Eigenvalue accumulation for operator convolutions on locally compact groups

This paper establishes that the asymptotic accumulation of eigenvalues near unity for convolutions of indicator functions and density operators on locally compact groups occurs if and only if the group is unimodular and the underlying sets form a Følner sequence, thereby extending known results for the Heisenberg group to broader classes such as nilpotent and homogeneous Lie groups.

The Gist

Imagine you are a sound engineer trying to understand the "shape" of a massive, complex room (a mathematical group) by bouncing sound waves off its walls. In this paper, the author, Florian Schroth, is investigating how these sound waves behave when they hit specific patches of the wall, and how the "echoes" (mathematical eigenvalues) tell us about the room's fundamental geometry.

Here is the story of the paper, broken down into simple concepts and analogies.

1. The Setup: The Room and the Echo

• The Room (The Group): Think of the mathematical object $G$ as a giant, infinite room. Some rooms are perfectly symmetrical (like a sphere), while others are lopsided (like a funnel). In math, the symmetrical ones are called unimodular, and the lopsided ones are non-unimodular.
• The Sound (The Operator): Instead of a microphone, we have a "density operator." Think of this as a specific, high-quality sound source placed in the room.
• The Patches (The Sets): We are looking at specific areas on the wall, called $E_k$. We are interested in what happens when we make these patches bigger and bigger (like inflating a balloon inside the room).
• The Echoes (Eigenvalues): When the sound hits these patches, it creates a pattern of echoes. Some echoes are very loud (close to 1), and some are quiet. The author is counting how many "loud" echoes appear as the patches get huge.

2. The Big Question: How Many Loud Echoes?

The author asks: If I keep making my wall patches bigger and bigger, how does the number of "loud" echoes grow?

Previously, a mathematician named Simon Halvdansson made a guess. He thought that no matter what kind of room you were in, the number of loud echoes would grow at a predictable rate, simply proportional to the size of the patch.

Schroth's Discovery: "Wait a minute," Schroth says. "That guess is wrong for lopsided rooms."

He proves that the "predictable growth" only happens if two very specific conditions are met:

1. The Room must be Symmetrical (Unimodular): The room cannot be a funnel; it must have a balanced geometry where the rules of volume work the same way in every direction.
2. The Patches must be "Well-Behaved" (Følner Sequences): The patches you are inflating must expand in a very specific, smooth way. They can't be jagged, spiky, or weirdly shaped. They need to be "nice" shapes that fill the room evenly, like a growing circle or a growing cube.

If the room is lopsided or the patches are weird, the math breaks down, and the number of loud echoes doesn't follow the simple rule.

3. The "Magic" of Special Rooms (Nilpotent and Homogeneous Groups)

The paper then zooms in on a special class of rooms called Nilpotent Lie Groups (and a subset called Homogeneous Groups).

• The Analogy: Imagine a room built out of Lego blocks where the structure is rigid and predictable. These rooms are always symmetrical (unimodular).
• The Result: Because these rooms are symmetrical, Schroth's rule works perfectly here. If you take a standard shape (like a ball) and keep making it bigger, the number of loud echoes grows exactly as the size of the ball grows.

4. The Special Case: The Heisenberg Group

The paper ends with a famous example: the Heisenberg Group.

• The Analogy: Think of this as a specific type of 3D space used in quantum physics (the study of tiny particles). It's a bit twisted, but it's one of those "Lego" rooms that is perfectly symmetrical.
• The Payoff: Schroth shows that his new, general rule perfectly explains a known result about this specific quantum room. It's like finding a master key that opens a specific lock you already knew about, but now you know why it works and that it works for many other locks too.

Summary: The "Aha!" Moment

• Old Idea: "If you make the patch bigger, the echoes grow predictably, no matter the room."
• New Truth: "The echoes only grow predictably if the room is perfectly symmetrical AND the patch expands smoothly."
• Why it matters: This helps mathematicians and physicists understand how quantum systems behave in different types of spaces. It tells us that the geometry of the universe (or the mathematical model of it) dictates how information (like eigenvalues) accumulates.

In a nutshell: This paper is about realizing that you can't just throw a ball into any room and expect it to bounce the same way. The shape of the room matters, and the shape of the ball matters. If they are both "nice" (symmetrical and smooth), the math works out beautifully. If they are messy, the math gets complicated.

Technical Summary

Here is a detailed technical summary of the paper "Eigenvalue accumulation for operator convolutions on locally compact groups" by Florian Schroth.

1. Problem Statement

The paper investigates the asymptotic distribution of eigenvalues for a specific class of operators known as operator convolutions on locally compact groups ($G$). Specifically, it analyzes the sequence of operators $E_k * S$, where:

• $S$ is a fixed density operator (a positive trace-class operator with specific normalization) acting on a separable Hilbert space $H$.
• $E_k$ is a sequence of measurable subsets of $G$ with finite, non-zero Haar measure.
• The convolution is defined as $E_k * S := \chi_{E_k} * S$, where $\chi_{E_k}$ is the indicator function of the set $E_k$.

The core question is: Under what conditions on the group $G$ and the sequence of sets $(E_k)$ does the number of eigenvalues of $E_k * S$ lying in the interval $(1-\delta, 1]$ (for small $\delta > 0$) asymptotically equal the trace of the operator?

Mathematically, the paper seeks to determine when the following limit holds for all $\delta \in (0, 1)$:
$$\lim_{k \to \infty} \frac{\#\{n \mid \lambda_n^{(k)} > 1 - \delta\}}{\text{tr}(S)\mu(E_k)} = 1$$
where $\lambda_n^{(k)}$ are the eigenvalues of $E_k * S$.

