One-Dimensional Stochastic Differential Equations with Generalized Drift

• Published in: Theory of probability and its applications 2014-01, Vol.58 (3), p.345-357
• Main Authors: Blei, S, Engelbert, H-J
• Format: Article
• Language: English
• Subjects: Borel sets; Decomposition; Differential equations; Drift; Mathematical analysis; Stochasticity; Uniqueness
• ISSN: 0040-585X; 1095-7219
• DOI: 10.1137/S0040585X97986655

We consider one-dimensional stochastic differential equations with generalized drift which involve the local time $L^X$ of the solution process $ X_t = X_0 + \int_0^t b(X_s) \, {\rm d} B_s + \int_{\bf R} L^X(t,y)$ where $b$ is a measurable real function, $B$ is a Wiener process, and $\nu$ denotes a set function which is defined on the bounded Borel sets of the real line ${\bf R}$ such that it is a finite signed measure on ${\cal B}([-N,N])$ for every $N \in {\bf N}$. This kind of equation is, in dependence of using the right, the left, or the symmetric local time, usually studied under the atom condition $\nu(\{x\}) < 1/2$, $\nu(\{x\}) > -1/2$, and $|\nu(\{x\})| < 1$, respectively. This condition allows us to reduce an equation with generalized drift to an equation without drift and to derive conditions on existence and uniqueness of solutions from corresponding results for equations without drift. The main aim of the present paper is to treat the cases $\nu(\{x\}) \ge 1/2$, $\nu(\{x\}) \le -1/2$, and $|\nu(\{x\})| \ge 1$, respectively, for some $x \in {\bf R}$, and we give a complete description of the features of equations with generalized drift and their solutions in these cases. [PUBLICATION ABSTRACT]