Search for electroweak scale dijet resonances in pile-up collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

This paper presents the first search for electroweak-scale dijet resonances in the 100–250 GeV mass range using pile-up collisions recorded by the ATLAS detector at $\sqrt{s}=13$ TeV, a novel strategy that bypasses high jet trigger thresholds to probe low-mass hadronic resonances without observing any significant excess over Standard Model expectations.

The Gist

Imagine the Large Hadron Collider (LHC) as the world's most powerful, high-speed particle collider. Every second, it smashes protons together, creating a chaotic storm of debris. Usually, scientists are looking for a specific, rare "treasure" hidden in that storm—a new particle that might explain the universe.

However, there's a problem. The storm is so loud and crowded with ordinary debris (called "background noise") that the detectors' "security guards" (triggers) have to set the alarm threshold very high. They only let in events with massive amounts of energy to avoid being overwhelmed. This means they miss the smaller, quieter, but potentially exciting events happening in the low-energy range. It's like trying to hear a whisper in a rock concert by only listening for people screaming.

The New Strategy: Listening to the "Crowd Noise"

This paper describes a clever trick the ATLAS collaboration used to hear those whispers.

Normally, when the LHC smashes protons, it doesn't just happen once per second. It happens in "bunches" of collisions. Sometimes, multiple collisions happen at the exact same time. Scientists call this "pile-up."

Think of it like a busy train station:

• The Triggered Event: A specific VIP passenger (a single electron or muon) gets off the train. The station security (the trigger) sees them, stops the train, and records everything about that VIP.
• The Pile-up: While the VIP is being checked, dozens of other regular passengers (other protons colliding) are also getting off the train in the same second.

In the past, scientists mostly ignored these "regular passengers" because they were just considered background noise. But in this study, the ATLAS team decided to look at them. They realized that even though the security guard was busy watching the VIP, the cameras were still recording the regular passengers.

How They Did It

1. The VIP Filter: They selected data where a "VIP" (a high-energy electron or muon) was detected. This ensured they had a valid recording of that moment.
2. The Crowd Scan: Instead of just studying the VIP, they went back into the recording and looked at all the other collisions happening in that same split second. They treated these "pile-up" collisions as their own separate events.
3. The Search: They looked for pairs of "jets" (sprays of particles) in these pile-up collisions that might have come from a new, low-mass particle.

Why This Matters

This is like realizing that while you were interviewing the CEO of a company, you could also analyze the conversations happening in the breakroom right next door. You get a huge amount of extra data without needing to set up a new interview.

By using this method, they effectively created a new dataset of 1.30 inverse picobarns of data. While this sounds small compared to the total data ATLAS collects, it is a massive amount of low-energy data that was previously inaccessible because the "security guards" would have blocked it.

What They Found

They scanned this new dataset for a mass range between 100 and 250 GeV (a relatively low energy scale). They were looking for:

• Standard Model particles: Like the W and Z bosons (which they expected to see but didn't find clearly).
• New Physics: Specifically, a hypothetical particle called a Z-prime (Z') that could be a bridge to "Dark Matter," or other generic new particles.

The Verdict

The result? No new treasure was found.

The data looked exactly like what the Standard Model (our current best theory of physics) predicts. There were no strange spikes or "bumps" in the data that would indicate a new particle.

However, this isn't a failure. It's a success in a different way. Because they didn't find anything, they can now say with high confidence: "If a new particle like a Z' exists in this specific mass range, it must be very rare or interact very weakly." They set strict limits on how heavy or how strongly it could interact, effectively narrowing the search area for future experiments.

In Summary
The ATLAS team used a clever "recycling" strategy to look at the "trash" (pile-up collisions) that usually gets thrown away. They turned it into a new, clean dataset to search for low-energy particles. They didn't find any new particles, but they successfully proved that this new method works and ruled out several possibilities for what new physics might look like in that specific energy range.

Technical Summary

Technical Summary: Search for Electroweak Scale Dijet Resonances in Pile-up Collisions

Problem Statement
Conventional searches for hadronic resonances in the sub-TeV mass regime at the Large Hadron Collider (LHC) are severely constrained by the overwhelming rate of Standard Model (SM) multijet background processes described by Quantum Chromodynamics (QCD). To manage data rates, experiments typically impose high jet transverse momentum ($p_T$) trigger thresholds, which effectively exclude low-mass resonances (100–250 GeV) from analysis. While alternative strategies exist—such as storing reduced information for trigger-level analysis or requiring associated high-$p_T$ initial-state radiation (ISR) objects like photons or jets—these approaches suffer from intrinsic limitations. Trigger-level analyses lack full offline reconstruction capabilities, and ISR-based searches suffer from reduced acceptance and rely on specific topological assumptions (e.g., $p_T$-balanced decays) that may not hold for all new physics models, such as those involving dark QCD or gluon-initiated production.

