Renormalisation in the flow approach for singular SPDEs

This paper establishes that the renormalisation of singular stochastic partial differential equations (SPDEs) within Duch's flow approach, utilizing a recursive decorated tree ansatz with local extractions, yields a scheme identical to the BPHZ renormalisation found in regularity structures.

The Gist

Imagine you are trying to predict the weather, but your data is so noisy and chaotic that the math breaks down. The numbers blow up to infinity, making the equations useless. This is the problem of "singular Stochastic Partial Differential Equations" (SPDEs). They describe systems like heat spreading through a material with random, jagged noise, or how a surface grows unevenly.

For the last decade, mathematicians have had two main "toolkits" to fix these broken equations: Regularity Structures and Paracontrolled Calculus. These toolkits use complex algebraic tricks to "renormalize" the equations—basically, subtracting out the infinite noise to reveal the meaningful signal underneath.

Recently, a new method called the Flow Approach (developed by Duch) appeared. Instead of fixing the noise all at once, it imagines a "flow" of time where you slowly smooth out the noise, starting from the very small scales and moving up. It's like watching a blurry photo slowly come into focus.

The Problem:
While the Flow Approach works, it was a bit of a "black box." People knew it worked, but they didn't fully understand the hidden algebraic machinery inside it. It was like having a car that drives perfectly, but no one knew exactly how the engine was built.

The Solution (This Paper):
Yvain Bruned and Aurélien Minguella decided to open the hood. Their goal was to take the Flow Approach and rebuild its engine using the same blueprints as the older, well-understood "Regularity Structures" method.

Here is how they did it, using some everyday analogies:

1. The "Tree" of Possibilities

To handle the chaos of the equations, the authors use Decorated Trees. Imagine a family tree, but instead of people, the branches represent different ways the noise can interact with the system.

• The Roots: The starting point of the noise.
• The Branches: How the noise spreads and interacts.
• The Leaves: The final result.

In the old "Regularity Structures" method, these trees were very rigid. In the new "Flow Approach," the trees are a bit more flexible, allowing the "noise" to be spread out over space rather than pinned to a single spot.

2. The "Flow" vs. The "Tree"

The Flow Approach is like a river. You start with a rough, rocky riverbed (the raw noise) and slowly smooth it out as the water flows downstream.

• The Old Way: You looked at the whole river at once and tried to calculate the smoothness.
• The New Way (This Paper): The authors show that you can actually build the river's path by looking at the individual "trees" (the interactions) and rearranging them. They proved that if you arrange these trees correctly, they naturally follow the rules of the "Flow."

3. The "Renormalization" (The Magic Eraser)

The core of the paper is about Renormalization.

• The Analogy: Imagine you are trying to draw a picture, but someone keeps spraying random paint splatters on it. To see the picture, you have to wipe away the splatters.
• The Trick: In math, you can't just "wipe" them away; you have to subtract them algebraically. The authors introduced a specific "map" (called an Evaluation Map) that tells you exactly which splatters to wipe away and how much to subtract.

They proved that the "Flow Approach" uses the exact same wiping rules as the older "Regularity Structures" method. It's like discovering that two different chefs, using different recipes, are actually using the exact same secret spice blend to make their soup taste right.

4. The "Local" vs. "Global" View

One of the biggest differences the authors highlight is how they handle location.

• Regularity Structures: It's like looking at a map where every point is labeled with its exact address. You know exactly where you are.
• Flow Approach: It's like looking at a map where the addresses are a bit blurry; you know you are in a general area, but the details are smeared out by the "flow."

The authors showed that even though the Flow Approach starts with this "blurry" view, they can mathematically "sharpen" it at the end to match the precise "address" system of the older method. They proved that the "blur" is just a temporary step in the process, not a fundamental difference in the math.

The Big Takeaway

The paper doesn't invent a new way to solve these equations or claim it will solve climate change or cure diseases. Instead, it does something more fundamental: It connects the dots.

It proves that the new, modern "Flow Approach" is mathematically identical to the established "Regularity Structures" approach. It shows that the complex, recursive steps in the Flow Approach are just a different way of organizing the same algebraic trees.

In short: They took a new, mysterious method, took it apart, and showed that inside, it's built with the same bricks as the old, trusted method. This gives mathematicians confidence that the Flow Approach is solid, reliable, and fully understood.

Technical Summary

Technical Summary: Renormalisation in the flow approach for singular SPDEs

Problem Statement
This work addresses the algebraic structure of renormalisation for singular stochastic partial differential equations (SPDEs) within the "flow approach" recently proposed by Duch. While the flow approach, inspired by Polchinski's continuous renormalisation group, has successfully established well-posedness for subcritical SPDEs (such as the $\phi^4_3$ and generalized KPZ equations) by constructing stochastic estimates inductively, its algebraic underpinnings remained less transparent compared to the theory of regularity structures. Specifically, the recursive construction of coefficients in the flow approach was previously defined implicitly through the flow equation itself. This paper seeks to clarify the combinatorics and algebraic machinery behind this method, explicitly defining the renormalisation procedure and demonstrating its equivalence to the established BPHZ renormalisation scheme used in regularity structures.

Methodology
The authors employ a framework based on decorated trees to formalize the solution ansatz for the flow equation. The methodology diverges from previous flow-based works by defining the evaluation map (which transforms abstract trees into distributions) explicitly and recursively, rather than deriving it solely from the flow equation.

