From path integral quantization to stochastic quantization: a pedestrian’s journey
Dario Benedetti, Ilya Chevyrev, Razvan Gurau · 2026
Read From path integral quantization to stochastic quantization: a pedestrian’s journey by Dario Benedetti, Ilya Chevyrev, Razvan Gurau (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-10761.txt.
Introduction
Dario Benedetti, Ilya Chevyrev, and Razvan Gurau give two new proofs that path integral and stochastic quantizations of scalar Euclidean quantum field theories agree. Both proofs use Taylor interpolations indexed by forests, in the style of constructive field theory. The first works term by term in the Feynman expansion, where forests appear as spanning subgraphs of Feynman diagrams. The second works at the level of the whole path integral and avoids expanding the full perturbation series. Together they offer a pedestrian route between the two quantization schemes.
Catalog record
| CREATOR | Dario Benedetti, Ilya Chevyrev, Razvan Gurau |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 121798 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.10761
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
From path integral quantization to stochastic quantization: a pedestrian’s journey Dario Benedetti, Ilya Chevyrev, Razvan Gurau Abstract We give two novel proofs that the path integral and stochastic quantizations of generic scalar Euclidean quantum field theories are equivalent. Our proofs rely on Taylor interpolations indexed by forests, in the fashion of constructive field theory. The first proof works at the level of individual terms in the Feynman expansion, with the forests appearing as spanning forests in Feynman graphs. The second one works at the level of the path integral and avoids the full expansion of the Feynman perturbation series. 1 Introduction In a seminal paper [ 2 ] , Parisi and Wu argued that the path integral formulation of Euclidean quantum field theory (EQFT) can be obtained as the equilibrium limit of a stochastic dynamic of Langevin type with respect to a fictitious time, in line with the interpretation of EQFT as equilibrium statistical mechanics. This idea stimulated a great deal of activity in the physics community, mainly in the 1980s [ 3 ] , and has received renewed attention in the mathematics community over the last decade. In particular, with the advent of new tools in the study of stochastic partial differential equations, such as regularity structures [ 4 ] , paracontrolled distributions [ 5 ] , and renormalisation group [ 6 , 7 ] , the way was opened to a new approach to the rigorous construction and analysis of the EQFT measure for various models, such as \Phi^{4}_{3} [ 8 , 9 , 10 , 11 , 12 ] . The stochastic methods are complementary to the constructive field theory approach based on the path integral formulation [ 13 , 14 ] and, at a technical level, these two rigorous formulations of EQFT have remained somewhat separated. One of the aims of this paper is to bring these two approaches closer together. The contributions of this paper are two novel proofs that the path integral and stochastic quantizations are equivalent. The first proof works at the level of the Feynman graph expansion in the path integral formulation of a scalar EQFT [ 14 ] and shows that it is equivalent to the tree expansion in stochastic quantization [ 2 ] . Although this result is known in the literature, the currently available proofs are either not direct (using a Fokker–Planck equation) or somewhat sketchy (see [ 3 , Sec. 3.3] and references therein). In particular, the classical proof of [ 15 ] relies on a momentum representation of the covariance, momentum conservation and an induction. Besides the fact that the induction is not trivial, the proof in [ 15 ] does not apply to theories in which the momentum is not conserved. In contrast, our first proof proceeds by a Taylor expansion at the level of Feynman amplitudes, leading to an explicit representation of each graph contribution as a sum over spanning forests weighted by stochastic amplitudes. Our second proof of equivalence of the two quantization procedures relies on a novel strategy involving a tree-indexed Taylor interpolation directly at the level of the path integral without expressing it as a sum over Feynman graphs. Such Taylor interpolations are the backbone of constructive field theory [ 14 , 16 ] and ours is