Bilinear forms with trace functions

This paper establishes non-trivial bounds for bilinear sums of trace functions below the Pólya-Vinogradov range by leveraging a general stratification theorem and a robust Goursat-Kolchin-Ribet criterion, thereby extending previous results from special cases to a broader class of $\ell$-adic sheaves with specific geometric monodromy properties.

The Gist

Here is an explanation of the paper "Bilinear Forms with Trace Functions" using simple language, analogies, and metaphors.

The Big Picture: Finding Patterns in Chaos

Imagine you are a detective trying to solve a mystery in a massive, noisy city. The city is full of people (numbers) walking around, and they are all wearing different colored hats (mathematical sequences). Your job is to figure out if two groups of people are secretly communicating with each other, even though they are trying to hide it.

In mathematics, this "communication" is called correlation. The authors of this paper are trying to measure how much two random-looking lists of numbers "wiggle" together. If they wiggle in sync, there is a hidden pattern. If they wiggle randomly, there is no pattern.

The specific tool they are using is a Bilinear Form. Think of this as a giant mixing bowl. You pour in two ingredients (two lists of numbers, let's call them List A and List B) and a secret sauce (a complex function called a "kernel"). The goal is to see if the mixture cancels itself out (meaning the ingredients are random) or if a strong flavor remains (meaning there is a hidden connection).

The Problem: The "Too Small" Puzzle

For a long time, mathematicians could only solve this puzzle if the lists of numbers were huge. Imagine you need a stadium full of people to hear a whisper. If you only have a few people in a small room, the noise is too loud, and you can't tell if they are whispering or just shouting randomly.

In math terms, previous methods only worked when the lists of numbers were very long compared to the total number of possibilities. The authors wanted to solve the puzzle even when the lists were short (like a small room). This is much harder because there is less data to work with.

The New Tool: "Gallant" Sheaves

The authors introduce a new way to look at the "secret sauce" (the kernel function). They call the functions that work with their new method "Gallant Sheaves."

• The Metaphor: Imagine the secret sauce is a magical potion. In the past, you could only use potions made from specific, rare herbs (like Hyper-Kloosterman sums). If you tried to use a different herb, the potion would fail.
• The Innovation: The authors discovered that the potion doesn't actually care which specific herb you use, as long as the herb has a certain structural backbone. They call these "Gallant" herbs.
  • A "Gallant" herb is one that is strong, symmetrical, and doesn't break easily.
  • They proved that if the backbone of the potion is "Gallant," the mixing bowl will work, even with a small crowd of people.

This is a huge deal because "Gallant" herbs are everywhere! They aren't rare anymore. This means the method works for a massive variety of mathematical problems, not just the few special cases we knew before.

How They Did It: The Three Magic Tricks

To prove this, the authors combined three powerful ideas:

1. The "Soft" Stratification (The Map):
   Imagine you are looking at a foggy landscape. You can't see the whole picture at once. The authors used a technique (based on an idea by Junyan Xu) to draw a map of the fog. They realized that while the fog is messy, it has "safe zones" and "danger zones."

   • The Trick: They proved that for most points in the landscape, the noise cancels out perfectly (square-root cancellation). The "danger zones" where the noise doesn't cancel are very small and rare. By mapping these zones, they could ignore the noise and focus on the signal.

2. The "Goursat-Kolchin-Ribet" Filter (The Bouncer):
   This is a fancy mathematical rule that acts like a bouncer at a club. It checks if two groups of numbers are "friends" (connected) or "strangers" (independent).

   • The Innovation: Previous versions of this bouncer were picky; they only let in specific types of groups. The authors built a new, tougher bouncer that checks the structure of the group rather than just its name. If the group is "Gallant" (structurally sound), the bouncer lets it pass and confirms that the noise will cancel out.

3. The "Shift" Technique (The Detective's Move):
   This is a classic move in detective work. If you want to find a pattern, you shift the timeline slightly and compare the new version to the old one.

   • The Application: The authors shifted their lists of numbers by a small amount (like $+uv$). This created a new equation that was easier to solve. By combining this shift with their "Gallant" filter, they could prove that the correlation was tiny, even for short lists.

Why Does This Matter? (The "Cubic Toroidal Moments")

The paper ends by showing why this matters in the real world of numbers. They apply their new method to study Dirichlet L-functions.

