🔢 mathematics

Minimal Weierstrass models and regular models of hyperelliptic curves

This paper investigates the properties of minimal Weierstrass models for hyperelliptic curves over discrete valuation fields, specifically characterizing stable reduction for genus 2 curves and demonstrating how multiple minimal models enable the computation of the Euler factor and volume form of the Jacobian's Néron model.

Original authors: Qing Liu

Published 2026-03-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Qing Liu

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 trying to understand a complex, ancient building (a Hyperelliptic Curve) that sits on a piece of land with very specific, tricky soil conditions (a Discrete Valuation Field).

Your goal is to understand the building's true nature, its stability, and how it behaves when the ground shifts slightly (reduction modulo a prime). To do this, you need a perfect blueprint.

The Problem: One Blueprint Isn't Always Enough

Usually, mathematicians try to find the "minimal" blueprint—a single, simplest equation that describes the curve perfectly. This is like finding the one best way to draw a map of a city.

However, Qing Liu's paper discovers a fascinating twist: Sometimes, there isn't just one best blueprint. In fact, for certain curves (specifically those with an "even" complexity, or genus), there can be a whole chain of equally good, minimal blueprints.

Think of it like this: You are trying to describe a mountain range.

Blueprint A shows the mountain from the North.
Blueprint B shows it from the South.
Blueprint C shows it from the East.

In this paper, Liu shows that sometimes you need all these different views (models) to understand the whole picture. If you only look at one, you miss the connections between them.

The "Chain" of Models

When there are multiple minimal models, they aren't random. They are linked in a specific chain.

Imagine a row of houses. The first house ( $W_0$ ) and the last house ( $W_n$ ) are the "extremal" models.
Between them, there are other houses ( $W_1, W_2, \dots$ ).
To get from one house to the next, you have to "zoom in" or "zoom out" using a special mathematical lens (multiplying or dividing by a prime number $\pi$ ).
The paper provides a recipe (an Algorithm) to find every house in this chain if you start with just one. It's like having a GPS that tells you, "If you are at House A, the next best view is House B, and the one after that is House C."

The "Regular Model": The Ultimate Map

Mathematicians want to know what the curve looks like when the ground gets muddy (the "closed fiber"). The "Minimal Regular Model" is the most detailed, smooth, and accurate map of this muddy terrain.

Liu's big discovery is that you don't need to build the whole complex map from scratch.

If you have the two "extremal" blueprints ( $W_0$ and $W_n$ ) from the chain, you can essentially "stitch them together."
The space between them is filled with a chain of simple loops (like a string of pearls or a chain of projective lines).
By looking at just these two extreme models, you can reconstruct the entire complex structure of the curve's reduction. It's like realizing that if you know the shape of the two ends of a bridge, you can figure out the shape of the whole bridge, even if the middle is hidden in fog.

The Special Case: Genus 2 (The "Two-Humped" Curve)

The paper gets even more practical when looking at curves of Genus 2 (curves that look like a double-humped camel or a figure-8).

The Stable Reduction: This is the "ultimate shape" the curve settles into after all the mathematical dust settles.
Liu proves that if a Genus 2 curve has multiple minimal models, its stable shape is two separate loops (like two circles) touching at a single point.
If it has only one minimal model, the shape is different (maybe one big loop with a kink).
Why does this matter? This tells us exactly how the curve behaves in number theory. It helps calculate things like the Euler Factor (a number that predicts how the curve interacts with prime numbers) and the Volume Form (a way to measure the "size" of the curve's associated group).

The "Volume" and "Euler" Analogy

Imagine the curve is a factory producing numbers.

The Volume Form is the size of the factory's output container.
The Euler Factor is the efficiency rating of the factory.

Liu shows that if you have those multiple minimal models (the chain of blueprints), you can calculate the factory's output size and efficiency just by looking at the two extreme blueprints ( $W_0$ and $W_n$ ). You don't need to inspect every single brick of the factory; the two ends tell you everything you need to know.

The Real-World Example: The Modular Curve $X_0(22)$

To prove this works, the author applies the theory to a famous curve called $X_0(22)$ (related to modular forms, which are deep objects in number theory).

