🔢 mathematics

On the Iitaka conjecture for anticanonical divisors in positive characteristic

This paper establishes the validity of an Iitaka-type inequality for anticanonical divisors in positive characteristic under specific conditions involving threefolds, curves, and strongly F-regular pairs, while also providing counterexamples in characteristics 2 and 3 for fibrations with non-normal fibres derived from Tango–Raynaud surfaces.

Original authors: Marta Benozzo

Published 2026-03-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Marta Benozzo

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 trying to understand the complexity of a massive, multi-story building (let's call it The Source). You want to know how "complicated" the whole building is.

In the world of algebraic geometry (the math behind shapes and spaces), mathematicians use a number called the Kodaira dimension to measure this complexity. Think of it like a "complexity score." A simple point has a low score; a wild, twisting fractal has a high score.

There is a famous rule of thumb called the Iitaka Conjecture. It says:

If you have a building (Source) made of many floors (Fibres) stacked on top of a foundation (Target), the complexity of the whole building should be at least the sum of the complexity of the foundation plus the complexity of a single floor.

$\text{Complexity(Whole)} \ge \text{Complexity(Floor)} + \text{Complexity(Foundation)}$

This rule works perfectly in "Characteristic 0" (our normal, familiar mathematical world). But this paper asks: Does this rule still work in "Positive Characteristic"?

What is "Positive Characteristic"?

Think of "Characteristic 0" as a smooth, continuous world like a video game with infinite resolution.
"Positive Characteristic" is like a video game with a low frame rate or a pixelated grid. In this world, math behaves differently. Some rules that work smoothly in the real world break when you try to apply them to this "pixelated" grid. Specifically, things that usually cancel out or disappear (like certain types of "noise" in the math) might get stuck and cause problems.

The New Rule: The "Anti-Complexity" Conjecture

The author, Marta Benozzo, isn't looking at the standard complexity score. She is looking at the Anticanonical score.

Standard Complexity ( $K$ ): Measures how "curved" or "twisted" a shape is.
Anti-Complexity ( $-K$ ): Measures how "straight" or "simple" a shape is. Think of it as the "Simplicity Score."

The Anticanonical Conjecture (the one she is testing) flips the inequality:

The Simplicity of the Whole Building should be less than or equal to the Simplicity of the Floor plus the Simplicity of the Foundation.

$\text{Simplicity(Whole)} \le \text{Simplicity(Floor)} + \text{Simplicity(Foundation)}$

The Main Discovery: "It Depends on the Floor"

Benozzo proves that this new rule does hold true in the "pixelated" world, BUT only under very specific conditions:

The Floor Must Be "Regular": The individual floors of the building must be smooth and well-behaved. They can't be crumpled or broken.
The Floor Must Be "Strongly F-Regular": This is a fancy math term that essentially means the floor is "sturdy" and doesn't have any hidden cracks that the "pixelated" math would exploit.
The Building Size: It works if the building is a 3D structure (a threefold) or if the foundation is just a line (a curve).

The Analogy:
Imagine building a tower out of Lego bricks.

If the bricks are perfect (Strongly F-regular), the tower's stability follows the rules.
If the bricks are warped or broken (non-normal), the tower might collapse in weird ways that break the rules.

The "Gotcha": When the Rules Break

The most exciting part of the paper is the Counterexamples.

Benozzo shows that if you remove the condition that the floors must be "regular" (smooth), the rule fails completely in the pixelated world (specifically in "Characteristics 2 and 3").

How did she break it?
She used a special mathematical object called a Tango–Raynaud Surface.

The Metaphor: Imagine a special type of Lego brick that looks normal from the outside but has a hidden internal mechanism. In the "pixelated" world, these bricks have a weird property: they can hold "ghost energy" (mathematically, non-zero cohomology) that shouldn't exist.
The Result: By stacking these weird bricks, she built a tower where the "Simplicity Score" of the whole tower is 0 (very simple), but the "Simplicity Score" of the foundation is negative infinity (chaotic).
The Violation: The rule said the Whole should be less than or equal to Floor + Foundation. But here, $0 \not\le \text{Floor} + (-\infty)$ . The math breaks!

Why Does This Matter?

In the "real" world (Characteristic 0), we have powerful tools (like the "Canonical Bundle Formula") that act like a safety net, catching these errors. In the "pixelated" world (Positive Characteristic), that safety net is full of holes.

This paper is crucial because:

It tells us exactly where the safety net works (when floors are smooth).
It shows us exactly where it fails (when floors are broken), using the Tango–Raynaud surfaces as the "smoking gun."

Summary in One Sentence

Marta Benozzo proved that a rule about measuring the "simplicity" of geometric shapes works in the weird, pixelated world of positive characteristic only if the shapes are perfectly smooth, but if you try to use broken shapes, the rule collapses, revealing deep and surprising differences between our normal math world and this exotic one.

1. Problem Statement

The paper investigates the validity of the Iitaka conjecture for anticanonical divisors ( $C^-_{n,m}$ ) over fields of positive characteristic ( $p > 0$ ).

The Conjecture: Let $f: X \to Z$ be a fibration between normal projective varieties. Let $\Phi$ be a general fiber. The conjecture posits the inequality:
$\kappa(X, -(K_X + \Delta)) \leq \kappa(\Phi, -(K_\Phi + \Delta_\Phi)) + \kappa(Z, -K_Z)$
where $\kappa$ denotes the Iitaka dimension, $K$ is the canonical divisor, and $\Delta$ is an effective divisor.
Context:
- In characteristic 0, this was proven by Chang [Cha23] under specific conditions (klt singularities, stable base locus not dominating the target).
- In positive characteristic, the classical Iitaka conjecture (for $K_X$ ) is known to fail generally due to the failure of Kodaira vanishing and the existence of inseparable morphisms.
- The paper asks whether the anticanonical version holds in positive characteristic and identifies the precise conditions under which it succeeds or fails.

