Passive tracer in a slowly decorrelating random flow with a large mean

Abstract

We consider the movement of a particle advected by a random flow of the form $\vv+\delta \bF(\vx)$, with $\vv\in\R^d$ a constant drift, $\bF(\vx)$ -- the fluctuation -- given by a zero mean, stationary random field and $\delta\ll 1$ so that the drift dominates over the fluctuation. The two-point correlation matrix $\bR(\vx)$ of the random field decays as $|\vx|^{2\alpha-2}$, as $|\vx|\to+\infty$ with $\alpha<1$. The Kubo formula for the effective diffusion coefficient obtained in \cite{kp79} for rapidly decorrelating fields diverges when $1/2\le\alpha<1$. We show formally that on the time scale $\delta^{-1/\alpha}$ the deviation of the trajectory from its mean $\by(t)=\vx(t)-\vv t$ converges to a fractional Brownian motion $B_\alpha(t)$ in this range of the exponent $\alpha$. We also prove rigorously upper and lower bounds which show that $\E[|\by(t)|^2]$ converges to zero for times $t\ll\delta^{-1/\alpha}$ and to infinity on time scales $t\gg \delta^{-1/\alpha}$ as $\delta\to 0$ when $\alpha\in(1/2,1)$. On the other hand, when $\alpha<1/2$ non-trivial behavior is observed on the time-scale $O(\delta^{-2})$.