Bounds on axion-like particles from fully merged photons at the LHC

Technical Summary: Bounds on Axion-like Particles from Fully Merged Photons at the LHC

Problem Statement
The paper addresses the search for Axion-like particles (ALPs) in the mass range $0.1 \text{ GeV} < m_a < 6 \text{ GeV}$ that couple predominantly to photons. In this specific mass regime, ALPs produced at the Large Hadron Collider (LHC) via the process $q\bar{q} \to a\gamma$ decay into two photons ($a \to \gamma\gamma$). Due to the high boost of the ALP at LHC energies, the decay photons are highly collimated. Consequently, standard detector reconstruction algorithms often merge these two photons into a single "photon" candidate. The author investigates whether the high-energy tail of the diphoton invariant mass ($m_{\gamma\gamma}$) distribution, where the signal mimics a single high-energy photon, can provide competitive constraints on the ALP-photon coupling ($g_{a\gamma\gamma}$) compared to existing limits from lower-energy experiments and other LHC searches.

Methodology
The analysis recasts the ATLAS search for high-mass diphoton resonances using the full 139 fb$^{-1}$ of 13 TeV proton-proton collision data. The theoretical framework assumes an effective field theory where the ALP interacts with the Standard Model via dimension-five operators involving the hypercharge ($B$) and $SU(2)$($W$) field strength tensors. The author assumes a scenario where the hypercharge coupling dominates ($c_B \gg c_W$), leading to a specific relation between the ALP-photon and ALP-photon-Z couplings.

Key methodological steps include:

1. Signal Generation: Signal events ($pp \to a\gamma$) were generated at leading order using MadGraph, applying a $k$-factor of 1.3 derived from NLO calculations for the similar $pp \to \gamma\gamma$ process.
2. Merged Photon Assumption: The ALP is treated as a single photon in the detector simulation. Crucially, the analysis imposes a requirement that the ALP decays before the inner radius of the electromagnetic calorimeter (ECAL) and that the resulting photon pair is sufficiently collimated to pass "tight" photon identification criteria.
3. Background Modeling: The dominant background is the Standard Model $q\bar{q} \to \gamma\gamma$ process. The author utilizes the background model provided by ATLAS, which is derived from a fit to Monte Carlo simulations and uncorrelated control samples, ensuring it is not contaminated by the specific signal considered.
4. Statistical Analysis: A profile-likelihood ratio test was performed on all bins with $m_{\gamma\gamma} > 1.5$ TeV using the pyhf package. The analysis assumed negligible uncertainties on the background prediction and utilized asymptotic formulas validated by toy simulations.

Key Contributions and Results
The primary contribution is the derivation of the strongest current bounds on ALP-photon couplings for the mass range $0.1 \text{ GeV} \lesssim m_a \lesssim 6 \text{ GeV}$.

• Exclusion Limits: The analysis sets a 95% confidence level upper limit on the coupling $g_{a\gamma\gamma}$. For $m_a \approx 1$ GeV, the bound reaches approximately $g_{a\gamma\gamma} \approx 4 \times 10^{-4} \text{ GeV}^{-1}$.
• Mass Dependence: The limits weaken for $m_a \lesssim 0.2$ GeV because the ALP decay length becomes too large (displaced), causing the decay to occur outside the ECAL. The limits also weaken for $m_a \gtrsim 1$ GeV due to the conservative application of "tight" photon selection criteria, which reduces signal efficiency.
• Comparison with Other Experiments: The derived ATLAS bounds surpass existing limits from LEP I, FASER, beam dump experiments, and heavy-ion runs (ATLAS and CMS Pb-Pb) across the specified mass window. The author notes that while Belle II currently has weaker bounds due to limited integrated luminosity (445 pb$^{-1}$ used in the comparison), its future dataset (aiming for 50 ab$^{-1}$) is expected to surpass the High-Luminosity LHC (HL-LHC) projections.
• Alternative Scenarios: The paper briefly explores the $c_W \approx c_B$ scenario. While this suppresses the $a\gamma Z$ coupling, the author notes that constraints from $B \to Ka$ decays (measured by BaBar and Belle) would impose a much stronger bound ($g_{a\gamma\gamma} \lesssim 5 \times 10^{-6} \text{ GeV}^{-1}$) in this specific case, rendering the LHC search less competitive.

Significance and Claims
The paper claims that the high-energy tail of the diphoton spectrum at the LHC, often analyzed for heavy resonances, provides a powerful and previously underutilized probe for light ALPs that decay into collimated photons. The author emphasizes that this approach yields the most stringent constraints in the 0.1–6 GeV mass range under the assumption of a UV completion involving heavy charged particles.

The author modestly suggests that the bounds could be improved by relaxing the photon selection criteria to include "loose" photons, potentially reducing the signal loss by ~10% without significantly altering the background estimate. Furthermore, the paper identifies the potential for future improvements at the HL-LHC (3 ab$^{-1}$) and highlights that the current analysis assumes a specific UV completion (heavy charged particles); if the UV completion involves lighter particles that evade detection, the constraints would differ. The work serves as a proof-of-concept that "merged photon" signatures in high-mass resonance searches are a viable channel for constraining light new physics.