Markov field property of stochastic differential equations (Q1902951)

In the recent time several authors (Ocone/Pardoux, Nualart/Pardoux, Donati-Martin, Ferrante/Nualart) have studied conditions on the functions \(f\) and \(g\) under which the, in general, nonadapted solution of the stochastic boundary value problem \[ dX_t = f(t,X_t)dt + \sum^k_{i = 1} g_i(t, X_t) \circ W^i_t,\quad 0 \leq t \leq 1,\qquad h(X_0,X_1) = 0 \] is a Markov field (or germ Markov field, resp.). Here \(W = (W^1,\dots,W^k)\) denotes a \(k\)-dimensional Wiener process, the circle in the diffusion term denotes the (extended) Stratonovich integral. The method for proving the Markov property has been based in all cases studied on a change of measure argument using an extended version of the Girsanov theorem. The main difficulty in this method is that the Carleman-Fredholm determinant of a certain integral operator must be computed explicitly. The authors provide a new approach based on the characterization of the conditional independence of two independent random variables \(Z_1\) and \(Z_2\) given some function \(g(Z_1,Z_2)\). This allows them to give shorter proofs of known results and to treat new problems.