Abundance theorem for semi log canonical surfaces in positive characteristic (Q294177): Difference between revisions

Let \(X\) be a connected equi-dimensional algebraic variety over an algebraically closed field \(k\). Assume that \(X\) is \(S_2\) and the local ring of each codimension one point is either regular or nodal. Given a pair \((X, \Delta)\), let \(\nu: Y\rightarrow X\) be the normalization of \(X\) and \(\Delta_Y\) be the \(\mathbb{Q}\)-divisor on \(Y\) such that \(K_Y+\Delta_Y=\nu^*(K_X+\Delta)\). The pair \((X, \Delta)\) is called semi log canonical (slc) if \((Y, \Delta_Y)\) is log canonical (lc) provided log resolutions of singularities exist. In the paper under review, the author proves that for any projective slc surface \((X, \Delta)\) over \(k\) with \(\text{char}(k)>0\), if \(K_X+\Delta\) is nef, then \(K_X+\Delta\) is semi-ample. The abundance theorem for lc surfaces in arbitrary characteristic was proved by \textit{T. Fujita} in [J. Fac. Sci., Univ. Tokyo, Sect. IA Math, No. 30, 685--696 (1984; Zbl 0543.14004)]. In characteristic zero, the abundance theorem was established for slc surfaces by [\textit{D. Abramovich, L.-Y. Fong, S. Keel, J. Kollár}, and \textit{J. McKernan}, in Flips and Abundance for Algebraic Threefolds, Soc. Math. France, Astérisque No.211, 95-101, 139--158 (1992; Zbl 0799.14017)] , and for slc threefolds by [\textit{O. Fujino}, Duke Math. J., No.102, 513--532 (2000; Zbl 0986.14007)]. The proof follows the strategy in [\textit{O. Fujino}, Duke Math. J., No.102, 513--532 (2000; Zbl 0986.14007)] and uses the MMP for klt surfaces in positive characteristic. After taking the normalization \(Y\) of \(X\) and a projective birational morphism from a dlt surface \(\mu: (Z, \Delta_Z)\rightarrow (Y, \Delta_Y)\) with \(K_Z+\Delta_Z=\mu^*(K_Y+\Delta_Y)\), one needs to show the set of ``gluable'' sections of \(\mathscr{O}_{Z}(m(K_Z+\Delta_Z))\) generates \(\mathscr{O}_{Z}(m(K_Z+\Delta_Z)\) for a sufficiently large and divisible integer \(m\). This is reduced to one lower dimensional assertion: let \(S=\left\lfloor \Delta_Z\right\rfloor\), the integral part of \(\Delta_Z\), then the space of ``\(B\)-invariant'' sections of \(\mathscr{O}_{S}(m(K_Z+\Delta_Z)|_S)\) generates \(\mathscr{O}_{S}(m(K_Z+\Delta_Z)|_S)\). The assertion follows from the finiteness theorem, that is, for any lc curve (possibly reducible) \((C, \Delta)\) and positive integer \(m\) such that \(m(K_C+\Delta)\) is Cartier, the representation \(\text{Aut}(C, \Delta)\rightarrow \text{GL}(H^0(C, m(K_C+\Delta)))\) has finite image.