⚛️ phenomenology

First look at the evaluation of two-loop Feynman integrals for radiative return processes

This paper presents the calculation of planar two-loop four-point Feynman integrals with full electron mass dependence for radiative return processes, overcoming analytical complexities involving nested square roots and elliptic geometries to provide stable numerical evaluations essential for next-to-next-to-leading order QED predictions.

Original authors: Mattia Pozzoli, William J. Torres Bobadilla

Published 2026-07-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Mattia Pozzoli, William J. Torres Bobadilla

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

Imagine you are trying to predict exactly how a billiard ball will bounce off a cushion, but this isn't a normal table. The balls are electrons, the cushions are other particles, and the physics is so complex that you need to calculate the path of every single ripple in the fabric of space-time that the collision creates.

This paper is a "first look" at a massive, two-year-long calculation (in terms of complexity, not time) required to make these predictions with extreme precision. Here is the breakdown of what the authors did, using simple analogies.

The Big Picture: The "Radiative Return"

Scientists at low-energy particle colliders (like those used to study the muon's magnetic properties) are trying to measure a process called "radiative return."

The Analogy: Imagine you are throwing a ball at a target. Sometimes, the ball hits the target and bounces back perfectly. Other times, the ball hits the target, but it also accidentally knocks off a piece of its own skin (a photon) on the way out. This "skin" is the radiation.
The Goal: To understand the target perfectly, scientists need to account for every possible way that "skin" could be knocked off. The authors are calculating the rules for the most complicated version of this event: where the collision happens twice (two loops) and involves the mass of the electron.

The Problem: A Maze of Math

To predict these events, physicists use "Feynman integrals." Think of these as a giant, multi-dimensional maze.

The Challenge: Usually, these mazes are like a grid in a city: you can walk straight, turn left, turn right, and eventually find the exit. But in this specific calculation, the maze has twisted tunnels (elliptic geometries) and nested traps (nested square roots).
The "Twisted Tunnels": These are shapes so complex they aren't just lines or circles; they are like the surface of a donut with extra holes. Standard math tools break down here.
The "Nested Traps": This is a square root inside another square root, making the math incredibly messy and prone to errors when you try to plug in numbers.

The Solution: Building a Better Map

Instead of trying to solve the maze by walking through it blindly, the authors built a new kind of map called Differential Equations.

The Strategy: Imagine you are lost in a forest. Instead of guessing which way to go, you build a system of signs (equations) that tell you exactly how the terrain changes as you move.
The Innovation: The authors didn't try to force the twisted tunnels (elliptic curves) into a simple grid. Instead, they created a map that acknowledges the twists but keeps the instructions simple. They made the instructions "polynomial," meaning they are made of simple building blocks (like $x^2 + 3x$ ) rather than impossible-to-calculate infinite series.
The "Nested Square Root" Issue: They found a way to navigate the "nested traps" without actually having to solve the trap itself inside the instructions. They kept the instructions clean (using only simple square roots) so a computer could follow them without getting confused.

The Result: A Stable Computer Program

The authors didn't just write the math on paper; they turned it into a computer program (written in the Julia language) that can actually calculate the answers.

The Test: They tested this program on 1,000 different scenarios (like throwing the ball at the target from 1,000 different angles).
The Outcome: The program worked! It successfully calculated the answers across the entire "physical region" (the range of real-world conditions).
The Catch: Because the math is so hard (especially the "twisted tunnels"), the computer sometimes gets a little fuzzy. In the easiest cases, it got 5 digits of precision (very accurate). In the hardest cases, it got 2 digits (good enough to know the ball is going roughly the right way, but not the exact millimeter).

Why This Matters (According to the Paper)

This paper doesn't claim to have solved the whole mystery of the universe or fixed the muon's magnetic moment yet. Instead, it claims to have built the essential building blocks.

The Metaphor: If the final goal is to build a skyscraper (a complete, ultra-precise prediction for particle physics), this paper is the team that successfully poured the concrete foundation and framed the first floor. They proved that the foundation is stable and that the materials (the math) can be handled by modern computers.
The Next Step: The authors state that the next step is to build the rest of the skyscraper (calculating the non-planar parts and assembling the full amplitude), which they are currently working on.

In summary: The authors took a mathematically terrifying, twisted maze of particle physics, built a simplified map for it, and proved that a computer can navigate that map successfully. This paves the way for much more accurate predictions of how particles behave in low-energy experiments.

Technical Summary: First Look at the Evaluation of Two-Loop Feynman Integrals for Radiative Return Processes

Problem Statement
Precision tests of the Standard Model at low-energy electron–positron colliders (e.g., BaBar, Belle II, BESIII, KLOE) rely on accurate theoretical predictions for radiative return processes, such as $e^+e^- \to \pi^+\pi^-\gamma$ and $e^+e^- \to \mu^+\mu^-\gamma$ . These processes are critical for determining the hadronic vacuum polarisation contribution to the muon anomalous magnetic moment ( $(g-2)_\mu$ ). To resolve persistent tensions between experimental measurements and theoretical predictions, Next-to-Next-to-Leading Order (NNLO) QED corrections are required.

