⚛️ high-energy experiments

Measurement of the Higgs boson decay to a low-mass dilepton system and a photon in $pp$ collisions at $\sqrt{s} =$ 13 and 13.6 TeV with the ATLAS detector

Using ATLAS detector data from 13 and 13.6 TeV proton-proton collisions, this study measures the Higgs boson decay to a photon and a low-mass dilepton pair, combining the results to achieve a 3.4 standard deviation observation of the signal with a strength consistent with Standard Model predictions.

Original authors: ATLAS Collaboration

Published 2026-08-05

📖 3 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, cosmic kitchen where the most famous chef, the Higgs boson, is constantly cooking up new particles. For years, scientists have been trying to figure out exactly how this chef works. They know the Higgs exists—it was discovered in 2012—but they want to see it doing its most delicate tricks. One of these tricks is a very rare recipe: the Higgs decaying (or breaking apart) into a photon (a particle of light) and a pair of lighter particles called leptons (like electrons or muons). It's like watching a magician pull a rabbit out of a hat, but the rabbit is actually made of light and tiny, ghostly particles, and the trick happens so rarely that you have to watch millions of shows to see it once. Why does this matter? Because if the Higgs does this trick slightly differently than our best recipes (the Standard Model) predict, it might mean there are secret ingredients or new physics hiding in the kitchen that we haven't discovered yet.

This paper is the latest report from the ATLAS team, a massive group of scientists working with a giant particle detector called ATLAS at the Large Hadron Collider (LHC) in Europe. They decided to throw a massive party, smashing protons together at record-breaking speeds to see if they could catch the Higgs boson performing this specific "photon plus two leptons" dance. They looked at data collected between 2022 and 2024, which corresponds to a huge amount of collision data (164 fb⁻¹). To make the search easier, they split the events into nine different categories, like sorting guests at a party by what they are wearing or how they are dancing. They also developed a new, super-smart computer algorithm (a "boosted decision tree") to help spot a tricky type of event where two electrons crash into each other so hard they look like a single blob of energy.

The results are a bit like finding a needle in a haystack, but the haystack is the size of a mountain. When they looked at just the new data from 2022–2024, they found a hint of the signal, but it wasn't quite loud enough to be certain. The "signal strength" they measured was about 0.64, which is lower than the perfect score of 1.00 predicted by the Standard Model, but the uncertainty is large enough that it could still be a fluke. The statistical significance was 1.7 standard deviations, which is like hearing a whisper in a noisy room—you think you heard something, but you aren't sure.

However, the real magic happened when they combined this new data with a previous search they did between 2015 and 2018. By merging the two datasets, they got a much clearer picture. The combined result showed a signal strength of 1.03, which is almost exactly what the Standard Model predicted! The statistical significance of this combined result is 3.4 standard deviations. In the world of particle physics, this is considered "evidence." It's not quite a "discovery" (which usually requires 5 standard deviations, or a shout that can be heard across the room), but it is a very strong hint that the Higgs boson is indeed performing this rare dance just as the theory says it should. The team didn't find any new, weird physics breaking the rules, but they did successfully confirm that the Higgs can decay into a photon and a low-mass pair of electrons or muons, adding another successful chapter to our understanding of how the universe works.

Technical Summary: Measurement of the Higgs boson decay to a low-mass dilepton system and a photon in $pp$ collisions at $\sqrt{s}=13$ and 13.6 TeV with the ATLAS detector

Problem and Motivation
The Standard Model (SM) predicts that the Higgs boson can decay into a photon and a pair of leptons ( $H \to \ell\ell\gamma$ , where $\ell = e, \mu$ ) primarily through loop-induced processes. This decay channel, specifically in the low invariant mass region of the dilepton system ( $m_{\ell\ell} < 30$ GeV), is dominated by the $H \to \gamma^*\gamma \to \ell\ell\gamma$ process. While the branching ratio is approximately 5% of the $H \to \gamma\gamma$ rate, the experimental observation is challenging due to the low mass of the dilepton pair and the high transverse momentum of the system, which can cause electromagnetic showers from electrons to merge in the calorimeter.

This analysis aims to measure the signal strength of the $H \to \ell\ell\gamma$ process in the region $m_{\ell\ell} < 30$ GeV. This measurement complements previous searches for $H \to Z\gamma$ (where $m_{\ell\ell} > 30$ GeV) and provides a unique probe of the loop-induced Higgs-photon interaction, independent of the Higgs production cross-section. Furthermore, the non-zero virtuality of the intermediate photon ( $\gamma^*$ ) offers sensitivity to potential new physics effects and allows for the study of the CP properties of the Higgs boson via forward-backward asymmetry.

Methodology
The analysis utilizes proton-proton collision data recorded by the ATLAS detector at the Large Hadron Collider (LHC).

