Search for light pseudoscalar bosons, pair-produced in Higgs boson decays in the four-electron final state in proton-proton collisions at $\sqrt{s}$ = 13 TeV

Using 138 fb$^{-1}$ of 13 TeV proton-proton collision data collected by the CMS detector, this study presents the first LHC search for Higgs boson decays into pairs of light pseudoscalar bosons that subsequently decay into four electrons, finding no significant excess and setting stringent upper limits on the branching fraction down to $10^{-5}$ for pseudoscalar masses between 10 and 100 MeV.

The Gist

The Big Picture: Hunting for Invisible Ghosts in a Giant Collision

Imagine the Large Hadron Collider (LHC) at CERN as a massive, high-speed racetrack where protons (tiny subatomic particles) are smashed together at nearly the speed of light. Usually, these collisions create a chaotic explosion of known particles, like a car crash scattering debris everywhere.

Physicists are looking for something new hiding in that debris: Axion-Like Particles (ALPs). Think of these ALPs as "ghosts." They are very light, very shy, and interact very weakly with normal matter. The Standard Model of physics (our current rulebook for how the universe works) doesn't fully explain things like dark matter or why the universe behaves the way it does, so scientists suspect these "ghosts" might be the missing pieces.

The Specific Hunt: The "Four-Electron" Trail

This paper describes a specific search conducted by the CMS experiment (one of the giant detectors at the LHC). Here is the strategy they used, broken down simply:

1. The Source: The Higgs Boson
Scientists know the Higgs boson exists (it's the particle that gives other particles mass). They hypothesize that sometimes, instead of decaying into the usual suspects, a Higgs boson might decay into two of these "ghost" ALPs.

• Analogy: Imagine a heavy bowling ball (the Higgs) rolling down a lane. Usually, it hits a pin and stops. But in this theory, sometimes it splits into two tiny, invisible marbles (the ALPs) that zoom away.

2. The Decay: The "Ghost" Turns Visible
These ALPs are unstable. They don't last long. They quickly decay into pairs of electrons and positrons (anti-electrons).

• The Catch: Because these ALPs are so light and moving so fast, the electron and positron they produce are squeezed incredibly close together. They are so close that they look like a single, merged blob to the detector.
• Analogy: Normally, if a firecracker explodes, you see two sparks flying apart. But if the explosion happens inside a super-tight tube, the two sparks fly out so close together they look like one single, bright streak of light.

3. The Challenge: Seeing the Unseeable
The CMS detector is amazing, but it's not perfect. Usually, when two particles are this close, the detector's "eyes" (specifically the calorimeter, which measures energy) can't tell them apart. It just sees one big electron.

• The Innovation: The team developed a new, super-smart computer algorithm (a "multivariate algorithm") that acts like a high-powered microscope. Instead of just looking at the energy blob, it looks at the tiny tracks left by the particles in the silicon tracker. It can tell, "Hey, this isn't one electron; it's two electrons hugging each other so tightly they look like one." They call these merged pairs MEPs (Merged Electron-Positron pairs).

4. The Search Strategy
The scientists looked at 138 "years" of collision data (a massive amount of information). They asked the computer to find events where:

1. A Higgs boson was created.
2. It decayed into two ALPs.
3. Each ALP decayed into a merged electron-positron pair.
4. Result: They were looking for a total of four electrons in the final event, but arranged in two tight, merged pairs.

The Results: The "Silence" is the News

After sifting through the data, the team found no evidence of these ALPs.

• The Analogy: Imagine you are listening for a specific, rare bird song in a noisy forest. You have the best microphones and the smartest software to filter out the wind and other birds. You listen for months. You don't hear the song.
• What this means: While they didn't find the "ghosts," the fact that they didn't find them is actually a huge success. It tells us that if these ghosts exist, they are even more elusive than we thought.

The New Limits: Drawing the Map

Because they didn't find the particles, they set a "boundary line" on the map of the universe.

• They proved that if these ALPs exist with masses between 10 and 100 MeV (very light), they cannot be produced by the Higgs boson more than a tiny fraction of the time (less than 1 in 100,000 times).
• They also ruled out certain "lifetimes" for these particles. If the particles lived too long or decayed too quickly, they would have been seen.

Why This Matters

This is the first time anyone has looked for this specific "four-electron" signature at the LHC.

• Previous searches looked for photons (light particles) or heavier particles.
• This search pushed the boundary down to very low masses (10 MeV), a region that was previously "blind" to the LHC.
• By developing the new algorithm to see these "merged" electron pairs, they have built a better net for catching these elusive particles in the future.

In summary: The scientists built a super-advanced net to catch a specific type of "ghost" particle that might be hiding in Higgs boson collisions. They cast the net wide, but the net came up empty. However, by coming up empty, they have proven that these ghosts are either not there, or they are even harder to catch than we hoped, effectively narrowing the search area for future experiments.

