Probabilistic Weyl Law for Twisted Toeplitz Matrices with Rough Symbols

This paper establishes that the empirical spectral measure of twisted Toeplitz matrices with symbols that are smooth in frequency and piecewise Hölder continuous in position converges weakly in probability to the push-forward of the Lebesgue measure under the symbol, even when the matrices are subject to small random perturbations.

The Gist

The Big Picture: Predicting the Chaos

Imagine you have a giant, complex machine made of thousands of tiny gears (a matrix). You want to know how this machine behaves. Specifically, you want to know where its "energy" (its eigenvalues or spectrum) will end up.

In the world of math, these machines are often built from a blueprint called a symbol. Think of the symbol as the instruction manual or the recipe for the machine.

• Smooth Symbols: If the recipe is written in perfect, flowing cursive, the machine behaves very predictably.
• Rough Symbols: In this paper, the author deals with recipes that are "rough." They have jagged edges, sudden jumps, or are written in a messy, broken handwriting. These are called rough symbols.

The problem is: When you build a machine from a messy recipe, it's chaotic. You can't easily predict where the energy will go. It might clump in one corner or scatter randomly.

The Twist: Adding a Little "Noise"

The author asks a fascinating question: What happens if we shake the machine a little bit?

He takes this messy machine (the Twisted Toeplitz Matrix) and adds a tiny amount of random noise (a random perturbation). Imagine sprinkling a little bit of static electricity or shaking the table slightly.

The Discovery:
Surprisingly, even though the original recipe was messy and the machine was chaotic, adding that tiny bit of random noise acts like a magic smoothing agent.

• Suddenly, the energy stops clumping in weird places.
• It spreads out perfectly evenly across the shape defined by the original messy recipe.
• The chaos organizes itself into a perfect pattern.

This is the Probabilistic Weyl Law: A rule that says, "If you add a little randomness to a rough system, the average behavior becomes perfectly predictable and smooth."

The Characters in Our Story

1. The Twisted Toeplitz Matrix (The Machine):
   Think of this as a giant grid of numbers. In a normal "Toeplitz" machine, the numbers repeat in diagonal stripes (like a tiled floor). But this one is "Twisted" because the numbers change as you move across the grid, following a specific, sometimes messy, rule.

2. The Rough Symbol (The Messy Blueprint):
   This is the rule the machine follows.

   • Smooth part: It flows nicely in one direction (like a river).
   • Rough part: It has "jump cuts" or sudden breaks in the other direction (like a movie with bad editing).
   • Analogy: Imagine a map of a city where the roads are straight and smooth, but the street names suddenly change or disappear at certain blocks. That's a rough symbol.

3. The Random Perturbation (The Shaker):
   This is the tiny bit of randomness added to the machine.

   • Analogy: Imagine a choir singing a song. If one person sings slightly off-key (the rough symbol), the whole song sounds messy. But if you add a tiny bit of random background noise (static), the human ear (or the math) actually starts to hear the true melody of the song much more clearly. The noise washes out the specific errors and reveals the general shape.

4. The Empirical Spectral Measure (The Crowd):
   This is just a fancy way of saying "where are all the energy points?"

   • Analogy: Imagine a stadium full of people (the eigenvalues). The "measure" is a map showing where the people are standing.
   • Without noise: The people might be huddled in a few weird corners because the instructions were confusing.
   • With noise: The people spread out evenly to fill the entire stadium, exactly matching the shape of the stadium's blueprint.

Why Does This Matter?

1. Stability in a Messy World:
In real life, nothing is perfectly smooth. Materials have cracks, signals have static, and data has errors. This paper proves that if you have a system with these "rough" imperfections, adding a tiny bit of randomness doesn't break it; it actually stabilizes it. It forces the system to reveal its true, underlying shape.

2. The "Gaps" are Real:
The paper shows that even with the noise, the energy doesn't go everywhere. It respects the "gaps" in the blueprint.

