The diminished base locus is not always closed

Abstract

We exhibit a pseudoeffective R-divisor D_\lambda on the blow-up of P^3 at nine very general points which lies in the closed movable cone and has negative intersections with a set of curves whose union is Zariski dense. It follows that the diminished base locus B_-(D_\lambda) = \bigcup_{A ample}} B(D_\lambda+A) is not closed and that D_\lambda does not admit a Zariski decomposition in even a very weak sense. By a similar method, we construct an R-divisor on the family of blow-ups of P^2 at ten distinct points, which is nef on a very general fiber but fails to be nef over countably many prime divisors in the base.Comment: Revamped introduction, various simplifications. To appear in Compositio Mathematica (with minor changes

On positivity and base loci of vector bundles

“…Fix an ample line bundle A and a rational number ε > 0. We define Furthermore, as we mentioned earlier, Lesieutre [13] proved that there exist line bundles which are pseudo-effective but not weakly positive. In particular, B − (L) is not necessarily closed.…”

A Mehta-Ramanathan theorem for linear systems with basepoints

“…7] gave an example of such behavior in characteristic . Lesieutre [, Thm. 1.2] showed that this is also possible over the complex numbers if one allows ‐divisors instead of line bundles.…”

Comparing multiplier ideals to test ideals on numerically ℚ‐Gorenstein varieties

“…Said another way, given the behavior of nefness in families (cf. [10,11]), it is unclear if numerical pull-back can be preserved after reduction to positive characteristic. In a sense, this is the difficulty that must be avoided in proving Theorem 1, as otherwise one could use Proposition 17 in order to verify (3.1).…”