They used the algorithm to find that this curve has two minimal models over the 2-adic numbers.
By stitching these two models together, they successfully calculated the exact "Tamagawa number" (a measure of the curve's complexity at the prime 2) and the Euler factor.
It's like using a new, simpler tool to solve a puzzle that previously required a massive, complicated machine.

Summary

In simple terms, Qing Liu's paper says:

Sometimes, one equation isn't enough. Some curves have a family of equally good equations.
These equations form a chain. You can move from one to the next in a predictable way.
The ends tell the whole story. By looking at the first and last equations in the chain, you can reconstruct the entire geometry of the curve's "muddy" reduction.
It makes calculations easy. This allows mathematicians to quickly compute important numbers (like volume and efficiency) for these curves without doing heavy, messy work.

It's a bit like realizing that to understand a long, winding road, you don't need to walk every step. If you know the start and the end, and you know the road is a straight chain of segments, you can figure out the whole journey.

1. Problem Statement

The paper addresses the arithmetic geometry of hyperelliptic curves $C$ of genus $g \ge 2$ over a discrete valuation field $K$ with a perfect residue field $k$ . The central problem is the determination of local arithmetic invariants (such as the Euler factor of the Jacobian, the Tamagawa number, and volume forms) and the structure of the minimal regular model and stable reduction of $C$ .

While regular models provide the necessary geometric framework to compute these invariants, they are often computationally difficult to construct because their special fibers can have a large number of irreducible components with complex intersection patterns. Conversely, Weierstrass models (defined by a single equation $y^2 + Q(x)y = P(x)$ ) are computationally simple but may be singular and do not always capture the full regular structure.

The paper investigates the relationship between these two types of models, specifically focusing on the scenario where a curve possesses more than one minimal Weierstrass model (up to isomorphism). This phenomenon, which occurs primarily when the genus $g$ is even, creates a specific geometric structure that allows for the explicit computation of invariants without constructing the full minimal regular model.

2. Methodology

The author employs a combination of algebraic geometry, rigid analytic geometry, and algorithmic number theory:

Chains of Models: The paper introduces the concept of a "chain of Weierstrass models" connecting two non-isomorphic minimal models. This involves studying the smooth models of $\mathbb{P}^1_K$ (the quotient $C/\langle \sigma \rangle$ ) and their liftings to $C$ .
Multiplicity Analysis: A rigorous analysis of the multiplicities $\lambda(p)$ and singularity measures $\delta(p)$ of points on the special fiber is used to classify minimal models and determine their relationships.
Blow-ups and Desingularization: The author analyzes the resolution of singularities by blowing up points on Weierstrass models, relating the resulting geometry to Tate's algorithm for elliptic curves.
Rigid Analytic Geometry: For complete fields, the curve $C$ is decomposed into affinoid domains and an open annulus. This decomposition allows the curve to be viewed as a union of two curves of genus $g/2$ connected by a chain of projective lines.
Canonical Models: The paper utilizes the canonical model $C_{can}$ (obtained by contracting $(-2)$ -curves in the minimal regular model) as a bridge between the singular Weierstrass models and the smooth minimal regular model.

3. Key Contributions and Results

A. Structure of Minimal Weierstrass Models (Even Genus)

When $g$ is even and $C$ has more than one minimal Weierstrass model, the models form a specific chain structure:

Theorem 0.1 & 1.14: There exists a chain of Weierstrass models $W_0, W_1, \dots, W_n$ (where $n$ is even) such that the minimal models are exactly $W_i$ for even $i$ .
Extremal vs. Inner Models: $W_0$ and $W_n$ are the extremal models. The intermediate models $W_i$ (for even $0 < i < n$ ) are inner.
Equations: If $W_0$ is defined by $y^2 + Q(x)y = P(x)$ , then the other minimal models $W_i$ are given by explicit coordinate transformations involving powers of the uniformizer $\pi$ .
Count: The number of minimal models $m_C$ is bounded by $1 + \frac{\nu(\Delta_C)}{2g(g+1)}$ , where $\nu(\Delta_C)$ is the minimal discriminant valuation.

