⚛️ phenomenology

Probing light axion-like particles in vector boson fusion at CMS with data parking and scouting

This paper proposes a novel strategy using CMS data parking and scouting techniques combined with tracker-based reconstruction of merged photon pairs to overcome trigger thresholds and resolve boosted decays, thereby enabling the first sensitivity to light axion-like particles in the 10 MeV–10 GeV mass range through vector boson fusion.

Original authors: Sena Durgut, Mariel Peczak, Gonzalo Alonso-Álvarez, Chiara Amendola, Matteo Cremonesi, Joerg Jaeckel, Matteo Marchegiani

Published 2026-09-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sena Durgut, Mariel Peczak, Gonzalo Alonso-Álvarez, Chiara Amendola, Matteo Cremonesi, Joerg Jaeckel, Matteo Marchegiani

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

The universe is filled with particles that we can see and measure, from the heavy atoms that make up our world to the fleeting flashes of light that reveal the inner workings of stars. But physicists suspect there is a whole hidden layer of nature sitting in a difficult middle ground. Imagine a particle that is too light to be caught by the massive, high-energy collisions designed to find heavy new physics, yet too heavy and short-lived to be spotted by the ultra-sensitive, low-energy experiments that hunt for the faintest whispers of new forces. This "blind spot" exists for particles with masses between the weight of a tiny speck of dust and a heavy atom, interacting so weakly with ordinary matter that they slip through the cracks of current detection methods. One leading candidate for such a particle is the axion-like particle, a ghostly entity that could help explain why the universe looks the way it does and what makes up the invisible dark matter holding galaxies together. Finding these particles would be a monumental step in understanding the fundamental rules of reality, but catching them requires a strategy that is as clever as it is precise.

A team of researchers has proposed a new way to hunt for these elusive particles using the world's largest particle collider, the Large Hadron Collider, and specifically the Compact Muon Solenoid, or CMS, detector. The challenge is that when these light particles are created in collisions, they decay almost instantly into two beams of light, or photons. Because the parent particle is moving so fast, these two beams of light are squeezed so tightly together that they look like a single beam to the giant sensors designed to catch them. Furthermore, the energy of these beams is often too low to trigger the computer systems that decide which collisions are worth recording, meaning most of these events are discarded before scientists can even look at them. The researchers realized that while the sensors might see a single merged beam, the silicon tracker inside the detector, which is like a high-resolution digital camera, might still be able to see the two separate paths if the light beams hit a piece of material and split into pairs of electrons and positrons. By tracking these split paths, they can pinpoint exactly where the particle decayed, which is slightly away from the main collision point, allowing them to distinguish their signal from the overwhelming noise of ordinary particle interactions.

To make this work, the team combined two specific data-collection strategies already available at the CMS experiment. The first, used with data already collected in 2022 and 2023, involves a technique called data parking. Normally, the experiment records a massive amount of data but immediately throws away anything that doesn't meet strict energy requirements. With data parking, the experiment saves the raw information of collisions that look like they might contain the right type of particle jets, even if the light beams are too faint to trigger the usual alarms. This allows scientists to go back later, when they have more computing power, to reconstruct the events in full detail and look for those faint, merged light beams. The second strategy, planned for the future High-Luminosity LHC, is called data scouting. This method records a compact summary of every single collision, stripping away the heavy raw data but keeping the essential information about particle tracks and energy. This allows the experiment to keep its eyes open for rare events without being overwhelmed by the sheer volume of data, effectively removing the energy thresholds that usually block these soft signals.

The researchers simulated how well these methods would work, focusing on a specific way these particles are made, where two protons exchange force-carrying particles to create the new particle, leaving behind two distinct jets of debris. They found that by looking for the specific signature of a particle decaying slightly away from the collision point and using the high-resolution tracker to separate the merged light beams, they could filter out the background noise of ordinary particles. A major hurdle they addressed was a long-lived particle called the neutral kaon, which can also decay into two light beams and mimic the signal. In the data already collected, they showed that they could distinguish the new particle from this background by measuring the exact mass of the light beams, which would be different for the new particle. For the future high-luminosity data, where such a precise mass measurement is harder to perform, they demonstrated that the speed and direction of the light beams, combined with the surrounding particle activity, would be enough to separate the signal from the background.

The results of this study suggest that the CMS experiment can now explore a region of the universe that has never been tested before. The data already sitting in the archives from the recent run could reveal these particles if they exist within a specific range of masses and interaction strengths. If they are not found there, the future data from the High-Luminosity LHC, collected with the new scouting methods, could push the search even further, potentially reaching particles that are ten times lighter or interact ten times more weakly than what is currently possible. This approach does not require building a new machine or waiting for a new technology; it relies on rethinking how to look at the data that is already being produced. By turning the detector's ability to see fine details and the experiment's ability to save more data into a powerful combination, the researchers have opened a new window into the hidden sector of physics, offering a realistic path to discovering particles that have so far remained just out of reach.

Technical Summary: Probing Light Axion-like Particles in Vector Boson Fusion at CMS with Data Parking and Scouting

Problem Statement
Axion-like particles (ALPs) with masses between 10 MeV and 10 GeV and moderately small couplings ( $10^{-3}$ – $10^{-5}$ GeV $^{-1}$ ) occupy an experimental "blind spot." They decay too rapidly for intensity-frontier experiments (where decay products are absorbed by shielding) and are too light and feebly coupled for conventional collider searches. At the Large Hadron Collider (LHC), electromagnetically coupled ALPs are produced at substantial rates via Vector Boson Fusion (VBF), the dominant production mode. However, two primary obstacles prevent their detection:

Trigger Thresholds: The produced ALPs are soft (low transverse momentum, $p_T \sim 5\text{--}10$ GeV), causing standard trigger protocols to discard the events.
Merged Photons: Due to the high Lorentz boost of light ALPs, the two decay photons are highly collimated. For masses below $\sim 1$ GeV, their angular separation falls below the granularity of the electromagnetic calorimeter (ECAL), causing them to merge into a single energy deposit that mimics a single photon.

