The supersingular locus of the Shimura variety for \(\mathrm{GU}(1, n-1)\) over a ramified prime (Q2454436)

To study arithmetic properties of Shimura variety, it is a standard strategy to define a model over (a localization or completion at some prime ideal of) the ring of integers of its field of definition, and then to analyze the geometric structure of the special fiber, a variety defined over a finite field. One can then exploit the specifics of varieties over fields of positive characteristic. For example, assuming that the variety in question is a moduli space of abelian varieties, a typical situation in the setting of Shimura varieties, one has the Newton stratification, whose strata consist of points where the isogeny class of the \(p\)-divisible group of the corresponding abelian variety (with additional structure) is fixed. There is a unique closed Newton stratum, called the basic locus. While it is too much to hope for an explicit description of the whole special fiber, an explicit description of the basic locus is possible in favorable situations. The purpose of the paper at hand is to give an explicit description of the basic locus (which in this case coincides with the supersingular locus) of the Shimura varieties for the group \(GU(1, n-1)\) at a ramified prime. Via the uniformization theorem of Rapoport and Zink, the problem can be translated into describing a certain moduli space of \(p\)-divisible groups, the corresponding ``Rapoport-Zink space''. In this way, the problem can be attacked using Dieudonné theory and related methods. As their main result, the authors obtain a description of the underlying reduced scheme of the basic Rapoport-Zink space along the following lines: It is a union of strata, and each stratum is isomorphic to a Deligne-Lusztig variety for a symplectic group (of varying rank) over \(\mathbb F_p\), attached to a Coxeter element. The closure of each stratum is a union of strata. The index set of the stratification and the closure relations can be described explicitly in terms of the Bruhat-Tits building of a special unitary group over \(\mathbb Q_p\) (a twisted form of the derived group of the original unitary group). The closures of strata are normal with isolated singularities. For example, if \(n=3\), then the basic locus is \(1\)-dimensional. All its irreducible components are projective lines. Each projective line contains \(p+1\) ``special points'', and each special point lies on \(p+1\) projective lines (and the projective lines intersect only in special points, and any two projective lines intersect in at most one point). The special points are exactly the \(0\)-dimensional strata; each projective line minus its special points is a \(1\)-dimensional stratum. For more precise statements, we refer to the introduction of the paper. These results are useful to study special subvarieties and their intersections, e.g., in the Kudla program, and for the Arithmetic Gan-Gross-Prasad conjecture and the Arithmetic Fundamental Lemma conjecture of W. Zhang. A similar description in the unramified case was previously given \textit{I. Vollaard} and \textit{T. Wedhorn} [Invent. Math. 184, No. 3, 591--627 (2011; Zbl 1227.14027)]. Compare also the paper [``Basic loci of Coxeter type in Shimura varieties'', \url{arXiv:1311.6263}], by \textit{X. He} and the reviewer which describes a general group-theoretic approach to the analogous problem in the equal characteristic case.