🔢 mathematics

Brauer groups of smooth loci in linear systems and torsors over Jacobians of plane curves

This paper establishes that the Brauer groups of smooth loci in linear systems on simply connected smooth projective varieties (such as the projective plane, K3 surfaces, and cubic fourfolds) are at most $\mathbb{Z}/2\mathbb{Z}$ under suitable ampleness conditions, a result derived from studying the 2-nodal locus and applied to compute the Tate–Shafarevich group of torsors over relative Jacobians of universal smooth plane curves.

Original authors: Moritz Hartlieb, Weite Pi

Published 2026-06-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Moritz Hartlieb, Weite Pi

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 an architect designing a massive gallery of art. In this gallery, every single painting represents a different shape drawn on a canvas (mathematicians call these "plane curves"). Most of your paintings are perfect, smooth, and beautiful. But some have flaws: a tiny crack here, a jagged tear there.

The Discriminant Locus is the special section of your gallery where you display only the "flawed" paintings. The Smooth Locus is the rest of the gallery, filled only with the perfect, smooth ones.

This paper is a mathematical investigation into the "hidden structure" or "twistedness" of the space containing these perfect paintings. Specifically, the authors are asking: If you walk around the gallery of perfect curves, is there any hidden, non-trivial "twist" in the fabric of the space itself?

In mathematics, this hidden twist is called the Brauer Group. Think of the Brauer Group like a "mystery box" attached to your gallery.

If the group is zero, the gallery is perfectly flat and simple; there are no hidden twists.
If the group is Z/2Z, it means there is exactly one type of twist possible (like a Möbius strip).
If the group is Z/6Z, there are six different ways the space can twist.

Here is what the authors discovered, broken down into simple concepts:

1. The Main Rule of Thumb

The authors found a general rule for when these galleries are "twisted" or "untwisted." It depends on how "ample" (how big and powerful) the tools are used to create the curves.

The Finding: If the tools are "sufficiently strong" (mathematically, "sufficiently ample"), the hidden twist in the gallery of perfect curves is usually very small. It is either nothing at all or just a simple binary twist (like a light switch: on or off).
The Analogy: Imagine trying to draw a circle with a weak pencil (low power). You might get a wobbly, twisted line. But if you use a super-strong, precise laser cutter (high power), the resulting shape is so rigid that the space around it becomes perfectly smooth and untwisted.

2. The Case of Plane Curves (Drawing on a Flat Sheet)

The authors tested this rule on curves drawn on a standard flat sheet (the projective plane). The "power" of the drawing tool is determined by the degree of the curve (how complex the shape is).

Odd Degrees (3, 5, 7...): If you draw curves with an odd number of "loops" or complexity, the gallery is perfectly smooth. The Brauer group is zero. There are no hidden twists.
Even Degrees (4, 6, 8...): If you draw curves with an even number of loops, the gallery has a single, simple twist. The Brauer group is Z/2Z. It's like the gallery is a Möbius strip; you can walk around it, but you end up on the "other side."

3. The "Jacobian" and the "Torsor" (The Map and the Travelers)

The paper also looks at something called the Jacobian.

The Analogy: Imagine the "Jacobian" is a map of all the possible ways to travel around a specific curve.
The Torsor: A "torsor" is like a traveler who wants to go on a trip but doesn't have a starting point on the map. They are "twisted" relative to the map.
The Discovery: The authors calculated exactly how many different types of "lost travelers" (torsors) exist for these curves.
- For curves with degree 4 or higher, the number of lost travelers is exactly equal to the degree of the curve (e.g., for degree 4, there are 4 types; for degree 5, there are 5 types).
- For cubic curves (degree 3), the situation is special. Because of a specific geometric quirk (the "2-nodal locus" breaks into pieces), the number of lost travelers is 6.

4. Other Shapes: K3 Surfaces and Cubic Fourfolds

The authors didn't just stop at flat sheets. They applied their rule to more complex, 3D-like shapes called K3 surfaces and Cubic Fourfolds.

