Carleman inequality for backward stochastic parabolic equations with general coefficients (Q1763541)

By generalization of the method of backward stochastic differential equations to stochastic parabolic partial differential equations (SPDEs, for short) backward SPDEs have been introduced by \textit{G. Tessitore} [Stochastic Anal. Appl. 14, No.~4, 461-486 (1996; Zbl 0876.60044)] and \textit{J. Ma} and \textit{J. Yong} [Stochastic Processes Appl. 70, No.~1, 59--84 (1997; Zbl 0911.60048)]; the reader interested in more recent progress in this subject is referred to \textit{Y. Hu, J. Ma} and \textit{J. Yong} [Probab. Theory Relat. Fields 123, No.~3, 381--411 (2002; Zbl 1011.60046)]. The authors of the present short note establish a Carleman inequality for linear backward SPDEs. Their result improves considerably that obtained by \textit{V. Barbu, A. Răşcanu} and \textit{G. Tessitore} [Appl. Math. Optimization 47, No.~2, 97--120 (2003; Zbl 1087.93011)].\ Indeed, while Barbu, Răşcanu and Tessitore supposed that the second-order differential operator in the backward SPDE is the Laplacian and the coefficients are independent of the space variable, in the present paper the second-order differential operator is a general strictly elliptic divergence operator with deterministic \(C^{1,2}\)-coefficients, and the coefficients of the backward SPDE are general adapted processes. In application of the Carleman inequality an observability theorem for linear backward SPDEs is deduced with the help of which the authors study the exact null controllability of linear (forward) SPDEs.