Estimation of the density of hypoelliptic diffusion processes with application to an extended Itô's formula (Q5960461)

The authors establish the uniform bound \(p_t(x)\leq C/t^{(2n+1)/2)}\) for the density of the solution at time \(t\in (0,1]\) of a one-dimensional stochastic differential equation \[ X_t=x_0+\int_0^tb(X_s)ds+\int_0^t \sigma(X_s)dW_s, \] under the hypoellipticity condition \(\sigma(x_0)=\dots =\sigma^{(n-1)}(x_0)=0\) and \(b(x_0)\sigma^{(n)}(x_0)\not=0\). This bound is proved using the techniques of Malliavin calculus. The same bound is obtained for \(\left(\int_{\mathbb R} \left| {p'_t(x)\over p_t(x)}\right| p_t(x) dx\right)^{1/r}\). These bounds imply that the time-reversed process \(\overline X_t:=X_{1-t}\) is also a diffusion process. As an application of these results, the authors are able to use the method introduced by \textit{H. Föllmer, P. Protter} and \textit{A. N. Shiryaev} [Bernoulli 1, No. 1--2, 149-169 (1995; Zbl 0851.60048)] to prove an extended Itô formula for \(F(X_t,t)\). The function \(F(x,t)\) is such that, for some \(\beta> {2n+1\over 2}\vee 2\), the partial derivative \({\partial\over\partial x} F(x,t)\) belongs to \(L^\beta_{\text{loc}} ({\mathbb R})\) as a function of \(x\), for all \(t\in [0,1]\), and for any compact set \(T\subset {\mathbb R}\), the function \(t\rightarrow \int_T | {\partial\over\partial x} F(x,t)| ^\beta dx\) is continuous in \([0,1]\).