On the simulated kinematic distributions of semileptonic $B$ decays

This paper identifies unphysical features in the kinematic distributions of semileptonic $B$ decays involving resonances caused by neglected phase-space factors in the EvtGen Monte-Carlo generator's sampling algorithm and proposes a reweighting method to correct these affected simulated samples.

The Gist

Imagine you are a chef trying to bake the perfect cake (a subatomic particle decay) for a very picky judge (a particle physics experiment). To get the recipe right, you need to know exactly how much of each ingredient (energy, mass, angles) goes into the mix.

In the world of high-energy physics, scientists use computer programs called Monte Carlo event generators to simulate these "cakes" before they even bake them in real life. One of the most popular "chefs" in this kitchen is a program called EvtGen. It's the go-to tool for simulating how heavy particles (like B-mesons) break apart into smaller pieces.

However, this paper by Florian Herren and Raynette van Tonder discovers that EvtGen has been using a broken measuring cup.

The Broken Measuring Cup: What Went Wrong?

When a heavy particle decays into a resonance (a short-lived, wobbly particle that quickly falls apart again), the computer needs to decide how much "room" (phase space) is available for the pieces to move around.

Think of it like this:

• The Real World: If you have a big, floppy balloon (a broad resonance) and you let it go, it can stretch and move in many ways, but it's still constrained by the size of the room.
• The EvtGen Mistake: EvtGen was acting like it forgot the walls of the room existed for certain types of balloons. It was generating "balloons" that were too stretched out or had weird, impossible shapes.

Specifically, the program was ignoring a mathematical rule called a "phase-space factor." In everyday terms, this is like ignoring the fact that you can't fit a 10-foot ladder through a 3-foot door. Because EvtGen ignored this, it was creating fake data where particles had impossible amounts of energy or mass.

The Symptoms: A Cake with a Weird Texture

Because of this broken measuring cup, the simulated data looks "off" in two main ways:

1. The "Long Tail" Effect: For broad, wobbly resonances (like the $D^*_0$ or $D'_1$ particles), the simulation creates a "tail" of events that shouldn't be there. Imagine a bell curve (the normal shape of a distribution) that suddenly has a long, flat tail stretching out to the side. This makes it look like there are more high-energy particles than there actually are.
2. The "Step" in the Floor: At the very edge of what is physically possible, the data doesn't fade away smoothly like a sunset. Instead, it hits a wall and stops abruptly, like a staircase. This is physically impossible in nature.

Why Should We Care? (The Impact)

You might think, "So the computer made a small mistake in the simulation. Who cares?"

Well, imagine you are trying to measure the exact weight of a gold bar (a physical constant) by weighing it against a known standard. If your scale is slightly off, your measurement of the gold bar is wrong.

In physics, these simulations are used to:

• Find new physics: If the simulation is wrong, scientists might think they found a new particle when they actually just found a bug in the code.
• Measure the Standard Model: Experiments at places like LHCb, Belle II, and ATLAS rely on these simulations to calculate things like $R(D^*)$ (a ratio that tests if the universe treats different types of matter equally). If the simulation is wrong, the ratio is wrong, and we might miss a clue about why the universe exists.

The authors show that for certain decays involving heavy particles, the error can be huge—sometimes changing the results by 50% or even 100%.

The Quick Fix: The "Reweighting" Spell

The authors know that fixing the core code of EvtGen will take a long time (like rewriting the entire recipe book). But they don't want to wait.

So, they offer a short-term patch called reweighting.

Think of it like this:

1. The computer generates a batch of "bad cakes" using the broken measuring cup.
2. The scientists look at each cake and say, "This piece of cake is too big; let's pretend it's 20% smaller." Or, "This piece is too small; let's pretend it's 50% bigger."
3. They assign a weight to every single simulated event. If an event looks like it came from the "broken" part of the distribution, they give it a low weight (make it count less). If it looks right, they give it a high weight.

By applying these weights, the "bad" data is mathematically reshaped to look like the "good" data. It's like using a photo filter to fix a blurry picture.

The Bottom Line

This paper is a warning label for the physics community. It says: "Hey, the tool you've been using for decades has a hidden flaw that makes the data look weird for specific types of particle decays."

They have provided a temporary "filter" (reweighting) to fix the data for current experiments, but they urge everyone to eventually upgrade the "measuring cup" (the sampling algorithm) so that future experiments don't have to rely on a patch.

In short: The universe is playing by strict rules, and for a while, our computer simulations were breaking those rules. Now we know how to fix the simulation so it matches reality again.

Technical Summary

1. Problem Statement

The paper identifies a critical flaw in EvtGen, a specialized Monte Carlo (MC) event generator widely used for simulating heavy hadron decays in experiments like Belle II, LHCb, and ATLAS/CMS.

• The Core Issue: EvtGen's phase-space sampling algorithm for decays involving resonances (specifically semileptonic $B \to R \ell \nu$ and nonleptonic decays into two resonances) neglects essential phase-space factors and incorrectly handles the factorization of kinematic variables.
• The Mechanism of Failure:
  • In standard rejection sampling, the algorithm should generate kinematic variables (like invariant mass $M$ and momentum transfer $q^2$) according to the full differential decay rate.
  • EvtGen optimizes the generation of the resonance invariant mass $M$ based on a simplified probability density function (PDF) that ignores the momentum dependence of the primary decay vertex ($|\vec{p}_R|$).
  • Crucially, if the generation of the remaining kinematics ($q^2$, angles) fails for a fixed $M$, EvtGen does not restart the sampling of $M$. Instead, after a fixed number of failed attempts (10,000), it accepts the last generated point.
  • Consequence: This results in a sampling distribution that lacks the necessary phase-space suppression factors (powers of $|\vec{p}_R|$). This leads to unphysical features, such as non-vanishing distributions at kinematic boundaries, artificial "steps," and significantly enhanced tails in the hadronic invariant mass spectrum.

