⚛️ high-energy experiments

Search for emerging jets in $pp$ collisions at $\sqrt{s} = 13$ TeV with the ATLAS experiment

Using 140 fb $^{-1}$ of 13 TeV proton-proton collision data, the ATLAS experiment conducted a search for emerging jets in four-jet topologies and found no significant excess over the Standard Model background, setting exclusion limits on mediator masses between 1 and 2 TeV for specific dark sector parameters.

Original authors: ATLAS Collaboration

Published 2026-07-16

📖 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, bustling city where most of the buildings are made of a familiar, visible material called "Standard Model" matter. We know this city well; we've mapped its streets and named its citizens. But astronomers have long suspected that this visible city is just the tip of the iceberg. Hidden beneath the streets, in a vast, invisible underground network, lies a "Dark Sector." We can't see it directly because its inhabitants don't interact with light or the forces we use to build our world. However, we know this dark city must exist because its gravity holds the visible city together, preventing it from flying apart. The big mystery is: what does this dark city look like? Is it a quiet, empty void, or is it a chaotic metropolis with its own complex laws of physics, its own particles, and its own strange architecture?

Scientists are particularly interested in a theory called "Dark QCD." In our visible world, a force called Quantum Chromodynamics (QCD) acts like a super-strong glue, binding tiny particles called quarks together to form protons and neutrons. The Dark QCD theory suggests that the hidden dark sector has its own version of this glue, creating its own "dark quarks" that clump together into "dark hadrons." The exciting part is that these dark particles might not stay hidden forever. They could occasionally decay, turning into our familiar particles and leaving behind a very specific, weird signature: a jet of particles that seems to pop into existence in the middle of a detector, far away from where the collision started. These are called "emerging jets," and finding them would be like discovering a secret subway station that only opens its doors for a split second before vanishing again.

This paper is a report from the ATLAS experiment, a massive particle detector at the Large Hadron Collider (LHC) in Switzerland. The team acted like cosmic detectives, sifting through a mountain of data from 140 billion billion proton collisions (140 fb⁻¹) that happened between 2015 and 2018. They were hunting for a specific scenario: a pair of heavy, invisible "mediator" particles (think of them as dark messengers) being created in a collision. These messengers would instantly decay into a mix of normal particles and dark particles. The dark particles would then travel a short distance inside the detector before turning into a spray of normal particles, creating a "jet" that looks like it's emerging from thin air, complete with multiple "displaced vertices" (points where the decay happened away from the center).

To find these elusive signals, the scientists didn't just look for one thing; they built a sophisticated two-stage "digital detective" using machine learning. First, they trained a computer to recognize the unique shape and structure of a single jet that might be hiding dark secrets. Then, they trained a second computer to look at the whole event, checking if the four jets in the collision matched the pattern of a dark messenger pair decaying. They compared their findings against a massive background of ordinary particle collisions, which are like the noisy crowd at a concert that can easily drown out a whisper.

The result of this massive search? The detectors remained silent. The team found no evidence of emerging jets. The number of events they saw matched perfectly with the predictions for ordinary physics, with no mysterious excesses popping up. Because they didn't find the signal, they couldn't prove the dark messengers exist, but they did something equally important: they drew a very tight map of where these messengers cannot be. They ruled out the existence of these specific dark mediators if they have masses between 1 and 2 TeV (a unit of mass energy) and if the dark particles they create have a specific lifetime of 20 mm. In other words, if these dark messengers exist, they are either much heavier than 2 TeV, or they behave in a way that is totally different from the model the scientists were testing. The search continues, but for now, this particular version of the dark city remains hidden.

Technical Summary: Search for Emerging Jets in $pp$ Collisions at $\sqrt{s}=13$ TeV with the ATLAS Experiment

Problem and Motivation
Despite extensive evidence for Dark Matter (DM), its particle nature remains unknown. Theoretical models propose a "Dark Sector" (DS) governed by a confining gauge group analogous to Quantum Chromodynamics (QCD), often termed "Dark QCD." In such models, dark quarks ( $Q_d$ ) hadronize into dark mesons (dark pions, $\pi_d$ ) and baryons. While stable dark baryons could constitute DM, unstable dark mesons may decay back to Standard Model (SM) particles via a heavy scalar mediator ( $\phi$ ). Depending on the lifetime of the dark pions, these decays produce distinct signatures: prompt dark jets, semi-visible jets (containing missing transverse energy), or "emerging jets" (EJs). EJs are characterized by a high multiplicity of long-lived dark hadrons that decay within the detector volume, producing multiple displaced vertices (DVs) contained within a single jet cone. This paper presents a search for EJs arising from the pair-production of dark scalar mediators, a scenario not fully covered by previous searches for prompt or semi-visible jets.

Methodology
The analysis utilizes 140 fb $^{-1}$ of proton-proton collision data at $\sqrt{s}=13$ TeV collected by the ATLAS detector between 2015 and 2018 (Run 2).

