First evidence of the decay B+ to pi+ e+ e-

arXiv:2604.26784 · hep-ex · Submitted 2026-04-29 · Read on arXiv

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: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "First evidence of the decay B+ to pi+ e+ e-".

Mira: Here's the gist of the paper. The first evidence for the decay B+→ π+e+e− is reported using proton-proton collision data recorded by the LHCb experiment at centre-of-mass energies of 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 <ref:2604.26784#pg2>.

Kai: Signal Observation and Measurement A signal excess with a significance of 3.2σ is observed and the branching fraction is measured to be B(B+→ π+e+e−) = (2.4 +0.9 −0.8 +0.4 −0.2) × 10−8, where the first set of uncertainties is statistical and the second is systematic <ref:2604.26784#pg2>.

Mira: This result is consistent with the Standard Model expectation <ref:2604.26784#pg2>.

Kai: The measurement was performed using data from pp collisions at centre-of-mass energies of √s = 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 <ref:2604.26784#pg4>.

Paper summary: Kai: We’re wrapping up this discussion on First evidence of the decay B plus to pi plus electron plus electron minus. The authors are listing

Mira: The paper is titled First evidence of the decay B plus to pi plus electron plus electron minus. They are highlighting how they found a signal excess with a significance of three point two sigma, and they measured the branching fraction to be two point four plus zero point nine minus zero eight plus zero four minus zero two times ten to the negative eight one <ref:2604.26784#pg1>.

Lev: It’s important to note that they found this result is consistent with the Standard Model expectation one <ref:2604.26784#pg3>.

Kai: Yeah, and they also looked at how this compares to the B plus to pi plus mu plus mu minus decay which was measured at one point eight three plus zero two five times ten to the negative eight twenty <ref:2604.26784#pg3>.

Mira: So, for someone just listening, what this means is that while they found some evidence for a new process involving the b quark transitioning to a d quark and electron pair, it’s currently sitting right on the edge of what we expect from the Standard Model.

Lev: It means they’re setting up a measurement that could probe physics beyond the Standard Model if those systematic uncertainties turn out to be wrong.

Kai: Exactly, and they are doing this by looking at two different regions in the decay mass squared, one low q square and one high q square. This allows them to test different parts of the theory.

Mira: That’s a smart way to handle it; testing both regions helps confirm that any signal isn't just an artifact of where they looked in the detector data.

Lev: It shows they are being thorough, but their limitation is that they have to extrapolate beyond the measured mass ranges to get the total branching fraction across all possible decay modes.

Kai: So, it’s first evidence for a b to d lepton lepton transition reported with a three point two sigma significance, which is consistent with what we expect from the Standard Model fourteen <ref:2604.26784#pg3>.

Conclusion: Kai: So we’re wrapping up this discussion on "First evidence of the decay B plus to pi plus electron plus electron minus". The authors are listing

Mira: This work is about finding a signal for a specific decay of B hadrons, and they use data from the LHCb experiment to do it. They’re looking at how the b quark changes into a d quark and an electron pair.

Lev: It shows that while we haven't found anything entirely new, this is important because it tests how well our current understanding of the Standard Model holds up when we look at these flavor-changing neutral current decays.

Kai: Exactly. They found a signal excess with a significance of three point two sigma, and they measured the branching fraction to be two point four plus zero point nine minus zero eight plus zero four minus zero two times ten to the negative eight.

Mira: That measurement is consistent with what the Standard Model predicts, which is good news for that part of physics. They also looked at a related decay, B plus to pi plus mu plus mu minus, and found that was also in line with expectations there.

Lev: So they are using these comparisons to build a picture of how solid our framework is for these b hadron decays right now. It's checking if the rules we have work when we test them on real data.

Kai: They used some pretty careful methods for background suppression, including multivariate classifiers and checks using particle identification information from things like RICH detectors and calorimeters.

Mira: And they accounted for things like how electrons lose energy through something called bremsstrahlung by adjusting the electron momentum to include energy from those specific detector clusters.

Lev: They also had specific vetoes to cut out backgrounds where a kaon gets mistaken for a pion, and they used mass cuts on pi plus electron minus to keep things clean.

Kai: In the end, they determined that these selection criteria lead them to a final branching fraction measurement of two point four plus zero point nine minus zero eight plus zero four minus zero two times ten to the negative eight.

Mira: The overall message is that this finding is consistent with the Standard Model expectation, and it represents the first evidence for a b to d lepton lepton transition. That’s a small step, but it opens up avenues for future searches for physics beyond the Standard Model in these areas.