Data Samples: The primary dataset corresponds to an integrated luminosity of 164 fb $^{-1}$ collected at $\sqrt{s} = 13.6$ TeV during 2022–2024 (Run 3). These results are combined with a previous analysis using 139 fb $^{-1}$ at $\sqrt{s} = 13$ TeV (Run 2, 2015–2018).
Event Selection: Events are selected requiring a photon and a pair of same-flavor, opposite-charge leptons ( $e^+e^-$ or $\mu^+\mu^-$ ). The dilepton invariant mass is constrained to $m_{\ell\ell} < 30$ GeV, and the three-body invariant mass $m_{\ell\ell\gamma}$ is required to be within the Higgs mass window (110–160 GeV).
Object Reconstruction:
- Merged Dielectrons: A significant challenge in the $ee\gamma$ channel is the overlap of electron showers. A dedicated Boosted Decision Tree (BDT) algorithm, trained using LightGBM, was developed to identify "merged dielectron" candidates where two tracks are matched to a single calorimeter cluster. This algorithm improves signal efficiency by approximately 5% relative to previous cut-based methods while maintaining background rejection.
- Triggers: A combination of single-lepton, dilepton, single-photon, diphoton, and lepton+photon triggers was employed, with specific low-threshold triggers supplemented by higher $p_T$ triggers to maintain efficiency at high instantaneous luminosity.
Categorization: Selected events are divided into nine mutually exclusive categories based on lepton flavor ( $\mu\mu$ $μμ$ , resolved $ee$, merged $ee$) and event topology:
1. VBF: Events with two forward jets satisfying specific kinematic criteria ( $m_{jj} > 500$ GeV, $|\Delta\eta_{jj}| > 2.7$ ).
2. High $p_{Tt}$ : Events failing VBF criteria but exhibiting high transverse momentum projection onto the $\ell\ell\gamma$ thrust axis ( $p_{Tt} > 100$ GeV).
3. Low $p_{Tt}$ : All remaining events.
Signal and Background Modeling:
- Signal: Modeled using a double-sided Crystal Ball (DSCB) function derived from Monte Carlo (MC) simulations of $H \to \gamma^*\gamma \to \ell\ell\gamma$ .
- Background: The dominant background is irreducible non-resonant $\ell\ell\gamma$ production. Its shape is modeled using parametric functions (exponentials, polynomials, or power laws) selected via a spurious signal study. Resonant backgrounds from $H \to \gamma\gamma$ (where one photon converts) and fake leptons/photons are estimated using data-driven techniques and MC templates.
Statistical Analysis: A simultaneous unbinned maximum-likelihood fit is performed on the $m_{\ell\ell\gamma}$ distributions across all nine categories. Systematic uncertainties (experimental and theoretical) are incorporated as nuisance parameters.

Key Contributions

Run-3 Measurement: This paper presents the first measurement of $H \to \ell\ell\gamma$ using the full 164 fb $^{-1}$ dataset at $\sqrt{s} = 13.6$ TeV.
Merged Dielectron Identification: The development and application of a BDT-based identification algorithm for merged dielectron candidates significantly enhances the sensitivity of the $ee\gamma$ channel, improving the expected significance by 5% compared to previous cut-based approaches.
Combined Analysis: The results are combined with the Run-2 analysis, providing the most precise measurement of this decay mode to date, with improved statistical power and reduced systematic uncertainties through the combination of datasets with different center-of-mass energies and detector conditions.

Results

Run-3 Only (13.6 TeV):
- The observed signal strength parameter ( $\mu$ ), defined as the ratio of observed yield to the SM expectation, is measured to be $\mu = 0.64^{+0.44}_{-0.39}$ .
- The expected signal strength is $\mu = 1.00^{+0.47}_{-0.41}$ .
- The observed (expected) significance over the background-only hypothesis is 1.7 $\sigma$ (2.5 $\sigma$ ).
- The measured cross-section times branching ratio is $4.1^{+2.8}_{-2.5}$ fb.
Combined Run-2 and Run-3:
- The combined observed signal strength is $\mu = 1.03^{+0.35}_{-0.32}$ .
- The combined expected signal strength is $\mu = 1.00^{+0.34}_{-0.32}$ .
- The combined observed (expected) significance is 3.4 $\sigma$ (3.3 $\sigma$ ).
- The result is consistent with the Standard Model prediction ( $p$ -value of 0.41 for the Run-3 fit).

Significance
The paper reports a Run-2 result with a significance of 3.2 $\sigma$ as "evidence" for the $H \to \ell\ell\gamma$ process. The new Run-3 measurement yields an observed significance of 1.7 $\sigma$ , while the combination of Run-2 and Run-3 data results in an observed significance of 3.4 $\sigma$ . This combined result is consistent with the Standard Model prediction and represents the most precise measurement of this decay mode to date. The measurement validates the SM prediction for this rare, loop-induced decay channel and demonstrates the capability of the ATLAS detector to reconstruct complex final states involving merged electromagnetic showers. Furthermore, it offers a complementary test of the Higgs-photon interaction loop structure and establishes a baseline for future precision measurements and searches for new physics in the low-mass dilepton spectrum.

More like this