Werner's stochastic Green formula for planar Brownian loops: absolute winding mass in the Wiener sausage, cutoff-profile robustness, and the class of admissible loops

Werner's 1995 paper 'Formule de Green, lacet brownien plan et aire de Lévy' proved a stochastic analogue of Green's formula for the planar Brownian loop: the integral of the winding number over the complement of a thin Wiener sausage converges in probability to Lévy's stochastic area, while the signed mass of the $\varepsilon$-regularized winding field inside the sausage tends to $0$. The paper left three questions open: (1) whether the total absolute regularized winding mass inside the $\varepsilon$-Wiener sausage converges in probability to a positive constant; (2) whether the formula extends to indices regularized by general spatial cutoff profiles, with a profile-independent limit; (3) the maximal class of stochastic loops and the weakest integrand regularity for which polygonal approximations converge to a mesh-independent generalized Stratonovich integral agreeing with the classical Stratonovich integral when the latter exists. The successor program of Sauzedde (2021-2026) established regularization robustness for the signed functional and for the renormalized squared winding, and introduced intrinsic constructions that avoid polygonal approximation, but does not address the absolute-mass conjecture, nor the general-profile or maximal-class questions in their stated forms. The problem remains open: component (1) is untouched, and components (2)-(3) are narrowed but not settled.

Background

Let $Z=(Z^1,Z^2)$ be a planar Brownian loop, that is, a Brownian bridge: a standard planar Brownian motion on a fixed time interval conditioned to return to its starting point (equivalently for most purposes, a Brownian path closed by appending a straight line segment). For a point $z$ off the trajectory, the winding number, or index, $n_z$ of the loop around $z$ is the net number of turns of $Z$ around $z$: the continuous angular variation of the vector $Z_s-z$ divided by $2\pi$. For a smooth closed curve $\gamma$, Green's formula reads $\int_{\mathbb{R}^2}n_z\,(\operatorname{curl}F)(z)\,dz=\oint_\gamma F\cdot d\gamma$ for every $C^2$ vector field $F$; in particular $\int_{\mathbb{R}^2}n_z\,dz$ is the signed area enclosed by $\gamma$, counted with multiplicity. A natural question, which underlies the present problem, is which parts of this identity survive when $\gamma$ is a Brownian loop, whose trajectory is a fractal set that comes arbitrarily close to every point it surrounds.

For the Brownian loop the naive identity fails in a strong way. Werner proved that $\int_{\mathbb{R}^2}|n_z|\,dz=\infty$ almost surely, and that the Amperean area $\int_{\mathbb{R}^2}(n_z)^2\,dz$ is also almost surely infinite, even though Lévy's stochastic area $\mathcal{A}=\frac{1}{2}\int_0^1(Z^1_s\,dZ^2_s-Z^2_s\,dZ^1_s)$, a Stratonovich-type stochastic integral that plays the role of the signed enclosed area, is an almost surely finite random variable. (The Stratonovich integral is defined as a limit of Riemann sums evaluated at midpoints of the time partition and, unlike the Itô integral, satisfies the classical chain rule.) The distributional shape of the winding field is classical: Spitzer proved in 1958 that the winding of planar Brownian motion around a fixed point, divided by $\log t$, converges to a Cauchy law (the symmetric heavy-tailed distribution with density $1/(\pi(1+x^2))$ on $\mathbb{R}$), and Yor determined in 1980 the exact law of $n_z$ for the Brownian loop around a fixed point in terms of the Hartman-Watson distribution, an explicit probability law arising in Yor's computation of exponential functionals of Brownian motion; Werner showed in 1994 that the Lebesgue area of the set $D_N=\{z:n_z\ge N\}$ of points around which the loop winds at least $N$ times in the positive direction satisfies $2\pi N|D_N|\to 1$ in $L^2$. The tail $|D_N|\sim 1/(2\pi N)$ is the analytic source of the divergences above: the sets $\{z:|n_z|\ge N\}$ have areas of order $1/N$, so $\int_{\mathbb{R}^2}|n_z|\,dz=\sum_{N\ge1}|\{z:|n_z|\ge N\}|$ diverges like the harmonic series, while cancellation between positively and negatively wound points can still leave signed quantities finite.