LHCb collaboration

European Organization for Nuclear Research (CERN)

hep-ex

Submitted: 2026-04-29

Updated: 2026-11-02

Comments: All figures and tables, along with any supplementary material and additional information, are available at https://lbfence.cern.ch/alcm/public/analysis/full-details/4635 (LHCb public pages)

Journal ref: Phys. Rev. Lett. 137 (2026) 071804

DOI: 10.1103/97c9-jl1m

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 81/100

The gist: The first evidence for the decay B+→ π+e+e− is reported using proton-proton collision data recorded by the LHCb experiment at centre-of-mass energies of 7, 8 and 13 TeV…

Key concepts

Branching Fraction
This value represents the probability that a specific B+ particle will decay into the B+ to pi+ e+ e- mode. It is calculated by taking the observed number of signal events and dividing it by all other expected events, like background processes. The final result is expressed as a fraction.
Background Suppression Techniques
To find a rare signal, researchers used several methods to filter out unwanted noise from other particle decays. This included using advanced classifiers based on detector information (like RICH detectors) and vetoing events that look like other known decay processes, such as those involving D0 mesons.
Dielectron Mass Squared (q²)
This is a kinematic variable related to the energy and momentum of the two electrons in the decay. The analysis was split into two regions: low-q² and high-q², allowing researchers to examine different aspects of the decay process and check for consistency across these energy ranges.

Terminology

Summary

The first evidence for the decay B+→ π+e+e− is reported using proton-proton collision data recorded by the LHCb experiment at centre-of-mass energies of 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 <ref:2604. p. 0

Where the paper says it

EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN) CERN-EP-2026-055 LHCb-PAPER-2025-050 April 29, 2026 First evidence of the decay B+ → π+e+e− LHCb collaboration† Abstract The first evidence for the decay B+→ π +e +e − is reported using proton-proton collision data recorded by the LHCb experiment at centre-of-mass energies of 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 . A signal excess with a significance of 3.

26784#pg2>.

Signal Observation and Measurement

A signal excess with a significance of 3.2σ is observed and the branching fraction is measured to be B(B+→ π+e+e−) = (2. p. 0

Where the paper says it

A signal excess with a significance of 3. 2σ is observed and the branching fraction is measured to be B(B+→ π +e +e −) = (2.

4 +0.9 −0.8 +0.4 −0.2) × 10−8, where the first set of uncertainties is statistical and the second is systematic <ref:2604. p. 0
Where the paper says it

2σ is observed and the branching fraction is measured to be B(B+→ π +e +e −) = (2. 2 ) × 10−8 , where the first set of uncertainties is statistical and the second is systematic.

26784#pg2>. This result is consistent with the Standard Model expectation <ref:2604. p. 0
Where the paper says it

2 ) × 10−8 , where the first set of uncertainties is statistical and the second is systematic. The result is consistent with the Standard Model expectation.

26784#pg2>. The measurement was performed using data from pp collisions at centre-of-mass energies of √s = 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 <ref:2604. p. 2
Where the paper says it

4 × 10−8 at 90% confidence level for the B+→ π +e +e − decay [24]. In this Letter, a search for the B+→ π +e +e − decay is presented using data from pp collisions at centre-of-mass energies of √ s = 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 , recorded by the LHCb experiment during Run 1 (2011–2012) and Run 2 (2015–2018) of the LHC.

26784#pg4>.

Background Suppression Techniques

To manage the expected background from various processes, several stringent requirements are imposed on the signal candidates <ref:2604.26784#pg6>. These include:

  1. Electron, pion and kaon candidates are identified using multivariate classifiers that employ particle-identification (PID) information from the two ring-imaging Cherenkov (RICH) detectors, the calorimeters and the muon chambers <ref:2604. p. 3

    Where the paper says it

    Kaons, pions and electrons are identified using multivariate classifiers that employ particle-identification (PID) information from the two ring-imaging Cherenkov (RICH) detectors, the calorimeters and the muon chambers. Electron, pion and kaon candidates are also required to have sizeable pT and to be inconsistent with originating from a PV.

    26784#pg5>.

  2. Electrons lose a significant fraction of their energy through bremsstrahlung in the detector, and this effect is accounted for by correcting the electron momentum to include energy from specific ECAL clusters <ref:2604. p. 3

    Where the paper says it

    Electrons lose a significant fraction of their energy through bremsstrahlung in the detector. To account for this effect, the electron momentum is corrected to include the energy of any reconstructed clusters in the ECAL that match the position of the extrapolated electron track before the LHCb dipole magnet and are not associated with any other track in the event.

    26784#pg6>.

  3. Backgrounds where a kaon is misidentified as a pion are reduced using pion PID requirements, such as requiring m(π+→Ke−) > mD0 <ref:2604. p. 3

    Where the paper says it

    These background contributions, along with semileptonic B+→ h +h −e +νe decays, are suppressed to manageable levels with stringent electron PID requirements. Backgrounds where a kaon is misidentified as a pion, from both fully and partially reconstructed B→ K+e +e −(X) decays, contaminate the region below the B+ mass and are reduced using pion PID requirements.

    26784#pg6>.

  4. Decays mediated by b→ c transitions are significantly suppressed by vetoing candidates with 2m(π+e−→h) around the known D0 mass <ref:2604. p. 3

    Where the paper says it

    The contribution from the B+→ D0 (π +e −νe)e +νe decay is also suppressed by this veto, and further reduced by the requirements introduced to suppress combinatorial background. Similar decays where a hadron is misidentified as an electron, such as B+→ D0 (h +π −)e +νe, are suppressed by vetoing candidates with 2

    26784#pg6>.

