Test of lepton flavor universality with measurements of $R(D^{+})$ and $R(D^{*+})$ using semileptonic $B$ tagging at the Belle II experiment

Using a 365 fb⁻¹ data sample from the Belle II experiment, the authors measured the lepton flavor universality ratios $R(D^+)$ and $R(D^{*+})$ via semileptonic $B$ tagging, obtaining results consistent with world averages that collectively deviate from Standard Model predictions by 1.7 standard deviations.

The Gist

Imagine the universe as a massive, high-stakes cooking competition. In this competition, the "Standard Model" is the rulebook that says all chefs (particles) must follow the exact same recipe, regardless of who they are. Specifically, it says that the "flavor" of a chef shouldn't matter: a chef named "Electron" and a chef named "Muon" should be able to cook the exact same dish with the exact same efficiency. This is called Lepton Flavor Universality.

However, in recent years, some chefs have started whispering that the rulebook might be wrong. They suspect that the "Tau" chef (a heavier, more exotic cousin of the electron and muon) might be getting special treatment or, conversely, struggling more than the others.

This paper is a report from the Belle II experiment, a giant particle detector in Japan, acting as the ultimate food critic. They went to the "kitchen" (a particle collider called SuperKEKB) to taste-test a specific dish: B-mesons decaying into D-mesons and a lepton.

Here is the breakdown of their investigation in simple terms:

1. The Setup: Finding a Needle in a Haystack

The experiment creates millions of collisions, producing pairs of "B-mesons" (let's call them B-Pairs).

• The Challenge: When a B-meson decays, it's messy. It's like trying to identify a specific flavor of ice cream in a blizzard.
• The Trick (The "Tag"): To make sense of the chaos, the scientists use a clever trick. They look at the other B-meson in the pair (the "companion"). If they can identify what the companion did (the "Tag"), they know exactly what the partner (the "Signal") should have done.
• The Analogy: Imagine a dance hall where couples always enter together. If you see one partner leave the dance floor doing a specific move (the Tag), you can predict exactly what the other partner is doing, even if you can't see them clearly yet.

2. The Recipe: Two Types of Leptons

The scientists were looking for two specific outcomes in the "Signal" B-meson:

1. The Light Version: The B-meson decays into a D-meson and a light lepton (an Electron or a Muon).
2. The Heavy Version: The B-meson decays into a D-meson and a heavy Tau lepton.

The Tau is heavy and unstable. It almost immediately breaks down into a lighter lepton (electron or muon) plus some invisible neutrinos. This makes it hard to spot because it looks very similar to the "Light Version" decay, just with a few more invisible particles sneaking away.

3. The Detective Work: Sorting the Clues

To tell the difference between the "Light" and "Heavy" decays, the scientists used a Digital Detective (Machine Learning).

• They fed a computer five key clues about every event:
  1. The Angle: How the particles flew out (like the trajectory of a thrown ball).
  2. The Missing Energy: How much energy was "lost" to invisible neutrinos (like a bank account with missing funds).
  3. The Shape: How the event looked overall.
  4. Momentum: How fast the particles were moving.
• The computer learned to distinguish between the "Light" decays, the "Heavy" decays, and background noise (like random static on a radio).

4. The Results: Did the Tau Get Special Treatment?

The scientists calculated a ratio called R(D) and R(D)*.

• What it means: This is the ratio of "Heavy Tau dishes" to "Light Electron/Muon dishes."
• The Rulebook Prediction: According to the Standard Model, the Tau should be slightly less common because it's heavier, but the ratio should be very precise (around 0.30 for D* and 0.30 for D).
• The Measurement:
  • They found R(D) = 0.306* and R(D+) = 0.418.
  • These numbers are slightly higher than the rulebook predicts, but not by enough to declare a revolution.

5. The Verdict: "Consistent, but with a Hint of Mystery"

• The Significance: The results differ from the Standard Model prediction by 1.7 standard deviations.
• The Analogy: Imagine the rulebook says a coin should land on heads 50% of the time. You flip it 1,000 times and get 520 heads. That's interesting, but it's not enough to say the coin is rigged. You'd need to get 600 heads to be sure.
• The Conclusion: The Belle II team says, "Our results are consistent with the Standard Model, but they are also consistent with the 'World Average' of previous experiments which showed a slight tension."

