Stochastic variational formulas for solutions to linear diffusion equations (Q742887)

The paper addresses a careful study on the properties of the solution of the equation \[ \frac{\partial u_\epsilon}{\partial t}+b(y,t)\frac{\partial u_\epsilon}{\partial y}+\frac{\epsilon}{2} \frac{\partial^2 u_\epsilon}{\partial y^2}=0,\quad y\in\mathbb{R},\;t<T, \] with terminal conditions \[ \lim_{t\to T}u_\epsilon(x,y,t)=0,\;y<x \quad\text{and}\quad \lim_{t\to T}u_\epsilon(x,y,t)=1,\;y>x. \] Among the main properties obtained by the authors let us highlight the following: the probabilistic representation of \(u_\epsilon\), convergence as \(u_\epsilon\to 0\) by means of the use of its associated logarithm (large deviation result), and some useful bounds on the fundamental solution associated to \(u_\epsilon\). The analysis is possible due to a smart characterisation of the logarithm of \(u_\varepsilon\) expressed as the optimal value of a given stochastic control problem.