🔢 mathematics

On spectral sequences for semiabelian varieties over non-closed fields

This paper provides a new, concise proof for the first potentially non-zero differential of the Hochschild–Serre spectral sequence for semiabelian varieties over non-closed fields, establishing specific non-vanishing conditions for Jacobians, proving degeneration when the Albanese torsor is trivial, and deriving formulas for tori and smooth projective curves.

Original authors: Alexander Petrov, Alexei Skorobogatov

Published 2026-06-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Alexander Petrov, Alexei Skorobogatov

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 trying to understand a complex machine, like a high-end camera or a sophisticated clock. To understand how it works, you might try to take it apart and look at each gear, spring, and lens individually. In mathematics, specifically in a field called algebraic geometry, researchers study shapes called varieties (which can be thought of as multi-dimensional curves and surfaces) over different types of number systems (fields).

This paper by Alexander Petrov and Alexei Skorobogatov is about a specific tool they use to understand these shapes: a mathematical "spectral sequence." You can think of a spectral sequence as a multi-layered map or a step-by-step translation guide. It takes information about a shape defined over a simple, "closed" world (where all numbers exist) and tries to translate it back to a "non-closed" world (like the rational numbers, where some roots don't exist).

Usually, this translation process is straightforward. It's like taking a photo, printing it, and then scanning it back; the image stays the same. However, the authors discovered that sometimes, this process distorts the image. There are "differentials" (mathematical operations) in this map that don't just pass information along; they actively change it, creating a mismatch between the original shape and its translated version.

Here is a breakdown of their main discoveries using everyday analogies:

1. The "Glitch" in the Translation (The Main Discovery)

The authors proved that for certain shapes called semiabelian varieties (which are like hybrid machines combining wheels and engines), this translation map isn't always perfect.

The Analogy: Imagine you have a set of instructions for building a house. Usually, if you follow the steps, you get the house. But for these specific shapes, there is a "glitch" in the instructions. A step that should just move a brick from one pile to another actually changes the brick into something else.
The Result: They found a specific example of a shape (an abelian surface) over the rational numbers where this glitch is real and non-zero. This means the shape has a hidden "twist" that only appears when you look at it through this specific mathematical lens.

2. The "Theta-Characteristic" Puzzle

Why does this glitch happen? The authors linked it to a specific property of curves (shapes like loops or figure-eights).

The Analogy: Think of a curve as a necklace with beads. Some beads are special "theta-characteristics." The authors found that if the arrangement of these special beads cannot be evenly divided by 2 (a mathematical concept called "divisibility"), the translation map breaks.
The Proof: They showed that if you have a curve where the "theta-beads" are stuck in a position that can't be split in half, the mathematical map will definitely produce a non-zero error (a non-zero differential). They provided concrete examples of such curves (genus 2 curves) to prove this happens in the real world of numbers.

3. When the Map Does Work (The "Good" News)

Not all shapes suffer from this glitch.

The Analogy: If the shape has a "rational point" (a specific spot that exists in the number system you are using, like a handle on a door that you can actually grab), the translation works perfectly.
The Result: They proved that if a curve has a "degree 1" rational divisor (a fancy way of saying it has a specific, accessible point), the spectral sequence collapses into a perfect, non-distorted sum. The "glitch" disappears, and the map is a faithful translation.

4. The "Torus" and the "Brauer Group"

The paper also looks at shapes called tori (which are like donuts or multi-dimensional rings).

The Analogy: Think of a torus as a hollow tube. The authors figured out how to calculate a specific "safety rating" for these tubes, called the Brauer group. This group measures how "twisted" the tube is in a way that prevents you from filling it with certain fluids.
The Result: They gave a new, shorter formula to calculate this twisting. They showed that for most "nice" tori (called quasi-trivial tori), the safety rating is actually zero (no hidden twists), meaning the translation map is perfect for them too.

5. The "Torsor" (The Missing Handle)

Finally, they looked at torsors.