Branching Fraction Extraction and Uncertainty

The branching fraction of the B+→ π+e+e− decay is calculated using a formula that incorporates the observed yield and reconstruction and selection efficiency for the corresponding decay mode <ref:2604.26784#pg7>. The final result is determined to be B(B+→ π+e+e−) = (2. p. 7

Where the paper says it

2σ and the branching fraction is determined for the first time. The result is in good 6

4 +0.9 −0.8 +0.4 −0.2) × 10−8, where the first uncertainties are statistical and the second are systematic <ref:2604. p. 0
Where the paper says it

2σ is observed and the branching fraction is measured to be B(B+→ π +e +e −) = (2. 2 ) × 10−8 , where the first set of uncertainties is statistical and the second is systematic.

p. 0
Where the paper says it

2σ is observed and the branching fraction is measured to be B(B+→ π +e +e −) = (2. 2 ) × 10−8 , where the first set of uncertainties is statistical and the second is systematic.

p. 0
Where the paper says it

2σ is observed and the branching fraction is measured to be B(B+→ π +e +e −) = (2. 2 ) × 10−8 , where the first set of uncertainties is statistical and the second is systematic.

26784#pg9>. The significance of the signal contribution is estimated using Wilks’ theorem and found to be 3. p. 7
Where the paper says it

The extrapolation to the full q 2 range is performed using B+→ π +e +e − simulation in which cc¯ resonances are not included [35]. The significance of the signal contribution is estimated using Wilks’ theorem [45] and found to be 3.

p. 7
Where the paper says it

The extrapolation to the full q 2 range is performed using B+→ π +e +e − simulation in which cc¯ resonances are not included [35]. The significance of the signal contribution is estimated using Wilks’ theorem [45] and found to be 3.

2σ, taking into account systematic uncertainties <ref:2604. p. 7
Where the paper says it

The significance of the signal contribution is estimated using Wilks’ theorem [45] and found to be 3. 2σ, taking into account systematic uncertainties.

p. 7
Where the paper says it

The significance of the signal contribution is estimated using Wilks’ theorem [45] and found to be 3. 2σ, taking into account systematic uncertainties.

26784#pg11>.

Region-Specific Analysis

The analysis was performed in two regions of dielectron mass squared, q2: a low-q2 region, q2 ∈ [0.045, 6] GeV2/c4 <ref:2604. p. 3

Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

26784#pg6>, and a high-q2 region, q2 ∈ [15, 25] GeV2/c4 <ref:2604. p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

26784#pg6>. The significance of the signal in these two regions is 3. p. 4
Where the paper says it

Two different BDT-output regions, referred to as LowBDT and HighBDT, are retained for the final analysis to help control the modelling of the combinatorial background in the final fit to data. The definition of the BDT regions is optimised independently in each q 2 and bremsstrahlung category to maximise the signal significance, computed using simulated pseudoexperiments.

p. 4
Where the paper says it

Two different BDT-output regions, referred to as LowBDT and HighBDT, are retained for the final analysis to help control the modelling of the combinatorial background in the final fit to data. The definition of the BDT regions is optimised independently in each q 2 and bremsstrahlung category to maximise the signal significance, computed using simulated pseudoexperiments.

2σ and 0.8σ, respectively <ref:2604. p. 7
Where the paper says it

The significance of the signal in these two regions is 3. 8σ, respectively.

p. 7
Where the paper says it

The significance of the signal in these two regions is 3. 8σ, respectively.

26784#pg12>. An upper limit to the branching fraction in the high-q2 region was computed using the CLs method to be B(B+→ π+e+e−) [15,25] GeV2/c4 <ref:2604. p. 7
Where the paper says it

8σ, respectively. As no signal is observed in the high-q 2 region, an upper limit to the branching fraction is computed using the CLs method [46] to be B(B+→ π +e +e −) [15,25] GeV2/c4 < 1.

26784#pg12>.

Consistency with Standard Model

All of the results above are in good agreement with the SM expectation <ref:2604. p. 7

Where the paper says it

49) × 10−8 at 90% (95%) confidence level. All of the results above are in good agreement with the SM expectation [17].

26784#pg10>. The profile-likelihood function is shown in Fig. p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

3 for the integrated q2 region and in Fig. p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

p. 9
Where the paper says it

End matter The profile likelihood as a function of the B+→ π +e +e − branching fraction is shown in Fig. 3 for the integrated q 2 region and in Fig.

4 for the two separate q2 regions <ref:2604. p. 9
Where the paper says it

3 for the integrated q 2 region and in Fig. 4 for the two separate q 2 regions.

p. 9
Where the paper says it

3 for the integrated q 2 region and in Fig. 4 for the two separate q 2 regions.

p. 9
Where the paper says it

3 for the integrated q 2 region and in Fig. 4 for the two separate q 2 regions.

p. 9
Where the paper says it

3 for the integrated q 2 region and in Fig. 4 for the two separate q 2 regions.

p. 9
Where the paper says it

3 for the integrated q 2 region and in Fig. 4 for the two separate q 2 regions.

