Search for GeV-scale axion-like particles in PbPb collisions
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.
Kai: We're looking at the paper: "Search for GeV-scale axion-like particles in PbPb collisions".
Mira: Here's the gist of the paper. A search is performed for axion-like particles with masses in the range 3–10 GeV using PbPb collision data recorded by the LHCb experiment This result provides the most stringent limits to date on the production of axion-like particles from exclusive photon-photon fusion processes with masses between 3 and 5 GeV Search Scope and Sensitivity The search targets ALP masses in the range 3–10 GeV using PbPb collision data recorded by the LHCb experiment at a nucleon-nucleon centre-of-mass energy of √sNN = 5.02 TeV This is the first study of this process in the unique forward acceptance of LHCb The lower boundary is determined by the trigger requirements applied during data taking In this region, the analysis has an expected sensitivity that can improve limits from exclusive photon-photon fusion processes relative to those from previous experiments, particularly in the 3–5 GeV range In the search for resonances decaying into two photons in UPC events, three primary sources of background must be considered: light-by-light scattering, γγ → γγ (LbL); gluon-gluon fusion, gg → γγ (ggF); and the Breit-Wheeler process, γγ → e +e− (BWh) In the 3–10 GeV mass interval, J/ψ→ e+e− production in UPC is also accounted for as a source of contamination Both the BWh and J/ψ→ e+e− channels contribute to the background through electron to photon misidentification Event Selection Criteria Candidates are selected by online and offline selections, requiring no more than 50 SPD deposits and at least one photon candidate with transverse energy ET > 960 MeV for the hardware trigger The offline selection requires two reconstructed photons each with ET > 1300 MeV in a low track-multiplicity event, having no reconstructed charged particle tracks in the VELO and fewer than ten clusters in the VELO To suppress background from inelastic PbPb collisions, a further multiplicity veto requirement that combines the signals in all HeRSCheL stations is imposed Requirements Aϕ < 0.02 and p2T(γγ) < 0.03 GeV2 are imposed to maintain good signal to background ratio for masses in the range 3–5 GeV Limit Determination The resulting upper limits with 95% CLs on the ALP production cross-section vary between 1.1 pb and 3.1 pb for masses between 3 and 10 GeV The corresponding limits on the coupling gaγ vary between 3.2 × 10−4 GeV−1 and 9.5 × 10−4 GeV−1 over the same mass interval These exclusion regions are compared to other experiments in Figs.
Kai: 3 and 4 using the AxionLimits package Systematic Uncertainties Systematic uncertainties on the upper limit calculation arise from the efficiency determination, cross-section determination for the background processes, and knowledge of the integrated luminosity The relative uncertainty on the integrated luminosity is 4.2% The uncertainty associated with the HeRSCheL efficiency is 1.4% A common photon reconstruction systematic uncertainty of 4.5%, determined from B+→ J/ψK∗+ and B+→ J/ψK+ decays, is assigned to all samples The uncertainty on the efficiency associated with the multiplicity requirements on the number of VELO tracks, VELO clusters, additional reconstructed photons, and SPD hits is dominated by the limited size of the simulated samples The resulting relative systematic uncertainty on the J/ψ→ e+e− cross-section is 4% The total uncertainty is ±32% + 8.3% − 9.2% This analysis uses data corresponding to an integrated luminosity of 0.228 nb−1 collected in 2018 at a nucleon-nucleon centre-of-mass energy of √sNN = 5.02 TeV It represents the first search based on the production of two photons in the pseudorapidity interval 2.0 < η(γ) < 4.
Paper summary: Kai: So we're looking at this paper called "Search for GeV-scale axion-like particles in PbPb collisions". It's about searching for these particles, which we call axion-like particles, with masses between three and ten gigaelectron volts.
Mira: Yeah, the paper is focused on using data from lead lead collisions recorded at the LHCb experiment. The main claim is that they've set some of the most stringent limits yet on how often these axion-like particles can be produced through exclusive photon-photon fusion processes with masses between three and five gigaelectron volts.
Lev: That sounds like a specific kind of search, focusing on a certain mass range. What does that mean for the actual experimental setup?
Kai: Well, they are using data from PbPb collisions recorded at a nucleon-nucleon center of mass energy of five point zero two tera electron volts. This is the first study they're doing for this specific process in LHCb's unique forward acceptance.
