💻 computer science

Multi-indices coproducts from ODEs to singular SPDEs

This paper introduces explicit formulae for multi-index coproducts in ODEs and singular SPDEs by leveraging their relationship as adjoints to dual products defined via easily computable symmetry factors.

Original authors: Yvain Bruned, Yingtong Hou

Published 2026-01-27

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yvain Bruned, Yingtong Hou

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 bake a very complex cake, but the recipe is written in a language so dense and tangled that no one can read it. This is the situation mathematicians face when trying to solve certain types of equations that describe chaotic, "rough" systems in nature, like turbulent fluids or financial markets. These are called Singular Stochastic Partial Differential Equations (SPDEs).

For a long time, mathematicians used a tool called "decorated trees" (think of a family tree with extra labels on the branches) to break these equations down into manageable pieces. However, this paper introduces a new, more efficient way to do the same thing using Multi-indices.

Here is a simple breakdown of what the authors, Yvain Bruned and Yingtong Hou, have achieved:

1. The New "Alphabet": Multi-indices

Instead of drawing complex trees, the authors use Multi-indices.

The Analogy: Imagine a tree is a detailed architectural blueprint. A multi-index is like a simple shopping list.
How it works: Instead of tracking the shape of every branch, they just count how many times specific ingredients (derivatives) appear. For example, instead of drawing a tree with three leaves, they just write "3 apples."
The Goal: This makes the math much more compact and easier to manipulate, especially for the specific type of equations found in physics and finance.

2. The Two Magic Tools: Coproducts

The paper focuses on two specific mathematical operations called Coproducts. In the world of these equations, a coproduct is like a scissors and glue operation.

The Butcher-Connes-Kreimer Coproduct (The "Cut"): This operation takes a complex expression and cuts it into smaller, simpler pieces. It's like taking a big Lego structure and breaking it down into individual bricks so you can study them.
The Extraction-Contraction Coproduct (The "Swap"): This is a more complex operation where you pull a piece out, do something to it, and then put it back in a different way. It's like taking a specific gear out of a clock, polishing it, and putting it back in to see how the whole clock runs differently.

3. The Problem: The "Black Box"

Until now, while mathematicians knew these "scissors and glue" tools existed, they didn't have a clear, step-by-step manual (an explicit formula) for how to use them on these new "shopping lists" (multi-indices). They had to translate the problem back and forth between the "shopping list" and the "blueprint" (trees) to get the answer, which was slow and confusing.

4. The Solution: The "Mirror" Trick

The authors' main breakthrough is finding a direct way to calculate these cuts and swaps without ever needing to draw a tree.

The Analogy: Imagine you want to know how a mirror reflects an object, but you don't want to look at the mirror directly. Instead, you look at the object's "shadow" (the dual product) and realize that if you know the shadow perfectly, you can instantly know the reflection.
The Method: The authors used a mathematical concept called an Inner Product (a way of measuring the "distance" or "similarity" between two lists). They discovered that the "scissors and glue" tools are just the mirror image (the adjoint) of simpler multiplication tools.
The Result: Because the multiplication tools are easy to write down, the authors could simply "flip" them over to get the exact formula for the complex cuts and swaps.

5. Why This Matters (According to the Paper)

The paper provides explicit formulae. This means they have written down the exact recipe for these operations.

Symmetry Factors: They introduced a way to count "symmetry" (how many ways you can rearrange the ingredients without changing the result). This is crucial because it prevents double-counting errors.
Versatility: They solved this for two types of problems:
1. ODEs (Ordinary Differential Equations): Simpler systems, like a single pendulum swinging.
2. SPDEs (Singular SPDEs): The complex, chaotic systems mentioned earlier.

Summary

Think of this paper as the instruction manual for a new, more efficient way to solve the world's messiest math problems.

Before: You had to translate the problem into a complex drawing (trees), use a vague set of rules, and then translate it back.
Now: You can work directly with a simple list of ingredients (multi-indices) using a clear, step-by-step recipe (the explicit formulae) to cut, swap, and solve the equations.

The authors didn't just say "it's possible"; they wrote down the exact math so anyone can follow the steps to break down these difficult equations into their simplest parts.

Technical Summary: Multi-indices Coproducts from ODEs to Singular SPDEs

Problem Statement

The theory of Regularity Structures, introduced by Martin Hairer, provides a framework for solving singular stochastic partial differential equations (SPDEs). Traditionally, this theory relies on decorated rooted trees to code the local expansions of solutions. However, an alternative approach using multi-indices was proposed to index these expansions via elementary differentials rather than iterated stochastic integrals.

While the Hopf algebraic structures underlying multi-indices (specifically for recentering and renormalization) have been established, a significant gap remained: explicit formulae for the coproducts acting on these multi-indices were largely unknown. Previous attempts, such as those in [20], derived formulae by comparing multi-indices to rooted trees and guessing the result, rather than deriving them directly from the multi-index structure. This paper addresses the need for direct, explicit formulae for the two primary coproducts in this context: the Butcher-Connes-Kreimer (BCK) type coproduct (dual to the Grossman-Larson product) and the extraction-contraction coproduct (dual to the insertion product).

