The EKOR-stratification on the Siegel modular variety with parahoric level structure

This paper investigates the arithmetic geometry of the reduction modulo $p$ of the Siegel modular variety with parahoric level structure by realizing its EKOR-stratification as the fibers of a smooth morphism into an algebraic stack parametrizing homogeneously polarized chains of truncated displays.

The Gist

Imagine you are a cartographer trying to draw a map of a mysterious, foggy island. This island is the Siegel Modular Variety. In the world of mathematics, this isn't a physical island, but a vast, complex geometric space that helps us understand the behavior of special shapes called Abelian varieties (which are like multi-dimensional doughnuts or toruses).

The author, Manuel Hoff, is trying to solve a specific problem: How do we navigate the "foggy" parts of this island?

Here is the breakdown of the paper using simple analogies:

1. The Setting: The Foggy Island

The island has two sides:

• The Sunny Side (Generic Fiber): This is the part of the island where the weather is clear. The geometry is smooth, predictable, and easy to walk on.
• The Foggy Side (Special Fiber): This is the part of the island that exists in "characteristic $p$" (a specific mathematical universe related to prime numbers). Here, the ground is cracked, jagged, and full of singularities (sharp points where the geometry breaks down). This is the "bad reduction" mentioned in the title.

Mathematicians have known for a long time that the foggy side is messy. They want to slice this messy terrain into neat, smooth strips so they can study it piece by piece. These strips are called strata.

2. The Old Map: The KR Strata

Previously, mathematicians (Rapoport and Zink) had a way to slice the island. They called these slices KR strata (Kottwitz-Rapoport).

• The Analogy: Imagine the island is a giant, crumpled piece of paper. The KR method draws lines on the paper based on how the paper is folded. It works, but it's a bit rigid. It tells you where you are, but it doesn't give you a smooth path to walk between the folds.

3. The New Map: The EKOR Strata

The paper focuses on a newer, more refined set of slices called EKOR strata (Ekedahl-Kottwitz-Oort-Rapoport).

• The Analogy: Think of the KR strata as dividing the island into broad "zones." The EKOR strata are like dividing those zones into specific "neighborhoods" based on the exact shape of the local terrain (specifically, the shape of the "p-torsion" of the doughnuts).
• The Goal: The author wants to prove that these EKOR neighborhoods are not just random patches of fog, but are actually smooth, well-behaved streets. If they are smooth, we can do calculus on them, which unlocks a lot of powerful mathematical tools.

4. The Magic Tool: The "Display" Machine

To prove these neighborhoods are smooth, the author invents a new machine. He calls it a Display (specifically, a "homogeneously polarized chain of displays").

• The Analogy: Imagine you have a complex, 3D sculpture (the Abelian variety) that is hard to photograph because it's foggy. You need a way to turn it into a flat, 2D blueprint that is easy to read.
• The Machine: The "Display" is a translation device. It takes the messy, foggy 3D sculpture and converts it into a clean, algebraic blueprint (a chain of modules and maps).
• The Innovation: Previous machines (like "F-zips" or "local shtukas") worked well only when the island was in a "hyperspecial" state (a very perfect, ideal version of the island). When the island was in a "parahoric" state (a more general, slightly broken version), those machines broke down or gave bad maps.
• Hoff's Fix: Hoff builds a new, upgraded machine (the $(m,n)$-truncated display) that works even when the island is broken or "parahoric." It's like a translator that can speak the dialect of the broken terrain perfectly.

5. The Grand Achievement: The Smooth Bridge

The main result of the paper is the construction of a smooth bridge (a mathematical morphism).

• The Bridge: Hoff builds a bridge that connects the messy, foggy island ($Ag,J,N$) directly to a clean, organized warehouse of blueprints (the algebraic stack of displays).
• Why it matters:
  1. Smoothness: Because the bridge is "smooth," it proves that the EKOR neighborhoods on the island are also smooth. The fog clears up!
  2. Navigation: It gives mathematicians a new way to navigate the island. Instead of stumbling through the fog, they can walk across the bridge, look at the blueprints, and know exactly what the terrain looks like.
  3. Closure: It helps them understand how these neighborhoods touch each other (the "closure relations").

