🔢 mathematics

$\mathrm{PGL}_n(\mathbb{C})$ -character stacks and Langlands duality over finite fields

This paper proposes and partially proves a conjectural formula for the mixed Poincaré polynomial of generic $\mathrm{PGL}_n(\mathbb{C})$ -character stacks, demonstrating that it interpolates the structure coefficients of specific class function bases for $\mathrm{PGL}_n(\mathbb{F}_q)$ and $\mathrm{SL}_n(\mathbb{F}_q)$ in a manner analogous to a non-abelian Fourier transform.

Original authors: Emmanuel Letellier, Tommaso Scognamiglio

Published 2026-05-28

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Emmanuel Letellier, Tommaso Scognamiglio

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 trying to understand the shape of a very complex, multi-dimensional landscape. In mathematics, this landscape is called a character stack. It's a place where you can arrange a set of "loops" (like rubber bands) on a surface (like a donut or a sphere with holes) such that when you travel around the holes, the loops twist in specific, pre-determined ways.

This paper, written by Emmanuel Letellier and Tommaso Scognamiglio, is like a detailed map and a set of instructions for measuring the "volume" and "texture" of these landscapes, specifically when the loops belong to a group of matrices called PGLn.

Here is a breakdown of their journey, using simple analogies:

1. The Landscape: Character Stacks

Think of a character stack as a giant playground where you build structures out of mathematical "blocks."

The Blocks: These are matrices (grids of numbers) that represent how things transform.
The Rules: You have a surface with holes (punctures). You must arrange your blocks so that if you walk around a hole, the block you pick up matches a specific "fingerprint" (a conjugacy class).
The Twist: Usually, mathematicians study these playgrounds when the blocks are "nice" and simple (connected). This paper looks at the messy, complicated cases where the blocks have "disconnected" stabilizers. Imagine a puzzle piece that can be flipped or rotated in ways that don't quite line up perfectly with its neighbors. This creates a landscape with multiple separate islands (components) rather than one smooth continent.

2. The Goal: Counting the "Atoms" of the Landscape

The authors want to calculate the Mixed Poincaré polynomial.

The Analogy: Imagine the landscape is a building made of different types of bricks. Some bricks are heavy (high "weight"), some are light. The Poincaré polynomial is a formula that counts how many bricks of each weight exist in the building.
The Challenge: For the messy, "disconnected" cases, no one knew the formula to count these bricks. The authors propose a conjecture (a very educated guess) for this formula.

3. The "Magic Mirror": Langlands Duality

The most exciting part of the paper is the connection to Langlands Duality.

The Analogy: Imagine you have two different languages describing the same object.
- Language A (PGLn): Describes the object using "loops" and "convolutions" (mixing things together like ingredients in a soup).
- Language B (SLn): Describes the same object using "point-by-point" multiplication (like stacking blocks directly on top of each other).
The Discovery: The authors show that the formula they guessed for the PGLn landscape acts as a translator between these two languages.
- If you take the "soup" of PGLn loops and mix them, the result tells you exactly how many ways you can stack the "blocks" in the SLn language.
- It's like realizing that the recipe for a cake in one country is mathematically identical to the recipe for a pie in another country, once you adjust the units of measurement.

4. The "Fourier Transform" Connection

The paper mentions a Fourier inversion formula.

The Analogy: In music, a Fourier transform takes a complex sound wave and breaks it down into simple notes (frequencies). The "inversion" takes the notes and rebuilds the sound wave.
The Paper's Claim: The authors show that their geometric formula is a "multiplicative version" of this. They are taking a complex geometric shape, breaking it down into its "notes" (character sheaves), and showing that multiplying those notes together perfectly reconstructs the original shape's properties.

5. What Did They Prove?

The Conjecture: They proposed a specific formula for the "brick count" (Mixed Poincaré polynomial) of these messy landscapes.
The Proof: They couldn't prove the entire formula in all its complexity yet, but they proved it works when you look at a specific "slice" of the data (called the Euler specialization). This is like proving a recipe works by tasting the final dish, even if you haven't written down every single step of the cooking process yet.
The Result: They demonstrated that this formula perfectly bridges the gap between the "soup" (PGLn convolution) and the "stacking" (SLn multiplication).

Summary

In simple terms, this paper is about finding a universal translator. The authors discovered a mathematical formula that connects two seemingly different ways of counting and organizing complex geometric shapes. They showed that the "messy" shapes formed by one group of matrices (PGLn) are secretly the same as the "clean" shapes formed by a related group (SLn), provided you look at them through the right lens.

They didn't just find a new shape; they found the dictionary that allows mathematicians to speak between two different branches of geometry and representation theory, proving that deep down, they are describing the same underlying reality.

