Local time-space calculus for symmetric Lévy processes (Q554450)

Let \((X_t)_{t\in[0,\infty)}\) be a Lévy process with characteristic exponent \(\psi(\xi) = \int_{-\infty}^{\infty}(1-e^{i\xi x}+i\xi x \mathbf{1}_{\{|x|\leq 1\}})\nu (dx)\), that is, without drift and Brownian component. Moreover, assume that \(\psi\) satisfies \[ \int_{-\infty}^{\infty} \frac{d\xi}{1+\psi(\xi)}<\infty, \] and that \(\beta : (0,1] \rightarrow [0,\infty)\), defined by \[ \beta(t) = \bigg(\int_0^{\infty} e^{-2 t \psi(\xi)}\psi(\xi) d\xi\bigg)^{1/2}, \] satisfies \(\int_0^1 \beta(t) <\infty\). The paper establishes generalized Itô-Tanaka formulae for the process \((F(X_t,t))_{t \in [0,1]}\), where \(F : \mathbb R \times [0,1] \rightarrow \mathbb R\) is a suitable continuous function. More precisely, the main result of the paper states that, if \(\frac{\partial F}{\partial t}\) exists as a square-integrable Radon-Nikodým derivative with respect to the Lebesgue measure and \[ \int_0^1 \beta(t) \int_{-\infty}^{\infty}\int_{-\infty}^{\infty} (F(x+y,t)-F(x,t))^2\, dx \nu (dy) dt < \infty, \] then \((F(X_t,t))_{t \in [0,1]}\) is a Dirichlet process, given as a sum of a square-integrable martingale and a continuous additive functional of zero quadratic variation. Explicit expressions for these two processes are presented. The proof relies on a novel construction of stochastic integration with respect to the local time process of \(X\), which also allows one to define generalized local times of \(X\) on measurable curves. Additionally, a localized version of the Itô-Tanaka formula is given.