🔢 mathematics

Ekedahl-Oort types of stable curves

This paper extends the definition of Ekedahl-Oort types from smooth curves to all stable curves by utilizing Hasse-Witt triples, demonstrating their consistency with the generalized Jacobian approach, and applying this framework to compute the dimensions of specific Ekedahl-Oort loci and generalize existing results on $p$ -rank and $a$ -number loci.

Original authors: Dušan Dragutinović

Published 2026-01-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Dušan Dragutinović

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a cartographer trying to map a vast, mysterious landscape called the "Moduli Space of Curves." This isn't a landscape of mountains and rivers, but a mathematical universe where every possible shape of a smooth, closed loop (a "curve") of a certain complexity (genus) lives.

In this paper, the author, Dušan Dragutinović, is trying to solve a specific problem: How do we classify these curves based on how they behave in a specific mathematical "weather" called characteristic $p$ ?

Here is a breakdown of the paper's journey, using simple analogies.

1. The Problem: Classifying Shapes by Their "Fingerprint"

In this mathematical world, every curve has a hidden "fingerprint" called its Ekedahl-Oort type. Think of this fingerprint as a unique ID card that tells you how the curve reacts to a specific mathematical operation called the Frobenius operator.

The Old Way: Previously, to find this fingerprint, mathematicians had to build a complex machine called a "Jacobian" for every curve. It was like trying to identify a person by first building a full-scale robot replica of them and then checking the robot's ID. This worked for perfect, smooth curves, but it got messy and confusing when the curves had "cracks" or "knots" (mathematicians call these stable curves or singularities).
The New Way: The author introduces a new, "intrinsic" method. Instead of building a robot replica, he looks directly at the curve's own internal structure (specifically, a tool called a Hasse-Witt triple). It's like identifying a person just by looking at their face, without needing the robot.

2. The Big Discovery: The "Normalizing" Trick

The paper's first major breakthrough (Theorem A) is a revelation about curves with cracks.

Imagine a broken necklace (a singular curve). If you take it apart and straighten it out into a perfect, unbroken string (the "normalization"), you might think the ID card changes.

The Paper's Claim: Surprisingly, the author proves that the ID card stays exactly the same. Whether the curve is broken or perfect, its "Ekedahl-Oort type" is identical to that of its straightened-out version.
Why it matters: This means we can study the messy, broken curves by simply studying their clean, smooth versions. It simplifies the whole map.

3. The Inductive Ladder: Climbing the Mountain

Once the author established this new, simpler way to look at curves, he used it to climb a mathematical mountain. He wanted to know: "How big are the regions on the map where curves have specific fingerprints?"

The Analogy: Imagine the map is divided into zones. Some zones contain curves that are "very broken" (low p-rank), and others contain curves that are "very smooth" (high a-number).
The Technique: The author uses an "inductive technique." Think of it like climbing a ladder. If you know the size of a room on the 3rd floor, you can calculate the size of the room on the 4th floor without measuring it from scratch.
The Result: He created a formula to calculate the maximum possible size (dimension) of these zones for curves of any complexity ( $g$ ). Before this, mathematicians could only guess the sizes for very simple curves. Now, they have a precise ruler for much more complex ones.

4. New Boundaries and "Forbidden" Zones

Using this new ruler, the author found some surprising things:

New Limits: He calculated strict upper limits for the size of zones where curves have a "high a-number" (a specific measure of complexity). For example, he proved that the zone of curves with a specific "super-complex" fingerprint (type $[3,2,1]$ ) cannot be as big as people might have hoped. It's smaller than the "expected" size.
The "Empty" Zones: In specific mathematical "weathers" (characteristic 2 and 3), he found that certain fingerprints simply do not exist for smooth curves of certain sizes.
- Example: In characteristic 2, there are no smooth curves of genus 4 with a specific fingerprint $[4,3]$ . It's like trying to find a square circle; the math says it's impossible in that specific environment.
- Example: In characteristic 3, there is exactly one hyperelliptic curve (a specific type of curve) with a certain fingerprint. It's a unique, one-of-a-kind shape.

5. The "Hyperelliptic" Shortcut

The author also looked at a special family of curves called hyperelliptic curves (think of them as curves that look like a figure-eight or have a specific symmetry).

He showed that by studying these simpler, symmetric curves, you can deduce the rules for the entire, messy landscape.
In characteristic 3, he used the properties of these symmetric curves to prove that certain "forbidden" fingerprints are indeed impossible for larger curves.

Summary

In plain English, this paper does three main things:

Simplifies the Rules: It proves you don't need complex machinery to identify the "fingerprint" of a broken curve; you can just look at its smooth version.
Builds a Calculator: It provides a method to calculate the exact size of regions on the mathematical map where curves have specific properties.
Maps the Edges: It identifies exactly which "fingerprints" are possible and which are impossible in specific mathematical environments (like characteristic 2 and 3), revealing that some shapes simply cannot exist in those worlds.

The paper doesn't talk about real-world applications like engineering or medicine. It is purely about understanding the fundamental geometry and classification of these abstract mathematical shapes.

Technical Summary: Ekedahl-Oort Types of Stable Curves

Problem Statement
The paper addresses the interpretation and computation of Ekedahl-Oort types for stable curves in characteristic $p > 0$ . While the Ekedahl-Oort type of a smooth curve is classically defined via its Jacobian (a principally polarized abelian variety), extending this definition to stable curves (which have generalized Jacobians that are semi-abelian varieties) requires careful handling. The author seeks to establish an intrinsic definition of these types that does not rely solely on the generalized Jacobian, thereby facilitating the application of inductive techniques to compute the dimensions of Ekedahl-Oort loci within the moduli space of stable curves, $\mathcal{M}_g$ . Specifically, the paper aims to generalize existing results on the dimensions of $p$ -rank and $a$ -number loci and to compute bounds for new, more complex Ekedahl-Oort strata.

