Canonical differential equations and intersection matrices

This paper elaborates on a method to derive polynomial relations between iterated integrals in multi-loop Feynman integrals by utilizing properties of the canonical intersection matrix to convert non-linear constraints into linear ones, illustrated through examples involving Calabi-Yau varieties and higher-genus Riemann surfaces.

The Gist

Imagine you are trying to solve a massive, multi-layered puzzle. In the world of particle physics, these puzzles are called Feynman integrals. They are the mathematical recipes physicists use to calculate how particles collide and interact. The more complex the collision (the more "loops" in the diagram), the harder the puzzle becomes.

For decades, physicists have used a powerful tool called Differential Equations to solve these puzzles. Think of a differential equation as a set of instructions: "If you change the speed of the car, how does the distance change?" By following these instructions, you can figure out the final answer.

However, there's a catch. Sometimes the instructions are written in a messy, complicated language that is incredibly hard to follow. Physicists wanted to find a "canonical" (perfectly clean and organized) way to write these instructions, where the math becomes simple and predictable.

This paper is about a new set of tools to help physicists find that perfect, clean way of writing the instructions, even for the most difficult puzzles.

Here is the breakdown of their discovery using simple analogies:

1. The Problem: The "Messy Kitchen"

Imagine you are a chef trying to bake a complex cake (the Feynman integral). You have a recipe, but it's written in a chaotic way. The ingredients are mixed up, the steps are out of order, and the instructions depend on a mysterious variable called $\epsilon$ (which is like a "tweak" in the recipe to make the math work).

Physicists have found a way to organize the recipe so that the "tweak" ($\epsilon$) is neatly separated from the rest. This is called an $\epsilon$-factorised basis. It's like having a recipe where you first list all the "tweaks" in a separate column, making the rest of the cooking much easier.

But here's the snag: For simple cakes, this works perfectly. For complex, multi-layered cakes (involving exotic shapes like Calabi-Yau varieties or higher-genus surfaces), the "tweaks" get tangled with new, weird ingredients that nobody knows how to handle yet. These new ingredients are called $\epsilon$-functions.

2. The Discovery: The "Magic Mirror"

The authors of this paper realized that these messy recipes have a hidden structure. They looked at the problem through the lens of Twisted Cohomology.

• The Analogy: Imagine you have a mirror. If you look at your reflection, you see yourself. But in this math world, there's a "twisted" mirror. If you look at the recipe in the mirror, it doesn't just show you a copy; it shows you a dual version of the recipe.
• The Intersection Matrix: When you compare the original recipe and the mirror image, you get a grid of numbers called an Intersection Matrix.
• The Big Reveal: The authors found that if you organize your recipe correctly (into a "canonical" basis), this grid of numbers becomes constant. It stops changing as you move through the puzzle. It's like finding that the blueprint of a building is the same no matter which floor you are on.

3. The Solution: The "Deconstruction Machine"

Knowing that the grid is constant is great, but it doesn't immediately tell you what the weird new ingredients ($\epsilon$-functions) are. The authors built a mathematical "deconstruction machine" to solve this.

They realized that the rotation needed to clean up the recipe (turning the messy one into the clean one) can be split into two parts:

1. The Symmetric Part (The Knowns): This part is made of ingredients we already know how to handle (like standard algebraic functions and periods).
2. The Orthogonal Part (The Unknowns): This part contains the truly new, mysterious ingredients ($\epsilon$-functions).

The Magic Trick:
Usually, figuring out how these ingredients mix is like solving a non-linear equation (a tangled knot that is very hard to untie). The authors proved that by splitting the recipe into these two parts, the "tangled knot" becomes a straight line.

• Instead of solving a complex, non-linear puzzle, they reduced it to a linear system (like simple addition and subtraction).
• This allows them to instantly identify which parts of the recipe are just standard ingredients and which parts are the new, mysterious ones.

4. The Results: Solving Real Puzzles

The team tested their new machine on several difficult "cakes":

• Calabi-Yau Varieties: These are complex geometric shapes that appear in string theory.
• Higher-Genus Surfaces: Imagine a donut with multiple holes (a pretzel).
• Four-Loop Banana Integral: A specific, very complex particle collision diagram.

