From path integral quantization to stochastic quantization: a pedestrian's journey

This paper establishes the equivalence between path integral and stochastic quantizations for generic scalar Euclidean quantum field theories by providing two novel proofs based on Taylor interpolations indexed by forests: one operating at the level of individual Feynman expansion terms and the other directly at the path integral level without requiring a full perturbative expansion.

The Gist

Here is an explanation of the paper "From path integral quantization to stochastic quantization: a pedestrian's journey," translated into simple language with creative analogies.

The Big Picture: Two Ways to Cook the Same Dish

Imagine you are trying to bake a very complex cake (representing a Quantum Field Theory). You want to know exactly how the ingredients interact to create the final flavor (the correlation functions or predictions of the theory).

In the world of physics, there are two famous, very different recipes for baking this cake:

1. The Path Integral Recipe (The "All-At-Once" Approach):
   Imagine you are a time-traveling baker. You don't just bake one cake; you bake every possible version of the cake simultaneously in a giant multiverse. Some versions have too much sugar, some have no flour, some are burnt. You then take a "weighted average" of all these infinite possibilities to find the true flavor. This is the standard Path Integral method. It's powerful but mathematically messy because you are summing up an infinite number of chaotic scenarios.

2. The Stochastic Recipe (The "Random Walk" Approach):
   Imagine you are baking a cake in a kitchen where the oven is shaking violently (representing random noise). You start with a raw dough and let it sit in this shaking oven for a long time. As time passes, the random shaking mixes the ingredients. Eventually, the cake settles down into a stable, perfect shape. This is Stochastic Quantization. Instead of summing all possibilities at once, you let a random process evolve over "fictitious time" until it reaches equilibrium.

The Problem:
For decades, physicists have known these two recipes should produce the exact same cake. However, proving they are mathematically identical has been like trying to prove two different languages describe the same poem. The existing proofs were either too complicated, relied on specific assumptions (like momentum conservation), or were just "sketchy" hand-waving.

The Solution (This Paper):
The authors, Dario Benedetti, Ilya Chevyrev, and Razvan Gurau, have provided two new, crystal-clear proofs that these two recipes are indeed identical. They didn't just say "it works"; they showed you exactly how the ingredients in one recipe transform into the ingredients of the other.

---

The Secret Ingredient: The "Forest" Analogy

To connect these two recipes, the authors use a clever mathematical tool called Taylor Interpolation indexed by Forests.

Let's break this down with an analogy:

Imagine you have a tangled ball of yarn (representing the complex interactions in the Path Integral). You want to untangle it to see the structure underneath.

• The Path Integral sees the yarn as a giant, messy knot where every thread is connected to every other thread in a chaotic web.
• The Stochastic Method sees the yarn as a series of trees growing over time.

The authors' trick is to take that messy ball of yarn and systematically cut it into Forests (collections of trees).

1. The "Tree Picking" Algorithm:
   Imagine you are walking through a dense forest (the Feynman graph). You start at the root (the external observation point). You decide to walk a specific path, always choosing the "rightmost" available path that doesn't create a loop. This path becomes a Tree.

   • The paths you didn't walk become Noise Edges (the "background static" or random fluctuations).

2. The Transformation:
   The authors show that if you take the messy "All-At-Once" sum (Path Integral) and break it down into these specific "Tree" paths plus "Noise" connections, you get exactly the same mathematical result as the "Random Walk" method (Stochastic Quantization).

The Two Proofs: A Pedestrian's Journey

The paper offers two ways to walk this path, like two different hiking trails leading to the same mountain peak.

Trail 1: The "Graph-by-Graph" Hike (Perturbative Proof)

• The Method: This proof looks at the cake recipe one tiny ingredient at a time. In physics, we often calculate things by looking at specific diagrams (Feynman graphs).
• The Analogy: Imagine you have a specific, complex Lego structure. The authors show you how to take that specific Lego structure apart, piece by piece, and reassemble it into a set of trees growing from a base.
• The Result: They prove that for every single diagram in the Path Integral, there is a corresponding sum of "tree diagrams" in the Stochastic method. It's like showing that every specific puzzle piece in one box fits perfectly into a specific slot in the other box.

