Stochastic differential equations with coefficients in Sobolev spaces (Q984414)

The paper deals with the stochastic differential equation \[ dX_t = \sum_{i=1}^m A_i(X_t)dW_t^i + A_0(X_t)dt \] on \({\mathbb R}^d\). Here, \(W_t\) is a standard Brownian motion in \({\mathbb R}^d\) and the equation is considered in the Itô sense. The diffusion coefficients \(A_1,\,\dots,\,A_m\) belong to the Sobolev space \(W_{\text{loc}}^{1,p}({\mathbb R}^d)\) (\(p>q\)) have linear growth. For the drift coefficient \(A_0\), two cases are considered in the paper. In the first case, in which \(A_0\) is considered to be a continuous vector field whose distributional divergence with respect to the Gaussian measure exists, the authors study the absolute continuity of the push-forward measure with respect to the Lebesgue measure on \({\mathbb R}^d\). It is shown that \(X_t\) leaves the Lebesgue measure on \({\mathbb R}^d\) quasi-invariant, provided that the coefficients are bounded Lipschitz continuous. In the second case, when the coefficient \(A_0\) has Sobolev regularity \(W_{\text{loc}}^{1,p'}\) for some \(p'>1\), the authors establish the existence and uniqueness of stochastic flow maps.