K3 Surfaces: These are like complex, doughnut-shaped surfaces but with special properties.
- If the surface is "large enough" (degree 20 or higher), the gallery of perfect curves on it is perfectly smooth (Brauer group = 0).
- However, if the surface is "small" (low degree), the gallery does have twists. The authors even found specific examples where the twist is non-zero.
Cubic Fourfolds: These are even more complex 4-dimensional shapes. The authors proved that for a "general" (typical) cubic fourfold, the gallery of perfect curves is perfectly smooth (Brauer group = 0).

5. How They Did It (The "2-Nodal" Detective Work)

How did they figure this out? They didn't just look at the perfect curves. They looked at the "flawed" ones, specifically the ones with two distinct cracks (called the "2-nodal locus").

The Metaphor: Imagine you want to know if a room is empty. Instead of looking at the empty room, you look at the doorway where people with two bags (two cracks) are standing. By counting how these "double-baggers" are arranged and how they overlap, the authors could deduce the hidden structure of the empty room.
They found that for even-degree curves, the way these "double-cracked" curves overlap creates a specific "shadow" that forces the space to have that single binary twist (Z/2Z). For odd degrees, the shadows cancel each other out, leaving the space perfectly smooth.

Summary

In short, this paper maps out the "hidden topology" of spaces containing perfect mathematical curves.

Simple Rule: Stronger tools (higher degrees) usually mean simpler spaces.
The Twist: Even-degree curves on a flat plane have a simple "Möbius strip" twist. Odd-degree curves do not.
The Travelers: The number of "lost travelers" (torsors) on these maps is directly tied to the complexity of the curve, with a special exception for cubic curves.

The paper provides a precise mathematical "inventory" of these hidden twists for several important types of geometric shapes, showing that while the spaces can be complex, their hidden structures are surprisingly small and predictable.

Technical Summary: Brauer Groups of Smooth Loci in Linear Systems and Torsors over Jacobians of Plane Curves

Problem Statement
The paper investigates the cohomological Brauer group, $Br(|L|_{sm}) := H^2_{\text{\'et}}(|L|_{sm}, \mathbb{G}_m)$ , of the smooth locus $|L|_{sm}$ within a linear system $|L|$ on a simply connected smooth projective variety $X$ . Here, $|L|_{sm}$ parametrizes smooth divisors, while its complement is the discriminant locus $\Delta_L$ of singular divisors. A primary motivation is to understand the structure of the Tate–Shafarevich group $\text{X}(\mathcal{J}) := H^1(B, \text{Pic}^0_{\mathcal{C}/B})$ , which classifies torsors over the relative Jacobian $\mathcal{J} \to B$ of the universal smooth curve $\mathcal{C} \to B$ (where $B = |L|_{sm}$ ). The authors aim to determine these groups for specific geometric settings: plane curves, primitively polarized K3 surfaces, and cubic fourfolds.

Methodology
The approach relies on a stratification of the discriminant locus $\Delta_L$ based on the number of singular points. The authors define $\Delta_k$ as the locus of hypersurfaces with exactly $k$ singular points (specifically ordinary nodes in the surface case) and analyze the universal singular locus $\Sigma \subset |L| \times X$ .

