Singular foliations with trivial canonical class (Q1671171)

Let \(X\) be a complex projective manifold, and let \(\mathcal F \subset T_X\) be a (singular) foliation, that is a saturated subsheaf that is closed under the Lie bracket. The locus where \(\mathcal F \subset T_X\) is not a subbundle is called the singular locus \(\mathrm{sing}(\mathcal F)\) of the foliation, the foliation is regular when \(\mathrm{sing}(\mathcal F)=\emptyset\). The canonical bundle \(K_{\mathcal F}\) of the foliation \(\mathcal F\) is defined as the dual of \(\det(\mathcal F)\). In analogy with the classification of projective manifolds in terms of the canonical bundle, one expects that the positivity properties of \(K_{\mathcal F}\) govern the geometry of the foliation. For foliations on surfaces this program is fully established by the work of \textit{M. Brunella} [Birational geometry of foliations. Reprint of the 2000 edition with new results. Cham: Springer (2015; Zbl 1310.14002)] and \textit{M. McQuillan} [Pure Appl. Math. Q. 4, No. 3, 877--1012 (2008; Zbl 1166.14010)]. This paper contributes to the MMP for foliations on higher-dimensional manifolds by establishing several fundamental results in the case where the canonical bundle is numerically trivial: The first main theorem states that if the anticanonical bundle \(K_{\mathcal F}^*\) is pseudoeffective and the singular locus \(\mathrm{sing}(\mathcal F)\) is not empty, then the manifold \(X\) is uniruled. If \(K_{\mathcal F}^*\) is not numerically trivial the statement also follows from the recent result of \textit{F. Campana} and \textit{M. Păun} [Publ. Math., Inst. Hautes Étud. Sci. 129, 1--49 (2019; Zbl 1423.14109)], but the case where \(c_1(K_{\mathcal F})=0\) shows a surprising interaction between the singularities of the foliation and the global geometry of \(X\). This interaction can be made more explicit for a foliation that has canonical singularities in the sense of McQuillan: under this assumption the authors prove that the leaves of \(\mathcal F\) are dominated by rational curves if and only if the canonical bundle \(K_{\mathcal F}\) is not pseudoeffective. For the proof of the first main theorem the authors extend a technique introduced by \textit{J.-P. Demailly} [in: Complex geometry. Collection of papers dedicated to Hans Grauert on the occasion of his 70th birthday. Berlin: Springer. 93--98 (2002; Zbl 1011.32019)] to show that if \(K_X\) is pseudoeffective, then the foliation \(\mathcal F\) admits a supplement \(\mathcal G \subset T_X\) such that \(T_X=\mathcal F \oplus \mathcal G\). In particular both subsheaves are subbundles. The second main theorem gives a classification of codimension one foliations with \(c_1(K_{\mathcal F})=0\). If the foliation is regular, a complete description was established by \textit{F. Touzet} [Ann. Sci. Éc. Norm. Supér. (4) 41, No. 4, 657--670 (2008; Zbl 1166.32014)]. In view of the first main theorem one can thus assume that \(X\) is uniruled and the foliation \(\mathcal F\) has canonical singularities. In this case the authors prove that there exists an étale cover \(\mu: Y \times Z \rightarrow X\) such that the canonical bundle \(K_Y\) is trivial, and the foliation \(\mu^{-1}(\mathcal F)\) is a pull-back of a foliation \(\mathcal G\) on \(Z\) with {\em trivial tangent sheaf}. If the general leaf of the foliation \(\mathcal F\) is not algebraic, the variety \(Z\) is a projective equivariant compactification of a complex abelian Lie group and \(\mathcal G\) is induced by the action of a codimension one subgroup. The proof of this result involves an impressive variety of tools including deformation theory of rational curves along a foliation, the study of transversely projective structures and the reduction of the foliation to fields of positive characteristic. Parts of this very inspiring paper have been generalised recently by \textit{S. Druel} [``Codimension one foliations with numerically trivial canonical class on singular spaces'', Preprint, \url{arXiv:1809.06905}] to foliations on singular spaces.