Apéry-Fermi pencil of \(K3\)-surfaces and 2-isogenies (Q2177024)

The equation \[ X + \frac 1X + Y + \frac 1Y + Z + \frac 1Z = k, \quad k \in \mathbb{C} \] defines a pencil of \(K3\) surfaces \(Y_k\), called the Apéry-Fermi pencil, which also appears in quantum field theory. By results of \textit{C. Peters} and \textit{J. Stienstra} [Lect. Notes Math. 1399, 110--127 (1989; Zbl 0701.14037)], the transcendental lattice of the generic member \(Y_k\) of the pencil is known to be isometric to \(T = U \oplus \langle 12 \rangle\). The present paper contains a careful case-by-case analysis of the elliptic fibrations on \(Y_k\), with a special view towards their \(2\)-torsion sections. Such sections are indeed interesting because their associated fibrewise translation is a symplectic involution \(i\colon Y_k \rightarrow Y_k\), called a van Geemen-Sarti involution, which gives rise to a \(2\)-isogeny (a rational map of degree \(2\)) between \(Y_k\) and the resolution of the quotient \(Y_k/i\). One says that \(i\) induces a Shioda-Inose structure, or that \(i\) is a Morrison-Nikulin involution, if the resolution of \(Y_k/i\) is a Kummer surface of transcendental lattice \(T(2)\). In Section 2 and 3, the authors enumerate all \(27\) elliptic fibrations on \(Y_k\) together with their singular fibers, Mordell-Weil rank, torsion sections and Weierstraß equations. The study of the \(12\) van Geemen-Sarti involutions is carried out in Section 5. As it turns out, \(7\) of them are Morrison-Nikulin involutions, while the other \(5\) induce a \(2\)-isogeny onto a \(K3\) surface of transcendental lattice \(\langle -2 \rangle \oplus \langle 2 \rangle \oplus \langle 6 \rangle\). In Section 6 the same classification is carried out for the surface \(Y_2\), which is a singular \(K3\) surface of transcendental lattice \(\langle 2 \rangle \oplus \langle 4 \rangle\), previously studied by the authors [Fields Inst. Commun. 67, 153--199 (2013; Zbl 1312.14090)]. There are \(30\) elliptic fibrations and \(20\) van Geemen-Sarti involutions, of which \(13\) are Morrison-Nikulin involutions. In other \(5\) cases the resolution of \(Y_2/i\) is isomorphic to \(Y_2\) itself, giving rise to a `self \(2\)-isogeny', a rare phenomenon that only appears for \(k = 2\) or \(k = 10\) if \(k \in \mathbb Z\).