Invariant measures related with Poisson driven stochastic differential equation. (Q2574577): Difference between revisions

A stochastic differential equation \[ d\xi (t) = a(\xi (t))\, dt + \int _ \Theta \sigma (\xi (t),\theta )N(dt,d \theta ) \tag{1} \] in \(\mathbb R^ {m}\) is studied. It is supposed that \(N\) is a Poisson random measure with the characteristic measure \(n\), \(a:\mathbb R^ {m}\to \mathbb R^ {m}\) is Lipschitz continuous, and \(\sigma : \mathbb R^ {m} \times \Theta \to \mathbb R^ {m}\) is a measurable mapping satisfying \(\| \sigma (x,\cdot )-\sigma (y,\cdot )\| _ {L^ 2(n)} \leq L_ \sigma \| x-y\| \) for some \(L_ \sigma <\infty \) and all \(x,y\in \mathbb R^ {m}\). Under these assumptions the equation (1) defines a Markov process with a transition semigroup \((P_ {t})\). Moreover, trajectories of any solution \(\xi \) to (1) are piecewise continuous, with a countable number of jumps at times \(0<t_ 1<t_ 2<\cdots \nearrow \infty \). There exists a Markov operator \(P\) such that the law of \(\xi (t_ {k})\) is \(P^ {k}\mu \), provided \(\mu \) is the law of \(\xi (0)\). The authors prove that every invariant measure for \(P\) corresponds to an invariant measure for \((P_ {t})\) and vice versa, and use this result to find sufficient conditions for existence of invariant measures. Finally, they apply the result to estimate Hausdorff and concentration dimensions of measures invariant for \((P_ {t})\) or \(P\).