• Analogy: If your blueprint says "No buildings allowed in the park," the energy (the people) won't go there, even with the noise. The noise fills the allowed areas perfectly, but it respects the forbidden zones.

3. A New Tool for Mathematicians:
The author uses a technique called Semiclassical Analysis.

• Analogy: This is like using a microscope that can zoom in and out. Sometimes you look at the individual gears (quantum view), and sometimes you look at the whole machine (classical view). The author figured out how to switch between these views to prove that the "roughness" disappears when you look at the big picture with a little noise.

The "Takeaway" Metaphor

Imagine you are trying to draw a picture of a mountain range using a broken crayon. The lines are jagged, and the shading is uneven (the Rough Symbol). If you look at the drawing, it looks messy and hard to understand.

Now, imagine you take a piece of sandpaper and gently rub the paper (the Random Perturbation).

• The sandpaper doesn't erase the mountain.
• Instead, it smooths out the jagged crayon lines.
• Suddenly, the true, majestic shape of the mountain range emerges clearly. The "roughness" of the crayon is gone, and the "smoothness" of the mountain is revealed.

The paper proves that for a huge class of mathematical machines, a little bit of "sandpaper" (randomness) is all you need to see the beautiful, smooth pattern hidden inside the mess.

Technical Summary

1. Problem Statement

The paper investigates the asymptotic behavior of the empirical spectral measure of large twisted Toeplitz matrices subject to small random perturbations. Specifically, the author considers matrices $M_N(p)$ associated with a symbol $p(x, \xi)$ defined on the phase space $[0, 1]^d \times \mathbb{T}^d$.

The core challenges addressed are:

• Rough Symbols: Unlike previous works that assumed smooth symbols, this paper allows symbols that are piecewise Hölder continuous with respect to the position variable $x$ and smooth with respect to the frequency variable $\xi$. These symbols may possess jump-type discontinuities (singularities).
• Infinite Bandwidth: The construction generalizes finite-band Toeplitz matrices to potentially infinite-band matrices (depending on the Fourier support of the symbol).
• Random Perturbations: The study focuses on the matrix $M_N(p) + \delta Q_N$, where $\delta$ is a small parameter decaying polynomially with $N$, and $Q_N$ is a random matrix satisfying specific anti-concentration properties.

The goal is to prove that the empirical spectral measure of this perturbed matrix converges weakly in probability to the push-forward of the Lebesgue measure by the symbol $p$ (a probabilistic Weyl law).

2. Methodology

The author employs tools from semiclassical analysis and random matrix theory, adapting techniques developed by Vogel, Christiansen, Zworski, and Hager-Sjöstrand. The proof strategy involves several key steps:

A. Quantization and Symbol Regularization

• Quantization: The paper defines a quantization procedure $Op_h(p)$ mapping symbols to operators on sequences with polynomial growth. For the discrete matrix $M_N(p)$, this corresponds to a truncation of an infinite matrix representation.
• Handling Roughness: Since $p$ is not smooth, the author extends $p$ periodically and regularizes it using a mollifier $\psi_h$ to create a smooth symbol $\tilde{p}$.
• Phase Space Dilation: To handle the singularities and the specific scaling of the problem, a phase space dilation is introduced. This transforms the problem into a setting where the semiclassical parameter $\tilde{h}$ and dilation parameters $\alpha_1, \alpha_2$ allow for the application of standard semiclassical calculus on a torus.

B. Functional Calculus and Trace Estimates

• Grushin Problem: A central tool is the construction of a Grushin problem for the operator $\tilde{p}_N - z$. This involves augmenting the operator with auxiliary finite-dimensional spaces to make it invertible. This technique allows the authors to relate the determinant of the perturbed operator to the integral of the logarithm of the symbol.
• Logarithmic Potential: The convergence of the spectral measure is studied via the convergence of its logarithmic potential $\phi_{\mu_N}(z) = \frac{1}{N^d} \log |\det(M_N(p) + \delta Q_N - z)|$.
• Trace Formulas: The paper derives precise trace and log-determinant formulas for the regularized operators, showing that they converge to the integral of the symbol over the phase space, with error terms controlled by the Hölder exponent $\varrho$ and the geometry of the singularities.

