🔢 mathematics

$L^4$ norm of spectral projectors on polynomially small frequency intervals for $S^1$ -symmetric surfaces

This paper improves the upper bound on the $L^2 \to L^4$ norm of spectral projectors for the Euclidean disk and other $S^1$ -symmetric surfaces in the regime of polynomially small frequency intervals by leveraging the joint eigenbasis of Bessel functions, nonstationary phase estimates, and a novel arithmetic summation technique.

Original authors: Ambre Chabert

Published 2026-06-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ambre Chabert

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are standing in a perfectly round, flat room (a disk) and you clap your hands. The sound waves bounce around the walls, creating a complex pattern of vibrations. In mathematics, these vibrations are called eigenfunctions, and the specific "pitch" or frequency of the clap is represented by a number called $\lambda$ .

This paper is about understanding how "loud" or concentrated these sound waves can get in specific spots, but with a very specific twist: we aren't just looking at one single pitch. Instead, we are looking at a tiny, narrow band of pitches right next to each other (a range from $\lambda - \delta$ to $\lambda + \delta$ ).

Here is the breakdown of what the author, Ambre Chabert, discovered, explained simply:

1. The Goal: Measuring the "Loudness"

In the world of waves, there's a standard rule (Sogge's bound) that tells us how loud a wave can get if we look at a wide range of frequencies. It's like saying, "If you listen to a whole orchestra, the loudest note can't be louder than X."

However, this paper asks: What happens if we only listen to a very tiny slice of the orchestra? (A "polynomially small" slice).

The Challenge: When you narrow the range of frequencies, the waves can sometimes bunch up in strange ways, creating "hotspots" of extreme loudness.
The Discovery: The author proves that even with this tiny slice of frequencies, the waves on a flat disk don't get too crazy. They are bounded by a specific limit that is slightly better (smaller) than the old, general rules.

2. The "Whispering Gallery" Problem

The paper has to be careful about one specific place: the very edge of the disk.

The Metaphor: Imagine a whispering gallery in a cathedral. If you stand near the curved wall and whisper, the sound travels along the wall and can be heard clearly on the other side. These are called "whispering modes."
The Issue: On a disk, these waves hug the edge and get incredibly loud (louder than the standard rules predict).
The Fix: The author's math works perfectly for the inside of the room, away from the walls. To make the math work, they had to draw an invisible line and say, "We are only measuring the volume in the middle of the room, not right against the wall."

3. The Secret Weapon: "Bessel Functions" and "Caustics"

To solve this, the author used the specific mathematical shapes of the waves on a disk, known as Bessel functions.

The Analogy: Think of these waves as ripples in a pond.
- Inside the ripple: The water is calm (exponentially small).
- At the edge of the ripple: The water gets choppy and messy (this is the caustic, or the "focal point" where waves crash together).
- Outside the ripple: The water is rippling smoothly and predictably.
The Strategy: The author broke the problem into three parts:
1. The Calm Zone: Where the waves are too weak to matter.
2. The Messy Zone (Caustic): Where the waves crash. This is hard to calculate, but the author showed that even here, the "loudness" is manageable.
3. The Smooth Zone: Where the waves are just oscillating. Here, the author used a clever trick: because the waves are oscillating at slightly different speeds, they cancel each other out when you add them up (like noise-canceling headphones).

4. The "Convexity" Trick

The most clever part of the math involves a concept called convexity.

The Metaphor: Imagine you are trying to balance four stones on a curved hill. If the hill is perfectly flat, the stones might slide off easily. But if the hill is curved (convex), the stones naturally settle into a stable position.
The Math: The author showed that the "frequencies" of these waves behave like stones on a curved hill. Because of this curve, the waves cannot line up perfectly to create a massive, infinite explosion of loudness. They are forced to spread out just enough to keep the total volume under control.

5. The Result: A New "Volume Limit"

The paper concludes with a new formula for the maximum volume (the $L^4$ norm) of these waves.

The Old Rule: If you listen to a wide range, the volume limit is $X$ .
The New Rule: If you listen to a tiny, precise range, the volume limit is slightly lower (specifically involving a factor of $\lambda^{1/8}$ and the width of the range $\delta^{1/8}$ ).
Why it matters: It proves that even when you zoom in very closely on a specific frequency on a round surface, the waves behave in a predictable, controlled way, provided you stay away from the walls.

Summary

The paper is a sophisticated mathematical proof that says: "On a flat, round surface, if you listen to a very narrow band of high-pitched sounds, the sound won't get infinitely loud in the middle of the room. It has a strict limit, and we can calculate exactly what that limit is, as long as we ignore the weird echoes happening right against the wall."

The author also suggests that this method works for other round shapes (like spheres or donuts) that have similar symmetries, provided you stay away from their "focal points" and edges.

Technical Summary: $L^4$ Norm of Spectral Projectors on Polynomially Small Frequency Intervals for $S^1$ -Symmetric Surfaces

Problem Statement
The paper investigates the operator norm of spectral projectors $P_{\lambda, \delta}$ associated with the Laplace-Beltrami operator $\Delta$ on a compact Riemannian manifold $(M, g)$ . Specifically, it addresses the mapping properties of $P_{\lambda, \delta}$ from $L^2(M)$ to $L^4(M)$ , where the projector is defined on a frequency interval $[\lambda - \delta, \lambda + \delta]$ . While the case of unit bandwidth ( $\delta = 1$ ) is well-understood via Sogge's local estimates, the regime where the bandwidth $\delta$ is polynomially small relative to the target frequency $\lambda$ (i.e., $\delta \ll 1$ ) depends on global geometry and remains largely open. The authors focus on the Euclidean disk $\mathbb{D}^2$ and, more generally, surfaces isometric under an $S^1$ action, in the specific regime where $\delta \geq \lambda^{-1/3 + \eta}$ .

