🔢 mathematics

Elliptic matroids and modular curves

This paper establishes a natural isomorphism between the open modular curve $X_1(n)^\circ$ and the matroid realization space of the elliptic matroid $T_n$ over $\mathbb{Z}[1/n]$ , thereby providing an algebraic link between matroid theory and the classification of rational torsion points on elliptic curves.

Original authors: Matthew Baker

Published 2026-08-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Matthew Baker

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 the universe of mathematics as a giant, invisible playground where shapes and numbers dance together. In one corner of this playground, there's a game called "Matroid Theory." Think of a matroid not as a physical object, but as a rulebook for how points can line up. It's like a set of instructions that says, "If you have three points, they can only stand in a straight line if their secret numbers add up to zero." It's a way of describing geometry without needing to draw the picture first. In another corner, there's the world of "Modular Curves." These are like magical maps that track the behavior of special shapes called elliptic curves—think of them as donuts with a very specific, twisty geometry that shows up in everything from cryptography to the deepest mysteries of number theory.

For a long time, mathematicians have wondered if these two corners of the playground are actually the same place. If you take the rulebook of the matroid game and try to build it with real points on a flat surface, does it automatically create one of those magical elliptic curve maps? This question is tricky because it involves translating between a rigid set of logical rules (the matroid) and a fluid, geometric shape (the curve). Solving this isn't just about winning a game; it helps us understand the fundamental building blocks of numbers and shapes, and it has surprising connections to how we secure digital information.

This paper, written by Matthew Baker, acts as a master key that finally unlocks the door between these two worlds, but only for a specific size of the game. The author proves that for any number $n$ that is 10 or larger, the "rulebook" of the matroid game (called the elliptic matroid $T_n$ ) and the "map" of the modular curve (called $X_1(n)^\circ$ ) are actually identical. It's like discovering that two different languages, which look completely different on the surface, are actually just different dialects of the exact same language.

The paper does something remarkable: it proves this identity not just for the complex numbers (the usual playground for these shapes), but for any field of numbers, as long as the number $n$ doesn't share any factors with the "characteristic" of the field (a fancy way of saying the rules of arithmetic don't break in a weird way). The author shows that if you have a collection of points that follow the matroid's line-up rules, you can always find a unique, smooth, or slightly cracked (nodal) cubic curve that passes through all of them, and that curve is exactly the one described by the modular map.

The author is very careful to note that this perfect match only works when $n \ge 10$ . If you try to play the game with fewer points (like $n=5$ or $n=6$ ), the rulebook is too simple—it only allows for one single arrangement of points—while the map is still a complex, winding curve. In those small cases, the two things are definitely not the same. The paper also explicitly rules out the idea that this works for $n$ values that divide the characteristic of the field (like trying to do the math in a system where $1+1=0$ when $n$ is even), because the rules of the game break down there.

To prove this, the author uses a clever trick involving "seeds" and "propagation." Imagine you have nine points that form a specific grid. You can draw two different sets of three lines that connect them, creating two "reducible" shapes (like a triangle made of three sticks). Where these two shapes cross, you get exactly nine points. The paper uses a classic geometric rule (Chasles' theorem) to say that if you have a tenth point that fits the pattern, it forces a unique curve to pass through all ten. Once you have that curve, the author uses a "deformation" argument—a method of checking if the shape holds up even when you wiggle the numbers slightly—to prove that the connection between the points and the curve is unbreakable, not just a lucky coincidence.

The result is a complete, algebraic proof that for $n \ge 10$ , the abstract logic of the matroid and the geometric reality of the modular curve are one and the same. This isn't just a guess or a simulation; it's a rigorous mathematical proof. As a bonus, the paper connects this finding to a famous theorem by Barry Mazur about the "torsion points" (special repeating points) on elliptic curves. It shows that the fact that you can't build the matroid $T_p$ with rational numbers for any prime $p \ge 11$ is exactly the same as Mazur's famous result that there are no rational points of order $p$ on elliptic curves for those primes. In short, the paper proves that the logic of the game and the geometry of the map are perfectly synchronized, but only when the game is big enough to be interesting.

Technical Summary: Elliptic Matroids and Modular Curves

Problem Statement
The paper addresses the relationship between the combinatorial structure of the "elliptic matroid" $T_n$ and the geometry of the modular curve $X_1(n)$ . Specifically, for an integer $n \ge 4$ , $T_n$ is defined as the rank-3 matroid on the ground set $\mathbb{Z}/n\mathbb{Z}$ where the non-bases are the three-element subsets $\{a, b, c\}$ satisfying $a+b+c \equiv 0 \pmod n$ . This structure arises naturally from the collinearity relations of $n$ -torsion points on an elliptic curve embedded as a plane cubic.

The central problem is to determine whether the space of realizations of $T_n$ over a field $k$ (up to rescaling and projective equivalence) is isomorphic to the moduli space of elliptic curves with a point of order $n$ , denoted $X_1(n)^\circ$ . Here, $X_1(n)^\circ$ is the open subscheme of the compactified modular curve $X_1(n)$ where the universal generalized elliptic curve is irreducible (i.e., smooth elliptic curves or irreducible nodal cubics).