Mira: And they're looking at the production exclusively through two photons going into an axion-like particle, using that gamma gamma to axion notation, which is written as gamma gamma arrow a arrow gamma gamma.
Lev: That means they are looking for a specific final state where two photons are produced, and then those photons decay into the particle they are searching for.
Kai: Exactly. The abstract says this search extends sensitivity to previously unexplored regions of the parameter space, probing axion-like particle masses and couplings down to three point zero gigaelectron volts and three point two times ten to the minus four gigaelectron volt inverse, respectively.
Mira: And they state that this result provides the most stringent limits to date on axion-like particle production from exclusive photon-photon fusion processes with masses between three and five gigaelectron volts.
Lev: That's a comparison against other experiments, right? How does this place itself in the broader context of ALP searches?
Kai: They compare their exclusion regions to other experiments using the AxionLimits package. The results show upper limits on the axion like particle production cross section between one point one picobarn and three point one picobarn for masses between three and ten gigaelectron volts.
Mira: And the corresponding limits on the coupling ga gamma vary from three point two times ten to the minus four gigaelectron volt inverse to nine point five times ten to the minus four gigaelectron volt inverse over that same mass interval.
Lev: Those are some very specific numbers, so you can see how tight these constraints are. What's a real concern here, though? The paper mentions background processes they have to consider before they can claim anything.
Paper summary: Kai: Right. They have to account for a few things in the search for resonances that decay into two photons in those ultraperipheral collisions, which include light by light scattering, or LbL.
Mira: And gluon gluon fusion, gg arrow gamma gamma, which is ggF. But they also have to deal with the Breit Wheeler process and the J psi production channel as sources of contamination.
Lev: That sounds like a lot of background noise to filter out in a collision environment where you're looking for something very rare. What kind of cuts are they using?
Kai: They use a few selection criteria to filter candidates. First, they require no more than fifty signal deposits and at least one photon candidate with transverse energy greater than nine hundred sixty megaelectron volts for the hardware trigger.
Mira: Then offline selection gets stricter: two reconstructed photons each with transverse energy greater than one thousand three hundred megaelectron volts in a low track multiplicity event.
Lev: And they also require no reconstructed charged particle tracks in the VELO and fewer than ten clusters in the VELO. That's trying to keep the event relatively clean before they even start looking at the physics signal.
Kai: Plus, they impose a further multiplicity veto requirement combining signals from all HeRSCheL stations to suppress background from inelastic lead lead collisions.
Mira: And for masses between three and five gigaelectron volts, they put in requirements of alpha less than zero point zero two and p squared times t gamma gamma less than zero point zero three giga electron volt squared to maintain a good signal to background ratio.
Lev: It sounds like they are trying to isolate a very specific kinematic region where the signal is expected to be strongest. What’s the overall uncertainty they report? Because that's where you find out how much faith we can put in these limits.
Kai: The systematic uncertainties on the upper limit calculation come from efficiency determination, cross section determination for those background processes, and knowledge of the integrated luminosity.
Mira: They say the relative uncertainty on the integrated luminosity is four point two percent, and they assign a common photon reconstruction systematic uncertainty of four point five percent to all samples.
Lev: And the uncertainty associated with those multiplicity requirements—the number of VELO tracks, VELO clusters, additional reconstructed photons, and SPD hits—is dominated by the limited size of the simulated samples.
Paper summary: Kai: The resulting relative systematic uncertainty on the J psi production cross section is four percent. So they total a significant uncertainty, about thirty two percent plus eight point three minus nine point two percent overall.
Mira: So they're telling us that while these limits are strong, we still have a substantial margin of error in how tight those constraints actually are.
Lev: That makes sense from an engineering standpoint. If you were to run this search on real hardware, you'd need to figure out how much overhead that multiplicity veto adds to your event reconstruction pipeline.
Kai: Exactly. And the paper itself is based on an integrated luminosity of zero point two two eight nanobarn minus one collected in two thousand eighteen at a nucleon-nucleon center of mass energy of five point zero two tera electron volts. This is the first search based on the production of two photons in the pseudorapidity interval from two point zero to four.
Mira: It’s important to remember this whole paper, "Search for GeV-scale axion-like particles in PbPb collisions," is about setting limits on these particles using data collected at LHCb.
Lev: So what does this mean for the world if we take these results seriously? What is the bigger picture here?
