🔢 mathematics

On generalized canonical bundle formula and boundedness of complements in complex analytic setting

This paper establishes the generalized canonical bundle formula for generalized lc-trivial fibrations with irrational coefficients over non-compact bases in the complex analytic setting, demonstrating the compatibility of discriminant and moduli b-divisors with restriction to arbitrary open subsets while also addressing the boundedness of complements.

Original authors: Kenta Hashizume

Published 2026-03-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Kenta Hashizume

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 massive, complex building. In the world of mathematics, this building is a complex analytic variety (a shape defined by complex numbers), and the "blueprints" are equations describing its geometry.

For decades, mathematicians have had a powerful tool called the Canonical Bundle Formula. Think of this formula as a "master key" that lets you understand the whole building by looking at its foundation and its roof separately. It tells you how the "curvature" (a measure of how bent or twisted the shape is) of the whole structure relates to the curvature of its base and its fibers (the vertical slices).

However, until now, this key only worked well for buildings that were "compact" (finite and closed, like a sphere) or had very specific, "rational" numbers in their blueprints. Real-world shapes in complex analysis are often "non-compact" (infinite or open-ended) and involve "irrational" numbers (like $\pi$ or $\sqrt{2}$ ), which made the old key useless.

Kenta Hashizume's paper is like inventing a new, super-versatile master key that works for any building, no matter how open-ended or mathematically messy its blueprints are.

Here is a breakdown of the paper's two main achievements using simple analogies:

1. The Generalized Canonical Bundle Formula (The "Universal Translator")

The Problem:
Imagine you are trying to describe a long, winding river (the "fibration") that flows from a mountain (the "base") down to the sea.

Old Method: You could only describe the river if the map was a closed, finite island. If the river went on forever, or if the water contained "weird" ingredients (irrational coefficients), the old math broke down.
The New Discovery: Hashizume created a new way to translate the river's properties. He proved that even if the river flows infinitely and has weird ingredients, you can still break it down into two parts:
1. The Discriminant (The "Terrain"): This describes the bumps and dips of the riverbed caused by the landscape.
2. The Moduli (The "Flow"): This describes the actual movement of the water, independent of the terrain.

The Magic Trick:
The most difficult part of this paper was proving that if you zoom in on just a small section of the river (an "open subset"), the math still holds up.

Analogy: Imagine taking a photo of a tiny part of a massive tapestry. Usually, if you zoom in, the pattern gets blurry or changes because you lose the context of the whole. Hashizume proved that for these specific mathematical shapes, the pattern on the tiny photo is exactly compatible with the pattern on the whole tapestry. You can study a small piece and be 100% sure it represents the whole.

2. The Boundedness of Complements (The "Safety Net")

The Problem:
In geometry, "singularities" are like sharp corners, tears, or knots in the fabric of the shape. These are dangerous spots where the math gets messy.

Complements: Think of a "complement" as a safety net or a patch you can sew onto the fabric to smooth out the sharp corners.
Boundedness: The big question was: "Is there a limit to how big or complex this safety net needs to be?" If the knots get infinitely complex, do we need an infinitely large net?

The Discovery:
Hashizume proved that no, there is a limit.

Analogy: Imagine you are a tailor fixing torn clothes. You might worry that some tears are so weird you'll need a patch the size of a house. Hashizume proved that no matter how weird the tear (as long as it's a certain type of "complex analytic" tear), you only ever need a patch from a finite, pre-defined catalog of sizes.
He showed that for any shape of a certain dimension (say, 3D), there is a specific list of "patch sizes" (integers) that will always work to fix the shape, making it smooth and manageable.

Why Does This Matter?

This paper is a bridge.

It expands the territory: It allows mathematicians to apply powerful tools to "open" and "infinite" shapes, not just closed, finite ones. This is crucial for complex analysis, where shapes are often open.
It unifies the language: It uses "Generalized Pairs," a modern language that treats the "shape" and its "extra ingredients" (the nef part) as a single unit. This makes solving problems much easier.
It solves a long-standing puzzle: By proving that "complements" are bounded, it confirms that these complex shapes, while wild, are actually tame in a fundamental way. They don't spiral out of control; they follow strict, predictable rules.

