Evidence for the decay B 0 to mu+ mu- and improved measurements of the decay B 0 s to mu+ mu-

arXiv:2611.00230 · hep-ex · Submitted 2026-10-30 · 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: We're looking at the paper: "Evidence for the decay B 0 to mu+ mu- and improved measurements of the decay B 0 s to mu+ mu-".

Mira: Here's the gist of the paper. The first single-experiment evidence for the rare decay B0 → µ+µ− and improved measurements of the B0s → µ+µ− branching fraction and effective lifetime are reported Evidence for the decay B0 → µ+µ− and improved measurements of the decay B0s → µ+µ− branching fraction and effective lifetime are reported using Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025 The results are obtained using Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025 All measurements are found to be in agreement with the Standard Model expectations Experimental Setup and Data Combination The analysis utilizes Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025.

Kai: The results are reported by combining the Run 3 data with the previously analysed Run 1 and Run 2 data, which were recorded at √s = 7 TeV (2011), 8 TeV (2012) and 13 TeV (2015–2018), with a total integrated luminosity of 9 fb−1.

Mira: The LHCb Upgrade I detector is a single-arm forward spectrometer covering the pseudorapidity range 2 < η < 5, designed for the study of particles containing b or c quarks.

Kai: This detector represents a major enhancement of the LHCb experiment, allowing operation at an instantaneous luminosity of 2 × 10 cubed cm−2 s−1, five times that of the previous detector.

Title and authors: Kai: This paper, "Evidence for the decay B zero to mu+ mu- and improved measurements of the decay B zero s to mu+ mu-", is essentially reporting that they have found some single experiment evidence for a very rare process we haven't seen before.

Mira: And at the same time, they’re pushing the precision limits for two other important decays: the B zero s to two muons and the B zero to two muons.

Lev: What interests me is that they are using Run three proton-proton collision data with an integrated luminosity of eighteen femtobarns recorded between two thousand twenty-four and two thousand twenty-five which is a major experimental feat <ref:2611.00230#pg1,an integrated luminosity of 18>.

Kai: Exactly, and it shows how combining these different datasets from different runs helps you improve the precision of these measurements significantly.

Mira: It sets the stage for how we can use this precision to test theories about particle physics beyond what we currently know in the Standard Model by looking at flavor structures.

The paper's summary: Kai: The paper summarizes how they used Run three proton-proton collision data with an integrated luminosity of eighteen femtobarns recorded between two thousand twenty-four and two thousand twenty-five to look for this decay <ref:2611.00230#pg1,an integrated luminosity of 18>.

Mira: They describe their methodology where they use a boosted decision tree classifier to filter out the background noise when identifying candidate B zero s decays.

Lev: And they determine the signal yields by doing an extended unbinned maximum likelihood fit on the dimuon invariant mass, which is done in five different subsamples based on that classifier's performance.

Kai: They also measure the effective lifetime using candidates where their decision tree score is greater than zero point five, which they specifically optimized to give them the best expected signal significance for those B zero s decays.

Mira: For the branching fractions, they extracted them from an extended unbinned maximum likelihood fit over the dimuon invariant mass range of four thousand nine hundred to five thousand nine hundred mega electron volts per speed of light squared.

The paper's improvements: Kai: Now, let's talk about what they actually improved on in this paper. They report new signal branching fractions for both the B zero s to two muons and the B zero to two muons decays.

Mira: And they also give a measurement for the ratio of these two branching fractions called R mu mu, which is a theoretically clean observable because it relates those two decays directly.

Lev: They also determined an effective lifetime for the B zero s decay, which they found to be one point four two plus or minus zero point eighteen plus or minus zero point zero two picoseconds.

Kai: And they did this by fitting the background-subtracted decay-time distribution in a single bin from the decision tree response, and they also combined this with results from previous work to get a more precise lifetime of one point six plus or minus zero point sixteen plus or minus zero point zero one picoseconds.

Conclusion: Mira: So, to wrap up, the main thing is that they found evidence for the B zero decaying into two muons and got what they call world's most precise measurements for the B zero s branching fraction and lifetime.

Kai: And they also managed to measure R mu mu with a value of zero point three eight plus or minus zero point fourteen, which is really telling about new flavor structures beyond the Standard Model.

Lev: From an engineering standpoint, they did this by making use of the LHCb Upgrade I detector, and that detector is what allowed them to get these measurements so precise.

