Quadratic covariation estimates in non-smooth stochastic calculus (Q468746)

The authors study the asymptotic behaviour of the quadratic covariation between \(f(\varepsilon W)\) and \(W\), as \(\varepsilon \to 0\), where \(W\) is a standard Brownian motion and \(f\) is a non-smooth function. It is shown in particular that, in the case of \(f \in C^{\alpha}\), the speed of this decay is not exponential as expected, but at least polynomial in \(\varepsilon\). The authors use a recent representation as a backward-forward Itō integral of \([f(\varepsilon W), W]\) to prove an \(\varepsilon\)-dependent approximation scheme which is of independent interest. The result is obtained by means of providing estimates to this approximation. The results are then adapted to generalize the ones of [\textit{S. A. Almada Monter} and \textit{Y. Bakhtin}, Nonlinearity 24, No. 6, 1883--1907 (2011; Zbl 1239.60043)] and [\textit{Y. Bakhtin}, Probab. Theory Relat. Fields 150, No. 1--2, 1--42 (2011; Zbl 1231.34100)] related to the small noise exit from a domain problem for the saddle case.