⚛️ high-energy experiments

Measurement of the Higgs boson production in association with top quarks in multilepton final states in $pp$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

Using 140 fb $^{-1}$ of 13 TeV proton-proton collision data collected by the ATLAS detector, this study measures the $t\bar{t}H$ production cross-section in multilepton final states, finding a signal strength of $0.63^{+0.20}_{-0.19}$ with 3.3 $\sigma$ observed significance, while also performing differential measurements, constraining $tH$ production, and excluding large $CP$-odd mixing in the top-Higgs Yukawa coupling.

Original authors: Atlas Collaboration

Published 2026-07-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Atlas Collaboration

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, invisible web of forces holding everything together. For decades, scientists have been trying to map this web, looking for the "glue" that gives particles their mass. In 2012, they found a crucial piece of this puzzle: a particle called the Higgs boson. Think of the Higgs boson as a cosmic snowstorm; as other particles move through it, they get "stuck" in the snow, which is what we experience as mass. The heavier the particle, the more snow it seems to catch.

Now, meet the heavyweight champion of the particle world: the top quark. It's so massive that it should be grabbing a huge handful of that cosmic snow. Because it's so heavy, it's the perfect test subject to see if the Higgs boson really acts the way our best theories predict. If the top quark and the Higgs boson are dancing together in a specific way, it confirms our understanding of how the universe got its mass. But if they dance differently, or if the Higgs has a secret "twist" in its moves, it could mean there's new, unknown physics hiding in the shadows. This is why physicists are so eager to watch the top quark and the Higgs boson interact.

The Great Cosmic Catch: Hunting the Top-Higgs Dance

In a massive experiment at the Large Hadron Collider (LHC) in Europe, the ATLAS collaboration acted like a team of ultra-precise detectives. They smashed protons together at nearly the speed of light, creating a chaotic storm of particles. Out of trillions of these collisions, they were looking for a very rare, specific event: a top quark and an anti-top quark (a "top pair") being born at the same time as a Higgs boson. This is the "top-Higgs" party, or as physicists call it, $t\bar{t}H$ .

Finding this party is like looking for a specific, shy guest at a crowded, noisy concert. The Higgs boson is unstable and disappears almost instantly, turning into other particles. The top quarks are also short-lived. To catch them, the ATLAS team had to look for the "footprints" they leave behind: a specific mix of electrons, muons (heavy cousins of electrons), and tau particles. They sifted through data equivalent to 140 inverse femtobarns of collisions—a massive dataset collected over several years—to find these rare signatures.

What They Found

After crunching the numbers, the team found a signal! They observed the top-Higgs party happening more often than random noise would allow. The result is an "observed significance" of 3.3 sigma. In the world of particle physics, this is a strong hint—like hearing a distinct melody in a noisy room—but it's not quite the "smoking gun" (which usually requires 5 sigma). If you were to run this experiment a million times, the team expected to see a signal this strong about 5.3 times out of 100 purely by chance. So, while they are very confident they are seeing something real, they are still gathering more evidence to be absolutely certain.

When they measured how often this happens (the cross-section), they found a rate of 321 fb. The Standard Model (our best textbook of physics) predicted it should be 507 fb. The measured value is lower than expected, sitting at about 63% of the predicted rate ( $0.63^{+0.20}_{-0.19}$ ). However, the "fuzziness" or uncertainty in the measurement is quite large. Because the range of possible values overlaps with the textbook prediction, the result is considered "compatible" with the Standard Model. It's like guessing the weight of a mystery box: your guess was a bit light, but the scale was wobbly enough that the box could still be exactly what you thought it was.

The Secret Twist: Is the Higgs a Shape-Shifter?

One of the most exciting parts of this paper is the search for a "secret twist" in the Higgs boson's personality. In the Standard Model, the Higgs is a "CP-even" particle, which is a fancy way of saying it has a specific symmetry, like a perfect sphere. But some theories suggest it could be a mix of "CP-even" and "CP-odd" (like a sphere that's slightly squashed or twisted).

The team analyzed the angles and directions of the particles flying out of the collision to see if the Higgs was behaving like a twisted shape-shifter. They found no evidence of a twist. In fact, they were able to rule out (exclude) the possibility that the Higgs is a "purely twisted" (CP-odd) particle with a high degree of confidence. They also set a limit on how much "twist" could be hiding: if the mixing angle (the amount of twist) is larger than 62 degrees, it's not there. This result aligns perfectly with the idea that the Higgs is the standard, non-twisted particle we thought it was.

