Delay differential equations driven by Lévy processes: stationarity and Feller properties (Q855685)

The authors study a stochastic delay differential equation \[ \begin{align*}{ dX(t)&=\left(\,\int_{[-\alpha,0]}X(t+s)\mu(ds)\right)dt+F(X)(t-)dL(t),\qquad t\ge 0,\cr X(u)&=\Phi(u)\quad \text{ for } u\in[-\alpha,0],}\end{align*} \] where \(L\) is a Lévy process adapted to a filtration \(({\mathcal F}_t)_{t\geq 0}\), \(\mu\) is a signed finite measure, the initial process \(\Phi\) is \({\mathcal F}_0\)-measurable with paths in the Skorokhod space \(D([-\alpha,0])\), and \(F: D([-\alpha,\infty))\to D([-\alpha,\infty))\) is deterministic such that \[ | F(\varphi_1)(t)-F(\varphi_2)(t)| \leq K\sup_{t-\alpha\leq s\leq t}| \varphi_1(s)-\varphi_2(s)| ,\quad t\geq 0, \] and \[ F(\varphi(s+\cdot))(t)=F(\varphi)(t+s),\quad t,s\geq 0. \] The unique strong solution \(X=(X(t))_{t\geq -\alpha}\) of this equation determines the segment process \((X_t)_{t\geq 0}\) with values in \(D([-\alpha,0])\), defined by \(X_t(s)=X(t+s)\), \(s\in[-\alpha,0]\). It is proved that this segment process is Markov, Feller for \(t\geq\alpha\), but not strong Feller. The main result of the paper is the existence of a stationary measure for this process provided that \(F\) is bounded, the Lévy measure \(\nu\) of \(L\) has the property \(\int_{| x| >1}\log| x| \nu(dx)<\infty\) and \(\sup\{\text{Re}(\lambda): \lambda\in{\mathbb C}, \lambda=\int_{[-\alpha,0]}e^{\lambda s}\mu(ds)\}<0\). An important auxiliary result is the theorem on tightness of \(\{{\mathcal L}(X_t)\}_{t\geq\alpha}\). The problem of uniqueness of the stationary distribution is also discussed.