Methodology
The authors propose a strategy combining theoretical reconstruction techniques with specific CMS data-acquisition modes to overcome these obstacles. The approach relies on three core components:

Photon Conversion and Tracker Reconstruction:
Instead of relying on the ECAL, the strategy exploits the fraction of photons that convert into $e^+e^-$ pairs within the silicon tracker material. The tracker provides an angular resolution two orders of magnitude finer than the ECAL granularity. By reconstructing the conversion tracks, the two collimated photons can be separated, and the displaced ALP decay vertex can be reconstructed. This displacement ( $\tilde{b} > 1$ cm) serves as a primary discriminant against prompt Standard Model backgrounds (e.g., $\pi^0 \to \gamma\gamma$ ).
Data Acquisition Strategies:
- Run 3 (Data Parking): For the 312 fb $^{-1}$ dataset, the authors utilize "data parking." Events are selected based solely on the VBF jet topology (forward jets, large dijet mass, and pseudorapidity gap) without imposing thresholds on the ALP decay products. The raw data is stored for later offline reconstruction. This allows for the reconstruction of the merged diphoton invariant mass ( $m_{\gamma\gamma}$ ) using machine learning techniques (e.g., ClusTEX) applied to ECAL showers and tracks.
- HL-LHC (Data Scouting): For the projected 3 ab $^{-1}$ dataset, the authors propose "Level-1 (L1T) data scouting." This records a compact summary of events (tracks, calorimeter clusters) at the trigger level with virtually no thresholds. While this precludes full offline invariant mass reconstruction, it retains the full event rate for soft signals. Discrimination in this stream relies on the diphoton $p_T$ , conversion track kinematics, and hadronic activity.
Background Suppression:
- Prompt Backgrounds: Removed by the requirement of a displaced vertex ( $\tilde{b} > 1$ cm).
- Long-lived Background ( $K_L \to \gamma\gamma$ ): This is the dominant surviving background.
  - In Run 3, the $K_L$ is suppressed by reconstructing the invariant mass; $K_L$ decays cluster at $m_{K_L} \approx 0.5$ GeV, allowing separation from ALPs with $m_a \neq m_{K_L}$ .
  - In HL-LHC scouting, where mass reconstruction is unavailable, suppression relies on the diphoton $p_T$ (signal is harder than the soft $K_L$ spectrum), the isolation of the photon pair (ALPs recoil against VBF jets; $K_L$ are produced inside jets), and quark-jet tagging (using algorithms like UParT).

Key Contributions

Feasibility Study: The paper demonstrates that the CMS experiment can access the 10 MeV–10 GeV ALP parameter space by combining VBF topology triggers with tracker-based photon conversion reconstruction.
Strategic Integration: It integrates the methodology of Ref. [38] (tracker-based reconstruction of merged photons) with CMS-specific data acquisition tools (parking and scouting) that were previously unutilized for this specific mass range.
Background Handling: It provides a detailed analysis of the $K_L \to \gamma\gamma$ background, distinguishing between strategies for Run 3 (mass-based rejection) and HL-LHC (kinematic and topological rejection).
Sensitivity Projections: The authors project the reach in the $(m_a, g_{a\gamma\gamma})$ plane for both the existing Run 3 parked data and the future HL-LHC scouting stream.

Results

Run 3 Reach: The analysis of the 312 fb $^{-1}$ parked dataset is projected to probe a region of the parameter space that has not been explored by previous measurements. This reach is achieved without new data taking, utilizing events already recorded.
HL-LHC Reach: The L1T data scouting projection at 3 ab $^{-1}$ extends the sensitivity by roughly an order of magnitude in both coupling ( $g_{a\gamma\gamma}$ ) and mass ( $m_a$ ) compared to Run 3.
Background-Free Assumption: The projections assume that the $K_L \to \gamma\gamma$ background can be suppressed to negligible levels. For Run 3, this relies on the mass resolution of the merged diphoton system. For HL-LHC, it relies on the $p_T$ separation and topological isolation. The authors note that the $K_L$ background is significant near $m_a \approx 0.5$ GeV, where mass separation is ineffective.

Significance and Claims
The paper claims that the breadth of CMS's trigger and data-acquisition strategies is the critical enabler for this search. By relaxing photon thresholds through VBF-based selection and utilizing the tracker's superior angular resolution, the experiment can overcome the "softness" and "collimation" barriers that have rendered this mass range inaccessible.

The authors modestly state that while the projected reach is significant, the suppression of the $K_L$ background in the HL-LHC scouting stream requires further development of trigger-level mass estimators or multivariate discriminants. They emphasize that the parked Run 3 data represents an immediate opportunity to produce the first LHC result on VBF-produced light ALPs, as the necessary data is already recorded. The study does not claim a discovery but rather defines a viable path to probe a previously untested region of the ALP parameter space.

Technical Summary: Probing Light Axion-like Particles in Vector Boson Fusion at CMS with Data Parking and Scouting

More like this