Mira: And it confirms that all their results agree with what we expect from the Standard Model, which gives us confidence in our current understanding of particle interactions.

Kai: It’s a solid result for the B zero s decay, and it opens a new door for using this precision to look for new physics contributions at energy scales that are too high to reach directly at the LHC.

Lev: For me, the real constraint is how much overhead you need on real hardware to run something like this effectively; that’s where the engineering challenge lies.

Kai: That's a solid result for the B zero s decay, and it opens a new door for using this precision to look for new physics contributions at energy scales that are too high to reach directly at the LHC.

European Organization for Nuclear Research (CERN)

hep-ex

Submitted: 2026-10-30

Updated: 2026-11-13

Comments: All figures and tables, along with machine-readable versions and any supplementary material and additional information, are available at https://lhcb-glance.cern.ch/alcm/public/analysis/full-details/6370/ (LHCb public pages)

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

Importance score: 85/100

The gist: The first single-experiment evidence for the rare decay B0 → µ+µ− and improved measurements of the B0s → µ+µ− branching fraction and effective lifetime are reported Evidence for the decay…

Key concepts

B0(s) → μ+μ− Decay
This is a rare decay process where a B0 meson (a specific type of subatomic particle containing a bottom quark) decays into two muons. Measuring this decay helps physicists understand the properties and behavior of these heavy particles, which are central to testing the Standard Model.
Boosted Decision Tree (BDT) Classifier
A machine learning tool used to separate signal events from background noise. It is trained using simulated B0s decays as the 'signal' and other particle interactions as 'background'. The BDT uses various features, such as track isolation and angular differences, to accurately identify the rare decay events.
Effective Lifetime (τμμ)
This measures how long a B0s meson exists before decaying. It is determined by fitting the distribution of decay times. The measurement helps constrain New Physics contributions by encoding information about the relative contributions of light and heavy mass eigenstates in the decay process.

Terminology

Summary

The first single-experiment evidence for the rare decay B0 → µ+µ− and improved measurements of the B0s → µ+µ− branching fraction and effective lifetime are reported

Evidence for the decay B0 → µ+µ− and improved measurements of the decay B0s → µ+µ− branching fraction and effective lifetime are reported using Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025

The results are obtained using Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025

All measurements are found to be in agreement with the Standard Model expectations

Experimental Setup and Data Combination

The analysis utilizes Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025. p. 0

Where the paper says it

The results are obtained using Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb−1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13. 6 TeV between 2024 and 2025.

The results are reported by combining the Run 3 data with the previously analysed Run 1 and Run 2 data, which were recorded at √s = 7 TeV (2011), 8 TeV (2012) and 13 TeV (2015–2018), with a total integrated luminosity of 9 fb−1. p. 4
Where the paper says it

6 TeV, corresponding to an integrated luminosity of 18 fb−1 . In addition, results are reported combining the Run 3 data with the previously analysed Run 1 and 2 data, which were recorded at √ s = 7 TeV (2011), 8 TeV (2012) and 13 TeV (2015–2018), with a total integrated luminosity of 9 fb−1 .

The LHCb Upgrade I detector is a single-arm forward spectrometer covering the pseudorapidity range 2 < η < 5, designed for the study of particles containing b or c quarks. p. 4
Where the paper says it

The results are the most precise to date by a single experiment and supersede those presented in Refs. Detector and simulation The LHCb Upgrade I detector [47] is a single-arm forward spectrometer covering the pseudorapidity range 2 < η < 5, designed for the study of particles containing b or c quarks.

This detector represents a major enhancement of the LHCb experiment, allowing operation at an instantaneous luminosity of 2 × 10 cubed cm−2 s−1, five times that of the previous detector. p. 4
Where the paper says it

The identification of particles [49] is performed with two ring-imaging Cherenkov (RICH) detectors, an electromagnetic calorimeter, a hadronic calorimeter and a muon system. The Upgrade I detector represents a major enhancement of the LHCb experiment, allowing operation at an instantaneous luminosity of 2 × 1033 cm−2 s −1 , five times that of the previous detector.

