On canonical bundle formula for fibrations of curves with arithmetic genus one

This paper establishes canonical bundle formulas for fibrations of curves with arithmetic genus one in characteristic $p>0$, distinguishing between separable and inseparable cases, and applies these results to prove that a klt pair with a nef anti-log canonical divisor and a relative dimension one Albanese morphism is a fiber space over its Albanese variety.

The Gist

Imagine you are an architect trying to understand the blueprint of a massive, complex building (a mathematical "variety"). You know the building has a specific structural rule: its "anti-gravity" (the anti-canonical divisor) is pushing outward, keeping the structure stable.

In the world of complex numbers (like our familiar 3D space), mathematicians have long known how to take this building apart. They can say, "This building is essentially a stack of identical floors (fibers) sitting on a flat, perfect foundation (an abelian variety)." This is done using a tool called the Canonical Bundle Formula. Think of this formula as a master key that translates the complex shape of the whole building into a simple description of its foundation and the "glue" (singular fibers) holding it together.

The Problem:
This paper tackles the same problem, but in a very strange, "twisted" universe called Characteristic $p > 0$. In this universe, the rules of geometry are different. It's like trying to build with LEGO bricks that sometimes melt together or disappear when you try to pull them apart. Specifically, the authors are looking at buildings where the "floors" are curves (like circles or figure-eights) that have a special property called arithmetic genus one.

In this twisted universe, two big problems arise:

1. The "Melting" Problem: Sometimes, the connection between the building and its foundation is "inseparable." Imagine trying to pull a thread out of a sweater, but the thread and the sweater have fused into a single, unbreakable blob. Standard math tools break here.
2. The "Singular" Problem: The floors themselves might be weird, crumpled shapes (singular curves) rather than smooth circles.

The Solution: A New Toolkit
The authors, Chen, Wang, and Zhang, have developed a new set of tools to fix the master key (the Canonical Bundle Formula) so it works in this twisted universe.

Here is how they did it, using some creative analogies:

1. The "Leaf" Analogy (Foliations)

When the connection between the building and the foundation is "inseparable" (the melted thread), the authors use a concept called Foliations.

• Imagine: A forest where the trees are growing in a specific direction, forming invisible "sheets" or layers.
• The Trick: Instead of trying to pull the thread apart, they trace the direction of the "leaves" (the foliation). This allows them to navigate the "melted" parts of the geometry without getting stuck. It's like using a map of wind currents to sail through a storm that would otherwise capsize a normal boat.

2. The "Shadow" Analogy (Base Changes)

To handle the weird, crumpled floors, they use a technique called Base Change.

• Imagine: You are looking at a shadow on a wall, but the object casting the shadow is distorted. You can't tell what the object really looks like just by staring at the shadow.
• The Trick: They take the object and shine a different kind of light on it (a "purely inseparable morphism"). This changes the shadow, revealing the true shape of the object underneath. They do this step-by-step, peeling back layers of distortion until the "floor" becomes smooth and understandable again.

3. The "Genus One" Mystery

The specific floors they are studying have "arithmetic genus one."

• In our world: This is like a perfect circle (an elliptic curve).
• In their world: These floors can be "quasi-elliptic." Imagine a circle that has been squashed so hard it looks like a figure-eight or a crumpled ring.
• The Discovery: They found that even when these floors are squashed and the connection to the foundation is "melted," the formula still works, provided the foundation (the base $S$) has a special property called Maximal Albanese Dimension.
  • Analogy: Think of the foundation as a giant, flat, infinite parking lot. If the building sits on this specific type of lot, the authors can prove that the "squashed" floors don't break the rules.

The Big Result (The "Aha!" Moment)

The ultimate goal of this paper is to prove a structural theorem about these buildings.

Theorem: If you have a building with "anti-gravity" pushing out (nef anti-canonical divisor) and the floors are curves (relative dimension one), then the building is a perfect stack of floors over a flat, perfect foundation.

