Density estimate in small time for jump processes with singular Lévy measures (Q5949602)

Consider a possibly singular \(d\)-dimensional Lévy measure \(\mu\) such that as \(\rho \to 0\), \[ C_1 \rho^{2-\beta} I \leq \int_{|\zeta |\leq \rho } \zeta \zeta^* \mu (d\zeta) \leq C_2 \rho^{2-\beta} I, \] for some constants \(C_1,C_2 > 0\), where \(\beta \in (0,2)\) and \(I\) denotes the identity matrix. The author obtains small time estimates for the transition density \(p(t,x,y)\) of a jump process \((x_t)_{t\in {\mathbb R}}\) given as the solution of a stochastic differential equation driven by a Lévy process with \(\mu\) as Lévy measure. In particular he shows that \(p(t,x,y) \asymp c t^{\alpha (x,y) - d/\beta }\) when \(y\) is accessible, i.e. \(y\) can be reached from \(x\) in a finite number of jumps along the trajectory of \((x_t)_{t\in {\mathbb R}}\), and \(\alpha (x,y)\) represents the minimum number of such jumps. If \(y\) is not accessible, the decay of \(p(t,x,y)\) is not polynomial, but may be exponential. This extends similar results obtained by \textit{J. Picard} [ESAIM, Probab. Stat. 1, 357-389 (1997; Zbl 0899.60065) and Stochastic Processes Appl. 67, No.~2, 251-279 (1997; Zbl 0889.60088)].