Vector fields on factorial schemes (Q1891731)

Let \(f,g \in \mathbb{C} [x,y]\) be two polynomials such that their Jacobian is a nonzero constant. The famous Jacobian problem in dimension 2 asks whether or not \(\mathbb{C} [f,g] = \mathbb{C} [x,y]\) in this case. If the Jacobian problem has an affirmative answer then the \(\mathbb{C}\)-derivation \[ \delta : = (\partial f/ \partial x) \cdot \partial/ \partial y - (\partial f/ \partial y) \cdot \partial/ \partial x = \partial/ \partial g \] is locally nilpotent on \(\mathbb{C} [x,y]\), i.e. for every \(h \in \mathbb{C} [x,y]\) we have \(\delta^n (h) = 0\) for \(n\) sufficiently large. Conversely, one can show that if \(\delta\) is locally nilpotent then \(\mathbb{C} [f,g] = \mathbb{C} [x,y]\) [see, for example, \textit{H. Bass}, \textit{E. H. Connell} and \textit{D. Wright}, Bull. Am. Math. Soc., New Ser. 7, 287-330 (1982; Zbl 0539.13012)]. The present paper contains a criterion of local nilpotency for such a derivation \(\delta\) in terms of integral curves of \(\delta\). This criterion is a consequence of some general results concerning derivations on factorial domains. Central notions are the \(\delta\)-integral ring (:= the subalgebra generated by the polynomials defining \(\delta\)-integral curves) and the monoid of \(\delta\)-integral factors.