The Verdict

The ATLAS team successfully caught a glimpse of the top quark and Higgs boson hanging out together. While the number of times they saw it was a bit lower than the perfect textbook prediction, the difference isn't statistically significant enough to say the textbook is wrong. The Higgs boson continues to behave like the standard, non-twisted particle, and the top quark seems to be grabbing its share of the cosmic snow just as expected. The search continues, and with more data, scientists hope to tighten the net and see if there are any hidden secrets left to discover in this cosmic dance.

Technical Summary: Measurement of the Higgs boson production in association with top quarks in multilepton final states in $pp$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

Problem and Motivation
Following the 2012 discovery of the Higgs boson, a primary objective of the Large Hadron Collider (LHC) physics program has been the precise measurement of its properties. The interaction between the Higgs boson and the top quark is of particular interest due to the top quark's large mass ( $\approx 173$ GeV), which implies a strong coupling to the Higgs field in the Standard Model (SM). This coupling is crucial for understanding electroweak symmetry breaking. While the inclusive cross-section of associated top-quark pair and Higgs boson production ( $t\bar{t}H$ ) allows for a direct determination of the top-quark Yukawa coupling with fewer theoretical assumptions than loop-induced processes, previous measurements at $\sqrt{s}=13$ TeV using datasets of 36 fb $^{-1}$ and 137 fb $^{-1}$ showed signal strengths compatible with, but slightly above, SM predictions. Furthermore, the $CP$ nature of the top-Higgs coupling remains an open question; while the SM predicts a $CP$-even scalar, extended Higgs sectors could introduce a $CP$-mixed state, providing a source of $CP$-violation necessary for baryogenesis. This paper presents a measurement using the full Run 2 dataset (140 fb $^{-1}$ ) to improve the precision of the inclusive and differential $t\bar{t}H$ cross-sections, search for single top-Higgs production ($tH$), and probe the $CP$ structure of the top-Higgs Yukawa coupling.

Methodology
The analysis utilizes proton-proton collision data recorded by the ATLAS detector at $\sqrt{s}=13$ TeV, corresponding to an integrated luminosity of 140 fb $^{-1}$ . The study focuses on multilepton final states where the Higgs boson decays via $H \to WW^*$ , $H \to \tau^+\tau^-$ , or $H \to ZZ^*$ , and top quarks decay leptonically.

Event Selection and Categorization: Six distinct channels are defined based on the number and flavor of reconstructed charged leptons ( $\ell = e, \mu$ ) and hadronically decaying $\tau$ -leptons ( $\tau_{had}$ ):
1. $2\ell SS0\tau_{had}$ (two same-sign leptons, no $\tau_{had}$ )
2. $3\ell0\tau_{had}$ (three leptons, no $\tau_{had}$ )
3. $4\ell$ (four leptons)
4. $2\ell SS1\tau_{had}$ (two same-sign leptons, one $\tau_{had}$ )
5. $1\ell2\tau_{had}$ (one lepton, two $\tau_{had}$ )
6. $2\ell2\tau_{had}$ (two leptons, two $\tau_{had}$ )
  These channels are statistically independent. Object reconstruction employs strict identification criteria, including "Tight" and "Medium" working points for leptons and $\tau_{had}$ , and advanced $b$ -tagging algorithms (DL1r).
Background Estimation: Backgrounds are categorized as irreducible (prompt leptons from $t\bar{t}W$ , $t\bar{t}Z$ , dibosons) or reducible (non-prompt leptons, misidentified $\tau_{had}$ , charge misidentification). Irreducible backgrounds are estimated using Monte Carlo (MC) simulations normalized via a simultaneous likelihood fit to data in dedicated control regions (CRs). Reducible backgrounds, particularly non-prompt leptons and misidentified $\tau_{had}$ , are estimated using data-driven techniques, including the fake-factor method and specific CRs enriched in these processes.
Signal Extraction and Discrimination: Multivariate analysis (MVA) techniques, specifically Boosted Decision Trees (BDTs) and Deep Neural Networks (DNNs), are employed to separate signal from background.
- For the $t\bar{t}H$ cross-section measurement, events are categorized by the output scores of these MVAs.
- For the differential measurement in bins of the Higgs boson transverse momentum ( $p_T^H$ ) within the Simplified Template Cross-Section (STXS) framework, a Graph Neural Network (GNN) is trained to estimate $p_T^H$ using full event information (leptons, jets, $E_T^{miss}$ ) in channels where full reconstruction is ambiguous.
- For the $CP$ interpretation, dedicated $CP$-sensitive BDTs are trained to distinguish between $CP$-even and $CP$-odd hypotheses.
Statistical Analysis: A combined profile-likelihood fit is performed across all signal and control regions. The fit extracts the signal strength ( $\mu$ ), normalizes background processes, and constrains systematic uncertainties (nuisance parameters). Systematic uncertainties include experimental factors (jet energy scale, lepton identification, luminosity) and theoretical modeling (parton shower, PDFs, higher-order corrections).

