Conjectural decomposition of symmetric powers of automorphic representations for $\mathrm{GL}(n)$

This paper establishes a conditional, $k$-independent upper bound on the number of cuspidal isobaric summands in the symmetric $k$-th power lift of a cuspidal automorphic representation for $\mathrm{GL}(n)$, assuming specific Langlands functoriality conjectures and extending the analysis to cases with relaxed cuspidality conditions.

The Gist

Imagine you are a master chef working in a high-end kitchen. You have a special, complex ingredient called $\pi$ (pi). In the world of mathematics, this isn't a number; it's a "cuspidal automorphic representation," which we can think of as a unique, irreducible flavor profile that exists across a vast network of number fields (let's call them "culinary regions").

The paper by Kin Ming Tsang is about what happens when you take this single ingredient and blend it with itself multiple times.

The Main Dish: Symmetric Powers

In math, if you take your ingredient $\pi$ and mix it with itself $k$ times, you create a new, massive dish called the Symmetric $k$-th Power ($Sym_k(\pi)$).

• The Problem: When you blend these ingredients, the resulting dish doesn't always stay as one single, unified flavor. Sometimes, it breaks apart into smaller, distinct sub-dishes (called isobaric summands).
• The Goal: The author wants to answer a simple question: "If I blend my ingredient $k$ times, what is the maximum number of distinct sub-dishes I could possibly end up with?"

The Ingredients and Tools

To solve this, the author uses a few key tools:

1. The Recipe Book (L-functions): Mathematicians use something called "L-functions" to describe the flavor profile of these dishes. If two dishes have the same L-function, they are essentially the same dish.
2. The Blender (Functoriality): There's a famous conjecture (Langlands' principle) that says if you blend your ingredients correctly, the result should still be a valid, "automorphic" dish (a dish that fits the rules of the universe).
3. The Decomposition: Sometimes, the blender doesn't just mix; it separates. The big dish might split into a salad of smaller, pure flavors (cuspidal representations).

The Big Discovery: The "Ceiling" on Chaos

The paper establishes a conditional upper bound. Think of this as a safety net or a ceiling.

• The Assumption: The author assumes that for all the smaller blends (from 1 up to $k-1$), the dishes stayed pure and didn't break apart.
• The Result: Even if the $k$-th blend is huge and complex, the author proves that it cannot break into an infinite number of tiny pieces. There is a strict limit to how many pieces it can split into.

The Analogy of the Tower:
Imagine building a tower out of blocks.

• If you stack 2 blocks, it's stable.
• If you stack 3, it's stable.
• The author asks: "If I stack $k$ blocks, how many separate towers could this one giant pile accidentally collapse into?"
• The paper says: "Don't worry, it won't collapse into a million tiny piles. Even for a huge $k$, the number of separate towers is limited by a specific formula that depends on the size of your original ingredient ($n$) and how many times you blended it ($k$)."

The "Relaxed" Rules

In the first part of the paper, the author assumes the previous blends were perfect (pure). But what if they weren't? What if some earlier blends were already a bit messy?

The author extends the study to these "messier" scenarios. They show that even if you relax the rules and allow some earlier blends to be imperfect, you can still put a limit on how messy the final $k$-th blend gets. However, the more "messy" the previous steps are, the higher the limit on the number of pieces becomes.

The "Sharpness" (Why the Limit is Real)

A crucial part of the paper is proving that these limits aren't just made-up numbers; they are sharp. This means there are specific, real-world examples (using things like "quasi-icosahedral" representations, which are like rare, exotic spices) where the dish actually breaks apart into exactly that maximum number of pieces. The author shows that you can't lower the ceiling any further because nature (math) actually hits that ceiling in specific cases.

Summary in Plain English

1. The Setup: You have a complex mathematical object. You keep mixing it with itself.
2. The Fear: You worry that after many mixes, it might shatter into too many tiny, unmanageable pieces.
3. The Solution: The author proves that no matter how many times you mix it, there is a strict, calculable limit on how many pieces it can break into.
4. The Twist: This limit holds true even if the mixing process wasn't perfect in the earlier steps, though the limit gets slightly higher the messier the earlier steps were.
5. The Proof: The author didn't just guess this limit; they found specific examples where the object breaks into exactly that many pieces, proving the limit is the best possible answer.