This problem generalizes previous work on localization operators (where $S$ is a rank-one tensor) and addresses a specific claim made by Halvdansson regarding the affine group, which the author aims to verify or refute.

2. Methodology

The author employs a blend of Quantum Harmonic Analysis, Abstract Harmonic Analysis, and Operator Theory.

• Operator Convolutions: The paper utilizes the framework developed by Halvdansson and Werner, defining two types of convolutions:

  1. Function-Operator Convolution: $f * T = \int_G f(x) \pi(x)^* T \pi(x) d\mu(x)$.
  2. Operator-Operator Convolution: $(S * T)(x) = \text{tr}(S \pi(x)^* T \pi(x))$.
     The analysis relies heavily on the properties of the Duflo-Moore operator $D$, which is required for the existence of square-integrable irreducible representations.

• Følner Sequences and Amenability: The concept of Følner sequences (sequences of sets that become increasingly invariant under group translation) is central. The paper establishes a rigorous link between the asymptotic behavior of the eigenvalues and the Følner property of the sets $E_k$.

• Trace Class Analysis: The author uses the spectral properties of trace-class operators. A key technique involves using functional calculus on the eigenvalues. By constructing specific bounded Borel functions $\rho(t)$ (e.g., $\rho(t) = t(1-t)$), the author relates the trace of $\rho(E_k * S)$ to the count of eigenvalues near 1.

• Handling Non-Unimodular Groups: A significant methodological hurdle is the non-unimodular case (where the modular function $\Delta_G \neq 1$). The paper proves that the desired asymptotic behavior fails in non-unimodular settings unless specific, restrictive conditions are met.

• Extension to Nilpotent Groups: Since non-trivial nilpotent Lie groups do not admit square-integrable irreducible representations (due to non-compact centers), the author adapts the theory to representations that are square-integrable modulo the center ($Z$). This involves defining convolutions on the quotient group $G/Z$ and utilizing a smooth section $s: G/Z \to G$.

3. Key Contributions

A. Refutation of a Previous Claim

The paper explicitly disproves a claim by Halvdansson [12, Theorem 5.14] regarding the affine group. Halvdansson suggested that for the affine group, the eigenvalue count grows asymptotically like $\text{tr}(S)\mu(E_k)$ under dilation. Schroth proves this is false because the affine group is non-unimodular. The asymptotic behavior described in the main theorem only holds if the group is unimodular.

B. Necessary and Sufficient Conditions (The Main Theorem)

Theorem 5.4 establishes the equivalence:
For a connected locally compact group $G$ with a square-integrable irreducible representation $\pi$, and a sequence of sets $(E_k)$, the limit
$$\lim_{k \to \infty} \frac{\#\{n \mid \lambda_n^{(k)} > 1 - \delta\}}{\text{tr}(S)\mu(E_k)} = 1$$
holds for all $\delta > 0$ if and only if:

1. $G$ is unimodular ($\Delta_G \equiv 1$).
2. The sequence of inverse sets $(E_k^{-1})$ is a Følner sequence.

This result unifies the geometric property of the group (unimodularity) and the analytic property of the sets (Følner) with the spectral properties of the operators.

C. Generalization to Nilpotent and Homogeneous Groups

The paper successfully extends these results to:

• Nilpotent Lie Groups: By utilizing representations square-integrable modulo the center.
• Homogeneous Groups: Including stratified Lie groups like the Heisenberg group.
  The author constructs specific Følner sequences for these groups (growing balls in word metrics and dilated sets) and proves the eigenvalue accumulation theorem holds for them.

4. Key Results

1. Theorem 5.4 (Main Result): The asymptotic density of eigenvalues near 1 is exactly 1 (normalized by volume and trace) if and only if the group is unimodular and the sets form a Følner sequence.
2. Theorem 6.1 (Nilpotent Case): For a simply connected, connected nilpotent Lie group $G$, if $V$ is a compact symmetric neighborhood in $G/Z$, then for the sequence of sets $V^k$ (powers of the set), the eigenvalue accumulation limit holds.
3. Theorem 6.2 (Homogeneous Case): For homogeneous groups, if sets are scaled by dilations $\Gamma_{r_k}$ with $r_k \to \infty$, the limit holds.
4. Recovery of Heisenberg Results: The paper recovers established results for the Heisenberg group ($H_1$) as a special case, confirming that for the Schrödinger representation, the number of eigenvalues near 1 scales with the area of the dilated set in the phase space ($R^2$).

5. Significance

• Theoretical Rigor: The paper provides a definitive characterization of when "quantum" localization operators (generalized to group convolutions) exhibit classical volume-scaling behavior. It clarifies that unimodularity is a strict requirement for this specific type of eigenvalue accumulation, correcting previous literature that may have overlooked the modular function's impact.
• Bridge Between Analysis and Geometry: It connects the spectral theory of operators (eigenvalue distribution) directly to the geometric structure of the underlying group (amenability/Følner sequences and unimodularity).
• Applicability to Physics: Since the framework is rooted in Quantum Harmonic Analysis (generalizing the Heisenberg group), these results are relevant for understanding time-frequency analysis, signal processing on groups, and the semi-classical limits of quantum systems on non-abelian groups.
• Methodological Extension: The adaptation of operator convolutions to representations that are only square-integrable modulo the center allows the theory to be applied to the broad class of nilpotent Lie groups, which are fundamental in representation theory but previously excluded from strict square-integrable frameworks.

In summary, Schroth's work establishes that the "classical" behavior of eigenvalue counts for operator convolutions is a signature of unimodular amenability, providing a precise mathematical boundary for when these asymptotic laws apply and when they fail.