Combinatorial degenerations of surfaces and Calabi-Yau threefolds (Q2520598)

This paper studies combinatorial degenerations of minimal surfaces of Kodaira dimension \(0\) over local fields. The main results are that the ``type'' of the degenerations can be read off from the monodromy operator acting on certain cohomology group. Let \(R\) be a complete discrete valuation ring with perfect residue field \(k\) of characteristic \(p>3\) and the field \(F\) of fractions. Let \(\pi\) be a uniformizer of \(R\). Let \(X\) be a smooth and projective scheme over \(F\). A \textit{model} of \(X\) over \(R\) is a regular algebraic space \({\mathcal{X}}\), proper and flat over \({\mathcal{X}}\) over \(R\), whose generic fibre is isomorphic to \(X\), and whose special fibre is a scheme. A model is said to be \textit{semistable} (resp. \textit{strictly semistable}) if it is étale locally smooth over \(R[x_1,\cdots, x_d](x_1\ldots x_r-\pi)\) (resp. if furthermore the irreducible components of the special fibre \(Y\) are smooth over \(k\)). The main problems discussed in this article are (1) Does there exist a model \({\mathcal{X}}\) of \(X\) which is smooth over \(R\)? (``abstract'' good reduction), and (2) Given a semistable model \({\mathcal{X}}\) of \(X\), can one tell whether or not \({\mathcal{X}}\) is smooth? (``concrete'' good reduction) The paper studies these problems for minimal smooth projective surfaces over \(F\) of Kodaira dimension \(\kappa=0\), namely, \(K3\) surfaces, Enriques surfaces, abelian surfaces and bielliptic surfaces. The prototypical result is formulated as follows. {Theorem 1}: Let \(X/F\) be a minimal surface with Kodaira dimension \(\kappa=0\). Let \(\ell\) be a prime (possibly equal to \(p\)). Let \({\mathcal{X}}/R\) be a ``minimal'' model of \(X\). Then the special fibre \(Y\) of \({\mathcal{X}}\) is ``combinatorial'', and moreover, there exists an ``\(\ell\)-adic local system'' \(V_{\ell}\) of \(X\) such that \(Y\) is of Type I, II or III as the nilpotency index of a certain monodromy operator on \(H^i(X,V_{\ell})\) is \(1, 2\) or \(3\), respectively. In other words, the degeneration ``type'' can be read off from the monodromy operator acting on \(H^i(X, V_{\ell})\). This can be viewed as an arithmetic analogue of results of \textit{U. Persson} [Mem. Am. Math. Soc. 189, 144 p. (1977; Zbl 0368.14008)] and \textit{V. S. Kulikov} [Math. USSR, Izv. 11, 957--989 (1977; Zbl 0387.14007)] on degenerations of complex surfaces. The hypothesis of ``minimality'' of the model \({\mathcal{X}}\) is more or less the assumption that the canonical divisor of \({\mathcal{X}}\) is numerically trivial. Toward higher dimensions, Calabi-Yau threefolds are studied looking at ``maximally unipotent'' degenerations. A morphism \(f: X\to S\) of algebraic varieties is said to be a \textit{Mori fiber space} if it is projective with connected fibres, and the anticanonial divisor \(-K_X\) is \(f\)-ample. If \(Y=\bigcup_i V_i\) is a simple normal crossings variety over \(k\) of dimension \(3\). \(Y\) is said to be a \textit{combinatorial Calabi-Yau of Type IV} if geometrically \(\bullet\) each component \(V_i\) is birational to a Mori fibre space over a unirational base; \(\bullet\) each connected component of every double surfaces \(S_{ij}\) is rational; \(\bullet\) each connected component of every triple curve \(C_{ijk}\) is rational; \(\bullet\) the dual graph \(\Gamma\) of \(Y\) is a triangulation of the \(3\)-sphere \(S^3\). The result is formulated as follows. {Theorem 2}: Let \({\mathcal{X}}\) be a strictly semistable \(R\)-scheme with generic fibre \(X\) a Calabi-Yau threefold. Assume moreover that the sheaf of logarithmic \(3\)-forms \(\omega_{{\mathcal{X}}}\) on \({\mathcal{X}}\) relative to \(R\) is trivial, and that \(N^3\neq 0\) on \(H^3(X)\) where \(N\) is a monodromy operator acting on \(H^3(X)\) such that \(N^4=0\). Then the special fibre \(Y\) of \({\mathcal{X}}\) is a combinatorial Calabi-Yau of Type \(IV\).