Lyapunov exponents of linear stochastic functional differential equations driven by semimartingales. I: The multiplicative ergodic theory (Q1908228)

An almost sure Lyapunov spectrum consisting of a discrete non-random set of Lyapunov exponents \(\{\lambda_i\}^\infty_{i=1}\subset\mathbb{R}\cup\{-\infty\}\) is shown for a linear stochastic functional differential equation (FDE): \[ dx(t)=\Biggl\{\int_{[-r,0]}\mu(t)(ds)x(t+s)\Biggr\}dt+dN(t)\int^0_{-r} K(t)(s)x(t+s)ds+ dL(t)x(t^-), \] \[ t>0,\quad x(0)=v\in\mathbb{R}^n,\quad x(s)=\eta(s),\quad -r\leq s\leq 0,\quad r>0, \] \[ (x(0),x_0)=(v,\eta)\in M_2:=\mathbb{R}^n\times L^2([-r,0],\mathbb{R}^n). \] Here, the helix semimartingales \(N\), \(L=M+V\) are \(n\times n\)-matrix valued processes with jointly stationary increments, \(M\) is a continuous helix-local martingale and \(V\) is a helix whose sample paths are all right continuous and of bounded variation on compact sets of \(\mathbb{R}^+\). On \(M_2\), a robust and compactifying version of the stochastic flow \(X\) of this FDE is constructed so that the cocycle property and Ruelle integrability condition [\textit{D. Ruelle}, Publ. Math., Inst. Hautes Étud. Sci. 50, 27-58 (1979; Zbl 0426.58014) and Ann. Math., II. Ser. 115, 243-290 (1982; Zbl 0493.58015)] are read off from the random integral equation equivalent to the FDE. Then the Ruelle infinite-dimensional multiplicative ergodic theorem, the existence of the Lyapunov spectrum and a saddle-point property in the hyperbolic case are established.