🔢 mathematics

On the superadditivity of anticanonical Iitaka dimension

This paper establishes an Iitaka-type inequality for the anticanonical Iitaka dimension of a fibration over a field of arbitrary characteristic, proving that $\kappa(X,-K_X) \leq \kappa(X_y,-K_{X_y}) + \kappa(Y,-K_Y)$ under the condition that the $\mathbb{Q}$ -linear series $|-K_X|_{\mathbb{Q}}$ exhibits good singularities on the normal general fibre.

Original authors: Marta Benozzo, Iacopo Brivio, Chi-Kang Chang

Published 2026-03-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Marta Benozzo, Iacopo Brivio, Chi-Kang Chang

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 the "size" and "complexity" of a massive, multi-story building. In mathematics, specifically in a field called Algebraic Geometry, these buildings are called varieties, and their complexity is measured by something called the Iitaka dimension.

Think of the Iitaka dimension as a "complexity score." A flat sheet of paper has a low score. A twisting, turning, multi-dimensional sculpture has a high score.

The Big Picture: The "Superadditivity" Puzzle

For decades, mathematicians have been trying to solve a puzzle about how these complexity scores behave when you build a big structure out of smaller parts.

Imagine your big building (let's call it $X$ ) is actually a stack of smaller, identical rooms (the fibres, or $X_y$ ) arranged along a hallway (the base, or $Y$ ).

The Old Rule (Canonical): If you look at the "standard" complexity of the whole building, it turns out to be at least the sum of the complexity of a single room plus the complexity of the hallway. (Room + Hallway $\le$ Whole Building). This is like saying a skyscraper is at least as complex as a single floor plus the elevator shaft.
The New Rule (Anticanonical): The authors of this paper are investigating the opposite kind of complexity. Instead of measuring how "bent" or "curved" the building is (which is the standard way), they are measuring how "straight" or "open" it is. They call this the Anticanonical dimension.

Their big discovery is a new inequality:

The complexity of the whole building (Anticanonical) $\le$ The complexity of one room + The complexity of the hallway.

Wait, why is this surprising? Usually, when you add things up, the whole is bigger than the parts. Here, they are proving that for this specific type of "openness," the whole building cannot be more complex than the sum of its parts. It's like saying a giant, open-air stadium cannot be more "open" than a single seat plus the walkway leading to it.

The Problem: The "Rough" Terrain

This rule works perfectly in the world of Complex Numbers (think of the smooth, perfect geometry of the classical world). But the authors wanted to know: Does this rule work in "Positive Characteristic"?

In math, "Positive Characteristic" is like a world where the rules of arithmetic wrap around, like a clock. It's a bit like a video game where gravity works differently. In this weird world, things get "rough."

Sometimes, a "room" that looks smooth from the outside is actually crumpled or broken on the inside.
Sometimes, the hallway doesn't connect to the rooms in a clean way.

If the rooms are broken or the connection is messy, the "complexity score" calculation breaks down. The old rules fail.

The Solution: The "Smoothness" Check

The authors realized that the rule does hold in this rough, video-game world, BUT only if the rooms are "well-behaved."

They introduced a special condition called $Q^0_y(X, 0) = k(y)$ .

The Analogy: Imagine you are inspecting the rooms before you calculate the score. You need to make sure the "Frobenius" (a magical force in this video-game world that tests the structural integrity of the walls) doesn't find any hidden cracks.
If the room passes this "Frobenius stress test" (meaning the singularities are "good"), then the rule holds: Whole $\le$ Part + Hallway.
If the room fails the test (it's too crumpled), the rule might fail.

How They Proved It (The Detective Work)

The paper is essentially a detective story where they use three main tools to solve the case:

The "Descent" Trick: They showed that if the whole building has a certain property, that property must "flow down" to the hallway. If the building is "open," the hallway must be "open" too.
The "Injectivity" Theorem: They proved that if you try to map the "openness" of the whole building down to a single room, you don't lose any information. It's like taking a high-resolution photo of the whole stadium and zooming in on one seat; you still see the seat clearly. This ensures the math doesn't collapse.
The "Frobenius" Base Change: In the rough video-game world, they had to use a special technique called "Frobenius base change."
- Analogy: Imagine the hallway is a wobbly, unstable bridge. To fix it, they didn't just patch it; they ran the bridge through a "time machine" (the Frobenius map) multiple times. This process, while strange, smoothed out the wrinkles and made the bridge (the general fibre) stable and normal. Once the bridge was stable, they could apply their standard math tools to prove the rule.

