⚛️ high-energy experiments

Search for a resonance in events with four top quarks decaying into two leptons and jets in proton-proton collisions

Using proton-proton collision data from the CMS experiment at 13 and 13.6 TeV, this study presents the first dedicated search for heavy resonances decaying into four top quarks in the two-lepton plus jets channel, finding no significant excess and setting upper limits on production cross sections for various vector, scalar, and pseudoscalar boson models.

Original authors: CMS Collaboration

Published 2026-08-12

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: CMS 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 LEGO set. For decades, physicists have been building a model called the Standard Model to explain how all the tiny pieces fit together. It's a brilliant blueprint that has predicted almost everything we've seen so far. But there's a problem: the model is missing some crucial pieces. It doesn't explain things like dark matter, the invisible glue holding galaxies together, or why the heaviest LEGO brick in the set—the top quark—is so incredibly heavy. This heavy brick is so massive that it might be the key to unlocking a secret door to a whole new world of physics. Scientists are desperate to find out if there are new, hidden forces or particles that only talk to this heavy top quark, acting like a secret handshake that the rest of the universe doesn't know about.

In this paper, the CMS collaboration at CERN's Large Hadron Collider (LHC) acts like a team of cosmic detectives hunting for a specific, rare crime scene. They are looking for a heavy "resonance"—think of it as a giant, unstable drum that briefly appears and then shatters into four top quarks at once. This is an incredibly rare event; usually, top quarks are produced in pairs, so finding four of them together is like finding four rare golden tickets in a single chocolate bar. The team smashed protons together at mind-boggling speeds, creating a storm of debris, and sifted through the wreckage looking for a specific pattern: two electrons or muons (light, ghostly particles) and a bunch of jets (sprays of particles) that look like they came from two top quarks that exploded into pure energy. They used a clever machine-learning algorithm, like a super-smart security guard, to spot which jets were actually top quarks and which were just ordinary noise.

The search covered a massive amount of data, equivalent to 138 and 35 inverse femtobarns of collisions at energies of 13 and 13.6 TeV. The scientists reconstructed the mass of the potential new particle by looking at pairs of jets, using a special "variable-radius" technique that acts like a zoom lens, adjusting its focus depending on how fast the particles were moving. They checked for new particles ranging from 500 GeV to 4 TeV in mass, with widths (how quickly they decay) between 4% and 50%. They also looked for different types of mediators: a vector boson (like a heavy Z' particle), a scalar, or a pseudoscalar (like a heavy Higgs or an axion-like particle).

The verdict? The detectives found no evidence of the crime. The data matched the Standard Model predictions perfectly, with no mysterious excess of events that would signal a new particle. While they didn't find the new physics they were hoping for, they did set strict boundaries. They ruled out the existence of these specific heavy resonances up to certain masses. For example, if a heavy vector particle with a 50% width existed, it would have to be heavier than 850 GeV, or it simply doesn't exist at all (the team expected to see it up to 1000 GeV). They also set limits on how strongly these hypothetical particles could interact with top quarks. While this search didn't discover a new particle, it successfully closed the door on several specific theories, telling us that if these heavy resonances exist, they are hiding in a part of the universe we haven't looked at closely enough yet. This is the first time this specific search has been done in the "two-lepton" channel, making it a unique and important step in the ongoing quest to understand the heavy top quark and the secrets of the universe.

Technical Summary: Search for a Resonance in Events with Four Top Quarks Decaying into Two Leptons and Jets

Problem and Motivation
The observation of four-top quark ( $tt\bar{t}\bar{t}$ ) production by the ATLAS and CMS Collaborations marks a significant milestone in testing the Standard Model (SM). While the SM is successful, it lacks an explanation for dark matter, and many Beyond the Standard Model (BSM) theories propose new particles interacting dominantly with top quarks due to the top's large Yukawa coupling. Previous searches for "top-philic" heavy resonances (vector $Z'$ , scalar $\phi$ , or pseudoscalar $a$ bosons) in four-top events have focused on final states with a single lepton, targeting high-mass resonances ( $m_{Z'} > 1$ TeV) where top quarks are highly boosted. This paper addresses a gap in the search strategy by focusing on a new final state: events with two leptons (electrons or muons) and jets. This channel is sensitive to resonances with masses below 1 TeV, where the decay products of the hadronically decaying top quarks are less collimated, requiring different reconstruction techniques.

Methodology
The analysis utilizes proton-proton collision data collected by the CMS experiment at center-of-mass energies of 13 TeV (2016–2018) and 13.6 TeV (2022), corresponding to integrated luminosities of 138 fb $^{-1}$ and 35 fb $^{-1}$ , respectively.