Two regularizations make the Green formula meaningful for the Brownian loop. The Wiener sausage $S_\varepsilon$ of thickness $\varepsilon$ is the $\varepsilon$-neighborhood of the trajectory; classical estimates of Spitzer give $\mathbb{E}[|S_\varepsilon|]$ of order $1/|\log\varepsilon|$, so the sausage is a very thin neighborhood of the curve. The $\varepsilon$-regularized index is $n_z^\varepsilon=\frac{1}{2\pi i}\int_0^1\frac{\mathbf{1}_{\{|Z_s-z|>\varepsilon\}}}{Z_s-z}\,\delta'Z_s,$ where $\delta'Z_s$ denotes the Stratonovich-type differential used in the source; it agrees with the raw index $n_z$ on the complement of $S_\varepsilon$ and is cut off inside the sausage. Werner's main results are that $\int_{\mathbb{R}^2\setminus S_\varepsilon}n_z\,dz\to\mathcal{A}$ in probability, that the signed sausage mass $X_\varepsilon=\int_{S_\varepsilon}n_z^\varepsilon\,dz\to 0$ in probability with $\mathbb{E}[X_\varepsilon^2]\le C/|\log\varepsilon|$ and almost surely along sequences $\varepsilon_p$ decaying faster than $e^{-p^8}$, and, in the 1994 companion paper, that $\frac{1}{|\log\varepsilon|}\int_{\mathbb{R}^2}(n_z^\varepsilon)^2\,dz\to\frac{1}{2\pi}$ in probability. Together these form a stochastic Green formula: they recover the Lévy area of the loop from its spatial winding field.

The same 1995 paper develops a second, dual tool: a generalized Stratonovich integral $\int_0^1G(Z_s)\,\delta'Z_s$ for locally bounded Borel integrands $G$, defined along the Brownian bridge as a limit of integrals along polygonal approximations of the path (line integrals along polygonal interpolations of $Z$ on finer and finer time meshes), and shown to coincide with the classical Stratonovich integral when the latter exists; this is what gives meaning to the differential $\delta'Z_s$ above, and the construction extends to loops given by reversible semimartingales, that is, processes that are semimartingales (decomposable into a local martingale and a finite-variation drift) and remain so under time-reversal. The whole circle of ideas has a concrete physical reading: the regularized index models the Aharonov-Bohm phase, the quantum-mechanical phase shift accumulated by a charged particle encircling a magnetic impurity of finite core size, and the decomposition of the Lévy area into winding sectors (the level sets $\{z:n_z=k\}$ of the winding field) is used in the statistical mechanics of anyon gases (anyons being two-dimensional particles with fractional exchange statistics) and two-dimensional charged systems.

The program lay mostly dormant until its systematic revival by Sauzedde (2021-2026, including a Sorbonne thesis): strengthened $L^p$ and almost-sure asymptotics for $D_N$ with second-order terms; the theorem that planar Brownian motion 'winds evenly', in the sense that the measures $2\pi N\mathbf{1}_{D_N}\,dz$ converge almost surely to the occupation measure, the random measure recording the time spent by the path in each Borel set; a construction of the Lévy area without any approximation, as a Cauchy limit of average windings around Poisson-distributed points; a stochastic Green formula for general $C^{1+\beta}$ one-forms (differential forms with $\beta$-Hölder first derivatives) with symmetric cut-offs on the winding value, with applications to inhomogeneous magnetic impurities and a 1-stable central limit theorem; and a mollifier-independent construction of the renormalised Amperean area. Against this backdrop, the present problem records the three questions that Werner's original paper explicitly left open, which together delineate what is still missing for a complete, robust and quantitatively controlled stochastic Green formula: control of the absolute (unsigned) winding mass near the trajectory, independence of the spatial cutoff profile, and the maximal domain of validity in loops and integrands.

Problem Statement

Let $Z$ be a planar Brownian loop, $S_\varepsilon=\{z:\inf_{0\le s\le1}|Z_s-z|\le\varepsilon\}$ its Wiener sausage of thickness $\varepsilon$ (the $\varepsilon$-neighborhood of the trajectory), and $n_z^\varepsilon=\frac{1}{2\pi i}\int_0^1\frac{\mathbf{1}_{\{|Z_s-z|>\varepsilon\}}}{Z_s-z}\,\delta'Z_s$ the $\varepsilon$-regularized winding number, with $\delta'Z_s$ the Stratonovich-type differential of the source's generalized stochastic integral; $n_z^\varepsilon$ agrees with the ordinary winding number $n_z$ (the net number of turns of $Z$ around $z$) off the sausage and is cut off inside it. Werner proved that $\int_{\mathbb{R}^2\setminus S_\varepsilon}n_z\,dz\to\mathcal{A}$ in probability, where $\mathcal{A}=\frac{1}{2}\int_0^1(Z^1_s\,dZ^2_s-Z^2_s\,dZ^1_s)$ is Lévy's stochastic area, and that the signed sausage mass $X_\varepsilon=\int_{S_\varepsilon}n_z^\varepsilon\,dz\to 0$ in probability. The problem, as explicitly left open in the source and not closed by the subsequent literature, is to complete the stochastic Green formula along three linked axes, with the common objective of a robust and quantitatively controlled Green formula for the Brownian winding field:

