⚛️ high-energy experiments

Study of $t\bar{t}H$ and $tH$ production in the $H\to\tau\tau$ channel in $pp$ collisions at $\sqrt{s}=13$ TeV and 13.6 TeV with the ATLAS detector

This paper presents a study of $t\bar{t}H$ and $tH$ production in the $H\to\tau\tau$ channel using ATLAS data from proton-proton collisions at 13 and 13.6 TeV, reporting a signal strength of $\mu_{t\bar{t}H}=1.51^{+0.71}_{-0.62}$ and providing differential cross-section measurements for $t\bar{t}H$ .

Original authors: ATLAS Collaboration

Published 2026-07-31

📖 7 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, cosmic kitchen where the most fundamental ingredients are constantly being mixed and cooked. For decades, physicists have been trying to figure out the recipe book of the universe, known as the Standard Model. In this kitchen, there's a special ingredient called the Higgs boson, often nicknamed the "God particle" (though scientists prefer just "Higgs"). Think of the Higgs as the universe's sticky syrup; it's the thing that gives other particles their mass, or "heaviness," as they move through it. Without this syrup, particles would zip around at the speed of light and never stick together to form atoms, stars, or you.

But the Higgs doesn't just sit there; it interacts with other ingredients. One of the most important interactions is with the top quark, which is the heaviest particle in the known universe. If you imagine the top quark as a massive, grumpy boulder, the Higgs is the only thing heavy enough to really push it around. Scientists are obsessed with measuring exactly how strong this push is. Why? Because if the push is slightly different than the recipe book predicts, it could mean there's a secret new ingredient or a hidden rule of physics we haven't discovered yet. This paper is like a team of master chefs using the world's most powerful oven to see if they can catch the Higgs and the top quark interacting in a very specific, messy, and difficult-to-spot way.

The Big Hunt: Catching the Higgs and the Top Quark in a "Full-House" Mess

This paper is a report from the ATLAS experiment at CERN, the home of the Large Hadron Collider (LHC). The LHC is a 27-kilometer ring of superconducting magnets that smashes protons together at nearly the speed of light. It's like a cosmic particle accelerator that creates a mini-Big Bang in a tube, generating a shower of new particles. The ATLAS detector is a giant, multi-layered camera surrounding the collision point, designed to take a picture of every single particle that flies out.

In this study, the scientists were looking for a very specific event: the production of a Higgs boson sitting right next to either a pair of top quarks (called $t\bar{t}H$ ) or a single top quark (called $tH$). To make things even trickier, they only looked for cases where the Higgs boson decayed (broke apart) into a pair of tau leptons ( $\tau\tau$ ). And here is the real kicker: they only counted the events where everything—the top quarks and the tau leptons—decayed into hadrons (particles made of quarks, like pions and protons).

Why is this so hard? Imagine trying to find a specific, rare flavor of ice cream in a blizzard of snow, but the ice cream is melting into the snow, and the snow is also turning into other random flavors. The "fully hadronic" final state is incredibly messy. It's a chaotic pile of jets (sprays of particles) that looks very similar to the background noise of the universe. Most other experiments try to find these events by looking for cleaner signals, like electrons or muons, but this team decided to tackle the messy pile head-on.

The New Tool: A Smart Filter for Tau Particles

To cut through the noise, the team used a new, super-smart tool called GNTau. Think of the tau lepton as a shy ghost that leaves a very faint, specific footprint. In the past, the detector used a standard neural network to spot these footprints, but it sometimes got confused by fake footprints left by regular particles (jets).

The new GNTau tool is like a detective with a magnifying glass and a crystal ball. It uses a "Graph Neural Network," which is a type of artificial intelligence that looks at the relationships between all the pieces of the puzzle at once. Instead of just looking at one particle, it understands how the tracks, energy clusters, and jets talk to each other. This new detective is much better at telling the difference between a real tau ghost and a fake one, allowing the scientists to spot the signal in the noise much more clearly than before.

The Results: A Glimpse of the Signal

The team analyzed a massive amount of data: 140 inverse femtobarns ( $fb^{-1}$ ) from 2015–2018 (Run 2) and 161 $fb^{-1}$ from 2022–2024 (Run 3). To put that in perspective, they looked at trillions of collisions.

