Weak convergence of stochastic integrals with applications to SPDEs

Xavier Bardina, Salim Boukfal · 2025

Read Weak convergence of stochastic integrals with applications to SPDEs by Xavier Bardina, Salim Boukfal (2025) 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-2504-08317.txt.

Introduction

Xavier Bardina and Salim Boukfal prove limit results for stochastic integrals in a 2025 paper. Sequences of random fields built from integrals are studied. Conditions are given for them to converge weakly to integrals against the Brownian sheet.

The domain is a rectangle in multi-dimensional space. The kernel function must meet stated integrability conditions. The driving processes need only converge in finite-dimensional law. That makes the result widely usable in practice. An application to the stochastic Poisson equation follows. Readers get a reusable convergence tool for equations driven by noise.

Catalog record

CREATORXavier Bardina, Salim Boukfal
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-08317

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Weak convergence of stochastic integrals with applications to SPDEs Xavier Bardina, Salim Boukfal Abstract In this paper we provide sufficient conditions for sequences of random fields of the form ∫ D f ⁡ ( x , y ) ​ θ n ​ ( y ) ​ 𝑑 y \int_{D}f(x,y)\theta_{n}(y)dy to weakly converge, in the space of continuous functions over D D , to integrals with respect to the Brownian sheet, ∫ D f ⁡ ( x , y ) ​ W ​ ( 𝑑 y ) \int_{D}f(x,y)W(dy) , where D ⊂ ℝ d D\subset\mathbb{R}^{d} is a rectangular domain, x ∈ D x\in D , f f is a function satisfying some integrability conditions and { θ n } n \{\theta_{n}\}_{n} is a sequence of stochastic processes whose integrals ∫ [ 0 , x ] θ n ​ ( y ) ​ 𝑑 y \int_{[0,x]}\theta_{n}(y)dy converge in law to the Brownian sheet (in the sense of the finite dimensional distribution convergence). We then apply these results to stablish the weak convergence of solutions of the stochastic Poisson equation. Keywords: Brownian sheet, stochastic integral, random walk, Poisson process, Kac-Stroock, weak convergence, stochastic Poisson equation MSC 2020: 60F05, 60F17, 60G50, 60G60, 60H05, 60H15 1 Introduction Let d\geq 1 be any integer and T=(T_{1},...,T_{d})\in\mathbb{R}_{+}^{d} be fixed and consider in D=[0,T]=\prod_{i=1}^{d}[0,T_{i}]=[0,T_{1}]\times...\times[0,T_{d}] the usual partial ordering. In, for instance, [ 15 , Corollary 1, p. 683] or [ 1 , Theorem A.1] , and [ 2 ] , the authors exhibited examples of sequences of processes \zeta_{n}=\{\zeta_{n}(x)\colon x\in D\} , converging weakly to the Brownian sheet in the space of continuous functions over D , \mathcal{C}(D) , generalizing the well-known Donsker’s invariance principle and the results proved by Stroock in [ 14 ] in the multidimensional parameter set case, respectively. A natural question to ask ourselves then is if the integrals with respect to such processes converge, in the same sense, as stochastic processes, towards integrals with respect to the Brownian sheet. Of course, this problem has already been addressed in [ 9 ] , [ 10 ] (when a single parameter is considered) and, partially, in [ 3 ] , where two parameters are considered and the convergence of the stochastic integrals of a particular integrand (the Green function associated to the heat equation with Dirichlet boundary conditions) is then used to show the convergence of solutions of the stochastic heat equation. Mostly motivated by the latter, in [ 1 ] , the authors consider sequences of kernels \{\theta_{n}\}_{n\in\mathbb{N}} so that the processes \zeta_{n} can be written as \zeta_{n}(x)=\int_{[0,x]}\theta_{n}(y)dy and provide sufficient conditions for the processes X_{n}\coloneqq\left\{X_{n}(x)=\int_{[0,x]}f(y)\theta_{n}(y)dy\colon x\in D\right\} to weakly converge, in \mathcal{C}(D) , to the process X\coloneqq\left\{X(x)=\int_{[0,x]}f(y)W(dy)\colon x\in D\right\}. Where W=\{W(x)\colon x\in D\} denotes a Brownian sheet and f is some function in L^{2q}(D) for some real q\geq 1 . However, in many applications (for instance, in stochastic partial differential equations) the integrands involved might depend on the time/space parameter as well, which motivates the study of the same problem when the integrand depends on it. Hence, the aim of this paper is to continue with the study of the weak approximation of stochastic integrals in [ 1 ] to extend the results to integrals of the form \int_{[0,x]}f(x,y)W(dy)\hskip 9.24994pt\text{and}\hskip 