2. Methodology

The authors employed a comparative analysis between the actual EvtGen implementation and a theoretically correct sampling method derived from first principles.

• Theoretical Derivation:
  • They derived the correct normalized differential decay rate for semileptonic decays ($B \to R \ell \nu \to P_1 P_2 \ell \nu$) and nonleptonic decays ($B \to R_1 R_2$).
  • They explicitly formulated the PDF used by EvtGen (Eq. 8) versus the correct PDF (Eq. 2), highlighting the missing Jacobian and phase-space factors in the EvtGen treatment.
• Simulation and Comparison:
  • They generated event samples using EvtGen for various decay modes, including broad resonances ($D_0^*(2300)$, $D_1'(2430)$) and narrow resonances ($D_1(2420)$, $D_2^*(2460)$).
  • They generated "ground truth" samples by sampling directly from the correct analytical PDFs (Eq. 2 and Eq. 13) using the same form factor models (ISGW2, LLSW) as EvtGen.
  • They compared kinematic distributions ($M^2$, $q^2$, lepton energy, recoil $w$) between the two methods.
• Correction Strategy:
  • They developed a reweighting method to correct existing MC samples without re-running the full simulation.
  • They derived event-wise correction weights $w(\vec{v})$ defined as the ratio of the correct differential rate to the EvtGen differential rate (Eq. 17).

3. Key Contributions

1. Identification of a Systematic Flaw: The paper demonstrates that the EvtGen algorithm is fundamentally flawed for resonant decays, not just a minor bug, affecting both broad and narrow resonances.
2. Quantification of Distortions:
   • Broad Resonances: The distortion is severe. The hadronic invariant mass spectrum exhibits a long, unphysical tail and a step at the kinematic boundary ($M = M_B - m_\ell$). This causes a >50% difference in $q^2$ and recoil distributions.
   • Narrow Resonances: While less affected due to the narrow width, the algorithm still introduces a cutoff at $M_R \pm 15\Gamma_R$, leading to ~2-5% distortions in $q^2$ spectra.
   • Tau Leptons: The effects are amplified for $B \to R \tau \nu$ decays due to the reduced phase space, with differences exceeding 100% in certain $q^2$ regions.
3. Impact on Nonleptonic Decays: The authors extend the analysis to nonleptonic decays into two resonances (e.g., $D^0 \to \bar{K}^* \rho$), showing that EvtGen creates "fake" phase-space boundaries, leading to unphysical steps in invariant mass spectra.
4. Practical Solution: A concrete, short-term reweighting formula is provided to correct existing datasets, allowing experimental collaborations to mitigate these biases immediately.

4. Results

• Kinematic Distortions:
  • $M^2$ Spectrum: EvtGen fails to suppress events near the kinematic limit, creating a "step" where the distribution should vanish smoothly. For broad resonances like $D_1'(2430)$, the peak of the distribution shifts, and the tail is significantly enhanced.
  • $q^2$ Spectrum: Due to the correlation between $M^2$ and $q^2$, the distortion in $M^2$ causes an excess of events at low $q^2$ and a deficit at high $q^2$.
  • Moments: The distortion significantly alters the moments of the hadronic mass spectrum. For $B \to D_1' \mu \nu$, the second and third moments are calculated to be 4x and 11x larger than expected, respectively.
• Impact on Experimental Measurements:
  • $R(X)$ and $R(D^{**})$: The mismodeling of the $M_X$ spectrum (hadronic mass) and $q^2$ spectrum directly impacts measurements of lepton universality ratios ($R(X)$) and exclusive $B \to D^{**} \tau \nu$ branching fractions. The authors suggest that the "deficit at low $q^2$" and "enhancement at high $M_X$" observed in recent Belle II and LHCb analyses could be partially attributed to this EvtGen artifact.
  • Systematic Uncertainties: The shifts in kinematic moments exceed the total systematic uncertainties of previous Belle and BaBar measurements and are comparable to the precision expected from Belle II.
• Reweighting Performance:
  • The proposed reweighting successfully restores the correct kinematic shapes.
  • However, for nonleptonic decays, the weights can become singular near the fake phase-space boundaries, leading to large statistical fluctuations in the reweighted samples.

5. Significance

• Immediate Impact on Precision Physics: The findings challenge the reliability of current MC samples used in precision flavor physics at Belle II, LHCb, and ATLAS/CMS. Measurements of inclusive semileptonic branching fractions, lepton universality tests ($R(X)$), and searches for new physics in $B \to D^{**} \tau \nu$ may be biased.
• Call to Action: The paper urges experimental collaborations to:
  1. Apply the provided reweighting corrections to existing data immediately.
  2. Cross-check results using alternative general-purpose generators (e.g., Sherpa, Herwig) which may handle phase space differently.
  3. Prioritize the implementation of a new, correct phase-space sampling algorithm in the core EvtGen codebase.
• Broader Context: This work highlights the necessity of rigorous validation of MC generators, especially as experiments move toward higher precision where subtle algorithmic artifacts can mimic or obscure new physics signals. It also points out that similar issues likely exist in decays involving sub-threshold resonances or resonance-to-resonance transitions.