Joint LHCb--Belle II Prospects to Constrain New Physics in $B\to D^{(*)}\tau\nu$

This paper presents a first sensitivity study demonstrating that a simultaneous joint fit of LHCb and Belle II data for $B\to D^{(*)}\tau\nu$ decays, utilizing event-by-event reweighting to handle non-SM biases and shared form-factor parameters, significantly reduces model-induced biases and improves constraints on New Physics Wilson coefficients compared to independent analyses.

The Gist

Imagine you are trying to solve a giant, cosmic jigsaw puzzle. The picture on the box is the Standard Model, our current best theory of how the universe works. But for a few years, scientists have noticed that some pieces—specifically those involving heavy particles called B-mesons decaying into tau particles—just don't seem to fit the picture perfectly. They are wobbling, suggesting there might be a hidden piece of the puzzle we haven't found yet: New Physics.

This paper is a blueprint for how two giant particle detectives, LHCb (at CERN in Europe) and Belle II (in Japan), can work together to find that hidden piece.

Here is the breakdown of their strategy, using some everyday analogies:

1. The Problem: Two Different Maps, One Destination

Currently, LHCb and Belle II are like two different cartographers drawing maps of the same territory.

• LHCb is like a high-speed highway camera; it sees millions of cars (particles) zooming by, but the view is a bit blurry and crowded.
• Belle II is like a pristine, quiet park; it sees fewer cars, but the view is crystal clear and the environment is controlled.

Both are trying to measure the same thing: how often these B-mesons turn into tau particles. But here's the catch: to interpret their data, they both have to rely on a "theoretical map" (called form factors) that describes how the particles behave inside the atom.

The Trap: If LHCb and Belle II use slightly different versions of this theoretical map, or if they try to combine their results after they have already finished their individual analyses, they might get a distorted picture. It's like trying to merge two GPS routes where one assumes the roads are straight and the other assumes they are curved. The final route might look fine, but it could lead you off a cliff.

2. The Solution: The "Live Rewind" Button

The authors propose a new way to combine these experiments. Instead of finishing the puzzle separately and then taping the pictures together, they suggest building a single, giant puzzle from the start.

To do this, they invented a tool called REDIST. Think of this as a "Live Rewind" button for particle simulations.

• Usually, if scientists want to test a new theory (like "What if there is a hidden force?"), they have to run the entire computer simulation from scratch, which takes weeks.
• With REDIST, they can take a single simulation and instantly "rewind" and "replay" it with different rules. They can say, "Okay, let's pretend the hidden force is 10% stronger," and the computer instantly reshapes the data to match that new rule without re-running the whole engine.

This allows them to test thousands of "New Physics" scenarios instantly and see which one fits the data best.

3. The Strategy: A Unified Team Huddle

The paper compares two ways of combining the data:

• The Old Way (Post-Fit Average): Imagine LHCb and Belle II each solve their half of the puzzle, write down their best guess for the hidden piece, and then meet to average their answers.

  • The Flaw: If they used slightly different assumptions about the background noise (the "form factors"), their averages might cancel out the truth or create a fake signal. It's like two chefs tasting a soup separately, adjusting the salt based on their own spoons, and then mixing the soups. The final taste might be weird.

• The New Way (Simultaneous Fit): Imagine LHCb and Belle II sit at the same table with one giant bowl of soup. They agree on the exact recipe (the theoretical map) and taste the soup together, adjusting the salt simultaneously for both halves.

  • The Benefit: Because they are sharing the "nuisance parameters" (the background noise and theoretical maps) in real-time, they can cancel out errors much better. If LHCb sees a wobble that looks like a hidden force, but Belle II sees it's just a quirk of their specific camera, the joint model knows to ignore it.

4. The Result: Sharper Vision

By using this "Live Rewind" tool and the "Unified Team Huddle" strategy, the paper predicts that by the year 2030 or 2040, the combined power of LHCb and Belle II will be incredibly sharp.

