🔢 mathematics

Rainbow percolation

This paper establishes a genuine phase transition for the weight-dependent random connection model on a Poisson point process, proving that finite components exist when $\lambda\beta<1$ while an infinite component emerges when $\lambda\beta\geq 31$ , and extends these results to the age-dependent random connection model using novel arguments involving "rainbow" disconnections and hyperbolic Peierls arguments.

Original authors: Peter Gracar, Benjamin Lees

Published 2026-10-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Peter Gracar, Benjamin Lees

Original paper licensed under CC BY 4.0 (https://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 a vast, empty line stretching forever in both directions. Now, scatter dots along this line at random. Some dots are ordinary, but others are special: they carry a hidden "reach" that determines how far they can stretch to grab a neighbor. In the world of mathematics, this setup is a way to study how connections form in a network. If the dots are too sparse or their reach too short, the network breaks into tiny, isolated islands. But if the reach is long enough, or the dots dense enough, a single, massive chain can form that stretches infinitely, linking everything together. This tipping point, where a system suddenly shifts from being broken into pieces to becoming a single, unbroken whole, is called a phase transition. It is a fundamental concept in physics and mathematics, appearing in everything from magnets to the spread of diseases.

For decades, mathematicians have studied a specific version of this problem where the dots are placed randomly, but their reach follows a very particular rule: the farther apart two dots are, the less likely they are to connect, but the chance of connection drops off at a very specific, critical speed. If the connection chance drops just a little slower, an infinite chain is guaranteed. If it drops a little faster, the chain is guaranteed to break. But right at that critical speed, the answer has been elusive. The difficulty is compounded when the dots are not independent; instead, their ability to reach out depends on a shared, random property, making the connections in the network tangled and dependent on one another. For a long time, it was unclear whether this specific, tangled setup would ever form an infinite chain or if it would always remain fragmented.

In a new study, researchers Peter Gracar and Benjamin Lees from the University of Leeds have finally settled this question. They proved that this specific network model does indeed undergo a genuine phase transition. They showed that if the average reach of the dots is kept below a certain threshold, the network will almost certainly remain broken into finite pieces, no matter how long you wait or how far you look. However, if the reach is increased just enough to cross a specific value, an infinite chain is guaranteed to appear. The researchers did not just guess this; they provided a rigorous mathematical proof that the transition exists, locating the exact point where the shift happens between a value of one and a value of thirty-one.

To reach this conclusion, the team had to navigate a tricky obstacle. They first examined a simplified version of the problem where the dots were forced to sit at fixed, evenly spaced intervals, like beads on a string. In this rigid setup, they discovered something surprising: the network never forms an infinite chain, no matter how large the reach becomes. It stays broken, jumping instantly from total fragmentation to a state where only the immediate neighbors are connected. This result was counterintuitive because, even in this broken state, there are still infinitely many long connections crossing every single point on the line. The reason the chain never forms is that the fixed spacing limits how many dots can fit into any given area, capping the density of connections.

The real breakthrough came when the researchers returned to the original, more fluid model where the dots can appear anywhere along the line. Here, the spacing is not fixed; the dots can cluster together. The authors demonstrated that this ability to cluster is the key. In the continuous version, the density of dots can become arbitrarily high in certain regions, allowing the network to overcome the limitations that kept the fixed version broken. They proved that once the reach parameter is large enough, these dense clusters link up across different scales, creating a pathway that extends forever. They used a clever method of breaking the line into nested windows of different sizes, showing that if the windows are dense enough, they act like stepping stones that inevitably connect to form a giant, unbroken structure.

The study also included a numerical investigation to pinpoint the exact location of the tipping point more precisely. By running thousands of computer simulations, the researchers found that the critical value where the infinite chain first appears is likely much closer to two than to the upper limit of their proof. The simulations suggested that at this critical point, the size of the largest connected group jumps suddenly, rather than growing slowly. This behavior is distinct from other types of networks where the change is gradual. The findings confirm that the model behaves differently depending on whether the dots are fixed in place or free to move, highlighting how the freedom of movement in a system can fundamentally alter its ability to connect.

This work closes a significant gap in our understanding of how networks form in one dimension. It resolves a question that had been left open by earlier research, showing that the boundary between a broken world and a connected one is sharp and real, even when the connections are complex and dependent on each other. The results provide a clear map for when a system will hold together and when it will fall apart, offering a precise mathematical description of a phenomenon that appears in many natural and engineered systems. The researchers' proof stands as a definitive answer to a long-standing puzzle, demonstrating that even in a world of randomness and dependence, order can emerge from chaos, provided the conditions are just right.

Technical Summary: Rainbow Percolation

Problem Statement
This paper investigates the weight-dependent random connection model on the real line ( $\mathbb{R}$ ) at the critical decay of its connection kernel. The model is defined on a Poisson point process of intensity $\lambda$ on $\mathbb{R} \times (0, 1)$ . Vertices $(x, t)$ and $(y, s)$ are connected if $(t \vee s)|x - y| \le \beta$ , where $\beta > 0$ is a parameter. This connection rule implies that vertices behave as balls with radii distributed according to a Pareto law with tail exponent 1.

