🔢 mathematics

Non-existence of Shimura curves of Mumford type generically in the non-hyperelliptic locus

The paper proves that there are no Shimura curves with strictly maximal Higgs fields generically located in the Torelli locus of non-hyperelliptic curves of genus $g \geq 4$ , thereby establishing that Shimura curves of Mumford type do not generically exist within this locus.

Original authors: Xin Lu, Shengli Tan, Kang Zuo

Published 2026-05-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Xin Lu, Shengli Tan, Kang Zuo

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

The Big Picture: A Map of Shapes and Special Roads

Imagine the mathematical world as a massive, infinite library called $\mathcal{A}_g$ . This library doesn't contain books, but rather shapes (specifically, complex geometric objects called abelian varieties).

Inside this library, there is a special, exclusive section called the Torelli Locus ( $T_g$ ). Think of this section as a "Gallery of Curves." Every object in this gallery is a "Jacobian," which is a special shape built directly from a simpler, curvy line (an algebraic curve).

The authors of this paper are investigating a specific type of "road" or path that can be drawn through this library. These roads are called Shimura curves.

Normal roads: Most paths in the library are just random.
Shimura roads: These are "Golden Roads." They are perfectly straight (in a mathematical sense called "totally geodesic") and connect to very special, highly symmetric points called CM points (Complex Multiplication points).

The paper asks a specific question: Can we find a "Golden Road" that stays entirely inside the "Gallery of Curves," but specifically avoids the "Hyperelliptic" section?

The Cast of Characters

The Curves (The Art):
- Hyperelliptic Curves: These are like symmetrical, double-sided mirrors. If you fold them in half, the two sides match perfectly. They are very common and easy to recognize.
- Non-Hyperelliptic Curves: These are the "wild" ones. They are asymmetrical and more complex. They don't have that perfect mirror symmetry.
- The Gallery ( $T_g$ ): Contains both types.
- The Wild Gallery ( $T_g \setminus TH_g$ ): Contains only the non-hyperelliptic (asymmetrical) curves.
The Special Roads (Shimura Curves of Mumford Type):
- These are a specific, rare breed of Golden Roads. They are famous because the objects they pass through have almost no internal symmetries (their "endomorphism ring" is just the integers, $\mathbb{Z}$ ). They are the "lonely" roads of the library.
- They are known to exist in the library, but the question is: Do they ever wander into the "Wild Gallery"?
The Higgs Field (The Engine):
- To prove a road is a "Golden Road," mathematicians look at its engine, called the Higgs field.
- If the engine is running at "Strictly Maximal" speed, the road is definitely a Golden Road of the "Mumford" type.
- The paper focuses on roads where this engine is running at maximum capacity.

The Story of the Proof

The authors, Xin Lu, Shengli Tan, and Kang Zuo, set out to prove a negative: No such Golden Road exists in the Wild Gallery for curves of genus 4 or higher.

Here is how they did it, using a construction analogy:

1. The Setup: Building a Bridge
Imagine you are trying to build a bridge (a family of curves) that represents a Golden Road.

If the road stays in the "Hyperelliptic" section (the mirror section), the bridge is easy to build and stable.
But if the road tries to enter the "Wild Gallery" (non-hyperelliptic), the bridge has to twist and turn. The authors show that if the road is a "Mumford type" (strictly maximal engine), the bridge must be built in a very specific, rigid way.

2. The Conflict: The Slope Inequality
The authors use a mathematical tool called a Slope Inequality. Think of this as a physics law for bridges.

The Law: "If a bridge is built this way (representing a Mumford curve in the wild), it must be incredibly steep and heavy to stay standing."
The Reality: When they calculate the actual weight and steepness required by the geometry of the "Wild Gallery," they find it's impossible to build. The bridge would collapse.

3. The "Double Cover" Twist
The proof gets tricky because of a "double cover" (a mirror trick).

When a road enters the Wild Gallery, the path it takes is actually a "double loop" around a simpler path.
The authors analyze the "ramification points" (where the loop twists). They found that for the road to be a "Mumford type" (strictly maximal), the twisting points create a mathematical contradiction.
It's like trying to fit a square peg into a round hole, but the hole is also shrinking as you push the peg in. The math simply doesn't add up.

The Conclusion

The paper concludes that it is impossible to draw a "Mumford type" Golden Road that stays strictly inside the "Wild Gallery" (non-hyperelliptic curves) for any curve with 4 or more "holes" (genus $g \ge 4$ ).

What this means: If you see a Golden Road (Shimura curve) with a "Strictly Maximal" engine, and it is located in the world of complex curves, it must pass through the "Hyperelliptic" (mirror) section. It cannot hide entirely in the "Wild" section.

Why Should You Care? (Without the Jargon)

This paper is a bit like a detective story in the world of shapes.

The Mystery: "Where do these special, lonely roads live?"
The Clue: We knew they lived in the library, but we didn't know if they avoided the "Wild" section.
The Verdict: The authors proved they cannot avoid the "Wild" section entirely. If they are there, they must touch the "Mirror" section.

This helps mathematicians understand the "map" of the universe of shapes better. It tells us that the most special, symmetric roads and the most complex, asymmetrical shapes are actually more intertwined than we thought. You can't have one without eventually bumping into the other.

In short: The paper proves that a specific type of "perfectly straight, lonely path" cannot exist only among the "asymmetrical, wild shapes" of the mathematical world. It must visit the "symmetrical" neighborhood at some point.