The primary bottleneck in achieving NNLO accuracy for these processes, while retaining full dependence on the electron mass, is the evaluation of two-loop four-point Feynman integrals. The presence of massive propagators introduces significant algebraic complexity in Integration-By-Parts (IBP) reductions and Differential Equations (DEs). Furthermore, these integrals involve special functions associated with higher-genus geometries, specifically elliptic curves and, in some cases, nested square roots. Constructing a canonical basis for these integrals—where DEs are $\epsilon$ -factorised and expressed in terms of dlog one-forms—is mathematically challenging and often not strictly necessary for numerical phenomenology.

Methodology
The authors focus on the initial-state radiation (ISR) contribution to the process $e^+e^- \to \gamma\gamma^*$ , which reduces the problem to evaluating genuine four-point two-loop amplitudes. They consider three planar integral families, denoted PL1, PL2, and PL3, which contain a total of 212 master integrals (MIs), including those associated with five non-isomorphic elliptic curves and one sector with a nested square root.

Instead of attempting to construct a fully canonical basis for all sectors (which is difficult for elliptic and nested-root geometries), the authors adopt a strategy developed in previous works [48, 49]. Their approach involves:

Basis Construction: They construct a basis of MIs such that the resulting system of DEs is polynomial in the dimensional regulator $\epsilon$ . For polylogarithmic sectors, they aim for canonical forms. For elliptic sectors, they identify MIs associated with elliptic differentials of the first, second, and third kind.
Handling Elliptic Geometries: For sectors where the elliptic nature is not manifest in the corner integral (e.g., four-point kite sectors), they utilize reducible super-sectors and analyze Picard-Fuchs operators to identify the geometry and construct appropriate bases.
Handling Nested Square Roots: In a specific sector of PL1 involving a nested square root, they demonstrate that while a transformation exists to $\epsilon$ -factorise the DEs (involving a duplet of nested roots), they prefer to work with a basis that keeps the DEs linear in $\epsilon$ and avoids introducing complex transcendental structures, relying instead on rational functions and simple square roots.
Numerical Integration: The DEs are solved numerically. Boundary conditions are determined using the AMFlow package at a specific phase-space point chosen to mimic low-energy kinematic configurations. The integration is performed using an in-house implementation in the Julia programming language, employing non-stiff integrators (Tsit5 and Vern).

Key Contributions and Results

Systematic Framework: The paper presents the first systematic construction of DEs for the planar integral families relevant to NNLO radiative return processes with full electron mass dependence.
Elliptic Sector Treatment: The authors successfully identify and treat sectors associated with five distinct elliptic curves. They provide explicit bases for these sectors, including the challenging four-point kite topologies where elliptic behavior arises from integrals with dots (propagator powers $>1$ ) rather than corner integrals.
Nested Square Root Analysis: The paper details the treatment of a sector in PL1 containing a nested square root. It constructs a basis that satisfies DEs polynomial in $\epsilon$ without requiring the explicit introduction of the nested root structure in the differential equations, facilitating numerical stability.
Numerical Validation: The authors demonstrate stable numerical evaluations across the physical region relevant for low-energy experiments ( $s \in \{1.0404, 16, 100\} \text{ GeV}^2$ $s \in {1.0404, 16, 100} GeV^{2}$ ).
- Performance: Evaluation times range from 40 to 800 ms per phase-space point.
- Precision: Comparisons with AMFlow show agreement of at least eight significant figures at specific test points. Comparisons with the DiffExp package (using double precision) show approximately 5 digits of agreement for PL1 and 2 digits for the more complex PL2 and PL3 families in worst-case scenarios.
- Robustness: The method remains stable despite large hierarchies of scales ( $m_e^2/s \sim 10^{-7} - 10^{-8}$ ).

Significance
The paper claims to provide "essential building blocks for NNLO predictions for radiative return processes." By successfully constructing and numerically integrating the DEs for the planar sectors, the authors have overcome the primary computational hurdle for the initial-state radiation component of the NNLO amplitude.

The work is modest in its scope, explicitly stating that it covers only the planar families. The extension to non-planar families and the assembly of the complete scattering amplitude are noted as ongoing work. The significance lies in the demonstration that a polynomial-in- $\epsilon$ approach, combined with numerical integration, is a viable and efficient strategy for handling the complex geometries (elliptic curves and nested roots) inherent in massive two-loop calculations, bypassing the need for fully analytic canonical forms that are currently difficult to obtain for such geometries.