In every case, their method successfully identified the "known" ingredients and isolated the "unknown" ones. In some cases, they found that the "unknown" ingredients were actually just combinations of known things, simplifying the problem significantly. In others, they confirmed that the new ingredients were genuinely new and couldn't be simplified further.

Summary

Think of this paper as a new organizing system for a chaotic library.

• Before: Books (integrals) were scattered everywhere, and finding the right one was a nightmare.
• The Old Way: You could organize some books, but the complex ones remained a mess.
• This Paper: The authors found a "magic key" (the constant intersection matrix) and a "sorting algorithm" (the decomposition into symmetric and orthogonal parts). This allows them to instantly separate the books that belong in the "Standard Section" from the ones that need a "Special Section."

This doesn't just make the math prettier; it makes it possible to calculate the behavior of particles in experiments that were previously too difficult to solve, bringing us closer to understanding the fundamental laws of the universe.

Technical Summary

1. Problem Statement

The evaluation of multi-loop Feynman integrals is a cornerstone of precision predictions in particle physics and gravitational wave astronomy. The standard approach involves:

1. IBP Reduction: Reducing a large set of integrals to a finite basis of "master integrals."
2. Differential Equations (DEs): Solving a system of DEs for these master integrals with respect to kinematic variables.

A major bottleneck arises when seeking a canonical basis (or $\varepsilon$-factorised basis). In such a basis, the DE system takes the form $d\vec{J} = \varepsilon \mathbf{A}(x) \vec{J}$, where the matrix $\mathbf{A}(x)$ depends only on kinematics $x$ and not on the dimensional regulator $\varepsilon$.

• The Challenge: While canonical bases involving only $d\log$-forms are well-understood (leading to Multiple Polylogarithms), many modern problems involve non-trivial geometries (elliptic curves, Calabi-Yau varieties, higher-genus Riemann surfaces). In these cases, the DEs involve $\varepsilon$-functions: new transcendental functions defined as iterated integrals over algebraic functions and periods of the underlying geometry.
• The Open Question: It is currently unclear which of these $\varepsilon$-functions are genuinely new (cannot be expressed in terms of known functions like periods and their derivatives) and which can be reduced to simpler forms. Furthermore, the constraints relating these functions are often highly non-linear and difficult to solve.

2. Methodology

The authors leverage Twisted Cohomology Theory, the mathematical framework describing Feynman integrals in dimensional regularisation. Their methodology rests on three pillars:

A. The Canonical Intersection Matrix

In a canonical basis, the cohomology intersection matrix $\mathbf{C}_c(x, \varepsilon)$ (which pairs basis elements with their duals) becomes remarkably simple: it is constant in $x$ and takes the form $\mathbf{\Delta} M(\varepsilon)$, where $\mathbf{\Delta}$ is a constant matrix independent of kinematics.

• The authors show that $\mathbf{\Delta}$ relates the differential equation matrices of the basis and its dual via a specific transposition operation defined by $\mathbf{\Delta}$.
• They propose a method to determine $\mathbf{\Delta}$ without explicitly computing the intersection matrix in the original basis, by solving linear systems derived from the DE structure.

B. Decomposition of the Rotation Matrix

The transformation to a canonical basis involves a rotation matrix $\mathbf{U}_t(x, \varepsilon)$. The authors focus on the $\varepsilon^0$ part of this matrix, denoted $\mathbf{U}^{(0)}_t$, whose entries are the $\varepsilon$-functions.

• They observe that $\mathbf{U}^{(0)}_t$ belongs to a specific group of unipotent lower-triangular matrices ($G_{par}$) preserving the mixed Hodge structure of the geometry.
• Key Theorem (Proposition 1): Any matrix $\mathbf{U} \in G_{par}$ can be uniquely decomposed as:
  $$\mathbf{U} = \mathbf{O} \mathbf{R}$$
  where:
  • $\mathbf{O}$ is $\Delta$-orthogonal ($\phi_\Delta(\mathbf{O}) = \mathbf{O}^{-1}$).
  • $\mathbf{R}$ is $\Delta$-symmetric ($\phi_\Delta(\mathbf{R}) = \mathbf{R}$).
    Here, $\phi_\Delta$ is a generalized transposition operation.

