🔢 mathematics

Dichotomy for the p-primary Brauer-Manin obstruction in characteristic p

This paper establishes a dichotomy for the p-primary Brauer-Manin obstruction on smooth, projective, geometrically integral varieties over global function fields of characteristic p, demonstrating that the set of potentially relevant places is either finite or almost all places depending on whether the unipotent Brauer group vanishes, with the set of exceptional places being empty if the variety is defined over a finite field.

Original authors: Christopher D. Lazda, Alexei N. Skorobogatov

Published 2026-06-16

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Christopher D. Lazda, Alexei N. 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 solve a massive, cosmic puzzle. The pieces of this puzzle are points on a geometric shape (let's call it a "variety") that exists over a specific type of number system called a "global function field." Think of this number system like a vast, interconnected web of roads, where every intersection is a "place" (or a location).

The goal of the puzzle is to find a solution that works everywhere on the web simultaneously. However, there are "obstacles" or "roadblocks" that might prevent a solution from existing. In mathematics, these obstacles are called the Brauer–Manin obstruction.

This paper, written by Christopher Lazda and Alexei Skorobogatov, investigates a specific type of roadblock that only appears in a world with "positive characteristic" (a mathematical setting where numbers wrap around like a clock, but with a twist). They discover a sharp dichotomy (a split into two completely different outcomes) based on a hidden feature of the geometric shape.

Here is the breakdown of their discovery using simple analogies:

The Two Worlds: The "Silent" vs. The "Noisy"

The authors look at a hidden property of the shape called the Unipotent Brauer Group (let's call it the "Noise Level"). They find that the behavior of the roadblocks depends entirely on whether this Noise Level is zero or not.

Case 1: The Silent World (Noise Level = 0)

If the shape has no noise (the unipotent Brauer group is zero) and its "Picard scheme" (think of this as the shape's internal structural blueprint) is free of a specific type of mathematical glitch (p-torsion free), then the situation is very calm.

The Finding: Only a tiny, finite number of road intersections (places) on the web are actually relevant to the roadblocks.
The Analogy: Imagine you are trying to drive across a continent. In this "Silent World," you only need to worry about traffic jams at 5 specific cities. Everywhere else, the roads are clear. If you avoid those 5 cities, the rest of the journey is smooth.
Special Case: If the shape was originally built on a simple, finite field (like a shape that was "born" in a small, closed village and then moved to the big web), then there are zero relevant cities. The roadblocks simply don't exist. The puzzle is solvable everywhere.

Case 2: The Noisy World (Noise Level > 0)

If the shape has noise (the unipotent Brauer group is not zero), the situation changes dramatically.

The Finding: Almost every intersection on the web is a potential roadblock.
The Analogy: In this "Noisy World," it's as if the entire continent is under construction. Almost every single city you try to drive through has a roadblock. You can't just avoid a few; the obstruction is everywhere.
Special Case: If the shape was "born" in a small village and moved to the web, then every single city is a roadblock. The obstruction is total.

Why Does This Happen? (The Mechanics)

The authors explain this using two main tools:

The "Spreading Out" Map: They imagine the shape as a flower that grows along a curve (a line of numbers). They show that if the "noise" is zero, the flower's roots (the mathematical structures causing the roadblocks) are small and finite. They only grow in a few specific spots.
The "Filter" Test: If the "noise" is present, the roots are wild and uncontrolled. They spread out so much that they touch almost every part of the web.

The "Constant" Exception

The paper highlights a special scenario where the shape is "constant" (it doesn't change its nature as you move across the web).

If the shape is "quiet" (No Noise), the roadblocks vanish completely. The puzzle is solved.
If the shape is "loud" (Has Noise), the roadblocks are everywhere. The puzzle is impossible to solve.

Summary

The paper draws a clear line in the sand:

No Noise: The obstacles are rare and manageable (or non-existent).
Noise: The obstacles are everywhere and unavoidable.

This helps mathematicians understand when they can expect to find solutions to these complex geometric puzzles and when they should expect the "Brauer–Manin obstruction" to block them at almost every turn. It turns a chaotic problem into a predictable one based on a single, measurable property of the shape.

Technical Summary: Dichotomy for the p-primary Brauer–Manin Obstruction in Characteristic p

Problem Statement
The paper investigates the behavior of the Brauer–Manin obstruction for smooth, projective, geometrically integral varieties $X$ over a global function field $k$ of characteristic $p > 0$ . Specifically, it addresses the question of which places $v$ of $k$ are "potentially relevant" to the $p$ -primary Brauer–Manin obstruction. A place $v$ is potentially relevant if there exists a finite extension $L/k$ , a place $w$ of $L$ over $v$ , and an element $A \in \text{Br}(X_L)$ such that the evaluation map $\text{ev}_A: X(L_w) \to \text{Br}(L_w)$ is non-constant.

