Brownian motion with singular drift (Q1394527)

A stochastic differential equation in \(\mathbb R^d\), driven by a \(d\)-dimensional additive Wiener process \(W\), can be written as \(dX_t=dW_t+dA_t\), where \(A_t=(A^1_t,\dots, A^d_t)\) and \(A^i_t=\int_0^t f_i(X_s) ds\). If the role of \(f_i(x) dx\) is replaced by a more general signed measure \(\pi(dx)\), the solution can be defined by means of an approximation with regularised drifts: Given Radon measures \(\pi^i(dx)\) on \(\mathbb R^d\) and approximations of the identity \(\psi_{\varepsilon}(x)=\varepsilon^{-d}\psi(x/\varepsilon)\), define \(\pi^i_n(dx):=G^i_n dx\), with \(G^i_n(x):=\int_{\mathbb R^d} \psi_{2^{-n}}(x-y) \pi^i(dy)\), and denote \(G_n(x):=(G^1_n(x),\dots,G^d_n(x))\). A weak solution relative to the generalised drift \(\pi\) and initial condition \(x_0\) is a probability under which \(X_t=x_0+ W_t + A_t\), where a) \(A_t\) is the limit in probability, uniformly on compacts, of \(\int_0^t G_n(X_s)\), b) there is a subsequence along which \(\sup_n \int_0^t G_{n}(X_s) ds<\infty\) a.s. for each \(t\), c) \(W_t\) is an adapted Wiener process. In the situation above, if \(d\geq 2\), and \[ \limsup_{\varepsilon\to 0}\sup_{x\in\mathbb R^d} \int_{B(x,\varepsilon)} |x-y|^{-(d-1)} |\pi^i|(dy) =0, \] then there is a unique weak solution, which is conservative, and the family of solutions corresponding to different initial conditions forms a strong Markov process. Moreover, each component of \(A\) is a continuous additive functional of \(X\) of finite variation. According to the authors, although in the absolutely continuous case their approach does not recover all known results, it includes most of them and many more which are new.