Ergodicity for stochastic equations of Navier-Stokes type (Q2078114)

Let \(W\) be a one-dimensional Wiener process. Then the equation \begin{align*} \textrm{d}u_1&=[-\lambda_1u_1-(u_2^2+u_3^2)+\kappa]\,\textrm{d}t+\sigma\,\textrm{d}W(t) \\ \textrm{d}u_2&=[-\lambda_2u_2+u_1u_2]\,\textrm{d}t \\ \textrm{d}u_3&=[-\lambda_3u_3+u_1u_3]\,\textrm{d}t \end{align*} has an invariant probability measure for any choice of the real parameters \(\lambda_1,\lambda_2,\lambda_3,\kappa\) and \(\sigma\). If \(\lambda_1>0\) then there exists a Gaussian invariant probability measure (denote it by \(\theta\)). If \(\lambda_1>0\) and \(\kappa<\lambda_1\min\,\{\lambda_2,\lambda_3\}\) then \(\theta\) is the unique invariant probability measure and it is stochastically stable (i.e. if \(u^v\) denotes the solution with an initial datum \(v\in\mathbb R^3\) then the laws of \(u^v(t)\) converge weakly to \(\theta\) as \(t\to\infty\), for every initial datum \(v\in\mathbb R^3\)). If \(\lambda_1>0\) and \(\kappa>\lambda_1\min\,\{\lambda_2,\lambda_3\}\) then there exists another invariant probability measure \(\mu\) distinct from \(\theta\).