Global attractors for 3-dimensional stochastic Navier-Stokes equations (Q702417)

The approach of \textit{G. R. Sell} [J. Dyn. Differ. Equations 8, 1--33 (1996; Zbl 0855.35100)] is extended to stochastic Navier-Stokes equations of the form \[ du= \bigl(\nu\Delta u-\langle u,\nabla\rangle u +f(u)-\nabla p\bigr)\,dt+g(u)\,dW(t), \quad \text{div}\,u=0,\tag{1} \] on a bounded domain \(D\subset\mathbb R^3\), where \(W\) is a one-dimensional Wiener process, and \(f\) and \(g\) are continuous and satisfy linear growth conditions with rates dominated by the dissipativity of the generator given by the linear part. A certain space of \(L^2\) stochastic processes with values in \(H\), where \(H\) is the standard Hilbert space associated with (1), is taken as state space, on which a semiflow, arising from a skew product of the time shift and a measure preserving semiflow on the probability space, is acting. An attractor is defined to be a strictly invariant and closed (with respect to the \(L^2\) topology) set in the state space, such that its law is a classical attractor for the induced action on the space of probability measures on the function space, and such that every \(L^2\) bounded set is attracted in a certain sense, weaker than the usual one. Notably, the attractor on the space of stochastic processes is not assumed to be compact. The main theorem states existence of both a filtered probability space and a class of processes \(X\) in the state space, consisting of adapted weak solutions of (1) with a semiflow leaving it invariant, such that for any \(L^2\) initial condition there is a solution in \(X\), and such that an attractor exists in \(X\). The construction uses non-standard analysis, in particular establishing existence of \(X\) over a filtered Loeb probability space. It is shown that the attractor consists of solutions which can be extended to the whole time axis, and that it fails to be compact in general, where the latter is essentially due to the richness of Loeb probability spaces.