On the generalized stochastic Dirichlet problem. I: The stochastic weak maximum principle (Q964220)

The authors study the following stochastic Dirichlet problem \[ \begin{aligned} &Lu(x,\omega) = h(x,\omega) + \nabla f(x,\omega),\quad x\in I,\; \omega\in\Omega\\ &u(\cdot,\omega)|_{\partial I} = g(\cdot,\omega) \end{aligned} \] where \(L\) is a linear elliptic second-order differential operator, \(I\subset\mathbb R^n\) is an open and bounded domain, \(h,f\) are generalized processes. The main aim of this paper is to adapt classical methods from the theory of second-order elliptic PDE (e.g.\ as in Gilbarg-Trudinger) to the above stochastic Dirichlet problem where the coefficients of \(L\) are generalized processes (in the sense of \textit{S. Pilipović} and \textit{D. Seleši} [Infin. Dimens. Anal. Quantum Probab. Relat. Top. 10, No. 1, 79--110 (2007; Zbl 1115.60068)]). To overcome some obvious problems how to interpret multiplications of generalized processes, the authors understand the above problem using Wick products and recasting it as \(L\lozenge u = h + \nabla f\) in the framework of white noise analysis. This is combined with Sobolev space methods. Here \(L\lozenge u= \nabla(A\lozenge\nabla u + b\lozenge u) + c\lozenge\nabla u + d\lozenge u\) with all coefficients being sufficiently regular (in a Kondratiev-Sobolev space sense) generalized processes. The main result of the paper is a stochastic weak maximum principle for the operator \(L\) which, in turn, implies e.g.\ the uniqueness of the solution of \(L\lozenge u = h+\nabla f\).