Stability of BSDEs with random terminal time and homogenization of semilinear elliptic PDEs (Q1266271)

The present paper extends the probabilistic method of the homogenization of parabolic partial differential equations (PDEs) to elliptic PDEs. In 1996, by using a probabilistic method based on the nonlinear Feynman-Kac formula developped by \textit{E. Pardoux} and \textit{S. Peng} [in: Stochastic partial differential equations and their applications. Lect. Notes Control Inf. Sci. 176, 200-217 (1992; Zbl 0766.60079)], Hu/Peng/Buckdahn [HPB] have studied the homogenisation for parabolic PDEs. The key for this approach is the deduction of a certain stability result for backward stochastic differential equations (BSDE) with a finite, deterministic time horizon, and its combination with the well-known results of \textit{A. Bensoussan}, \textit{J. L. Lions} and \textit{G. Papanicolaou} [in: Proc. int. Symp. on stochastic differential equations, Kyoto 1976, 21-40 (1978; Zbl 0411.60078)] on homogenization. \textit{S. Peng} [Stochastics Stochastics Rep. 37, No. 1/2, 61-74 (1991; Zbl 0739.60060)] and, under more general conditions, \textit{R. W. R. Darling} and \textit{E. Pardoux} [Ann. Probab. 25, No. 3, 1135-1159 (1997; Zbl 0895.60067)] have shown that the solution of an elliptic PDE can be described by a BSDE with random time horizon. The authors of the present paper have applied this method to study the homogenisation of elliptic PDEs. But the fact that the random time horizon is not bounded anymore, makes impossible to apply directly the approach by [HPB]. The method of Briand/Hu is based on an approximation procedure introduced by Darling and Pardoux (loc. cit.) which is extended here to the study of the stability of BSDEs with random time horizon. The case of BSDEs and PDEs in dimension 1 is remarkable, where the authors can partially renounce the standard monotonicity hypothesis used by Peng and by Darling and Pardoux (loc. cit.).