Analysis Methodology

The analysis follows a similar strategy to that adopted with the previous version of the LHCb experiment for Run 1 and 2 data, but introduces developments enabled by the Upgrade I detector and Run 3 data-taking conditions. Candidate B0(s)→µ+µ− decays are identified using a boosted decision tree (BDT) classifier trained to reduce combinatorial background. p. 5

Where the paper says it

Analysis method The analysis follows a similar strategy to that adopted with the previous version of the LHCb experiment for Run 1 and 2 data [43, 44], but introduces developments enabled by the Upgrade I detector and Run 3 data-taking conditions. Following the trigger and preselection, candidate B0 (s)→ µ +µ − decays are identified using a boosted decision tree (BDT) classifier [60, 61] trained to reduce combinatorial background.

Signal yields of B0s → µ+µ− and B0→ µ+µ− decays are determined with an extended unbinned maximumlikelihood fit to the dimuon invariant mass, performed simultaneously in five subsamples split according to the response of the BDT classifier. p. 5
Where the paper says it

Following the trigger and preselection, candidate B0 (s)→ µ +µ − decays are identified using a boosted decision tree (BDT) classifier [60, 61] trained to reduce combinatorial background. The signal yields of B0 s → µ +µ − and B0→ µ +µ − decays are determined with an extended unbinned maximumlikelihood fit to the dimuon invariant mass, performed simultaneously in five subsamples split according to the response of the BDT classifier.

The effective lifetime is measured using candidates passing the requirement BDT > 0. p. 6
Where the paper says it

25 is not included in the fit. The effective lifetime is measured using candidates passing the requirement BDT > 0.

5, which is optimised to give the maximum expected signal significance for B0s → µ+µ− decays. p. 6
Where the paper says it

The effective lifetime is measured using candidates passing the requirement BDT > 0. 5, which is optimised to give the maximum expected signal significance for B0 s → µ +µ − decays.

Branching Fraction Extraction

The branching fractions of B0→ µ+µ− and B0s → µ+µ− decays are extracted from an extended unbinned maximum-likelihood fit to the dimuon invariant mass (mµ+µ−) in the range [4900, 5900] MeV/c2. Three types of backgrounds are considered in the fit: B0(s)→ h+h′− decays, semileptonic decays, and combinatorial background. p. 7

Where the paper says it

The branching fractions of B0→ µ +µ − and B0 s → µ +µ − decays are extracted from an extended unbinned maximum-likelihood fit to the dimuon invariant mass (mµ+µ− ) in the range [4900, 5900] MeV/c 2 . Three types of backgrounds are considered in the fit: B0 (s)→ h +h ′− decays, semileptonic decays, and combinatorial background.

The expected yield for each background source is estimated using simulation. p. 8
Where the paper says it

normalisation mode. The expected yields of each background source in the mass and BDT regions used for the fit are estimated to be 37 ± 2 B0 (s)→ h +h ′−, 250 ± 34 B0 → π −µ +νµ, 18 ± 6 B0 s → K−µ +νµ, 32 ± 13 Λ 0 b → pµ−νµ, 123 ± 12 B0(+) → π 0(+)µ +µ −, and 90 ± 8 B+ c → J/ψµ+νµ decays.

The contribution of B0s → µ+µ−γ decays is found to be negligible. p. 8
Where the paper says it

The expected yields are Gaussian-constrained in the fit according to the values and uncertainties of their estimated efficiencies and branching fractions [63]. The contribution of B0 s → µ +µ −γ decays is found to be negligible.

Effective Lifetime Determination

The effective lifetime is measured in a single bin of the BDT response. A fit to the background-subtracted decay-time distribution is then performed in order to determine the effective lifetime. p. 8

Where the paper says it

Combinatorial background, B0→ µ +µ − decays and misidentified B0 (s)→ h +h ′− decays are modelled in the fit, with the B0 (s)→ h +h ′− yield Gaussian-constrained to the expected value, determined using the method described above, while all other yields are allowed to float freely. A fit to the background-subtracted decay-time distribution is then performed in order to determine the effective lifetime.

The efficiency to reconstruct and select B0s → µ+µ− decays as a function of decay time is parametrised as the exponential of a cubic spline, which is flexible enough to capture the features of the efficiency curve while being guaranteed to be positive. p. 8
Where the paper says it

A fit to the background-subtracted decay-time distribution is then performed in order to determine the effective lifetime. The efficiency to reconstruct and select B0 s → µ +µ − decays as a function of decay time is parametrised as the exponential of a cubic spline, which is flexible enough to capture the features of the efficiency curve while being guaranteed to be positive.

