[Paper Review] Effect of Noise on Front Propagation in Reaction-Diffusion equations of KPP type

Carl Mueller, Leonid Mytnik+1 more|ArXiv.org|Feb 19, 2009

Stochastic processes and statistical mechanics19 references4 citations

TL;DR

This paper investigates the effect of additive space-time white noise on reaction-diffusion equations of KPP type, proving the Brunet-Derrida conjecture that the front speed in the stochastic KPP equation converges asymptotically to $2 - \pi^2 |\log \epsilon^2|^{-2}$, up to a logarithmic correction term of order $(\log|\log\epsilon|)|\log\epsilon|^{-3}$, under mild conditions on the noise and initial data. The analysis combines stochastic PDE theory, coupling arguments, and large deviation estimates to establish the speed shift due to noise.

ABSTRACT

We consider reaction-diffusion equations of KPP type in one spatial dimension, perturbed by a Fisher-Wright white noise, under the assumption of uniqueness in distribution. Examples include the randomly perturbed Fisher-KPP equations $ \partial_t u = \partial_x^2 u + u(1-u) + ε\sqrt{u(1-u)}\dot W, $ and $ \partial_t u = \partial_x^2 u + u(1-u) + ε\sqrt{u}\dot W, $ where $\dot W= \dot W(t,x)$ is a space-time white noise. We prove the Brunet-Derrida conjecture that the speed of traveling fronts is asymptotically $ 2-π^2 |\log ε^2|^{-2} $ up to a factor of order $ (\log|\logε|)|\logε|^{-3}$.

Motivation & Objective

• To rigorously establish the Brunet-Derrida conjecture on the speed of traveling fronts in stochastic KPP equations perturbed by space-time white noise.
• To analyze the asymptotic behavior of front propagation in reaction-diffusion equations with multiplicative noise, particularly when the noise intensity $\epsilon$ is small.
• To prove the existence and uniqueness of random traveling front solutions in the stochastic setting, and to characterize the limiting front speed as $\epsilon \to 0$.
• To develop a stochastic coupling framework that compares the stochastic solution to a deterministic reference solution, enabling precise estimates on the front location and speed.
• To extend the classical KPP theory to the stochastic case by establishing convergence to a random traveling front and deriving the leading-order correction to the front speed due to noise.

Proposed method

• The authors study the stochastic PDE $\partial_t u = \partial_x^2 u + f(u) + \epsilon \sigma(u) \dot{W}$ on $\mathbb{R} \times \mathbb{R}_+$, where $\dot{W}$ is space-time white noise and $f(u) = u(1-u)$, with $\sigma^2(u)$ satisfying Lipschitz and positivity conditions on $[0, u^*]$.
• They define mild solutions via the integral equation involving the heat kernel $G(t,x)$, and use the stochastic integral representation to control the noise term via Gaussian chaos estimates.
• A key technique is the construction of a coupling between the stochastic solution $u$ and a deterministic reference solution $\varrho$, using a comparison argument based on the Lipschitz property of $f$ and the noise term's structure.
• The proof relies on a localization argument in a moving frame, where the front is tracked relative to the position $r(t) = \sup\{x : u(t,x) > 0\}$, and the process $\tilde{u}(t,x) = u(t, x + r(t))$ is shown to converge to a stationary law.
• Large deviation estimates are used to bound the probability that the stochastic solution exceeds a deterministic reference profile, with bounds depending on $\epsilon$ and $\gamma$, a small parameter controlling the size of the domain.
• The authors derive precise estimates on the noise-induced shift in front speed by analyzing the solution in a moving frame and using exponential tail bounds on the stochastic integral $Z(t,x)$.

Experimental results

Research questions

• RQ1 What is the asymptotic speed of the traveling front in the stochastic KPP equation with additive space-time white noise as the noise intensity $\epsilon \to 0$?
• RQ2 Does the Brunet-Derrida conjecture hold for the stochastic KPP equation with multiplicative noise of the form $\epsilon \sqrt{u(1-u)} \dot{W}$?
• RQ3 How does the presence of noise affect the front propagation speed compared to the deterministic KPP case, where the minimal speed is $2$?
• RQ4 Can the speed shift due to noise be quantified with explicit asymptotic corrections, and what is the order of magnitude of the correction term?
• RQ5 Is there a unique non-degenerate stationary distribution for the process viewed in the moving frame of the front, and does the front speed converge almost surely?

Key findings

• The front speed in the stochastic KPP equation converges asymptotically to $2 - \pi^2 |\log \epsilon^2|^{-2}$ as $\epsilon \to 0$, confirming the Brunet-Derrida conjecture.
• The correction to the front speed is of order $|\log \epsilon|^{-2}$, with a secondary logarithmic correction of order $(\log|\log \epsilon|)|\log \epsilon|^{-3}$.
• The existence and uniqueness of a random traveling front solution is established, with the front position $r(t)$ satisfying $\lim_{t \to \infty} t^{-1} r(t) = v_\epsilon$ almost surely.
• The stochastic solution remains close to the deterministic solution in a moving frame, with high probability, via coupling and large deviation estimates on the noise term.
• The proof relies on precise bounds on the stochastic integral $Z(t,x)$ and its impact on the solution, using exponential tail estimates and heat kernel localization.
• The result holds under general conditions on $f$ and $\sigma^2$, including the key examples $f(u) = u(1-u)$ and $\sigma^2(u) = u(1-u)$ or $\sigma^2(u) = u$, with initial data in $\mathcal{C}_{\rm exp}$ satisfying integrability and positivity conditions.