• Sensitivity: They will be able to detect even tiny deviations from the Standard Model.
• Bias Reduction: They will avoid the "fake signals" that happen when theories don't match up perfectly between experiments.

The Bottom Line

This paper isn't just about crunching numbers; it's about changing the rules of the game. It argues that to find the next big discovery in physics, we can't just add more data; we have to change how we combine that data.

By building a single, flexible, and shared statistical model, LHCb and Belle II can stop guessing and start knowing. They are essentially building a super-microscope that doesn't just look at the particles, but looks at the relationship between the particles and the laws of physics, ensuring that when they finally find the "New Physics," they know it's real and not just an illusion caused by a bad map.

Technical Summary

1. Problem Statement

The paper addresses two critical challenges in interpreting semileptonic $b \to c \tau \bar{\nu}_\tau$ decays within the Weak Effective Theory (WET) framework, specifically regarding the persistent $3.8\sigma$ anomaly in $R(D^{(*)})$ ratios:

• Inconsistent Treatment of Shared Theory Inputs: Current combination strategies often rely on "post-fit averages" of results from different experiments (LHCb and Belle II). This approach fails to properly correlate shared hadronic inputs, specifically the form-factor parameters ($\vec{\alpha}$) that describe the $B \to D^{(*)}$ transition. Treating these as independent or partially correlated nuisance parameters obscures information propagation between datasets, leading to inconsistent constraints and potential biases.
• Model Dependence in Signal Extraction: Standard analyses generate Monte Carlo (MC) templates assuming Standard Model (SM) couplings. If New Physics (NP) exists, it alters kinematic distributions (e.g., $q^2$), acceptance, and resolution effects. Reinterpreting SM-derived templates using NP hypotheses introduces model-induced biases because the detector response and template shapes change coherently with the physics hypothesis.

2. Methodology

The authors propose a joint, likelihood-level combination strategy using a novel framework called REDIST.

• Theoretical Framework:

  • The analysis uses the Weak Effective Theory (WET) to parameterize NP effects via Wilson coefficients ($\vec{C} = C_{VL}, C_{VR}, C_{SL}, C_{SR}, C_T$).
  • Hadronic inputs are parameterized via form factors ($\vec{\alpha}$), using parameterizations like CLN, BGL, or BLPR(XP).

• The REDIST Framework:

  • Event-by-Event Reweighting: Instead of regenerating detector simulations for every NP hypothesis, the authors use the HAMMER library to reweight fully simulated SM events.
  • Likelihood Construction: The reweighting is embedded directly into a binned likelihood (implemented via PYHF and HISTFACTORY). The weight for an event $e$ is calculated as the ratio of differential rates:
    $$w_e(\theta) = \frac{d\Gamma/d\Omega(\theta; k_e)}{d\Gamma/d\Omega(\theta_0; k_e)}$$
    where $\theta = (\vec{C}, \vec{\alpha})$. This ensures that variations in Wilson coefficients and form factors propagate coherently through detector acceptance, resolution, and reconstruction.
  • Backend Agnosticism: The framework allows swapping theory engines (e.g., HAMMER, EOS) without altering the statistical model, facilitating robust cross-checks.

• Analysis Configurations:

  • LHCb-like: Simulated using RapidSim. Includes channels for $B^0 \to D^{(*)-}\tau^+\nu_\tau$ with $\tau \to \mu \bar{\nu}_\mu \nu_\tau$ and $\tau \to 3\pi \nu_\tau$. Observables include $m^2_{miss}$, $q^2$, and lepton energy.
  • Belle II-like: Simulated using EVTGEN with kinematic constraints from hadronic $B$-tagging. Includes $B \to D^{(*)}\tau\nu$ with leptonic and hadronic ($\tau \to \pi \nu$) $\tau$ decays. Observables include $m^2_{miss}$ and extra energy ($E_{extra}$).
  • Datasets: Asimov datasets are generated for 2030 and 2040 benchmark scenarios, reflecting projected experimental precisions.