Kai: It means we are getting better constraints on theories that predict new pseudo scalar particles, which share the quantum numbers of axions. These could be from dark sector portals, QCD axion models, or even string theory frameworks.
Mira: The implication is that if these particles exist in those specific mass and coupling ranges, this search helps narrow down where those theoretical models are most likely to be correct.
Lev: For us building error correction hardware, this tells us the scale of the signal we're looking for in real experiments. It gives a target for what kind of rare event signature we need to see on our machines.
Kai: The paper sets limits down to three point zero gigaelectron volts and three point two times ten to the minus four gigaelectron volt inverse coupling. These are really probing some specific, low-mass regimes that haven't been looked at this way before.
Mira: So it’s less about finding the particle and more about ruling out certain parameter spaces for these new physics candidates based on what we don't see in the data yet.
Lev: That’s a very common outcome in high energy physics searches, I guess. Ruling things out is often just as important as finding them.
Kai: Absolutely. The constraints from this specific search are what it provides right now for this particular process at these energies and experiments. It's a concrete piece of the puzzle for the broader search program.
Conclusion: Kai: So this paper, "Search for GeV-scale axion-like particles in PbPb collisions," is basically showing us what they found when they looked for these new particles in lead lead crashes at the LHCb experiment.
Mira: Right, it’s about setting limits on those axion-like particles with masses between three and ten gigaelectron volts. The authors are trying to constrain where these theoretical models for new physics might be correct.
Lev: It's important to remember this is a search based on photon-photon fusion processes, which means they are looking at a very specific way these particles might be produced in those collisions.
Kai: Exactly. The main thing they’ve done is provide the most stringent limits yet for axion-like particle production from exclusive photon-photon fusion with masses between three and five gigaelectron volts.
Mira: That means if a particle exists in that mass range and has a certain coupling strength, this study tells us how unlikely it is to see it in those specific collisions.
Lev: The numbers they report are on the production cross section, which is how often you expect to see this interaction happen per collision.
Kai: They’ve put upper limits between one point one picobarn and three point one picobarn for masses between three and ten gigaelectron volts.
Mira: And that translates into limits on the coupling constant, ga gamma, which ranges from three point two times ten to the minus four gigaelectron volt inverse to nine point five times ten to the minus four gigaelectron volt inverse over that mass interval.
Lev: The caveat is that these limits come with some systematic uncertainties, like how well they know their efficiency and cross section for the background processes.
Kai: They say the total uncertainty on those limits is around thirty two percent plus eight point three minus nine point two percent. That’s a pretty big range we have to work with.
Mira: So, what this means for us is that if these particles are in that mass and coupling range, they probably aren't produced as frequently as the upper limits suggest in these collisions.
Lev: It means it helps narrow down the parameter space for those theoretical models we’ve been discussing, like dark sector portals or string theory frameworks.
Kai: Yeah, it’s less about finding a particle and more about ruling out certain possibilities based on what we don't see in the data yet.
Mira: So, this search is a step in constraining the landscape of new physics candidates that share quantum numbers with axions.
LHCb collaboration
European Organization for Nuclear Research (CERN)
hep-ex, nucl-ex
Submitted: 2026-10-29
Updated: 2026-11-03
Comments: All figures and tables, along with any supplementary material and additional information, are available at https://lhcb-glance.cern.ch/alcm/public/analysis/full-details/6090 (LHCb public pages)
Project page: https://cajohare.github.io/AxionLimits
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 75/100
The gist: A search is performed for axion-like particles with masses in the range 3–10 GeV using PbPb collision data recorded by the LHCb experiment This result provides the most stringent limits to date on…
Key concepts
- Axion-like Particles (ALPs)
- These are hypothetical subatomic particles that are lighter than axions. The search looks for them by looking for events where two photons are produced, which could be the decay signature of an ALP. They have a mass range being investigated between 3 and 10 GeV.
- Exclusive Photon-Photon Fusion
- This refers to a specific physical process where two photons interact to produce a new particle, such as an ALP. The LHCb experiment uses the unique forward acceptance of its detectors to look for these rare events in heavy ion collisions.
- Background Processes
- When searching for a signal, researchers must account for other known processes that can mimic the signal. This search considers light-by-light scattering ($\gamma\gamma \rightarrow \gamma\gamma$), gluon-gluon fusion ($\text{gg} \rightarrow \gamma\gamma$), and the Breit-Wheeler process, all of which can produce two photons.
