Flow decomposition and large deviations (Q1923691)

The purpose of the paper is to show how the decomposition for flows of stochastic equations [\textit{J. M. Bismut}, Z. Wahrscheinlichkeitstheorie Verw. Geb. 55, 331-350 (1981; Zbl 0456.60063); \textit{H. Kunita}, in: Stochastic integrals. Lect. Notes Math. 851, 213-255 (1981; Zbl 0474.60046)] can be used to obtain new large deviation principles for the diffusions generated by \(\varepsilon^2L_1+L_2\), where \(L_1\) and \(L_2\) are two second-order differential operators, and when \(\varepsilon\to 0\). The main result of the paper contains the results of \textit{H. Doss} and \textit{D. W. Stroock} [J. Funct. Anal. 101, No. 2, 370-391 (1991; Zbl 0736.60050)] and \textit{J. T. Rabeherimanana} [Thèse de l'Univ. Paris VII (1992)], and extends them to the general case with no hypothesis at all on the Lie algebra. Another applications of the flow decomposition to a nonlinear filtering is given, extending earlier results of \textit{H. Doss} [Ann. Inst. Henri Poincaré, Probab. Stat. 27, No. 3, 407-423 (1991; Zbl 0739.60030)] and \textit{J. T. Rabeherimanana} (loc. cit.).