Trail 2: The "Whole Landscape" Hike (Non-Perturbative Proof)

• The Method: This is the more advanced proof. Instead of looking at individual Lego pieces, they look at the entire box of Legos at once.
• The Analogy: Imagine you have a giant, continuous sheet of fabric (the Path Integral). Instead of cutting it into pieces, they use a special "magic scissors" (Taylor Interpolation) to slice the fabric in a way that reveals the underlying tree structure without ever breaking the fabric apart.
• The Result: They show that the entire mathematical formula for the Path Integral can be rewritten directly as the formula for the Stochastic process. This is a stronger proof because it doesn't rely on breaking the problem down into infinite small steps; it treats the whole system as a single, flowing entity.

Why Does This Matter?

1. It's More General: Previous proofs only worked for simple cases where particles conserved momentum (like billiard balls hitting each other). These new proofs work even for "weird" spaces where momentum isn't conserved, such as curved spaces or specific quantum models (like the Gross–Wulkenhaar model).
2. It's Constructive: They don't just say "it's true." They give you a step-by-step map (the forest algorithm) to convert one method into the other.
3. Bridging the Gap: This brings together two major schools of thought in mathematical physics. The "Constructive Field Theory" crowd (who love Path Integrals) and the "Stochastic Analysis" crowd (who love Random Processes) can now speak the same language.

Summary

Think of this paper as a Rosetta Stone for quantum field theory.

• Left Side: The chaotic, all-at-once "Path Integral" (The Multiverse Bakery).
• Right Side: The evolving, time-based "Stochastic Quantization" (The Shaking Oven).
• The Translation: A set of rules (Forests and Trees) that proves, beyond a shadow of a doubt, that both methods bake the exact same cake.

The authors have taken a very high-level, abstract mathematical problem and provided a "pedestrian's journey"—a clear, walkable path—for anyone to understand how these two worlds are connected.

Technical Summary

Here is a detailed technical summary of the paper "From path integral quantization to stochastic quantization: a pedestrian's journey" by Dario Benedetti, Ilya Chevyrev, and Razvan Gurau.

1. Problem Statement

The paper addresses the theoretical disconnect between two rigorous formulations of Euclidean Quantum Field Theory (EQFT):

1. Path Integral Quantization: Based on the functional integral measure $d\nu(\phi) \propto e^{-S(\phi)}[d\phi]$, where correlation functions are computed via perturbative expansions (Feynman diagrams) or constructive field theory methods.
2. Stochastic Quantization (Parisi-Wu): Based on a Langevin stochastic differential equation with a fictitious time $t$. Correlation functions are obtained as the equilibrium limit ($t \to \infty$) of stochastic averages.

While it is widely accepted that these two approaches yield the same physical results, existing proofs of their equivalence are often:

• Indirect: Relying on the Fokker-Planck equation.
• Restrictive: Depending on momentum conservation or specific momentum representations, making them inapplicable to theories with non-conserved momentum (e.g., on curved spaces or with harmonic oscillator potentials like the Gross-Wulkenhaar model).
• Sketchy: Lacking explicit, constructive derivations at the level of individual graph amplitudes or the full path integral.

The authors aim to provide two novel, direct, and general proofs establishing the equivalence of these quantization schemes for generic scalar Euclidean QFTs, without relying on momentum conservation or inductive arguments.

2. Methodology

The authors employ Taylor interpolations indexed by forests, a technique rooted in constructive field theory. They develop two distinct proofs:

A. Perturbative Proof (Level of Individual Graphs)

• Approach: They compare the Feynman amplitude of a specific graph in the path integral expansion with the sum of stochastic amplitudes derived from the same underlying combinatorial map.
• Mechanism:
  • They utilize a "tree picking" algorithm (specifically, a "keep to the right" or "first available successor" algorithm) to select spanning forests within a Feynman graph.
  • They apply a fundamental theorem of calculus (Taylor expansion with integral remainder) to the covariance terms. Specifically, they replace the static covariance $1/A$with an integral over fictitious time involving the heat kernel$e^{-tA}$.
  • By iteratively interpolating the time parameters of internal vertices, they decompose a Feynman amplitude (a sum over all edges) into a sum over spanning forests (tree edges) and noise edges (non-tree edges).
  • The tree edges acquire time-ordered propagators $\theta(t_v - t_w)e^{-(t_v-t_w)A}$, while the noise edges acquire the equilibrium propagator $e^{-|t_v-t_w|A}/A$.