C. Linearization of Constraints

The authors prove that the entries of the $\Delta$-symmetric part ($\mathbf{R}$) are entirely determined by the known functions in the field $F_{ss}$ (algebraic functions and periods).

• Consequently, the non-linear polynomial constraints on the $\varepsilon$-functions (derived from the constancy of the intersection matrix) can be reduced to linear constraints for the entries of $\mathbf{R}$.
• The $\Delta$-orthogonal part ($\mathbf{O}$) contains the genuinely new $\varepsilon$-functions. The dimension of the group of such orthogonal matrices provides an upper bound on the number of independent new functions required.

3. Key Contributions

1. Decomposition Theorem: The proof that the rotation to a canonical basis can be split into a "known" symmetric part (solvable via linear algebra) and an "unknown" orthogonal part (containing the new transcendental functions).
2. Algorithmic Reduction: A method to convert the difficult problem of solving non-linear polynomial relations between $\varepsilon$-functions into a tractable linear algebra problem.
3. Analytic Determination of $\mathbf{\Delta}$: A novel technique to determine the canonical intersection matrix $\mathbf{\Delta}$ directly from the differential equation structure, bypassing the need to compute complex intersection numbers in the original basis.
4. Extension to Non-Self-Dual Cases: While the primary results apply to self-dual scenarios (like maximal cuts), the authors demonstrate how to extend these ideas to non-self-dual cases (non-maximal cuts) by relating them to self-dual limits or adding regulators.

4. Results and Examples

The paper validates the methodology on several complex examples:

• Deformed Calabi-Yau Operators:

  • Analyzed hypergeometric functions (${}_3F_2$ and ${}_4F_3$) associated with families of K3 surfaces and Calabi-Yau threefolds.
  • Demonstrated that the number of new $\varepsilon$-functions matches the dimension of the orthogonal group $O_{\Delta}$, confirming the upper bound.
  • Showed that for $N=3$ (K3), only 1 new function is needed; for $N=4$ (CY 3-fold), only 2 are needed, despite the rotation matrix having 6 potential generators.

• Higher-Genus Riemann Surfaces:

  • Applied the method to Lauricella functions associated with hyperelliptic curves of genus $g$.
  • Derived that for odd curves, the number of new generators is bounded by $g(g-1)/2$, and for even curves by $g(g+1)/2$.

• Four-Loop Banana Integral:

  • Applied the formalism to a four-loop banana integral with two distinct masses (a problem involving a two-parameter family of CY threefolds).
  • Successfully reduced the set of generators for the function field $F_\varepsilon$ from 9 potential functions to 6 independent functions, all expressible as iterated integrals over the periods of the underlying geometry.

• Non-Self-Dual Scenarios:

  • Illustrated how to handle cases where self-duality is broken (e.g., by unequal parameters in hypergeometric functions or unregulated poles) by constructing associated self-dual systems or taking specific limits.

5. Significance

This work provides a systematic framework for understanding the transcendental structure of multi-loop Feynman integrals beyond the realm of Multiple Polylogarithms.

• Efficiency: By reducing non-linear constraints to linear ones, it significantly simplifies the construction of canonical bases for complex geometries.
• Classification: It offers a rigorous way to count and identify genuinely new transcendental functions ($\varepsilon$-functions) versus those that are merely combinations of known periods.
• Foundation: It bridges the gap between the algorithmic construction of DEs and the deep mathematical structures of twisted cohomology and Hodge theory, suggesting that the "canonical" nature of these bases is rooted in the geometry of the intersection matrix.

In summary, the paper establishes that the constancy of the intersection matrix in a canonical basis is not just a property but a powerful tool to linearize and solve the constraints on the transcendental functions appearing in multi-loop calculations.