Abstract
This paper investigates wave propagation in randomly layered media characterized by long-range correlations (LRC). Unlike standard random media, which typically cause pulses to spread and blur, LRC media lead to an asymptotic travel time described by complex stochastic processes. We introduce a general framework to study the asymptotic transmitted pulse, proving that the resulting time-shift can be described in terms of non-Gaussian and/or multifractal processes, while the pulse itself retains its original shape.
Introduction
In classical wave propagation theory, researchers frequently rely on "mixing" or short-range correlation models. In these environments, random fluctuations in the medium—such as density or pressure—average out quickly over distance. This leads to the well-known O’Doherty–Anstey result: as a pulse travels, it undergoes deterministic spreading (blurring) and experiences a random delay following standard Brownian motion.
However, many real-world environments, including the Earth's crust and turbulent atmospheres, exhibit long-range correlations (LRC). In these settings, fluctuations at one point are statistically linked to points at significant distances. Previous research indicated that in specific LRC cases, the pulse does not blur but instead shifts in time according to fractional Brownian motion. This paper generalizes these findings, exploring what happens when the medium is non-Gaussian or multifractal—where the "roughness" of the medium varies as the wave progresses.
Methodology
The study employs an asymptotic analysis of one-dimensional (1D) acoustic wave equations (Euler equations). The properties of the medium are modeled as a centered random process with specific scaling parameters ($\epsilon$).
A critical component of the mathematical proof is the application of Rough Paths Theory, specifically Terry Lyons’s Continuity Theorem. This framework allows the authors to handle the convergence of stochastic integrals that are significantly "rougher" than standard Brownian motion. By decomposing the wave into right- and left-going components and analyzing the propagator—the operator responsible for moving the wave through the medium—the authors demonstrate that the pulse's evolution is dominated by the asymptotic travel time.
Key Insights and Theoretical Framework
Shape Preservation
The most significant finding is the "dramatic contrast" between short-range and long-range media. In LRC media, the pulse retains its initial shape $f$. The randomness of the medium does not cause the signal to dissipate or spread; instead, the stochasticity is entirely "pushed" into the travel time.
The Travel-Time Link
The paper establishes a direct mathematical link (Theorem 3.1) between the statistical properties of the medium and the resulting time-shift. By identifying the correlation structure of the medium, one can predict whether the pulse delay will follow a Gaussian, Hermite, or multifractal process.
Non-Gaussianity and Multifractality
If the medium's fluctuations are non-Gaussian—modeled via Hermite polynomials—the pulse shift follows a Hermite process. This represents a major departure from the standard Gaussian "bell curve" assumptions. Furthermore, the authors introduce a model where the correlation decay rate varies along the propagation direction, leading to a multifractional time-shift that reflects the non-homogeneity of the environment.
Formal Results
- Theorem 3.1 (General Convergence): Proves that under general LRC assumptions, the transmitted pulse $a(Z, s)$ converges to $f(s - \frac{1}{2}V(Z))$, where $V(Z)$ represents the asymptotic travel time.
- Theorem 4.1 (Non-Gaussian Media): Defines the specific shift for non-Gaussian media using the $K$-th Hermite process.
- Theorem 6.1 (Multifractal Framework): Combines non-Gaussianity and multifractality into a single framework, providing a model for the most complex media types.
Qualitatively, these results confirm that the "roughness" of a medium translates directly into the "roughness" of the signal's arrival time without sacrificing signal clarity.
Discussion and Limitations
While this framework provides a rigorous basis for understanding LRC media, it is currently limited to one-dimensional (layered) models. While 1D models are highly relevant for certain geophysical applications, extending this to 2D or 3D environments remains a significant mathematical challenge. Additionally, the results rely on specific asymptotic scaling ($\epsilon \to 0$); the behavior of these waves in "intermediate" regimes requires further numerical investigation.
This framework offers promising applications for modeling the multi-scale crust of the sedimentary Earth and turbulent atmospheres, suggesting that the next step in this research involves applying these theorems to empirical physical sensor data.
Summary updated