In a nutshell: This paper is a safety guide for mathematicians, assuring them that even when they perform complex, high-power operations on these abstract objects, the universe won't let the result explode into chaos. There is always a predictable, finite structure waiting at the end.

Technical Summary

1. Problem Statement

The paper addresses the decomposition of symmetric power lifts of cuspidal automorphic representations. Let $\pi$ be a cuspidal automorphic representation of $GL(n)$ over a number field $F$. The Langlands functoriality principle predicts the existence of an automorphic representation $\Pi$ (denoted $Sym^k(\pi)$) on $GL(m)$, where $m = \binom{n+k-1}{k}$, such that its $L$-function matches the symmetric $k$-th power $L$-function of $\pi$.

The central problem is to bound $N(Sym^k(\pi))$, the number of cuspidal isobaric summands in the decomposition of $Sym^k(\pi)$.

• If $Sym^k(\pi)$ is cuspidal, $N=1$.
• If it decomposes into a sum of cuspidal representations, $N > 1$.
• While bounds are known for small $n$ (e.g., $n=2, 3, 4$) and specific $k$, general bounds for $n \ge 5$ and arbitrary $k$ under conditional functoriality assumptions were lacking.

The author aims to establish conditional upper bounds on $N(Sym^k(\pi))$ assuming the automorphy of lower symmetric powers and exterior power lifts, and to determine how these bounds behave as $k$ grows large.

2. Methodology

The proof strategy relies on a combination of Langlands functoriality, Rankin–Selberg convolution analysis, and combinatorial identities involving Schur polynomials.

A. Schur Polynomial Identities

The core algebraic tool is a specific identity relating products of Schur polynomials. The author utilizes the Littlewood–Richardson rule to derive the following identity (Lemma 2.3):
$$\sum_{i=0}^{\lfloor n/2 \rfloor} S_{(k-2i)} \cdot S_{(1^{2i})} = \sum_{i=0}^{\lceil n/2 \rceil - 1} S_{(k-2i-1)} \cdot S_{(1^{2i+1})}$$
Here, $S_{(k)}$ corresponds to the symmetric power $Sym^k$, and $S_{(1^j)}$ corresponds to the exterior power $\Lambda^j$. This identity translates into a relationship between $L$-functions:
$$\prod_{i=0}^{\lfloor n/2 \rfloor} L(s, Sym^{k-2i}(\pi) \times \Lambda^{2i}(\pi)) = \prod_{i=0}^{\lceil n/2 \rceil - 1} L(s, Sym^{k-2i-1}(\pi) \times \Lambda^{2i+1}(\pi))$$

B. Pole Analysis and Degree Bounds

The author employs a "pole counting" argument:

1. Assume $Sym^k(\pi)$ is not cuspidal and has a cuspidal summand $\tau$ of degree $r$.
2. Consider the Rankin–Selberg $L$-function $L(s, Sym^k(\pi) \times \tilde{\tau})$. If $\tau$ is a summand, this $L$-function has a pole at $s=1$.
3. Using the Schur identity, the pole of the left-hand side (involving $Sym^k$) must be balanced by poles on the right-hand side.
4. By assuming the automorphy of exterior powers $\Lambda^j(\pi)$ and the non-vanishing of certain Rankin–Selberg products at $s=1$, the author forces the existence of a pole in a term involving a lower symmetric power $Sym^{k-2i-1}(\pi)$.
5. This creates a recursive lower bound on the degree $r$ of any cuspidal summand. If the degree $r$ is too small, a contradiction arises regarding the existence of poles.

C. Inductive Relaxation

The paper introduces a parameter $\gamma$ to relax the assumption that all lower symmetric powers are cuspidal.

