Compact moduli of elliptic K3 surfaces (Q6138570)

The authors study compactifications of moduli spaces of elliptic \(K3\) surfaces. These moduli spaces are constructed via the framework of KSBA stable pairs, using the earlier work [\textit{K. Ascher} and \textit{D. Bejleri}, Proc. Lond. Math. Soc. (3) 122, No. 5, 617--677 (2021; Zbl 1464.14037)], where they constructed compactifications of moduli spaces of log canonical models of Weierstrass elliptic surface pairs. There are other compactifications of moduli spaces of \(K3\) surfaces (e.g. GIT or the Baily-Borel compactification) but, unlike the spaces constructed via the KSBA framework, such compactifications lack an intrinsic modular interpretation at the boundary. The authors construct explicit maps between their KSBA spaces and GIT and Satake-Baily-Borel (SBB) compactifications to endow their boundaries with a modular meaning. The main results of the paper are as follows. If \(C\) is a smooth curve and \(f:X \rightarrow C\) is an elliptic \(K3\) surface with a section \(S\), the authors define a KSBA compactification \(\overline{\mathcal{W}}(\mathcal{A})\) of the moduli space of pairs \((f:X \rightarrow C, S+F_{\mathcal{A}})\), where \[ F_{\mathcal{A}} = \sum a_i F_i \] is a \(\mathbb{Q}\)-divisor defined for \(a_i \in \mathbb{Q} \cap [0,1]^{24}\), where \(\mathcal{A} = (a_1, \ldots, a_{24})\) and \(F_i\) are the 24 nodal fibres of \(f\). For \(\mathcal{A} = (a, \ldots, a)\), they let \(\overline{\mathcal{W}}(a)\) denote \(\overline{\mathcal{W}}(\mathcal{A})\) and define \(\overline{\mathcal{W}}_{\sigma}(a)\) as the quotient of \(\overline{\mathcal{W}}(a)\) by the symmetric group \(S_{24}\) acting on the fibres \(F_i\). If \(\mathcal{W}\) is the moduli space of elliptic \(K3\) surfaces with a section, constructed as a locally symmetric variety with coarse space \(W\) and \(\overline{W}^*\) is the Baily-Borel compactification of \(W\), the authors prove that: ``(Theorems 6.13, 6.15 and 6.14 and Figure 1) The proper Deligne-Mumford stacks \(\overline{\mathcal{W}}(a)\) for \(a \in \mathbb{Q} \cap [0,1]\) give modular compactifications of \(\mathcal{W}\). There is an explicit classification of the broken elliptic \(K3\) surfaces parametrized by \(\overline{\mathcal{W}}_{\sigma}(\epsilon)\), and an explicit morphism from the coarse space \(\overline{W}_{\sigma}(\epsilon)\) to \(\overline{W}^*\), the SBB compactification of \(W\). Furthermore, the surfaces parametrized by \(\overline{\mathcal{W}}_{\sigma}(\epsilon)\) satisfy \(H^1(X, \mathcal{O}_X)=0\) and \(\omega_X \cong \mathcal{O}_X\).'' They also consider the space \(\overline{\mathcal{K}}_{\epsilon}\), obtained as a compactification of the moduli space of pairs \((f:X \rightarrow \mathbb{P}^1, S + \epsilon F)\) where \(F\) is a single nodal fibre, and the associated coarse moduli space \(\overline{K}_{\epsilon}\). They prove that: ``(Theorems 8.1 and 8.2, and Figure 1) The compact moduli space \(\overline{\mathcal{K}}_{\epsilon}\) parametrizes irreducible semi-log canonical Weierstrass elliptic \(K3\) surfaces satisfying \(H^1(X, \mathcal{O}_X) = 0 \) and \(\omega_X \cong \mathcal{O}_X\). Moreover, there is an explicit generically finite morphism from the coarse space \(\overline{K}_{\epsilon}\) to \(\overline{W}^*\).'' The final moduli space they consider is \(\overline{\mathcal{F}}_{\epsilon}\), defined as for \(\overline{\mathcal{K}}(\epsilon)\), but where \(F\) is any fibre of \(f\) (i.e. not necessarily singular) with weight \(\epsilon \ll 1\). They prove that: ``(Theorem 8.8 and Figure 1) There exists a smooth proper Deligne-Mumford stack \(\overline{\mathcal{F}}_{\epsilon}\) parametrizing semi-log canonical elliptic \(K3\) surfaces with a single marked fiber. Its coarse space is isomorphic to an explicit GIT quotient \(\widetilde{W}^G\) of Weierstrass \(K3\) surfaces and a chosen fiber. Furthermore, the surfaces parametrized by \(\overline{\mathcal{F}}_{\epsilon}\) satisfy \(H^1(X, \mathcal{O}_X) = 0 \) and \(\omega_X \cong \mathcal{O}_X\).'' Much of the paper is dedicated to the results of \S5 and \S6, in which a detailed analysis of the surfaces parametrized by \(\overline{\mathcal{W}}(a)\) is performed as \(a\) varies. This is via an explicit description of wall crossings using MMP techniques. In \S7, the authors enumerate the boundary strata of \(\overline{\mathcal{W}}\), in the spirit of Kulikov models. Stable pairs compactification of the space of elliptic \(K3\) surfaces were also constructed by Brunyate using Kulikov models: these results offer a new approach.