Prior work by Borisov and Roulleau established a set-theoretic bijection between these spaces over $\mathbb{C}$ for $n \ge 10$ , relying on complex-analytic methods and weight-one modular forms. The paper seeks to extend this correspondence to arbitrary fields of characteristic not dividing $n$ and to upgrade the correspondence from a set-theoretic bijection to an isomorphism of schemes over $\mathbb{Z}[1/n]$ .

Methodology
The paper employs a purely algebraic and incidence-theoretic approach, avoiding modular forms and complex analysis. The methodology is divided into two parts:

Field-Valued Correspondence (Part 1):
- Seed Pencil Construction: Given a realization of $T_n$ , the author constructs a "seed pencil" of cubic curves. By selecting specific indices, they form two reducible cubics (unions of three lines) whose scheme-theoretic intersection consists of nine distinct marked points.
- Chasles Propagation: Using a classical theorem of Chasles (a specific case of the Cayley-Bacharach theorem), the author shows that if a cubic passes through eight of the nine intersection points of two reducible cubics, it must pass through the ninth. By arranging the indices of $T_n$ into "zero-sum grids" (3x3 arrays where row and column sums are zero), they propagate the existence of a unique cubic $C$ containing all marked points $P_i$ .
- Reconstruction of Group Law: Once all points lie on an irreducible cubic $C$ , the author proves $C$ is either smooth or nodal (excluding the cuspidal case via the characteristic assumption). They then recover the group law on the smooth locus $C_{sm}$ with identity $P_0$ , demonstrating that the collinearity relations of the matroid force the points to satisfy $P_i = i P_1$ and that $P_1$ has exact order $n$ .
- Small Levels: Special arguments using combinatorial lemmas on zero-sum triples handle the cases $n=10$ and $n=11$ , where the general propagation argument requires slight modification.
Scheme-Theoretic Correspondence (Part 2):
- Artinian-Point Criterion: To establish an isomorphism of schemes, the author utilizes a criterion stating that a morphism of finite presentation between Noetherian schemes is an isomorphism if and only if it induces a bijection on points valued in every local Artinian algebra.
- Deformation Theory: The field-valued reconstruction is lifted to the relative setting over local Artinian rings. Using Nakayama's lemma and base change arguments, the "seed pencil" and "Chasles propagation" are shown to hold for deformations. This allows the reconstruction of the relative cubic and the group law over Artinian rings, proving that the map is a bijection on $A$ -valued points for all such $A$ .

Key Contributions and Results

Theorem 1.1 (Field-Valued Correspondence): For $n \ge 10$ and any field $k$ with $\text{char}(k) \nmid n$ , the map $\beta_k: X_1(n)^\circ(k) \to RT_n(k)$ is a bijection. This recovers the Borisov–Roulleau theorem over $\mathbb{C}$ using elementary projective geometry and Bézout's theorem.
Theorem 1.2 (Scheme-Theoretic Correspondence): For $n \ge 10$ , the modular morphism $\beta: X_1(n)^\circ \to R_n$ is an isomorphism of schemes over $\mathbb{Z}[1/n]$ , where $R_n$ is the realization scheme of the matroid $T_n$ . This provides a natural integral model for the modular curve as a matroid realization space.
Connection to Mazur's Theorem: The paper establishes that for primes $p \ge 11$ , the non-representability of the matroid $T_p$ over $\mathbb{Q}$ is equivalent to the statement that $X_1(p)^\circ(\mathbb{Q}) = \emptyset$ . Since $X_1(p)^\circ(\mathbb{Q})$ parametrizes elliptic curves over $\mathbb{Q}$ with a rational point of order $p$ , this equivalence links the combinatorial non-representability of $T_p$ directly to Mazur's celebrated theorem on rational torsion points.
Appendix on Band Schemes: The author constructs a canonical "realization band scheme" for matroids over the initial band $\mathbb{F}_{\pm 1}$ , upgrading the affine scheme realization to a structure over $\mathbb{F}_{\pm 1}$ in the sense of Baker–Jin–Lorscheid.

Significance and Claims
The paper claims to provide a "purely algebraic and incidence-theoretic" proof of the Borisov–Roulleau correspondence, removing the reliance on complex analysis and modular forms. By upgrading the result to an isomorphism of schemes over $\mathbb{Z}[1/n]$ , the work endows the modular curve $X_1(n)^\circ$ with a new geometric interpretation as a matroid realization space.

The author notes that the hypothesis $n \ge 10$ is partly an artifact of their method; for $n \le 6$ , the realization space is a point while $X_1(n)^\circ$ is a curve, making a bijection impossible. For $7 \le n \le 9$ , the bijection holds, but their specific propagation method fails due to the lack of a "tenth point" to select a unique cubic from the pencil.

The paper explicitly states that the technical scheme-theoretic arguments in Part 2 were largely developed with the assistance of AI (specifically GPT-5.5), which generated the precise formalism for the 3x3 zero-sum grids and the deformation arguments, while the author provided the core algebraic strategy and verified all content. The work serves as a bridge between matroid theory, algebraic geometry, and the arithmetic of elliptic curves, offering a new perspective on the moduli of elliptic curves via combinatorial realizability.

Technical Summary: Elliptic Matroids and Modular Curves

More like this