Classification of degenerations of codimension \({\le}\,5\) and their Picard lattices for Kählerian \(K3\) surfaces with the symplectic automorphism group \((C_2)^2\) (Q6156021)

The present paper continues the author's work on symplectic automorphisms of complex \(K3\) surfaces in the greatest generality, i.e. including non-projective surfaces (which are Kähler by \textit{Y. T. Siu} [Invent. Math. 73, 139--150 (1983; Zbl 0557.32004)] ). A finite group \(G\) acts symplectically on some \(K3\) surface \(X\) if it leaves the regular 2-form invariant. Since much of the theory is goverened by lattice theory thanks to the Torelli theorem (and the surjectivity of the period map), the symplectic \(G\)-action can be understood in terms of the invariant lattice \(S^G=\mathrm{Pic}(X)^G\) and particularly the coinvariant lattice \[ S_G = (S^G)^\perp\subset\mathrm{Pic}(X). \] Remarkably, this is a negative-definite even lattice without roots (the author cals it 'of Leech-type'). Moreover, \(S_G\) is uniquely determined by \(G\) if \(G\) is abelian by \textit{V. V. Nikulin} [Trans. Mosc. Math. Soc. 2, 71--135 (1980; Zbl 0454.14017)], and also for most other finite groups by \textit{K. Hashimoto} [Nagoya Math. J. 206, 99--153 (2012; Zbl 1243.14029)]. In any case, it follows that \(K3\) surfaces with a symplectic \(G\)-action move in families of dimension \(20-\mbox{rk } S_G\) (and one dimension less for projective \(K3\) surfaces). The degenerations considered by the author are roughly characterised by acquiring additional smooth rational curves (also referred to as \((-2)\)-curves for their self-intersection number in \(\mathrm{Pic}(X)\)). The precise condition is as follows: Let \(C_1,\hdots, C_t\subset X\) be smooth rational curves such that their \(G\)-orbits \(G(C_1), \hdots, G(C_t)\) are distinct. Consider the sublattice \[ L_0 = \langle S_G, C_1,\hdots, C_t\rangle\subset\mathrm{Pic}(X) \] generated by \(S_G\) and those \((-2)\)-curves and let \(L = (L\otimes\mathbb Q)\cap\mathrm{Pic}(X)\) denote the primitive closure of \(L_0\). Then the condition requires that \(L\) still is negative definite. In practice, this implies that all curves in the orbits \(G(C_1),\hdots, G(C_t)\) can be contracted to ADE singularities on a singular surface \(X_0\) which still admits a \(G\)-action. Such surfaces are considered as codimension \(t\) degenerations of the original smooth surfaces. Indeed, since \(\mathrm{rank} L = \mathrm{rank} S_G + t\), these surfaces move in families of dimension \(20 - \mathrm{rank} S_G - t\), and the very general member will have Picard lattice \(L\). Building on previous work for groups of higher order, the author works out the complete classification of codimension \(t\leq 5\) degenerations for the case \(G=(\mathbb Z/2\mathbb Z)^2\). This is achieved purely in lattice theoretic terms by some massive tables and lists. For instance, already for \(t=1\), there are three cases to consider, depending on the orbit length of \(G(C_1)\); for larger \(t\), the classification becomes increasingly complicated.