In simpler terms: Even in this weird, twisted universe where things melt and crumple, if the building is stable enough, it turns out to be a very orderly structure. It's not a chaotic mess; it's a clean, organized tower sitting on a perfect, flat base (an abelian variety).

Why Does This Matter?

This is like discovering a new law of physics. Before this, we knew how to build these towers in "normal" math (Complex numbers). We had some hints about how they worked in "twisted" math (Positive characteristic), but the rules were messy and incomplete.

This paper provides the instruction manual for building these structures in the twisted universe. It tells mathematicians:

1. How to calculate the "weight" of the building (the canonical bundle).
2. How to handle the "melted" connections.
3. How to prove that the building is actually a nice, orderly stack.

This paves the way for future discoveries, helping mathematicians classify all possible shapes that can exist in this strange, high-dimensional, "melting" universe. It's a foundational step toward understanding the very fabric of algebraic geometry in positive characteristic.

Technical Summary

Here is a detailed technical summary of the paper "On Canonical Bundle Formula for Fibrations of Curves with Arithmetic Genus One" by Jingshan Chen, Chongning Wang, and Lei Zhang.

1. Problem Statement

The paper addresses the development of canonical bundle formulas for fibrations of relative dimension one in positive characteristic ($p > 0$).

• Context: Over the complex numbers ($\mathbb{C}$), the canonical bundle formula relates the canonical divisor of the total space $X$ to the base $S$ via $K_X + \Delta \sim_{\mathbb{Q}} f^*(K_S + \Delta_S + M)$, where $\Delta_S$ encodes singular fibers and $M$ is a moduli part. This relies heavily on Hodge theory and moduli spaces, which are not available in positive characteristic.
• Specific Challenge: While formulas exist for smooth generic fibers or specific cases (like elliptic fibrations), the situation becomes highly complex when the generic fiber is a singular curve of arithmetic genus one (e.g., quasi-elliptic fibrations occurring in $p=2, 3$).
• Gap: Existing results (e.g., by Witaszek) handle cases where the generic fiber is smooth or the fibration is separable. There is a lack of comprehensive formulas for inseparable fibrations with singular generic fibers (geometrically non-reduced or non-normal) in positive characteristic.
• Application Goal: To understand the structure of varieties with nef anti-canonical divisors ($-(K_X + \Delta)$ is nef), specifically investigating whether the Albanese morphism is a fibration when the relative dimension is one.

2. Methodology

The authors employ a combination of foliation theory, inseparable base changes, and adjunction formulas adapted to positive characteristic.

• Foliations and Inseparable Morphisms: The authors utilize the correspondence between foliations and finite purely inseparable morphisms of height one. This allows them to compare canonical divisors under base changes that are not separable. They define "pushing down" foliations along fibrations to analyze the behavior of relative canonical divisors.
• Analysis of Singular Curves: A significant portion of the work (Section 4) is dedicated to the geometry of regular but non-smooth projective curves of arithmetic genus one over imperfect fields. They classify these curves (conics, non-reduced curves) and analyze the behavior of their normalization and conductor divisors under purely inseparable extensions.
• Iterative Base Change: For inseparable fibrations, the authors construct a sequence of base changes (using Frobenius morphisms and normalization) to reduce the problem to cases where the generic fiber becomes normal or the base becomes separable over its Albanese image.
• Moduli and Linear Systems: They utilize movable linear systems and the theory of stable rational curves (where applicable) to bound the coefficients of the discriminant divisor.

