Affine Deligne-Lusztig Varieties of Positive Coxeter Type

Imagine you are trying to navigate a massive, multi-dimensional city called The Group. This city is built on a grid that stretches infinitely in all directions (the "Affine" part). In this city, there are two different ways to organize the buildings and streets:

The "Shape" Map: This groups buildings based on their fundamental geometric shape and how they twist under a special wind called the Frobenius (a mathematical force that shuffles things around).
The "Address" Map: This groups buildings based on their specific coordinates and how they relate to a central hub (the "Iwahori" subgroup).

Affine Deligne-Lusztig Varieties are the specific neighborhoods where these two maps overlap. Mathematicians have been trying to understand the layout of these neighborhoods for decades. The question is: If I give you a specific coordinate (an element $x$ ) and a specific shape (an element $b$ ), what does the neighborhood look like? Is it empty? Is it a single point? Is it a complex maze? How big is it?

For a long time, we only had a good map for a very specific, small district of this city. A team of mathematicians (He, Nie, and Yu) figured out that if the coordinate $x$ had a special "finite Coxeter" property, the neighborhood was surprisingly simple: it looked like a stack of familiar, classical shapes (like spheres or tori) with some extra "tubes" attached.

The New Discovery: "Positive Coxeter Type"

This paper, by Felix Schremmer, Ryosuke Shimada, and Qingchao Yu, introduces a much broader category of coordinates called "Positive Coxeter Type."

Think of the old "Finite Coxeter" rule as a strict dress code: "You must wear a suit to enter the VIP lounge." This paper says, "Actually, you can enter if you wear a suit, a tuxedo, or even a very specific, well-tailored t-shirt." The "Positive Coxeter Type" is a more flexible, yet still very structured, rule that covers many more neighborhoods in the city.

The Main Results (The "What")

The authors prove three amazing things about these new neighborhoods:

The "No Surprises" Rule (Non-emptiness): They found a simple checklist to see if a neighborhood exists. If your coordinate passes the check, the neighborhood is there. If not, it's empty. It's like having a perfect GPS that tells you exactly which streets are paved and which are dead ends.
The "Perfect Size" Formula: They gave a precise formula to calculate the dimension (the number of directions you can move) of these neighborhoods. It's like knowing exactly how many rooms a house has just by looking at its address.
The "Lego" Structure (Geometry): This is the coolest part. They proved that these complex neighborhoods are actually built like Lego towers.
- The base is a classic, well-understood shape (a "Classical Deligne-Lusztig variety").
- Stacked on top of this base are simple "tubes" (mathematical lines or planes, like $A^1$ or $G_m$ ).
- Crucially, they proved that for the most basic cases, these tubes are just straight, un-twisted pipes. The whole thing is a simple product: Base $\times$ Tubes. It's not a twisted knot; it's a clean, straight stack.

The "Reverse" Discovery (The "Converse")

The paper also asks the reverse question: If a neighborhood looks like this simple Lego tower, does that mean the coordinate must be of "Positive Coxeter Type"?

The answer is yes. They proved that this simple, clean geometric structure is a unique fingerprint. If you see a neighborhood that is a clean stack of tubes over a base, you know immediately that the coordinate generating it belongs to this special class. This allows mathematicians to identify these special coordinates just by looking at the shape of the resulting variety.

Why Does This Matter? (The "So What?")

You might ask, "Who cares about these infinite cities?"

These varieties are the secret keys to understanding Shimura Varieties, which are objects at the heart of modern number theory and the Langlands Program (a grand unification theory of mathematics).

The Old Way: Previously, we could only describe the "Basic Loci" (the most fundamental parts of these Shimura varieties) for very specific, restrictive cases. It was like only being able to map the downtown area of a country.
The New Way: With "Positive Coxeter Type," we can now map vast new territories. The authors show that many important cases (like those involving $GL_n$ or $GSp_6$ ) that were previously too messy to understand are actually just "Positive Coxeter Type."

The Analogy of the "Critical Strip"

The authors mention that the old definition had "pathologies" (glitches) in certain areas of the city called "critical strips." Imagine a map that works perfectly in the city center but suddenly says "Road Closed" or "Infinite Loop" when you get to the suburbs, even though the road is fine.

The new "Positive Coxeter Type" definition fixes these glitches. It is stable and consistent, even when you twist the city or look at it from different angles. It's a more robust, reliable map for the entire mathematical landscape.