Why Does This Matter?

If the ratio were significantly higher than expected, it would mean the Standard Model is broken. It would imply that a new, undiscovered force or particle is helping the Tau lepton cheat the rules. This would be the biggest discovery in physics in decades, potentially opening the door to "New Physics."

In summary:
The Belle II team acted like super-precise food critics in a chaotic kitchen. They used a clever "tagging" trick and a smart computer to count how often the heavy Tau lepton appears compared to the lighter ones. They found that the Tau is appearing slightly more often than the rulebook predicts, but not enough to prove the rulebook is wrong yet. The mystery remains, and the search for the "New Physics" continues!

Technical Summary

1. Problem and Motivation

The Standard Model (SM) of particle physics posits Lepton Flavor Universality (LFU), the principle that the electroweak gauge couplings are identical for all three generations of leptons ($e, \mu, \tau$), differing only by mass effects. However, previous measurements of the ratios $R(D^{(*)})$ by experiments like BaBar, Belle, and LHCb have shown tensions with SM predictions, suggesting potential LFU violation (LFUV) and physics beyond the Standard Model (BSM).

The specific observables are defined as:
$$R(D^{(*)}) = \frac{\mathcal{B}(B^0 \to D^{(*)+} \tau^- \bar{\nu}_\tau)}{\mathcal{B}(B^0 \to D^{(*)+} \ell^- \bar{\nu}_\ell)}$$
where $\ell$ represents an electron ($e$) or muon ($\mu$).

The Challenge: Measuring these ratios is experimentally difficult because the signal ($B \to D^{(*)} \tau \nu$) and the normalization channel ($B \to D^{(*)} \ell \nu$) share the same visible final state ($D^{(*)} \ell \nu$) when the $\tau$ decays leptonically ($\tau \to \ell \nu \nu$). The primary difference is the number of neutrinos (3 in the signal vs. 1 in the normalization), which affects kinematic distributions but makes event separation challenging. Previous Belle II analyses relied on hadronic $B$-tagging, which suffers from lower efficiency. This paper introduces a new approach using semileptonic $B$-tagging to increase the dataset and improve precision.

2. Methodology

Dataset and Detector

• Experiment: Belle II at the SuperKEKB $e^+e^-$ collider.
• Data: An integrated luminosity of 365 fb$^{-1}$ collected between 2019 and 2022 at the $\Upsilon(4S)$ resonance (10.58 GeV), corresponding to approximately $(387 \pm 6) \times 10^6$ $B\bar{B}$ events.
• Background Control: 42.3 fb$^{-1}$ of off-resonance data used to model $q\bar{q}$ continuum backgrounds.

Event Reconstruction Strategy

The analysis employs a "tag-and-signal" approach where the $\Upsilon(4S)$ decays into a $B^0\bar{B}^0$ pair.

1. $B_{tag}$ Reconstruction: The companion $B$ meson is reconstructed in semileptonic decay modes ($B^0 \to D^{(*)} \ell \nu$) using the Full Event Interpretation (FEI) algorithm. This hierarchical approach uses multivariate classifiers to identify $D$ mesons and leptons.
   • Constraint: The lepton momentum in the center-of-mass frame must exceed 1 GeV to suppress signal-side semitauonic decays in the tag.
   • Kinematic Cut: $\cos \theta_{BY}$ (angle between $B$ momentum and visible decay products) is restricted to $[-1.75, 1.1]$ to ensure high purity.
2. $B_{sig}$ Reconstruction: The signal $B$ meson is reconstructed in the final states $D^+ \ell^-$ and $D^{*+} \ell^-$ (where $\ell = e, \mu$).
   • The signal-side lepton must have a charge opposite to the tag-side lepton.
   • Strict particle identification (PID) criteria are applied for electrons (ECL energy/momentum ratio) and muons (KLM penetration).
   • $D$ and $D^*$ candidates are reconstructed via various hadronic decay modes with tight mass window constraints.

Signal Extraction via Multivariate Classification

To distinguish between the signal ($B \to D^{(*)} \tau \nu$), the normalization channel ($B \to D^{(*)} \ell \nu$), and backgrounds (including $B \to D^{**} \ell \nu$ and continuum), a multiclass Gradient Boosted Decision Tree (BDT) is trained.