Signal Topology: The search targets the production of a heavy resonance ( $Z'$ , $\phi$ , or $a$ ) in association with a top-antitop pair ( $pp \to ttZ'$ , $tt\phi$ , $tta$). The resonance decays into a second $tt$ pair. The analysis selects events where the associated $tt$ pair decays leptonically ( $t \to b\ell\nu$ ), yielding two opposite-sign leptons, while the resonance-decaying $tt$ pair decays fully hadronically ( $t \to bqq'$ ).
Object Reconstruction:
- Leptons: Selected electrons and muons must satisfy specific $p_T$ and isolation criteria.
- Jets: Standard fixed-radius ( $R=0.4$ , AK4) jets are used to identify $b$ -quarks from the leptonic top decays.
- Variable-Radius (VR) Jets: To reconstruct the hadronically decaying top quarks from the resonance, a variable-radius jet clustering algorithm (HOTVR) is employed. This algorithm uses a $p_T$ -dependent distance parameter ( $R_{eff} = 600 \text{ GeV}/p_T$ ) to consistently capture decay products of moderately boosted top quarks (200–600 GeV).
Machine Learning: A Boosted Decision Tree (BDT) classifier is trained to identify (tag) VR jets originating from hadronically decaying top quarks. The BDT utilizes inputs such as jet mass, subjet multiplicity, and $N$ -subjettiness ratios.
Event Categorization: Events are categorized into Signal Regions (SRs) requiring exactly two leptons, at least two AK4 jets, and at least two VR jets that are $t$ -tagged. Two SRs are defined based on the number of $b$ -tagged AK4 jets ($SR1b2T$ and $SR2b2T$). Control Regions (CRs) and Transfer Regions (TRs) are defined to estimate backgrounds from data.
Background Estimation: The dominant background arises from $t\bar{t}$ production with additional jets where light-flavor jets are misidentified as top jets. This background is estimated using a data-driven method. Transfer factors, derived from simulated samples, relate the yield of events with misidentified top jets in the SRs to events with no $t$ -tagged jets in the TRs. The method is validated in CRs enriched in $Z/\gamma^* + \text{jets}$ events.
Statistical Analysis: A simultaneous maximum likelihood fit is performed on the dijet invariant mass ( $m_{JJ}$ ) distribution in the SRs. Systematic uncertainties (luminosity, jet energy scale, $b$ -tagging, $t$ -tagging, theoretical cross sections, and background modeling) are incorporated as nuisance parameters.

Key Contributions

First Dedicated Search: This is the first dedicated search for top-philic resonances in the four-top quark final state requiring two leptons.
Extended Mass Sensitivity: By utilizing VR jets and a machine-learning $t$ -tagger, the analysis extends sensitivity to mediator masses below 1 TeV, complementing previous searches that focused on highly boosted topologies.
Data-Driven Background Estimation: The paper implements a robust data-driven technique to estimate the dominant background of misidentified jets, validated in sideband regions, minimizing reliance on simulation for the background shape.
Multi-Energy Combination: The analysis combines data from 13 TeV and 13.6 TeV collisions, providing the first results for this signature incorporating the 13.6 TeV dataset.
ALP Interpretation: The results are reinterpreted to set limits on Axion-Like Particles (ALPs) coupling to top quarks, bridging the gap between $tt$ resonance searches and missing transverse momentum searches.

Results

Observation: No significant excess of events over the SM background expectation is observed. The data agree with the background-only hypothesis, with a $p$ -value of 0.21 for the background-only fit.
Local Significance: The largest deviation from the background hypothesis occurs for a 500 GeV vector mediator with a 4% relative width, showing a local significance of 2.1 standard deviations. Similar deviations (2.0–2.4 $\sigma$ ) are observed for scalar and pseudoscalar hypotheses at the same mass. These are consistent with statistical fluctuations.
Exclusion Limits: Upper limits at 95% confidence level (CL) are set on the production cross section for resonance masses between 500 GeV and 4 TeV and relative widths of 4%, 10%, 20%, and 50%.
- For the vector-mediator benchmark ( $Z'$ ), resonances with a 50% relative width are excluded up to masses of 850 GeV. The expected exclusion limit is 1000 GeV.
- No mediators with 10% or 20% widths are excluded by the observed data.
- For scalar ( $\phi$ ) and pseudoscalar ( $a$ ) mediators, no signal hypotheses are excluded by the observed data, though the pseudoscalar model is narrowly expected to be excluded for a 500 GeV signal with 50% width.
ALP Limits: Limits on the ALP-top coupling ( $c_t/f_A$ ) are derived. For a 500 GeV ALP, couplings above 0.025 GeV $^{-1}$ are excluded.

Significance
The paper claims significance primarily as a pioneering search in a specific kinematic regime (two-lepton channel, sub-TeV masses) that was previously unexplored for top-philic resonances. It demonstrates the effectiveness of variable-radius jet clustering and machine-learning classifiers in reconstructing moderately boosted top quarks. While no new physics is discovered, the results provide the first dedicated constraints in the two-lepton channel for these models and offer a complementary interpretation for ALP searches, distinguishing potential ALP signals from toponium enhancements in the $tt$ threshold region. The analysis sets the stage for future searches with increased luminosity and improved reconstruction techniques.

More like this