Technical Summary

Technical Summary: Search for Light Pseudoscalar Bosons in Higgs Decays to Four Electrons

Problem Statement
The Standard Model (SM) of particle physics, while successful, leaves fundamental questions unanswered, including the strong CP problem and the nature of dark matter (DM). The Peccei-Quinn mechanism proposes the existence of the axion, and recent theoretical studies suggest that axionlike particles (ALPs) with masses in the tens-of-MeV range are viable. These light ALPs could serve as mediators between DM and SM particles or as DM candidates themselves. While previous LHC searches have constrained ALPs originating from Higgs boson decays ($H \to AA$) down to masses of approximately 100 MeV, the region below 100 MeV, particularly down to 10 MeV, remains largely unexplored. The primary experimental challenge in this low-mass regime is that ALPs with momenta above 30 GeV decay into highly collimated electron-positron pairs ($e^+e^-$). Standard reconstruction algorithms typically merge these pairs into a single electron or misidentify them as converted photons, rendering them invisible to conventional triggers and identification methods.

Methodology
This analysis utilizes proton-proton collision data collected by the CMS detector at $\sqrt{s} = 13$ TeV during 2016–2018, corresponding to an integrated luminosity of 138 fb$^{-1}$. The search targets the process $H \to AA \to 4e$, where the ALP ($A$) decays into an electron-positron pair.

• Reconstruction of Merged Electron-Positron Pairs (MEPs): To address the collimation challenge, the analysis develops a dedicated reconstruction algorithm for MEPs. This involves matching two Gaussian-sum filter (GSF) tracks from the silicon tracker to the same electromagnetic calorimeter (ECAL) cluster. The tracker's superior angular resolution compared to the ECAL allows for the resolution of pairs with opening angles that would otherwise be merged. Selection criteria include transverse momentum ($p_T > 5$ GeV for tracks, $E > 25$ GeV for clusters) and vertex fitting requirements.
• Identification via Machine Learning: A Gradient Boosted Decision Tree (GBDT) classifier, implemented using XGBoost, is employed to distinguish signal MEPs from backgrounds. Backgrounds include photon conversions, electrons with nearby tracks, and jets mimicking the MEP signature. The classifier utilizes observables such as ECAL shower shapes, angular separation between tracks, and vertex displacement relative to the primary vertex. The model is trained separately for different data-taking periods to account for detector condition variations.
• Event Selection and Background Modeling: Events are selected requiring two identified MEPs. The Higgs boson candidate is reconstructed from the invariant mass of the two MEPs ($m_{4e}$), assuming the ALPs are massless relative to the Higgs mass. The background is modeled using a data-driven approach:
  • Non-resonant background: Modeled by fitting the sidebands of the 125 GeV Higgs mass peak (100–180 GeV, excluding 115–135 GeV) using various parametric functions (polynomials, exponentials, etc.), with model uncertainty handled via the discrete profiling method.
  • Resonant background: Primarily from $H \to \gamma\gamma$ where photons convert to $e^+e^-$ pairs. This is modeled using simulated events fitted with parametric shapes (sum of Gaussians or double-sided Crystal Ball functions).
• Statistical Analysis: An unbinned profile-likelihood fit is performed on the $m_{4e}$ spectrum using the COMBINE package. Systematic uncertainties, including identification efficiencies, trigger efficiencies, and luminosity, are incorporated as nuisance parameters.

Key Contributions

• Novel Reconstruction Technique: The development of a multivariate algorithm specifically designed to identify highly collimated electron-positron pairs using combined tracker and calorimeter information. This enables sensitivity to ALP masses as low as 10 MeV, a regime previously inaccessible due to the merging of decay products.
• First LHC Search in this Channel: This work represents the first search for Higgs boson decays to four electrons via light pseudoscalars at the LHC.
• Extended Mass Reach: The analysis significantly extends the experimental sensitivity to ALPs with masses below 100 MeV, probing a parameter space where theoretical motivations (such as the 17 MeV anomaly) are active.

Results
No significant excess above the Standard Model background predictions was observed in the four-electron invariant mass spectrum. Consequently, upper limits on the branching fraction for $H \to AA \to 4e$ were set at the 95% confidence level (CL).

• Sensitivity: The analysis achieves branching fraction sensitivities as low as $10^{-5}$ for ALP masses between 10 and 100 MeV and proper decay lengths ($c\tau$) below 100 $\mu$m.
• Limits: The observed limits are generally consistent with expected limits, with deviations remaining within the 1-standard-deviation uncertainty band. The sensitivity degrades at higher masses (due to wider opening angles) and longer lifetimes (due to track reconstruction failures).
• Interpretation: The results are interpreted within an effective ALP-Higgs coupling model, providing constraints on the coupling strength relative to the energy scale of the effective theory.

Significance
This paper establishes the first direct limits on the exotic decay $H \to AA \to 4e$ for ALPs with masses of order 10 MeV. By extending collider coverage to masses as low as 10 MeV for the first time, the search significantly improves the experimental sensitivity to axionlike particles in this low-mass region. The results provide the most stringent constraints to date on this specific model and establish a new benchmark for future searches at the LHC. The limiting factor for the sensitivity remains the number of events in the data, while the leading systematic uncertainty arises from the identification efficiency of merged electron-positron pairs.