Methodology
The approach relies on the explicit spectral decomposition available due to the $S^1$ symmetry. The methodology proceeds through the following stages:

Joint Eigenbasis Decomposition: The problem is reduced to estimating the $L^4$ norm of a linear combination of joint eigenfunctions of $\sqrt{-\Delta}$ and the rotation generator $\frac{1}{i}\partial_\theta$ . On the disk, these eigenfunctions are expressed in terms of Bessel functions $J_n(\lambda_{k,n} r)e^{in\theta}$ . The $L^4$ norm is expanded into a sum of coefficients $C_{\mathbf{n}}$ involving products of four eigenfunctions indexed by quadruplets $\mathbf{n} = (n_1, n_2, n_3, n_4)$ .
Resonance and Zero-Sum Sets: The angular integration imposes a resonance condition $n_1 + n_4 = n_2 + n_3$ , reducing the effective dimension of the summation. The authors define "zero-sum sets" and further classify them based on the spacing of the indices.
Microlocal Trisection: The analysis of the coefficients $C_{\mathbf{n}}$ $C_{n}$ is divided into three regimes based on the behavior of the Bessel functions relative to their caustic set (where the amplitude transitions from exponential decay to oscillation):
- Interior: Deep inside the classically forbidden region (exponentially small).
- Caustic: Near the caustic (approximated by Airy functions).
- Exterior: In the classically allowed region (oscillatory, approximated by stationary phase).
Oscillatory Integral Estimates: For the exterior and caustic regimes, the coefficients are rewritten as oscillatory integrals with a large parameter $\lambda$ . The authors utilize integration by parts, relying on lower bounds for the gradient of the phase functions. A crucial step involves proving that the phase gradient does not vanish, which is established using the strict convexity of the function $f(r, \nu) = \nu \arccos(\nu/r) - \sqrt{r^2 - \nu^2}$ .
Arithmetic Counting: The summation over the quadruplets is controlled using a new arithmetic estimate. This involves counting the distribution of lattice points (rescaled by $\lambda$ ) lying on or near a strictly convex curve. The authors prove a "Quadrilateral Jarnik's Lemma," extending classical results on triangles to quadrilaterals satisfying the zero-sum condition, to bound the number of contributing terms.

Key Contributions and Results

Main Theorem (Theorem 1): For the Euclidean disk $\mathbb{D}^2$ , away from the boundary and the center (specifically, microlocalized away from the whispering modes near the boundary), the authors prove the following bound for the spectral projector norm:
$\| P_{\lambda, \delta} \chi(x, D) \|_{L^2 \to L^4} \lesssim_{\varepsilon, \eta} \lambda^{1/8} \delta^{1/8} \ln(\lambda)^{1/4}$
for any bandwidth $\delta \geq \lambda^{-1/3 + \eta}$ . Here, $\chi(x, D)$ is a pseudodifferential microlocalizer excluding a small cone near the boundary where whispering modes concentrate.
Optimality and Logarithmic Loss: The exponent $1/8$ on $\lambda$ matches Sogge's exponent $\gamma(4)$ for the case $\delta=1$ , indicating the estimate is nearly coherent with the local theory, with only a logarithmic loss. The authors note that while the exponent on $\delta$ (currently $1/8$ ) may not be optimal, any positive exponent represents a polynomial improvement over the $\delta=1$ case.
Extension to $S^1$ -Symmetric Surfaces (Theorem 3): The method is extended to general compact Riemannian surfaces with an $S^1$ isometric action and isolated fixed points. The result holds for microlocalizations on "regular domains" in the cotangent space, defined by conditions on the curvature of the momentum map and the nature of caustic singularities (specifically fold-type).
Corollaries on Eigenfunctions: The results imply improved bounds for individual $L^2$ -normalized eigenfunctions $\phi_\lambda$ in the interior of the disk: $\|\phi_\lambda\|_{L^4} \lesssim \lambda^{1/12 + \eta}$ . This improves upon the standard Sogge bound of $\lambda^{1/6}$ (which is saturated by whispering modes near the boundary).

Significance and Claims
The paper claims to provide a significant improvement over existing bounds for spectral projectors on thin frequency intervals for manifolds with symmetry. By explicitly utilizing the joint eigenbasis and the specific geometry of the disk (and $S^1$ surfaces), the authors demonstrate that the concentration of mass on caustics, rather than points or geodesics, dictates the $L^4$ behavior.

The authors highlight that the $L^4$ case is singular compared to other $L^p$ norms on the disk; unlike the flat torus where $L^4$ norms are bounded (Cooke-Zygmund), the disk exhibits logarithmic growth due to caustics. The work bridges the gap between the fully understood local case ( $\delta=1$ ) and the open problem of eigenfunction norms ( $\delta=0$ ).

The paper explicitly states that the lower bound on the bandwidth $\delta \geq \lambda^{-1/3 + \eta}$ arises from two sources: the arithmetic distribution of eigenvalues (linked to the circle problem/Weyl law remainder) and the spacing of the joint spectrum. The authors acknowledge that while the method could potentially be extended to smaller $\delta$ using Jarnik's Lemma for exact eigenfunctions, this is not pursued in the current work. Furthermore, the exclusion of the boundary cone is necessary to avoid the "whispering modes" which have a larger $L^4$ norm ( $\lambda^{1/6}$ ).

The methodology is presented as robust, relying on convexity properties of the phase function and new arithmetic counting estimates, suggesting applicability to other completely integrable systems satisfying similar geometric conditions.