The Analogy: Imagine a shape that is identical to a machine but is missing its "handle" (a rational point). It's a machine that could be turned on, but you can't find the switch.
The Result: They proved that the "glitch" in the translation map is directly caused by the fact that the handle is missing. If the handle (the rational point) exists, the map is smooth. If the handle is missing, the map creates a specific error term. This error term is a mathematical fingerprint of the missing handle.

Summary

In simple terms, this paper is about finding the exceptions to the rule.

The Rule: Usually, you can translate complex geometric shapes from a "perfect" world to a "real" world without losing information.
The Exception: The authors found that for certain shapes (specifically those without a "handle" or with specific "theta-bead" arrangements), the translation process introduces a specific, measurable error.
The Contribution: They didn't just say "it happens"; they gave a precise formula for how it happens and provided concrete examples of shapes where this error is guaranteed to exist. They also clarified when the error doesn't happen, giving mathematicians a clear guide on which shapes are "safe" to translate and which ones require extra care.

Technical Summary: On Spectral Sequences for Semiabelian Varieties over Non-Closed Fields

Problem Statement
The paper investigates the structure of the étale cohomology complex $R\Gamma_{\text{\'et}}(X_{k^s}, \mathbb{Z}_\ell)$ for algebraic varieties $X$ over a non-closed field $k$ . While Deligne's decomposition theorem ensures that for smooth projective varieties, the rational étale cohomology complex is quasi-isomorphic to the direct sum of its individual cohomology groups, this property fails for integral coefficients. This failure is manifested in the Hochschild–Serre spectral sequence:
$E_2^{p,q} = H^p(k, H^q_{\text{\'et}}(X_{k^s}, \mathbb{Z}_\ell)) \Rightarrow H^{p+q}_{\text{\'et}}(X, \mathbb{Z}_\ell)$
which, in general, possesses non-zero differentials. The authors aim to provide a precise description of the first potentially non-zero differentials for semiabelian varieties (extensions of abelian varieties by tori) and to determine conditions under which these differentials vanish or are non-zero.

Methodology
The authors employ a combination of derived category techniques, group cohomology, and arithmetic geometry:

Derived Category Analysis: They analyze the truncation $\tau^{[i, i+1]}R\Gamma_{\text{\'et}}(X_{k^s}, \mathbb{Z}/n)$ to identify the extension classes corresponding to the differentials in the spectral sequence.
Reduction to Group Cohomology: Utilizing the fact that semiabelian varieties are étale $K(\pi, 1)$ -spaces, they identify the étale cohomology complex with the continuous group cohomology complex of the Tate module (or character module for tori).
Explicit Construction of Extensions: They construct explicit 2-extensions involving quadratic functions and tensor products to represent the extension classes in $\text{Ext}^2_\Gamma$ . Specifically, they relate the differential $\delta_2^{p,2}$ to the extension defined by the sequence of quadratic functions $0 \to M \to Q(M) \to M \otimes 2 \to \wedge^2 M \to 0$ .
Arithmetic Pairings: For abelian varieties, particularly Jacobians, the authors utilize the Cassels–Tate pairing and the Poitou–Tate duality to test the divisibility of specific cohomology classes (related to the torsor of theta-characteristics) by 2.
Brauer–Manin Obstruction: They interpret certain cohomological obstructions via the Brauer–Manin pairing on locally soluble varieties to construct explicit counterexamples.

