Conservation laws and effective hadronization models

1. Problem Statement

Hadronization—the process by which asymptotically free quarks and gluons form bound hadronic states—is a major source of theoretical uncertainty in particle physics. Current Monte Carlo event generators (e.g., Pythia, Herwig) rely on phenomenological models (string or cluster fragmentation) that are typically Markovian (memoryless) by construction.

• The Tension: While the microscopic string-breaking dynamics are local and independent, physical hadronization must satisfy global constraints (exact energy-momentum conservation, flavor conservation) that couple the entire fragmentation trajectory.
• Current Limitations: Existing implementations resolve this tension via ad-hoc "endpoint corrections" (forcing the final remnant to decay) or rejection sampling. These methods either sacrifice theoretical rigor for efficiency or incur prohibitive computational costs for high-multiplicity events. There is no systematically improvable framework connecting microscopic dynamics to macroscopic observables while respecting global conservation laws.

2. Methodology

The author recasts hadronization as a conditioned stochastic diffusion process. The core mathematical tool is the Doob $h$-transform, which allows global constraints to be absorbed into a local renormalization of the dynamics.

A. Stochastic Framework & Doob Transform

• Bare Dynamics: The unconditioned string fragmentation is modeled as a Markov chain governing the evolution of the remaining string mass $M$. The transition kernel $K_{bare}$ is local and factorized.
• Conditioning: Physical hadronization requires the trajectory to terminate in a "success band" $S = [M_\star, M_{cut}]$ rather than failing by undershooting into $F = (0, M_\star)$.
• The Transform: By conditioning the bare process on the global constraint $C$ (successful termination), the transition kernel is renormalized:
  $$K_h(x \to y) = K_{bare}(x \to y) \frac{h(y)}{h(x)}$$
  where $h(x) = P_{bare}(C | X_t = x)$ is the success probability (the probability that a cascade starting at mass $x$ terminates successfully).
• Emergent Force: In the continuum limit, this reweighting induces an additional drift term (an "emergent force") proportional to $\sigma^2(M) \partial_M \ln h(M)$, biasing the evolution toward viable trajectories without breaking the Markovian property.

B. Effective Field Theory (EFT) Tower

The paper constructs a tower of effective theories organized by the remaining string mass $M$, exploiting the chiral limit ($m_\pi \to 0$) where the Lund kernel exhibits exact longitudinal-transverse factorization.

1. UV Regime ($M \gg M_{cut}$): The dynamics are scale-invariant. Transverse phase space effects are suppressed, and transport coefficients (drift $\mu$, diffusion $\sigma^2$) are constant (fixed point).
2. Intermediate Running Regime: Transverse phase space induces controlled scale dependence (running) in the Wilson coefficients.
3. IR/Boundary Regime ($M \sim M_{cut}$): Local diffusion competes with rare, order-one jumps. The continuum approximation breaks down, requiring non-local tail operators to describe direct transitions across the termination boundary.

C. Global Observables

The success probability $h(M)$ is computed by solving an integral backward equation (or a differential equation with non-local source terms). This function serves as the matching quantity connecting the different EFT regimes and acts as the renormalization factor for the dynamics.

3. Key Contributions

• Mathematical Resolution: Provides a precise mathematical resolution to the conflict between local Markovian dynamics and global conservation laws using the Doob $h$-transform.
• Wilsonian Structure: Demonstrates that hadronization admits a genuine Wilsonian EFT structure, complete with $\beta$-functions, anomalous dimensions, and systematic matching conditions.
• Factorization: Achieves a clean factorization of universal microscopic fragmentation dynamics (UV) from infrared constraint effects (IR).
• Tail Operators: Introduces explicit non-local operators to handle rare fragmentation events (large jumps) that cannot be captured by local diffusion approximations, essential for accurate global observables.
• Hybrid Sampling Scheme: Proposes a computationally efficient algorithm that applies the $h$-transform only in the IR boundary layer where it is dynamically active, eliminating the extensive rejection overhead of naive sampling.

4. Key Results

• Renormalized Dynamics: The conditioned dynamics are exactly described by a local generator with a modified drift:
  $$\mu_h(M) = \mu(M) + \sigma^2(M) \partial_M \ln h(M)$$
  This "budget awareness" force slows the cascade when it risks undershooting the success band and accelerates it when a successful termination is likely.
• Success Probability ($h(M)$):
  • In the UV, $h(M)$ approaches a constant plateau ($h_\infty$).
  • In the IR boundary layer, $h(M)$ exhibits a rapid exponential approach to the plateau, governed by the competition between diffusion and tail operators.
  • The paper provides analytic approximations and exact numerical solutions (via matrix inversion of the integral equation) for $h(M)$.
• Computational Efficiency:
  • Bare Sampling: Cost scales as $L/h_\infty$ (where $L$ is trajectory length), becoming inefficient as $h_\infty \to 0$.
  • Fully Conditioned: Same asymptotic cost as bare sampling due to rejection overhead in the UV.
  • Hybrid Scheme: By switching to the $h$-transform only near $M_{cut}$, the cost reduces to $L + O(1)$. This yields a speedup of $\sim 1/h_\infty$ (up to $6\times$ in tested scenarios) without sacrificing accuracy.
• Validation: Numerical comparisons show the hybrid scheme and EFT approximations reproduce exact Monte Carlo results for hadron multiplicity and success rates with sub-percent to few-percent accuracy.

5. Significance and Outlook

• Theoretical Rigor: Moves hadronization modeling from heuristic tuning to a systematic expansion scheme with controlled power counting and renormalization group flow.
• Practical Application: The hybrid sampling scheme offers a path to implement exact conservation laws in production Monte Carlo generators (like Pythia) without the computational penalty of current rejection methods.
• Future Directions:
  • Machine Learning: Enables ML models to learn the bare UV kernel directly, separating microscopic dynamics from IR constraints.
  • Neutrino Physics: Provides a systematic framework for low-energy hadronization ($W \sim \text{few GeV}$), critical for oscillation experiments where current generators rely on uncontrolled approximations.
  • Extensions: The formalism can be extended to include finite hadron masses, gluonic strings, and transverse recoil dynamics (2D conditioned processes).

In summary, this work establishes that global conservation laws do not require abandoning local stochastic dynamics; instead, they induce a calculable, local renormalization (an emergent force) that can be systematically integrated into an Effective Field Theory framework, offering both theoretical clarity and practical computational advantages.