⚛️ high-energy experiments

Search for soft unclustered energy patterns in proton-proton collisions at $\sqrt{s}$ = 13 TeV using data scouting

Using 127 fb $^{-1}$ of 13 TeV proton-proton collision data collected via the CMS data scouting stream, this study searches for soft unclustered energy patterns predicted by hidden-valley models, finds no evidence beyond standard model backgrounds, and sets the most stringent limits to date on gluon fusion production of heavy scalar mediators.

Original authors: CMS Collaboration

Published 2026-07-13

📖 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 Large Hadron Collider (LHC) as the world's most energetic particle smasher, where protons zoom around at nearly the speed of light and crash into each other like microscopic billiard balls. Usually, when these balls collide, they shatter into neat, predictable piles of debris—like a car crash that leaves recognizable fenders and wheels. Physicists call these "jets."

But what if, instead of a neat crash, the collision created a chaotic, fuzzy cloud of thousands of tiny, slow-moving particles, spreading out in every direction like a dandelion seed head exploding in a gentle breeze? This is the "Soft Unclustered Energy Pattern," or SUEP for short.

According to some theories about a "Hidden Valley" of the universe—a secret sector of particles that don't play by our usual rules—these fuzzy clouds should exist. They would be the result of a heavy, invisible "mediator" particle decaying into a swarm of dark particles that then turn into the ordinary stuff we can see.

The Great Hunt
The CMS experiment at CERN decided to go on a treasure hunt for these fuzzy clouds. They looked at a massive pile of data: 127 fb⁻¹ of proton-proton collisions at an energy of 13 TeV. That's a lot of collisions!

However, there was a catch. These fuzzy clouds are made of very low-energy particles. In a normal search, the experiment's "security guards" (the triggers) would ignore them, thinking they were just background noise or a glitch. To catch them, the CMS team used a special trick called data scouting.

Think of data scouting like a bouncer at a club who usually checks everyone's ID and takes a full photo. With data scouting, the bouncer only takes a quick snapshot of the most important details and lets the party start faster. This allowed the scientists to lower the "entry threshold" for the particles they were looking for, dropping the requirement from 1 TeV down to 410 GeV. This was crucial because it let them see the faint, fuzzy signals that would have otherwise been missed.

The Search Strategy
The scientists looked for events where a big jet of energy (from the crash) recoiled against a second, strange jet. They didn't just look at the second jet; they looked inside it. They asked: "Is this jet a neat, two-pronged stick, or is it a perfectly round, isotropic ball of energy?"

They used a mathematical tool called "sphericity" to measure this. A value of 0 means a flat stick; a value of 1 means a perfect sphere. They were hunting for jets with a sphericity greater than 0.5 and more than 50 tiny particles inside them.

The Result: The Quiet Cloud
After sifting through the data, the team found something very specific: nothing.

The number of fuzzy, spherical events they saw matched exactly what they expected from the Standard Model (the known rules of physics). There were no extra fuzzy clouds hiding in the data. The observed results were consistent with the background prediction.

What This Means
Because they didn't find the fuzzy clouds, they didn't discover a new hidden world. Instead, they did something just as important: they drew a very tight fence around where that hidden world could be.

They set the most stringent limits to date on the production of these heavy scalar mediators. In simple terms, they proved that if these heavy mediators exist, they are either much heavier than they thought, or they are much rarer than their models predicted. They ruled out a huge range of possibilities, especially for models where the dark particles have a "temperature" and mass around a few GeV.

The Bottom Line
The paper doesn't claim to have found the Hidden Valley. Instead, it says, "We looked very hard with our new, sensitive net, and we didn't find the fish we were hoping for." This tells future physicists that if the Hidden Valley is real, it's hiding in a much deeper, darker corner than the models suggested. The search continues, but the map of where to look has just gotten a lot more precise.

Technical Summary: Search for Soft Unclustered Energy Patterns in Proton-Proton Collisions at $\sqrt{s} = 13$ TeV Using Data Scouting

Problem Statement
Hidden-valley models featuring a strongly coupled dark sector (DS) offer a compelling framework for physics beyond the Standard Model (SM), potentially addressing the hierarchy problem and providing dark matter candidates. A specific signature predicted by these models, particularly those with a large 't Hooft coupling ( $\lambda \gg 1$ ), is the Soft Unclustered Energy Pattern (SUEP). SUEPs manifest as high-multiplicity events with diffusely distributed, low-momentum charged particles that are isotropic in the center-of-mass frame.

The primary experimental challenge in detecting SUEPs at the Large Hadron Collider (LHC) is their kinematic similarity to SM backgrounds, specifically quantum chromodynamics (QCD) multijet events. Standard trigger strategies, which rely on high transverse momentum ( $p_T$ ) thresholds to manage data rates, often fail to capture SUEP events because the constituent particles are soft and unclustered. Previous searches utilizing full offline reconstruction required high scalar sums of jet transverse momenta ( $H_T > 1$ TeV), resulting in significant signal loss for models with lower mediator masses or specific temperature parameters.

