Numerically trivial involutions of Kummer type of an Enriques surface (Q624655)

A non-trivial involution on a complex \(K3\) surface acts non-trivially on its cohomology (i.e.~on \(H^2\)) by the Torelli theorem. It is a remarkable fact that the same does not hold true for Enriques surfaces. In this context one distinguishes two kinds of involutions \(\imath\) on an Enriques surface \(S\). 1) If \(\imath^*\) acts trivially on \(H^2(S,\mathbb{Z})\), then \(\imath\) is called cohomologically trivial. 2) If \(\imath^*\) acts trivially on \(H^2(S,\mathbb{Q})\), then \(\imath\) is called numerically trivial. In [\textit{S. Mukai, Y. Namikawa}, Invent. Math. 73, 383--411 (1983; Zbl 0518.14023)] \textit{W. Barth} and \textit{C. Peters} exhibited a two-dimensional family of Enriques surfaces with a cohomologically trivial involution. Here the covering \(K3\) surfaces are generally not of Kummer type. In [\textit{S. Mukai, Y. Namikawa}, Invent. Math. 77, 383--397 (1984; Zbl 0559.14038)] it was claimed that a second family arises from Kummer surfaces of product type as Liebermann involutions were claimed to be cohomologically trivial. This is corrected in the present paper: arguing with some natural elliptic fibrations, it is shown that involutions of Liebermann type are numerically trivial, but not cohomologically trivial. The author then continues by classifying Enriques surfaces of Kummer type with numerically trivial involution. Next to the above family, there is one further construction that was discovered by \textit{S. Kondo} (see [Jap. J. Math., New Ser. 12, 191--282 (1986; Zbl 0616.14031)]) after going unnoticed in [Zbl 0559.14038]. The present paper derives a numerically trivial involution from standard Cremona involutions of the quadric surface. This also reveals relations to quartic del Pezzo surfaces. The main result of the present paper is the following: any Enriques surface of Kummer type with a numerically trivial involution arises through one of the above two constructions.