A unifying class of Skorokhod embeddings: connecting the Azéma-Yor and Vallois embeddings (Q880476)

The authors contribute to the Skorokhod embedding problem which can be described in the following way: Given a one-dimensional Brownian motion \(B_t\) and a centered probability law \(\mu\). Is there a stopping time \(T\) such that \(B_T\) has the distribution \(\mu\)? Skorokhod posed and solved this problem in 1965 and thereafter many different solutions have been obtained, see the survey by \textit{J. Obłoj} [The Skorokhod embedding problem and its offspring. Probability Surveys 1, 321--390 (2004)]. In the present paper the authors show a new solution which connects the Azema-Yor and Vallois embeddings. The main result is as follows: Given a centered probability law \(\mu\), let \(x\geq 0\) and denote by \(L^x\) be the local time of the one-dimensional Brownian motion \(B_t\) at \(x\). Then there exists an explicitly given) stopping time \(T\) such that \(B_T\) is distributed like \(\mu\), \(T\) maximises \(\tau\mapsto\mathbb E\Psi(L_\tau^x)\) for any convex function \(\Psi\) over all embeddings \(\tau\) of \(\mu\) in \(B_t\) such that \(B_{t\wedge\tau}\) is uniformly integrable. The stopping time \(T\) is the first exit time of an auxiliary process \(X_t\) from an interval depending on its local time \(L_t^X\) at zero; \(X_t\) is a variably skewed Brownian motion solving the equation \(X_t = B_t - G(L_t^X)\) where \(G\) is some Lipschitz continuous function with Lipschitz constant less than \(1\).