Good Reduction of Enriques Surfaces

• Good reduction of Enriques surfaces is defined by the existence of a smooth proper model over discrete valuation rings and a K3 double cover with unramified ℓ-adic cohomology.
• The study employs lattice-theoretic, crystalline, and automorphism extendability criteria to characterize the reduction behavior and invariant subspaces.
• These results bridge geometric and arithmetic aspects, underpinning finiteness theorems and moduli implications for Enriques surfaces and their K3 covers.

An Enriques surface is a minimal smooth projective surface of Kodaira dimension zero, with $b_2 = 10$ and numerically trivial canonical bundle $\omega_X$, but such that $\omega_X \not\cong \mathcal{O}_X$ and $\omega_X^{\otimes 2} \cong \mathcal{O}_X$. Over any field of characteristic not 2, Enriques surfaces admit a canonical finite étale double cover by a K3 surface. The study of good reduction for Enriques surfaces investigates the existence and properties of smooth proper models over discrete valuation rings and relates these to the behavior of their K3 double covers, ℓ-adic cohomological invariants, and automorphisms, with particular focus on implications for moduli and arithmetic finiteness results.

1. Classical and Cohomological Notions of Good Reduction

Given a discretely valued field $K$ with ring of integers $O_K$ and residue characteristic $p \ne 2$, an Enriques surface $X/K$ is said to have good reduction if it admits a smooth, proper model $\mathcal{X}/O_K$ with generic fiber isomorphic to $X$ (Takamatsu, 2019, Zhao, 26 Nov 2025). For Enriques surfaces, this notion is equivalent to the existence of a smooth, projective $O_K$-scheme with generic fiber $X$ (as opposed to more general algebraic spaces necessary for some K3 surfaces).

Cohomological refinements are central in modern treatments. If $X/K$ is an Enriques surface, a K3 double cover $\widetilde{X} \to X$ exists and is unique up to isomorphism determined by a choice of isomorphism $\omega_{X/K} \simeq \mathcal{O}_X$. The surface $X$ is said to admit a cohomological good K3 cover if, for some K3 cover $\pi: \widetilde{X} \to X$, the Galois representation $$H^2_{\text{et}}(\widetilde{X}_{\overline K}, \mathbb{Q}_\ell)$$ is unramified for all $\ell \ne p$, i.e., the inertia group $I_K$ acts trivially [(Takamatsu, 2019), Definition 3.2]. This is a strictly weaker condition than the existence of a smooth proper model for $X$, as it only requires the unramifiedness of the cohomology of the K3 cover, not that $X$ itself has a smooth model.

A third, intermediate notion—flower-pot reduction—describes the scenario where $X$ fails to have a smooth model, but its K3 cover $\widetilde{X}$ does, so its cohomology is unramified, but $X$ does not have good reduction in the classical sense (Takamatsu, 2019).

2. Characterizations via K3 Covers and ℓ-adic Cohomology

Let $E/K$ be an Enriques surface, $\pi: X \to E$ its canonical K3 double cover, and $R$ a DVR with residue field of characteristic $p \ne 2$. Good reduction of $E$ can be characterized through the interaction of $X$ and the fixed-point-free involution $\iota \in \operatorname{Aut}(X)$ corresponding to the deck transformation of the double cover (Zhao, 26 Nov 2025):

• $E$ has good reduction over $R$ if and only if $X$ has good reduction over $R$ and $\iota$ extends to an involution of the smooth model $\mathcal{X}/R$ such that the special fiber involution remains fixed-point-free.
• This fixed-point-freeness is equivalent to the condition that the subspace of $\ell$-adic étale cohomology $$H^2_{\text{et}}(X_{\overline{k}}, \mathbb{Q}_\ell)^{\iota_k = \mathrm{id}}$$ has dimension $10$. This cohomological invariant is stable under reduction if and only if the involution extends correctly (Zhao, 26 Nov 2025).

Thus, once a smooth model of $X$ is established, the extendability of the Enriques structure is reduced to a check on $\ell$-adic cohomology invariants. For group actions more generally, the extendability to smooth models is characterized through the equivariance of the specialization isomorphism on second cohomology [(Zhao, 26 Nov 2025), Section 3.1].