(1) Absolute winding mass. Prove or disprove that the total absolute regularized winding mass inside the sausage, $\int_{S_\varepsilon}|n_z^\varepsilon|\,dz,$ converges in probability, as $\varepsilon\to 0$, to a positive finite constant (and identify the constant). The signed result $X_\varepsilon\to 0$ relies on cancellation between positively and negatively wound points; the absolute version has no such mechanism, and the known second-moment bounds do not control first absolute moments of the Cauchy-tailed winding variables.

(2) Cutoff-profile robustness. Extend the definition to regularized indices with a general cutoff profile $h$, $n_z^\varepsilon(h)=\frac{1}{2\pi i}\int_0^1\frac{h((Z_s-z)/\varepsilon)}{Z_s-z}\,\delta'Z_s$

Here the admissible profile $h$ is a bounded measurable function that vanishes near $0$ and equals $1$ outside a bounded set; the indicator $\mathbf{1}_{|r|>1}$ of (1) is the model case, for which $h((Z_s-z)/\varepsilon)=\mathbf{1}_{|Z_s-z|>\varepsilon}$ and $n_z^\varepsilon(h)$ reduces to $n_z^\varepsilon$. Determine whether the limits that the statement attaches to the stochastic Green formula are independent of the profile $h$: the off-sausage recovery of Lévy's area $\int_{\mathbb{R}^2\setminus S_\varepsilon}n_z^\varepsilon(h)\,dz\to\mathcal{A}$, the vanishing of the signed sausage mass $\int_{S_\varepsilon}n_z^\varepsilon(h)\,dz\to0$, and the renormalized squared winding $\frac{1}{|\log\varepsilon|}\int_{\mathbb{R}^2}(n_z^\varepsilon(h))^2\,dz\to\frac{1}{2\pi}$; if not, exhibit a profile for which one of these limits fails. This fixes the scope of (2) to the established limits and does not include the open absolute winding mass of (1), so (2) can be resolved independently of (1).

(3) Maximal domain of validity. Determine the largest class of stochastic loops (beyond reversible semimartingales, that is, loops that are semimartingales and remain so under time-reversal) and the weakest regularity of integrands $G$ for which the polygonal approximations of the path on finer and finer time meshes converge to a mesh-independent limit whose value is the generalized Stratonovich integral of $G$ along the loop, and this limit agrees with the classical Stratonovich integral when the latter exists, so that the Green formula applies; in particular, settle whether polygonal convergence can fail for a weaker integrand class than the locally bounded Borel integrands for which the source constructs it, such as an unbounded, merely measurable $G$. Resolving (3) requires settling both optima jointly, over the ambient class of stochastic loops (the reversible semimartingales being the known case) and the scale of integrand regularity descending from the locally bounded Borel class the source uses, the reachable loop class possibly depending on the regularity of $G$; no optimum is asserted to be attained. The 'in particular' probe of a weaker-than-locally-bounded-Borel $G$ bears on the regularity optimum.

Components may also be resolved individually. Resolving (1) requires controlling the spatially regularized unsigned mass at sausage scale.

