⚛️ high-energy experiments

Search for Lorentz-boosted di- $\tau$ resonances produced in association with top quark pairs in $\sqrt{s}=13$ TeV pp collisions with the ATLAS detector

Using 140 fb $^{-1}$ of 13 TeV proton-proton collision data collected by the ATLAS detector, this study presents a search for Lorentz-boosted pseudoscalar resonances decaying into hadronic $\tau$ -lepton pairs in association with top quark pairs, finding no significant deviation from the Standard Model and setting upper limits on the production cross-section for masses between 20 and 85 GeV.

Original authors: ATLAS Collaboration

Published 2026-07-31

📖 4 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 puzzle. For decades, scientists have been trying to fit the pieces together to understand how everything works, from the tiniest specks of dust to the massive stars. The "Standard Model" is the picture on the puzzle box that we've mostly solved; it explains how particles like electrons and quarks interact. But there are still gaps in the picture. We know there's something called "Dark Matter" holding galaxies together, and we know the rules of the universe might be a bit more complicated than our current puzzle box suggests. One of the biggest missing pieces is the "Higgs field," which gives particles their mass. We found the main Higgs particle in 2012, but physicists suspect there might be other, lighter, "hidden" cousins of this particle lurking in the shadows. Finding them would be like discovering a secret room in a house we thought we knew perfectly. If these light particles exist, they could explain why the universe looks the way it does and might even reveal the nature of Dark Matter.

To find these elusive particles, scientists need a giant microscope. Enter the Large Hadron Collider (LHC) at CERN, a 27-kilometer ring buried underground that smashes protons together at nearly the speed of light. It's like a cosmic crash test where the debris from the collision can briefly form new, exotic particles that usually don't exist. The ATLAS detector is one of the giant cameras watching these crashes, recording every splinter and spark to see if anything unusual pops out.

In this specific study, the ATLAS team is hunting for a very specific type of particle: a light particle (let's call it particle a) that is produced alongside a pair of heavy top quarks. When this particle a is created, it almost immediately falls apart into a pair of tau particles (a heavy cousin of the electron). The tricky part is that in the high-energy collisions of the LHC, this particle a is often moving so fast that it gets "Lorentz-boosted." Imagine throwing a snowball so hard that it stretches out; the two tau particles it breaks into get squished so close together that they look like a single, messy blob rather than two separate snowballs.

The team analyzed a massive amount of data—140 inverse femtobarns of proton-proton collisions collected between 2015 and 2018. That's a huge pile of cosmic crash footage. They developed a special, custom-made "net" to catch these squished, boosted tau pairs. Usually, when particles are squished this tight, it's very hard to tell them apart from the background noise of regular jets (which are just sprays of ordinary particles). But the ATLAS team built a new algorithm that acts like a super-powered magnifying glass, looking deep inside those messy blobs to see if they really contain two taus that have decayed into visible particles (specifically, hadrons like pions).

They looked for these events in the mass range between 20 GeV and 85 GeV. The result? They found nothing unusual. The number of "squished tau blobs" they saw was consistent with the Standard Model prediction, matching the expected background within the statistical uncertainty of about ±6 events. It's like searching a crowded room for a specific person wearing a red hat, and after checking everyone, you realize the number of people wearing that hat is exactly what you'd expect if no one special was hiding there.

Because they didn't find the particle, they couldn't say "it exists." Instead, they set a strict limit on how heavy or how common it could be. They calculated that if this particle a does exist, it must be so rare or so weakly interacting that the chance of it being produced is less than 0.19 femtobarns (a tiny unit of probability). For a specific theoretical model called the Two-Higgs-Doublet Model, they ruled out the particle's existence for certain masses, narrowing down the search area. Essentially, they didn't find the treasure, but they successfully mapped out a large area of the map where the treasure definitely isn't hiding, forcing scientists to look in new places. This search is the first time ATLAS has looked for this specific "boosted" signature in this way, proving that their new "magnifying glass" works, even if the ghost they were chasing stayed hidden this time.

Technical Summary: Search for Lorentz-boosted di- $\tau$ resonances produced in association with top quark pairs

Problem and Motivation
The discovery of the Higgs boson established the Standard Model (SM) mechanism for electroweak symmetry breaking but did not preclude an extended Higgs sector. Many Beyond the Standard Model (BSM) scenarios, including Dark Matter models, neutral naturalness, and axion-like particle scenarios, predict the existence of light Higgs-like bosons (masses below 125 GeV). These particles often couple preferentially to heavy fermions, making the associated production with top-antitop ( $t\bar{t}$ ) pairs a critical probe. Specifically, the decay of a light pseudoscalar resonance ( $a$ ) into a $\tau^+\tau^-$ pair offers a sensitive channel. While previous searches have focused on resolved $\tau$ decay modes or different production mechanisms, the specific topology of a low-mass $a$ resonance produced in association with $t\bar{t}$ and decaying into a fully hadronic di- $\tau$ final state ( $a \to \tau_{\text{had}}\tau_{\text{had}}$ ) in the highly boosted regime remained largely unexplored. In this regime, the decay products are highly collimated, rendering traditional resolved reconstruction techniques inefficient.

