Algebraic geometers study special geometric objects called K3 surfaces, which are smooth two-dimensional shapes with particularly nice mathematical properties. One well-known way to build a K3 surface is to start with an abelian surface, which is a higher-dimensional version of a donut shape (technically a torus with extra algebraic structure), then fold it by a finite symmetry group and smooth out the resulting sharp points. The surfaces built this way are called generalised Kummer surfaces, named after a nineteenth-century mathematician. The central question of this paper is: exactly which symmetry groups can be used in this construction, and in what geometric settings does it work?
The main challenge is that the answer depends on the underlying number system, or "characteristic," in which the geometry lives. Most classical results assume you are working over the ordinary real or complex numbers, but geometry can also be done over number systems built from prime numbers, called positive characteristic. In those settings, strange phenomena can occur that have no classical analogue. The authors systematically work through the cases where the size of the symmetry group is divisible by the characteristic prime (2, 3, or 5), which is precisely where things get complicated. They use a technical toolkit involving singularity theory and a structure called the Tate module, which tracks how the symmetry group acts on the abelian surface, to pin down exactly what types of singularities can appear and which groups are allowed.
A key finding is that when the abelian surface is "supersingular," a special degenerate case that only occurs in positive characteristic, the construction simply never works if the group size shares a factor with the characteristic. This rules out a large family of potential examples. On the positive side, the authors construct explicit working examples by taking products of two elliptic curves (donut-shaped curves) and quotienting by carefully chosen groups. Together these results complete a full classification, extending prior partial results by Katsura and Rybakov, and providing a definitive map of all generalised Kummer surfaces across every possible characteristic.