Algorithms on the Pyasetskii involution on local Langlands parameters of classical groups

This paper presents an algorithm for computing the Pyasetskii involution on local Langlands parameters for the classical groups $\mathrm{Sp}_{2n}$, $\mathrm{SO}_{2n+1}$, and $\mathrm{O}_{2n}$ by combining Moeglin-Waldspurger's method for $\mathrm{GL}_n$ with Lanard-Mínguez's approach for Aubert-Zelevinsky involutions, while also providing a geometric interpretation for the bad parity case.

The Gist

Imagine you are a master archivist working in a massive, ancient library. This library doesn't hold books about history or fiction; it holds the "blueprints" for every possible shape and symmetry in a specific mathematical universe. These blueprints are called L-parameters.

The authors of this paper, Alexander Hazeltine and Chi-Heng Lo, have built a new instruction manual (an algorithm) to solve a specific puzzle in this library: how to find the "mirror image" of any given blueprint.

Here is the breakdown of their work using simple analogies:

1. The Library and the Blueprints

Think of the Local Langlands Correspondence as a giant translation dictionary. It links two different worlds:

• World A: The world of abstract symmetries (represented by the blueprints/L-parameters).
• World B: The world of actual functions and waves (represented by mathematical representations).

The paper focuses on a specific type of symmetry group called Classical Groups (like rotating spheres or preserving specific distances). The authors want to know: If I have a blueprint for a shape, what is its "dual" or "mirror" blueprint?

2. The Mirror Trick: The Pyasetskii Involution

In this library, there is a special rule called the Pyasetskii involution. Think of it as a magical mirror.

• If you hold up a blueprint (let's call it Blueprint A) to this mirror, it doesn't just flip left-to-right. It rearranges the internal structure to create Blueprint B.
• This isn't a random flip. It's a precise transformation that preserves the "essence" of the shape but changes its "position" in the library's hierarchy.
• The authors call this process an involution because if you look in the mirror twice, you get back to where you started ($A \to B \to A$).

3. The Problem: The Library is Too Big

For a long time, mathematicians knew how to use this mirror for one specific type of blueprint (those related to General Linear Groups, which are like simple, straight-line structures). They had a perfect recipe for it, developed by Mœglin and Waldspurger.

However, the library also contains Classical Groups (like spheres and orthogonal shapes). These are more complex.

• Some of these blueprints are "well-behaved" (Good Parity).
• Some are "tricky" or "broken" (Bad Parity).

Until this paper, no one had a clear, step-by-step recipe to use the mirror on the tricky, "Bad Parity" blueprints. It was like having a map for the easy parts of a maze but getting lost in the dark, twisty tunnels.

4. The Solution: A Hybrid Recipe

The authors created a new algorithm by combining two existing tools:

1. The Old Recipe (Mœglin-Waldspurger): Used for the "easy" blueprints.
2. The New Recipe (Lanard-M´ınguez): A tool originally designed for a different problem (the Aubert-Zelevinsky involution) that the authors realized could be adapted for the "tricky" blueprints.

The Strategy:
They realized they could break any complex blueprint down into smaller, independent pieces (like taking apart a Lego castle into individual bricks).

• Piece 1 (Non-Selfdual): These are bricks that don't look like their own mirror image. They use the Old Recipe.
• Piece 2 (Good Parity): These are tricky but stable. They use a clever trick involving the "closure order" (a way of ranking how "close" blueprints are to each other) to prove the mirror works simply.
• Piece 3 (Bad Parity): These are the truly messy bricks. Here, they adapted the New Recipe. They proved that for these specific messy cases, the "Mirror Trick" (Pyasetskii) is exactly the same as the "Transformation Trick" (Aubert-Zelevinsky) used in representation theory.

5. The "Aha!" Moment: Why It Matters

The most exciting part of the paper is the connection to ABV-packets.

• Imagine that every blueprint doesn't just represent one object, but a whole family of related objects (a packet).
• Mathematicians have a strong hunch (a conjecture) that if you take a whole family of objects and apply the "Mirror Trick" to the whole family, you get the family of the mirror images.
• The authors proved their algorithm works. By showing that their algorithm correctly calculates the mirror image for the "Bad Parity" cases, they provided strong evidence that this hunch is true.

Summary Analogy

Imagine you are a chef trying to reverse-engineer a complex recipe.

• You know how to reverse simple soups (General Linear Groups).
• You have a new, complicated stew (Classical Groups) with some ingredients that are hard to handle (Bad Parity).
• You realize that the technique used to deconstruct a cake (Aubert-Zelevinsky) actually works perfectly for the hard ingredients in your stew.
• You write down a new cookbook chapter that tells you exactly how to reverse any stew, no matter how complex.
• By doing this, you prove that your kitchen's "flavor theory" (ABV-packets) is consistent: if you reverse the ingredients, you get the reverse of the dish.

In short: Hazeltine and Lo have written a user-friendly guide that allows mathematicians to take any complex symmetry blueprint in this specific universe, flip it inside out using a precise mathematical mirror, and be confident that the result is correct. This helps solve bigger mysteries about how numbers and shapes relate to each other.

Technical Summary

1. Problem Statement

The paper addresses the computational problem of determining the Pyasetskii involution (denoted $\hat{\phi}$) for Local Langlands Parameters (L-parameters) of classical groups over non-Archimedean local fields.