Ampleness Conditions: The authors introduce a set of "ampleness conditions" (denoted $\dagger$ ) on the line bundle $L$ . These conditions ensure that the discriminant locus is irreducible, that the locus of curves with two nodes ( $\Delta_2$ ) is an irreducible divisor, and that the universal 2-nodal locus $\Sigma_2$ is irreducible. These conditions are verified for specific cases using results on Severi varieties and jet ampleness (e.g., 5-very ampleness).
Homological Computation: Using the purity of the Brauer group and the Gysin sequence, the authors relate $Br(|L|_{sm})[n]$ to the Borel–Moore homology of the discriminant. Specifically, they reduce the computation to determining whether the cohomology class $[\Sigma_2]$ vanishes in $H^2(\Sigma, \mathbb{Z}/2\mathbb{Z})$ .
Relative Symmetric Products: To compute the Tate–Shafarevich group, the authors study the Brauer groups of relative symmetric products $C^{(n)}_B$ . They utilize the fact that for sufficiently large $n$ , the Abel–Jacobi map $C^{(n)}_B \to \mathcal{J}$ is a Brauer–Severi variety (a fibration in projective spaces). By establishing isomorphisms between the Brauer groups of $C^{(n)}_B$ for different $n$ (via closed immersions induced by adding points), they relate the period of the associated Brauer class to the geometry of the curves.
Exact Sequences: The final computation of $\text{X}(\mathcal{J})$ is derived by combining the computed Brauer groups with exact sequences arising from the Leray spectral sequence and the short exact sequence of étale sheaves $0 \to \text{Pic}^0_{\mathcal{J}/B} \to \text{Pic}_{\mathcal{J}/B} \to \text{NS}_{\mathcal{J}/B} \to 0$ .

Key Results

General Theorem on Brauer Groups: For a simply connected smooth projective variety $X$ and a "sufficiently ample" line bundle $L$ satisfying condition $\dagger$ , the Brauer group $Br(|L|_{sm})$ is at most $\mathbb{Z}/2\mathbb{Z}$ . Specifically, it vanishes if the class $[\Sigma_2]$ is non-zero in $H^2(\Sigma, \mathbb{Z}/2\mathbb{Z})$ , and is $\mathbb{Z}/2\mathbb{Z}$ otherwise.
Plane Curves:
- For the linear system of plane curves of degree $d \ge 3$ , $Br(|L|_{sm}) \cong 0$ if $d$ is odd, and $\cong \mathbb{Z}/2\mathbb{Z}$ if $d$ is even.
- For $d \ge 4$ , the Tate–Shafarevich group of the relative Jacobian is generated by the torsor $\text{Pic}^1_{\mathcal{C}/B}$ , yielding $\text{X}(\mathcal{J}) \cong \mathbb{Z}/d\mathbb{Z}$ .
- For $d=3$ (cubic curves), the Tate–Shafarevich group is $\text{X}(\mathcal{J}) \cong \mathbb{Z}/6\mathbb{Z}$ . This case requires separate treatment due to the reducibility of the 2-nodal locus.
K3 Surfaces:
- For a primitively polarized K3 surface $(S, L)$ of Picard rank one and degree $L^2 \ge 20$ , $Br(|L|_{sm}) = 0$ . The condition $L^2 \ge 20$ ensures $L$ is 5-very ample.
- The paper provides counterexamples for low-degree K3 surfaces: for degree 2, the Brauer group is large (at least 343-dimensional over $\mathbb{F}_p$ for any prime $p$ ); for degree 4, $Br(|L|_{sm}) \cong \mathbb{Z}/2\mathbb{Z}$ .
Cubic Fourfolds: For a general smooth cubic fourfold $X$ , the Brauer group of the smooth locus of the linear system $|\mathcal{O}_X(1)|$ vanishes, i.e., $Br(|\mathcal{O}_X(1)|_{sm}) = 0$ .

Significance and Claims
The authors frame their work as providing "surprisingly effective control" on the Brauer group of the base of linear systems, a technical hurdle often present in the study of K3 surfaces and hyperkähler varieties. The primary contribution is the explicit computation of the Tate–Shafarevich group for the relative Jacobian of universal smooth plane curves, identifying its generators and structure.

The paper notes that these results apply to the smooth loci of Lagrangian fibrations associated with compact hyperkähler varieties of $K3^{[n]}$ and $OG10$ types (specifically the Laza–Saccà–Voisin fibration). The vanishing of the Brauer group in these high-degree or general cases suggests that the smooth loci of general Lagrangian fibrations may often have trivial Brauer groups, a property that facilitates the study of twisted sheaves and moduli spaces. The authors explicitly state that they do not treat the geometric realization of the non-trivial Brauer–Severi varieties for even degrees or the specific 2-torsion twists for cubic curves, leaving these as open problems.