Some remarks on the divisorial Zariski decomposition on varieties whose first Chern class is zero (Q627488)

Let \(X\) be a complex projective manifold and \(D\) a pseudoeffective \(\mathbb R\)-divisor on \(X\), that is, a divisor whose numerical class lies in the closure of the cone of effective divisors. \textit{S. Boucksom} [Ann. Sci. Éc. Norm. Supér. (4) 37, No. 1, 45--76 (2004; Zbl 1054.32010)] and, independently, \textit{N. Nakayama} [Zariski-decomposition and abundance. MSJ Memoirs 14. Tokyo: Mathematical Society of Japan (2004; Zbl 1061.14018)] introduced the notion of divisorial Zariski decomposition \(D=P(D)+N(D)\), where the ``negative part'' \(N(D)\) is an effective \(\mathbb R\)-divisor and \(P(D)\) an \(\mathbb R\)-divisor that is nef in codimension one (cf. ibid. for precise definitions). In the paper under review, the author establishes some basic properties of the negative part \(N(D)\) if \(X\) is a projective manifold with numerically trivial canonical class: if \(D\) is effective, there exists a birational map \(X \dashrightarrow X'\) which contracts all the irreducible components of \(N(D)\). Moreover, if \(D\) is a big \(\mathbb Q\)-divisor, \(N(D)\) is also a \(\mathbb Q\)-divisor. In fact, the author shows more generally that these properties hold for any pseudoeffective \(K_X+\Delta\) where \((X, \Delta)\) is a klt pair. The proof is based on running a minimal model program with scaling, a technique introduced in the famous work by \textit{C. Birkar, P. Cascini, C. D. Hacon} and \textit{J. McKernan} [J. Am. Math. Soc. 23, No. 2, 405--468 (2010; Zbl 1210.14019)]. In the case of a holomorphic symplectic manifold \(X\), the author obtains the following application: an irreducible divisor \(E \subset X\) is exceptional (i.e. \(E=N(E)\)) if and only if it is dominated by rational curves \(l\) such that \(E \cdot l<0\).