Observation of the rare decay eta to mu+ mu- e+ e-
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: "Observation of the rare decay eta to mu+ mu- e+ e-".
Mira: Here's the gist of the paper. A first observation of the rare decay η → µ+µ-e+e- is reported by the CMS Collaboration at the CERN LHC <ref:2605.00615#pg2>.
Kai: The branching fraction is measured to be B(η → µ+µ-e+e-) = (2.4 ± 0.8) × 10−6, with the uncertainty including statistical and systematic sources as well as the B(η → µ+µ-γ) uncertainty <ref:2605.00615#pg2>.
Mira: This result is close to two orders of magnitude smaller than the existing limit, and is consistent with recent theoretical predictions <ref:2605.00615#pg2>.
Kai: Experimental Setup and Measurement Strategy The measurement of the rate for rare decays of these particles is motivated by several considerations <ref:2605.00615#pg4>.
Paper summary: Kai: To sum up this paper, it reports a new observation of the eta to mu plus mu minus e plus e minus decay with a branching fraction of about two point four times ten to the power negative six <ref:2605.00615#pg1>.
Mira: It shows that this new measurement is consistent with theoretical predictions for these rare decays, which are predicted to be around two point three to two point four times ten to the power negative six <ref:2605.00615#pg10>.
Lev: The authors acknowledge that their method relies on data-driven approaches like vector meson dominance or the light front quark model for calculating the necessary transition form factors <ref:two thousand six hundred five point zero zero six one five#pg3.
Kai: It means that we have a new data point in the picture for how these specific decays happen, which helps refine those theoretical calculations about light by light hadronic contributions to the muon anomalous magnetic moment <ref:2605.00615#pg3>.
Mira: For someone who only listens to this show, it means that we've measured something new that aligns with what physicists expect from their models, which is good for testing those models about lepton flavor universality <ref:two thousand six hundred five point zero zero six one five#pg3.
Lev: The limitation they state is that the calculation of the widths depends on theoretically unknown transition form factors which model those low energy quantum chromodynamic interactions <ref:two thousand six hundred five point zero zero six one five#pg3.
Kai: So, for now, it's a first observation that fits the standard model predictions, but it still leaves room for those theoretical uncertainties about the form factors <ref:two thousand six hundred five point zero zero six one five#pg3.
Mira: That means we need more measurements to constrain those unknown parts of the theory and really pin down what's going on with these decays <ref:two thousand six hundred five point zero zero six one five#pg3.
Conclusion: Kai: So, we've seen this new observation of that rare decay eta to mu plus mu minus e plus e minus, and now we're looking at what that actually means for the broader picture.
Mira: It’s just a first look at this specific channel, but the authors are suggesting it fits right into what the standard model predicts for these kinds of decays.
Lev: From an error correction standpoint, fitting two point four times ten to the negative six branching fraction in is tight; we need very high fidelity measurements to confirm that signal versus background separation.
Kai: Exactly, and the paper points out that while this result is close to theory, there's still a bit of room left for those theoretical uncertainties we talked about earlier, especially with the form factors.
Mira: Right. The main thing here is that getting this data point helps us test if lepton flavor universality holds true across different particle decays.
Lev: It’s a data-driven approach, so they’re using the eta to mu plus mu minus gamma mode as a reference to calculate this new ratio, which is a clever way to reduce some of the messy experimental uncertainties.
Kai: So basically, they've found something new that lines up with the math we've been working on for a while.
Mira: It opens the door for us to constrain those theoretical calculations about how these particles interact at low energies, which is where a lot of our modeling comes into play.
CERN
hep-ex
Submitted: 2026-05-01
Updated: 2026-11-08
Comments: Replaced with the published version. Added the journal reference and the DOI. All the figures and tables can be found at http://cms-results.web.cern.ch/cms-results/public-results/publications/BPH-24-001 (CMS Public Pages)
Journal ref: Phys. Rev. Lett. 137 (2026) 191902
DOI: 10.1103/dv54-jq8p
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 76/100
The gist: A first observation of the rare decay η → µ+µ-e+e- is reported by the CMS Collaboration at the CERN LHC <ref:2605.00615#pg2>.
Key concepts
- Branching Fraction (BF)
- The BF measures the probability that a specific particle decay will occur relative to all possible decays of that particle. In this case, it quantifies how often an eta meson decays into the specific final state of two muons and two electrons.
- Chiral Anomalies
- These are quantum mechanical effects related to how particles interact with fundamental forces like electromagnetism. The paper suggests that single- and double-Dalitz decays of the eta meson to leptonic states happen through these anomalies involving virtual photons.
- Signal Modeling and Background Subtraction
- Researchers used simplified Monte Carlo simulations to distinguish the rare signal from unwanted background events. They modeled how different decay modes look in mass spectra, identifying that the main background mimicking the signal was actually the eta to mu+ mu-gamma decay.
