Harnack inequalities for Ornstein-Uhlenbeck processes driven by Lévy processes (Q553105)

Let \(L=(L_t)_{t\geq 0}\) be a \(d\)-dimensional Lévy process, \(A\) a \(d\times d\) matrix, and \(X=(X_t)_{t\geq 0}\) the Ornstein-Uhlenbeck processes defined by the stochastic differential equation \(dX_t=A X_t+dL_t\). The author studies the so-called Harnack and log-Harnack inequalities for the the semigroup \((P_t)_{t\geq 0}\) of the process \(X\). More precisely, he proves the following two results: (i) If the Lévy measure of \(L\) dominates the Lévy measure of a rotationally invariant \(\alpha\)-stable process for some \(\alpha\in (0,2)\), then there exists a constant \(C=C(d,\alpha,A)\) such that for all \(t\geq \) all \(x,y\in \mathbb R^d\) and all non-negative, bounded Borel functions \(f\), \[ P_t f(x) \leq C P_t f(y)\left(1+\frac{|x-y|}{(t\wedge 1)^{1/\alpha}}\right)\, , \] and (ii) If the Lévy measure of \(L\) dominates the Lévy measure of a truncated rotationally invariant \(\alpha\)-stable process for some \(\alpha\in (0,2)\), then there exists a constant \(C=C(d,\alpha,A)\) such that for all \(t\geq \) all \(x,y\in \mathbb R^d\) and all non-negative, bounded Borel functions \(f\) with \(f\geq 1\), \[ P_t (\log f)(x)\leq \log P_t f(y)+C \left(1+|x-y|\right) \log\left(\frac{2+|x-y|}{t\wedge 1}\right)\, . \] The proofs are based on the estimates of the ratio of transition densities \(p_t(x,y)/p_t(y,z)\), which are derived from known sharp estimates of \(p_t\).