The jumping coefficients of non-\(\mathbb{Q}\)-Gorenstein multiplier ideals (Q906869)

The theory of multiplier ideals has found many applications in higher-dimensional algebraic geometry. In one of its forms, the theory deals with a formal pair \((X, cZ)\), where \(X\) is a \(\mathbb{Q}\)-Gorenstein normal variety, \(Z\subseteq X\) is a closed subscheme, and \(c\) is a non-negative real number. \textit{T. de Fernex} and \textit{C. Hacon} [Compos. Math. 145, 393--414 (2009; Zbl 1179.14003)] extended the theory of multiplier ideals into pairs on arbitrary normal varieties. Therefore for a normal variety \(X\), a closed subscheme \(Z\) of \(X\), and \(c\geq 0\), one can define a sheaf of ideals \(\mathcal{J}(X, cZ)\), and hence can define jumping numbers associated to the pair \((X, Z)\) as usual. de Fernex and Hacon [loc. cit.] asked if in this generality the set of jumping numbers is still a discrete set of rational numbers. Rationality is known to be false by [\textit{S. Urbinati}, Mich. Math. J. 61, No. 2, 265--277 (2012; Zbl 1260.14042)]. Concerning discreteness, several special cases have been considered by S. Urbinati and others. In the paper under review, the author proves the discreteness of jumping numbers under the additional assumption that the non-\(\mathbb{Q}\)-Gorenstein locus of \(X\) is zero dimensional. Moreover, he shows there exists a closed subset \(W\) on \(X\) with codimension at least three, such that the set of jumping numbers of \((X\backslash W, Z|_{X\backslash W})\) is a discrete set of rational numbers for any closed subscheme \(Z\subset X\). Along the way, some other interesting results are obtained. For instance, following \textit{H. Rossi} [Rice Univ. Stud. 54, No. 4, 63--73 (1968; Zbl 0179.40103)], the author proves that given a coherent sheaf \(\mathscr{F}\) on a normal variety \(X\), there exists a minimal proper birational morphism \(f: Y\rightarrow X\) such that \(Y\) is normal and \(f^*{\mathscr{F}}\) modulo its torsion subsheaf is locally free. And if \(\mathscr{F}\) has rank one, then the resulting invertible sheaf is \(f\)-ample. The ACC property of the jumping numbers follows from the following simple observation. Suppose \(\{t_k\}_{k\geq 1}\) is an increasing sequence of jumping numbers of \((X, Z)\) converging to \(t_0\). If there exists an invertible sheaf \(\mathscr{L}\) on \(X\) such that \(\mathscr{L}\otimes \mathcal{J}(X, t_kZ)\) is globally generated for all \(k\geq 1\), then \(\{\mathcal{J}(X, t_kZ)\}\) stabilizes. This will contradict the definition of jumping number. To find such an \(\mathscr{L}\), the author, under the extra assumption that the non-\(\mathbb{Q}\)-Gorenstein locus has dimension zero, proves a slightly weaker form of a conjecture of \textit{S. Urbinati} [``Divisorial models of normal varieties'', \url{arXiv:1211.1692}], which asserts that for any normal projective variety \(X\) and a Weil divisor \(D\) on \(X\), there exists an ample divisor \(H\) on \(X\) such that for any positive integer \(m\), the sheaf \(\mathscr{O}_X(m(D+H))\) is globally generated. This enables him to construct an \(m\)-compatible divisor \(\Delta_m\) for \((X, t_kZ)\) for each \(k\) and rephrases the generalized multiplier ideal \(\mathcal{J}(X, t_kZ)\) as a usual one \(\mathcal{J}((X, \Delta_m), t_kZ)\) for a sufficiently divisible \(m\). The key point is that the class of \(\Delta_m\) in \(NS^1_{\mathbb{Q}}(X)\) is independent of \(m\), and hence of \(k\). Then a combined application of Nadel vanishing and Castelnuovo-Mumford regularity guarantees the existence of such \(\mathscr{L}\).