Stochastic PDEs driven by nonlinear noise and backward doubly SDEs (Q5960452)

The authors study the semilinear parabolic stochastic PDE \[ \begin{multlined} \{(\partial_t+{\mathcal L})u(t,x)+ f(t,x,u,\nabla u\cdot\sigma)\} dt+ (d_tB)(t,g(t,x,u,\nabla u\cdot\sigma)=0,\\ (t,x)\in [0,T]\times \mathbb{R}^d,\;u(T,x)= h(x),\quad x\in\mathbb{R}^d,\end{multlined} \] with \({\mathcal L}={1\over 2}\cdot\text{tr}(\sigma\sigma^* D^2_x)+ bD\), \(B(t,y)\) is some field of continuous backward local martingales and \(\int^T_s(d_t B)(t,g(t,x,u,\nabla u\cdot\sigma)\) is defined in the sense of the backward Itô-Kunita integral. \textit{E. Pardoux} and \textit{S. Peng} [Probab. Theory Relat. Fields 98, No. 2, 209-227 (1994; Zbl 0792.60050)] solved this equation in the classical case \((dB)(t,g(t,.))= g(t,.) dB_t\), where \(B\) is multidimensional Brownian motion, and they gave a stochastic interpretation of the solution by introducing a backward doubly stochastic differential equation. Later, \textit{V. Bally} and \textit{A. Matoussi} [J. Theor. Probab. 14, No. 1, 125-164 (2001; Zbl 0982.60057)] showed that the same probabilistic interpretation holds for the variational formulation of the stochastic PDE. The present paper generalizes the work of Pardoux and Peng (loc. cit.) and that of \textit{H. Kunita} [``Stochastic flows and stochastic differential equations'' (1990; Zbl 0743.60052)] who has studied the case \(f=0\), and \(B\) linear in the solution and its first derivative. To give a stochastic interpretation to the classical and the Sobolev solution of the above stochastic PDE, they generalize Pardoux-Peng's backward doubly SDE and solve it in the classical and the ``weak'' sense.