Technical Summary: $PGL_n(\mathbb{C})$ -character stacks and Langlands duality over finite fields

Problem Statement
This paper investigates the mixed Poincaré polynomials of generic $PGL_n(\mathbb{C})$ -character stacks associated with a punctured Riemann surface. While the cohomology of $GL_n$ and $SL_n$ character stacks has been extensively studied (notably by Hausel, Letellier, Rodriguez-Villegas, Mellit, and Schiffman), the case of $PGL_n$ presents unique challenges. Specifically, $PGL_n$ conjugacy classes can possess non-connected stabilizers, leading to non-trivial irreducible equivariant local systems. The authors aim to compute the intersection cohomology of these stacks with coefficients in these local systems and to establish a connection between these geometric invariants and the representation theory of finite groups $PGL_n(\mathbb{F}_q)$ and $SL_n(\mathbb{F}_q)$ .

Methodology
The authors employ a combination of geometric representation theory, intersection cohomology, and the theory of character sheaves. The core methodology involves:

Geometric Setup: Defining the character stack $\mathcal{M}_C$ as the quotient of a representation variety $X_C$ by the $PGL_n$ action. The authors analyze the geometry of these stacks, particularly the stratification induced by the Zariski closure of conjugacy classes and the action of finite groups arising from the disconnected components of stabilizers.
Twisted Intersection Cohomology: To handle the disconnected stabilizers, the authors introduce a "twisted" setting involving the action of a finite group $H(C)$ on generic $GL_n$ -character stacks. They generalize Katz's results on counting points over finite fields to relate the Euler characteristic of the intersection cohomology of the stack to the trace of Frobenius on the cohomology of the corresponding variety over $\mathbb{F}_q$ .
Langlands Duality and Character Sheaves: The paper utilizes Lusztig's theory of character sheaves. It establishes a correspondence between local systems on $PGL_n$ -conjugacy classes and character sheaves on the dual group $SL_n$ . This involves analyzing the geometric induction functors and the decomposition of perverse sheaves under restriction to the dual group.
Combinatorial Formulas: The authors utilize symmetric functions, specifically Macdonald polynomials and hook functions ( $H_{g,\omega}$ ), to formulate conjectural expressions for the mixed Poincaré series. They derive explicit formulas for the "pure part" (specialization $t \to -1$ ) and the Euler specialization.

Key Contributions and Results

Conjectural Formula for Mixed Poincaré Series: The authors propose a conjectural combinatorial formula (Conjecture 5.5.12) for the mixed Poincaré series $IH_c(\mathcal{M}_C, \mathcal{E}_\chi; q, t)$ of the $PGL_n$ -character stack with coefficients in a local system $\mathcal{E}_\chi$ arising from a character $\chi$ of the stabilizer group $A(C)$ . The formula involves a sum over specific combinatorial types weighted by rational functions $H_{g, \omega_r}$ and coefficients $\Delta^s_\chi$ .
Proof under Euler Specialization: The main theorem (Theorem 1.1.2 and Theorem 5.5.11) proves that the conjectured formula holds true under the Euler specialization ( $t \to -1$ ). This result connects the geometry of the stack to the number of points over finite fields.
Interpolation of Structure Coefficients: A central result (Theorem 1.2.1 and Theorem 8.5.1) demonstrates that the Euler polynomial of the character stack interpolates the structure coefficients of two distinct based rings:
1. The ring of class functions on $PGL_n(\mathbb{F}_q)$ equipped with the convolution product, with a basis given by the characteristic functions of intersection cohomology complexes of local systems on conjugacy classes.
2. The ring of class functions on $SL_n(\mathbb{F}_q)$ equipped with pointwise multiplication, with a basis given by the characteristic functions of Lusztig's character sheaves on $SL_n$ .
Non-Abelian Fourier Transform Analogy: The relationship established between the convolution product on $PGL_n(\mathbb{F}_q)$ and the tensor product on $SL_n(\mathbb{F}_q)$ is described as a multiplicative version of the Fourier inversion formula. This generalizes the known $GL_n$ picture, where character sheaves correspond directly to irreducible characters, to the $SL_n/PGL_n$ duality where the correspondence is more intricate due to disconnected stabilizers.
Explicit Verification: The conjecture is verified explicitly for the case $n=2$ , $k=4$ (four punctures) with semisimple regular conjugacy classes (Theorem 1.1.3 and Theorem 9.5.2).

Significance
The paper claims to provide a unified geometric framework that bridges the cohomology of character stacks with the representation theory of finite reductive groups in the context of Langlands duality. By extending the results from $GL_n$ to $PGL_n$ , the authors address the complexity introduced by non-connected stabilizers. The work suggests that the mixed Hodge structure of these moduli spaces encodes deep arithmetic information regarding the structure constants of character rings for finite groups. The authors note that while the full conjecture for the mixed Poincaré polynomial remains open, the proven Euler specialization and the explicit $n=2$ case provide strong evidence for the validity of the proposed formulas and the underlying duality picture. The results are presented as a generalization of previous work by Hausel, Letellier, and Rodriguez-Villegas, specifically adapting their methods to the non-generic and non-connected stabilizer setting.