🔢 mathematics

Courbes et fibrés vectoriels en théorie de Hodge $z$ -adique globale

This paper establishes foundational results on the moduli of $G$ -bundles over the global analogue of the Fargues-Fontaine curve, using it to reformulate the global Langlands conjecture in terms of categorical local Langlands and verifying this framework for commutative groups, while also proving a GAGA theorem for smooth proper schemes over sousperfectoid spaces.

Original authors: Siyan Daniel Li-Huerta

Published 2026-02-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Siyan Daniel Li-Huerta

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 the universe of mathematics as a vast, intricate city. For a long time, mathematicians have been studying the "local" neighborhoods of this city—tiny, self-contained districts where the rules are well-understood. One of the most famous neighborhoods is the Fargues–Fontaine curve. Think of this as a magical, infinitely detailed map of a single local district that has revolutionized how we understand the relationship between two major languages of math: Geometry (shapes and spaces) and Number Theory (equations and primes).

This paper, by Siyan Daniel Li-Huerta, asks a bold question: What happens if we try to build a "Global" version of this magical map? Instead of looking at one tiny neighborhood, can we create a map for the entire country (a "global function field")?

Here is a breakdown of the paper's journey, using everyday analogies:

1. The Problem: Trying to Stitch the World Together

The author tries to construct this "Global Hartl–Pink curve" by taking a global field (like the entire country) and trying to glue it to a special mathematical space.

The Obstacle: It's like trying to stitch a giant, complex tapestry (the global field) to a piece of glass (the local space) without the tapestry tearing or the glass shattering. The math gets messy because the "glue" (a mathematical operation called Frobenius) doesn't behave nicely when applied to the whole country at once.
The Workaround: Instead of building the map itself (which is too hard), the author decides to build the catalog of all possible shapes that can exist on this map. In math, these shapes are called Vector Bundles or G-bundles. Think of these as different types of "wrapping paper" or "fabric" that can be draped over the invisible map.

2. The Solution: The "Igusa" Catalog

The author successfully builds a catalog called $Bun_{G,F}$ .

What is it? Imagine a massive library where every book represents a different way to wrap a shape around the global field.
The "Igusa" Connection: The paper shows this library acts like a "master key" or a central hub (called an Igusa stack). Just as a master key can open many different doors, this catalog organizes the geometry of the entire field.
The "Local" vs. "Global" Link: The author proves you can look at a specific page in this library (a local view) and instantly know how it relates to the whole book (the global view). This is done through a "localization map," which is like a translator that converts a local dialect into the global language.

3. The Big Discovery: The "Shtuka" Bridge

The paper introduces a concept called Shtukas.

The Analogy: Imagine a Shtuka as a "traveling merchant" who carries goods (mathematical data) between different cities. The merchant has a rule: they must change their cargo slightly when they cross a border.
The Fiber Product Conjecture: The author proves a famous guess (the "Fiber Product Conjecture") that says: The entire global catalog ( $Bun_{G,F}$ ) is exactly what you get if you take all the local merchants, look at their cargo, and see how they fit together.
Why it matters: This confirms that the "Global" catalog isn't just a random collection; it is perfectly structured by the "Local" pieces. It's like proving that a giant mosaic is made of perfect, interlocking tiles.

4. The Langlands Connection: The Ultimate Dictionary

The paper's ultimate goal is to update the Langlands Conjecture.

The Analogy: The Langlands Conjecture is often called the "Grand Unified Theory" of math. It proposes a secret dictionary that translates between two completely different languages:
1. The Automorphic Language: Describes patterns in shapes and spaces (like the "wrapping paper" bundles).
2. The Galois Language: Describes the hidden symmetries of numbers (like the "roots" of equations).
The Paper's Contribution: The author uses the new "Global Catalog" ( $Bun_{G,F}$ ) to rewrite this dictionary. They propose a new way to translate the entire global field, refining previous attempts by other giants in the field.
The "Categorical" Twist: Instead of just translating single words, this new dictionary translates entire sentences and stories (categories of objects). It suggests that the "Global Catalog" is the perfect stage where these two languages meet and converse.

5. The Proof: When the Math is Simple

To prove their new dictionary works, the author tests it on the simplest possible case: Commutative Groups (think of a simple, straight line rather than a complex, twisting knot).

The Result: They prove that for these simple cases, the new dictionary is 100% accurate. The "Global Catalog" perfectly matches the "Galois Symmetries." This gives strong evidence that the theory works for the complex cases too, even though the full proof for the complex cases is still a work in progress.

6. A Side Quest: The "GAGA" Theorem

Along the way, the author proves a side result called a GAGA theorem.

The Analogy: Imagine you have a sculpture made of clay (algebraic geometry) and you want to know if it looks the same when you paint it with watercolors (analytic geometry).
The Finding: The author proves that for these specific types of "sculptures" (smooth, proper schemes), the clay version and the watercolor version are identical. You can switch between them without losing any information. This is a powerful tool that helps them build the main catalog.

Summary

In simple terms, this paper builds a universal catalog for a complex mathematical landscape. It proves that this catalog is perfectly organized by local pieces, uses it to rewrite the "Grand Dictionary" (Langlands Conjecture) that connects shapes and numbers, and verifies that this new dictionary works perfectly for simple cases. It's a foundational step toward understanding the deep, hidden architecture of the mathematical universe.

