🔢 mathematics

Prym-Brill-Noether Theory for General Covers

This paper establishes new dimension bounds for Prym-Brill-Noether varieties of general étale double covers of k-gonal curves, disproving a conjecture by Creech et al., by utilizing a complete combinatorial description of these varieties on a specific tropical "loop of loops" curve and applying Coxeter group theory to prove lifting results.

Original authors: David Jensen

Published 2026-07-02

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: David Jensen

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 a master architect trying to design a very specific type of building. In the world of mathematics, this "building" is a geometric shape called a Prym-Brill-Noether variety.

To understand what the author, David Jensen, has done, we first need to understand the materials he's working with and the problem he's trying to solve.

The Setting: A Double-Decker City

Imagine a city called Curve City ( $C$ ). Now, imagine a second city, Double-City ( $\tilde{C}$ ), built directly on top of it. Every street in Double-City has a twin street in Curve City below it, and they are connected by invisible bridges. This is what mathematicians call an "étale double cover."

In this Double-City, there are special "divisors." Think of a divisor as a collection of specific landmarks (like street corners or parks) that you can visit. The rules of the city say that if you visit a certain set of landmarks in Double-City, you must be able to "see" a specific pattern when you look down at Curve City. This pattern is called the Prym condition.

The Prym-Brill-Noether variety is essentially a "map" or a "directory" that lists all the possible ways you can arrange these landmarks in Double-City so that they satisfy the Prym condition and have a certain level of "complexity" (called rank).

The Problem: How Big is the Directory?

For a long time, mathematicians knew how big this directory was for a "generic" city (a city with no special features). They had a formula for the size (dimension) of this directory.

However, they wanted to know: What happens if the city has special features?
Specifically, what if the city is built on a loop (like a donut shape, or an elliptic curve) or if the city has a specific "gonal" structure (meaning it can be mapped to a line in a specific way)?

Previous researchers had made guesses (conjectures) about the size of these directories for these special cities. Some of these guesses were wrong.

The Solution: The "Loop of Loops" Model

To solve this, Jensen didn't try to build the actual complex cities. Instead, he built a skeleton model using tropical geometry.

Think of tropical geometry as a way of studying shapes by turning them into wireframes or stick figures. Instead of smooth curves, you have straight lines and sharp corners.

Jensen chose a very specific, somewhat strange-looking wireframe model called the "Loop of Loops."

Imagine a chain of loops (like a chain of rings).
Now, imagine a "Loop of Loops" where the loops themselves are made of smaller loops. It looks like a fractal of rings.

Jensen used this model because it acts like a perfect test case. If you can figure out the rules for this wireframe model, you can often figure out the rules for the real, smooth cities.

The Discovery: Counting with "Lingering Words"

Here is the clever part. Jensen realized that every valid arrangement of landmarks in his wireframe model could be translated into a word made of letters.

The Letters: These letters come from a mathematical system called Coxeter groups (think of them as a set of rules for swapping things around).
The Lingering: Sometimes, in these words, a letter might be an "empty space" or a "pause." Jensen calls these "lingering words." It's like writing a sentence where some words are optional, but the overall meaning (the structure of the city) must remain intact.

He discovered that the size of the directory (the dimension of the variety) depends entirely on how many of these "lingering words" exist that follow specific rules.

The Main Results

1. The "k-Elliptic" Cities (Cities with a Donut Connection)
Jensen looked at cities that have a special connection to a donut shape (genus 1). He found that the size of the directory depends on how "tight" this connection is (a number $k$ ).

The Result: He proved a new, tighter formula for the size of the directory.
The Correction: He showed that a previous guess by other mathematicians (Creech, Len, Ritter, and Wu) was wrong. Their guess was too optimistic; the directory is actually smaller than they thought in many cases.

2. The "k-Gonal" Cities (Cities with a Specific Map)
He also looked at cities that can be mapped to a line in a specific way ( $k$ -gonal).

The Result: He provided a new, better upper bound (a maximum limit) for the size of the directory.
The New Guess: He didn't just stop at the limit; he proposed a new, more precise formula for what the size should be, based on a concept called "orthogonal splitting types" (which is like checking if the city's streets can be split into perfect, symmetrical pairs).

The "Lifting" Trick

One of the most powerful tools Jensen used is called a lifting theorem.

The Metaphor: Imagine you have a shadow of a 3D object cast on a wall (the tropical wireframe model). You know the shape of the shadow perfectly. Jensen proved that if the shadow has a certain structure, you can be 100% sure that the real 3D object (the actual mathematical city) exists and has the exact same structure.
Why it matters: This allowed him to take his results from the simple wireframe model and apply them directly to the complex, real-world mathematical curves, proving his formulas are correct for the "general" cases.

Summary

David Jensen took a difficult problem about the size of mathematical directories for special types of curves. He solved it by:

Building a simplified "wireframe" model called the Loop of Loops.
Translating the geometry of the model into lingering words (a type of mathematical code).
Using the rules of these words to calculate the exact size of the directory.
Proving that these results "lift" back up to the real mathematical world, correcting previous errors and providing new, more accurate formulas.

In short, he used a stick-figure model and a code of "lingering words" to fix a broken map of a complex mathematical city.