Methodology
This analysis employs the "data scouting" strategy of the CMS experiment to lower the trigger threshold for hadronic activity while maintaining a manageable data rate.

Data Scouting Strategy: Instead of recording raw event data, the High-Level Trigger (HLT) records only reconstructed objects (Particle-Flow candidates, jets, and primary vertices). This reduces the event size by a factor of $\approx 100$ , allowing the $H_T$ trigger threshold to be lowered from 1 TeV (used in previous offline searches) to 410 GeV. This significantly increases the acceptance for SUEP signals, particularly those with temperatures ( $T_D$ ) and dark-meson masses ( $m_\phi$ ) $\lesssim 2$ GeV.
Dataset: The search utilizes proton-proton collision data collected at $\sqrt{s} = 13$ TeV during 2016–2018, corresponding to an integrated luminosity of 127 fb $^{-1}$ .
Signal Model: The analysis targets the production of a heavy scalar mediator ( $S$ ) via gluon-gluon fusion (ggF), which decays into a SUEP. The dark sector showering is modeled using a relativistic Boltzmann distribution characterized by a dark-meson mass ( $m_\phi$ ) and a temperature ( $T_D$ ). Dark mesons decay into dark photons ( $A'$ ), which subsequently decay into SM particles ( $e^+e^-$ , $\mu^+\mu^-$ , or $\pi^+\pi^-$ ).
Event Selection:
- Events must pass a high- $H_T$ trigger (threshold 410 GeV) and a preselection requiring $H_T > 560$ GeV based on anti- $k_T$ jets ( $R=0.4$ ).
- Two large-radius jets ( $R=1.5$ , AK15) are reconstructed. The jet with the higher constituent multiplicity is identified as the SUEP candidate, recoiling against the Initial State Radiation (ISR) system.
- To recover the isotropic event shape, constituents of the SUEP candidate are boosted to the jet rest frame. The generalized sphericity ( $S_{\text{boosted}}$ ) is calculated.
- Signal regions (SR) are defined by requiring the number of constituents ( $n_{\text{constituent}}$ ) $> 50$ and $S_{\text{boosted}} > 0.5$ .
Background Estimation: The dominant background (QCD multijets) is estimated using an extended ABCD method. The phase space is partitioned into a $3 \times 3$ grid based on two uncorrelated variables: $n_{\text{constituent}}$ and $S_{\text{boosted}}$ . This approach improves the treatment of higher-order correlations between these variables compared to standard ABCD methods. A correction factor ( $c_{\text{bin}}$ ) is derived from control regions to account for shape deviations in the signal region.
Statistical Analysis: A binned maximum likelihood fit is performed on the $n_{\text{constituent}}$ distribution across control and signal regions, treating data-taking years separately. Systematic uncertainties, including tracking efficiency (up to 50% for high-multiplicity bins) and luminosity, are incorporated.

Key Contributions

Lowered Trigger Thresholds: By utilizing data scouting, the analysis successfully lowers the $H_T$ threshold to 410 GeV, extending sensitivity to SUEP models that were previously inaccessible due to trigger inefficiencies.
Extended ABCD Method: The implementation of a $3 \times 3$ grid in the ABCD method allows for a more robust estimation of QCD backgrounds in the presence of correlations between event multiplicity and sphericity.
Comprehensive Parameter Scan: The search covers a broad range of mediator masses ( $m_S$ from 125 GeV to 2 TeV), dark-meson masses ( $m_\phi$ from 1 to 8 GeV), and temperatures ( $T_D$ ), including scenarios with leptonic and hadronic dark photon decays.

Results

Observation: The observed data in the signal region are consistent with the Standard Model background prediction. No significant excess of events is observed.
Limits: The analysis sets the most stringent limits to date on the production cross-section of heavy scalar mediators decaying into SUEP-like signals.
- For mediator masses $m_S$ between 125 GeV and 2 TeV, the observed limits improve upon previous offline reconstruction searches by up to one order of magnitude in most of the parameter space.
- Exclusion limits are presented in the $(T_D, m_\phi)$ plane for various $m_S$ values. The region below the exclusion curves is ruled out for the tested signal models.
- Sensitivity is reduced for signals with very low $n_{\text{constituent}}$ due to the lower tracking efficiency inherent to the scouting trigger compared to offline reconstruction.

Significance and Claims
The paper claims that this search represents the most sensitive probe to date for SUEP signatures arising from hidden-valley models with large 't Hooft couplings. By leveraging the data scouting stream, the CMS collaboration has successfully mitigated the trigger bias against soft, high-multiplicity final states. The results provide robust constraints on simplified models of dark sector showers, excluding large portions of the parameter space for scalar mediators with masses above 125 GeV and dark sector temperatures and masses in the few GeV range. The analysis demonstrates the efficacy of data scouting strategies in uncovering unconventional signatures that would otherwise remain hidden behind standard trigger thresholds.

More like this