[Paper Review] Convergence of Passive Scalars in Ornstein-Uhlenbeck Flows to Kraichnan's Model

Albert Fannjiang|ArXiv.org|Sep 5, 2002

Stochastic processes and financial applications6 references4 citations

TL;DR

This paper establishes the convergence of passive scalar fields in Ornstein-Uhlenbeck velocity flows to Kraichnan’s model under a white-noise limit, showing that the limiting scalar dynamics correspond to a Kraichnan model with enhanced spatial regularity characterized by a Hurst exponent $η = \alpha + \beta - 1$. The convergence holds in distribution under specific scaling conditions on the correlation time and diffusivity, with precise moment conditions ensuring the vanishing of correction terms.

ABSTRACT

We prove that the passive scalar field in the Ornstein-Uhlenbeck velocity field with wave-number dependent correlation times converges, in the white-noise limit, to that of Kraichnan's model with higher spatial regularity.

Motivation & Objective

• To establish a rigorous connection between the passive scalar dynamics in finite-correlation-time Ornstein-Uhlenbeck (OU) velocity fields and Kraichnan’s white-in-time velocity model.
• To identify the limiting spatial regularity of the scalar field in the white-noise limit, showing it corresponds to a Kraichnan model with increased spatial Hurst exponent $\eta = \alpha + \beta - 1$.
• To determine the precise scaling conditions on the correlation time $\varepsilon^{-2}a^{-1}|k|^{-2\beta}$ and diffusivity $\kappa(\varepsilon)$ under which convergence occurs, especially in the limit $\ell_1 \to 0$.
• To resolve the issue of Stratonovich correction terms and non-Lipschitz velocity fields in the compressible case by identifying sufficient conditions for convergence in distribution.

Proposed method

• Uses a spectral representation of the two-time structure function for the OU velocity field with wave-number-dependent correlation time $a^{-1}|k|^{-2\beta}$, ensuring spatial power spectrum $\mathcal{E}(\alpha,k)$ with $1 < \alpha < 2$.
• Applies the martingale problem formulation to characterize the limiting scalar dynamics, replacing the Stratonovich SDE with a well-defined martingale problem in the limit.
• Employs Itô-Stratonovich correction analysis and Gaussian chaos estimates to control the higher-order terms arising from the non-Markovian OU flow.
• Derives bounds on the residual terms $R^\varepsilon_3$, $R^\varepsilon_2$, and $R^\varepsilon_1$ using moment estimates and spectral decay, showing their vanishing under specific scaling of $\varepsilon$, $\ell_1$, and $\kappa(\varepsilon)$.
• Imposes conditions on $\alpha + 2\beta$ and the behavior of $\kappa(\varepsilon)$, $\ell_1(\varepsilon)$ to ensure convergence in the space $D([0,t_0); L^\infty_{w^*} \cap L^2_w)$.
• Uses the Skorokhod topology on càdlàg processes with values in weak* $L^\infty$ and weak $L^2$ to define convergence in distribution of the scalar field $T_t^\varepsilon$.

Experimental results

Research questions

• RQ1 Under what conditions does the passive scalar field in an Ornstein-Uhlenbeck velocity field converge to Kraichnan’s model in the white-noise limit?
• RQ2 How does the spatial regularity of the limiting scalar field depend on the parameters $\alpha$ and \beta of the OU velocity field’s correlation structure?
• RQ3 What are the precise scaling conditions on $\varepsilon$, $\ell_1$, and $\kappa(\varepsilon)$ that ensure convergence when the viscous cutoff $\ell_1$ vanishes?
• RQ4 Why does the Stratonovich correction term in the limiting SDE remain well-defined only when $\alpha + \beta > 3/2$?
• RQ5 How do the moment estimates of the velocity field’s increments control the convergence of the scalar field’s dynamics?

Key findings

• The solution $T_t^\varepsilon$ of the advection-diffusion equation with OU velocity field converges in distribution to the solution of Kraichnan’s model with spatial power spectrum $2a^{-1}\mathcal{E}(\alpha + \beta, k)$, corresponding to a Hurst exponent $\eta = \alpha + \beta - 1$.
• For the case $\ell_1 = \ell_1(\varepsilon) \to 0$, convergence holds under conditions such as $\alpha + 2\beta > 4$, or $\alpha + 2\beta = 4$ with $\lim_{\varepsilon \to 0} \kappa \varepsilon^2 \sqrt{\log(1/\ell_1)} = 0$, ensuring the vanishing of correction terms.
• When $\alpha + 2\beta < 3$, convergence requires $\lim_{\varepsilon \to 0} \varepsilon \ell_1^{\alpha + 2\beta - 3} = 0$, and for $\alpha + 2\beta = 3$, the condition $\lim_{\varepsilon \to 0} \varepsilon \sqrt{\log(1/\ell_1)} = 0$ is sufficient.
• The most singular correction terms are controlled via Gaussian moment estimates, with bounds scaling as $\varepsilon \ell_1^{\alpha + 2\beta - 3}$ or $\kappa \varepsilon^2 \ell_1^{\alpha + 2\beta - 4}$, which vanish under the stated conditions.
• In the limit $\varepsilon, \ell_1 \to 0$, the limiting scalar process satisfies the SDE $dT_t = \left(\frac{\kappa_0}{2}\Delta + \frac{1}{a}\bar{\mathcal{B}}\right)T_t dt + \sqrt{2}a^{-1/2} \nabla T_t \circ d\bar{W}_t^{(1)}$, with $\bar{\mathcal{B}}$ defined via the limit of $A_3^{(1)}$.
• The convergence is established in the space $D([0,t_0); L^\infty_{w^*} \cap L^2_w)$, ensuring weak* and weak $L^2$ convergence of the scalar field over finite time intervals.