🔢 mathematics

On Moy-Prasad quotients over Laurent series fields

This paper defines and calculates a stratification by conjugacy classes of twisted Levi subgroups on Moy-Prasad quotients of a connected reductive group over an algebraically closed Laurent series field, providing a foundational result for the local geometric Langlands program.

Original authors: David Yang

Published 2026-03-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: David Yang

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

The Big Picture: Mapping the Unseen City

Imagine you are trying to understand a massive, invisible city called $G((t))$ . This isn't a city of buildings, but a "city" of mathematical symmetries (a group) that exists over a field of infinite series (like $k((t))$ ).

Mathematicians are trying to solve the Local Geometric Langlands Program, which is essentially a grand project to classify all the possible "languages" or "stories" (representations) that can be told within this city.

The main discovery in recent years is that you don't need to understand the whole chaotic city at once. Instead, every story in this city is built by stitching together smaller, simpler stories coming from specific neighborhoods called Twisted Levi Subgroups. Think of these as the "foundational districts" of the city.

The Problem: We know these districts exist, but we don't have a good map telling us exactly which district a specific story comes from. If you hand me a random "story" (a mathematical object), how do I know which foundational district it belongs to?

The Solution: This paper builds that map. It creates a system to sort every possible object in the city into the correct district.

The Tools: The "Moy-Prasad" Filter

To navigate this city, the author uses a tool called the Moy-Prasad filtration.

The Analogy: The Onion Layers
Imagine the city is an onion.

The center is the core.
As you peel back layers, you get further out.
Each layer represents a level of "depth" or complexity.

Mathematicians look at the space between two specific layers (called $k_{x,r}/k_{x,r+}$ ). This space is like a slice of the onion. It contains a mix of different types of "stuff" (mathematical vectors).

The goal of the paper is to look at this slice and ask: "If I pick a random piece of stuff from this slice, which foundational district (Twisted Levi subgroup) does it belong to?"

The Core Mechanism: The "DNA Test"

The paper introduces a method to identify the "DNA" of any object in this slice.

The Scanner ( $q_{x,r}$ ): The author defines a special map (a scanner) that takes an object from the onion slice and projects it onto a simpler, abstract landscape called $C$ (the GIT quotient).
- Analogy: Imagine you have a complex, 3D sculpture. You shine a light on it to cast a 2D shadow. The shape of the shadow tells you something about the sculpture's core structure.
The Shadow's Shape: In this abstract landscape $C$ , every point corresponds to a specific type of symmetry.
- If the shadow falls in a specific region, it means the object belongs to a specific "district" (a Twisted Levi subgroup).
- If the shadow is at the very center (zero), the object is "unstable" (it's too chaotic to belong to a clean district).
The Stratification (The Map): The paper proves that you can divide the entire onion slice into distinct zones (strata).
- Zone A: Objects here belong to District A.
- Zone B: Objects here belong to District B.
- Zone C: Objects here belong to District C.

The most important result (Theorem 3.19) says: This map is perfect. It is a one-to-one correspondence. If you know which zone an object is in, you know exactly which district it comes from, and vice versa.

The "Twisted" Part

Why are they called Twisted Levi subgroups?

The Analogy: The Rotating Mirror
In a standard city, the districts are straight and aligned. But in this mathematical city, the rules are "twisted" by a Galois action (a kind of rotation or reflection of the field).

Imagine looking at a building in a funhouse mirror. It looks like a standard building, but it's slightly skewed.
A "Twisted Levi" is a subgroup that looks like a standard district only if you step outside the city and look at it from a different angle (an algebraic closure). Inside the city, it looks twisted.

The paper handles these twisted districts carefully, ensuring the map works even when the geometry is skewed.

Why Does This Matter?

The author mentions this is an "input" for future papers.

The Analogy: The Recipe Book
Imagine you are a chef trying to recreate a complex dish (the representation of the whole group).

You know the dish is made of ingredients from specific farms (the Twisted Levi subgroups).
But you don't know which farm provided the specific spice you are holding.
This paper is the labeling machine. It takes the spice, scans it, and stamps it with the name of the farm it came from.

Once you have this labeling machine, you can systematically reconstruct the entire dish by gathering the right ingredients from the right farms. This is crucial for the Geometric Langlands Program, which aims to unify two different areas of mathematics (number theory and geometry) by understanding these "recipes."

Summary in One Sentence

David Yang has created a precise mathematical "sorting machine" that takes complex, layered mathematical objects and instantly identifies which foundational symmetry group they belong to, providing the essential map needed to solve a major unsolved problem in modern mathematics.

1. Problem Statement and Context

The paper addresses a foundational problem in the local geometric Langlands program. Specifically, it seeks to classify categorical representations of the loop group $G((t))$ , where $G$ is a connected reductive group over the field of Laurent series $k((t))$ (with $k$ algebraically closed).