Summary

In short, this paper:

Expands the Map: It defines a new, larger class of mathematical objects ("Positive Coxeter Type") that behave nicely.
Solves the Geometry: It proves these objects are built from simple, clean Lego blocks (a base shape plus straight tubes).
Fixes the Glitches: It removes the inconsistencies of previous definitions.
Opens New Doors: It allows mathematicians to solve problems about Shimura varieties (and by extension, deep number theory problems) that were previously impossible to tackle.

It's a tour de force that turns a confusing, tangled mess of infinite possibilities into a clean, orderly, and predictable structure.

1. Problem Statement and Background

Context:
Affine Deligne–Lusztig varieties (ADLVs), denoted $X_x(b)$ , are geometric objects associated with a reductive group $G$ over a local field $F$ . They arise in the study of the reduction of Shimura varieties and moduli spaces of local $G$ -shtukas. An ADLV is defined as the reduced subscheme of the affine flag variety $G(\breve{F})/I$ (where $I$ is an Iwahori subgroup and $\breve{F}$ is the completion of the maximal unramified extension of $F$ ) whose geometric points are:
$X_x(b) = \{ gI \in G(\breve{F})/I \mid g^{-1}b\sigma(g) \in I\dot{x}I \}$
where $x$ is an element of the extended affine Weyl group $\tilde{W}$ , $b \in G(\breve{F})$ , and $\sigma$ is the Frobenius automorphism.

The Challenge:
The geometry of $X_x(b)$ is notoriously difficult to understand in general. Key open problems include:

Determining for which pairs $(x, [b])$ the variety is non-empty.
Calculating the dimension of $X_x(b)$ .
Describing the structure of its irreducible components and the action of the $\sigma$ -centralizer $J_b(F)$ .

Previous Work:
He, Nie, and Yu (HNY) recently made significant progress for a specific class of elements called those with "finite Coxeter part." In this case, the finite part of $x$ (under a specific decomposition) is a partial $\sigma$ -Coxeter element. They proved that for these elements, $X_x(b)$ has a simple structure: it is equi-dimensional, and its irreducible components are iterated fiber bundles over classical Deligne–Lusztig varieties of Coxeter type.

Limitation:
The "finite Coxeter part" condition is too restrictive. It is not stable under Deligne–Lusztig reduction (a key tool for inductive proofs) and fails to capture many cases relevant to Shimura varieties (e.g., certain Ekedahl-Oort strata).

Goal:
The authors aim to generalize the HNY results to a broader class of elements, termed "positive Coxeter type," which is stable under reduction and covers cases where the finite Coxeter part notion fails.

2. Methodology and Key Definitions

The paper relies on a combination of combinatorial group theory (affine Weyl groups), geometric reduction techniques, and the theory of Newton points.

A. Length Positive Elements

The authors utilize the concept of length positive elements, introduced by the first author in previous work. For an element $x = w\epsilon^\mu \in \tilde{W}$ , a subset $LP(x) \subseteq W$ is defined. An element $v \in W$ is in $LP(x)$ if the length functional $\ell(x, v\alpha) \geq 0$ for all positive roots $\alpha$ .

This set $LP(x)$ is non-empty and contains the minimal length element $v$ such that $v^{-1}\mu$ is dominant.
Crucially, $LP(x)$ behaves well under Deligne–Lusztig reduction, unlike the specific "minimal dominant" element used in the HNY framework.

B. Positive Coxeter Type

Definition: An element $x = w\epsilon^\mu \in \tilde{W}$ is of positive Coxeter type if there exists a $v \in LP(x)$ and a $\sigma$ -stable subset $J \subseteq S$ (simple reflections) such that $v^{-1}\sigma(wv)$ is a partial $\sigma$ -Coxeter element in the parabolic subgroup $W_J$ .

A pair $(x, v)$ satisfying this is called a positive Coxeter pair.
This definition generalizes the "finite Coxeter part" condition. While HNY required $v$ to be the specific minimal element making $v^{-1}\mu$ dominant, the new definition allows any $v \in LP(x)$ , provided the resulting conjugate is a partial $\sigma$ -Coxeter element.

C. Deligne–Lusztig Reduction

The authors employ the Deligne–Lusztig reduction method. This involves constructing a "reduction tree" for $x$ by applying simple affine reflections $s_a$ .

Type I Reduction: Preserves length ( $\ell(s_a x \sigma(s_a)) = \ell(x)$ ).
Type II Reduction: Decreases length by 2 ( $\ell(s_a x \sigma(s_a)) = \ell(x) - 2$ ).
The geometry of $X_x(b)$ is related to the geometry of the endpoints of these reduction paths. The core technical innovation is proving that the property of being "positive Coxeter type" is preserved under these reductions (Lemmas 5.1 and 5.2).

