Sharp Eigenfunction Bounds on the Torus for large $p$

This paper establishes the first sharp $L^p$ bounds for eigenfunctions of the Laplacian on the square torus for $p > \frac{2d}{d-4}$ with $d \geq 5$, proving the discrete restriction conjecture without loss by refining the circle method and extending these sharp results to spectral projectors and additive energy.

The Gist

Imagine you are standing in a giant, perfectly square room where the walls, floor, and ceiling are all mirrors. This is a Torus (specifically, a "square torus"). If you shout a sound wave in this room, the sound bounces around forever, creating a complex pattern of echoes.

In mathematics, these sound waves are called eigenfunctions. They are special because they vibrate at specific, pure frequencies, like a perfect musical note.

The big question this paper answers is: How loud can these waves get in specific spots?

The Problem: The "Loudness" Puzzle

Mathematicians have been trying to figure out the maximum volume (or "peak") these waves can reach compared to their average volume.

• The Average Volume: This is easy to measure. It's like the average energy of the sound in the room.
• The Peak Volume: This is the loudest point the sound reaches.

For a long time, mathematicians (like Bourgain and Demeter) could predict the peak volume, but their prediction had a tiny "safety margin" attached to it. It was like saying, "The wave will be at most 100 decibels, plus maybe a tiny bit of static noise." They couldn't get rid of that "static noise" (mathematically called an $\epsilon$-loss) for certain high-pitched notes.

The Breakthrough: Cutting the Static

Daniel Pezzi, the author of this paper, has found a way to remove that static noise completely for very high-pitched notes (large values of $p$) in rooms with 5 or more dimensions.

He proves that for these specific conditions, the peak volume is exactly what the math predicts, with no extra "static" needed. It's the sharpest, cleanest prediction possible.

How Did He Do It? The "Circle Method" Analogy

To understand his method, imagine you are trying to predict the weather by looking at the stars.

• The Old Way: You look at the stars and say, "It's probably going to rain, but there's a 1% chance of error."
• Pezzi's Way: He uses a technique called the Circle Method. Imagine the "time" of the wave is a circle. Most of the time, the wave behaves nicely and is quiet. But at specific moments—when the time aligns perfectly with simple fractions (like 1/2, 1/3, 1/4)—the wave gets very loud because all the tiny ripples line up perfectly (constructive interference).

Pezzi realized that to get a perfect prediction, you have to be incredibly precise about when these loud moments happen.

• He split the problem into three parts:
  1. The "Local" Noise: What happens right at the start (time = 0).
  2. The "Rational" Moments: The specific times when the wave gets loud because of simple fractions.
  3. The "Error" Term: Everything else, which turns out to be very quiet.

By using advanced number theory (the study of whole numbers and their patterns), he was able to calculate the "Rational Moments" so precisely that the "static noise" vanished.

Why Does This Matter? (The "Additive Energy" Analogy)

The paper also applies this to a concept called Additive Energy.

Imagine you have a bag of marbles, each with a number on it. You want to know: "How many ways can I pick a group of marbles that add up to the same total in two different ways?"

• If the marbles are random, there aren't many ways.
• If the marbles are arranged in a special pattern (like points on a sphere), there are many ways.

Pezzi's new, sharper math allows him to count these patterns much more accurately. This helps mathematicians understand the hidden structure of numbers and how they fit together on spheres in high-dimensional space.

The "Logarithmic" Compromise

The paper also offers a "Plan B." If the room is slightly smaller (6 dimensions) or the notes aren't quite as high, he can't remove the static completely, but he can shrink it down to a tiny, manageable whisper (a "logarithmic loss"). This is still a huge improvement over the old "static noise."

Summary

• The Goal: Predict the maximum loudness of waves in a multi-dimensional square room.
• The Old Result: Good, but had a tiny bit of "fuzziness" or error.
• The New Result: For high-pitched notes in 5+ dimensions, the prediction is perfectly sharp. No fuzziness.
• The Method: A refined version of an old number theory trick (the Circle Method) that isolates the exact moments when waves line up perfectly.
• The Impact: This gives mathematicians the most precise tools possible to study waves, numbers, and the geometry of high-dimensional spaces.

In short, Daniel Pezzi took a blurry, slightly fuzzy photo of a mathematical phenomenon and used a new lens to make it crystal clear.

Technical Summary

Here is a detailed technical summary of the paper "Sharp Eigenfunction Bounds on the Torus for Large $p$" by Daniel Pezzi.

1. Problem Statement

The paper addresses the Discrete Restriction Conjecture for eigenfunctions of the Laplacian on the square torus $\mathbb{T}^d = \mathbb{R}^d/\mathbb{Z}^d$. Specifically, it seeks to establish optimal $L^p$ bounds for eigenfunctions $e_N$ associated with the eigenvalue $\lambda = N^2$ (where $N$ is an integer radius).

The conjecture posits that for $d \ge 2$ and $p \ge 2$:
$$\|e_N\|_{L^p(\mathbb{T}^d)} \lesssim_\epsilon N^{\epsilon} N^{\frac{d-2}{2} - \frac{d}{p}} \|e_N\|_{L^2(\mathbb{T}^d)}$$
The central challenge is the presence of the subpolynomial loss factor $N^\epsilon$. While previous results by Bourgain and Demeter established this bound with an $N^\epsilon$ loss for various ranges of $p$, the goal is to remove this loss entirely (achieving a "sharp" bound) for large $p$ and high dimensions ($d \ge 5$).