Why Does This Matter?

This paper is a bridge between two different universes:

The Smooth World (Characteristic 0): Where everything is perfect and predictable.
The Rough World (Characteristic $p$ ): Where arithmetic wraps around and things get messy.

The authors are showing us that even in the messy, rough world, if we check the "structural integrity" of the rooms (using F-singularities), the same beautiful laws of geometry that hold in the smooth world still apply.

In simple terms:
They proved that for a specific type of geometric building, the total "openness" is limited by the sum of the room's openness and the hallway's openness. This is true even in a weird, mathematically "rough" universe, provided the rooms aren't too broken. It connects the rules of the smooth, classical world to the chaotic, modular world of modern number theory.

1. Problem Statement and Context

The paper addresses the superadditivity of the anticanonical Iitaka dimension in the context of fibrations $f: X \to Y$ . This is the "anticanonical" counterpart to the famous Iitaka Conjecture ( $C_{n,m}$ ), which posits that for a fibration of smooth projective varieties, the Kodaira dimension satisfies $\kappa(X, K_X) \geq \kappa(X_y, K_{X_y}) + \kappa(Y, K_Y)$ .

While the original conjecture for the canonical divisor is largely established (via the Minimal Model Program), the inequality for the anticanonical divisor ( $-K_X$ ) behaves differently:

In Characteristic 0: The inequality $\kappa(X, -K_X) \leq \kappa(X_y, -K_{X_y}) + \kappa(Y, -K_Y)$ holds under specific conditions regarding the base locus and singularities.
In Positive Characteristic ( $p > 0$ ): The situation is more complex. The failure of generic smoothness (fibers may be non-normal or non-reduced) and the lack of a canonical bundle formula analogous to the characteristic zero case (Ambro's formula) have historically obstructed proofs.

The Core Problem: The authors aim to establish the inequality $\kappa(X, -K_X) \leq \kappa(X_y, -K_{X_y}) + \kappa(Y, -K_Y)$ for fibrations in positive characteristic, identifying the precise arithmetic conditions on the general fiber required for the inequality to hold.

2. Methodology

The authors employ a strategy that blends techniques from the Minimal Model Program (MMP), the theory of F-singularities (singularities defined via the Frobenius morphism), and foliation theory.

A. Characteristic Zero Approach (The Model)

The paper first refines the proof for characteristic zero (Theorem 1.4), following the structure of Chi-Kang Chang's previous work. The proof relies on four main steps:

Descent of Q-effectivity: Using the condition that the asymptotic multiplier ideal $\mathcal{I}(X_y, 0; \|-K_X\|_y)$ is trivial, they construct a "lc-trivial" fibration. This allows the use of the canonical bundle formula to descend the positivity of $-K_X$ to $-K_Y$ .
Precise Descent: A perturbation argument ensures that if $-K_X - f^*E$ is effective, then $-K_Y - \epsilon E$ is effective.
Injectivity Theorem: They prove that if $\kappa(Y, -K_Y) = 0$ , the restriction map on sections $H^0(X, -lK_X) \to H^0(X_y, -lK_{X_y})$ is injective.
Reduction to Calabi-Yau Base: By resolving the Iitaka fibration of $-K_Y$ , they reduce the problem to a base where $-K_Y$ is numerically trivial, allowing the application of the injectivity result.