Methodology
This paper presents a novel search strategy utilizing a unique dataset where multiple proton-proton ($pp$) interactions (pile-up) occurring within a single bunch crossing are reconstructed separately. The analysis uses data collected by the ATLAS detector during 2016–2018 at a center-of-mass energy of $\sqrt{s} = 13$ TeV.

• Dataset Construction: The dataset is derived from events triggered by single-electron or single-muon triggers (with a 26 GeV $p_T$ requirement). For each recorded bunch crossing (RBC), the "triggering primary vertex" (TPV)—associated with the lepton that fired the trigger—is identified and removed. The remaining "pile-up primary vertices" (PPVs) are analyzed as independent collision events. This method yields a trigger-unbiased dataset with an effective integrated luminosity of 1.30 pb$^{-1}$.
• Object Reconstruction: The analysis employs standard ATLAS object reconstruction extended to all primary vertices. Jets are reconstructed using the anti-$k_t$ algorithm ($R=0.4$) with Particle Flow Objects (PFOs) as inputs. A Jet-Vertex Tagger (JVT) is applied to ensure jets are consistent with their associated vertex.
• Selection Criteria: To ensure a trigger-unbiased sample, the TPV and any vertices within $\Delta z < 2$ mm are vetoed. The analysis selects PPVs containing at least two jets with $p_T > 17$ GeV and $|\eta| < 2.0$. Additional requirements include geometrical isolation between jets from different vertices, timing cuts to reject out-of-time pile-up, and a cut on the rapidity difference $|y^*| < 0.8$ to enhance resonant signal topologies over the $t$-channel QCD background.
• Background Estimation: The non-resonant QCD multijet background is modeled using a smoothly falling functional form fitted to the dijet invariant mass ($m_{jj}$) spectrum. The fit includes a turn-on term to account for threshold effects. The compatibility of the data with the background-only hypothesis is tested using $\chi^2$ $p$-values and the BumpHunter algorithm.
• Signal Models: The search interprets results in the context of:
  1. A simplified dark matter model featuring an axial-vector mediator ($Z'$) coupling to quarks.
  2. A generic Gaussian signal model.
  3. SM $W/Z$ boson production (modeled as double-Gaussian templates).

Key Contributions

• First Application of Pile-up for Dijet Resonances: This work constitutes the first application of pile-up collisions to probe low dijet mass scales, demonstrating that full offline reconstruction capabilities can be retained while accessing a regime typically forbidden by trigger thresholds.
• Trigger-Unbiased Low-Mass Access: The analysis provides a dataset with competitive statistical power for signatures with low jet trigger acceptance, offering a factor of 30 higher luminosity compared to the ATLAS ZeroBias stream and significantly higher statistics than prescaled single-jet triggers in the 100–140 GeV mass range.
• Model Independence: Unlike ISR-based searches, this resolved dijet strategy does not rely on substructure selections or assumptions about $p_T$-balanced decays, retaining sensitivity to models such as dark QCD where part of the shower may be invisible, and to resonances coupling predominantly to gluons.

Results

• Background Compatibility: The observed dijet invariant mass spectrum is consistent with the background-only hypothesis. The fit yields a $\chi^2$ $p$-value of 0.08. No significant excess is observed above the background expectation.
• Local Excesses: The BumpHunter algorithm identified a local excess between 106 and 131 GeV with a local significance of $1.7\sigma$ (global $p$-value 0.67). A specific $Z'$ signal hypothesis at $m_{Z'} = 112$ GeV showed a local significance of $3.0\sigma$, corresponding to a global significance of $2.3\sigma$ after accounting for the look-elsewhere effect; this is not considered a discovery.
• Exclusion Limits:
  • $Z'$ Model: For an axial-vector mediator with coupling $g_q = 0.6$, masses between 135 and 185 GeV are excluded at 95% confidence level (CL). The lowest excluded coupling is $g_q = 0.28$ at a mass of approximately 150 GeV.
  • Gaussian Model: Upper limits on the cross-section times acceptance times branching ratio ($\sigma \times A \times B$) are set for Gaussian resonances with relative widths of 10–25%. Limits range from 1.5 nb to 15 nb depending on the mass and width.
  • SM $W/Z$: While the analysis is sensitive to hadronic $W/Z$ decays, no significant signal is observed. The 95% CL upper limit on the yield is consistent with the expected SM contribution, though the sensitivity is near the threshold required for detection.

Significance
The paper claims that this analysis complements the existing ATLAS low-mass search programme by providing an inclusive search with full object reconstruction and without assumptions regarding initial-state radiation. It successfully demonstrates the feasibility of using pile-up collisions to probe electroweak scale resonances, a mass region previously inaccessible to standard trigger-based dijet searches due to prohibitive background rates. The results set competitive exclusion limits for simplified dark matter models and generic resonances, validating the pile-up reconstruction technique as a viable tool for extending the physics reach of the LHC into the low-mass regime.