3. Key Contributions and Results

A. Canonical Bundle Formulas

The paper establishes new formulas for $K_X + \Delta$ in terms of the base $S$ under various conditions:

1. Separable Fibrations (Theorem 1.2 & 6.2):

   • For a separable fibration $f: (X, \Delta) \to S$ with lc generic fiber, the authors prove a formula similar to Witaszek's but extended to cases where the generic fiber might be singular (non-smooth).
   • Result: There exist finite purely inseparable morphisms $\tau_1, \tau_2$ and rational numbers $a, b, c \ge 0$ such that:
     $$\tau_2^* \tau_1^* D \sim_{\mathbb{Q}} a K_{\bar{T}'} + b \tau_2^* K_{\bar{T}} + c \tau_2^* \tau_1^* K_S + E_{\bar{T}'}$$
   • Crucially, if the generic fiber is klt, the coefficient $c$ is strictly positive ($c \ge c_0 > 0$), depending on the maximal coefficient of $\Delta$.

2. Inseparable Fibrations with Maximal Albanese Dimension (Theorem 1.3 & 7.2, 7.3):

   • The authors restrict the base $S$ to be of maximal Albanese dimension (m.A.d.).
   • Case 1 (Separable Albanese): If the Albanese morphism $a_S: S \to A$ is separable, they prove $D \succeq_{\mathbb{Q}} \frac{1}{2p} K_S$, implying $\kappa(S, D) \ge \kappa(S)$.
   • Case 2 (Inseparable Albanese): If $a_S$ is inseparable, they prove $\kappa(S, D) \ge 0$. Furthermore, if the generic fiber is klt and $a_S$ is finite, they show $\kappa(S, D) \ge 1$ (unless specific degenerate conditions occur).

B. Structure of Varieties with Nef Anti-Canonical Divisors

The paper applies these formulas to study varieties where $-(K_X + \Delta)$ is nef.

• Theorem 1.5 (Theorem 8.1): Let $(X, \Delta)$ be a projective normal $\mathbb{Q}$-factorial klt pair with $-(K_X + \Delta)$ nef. If the Albanese morphism $a_X: X \to A$ has relative dimension one, then $a_X$ is a fibration (i.e., it is surjective and has connected fibers).
• Significance: This resolves a specific case of the structure theory for varieties with nef anti-canonical divisors in positive characteristic, confirming that the Albanese map behaves like a fiber space even when inseparable phenomena are present, provided the relative dimension is one.

C. Technical Lemmas on Curves of Genus One

• Proposition 4.3 & 4.5: Detailed classification of non-smooth curves of arithmetic genus one. They show that such curves only exist in characteristic $p=2$ or $3$. They analyze the normalization$Y \to X$and the conductor divisor, proving that the normalization is isomorphic to$\mathbb{P}^1$ and characterizing the ramification indices and residue field extensions.

4. Significance and Impact

1. Extension of Birational Geometry: The results extend the canonical bundle formula, a cornerstone of the Minimal Model Program (MMP), to the challenging setting of inseparable fibrations with singular fibers in positive characteristic.
2. Handling Wild Singularities: The paper provides a robust framework for dealing with "wild" singularities (quasi-elliptic fibrations) that do not occur in characteristic zero, utilizing foliation theory to bypass the lack of Hodge theory.
3. Structure Theory: The proof that the Albanese morphism is a fibration for varieties with nef anti-canonical divisors (Theorem 1.5) is a critical step toward a full classification of such varieties in positive characteristic. It suggests that these varieties are built from abelian varieties and elliptic/rational curves via group actions and foliations.
4. Methodological Innovation: The integration of foliation theory (specifically the correspondence between foliations and height-one inseparable morphisms) with the canonical bundle formula offers a new toolkit for researchers working in positive characteristic algebraic geometry.

5. Conclusion

Chen, Wang, and Zhang successfully generalize the canonical bundle formula to inseparable fibrations of curves with arithmetic genus one in positive characteristic. By distinguishing between separable and inseparable cases and leveraging the geometry of the Albanese map, they establish that the base of such fibrations has non-negative Kodaira dimension and that the Albanese morphism itself is a fibration under nef anti-canonical conditions. This work bridges a significant gap between the well-understood complex case and the more intricate positive characteristic landscape.