⚛️ phenomenology

Lévy structure of the forward fixed-coupling BFKL kernel and a fixed-order obstruction to positivity in the symmetric scheme

This paper establishes a Lévy jump-process interpretation for the leading-order fixed-coupling BFKL kernel but demonstrates that fixed-order next-to-leading corrections in the symmetric scheme inherently violate positivity due to a dominant negative cubic collinear pole, a property that persists across various resummation attempts and remains unresolved for a probabilistic interpretation.

Original authors: Alex Prygarin, Claudelle Capasia Madjuogang Sandeu, Karam Shekh Yusuf

Published 2026-09-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Alex Prygarin, Claudelle Capasia Madjuogang Sandeu, Karam Shekh Yusuf

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the subatomic world, protons and other particles are not solid spheres but swirling clouds of gluons, the carriers of the strong nuclear force. When these particles collide at nearly the speed of light, the gluons inside them multiply rapidly, creating a dense, chaotic environment. Physicists study this behavior by looking at how the particles' transverse momentum—their motion sideways relative to the collision direction—spreads out as the energy of the collision increases. This spreading is often visualized as a random walk, where a particle takes a series of unpredictable steps, sometimes moving slightly and sometimes jumping far away. For decades, a specific mathematical framework known as the Balitsky-Fadin-Kuraev-Lipatov, or BFKL, equation has been the standard tool for describing this random walk. At its most basic level, this equation suggests that the gluons move in a way that can be described as a pure probability, where every possible step has a positive chance of occurring, much like a fair coin toss.

However, as scientists have pushed their calculations to higher levels of precision to match modern experimental data, a problem has emerged. When they include the next layer of complexity in their equations, the mathematical description begins to break down. The steps in the random walk no longer look like fair probabilities; instead, some steps appear to have negative chances, which is physically impossible. This creates a paradox: the theory predicts a behavior that cannot exist in reality. A new study by researchers at Ariel University in Israel has investigated this breakdown to understand exactly where and why the standard description fails, and whether it can be fixed.

The researchers began by confirming that the simplest version of the theory, known as the leading-order approximation, works perfectly. They showed that at this basic level, the movement of gluons is indeed a valid random walk. They mapped out the steps the gluons take on a mathematical cylinder that represents both the size of the momentum change and the direction of the turn. They found that this process is "pure jump," meaning the gluons do not drift smoothly but instead make discrete jumps. The rate at which these jumps occur is infinite for very small steps, but the total distance covered remains finite, creating a stable, predictable pattern. This confirmed that the foundation of the theory is sound and that the random walk picture is a real, physical description of how gluons behave when the calculation is kept simple.

The trouble appears when the scientists add the next level of detail, known as the next-to-leading order, which is necessary for high-precision predictions. In this more complex version, the researchers discovered a fundamental obstruction. They found that the mathematical rules governing the steps of the random walk produce a negative value for the probability of certain steps. This is not a minor error or a rounding issue; it is a structural flaw in the equation itself. Specifically, the equation contains a term that grows too quickly and becomes negative near the edges of the allowed energy range. This negative value means that the equation cannot describe a real probability law. No matter how the researchers tried to adjust the mathematical frame of reference or the way they defined the energy scale, this negative probability persisted. The study proves that the standard, fixed-order version of the equation, which is widely used in current physics, cannot represent a real physical process where probabilities are always positive.

The team also tested several advanced methods that physicists use to fix such problems, known as resummations. These methods attempt to rearrange the infinite series of calculations to smooth out the errors. The researchers examined a popular family of these fixes, which involves summing up all the problematic terms in a specific way. They found that even these sophisticated corrections failed to restore the positive probabilities. The mathematical function describing the steps still crossed into negative territory, creating a "negative share" of the total movement. While the amount of negative probability was small, its existence meant that the process could not be described as a valid random walk. The study showed that this failure happens at every positive coupling strength, which is a measure of how strongly the particles interact, meaning the problem is universal and not limited to a specific energy range.

Interestingly, the researchers did find that a different version of the theory, using a different way of organizing the energy scales, did produce a valid, positive probability law. This suggests that the problem is not with the physics of the gluons themselves, but with the specific mathematical recipe used to calculate them in the standard symmetric scheme. The obstruction is a property of the truncated, or cut-off, version of the equation. The study concludes that while the simple, leading-order picture of the gluon random walk is a real and valid process, the more precise, fixed-order version currently in use is broken. To move forward, physicists will need to find a new way to resum the equations that goes beyond simply matching the current calculations, one that can restore the positive probabilities required for a physical description of the universe.

Technical Summary: Lévy Structure of the Forward Fixed-Coupling BFKL Kernel and a Fixed-Order Obstruction to Positivity

Problem Statement
The paper addresses the probabilistic interpretation of the forward Balitsky-Fadin-Kuraev-Lipatov (BFKL) kernel in Quantum Chromodynamics (QCD) at fixed coupling. While the leading-order (LO) evolution of the gluon density is often visualized as a random walk in the logarithm of transverse momentum ( $\ln k^2$ ), a rigorous stochastic interpretation requires the existence of a non-negative step measure (a Lévy measure) that generates the evolution. The authors investigate whether such a probability law exists for the LO kernel and, crucially, whether it persists at fixed-order next-to-leading logarithmic (NLL) accuracy within the symmetric renormalization scheme. A positive step measure is essential for unweighted Monte Carlo sampling and for motivating the transition from linear to nonlinear evolution in saturation physics.