- Systematic Uncertainties
- These are errors in the measurement that affect the final result, such as uncertainties in how well detector efficiency is known or how accurately background cross-sections are calculated. The total uncertainty on the limits is estimated to be around $\pm 32\% + 8.3\% - 9.2\%$.
Terminology
Summary
A search is performed for axion-like particles with masses in the range 3–10 GeV using PbPb collision data recorded by the LHCb experiment This result provides the most stringent limits to date on the production of axion-like particles from exclusive photon-photon fusion processes with masses between 3 and 5 GeV
Search Scope and Sensitivity
The search targets ALP masses in the range 3–10 GeV using PbPb collision data recorded by the LHCb experiment at a nucleon-nucleon centre-of-mass energy of √sNN = 5.02 TeV This is the first study of this process in the unique forward acceptance of LHCb The lower boundary is determined by the trigger requirements applied during data taking In this region, the analysis has an expected sensitivity that can improve limits from exclusive photon-photon fusion processes relative to those from previous experiments, particularly in the 3–5 GeV range
Background Processes Considered
In the search for resonances decaying into two photons in UPC events, three primary sources of background must be considered: light-by-light scattering, γγ → γγ (LbL); gluon-gluon fusion, gg → γγ (ggF); and the Breit-Wheeler process, γγ → e +e− (BWh) In the 3–10 GeV mass interval, J/ψ→ e+e− production in UPC is also accounted for as a source of contamination Both the BWh and J/ψ→ e+e− channels contribute to the background through electron to photon misidentification
Event Selection Criteria
Candidates are selected by online and offline selections, requiring no more than 50 SPD deposits and at least one photon candidate with transverse energy ET > 960 MeV for the hardware trigger The offline selection requires two reconstructed photons each with ET > 1300 MeV in a low track-multiplicity event, having no reconstructed charged particle tracks in the VELO and fewer than ten clusters in the VELO To suppress background from inelastic PbPb collisions, a further multiplicity veto requirement that combines the signals in all HeRSCheL stations is imposed Requirements Aϕ < 0.02 and p2T(γγ) < 0. p. 0 228 nb−1
and a nucleon-nucleon centreof-mass energy of √
sNN = 5. This is the first search for axion-like particles
produced exclusively by two photons (γγ → a → γγ) in the forward direction.Where the paper says it
Limit Determination
The resulting upper limits with 95% CLs on the ALP production cross-section vary between 1.1 pb and 3.1 pb for masses between 3 and 10 GeV The corresponding limits on the coupling gaγ vary between 3. p. 8 1 pb for
masses between 3 and 10 GeV. The corresponding limits on the coupling gaγ vary between
3. EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN)
CERN-EP-2026-273
LHCb-PAPER-2026-029
October 29, 2026
Search for GeV-scale axion-like
particles in PbPb collisions
LHCb collaboration†
Abstract
A search is performed for the production of axion-like particles with masses in
the range 3–10 GeV using PbPb collision data recorded by the LHCb experiment,
corresponding to an integrated luminosity of 0. 228 nb−1
and a nucleon-nucleon centreof-mass energy of √
sNN = 5. 5 × 10−4 GeV−1
over the same mass interval. The exclusion region
set by this paper is compared to other experiments in Figs.Where the paper says it
Where the paper says it
Where the paper says it
Systematic Uncertainties
Systematic uncertainties on the upper limit calculation arise from the efficiency determination, cross-section determination for the background processes, and knowledge of the integrated luminosity The relative uncertainty on the integrated luminosity is 4.2% The uncertainty associated with the HeRSCheL efficiency is 1. p. 7 The relative uncertainty on the integrated luminosity is 4. The uncertainty
associated with the HeRSCheL efficiency is 1. The uncertainty
associated with the HeRSCheL efficiency is 1. A common photon reconstruction systematic uncertainty of 4. 3% relative
to the nominal value; this difference is assigned as a systematic uncertainty. The uncertainty on the efficiency associated with the multiplicity requirements on
the number of VELO tracks, VELO clusters, and additional reconstructed photons is
dominated by the limited size of the simulated samples, reaching up to 39% (50%) for
the BWh (J/ψ→ e
+e
−) channels.Where the paper says it
Where the paper says it
Where the paper says it
Conclusion
This analysis uses data corresponding to an integrated luminosity of 0.228 nb−1 collected in 2018 at a nucleon-nucleon centre-of-mass energy of √sNN = 5. p. 8 In summary, this analysis uses data corresponding to an integrated luminosity of
0. 228 nb−1
collected in 2018 at a nucleon-nucleon centre-of-mass energy of √
sNN = 5. 228 nb−1
collected in 2018 at a nucleon-nucleon centre-of-mass energy of √
sNN = 5. It represents the first search based on the production of two photons in the pseudorapidity
interval 2.Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI can perform searches for axion-like particles in PbPb collisions using forward photon detection by applying trigger requirements such as "at least one photon candidate with transverse energy ET > 960 MeV" and defining a photon candidate based on specific calorimeter hits.