• Context: Let $G$ be a connected reductive group (specifically $Sp_{2n}$, $SO_{2n+1}$, or $O_{2n}$) over a non-Archimedean local field $F$. The set of L-parameters $\Phi(G)$ is in bijection with the set of orbits $V_\lambda / H_\lambda$ of a Vogan variety $V_\lambda$ under the action of a centralizer group $H_\lambda$.
• The Involution: The Pyasetskii involution is a geometric involution on the set of orbits $V_\lambda / H_\lambda$. It plays a crucial role in the study of local Arthur packets and the recently developed ABV-packets (Adams-Barbasch-Vogan).
• The Gap: While the Pyasetskii involution is well-understood for general linear groups ($GL_n$) via the Mœglin-Waldspurger algorithm (based on multi-segments), an explicit combinatorial algorithm for classical groups ($Sp_{2n}, SO_{2n+1}, O_{2n}$) was missing from the literature. Furthermore, the behavior of this involution in the "bad parity" case for classical groups was not explicitly computed or geometrically interpreted.

2. Methodology

The authors develop a unified algorithm by decomposing the problem into manageable components based on the structure of the infinitesimal parameter $\lambda$.

A. Decomposition of the Vogan Variety

The authors observe that the Vogan variety $V_\lambda$ and the centralizer $H_\lambda$ naturally decompose into products based on the isomorphism classes of representations appearing in the infinitesimal parameter.

• They decompose the parameter $\lambda$ into components $\lambda_{[\rho]}$ associated with equivalence classes $[\rho]$ of representations of the Weil group.
• This induces a decomposition of the L-parameter $\phi$ into isotypic parts $\phi_{[\rho]}$.
• Key Reduction: The computation of the Pyasetskii involution $\hat{\phi}$ reduces to computing the involution for each isotypic part $\hat{\phi}_{[\rho]}$ separately (Lemma 3.2).

B. Classification of Cases

The algorithm treats the isotypic components based on the properties of the representation $\rho$:

1. Non-selfdual case ($[\rho] \neq [\rho^\vee]$): The problem reduces to the $GL_n$ case. The Pyasetskii involution is computed using the standard Mœglin-Waldspurger algorithm.
2. Good parity case ($[\rho] = [\rho^\vee]$ and parity condition met): The authors prove that the Pyasetskii involution on the classical group coincides with the involution induced by the standard embedding into $GL_N$. This relies on the fact that the involution on $GL_N$ preserves the subset of parameters corresponding to the classical group.
3. Bad parity case ($[\rho] = [\rho^\vee]$ and parity condition failed): This is the most intricate case. The authors adapt the proof of Mœglin-Waldspurger (specifically Proposition II.6 from [MW86]) to the classical group setting.

C. The "Bad Parity" Algorithm

For the bad parity case, the authors utilize the Lanard-Minguez algorithm (from [LM25]), which was originally designed to compute the Aubert-Zelevinsky (AZ) involution for representations.

• They establish a geometric interpretation of the Lanard-Minguez algorithm in this context.
• They prove that for bad parity parameters, the Pyasetskii involution $\hat{\phi}$ is identical to the AZ involution $AZ(\phi)$.
• Technical Tool: They employ a lemma on finite partially ordered sets (Lemma 4.1). If two involutions $\iota_1, \iota_2$ on a poset satisfy $\iota_1(s) \geq \iota_2(s)$ for all $s$, then $\iota_1 = \iota_2$. This allows them to avoid re-proving the most technical half of the Mœglin-Waldspurger argument; they only need to show one inequality holds, and the involution property forces equality.

3. Key Contributions and Results

Main Theorem (Theorem 1.1)

The paper provides a complete algorithm to compute the Pyasetskii involution $\hat{\phi}$ for an L-parameter $\phi$ of a classical group $G_n$.

• Decomposition: $\phi$ is decomposed into isotypic parts $\phi_{[\rho]}$.
• Computation:
  • If $[\rho]$ is non-selfdual or good parity: Use the Mœglin-Waldspurger algorithm.
  • If $[\rho]$ is bad parity: Use the Lanard-Minguez algorithm (Algorithm 7.4) for the Aubert-Zelevinsky involution.
• Result: The global involution is the product of the involutions of the parts.

Geometric Interpretation

The paper provides a geometric justification for the Lanard-Minguez algorithm. It demonstrates that the combinatorial steps of the Lanard-Minguez algorithm (operating on multi-segments) precisely realize the geometric Pyasetskii involution on the Vogan variety orbits for the bad parity case.

Connection to ABV-Packets

The authors prove that their unconditional result (Theorem 1.1) provides strong evidence for a conjecture regarding ABV-packets (Conjecture 8.3).

• Conjecture: $\Pi^{ABV}_{\hat{\phi}}(G) = \{ \hat{\pi} \mid \pi \in \Pi^{ABV}_{\phi} \}$.
• Implication: In the bad parity case, L-packets are singletons. The authors show that if the Pyasetskii involution matches the AZ involution (as their algorithm proves), then the ABV-packet compatibility holds for these specific parameters.

4. Significance

• Algorithmic Completeness: This work fills a critical gap in the literature by providing the first explicit, combinatorial algorithm for the Pyasetskii involution on classical groups, extending the well-known $GL_n$ results.
• Unification of Concepts: It bridges the gap between the geometric definition of the Pyasetskii involution (via Vogan varieties) and the representation-theoretic Aubert-Zelevinsky involution (via multi-segments and AZ duality).
• Evidence for Deep Conjectures: By proving the algorithm unconditionally, the paper offers concrete evidence for the compatibility of ABV-packets with the Aubert-Zelevinsky involution, a central theme in the modern Local Langlands Program.
• Methodological Innovation: The use of the poset lemma (Lemma 4.1) to simplify the proof of the bad parity case is a significant technical contribution, streamlining the adaptation of Mœglin-Waldspurger's complex arguments to classical groups.

In summary, Hazeltine and Lo have successfully translated a complex geometric involution into a computable combinatorial algorithm for classical groups, unifying the theories of local Langlands parameters, Vogan varieties, and Arthur packets.