2. Methodology

The author employs a refinement of the Circle Method from analytic number theory, originally applied to this problem by Bourgain. The methodology involves several key technical innovations:

• Continuous Representation: The discrete sum of exponentials (the convolution kernel $K(x)$) is represented as a continuous integral involving the 1-dimensional Schrödinger propagator $G(t, x)$. This allows the use of tools from harmonic analysis and exponential sum estimates.
• Kernel Decomposition: The kernel $K$ is decomposed into three parts based on the rational approximation of the time variable $t$:
  1. Local Term ($K_0$): Supported near $t=0$.
  2. Major Arc Terms ($K_{Q,s}$): Supported near rationals $a/q$ with denominators $q \sim Q$.
  3. Error Term ($K_{err}$): Supported on times not well-approximated by rationals with small denominators.
• Refined Exponential Sum Estimates:
  • The proof relies on sharp bounds for Generalized Quadratic Gauss Sums $S(a, m, q)$.
  • Crucially, the author avoids the standard "triangle inequality" approach used in previous works (which introduces an $\epsilon$-loss) by utilizing square root cancellation properties of Gauss sums.
  • The author carefully analyzes the singular series (sums over $a$) involving Kloosterman and Salié sums. The paper demonstrates that for small denominators ($Q \sim 1$), the trivial bound is unavoidable due to the potential divisibility of $\lambda$ by many small integers, which limits the range of $p$ for which the sharp result holds.
• Interpolation: The author derives $L^2 \to L^2$ and $L^1 \to L^\infty$ bounds for the decomposed kernels and interpolates them to obtain $L^p$ bounds. The sharpness of the result depends on obtaining estimates without $N^\epsilon$ factors in the parameters $N$ and $2^s$.

3. Key Contributions and Results

A. Main Theorem (Sharp Bounds)

Theorem 1.2: For dimension $d \ge 5$ and $p > \frac{2d}{d-4}$, the Discrete Restriction Conjecture holds without the $N^\epsilon$ loss:
$$\|e_N\|_{L^p(\mathbb{T}^d)} \lesssim N^{\frac{d-2}{2} - \frac{d}{p}} \|e_N\|_{L^2(\mathbb{T}^d)}$$
This is the first sharp result for toral eigenfunctions since the work of Cooke and Zygmund (1970s), which was limited to $d=2$.

B. Logarithmic Loss Result

Theorem 1.3: For a wider range of $p$ (specifically $d \ge 6, p > \frac{2d}{d-3}$ or $d=5, p > 6$), the author proves a bound with only a logarithmic loss:
$$\|e_N\|_{L^p(\mathbb{T}^d)} \lesssim (\log N)^{\frac{d-2}{p(d-4)}} N^{\frac{d-2}{2} - \frac{d}{p}} \|e_N\|_{L^2(\mathbb{T}^d)}$$
This improves upon the $N^\epsilon$ loss of Bourgain and Demeter.

C. Applications

The paper extends these results to two other areas:

1. Spectral Projectors: Sharp $L^2 \to L^p$ bounds for spectral projectors onto annuli (quasimodes) are established for $p > \frac{2d}{d-4}$.
2. Additive Energy: The results are applied to the additive energy of lattice points on spheres ($E_n(F_{N,d})$). The author proves that the conjectured upper bound $E_n(F_{N,d}) \lesssim N^{2n(d-2)-d}$ holds for specific high-dimensional ranges (e.g., $d \ge 9$, or $d=5, n \ge 6$), confirming the conjecture for large $d$ and $n$.

4. Significance and Limitations

• Significance:

  • Removal of $\epsilon$-loss: The paper breaks the barrier of subpolynomial losses for high-dimensional tori, providing the first "truly sharp" bounds in dimensions $d \ge 5$ for large $p$.
  • Methodological Refinement: It demonstrates that the Circle Method, when combined with precise control over exponential sums (specifically avoiding the accumulation of divisor factors in the singular series), can yield sharp results where decoupling methods (which typically yield $N^\epsilon$) cannot.
  • Bridging Fields: It strengthens the connection between harmonic analysis (restriction estimates) and additive combinatorics (additive energy).

• Limitations:

  • Dimensional Constraint: The sharp result is currently limited to $d \ge 5$. The paper notes that for $d=4$, the number of lattice points on spheres behaves differently (requiring an $\epsilon$ factor), preventing a sharp bound via this method.
  • Range of $p$: The sharp result requires $p > \frac{2d}{d-4}$. The method fails to remove the loss for smaller $p$ because the "trivial" bounds for exponential sums near small denominators (where $Q \sim 1$) dominate the estimate. The author argues that improving this would require new techniques to handle the cancellation in the singular series for small $Q$.
  • Square Torus: The results are specific to the square torus $\mathbb{R}^d/\mathbb{Z}^d$ due to the reliance on the specific arithmetic properties of integer lattice points.

5. Conclusion

Daniel Pezzi's work represents a significant advancement in the spectral theory of the Laplacian on the torus. By refining the Circle Method and rigorously controlling the error terms in exponential sum estimates, the author achieves the first sharp $L^p$ bounds for eigenfunctions in high dimensions, removing the long-standing $N^\epsilon$ barrier for large $p$. This work not only resolves a specific conjecture but also provides a framework for understanding the interplay between the geometry of lattice points and the analytic properties of eigenfunctions.