3. The Shafarevich Finiteness Theorem

The Shafarevich finiteness theorem for Enriques surfaces asserts:

Given a finitely generated field $F$ over $\mathbb{Q}$ and a finite set of primes $\Sigma$, the set $\mathrm{Shaf}$ of $F$-isomorphism classes of Enriques surfaces $X/F$ which admit a cohomological good K3 cover at every $p \in \Sigma$ is finite [(Takamatsu, 2019), Theorem 3.7].

If one restricts to honest (classical) good reduction at every $p \in \Sigma$, the resulting subset is finite as well [(Takamatsu, 2019), Corollary 3.8].

The proof proceeds by associating to each Enriques surface $X$ its K3 cover $\widetilde{X}$, using Picard triviality to globalize twist data across $\Sigma$, and invoking the Shafarevich finiteness for K3 surfaces with unramified $H^2$, along with the finiteness of Enriques quotients up to isomorphism for a fixed K3 surface [(Takamatsu, 2019), Lemma 3.6].

4. Lattice-Theoretic and Crystalline Criteria

Explicit cases demonstrate the lattice-theoretic and crystalline torsion characterization of good reduction. For instance, the elliptic modular surface of level $4$, a complex K3 surface with Picard number $20$, admits a fixed-point-free involution—an Enriques involution of type IV (Shimada, 2018). The reduction modulo $3$ specializes to the Fermat quartic, a supersingular K3 surface of Artin invariant $1$, with a primitive embedding of the Néron–Severi lattice.

Necessary and sufficient conditions for the involution to yield a good reduction of an Enriques structure are:

• The invariant sublattice $\mathrm{NS}(X_0)^+$ is even unimodular of rank $10$ (isomorphic to $\mathrm{NS}(Y_0)(2)$),
• The anti-invariant part $\mathrm{NS}(X_0)^-$ is root-free,
• These properties are preserved under specialization due to the primitive embedding $\mathrm{NS}(X_0) \hookrightarrow \mathrm{NS}(X_3)$, so the involution remains fixed-point-free on the special fiber.

Crystalline Torelli theorems (Ogus–Madapusi Pera) guarantee that the period-condition for the involution remains intact under specialization, ensuring the involution lifts across the deformation and provides good reduction to the quotient Enriques surface (Shimada, 2018).

5. Extendability and Automorphisms

Any projective automorphism of the generic fiber of a smooth, proper K3 surface over a DVR that preserves a relatively ample line bundle extends to the whole family. Explicitly, the automorphism groups, including the $\mathbb{Z}/2^5$-Galois automorphism group arising from the double-plane and involutory structure, as well as additional involutions, all extend over DVRs of mixed characteristic (Shimada, 2018). In the Enriques context, the involutive structure essential for the quotient and the invariants controlling good reduction are stable under such descent.

For finite group actions $G$ (including symplectic groups on K3 surfaces), the action extends to the smooth model as long as the residue characteristic does not divide $|G|$ and the associated cohomological criteria are fulfilled [(Zhao, 26 Nov 2025), Section 5.1].

6. Special Phenomena: Flower-Pot Reduction

Flower-pot reduction denotes the phenomenon where an Enriques surface over a local field lacks a smooth model, but its K3 cover does admit one. In such cases, the cohomological good K3 cover condition holds but not classical good reduction [(Takamatsu, 2019), Section 5]. The finiteness theorem thus applies in both settings, but flower-pot reduction marks a strict weakening of classical reduction, interpolating between reduction behaviors of Enriques and their K3 covers.

7. Implications and Further Directions

The bridge between the good reduction of Enriques surfaces and their K3 double covers enables leveraging advanced results for K3 surfaces (such as the Shafarevich theorem and the crystalline Torelli theorem) to settle questions for Enriques surfaces. The precise cohomological, lattice-theoretic, and automorphism-extendability criteria yield effective practical tools for arithmetic and deformation theory. Furthermore, examining reductions in low characteristics exposes subtle phenomena (e.g., wild ramification, failure of extendability in characteristic $2$) and confirms the sharpness of such hypotheses [(Zhao, 26 Nov 2025), Section 5.2].

The interaction of automorphism groups, moduli, and reduction phenomena remains central in ongoing research on arithmetic surfaces and their moduli, with Enriques surfaces providing a crucial test case for the limits of cohomological and geometric reduction techniques.