The Challenge: Previous work (e.g., by Moy and Prasad) established that irreducible representations of $G((t))$ over local fields are classified by "unrefined minimal K-types." These types consist of a point $x$ in the Bruhat-Tits building and a character of the vector space quotient $K_{x,r}/K_{x,r+}$ (or its Lie algebra analogue $\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+}$ ).
The Gap: In the geometric setting, representations are not necessarily irreducible in the classical sense. The author's follow-up papers aim to show that all categorical representations of $G((t))$ are built via an induction process from twisted Levi subgroups (subgroups that become Levi subgroups over the algebraic closure).
The Specific Goal: To define this induction process, one must relate the "unrefined minimal K-types" of $G$ to those of its twisted Levi subgroups. This requires a precise geometric understanding of the stratification of the vector space $\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+}$ by the conjugacy classes of twisted Levi subgroups that "generate" the orbits within it.

2. Methodology

The paper employs a synthesis of Bruhat-Tits theory, Moy-Prasad filtration, and Geometric Invariant Theory (GIT).

Non-Split Generalization: Unlike previous work restricted to split groups, this paper handles quasi-split (and potentially non-split) groups. It utilizes the description of $G((t))$ as a twisted loop group via a Galois 1-cocycle acting on a split form $H$ . This allows the use of a "twisted" Lie algebra structure.
Moy-Prasad Filtration: The author analyzes the graded pieces $\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+}$ of the Lie algebra filtration associated with a point $x$ in the Bruhat-Tits building. These spaces carry an action of the reductive quotient group $L_x = P_x/P_x^+$ .
GIT and Semistability: The core analysis focuses on the semistable locus of the $L_x$ -action on $\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+}$ . The author utilizes the Kostant section and the Chevalley restriction theorem to map these spaces to a GIT quotient $C((t))$ , which parametrizes conjugacy classes of semisimple elements.
Stratification Strategy: The paper constructs a stratification of the vector space based on the "centralizer type" of the semisimple elements. It relates the orbits in the $G$ -setting to orbits in the $G_0$ -setting (where $G_0$ is a twisted Levi) by analyzing the map $q_{x,r}$ and its interaction with the building of $G_0$ .

3. Key Contributions and Results

A. The Stratification by Twisted Levi Subgroups

The central result is Theorem 3.19, which establishes a precise geometric correspondence between the semistable orbits of $G$ and those of its twisted Levi subgroups.

Definition of the Stratification: The author defines loci $(\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+})_{\geq G_0}$ and $(\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+})_{= G_0}$ based on the conjugacy class of the centralizer of the semisimple part of an element.
The Main Isomorphism: The theorem states that the locus of elements in $\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+}$ whose semisimple part has a centralizer conjugate to a specific twisted Levi $G_0$ is isomorphic (as an Artin stack) to a quotient of the corresponding locus in the twisted Levi's own Moy-Prasad space.
$\bigsqcup_{s \in S} (G_x/K_{x,0+}) \times_{(N^x_s/K^{G_s}_{x,0+})} (\mathfrak{k}^{G_s}_{x,r}/\mathfrak{k}^{G_s}_{x,r+})_{=G_s, \text{gen}} \xrightarrow{\sim} (\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+})_{=G_0}$
Here, $S$ represents the set of $G_x$ -conjugacy classes of subgroups $G_s$ conjugate to $G_0$ such that $x$ lies in the building of $G_s$ .

B. Technical Lemmas and Tools

Lemma 3.2 & Theorem 3.1: Characterizes the unstable locus as the set of elements that can be "pushed" into a deeper Moy-Prasad filtration level $K_{y,r+}$ for some $y$ in the building.
The Map $q_{x,r}$ : Constructs a canonical map from the Moy-Prasad quotient to the graded component of the GIT quotient $C((t))$ . This map is shown to be compatible with the filtration and the Galois twist.
Bijectivity on Semisimple Elements (Section 4.1): Proves that every semisimple element in the quotient lifts uniquely (up to conjugation) to a semisimple element in the Lie algebra, and that the centralizer of this lift determines the stratum.
Bijectivity on All Elements (Section 4.2): Extends the result from semisimple elements to all elements using the Jordan decomposition, showing that the centralizer of the semisimple part controls the entire element's position in the stratification.

4. Significance

Foundation for Local Geometric Langlands: This paper provides the necessary "local" geometric input for the global classification of categorical representations. It rigorously defines how representations of $G((t))$ are induced from twisted Levi subgroups, generalizing the classical theory of minimal K-types to the geometric, non-split setting.
Handling Non-Split Groups: By explicitly working with twisted loop algebras and Galois twists, the paper removes the restriction to split groups, which is crucial for the full scope of the Langlands program.
Geometric Structure of Moy-Prasad Quotients: The paper reveals that the seemingly complex vector spaces $\mathfrak{k}_{x,r}/\mathfrak{k}_{x,r+}$ possess a highly structured stratification governed by the combinatorics of the Bruhat-Tits building and the conjugacy classes of Levi subgroups.
Future Applications: The results are explicitly designed as a prerequisite for subsequent papers (joint with Dhillon and Varshavsky) that will construct the induction functors and prove the classification of categorical representations.

In summary, David Yang's paper solves a critical structural problem in the representation theory of loop groups over Laurent series fields, providing the geometric machinery required to decompose complex categorical representations into manageable pieces induced from twisted Levi subgroups.