Limit Theorems for One-Dimensional Homogenized Diffusion Processes
Jaroslav I. Borodavka, Sebastian Krumscheid · 2025
Read Limit Theorems for One-Dimensional Homogenized Diffusion Processes by Jaroslav I. Borodavka, Sebastian Krumscheid (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-2503-06691.txt.
Introduction
Random motion smoothed across fine scales follows laws of its own, and two limit theorems pin them down here. Jaroslav Borodavka and Sebastian Krumscheid study one-dimensional diffusion with a tiny scale parameter. As the scale shrinks, the motion nears a homogenized limit. They let the time horizon grow at the same time, which breaks older proofs because many terms blow up. Careful tracking of every scale-dependent term yields an ergodic theorem and a central limit theorem.
The pair then apply the results to a distance-based estimator for the limit equation's parameters. Its long-run behavior follows from the theorems. Such groundwork supports sound inference from data gathered at fine scales, from materials to finance.
Catalog record
| CREATOR | Jaroslav I. Borodavka, Sebastian Krumscheid |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 109 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-06691
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Limit Theorems for One-Dimensional Homogenized Diffusion Processes Jaroslav I. Borodavka, Sebastian Krumscheid Abstract We present two limit theorems, a mean ergodic and a central limit theorem, for a specific class of one-dimensional diffusion processes that depend on a small-scale parameter ε \varepsilon and converge weakly to a homogenized diffusion process in the limit ε → 0 \varepsilon\rightarrow 0 . In these results, we allow for the time horizon to blow up such that T ε → ∞ T_{\varepsilon}\rightarrow\infty as ε → 0 \varepsilon\rightarrow 0 . The novelty of the results arises from the circumstance that many quantities are unbounded for ε → 0 \varepsilon\rightarrow 0 , so that formerly established theory is not directly applicable here and a careful investigation of all relevant ε \varepsilon -dependent terms is required. As a mathematical application, we then use these limit theorems to prove asymptotic properties of a minimum distance estimator for parameters in a homogenized diffusion equation. 1 Introduction In probability theory, limit theorems are of fundamental importance and provide indispensable tools for the asymptotic analysis of statistical estimators. One particular instance of such limit theorems can be found in the field of stochastic processes and, especially, diffusion processes which are solutions of a stochastic differential equation (SDE). Consider the solution X of the one-dimensional Itô SDE dX(t)=b(X(t))dt+\overline{\sigma}(X(t))d\overline{W}(t),\quad t\in[0,T], (1.1) and suppose X has an invariant measure m on \mathscr{B}(\mathbb{R}) . Then, under various assumptions, there are limit theorems in the existing literature, cf. [ 11 , 3 ] , of the type \frac{1}{T}\int_{0}^{T}h(X(t))\,dt\rightarrow\int_{\mathbb{R}}h(x)\,dm(x)=:\tau^{2},\quad T\rightarrow\infty, (1.2) for a suitable test function h , where the convergence is understood either in the mean, almost sure, or stochastic sense. Another closely related limit result is the central limit theorem (CLT) \frac{1}{\sqrt{T}}\int_{0}^{T}h(X(t))\,dt\overset{\mathcal{D}}{\longrightarrow}\mathcal{N}(0,\tau^{2}),\quad T\rightarrow\infty, (1.3) for a m -mean zero function h , where \mathcal{N}(0,\tau^{2}) is a centered normal distribution with variance \tau^{2} . The work at hand, which we will motivate now, is mainly devoted to these two types of limit theorems for a particular class of diffusion processes. Consider the solution X_{\varepsilon} of the one-dimensional SDE dX_{\varepsilon}(t)=\left[f_{0}(X_{\varepsilon}(t))+\frac{1}{\varepsilon}f_{1}\left(\frac{X_{\varepsilon}(t)}{\varepsilon}\right)\right]dt+\sigma(X_{\varepsilon}(t))\,dW(t),\quad t\in[0,T], (1.4) where \varepsilon>0 is a small-scale parameter. Following the seminal work of [ 14 , 2 ] , the later works of [ 23 , 16 , 18 ] , and possibly the most recent contribution of [ 21 ] , this puts us, under appropriate assumptions on the coefficient functions, into the setting of homogenization for SDEs. Homogenization for SDEs is a type of diffusion approximation in which the process X_{\varepsilon} converges weakly in C([0,T];\mathbb{R}) to the process X as \varepsilon\rightarrow 0 for fixed T>0 , where X is the solution to ( 1.1 ). The homogenized coefficient functions b and \overline{\sigma} in ( 1.1 ) depend on f,g,\sigma , and the invariant measure of the fast process defined through Y_{\varepsilon}:=X_{\varepsilon}/\varepsilon . In this framework, we aim to prove the following two limit theorems. Firstly, a mean ergodic theorem (MET) of the type \frac{1}{T_{\varepsilon}}\int_{0}^{T_{\varepsilon}}h_{\varepsilon}(X_{\varepsilon}(t))\,dt\rightarrow\int_{\mathbb{R}}h(x)\,dm(x)=\tau^{2},\quad\varepsilon\rightarrow 0, (1.5) for a suitable sequence of test function h_{\varepsilon} depending on \varepsilon such that h_{\varepsilon}\rightarrow h , and exploding time horizon such that T_{\varepsilon}\rightarrow\infty as \varepsilon\rightarrow 0 . Secondly, we want to prove a CLT of the type \frac{1}{\sqrt{T_{\varepsilon}}}\int_{0}^{T_{\varepsilon}}h_{\varepsilon}(X_{\varepsilon}(t))\,dt\overset{\mathcal{D}}{\longrightarrow}\mathcal{N}(0,\tau^{2}),\quad\varepsilon\rightarrow 