This inquiry is motivated by recent results in the number field case (Bright and Newton; Ambrosi, Newton, and Pagano), which show that if $H^0(X, \Omega^2) \neq 0$ , then almost every place is potentially relevant. The authors seek to determine if an analogous dichotomy exists in positive characteristic, particularly regarding the role of the unipotent part of the Brauer group.

Mathematical Framework and Methodology
The analysis relies on the structure of the $p$ -primary torsion subgroup of the geometric Brauer group, $\text{Br}(X_{\bar{k}})[p^\infty]$ . Following Milne and Serre, the quotient of this group by its divisible subgroup is identified with the $\bar{k}$ -points of a quasi-algebraic group $G_X$ . This group $G_X$ is an extension of a finite étale group $D_X$ by a smooth connected unipotent group $U_X$ . The authors define $U_X$ as the unipotent Brauer group of $X$ .

The methodology combines two primary technical tools:

Representability of Flat Cohomology: The paper utilizes the generic representability results of Bragg and Olsson [BO] for flat cohomology sheaves $R^q\pi_*\mu_{p^n}$ associated with a spreading out $\pi: \mathcal{X} \to C$ of the variety $X$ .
Evaluation Filtration: For the case where $U_X \neq 0$ , the authors employ recent results by Krishna and Majumder [KM] concerning Kato's filtration on the local Brauer group in equicharacteristic $p$ . This filtration allows for the characterization of elements in the Brauer group that yield non-constant evaluation maps.

The proofs involve constructing smooth models $\pi: \mathcal{X} \to C$ over a curve $C$ with function field $k$ and analyzing the behavior of the sheaves $R^2\pi_*\mu_{p^n}$ and $R^3\pi_*\mu_{p^n}$ over $C$ . The dimension of $U_X$ is related to the slopes of the crystalline cohomology groups of $X_{\bar{k}}$ .

Key Contributions and Results
The central result is a dichotomy based on the vanishing of the unipotent Brauer group $U_X$ :

Theorem A (i): If $U_X = 0$ and the Picard scheme of $X$ is $p$ -torsion free, then only finitely many places of $k$ are potentially relevant to the $p$ -primary Brauer–Manin obstruction. In this case, the set of potentially relevant places is contained within the set of places where a specific smooth spreading out $\pi: \mathcal{X} \to C$ fails to have the property that $R^2\pi_*\mu_p$ is a finite flat group scheme and $R^3\pi_*\mu_p$ is an affine group scheme.
Theorem A (ii): If $U_X \neq 0$ , then almost all places of $k$ are potentially relevant to the $p$ -primary Brauer–Manin obstruction. The proof demonstrates that for any place $v$ in the support of a suitable model, one can construct a Brauer class whose evaluation map is non-constant by exploiting the non-triviality of the unipotent group scheme $R^2\pi_*\mu_p$ .

Corollaries and Special Cases:

Corollary B: Combining the $p$ -primary result with known $\ell$ -primary results ( $\ell \neq p$ ), if $U_X = 0$ and the Picard scheme is torsion-free, only finitely many places are relevant to the full Brauer–Manin obstruction.
Constant Varieties (Theorem C): If $X$ $X$ is defined over a finite field $\mathbb{F}$ $F$ (i.e., $X$ $X$ is a constant variety over $k$ $k$ ):
- If $U_X = 0$ and the Picard scheme is torsion-free, the Brauer–Manin pairing is identically zero.
- If $U_X \neq 0$ , every place of $k$ is potentially relevant.
Geometrically $p$ -torsion free Brauer Group (Proposition 4.1): If $\text{Br}(X_{\bar{k}})[p^\infty] = 0$ (which implies $U_X=0$ ), any place where $X$ has good reduction is not potentially relevant. The paper provides examples of non-isotrivial K3 surfaces satisfying these conditions, showing this case is distinct from the constant variety case.

Significance and Claims
The paper claims to establish a fundamental dichotomy in the behavior of the Brauer–Manin obstruction in positive characteristic, driven by the presence or absence of the unipotent Brauer group.

When $U_X = 0$ , the obstruction is "small" in the sense that only a finite set of places can obstruct the Hasse principle via the $p$ -primary part.
When $U_X \neq 0$ , the obstruction is "large," with almost all places being potentially relevant.

The authors note that while the existence of the curve $C$ required for the proofs is guaranteed by Bragg–Olsson, the precise determination of the exceptional set of places is generally difficult to make explicit, except in specific cases like constant varieties or those with good reduction. The work extends the understanding of the Brauer–Manin obstruction from number fields to global function fields, highlighting the distinct role played by unipotent phenomena in characteristic $p$ .