BSDEs, convergence in law and homogenization of semilinear parabolic SDEs (Q5929090)

The authors study the convergence in law of backward stochastic differential equations (BSDEs) associated to systems of semilinear parabolic partial differential equations (PDEs), in order to apply then their results to the problem of homogenization of PDEs. To be more precise we denote by \(X^\varepsilon\) the diffusion process generated by the second-order operator \({\mathcal L}^\varepsilon\) driving the PDE, and by \(M(X^\varepsilon)\) the martingale part of \(X^\varepsilon\). The associated BSDE, the solution \(Y^\varepsilon\) of which describes the solution of the PDE (generalized Feynman-Kac formula), is introduced in a nonclassical way, namely the martingale representation is made w.r.t. \(M(X^\varepsilon)\). Denoting the martingale part of the BSDE by \(M^\varepsilon\) and assuming the convergence in law of \(X^\varepsilon\) as \(\varepsilon\downarrow 0\), the authors study the convergence in law of the quadruplet \((X^\varepsilon,\langle X^\varepsilon, M^\varepsilon\rangle, Y^\varepsilon, M^\varepsilon)\) and determine the BSDE which is satisfied by the limit. One should mention that the reviewer, \textit{Y. Hu} and \textit{S. Peng} [NoDEA, Nonlinear Differ. Equ. Appl. 6, No. 4, 395-411 (1999; Zbl 0953.35017)] and \textit{P. Briand} and \textit{Y. Hu} [J. Funct. Anal. 155, No. 2, 455-494 (1998; Zbl 0912.60081)] have studied homogenization of PDE by a BSDE method which avoids convergence in law and allows to treat general nonlinear terms (of at most linear growth) depending on the gradient. Although, additionally to its own importance, the method of convergence in law developed by the authors allows also to treat PDE, and certainly also SPDE with random coefficients.