Length spectrum of large genus random metric maps

Simon Barazer, Alessandro Giacchetto, Mingkun Liu

Abstract: We study the length of short cycles on uniformly random metric maps (also known as ribbon graphs) of large genus using a Teichmüller theory approach. We establish that, as the genus tends to infinity, the length spectrum converges to a Poisson point process with an explicit intensity. This result extends the work of Janson and Louf to the multi-faced case.

Introduction

In the study of random geometry, researchers often investigate "maps"—graphs embedded in surfaces. A "metric map" (or ribbon graph) assigns a specific length to every edge, effectively defining a metric on the underlying surface. When the genus $g$ (the number of "holes" in the surface) becomes very large, the geometry of these maps becomes incredibly complex. This paper addresses the length spectrum: the collection of lengths of all possible cycles within these random maps.

Figure 3: Examples of metric ribbon graphs representing surfaces of different topological complexity.

While previous work by Janson and Louf (2023) solved this for "unicellular" maps (maps with only one face), this research tackles the more general and difficult "multi-faced" case. The challenge lies in bridging the gap between discrete combinatorics and the continuous world of Teichmüller theory—the study of how different geometric structures can be placed on a surface.

Methodology: From Combinatorics to Teichmüller Theory

The authors move away from purely combinatorial bijections and instead adopt a Teichmüller theory approach. Their methodology relies on several sophisticated tools:

  • Moduli Space Isomorphism: They exploit a known natural isomorphism between the space of metric ribbon graphs and the moduli space of Riemann surfaces with punctures.
  • Integration Formulas: They use a combinatorial version of Mirzakhani’s integration formula. This allows them to compute the expected number of cycles by integrating over the "volume" of the space of all possible maps.
  • Intersection Numbers: A key ingredient is the recent breakthrough by Aggarwal (2021) regarding the large-genus asymptotics of $\psi$-class intersection numbers, which provides the necessary estimates for the volumes of these moduli spaces.
  • Method of Moments: To prove convergence to a Poisson process, the authors calculate "factorial moments"—essentially checking if the average number of pairs, triples, etc., of cycles in specific length intervals matches what a Poisson distribution would predict.

Key Insights and Results

The most striking finding is that the intensity function $\lambda(\ell) = \frac{\cosh(\ell) - 1}{\ell}$ is identical to the one derived by Mirzakhani and Petri for hyperbolic surfaces. This suggests that at high genus, the "discrete" nature of a graph and the "continuous" nature of a hyperbolic manifold become indistinguishable from the perspective of short cycles. This "Miraculous Match" reveals a deep structural universality between discrete graphs and complex geometric surfaces.

Theorem A & B: Proves that the multiset of cycle lengths converges in distribution to a Poisson point process with intensity $\lambda(\ell) = \frac{\cosh(\ell) - 1}{\ell}$ as $g \to \infty$.

Corollary C (Girth/Systole): The length of the shortest cycle (the girth) converges to a non-homogeneous exponential distribution. Specifically, the probability that the shortest cycle is longer than $t$ is $\exp(-\int_0^t \lambda(\ell) d\ell)$.

Theorem D: Rigorously proves that the expectation of the number of simple closed curves (not restricted to cycles) is infinite for any fixed genus, explaining why the "cycle" restriction (visiting edges at most once) is mathematically necessary in the discrete setting.

Numerical Validation

The paper validates these theoretical results through numerical simulations. The authors sampled $10^3$ units of uniform random one-faced metric ribbon graphs for genera $g = 2, 8, \text{and } 64$. They used Rémy’s algorithm to generate random trees and the Chapuy–Féray–Fusy bijection to transform them into high-genus maps.

Figure 4: Histograms of cycle lengths for increasing genus $g$. As $g$ increases, the experimental distribution (blue) aligns more closely with the theoretical intensity curve (green).

As shown in the simulations, the experimental cycle counts closely follow the predicted theoretical intensity curve as the genus increases, providing strong empirical support for the Poisson paradigm in high-genus random graphs.

Limitations and Future Work

While these results hold under a specific scaling condition where the total boundary length is proportional to the genus ($L \sim 12g$), the behavior might change under more extreme regimes. Future research could explore the Laplacian Spectrum to understand the "spectral gap" and connectivity of these random high-genus surfaces. Additionally, the authors conjecture that the local limit of these maps is an infinite trivalent tree with exponentially distributed edge lengths (the Benjamini-Schramm limit), which remains to be fully proven for the multi-faced case.

Summary updated

Abstract

We study the length of short cycles on uniformly random metric maps (also known as ribbon graphs) of large genus using a Teichmüller theory approach. We establish that, as the genus tends to infinity, the length spectrum converges to a Poisson point process with an explicit intensity. This result extends the work of Janson and Louf to the multi-faced case.
BibTeX
@inproceedings{barazer2025length,
  author = {Simon Barazer and Alessandro Giacchetto and Mingkun Liu},
  title = {{Length spectrum of large genus random metric maps}},
  booktitle = {Forum},
  year = {2025},
  doi = {10.1017/fms.2025.31},
}
Figure