26784#pg11>. The fit is also repeated by allowing the branching fraction to vary independently in the low-q2 and high-q2 regions, and the results are found to be consistent with each other <ref:2604. p. 7
Where the paper says it

The profile-likelihood function is shown in the End Matter. The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other.

p. 7
Where the paper says it

The profile-likelihood function is shown in the End Matter. The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other.

p. 7
Where the paper says it

The profile-likelihood function is shown in the End Matter. The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other.

26784#pg11>. The branching fractions determined for both regions are B(B+→ π+e+e−) [0. p. 7
Where the paper says it

The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other. The branching fractions are determined to be B(B+→ π +e +e −) [0.

045,6] GeV2/c4 = (0. p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

76 +0.30 −0.27 +0.14 − 0.28) × 10−8 and B(B+→ π+e+e−) [15,25] GeV2/c4 = (0. p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

p. 3
Where the paper says it

Signal candidates are considered in two regions of dielectron mass squared, q 2 : a low-q 2 region, q 2 ∈ [0. 045, 6] GeV2 /c4 , which is above the region contaminated by photonto-electron conversions but below the charmonium resonances, and a high-q 2 region, q 2 ∈ [15, 25] GeV2 /c4 , above the ψ(2S) resonance.

40 + 0.52 − 0.45 + 0.24 − 0.22) × 10−8 <ref:2604. p. 0
Where the paper says it

†Authors are listed at the end of this Letter. arXiv:2604.

p. 0
Where the paper says it

†Authors are listed at the end of this Letter. arXiv:2604.

p. 0
Where the paper says it

†Authors are listed at the end of this Letter. arXiv:2604.

26784#pg14>. The result should be interpreted as a measurement of the total branching fraction in the entire q2 range accessible, including the regions of q2 not measured in this analysis <ref:2604. p. 7
Where the paper says it

2 ) × 10−8 , where the first uncertainties are statistical and the second are systematic. This result should be interpreted as a measurement of the total branching fraction in the entire q 2 range accessible, including the regions of q 2 not measured in this analysis.

26784#pg14>. The extrapolation to the full q2 range is performed using B+→ π+e+e− simulation in which cc¯ resonances are not included <ref:2604. p. 7
Where the paper says it

This result should be interpreted as a measurement of the total branching fraction in the entire q 2 range accessible, including the regions of q 2 not measured in this analysis. The extrapolation to the full q 2 range is performed using B+→ π +e +e − simulation in which cc¯ resonances are not included [35].

26784#pg14>. The significance of the signal contribution is estimated using Wilks’ theorem and found to be 3.2σ, taking into account systematic uncertainties <ref:2604.26784#pg14>.

Conclusion

A search for the B+→ π+e+e− decay has been performed using data collected by the LHCb experiment during Run 1 and Run 2 of the LHC, corresponding to an integrated luminosity of 9 fb−1 <ref:2604.26784#pg14>. Evidence for the signal is found with a significance of 3. p. 7

Where the paper says it

In summary, a search for the B+→ π +e +e − decay has been performed using data collected by the LHCb experiment during Run 1 and Run 2 of the LHC, corresponding to an integrated luminosity of 9 fb−1 . Evidence for the signal is found with a significance of 3.

2σ and the branching fraction is determined for the first time <ref:2604. p. 7
Where the paper says it

Evidence for the signal is found with a significance of 3. 2σ and the branching fraction is determined for the first time.

26784#pg14>. The result is in good agreement with the SM expectation and represents the first evidence for a b→ de+e− transition <ref:2604. p. 8
Where the paper says it

agreement with the SM expectation and represents the first evidence for a b→ de+e − transition. Acknowledgements We express our gratitude to our colleagues in the CERN accelerator departments for the excellent performance of the LHC.

26784#pg14>.

Acknowledgements

The authors express gratitude to their colleagues in the CERN accelerator departments for the excellent performance of the LHC <ref:2604.26784#pg20>. They also acknowledge support from numerous national agencies and computing resources provided by various institutions <ref:2604. p. 8

Where the paper says it

We acknowledge support from CERN and from the national agencies: ARC (Australia); CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 and CEA (France); BMFTR, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MNiSW and NCN (Poland); MEC/IFA (Romania); MICIU and AEI (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by ARDC (Australia), CBPF (Brazil), CERN, IHEP and LZU (China), IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), Polish WLCG (Poland), IFIN-HH (Romania), PIC (Spain), CSCS (Switzerland), GridPP (United Kingdom), and NSF (USA).

26784#pg20>. The work relies on multiple open-source software packages <ref:2604. p. 8
Where the paper says it

We acknowledge the computing resources that are provided by ARDC (Australia), CBPF (Brazil), CERN, IHEP and LZU (China), IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), Polish WLCG (Poland), IFIN-HH (Romania), PIC (Spain), CSCS (Switzerland), GridPP (United Kingdom), and NSF (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend.