Methodology
The core methodological contribution is the extension of Moonen's concept of Hasse-Witt triples from smooth curves to all stable curves.

Intrinsic Definition: The author defines the Ekedahl-Oort type of a stable curve $C$ directly via the Hasse-Witt triple $(Q, \Phi, \Psi)$ , where $Q = H^1(C, \mathcal{O}_C)$ , $\Phi$ is the Frobenius operator, and $\Psi$ is a specific $\sigma$ -linear bijection. This definition is intrinsic to the curve's cohomology and does not require passing through the generalized Jacobian.
Equivalence Proof: A central technical step involves proving that this intrinsic definition coincides with the classical definition derived from the generalized Jacobian $J_C$ . This is achieved by analyzing the normalization $h: \tilde{C} \to C$ and establishing that the Dieudonné module of $J_C$ decomposes into the direct sum of the Dieudonné module of the torus part and the Dieudonné module of the Jacobian of the normalization, $\tilde{C}$ .
Inductive Technique: The paper leverages the "clutching" morphisms (identifying points on lower-genus curves to form higher-genus curves) and the boundary divisor $\Delta_0$ of $\mathcal{M}_g$ . By utilizing the intrinsic definition, the author applies an inductive argument (generalizing techniques from [FvdG04]) to bound the dimensions of Ekedahl-Oort strata. The logic relies on the fact that if a stratum intersects the boundary $\Delta_0$ , its components can be related to strata in $\mathcal{M}_{g-1}$ .
Combinatorial Constraints: The inductive bounds are established under specific technical conditions on the Young diagrams $\mu$ defining the strata. These conditions relate to the inclusion relations of Ekedahl-Oort loci in the moduli space of abelian varieties $\mathcal{A}_g$ and the behavior of the Weyl group $W_g$ .

Key Contributions and Results

Theorem A (Equivalence): The paper proves that for a stable curve $C$ , its generalized Jacobian $J_C$ , and its normalization $\tilde{C}$ , the Ekedahl-Oort types are identical: $\mu(J_C) = \mu(C) = \mu(\tilde{C})$ . Here, $\mu(C)$ is defined via the intrinsic Hasse-Witt triple, while $\mu(J_C)$ is the classical definition. This validates the use of the intrinsic approach for stable curves.
Theorem B (Dimension Bounds): The author establishes a general inductive formula for the codimension of Ekedahl-Oort strata. If a Young diagram $\mu$ satisfies a specific technical condition (related to the inclusion of strata in $\mathcal{A}_g$ and $\mathcal{A}_{g+1}$ ), then any component $\Gamma$ of the locus $Z_\mu(\mathcal{M}_{g+1})$ that intersects the boundary $\Delta_0$ has a codimension in $\mathcal{M}_{g+1}$ bounded below by a value determined by the maximum dimension of components in $Z_\mu(\mathcal{M}_g)$ .
Generalization of Known Results: This framework recovers and generalizes previous results regarding the dimensions of loci with $p$ -rank $\le f$ and $a$ -number $\ge a$ .
New Bounds for Higher $a$ -numbers: The technique allows for the computation of upper bounds for loci that were previously out of reach, specifically the locus $T_3(\mathcal{M}_g)$ of curves with $a$ -number $\ge 3$ . The paper proves that $\dim T_3(\mathcal{M}_g) \le 3g - 8$ .
Hyperelliptic Curves and Specific Characteristics:
- The paper provides a criterion (Theorem 5.6) to compute the dimensions of Ekedahl-Oort loci $Z_\mu(\mathcal{M}_4)$ based on the properties of hyperelliptic curves.
- In characteristic $p=2$ , it is shown that $Z_{[4,3]}(\mathcal{M}_4)$ is empty, while $Z_{[3,2]}(\mathcal{M}_4)$ is non-empty and has the expected dimension.
- In characteristic $p=3$ , the paper demonstrates the existence of a unique hyperelliptic curve with Ekedahl-Oort type $[4,3]$ . This finiteness result is used to prove that for $\mu > [4,3]$ , the locus $Z_\mu(\mathcal{M}_4)$ is non-empty and pure of the expected codimension.
- It is shown that for $g \ge 5$ in characteristic 3, no smooth hyperelliptic curves satisfy $F_C^2 = 0$ , leading to the emptiness of certain Ekedahl-Oort loci.

Significance
The paper claims significance primarily through the unification of intrinsic and classical definitions, which unlocks the full potential of inductive techniques for studying the geometry of moduli spaces of stable curves. By moving away from a reliance on the generalized Jacobian for the definition of the type, the author can directly relate the cohomological properties of singular curves to their normalizations. This shift enables the derivation of dimension bounds for Ekedahl-Oort strata that were previously inaccessible, particularly for higher $a$ -numbers. The results provide a more complete "geography" of Ekedahl-Oort stratifications in $\mathcal{M}_g$ , clarifying the dimensions of components and the existence of curves with specific $p$ -torsion structures in low characteristics. The work serves as a bridge between the combinatorial theory of Ekedahl-Oort types in $\mathcal{A}_g$ and the geometric reality of stable curves in $\mathcal{M}_g$ .