B. Non-Perturbative Proof (Level of the Path Integral)

• Approach: They prove the equivalence directly at the level of the functional integral without expanding into a series of Feynman graphs.
• Mechanism:
  • They introduce field copies ($\phi_0, \phi_1, \dots$) to separate the observable from the interaction terms within the Gaussian measure.
  • They perform a Taylor interpolation on the fictitious time variables associated with these field copies.
  • This process generates a forest-indexed sum of path integrals.
  • Crucially, they apply a Hubbard-Stratonovich transformation to the resulting Gaussian integral over field copies. This transformation introduces a white noise field $\xi$.
  • The resulting expression is identified exactly as the expectation value of a product of stochastic fields $\Phi(t, x)$, where $\Phi$ is the solution to the Langevin equation expanded as a sum over rooted plane trees.

3. Key Contributions

1. Novel Proofs of Equivalence: The paper provides two rigorous, constructive proofs that the $n$-point functions of the path integral and the large-time limit of stochastic quantization are identical.
2. General Applicability: Unlike previous proofs, these arguments:
   • Do not rely on momentum conservation.
   • Work for any theory with a positively defined covariance (e.g., $-\Delta + x^2$ in the Gross-Wulkenhaar model).
   • Are directly generalizable to curved spaces.
3. Explicit Combinatorial Mapping:
   • The first proof explicitly maps Feynman graphs to sums over spanning forests, clarifying the combinatorial origin of the stochastic "noise edges."
   • It demonstrates that the Feynman amplitude $A(G)$ is the sum of stochastic amplitudes $A(G, F)$ over all spanning forests $F \prec G$.
4. Non-Perturbative Construction: The second proof bypasses the divergent perturbation series entirely, establishing the equivalence at the level of the formal path integral using tree-indexed Taylor interpolations. This is a significant step toward non-perturbative constructions of the EQFT measure.
5. Constructive Nature: Both proofs avoid inductive arguments, relying instead on explicit Taylor expansions, making the link between Feynman-type contributions and forest/tree expansions transparent.

4. Results

• Theorem 1 (Main Theorem): The quantum field theoretical $n$-point functions coincide with the stochastic $n$-point functions in the limit of large, coincident fictitious times:
  $$\langle \phi_{x_1} \dots \phi_{x_n} \rangle = \lim_{t \to \infty} \langle \Phi(t, x_1) \dots \Phi(t, x_n) \rangle_\xi$$
• Theorem 2 (Perturbative Reformulation): For any Feynman graph $G$, its amplitude $A(G)$ is exactly equal to the sum of the amplitudes of all stochastic diagrams based on $G$ with equal external times:
  $$A(G) = \sum_{F \in \mathcal{F}(G)} A(G, F)$$
  where the sum runs over all spanning forests $F$ of $G$.
• Explicit Examples: The authors provide low-order calculations (up to order $g^3$) for the $\phi^4$ model, explicitly showing how tadpole and sunset diagrams in the path integral decompose into sums of tree and noise-edge diagrams in the stochastic expansion, recovering the correct combinatorial factors.

5. Significance

• Bridging Communities: The paper unifies the "constructive field theory" community (focused on path integrals and convergence) with the "stochastic analysis" community (focused on SPDEs and regularity structures).
• Technical Robustness: By removing the reliance on momentum conservation, the results open the door to applying stochastic quantization techniques to a broader class of models, including those on curved manifolds and non-commutative geometries (like the Gross-Wulkenhaar model), where momentum space methods fail.
• Foundational Insight: The use of tree-indexed Taylor interpolations provides a new perspective on how the "noise" in stochastic quantization emerges from the "interaction" in path integrals. It suggests that the stochastic expansion is a natural reorganization of the path integral expansion based on the causal structure of the fictitious time.
• Renormalization Context: While the paper treats the equivalence formally (assuming regularization), the constructive nature of the proofs implies that renormalization procedures (like BPHZ) can be consistently applied to both frameworks, potentially simplifying the analysis of divergences in complex models.

In summary, the paper offers a "pedestrian's journey" (a direct, step-by-step derivation) that demystifies the connection between two pillars of modern quantum field theory, providing tools that are both mathematically rigorous and physically transparent.