26784#pg20>.

References

The paper references a wide array of prior work concerning B hadron decays, lepton flavor universality tests, and detector performance <ref:2604. p. 2

Where the paper says it

Flavour-changing neutral-current (FCNC) decays of b hadrons are suppressed in the Standard Model (SM) and therefore constitute a good laboratory to search for the effects of new physics. Several tensions with respect to the SM predictions have been observed in the measurements of branching fractions [1–3] and angular observables [4–12] of decays mediated by the b→ sℓℓ transition, whereas the most recent tests of lepton flavour universality in these processes are consistent with the SM [13–15].

26784#pg20>. Key references include those detailing previous measurements of B+→ π+µ+µ− decay <ref:2604. p. 2
Where the paper says it

Several tensions with respect to the SM predictions have been observed in the measurements of branching fractions [1–3] and angular observables [4–12] of decays mediated by the b→ sℓℓ transition, whereas the most recent tests of lepton flavour universality in these processes are consistent with the SM [13–15]. The decay B+→ π +ℓ +ℓ − is mediated by the b→ dℓℓ quark-level transition, which is further suppressed by the CKM matrix element ratio |Vtd| 2/|Vts| 2 in the SM and is also sensitive to new physics effects.

p. 2
Where the paper says it

Several tensions with respect to the SM predictions have been observed in the measurements of branching fractions [1–3] and angular observables [4–12] of decays mediated by the b→ sℓℓ transition, whereas the most recent tests of lepton flavour universality in these processes are consistent with the SM [13–15]. The decay B+→ π +ℓ +ℓ − is mediated by the b→ dℓℓ quark-level transition, which is further suppressed by the CKM matrix element ratio |Vtd| 2/|Vts| 2 in the SM and is also sensitive to new physics effects.

26784#pg4>, and those providing context on lepton flavor universality in b→ sll transitions <ref:2604. p. 5
Where the paper says it

These include partially reconstructed b-hadron decays via b→ se+e − and b→ de+e − transitions, such as B+,0→ K∗ (892)+,0 e +e − or B+,0→ ρ(770)+,0 e +e −, where one of the hadrons is not reconstructed. Some of these decays have never been observed, hence their yields are estimated from the branching fractions of the corresponding b→ sµ+µ − or b→ dµ+µ − transitions, assuming lepton universality, and a systematic uncertainty is evaluated to account for this assumption.

p. 5
Where the paper says it

These include partially reconstructed b-hadron decays via b→ se+e − and b→ de+e − transitions, such as B+,0→ K∗ (892)+,0 e +e − or B+,0→ ρ(770)+,0 e +e −, where one of the hadrons is not reconstructed. Some of these decays have never been observed, hence their yields are estimated from the branching fractions of the corresponding b→ sµ+µ − or b→ dµ+µ − transitions, assuming lepton universality, and a systematic uncertainty is evaluated to account for this assumption.

26784#pg4>. The methodology utilizes various tools such as Pythia for event generation <ref:2604.26784#pg12>, Geant4 for detector simulation <ref:2604. p. 2
Where the paper says it

Decays of unstable particles are described by EvtGen [31], in which final-state radiation is generated using Photos [32]. The interaction of the generated particles with the detector, and its response, are implemented using the Geant4 toolkit [33] as described in Ref.

p. 2
Where the paper says it

Decays of unstable particles are described by EvtGen [31], in which final-state radiation is generated using Photos [32]. The interaction of the generated particles with the detector, and its response, are implemented using the Geant4 toolkit [33] as described in Ref.

26784#pg14>, and decision-theoretic classifiers like TMVA for background modeling <ref:2604. p. 4
Where the paper says it

The expected contribution from misidentified B+→ h +h −h + and B+→ h +h −e +νe backgrounds is also validated using a data-driven method developed for a similar LHCb analysis of b→ se+e − decays [37], based on a control sample selected by inverting the PID requirements. Combinatorial background is suppressed using a boosted decision tree (BDT) classifier [38, 39], implemented in the TMVA toolkit [40].

p. 4
Where the paper says it

The expected contribution from misidentified B+→ h +h −h + and B+→ h +h −e +νe backgrounds is also validated using a data-driven method developed for a similar LHCb analysis of b→ se+e − decays [37], based on a control sample selected by inverting the PID requirements. Combinatorial background is suppressed using a boosted decision tree (BDT) classifier [38, 39], implemented in the TMVA toolkit [40].

26784#pg6>. The systematic uncertainties are detailed in Table 1, which summarizes the relative uncertainties on the branching fraction measurement <ref:2604. p. 7
Where the paper says it

Table 1: Relative systematic uncertainties on the B+→ π +e +e − branching fraction measurement in the low- and high-q 2 regions, and on the integrated measurement. Source Low-q 2 [%] High-q 2 [%] Integrated [%] Efficiency estimation 13.

p. 7
Where the paper says it

Table 1: Relative systematic uncertainties on the B+→ π +e +e − branching fraction measurement in the low- and high-q 2 regions, and on the integrated measurement. Source Low-q 2 [%] High-q 2 [%] Integrated [%] Efficiency estimation 13.

