Fatou limits of stochastic integrals
Vasily Melnikov
Abstract
The convergence of stochastic integrals is essential to stochastic analysis, especially in applications to mathematical finance, where they model the gains associated with a self-financing strategy. However, Fatou convergence of $(X^{n})_{n=1}^{\infty}$ $\unicode{x2014}$a notion introduced for its amenability to compactness principles$\unicode{x2014}$implies little about the sequence of Itô integrals $\left(\int_{0}^{\cdot}YdX^{n}\right)_{n=1}^{\infty}$ for a fixed integrand $Y$. Under a boundedness condition, we find convex combinations $(\widetilde{X}^{n})_{n=1}^{\infty}$ of $(X^{n})_{n=1}^{\infty}$ with Fatou limit $\widetilde{X}$, such that $\left(\int_{0}^{\cdot}Yd\widetilde{X}^{n}\right)_{n=1}^{\infty}$ converges in a Fatou-like sense to $\int_{0}^{\cdot}Yd\widetilde{X}$ for all continuous semimartingales $Y$. The result is sharp, in the sense that continuity of $Y$ cannot be relaxed to being the left limits process of a semimartingale.