lepton (electron or muon) with at least five $b$-quarks, (ii) at least two $b$-quarks accompanied by two leptons with the same electric charges or multiple leptons, and (iii) at least three $b$-quarks with two photons. The $t\bar{t}HH$ production cross-section, relative to its Standard Model prediction, is measured to be $μ_{t\bar{t}HH}=-3^{+11}_{-12}$. This result corresponds to a 95$\% $ confidence-level upper limit of 20 times the Standard Model prediction for the $t\bar{t}HH$ production cross-section, with an expected limit of 21. The Higgs effective field theory Wilson coefficient $c_{t\bar{t}HH}$ is also constrained at the same confidence level to the range of $-3.9<c_{t\bar{t}HH}<3.3$, compared with the expected range of $-4.0<c_{t\bar{t}HH}<3.5$.

Figures (11)

• Figure 1: Examples of leading-order Feynman diagrams for the non-resonant $t\bar{t}HH\xspace$ process. Each diagram depicts a distinct subprocess arising from different Higgs boson couplings: (a) the Higgs--top Yukawa coupling, (b) the Higgs--top Yukawa coupling and Higgs boson trilinear self-coupling, and (c) the non-SM $t\bar{t}HH\xspace$ quartic coupling (highlighted with a circle), as described by the Higgs effective field theory.
• Figure 2: BDT score distributions obtained from the combined Run 2 and Run 3 data samples, used to discriminate the $t\bar{t}HH\xspace$ signal from the main background processes: (a) $\gamma\gamma b\bar{b}$, (b) $t\bar{t}\gamma\gamma$, (c) $t\bar{t}H$, and (d) single Higgs boson and di-Higgs boson production. All histograms are normalised to unity. Vertical dashed lines indicate the BDT score thresholds used to define the signal regions, whose boundaries are determined simultaneously across the four BDT scores.
• Figure 3: Distributions of the sum of jet pseudo-continuous $b$-tagging scores in the $4b_\text{Hi}$ and $5b_\text{Lo}$ control regions of the 1L channel for the (a, c) Run 2 and (b, d) Run 3 data samples, comparing the data with the post-fit background predictions. The lower panels display the relative differences between the data and the background predictions. The hatched bands indicate the combined statistical and systematic uncertainties in the background predictions. For reference, the pre-fit background predictions are overlaid as dashed histograms.
• Figure 4: Distributions of the fit variables of the SSML control regions, using the same binning as in the fit, for the (a) Run 2 and (b) Run 3 data samples, comparing the data with the post-fit background predictions. The lower panels display the relative differences between the data and the background predictions. The hatched bands indicate the combined statistical and systematic uncertainties in the background predictions. For reference, the pre-fit background predictions are overlaid as dashed histograms.
• Figure 5: Distributions of the $D_{\texttt{1L}}$ discriminant in the $5b_{\text{Hi}}$ and $6b$ signal regions of the 1L channel for the (a, c) Run 2 and (b, d) Run 3 data samples, comparing the data with the post-fit signal-plus-background predictions. The observed $t\bar{t}HH\xspace$ signal in this channel, corresponding to a measured signal strength of $\mu^{\text{1L}}_{t\bar{t}HH\xspace}=-5^{+15}_{-15}$, is displayed as a negatively stacked contribution. The lower panels display the relative differences between the data and the background predictions. The hatched bands indicate the combined statistical and systematic uncertainties in the background predictions. For reference, the pre-fit background predictions are overlaid as dashed lines, and the SM $t\bar{t}HH\xspace$ signal distribution -- scaled up by a factor of 200 -- are shown as solid lines.