Invariant measure selection by noise: An Example

Jonathan C. Mattingly,Etienne Pardoux

DOI: https://doi.org/10.48550/arXiv.1403.3593

2014-03-14

Abstract:We consider a deterministic system with two conserved quantities and infinity many invariant measures. However the systems possess a unique invariant measure when enough stochastic forcing and balancing dissipation are added. We then show that as the forcing and dissipation are removed a unique limit of the deterministic system is selected. The exact structure of the limiting measure depends on the specifics of the stochastic forcing.

Probability,Dynamical Systems,Fluid Dynamics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in a deterministic system, when there are multiple invariant measures, how to select a unique invariant measure by adding appropriate stochastic forcing and dissipation. Specifically, the author has studied how the unique invariant measure of a deterministic system with two conserved quantities is selected after sufficient stochastic forcing and corresponding dissipation are introduced. As the forcing and dissipation gradually weaken to zero, the system will tend to a unique limit state, and the specific structure of this limit state depends on the specific form of the stochastic forcing.

Mathematically, this problem can be described as follows:

- The evolution equation of the deterministic system is:
\[
\dot{\xi}_t = B(\xi_t, \xi_t)
\]
where $\xi_t=(X_t, Y_t, Z_t)\in\mathbb{R}^3$, and $B$ is a quadratic nonlinear term that satisfies the conservation condition:
\[
B(\xi, \xi)\cdot\xi = 0
\]

- The evolution equation of the system after adding stochastic forcing and dissipation is:
\[
\dot{\xi}^\varepsilon_t = B(\xi^\varepsilon_t, \xi^\varepsilon_t)-\varepsilon\xi^\varepsilon_t+\sqrt{\varepsilon}\sigma\dot{W}_t
\]
where $\varepsilon > 0$ is the noise intensity parameter, $\sigma$ is the noise matrix, and $W_t$ is the standard Brownian motion.

The core problem of the paper is to explore whether the invariant measure of the system will converge to a unique limit measure when $\varepsilon\rightarrow0$, and whether this limit measure can provide important information about the original deterministic system.