No label defined (Q4220132)

Let \(X=(X_t; t\geq 0)\) be a regular one-dimensional diffusion on an interval \(I\subset {\mathbf R}\) with the speed measure \(m\). For \(t>0\) and \(x,y \in I\), let \(p(t;x,y)\) be the transition density of \(X\) with respect to \(m\), and let \(n_y(t,x)\) be the \({\mathbf P}_y\) density of the first hitting time \(H_x=\inf\{s: X_s=x\}\). If \(G_t^x= \sup\{u<t: X_u=x\}\) denotes the last exit time at \(x\) before time \(t\), the last exit decomposition says that for \(u<t\) \[ {\mathbf P}_x(G_t^x\in du, X_t\in dy) = p(u;x,x)n_y(t-u,x)du m(dy). \] The author proves this decomposition by using the symmetry properties of \(X\) (more precisely, the symmetry of the transition density: \(p(t;x,y)=p(t;y,x)\)) and the time reversal. The approach leads to complete characterizations of the laws of the processes \(\{X_u: 0\leq u\leq G_t^x\}\) and \(\{X_u: G_t^x\leq u\leq t\}\). The last part of the paper is devoted to proving the last exit decompositions at the life time \(\zeta\) of \(X\) (in case \(X\) is transient).