B. Positive Characteristic Adaptation

To extend this to characteristic $p > 0$ , the authors replace characteristic zero tools with their positive characteristic analogues:

Multiplier Ideals $\to$ Test Ideals/F-singularities: Instead of the multiplier ideal $\mathcal{I}$ , they utilize asymptotic test ideals and the subspace of Frobenius-stable sections denoted by $Q^0_y(X, B; M)$ .
Canonical Bundle Formula: Since Ambro's formula fails in general for $p>0$ , they utilize a result by Das and Schwede regarding globally F-split fibrations. They show that if the general fiber is globally F-split (or satisfies the condition $Q^0_y = k(y)$ ), a positive characteristic canonical bundle formula exists.
Handling Non-Normal Fibers: A major obstacle in $p>0$ is that the Iitaka fibration of $-K_Y$ may have non-normal fibers. The authors overcome this by performing a purely inseparable base change (via powers of the Frobenius morphism) and taking the normalization of the resulting spaces. This yields a new fibration with normal fibers while preserving the Iitaka dimension (via properties of foliations and inseparable maps).

3. Key Contributions and Results

Main Theorems

Theorem 1.3 (Positive Characteristic):
Let $f: X \to Y$ be a generically smooth fibration over an algebraically closed field of characteristic $p > 0$ . If:
- (a) $Q^0_y(X, 0) = k(y)$ (a global version of the asymptotic test ideal is trivial, implying good F-singularities on the fiber), and
- (b) The stable base locus of $-mK_X$ (for $p \nmid m$ ) does not dominate $Y$ ,
  Then:
  $\kappa(X, -K_X) \leq \kappa(X_y, -K_{X_y}) + \kappa(Y, -K_Y)$
  Furthermore, if $\kappa(Y, -K_Y) = 0$ , equality holds.
Theorem 1.4 (Refined Characteristic Zero):
A generalization of the characteristic zero result, stating that if the asymptotic multiplier ideal $\mathcal{I}(X_y, 0; \|-K_X\|_y)$ is trivial, the inequality holds.

Technical Innovations

Definition of $Q^0_y$ : The authors define a new invariant $Q^0_y(X, B; M)$ , which measures the "F-singularities" of the anticanonical linear series on a general fiber. This serves as the positive characteristic analogue to the triviality of the multiplier ideal in characteristic zero.
Injectivity Theorem in $p>0$ (Theorem 5.1): They prove an injectivity theorem for the restriction map of sections under the assumption of the existence of F-complements. This is the positive characteristic counterpart to the characteristic zero injectivity theorem.
Base Change via Frobenius: They develop a technique to resolve singularities of fibers in the Iitaka fibration of the base by using inseparable base changes and foliation theory (Theorem 6.6, Corollary 6.7), ensuring the fibers become normal without altering the Iitaka dimension.

4. Significance and Implications

Bridging Characteristics: The paper provides a robust framework for transferring results about anticanonical divisors from characteristic zero to positive characteristic. It demonstrates that "good" F-singularities (specifically global F-splitting or the triviality of $Q^0$ ) play the same role in positive characteristic as "good" singularities (klt/trivial multiplier ideals) do in characteristic zero.
Counterexamples and Necessity: The authors clarify that the inequality fails in general for $p>0$ without these conditions, primarily due to the failure of generic smoothness. Their work identifies exactly how the arithmetic properties of the fiber (Frobenius action) must behave to recover the inequality.
Conjectures on Reduction: The paper proposes Conjecture 8.2, suggesting that if the multiplier ideal is trivial in characteristic zero, then the corresponding F-singularity condition ( $Q^0 = k$ ) holds for infinitely many primes $p$ . This connects the geometry of complex varieties to the arithmetic of their reductions modulo $p$ .
K-Globally F-regular Varieties: The authors introduce the notion of K-globally F-regular varieties (interpolating between globally F-regular and globally F-split) and conjecture that log Fano/Calabi-Yau pairs in characteristic zero reduce to these types in positive characteristic for a dense set of primes.

Summary

This paper successfully establishes the superadditivity of the anticanonical Iitaka dimension in positive characteristic, provided the general fiber has sufficiently well-behaved F-singularities (measured by $Q^0_y$ ). By adapting the canonical bundle formula and injectivity theorems to the positive characteristic setting and utilizing inseparable base changes to normalize fibers, the authors provide a comprehensive solution to a problem that was previously obstructed by the lack of generic smoothness in characteristic $p$ .