Terminology
Summary
A first observation of the rare decay η → µ+µ-e+e- is reported by the CMS Collaboration at the CERN LHC <ref:2605. p. 0 EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN)
CERN-EP-2026-023
2026/05/04
CMS-BPH-24-001
Observation of the rare decay η → µ
+µ
−e
+e
−
The CMS Collaboration*
Abstract
A first observation of the rare decay η → µ
+µ
−e
+e
− is reported by the CMS Collaboration at the CERN LHC. The result is based on a proton-proton collision data sample
at √
s = 13. 0 fb−1
, acquired
in 2022 using a high-rate dimuon trigger. Using the η → µ
+µ
−γ decay channel
for normalization, the branching fraction is measured to be B(η → µ
+µ
−e
+e
−) =
(2. Using the η → µ
+µ
−γ decay channel
for normalization, the branching fraction is measured to be B(η → µ
+µ
−e
+e
−) =
(2. 8) × 10−6
, with the uncertainty including statistical and systematic sources
as well as the B(η → µ
+µ
−γ) uncertainty. 8) × 10−6
, with the uncertainty including statistical and systematic sources
as well as the B(η → µ
+µ
−γ) uncertainty. This result is close to two orders of
magnitude smaller than the existing limit, and is consistent with recent theoretical
predictions.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Experimental Setup and Measurement Strategy
The measurement of the rate for rare decays of these particles is motivated by several considerations <ref:2605.00615#pg4>. Single- and double-Dalitz decays of the η and η′ mesons to purely leptonic final states proceed via chiral anomalies with one or two virtual photons, respectively <ref:2605. p. 2 The measurement of the rate for rare decays of
these particles is motivated by several considerations: they assist in theoretical computations of
the light-by-light hadronic contributions to the muon anomalous magnetic moment [2], enable
novel tests of lepton flavor universality in particle physics [3], and are sensitive to beyondstandard-model theories that predict deviations in these rates [4, 5]. Single- and double-Dalitz decays of the η and η
′ mesons to purely leptonic final states proceed
via chiral anomalies with one or two virtual photons, respectively, and their widths depend
on theoretically unknown transition form factors that model the low-energy quantum chromodynamic interactions. 6 × 10−4 at the 90% confidence level [19]. The data used for this observation are from √
s = 13. The data used for this observation are from √
s = 13. 6 TeV proton-proton collisions in the CERN
LHC, collected in 2022 by the CMS experiment, corresponding to an integrated luminosity
of 38. Submitted to Physical Review Letters
© 2026 CERN for the benefit of the CMS Collaboration. 0 license
*See Appendix A for the list of collaboration members
arXiv:2605. Tabulated results are provided in the HEPData record for this analysis [20]. To
measure the BF for the η → µ
+µ
−e
+e
− signal, the η → µ
+µ
−γ mode, which has a measured
BF of (3. Removing this cut from both channels for the data and from the corresponding simulated events used to calculate the efficiency
results in an 8% change in the branching ratio, which is included as a systematic uncertainty. The uncertainty in the kinematic distribution in the reference channel is found by reweighting
the simulated events such that the form factor in the theoretical model is 1 standard deviation above or below the central value reported in Ref. The photon in the final state subsequently
converts to an electron-positron pair in the inner region of the CMS detector, meaning the signal
and reference channels have the same final-state particles. The BF can be calculated using
B2µ2e
B2µγ
=
N2µ2e
N2µγ
ϵ2µγ
ϵ2µ2e
, (1)
where BX is the branching fraction of η → X; N2µ2e is the total number of signal events observed; N2µγ is the number of reference channel events observed; and ϵ2µ2e and ϵ2µγ are theWhere the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Signal Modeling and Background Subtraction
To determine if the data in the vicinity of the η meson invariant mass peak represent solely η → µ+µ-e+e- events or if it is contaminated with background, a simplified MC simulation is used to model the invariant mass spectra for various η meson decays <ref:2605.00615#pg6>. The signal channel is shown in red squares, and is closely mimicked by the η → µ+µ-γ channel, shown in blue stars <ref:2605. p. 4 0 fb
CMS
Simulation
Figure 2: Summary of expected background contributions to the yield measured in the signal
channel based on a simplified MC simulation. The signal channel is shown in red squares, and
is closely mimicked by the η → µ
+µ
−γ channel, shown in blue stars. With one notable exception, potential backgrounds do not produce a significant number of
events in the η meson mass peak. This is because assigning the muon mass to a pion track
significantly shifts the reconstructed mass peak. The muons in such conversions receive a significant momentum kick from the recoiling
nucleus, which shifts and flattens the apparent mass peak, as seen clearly in the η → µ
+µ
−γ
distribution of Fig. As a result of these effects, the only background with a significant probability of mimicking the η → µ
+µ
−e
+e
− signal is the process η → µ
+µ
−γ, with the photon
converting to an electron pair in the innermost layer of the tracker.Where the paper says it