-
Improved AI can accurately model the detector response to diphoton candidates by simulating ALP signal samples using STARlight 3.13 for masses between 3 GeV and 10 GeV, while simultaneously simulating background processes like BWh and J/psi to e+e- using STARlight and SuperChic 2.04.
-
Improved AI can determine the efficiency of selection criteria by evaluating them using data-driven methods, such as
a tag-and-probe method similar to that developed in Ref. [62]
for photon-related requirements, and by usingbeam-empty events
to evaluate multiplicity requirements on VELO tracks and clusters. -
Improved AI can set upper limits on ALP mass hypotheses while avoiding potential mismodelling of the diphoton invariant-mass spectrum by evaluating limits
in mass windows centred on the mean reconstructed diphoton mass for each hypothesis
to retain 99.7 percent of the corresponding signal distribution.
Abstract
A search is performed for the production of axion-like particles with masses in the range 3-10 GeV using PbPb collision data recorded by the LHCb experiment, corresponding to an integrated luminosity of 0.228 nb-1 and a nucleon-nucleon centre-of-mass energy of sqrt s NN =5.02 TeV. This is the first search for axion-like particles produced exclusively by two photons (γγ to a toγγ) in the forward direction. It extends sensitivity to previously unexplored regions of the parameter space, probing axion-like particle masses and couplings down to 3.0 GeV and 3.2 times10-4 GeV-1, respectively. This result provides the most stringent limits to date on the production of axion-like particles from exclusive photon-photon fusion processes with masses between 3 and 5 GeV.
Sources
- Axions and Axion-Like Particles
- Aspects of Chiral Symmetry
- Dark Matter through the Axion Portal
- A taste of dark matter: Flavour constraints on pseudoscalar mediators
- Extending LHC Coverage to Light Pseudoscalar Mediators and Coy Dark Sectors
- Dark photon relic dark matter production through the dark axion portal
- Axion-like particle assisted strongly interacting massive particle
- ALP-portal majorana dark matter
- Dark Matter Through the Axion-Gluon Portal
- A Leptonic ALP Portal to the Dark Sector
- Light scalars in composite Higgs models
- Searching for axions and ALPs from string theory
- Axions In String Theory
- String Axiverse
- The R Axion From Dynamical Supersymmetry Breaking
- R Symmetry Breaking Versus Supersymmetry Breaking
- R-axion at colliders
- Measurement of light-by-light scattering and search for axion-like particles with 2.2 nb$^{-1}$ of Pb+Pb data with the ATLAS detector
- Evidence for light-by-light scattering and searches for axion-like particles in ultraperipheral PbPb collisions at $\sqrt{s_\mathrm{NN}} =$ 5.02 TeV
- Measurement of light-by-light scattering and the Breit-Wheeler process, and search for axion-like particles in ultraperipheral PbPb collisions at $\sqrt{s_\mathrm{NN}}$ = 5.02 TeV
Related papers
- DrSai: an agentic AI system toward autonomous physics analysis workflows at BESIII
- CANTON- mu Proposal: A Next-Generation Muon g-2 Measurement at Sub-0.1 ppm Precision
- Search for dark photon dark matter near 1,meV using a superconducting mixer
- ANUBIS detector and its sensitivity to neutral long-lived particles
- How to scale your HEP ML models: A recipe for robust architecture comparisons at scale
- Search for the 16 O(ppp) to 13 C pi+ pi+ e+ Decay Mode in Super-Kamiokande Using Machine Learning Techniques