C. Perturbation Analysis

• Deterministic vs. Random: The proof separates the deterministic error (arising from the roughness of the symbol and the discretization) from the random error.
• Anti-concentration: The random matrix $Q_N$ is assumed to satisfy an anti-concentration bound (Assumption 3), ensuring that the smallest singular value of the perturbed matrix does not vanish too quickly. This prevents eigenvalues from clustering too tightly near zero or singularities, which is crucial for the stability of the logarithmic potential.
• Low-Rank Perturbations: The paper also addresses the stability of the result under the addition of deterministic low-rank matrices $R_N$ (Corollary 1.3), showing that the limiting spectral distribution remains unchanged.

3. Key Contributions

1. Generalization of Symbol Class: The paper extends the probabilistic Weyl law from smooth symbols (or specific piecewise constant cases) to the class $C^{0, \varrho}_{pw} S^d$, which includes piecewise Hölder continuous symbols with jump discontinuities. This is a significant step toward modeling non-periodic boundary conditions and physical systems with interfaces.
2. Infinite-Band Matrices: The construction allows for symbols with infinite Fourier series, moving beyond the finite-banded Toeplitz matrices studied in previous literature (e.g., by Basak, Paquette, and Zeitouni).
3. Rigorous Error Bounds: The author provides explicit error estimates depending on the Hölder exponent $\varrho$, the dimension $d$, and the geometric properties of the singularity set $U_p$ (via Assumption 1 regarding the Minkowski dimension of the singularities).
4. Unified Framework: The work unifies the treatment of random perturbations for both smooth and rough symbols using semiclassical analysis, bridging the gap between deterministic spectral theory and random matrix theory.

4. Main Results

Theorem 1.2 (Main Theorem):
Let $p \in C^{0, \varrho}_{pw} S^d$ satisfy Assumptions 1 (geometric condition on singularities) and 2 (volume condition on level sets). Let $Q_N$ be a random matrix satisfying Assumption 3. For $\delta = N^{-(\kappa_3 + \delta_0)}$, the empirical spectral measure $\mu_N$ of $M_N(p) + \delta Q_N$ converges weakly in probability to $p_* L$, where $L$ is the normalized Lebesgue measure on $[0, 1]^d \times \mathbb{T}^d$.

Corollary 1.3 (Stability):
The limiting spectral distribution is robust against the addition of deterministic matrices $R_N$ with controlled norm and rank $O(N^{d-\kappa_4})$. This implies that "outliers" or specific structural perturbations do not alter the bulk spectral distribution.

Technical Estimates (Theorem 5.4):
The paper establishes precise asymptotic formulas for the trace of functions of the operator and the log-determinant, quantifying the error terms as $O(N^{-\eta \kappa_1})$ and $O(N^{-\varrho \eta})$, where $\eta$ is a regularization parameter.

5. Significance

• Mathematical Physics: The results are relevant for the study of quantum chaos and disordered systems where the underlying potential (symbol) is not smooth (e.g., piecewise constant potentials or potentials with interfaces).
• Numerical Linear Algebra: The paper explains the spectral instability of non-normal matrices (like Toeplitz matrices) under small perturbations. It confirms that for a wide class of "rough" symbols, the spectrum fills the numerical range of the symbol, a phenomenon often observed in simulations but difficult to prove for non-smooth cases.
• Methodological Advance: The successful application of semiclassical calculus to rough symbols via regularization and phase space dilation opens new avenues for analyzing spectral problems in non-smooth settings, potentially applicable to other areas like PDEs with discontinuous coefficients.

In summary, Lucas Noël's paper provides a rigorous theoretical foundation for the spectral distribution of randomly perturbed twisted Toeplitz matrices with rough symbols, proving that the "probabilistic Weyl law" holds even in the presence of jump discontinuities and infinite bandwidth.