Estimates for the density of a nonlinear Landau process (Q852598)

Consider the nonlinear Landau equation \[ X_t=X_0+\int_0^t\int_0^1\sigma(X_s-Y_s(\alpha)) W(d\alpha,ds)+\int_0^t\int_0^1b(X_s-Y_s(\alpha)) d\alpha\,ds \] where \((X_t,Y_t(\alpha);t\geq0,\alpha\in[0,1])\) is the unknown process, \(W\) is a white noise on \([0,1]\times \mathbb{R}_+\), and \(\sigma\) and \(b\) are some specific coefficients which are defined in terms of another function \(h\); another constraint is that \(X\) and \(Y(\alpha)\) must have the same law. Under suitable assumptions, this equation has a unique solution, and the law of \(X\) can be used to model collisions of particles in a plasma. The aim of this article is to prove lower and upper bounds for the density of \(X_t\). Known results of Malliavin's calculus cannot be directly applied to this framework, and one has to adapt previously known techniques to this equation. This is worked out in the article, and one deduces unusual estimates, due to the bad behaviour of the coefficients.