In a Nutshell:
Hashizume took a set of mathematical rules that only worked for "perfect, closed boxes" and upgraded them to work for "infinite, messy, real-world shapes." He showed that even in the most chaotic, open-ended mathematical landscapes, there is an underlying order, and we can always find a finite, manageable way to describe and fix them.

1. Problem Statement and Context

The paper addresses two fundamental problems in birational geometry within the complex analytic setting (specifically for non-compact bases and non-projective varieties):

The Generalized Canonical Bundle Formula: Extending the canonical bundle formula to generalized lc-trivial fibrations over non-compact bases in complex analytic spaces. A major difficulty here is defining the restriction of $b$ -divisors to open subsets, as the standard algebraic compactification arguments do not apply directly to open analytic varieties.
Boundedness of Complements: Proving that for generalized pairs with specific singularity and positivity conditions, there exists a bounded set of integers $n$ such that an $n$ -complement exists. This is a crucial step in the Minimal Model Program (MMP) and the classification of Fano varieties.

The work operates within the framework of generalized pairs (introduced by Birkar–Zhang), where the "nef part" $M$ is given by $b$ -nef $\mathbb{R}$ - $b$ -divisors. This contrasts with other analytic approaches that use currents; the author chooses $b$ -divisors to maintain compatibility with projective morphisms between analytic spaces.

2. Methodology

The author employs a combination of techniques from algebraic geometry adapted to the complex analytic category:

Generalized Pairs Framework: Utilizing the definitions of generalized sub-pairs, log discrepancies, and adjunction in the analytic setting (following Fujino and Birkar).
Weak Semistable Reduction: Adapting theorems by Kawamata, Hacon–Păun, and Enokizono–Hashizume to construct weak semistable reductions for fibrations over analytic bases. This is essential for analyzing the discriminant and moduli parts of the bundle formula.
Relative MMP (Minimal Model Program): Using the relative MMP for generalized klt pairs in the analytic setting (referencing Enokizono–Hashizume and Fujino) to reduce lc-trivial fibrations to klt-trivial cases or to run specific MMP steps (e.g., flipping, contracting) to simplify the boundary divisor.
Convex Geometry and Limit Arguments: To handle irrational coefficients in the nef part $M$ , the author approximates $M$ by rational combinations of $b$ -nef divisors, applies results for the rational case, and takes limits.
Extension of Complements: Using the Connectedness Principle (Kawamata–Viehweg vanishing) and Generalized Inversion of Adjunction to lift complements from a boundary divisor (or a subvariety) to the total space.

3. Key Contributions and Results

A. Generalized Canonical Bundle Formula (Section 3)

The paper establishes a robust version of the canonical bundle formula for generalized lc-trivial fibrations $\pi: (X, B+M) \to Z$ over a non-compact base $S$ .

Theorem 1.1 (Theorem 3.10): For a generalized lc-trivial fibration where $M$ is a finite $\mathbb{R}_{>0}$ -linear combination of $b$ -nef $\mathbb{Q}$ - $b$ -Cartier divisors:
1. The discriminant $\mathbb{R}$ - $b$ -divisor $G$ and the moduli $\mathbb{R}$ - $b$ -divisor $N$ are well-defined.
2. $N$ is a finite $\mathbb{R}_{>0}$ -linear combination of $b$ -nef $\mathbb{Q}$ - $b$ -Cartier divisors on $Z$ .
3. The pair $(Z, G_Z + N_Z)$ forms a generalized sub-pair with the nef part $N$ .
4. Compatibility with Restriction: A critical technical breakthrough is proving that $G$ and $N$ are compatible with restriction to arbitrary open subsets $\tilde{S} \subset S$ . This resolves the ambiguity of defining $b$ -divisor restrictions in the analytic setting (Lemma 3.3).
5. The singularities are preserved: If $(X, B+M)$ is generalized lc (resp. klt), then $(Z, G_Z + N_Z)$ is also generalized lc (resp. klt).
Significance: This extends previous results by Filipazzi (algebraic) and Fujino (analytic klt case) to the generalized lc case with irrational coefficients over non-compact bases.