The study focuses on a regime where classical one-dimensional percolation theory fails due to two specific features:

Critical Decay: The connection probability decays as $d^{-2}$ (the inverse square of distance), which is the critical exponent for one-dimensional long-range percolation. In models with independent edges, the existence of an infinite component at this decay depends on the multiplicative constant, not just the exponent.
Edge Dependence: Edges sharing a vertex are strongly dependent because they share a common mark (radius). This prevents the direct application of classical results for independent long-range percolation.

The central question is whether this model undergoes a genuine phase transition (i.e., whether there exists a critical $\beta_c \in (0, \infty)$ ) or if the threshold is degenerate (always 0 or always $\infty$ ).

Methodology
The authors employ a combination of geometric probability, renormalization, and combinatorial arguments, distinguishing between a continuum model and a simplified discrete skeleton.

The Simplified Discrete Model (Subcritical Proof):
- The authors first analyze a discrete skeleton where vertices are pinned to $\mathbb{Z}$ and carry independent Pareto radii $R_i$ with scale $x_m$ .
- Rainbow Construction: They define a "rainbow" as a nested pair of long edges $\{a, b\}$ and $\{c, d\}$ (with $a < c < d < b$ ) such that the diagonals are absent and the overhangs are balanced.
- Cut-Point Certificates: To prove non-percolation, they construct "cut-point certificates" (gaps) within the overhangs of these rainbows. These certificates are events where radii are sufficiently small to prevent any path from crossing a specific cut.
- Multi-Scale Argument: Using the Kochen–Stone lemma and Kolmogorov's zero–one law, they show that for $x_m < 1$ , such rainbows and their associated gaps occur at infinitely many independent scales, almost surely confining every connected component to a finite interval.
The Continuum Model (Subcritical Extension):
- The discrete results are transferred to the Poisson point process. The authors adapt the cut-point certificates to the continuum using void events of explicit regions in $\mathbb{R} \times (0, 1)$ .
- They utilize the Harris–FKG inequality for Poisson processes to handle the dependence structure, showing that the subcritical behavior ( $\lambda \beta < 1$ ) holds for the continuum model.
Renormalization and Peierls Argument (Supercritical Proof):
- To prove the existence of an infinite component for large $\lambda \beta$ , the authors decompose the mark space $(0, 1)$ into dyadic bands.
- Scale Invariance: Each band forms a dense, bounded-range Gilbert-type graph. Due to the min-rule, a chain of points in a band $k$ can deterministically "absorb" points from higher bands (larger radii) within its span.
- Binary Tiling: The incidence structure of these dyadic windows across scales maps to the binary tiling of the hyperbolic half-plane.
- Peierls Contour: The problem reduces to independent site percolation on this tiling. By establishing that the "open" site probability (local density of points) is sufficiently high, they use a Peierls argument (counting contours) to prove the existence of an infinite cluster.

Key Contributions and Results

Existence of a Phase Transition: The paper proves that the model undergoes a genuine phase transition.
- Theorem 1 (Subcritical): If $\lambda \beta < 1$ , almost surely every connected component is finite.
- Theorem 2 (Supercritical): If $\lambda \beta \ge 31$ , almost surely an infinite connected component exists.
- Consequently, at intensity $\lambda=1$ , the critical value satisfies $1 \le \beta_c \le 31$ .
The Discrete Skeleton's Degeneracy:
- Theorem 4: For the simplified model on $\mathbb{Z}$ , the phase transition is degenerate. For any admissible parameter $x_m < 1$ , the graph is totally fragmented (no infinite component). At $x_m \ge 1$ , the graph is trivially connected (all nearest neighbors are connected). There is no intermediate supercritical phase.
- This highlights a fundamental divergence between the lattice and continuum models: the continuum allows for arbitrarily high local densities (driven by $\lambda \beta$ ) which the lattice cannot achieve due to fixed vertex spacing.
Extension to Age-Dependent Models:
- Theorem 3: By comparing kernels, the authors extend the supercritical result to the age-dependent random connection model (kernel $(t \vee s)^{1-\gamma}(t \wedge s)^\gamma$ ) for $\gamma \in (0, 1)$ . They show this model has a non-degenerate phase transition for every $\gamma$ , closing a case left open in previous literature.
Numerical Estimation:
- Appendix A presents a numerical study placing the critical value $\beta_c$ (at $\lambda=1$ ) near 2. The simulations suggest a discontinuous jump in the percolation density at the threshold, with the density of the infinite cluster rising to approximately 0.9 immediately above $\beta_c$ .

Significance
The paper resolves the phase diagram for the weight-dependent random connection model at the critical decay exponent ( $\gamma=1$ ), a boundary case where standard decay-exponent criteria are silent.

It demonstrates that dependence through common marks does not prevent a phase transition, even at the critical decay of $d^{-2}$ .
It clarifies the distinction between lattice and continuum percolation in heavy-tailed regimes, showing that the continuum model can support an infinite cluster where the lattice counterpart remains fragmented due to density constraints.
It provides a rigorous framework for analyzing rainbow structures and cut-point certificates in dependent percolation models, offering a new toolset for studying inhomogeneous random graphs.
The work closes an open problem regarding the age-dependent random connection model, confirming the existence of a non-degenerate phase transition for all $\gamma \in (0, 1)$ .

More like this