close in spirit to the original Battle–Federbush formula [ 17 , 18 ] . In particular, tree-indexed series similar to those we consider are a first step towards non-perturbative constructions of the measure. However, contrary to the usual constructive field theory interpolations which compute the connected expectations, our interpolation computes directly the expectations (that is the full Schwinger functions) of the model. Both proofs share a number of features: - they work in an arbitrary theory with a positively defined covariance; in particular, contrary to the proof of [ 15 ] , our arguments do not rely on momentum conservation and accommodate covariances such as -\Delta+x^{2} , as in the Gross–Wulkenhaar model [ 19 ] , which has recently been treated in stochastic quantization [ 20 ] ; - they are directly generalizable to curved space; 1 1 1 The construction of EQFT on curved spaces is an important open problem that has recently seen some advances within the stochastic approach (see [ 21 ] and references therein). - they are constructive and avoid inductive arguments; in particular, both rely on explicit Taylor interpolations, which makes the link between Feynman-type contributions and forest/tree expansions transparent. We point out that we do not consider renormalization in this paper: our statements should be understood in a formal sense or in settings where the relevant quantities are well-defined. In theories with divergences, our results would need to be complemented by an appropriate analysis of regularization and subtractions. Plan of the paper. In Sec. 2 , we begin by recalling the basic settings of EQFT in the path integral and stochastic formulations for a generic scalar field theory and we state our main theorem. In Sec. 3 , we take a step back and introduce a number of prerequisites on combinatorial structures, Gaussian measures, and diffusion equations. In Sec. 4 , we give precise definitions of the perturbative expansions to all orders in the path integral and stochastic settings and we reformulate our main theorem at the perturbative level in terms of graph amplitudes. In Sec. 5 , we prove the perturbative formulation of our theorem. In Sec. 6 , we prove the main theorem as formulated in Sec. 2 non-perturbatively, directly at the level of the path integral. In the Appendices we provide explicit examples of the perturbative expansions and the Taylor interpolation at lowest orders. 2 Path integral and stochastic quantization We denote d the dimension of the Euclidean space \mathbb{R}^{d} , and for simplicity we only consider real scalar fields \phi:\mathbb{R}^{d}\to\mathbb{R} . In order to simplify the notation, we sometimes denote arguments of functions as indices, for instance \phi_{x}\equiv\phi(x) , C(x,y)=C_{xy} and \delta_{xy}=\delta^{d}(x-y) denotes the Dirac delta in the appropriate dimension; we sometimes denote \int d^{d}x\equiv\int dx\equiv\int_{x} and use the shorthand notation \phi A\phi\equiv\int_{xy}\;\phi_{x}A_{xy}\phi_{y} . Euclidean quantum field theory. EQFT is defined by a formal path integral measure of the form: \mathop{}\!\mathrm{d}\nu(\phi)=[\mathop{}\!\mathrm{d}\phi]\;\frac{1}{Z(0)}\;e^{-S(\phi)}\;,\qquad Z(0)=\int[\mathop{}\!\mathrm{d}\phi]\;e^{-S(\phi)}\;, where the partition function Z(0) is a normalization constant and the action S(\phi) is some real functional. The aim of EQFT is to compute the moments ( n -point functions or correlation functions) of the theory, which are related to physical observables. In constructive field theory [ 13 , 14 , 16 ] one is interested in making sense of the above definition and check that the moments have some specific properties, for instance that they obey the Osterwalder–Schrader axioms [ 22 , 23 ] . A common choice is to separate the action into a free part, quadratic in the field, and an interaction: S(\phi)=\frac{1}{2}\phi A\phi+V(\phi)\;, (2.1) where the operator A in the free part of the action is assumed to be a closed self-adjoint positive-definite operator, whose inverse is the free covariance of the theory C_{xy}=(A^{-1})_{xy} . The perturbation, or interaction, V(\phi) is a possibly nonlocal functional of \phi : V(\phi)=\sum_{q\geq 1}\frac{1}{q+1}\int_{x^{1},x^{2},\ldots,x^{q+1}}\;V(x^{1},x^{2},\ldots,x^{q+1})\;\phi_{x^{1}}\ldots\phi_{x^{q+1}}\;, (2.2) where by a slight abuse of notation we denote by V both the full potential and its kernels;