Existence and uniqueness of solutions to the inverse boundary crossing problem for diffusions (Q655577)

Given a real-valued process solution \(x_0\) with density \(p_0(\cdot)\) with nice coefficients of a stochastic differential equation \[ dx_t= \mu(x_t, t)\,dt+\sigma(x_t, t)\,dB_t, \] and a function \(b: \mathbb{R}_+\to \mathbb{R}\), define the stopping time \[ H_b= \text{inf}\{t> 0; x_t= b_t\}. \] It is a classical problem to study the quantity \[ p_b(t)= \text{P}(H_b\geq t). \] The inverse boundary crossing problem consists in finding a function \(b\) in such a way as to have a \(p_b\) on an a priori given function \(p\). The starting point of the present work is the remark that the function \[ f(x,t)= \text{P}(H_b\geq t, x_t> x) \] formally satisfies the variational inequality \[ \begin{cases} \max\{Lf,f-p\}= 0,\quad &\text{in }L^\infty(\mathbb{R}\times \mathbb{R}^{-n}_+),\\ f(.,0)= 1- p_0(.),\quad &\text{on }\mathbb{R},\end{cases}\tag{1} \] where \(L\) is the generator of the diffusion, so \(b(t)\) can be formally recovered by \[ b(t)= \text{inf}\{x; p_t- f(x,t)> 0\},\tag{2} \] for any \(t> 0\). Existence and uniqueness results for viscosity solutions of (1) were proved earlier, but no proof of the fact that (2) defines a solution of the inverse boundary crossing problem was available, as the lack of regularity of the viscosity solution of (1) makes this difficult. This is done here by a fine study of crossing times of upper semicontinuous boundaries and approximation arguments well-suited to the viscosity framework.