Where the paper says it
Where the paper says it
Efficiency and Systematic Uncertainties
The ratio of signal to reference channel efficiencies is found to be ϵ2µ2e/ϵ2µγ = 51.0 ± 11.8, where the uncertainty includes statistical and systematic components as detailed below <ref:2605. p. 6 Including all corrections, the ratio of signal to reference channel efficiencies is found to be
ϵ2µ2e/ϵ2µγ = 51. 8, where the uncertainty includes statistical and systematic components
as detailed below. This fit result is
N2µγ = 329 ± 22. Several sources of systematic uncertainty affect the result, while many others cancel when taking the ratio to the reference channel. For the signal channel, this reweighting is done only to estimate the
uncertainty and not as a correction on the result (which would have resulted in the measured
BF being 11% lower). Combined with the 14% uncertainty in the dielectron identification efficiency SF and 8% from the limited number of simulated events, these uncertainties add up to a
total systematic uncertainty of 27%.Where the paper says it
Where the paper says it
Where the paper says it
Conclusion and Comparison to Theory
The measured BF ratio is B2µ2e/B2µγ = [7.3 ± 0.9 (stat) ± 2. p. 8 (1), the measured BF ratio is
B2µ2e
B2µγ
= [7. 9 (stat) ± 2. 9 (stat) ± 2. 0 (syst)] × 10−3 = [7. Submitted to Physical Review Letters
© 2026 CERN for the benefit of the CMS Collaboration. 0 license
*See Appendix A for the list of collaboration members
arXiv:2605. 2] × 10−3
. Using the world average value of B2µγ = (3. The photon in the final state subsequently
converts to an electron-positron pair in the inner region of the CMS detector, meaning the signal
and reference channels have the same final-state particles. The BF can be calculated using
B2µ2e
B2µγ
=
N2µ2e
N2µγ
ϵ2µγ
ϵ2µ2e
, (1)
where BX is the branching fraction of η → X; N2µ2e is the total number of signal events observed; N2µγ is the number of reference channel events observed; and ϵ2µ2e and ϵ2µγ are the (1), the measured BF ratio is
B2µ2e
B2µγ
= [7. 9 (stat) ± 2. 9 (stat) ± 2. 0 (syst)] × 10−3 = [7. The photon in the final state subsequently
converts to an electron-positron pair in the inner region of the CMS detector, meaning the signal
and reference channels have the same final-state particles. The BF can be calculated using
B2µ2e
B2µγ
=
N2µ2e
N2µγ
ϵ2µγ
ϵ2µ2e
, (1)
where BX is the branching fraction of η → X; N2µ2e is the total number of signal events observed; N2µγ is the number of reference channel events observed; and ϵ2µ2e and ϵ2µγ are the Submitted to Physical Review Letters
© 2026 CERN for the benefit of the CMS Collaboration. 0 license
*See Appendix A for the list of collaboration members
arXiv:2605. 6 × 10−4 at 90% confidence level, set by the WASA
experiment at the CELSIUS storage ring [19]. Our measurement is consistent with the latest
standard model predictions of 2. Our measurement is consistent with the latest
standard model predictions of 2. 4 × 10−6
, based on data-driven approaches [11, 12], while
other models yield predictions as low as 0.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Acknowledgments and Data Availability
We congratulate our colleagues in the CERN accelerator departments for the excellent performance of the LHC and thank the technical and administrative staffs at CERN and at other CMS institutes for their contributions to the success of the CMS effort <ref:2605.00615#pg11>. The release and preservation of data used by the CMS Collaboration as the basis for publications is guided by the CMS data preservation, re-use and open access policy <ref:2605. p. 9 Finally, we acknowledge the enduring support for the construction and operation of the
LHC, the CMS detector, and the supporting computing infrastructure provided by the following funding agencies: SC (Armenia), BMBWF and FWF (Austria); FNRS and FWO (Belgium);
CNPq, CAPES, FAPERJ, FAPERGS, and FAPESP (Brazil); MES and BNSF (Bulgaria); CERN;
CAS, MoST, and NSFC (China); MINCIENCIAS (Colombia); MSES and CSF (Croatia); RIF
(Cyprus); SENESCYT (Ecuador); ERC PRG and PSG, TARISTU24-TK10 and MoER TK202 (Estonia); Academy of Finland, MEC, and HIP (Finland); CEA and CNRS/IN2P3 (France); SRNSF
(Georgia); BMFTR, DFG, and HGF (Germany); GSRI (Greece); MATE and NKFIH (Hungary);
DAE and DST (India); IPM (Iran); SFI (Ireland); INFN (Italy); MSIT and NRF (Republic of Korea); MES (Latvia); LMTLT (Lithuania); MOE and UM (Malaysia); BUAP, CINVESTAV, CONACYT, LNS, SEP, and UASLP-FAI (Mexico); MOS (Montenegro); MBIE (New Zealand); PAEC
(Pakistan); MSHE, NSC, and NAWA (Poland); FCT (Portugal); MESTD (Serbia); MICIU/AEI
and PCTI (Spain); MOSTR (Sri Lanka); Swiss Funding Agencies (Switzerland); MST (Taipei);
MHESI (Thailand); TUBITAK and TENMAK (Turkiye); NASU (Ukraine); STFC (United King- ¨
dom); DOE and NSF (USA). Data availability
Release and preservation of data used by the CMS Collaboration as the basis for publications
is guided by the CMS data preservation, re-use and open access policy.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
-
Precision measurement of rare decay branching fractions can be improved by utilizing data-driven approaches as described in Ref. [13, 14], which are
relied upon for muon anomalous magnetic moment calculations because of their assumed accuracy.
