Martingale nature and laws of the iterated logarithm for Markov processes of pure-jump type (Q2664530)

Consider an \({\mathbb R}^d\)-valued time-homogeneous Markov process \((X_t)_{t\geq 0}\) with Feller generator \(L\) of the form \[ Lu(x) = \int_{{\mathbb R}^d \setminus \{ 0 \} } ( u(x+z)-u(x) - \langle \nabla u(x),z\rangle {\mathbf 1}_{\{ |z|<1\}} ) N(x,dz), \] where the Lévy measure \(N(x,dz)\) satisfies \[ \sup_{x\in {\mathbb R}^d } \int_{{\mathbb R}^d \setminus \{ 0 \} } |z|^2 N(x,dz) < \infty. \] The authors show that under the condition \[ \int_{\{ |z| \geq 1\} } z^{(i)} N(x,dz) =0, \qquad 1 \leq i \leq d, \] the process \((X_t)_{t\geq 0}\) is a purely discontinuous martingale with finite second moments. A law of the iterated logarithm for the sample paths of \((X_t)_{t\geq 0}\) is deduced as an application.