🔢 mathematics

Arithmetic Fujita approximation over adelic curves

This paper proves a conjecture by Huayi Chen and Atsushi Moriwaki by establishing an analogue of Fujita's approximation theorem within the framework of Arakelov theory over adelic curves.

Original authors: Chunhui Liu

Published 2026-01-27

📖 5 min read🧠 Deep dive

CC0 1.0

Original authors: Chunhui Liu

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ 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: Smoothing Out a Rough Mountain

Imagine you are a geologist studying a very jagged, rough mountain range. This mountain represents a complex mathematical object called a line bundle (think of it as a specific way of wrapping a sheet of paper around a shape).

In the world of pure geometry, there is a famous rule called Fujita's Approximation. It says: "Even if your mountain is jagged and weird, you can always find a nearby, perfectly smooth, and round hill (called an ample bundle) that looks almost exactly like your mountain, provided you zoom out far enough."

This paper is about proving that this rule still works in a much more complicated, "super-charged" version of geometry called Arakelov theory over adelic curves.

The Setting: The "Adelic Curve" (The Universal Map)

To understand the setting, imagine a standard map of a country. Now, imagine a "Super-Map" that doesn't just show the land, but also includes every possible way of measuring distance on that land simultaneously—measuring in miles, kilometers, inches, and even abstract "mathematical units" all at once.

The Adelic Curve: This is the "Super-Map." It combines different types of number systems (like the numbers we use for counting, and the numbers used in complex analysis) into one giant framework.
The Challenge: In this Super-Map world, things get messy. The usual tools mathematicians use to smooth out the "jagged mountains" often break because the measurements are too chaotic or "noisy."

The Problem: The "Rough Mountain" in the Super-Map

The authors are tackling a specific question posed by two other famous mathematicians, Chen and Moriwaki:
"Does the 'smoothing rule' (Fujita's Approximation) still work when we are using this messy Super-Map?"

Previously, mathematicians knew how to do this for simple maps (like standard number fields). But for the Super-Map, the usual methods failed. It was like trying to smooth a crumpled piece of paper using a tool that only works on flat surfaces.

The Solution: A New Way to "Filter" the Mess

Chunhui Liu proves that yes, the smoothing rule does work, even in this messy Super-Map. Here is how the paper achieves this, using an analogy:

1. The "Harder-Narasimhan" Filter (The Sieve)

Imagine you have a bucket of mixed sand and rocks (the messy data). You want to separate the valuable rocks from the useless sand.

The paper uses a mathematical "sieve" called the Harder-Narasimhan filtration.
This sieve sorts the data based on how "strong" or "valuable" different parts of the mountain are. It separates the "best" parts from the "weakest" parts.

2. The "Spectral Norm" (The New Ruler)

In the messy Super-Map, the old rulers didn't work well. The author invents a new way to measure the "size" of the data, called the spectral norm.

Think of this as switching from a standard tape measure to a "magic ruler" that can measure the true potential of the mountain, ignoring the static and noise of the Super-Map.

3. The "Small Section" (The Anchor)

In the old days, to smooth a mountain, you needed a "perfect" piece of land. In this new Super-Map, perfect pieces are rare.

Instead, the author uses a concept called a "non-zero small section."
Analogy: Imagine you are trying to anchor a tent. You don't need a perfect, solid rock; you just need one small, sturdy twig that holds the tent down firmly enough. The paper proves that even in the chaotic Super-Map, you can always find at least one "sturdy twig" (a small section) to hold your approximation together.

The Main Result: The "Smooth Hill" Exists

The paper concludes with a powerful theorem (Theorem 1.1):

No matter how complex your "mountain" (the adelic line bundle) is, and no matter how much error you are willing to tolerate (epsilon), you can always:

Zoom out (multiply the bundle by a large number $p$ ).
Find a smoother path (a birational morphism, which is like taking a detour to a better view).
Decompose it into two parts:
- Part A: A perfectly smooth, round hill (an arithmetically ample bundle).
- Part B: A tiny, manageable anchor (a bundle with a non-zero small section).

The volume (the "amount of stuff") of this smooth hill is almost identical to the volume of your original rough mountain.

Why This Matters (According to the Paper)

It Answers a Conjecture: It confirms a guess made by Chen and Moriwaki that this smoothing process works even in the most general, complex mathematical settings.
It Unifies Math: It shows that the rules of geometry (how shapes behave) and the rules of arithmetic (how numbers behave) can be smoothed out together, even when the underlying "map" is incredibly complicated.
It Handles "Real" Numbers: The paper doesn't just work for simple fractions; it works for "Real" numbers (infinite decimals), which makes the result much more robust.

Summary in One Sentence

Chunhui Liu proves that even in the most chaotic and complex mathematical "Super-Maps," you can always find a way to approximate a rough, jagged shape with a smooth, perfect one, provided you use a new type of mathematical "sieve" and a "magic ruler" to filter out the noise.

Technical Summary: Arithmetic Fujita Approximation over Adelic Curves

Problem Statement
The paper addresses the extension of Fujita's approximation theorem to the framework of Arakelov geometry over adelic curves. In classical algebraic geometry, Fujita's theorem asserts that a big line bundle can be approximated arbitrarily closely by an ample line bundle tensored with an effective line bundle, up to a birational modification. In the arithmetic setting over number fields, this conjecture was previously established by Chen and Yuan using distinct methods (plurisubharmonic envelopes and $\mathbb{R}$ -filtrations, respectively).

