Diagonal estimates of transition probability densities of certain degenerate diffusion processes (Q2641011)

Let \(V_ 1,V_ 2,...V_ n\in C_ b^{\infty}(R^ d,R^ d)\), \(V_ 0=\sum_{I\in E^ 0(2)}\sigma^ IV_{[I]}\) and for \(x\in R^ d\), \(b\geq a\) and \(t>0\), \[ \lambda_ b(t,x)=\max \{t^{| H| /2}| \det (V_{[I]}(x))_{I\in H}|:\;H\subset G^ 0(b)\text{ with }^{\#}H=d\} \] (for details, see the paper). Let p(t,x,dy) be the transition probability of the diffusion process corresponding to the operator \(L=\sum^{n}_{i=1}\hat V_ i^ 2+\hat V_ 0\), where \(\hat V_ i=\sum^{d}_{j=1}V_ i^ j(x) \partial /\partial x^ j\), \(i=0,1,...,n\), and p(t,x,\(\cdot)\) is its smooth density. The author proves the following Theorem. There exist \(0<c<C<\infty\), \(b\geq a\) and \(0<t_ 0\leq 1\) such that for all \(x\in \Omega_{a,\nu}\) and \(0<t\leq t_ 0,\) \[ c/\lambda_ b(t,x)\leq p(t,x,x)\leq C/\lambda_ b(t,x). \] A probabilistic version of this theorem was given by \textit{S. Kusuoka} and \textit{D. W. Strook} [J. Fac. Sci., Univ. Tokyo, Sect. I A 34, 391-442 (1987; Zbl 0633.60078)].