🔢 mathematics

Symmetric Differentials on K3 Surfaces

This paper establishes that a K3 surface over an algebraically closed field admits a nonzero global symmetric differential of positive degree if and only if it is a supersingular surface with Artin invariant $\sigma_0=1$ in characteristic $p=2$ , where such differentials exist uniquely in every positive even degree, while also extending a theorem by Jang regarding the isomorphism of supersingular K3 surfaces to smooth quartics.

Original authors: Frank Gounelas, Christian Liedtke

Published 2026-08-19

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Frank Gounelas, Christian Liedtke

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

In the vast landscape of mathematics, where shapes are studied not just by their size or form but by the invisible flows that move across them, there exists a special class of surfaces known as K3 surfaces. These are smooth, closed shapes that behave in a very particular way: they have no holes that can be filled, and their internal geometry is perfectly balanced, much like a calm lake that reflects the sky without distortion. For decades, mathematicians have tried to understand what kinds of "flows" or patterns can exist on these surfaces. One way to measure these patterns is by looking for symmetric differentials, which are mathematical objects that describe how directions on the surface can be multiplied and combined to create new, stable directions. In the world of complex numbers, which governs most of classical geometry, it was already known that these surfaces are too rigid to support any such patterns. They are essentially empty of these specific flows.

However, mathematics also explores what happens when we change the fundamental rules of the game, specifically by switching from the familiar world of real and complex numbers to a world built on finite fields, where arithmetic wraps around after a certain number. This is known as working in positive characteristic. In this strange, discrete arithmetic universe, the usual laws of geometry can behave differently, and surfaces that are rigid in one setting might become flexible in another. The question that remained unanswered was whether these K3 surfaces, when placed in this finite arithmetic world, could suddenly develop these symmetric patterns. If they could, it would mean that the geometry of these surfaces is far more sensitive to the underlying number system than previously thought.

A team of researchers has now settled this question with a definitive answer. They proved that for a K3 surface to possess even a single non-zero symmetric pattern, two very strict conditions must be met simultaneously. First, the arithmetic of the world must be built on the number two. Second, the surface itself must belong to a very rare and specific category known as "supersingular," and within that category, it must have a specific numerical fingerprint called an Artin invariant of one. If the surface is in any other arithmetic world, or if it is a different type of supersingular surface in the world of two, it remains completely empty of these patterns, just as it is in the complex world.

The researchers did not just state this result; they mapped out exactly why it happens. They began by showing that if a K3 surface were to have such a pattern, it would have to be "uniruled," a property that implies the surface is covered by a family of straight lines, making it much less rigid than a typical K3 surface. This led them to a contradiction in most cases, because standard K3 surfaces are not covered by lines. They then had to deal with the tricky case where the arithmetic rules allow for patterns that are not separable, meaning they are deeply intertwined with the number two in a way that prevents standard separation techniques. By carefully peeling back layers of these patterns using a process of extracting roots, they showed that any such pattern would eventually reveal a hidden structure that simply cannot exist on a K3 surface, unless the surface is one of those rare, special cases.

For the exceptional cases where the surface is indeed supersingular in the world of two, the researchers had to look at the surface through a different lens. They found that when the surface has the highest level of complexity (an Artin invariant of one), it can be described as a smooth intersection of two specific shapes in a higher-dimensional space. In this specific configuration, the surface splits into two parts that are mirror images of each other. This splitting allows for exactly one non-zero pattern to exist for every even degree, while all odd degrees remain empty. It is a unique exception to the rule of emptiness. Conversely, they showed that if the surface is a different type of supersingular K3 surface in the world of two, or if it is a smooth quartic surface (a shape defined by a fourth-degree equation) in any other arithmetic world, it remains stubbornly empty of these patterns.

This work also clarifies a long-standing question about the geometry of these surfaces. The researchers demonstrated that the existence of these patterns is directly linked to a property called "pseudoeffectivity," which roughly measures whether the surface's internal directions can be used to build larger, stable structures. They proved that this property holds true if and only if the surface is one of those rare, special cases in the world of two. This finding connects the abstract existence of patterns to the broader geometric structure of the surface, showing that the ability to support these flows is a rare and delicate feature.

The study also sheds light on related shapes called Enriques surfaces, which are closely connected to K3 surfaces. The researchers found that the rules for these surfaces are slightly different. In most cases, they also lack these patterns, but in the specific arithmetic world of two, there are certain types of Enriques surfaces that do carry them. This distinction highlights how sensitive these geometric objects are to the precise nature of the number system they inhabit.

Ultimately, this paper provides a complete census of where these symmetric patterns can exist on K3 surfaces. It confirms that for the vast majority of these surfaces, across almost all mathematical universes, the answer is a firm no. The only time the answer is yes is in a very narrow, specific corner of the mathematical landscape: the world of arithmetic based on the number two, and only for the most singular and complex version of the K3 surface. This result closes a chapter on the classification of these surfaces, showing that their rigidity is the norm, and their flexibility is a rare, highly constrained exception.

Technical Summary: Symmetric Differentials on K3 Surfaces

Problem Statement
The paper addresses the classification of global symmetric differentials on K3 surfaces over an algebraically closed field $k$ of arbitrary characteristic $p \ge 0$ . Specifically, it seeks to determine the dimension of the vector spaces $H^0(X, \text{Sym}^n \Omega^1_X)$ for $n \ge 1$ . While the vanishing of these spaces is a known theorem for K3 surfaces in characteristic zero (due to Kobayashi, derived from transcendental methods, and recently proven algebraically by Ciliberto–Gounelas), the behavior in positive characteristic $p > 0$ remained open. The authors aim to compute these spaces for all characteristics and relate the existence of such differentials to the birational geometry of the projectivised cotangent bundle $P(\Omega^1_X)$ .

