Introduction to the toric Mori theory (Q1766489)

In 1983, \textit{M. Reid} [in: Arithmetic and geometry, Pap. dedic. I. R. Shafarevich, Vol. II: Geometry, Prog. Math. 36, 395--418 (1983; Zbl 0571.14020)] gave combinatorial proofs of the log minimal model program (MMP) for complete toric varieties, under some technical restrictions but with no restriction as to dimension. His proofs used combinatorial methods. At the time much of Mori's work lay in the future and this was the first case of MMP (for dimension \(>2\)) to be carried out fully. Now, however, it makes sense to return to the toric case equipped with the results of the immense progress made in birational geometry in the meantime. That is what this paper does. The methods here are not combinatorial, and in particular they apply to noncomplete toric varieties. Various other technical restrictions, imposed originally more for convenience than out of necessity, can also be dispensed with. After three sections of introduction and background we reach the results of the paper in sections 4--6. Section 4 contains proofs of the toric log MMP. The cone theorem, that \(\text{NE}(X/Y)\subset N_1(X/Y)\) is a convex polyhedral cone for \(f\colon X \to Y \) a proper toric morphism, is even accorded two proofs. The second, though simpler, requires \(X\) to be projective and \({\mathbb Q}\)-factorial, and the characteristic to be zero. Finite generation, the fact that \(\bigoplus f_*{\mathcal O}_X(mD)\) is a finitely-generated \({\mathcal O}_Y\)-algebra, becomes straightforward once interpreted in toric terms, and so does existence of flips. Termination of flips uses the observation that the 1-skeleton of the fan is unchanged, and determines \(X\) up to finitely many possibilities. Section 5 is devoted to Zariski decomposition in the toric context, which (in one of its versions) becomes a fairly easy consequence of toric Mori theory: indeed the results obtained are in some cases slightly stronger than what is already in the literature. In Section 6 the authors consider minimal, canonical and log canonical models of hypersurface singularities. They recover the existence results for such models obtained by \textit{S. Ishii} [in: Birational algebraic geometry, Baltimore, MD, USA, April 11--14, 1996. Providence, RI: American Mathematical Society. Contemp. Math. 207, 63--77 (1997; Zbl 0957.14011)], again without using combinatorial methods. Their results are slightly stronger than those stated by Ishii, but it is easy to adjust her arguments so as to achieve the same results by her methods too. The authors recommend that their paper should be compared with Chapter~14 of [\textit{K. Matsuki}, ``Introduction to the Mori program'', Universitext (New York, NY: Springer) (2002; Zbl 0988.14007)]. They also point out that, quite apart from the historical significance of Reid's paper, it and also Ishii's contain combinatorial results and ideas which are not replicated here, and that the present paper does not therefore supersede them.