• Case $\gamma=1$: Assumes $Sym^m(\pi)$ is cuspidal for all odd $m$ in a specific range.
• General $\gamma$: Assumes $Sym^m(\pi)$ is cuspidal only for $m$ in the range $[k-n-\gamma+1, k-\gamma]$.
  The author uses induction to show that even with relaxed cuspidality assumptions, the degree of any summand in $Sym^k(\pi)$ is bounded below by a function $\delta_{n,k}(\gamma)$.

3. Key Contributions and Results

A. Main Theorems

The paper establishes two primary theorems providing upper bounds for $N(Sym^k(\pi))$.

Theorem 1.1 (Strong Cuspidality Assumption):
Assuming $Sym^m(\pi)$ is automorphic for $k-n \le m \le k$ and cuspidal for all odd $m$ in the range $k-2i-1$ (for $0 \le i \le \lfloor \frac{n-1}{2} \rfloor$), and assuming exterior power lifts are automorphic:
$$N(Sym^k(\pi)) \le \left\lfloor \frac{\binom{n+k-1}{n-1}}{\delta_{n,k}(1)} \right\rfloor$$
where $\delta_{n,k}(1)$ is a specific lower bound on the degree of summands derived from the minimum of a ratio of binomial coefficients.

Theorem 1.2 (Relaxed Cuspidality Assumption):
If the cuspidality assumption is relaxed such that $Sym^m(\pi)$ is only guaranteed to be cuspidal for $m \in [k-n-\gamma+1, k-\gamma]$, the bound becomes:
$$N(Sym^k(\pi)) \le \left\lfloor \frac{\binom{n+k-1}{n-1}}{\delta_{n,k}(\gamma)} \right\rfloor$$
where $\delta_{n,k}(\gamma)$ incorporates the factor $n^{\gamma-1}$ in the denominator, reflecting the weaker assumptions.

B. Asymptotic Behavior

A significant contribution is the analysis of these bounds as $k \to \infty$ (specifically $k \ge n^2 + \epsilon$).

• The author proves that for sufficiently large $k$, the bound becomes independent of $k$.
• The asymptotic behavior of the bound is governed by the central binomial coefficient:
  $$N(Sym^k(\pi)) \sim \binom{n}{\lfloor n/2 \rfloor} \sim \frac{2^{n+1}}{\sqrt{\pi n}}$$
• This implies that regardless of how large the symmetric power $k$ is, the number of cuspidal summands does not grow indefinitely; it is capped by a constant depending only on $n$.

C. Sharpness and Examples

• GL(2) Case: The bound is shown to be sharp for $n=2, k=7$ using quasi-icosahedral representations. The paper constructs an example where $Sym^7(\pi)$ decomposes into summands of degree exactly 2, matching the theoretical lower bound derived.
• GL(n) Case: The paper generalizes Walji's work on $Sym^2(\pi)$ for $GL(3)$ to $GL(p^n)$, providing sharp bounds for the number of summands in the symmetric square lift based on the structure of the central character group.

4. Significance

1. Generalization: This work extends previous results by Ramakrishnan ($n=2$) and Walji ($n=3, 4$) to arbitrary $n \ge 5$, providing a unified framework for bounding symmetric power decompositions.
2. Independence from $k$: The discovery that the upper bound on the number of summands stabilizes as $k$ increases is a profound structural insight. It suggests that the "complexity" of the symmetric power lift (in terms of the number of distinct cuspidal components) is bounded by the dimension $n$ of the original representation, not the power $k$.
3. Conditional Framework: The results are conditional on standard Langlands functoriality conjectures (automorphy of exterior powers and Rankin–Selberg products). This provides a clear roadmap: if these conjectures are proven, these explicit bounds become theorems.
4. Methodological Innovation: The use of Schur polynomial identities to derive lower bounds on the degrees of isobaric summands offers a new algebraic technique for analyzing automorphic $L$-functions, distinct from traditional analytic number theory approaches.

In summary, Tsang provides a rigorous, conditional upper bound on the decomposition of symmetric powers of $GL(n)$ automorphic representations, demonstrating that for large $k$, the number of cuspidal components is bounded by a constant dependent only on $n$, and proving the sharpness of these bounds in specific cases.