• Input Variables:
  1. $\cos \theta_{BY}^{sig}$: Kinematic variable sensitive to the number of missing neutrinos.
  2. $E_{extra}$: Unassigned energy in the calorimeter (distinguishes signal from background).
  3. $\cos^2 \Phi_B$: A variable combining kinematics of both tag and signal sides.
  4. $p^*_D$: Momentum of the $D$ meson in the center-of-mass frame.
  5. $p^*_\ell$: Momentum of the lepton in the center-of-mass frame.
• Classification: The BDT outputs scores ($z_\tau, z_\ell, z_{bkg}$) for semitauonic, semileptonic, and background events.

Fitting Procedure

A binned two-dimensional maximum likelihood fit is performed on the variables $z_\tau$ and $z_{diff} = z_\ell - z_{bkg}$.

• The fit simultaneously determines the yields of semitauonic and semileptonic signals across four categories: $D^+e^-$, $D^+\mu^-$, $D^{*+}e^-$, and $D^{*+}\mu^-$.
• Systematic uncertainties are incorporated as nuisance parameters with Gaussian or Poisson constraints directly into the likelihood function.
• The ratio $R(D^{(*)})$ is derived from the fitted yields and efficiency corrections.

3. Key Contributions

• First Semileptonic Tagging Measurement at Belle II: This is the first measurement of $R(D^{(*)})$ by Belle II using semileptonic $B$-tagging, providing an orthogonal dataset to their previous hadronic-tagging results. This allows for cross-validation and reduced systematic correlations.
• Advanced Multivariate Analysis: The implementation of a multiclass BDT using five kinematic variables to separate signal, normalization, and complex backgrounds ($D^{**}$, gap modes) in a single fit framework.
• Comprehensive Systematic Treatment: A detailed evaluation of uncertainties, particularly the modeling of "semileptonic gap" processes (non-resonant $D^{**}$ contributions) and lepton identification efficiencies, which are the dominant systematic sources.

4. Results

The measured ratios are:

• $R(D^+) = 0.418^{+0.075}_{-0.073} (\text{stat}) ^{+0.049}_{-0.056} (\text{syst})$
• $R(D^{*+}) = 0.306^{+0.035}_{-0.033} (\text{stat}) ^{+0.016}_{-0.018} (\text{syst})$

Comparison with Standard Model:

• SM Predictions: $R(D) = 0.296 \pm 0.004$ and $R(D^*) = 0.254 \pm 0.005$.
• The measured values are consistent with the SM within 1.7 standard deviations when accounting for the correlation ($\rho = -0.24$) between the two measurements.
• The results are also consistent with the current World Average (HFLAV) within 0.6 standard deviations.

LFU Tests (Electron vs. Muon):
The analysis tested LFU within the semileptonic channels:

• $R(D^+)_{e/\mu} = 1.068 \pm 0.050 (\text{stat}) \pm 0.024 (\text{syst})$
• $R(D^{*+})_{e/\mu} = 1.079 \pm 0.037 (\text{stat}) \pm 0.020 (\text{syst})$
  Both are consistent with the SM expectation of unity (1.0) within $\sim 1.2\text{--}1.6\sigma$.

Uncertainties:

• Statistical uncertainties dominate the total error budget (approx. 17-18% for $R(D^+)$ and 11% for $R(D^{*+})$).
• The largest systematic uncertainty arises from the finite size of the Monte Carlo simulation samples used for efficiency determination and template shapes.
• The second largest systematic comes from the modeling of semileptonic gap processes ($B \to D^{**} \ell \nu$).

5. Significance

This paper represents a significant step in the global effort to resolve the $R(D^{(*)})$ anomaly.

1. Validation: By using a completely different tagging strategy (semileptonic vs. hadronic), the Belle II collaboration validates that their previous results were not driven by specific reconstruction biases.
2. Precision: While the current results do not confirm the LFU violation hinted at by older world averages, they provide a crucial independent measurement with competitive precision.
3. Future Outlook: The results highlight that statistical uncertainty is currently the limiting factor. As Belle II accumulates more data (targeting 50 ab$^{-1}$), the statistical precision will improve, potentially clarifying whether the tension with the SM is a statistical fluctuation or a sign of new physics. The current results, being consistent with the SM, suggest that if new physics exists, its effects may be smaller than previously hypothesized or require more precise measurements to detect.