Backward SDEs and Cauchy problem for semilinear equations in divergence form (Q1400828)

The authors considers the following Cauchy problem \[ \begin{cases} (\partial_t+ A_t) u(t,x)= -f(t, x,u(t,x),(\sigma\nabla u)(t,x)),\quad (t,x)\in [0,T)\times \mathbb{R}^d,\\ u(T, x)= \varphi(x),\end{cases} \] with \(A_t= {1\over 2} \sum^d_{i,j=1} D_j(a^{ij}(t, x)D_i)\), for \(a: [0,T]\times \mathbb{R}^d\to \mathbb{R}^d\otimes \mathbb{R}^d\) measurable, symmetric, such that, for some \(\lambda\), \(\Lambda> 0\), \[ \lambda|\xi|^2\leq \sum_{i,j} a^{ij}(t, x)\xi_i\xi_j\leq \Lambda|\xi|^2, \] and where \(\sigma\) is the symmetric square-root of \(a\), and \(\varphi\) and \(f\) satisfy classical assumptions. The notion of backward SDE is generalized in the following way: a couple of stochastic processes \((Y^{s,x}_t, Z^{s,x}_t)\) is a solution of the BSDE associated to \(A_t\) if it satisfies usual conditions of measurability and boundedness in \(L^2\) and if \[ \begin{multlined} Y^{s,x}_t= \varphi(X_T)+ \int^T_t f(r, X_r, Y^{s,x}_r, Z^{s,x}_r) dr-\\ \int^T_t \langle Z^{s,x}_r, \sigma^{-1}(r, X_r) dM_{s,r}\rangle,\qquad t\in [s,T],\;P_{s,x}\text{-a.s.},\end{multlined} \] where \((X,P_{s,x})\) is the diffusion associated to \(A_t\), \(M\) is the continuous local martingale involved in a generalized Fukushima's decomposition for \(X\). The main result of the paper is that there exists a unique solution to the above backward SDE. This solution is related to the weak solution of the Cauchy problem in the same way as in the case of classical BSDEs.