🔢 mathematics

Picard bundles and the degree of irrationality of Jacobians

This paper investigates the positivity properties of twisted rank- $g$ Picard bundles on the $g$ -fold symmetric product of a smooth projective curve to establish an upper bound of $2^g$ for the degree of irrationality of any genus $g$ Jacobian.

Original authors: Federico Moretti, Andrés Rojas

Published 2026-05-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Federico Moretti, Andrés Rojas

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 have a complex, multi-dimensional shape called a Jacobian. In the world of mathematics, this shape is built from a simpler, one-dimensional curve (like a loop or a twisted string). Mathematicians have long been trying to understand how "complicated" this Jacobian shape is.

One way to measure this complexity is called the degree of irrationality. Think of it as asking: "What is the simplest way to project this complicated shape onto a flat screen (like a standard 2D or 3D space) without losing too much information?" If you can project it easily with a low degree of distortion, the shape is "less irrational" (simpler). If you need a very complex, high-distortion projection, it's "more irrational."

For a long time, mathematicians knew a rough upper limit for this complexity, but they suspected the true limit was lower. This paper by Federico Moretti and Andrés Rojas proves a new, tighter limit: The complexity of a Jacobian built from a curve with $g$ holes is never more than $2g$ .

Here is how they did it, using some creative mental images:

1. The "Magic Map" and the "Bundle of Strings"

The authors start with a tool called a Picard bundle. Imagine the Jacobian as a vast landscape. On top of this landscape, they place a giant, flexible "blanket" made of many strings (a vector bundle).

The Strings: Each point on the blanket represents a specific piece of data about the original curve.
The Twist: The authors take this blanket and give it a specific "twist" (mathematically, they tensor it with a divisor). This twist is crucial because it makes the blanket globally generated.

What does "globally generated" mean?
Imagine you have a giant, flexible sheet. If it's "globally generated," it means that no matter where you stand on the sheet, you can always find a set of ropes (sections) tied to it that can pull the sheet taut in every direction. You never get stuck in a "dead zone" where the sheet goes slack. This property is the key that unlocks the rest of the proof.

2. The Symmetric Product: The "Group Photo"

To study this blanket, the authors move from the Jacobian to a place called the symmetric product ( $C^{(g)}$ ).

The Analogy: If your curve is a single person, the Jacobian is a way to describe all possible groups of people. The symmetric product is like a "group photo" station. If you have a curve with $g$ holes, this station takes $g$ points from the curve and arranges them into a single group.
The authors show that their twisted blanket (the bundle) looks very nice and orderly when viewed from this "group photo" station.

3. The "Top Chern Class": Counting the Knots

The authors calculate a specific number associated with their blanket, called the top Chern class.

The Analogy: Imagine the blanket is a piece of fabric. If you try to tie a knot in it using a specific number of strings, the number of ways you can tie that knot without the strings crossing in a messy way is the Chern class.
In this paper, they calculate that for a curve with $g$ holes, this number is exactly $2g$ .

4. The Final Projection: The "Shadow"

Here is the main trick. The authors use a general mathematical principle: If you have a "globally generated" blanket with a specific knot-count ( $2g$ ), you can use it to cast a shadow of the shape onto a flat screen.

The Process: They pick a random set of ropes from their blanket and use them to project the Jacobian onto a standard space ( $P^g$ ).
The Result: Because of the properties of the blanket they built, this projection is guaranteed to work (it covers the whole screen) and the "degree" of the projection (how many times the original shape wraps around the screen) is bounded by that knot-count number: $2g$ .

Why is this a big deal?

Before this paper, the best known guess for the complexity limit was much higher (roughly $g$ times the complexity of the curve itself). The authors proved that the limit is actually much lower—just twice the number of holes in the curve.

In summary:
The authors built a special mathematical "blanket" over a complex shape. They proved this blanket is strong enough to be pulled tight everywhere. By counting the "knots" in this blanket, they showed that the shape can be flattened onto a screen with a maximum distortion of $2g$ . This settles a long-standing question about how "irrational" or complex these specific mathematical shapes can be.

They also noted that this projection isn't just one rigid method; it's like a family of different camera angles (parametrized by a Grassmannian), all of which produce a view of the shape with this same low complexity limit.

Technical Summary: Picard Bundles and the Degree of Irrationality of Jacobians

Problem Statement
The paper addresses the problem of bounding the degree of irrationality ( $\text{irr}(X)$ ) of Jacobian varieties $J_C$ associated with smooth projective curves $C$ of genus $g \ge 2$ . The degree of irrationality is defined as the minimal degree of a rational dominant map from $X$ to the projective space $\mathbb{P}^{\dim(X)}$ . While lower bounds for abelian varieties of dimension $\ge 3$ are known, relevant upper bounds remain elusive. The standard upper bound for Jacobians, derived by symmetrizing a map $C \to \mathbb{P}^1$ , yields $\text{irr}(J_C) \le \text{gon}(C)^g \le \lfloor \frac{g+3}{2} \rfloor g$ . The authors aim to improve this bound by investigating the positivity properties of specific vector bundles.