9.24994pt\int_{D}f(x,y)W(dy), where f\colon D\times D\to\mathbb{R} will be some function satisfying some integrability conditions. Given that \int_{[0,x]}f(x,y)W(dy)=\int_{D}\mathbb{I}_{[0,x]}(y)f(x,y)W(dy), where \mathbb{I}_{A} denotes the indicator function of A\subset\mathbb{R}^{d} ( \mathbb{I}_{A}(y)=1 if y\in A and \mathbb{I}_{A}(y)=0 otherwise), we will focus our efforts on studying the latter. We will see, as well, how these results can be applied to prove the weak convergence of solutions to the stochastic Poisson equation with Dirichlet boundary conditions. The paper is organized as follows, Section 2 is devoted to briefly introduce the Brownian sheet, the stochastic integral with respect to it and to provide some results regarding the weak convergence of stochastic processes. In Section 3 we state and proof the main theorem and, in Section 4 we see how these results can be applied to prove the weak convergence of solutions of the Poisson equation. 2 Preliminaries In this section we shall provide the main definitions and tools we will be working with. 2.1 Brownian sheet and integrals with respect to it To define the Brownian sheet and the stochastic integral with respect to such process, we will make use of the isonormal Gaussian process over a real separable Hilbert space H with inner product \langle\cdot,\cdot\rangle_{H} . Definition 2.1 . We say that a stochastic process W=\{W(h)\colon h\in H\} defined in a complete probability space is an isonormal Gaussian process if it is a centered Gaussian process with covariance function \text{Cov}(f,g)=\langle f,g\rangle_{H} for all f,g\in H . From now on, we shall take H=L^{2}(D) with the usual inner product. A Brownian sheet (or d -parameter Wiener process) is then defined as the process \tilde{W}=\{\tilde{W}(x)\colon x\in D\} with \tilde{W}(x)=W\left(\mathbb{I}_{[0,x]}\right). For a given function f\in L^{2}(D) and x\in D , we then define the Wiener integral of f with respect to the Brownian sheet over [0,x] as W(f\mathbb{I}_{[0,x]}) and denote it by W(f\mathbb{I}_{[0,x]})=\int_{[0,x]}f(y)\tilde{W}(dy). To simplify the notation, we will write \int_{[0,x]}f(y){W}(dy) instead of \int_{[0,x]}f(y)\tilde{W}(dy) . One can easily check (via Kolmogorov’s continuity Theorem), that the Brownian sheet and the integral of an L^{2}(D) function with respect to it have a continuous version, so, when talking about these objects, we will be talking about the continuous versions. 2.2 Weak convergence of stochastic processes To stablish the weak convergence (or convergence in law or in distribution) of a sequence of stochastic processes \{X_{n}\}_{n\in\mathbb{N}} in \mathcal{C}(D) , one usually needs to prove the convergence of the finite dimensional distributions and that it is relatively compact or, by Prokhorov’s Theorem, that it is tight. The first result we introduce (whose proof can be found in [ 1 , Lemma 2.1] ) is a result regarding the convergence in law of random variables, which will be useful when proving the convergence of the finite dimensional distributions, while the second one is a tightness criterion which generalizes Proposition 2.3, p. 95, in [ 16 ] . Lemma 2.1 . Let (F,||\cdot||) be a normed vector space and \{J^{n}\}_{n\in\mathbb{N}} and J linear maps from F to L^{1}(\Omega) such that \sup_{n\geq 1}\mathbb{E}\left[\left|J^{n}(f)\right|\right]\leq C||f|| and \mathbb{E}\left[\left|J(f)\right|\right]\leq C||f|| for any f\in F and for some positive constant C and assume there is a dense subset D of F with respect to the norm ||\cdot|| such that J^{n}(f) converges in law to J(f) as n approaches infinity for any f\in D . Then, for any f\in F , J^{n}(f) converges in law to J(f) as n approaches infinity. Proposition 2.1 . Let \{X_{n}\}_{n\in\mathbb{N}} be a sequence of random fields in D whose sample paths vanish on \{x\in D\colon x_{1}\cdot...\cdot x_{d}=0\} almost surely and for which there are constants C,p>0 and \delta>d such that \sup_{n}\mathbb{E}\left[\left|X_{n}(x)-X_{n}(z)\right|^{p}\right]\leq C\left(\sum_{i=1}^{d}|x_{i}-z_{i}|\right)^{\delta}. (1) Then, for each n\in\mathbb{N} , X_{n} possesses a continuous version and the sequence of continuous versions is tight in \mathcal{C}(D) . Proof. The existence of a continuous version follows from Kolmogorov’s continuity Theorem. Hence, from now on, we shall consider that the processes X_{n} have a.s. continuous sample paths. In order to prove tightness, we will see that the hypotheses of [ 4 , Theorem 3, p. 1665] are

Read the complete text