26784#pg14>. The final results are consistent with SM predictions derived from lattice QCD calculations <ref:2604. p. 2
Where the paper says it

Flavour-changing neutral-current (FCNC) decays of b hadrons are suppressed in the Standard Model (SM) and therefore constitute a good laboratory to search for the effects of new physics. Several tensions with respect to the SM predictions have been observed in the measurements of branching fractions [1–3] and angular observables [4–12] of decays mediated by the b→ sℓℓ transition, whereas the most recent tests of lepton flavour universality in these processes are consistent with the SM [13–15].

p. 2
Where the paper says it

Flavour-changing neutral-current (FCNC) decays of b hadrons are suppressed in the Standard Model (SM) and therefore constitute a good laboratory to search for the effects of new physics. Several tensions with respect to the SM predictions have been observed in the measurements of branching fractions [1–3] and angular observables [4–12] of decays mediated by the b→ sℓℓ transition, whereas the most recent tests of lepton flavour universality in these processes are consistent with the SM [13–15].

26784#pg14>. The paper also includes detailed efficiency estimations and control sample calibrations to improve data-driven methods <ref:2604. p. 4
Where the paper says it

Residual contamination from any of these backgrounds is estimated using simulation. The expected contribution from misidentified B+→ h +h −h + and B+→ h +h −e +νe backgrounds is also validated using a data-driven method developed for a similar LHCb analysis of b→ se+e − decays [37], based on a control sample selected by inverting the PID requirements.

p. 4
Where the paper says it

Residual contamination from any of these backgrounds is estimated using simulation. The expected contribution from misidentified B+→ h +h −h + and B+→ h +h −e +νe backgrounds is also validated using a data-driven method developed for a similar LHCb analysis of b→ se+e − decays [37], based on a control sample selected by inverting the PID requirements.

26784#pg14>. The analysis incorporates various physical backgrounds, including semileptonic decays B+→ ρ(770)0(π+π−)e+νe and B+→ D0(π+e−νe)e+νe <ref:2604. p. 6
Where the paper says it

In some cases, the limited size of these simulation samples induces a large uncertainty on the expected background contribution, and the parametrisation of these components in the fit is impractical. These include the semileptonic decays B+→ ρ(770)0 (π +π −)e +νe and B+→ D0 (π +e −νe)e +νe, as well as b-hadron decays with photons in the final state that convert into electron-positron pairs in the detector, one of which is selected as one of 5

p. 6
Where the paper says it

In some cases, the limited size of these simulation samples induces a large uncertainty on the expected background contribution, and the parametrisation of these components in the fit is impractical. These include the semileptonic decays B+→ ρ(770)0 (π +π −)e +νe and B+→ D0 (π +e −νe)e +νe, as well as b-hadron decays with photons in the final state that convert into electron-positron pairs in the detector, one of which is selected as one of 5

26784#pg14>. The profile-likelihood function is shown in Fig. 3 for the integrated q2 region and in Fig. 4 for the two separate q2 regions <ref:2604.26784#pg14>. The systematic uncertainties are evaluated by generating pseudoexperiments with alternative configurations of the fit model or alternative values of the constrained parameters <ref:2604. p. 6
Where the paper says it

Systematic uncertainties arise from the evaluation of the efficiencies of signal and normalisation modes, from the modelling of the invariant-mass distributions and from the effect of possible residual backgrounds not included in the baseline fit model. Systematic uncertainties are evaluated by generating pseudoexperiments with alternative configurations of the fit model or alternative values of the constrained parameters.

p. 6
Where the paper says it

Systematic uncertainties arise from the evaluation of the efficiencies of signal and normalisation modes, from the modelling of the invariant-mass distributions and from the effect of possible residual backgrounds not included in the baseline fit model. Systematic uncertainties are evaluated by generating pseudoexperiments with alternative configurations of the fit model or alternative values of the constrained parameters.

26784#pg14>. The total uncertainty is obtained by adding all contributions in quadrature <ref:2604. p. 7
Where the paper says it

A summary of the systematic uncertainties is shown in Table 1. The total uncertainty is obtained by adding all contributions in quadrature.

p. 7
Where the paper says it

A summary of the systematic uncertainties is shown in Table 1. The total uncertainty is obtained by adding all contributions in quadrature.

26784#pg14>. The branching fractions are determined to be B(B+→ π+e+e−) = (2. p. 7
Where the paper says it

The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other. The branching fractions are determined to be B(B+→ π +e +e −) [0.

p. 7
Where the paper says it

The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other. The branching fractions are determined to be B(B+→ π +e +e −) [0.