Summary in One Sentence

Manuel Hoff has built a new, high-tech translation device that turns the messy, jagged geometry of a specific mathematical landscape into a clean, smooth blueprint, proving that the hidden neighborhoods within this landscape are actually perfectly smooth and navigable.

Why should a non-mathematician care?

While this sounds abstract, this kind of work is the "infrastructure" of modern number theory. It helps us understand the deep, hidden structures of numbers (like prime numbers). Just as civil engineers need to understand the soil before building a bridge, number theorists need to understand the "geometry of numbers" to solve problems related to cryptography, coding theory, and the fundamental laws of mathematics. This paper lays a new, stronger foundation for that infrastructure.

Technical Summary

Here is a detailed technical summary of the paper "The EKOR stratification on the Siegel modular variety with parahoric level structure" by Manuel Hoff.

1. Problem Statement and Context

The paper addresses the arithmetic geometry of the Siegel modular variety $A_{g,J,N}$ with parahoric level structure over the ring of $p$-adic integers $\mathbb{Z}_p$. Specifically, it focuses on the reduction of this variety modulo $p$.

• The Object of Study: $A_{g,J,N}$ is a quasi-projective scheme parametrizing polarized chains of $g$-dimensional Abelian varieties with full level $N$ structure, where the level structure is defined by a lattice chain indexed by a subset $J \subseteq \mathbb{Z}$.
• The Challenge: While the generic fiber is smooth, the special fiber (over $\mathbb{F}_p$) is typically singular. Understanding its geometry requires stratifying the special fiber.
• Existing Stratifications:
  • Kottwitz–Rapoport (KR) Stratification: Defined by fibers of a map to a quotient of the local model. These strata are smooth but do not capture the full isomorphism class of the $p$-torsion.
  • Ekedahl–Kottwitz–Oort–Rapoport (EKOR) Stratification: A refinement of the KR stratification. Two points lie in the same EKOR stratum if their associated $p$-torsion group schemes (or related Dieudonné data) are isomorphic in a specific sense.
• The Open Question: Can the EKOR stratification be realized as the fibers of a smooth morphism from the special fiber of the Shimura variety to a naturally defined algebraic stack? Previous results by Viehmann, Wedhorn, Shen, and Zhang established this for hyperspecial levels or specific cases, but a general construction for parahoric levels using "deperfection" (non-perfect) objects was lacking or incomplete.

2. Methodology

Hoff employs a strategy based on displays and truncated Witt vectors, moving away from perfectoid/perfect geometry to work with $p$-nilpotent rings.

A. Theoretical Framework: Displays and Truncation

• Displays: The author utilizes the theory of displays (introduced by Zink), which are characteristic $p$ analogues of Dieudonné modules but defined over non-perfect rings. A display is a tuple $(M, M_1, \Psi)$ where $M$ is a finite projective module over Witt vectors $W(R)$, $M_1$ is a specific submodule, and $\Psi$ is a "divided Frobenius."
• Truncated Displays: Hoff introduces $(m,n)$-truncated displays. These are variations where the Witt vectors and Frobenius maps are truncated to finite levels ($W_m(R)$ and $W_n(R)$). This is crucial because the standard "perfect" local shtuka approach (used by Xiao-Zhu and Shen-Yu-Zhang) fails to provide a smooth morphism in the general parahoric case without taking perfections.
• Chains and Polarization: The paper generalizes these concepts to chains of displays indexed by $J$, equipped with homogeneous polarizations. This structure mirrors the moduli problem of the Siegel modular variety.

B. Construction of the Target Stack

• The Stack $HPolChDisp_{g,J}$: The author constructs an algebraic stack parametrizing homogeneously polarized chains of $(m,n)$-truncated displays.
• Quotient Description: A key technical result (Proposition 1.4) shows that this stack admits a quotient stack description:
  $$HPolChDisp_{g,J}^{(m,n)} \cong [ (L^{(m)}G)_{\Delta} \backslash M_{loc,(n)} ]$$
  where $L^{(m)}G$ is the truncated positive loop group of the parahoric group $G$, and $M_{loc,(n)}$ is a torsor over the local model $M_{loc}$. This connects the moduli of displays directly to the geometry of the local model.