• Comparison Strategies:

  1. Combined Fit: A single likelihood where Wilson coefficients and shared form-factor parameters are floated simultaneously across all channels and experiments.
  2. Post-fit Sum: A point-by-point summation of profiled likelihoods from individual channels (the traditional approach).

3. Key Contributions

• First Joint Sensitivity Study: This is the first study to perform a combined extraction of Wilson coefficients using realistic LHCb and Belle II analysis configurations within a unified EFT framework.
• REDIST Implementation: Introduction of a portable, reproducible statistical model that integrates theory reweighting directly into the likelihood minimization, solving the "regeneration bottleneck" for EFT scans.
• Demonstration of Bias: The paper provides concrete evidence that post-fit combinations can be biased if different channels use inconsistent hadronic parameterizations (e.g., mixing BLPR and BGL forms), leading to spurious shifts in fitted Wilson coefficients.

4. Results

The study focuses on three representative Wilson coefficients: $C_{SL}$ (scalar), $C_{VR}$ (right-handed vector), and $C_T$ (tensor).

• Complementarity of Channels:

  • $C_{SL}$: Primarily constrained by $B \to D$ modes.
  • $C_{VR}$: Enhanced sensitivity in $B \to D^*$ due to vector meson helicity structure.
  • $C_T$: Shows strong shape leverage in $B \to D^*$, creating orthogonal degeneracy directions compared to other operators.
  • The combined fit significantly tightens constraints by intersecting these complementary degeneracy directions.

• Impact of Shared Hadronic Inputs:

  • Treating form-factor parameters as shared nuisance parameters in a simultaneous fit reduces model-induced bias and improves sensitivity.
  • Channels with strong normalisation handles (e.g., $\tau \to \mu$) calibrate nuisance directions that would otherwise mimic NP effects in other channels.

• Bias in Post-Fit Combinations:

  • A toy example demonstrates that if different channels use different form-factor parameterizations (e.g., BLPR vs. BGL), the "post-fit sum" yields a displaced minimum compared to a unified likelihood. This confirms that independent profiling on different nuisance manifolds creates artificial tension.

• Projected Precision (2030/2040):

  • The 2040 benchmarks show significant contraction of confidence regions compared to 2030.
  • Conditional 1D Intervals (SM injection, 2040 Combined):
    • $Re(C_{VL})$: $\pm 0.007$
    • $Re(C_{SL})$: $[-0.05, 0.02]$
    • $Re(C_{VR})$: $\pm 0.007$
    • $Re(C_T)$: $[-0.002, 0.003]$
  • The combined fit consistently outperforms individual LHCb or Belle II projections, particularly for imaginary parts where CP-averaged observables usually suffer from phase-reflection ambiguities.

5. Significance and Outlook

• Necessity of Joint Fits: The paper concludes that a joint likelihood fit is not just advantageous but necessary for a well-defined EFT interpretation. It prevents the loss of information flow between channels and eliminates inconsistencies arising from disparate hadronic modeling.
• Future Strategy: The authors advocate for the publication of portable, morphable likelihoods where dominant shared hadronic parameters are explicit. This allows the community to perform consistent EFT constraints without relying on error-prone post-fit sums.
• Next Steps: Future work should include:
  • Separating hadronic $\tau$ modes (e.g., $\tau \to \pi$ vs. $\tau \to \rho$) to reduce degeneracies.
  • Incorporating CP-odd observables (e.g., CP asymmetries, triple-product correlations) to break the $Im(C) \to -Im(C)$ symmetry and constrain imaginary Wilson coefficients linearly.
  • Refining feed-down modeling ($B \to D^{**}$) as correlated nuisances.

In summary, this work provides a robust, scalable strategy for the next generation of flavor physics analyses, ensuring that increasing experimental precision translates into reliable constraints on New Physics.