On the integrability of holomorphic vector fields (Q841387)

The aim of this paper is to give necessary and sufficient conditions ensuring that a generic germ of holomorphic vector field \(X\) in \(\mathbb{C}^3\) singular at the origin admits a holomorphic first integral. Here ``generic'' means that the linear part at the origin \(DX(O)\) of \(X\) is diagonalizable and invertible; and that the coordinate planes are \(X\)-invariant (this can be achieved for generic vector fields up to a holomorphic change of coordinates). The authors first prove that a generic (in the previous sense) germ of holomorphic vector field admitting a holomorphic first integral must satisfy the following condition \((\star)\): there exists a real line \(L\subset\mathbb{C}\) through the origin containing all the eigenvalues of \(DX(O)\), and such that one of the connected components of \(L\setminus\{O\}\) contains exactly one eigenvalue \(\lambda(X)\). Let \(S_X\) be the smooth \(X\)-invariant curve associated to \(\lambda(X)\). The main result of this paper says that if \(X\) is generic and satisfies condition \((\star)\), then the following implications hold: if the holonomy of \(S_X\) is periodic then \(X\) admits a holomorphic first integral; if \(X\) admits a holomorphic first integral then the leaves of \(X\) are closed off the singular set; and if the leaves of \(X\) are closed off the singular set then the holonomy of \(S_X\) has finite orbits. Actually, the authors claim that these four conditions are equivalent. Unfortunately, they rest the proof of the missing implication (holonomy with finite orbits implies periodic holonomy) on a statement (Theorem 3.6: every germ of holomorphic map tangent to the identity in \(\mathbb{C}^2\) has a parabolic curve) which is false (the easiest example is \((z,w)\mapsto(z,w+z^2)\); it is true that every holomorphic germ tangent to the identity in \(\mathbb{C}^2\) with an isolated fixed point has a parabolic curve [see \textit{M. Abate}, Duke Math. J. 107, No. 1, 173--207 (2001; Zbl 1015.37035)] but this is not enough to prove the authors' claim). I do believe that the missing implication is correct, but as it stands the proof is incomplete.