0. (1.6) The interest for these two limit theorems stems from the increasing emergence of parametric and nonparametric estimation procedures for homogenized equations like ( 1.1 ), but with observations coming from perturbed equations like ( 1.4 ). While this is an instance of model-misspecification, the laws of the processes are similar in a weak sense when \varepsilon\rightarrow 0 ; and, yet, it turns out that many estimators cannot recover homogenized parameters from perturbed observations, as it has been repeatedly demonstrated in [ 17 , 15 , 12 , 19 ] . In the aforementioned references, as well as in [ 9 , 1 ] , the authors utilize exclusively sequential limits to prove asymptotic properties of their estimators, that is, they first let T\rightarrow\infty and then \varepsilon\rightarrow 0 , or vice versa. To the best of our knowledge, the case of taking simultaneous limits where almost every relevant quantity depends on \varepsilon has not been considered yet. This can possibly be attributed to the difficulty, that many relevant quantities and bounds explode for \varepsilon\rightarrow 0 , so that classically established results in the literature for diffusion processes cannot be directly applied. Hence, with this work, we hope to lay the theoretical groundwork for the rigorous asymptotic analysis of estimators for homogenized models when confronted with perturbed data. The main contributions of our work are the proofs of the two limit theorems; a stochastic convergence result stringently connected to the proof of the CLT, and a simple application to parameter estimation in a continuous-time setting for homogenized models under perturbed data. At this point, it is important to emphasize that the carried out proof ideas are classical, and we will take great care in crediting the authors of the original proofs; however, the \varepsilon -dependence in all relevant quantities demands nonetheless a careful investigation of the existing proofs. The main content of the article starts with section 2 , which is itself divided into three subsections. The first subsection deals with the precise definition of the framework, the assumptions under which we will work, and the rigorous proof of the MET, which is stated in Theorem 2.9 . Subsection 2.2 is devoted to the proof of the required stochastic convergence result contained in Corollary 2.15 . The last subsection 2.3 presents the proof of the CLT whose statement can be found in Theorem 2.18 . Finally, an exemplary application to parameter estimation is discussed in section 3 , with the main result being Proposition 3.2 . In the conclusion, we summarize the main contributions again, and we also give remarks on potential extensions of the results to more general test functions and to the multidimensional case. 2 On \varepsilon -dependent limit theorems 2.1 An \varepsilon -dependent mean ergodic theorem Consider a probability space \left(\Omega,\mathcal{F},\mathbb{P}\right) with a filtration \left(\mathcal{F}_{t}\right)_{t\in[0,\infty)} that satisfies the usual conditions, and a one-dimensional Brownian motion W:=\left(W(t),\mathcal{F}_{t}\right)_{t\in[0,\infty)} on that probability space. Assume further that there is another one-dimensional Brownian motion \overline{W}:=\left(\overline{W}(t),\mathcal{F}_{t}\right)_{t\in[0,\infty)} that is independent of W and lives on the same probability space. This construction is feasible through a suitable extension of the original probability space. Fix x_{0}\in\mathbb{R} once and for all. Consider the following one-dimensional stochastic differential equation (SDE) in \mathbb{R} , depending on a small parameter \varepsilon>0 , dX_{\varepsilon}(t)=b_{\varepsilon}(X_{\varepsilon}(t))dt+\sigma(X_{\varepsilon}(t))dW(t),\quad t>0,\quad X_{\varepsilon}(0)=x_{0}, (2.1) with suitable Borel-measurable functions b_{\varepsilon}\colon\mathbb{R}\rightarrow\mathbb{R} , \sigma\colon\mathbb{R}\rightarrow\mathbb{R} , and assume, for the moment, that there exists a unique, strong solution to this SDE on the given probability space. We suppose a setting where X_{\varepsilon} converges weakly to another stochastic process X in C([0,T];\mathbb{R}) as \varepsilon\rightarrow 0 . Here, X is the unique, strong solution to the stochastic differential equation dX(t)=b(X(t))dt+\overline{\sigma}(X(t))d\overline{W}(t),\quad t>0,\quad X(0)=x_{0}, (2.2) with Borel-measurable functions b\colon\mathbb{R}\rightarrow\mathbb{R} and \overline{\sigma}\colon\mathbb{R}\rightarrow\mathbb{R} . Note that \sigma\neq\overline{\sigma} in general. Assume further that the following scale functions exist \displaystyle f_{\varepsilon}(x):=\int_{0}^{x}\exp\left(-\int_{0}^{z}\frac{2b_{\varepsilon}(y)}{\sigma(y)^{2}}\,dy\right)\,dz,\quad f(x):=\int_{0}^{x}\exp\left(-\int_{0}^{z}\frac{2b(y)}{\overline{\sigma}(y)^{2}}\,dy\right)\,dz,\quad x\in\mathbb{R}. (2.3) These functions are harmonic with respect to the differential operators \displaystyle\mathcal{A}_{\varepsilon}:=b_{\varepsilon}\partial_{x}+\frac{1}{2}\sigma^{2}\partial_{xx},\quad\mathcal{A}:=b\partial_{x}+\frac{1}{2}\overline{\sigma}^{2}\partial_{xx}, (2.4) respectively, that is \mathcal{A}_{\varepsilon}f_{\varepsilon}=0 and \mathcal{A}f=0 . Setting \xi_{\varepsilon}:=f_{\varepsilon}\circ X_{\varepsilon} and \xi:=f\circ X and applying the (generalized) Itô formula yields the transformed stochastic differential equations