On the structure of attractors and invariant measures for a class of monotone random systems (Q3155189)

Let \(V\) be a real separable Banach space with a cone \(V_+ \subset V\). A random dynamical system (RDS) with time \({\mathbb T}_+\) (\({\mathbb T}\) is \({\mathbb Z}\) or \({\mathbb R}\)), taking values in an admissible set \(X\subset V\), is a pair \((\theta ,\phi )\), where \(\theta\) is a family of measure preserving transformations \(\theta (t), t\in {\mathbb T}\), of a probability space \((\Omega ,{\mathcal F},P)\) with \(\theta (0)=\text{id}\), \(\theta (t)\circ \theta (s)=\theta (t+s)\), and \(\phi\) is a cocycle over \(\theta\) of continuous mappings of \(X, \phi :{\mathbb T}_+ \times \Omega \times X\) such that the mapping \(x\mapsto\phi (t,\omega )x\) is continuous for \(t\geq 0, \omega \in \Omega\) and \(\phi (0,\omega )=\text{id}\), \(\phi (t+s, \omega )= \phi (t,\theta (s)\omega )\circ \phi (s,\omega )\), \(t,s\geq 0, \omega \in \Omega\). An RDS \((\theta ,\phi )\) is order preserving if \(x\leq y\) implies \(\phi (t,\omega )x\leq \phi (t,\omega )y\), \(t\geq 0,\omega \in \Omega\). The main result of the paper is the following: If \((\theta ,\phi )\) is an order preserving RDS on some admissible \(X\subset V\) and there exists a probability measure \(\pi\) on Borel sets of \(X\) such that \(\lim_{t\rightarrow \infty}{\mathbb E}f(\phi (t,\cdot ),x)=\int_X f(y)\pi (dy)\), \(x\in X, f\in C_b (X)\), then there exists a unique invariant measure \(\mu\) for the RDS \((\theta ,\phi )\). This measure is generated by some stationary process. If the cone \(V_+\) of nonnegative elements of \(V\) is normal then this stationary process is a global random attractor with respect to convergence in probability. As examples, one-dimensional ordinary and retarded stochastic differential equations and a class of parabolic stochastic partial differential equations are considered.