3. Key Contributions and Results

Theorem A: Structure of ADLVs for Positive Coxeter Type

This is the main theorem (Theorem 5.7), generalizing the HNY results. Let $x$ be of positive Coxeter type with pair $(x, v)$ and support $J$ .

Non-emptiness and Saturation: The set of $\sigma$ -conjugacy classes $[b]$ for which $X_x(b) \neq \emptyset$ is a saturated interval $[b_{x, \min}, b_{x, \max}]$ . The minimal and maximal elements are explicitly described using the Newton points and the weight function of the quantum Bruhat graph.
Dimension Formula: For any $[b]$ in this interval, $X_x(b)$ is equi-dimensional with dimension:
$\dim X_x(b) = \frac{1}{2} \left( \ell(x) + \ell(v^{-1}\sigma(wv)) - \langle \nu(b), 2\rho \rangle - \text{def}(b) \right)$
where $\nu(b)$ is the Newton point and $\text{def}(b)$ is the defect.
Geometric Structure: The group $J_b(F)$ $J_{b} (F)$ acts transitively on the irreducible components of $X_x(b)$ $X_{x} (b)$ . Each irreducible component is an iterated fiber bundle (with fibers isomorphic to $\mathbb{A}^1$ $A^{1}$ or $\mathbb{G}_m$ $G_{m}$ ) over an irreducible component of a classical Deligne–Lusztig variety of Coxeter type.
- Note: In the "basic" case (where $b$ is basic), the bundle is trivial, meaning the component is a direct product of a classical DL variety and affine space.

Theorem B: Characterization (Converse)

Proposition 5.19: The authors provide a converse to Theorem A. If the dimension of $X_x(b)$ for the minimal $\sigma$ -conjugacy class matches the expected formula (involving a specific $v \in LP(x)$ ) and the finite part $w$ is $\sigma$ -conjugate to a partial $\sigma$ -Coxeter element, then $x$ is of positive Coxeter type. This characterizes the class via geometric properties.

Theorem 5.12: Minimal Length Characterization

For elements $x$ of minimal length in their $\sigma$ -conjugacy class, $x$ is of positive Coxeter type if and only if its finite part $w$ is $\sigma$ -conjugate to a partial $\sigma$ -Coxeter element. This bridges the gap between the combinatorial definition and the geometric reality for minimal elements.

Applications to Shimura Varieties

The paper demonstrates that the "positive Coxeter type" condition is necessary to describe the basic loci of Shimura varieties in cases where the "finite Coxeter part" condition fails.

Examples: The authors cite cases involving $GL_n$ and $GSp_6$ where Ekedahl-Oort strata are of positive Coxeter type but not finite Coxeter type.
Depth: The notion is crucial for studying Shimura data of depth $< 2$ (introduced by Viehmann and the first author), where previous methods were insufficient.

4. Technical Innovations

Stability under Reduction: The most significant technical hurdle was showing that the "finite Coxeter part" property is not stable under Deligne–Lusztig reduction, whereas the "positive Coxeter type" property (using the flexible set $LP(x)$) is stable. This allows for a clean inductive proof on the length $\ell(x)$ .
Quantum Bruhat Graph: The authors use the weight function of the Quantum Bruhat Graph to explicitly compute the generic Newton point (maximal element) for positive Coxeter type elements, refining previous descriptions.
Very Special Parahoric Subgroups: The geometric description of the components involves reducing to a "very special parahoric" subgroup, linking the affine geometry to classical Deligne–Lusztig varieties in a reductive quotient.

5. Significance

Generalization: The paper vastly extends the scope of known geometric structures for ADLVs. It moves from a narrow class of elements (finite Coxeter part) to a robust, stable class (positive Coxeter type) that covers almost all cases of interest in current Shimura variety research.
Unified Framework: It provides a unified dimension formula and structural description that works for both the "shrunken Weyl chamber" cases and more complex boundary cases.
Shimura Varieties: The results are immediately applicable to the study of the basic locus of Shimura varieties, particularly for groups like $GL_n$ and $GSp_{2n}$ , allowing for the description of strata that were previously inaccessible.
Conjectural Upper Bounds: The work supports and refines conjectures regarding the dimension of ADLVs, suggesting that the "virtual dimension" formula holds more broadly when using the correct set of length-positive elements.

In summary, Schremmer, Shimada, and Yu have successfully generalized the theory of affine Deligne–Lusztig varieties of Coxeter type, providing a powerful new tool for arithmetic geometry and the Langlands program.