On integral Zariski decompositions of pseudoeffective divisors on algebraic surfaces (Q256367)

Let us consider a smooth projective surface \(S\) (over an algebraic closed field). A negative (resp. \(-k\)) curve \(C\) on \(S\) is a reduced irreducible divisor \(C\) of negative self intersection (resp. \(-k\)). The Bounded Negativity Conjecture states that in characteristic \(0\) there exists an integer \(b(S)\) such that \(C^2 \geq b(S)\). This is known to be false in positive characteristic putting consequently the question of characterizing those surfaces with this property. On the other hand a, in principle, completely different question is that of upper bounding denominator in the Zariski decompositions of pseudoeffective integral divisors on \(S\). Rather surprisingly both questions are related (see Proposition 1.2 and references there): if \(C^2\geq -1\) for any reduced irreducible curve \(C\) then any pseudoeffective integral divisor has an integral Zariski decomposition. Naturally the converse question appears. In Theorem A the authors show that the answer is negative: there exists a smooth complex projective surface \(S\) (in fact a \(K3\) surface of Picard number \(2\)) with \(-2\)-curves for which any integral pseudoeffective divisor has an integral Zariski decomposition. But (see Theorem B) the answer is positive when adding on \(S\) the hypothesis that the determinant of the intersection form of its Néron-Severi lattice is \(1\) or \(-1\).