Key Contributions

Dataset Scale: This analysis utilizes the full ATLAS Run 2 dataset (140 fb $^{-1}$ ), representing a significant increase in statistics compared to previous multilepton $t\bar{t}H$ measurements.
Improved Modeling: The analysis benefits from improved jet flavor tagging, lepton identification, and a refined understanding of dominant backgrounds ( $t\bar{t}W$ , $t\bar{t}Z$ ) derived from dedicated measurements. It employs a new simulation and systematic uncertainty model, including in situ normalization for main backgrounds.
Advanced Reconstruction: The introduction of a Graph Neural Network (GNN) to estimate the Higgs boson transverse momentum ( $p_T^H$ ) allows for a differential cross-section measurement in the STXS framework, overcoming the difficulty of fully reconstructing the Higgs system in multilepton final states.
Combined Interpretation: The paper presents a simultaneous measurement of $t\bar{t}H$ and $tH$ (specifically the $tHqb$ mode) signal strengths and a dedicated $CP$-violation search using the same dataset and region definitions.

Results

Inclusive Cross-Section: The measured inclusive $t\bar{t}H$ cross-section is $\sigma_{t\bar{t}H} = 321^{+102}_{-99}$ fb. Normalized to the SM prediction ( $\sigma_{SM} = 507^{+35}_{-50}$ fb), the signal strength is $\sigma_{t\bar{t}H}/\sigma_{SM} = 0.63^{+0.20}_{-0.19}$ . This corresponds to an observed (expected) significance of $3.3\sigma$ ( $5.3\sigma$ ). The result is compatible with the SM hypothesis, with a compatibility probability of 7.2%.
Differential Cross-Section: The cross-section is measured differentially in three $p_T^H$ bins (0–120 GeV, 120–200 GeV, >200 GeV). The results show excellent agreement with the SM for the low and high $p_T$ bins, while the intermediate bin ( $120 < p_T^H < 200$ GeV) shows a lower value than predicted, attributed to data underfluctuations in specific categories and anti-correlations between bins. The STXS fit compatibility with the SM is 28%.
Combined $t\bar{t}H$ and $tH$ Measurement: A simultaneous fit yields signal strengths of $\mu_{t\bar{t}H} = 0.59^{+0.22}_{-0.20}$ and $\mu_{tHqb} = 7.2^{+4.6}_{-4.0}$ . The $tHqb$ result is consistent with previous ATLAS measurements combining multilepton and $H \to b\bar{b}$ channels.
$CP$ Interpretation: The analysis probes the $CP$-mixing angle $\alpha$ in the top-Higgs Yukawa coupling. The results are compatible with the SM ($CP$-even) hypothesis. The observed (expected) exclusion of the pure $CP$-odd hypothesis is at $1.8\sigma$ ( $3.1\sigma$ ). At the 68% confidence level, values of the mixing angle $|\alpha| > 62^\circ$ are excluded (expected exclusion $|\alpha| > 43^\circ$ ).

Significance
The paper claims that this measurement represents a significant improvement over previous analyses based on smaller datasets, offering a more precise determination of the top-quark Yukawa coupling and its $CP$ properties. By combining six multilepton channels and utilizing advanced machine learning techniques for background suppression and kinematic reconstruction, the analysis provides a robust test of the Standard Model in the top-Higgs sector. The results, while showing a central value below the SM prediction for the inclusive cross-section, remain statistically compatible with the SM. The exclusion of large $CP$-mixing angles further constrains models of extended Higgs sectors and potential new sources of $CP$-violation. The methodology, particularly the use of GNNs for $p_T^H$ estimation and the simultaneous treatment of $t\bar{t}H$ and $tH$, sets a benchmark for future precision measurements in this channel.

More like this