B. Relation to Regular and Canonical Models

Theorem 0.3 & 2.6: The model $W_0 \vee W_n$ (the smallest model dominating both extremal models) is semi-stable at the intersection point of their special fibers. The minimal regular model $\mathcal{C}$ dominates $W_0 \vee W_n$ .
Geometry of the Special Fiber: The special fiber of the minimal regular model consists of a chain of projective lines connecting the strict transforms of the special fibers of $W_0$ and $W_n$ . Specifically, the fiber is a chain of $(n/2) + 1$ components (including the normalized extremal fibers).
Canonical Model: The canonical model $C_{can}$ dominates $W_0$ and $W_n$ but contracts the inner components. Its special fiber consists of at most two irreducible components (the normalizations of $(W_0)_k$ and $(W_n)_k$ ).

C. Rigid Analytic Decomposition and Stable Reduction

Theorem 2.14 & Corollary 2.23: Over the completion $\hat{K}$ , the curve $C$ decomposes into two affinoid curves $U_0, U_n$ (associated with $W_0$ and $W_n$ ) and an open annulus.
Stable Reduction: The potential stable reduction of $C$ is the union of two stable curves of genus $g/2$ (the stable reductions of the compactifications of $U_0$ and $U_n$ ) intersecting transversely at a single point.
Genus 2 Case: For $g=2$ , the existence of multiple minimal Weierstrass models is equivalent to the special fiber having a reduction type $[K_1 - K_2 - m]$ (where $K_1, K_2$ are Kodaira types of elliptic curves and $m \ge 1$ ).

D. Computation of Arithmetic Invariants (Genus 2)

The paper provides explicit formulas for computing invariants using only the extremal minimal Weierstrass models $W_0$ and $W_n$ :

Volume Form (Proposition 3.12): A canonical isomorphism exists between the determinant of the Neron model's differential forms and a scaled version of the differential forms on $W_0$ :
$\det \omega_{A/\mathcal{O}_K} \simeq \pi^{n/2} \det H^0(W_0, \omega_{W_0/\mathcal{O}_K})$
Euler Factors (Theorem 3.15): If $k$ is finite, the Euler factor of the Jacobian $A$ is the product of the characteristic polynomials of the Frobenius acting on the Picard schemes of the normalizations of the special fibers of $W_0$ and $W_n$ :
$P(A^0_k, T) = P(\text{Pic}^0_{\Gamma_0/k}, T) \cdot P(\text{Pic}^0_{\Gamma_n/k}, T)$
where $\Gamma_0, \Gamma_n$ are the normalizations of $(W_0)_k$ and $(W_n)_k$ at the singular points.

E. Algorithms

Algorithm 1.34: An algorithm is provided to determine all minimal Weierstrass models of a curve given one minimal model. It relies on computing multiplicities $\lambda(p)$ and checking for the existence of specific points with $\lambda = g+1$ or $g+2$ .
Algorithm 3.11: A procedure to detect the existence of stable reduction and determine its type for genus 2 curves based on the minimal Weierstrass models.

4. Significance

Computational Efficiency: The paper demonstrates that for hyperelliptic curves with multiple minimal Weierstrass models, one can bypass the complex construction of the full minimal regular model. Instead, local invariants can be computed directly from the simpler Weierstrass equations.
Structural Insight: It reveals a deep connection between the arithmetic of the Jacobian and the geometry of the "chain" of Weierstrass models. The number of minimal models ( $n/2 + 1$ ) directly corresponds to the length of the chain of rational curves in the stable reduction.
Genus 2 Classification: The results provide a complete characterization of stable reduction types for genus 2 curves in terms of the uniqueness or multiplicity of their minimal Weierstrass models.
Practical Application: The inclusion of algorithms and a concrete example (the modular curve $X_0(22)$ ) demonstrates the practical utility of these theoretical results for computing Tamagawa numbers and Euler factors, which are essential for the Birch and Swinnerton-Dyer conjecture and $L$ -function computations.

In summary, Qing Liu's work bridges the gap between the computational simplicity of Weierstrass models and the geometric complexity of regular models, providing a powerful toolkit for the arithmetic study of hyperelliptic curves, particularly in the genus 2 case.