Evenly Divisible Rational Approximations of Quadratic Irrationalities

Dan Carmon

Abstract

In a recent paper of Blomer, Bourgain, Radziwiłł and Rudnick, the authors proved the existence of small gaps between eigenvalues of the Laplacian in a rectangular billiard with sides $\pi$ and $\pi/\sqrt{\alpha}$, i.e. numbers of the form $\alpha m^2+ n^2$, whenever $\alpha$ is a quadratic irrationality of certain types. In this note, we extend their results to all positive quadratic irrationalities $\alpha$.

Introduction

The study of spectral gaps in quantum billiards sits at the intersection of mathematical physics and number theory. For a rectangular billiard table with side lengths $\pi$ and $\pi/\sqrt{\alpha}$, the energy levels (eigenvalues of the Laplacian) are given by the values of the binary quadratic form:

$$Q(m, n) = \alpha m^2 + n^2$$

where $m$ and $n$ are integers. A fundamental question in spectral geometry is the distribution of these energy levels, specifically the behavior of the "minimal gap" between distinct eigenvalues as the energy increases. While a random distribution of numbers (a Poisson process) would suggest certain spacing statistics, the algebraic nature of $\alpha$ can lead to much smaller gaps than expected, a phenomenon known as "clustering."

In 2016, Blomer, Bourgain, Radziwiłł, and Rudnick demonstrated that if $\alpha$ is a quadratic irrational of a specific form, the minimal gap shrinks significantly faster than the average spacing. However, their proof was restricted to certain classes of quadratic irrationals. This paper by Dan Carmon completes the picture, proving that these ultra-small gaps are a universal feature for all positive quadratic irrationalities.

The Problem: Spectral Gaps and Rational Approximations

The core challenge is to bound the minimal gap function $\delta_{min}^{(\alpha)}(N)$, which represents the smallest difference between two distinct values of $\alpha m^2 + n^2$ for $m^2, n^2 \le N$.

To prove that this gap is small, one must find integers $m, n, m', n'$ such that: $$|(\alpha m^2 + n^2) - (\alpha m'^2 + n'^2)| \text{ is very small.}$$

This problem is intimately connected to the theory of Diophantine approximation. If one can find a rational approximation $p/q \approx \alpha$ such that $p$ and $q$ are "evenly divisible"—meaning they can be factored into products of integers of similar magnitude—then these factors can be used to construct the integers $m, n, m', n'$ required to produce a small spectral gap.

Methodology: Algebraic Number Theory and Twisted Traces

Carmon’s approach utilizes the arithmetic of real quadratic fields $K = \mathbb{Q}[\sqrt{D}]$. The construction of the necessary rational approximations relies on several key mathematical pillars:

1. Pell’s Equation and Units

The author uses solutions to the Pell equation $x^2 - Dy^2 = 1$ to identify a fundamental unit or a special algebraic integer $\zeta$ in the ring of integers of $K$. Powers of this $\zeta$ generate a sequence of approximations that grow exponentially, providing the raw material for the proof.

2. Traces and Twisted Traces

The standard "trace" of an element $\omega \in K$ is defined as $tr(\omega) = \omega + \bar{\omega}$. While standard traces are sufficient for some values of $\alpha$, they do not provide enough flexibility for the general case. Carmon introduces the "twisted trace": $$\tilde{tr}(\omega) = (\omega - \bar{\omega})\sqrt{D}$$ By alternating between standard and twisted traces, the author can construct sequences of numerators $P_k$ and denominators $Q_k$ that satisfy the required approximation properties for any quadratic irrational $\alpha$.

3. Generalized Chebyshev Polynomials

The divisibility properties of these traces mirror the behavior of Chebyshev polynomials. Specifically, just as $U_{mn-1}(x)$ is divisible by $U_{m-1}(x)$, the traces of powers of algebraic integers exhibit specific divisibility patterns. Carmon leverages these patterns to ensure that $P_k$ and $Q_k$ are "evenly divisible," allowing them to be split into the components necessary to define the billiard's energy states.

Key Results

The primary contribution of this work is Theorem 1.2, which establishes a universal bound for the minimal gap:

  • Universal Small Gaps: For any positive quadratic irrational $\alpha$, the minimal gap $\delta_{min}^{(\alpha)}(N)$ is bounded by $N^{-1+\epsilon}$ infinitely often for any $\epsilon > 0$.
  • Specific Stronger Bounds: In cases where $\alpha$ satisfies certain algebraic conditions (specifically, when it is a square within the quadratic field), the bound is even tighter, reaching $\ll N^{-1}$.

These results show that the clustering of eigenvalues is not a fluke of specific numbers like $\sqrt{2}$, but a fundamental property of the quadratic nature of the side-length ratio.

Discussion and Future Directions

By extending the work of Blomer et al. to all quadratic irrationalities, Carmon has provided a complete characterization of this phenomenon for the quadratic case. However, the research opens several new avenues:

  • Higher Degree Irrationals: The current proof relies heavily on the structure of quadratic fields and Pell's equation. Whether similar small gaps exist for cubic irrationals (like $\sqrt[3]{2}$) or higher-degree algebraic numbers remains an open and significantly more difficult question.
  • Gap Frequency: The current result proves these gaps occur "infinitely often" as $N$ increases. A more stringent requirement would be to prove these bounds for all $N$ beyond a certain threshold, which remains a challenge for the general case.
  • Physical Interpretation: While the mathematical proof is rigorous, the implications for the physical dynamics of the billiard—such as the transition to quantum chaos—offer fertile ground for future collaboration between mathematicians and physicists.

Summary updated

Abstract

In a recent paper of Blomer, Bourgain, Radziwiłł and Rudnick, the authors proved the existence of small gaps between eigenvalues of the Laplacian in a rectangular billiard with sides $π$ and $π/\sqrtα$, i.e. numbers of the form $αm^2+ n^2$, whenever $α$ is a quadratic irrationality of certain types. In this note, we extend their results to all positive quadratic irrationalities $α$.
BibTeX
@inproceedings{carmon2017evenly,
  author = {Dan Carmon},
  title = {{Evenly Divisible Rational Approximations of Quadratic Irrationalities}},
  booktitle = {Israel},
  year = {2017},
  doi = {10.1007/s11856-017-1624-6},
}
Figure