Reduction of Kummer surfaces modulo 2 in the non-supersingular case

This paper establishes necessary and sufficient conditions for the good reduction of Kummer surfaces associated with abelian surfaces having non-supersingular reduction over a perfect field of characteristic 2, demonstrating that such reduction admits an explicit scheme model equivalent to an algebraic space model.

The Gist

Imagine you are an architect trying to build a perfect, smooth house (a Kummer surface) based on a blueprint that comes from a very strange, twisted foundation (an abelian surface).

Usually, if your foundation is solid, your house is solid. But in this paper, the authors are dealing with a very specific, tricky kind of foundation: one that exists in a world where the number 2 behaves differently than in our normal world (a mathematical world called "characteristic 2").

Here is the story of what they discovered, explained simply.

1. The Setup: The Twisted House

Think of an Abelian Surface as a complex, multi-layered doughnut shape. If you take this shape and fold it in half (matching every point with its opposite), you get a quotient. Usually, this folding creates a few sharp, jagged corners (singularities).

To make a Kummer Surface, you have to smooth out those jagged corners. In normal math worlds (where 2 is normal), this is easy: you just blow up the corners, and you get a beautiful, smooth K3 surface (a type of hyper-complex shape).

But in this paper, the "ground" is wet and slippery (characteristic 2). When you try to smooth out the corners here, the geometry gets messy. Sometimes the smoothing process fails, and the house collapses or becomes a weird, non-smooth pile of rubble.

2. The Problem: When Does the House Stay Standing?

The authors asked: "Under what exact conditions can we build a smooth Kummer house over this slippery ground?"

They found that it depends on how the "twist" of the foundation behaves. Imagine the foundation has little "handles" (points of order 2) that you can grab.

• Ordinary Case: The foundation has 4 handles.
• Almost Ordinary Case: The foundation has 2 handles.
• Supersingular Case: The foundation has 0 handles (this is the "broken" case where the house can't be saved, so they ignored it).

The big question is: Do these handles stay put, or do they wiggle around when you look at them from different angles?

3. The Discovery: The "Splitting" Rule

The authors discovered a simple rule for when the house stays smooth:

• The Ordinary Case (4 handles): The house is smooth if and only if the "handles" can be neatly separated into two groups: those that are stuck to the ground and those that float freely. In math speak, the "exact sequence splits." If the handles are tangled together in a specific way, the house will have a crack.
• The Almost Ordinary Case (2 handles): The house is smooth only if none of the handles are floating around. They must all be glued firmly to the ground. If even one handle is wiggling (defined over a larger field), the house cracks.

The Analogy: Imagine trying to stack blocks.

• In the Ordinary case, you can stack them if the bottom layer and top layer are aligned perfectly (split). If they are misaligned, the tower falls.
• In the Almost Ordinary case, the tower is so unstable that you can only stack it if every single block is perfectly still. If any block moves, the whole thing collapses.

4. The Solution: Building the House Explicitly

The paper doesn't just say "it works if X happens." The authors actually built the house.

They showed you exactly how to construct a smooth model (a blueprint for the house) using a technique called blowing up.

• Think of "blowing up" as taking a sharp corner and gently inflating it into a smooth curve, like blowing a bubble at the tip of a needle.
• They proved that if you follow their specific instructions for inflating the corners (which depend on the "handles" being in the right place), you get a perfect, smooth house that works over the integers (the ring of integers).

5. The "Twisted" Version

Finally, they looked at Twisted Kummer Surfaces. Imagine you take your blueprint and twist it slightly before building.

• They found that even if the original foundation (the Abelian surface) is too broken to build a smooth house, the twisted version might work!
• It's like taking a crooked piece of wood, twisting it the right way, and suddenly it fits perfectly into the frame. This is a surprising result: sometimes, the "broken" version becomes "fixed" just by changing the perspective (the twist).

Summary

In plain English, this paper solves a puzzle about building smooth geometric shapes in a weird mathematical world where the number 2 acts strangely.

