The cone conjecture for Calabi-Yau pairs in dimension 2 (Q993595)

This paper proves the cone conjecture for klt Calabi-Yau pairs of dimension 2. Let \(f : X \rightarrow S\) be a projective morphism between two algebraic varieties \(X\) and \(S\). A 1-cycle over \(S\) is a formal linear combination \(C=\sum_ia_iC_i\) where \(a_i \in \mathbb{R}\), and \(C_i\) are irreducible and reduced curves in \(X\) that are mapped to points by \(f\). A 1-cycle is called effective if \(a_i \geq 0\), for all \(i\). Two 1-cycles \(C\) and \(C^{\prime}\) are called numerically equivalent if \(D\cdot C_1=D \cdot C_2=0\), for any Cartier divisor \(D\) in \(X\). Dually, two Cartier divisors \(D_1\), \(D_2\) are called numerically equivalent over \(S\) if \(D_1 \cdot C=D_2 \cdot C=0\), for any curve \(C\) on \(X\) that is mapped to a point by \(f\). 1-cycles modulo numerical equivalence and Cartier divisors modulo numerical equivalence over \(S\) form two real vector spaces that are denoted by \(\mathrm{N}_1(X/S)\) and \(\mathrm{N}^1(X/S)\), respectively. These vector spaces are naturally dual to each other by the intersection product. The set of all effective cycles is a cone in \(\mathrm{N}_1(X/S)\) and is denoted by \(\mathrm{NE}(X/S)\). Let \(\overline{\mathrm{NE}}(X/S)\) be its closure in \(\mathrm{N}_1(X/S)\). Its dual cone in \(\mathrm{N}^1(X/S)\) is the cone of numerically effective Cartier divisors \(\mathrm{Nef}(X/S)\), i.e., Cartier divisors \(D\) such that \(D \cdot C \geq 0\), for any irreducible curve \(C\) in \(X\) over \(S\). A fundamental result from the theory of minimal models says that the birational geometry of \(f : X \rightarrow S\) is closely related to the convex geometry of the cones \(\overline{\mathrm{NE}}(X/S)\) and \(\mathrm{Nef}(X/S)\). Consequently, the study of these two cones is of great importance to birational geometry. If \(S\) is a point and \(X\) is a smooth Fano variety (which means \(-K_{X}\) is ample), then by the cone theorem it follows that \(\overline{\mathrm{NE}}(X/S)\) and \(\mathrm{Nef}(X/S)\) are rational polyhedral cones. For general varieties this is not true anymore. These cones can have infinitely many isolated extremal rays or they can be round. Both of these phenomena appear in the case of Calabi-Yau varieties, which are the next simplest varieties after Fano's. The original cone conjecture says that if \(X\) is a smooth Calabi-Yau variety, then the action of the automorphism group of \(X\) on the nef cone has a rational polyhedral fundamental domain [\textit{D. R. Morrison}, Journées de géométrie algébrique, Orsay 1992, Astérisque 218, 243--271 (1993; Zbl 0824.14007)]. This was later generalized by \textit{Y. Kawamata} to the case of \(\mathbb{Q}\)-factorial terminal Calabi-Yau fiber spaces \(f : X \rightarrow S\) [Int. J. Math. 8, No.5, 665--687 (1997; Zbl 0931.14022)] and by the author to Calabi-Yau klt pairs \((X/S, \Delta)\) [Compos. Math. 144, No. 5, 1176--1198 (2008; Zbl 1157.13006)]. A pair \((X/S, \Delta)\) is a klt \(\mathbb{Q}\)-factorial Calabi-Yau pair if it has \(\mathbb{Q}\)-factorial klt singularities and \(K_X+\Delta\) is numerically trivial over \(S\). The generalized cone conjecture stated by the author says that if \((X/S, \Delta)\) is a klt \(\mathbb{Q}\)-factorial Calabi-Yau pair then there exists a rational polyhedral cone \(\Pi\) in \(\mathrm{Nef}(X/S)\) which is a fundamental domain for the action of the automorphism group \(\mathrm{Aut}(X/S,\Delta)\) of \((X/S,\Delta)\) over \(S\) on \(\mathrm{Nef}(X/S)\). If \(X\) is terminal and \(\Delta=0\), then this is exactly Kawamata's conjecture for Calabi-Yau fiber spaces. The conjecture has been proved for smooth Calabi-Yau surfaces by Looijenga, \textit{H. Sterk} and \textit{Y. Namikawa} [Math. Z. 189, 507--513 (1985; Zbl 0545.14032); Math. Ann. 270, 201--222 (1985; Zbl 0536.14024)] and by Kawamata for 3-dimensional Calabi-Yau fiber spaces. At the moment of this writing the conjecture is open in all other cases. In this paper, the author proves the cone conjecture for klt Calabi-Yau pairs \((X,\Delta)\) of dimension 2 over the complex numbers. As a first step he proves the conjecture for klt Calabi-Yau surfaces. His method of proving this is as follows. Let \(f : Z \rightarrow X\) be the global index 1 cover of \(X\). This is a Calabi-Yau surface with canonical singularities and \(X\) is the quotient of \(Z\) by \(\mathbb{Z}_d\), the group of automorphisms of \(Z\) over \(X\), where \(d\) is the global index of \(X\). Let \(g : Y \rightarrow Z\) be its minimal resolution. This is a smooth Calabi-Yau surface. Then by the results of Sterk, Looijenga and Namikawa the cone conjecture holds for \(Y\). By another result of Oguiso, it also holds for the quotient \(W=Y/\mathbb{Z}_d\) and there is a natural birational morphism \(W \rightarrow X\). The result follows after a careful study of the relation between the nef cones of \(W\) and \(X\). In order to treat the general case, the author first reduces to the case when \((X,\Delta)\) is terminal, and in particular \(X\) smooth. This is done by considering a terminal model of \((X,\Delta)\) and by using earlier results of Looijenga about actions of groups on cones in order to compare the nef cone of \((X,\Delta)\) with this of its terminal model. Then by taking a Zariski decomposition of \(-K_X\), he shows that \(-K_X\) has Iitaka dimension 0,1 or 2. The proof is completed by a detailed and careful study of each of these cases.