4 +0.9 −0.8 +0.4 −0.2) × 10−8, where the first uncertainties are statistical and the second are systematic <ref:2604.26784#pg14>. The result is in good agreement with the SM expectation <ref:2604. p. 7
Where the paper says it

49) × 10−8 at 90% (95%) confidence level. All of the results above are in good agreement with the SM expectation [17].

p. 7
Where the paper says it

49) × 10−8 at 90% (95%) confidence level. All of the results above are in good agreement with the SM expectation [17].

26784#pg14>. The profile-likelihood function is shown in Fig. 3 for the integrated q2 region and in Fig. 4 for the two separate q2 regions <ref:2604.26784#pg14>. The fit is also repeated by allowing the branching fraction to vary independently in the low-q2 and high-q2 regions, and the results are found to be consistent with each other <ref:2604.26784#pg14>. The branching fractions are determined to be B(B+→ π+e+e−) [0. p. 7
Where the paper says it

The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other. The branching fractions are determined to be B(B+→ π +e +e −) [0.

p. 7
Where the paper says it

The fit is also repeated by allowing the branching fraction to vary independently in the low-q 2 and high-q 2 regions, and the results are found to be consistent with each other. The branching fractions are determined to be B(B+→ π +e +e −) [0.

045,6] GeV2/c4 = (0.76 +0.30 − 0.27 + 0.14 − 0.28) × 10−8 and B(B+→ π+e+e−) [15,25] GeV2/c4 = (0.40 + 0.52 − 0.45 + 0.24 − 0.22) × 10−8 <ref:2604.26784#pg14>. The significance of the signal in these two regions is 3.2σ and 0.8σ, respectively <ref:2604.26784#pg14>. As no signal is observed in the high-q2 region, an upper limit to the branching fraction is computed using the CLs method to be B(B+→ π+e+e−) [15,25] GeV2/c4. p. 7
Where the paper says it

8σ, respectively. As no signal is observed in the high-q 2 region, an upper limit to the branching fraction is computed using the CLs method [46] to be B(B+→ π +e +e −) [15,25] GeV2/c4 < 1.

All of the results above are in good agreement with the SM expectation <ref:2604.26784#pg14>. The first evidence for a b→ de+e− transition is reported <ref:2604. p. 2
Where the paper says it

21) × 10−8 [17]. Evidence for other b→ dµ+µ − decays has also been reported [21–23], while no b→ de+e − quark-level transition has been observed so far.

26784#pg14>. The paper references a wide array of prior work concerning B hadron decays, lepton flavor universality tests, and detector performance <ref:2604.26784#pg20>. Key references include those detailing previous measurements of B+→ π+µ+µ− decay <ref:2604.26784#pg4>, and those providing context on lepton flavor universality in b→ sll transitions <ref:2604.26784#pg4>. The methodology utilizes various tools such as Pythia for event generation <ref:2604.26784#pg12>, Geant4 for detector simulation <ref:2604.26784#pg14>, and decision-theoretic classifiers like TMVA for background modeling <ref:2604.26784#pg6>. The systematic uncertainties are detailed in Table 1, which summarizes the relative uncertainties on the branching fraction measurement <ref:2604.26784#pg14>. The final results are consistent with SM predictions derived from lattice QCD calculations <ref:2604.26784#pg14>. The paper also includes detailed efficiency estimations and control sample calibrations to improve data-driven methods <ref:2604.26784#pg14>. The analysis incorporates various physical backgrounds, including semileptonic decays B+→ ρ(770)0(π+π−)e+νe and B+→ D0(π+e−νe)e+νe <ref:2604.26784#pg14>. The profile-likelihood function is shown in Fig. 3 for the integrated q2 region and in Fig. 4 for the two separate q2 regions <ref:2604.26784#pg14>. The systematic uncertainties are evaluated by generating pseudoexperiments with alternative configurations of the fit model or alternative values of the constrained parameters <ref:2604.26784#pg14>. The total uncertainty is obtained by adding all contributions in quadrature <ref:2604.26784#pg14>. The branching fractions are determined to be B(B+→ π+e+e−) = (2.4 +0.9 −0.8 +0.4 −0.2) × 10−8, where the first uncertainties are statistical and the second are systematic <ref:2604.26784#pg14>. The result is in good agreement with the SM expectation <ref:2604.26784#pg14>. The profile-likelihood function is shown in Fig. 3 for the integrated q2 region and in Fig. 4 for the two separate q2 regions <ref:2604.26784#pg14>. The fit is also repeated by allowing the branching fraction to vary independently in the low-q2 and high-q2 regions, and the results are found to be consistent with each other <ref:2604.26784#pg14>. The branching fractions are determined to be B(B+→ π+e+e−) [0.045,6] GeV2/c4 = (0.76 +0.30 − 0.27 + 0.14 − 0.28) × 10−8 and B(B+→ π+e+e−) [15,25] GeV2/c4 = (0.40 + 0.52 − 0.45 + 0.24 − 0.22) × 13>

Pruning redundant reasoning tokens

The paper presents the first evidence for the decay B+→ π+e+e− using data from pp collisions at centre-of-mass energies of 7, 8 and 13 TeV with an integrated luminosity of 9 fb−1 <ref:2604.26784#pg4>. This measurement yielded a signal excess with a significance of 3. p. 0

Where the paper says it

EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN) CERN-EP-2026-055 LHCb-PAPER-2025-050 April 29, 2026 First evidence of the decay B+ → π+e+e− LHCb collaboration† Abstract The first evidence for the decay B+→ π +e +e − is reported using proton-proton collision data recorded by the LHCb experiment at centre-of-mass energies of 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb−1 . A signal excess with a significance of 3.