Results and Conclusions

The signal branching fractions are determined to be B(B0s → µ+µ−) = 3.31 + 0.31 + 0.15 − 0.30 − 0.13 × 10−9, B(B0→ µ+µ−) = 1.26 + 0.51 + 0.06 − 0.47 − 0.47 ±- - - - × 10−10. An excess of B0→ µ+µ− events is observed with respect to the background expectation with a significance of 3. p. 6

Where the paper says it

This classifier is trained using simulated B0 s → µ +µ − decays as signal proxy and a large simulated sample of inclusive b ¯b → Xµ+µ − processes as background. It combines information on the track impact parameter (IP) with respect to the associated PV, the distance of closest approach of the tracks, and the B candidate IP, χ 2 IP, vertex fit χ 2 and the angle between its momentum vector and the line joining the PV to its decay vertex.

2 standard deviations, taking systematic uncertainties into account. p. 5
Where the paper says it

The resulting B meson candidates must have a good vertex-fit quality, a significant displacement from any primary vertex (PV) and have a momentum that is well aligned with the vector joining the decay vertex and its associated PV. Since this algorithm imposes no PID requirements on the tracks, it reconstructs both B0 (s)→ µ +µ − and B0 (s)→ h +h ′− decays (where h ( ′ ) = K, π), reducing systematic uncertainties associated with their selection efficiencies.

The evidence for the B0→ µ+µ− decay enables the ratio of branching fractions to be measured as Rµµ = 0. p. 3
Where the paper says it

The measurement of Rµµ can therefore reveal new flavour structures beyond the CKM paradigm. In addition to the branching fractions and their ratio, the effective lifetime of the B0 s → µ +µ − decay can be measured.

038 + 0.016 − 0.014 ± 0.47, where the first uncertainty is statistical and the second systematic. p. 8
Where the paper says it

The resulting distribution is described using Gaussian kernel density estimates [66]. The effect on the final results of using alternative models to parametrise the B0 (s)→ h +h ′− mass distribution and the statistical uncertainty on its shape are found to be negligible.

The mass fit used in the lifetime measurement and the resulting background-subtracted decay-time distribution are shown in Fig. p. 8
Where the paper says it

The parameters of this efficiency function are determined using a fit to the background-subtracted B0→ K+π − decay-time distribution in data, shown in Fig. 4, where the B0 lifetime is fixed to the known value [63] and the parameters of the acceptance function are allowed to float.

6. The best-fit value for the effective lifetime is τµµ = 1. p. 8
Where the paper says it

Combinatorial background, B0→ µ +µ − decays and misidentified B0 (s)→ h +h ′− decays are modelled in the fit, with the B0 (s)→ h +h ′− yield Gaussian-constrained to the expected value, determined using the method described above, while all other yields are allowed to float freely. A fit to the background-subtracted decay-time distribution is then performed in order to determine the effective lifetime.

42 ± 0.18 ± 0.02 ps, where the first uncertainty is statistical and the second systematic. p. 8
Where the paper says it

The resulting distribution is described using Gaussian kernel density estimates [66]. The effect on the final results of using alternative models to parametrise the B0 (s)→ h +h ′− mass distribution and the statistical uncertainty on its shape are found to be negligible.

Validation of Measurement

The measurement of the effective lifetime is statistically dominated and the systematic uncertainty is uncorrelated with that of the result presented in Refs. [43, 44]. The results are combined using an inverse-variance weighted average, yielding τµµ = 1. p. 9

Where the paper says it

Semileptonic backgrounds with a neutrino in the final state are grouped in the orange component. is performed using the resulting acceptance function multiplied by a single exponential e −t/τµµ .

60 ± 0.16 ± 0.01 ps, where the first uncertainty is statistical and the second systematic. p. 8
Where the paper says it

The resulting distribution is described using Gaussian kernel density estimates [66]. The effect on the final results of using alternative models to parametrise the B0 (s)→ h +h ′− mass distribution and the statistical uncertainty on its shape are found to be negligible.

Conclusion

Enabled by the unique capabilities of the LHCb Upgrade I detector, evidence for the B0→ µ+µ− decay is reported with a significance of over three standard deviations, along with the world’s most precise determinations of the B0s → µ+µ− and B0→ µ+µ− branching fractions and the B0s → µ+µ− effective lifetime. This work marks the first analysis of B0(s) meson decays into two muons performed with the new detector, demonstrating its capabilities to perform precision flavour physics at high luminosity. p. 4

Where the paper says it

The detector includes a high precision tracking system consisting of a silicon-pixel vertex detector (VELO) surrounding the interaction region, a large-area silicon-strip detector located upstream of a dipole magnet, and three stations of scintillating-fibre detectors located downstream of the magnet [48]. The identification of particles [49] is performed with two ring-imaging Cherenkov (RICH) detectors, an electromagnetic calorimeter, a hadronic calorimeter and a muon system.