Methodology
The authors employ a combination of analytical derivation, complex analysis, and numerical computation:

Lévy-Khintchine Correspondence: They analyze the BFKL evolution equation by conjugating the amplitude and subtracting the growth factor (Pomeron intercept). This transforms the evolution into a probability transport problem on a cylinder of logarithmic transverse momentum and azimuth. They apply the Lévy-Khintchine theorem to determine if the characteristic exponent corresponds to a non-negative step measure.
Analytic Continuation and Inequalities: To test for positivity at NLL accuracy, the authors derive necessary conditions for a characteristic exponent to correspond to a non-negative measure. These include the "strip-edge inequality" (requiring the eigenvalue to be a minimum at the symmetric point along the real direction) and convexity conditions derived from moment identities.
Laurent Expansion and Pole Analysis: They examine the behavior of the NLL eigenvalue near the collinear edges ( $\gamma \to 0, 1$ ). Specifically, they isolate the cubic collinear pole in the NLL eigenvalue, whose coefficient is independent of the number of colors ( $N_c$ ) and flavors ( $n_f$ ).
Resummation Tests: The paper tests several existing resummation prescriptions (including all-poles approximations, matched forms, and collinearly improved kernels) to see if they restore positivity. This involves inverting the resummed exponents to obtain step densities and checking for sign changes.
Numerical Verification: For finite-rapidity Green functions, the authors use the de Hoog inversion algorithm to compute the evolution kernel at specific couplings and rapidities. They test for positive definiteness using Bochner's theorem (checking eigenvalues of Toeplitz matrices) and verify the location of singularities relative to the inversion contour.

Key Contributions and Results

Leading Order (LO) Interpretation: The authors establish a closed-form Lévy interpretation for the normalized forward BFKL kernel at LO. After conjugation and growth subtraction, the evolution is described by a single, non-negative Lévy measure on the cylinder. The process is pure jump with infinite activity but finite variation. The Pomeron intercept is identified as the relaxation rate of the first azimuthal harmonic. The Gaussian diffusion observed at large rapidity emerges from the accumulation of these jumps, not from a Gaussian component in the generator.
NLL Obstruction in the Symmetric Scheme: The paper proves that at fixed-order NLL accuracy in the symmetric scheme, no non-negative step measure exists for any positive coupling and rapidity.
- The obstruction arises from the cubic collinear pole in the NLL eigenvalue. The leading Laurent coefficient of this pole is $d_{13} = -1/2$ , which is independent of $N_c$ and $n_f$ .
- Near the collinear edge, this negative cubic term dominates the positive LO simple pole, causing the eigenvalue to fall below its value at the symmetric point. This violates the necessary condition that the symmetric point must be a minimum for a positive measure.
- This failure is intrinsic to the fixed-order truncation; a "control density" constructed from a positive exponent truncated at first order exhibits the same failure.
Failure of Resummations: The authors demonstrate that standard resummation prescriptions do not automatically restore positivity:
- All-poles approximations: The inverted density contains a Bessel function factor that oscillates and changes sign. The first zero occurs at a distance proportional to $1/\sqrt{\bar{\alpha}}$ . Even "matched" forms that include NLL corrections fail to remove these sign changes.
- Finite-rapidity Green functions: Numerical tests of improved kernels (e.g., from Colferai, Li, and Stasto) at finite rapidity show negative eigenvalues in the Toeplitz matrices, indicating a failure of positive definiteness.
Scheme Dependence: The obstruction is specific to the symmetric scheme. In a different rapidity-factorization scheme (the $(+)$ scheme), the cubic pole is removed, and a leading-logarithmic kernel is found to have a non-negative step measure at tested couplings. However, this is not a resummation of the symmetric NLL exponent, and the transformation between schemes is not a simple multiplicative conjugation.

Significance and Claims
The paper claims to resolve the question of whether the forward BFKL kernel admits a probabilistic interpretation at fixed order.

LO Validity: It confirms that the LO evolution is a genuine stochastic process (a Lévy process) without approximation, providing a rigorous foundation for the "random walk" picture of gluon evolution.
NLL Limitation: It establishes a rigorous obstruction to a probability law at fixed-order NLL accuracy in the symmetric scheme. The negative cubic collinear pole prevents the existence of a non-negative step measure.
Implications for Stochastic Methods: The authors conclude that an unweighted Monte Carlo simulation cannot sample the fixed-order NLL kernel because the step measure is signed. Consequently, any theoretical passage from linear to nonlinear evolution that relies on a stochastic interpretation of the fixed-order NLL kernel loses its justification, though the nonlinear equations themselves remain valid.
Open Questions: The paper does not prove that no resummation of the symmetric NLL kernel can restore positivity. It leaves open whether a probabilistic resummation exists that satisfies both perturbative matching and positivity. It also notes that the obstruction is a property of the fixed-order truncation, not necessarily of the physical NLL kernel as a whole.

The work emphasizes that the "random walk" is a precise mathematical reality at LO but becomes unavailable at fixed-order NLL in the symmetric scheme, highlighting positivity as a distinct requirement beyond perturbative agreement.

Technical Summary: Lévy Structure of the Forward Fixed-Coupling BFKL Kernel and a Fixed-Order Obstruction to Positivity

More like this