🔢 mathematics

Closed geodesics in short intervals for random hyperbolic surfaces

This paper demonstrates that for random hyperbolic surfaces of large genus, the variance of the count of closed geodesics in short intervals asymptotically approaches $2H \log X$ as the genus tends to infinity, a result driven by the high spectral density of Laplace eigenvalues and their expected GOE statistics, which contrasts with the behavior of primes in short intervals.

Original authors: Zeev Rudnick

Published 2026-05-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zeev Rudnick

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 on a vast, bumpy landscape made of pure math. This landscape is a "hyperbolic surface," a shape that curves away from itself everywhere, like a saddle or a Pringles chip, but stretched out infinitely in a specific way. On this surface, you can draw lines that loop back on themselves without ever crossing. These are called closed geodesics.

Think of these geodesics like rubber bands you've snapped onto the surface. Some are short and tight; others are long and winding. The longer the rubber band, the more "energy" it has.

The Big Question: Counting the Rubber Bands

Mathematicians have long been fascinated by prime numbers (2, 3, 5, 7, 11...). There is a famous rule (the Prime Number Theorem) that tells us roughly how many primes exist up to a certain number. But what happens if we look at a tiny slice of numbers? Do the primes appear randomly, or do they clump together?

This paper asks the same question, but for our rubber bands (geodesics) instead of prime numbers.

The Prime Analogy: Instead of counting numbers, we count rubber bands.
The "Short Interval": Instead of looking at all numbers up to a billion, we look at a tiny window, say between a billion and a billion plus a thousand.
The Goal: We want to know: If we pick a random, weirdly shaped surface, how many rubber bands will we find in that tiny window? And how much does that number change if we pick a different random surface?

The Experiment: A Crowd of Surfaces

You can't just pick one surface and call it a day, because one weird shape might be an outlier. So, the author, Ze'ev Rudnick, imagines a giant crowd of random surfaces. He uses a specific mathematical "dice roll" (called the Weil–Petersson measure) to pick these surfaces.

He asks: "If I look at the number of rubber bands in a short window for every surface in this crowd, how much does that number vary?"

The Surprising Result

In the world of prime numbers, mathematicians (Goldston and Montgomery) have a famous guess about how much the count varies. They think the variation depends on a specific formula involving logarithms.

Rudnick's paper finds something different for these random surfaces:

The Variation is Bigger: The number of rubber bands in a short window varies twice as much as the prime number guess would suggest.
The Formula: The variation grows like $2 \times (\text{window size}) \times \log(\text{total size})$ .

Why Twice as Much? The "Symmetry" Analogy

To understand why the number is doubled, we need to look at the "music" of the surface.

The Prime Numbers (GUE): Prime numbers are thought to behave like the notes played by a very complex, chaotic orchestra where every instrument is unique and independent. In physics terms, this is called GUE (Gaussian Unitary Ensemble) statistics. It's like a crowd of people all walking in different directions, completely independent of each other.
The Random Surfaces (GOE): The rubber bands on a random surface behave differently. They act like a crowd of people walking in pairs or groups that mirror each other. In physics terms, this is called GOE (Gaussian Orthogonal Ensemble) statistics.

The Analogy:
Imagine you are counting how many people pass through a door in one minute.

If the people are GUE (primes), they arrive in a completely random, uncoordinated stream. The "noise" or variation in the count is $X$ .
If the people are GOE (surfaces), they tend to arrive in synchronized pairs or clusters. This synchronization creates more "clumping," which doubles the variation in the count to $2X$ .

Rudnick shows that for these random surfaces, the "music" of the rubber bands follows the GOE pattern, which is why the variation is exactly double what you'd expect if they followed the prime number pattern.

The "Infinite Degree" Connection

The paper also connects this to a deeper theory involving "L-functions" (a type of mathematical function used to study primes).

Usually, these functions have a "degree" (like a complexity rating).
For standard primes, the degree is 1.
For these random surfaces, the author argues they act like a function with infinite degree.

When you have an infinite degree, the "early time" behavior of the math takes over. In this specific "early time" regime, the math simplifies, and the only thing that matters is that symmetry difference (GUE vs. GOE). This explains why the result is so clean: it's the "infinite degree" limit of the prime number theory, but with the "double variation" of the surface symmetry.

Summary

What they did: They counted rubber bands (geodesics) on random, bumpy shapes in tiny windows.
What they found: The number of rubber bands fluctuates wildly, specifically twice as much as the famous prediction for prime numbers.
Why: It's because the geometry of these random surfaces forces the rubber bands to behave in a "paired" or symmetric way (GOE), whereas prime numbers behave in a more "independent" way (GUE).
The Takeaway: This proves that while primes and rubber bands are mathematical cousins, they dance to different rhythms. On random surfaces, the dance is twice as energetic.

Technical Summary: Closed Geodesics in Short Intervals for Random Hyperbolic Surfaces

Problem Statement
This paper investigates the statistical distribution of closed geodesics on random hyperbolic surfaces of large genus within short intervals. The study draws a direct analogy to the classical problem of the distribution of prime numbers in short intervals. Specifically, the author examines the random variable $\Psi_M(X; H)$ , which counts closed geodesics with norms in the interval $[X, X+H]$ (where $H = o(X)$ ), weighted by the length of the corresponding primitive geodesic. This quantity is the geometric analogue of the Chebyshev function $\psi(x; H)$ in prime number theory.