B. Boundedness of Complements (Section 4)

The paper proves the boundedness of complements for generalized pairs in the analytic setting, generalizing results by Birkar, Filipazzi–Moraga, and Chen–Han–He–Xie.

Theorem 1.2 (Theorem 4.2): Let $\pi: (X, B+M) \to Z$ be a generalized lc pair of dimension $d$ with coefficients in a DCC set $\Lambda$ (accumulation points rational) and $pM$ being $b$ -Cartier. If $X$ is of Fano type over $Z$ and $-(K_X+B+M)$ is nef, then there exists an integer $n$ (depending only on $d, p, \Lambda$ ) such that $(X, B+M)$ has an $n$ -complement locally around any point $z \in Z$ .
- Condition: The complement $B^+$ satisfies $B^+ \geq B$ .
Theorem 1.3 (Theorem 4.3): A more general result where $-(K_X+B+M)$ is not necessarily nef, but there exists an effective $\mathbb{R}$ -divisor $\Delta$ making the pair generalized klt with $K_X+B+\Delta+M \sim_{\mathbb{R}} 0$ . This implies the existence of an $n$ -complement for some $n$ in a finite set $N$ (depending on $d, p$ ).
Corollaries:
- Corollary 1.4: Recovers the classical boundedness of complements for standard lc pairs (setting $M=0, X=Z$ ).
- Corollary 1.5: Establishes the boundedness of complements for klt pairs, ensuring the existence of $B^+$ such that $n(K_X+B^+)$ is Cartier.

4. Technical Challenges Overcome

Restriction of $b$ -divisors: In the analytic setting, open subsets do not necessarily admit compactifications, making the definition of $b$ -divisor restrictions non-trivial. The author proves that for the specific context of canonical bundle formulas, the discriminant and moduli parts behave well under restriction (Lemma 3.3), allowing the use of relative MMP techniques locally.
Irrational Coefficients: The formula allows $M$ to have irrational coefficients. The author handles this by approximating $M$ with rational divisors and using convexity arguments to pass to the limit, ensuring the moduli part remains $b$ -nef.
Non-Compact Bases: Unlike algebraic geometry where global projectivity is often assumed, the analytic setting requires working over Stein compact subsets and shrinking neighborhoods. The proofs carefully manage these local-to-global transitions using properties of Stein spaces and the relative MMP.

5. Significance and Impact

Foundational for Analytic MMP: This paper solidifies the theoretical framework for the Minimal Model Program in the complex analytic category, particularly for generalized pairs. It bridges the gap between the algebraic theory (where boundedness of complements is a cornerstone) and the analytic theory.
Applications: The results are prerequisites for:
- Proving the Ascending Chain Condition (ACC) for minimal log discrepancies in the analytic setting.
- Addressing moduli problems for Calabi–Yau pairs and log Fano pairs in complex geometry.
- Investigating the Yau–Tian–Donaldson conjecture and the projectivity of K-moduli spaces in the analytic context.
Generalization: It generalizes the work of Birkar (algebraic) and Filipazzi (algebraic generalized) to the complex analytic realm, showing that the powerful machinery of complements and canonical bundle formulas is robust enough to handle non-compact and non-projective analytic varieties.

In summary, Hashizume successfully extends the "boundedness of complements" program to the complex analytic setting by first establishing a rigorous generalized canonical bundle formula that respects the topological constraints of analytic spaces. This provides the necessary tools to study singularities and moduli in complex geometry with the same depth as in algebraic geometry.