Signal Model: The search targets the pair-production of bifundamental scalar mediators ( $\phi$ ) via $q\bar{q}$ annihilation and gluon-gluon fusion. Each mediator decays into a dark quark and a SM quark. The dark quark undergoes a shower in the dark sector, producing a collimated spray of dark pions that decay to SM down quarks ( $\pi_d \to d\bar{d}$ ) with a macroscopic lifetime. The final state topology consists of four jets: two SM jets and two emerging jets. The analysis assumes an unflavored model where the mediator couples only to SM down quarks.
Event Reconstruction: Standard ATLAS track reconstruction is supplemented by a dedicated "Large Radius Tracking" (LRT) algorithm to recover tracks with large impact parameters ( $|d_0| < 300$ mm). Secondary displaced vertices (DVs) are reconstructed using both standard and LRT tracks, subject to strict selection criteria to reject background from material interactions and long-lived SM particles (e.g., $K^0_S$ ).
Event Selection: Events are triggered on at least four jets with $p_T > 120$ GeV. Offline selection requires the scalar sum of the four leading jet $p_T$ ( $H_T$ ) to exceed 1 TeV. A key variable, the Prompt Track Fraction (PTF), is calculated for jets to distinguish SM-like jets from those containing displaced tracks. Events must contain at least one DV matched to a jet.
Classification Strategy: A two-stage Boosted Decision Tree (BDT) classifier is employed. The first stage operates at the jet level using observables like jet mass, width, and energy. The second stage operates at the event level, utilizing the BDT scores of the four leading jets, $H_T$ , minimum PTF, jet multiplicity, and transverse sphericity ( $S_T$ ).
Background Estimation: The dominant background is QCD multijet production, which is difficult to model accurately in simulation. A fully data-driven approach based on a modified ABCD method is used. The ABCD plane is defined by the event-level BDT response ( $R$ ) and the number of jet-matched displaced vertices ( $N_{DV}$ ). The signal region (SR) is defined by high $R$ and high $N_{DV}$ , while control regions (CRs) are used to estimate the background yield in the SR. A linear fit is applied to correct for correlations between the variables. Two validation regions (Low- $H_T$ and Shifted-ABCD) confirm the background estimation method.

Key Contributions

Novel Search Strategy: This analysis extends the search for Dark QCD signatures to the emerging jet topology in the four-jet final state, specifically targeting pair-produced scalar mediators. It complements previous Run 2 and Run 3 searches that focused on resonant vector mediators or two-jet topologies.
Advanced Machine Learning: The implementation of a two-stage BDT classifier effectively separates the complex EJ signature from the overwhelming QCD multijet background, leveraging both jet substructure and global event shape variables.
Robust Background Estimation: The application of a modified ABCD method with linear extrapolation allows for a precise, data-driven background estimate in a region where simulation is unreliable, validated through multiple control and validation regions.
Systematic Uncertainty Handling: The analysis includes a comprehensive treatment of systematic uncertainties, particularly those related to the reconstruction of displaced tracks and vertices (LRT), which are the dominant source of uncertainty for long-lived signal models.

Results
No significant excess of events above the Standard Model background expectation was observed.

Observed vs. Expected: In the signal region, 103 events were observed compared to a predicted background of $112 \pm 6$ (stat) $\pm 4$ (syst).
Exclusion Limits: Upper limits at the 95% confidence level (CL) were set on the pair-production cross-section of the scalar mediator ( $\sigma(pp \to \phi^\dagger\phi)$ $σ (pp \to ϕ^{†} ϕ)$ ).
- For dark pions with a proper decay length ( $c\tau_{\pi_d}$ ) of 20 mm and a mass ( $m_{\pi_d}$ ) of 20 GeV, mediator masses are excluded in the range of 1 to 2 TeV.
- Specifically, for $m_{\pi_d} = 20$ GeV and $c\tau_{\pi_d} = 20$ mm, mediator masses up to 2.00 TeV are excluded.
- For $m_{\pi_d} = 10$ GeV and $c\tau_{\pi_d} = 20$ mm, masses up to 1.99 TeV are excluded.
- For $m_{\pi_d} = 5$ GeV and $c\tau_{\pi_d} = 20$ mm, masses up to 1.95 TeV are excluded.
Lifetime Dependence: Sensitivity is highest for intermediate lifetimes ( $2 \text{ mm} < c\tau_{\pi_d} < 100 \text{ mm}$ ). Sensitivity decreases for very short lifetimes (due to reduced displacement) and very long lifetimes (due to decays occurring outside the tracker acceptance).
Comparison with CMS: The results are compared to recent CMS searches for emerging jets. While CMS sets stricter limits at very low and very high lifetimes using a Graph Neural Network (GNN), this ATLAS analysis demonstrates competitive sensitivity in the intermediate lifetime range ($2-300$ mm).

Significance
This paper establishes the most stringent limits to date on the pair-production of new scalar mediators in the emerging jet topology for specific mass and lifetime configurations. By probing cross-sections a factor of two smaller than previous ATLAS searches in the four-jet topology, this analysis significantly constrains the parameter space of Dark QCD models. The null result places strong constraints on the existence of dark sectors with QCD-like symmetries and scalar mediators coupling to down-type quarks, particularly for mediator masses up to 2 TeV and dark pion lifetimes around 20 mm. The methodology, particularly the data-driven background estimation and the two-stage ML classifier, provides a robust framework for future searches for long-lived particles in complex jet environments.

More like this