Hörmander's theorem for semilinear SPDEs (Q2279327): Difference between revisions

In this interesting paper, a general class of nonlinear stochastic PDEs is considered which is formally stated by \[du_t=Lu_tdt+N(u_t)dt+\sum_{i=1}^dF_i(u_t)\circ dB_t^i.\] Here, the multiplicative noise of Stratonovich type involves a \(d\)-dimensional Brownian motion, \(L\) is a selfadjoint negative definite operator on a separable Hilbert space, and \(N\), \(F_i\), \(i=1,\ldots,d\) are smooth nonlinear functions. When \(N\), \(F_i\) are polynomial vector fields satisfying a Hörmander's non-degeneracy condition given by the positive definition of a symmetric operator, and for smooth enough global mild solutions \(u\), the authors prove in probability, and for arbitrary initial condition, an upper bound involving the Malliavin matrix of solutions. In that case, the law of finite dimensional projections of \(u\) has a smooth density with respect to the Lebesgue measure. This result is established in the context of rough paths theory, and the development of new tools thereof, like a mild Itô formula and a rough version of Fubini's theorem. Some applications are presented for the \(2\)-dimensional stochastic Navier-Stokes equation, and \(m\)-dimensional reaction-diffusion equations.