Cone theorem via Deligne-Mumford stacks (Q842370): Difference between revisions

Mori's Cone Theorem is a fundamental result in higher dimensional algebraic geometry. It states that if \(X\) is a smooth projective variety, then the cone of effective curves is given by \(\overline {NE}(X)=\overline {NE}(X)_{ K_X\geq 0}+\sum \mathbb R _{\geq 0}[C_i]\) where the \(C_i\) are a countable collection of curves locally finite in the half space \(\overline {NE}(X)_{K_X< 0}\) and such that \(0>C_i\cdot K_X \geq -(\dim X +1)\). Mori's proof depends on a technique (known as bend and break) involving the deformation theory of maps of curves in to smooth varieties. In characteristic zero, the cone theorem is known to generalize to Kawamata log terminal pairs. The proof however uses cohomological techniques and produces the weaker bound \(0>C_i\cdot K_X \geq -2\dim X \). Mori's bend and break technique has been generalized by Kollár to projective varieties with LCIQ singularities (quotients of locally complete intersections). In this paper, the authors generalize the bend and break technique to the context of twisted stable maps to LCI Deligne-Mumford stacks. When \(X\) is a projective variety with only tame LCIQ singularities, they show that \[ \overline {NE}(X)=\overline {NE}(X)_{K_X\geq 0}+\text{Im}[ \overline {NE}(X_{\text{sing}})\to \overline {NE}(X)]+\sum \mathbb R _{\geq 0}[C_i] \] where the \(C_i\) are a countable collection of rational curves, locally finite in the half space \(\overline {NE}(X)_{K_X< 0}\) and such that \(0>C_i\cdot K_X \geq -(\dim X +1)\). Moreover, when \((X,D)\) is a divisorially log terminal \(3\)-fold, they show that \(\overline {NE}(X)=\overline {NE}(X)_{K_X\geq 0}+\sum \mathbb R _{\geq 0}[C_i]\) where the \(C_i\) are a countable collection of rational curves, locally finite in the half space \(\overline {NE}(X)_{K_X+D< 0}\) and such that \(0>C_i\cdot (K_X+D) \geq -4\).