All results are consistent with SM expectations and constrain new physics contributions at energy scales beyond those accessible through direct production at the LHC. p. 2
Where the paper says it

However, this suppression makes these processes particularly sensitive to contributions from high-scale NP, which can alter their branching fractions, effective lifetimes, charge-parity asymmetries and angular observables. Measurements of these processes can probe energy scales up to 100 TeV [12], significantly above the direct reach of the LHC.

Pruning redundant reasoning tokens

The suppression of the B0→ µ+µ− decay rate, relative to that of the B0s → µ+µ− decay, is predominantly due to the hierarchy of CKM matrix elements. The ratio of branching fractions is a theoretically clean observable and in the SM is given by Rµµ ≡ B(B0→ µ+µ−) / B(B0s → µ+µ−) = τB0 ΓH / (fB0 fB0s squared Vtd Vts squared q M2 B0 − 4m2 µ q M2 B0s − 4m2 µ, (1). p. 3

Where the paper says it

The suppression of the B0→ µ +µ − decay rate, relative to that of the B0 s → µ +µ − decay, is predominantly due to the hierarchy of CKM matrix elements. The ratio of branching fractions is a theoretically clean observable and in the SM is given by Rµµ ≡ B(B0→ µ +µ −) B(B0 s → µ+µ−) = τB0 ΓH ℏ fB0 fB0 s 2 Vtd Vts 2 q M2 B0 − 4m2 µ q M2 B0 s − 4m2 µ , (1) where τB0 denotes the B0 lifetime, ΓH the decay width of the heavy B0 s mass eigenstate, MB0 and MB0 s the B0 (s) meson masses, fB0 and fB0 s their decay constants, and mµ the muon mass.

The SM prediction RSMµµ = 0. p. 2
Where the paper says it

By judicious choice of observables, it is possible to gain sensitivity to unobserved particles through their quantum mechanical interference with SM processes, even when their rest mass may be orders of magnitude higher than the energy scale of the collider. This strategy has a long historical pedigree: the suppression of strangeness-changing neutral-current decays led to the prediction of the charm quark [5] before its direct discovery [6, 7], while the rate of B0–B0 mixing [8] and precision electroweak measurements [9] provided early indirect evidence for a heavy top quark, prior to its direct observation in 1995 [10, 11].

0280 ± 0.0016 [33] also holds in all NP theories satisfying the Minimal Flavour Violation (MFV) hypothesis [36]. p. 3
Where the paper says it

The SM prediction RSM µµ = 0. 0016 [33] also holds in all NP theories satisfying the Minimal Flavour Violation (MFV) hypothesis [36].

How the training signal is built

The BDT classifier is trained using simulated B0s → µ+µ− decays as signal proxy and a large simulated sample of inclusive b¯b → Xµ+µ− processes as background that have passed the trigger and offline selection. The most discriminating features used in the BDT are the responses of two dedicated isolation classifiers trained using tracks from the underlying event that do not form the signal candidate, exploiting the tendency of B0(s)→ µ+µ− decays to be more isolated, i. p. 6

Where the paper says it

This background is controlled using a BDT classifier, trained using simulated B0 s → µ +µ − decays as signal and inclusive b ¯b → Xµ+µ − events as background that have passed the trigger and offline selection. The most discriminating features used in the BDT are the responses of two dedicated isolation classifiers trained using tracks from the underlying event that do not form the signal candidate, exploiting the tendency of B0 (s)→ µ +µ − decays to be more isolated, i.

e. to have fewer additional nearby tracks, than background candidates. p. 6
Where the paper says it

The most discriminating features used in the BDT are the responses of two dedicated isolation classifiers trained using tracks from the underlying event that do not form the signal candidate, exploiting the tendency of B0 (s)→ µ +µ − decays to be more isolated, i. to have fewer additional nearby tracks, than background candidates.

