Quadratic covariation and an extension of Itô's formula (Q1903608): Difference between revisions

Let \(X\) be a standard Brownian motion and \(F\) an absolutely continuous function with locally square integrable derivative \(f\). The main result is the generalization of Itô's formula: \[ F(X_t) = F(0) + \int^t_0 f(X_s) dX_s + (1/2) \bigl[ f(X), X \bigr ]_t. \] It is shown that the quadratic covariance \([f(X), X]\) always exists and can be calculated as a limit in probability of usual approximating sums or, alternatively, as a difference of the backward and forward integrals of \(f(X)\) with respect to \(X\). The latter observation plays a key role in the existence result. An extension for a time dependent case is also given. As an application, a construction of the Brownian local time at a continuous curve is considered.