Generalization of Itô's formula for smooth nondegenerate martingales. (Q1879509)

Let \(W\) be a \(d\)-dimensional Wiener process, \(X=(X_ {t}) _ {0\leq t\leq T}\) a \(d\)-dimensional continuous local martingale adapted to \(W\) and satisfying certain nondegeneracy hypotheses. It is proven that if \(f\in L^ {p}_ {\text{loc}}(\mathbb R^ {d})\) for some \(p>d\), then the quadratic covariation of the processes \(f(X)\) and \(X^ {i}\), \(1\leq i\leq d\), defined as the limit in probability \[ \bigl [f(X),X^ {i}\bigr ]_ {t} = \lim _ {n\to \infty }\sum _{\substack{ t_ {i}\in D(n)\\ t_ {i}< t}} \bigl (f(X_ {t_ {i+1}}) - f(X_ {t_ {i}})\bigr ) \bigl (X^ {i}_ {t_ {i+1}} - X^ {i}_ {t_ {i}}\bigr ) \] exists, whenever \(\{D(n)\}\) is a sequence of partitions of \([0,T]\), the mesh of which tends to 0 as \(n\to \infty \) and \(\sup _ {n} \sup _ {t_ {i}\in D(n)} t^ {-1}_ {i}t_ {i+1} <\infty \). Consequently, the following generalized Itô formula \[ F(X_ {t}) = F(X_ {0}) + \sum ^ {d}_ {i=1} \int ^ {t}_ 0 {\partial F\over \partial x_ {i}}(X_ {s})\,dX^ {i}_ {s} + \frac 12 \sum ^ {d}_ {i=1} \left [ {\partial F\over \partial x_ {i}} (X_ {s}),X^ {i}_ {s}\right ]_ {t} \] holds for all \(F:\mathbb R^ {d}\to \mathbb R\) which belong locally to the Sobolev space \(W^ {1,p}(\mathbb R^ {d})\) with \(p>d\). Proofs are based on Malliavin calculus methods.