A martingale characterization of the Wiener process (Q921726)

Let \(X=(X_ t)_{t\geq 0}\) be a real-valued process, and \((F_ t)_{t\geq 0}\) (resp. \((G_ t)_{t\geq 0})\) an increasing (resp. decreasing) family of \(\sigma\)-fields, such that for every \(t\geq 0\), \(X_ t\) is adapted to \(F_ t\) and to \(G_ t\). The author proves that if \((X_ t)_{t\geq 0}\) and \((X^ 2_ t-t)_{t\geq 0}\) are both \((F_ t)\)-martingales, and if \((X_ tt^{-1})_{t\geq 0}\) and \(((X^ 2_ t- t)t^{-2})_{t\geq 0}\) are both reverse \((G_ t)\)-martingales, then there is a version of X which is a Wiener process. The proof amounts to show by Kolmogorov's criterion that X possesses a continuous version, and then to apply Lévy's characterization of the Brownian law.