1. The Rule: You can build a smooth shape if the "handles" of the foundation are arranged in a very specific, non-tangled way.
2. The Proof: They didn't just guess; they drew the blueprints and showed exactly how to smooth out the rough spots.
3. The Twist: Sometimes, if the original shape is broken, twisting it can fix it, allowing you to build a smooth house where you thought it was impossible.

This is important because these shapes (K3 surfaces) are like the "atoms" of higher-dimensional geometry. Understanding how they behave in tricky environments helps mathematicians understand the fundamental structure of the universe of numbers.

Technical Summary

Here is a detailed technical summary of the paper "Reduction of Kummer surfaces modulo 2 in the non-supersingular case" by C. D. Lazda and A. N. Skorobogatov.

1. Problem Statement

The paper addresses the problem of good reduction for Kummer surfaces associated with abelian surfaces over a complete discretely valued field $K$ of characteristic 0, where the residue field $k$ is perfect of characteristic 2.

• Context: A Kummer surface $\text{Kum}(A)$ is the minimal desingularization of the quotient $A/\iota$, where $\iota(P) = -P$ is the antipodal involution on an abelian surface $A$.
• The Challenge: In characteristic $\neq 2$, good reduction of $\text{Kum}(A)$ is equivalent to the existence of a quadratic twist $A_\chi$ with good reduction. However, in characteristic 2, this equivalence breaks down because the singular subscheme of $A/\iota$ is no longer étale.
• Specific Question: Given an abelian surface $A$ with good, non-supersingular reduction over $K$, what are the necessary and sufficient conditions for $\text{Kum}(A)$ to have good reduction (i.e., admit a smooth proper model over the ring of integers $\mathcal{O}_K$)?
• Scope: The authors restrict their analysis to the non-supersingular cases:
  • Ordinary: 2-rank $r=2$ (2-torsion has 4 points).
  • Almost Ordinary: 2-rank $r=1$ (2-torsion has 2 points).
  • Note: The supersingular case ($r=0$) is excluded as the geometry of the singularity (an elliptic singularity) prevents the resolution techniques used here from working.

2. Methodology

The authors employ a combination of explicit algebraic geometry, formal group theory, and Galois representation theory.

A. Explicit Resolution of Singularities

The core geometric strategy involves constructing explicit smooth models by resolving the singularities of the quotient scheme $Y = A/\iota$.

• Blow-up Commutativity: A crucial technical tool (Proposition 5.1) establishes that blowing up a section $Z$ commutes with specialization to the special fiber, provided the section meets both fibers at rational double points. This allows the authors to lift resolution steps from the special fiber to the generic fiber.
• Local Analysis:
  • Ordinary Case: The quotient $A/\iota$ has four singular points of type $D_4$. The authors analyze the formal local ring to show the exceptional curves of the minimal resolution are indexed by $A^\vee[2] \times A[2]$.
  • Almost Ordinary Case: The quotient has two singular points of type $D_8$. The authors use the decomposition of the formal group of $A$ (Lemma 3.2) into height 1 and height 2 components to explicitly resolve the singularities via a sequence of blow-ups.

B. Galois Action and Cohomology

To determine when a smooth model exists over $K$ (rather than just a finite extension), the authors compare Galois representations.

• Canonical Reduction: They utilize the theory of "canonical reduction" (from [CLL19] and [LM18]). If $\text{Kum}(A)$ has good reduction, its special fiber must be isomorphic to the canonical reduction of the generic fiber.
• Cohomological Comparison: They compare the $\ell$-adic étale cohomology of the generic fiber $\text{Kum}(A)_K$ and the special fiber $\text{Kum}(A)_k$.
  • $H^2_{\text{ét}}(\text{Kum}(A)_K, \mathbb{Q}_\ell) \cong H^2_{\text{ét}}(A_K, \mathbb{Q}_\ell) \oplus \mathbb{Q}_\ell(-1)^{\oplus W}$, where $W = A[2](K)$.
  • $H^2_{\text{ét}}(\text{Kum}(A)_k, \mathbb{Q}_\ell) \cong H^2_{\text{ét}}(A_k, \mathbb{Q}_\ell) \oplus \mathbb{Q}_\ell(-1)^{\oplus V}$, where $V$ depends on the 2-torsion of the special fiber.
• Splitting Criteria: The existence of a good reduction model is shown to be equivalent to the isomorphism of these Galois representations, which reduces to the splitting of specific exact sequences of $\Gamma_K$-modules involving the 2-torsion points.