The central question is to determine the variance of this count when the surface $M$ is chosen randomly from the moduli space $\mathcal{M}_g$ of hyperbolic surfaces of genus $g$ , equipped with the Weil–Petersson measure, in the limit as $g \to \infty$ . The study aims to compare these geometric statistics with conjectural predictions from analytic number theory regarding the variance of primes, specifically the Goldston–Montgomery conjecture.

Methodology
The analysis proceeds through a probabilistic framework utilizing the geometry of moduli space:

Randomization: The surface $M$ is treated as a random point in the moduli space $\mathcal{M}_g$ . The expectation and variance are computed with respect to the Weil–Petersson probability measure.
Mirzakhani's Integration Formulas: The core computational tool is the integration formula developed by Mirzakhani and Petri. These formulas allow for the calculation of the expected value of sums over simple closed geodesics (specifically non-separating ones) by converting them into integrals over length space involving hyperbolic sine functions.
Decomposition of Variance: The variance is decomposed into diagonal and off-diagonal terms:
- Off-diagonal terms: Pairs of distinct geodesics. The paper demonstrates that for random surfaces of large genus, the contribution from non-simple or separating geodesics is negligible ( $O(x^{1/g})$ ). The contribution from pairs of simple non-separating geodesics is shown to converge to the square of the expected value, effectively canceling out in the variance calculation.
- Diagonal terms: The sum of squares of the weights for individual geodesics. This term provides the dominant contribution to the variance.
Analytic Number Theory Comparison: To interpret the geometric results, the paper analyzes the variance of short interval sums for automorphic $L$ -functions of degree $d > 1$ . It utilizes the explicit formula and pair correlation conjectures (form factors) to relate the variance to the spectral statistics of the zeros of these $L$ -functions.

Key Contributions and Results

Main Theorem (Geometric Variance):
The primary result establishes the asymptotic behavior of the variance for random hyperbolic surfaces as the genus tends to infinity. For $X \to \infty$ and $H = o(X)$ :
$\lim_{g \to \infty} \text{Var}_{WP}^g(\Psi_M(X; H)) \sim 2 H \log X$
This result contrasts with the Goldston–Montgomery conjecture for primes, which predicts a variance of $H \log(X/H)$ . The geometric result exhibits two distinct features:
- The logarithmic factor is $\log X$ rather than $\log(X/H)$ .
- There is an additional multiplicative factor of 2.
Theoretical Connection to $L$ -functions:
The paper proves a theorem regarding primitive automorphic $L$ -functions of degree $d > 1$ . Assuming the Riemann Hypothesis (RH) for such functions, the normalized form factor $K(\alpha)$ governing the pair correlation of zeros is shown to satisfy $K(\alpha) = \alpha$ for the "very short" interval regime $\alpha < 1/d$ . This recovers the Gaussian Unitary Ensemble (GUE) form factor in the early-time regime solely from RH, without needing the full pair correlation conjecture.
Explanation of the Discrepancies:
The author provides a heuristic explanation for the differences between the geometric result and the prime number conjecture:
- The $\log X$ factor: The spectrum of the Laplacian on a hyperbolic surface has a much higher density of eigenvalues than the zeros of any finite-degree $L$ -function. Random hyperbolic surfaces are viewed heuristically as an "infinite-degree" limit. In this limit, only the early-time regime of the form factor (where $K(\alpha) \sim \alpha$ ) survives, leading to the $\log X$ dependence rather than $\log(X/H)$ .
- The factor of 2: While the zeros of $L$ -functions are expected to follow GUE statistics (where the form factor behaves as $\alpha$ for small $\alpha$ ), the Laplace eigenvalues of generic hyperbolic surfaces are expected to follow Gaussian Orthogonal Ensemble (GOE) statistics. The GOE form factor behaves as $2\alpha$ for small $\alpha$ . This factor of 2 in the spectral statistics directly translates to the factor of 2 in the variance of the geodesic count.

Significance
The paper claims to establish a rigorous link between the spectral statistics of random hyperbolic surfaces and the distribution of primes in short intervals, mediated by the theory of $L$ -functions and Random Matrix Theory.

It confirms that for generic hyperbolic surfaces, the variance of geodesic counts in short intervals follows a specific asymptotic law ( $2H \log X$ ) that differs from the classical prime number prediction.
It demonstrates that the "infinite-degree" nature of the Laplacian spectrum on these surfaces isolates the early-time behavior of the pair correlation function.
It identifies the factor of 2 as a direct consequence of the transition from GUE statistics (associated with $L$ -function zeros) to GOE statistics (associated with the Laplacian spectrum of generic surfaces).

The work relies on the integration formulas of Mirzakhani and Petri to handle the geometry of moduli space and extends the methodology of Goldston, Montgomery, Bui, Keating, and Smith to the geometric setting. The results suggest that random hyperbolic surfaces serve as a geometric model for the "infinite-degree" limit of automorphic $L$ -functions, where the specific symmetry class (GOE vs. GUE) dictates the variance scaling.