• The Analogy: Think of L-functions as the "DNA" of numbers. They contain deep secrets about how prime numbers are distributed.
• The Application: The authors used their method to study the "cubic moments" of these DNA strands. This helps mathematicians prove that certain numbers (values of these functions) are not zero.
• The Result: Knowing that these values aren't zero is crucial. It's like confirming that a specific gene actually exists and is active. This has implications for understanding the distribution of prime numbers and solving other deep mysteries in number theory.

Summary

• Old Way: We could only find hidden patterns if we had a huge amount of data, and only for very specific types of mathematical functions.
• New Way: The authors found a way to find patterns in small amounts of data for a huge variety of functions.
• The Secret: They realized that as long as the mathematical function has a strong, symmetrical "Gallant" structure, the noise cancels out naturally.
• The Impact: This opens the door to solving many old, unsolved problems in number theory that were previously too difficult because the data sets were too small or the functions were too complex.

In short, they built a better microscope that works in the dark, allowing us to see patterns in numbers that were previously invisible.

Technical Summary

Here is a detailed technical summary of the paper "Bilinear Forms with Trace Functions" by Étienne Fouvry, Emmanuel Kowalski, Philippe Michel, and Will Sawin.

1. Problem Statement

The central problem addressed is the estimation of bilinear sums involving trace functions of $\ell$-adic sheaves over finite fields. Specifically, the authors seek non-trivial bounds for sums of the form:
$$S = \sum_{m \sim M} \sum_{n \sim N} \alpha_m \beta_n K(mb^n c)$$
where:

• $q$ is a large prime modulus.
• $K$ is the trace function of an $\ell$-adic sheaf $\mathcal{F}$ on the affine line $\mathbb{A}^1_{\mathbb{F}_q}$.
• $(\alpha_m)$ and $(\beta_n)$ are arbitrary complex sequences (with no specific structure assumed other than size constraints).
• $b, c$ are non-zero integers.
• The goal is to obtain bounds significantly better than the "trivial" bound (which is roughly $\sqrt{MN}$) for ranges of $M$ and $N$ as small as possible relative to $q$.

Context: Previous works (e.g., by Friedlander-Iwaniec, Fouvry-Michel, and Kowalski-Michel-Sawin) achieved strong results but were restricted to specific types of sheaves, primarily hyper-Kloosterman sheaves or specific hypergeometric sheaves. This paper aims to generalize these results to a much broader class of sheaves defined by structural properties of their monodromy groups.

2. Methodology

The authors develop a framework that combines three major technical ingredients:

A. "Gallant" Sheaves and Monodromy Groups

The core innovation is the definition of gallant sheaves. A sheaf $\mathcal{F}$ is gallant if its geometric monodromy group $G$ (a subgroup of $GL_r(\mathbb{C})$) satisfies specific structural conditions:

1. The action of $G$ on $\mathbb{C}^r$ is irreducible.
2. Either:
   • The identity component $G^0$ is a simple algebraic group (e.g., $SL_r, Sp_r$).
   • $G$ is finite and contains a quasisimple normal subgroup acting irreducibly.

This definition allows the authors to treat a vast array of sheaves (including many with finite monodromy groups) under a unified hypothesis, moving beyond the specific "formula-based" approaches of previous literature.

B. Generalized Goursat–Kolchin–Ribet (GKR) Criterion

To handle sums of products of trace functions (which arise when applying the "shift by $+uv$" method), the authors need to determine when the tensor product of representations has non-trivial invariants (coinvariants).

• They establish a robust, qualitative version of the GKR criterion.
• Unlike classical versions that required specific group types, their criterion works under the "gallant" hypothesis.
• They prove that if the monodromy groups of the individual sheaves are "gallant," the monodromy group of the product sheaf is essentially the full product group (e.g., $G \times G$), unless the parameters lie on a specific "diagonal" algebraic subvariety.
• This allows them to prove that "most" sums of products exhibit square-root cancellation (bounded by $q^{1/2}$), while the "diagonal" cases (where cancellation fails) are confined to low-dimensional algebraic varieties.

C. Xu's Stratification Idea + Quantitative Sheaf Theory

The authors utilize a powerful idea from Junyan Xu (formalized via Sawin's Quantitative Sheaf Theory):

• Idea: Bounds on the moments of a trace function imply a stratification of the parameter space.
• Mechanism: If the $2m$-th moment of a sum is bounded, one can construct a filtration of algebraic varieties$X(w)$such that outside these varieties, the sum is small (e.g.,$O(q^{w/2})$).
• Application: By combining the GKR results (which provide moment bounds) with Sawin's theory (which controls the complexity and degrees of the exceptional varieties), the authors can rigorously bound the "diagonal" contributions that would otherwise ruin the estimates.