3. Key Contributions and Results

A. Potential Good Reduction (Theorem 1)

The authors prove that for any abelian surface $A$ with good, non-supersingular reduction, $\text{Kum}(A)$ always attains potentially good reduction with a scheme model. This is achieved by constructing an explicit smooth model over a finite extension of $K$ that trivializes the Galois action on $A[2]$.

B. Necessary and Sufficient Conditions for Good Reduction

The paper provides precise criteria for when the reduction is good over the base field $K$ itself:

1. Ordinary Case (Theorem 2 & 6.3):
   $\text{Kum}(A)$ has good reduction over $K$ if and only if the connected-étale sequence of the 2-torsion group scheme splits:
   $$0 \to A[2]^\circ \to A[2] \to A[2]^{\text{ét}} \to 0$$
   Equivalently, the exact sequence of $\Gamma_K$-modules $0 \to A^\vee2 \to A2 \to A2 \to 0$ must split.

   • Significance: This condition is strictly stronger than $A[2](K)$ being unramified. The authors provide an example (Example 6.4) where $A[2](K)$ is unramified but the sequence does not split, meaning $\text{Kum}(A)$ only has good reduction over a non-trivial unramified extension.

2. Almost Ordinary Case (Theorem 3 & 6.3):
   $\text{Kum}(A)$ has good reduction over $K$ if and only if the $\Gamma_K$-module $A[2](K)$ is trivial (i.e., all 2-torsion points are defined over $K$).

   • Significance: This is a very restrictive condition, as it requires the full 2-torsion to be rational over $K$.

C. Twisted Kummer Surfaces (Section 7)

The authors extend their results to twisted Kummer surfaces $\text{Kum}(A_Z)$, associated with $A[2]$-torsors $Z$.

• They derive conditions for good reduction of twisted surfaces, which involve the unramified nature of the étale part of the torsor and the existence of sections for the map $Z \to Z^{\text{ét}}$.
• Counter-intuitive Result: They demonstrate (Examples 7.6 and 7.7) that a twisted Kummer surface $\text{Kum}(A_Z)$ can have good reduction over $K$ even if the standard Kummer surface $\text{Kum}(A)$ does not. This highlights the arithmetic subtlety of twisting in characteristic 2.

4. Significance and Impact

• Resolution of a Characteristic 2 Gap: Prior to this work, the criteria for good reduction of K3 surfaces (specifically Kummer surfaces) in characteristic 2 were not fully understood, particularly regarding the distinction between algebraic space models and scheme models. The authors prove that in these non-supersingular cases, scheme models suffice.
• Explicit Construction: Unlike previous results relying on Artin's simultaneous resolution (which is non-constructive), this paper provides explicit equations and blow-up sequences to construct the smooth models.
• Arithmetic Geometry: The results clarify the relationship between the arithmetic of the underlying abelian surface (specifically the Galois action on 2-torsion) and the geometry of the associated K3 surface. The finding that good reduction of the Kummer surface can fail even when the abelian surface has good reduction (due to non-splitting of the 2-torsion sequence) is a significant arithmetic insight.
• Twisting Phenomenon: The extension to twisted Kummer surfaces reveals that the obstruction to good reduction can be "cured" by twisting, a phenomenon not previously highlighted in this specific context.

In summary, this paper provides a complete characterization of the good reduction of non-supersingular Kummer surfaces in characteristic 2, bridging the gap between the geometry of singular quotients and the arithmetic of Galois representations on torsion points.