Asymptotic Behavior on a Linear Self-Attracting Diffusion Driven by Fractional Brownian Motion

Litan Yan, Wu Xue, Xiaoyu XiaFractal and Fractional 2022 · DOI

TL;DR

This paper investigates the long-term behavior of a specific type of stochastic process called a "linear self-attracting diffusion" driven by fractional Brownian motion. The authors demonstrate that this process, under certain conditions, converges to a normal distribution and establish laws of large numbers for its behavior over extended time scales. This work contributes to understanding complex stochastic systems with path-dependent interactions, relevant in fields like polymer physics and financial modeling.

What Problem Does This Paper Address?

The paper tackles the challenge of understanding the asymptotic (long-term) behavior of a class of stochastic differential equations (SDEs) that exhibit "self-attraction." These SDEs are characterized by a drift term that depends on the entire past trajectory of the process itself. Specifically, the authors focus on a linear self-attracting diffusion driven by fractional Brownian motion (fBm), a generalization of standard Brownian motion that allows for long-range dependence. While the behavior of similar diffusions driven by standard Brownian motion has been studied, extending these results to the more complex fBm setting, which lacks the convenient martingale and Markov properties, presents significant analytical hurdles. The core problem is to characterize how these processes behave as time goes to infinity, particularly when the interaction is attractive (pulling the process back to previously visited regions).

Methods

The authors employ a combination of advanced stochastic analysis techniques. Key methods include:

  • Variation of Constants Method: This is used to derive an explicit form for the solution of the SDE, expressing it in terms of the driving fractional Brownian motion and a kernel function.
  • Malliavin Calculus: This powerful tool is utilized to analyze the properties of random variables derived from the fBm, particularly their differentiability and variance. It's crucial for dealing with the non-Markovian nature of fBm.
  • Skorohod Integral: This is a generalization of the Itô integral, necessary for integrating with respect to fractional Brownian motion.
  • Convergence Theorems: Standard convergence theorems (e.g., Slutsky's theorem, dominated convergence theorem, Borel-Cantelli Lemma) are used to establish the asymptotic results.
  • Analysis of Kernel Functions: The properties of the derived kernel function, $h_{\theta,\sigma}(t,s)$, are extensively studied to understand its behavior as $t \to \infty$.

The novelty lies in applying these techniques to the specific linear self-attracting diffusion driven by fBm, which requires careful adaptation due to the properties of fBm.

Datasets

Not discussed in the paper. This paper focuses on theoretical analysis of a stochastic model rather than empirical data.

Key Insights and Ideas

  • Convergence to Normality: A significant finding is that the scaled process $t^{-\sigma/\theta} X_t^H$ converges in distribution to a normal random variable with a specific variance $\zeta(H, \theta) = H\theta^{-2H}\Gamma(2H)$. This suggests that despite the complex path-dependent interactions and the long-range dependence of fBm, the process exhibits a predictable Gaussian behavior in the long run.
  • Role of Parameters: The parameters $\theta$ (strength of self-attraction) and $H$ (Hurst index of fBm) play crucial roles in determining the asymptotic behavior. The condition $\theta > 0$ is essential for the established results.
  • Strong Laws of Large Numbers: The paper establishes two strong laws of large numbers for related quantities, providing insights into the average behavior of the process over long time intervals. This is important for understanding the ergodicity of such systems.
  • Analytical Tractability of fBm: The work demonstrates that even with the analytical challenges posed by fractional Brownian motion (lack of martingale and Markov properties), it's possible to derive precise asymptotic results using advanced stochastic calculus tools like Malliavin calculus.

Key Results

  • Theorem 1: The solution $X_t^H$ converges in $L^2$ and almost surely to a specific random variable $X = \int_0^\infty h_{\theta,\sigma}(s)dB_s^H + \nu \int_0^\infty h_{\theta,\sigma}(s)ds$ as $t \to \infty$.
  • Theorem 2: The scaled process $t^{-\sigma/\theta} (X_t^H - X_\infty^H)$ converges in distribution to a normal random variable $N(0, \zeta(H, \theta))$ as $t \to \infty$, where $\zeta(H, \theta) = H\theta^{-2H}\Gamma(2H)$.
  • Theorem 3: Two strong laws of large numbers are established for the integral of the squared process and the integral of the process itself, showing convergence to specific constants related to $\theta$, $H$, and $\nu$.

Limitations and Future Work

  • Parameter Restrictions: The results are primarily derived for $\theta > 0$ and $\frac{1}{2} < H < 1$. The paper notes that the behavior for $\theta \le 0$ or $H \le \frac{1}{2}$ might differ significantly and is not fully covered.
  • Generalizations: The authors mention that they will consider a more general form of the self-attracting diffusion in a future paper, suggesting that the current work is a specific case.
  • Specific Kernel Function: The analysis heavily relies on the properties of the kernel function $h_{\theta,\sigma}(t,s)$. While its asymptotic behavior is studied, understanding its implications for other types of self-attracting diffusions could be a future direction.
  • Higher Dimensions: The current analysis is for a 1-dimensional process. Extending these results to higher dimensions would be a natural next step.

Summary updated

Abstract

Let BH={BtH,t≥0} be a fractional Brownian motion with Hurst index 12≤H&lt;1. In this paper, we consider the linear self-attracting diffusion: dXtH=dBtH+σXtHdt−θ∫0tXsH−XuHdsdt+νdt with X0H=0, where θ&gt;0 and σ,ν∈R are three parameters. The process is an analogue of the self-attracting diffusion (Cranston and Le Jan, Math. Ann.303 (1995), 87–93). Our main aim is to study the large time behaviors. We show that the solution t−σθHXtH−X∞H converges in distribution to a normal random variable, as t tends to infinity, and obtain two strong laws of large numbers associated with the solution XH.
BibTeX
@inproceedings{yan2022asymptotic,
  author = {Litan Yan and Wu Xue and Xiaoyu Xia},
  title = {{Asymptotic Behavior on a Linear Self-Attracting Diffusion Driven by Fractional Brownian Motion}},
  booktitle = {Fractal and Fractional},
  year = {2022},
  doi = {10.3390/fractalfract6080454},
}
Figure