Integration par parties dans l'espace de Wiener et approximation du temps local. (Integration by parts in the Wiener space and approximation of local time) (Q803644)

Let \({\mathcal X}=(X_ t:\) \(t\in [0,T])\) be a real-valued stochastic process having a continuous local time L(u,t), \(u\in {\mathbb{R}}\), \(0\leq t\leq T\), and \(X_{\epsilon}(t)=(\Psi_{\epsilon}*X)(t)\), \(t\geq 0\), the regularization of X by means of the convolution with the approximation of unity \(\Psi_{\epsilon}\). The main theorem (Theorem 3.5) is a generalization of various results about the approximation (for fixed u) of the local time L(u,\(\cdot)\) by means of a convenient normalization of the number \(N^{X_{\epsilon}}(u;\cdot)\) of crossings of the process \(X_{\epsilon}\) with the level u. Especially, this theorem extends to a class of not necessarily Markovian continuous martingales, a result of this type for one-dimensional diffusions due to \textit{J. M. Azais} [Ann. Inst. Henri Poincaré, Probab. Stat. 25, No.2, 175-194 (1989; Zbl 0674.60032)]. The methods of proof combine some estimations of the moments of the number of crossings with a level of a regular stochastic process with stochastic analysis techniques based upon integration by parts in the Wiener space.