References

1. Wendelin Werner (1995). Formule de Green, lacet brownien plan et aire de Lévy. Stochastic Processes and their Applications 57(2):225-245. DOI: 10.1016/0304-4149(94)00052-u. https://doi.org/10.1016/0304-4149(94)00052-u
2. Wendelin Werner (1994). Sur les points autour desquels le mouvement brownien plan tourne beaucoup. Probability Theory and Related Fields 99(1):111-144. DOI: 10.1007/BF01199592. https://doi.org/10.1007/BF01199592
3. Wendelin Werner (1994). Rate of explosion of the Amperean area of the planar Brownian loop. Séminaire de Probabilités XXVIII, Lecture Notes in Mathematics, 153-163. DOI: 10.1007/BFB0073843. https://www.numdam.org/item/SPS_1994__28__153_0/
4. Frank Spitzer (1958). Some theorems concerning 2-dimensional Brownian motion. Transactions of the American Mathematical Society 87:187-197. DOI: 10.1090/s0002-9947-1958-0104296-5. https://doi.org/10.1090/s0002-9947-1958-0104296-5
5. Frank Spitzer (1964). Electrostatic capacity, heat flow, and Brownian motion. Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete 3:110-121. DOI: 10.1007/BF00535970. https://doi.org/10.1007/BF00535970
6. Marc Yor (1980). Loi de l'indice du lacet Brownien, et distribution de Hartman-Watson. Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete 53(1):71-95. DOI: 10.1007/BF00531612. https://doi.org/10.1007/BF00531612
7. Isao Sauzedde (2022). Lévy area without approximation. Annales de l'Institut Henri Poincaré, Probabilités et Statistiques 58(4):2165-2200. DOI: 10.1214/21-aihp1230. https://doi.org/10.1214/21-aihp1230 (preprint arXiv:2101.03992, https://arxiv.org/abs/2101.03992)
8. Isao Sauzedde (2023). Planar Brownian motion winds evenly along its trajectory. ALEA, Lat. Am. J. Probab. Math. Stat. 20(2):1359-1366. DOI: 10.30757/alea.v20-50. https://doi.org/10.30757/alea.v20-50 (preprint arXiv:2102.12372, https://arxiv.org/abs/2102.12372)
9. Isao Sauzedde (2021). Winding and intersection of Brownian motions. Preprint arXiv:2112.01645. https://arxiv.org/abs/2112.01645
10. Isao Sauzedde (2026). Brownian windings, stochastic Green's formula and inhomogeneous magnetic impurities. Electronic Journal of Probability 31:1-26. DOI: 10.1214/26-ejp1584. https://doi.org/10.1214/26-ejp1584 (preprint arXiv:2301.00551, https://arxiv.org/abs/2301.00551)
11. Isao Sauzedde (2024). Renormalised Amperean area of Brownian motions and Symanzik representation of the 2D Abelian Yang-Mills-Higgs field. Preprint arXiv:2412.16781. https://arxiv.org/abs/2412.16781
12. Isao Sauzedde (2024). Holonomy of the planar Brownian motion in a Poisson punctured plane. Communications in Mathematical Physics 405, Article 139. DOI: 10.1007/s00220-024-05019-1. https://doi.org/10.1007/s00220-024-05019-1 (preprint arXiv:2110.15816, https://arxiv.org/abs/2110.15816, titled 'Homotopy and holonomy of the planar Brownian motion in a Poisson punctured plane')
13. Isao Sauzedde (2021). Windings of the planar Brownian motion and Green's formula (Enlacements du mouvement brownien plan et formule de Green). PhD thesis, Sorbonne Université. DOI: 10.70675/e3eaf1edzca20z4392z9a91zb37974483989. https://theses.fr/2021SORUS437
14. Isao Sauzedde (2021). Integration and stochastic integration in Gaussian multiplicative chaos. Preprint arXiv:2105.01232. https://arxiv.org/abs/2105.01232
15. Christophe Garban and José A. Trujillo Ferreras (2006). The expected area of the filled planar Brownian loop is π/5. Communications in Mathematical Physics 264:797-810. DOI: 10.1007/s00220-006-1555-2. https://doi.org/10.1007/s00220-006-1555-2 (preprint arXiv:math/0504496, https://arxiv.org/abs/math/0504496)
16. Alain Comtet and Yves Tourigny (2015). Explicit formulae in probability and in statistical physics. In: In Memoriam Marc Yor — Séminaire de Probabilités XLVII, Lecture Notes in Mathematics, 505-519. DOI: 10.1007/978-3-319-18585-9_22. https://doi.org/10.1007/978-3-319-18585-9_22 (preprint arXiv:1409.3777, 2014, https://arxiv.org/abs/1409.3777)
17. Bruno Schapira and Robert Young (2011). Windings of planar random walks and averaged Dehn function. Annales de l'Institut Henri Poincaré, Probabilités et Statistiques 47(1). DOI: 10.1214/10-aihp365. https://doi.org/10.1214/10-aihp365 (preprint arXiv:0807.2192, 2008, https://arxiv.org/abs/0807.2192)
18. Yves Le Jan (2019). Brownian winding fields. Séminaire de Probabilités L, Lecture Notes in Mathematics, 487-492. DOI: 10.1007/978-3-030-28535-7_17. https://doi.org/10.1007/978-3-030-28535-7_17 (preprint arXiv:1811.02737, https://arxiv.org/abs/1811.02737)