Nonlinear parabolic SPDEs involving Dirichlet operators (Q2787148)

The authors first prove existence, uniqueness, an energy estimate and a comparison result for real-valued solutions \((Y,M)\) to a backward doubly stochastic differential equation NEWLINE\[NEWLINE Y_t=\xi+\int_t^Tf(r,Y_r)\,dr+\int_t^Tg(r,Y_r)d^\dag\beta_r-\int_t^T\,dM_r NEWLINE\]NEWLINE driven by an infinite-dimensional Wiener process \(\beta=(\beta^k)_{k\in\mathbb N }\) (here \(d^\dag\beta\) indicates the backward Itō integral), where \(f\) and \(g\) are allowed to be random, \(f\) is monotone and continuous and \(g\) is Lipschitz continuous in the last variable.NEWLINENEWLINEIn the second part of the paper, a locally compact separable metric space \(E\) endowed with an everywhere dense Borel measure \(m\) are considered as well as a family of regular Dirichlet forms \(\{B^{(t)}:t\in\mathbb R\}\) on \(H=L^2(E,m)\) with the sector constant independent of \(t\) and a common domain \(V\subseteq H\). The operators \(A_t\) are associated with the forms \((B^{(t)},V)\). The authors prove the existence of a unique solution to the equation NEWLINE\[NEWLINE du(t)=-(A_tu+f(t,x,u))-g(t,x,u)d^\dag\beta_ t,\qquad u(T)=\varphi NEWLINE\]NEWLINE and present various properties of this solution (e.\,g., quasi-continuity, a stochastic representation of the solution via a stochastic Feynman-Kac formula, spatial and temporal regularity, energy estimates, a variation-of-constants formula, etc.). In some cases, the non-linearities \(f\) and \(g\) may even depend on the gradient of the solution through a Lipschitz-continuous dependence. Finally, the authors prove a connection between probabilistic and mild solutions.