On the laws of the Oseledets spaces of linear stochastic differential equations (Q4357558)

Oseledets' multiplicative ergodic theorem describes the invariant subspaces (Oseledets spaces) of linear random dynamical systems. The paper deals with these spaces for stochastic differential equations in Stratonovich form as \(dx_t= \sum^m_{i= 0} A_ix_t\circ dW^i_t\), where \(A_0,\dots, A_m\) are \(d\times d\) real matrices, \((W^1,\dots, W^m)\) is a canonical two-sided \(m\)-dimensional Wiener process, and \(dW^0_t:= dt\). The Oseledets (random) subspaces \(E_1,\dots, E_r\), corresponding to the Lyapunov numbers \(\lambda_1>\cdots>\lambda_r\), can be described by two flags of nested random subspaces \(V^+_1\supset\cdots\supset V^+_r\), \(V^-_1\subset\cdots\subset V^-_r\), by \(E_i= V^+_i\cap V^-_i\), and the two flags are independent. The author deals with the forward \((+)\) flag, sufficient to obtain smoothness results for the Oseledets spaces. Subspaces are identified with orthogonal projectors, and the generic \(k\)-dimensional Oseledets subspace is studied by identifying it with the stationary state of a flow of diffeomorphisms on the Grassmannian \(G_k(d)\) of \(k\)-dimensional subspaccs. This flow is obtained by decomposing the original random dynamical system in a radial and an angular part. Malliavin calculus is then applied to prove, under suitable assumptions, that the law of the \(k\)-dimensional subspace has a \(C^\infty\)-density with respect to the Riemannian volume on \(G_k(d)\). For the entire collection see [Zbl 0870.00040].