Methodology
The authors employ a geometric approach centered on Picard bundles and their behavior on the symmetric product $C^{(g)}$ . The methodology proceeds in three main stages:

Construction of Twisted Picard Bundles:
The authors consider the rank- $g$ Picard bundle $F$ on the Jacobian $J_C$ , whose fiber over $\alpha \in J_C$ is identified with $H^1(C, \mathcal{O}_C(-p_0) \otimes \alpha)$ . They pull this bundle back to the $g$ -fold symmetric product $C^{(g)}$ via the Abel–Jacobi map $\pi: C^{(g)} \to J_C$ . A key step is twisting this pullback by the divisor $x = p_0 + C^{(g-1)} \subset C^{(g)}$ to define a new vector bundle $\tilde{F} := \pi^*F \otimes \mathcal{O}_{C^{(g)}}(x)$ .
Analysis of Global Generation and Sections:
Using results on the generation of Picard bundles (specifically referencing [Par24]), the authors prove that $\tilde{F}$ is globally generated. They analyze the linear systems generated by the determinant maps of sections of $\tilde{F}$ and $F(\Theta)$ . A crucial component of the analysis involves identifying the global sections of $\tilde{F}$ with those of $F(\Theta)$ and describing their vanishing loci in terms of divisors of the form $p + C^{(g-1)}$ .
Application of Kernel Bundle Theory:
The authors utilize a correspondence established in [Mor23] between rational dominant maps $X \dashrightarrow \mathbb{P}^n$ and "good pairs" consisting of a reflexive sheaf and a linear subspace of its global sections. Specifically, they apply Lemma 2.2, which states that if $X$ is a smooth projective variety of dimension $n$ carrying a globally generated vector bundle $E$ of rank $n$ with $c_n(E) > 0$ , then a general linear subspace $V^\vee \subset H^0(E)$ induces a rational dominant map of degree equal to the top Chern class $c_n(E)$ .

Key Contributions and Results

Global Generation and Chern Class Calculation:
The primary technical result is Proposition 3.3, which establishes that the twisted bundle $\tilde{F}$ on $C^{(g)}$ is globally generated. Furthermore, the authors compute the top Chern class of this bundle:
$c_g(\tilde{F}) = 2g$
This calculation relies on the splitting principle and known formulas for the cohomology classes of the pullback of the theta divisor and the divisor $x$ .
Upper Bound on Degree of Irrationality:
By combining the global generation of $\tilde{F}$ with the general principle from [Mor23] (Corollary 2.3), the authors derive the main theorem:
Theorem A: For any smooth curve $C$ of genus $g \ge 2$ , the degree of irrationality of its Jacobian satisfies:
$\text{irr}(J_C) \le 2g$
This improves upon the previously known bound of $\lfloor \frac{g+3}{2} \rfloor g$ for $g \ge 2$ .
Geometric Realization of the Maps:
The paper provides a concrete geometric description of the rational maps $J_C \dashrightarrow \mathbb{P}^g$ constructed via this method. These maps arise from linear projections of the Jacobian embedded in $|(g+1)\Theta|^\vee$ . Geometrically, for a general point $p_1 + \dots + p_g \in C^{(g)}$ , the map sends this point to the intersection of a fixed linear subspace $\mathbb{P}^g \subset \mathbb{P}^{2g-1}$ with the linear span $\langle j(p_1), \dots, j(p_g) \rangle$ , where $j: C \hookrightarrow \mathbb{P}^{2g-1}$ is the embedding defined by the linear system $|\omega_C((g+1)p_0)|$ .
Rigidity and Special Cases:
The authors note that the constructed rational maps are not rigid; they form a large family parametrized by an open subset of the Grassmannian $\text{Gr}(g+1, 2g)$ . In the specific case of genus $g=2$ , Lemma 4.1 demonstrates that for every choice of the subspace $V^\vee \in \text{Gr}(3, H^0(\tilde{F}))$ , the induced map is dominant of degree exactly 4. This suggests that the bound $2g$ is sharp for $g=2$ and that the degree does not drop for special choices of the subspace in this case.

Significance
The paper claims significance by providing a new, tighter upper bound for the degree of irrationality of Jacobian varieties, a problem that has seen limited progress regarding upper bounds for abelian varieties of dimension $\ge 3$ . The work bridges the study of Picard bundles, Brill–Noether loci, and the birational geometry of abelian varieties. By demonstrating that the positivity properties of twisted Picard bundles on symmetric products can be leveraged to construct low-degree rational maps, the authors offer a novel approach to bounding irrationality. The authors explicitly state that they will address the geometric perspective for arbitrary symmetric products and Brill–Noether loci in future work, indicating that the current results are a foundational step in this direction.