Methodology
The proof is divided into two primary phases: establishing the necessity of uniruledness for the existence of symmetric differentials, and then analyzing the specific cases of supersingular K3 surfaces using explicit projective models and lattice theory.

From Symmetric Differentials to Uniruledness:
- Spectral Divisors: A non-zero section $s \in H^0(X, \text{Sym}^n \Omega^1_X)$ defines a spectral divisor $\Sigma_s$ in the projectivised cotangent bundle $P(\Omega^1_X)$ .
- Semistability and Reduction: Assuming $X$ is not uniruled, the authors utilize the slope-semistability of Frobenius pullbacks of $\Omega^1_X$ (a result of Langer) to show that a minimal spectral divisor must be reduced, irreducible, and horizontal.
- Separability and Root Extraction: In characteristic $p > 0$ , the generic binary form defining the spectral divisor may be inseparable. The authors construct a global $p$ -th root extraction process (using Frobenius monomial homomorphisms) to factor the degree $n = p^e \cdot m$ and produce a new section $t$ with a separable, irreducible generic binary form.
- Étale Covers: If $m \ge 2$ , the separable spectral divisor yields a connected, finite, étale cover of $X$ . Since K3 surfaces have a trivial étale fundamental group, such a cover cannot exist, leading to a contradiction unless $m=1$ .
- Frobenius-Twisted Vector Fields: If $m=1$ , the section $t$ corresponds to a non-zero global section of a Frobenius pullback $(F^e_X)^* \Omega^1_X$ . The authors rule this out using Cartier's canonical connection and a descent argument, showing that such sections cannot exist on non-uniruled K3 surfaces.
- Conclusion: Thus, any K3 surface admitting a non-zero symmetric differential must be uniruled, and consequently, supersingular (Shioda-supersingular).
Analysis of Supersingular K3 Surfaces:
- Quartic Models: The authors extend a theorem of Jang to show that a supersingular K3 surface is isomorphic to a smooth quartic surface in $\mathbb{P}^3$ if and only if $p \neq 2$ or ( $p=2$ and the Artin invariant $\sigma_0 \ge 3$ ).
- Vanishing for Quartics: They prove that no smooth hypersurface of dimension $\ge 2$ in projective space admits non-zero symmetric differentials in any characteristic. This establishes vanishing for all supersingular K3 surfaces except those with $p=2$ and $\sigma_0 \le 2$ .
- Characteristic Two, $\sigma_0 = 2$ : For this case, the authors utilize a quasi-elliptic fibration with specific reducible fibers (one $I^*_0$ and sixteen type III). They construct a short exact sequence $0 \to L \to \Omega^1_X \to L^{-1} \to 0$ where $L^2 = -24$ . By analyzing the extension class under iterated Frobenius pullbacks on a rational curve within the fibration, they prove the extension is non-split and that no symmetric powers of $\Omega^1_X$ have global sections.
- Characteristic Two, $\sigma_0 = 1$ : This unique surface is realized as a smooth complete intersection of bidegrees $(1,2)$ and $(2,1)$ in $\mathbb{P}^2 \times \mathbb{P}^2$ (Dolgachev–Kondô). Here, the cotangent bundle splits as $\Omega^1_X \cong L \oplus L^{-1}$ . The authors compute the symmetric powers explicitly, finding that $H^0(X, \text{Sym}^n \Omega^1_X)$ is one-dimensional for even $n$ and zero for odd $n$ .

Key Contributions and Results

Theorem 1.1: A K3 surface $X$ over an algebraically closed field admits a non-zero global symmetric differential of positive degree if and only if the characteristic is $p=2$ and $X$ is supersingular with Artin invariant $\sigma_0 = 1$ . In this exceptional case, $\dim_k H^0(X, \text{Sym}^n \Omega^1_X) = 1$ for even $n$ and $0$ for odd $n$ .
Theorem 1.3: A strengthening of Jang's theorem characterizing when a supersingular K3 surface is isomorphic to a smooth quartic in $\mathbb{P}^3$ : this occurs if and only if $p \ge 3$ or ( $p=2$ and $\sigma_0 \ge 3$ ).
Pseudoeffectivity: The paper establishes that the tautological class of $P(\Omega^1_X)$ is pseudoeffective if and only if $p=2$ and $\sigma_0 = 1$ .
Enriques Surfaces: The results imply that for Enriques surfaces, symmetric differentials vanish unless $p=2$ and the surface is "exceptional" (non-ordinary), in which case they are non-zero.

Significance and Claims
The authors claim that their work provides a complete classification of symmetric differentials on K3 surfaces across all characteristics, resolving the positive characteristic case which was previously unknown. The paper offers a new algebraic proof for the characteristic zero case (vanishing) that avoids transcendental methods.

The significance lies in the sensitivity of symmetric differentials as an invariant: they distinguish between different types of surfaces of Kodaira dimension zero (separating abelian/bielliptic surfaces from K3/Enriques surfaces) and, within the K3 family, distinguish between the generic supersingular case and the unique $\sigma_0=1$ case in characteristic two. The authors note that their findings suggest that Church's construction of non-unirational supersingular surfaces using symmetric differentials cannot be applied to K3 surfaces, as the only K3 surfaces with such differentials are unirational (being $\sigma_0=1$ in char 2).

The methodology is described as "by hand" for the exceptional cases, relying on explicit projective models, lattice-theoretic criteria, and the interplay between quasi-elliptic fibrations and Frobenius pullbacks, rather than general moduli theory. The paper also highlights the utility of the universal discriminant of binary forms (defined over $\mathbb{Z}$ ) in controlling the étaleness of spectral covers in arbitrary characteristic.

More like this