Technical Summary: Prym-Brill-Noether Theory for General Covers

Problem Statement
The paper addresses the dimension of the Prym-Brill-Noether variety $V^r(C, f)$ , defined as the set of divisor classes $[D]$ in the Prym variety of an étale double cover $f: \tilde{C} \to C$ such that the rank of $D$ is at least $r$ . While the dimension for general covers (where $C$ has genus $g$ ) was established by Bertram and Welters as $g - 1 - \binom{r+1}{2}$ , the behavior of this dimension for covers where the base curve $C$ possesses special geometric properties—specifically $k$ -gonal curves or $k$ -elliptic curves—remains less understood. Previous work by Len and Ulirsch, and Creech, Len, Ritter, and Wu, provided bounds for these cases but left open the question of optimality and the validity of specific conjectures regarding the dimension.

Methodology
The author employs tropical geometry as the primary tool to analyze these varieties. The approach involves:

Tropicalization: Constructing a tropical analogue of the Prym-Brill-Noether variety, $V^r(\Gamma, \pi)$ , for a metric graph $\Gamma$ and a free double cover $\pi: \tilde{\Gamma} \to \Gamma$ .
Specific Tropical Models: Instead of the "chain of loops" used in prior literature, the paper utilizes a "loop of loops" metric graph. This graph consists of $2g-2$ vertices arranged in a specific cycle of loops.
Combinatorial Encoding: The paper establishes a bijection between divisor classes in the tropical Prym variety and "lingering words" in Coxeter groups (specifically type $A_r$ and type $I_2(r+1)$ ). A lingering word allows for identity elements, mirroring the flexibility of divisor positions on the graph.
Lifting Theorems: By proving that the tropical variety is a local complete intersection and utilizing Rabinoff's lifting theorem, the author demonstrates that divisor classes in the tropical setting lift to the algebraic setting over non-archimedean fields. This allows the transfer of dimension bounds from the tropical graph to the algebraic curve.

Key Contributions and Results

Tropical Description of $V^r(\Gamma, \pi)$ :
For an $r$ -generic loop of loops (where edge lengths satisfy specific genericity conditions), the paper completely describes the tropical Prym-Brill-Noether variety. It is shown to be a union of tori indexed by lingering words in the Coxeter group $A_r$ that contain a reduced subword for the longest element $w_0$ .
- The variety is proven to be pure dimensional and connected in codimension 1.
- When the dimension is zero (i.e., $g-1 = \binom{r+1}{2}$ ), the cardinality of the variety matches the number of reduced words for $w_0$ , which corresponds to the number of standard Young tableaux on a staircase partition.
Bounds for $k$ -Elliptic Covers (Theorem 1.1):
The paper establishes an upper bound on the dimension of $V^r(C, f)$ when $C$ is a $k$ -elliptic curve (admitting a degree- $k$ map to a genus 1 curve).
$\dim V^r(C, f) \leq \begin{cases} g - 1 - \frac{k(r+1)}{2} & \text{if } k < r \\ g - 1 - \frac{r(r+1)}{2} & \text{if } k \geq r \end{cases}$
Equality holds when $k \geq r$ . This result is derived by analyzing $k$ -uniform loops of loops and proving a lifting result for these specific tropical configurations.
Bounds for $k$ -Gonal Covers (Theorem 1.3):
For general $k$ -gonal curves (admitting a degree- $k$ map to $\mathbb{P}^1$ ), the paper provides an improved upper bound on the dimension, refining previous results by Len, Ulirsch, and others. The new bound is:
$\dim V^r(C, f) \leq \begin{cases} g - 1 - n(r, k) & \text{if } \ell(\ell+1) \leq r+1 \\ g - 1 - \frac{(k-1)(r+1)}{2} & \text{if } k+1 \leq r+1 < \ell(\ell+1) \\ g - 1 - \binom{r+1}{2} & \text{if } k \geq r+1 \end{cases}$
where $\ell = \lceil k/2 \rceil$ and $n(r, k)$ is a previously defined function.
Disproof of Conjecture 3.9:
The paper explicitly disproves Conjecture 3.9 from Creech, Len, Ritter, and Wu [CLRW22], which proposed a different dimension formula for $k$ -gonal covers. The counterexample arises from the analysis of the "loop of loops" and the specific combinatorics of the associated Coxeter groups.
New Conjecture (Conjecture 1.4):
Motivated by the splitting type loci of orthogonal bundles on $\mathbb{P}^1$ (following the work of H. Larson), the author proposes a new conjecture for the dimension of $V^r(C, f)$ in the $k$ -gonal case. This conjecture suggests the dimension is exactly $g - 1 - (k-1)(r+1) + \binom{k}{2}$ when $k < r+1$ , and $g - 1 - \binom{r+1}{2}$ otherwise.
Bielliptic Case ( $k=2$ ):
In the special case of bielliptic curves ( $k=2$ ), the paper shows that the tropical variety is indexed by lingering words in the dihedral group $I_2(r+1)$ . This structure preserves many topological properties of the generic case, including purity of dimension and connectivity in codimension 1.

Significance and Claims
The paper claims to provide a new proof of the Bertram-Welters theorem for general covers and offers a significant improvement on existing bounds for covers of $k$ -gonal and $k$ -elliptic curves. By utilizing the "loop of loops" tropical model, the author demonstrates that the combinatorics of Coxeter groups (specifically the structure of lingering words and the Bruhat order) can effectively encode the geometry of Prym-Brill-Noether varieties.

The lifting theorem (Theorem 1.2) is central to the work, establishing that for $r$ -generic loops of loops, every tropical divisor class of rank $r$ lifts to an algebraic divisor class. This allows the author to deduce properties of the algebraic variety (such as reducedness and dimension) directly from the tropical combinatorics. The paper concludes that while the upper bounds derived are not always optimal, they represent a substantial refinement over previous literature and provide a framework for a more precise conjecture regarding the dimension of these varieties.