On the boundedness of the denominators in the Zariski decomposition on surfaces (Q1683973): Difference between revisions

An important tool to study divisors on smooth projective surfaces is the so called Zariski decomposition of pseudoeffective integral divisors. Any such \(D\) can be written uniquely as sum \(D=P+N\) of \(\mathbb{Q}\)-divisors such that \(P\) is nef; \(N\) is effective with negative definite intersection matrix when non zero; and \(P\cdot C=0\) for any component \(C\) of \(N\). The question faced in the paper under review is that of the existence of an upper bound \(d(X)\) for the denominators of the Zariski decomposition terms of any pseudoeffective divisor \(D\), in fact just for the coefficients \(a_i\) of the negative part \(N=\sum a_i N_i\). Writing these \(a_i\)'s as the solution of a system of equations (\(D\cdot N_j=\sum a_i N_i|\cdot N_j\)) via Cramer's rule, the determinant of the intersection matrix \([N_i\cdot N_j]\) appears as a bound for denominators of this decomposition. This is showing a bound when the pseudo-effective cone is rational polyhedral (see Remark 1.1) and put the focus on the study of the intersection of negative curves on the surface. In fact, the main result of the paper is the equivalence between the existence of this bound \(d(X)\) and the existence of an upper bound \(-b(X)\) for the self intersection \(C^2\) of irreducible and reduced curves \(C\) on \(X\), that is, \(C^2 \geq -b(X)\) for such \(C\). This second statement is commonly known as the Bounded Negative Conjecture when working over the complex numbers and it is an interesting open problem (see the Introduction of the paper and references therein) even for the blow-up of the plane in \(s \geq 10\) general points. In Section 3 a collection of meaningful examples is presented.