2σ, and the measured branching fraction is B(B+→ π+e+e−) = (2. p. 0
Where the paper says it

A signal excess with a significance of 3. 2σ is observed and the branching fraction is measured to be B(B+→ π +e +e −) = (2.

4 +0.9 −0.8 +0.4 −0.2) × 10−8 <ref:2604.26784#pg2>. The result is consistent with the Standard Model expectation <ref:2604. p. 0
Where the paper says it

2 ) × 10−8 , where the first set of uncertainties is statistical and the second is systematic. The result is consistent with the Standard Model expectation.

26784#pg2>.

How the training signal is built

The combinatorial background is suppressed using a boosted decision tree (BDT) classifier implemented in the TMVA toolkit <ref:2604.26784#pg6>. The BDT is trained with simulated B+→ π+e+e− decays as a proxy for signal, while background is represented by data in the lower and upper mass sidebands defined by m(π+e+e−) 5. p. 4

Where the paper says it

Combinatorial background is suppressed using a boosted decision tree (BDT) classifier [38, 39], implemented in the TMVA toolkit [40]. The BDT is trained with simulated B+→ π +e +e − decays as a proxy for signal, while background is represented by data in the lower and upper mass sidebands, defined by m(π +e +e −) < 5.

5 GeV/c2 <ref:2604. p. 0
Where the paper says it

†Authors are listed at the end of this Letter. arXiv:2604.

26784#pg8>. Two different BDT-output regions, referred to as LowBDT and HighBDT, are retained for the final analysis to help control the modelling of the combinatorial background in the final fit to data <ref:2604. p. 4
Where the paper says it

The training features include kinematic and topological properties of the final-state particles, the dielectron candidate and the reconstructed B+ meson, as well as the fit quality of the B+ decay vertex and its displacement with respect to its associated PV, which is defined as the PV with which the candidate has the smallest impact parameter significance. Two different BDT-output regions, referred to as LowBDT and HighBDT, are retained for the final analysis to help control the modelling of the combinatorial background in the final fit to data.

26784#pg8>. The definition of these BDT regions is optimized independently in each q2 and bremsstrahlung category to maximize the signal significance computed using simulated pseudoexperiments <ref:2604. p. 4
Where the paper says it

Two different BDT-output regions, referred to as LowBDT and HighBDT, are retained for the final analysis to help control the modelling of the combinatorial background in the final fit to data. The definition of the BDT regions is optimised independently in each q 2 and bremsstrahlung category to maximise the signal significance, computed using simulated pseudoexperiments.

26784#pg8>.

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

  1. Improved Particle Physics Simulation for New Physics Searches This system can simulate B+ to pi+ee- decays using EvtGen and Geant4 to generate final-state radiation and detector responses, allowing researchers to test new physics models against experimental data.

  2. Enhanced Background Rejection Capabilities This AI system can implement a multivariate algorithm used in the LHCb analysis, which is described as a multivariate algorithm [27, 28] used to identify secondary vertices consistent with b-hadron decays.

  3. Optimized Signal Extraction Based on Kinematic Cuts This system can utilize the defined signal regions based on dielectron mass squared, specifically the low-q2 region, q2 ∈ [0.045, 6] GeV2/c4 and the high-q2 region, q2 ∈ [15, 25] GeV2/c4, to maximize signal significance.

  4. Data-Driven Background Modeling This system can employ a data-driven method developed for a similar LHCb analysis of b→ se+e− decays to estimate expected contributions from residual physical backgrounds by generating pseudoexperiments with alternative configurations of the fit model or alternative values of the constrained parameters.

  5. Uncertainty Quantification for Branching Fractions This AI system can calculate the branching fraction using the formula provided, accounting for statistical and systematic uncertainties, and extrapolate to a full q2 range using B+→ π+e+e− simulation in which cc¯ resonances are not included [35].

Abstract

The first evidence for the decay B+ toπ+ e+ e- is reported using proton-proton collision data recorded by the LHCb experiment at centre-of-mass energies of 7, 8 and 13 TeV, corresponding to an integrated luminosity of 9 fb-1. A signal excess with a significance of 3.2 σ is observed and the branching fraction is measured to be (B+ toπ+ e+ e-) = (2.4,+0.9-0.8,+0.4-0.2) times 10-8, where the first set of uncertainties is statistical and the second is systematic. The result is consistent with the Standard Model expectation.

Sources

Related papers