Technical Summary: Global Hartl–Pink Curves and the Langlands Correspondence over Function Fields

Problem Statement
This paper investigates the global analogue of the Fargues–Fontaine curve (also known as the Hartl–Pink curve in the function field setting) over a global function field $F$ . While the local theory, developed by Fargues and Fontaine, provides a geometric framework for the local Langlands correspondence via the moduli stack of $G$ -bundles on the curve $X_{F_v, K}$ , a direct global analogue faces significant obstructions. Specifically, naively attempting to form the quotient $(\text{Spec } F \times \text{Spa } K)/\text{Frob}_K$ fails because the product $\text{Spec } F \times \text{Spa } K$ lacks a well-behaved adic space interpretation, and the Frobenius action is not free or totally discontinuous. The paper aims to define and study the moduli stack of $G$ -bundles on this "global Hartl–Pink curve," denoted $\text{Bun}_{G,F}$ , and utilize it to reformulate the global Langlands conjecture for $G$ over $F$ in terms of categorical local Langlands.

Methodology
The author employs a descent-theoretic approach to circumvent the difficulties in defining the global curve itself. Instead of constructing the curve directly, the paper defines the moduli stack $\text{Bun}_{G,F}$ as a colimit of moduli stacks $\text{Bun}_{G,U}$ over dense open subschemes $U \subset C$ (where $C$ is the smooth proper curve associated with $F$ ).

Analytification and Descent: The core technical tool is the study of vector bundles on the analytification $U^{\text{an}}_S$ of $U$ over an affinoid perfectoid space $S$ . The paper establishes $v$ -descent for vector bundles on these analytifications and proves a GAGA theorem for smooth proper schemes over sousperfectoid spaces (Appendix A), showing that algebraic vector bundles are equivalent to their analytifications.
Localization and Shtukas: A central mechanism is the localization map $\text{loc}_Z: \text{Bun}_{G,U} \to \prod_{z \in Z} \text{Bun}_{G,F_z}$ , where $Z = C \setminus U$ . The paper relates the global moduli to local moduli stacks and the moduli of global shtukas.
Fiber Product Conjecture: The paper proves a function field analogue of Scholze's fiber product conjecture, establishing a cartesian square relating the moduli of global shtukas, local shtukas, and the moduli of $G$ -bundles. This allows the recovery of global shtuka cohomology from local data.
Motivic Sheaves: To handle the derived category of sheaves, the paper utilizes the theory of motivic sheaves (as opposed to classical $\ell$ -adic sheaves) to facilitate comparisons between different sheaf theories and to define Hecke operators in a derived setting.

Key Contributions and Results

Definition and Geometry of $\text{Bun}_{G,F}$ :
- Theorem A: The stack $\text{Bun}_{G,F}$ is a small Artin $v$ -stack over $\mathbb{F}_q$ . It is $\ell$ -cohomologically smooth, and its dualizing complex with $\mathbb{F}_\ell$ -coefficients is isomorphic to $\mathbb{F}_\ell$ . This mirrors foundational results for the local Fargues–Fontaine curve.
- Theorem B: The global semistable locus $\text{Bun}^{ss}_{G,F}$ is an open substack isomorphic to a disjoint union of classifying stacks $\bigsqcup_{b \in B(F,G)_{\text{basic}}} * / G_b(F)$ , where $B(F,G)$ is the global Kottwitz set. This provides a geometrization of the global Kottwitz set, though the paper notes that a natural bijection between the underlying topological space $|\text{Bun}_{G,F}|$ and $B(F,G)$ does not generally exist due to the positive dimension of fibers over non-basic elements (analogous to Igusa varieties).
- Theorem C: The set of isomorphism classes of $\mathbb{F}_q$ -points of $\text{Bun}_{G,F}$ is naturally in bijection with the global Kottwitz set $B(F,G)$ .
Relation to Shtukas and Igusa Stacks:
- Theorem D: The paper proves a fiber product theorem relating the moduli of global shtukas (bounded by representations of the dual group) to the product of local shtuka moduli and the moduli of $G$ -bundles. Specifically, the moduli of global shtukas encodes the function field analogue of Igusa stacks. This result generalizes previous work on disjoint legs to the case of colliding legs.
Langlands Correspondence Reformulation:
- Theorem E: The cohomology of the moduli of global shtukas at infinite level (with respect to a set of places $Z$ ) can be recovered by applying Hecke operators to the object $\text{loc}_{Z,!}\Lambda$ in the derived category of the product of local moduli stacks.
- Conjecture F: The paper proposes a global Langlands conjecture stating that, under the assumption of the Fargues–Scholze categorical local Langlands conjecture, the object $\text{loc}_{Z,!}\Lambda$ corresponds to the pushforward of the dualizing complex $\omega_{LS_{\hat{G},U}}$ along the restriction map of Galois representations. This refines conjectures by Arinkin–Gaitsgory–Kazhdan–Raskin–Rozenblyum–Varshavsky and Zhu.
- Theorem G: The conjecture is verified in the case where $G$ is a torus (commutative), utilizing Langlands' work on tori and explicit descriptions of the moduli of Galois representations.

Significance and Claims
The paper claims to provide a foundational geometric framework for the global Langlands program over function fields that parallels the success of the local theory. By defining $\text{Bun}_{G,F}$ and proving its geometric properties, the author establishes that this stack plays the role of Igusa stacks over function fields. The primary significance lies in the reformulation of the global Langlands conjecture: rather than seeking a direct correspondence between automorphic forms and Galois representations, the paper suggests that the global conjecture should be viewed as an identification of the "boundary theories" (local data) via the global moduli stack.

The paper explicitly states that it verifies this conjecture for commutative groups and proves a generalization of the Langlands–Rapoport conjecture for moduli spaces of global shtukas with arbitrary legs. It also highlights the independent interest of the GAGA theorem for smooth proper schemes over sousperfectoid spaces, which was necessary to control the geometry of the analytifications used throughout the proofs. The work is presented as a step towards a unified categorical understanding of the Langlands correspondence, bridging the gap between the geometric structures of shtukas and the spectral side of the correspondence.