Key algebraic tools introduced and utilized include:

1. Decorated Trees: A set of trees $T$ equipped with node decorations (polynomials, noise types $\Xi$, and integration types) and edge decorations. Crucially, the authors introduce "grey" integration symbols and a binary marker $v \in \{0,1\}$ on nodes. The marker $v$ controls where grafting operations (representing Fréchet derivatives) can occur, distinguishing between the non-local nature of the flow approach and the local nature of regularity structures.
2. Abstract Derivative ($\uparrow_\mu$): A linear operator acting on trees that changes edge decorations from integration kernels $(G - G_\mu)$ to their scale derivatives $\dot{G}_\mu$. This operator satisfies a commutation relation with the evaluation map.
3. Extraction-Contraction Coproduct ($\Delta_r$): A coproduct acting on trees that extracts sub-trees at the root. This is adapted from the root-extraction coproduct in regularity structures but modified to handle the non-local features of the flow approach.
4. Localization Map ($M_{loc}$): A map that performs abstract Taylor expansions on the nodes of a tree, inserting polynomial decorations. This allows the authors to separate the local functional dependence from the stochastic terms.
5. Evaluation Map ($\Pi^R_{\varepsilon, \mu}$): The central object of the paper, defined recursively as:
   $$\tilde{\Pi}^R_{\varepsilon, \mu}\tau = (\ell\tilde{\Upsilon} \otimes \Pi^{R\times}_{\varepsilon, \mu})(M_{loc} \otimes \text{id})\Delta_r$$
   Here, $\ell$ is a character vanishing on trees of non-negative degree, $\tilde{\Upsilon}$ represents elementary differentials (functionals of the solution $\phi$), and $\Pi^{R\times}$ is a multiplicative model.

Key Contributions
The paper provides an explicit, bottom-up definition of the renormalisation coefficients in the flow approach, contrasting with the top-down or implicit definitions in prior literature. The main technical contributions are:

• Explicit Definition of the Evaluation Map: The authors define the renormalised evaluation map $\Pi^R_{\varepsilon, \mu}$ using the coproduct $\Delta_r$ and the localization map $M_{loc}$. This definition is recursive in the depth of the trees, similar to the framework in regularity structures, but adapted to the non-local context of the Polchinski flow.
• Algebraic Properties: The paper establishes two critical algebraic properties:
  1. Commutation with the Derivative: $\partial_\mu \Pi^R_{\varepsilon, \mu} = \Pi^R_{\varepsilon, \mu} \uparrow_\mu$. This shows that the derivative with respect to the scale parameter $\mu$ commutes with the evaluation map, mirroring the behavior of the abstract derivative on trees.
  2. Pre-Lie Morphism Property: A relation connecting the bilinear operator $B$ (arising from the Fréchet derivative in the Polchinski equation) and the grafting operator $\curvearrowright$. Specifically, $B(\Pi^R_{\varepsilon, \mu}\sigma, \Pi^R_{\varepsilon, \mu}\tau) = \sum_a \Pi^R_{\varepsilon, \mu}(\sigma \curvearrowright_a \tau)$. This identity ensures that the algebraic structure of the flow equation is preserved under the evaluation map.
• Coherence Result: The authors prove that the ansatz defined by their explicit evaluation map satisfies the truncated Polchinski flow equation. This serves as a "sanity check," confirming that the explicit algebraic construction is consistent with the dynamical definition used in previous works.
• Equivalence to BPHZ: By localizing the evaluation map and choosing the character $\ell$ to be the BPHZ character (defined by the condition that the expectation of the renormalised model vanishes at the base point), the authors demonstrate that the renormalisation scheme in the flow approach is identical to the BPHZ renormalisation in regularity structures.

Main Results

1. Theorem 5.3 (Coherence): The general ansatz for the flow equation, constructed via the explicit evaluation map, satisfies the truncated Polchinski equation. This confirms that the explicit algebraic definition of coefficients is compatible with the flow recursion.
2. Theorem 5.5 (Renormalised Equation): The renormalisation of the coefficients at the level of the SPDE yields a counterterm of the form $\sum \frac{\ell(\sigma)}{S(\sigma)} \Upsilon[\sigma][\phi]$. This recovers the known renormalised SPDE structure found in regularity structures, proving that the flow approach produces the same physical counterterms.
3. Theorem 5.14 (BPHZ Renormalisation): The renormalisation character $\ell$ required to satisfy the stochastic estimates in the flow approach is precisely the BPHZ character used in regularity structures. This is established by localizing the evaluation map and showing that the choice of counterterms corresponds to the condition $E[(\hat{\Pi}^R_{\varepsilon, 1}\tau)(0)] = 0$.

Significance
The paper claims to shed new light on the "hidden combinatorics" behind the flow approach. By providing an explicit, tree-based definition of the renormalisation map, the authors bridge the gap between the flow approach and the theory of regularity structures. The work demonstrates that despite the different analytical starting points (continuous flow vs. model reconstruction), the underlying algebraic renormalisation machinery is identical. This unification suggests that the flow approach is not merely an alternative analytical tool but possesses a robust algebraic structure equivalent to the well-developed Hopf algebraic framework of regularity structures. The authors note that while they do not re-prove the well-posedness of SPDEs (as this was already established in Duch's works), their results clarify the algebraic consistency and the specific nature of the renormalisation counterterms, offering a "diagram-free" perspective that aligns with recent developments in regularity structures.