C. The Morphism

• The Map $\Upsilon$: Using the Serre–Tate theorem and the equivalence between $p$-divisible groups and displays (Theorem 1.2), Hoff constructs a natural morphism:
  $$\Upsilon: A_{g,J,N}^{\wedge} \to HPolChDisp_{g,J}$$
  where $A_{g,J,N}^{\wedge}$ is the $p$-completion of the Siegel modular variety.
• Relation to EKOR: The composition of this map with the projection to the stack of $(m, 1\text{-rdt})$-truncated displays (where $1\text{-rdt}$is a specific truncation level related to the reductive quotient) realizes the map$\upsilon$ whose fibers are the EKOR strata.

3. Key Contributions and Results

1. Realization of the EKOR Stratification via a Smooth Morphism

The main theorem (Theorem 1.5) states that for any integers $(m,n)$ (with $n \neq 1\text{-rdt}$), the natural morphism:
$$A_{g,J,N}^{\wedge} \to HPolChDisp_{g,J}^{(m,n)}$$
is smooth.
Furthermore, the induced morphism on the special fiber:
$$(A_{g,J,N})_{\mathbb{F}_p} \to HPolChDisp_{g,J}^{(m, 1\text{-rdt})}$$
is also smooth.

2. Proof Strategy for Smoothness

The proof relies on a deformation-theoretic argument:

• Formal Locus: By the Serre–Tate theorem, the smoothness of the morphism depends on the associated $p$-divisible groups. The morphism is shown to be smooth along the locus of formal $p$-divisible groups (where the display is $F$-nilpotent).
• Density Argument: The author proves that every point in the moduli space specializes to a point in the formal locus (Corollary 4.8). Since the morphism is smooth on a dense open set (the formal locus) and the target stack is well-behaved, the morphism is smooth everywhere.

3. Resolution of Previous Issues

The paper addresses gaps in previous literature (specifically regarding the work of Shen, Yu, and Zhang [SYZ21] and Xiao-Zhu [XZ17]):

• Previous attempts to define EKOR strata using local shtukas (which require perfect rings) encountered issues with commutativity of diagrams and the behavior of the morphism in the parahoric case.
• Hoff's use of truncated displays (non-perfect objects) allows for a construction that works directly over $\mathbb{Z}_p$ and $\mathbb{F}_p$ without needing to pass to the perfection of the special fiber, thereby avoiding the technical obstructions found in earlier proofs.

4. Quotient Stack Description

The paper provides a rigorous quotient stack description of the moduli of polarized chains of displays, linking them explicitly to the local model $M_{loc}$ and the loop groups. This clarifies the relationship between the EKOR stratification and the geometry of the local model.

4. Significance and Impact

• Geometric Understanding: The result provides a powerful new tool for studying the geometry of Shimura varieties at parahoric level. By realizing the EKOR stratification as fibers of a smooth morphism, one immediately deduces that the EKOR strata are smooth, locally closed subschemes and that their closure relations are governed by the geometry of the target stack.
• Generalization: This work generalizes the classical Ekedahl–Oort stratification (hyperspecial case) to the general parahoric setting, unifying the theory of $F$-zips, displays, and local shtukas.
• Methodological Advancement: The introduction and systematic use of $(m,n)$-truncated displays offers a robust framework for arithmetic geometry that avoids the complexities of perfectoid geometry while retaining enough structure to capture deep arithmetic invariants.
• Future Directions: The author suggests that this framework can be extended to general Shimura varieties of Hodge type and Abelian type at parahoric level, potentially defining stacks of $(G, \mu)$-displays for arbitrary parahoric data.

In summary, Manuel Hoff successfully constructs a smooth morphism from the Siegel modular variety with parahoric level structure to a stack of truncated displays, thereby providing a definitive geometric realization of the EKOR stratification and resolving smoothness questions that were previously open or only partially addressed in the literature.