no. 16 · Tuesday 17 November 2026
$P$-trivial MMP, Zariski decompositions and minimal models for generalised pairs
Abstract
We develop an inductive method for generalised pairs (g-pairs) without the NQC assumption. It is based on $P$-trivial MMPs, whose steps are trivial with respect to a given nef divisor $P$. For klt g-pairs, if $P$ is numerically equivalent to an effective divisor, we perturb the boundary so that any MMP on $K_X+B+M+αP$ is $P$-trivial for $α\gg0$. For lc g-pairs, the components with coefficient one cannot be perturbed, so we use induction on lc centres. If $K_X+B+M-P$ is a $\mathbb{Q}$-divisor, then local Cartier indices control the coefficients under divisorial adjunction, and we prove a special termination. This includes the case $P=0$, where $K_X+B+M$ is a $\mathbb{Q}$-divisor. As applications, we prove that a projective lc g-pair $(X,B+M)$ of dimension $3$ has a minimal model if $K_X+B+M$ birationally has a Nakayama-Zariski decomposition with nef positive part and either $K_X+B+M\equiv D\geq0$ or $M\equiv D\geq0$. We also prove this for klt g-pairs in arbitrary dimension. In dimension $4$, the last condition can be replaced by the log abundance of the nef data. For usual pairs, this is a theorem of Birkar and the author.
- comments
- 28 pages. All comments are welcome
- fields
- Algebraic Geometry