An approach to the abundance conjecture for 3-folds (Q755816)

The author approaches the following question: Is every minimal model good? (``abundance conjecture''). To be more precise, let X be a projective minimal n-fold (i.e. a projective n-fold over \({\mathbb{C}}\) having only terminal singularities and nef canonical divisor \(K_ X)\) then the numerical Kodaira dimension \(\nu\) (X) of X is by definition the largest integer a with \(K^ a_ X\cdot H^{n-a}\neq 0\) (H is an ample divisor on X). In general one has \(n\geq \nu (X)\geq \kappa (X)\) and X is called ``good'' if Kodaira dimension and numerical Kodaira dimension are equal. Equivalently, X is good if some multiple of \(K_ X\) is generated by global sections [see \textit{Y. Kawamata}, Invent. Math. 79, 567-588 (1985; Zbl 0593.14010)]. The abundance conjecture states that every minimal model has abundant canonical class, which means that some multiple of \(K_ X\) is globally generated. The present paper deals with the case \(n=3\), where the minimal model conjecture is known to be true [\textit{S. Mori}, J. Am. Math. Soc. 1, No.1, 117-253 (1988; Zbl 0649.14023)]. If X is as above and of dimension three then: - a theorem of Miyaoka states that \(\kappa\) (X)\(\geq 0\). This gives the abundance conjecture in the case \(\nu (X)=0,\) - the case \(\nu (X)=1\) is treated by \textit{Y. Miyaoka} [Compos. Math. 68, No.2, 203-220 (1988; Zbl 0681.14019)], - one easily deduces from the Kawamata-Viehweg vanishing theorem [cf. \textit{Y. Kawamata}, Math. Ann. 261, 43-46 (1982; Zbl 0476.14007) and \textit{E. Viehweg}, J. Reine Angew. Math. 335, 1-8 (1982; Zbl 0485.32019)], that in case \(\nu (X)=3\) the abundance conjecture holds. The author investigates the remaining case \(\nu (X)=2\) and proves the following theorem: If X is a projective minimal Gorenstein 3-fold such that there exists a multicanonical divisor having a component being not birationally equivalent to a ruled surface, then the abundance conjecture holds for X.