Additional training features for this BDT are: p∆ϕ2 + ∆η2 where ∆ϕ and ∆η are the differences between the azimuthal angle and pseudorapidity of the two tracks forming the B candidate, the minimum χ 2IP of the signal candidate tracks, the angle between its momentum vector and the line joining the PV to its decay vertex, and vertex fit χ squared of the B candidate. p. 6
Where the paper says it

to have fewer additional nearby tracks, than background candidates. Additional training features for this BDT are: p ∆ϕ2 + ∆η 2 where ∆ϕ and ∆η are the differences between the azimuthal angle and pseudorapidity of the two tracks forming the B candidate, the minimum χ 2 IP of the signal candidate tracks, the angle between the B candidate momentum vector and the line between the PV and its decay vertex, the vertex fit χ 2 of the B candidate and the χ 2 IP of the B candidate.

Where the gains come from

The LHCb Upgrade I detector is a single-arm forward spectrometer covering the pseudorapidity range 2 < η < 5, designed for the study of particles containing b or c quarks. The Upgrade I detector maintains or improves the excellent track reconstruction and particle identification (PID) performance of the original LHCb detector in the higher pile-up environment of Run 3. p. 4

Where the paper says it

To enable this, the tracking and RICH subsystems have been replaced with higher granularity detectors, while the existing muon system has been upgraded with finer readout granularity and improved shielding. The Upgrade I detector maintains or improves the excellent track reconstruction and particle identification (PID) performance of the original LHCb detector in the higher pile-up environment of Run 3 [50].

Furthermore, the initial hardware trigger stage has been replaced with an all-software trigger, leading to substantial improvements in efficiencies for hadronic and electron final states. p. 4
Where the paper says it

The Upgrade I detector maintains or improves the excellent track reconstruction and particle identification (PID) performance of the original LHCb detector in the higher pile-up environment of Run 3 [50]. Furthermore, the initial hardware trigger stage has been replaced with an all-software trigger [51], leading to substantial improvements in efficiencies for hadronic and electron final states.

The results are the most precise to date by a single experiment and supersede those presented in Refs. p. 4
Where the paper says it

In addition, results are reported combining the Run 3 data with the previously analysed Run 1 and 2 data, which were recorded at √ s = 7 TeV (2011), 8 TeV (2012) and 13 TeV (2015–2018), with a total integrated luminosity of 9 fb−1 . The results are the most precise to date by a single experiment and supersede those presented in Refs.

[43, 44].

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. Precision Measurement of Rare B Meson Decays: The improved AI system can perform world’s most precise measurements of the Bs to mu plus mu minus branching fraction and effective lifetime by utilizing Run 3 proton-proton collision data with an integrated luminosity of 18 fb minus one.

  2. New Physics Sensitivity via Ratio Measurement: The system can measure the ratio Rmu mu, which is a theoretically clean observable, revealing new flavour structures beyond the CKM paradigm by comparing B0 to mu plus mu minus and B0 s to mu plus mu minus decays.

  3. Effective Lifetime Determination with NP Constraints: The AI can determine the effective lifetime of the Bs to mu plus mu minus decay and constrain New Physics contributions via the quantity Amu mu delta Gamma s, which encodes the relative contributions of the B0 s light and heavy mass eigenstates to the decay.

  4. Enhanced Background Rejection in Signal Extraction: The system can improve signal purity by using a Boosted Decision Tree classifier trained on simulated B0 s to mu plus mu minus decays as signal and inclusive b bar b to X mu plus mu minus processes as background, exploiting features like the responses of two dedicated isolation classifiers trained using tracks from the underlying event that do not form the signal candidate.

  5. Uncertainty Quantification for Lifetime Measurements: The system can quantify systematic uncertainties in lifetime measurements by accounting for factors such as the effects of residual background contamination from unmodelled semileptonic decays (0.006 ps), a nonzero production asymmetry on the decay-time distribution (0.002 ps), and the decay-time acceptance on the relative admixture of light and heavy eigenstates as a function of Amu mu delta Gamma s (0.003 ps).

Abstract

The first single-experiment evidence for the rare decay B 0 to μ+ μ- and the world's most precise measurements of the B 0 s to μ+ μ- branching fraction and effective lifetime are reported. The results are obtained using Run 3 proton-proton collision data corresponding to an integrated luminosity of 18 fb-1 recorded with the LHCb Upgrade I detector at a centre-of-mass energy of 13.6 TeV between 2024 and 2025. Their combination with the Run 1 and Run 2 LHCb results further improves the precision. All measurements are found to be in agreement with the Standard Model expectations.

Sources

Related papers