Technical Summary: Non-Existence of Shimura Curves of Mumford Type Generically in the Non-Hyperelliptic Locus

Problem Statement
The paper addresses the intersection of Shimura curves with the Torelli locus of algebraic curves. Specifically, it investigates the existence of Shimura curves of "Mumford type" (a specific class of Shimura curves of Hodge type not of PEL type) that are contained generically in the Torelli locus of non-hyperelliptic curves.

The authors focus on the question posed by Oort regarding genus $g=4$ : Is there any Shimura curve of Mumford type contained generically in the Torelli locus $T_4$ ? More generally, the paper seeks to determine if such curves exist for any genus $g \ge 4$ within the non-hyperelliptic locus. This inquiry is situated within the broader context of the Coleman-Oort conjecture, which predicts that for sufficiently large $g$ , no positive-dimensional special subvarieties (Shimura curves) exist generically in the Torelli locus.

Methodology
The proof relies on a combination of Hodge theory, the geometry of semi-stable families of curves, and slope inequalities. The methodology proceeds through the following steps:

Characterization via Higgs Bundles: The authors utilize the characterization of Shimura curves established by Viehweg and Zuo. A curve $C_0 \subseteq \mathcal{A}_g$ is a Shimura curve of Mumford type if and only if the associated logarithmic Higgs bundle admits a strictly maximal Higgs field. This condition is equivalent to the Arakelov equality holding for the degree of the Hodge bundle: $\deg E^{1,0}_C = \frac{g}{2} \deg \Omega^1_C(\log \Delta_C)$ .
Lifting to Semi-Stable Families: For a Shimura curve $C_0$ contained generically in the Torelli locus, the authors construct a family of semi-stable curves $f: S \to B$ representing $C_0$ via the Torelli morphism.
- If $C_0$ is not contained in the hyperelliptic locus, the map from the normalization of the inverse image of the curve in the moduli space of curves to the curve itself is either an isomorphism or a double cover.
- The case where the map is an isomorphism is ruled out by previous results (LZ19). Thus, the analysis focuses on the case where the map $\pi: B \to C$ is a double cover with a ramification locus $\Lambda$ .
Slope Inequality Derivation: The core technical contribution involves deriving a refined slope inequality for the family $f: S \to B$ .
- The authors construct a specific unramified base change $\psi: \tilde{B} \to B$ that lifts the involutions associated with the double cover and the hyperelliptic involution on singular fibers.
- By analyzing the quotient of the family by these involutions, they derive a lower bound for the self-intersection number $\omega^2_{S/B}$ (Theorem 3.1). This bound includes terms dependent on the ramification locus $\Lambda$ and the geometry of the singular fibers (specifically the number of irreducible components of specific genera).
Contradiction via Upper Bounds: The authors compare the derived lower bound for $\omega^2_{S/B}$ with an existing upper bound established in their previous work (LZ19, Theorem 4.3).
- The upper bound depends on the degree of the Hodge bundle and the configuration of singular fibers.
- By combining the strictly maximal Higgs field condition (which fixes the degree of the Hodge bundle relative to the base) with the slope inequalities, the authors derive an inequality involving the number of specific types of singular fibers ( $\Delta_{ct,b}$ ).

Key Results

Theorem 1.2: There does not exist any Shimura curve of Mumford type contained generically in the Torelli locus of non-hyperelliptic curves of genus $g=4$ .
Theorem 1.4: There does not exist any Shimura curve with a strictly maximal Higgs field generically in the Torelli locus of non-hyperelliptic curves for any genus $g \ge 4$ .
Corollary 1.5: If $g > 4$ , no Shimura curve with a strictly maximal Higgs field is contained generically in the Torelli locus $T_g$ .

The proof of Theorem 1.4 proceeds by contradiction. Assuming the existence of such a curve leads to an inequality of the form:
$\frac{7-g}{6} |\Delta_{ct,b}| > \deg f^*\omega_{S/B} \ge \frac{(g-1)^2}{2g+1} |\Delta_{ct,b}|$
For $g \ge 4$ , the coefficient on the left becomes non-positive or insufficient to satisfy the inequality against the positive lower bound on the right, yielding a contradiction.

Significance and Claims
The paper claims to resolve Question 1.1 posed by Oort regarding genus 4 by proving the non-existence of such curves in the non-hyperelliptic locus. Furthermore, it generalizes this result to all genera $g \ge 4$ .

The significance lies in the following:

Extension of Previous Work: The authors previously proved the non-existence of such curves in the hyperelliptic locus for $g \ge 8$ (LZ17) and established finiteness results for superelliptic curves (CLZ21). This paper extends the non-existence result to the non-hyperelliptic locus for all $g \ge 4$ .
Support for Coleman-Oort Conjecture: The results provide strong evidence for the Coleman-Oort conjecture, specifically ruling out a major class of potential counterexamples (Shimura curves of Mumford type) in the non-hyperelliptic Torelli locus for all $g \ge 4$ .
Methodological Refinement: The paper introduces a refined slope inequality (Theorem 3.1) that accounts for the specific geometry of families representing Shimura curves of Mumford type, particularly handling the ramification of the double cover and the structure of singular fibers.

The authors do not claim to resolve the full Coleman-Oort conjecture for all special subvarieties or all genera, but rather establish a definitive negative result for the specific class of Shimura curves of Mumford type in the non-hyperelliptic locus.