A second order analysis of McKean-Vlasov semigroups (Q2240470)

The McKean-Vlasov diffusions discussed here belong to a class of nonlinear Markov processes, which differ from ordinary Markov processes particularly in two respects: The corresponding evolution semigroup is nonlinear with respect to the initial condition of the system and their state space is the convex set of probability measures, hence, the usual functional analysis and differential calculus on Banach spaces is not directly applicable. In this article a second-order differential calculus is used to study the regularity and stability properties of the distribution of the semigroup of McKean-Vlasov diffusions. Second-order Taylor-type expansions are provided with remainder for the evolution semigroup as well as the stochastic flow. Bismut-Elworthy-Li formulae for the gradient and the Hessian of the integro-differential operators associated with the expansions are given, so are explicit Dyson-Phillips expansions and an analysis of the norm of the operators with some exponential decay inequalities with respect to the time horizon. The impact of the results is illustrated with a second-order extension of the Alekseev-Gröbner Lemma to nonlinear measure valued semigroups and interacting diffusion flows. The method developed also leads to the proof of some properties of uniform propagation of chaos with respect to time. A detailed discussion of relations to the existing literature is also provided.