After running their complex statistical analysis, they found:

For the pair of top quarks ( $t\bar{t}H$ ): They measured a "signal strength" ( $\mu$ ) of $1.51^{+0.71}_{-0.62}$ .
- What does this mean? A signal strength of 1.0 would mean the universe behaves exactly as the Standard Model predicts. Their result is 1.51, which is slightly higher than expected, but the "error bars" (the uncertainty) are quite large. The range goes from roughly 0.89 to 2.22. Because this range includes 1.0, the result is consistent with the Standard Model. It's like guessing the temperature of a room is 72°F, but your thermometer has a margin of error that says it could be anywhere between 65°F and 80°F. You can't say for sure if it's hotter or colder, but it's definitely not freezing or boiling.
- The statistical significance of this finding is 2.6 standard deviations. In the world of particle physics, this is a "hint" or "evidence," but not a definitive "discovery" (which usually requires 5 standard deviations). It's like hearing a faint noise in the next room; you're pretty sure something is there, but you haven't seen it yet.
For the single top quark ($tH$): They measured a signal strength of $-0.4^{+5.6}_{-5.1}$ .
- What does this mean? This result is very uncertain. The number is negative, which physically doesn't make sense for a rate, but in statistics, it just means the data didn't show a clear signal above the background noise. The huge error bars (ranging from about -5.5 to +5.2) mean the measurement is compatible with zero. In other words, they didn't see a clear signal for this specific process in this channel yet. It's like trying to hear a whisper in a hurricane; the data is too noisy to tell if the whisper is there or not.

What They Did (and Didn't) Find

The paper explicitly states that they did not find a deviation from the Standard Model. The results are consistent with what we already know. They did not rule out new physics, but they also didn't find evidence for it. They successfully measured the production of the Higgs with top quarks in this difficult "fully hadronic" channel for the first time with this level of detail, including a differential measurement (breaking the data down by how fast the Higgs was moving).

They also confirmed that their new GNTau tool works great, improving the ability to spot these rare events. The study is a triumph of data analysis and machine learning, showing that even in the messiest corners of the particle zoo, we can still find the rare gems if we have the right tools.

The Bottom Line

This paper is a solid step forward in our understanding of the Higgs boson. It proves that we can study the Higgs interacting with the heaviest particles even when the final result is a chaotic pile of debris. While they didn't find a "smoking gun" for new physics, they tightened the screws on the Standard Model, confirming that the universe is still behaving mostly as expected. The measurement of the $t\bar{t}H$ process is now more precise, and the hunt for the elusive $tH$ process continues, armed with better tools and more data than ever before. It's a reminder that in science, sometimes the most important discovery is simply proving that the rules we wrote down are still holding up, even when the universe tries to hide them in a storm of particles.

Technical Summary: Study of $t\bar{t}H$ and $tH$ production in the $H \to \tau\tau$ channel in $pp$ collisions at $\sqrt{s}=13$ and 13.6 TeV with the ATLAS detector

Problem and Motivation
Since the discovery of the Higgs boson in 2012, characterizing its couplings to fermions has been a primary objective of the LHC physics program. While the Yukawa coupling to $\tau$ -leptons ( $H \to \tau\tau$ ) has been established, the top-quark Yukawa coupling remains of particular interest due to the top quark's large mass ( $\approx 173$ GeV), which suggests a dominant role in electroweak symmetry breaking. The associated production of the Higgs boson with a top-quark pair ( $t\bar{t}H$ ) allows for a direct determination of this coupling with fewer theoretical assumptions than loop-induced processes. Furthermore, the production of a Higgs boson in association with a single top quark ($tH$) is sensitive to the relative phase between the Higgs couplings to the top quark and the $W$ boson, offering a unique probe for potential $CP$-violation effects.

Previous measurements of $t\bar{t}H$ and $tH$ have utilized various decay channels, but the $H \to \tau\tau$ mode with fully hadronic final states (where both top quarks and $\tau$ -leptons decay hadronically) presents a challenging yet distinct signature. This channel benefits from an unambiguous Higgs boson candidate and low contamination from other Higgs decay modes, though it suffers from significant backgrounds. This paper presents a simultaneous measurement of $t\bar{t}H$ and $tH$ production in this specific final state, utilizing the full Run 2 dataset and the initial Run 3 data collected by the ATLAS detector.

Methodology
The analysis utilizes proton–proton collision data recorded by the ATLAS detector at center-of-mass energies of $\sqrt{s}=13$ TeV (Run 2, 140 fb $^{-1}$ ) and 13.6 TeV (Run 3, 161 fb $^{-1}$ ).