2. Methodology

The author employs a combination of techniques from birational geometry in positive characteristic, specifically focusing on F-singularities and weak positivity of sheaves.

F-Singularities: The paper relies heavily on the theory of Strongly F-regular (SFR) and Sharply F-pure pairs. These are positive-characteristic analogues of klt and log canonical singularities, respectively. The assumption that the general fiber is SFR is crucial for the positive results.
Weak Positivity: Following the work of Ejiri and Gongyo [EG19], the author utilizes the weak positivity of direct image sheaves (specifically $f_* \mathcal{O}_X(m(K_X+\Delta))$ ) to establish inequalities between Iitaka dimensions.
Canonical Bundle Formula: The author adapts results from Witt [Wit21] to handle the relative dimension 1 case, as the full canonical bundle formula (Ambro [Amb05]) is not available in full generality in positive characteristic.
Counterexample Construction: To demonstrate failure, the paper constructs specific fibrations using Tango–Raynaud surfaces. These surfaces exploit the failure of Kodaira vanishing in characteristics 2 and 3, specifically utilizing curves with non-trivial $H^1(C, \mathcal{O}_C(-D))$ to create non-split extensions that yield vector bundles with specific properties.

3. Key Contributions and Results

A. Positive Results (Theorems 0.3 and 0.4)

The paper proves the conjecture $C^-_{n,m}$ under specific geometric constraints:

Case $C^-_{n,1}$ (Target is a curve):
- Theorem 0.3: Proven for a fibration $f: X \to Z$ where $Z$ is a smooth curve.
- Conditions: The general fiber $\Phi$ must be regular, and the pair $(\Phi, \Delta_\Phi)$ must be strongly F-regular. Additionally, the stable base locus of $-(K_X+\Delta)$ must not dominate $Z$ .
- Method: Uses weak positivity of sheaves and the "Easy Additivity" theorems.
Case $C^-_{3,m}$ (Source is a threefold):
- Theorem 0.4: Proven for a fibration from a normal projective threefold $X$ to a normal projective variety $Z$ over a field of characteristic $p \geq 5$ .
- Conditions: Same as above (regular fiber, SFR pair, base locus condition).
- Key Insight: When $X$ $X$ is a threefold, the base $Z$ $Z$ or the fiber $\Phi$ $Φ$ must be a curve.
  - If $Z$ is a curve, Theorem 0.3 applies.
  - If $Z$ is a surface, the author analyzes the Iitaka fibration of $-K_Z$ . For $p \geq 5$ , the general fibers of the anticanonical Iitaka fibration on a surface are smooth elliptic curves (quasi-elliptic fibrations only exist in $p=2,3$ ). This smoothness allows the reduction to the curve case.

B. Negative Results (Counterexamples)

The paper demonstrates that the assumptions on the singularities of the fibers and the base locus are necessary.

Failure in Characteristics 2 and 3:
- Theorem 5.4: Constructs a counterexample to $C^-_{7,6}$ in characteristics $p=2$ and $p=3$ .
- Construction: Uses Tango–Raynaud surfaces derived from Tango curves (curves where the Tango invariant $n(C) > 0$ ).
- Mechanism:
  - A non-split extension in $H^1(C, \mathcal{O}_C(-D))$ defines a rank-2 vector bundle $E$ .
  - The projectivization $P(E)$ yields a surface with specific properties.
  - By taking fiber products $X(m) = S \times_C \dots \times_C S$ , the author constructs a fibration $X(6) \to X(5)$ .
- Result:
  - $\kappa(X(6), -K_{X(6)}) = 0$ .
  - $\kappa(X(5), -K_{X(5)}) = -\infty$ .
  - The inequality $0 \leq -\infty + \kappa(\dots)$ fails.
- Crucial Observation: The fibers of this counterexample are non-normal. This highlights that the "regular fiber" assumption in the positive theorems is essential.
Failure of Base Locus Assumption:
- The paper notes that if the stable base locus of $-K_X$ dominates the target $Z$ , counterexamples exist even in dimension 2 (ruled surfaces), regardless of characteristic.

4. Significance and Implications

Boundary of Validity: The paper precisely delineates the boundary where the Iitaka conjecture for anticanonical divisors holds in positive characteristic. It confirms that the conjecture is true for low-dimensional sources ( $n=3$ ) and curve targets, provided the fibers are "well-behaved" (regular and SFR) and the characteristic is not too small ( $p \geq 5$ for the threefold case).
Role of Singularities: It establishes that the Strongly F-regular condition on the fiber is a necessary substitute for the "klt" condition used in characteristic 0. Without this, or if the fibers are non-normal, the inequality breaks down.
Characteristic Dependence: The results highlight a sharp distinction between characteristics $p \geq 5$ and $p \in \{2, 3\}$ . The existence of quasi-elliptic fibrations and the specific behavior of Tango curves in low characteristics are the primary obstructions.
Methodological Advancement: The work successfully adapts the "weak positivity" strategy of [EG19] and the "canonical bundle formula" approach of [Cha23] to the positive characteristic setting, overcoming the lack of a general canonical bundle formula by restricting to cases where the geometry of the fibers is controlled.

Conclusion

Marta Benozzo's paper provides a comprehensive analysis of the anticanonical Iitaka conjecture in positive characteristic. It proves the conjecture for threefolds and curve targets under strong regularity and F-singularity assumptions, while simultaneously providing explicit counterexamples in characteristics 2 and 3 when fibers are non-normal. This work underscores the delicate interplay between the geometry of fibers, the characteristic of the field, and the validity of additivity formulas for Kodaira dimensions.