Methodology

The authors develop explicit formulae by leveraging the duality between products and coproducts within the Hopf algebra of multi-indices. The core methodological strategy involves:

Defining Dual Products: Instead of constructing coproducts directly (which is combinatorially complex), the authors define explicit associative products ( $\star_2$ and $\star_1$ ) on forests of multi-indices.
- The BCK-type coproduct ( $\Delta$ ) is defined as the adjoint of the product $\star_2$ , which involves a derivation $D$ (or its SPDE generalization $D^{(n)}$ ).
- The extraction-contraction coproduct ( $\Delta^-$ ) is defined as the adjoint of the "simultaneous insertion" product $\star_1$ , which generalizes the Guin-Oudom procedure.
Inner Products and Symmetry Factors: A crucial component of the methodology is the definition of an inner product $\langle \cdot, \cdot \rangle$ on the space of multi-indices and forests. This inner product is weighted by symmetry factors $S(\cdot)$ , which count the number of ways to permute children in the corresponding combinatorial structure.
- The authors highlight a key advantage of multi-indices over trees: the symmetry factor is multiplicative ( $S(z^\beta z^{\bar{\beta}}) = S(z^\beta)S(z^{\bar{\beta}})$ ), a property that simplifies the derivation of coefficients significantly compared to the tree-based approach.
Adjoint Maps: The derivation of explicit coproduct formulae relies on computing the adjoints of the derivations used in the products (e.g., $\bar{D}$ and $\bar{\partial}_k$ ). The authors provide explicit expressions for these adjoint maps, allowing the coproducts to be written in terms of these operators acting on the "remainder" of a partitioned multi-index.
Extension to SPDEs: The framework is extended from Ordinary Differential Equations (ODEs) to singular SPDEs. This involves:
- Introducing non-commutative derivatives to handle space-time noise and the chain rule.
- Defining SPDE multi-indices using words where letters representing spatial derivatives ( $b_i$ ) and solution derivatives ( $n$ ) do not commute.
- Adapting the derivations and symmetry factors to this non-commutative setting while maintaining the multiplicative property of the symmetry factors.

Key Contributions and Results

1. Explicit Formulae for ODE Multi-indices

The paper provides two equivalent explicit formulae for the Linares–Otto–Tempelmayr coproduct (BCK type) on ODE multi-indices:

Formula 1 (Proposition 3.1): Expresses the coproduct as a sum over partitions of the multi-index $\beta = \beta_1 + \dots + \beta_n + \hat{\beta}$ , involving the inner product of the derivation $D^n$ applied to a remainder term.
Formula 2 (Theorem 3.5): An alternative expression utilizing the adjoint map $\bar{D}$ . This formula is structurally cleaner, involving the term $\bar{D}^n z^{\hat{\beta}}$ and a symmetry factor $S_{ext}$ .

Similarly, explicit formulae are derived for the extraction-contraction coproduct ( $\Delta^-$ ):

Formula 1 (Proposition 3.8): Based on the product $\star_1$ and sums over admissible splittings.
Formula 2 (Theorem 3.10): An alternative formula using the adjoint map $\bar{D}$ , which simplifies the coefficient calculation.

2. Extension to Singular SPDEs

The authors generalize these results to the context of singular SPDEs:

SPDE Multi-indices: They define a new class of multi-indices incorporating non-commutative derivatives and specific populated conditions (Equation 4.4).
Generalized Coproducts: They derive explicit formulae for the BCK-type coproduct (Theorem 5.3) and the extraction-contraction coproduct (Theorem 5.6) in the SPDE setting. These formulae involve generalized derivations $D^{(n)}$ and $\partial_k$ , and their adjoints $\bar{D}^{(n)}$ and $\bar{\partial}_k$ .
Handling Non-Commutativity: The derivation accounts for the non-commutative relations between spatial and solution derivatives (Equation 4.2), ensuring the formulae are valid for the local expansions of singular SPDEs.

3. Computational Advantages

The paper demonstrates that using the inner product and symmetry factors allows for a direct derivation of these formulae without referencing decorated trees. The multiplicative nature of the symmetry factor for multi-indices is identified as a critical simplification that avoids the complexities inherent in the tree-based combinatorics.

Significance and Claims

The paper claims to advance the "programme" of finding explicit coproduct formulae for multi-indices, a task previously only partially achieved via indirect comparisons with trees.

Direct Derivation: The authors emphasize that their proofs refer only to multi-indices, avoiding the need to map back and forth to rooted trees. This provides a self-contained algebraic understanding of the multi-index Hopf algebras.
Utility for B-series: The explicit formulae are presented as essential tools for computing the composition and substitution of multi-index B-series (for ODEs) and Regularity Structures B-series (for SPDEs).
Novelty in SPDEs: The formulae for SPDE multi-indices are described as new contributions to the literature, offering a compact formalism (following [4]) that differs from previous chain-rule expansions (e.g., [26]).
Modesty: The paper does not claim to solve new analytical problems or propose new numerical schemes. Instead, it positions itself as a foundational algebraic work that provides the necessary combinatorial tools (explicit coproducts) to facilitate further analytical and probabilistic work in the field of singular SPDEs and numerical analysis.

In summary, the work provides a rigorous, self-contained algebraic framework for computing the fundamental coproducts in the theory of Regularity Structures using multi-indices, offering explicit formulae that are computationally accessible and theoretically distinct from the tree-based approaches.