Event Selection: Events are selected requiring exactly two hadronically decaying $\tau$ -leptons ( $\tau_{\text{had-vis}}$ ) with opposite charges and no isolated electrons or muons. Additional requirements include missing transverse momentum ( $E_T^{\text{miss}} > 20$ GeV), angular separation constraints between $\tau$ candidates, and a collinear approximation variable ( $x$ ) to reconstruct the $\tau$ momentum. To target the signal topology, events must contain at least five jets, with at least one $b$ -tagged jet.
Object Reconstruction: The analysis employs a new Graph Neural Network (GNN) based identification algorithm, GNTau, for $\tau_{\text{had-vis}}$ reconstruction. This algorithm utilizes transformer-based architectures to combine track, cluster, and jet-level information, offering improved discrimination against quark/gluon-initiated jets and enhanced robustness against pile-up compared to previous recurrent neural network approaches.
Classification and Background Estimation: A multiclass Boosted Decision Tree (BDT) is trained to distinguish between four classes: $t\bar{t}H$ $t \overset{ˉ}{t} H$ , $tH$, $Z(\tau\tau)$ $Z (τ τ)$ , and $t\bar{t}$ $t \overset{ˉ}{t}$ . The BDT input variables include kinematic properties of jets and $\tau$ $τ$ -leptons, invariant masses of reconstructed $W$ $W$ and top candidates, and the di- $\tau$ $τ$ invariant mass ( $m_{\tau\tau}$ $m_{τ τ}$ ) calculated via the Missing Mass Calculator (MMC) technique.
- Backgrounds: The dominant irreducible backgrounds ( $Z(\tau\tau)$ and $t\bar{t}$ ) are estimated using Monte Carlo (MC) simulations normalized to data in dedicated control regions defined by BDT scores. The reducible background from misidentified $\tau$ -leptons (jets faking $\tau$ ) is estimated using a fake-factor method derived from data control regions.
Statistical Analysis: A combined profile-likelihood fit is performed across signal and control regions for both Run 2 and Run 3. The fit simultaneously extracts the signal strengths ( $\mu$ ) for $t\bar{t}H$ and $tH$, as well as normalization factors for background processes. Systematic uncertainties, including those related to jet energy scale, $b$ -tagging, $\tau$ identification, and theoretical modeling (PDFs, scales, parton showers), are incorporated as nuisance parameters.

Key Contributions

First $tH$ Measurement in $H \to \tau\tau$ : This work presents the first measurement of single-top-associated Higgs production ($tH$) in the $H \to \tau\tau$ decay channel with fully hadronic final states.
Application of GNTau: The analysis marks the first use of the GNTau $\tau$ -lepton identification algorithm in an ATLAS physics analysis, leveraging advanced graph neural network techniques to improve signal efficiency and background rejection.
Simultaneous $t\bar{t}H$ and $tH$ Fit: By performing a simultaneous fit, the analysis exploits the correlation between the two processes to constrain the top-quark Yukawa coupling and probe $CP$-violation effects more effectively than separate measurements.
Differential Measurement: The $t\bar{t}H$ cross-section is measured differentially in the Simplified Template Cross-Section (STXS) framework, binned by the Higgs boson transverse momentum ( $p_T(H)$ ), providing sensitivity to new physics effects such as anomalous couplings.

Results
The analysis yields the following signal strength measurements ( $\mu$ ), defined as the ratio of the measured cross-section to the Standard Model (SM) prediction:

$t\bar{t}H$ : $\mu_{t\bar{t}H} = 1.51^{+0.71}_{-0.62}$ . The observed significance is 2.6 standard deviations (expected: 1.8).
**$tH $:**$ \mu_{tH} = -0.4^{+5.6}_{-5.1}$. The result is consistent with the SM prediction, though the large uncertainty reflects the difficulty of isolating this rare process in this channel.
Correlation: There is a $-34\%$ correlation between the measured signal strengths of $t\bar{t}H$ and $tH$.
STXS Results: The differential $t\bar{t}H$ cross-section measurements in $p_T(H)$ bins show good agreement with SM predictions within uncertainties. The compatibility of the STXS fit with the SM is 52%.
Overall Compatibility: The combined fit result is compatible with the SM prediction with a $p$ -value of 84%.

Significance
The paper claims that these results are consistent with existing measurements and SM predictions. The primary significance lies in the extension of the $t\bar{t}H$ measurement to the fully hadronic $H \to \tau\tau$ channel with an expanded dataset and improved reconstruction techniques, and the pioneering measurement of $tH$ in this specific channel. The results provide constraints on the top-quark Yukawa coupling and offer a new avenue for probing $CP$-violating effects in the top-Higgs interaction, albeit with current uncertainties that prevent a definitive observation of deviations from the SM. The successful application of the GNTau algorithm demonstrates the potential of advanced machine learning techniques for improving $\tau$ -lepton identification in high-pile-up environments.

More like this