Canonical models of foliations (Q998630)

The paper forms part of a special issue of the journal in honor of F. Bogomolov. The papers [\textit{F. A. Bogomolov}, Dokl. Akad. Nauk SSSR 236, 1041--1044 (1977; Zbl 0415.14013)] and [\textit{M. McQuillan}, Publ. Math., Inst. Hautes Étud. Sci. 87, 121--174 (1998; Zbl 1006.32020)] can be seen as a precedent of the paper. In the introduction, the author explains the line of the paper and why he follows it instead of that given in some successive preprints by him with a close purpose. As he says, once one accepts that its above cited paper is really the key point, then one leads to a truly satisfactory theory of general type objects which is properly speaking the generalization of the uniformization theorem to this generality and which goes beyond the limited details that algebra alone provides. The material of the paper is dense in many points and exhaustive in its details which constitute the most precise description of the canonical model that the theory allows. The basic objects of study are pairs \((X, \mathcal{F})\), where \(X\) is a normal projective variety over a field \(k\) and \(\mathcal{F}\) an integrable foliation. The paper contains five sections. The first one entitled singularities is an exposition of purely algebraic properties, with a second subsection devoted to dimension 2 and a third subsection that discusses residues around singularities in an algebraic way appropriate to the minimal model theorem. The second section treats the positive characteristic case and its main goal is the foliated bend and break lemma of Mikaoya which on surfaces yields the cone theorem. An alternative approach by Bogomolov and the author is also summarized in this section. The third section is devoted to minimal models although the object is not exactly a minimal model theorem but to get a model with \(K_\mathcal{F}\) as positive as possible (see Definition III.3.1 of the paper). Theorem 1 of the paper states the canonical model theorem. The central part of this article is section IV entitled, classification, where Theorems 2 and 3 are stated and proved. The classification takes into account the existence of minimal models and attends the numerical Kodaira dimension and the Kodaira dimension of the model. Thus when \(\nu(\mathcal{F})= -\infty \), this implies that the foliation is a conic pencil, although the converse is only true with canonical singularities. The main property in case \(\nu(\mathcal{F})= 2 \) is the lack of an algebraic description since the canonical ring need not be finitely generated. A complete list of possibilities for the case \(\nu(\mathcal{F})= 0\) is given in IV.3.6, that is Theorem 2 of the paper. The case \(\nu(\mathcal{F})= 1\) must be divided in two subcases: \(\kappa(\mathcal{F}) \geq 0\), here abundance holds and there is a Kodaira fibration; and \(\kappa(\mathcal{F})= - \infty\), where the canonical models transpire to be the Baily-Borel compactification of Hilbert-Modular surfaces in either of their natural foliations (Theorem 3 of the paper). Finally, more or less, by definition, a choice of metrization of \(T_\mathcal{F}\), allows us to consider certain functorial object \(\sigma: S \rightarrow [S/\mathcal{F}]\) as a varying family of conformal structures. The papers ends with a last section devoted to the study of these conformal structures.