Invariant measures of diffusions with gradient drifts (Q630261): Difference between revisions

For the stochastic differential equation with gradient drift of the form \(\nabla\nu\) on \(\mathbb{R}^d\) that yields a stationary Kolmogorov equation, the authors investigate the question of whether the corresponding invariant probability measure \(\mu\) has a density \(\rho\) of the form \(\rho=Ce^{\nu}\). The function \(\nu\) is supposed to belong to a certain Sobolev space so that the gradient is considered in the sense of generalized functions. Among the results one should mention: (i) if \(\mu\) satisfies the stationary Kolmogorov equation and \(e^\nu\) is a probability density, then \(\rho=e^\nu\) and \(\mu=E^nu dx\) is the unique probabilistic solution; (ii) if \(w\) is a bounded function and \(\rho=we^\nu\), then \(w\) is constant; if \(\int_{\mathbb{R}^d}|\nabla\nu|\rho dx<\infty\), there exists a number \(C\) such that \(\rho=Ce^\nu\). A certain condition for uniqueness of \(\mu\) is also found.