Zariski decomposition and canonical rings of elliptic threefolds (Q581494)

Let M be a 3-fold with \(\kappa =2\), hence birationally an elliptic fibre space \(f: M\to S\) over a surface. The author proves in two steps that the canonical ring of M is finitely generated: Step 1. A canonical bundle formula \((2.15)\quad K_ M=f^*(K_ S+J+D)+(E-X)\) where J is a divisor class coming from the modular invariant, and \(| J|\) is free; D is a \({\mathbb{Q}}\)-divisor with multiplicities \(\leq 1\) corresponding to singular fibres; and E, X are effective, but harmless, in the sense that \(f_*{\mathcal O}_ M(\ell E)={\mathcal O}_ S\) for every \(\ell >0\) and codim(f(X)\(\subset S)\geq 2.\) Step 2. The finite generation of the ring of \(K_ S+D\) is well known by results of Kawamata. The proof and its significance is surveyed by results by \textit{S. Mori}, ``Classification of higher-dimensional varieties'', (9.12) [in Algebraic Geometry, Bowdoin 1985, Proc. Summer Res. Inst., Part 1, Proc. Symp. Pure Math. 46, 269-331 (1987)]. {\S} 1 and the appendix make formal noises about Zariski decomposition in higher dimensions. The author's treatment is rather different from that of \textit{X. Benveniste} [C. R. Acad. Sci., Paris, Sér. I 295, 107-110 (1982; Zbl 0524.14008)], since he allows blowing up isolated curves. However, in the reviewer's opinion, it is just as hopeless as a method.