Methodology
The analysis utilizes proton-proton collision data collected by the ATLAS detector at $\sqrt{s} = 13$ TeV, corresponding to an integrated luminosity of 140 fb $^{-1}$ from Run 2 (2015–2018). The search targets the mass range $20 \text{ GeV} \le m_a \le 85 \text{ GeV}$ .

Event Topology: The signal process involves $t\bar{t}a$ production where the $t\bar{t}$ system decays semileptonically (one top to $b\ell\nu$ , the other to $b q\bar{q}'$ ) and the pseudoscalar $a$ decays to a fully hadronic di- $\tau$ pair.
Object Reconstruction:
- Leptons: Events are triggered and selected based on a single isolated electron or muon ( $p_T > 27$ GeV).
- Boosted Di- $\tau$ Objects: A dedicated algorithm reconstructs the di- $\tau$ system as a single large-radius jet ( $R=1.0$ ) containing two sub-jets ( $R=0.2$ ). This approach captures the collimated decay products of the boosted $\tau$ -leptons.
- Identification: A Boosted Decision Tree (BDT) discriminates between prompt boosted di- $\tau$ objects and misidentified jets (fakes). The BDT utilizes track and calorimeter variables, trained on signal ( $t\bar{t}a$ ) and background ( $t\bar{t}$ ) samples.
Selection and Categorization:
- Signal Region (SR): Requires exactly one lepton, exactly one boosted di- $\tau$ candidate, at least three jets (one $b$ -tagged), missing transverse momentum ( $E_T^{\text{miss}} > 30$ GeV), and a specific angular alignment between the di- $\tau$ and $E_T^{\text{miss}}$ . The di- $\tau$ sub-jets must have opposite charges ( $Q_{\tau\tau} = -1$ ) and a high BDT score ( $>0.5$ ).
- Control and Validation Regions (CR/VR): Orthogonal regions are defined by varying the BDT score, sub-jet charge sign (same-sign vs. opposite-sign), and $E_T^{\text{miss}}$ requirements. These regions are used to normalize the dominant $t\bar{t}$ background and validate the background modeling, particularly for fake leptons and fake di- $\tau$ objects.
Background Estimation:
- The dominant background ( $t\bar{t}$ with jets faking di- $\tau$ ) is estimated using MC simulations normalized to data in the $t\bar{t}$ control region (CR $_{t\bar{t}}$ ).
- Fake lepton backgrounds are estimated using a data-driven Matrix Method.
- Irreducible backgrounds (e.g., $t\bar{t}Z$ with $Z \to \tau\tau$ ) are modeled via MC.
Statistical Analysis: A binned likelihood fit is performed under background-only and signal-plus-background hypotheses. Systematic uncertainties (theoretical, experimental, and modeling) are treated as nuisance parameters.

Key Contributions

Novel Reconstruction Technique: This work presents the first ATLAS result utilizing a fully hadronic boosted di- $\tau$ reconstruction technique for $t\bar{t}$ -associated production. This method significantly improves sensitivity in the low-mass region ($20-85$ GeV) where standard resolved $\tau$ reconstruction fails due to the small angular separation of decay products.
Comprehensive Background Handling: The analysis employs a robust strategy combining data-driven normalization for the dominant $t\bar{t}$ background and dedicated validation regions to constrain systematic uncertainties related to fake objects and misidentification.
Model-Independent and Dependent Limits: The analysis provides both model-independent upper limits on the visible cross-section and model-dependent exclusion limits within a Two-Higgs-Doublet Model (2HDM) framework.

Results

Observation: No significant deviation from the Standard Model background expectation is observed in the signal region. The observed event yield (51 events) is consistent with the predicted SM background ( $51 \pm 6$ events).
Model-Independent Limits: A 95% confidence level (CL) upper limit of 0.19 fb is placed on the visible cross-section ( $\langle A\epsilon\sigma \rangle$ ).
Model-Dependent Limits: Assuming a 2HDM signal scenario, upper limits on the cross-section times branching ratio, $\sigma(t\bar{t}a) \times \text{BR}(a \to \tau\tau)$ , are set. These limits range from 0.4 pb at $m_a = 20$ GeV to 0.05 pb at $m_a = 85$ GeV. The sensitivity decreases at lower masses primarily due to the reduced acceptance and efficiency of the boosted di- $\tau$ reconstruction algorithm for softer $\tau$ -lepton signatures.

Significance
This paper claims to deliver the first ATLAS result in the boosted fully hadronic di- $\tau$ final state for $t\bar{t}$ -associated production. By leveraging the boosted topology, the analysis achieves the strongest LHC limits for $t\bar{t}(a \to \tau\tau)$ production in the hadronic di- $\tau$ channel at a mass of 20 GeV at the time of publication. The results complement recent ATLAS searches for $t\bar{t}a$ production in the $a \to \mu\mu$ and $a \to b\bar{b}$ channels, providing further constraints on potential extensions of the Higgs sector and the parameter spaces of various BSM scenarios.

More like this