However, the generalization to adelic curves—a framework developed by Chen and Moriwaki to unify arithmetic geometry over number fields, function fields, and other bases with adelic structures—presents significant challenges. Specifically, the "Minkowskian property" (which facilitates the use of successive minima in the number field case) is absent in general adelic curves. Furthermore, measurability and integrability issues arise due to the complexity of the adelic structure (e.g., non-discrete $\sigma$ -algebras). The paper aims to resolve a conjecture by Chen and Moriwaki regarding whether a full arithmetic Fujita approximation holds in this general setting, specifically for $\mathbb{R}$ -line bundles, rather than just a "relative" version involving asymptotic slopes.

Methodology
The author employs a multi-step strategy combining the theory of $\mathbb{R}$ -filtrations, Harder-Narasimhan filtrations, and the construction of graded linear series over trivially valued fields.

Framework and Assumptions: The work operates under the assumption that the base field $K$ is perfect and that the adelic curve $S$ satisfies specific technical conditions: either the $\sigma$ -algebra $\mathcal{A}$ is discrete, or $K$ contains a countable sub-field dense in $K_w$ for all $w \in \Omega$ . These conditions ensure that subspace norms and quotient norms remain measurable and integrable, avoiding obstructions in the construction of adelic vector bundles.
Harder-Narasimhan Filtration and $\mathbb{R}$ -filtrations: The paper utilizes the Harder-Narasimhan filtration of adelic vector bundles, formulated via $\mathbb{R}$ -filtrations. This allows for the definition of asymptotic invariants such as the maximal and minimal slopes ( $\hat{\mu}^{\text{asy}}_{\max}$ and $\hat{\mu}^{\text{asy}}_{\min}$ ).
Transition to Trivially Valued Fields: A crucial technique involves constructing a graded linear series over a trivially valued field ( $S_0$ ) from the original series over $S$ . By applying the Harder-Narasimhan filtration to the spaces of global sections $H^0(X, L^{\otimes n})$ , the author defines a spectral norm. This transforms the original series into a "filtrated approximable graded linear series" over $S_0$ .
Approximation of Graded Linear Series: The author proves that for an arithmetically big line bundle, the associated graded linear series over $S_0$ is approximable. This involves showing that for any $\lambda < \hat{\mu}^{\text{asy}}_{\max}(L)$ , the sub-series defined by sections with spectral norms bounded by $e^{-D\lambda}$ contains an ample divisor. This step relies on a geometric Kodaira decomposition argument adapted to the adelic setting, avoiding the direct application of arithmetic Kodaira decomposition which fails in general adelic curves (as demonstrated by a counterexample in Example 4.4).
Extension to $\mathbb{Q}$ and $\mathbb{R}$ -line bundles:
- For $\mathbb{Q}$ -line bundles, the result follows directly from the integer case by scaling.
- For $\mathbb{R}$ -line bundles, the author uses a continuity argument. Since the arithmetic volume function is continuous, an arithmetically big $\mathbb{R}$ -line bundle can be approximated by an arithmetically big $\mathbb{Q}$ -line bundle such that the difference has a non-zero small section.

Key Contributions and Results
The primary result is Theorem 1.1 (Theorem 5.5), which establishes the arithmetic Fujita approximation for arithmetically big $\mathbb{R}$ -line bundles over proper adelic curves.

Theorem Statement: Let $S$ be a proper adelic curve with a perfect field $K$ satisfying the measurability conditions. Let $X$ be a projective variety and $L$ an arithmetically big $\mathbb{R}$ -line bundle. For any $\epsilon > 0$ , there exist a sufficiently large integer $p$ and a birational morphism $v: X' \to X$ such that:
$v^*(L^{\otimes p}) = A \otimes M$
where:
1. $A$ is an arithmetically ample line bundle.
2. $M$ is an adelic $\mathbb{R}$ -line bundle that admits a non-zero small section (a global section $s$ with $\|s\|_{\sup, w} \leq 1$ for all $w$ ).
3. The volumes satisfy the approximation inequality:
  $p^{-(\dim(X)+1)} \widehat{\text{vol}}(A) \geq \widehat{\text{vol}}(L) - \epsilon$
Generalization: This result generalizes previous work by Chen and Yuan (who treated the number field case) and provides a positive answer to the conjecture posed by Chen and Moriwaki in their 2021 work [11, Remark 9.2.7]. It moves beyond the "relative Fujita approximation" (which only controls the maximal slope) to a full decomposition where the "effective" part is controlled by the existence of small sections.

Significance
The paper claims to provide a definitive affirmative response to the conjecture of Chen and Moriwaki regarding the validity of Fujita's approximation in the general setting of adelic curves. Its significance lies in:

Unification: It extends a fundamental theorem of arithmetic geometry to a unified framework covering number fields, function fields, and other adelic bases, despite the loss of the Minkowskian property.
Technical Resolution: It overcomes specific obstructions related to measurability and integrability by imposing conditions on the adelic curve and utilizing the spectral norm construction over trivially valued fields.
Scope: It successfully handles $\mathbb{R}$ -line bundles, which are essential for modern arithmetic intersection theory, by establishing a bridge between $\mathbb{R}$ -bundles and $\mathbb{Q}$ -bundles via small sections.

The author notes that the result is a direct analogue of the geometric Fujita approximation, asserting that an arithmetically big bundle can be approximated by an ample bundle along an effective direction (represented by the bundle with a non-zero small section), thereby confirming the robustness of arithmetic positivity concepts in the adelic curve framework.