Key Contributions and Results

General Description of Differentials for Semiabelian Varieties (Theorem 1.6, 1.13):
The authors provide a new, short proof describing the differential $\delta_2^{p,2}: H^p(k, \wedge^2 M) \to H^{p+2}(k, M)$ (where $M = H^1_{\text{\'et}}(A_{k^s}, \mathbb{Z}_2)$ ) for a semiabelian variety $A$ . They show this differential corresponds to the connecting map of a specific 2-extension involving quadratic functions. This description is generalized to coefficients $\mathbb{Z}/2^m$ via Bockstein homomorphisms.
Non-Degeneration for Abelian Surfaces (Theorem 1, Theorem 3.5):
Contrary to the behavior of Jacobians with rational divisor classes, the authors prove that the Hochschild–Serre spectral sequence does not necessarily degenerate for all abelian varieties. They construct an explicit principally polarized abelian surface $A$ over $\mathbb{Q}$ where the differential $\delta_2^{0,2}$ is non-zero. This occurs when the class of the torsor of theta-characteristics (or the Albanese torsor) is not divisible by 2 in the Galois cohomology group $H^1(k, J)$ . Two explicit examples of genus 2 hyperelliptic curves are provided where this condition holds, utilizing the Brauer–Manin obstruction to verify non-divisibility.
Degeneration for Jacobians with Rational Divisors (Theorem 3, Theorem 2.1):
The paper establishes that if a curve $C$ admits a $k$ -rational divisor class of degree 1, the étale cohomology complex of its Jacobian $J$ decomposes in the derived category $D(\Gamma, \mathbb{Z}/n)$ . Consequently, the Hochschild–Serre spectral sequence for such Jacobians degenerates at the second page. This result generalizes known facts but is presented with a proof using integral Künneth projectors defined over $k$ .
Torsors and the Albanese Torsor (Theorem 4, Theorem 5.1):
For a torsor $X$ of a semiabelian variety $A$ , the authors describe the obstruction to splitting the complex $\tau^{[0,1]}R\Gamma_{\text{\'et}}(X_{k^s}, \mathbb{Z}/n)$ . They prove that the class corresponding to this obstruction in $H^2(k, A[n])$ is exactly the image of the class of the torsor $[X] \in H^1(k, A)$ under the Bockstein homomorphism. This links the non-splitting of the cohomology complex directly to the non-triviality of the torsor.
The Brauer Group of Tori (Theorem 4.1):
The authors describe the differential $d_3^{0,2}$ in the Hochschild–Serre spectral sequence for a torus $T$ with coefficients in $\mathbb{G}_m$ . This differential computes the obstruction to the surjectivity of the map $Br(T) \to Br(T_{k^s})^\Gamma$ . They provide an explicit formula for this differential on the 2-primary torsion subgroup using Bockstein maps and the extension class of the character module. As a corollary, they confirm that for quasi-trivial tori or tori over local/global fields, this map is surjective.
Differentials for Curves (Proposition 6.1, 6.5, Corollary 6.6):
For smooth projective curves, the paper generalizes Suslin's lemma. It characterizes the differentials on the second page in terms of the Albanese torsor. If the Albanese torsor is divisible by $n$ , the second-page differentials vanish, and the authors describe the non-zero differentials on the third page. These third-page differentials are shown to be induced by a specific class in $H^3(k, \mu_n)$ derived from the Picard stack of the curve.

Significance and Claims
The paper claims to offer a "new, short proof" of the formula for the first non-zero differential of the Hochschild–Serre spectral sequence for semiabelian varieties, unifying the treatment of abelian varieties and tori. Its primary significance lies in:

Explicit Counterexamples: Providing the first known examples of abelian varieties (specifically surfaces) over number fields where the spectral sequence fails to degenerate, challenging the intuition that such degenerations might hold for all abelian varieties.
Arithmetic Interpretation: Connecting the non-vanishing of differentials to deep arithmetic properties, specifically the 2-divisibility of the class of the torsor of theta-characteristics and the Brauer–Manin obstruction.
Computational Tools: Offering explicit formulas for computing the Brauer group of tori and the differentials for curves, answering specific questions raised in prior literature (e.g., regarding the Brauer group of tori in [CTS21]).
Generalization: Extending results on the decomposition of cohomology complexes to direct factors of products of Jacobians with rational divisor classes, a result the authors note may be known to experts but was previously unavailable in the literature.

The authors remain modest regarding the non-vanishing of differentials for tori, noting that while they provide a formula, it is currently unknown whether non-zero examples exist for tori (Remark 4.2). Similarly, the construction of the abelian surface relies on existing arithmetic constructions by Creutz and Viray, which the authors synthesize to demonstrate the non-degeneration phenomenon.