This suggests that AI systems analyzing particle physics data should incorporate these empirically motivated models to better predict and interpret low-rate decay modes. -
Event selection criteria can be optimized by employing sophisticated machine learning discriminators, such as the
boosted decision tree from Ref. [26] that does not include isolation variables,
for electron identification in signal events. This allows AI systems to achieve higher discrimination between signal and background compared to simpler models by training on shower shape variables, energies, and track hit patterns. -
Background modeling can be enhanced by using empirical fitting functions like the
double-Gaussian model
for the signal channel andBreit–Wigner model
for the resonant background in invariant mass spectra. AI systems can be trained to automatically select or tune these models based on goodness-of-fit metrics, as demonstrated by the choice of models that providefits with the best chi square values.
-
Systematic uncertainty estimation can be refined by using advanced statistical techniques, such as redoing fits with
two alternative models that give similar goodness-of-fit metrics.
This approach helps in rigorously quantifying uncertainties arising from model dependence, which is crucial for reliable branching fraction measurements. -
Efficiency corrections can be improved by using data and simulation comparisons to derive factors like the
dielectron SF is found to be 0.60 ± 0.08.
AI systems can be used to automatically derive these scale factors by comparing yields in data and simulation before and after applying specific selection criteria, such as theadditional electron identification requirements.
Abstract
A first observation of the rare decay η to μ+μ- e+ e- is reported by the CMS Collaboration at the CERN LHC. The result is based on a proton-proton collision data sample at sqrt s = 13.6 TeV corresponding to an integrated luminosity of 38.0 fb-1, acquired in 2022 using a high-rate dimuon trigger. Using the η to μ+μ-γ decay channel for normalization, the branching fraction is measured to be B (η to μ+μ- e+ e-) = (2.4 plus or minus 0.8) times 10-6, with the uncertainty including statistical and systematic sources as well as the B (η to μ+μ-γ) uncertainty. This result is about a factor of three smaller than the existing limit, and is consistent with recent theoretical predictions.
Sources
- Scalar Meson Contributions to a_\mu from Hadronic Light-by-Light Scattering
- Observation of the rare decay of the $\eta$ meson to four muons
- Precision tests of fundamental physics with $\eta$ and $\eta^\prime$ mesons
- New Physics Searches at Kaon and Hyperon Factories
- Eta decays at and beyond $\mathbf{p^4}$ in Chiral Perturbation Theory
- Study of $\pi^{0}$ and $\eta$ decays containing dilepton
- A data-driven approach to $\pi^{0}, \eta$ and $\eta^{\prime}$ single and double Dalitz decays
- The radiative corrections to double-Dalitz decays revisited
- The anomalous magnetic moment of the muon in the Standard Model
- The anomalous magnetic moment of the muon in the Standard Model: an update
- Measurement of the branching fraction of $\eta \to \mu^+ \mu^-$ and search for $\eta \to e^+ e^-$
- Observation of the rare eta->e+e-e+e- decay with the KLOE experiment
- Observation of the double Dalitz decay $\eta'\to e^+e^-e^+e^-$
- Measurement of eta meson decays into lepton-antilepton pairs
- Development of the CMS detector for the CERN LHC Run 3
- Performance of the CMS Level-1 trigger in proton-proton collisions at $\sqrt{s} =$ 13 TeV
- The CMS trigger system
- Performance of the CMS high-level trigger during LHC Run 2
- Electron and photon reconstruction and identification with the CMS experiment at the CERN LHC
- Performance of the CMS muon detector and muon reconstruction with proton-proton collisions at $\sqrt{s}=$ 13 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