The Azéma-Yor embedding in Brownian motion with drift (Q4524744)

Let \(B=(B_t)_{t\geq 0}\) be a standard Brownian motion starting at zero, let \(\mu >0\) be given and fixed, and let \(\nu\) be a probability measure on \(R\) having a strictly positive density \(F'\). The main aim of this paper is to prove that there exists a stopping time \(\tau_*\) of \(B\) such that \((B_{\tau_*}+\mu\tau_*) \sim\nu\) if and only if the following condition is satisfied: \(D_\mu: =\int_R \exp (-2\mu x)\nu (dx)\leq 1\). The author proves that if \(C_\mu=- (2\mu)^{-1}\log (D_\mu)\), then the following explicit formula is valid: \[ \tau_*= \inf\Biggl \{t >0 \biggl|(B_t +\mu t)\leq h_\mu \Bigl(\max_{0 \leq r\leq t}(B_r+\mu r)\Bigr) \Biggr\}, \] where the map \(s\to h_\mu (s)\) for \(s>C_\mu\) is expressed through its inverse by \[ h_\mu^{-1}(x)= -{1\over 2\mu} \log\left( {1\over 1-F(x)} \int^\infty_x \exp(-2 \mu t)dF(t) \right), \quad x\in R, \] and \(h_\mu(s) =-\infty\) for \(s\leq C_\mu\). In addition, the author proves that \(\tau_*\) is pointwise the smallest possible stopping time satisfying \((B_{\tau_*} +\mu \tau_*) \sim\nu\) which generates stochastically the largest possible maximum of the process \((B_t+ \mu t)_{t\geq 0}\) up to the time of stopping. This minimax property characterizes \(\tau_*\) uniquely. For the entire collection see [Zbl 0948.00040].