3. Key Contributions

1. Generalization of Scope: The paper removes the restriction to specific exponential sums (like Kloosterman sums). It applies to any sheaf whose geometric monodromy group is "gallant." This includes:

   • Hypergeometric sheaves (generic cases).
   • Sheaves with finite monodromy groups (e.g., those arising from symmetric groups $S_n$ for $n \ge 6$).
   • Sheaves with orthogonal monodromy groups (specifically $O_4$ and $SO_4$, termed "oxozonic" and "sulfatic" in the paper).
   • Sheaves obtained via Tannakian operations, twists, and change of variables.

2. New GKR Framework: They provide a flexible criterion for the vanishing of coinvariants in tensor products of representations, applicable to both infinite (simple algebraic) and finite (quasisimple) monodromy groups.

3. Stratification for Bilinear Forms: They successfully integrate Xu's stratification theorem into the analytic number theory context of bilinear forms, allowing for precise control over the "exceptional" sets where cancellation fails.

4. Main Results

The paper provides non-trivial bounds for Type I (linear in $\beta$) and Type II (bilinear in $\alpha, \beta$) sums.

Theorem 1.1 (Simplified Main Result):
Let $K$ be the trace function of a gallant sheaf. For non-zero integers $b, c$, and parameters $M, N$ satisfying:
$$q^\delta \le M, \quad MN \ge q^{3/4+\delta}$$
there exists $\eta > 0$ such that for any sequences $\alpha, \beta$:
$$\sum_{m \sim M} \sum_{n \sim N} \alpha_m \beta_n K(mb^n c) \ll (MN)^{1/2 - \eta} \|\alpha\|_2 \|\beta\|_2$$
This recovers the best known ranges for Kloosterman sums but applies to a vastly wider class of functions.

Theorem 1.3 (Precise Estimates):
The authors provide explicit bounds depending on an integer parameter $l \ge 2$:

• Type I: Non-trivial bounds for $M^2 N \ge q^{1+\delta}$.
• Type II: Non-trivial bounds for $MN \ge q^{3/4+\delta}$.
  These bounds are uniform in the complexity of the sheaf (though the implied constant depends on it).

Theorem 1.4 (Trilinear Sums):
They extend the method to trilinear sums involving monomial arguments $K(j^a m^b n^c)$, obtaining non-trivial bounds under conditions like $J \ge q^\delta$ and $MN \ge q^{1/2+\delta}$.

Theorem 1.6 (Oxozonic Case):
They handle the exceptional case where the monodromy group is isomorphic to $O_4$ (oxozonic), showing that the results hold (with slight modifications) for these sheaves, which are relevant to cubic moments of L-functions.

5. Significance and Applications

• Cubic Moments of L-functions: The primary motivation for this work is the study of cubic toroidal moments of Dirichlet L-functions at the central point:
  $$M_{a,b,c}(q) = \sum_{\chi \pmod q} L(1/2, \chi^a) L(1/2, \chi^b) L(1/2, \chi^c)$$
  The authors prove that for almost all triples $(a,b,c)$, the associated sheaf is gallant or oxozonic, allowing them to derive asymptotic formulas and non-vanishing results for these moments.

• Non-Vanishing of L-values: The results lead to Corollary 1.9, establishing that a positive proportion of Dirichlet characters have non-vanishing products of L-values, a crucial step in understanding the distribution of zeros of L-functions.

• Flexibility in Analytic Number Theory: By decoupling the analytic method from the specific algebraic formula of the trace function, this paper opens the door to applying bilinear form techniques to new families of exponential sums arising in automorphic forms, random matrix theory, and arithmetic geometry.

• Handling Finite Monodromy: A novel contribution is the ability to handle sheaves with finite monodromy groups (previously difficult because they do not satisfy the "large group" assumptions of earlier GKR applications). This is achieved by identifying the quasisimple core subgroup within the finite group.

In summary, this paper represents a major theoretical advance in analytic number theory, unifying the treatment of bilinear sums for a broad spectrum of arithmetic functions through the lens of geometric monodromy and quantitative sheaf theory.