🔢 mathematics

Quantum percolation conjecture on regular trees

This paper proves that for Bernoulli bond percolation on regular trees of degree at least three, infinite clusters emerge strictly before the appearance of absolutely continuous spectrum, thereby establishing that the critical probability $p_c$ is strictly less than 1.

Original authors: Simon Becker, Izak Oltman

Published 2026-09-30

📖 8 min read🧠 Deep dive

CC0 1.0

Original authors: Simon Becker, Izak Oltman

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ 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, branching network where every connection is a potential path for a traveler. In the world of quantum physics, these networks are not made of roads or wires, but of mathematical possibilities where particles move. Scientists have long been interested in how these particles behave when the network itself is random, with some paths open and others blocked. This field, known as percolation, asks a simple question: if you randomly remove connections from a network, at what point does it break apart into small, isolated islands, and when does a giant, infinite path emerge? For a long time, researchers knew that once this giant path appeared, it would exist. But a deeper, more mysterious question remained: does the existence of this giant path also mean that quantum particles can flow freely through it, like water in a wide river, or do they get stuck, trapped in small pockets?

The answer depends on the type of movement the particles exhibit. In quantum mechanics, particles can move in two very different ways. Sometimes they flow smoothly and continuously, spreading out over the entire network; this is called having an "absolutely continuous" spectrum, and it represents a state of free movement. Other times, the particles become localized, vibrating in place or trapped in specific spots, unable to travel far. This is known as having a "pure point" spectrum. For decades, mathematicians and physicists wondered if the moment a giant path appeared in a random network was the exact same moment that free-flowing quantum states began to appear. It was a reasonable guess that these two events would happen together, but proving it required navigating a complex landscape of probability and infinite structures.

In a new study, Simon Becker and Izak Oltman have settled this question for a specific, highly structured type of network: a regular tree. Imagine a tree where every branch splits into the same number of smaller branches, forever, creating a perfectly symmetrical, infinite structure. The researchers investigated what happens when you randomly keep or remove the connections between these branches. They discovered that the appearance of the giant path and the appearance of free-flowing quantum states do not happen at the same time. Instead, there is a distinct, measurable gap between them. When the network first becomes large enough to contain an infinite path, the quantum particles are still trapped. They cannot flow freely. It is only after the network becomes even more connected, passing through a specific interval of increased density, that the particles finally break free and begin to move continuously.

To reach this conclusion, the authors developed a rigorous mathematical proof that rules out the possibility of free-flowing states in that initial gap. They focused on the behavior of the "resolvent," a mathematical tool that acts like a probe, revealing how a system responds to different energies. By analyzing how this probe behaves on the infinite branches of the tree, they showed that for a specific range of connection probabilities—just above the point where the giant path first appears—the mathematical signature of free flow is completely absent. In this range, the quantum states are not free; they lack an absolutely continuous component. The researchers proved that the imaginary part of this probe, which would indicate free movement, is zero almost everywhere. This means that even though the network is technically infinite and connected, it is effectively a series of dead ends for quantum particles in terms of continuous transport. It is important to note that while the study proves the absence of free-flowing states, it does not claim that the particles are strictly localized in the sense of being purely trapped in specific spots; the absence of the continuous component is a specific spectral property that does not automatically imply the spectrum is purely point-like or that the particles are dynamically localized.

The study also clarifies the limits of this finding. The researchers did not claim that free flow is impossible forever; they showed that it is impossible only in that specific, narrow window just after the network forms. They acknowledged that if the network becomes sufficiently dense, with connections kept at a very high probability, free flow does eventually return. Their work fills a critical gap in our understanding, proving that the transition from a broken network to a flowing one is not a single instant but a two-step process. First, the path opens. Second, the path becomes wide and connected enough to support free movement. Between these two moments, the network exists in a strange, intermediate state where it is large enough to be infinite, yet too fragmented to allow quantum particles to travel freely.

This result is significant because it challenges the intuition that connectivity and flow are the same thing. In the world of quantum percolation on these trees, being connected is not enough to guarantee movement. The researchers used a combination of probabilistic arguments and advanced analysis of how random branches interact to show that the randomness of the network itself acts as a barrier. Even when the giant path exists, the random variations in the size and shape of the smaller branches create a kind of friction that stops the particles. The proof relies on showing that the mathematical equations describing the system have no solution that allows for free flow in that specific interval. This is a definitive mathematical proof, not a simulation or an approximation, establishing a strict separation between the threshold for existence and the threshold for flow.

The work builds on earlier findings by other mathematicians who had shown that free flow does eventually appear when the network is very dense. Becker and Oltman's contribution is to show that this does not happen immediately. They identified a precise, explicit interval where the opposite is true: the spectrum lacks an absolutely continuous component. This means that for the energies within the range of the system, the particles do not exhibit continuous transport. The researchers calculated the size of this interval based on the degree of the tree, showing that for any tree with three or more branches at every node, this gap is guaranteed to exist. The size of the gap depends on the specific structure of the tree, but the phenomenon itself is universal for this class of networks.

By proving that these two thresholds are distinct, the study resolves a long-standing conjecture in the field. It demonstrates that the emergence of an infinite cluster and the emergence of extended quantum states are separate events. The paper does not suggest that the particles are moving in a chaotic or unpredictable way; rather, it proves they lack the specific spectral signature of free flow. The absence of free flow is a robust feature of the system in that intermediate zone. The researchers' method involved a careful analysis of how the random branches of the tree interact, using a technique that compares the behavior of the system under different conditions to show that a contradiction would arise if free flow were present. This contradiction confirms that the free-flowing states simply do not exist in that region.

The implications of this work extend beyond the specific mathematical model. It provides a clearer picture of how disorder affects quantum transport. In many physical systems, from materials to biological networks, understanding when and how flow begins is crucial. This study shows that the onset of connectivity is not the same as the onset of transport. There is a hidden phase where the structure is there, but the function is not yet active. The researchers have mapped out this phase with mathematical precision, showing exactly where it begins and ends. Their findings offer a new perspective on the relationship between structure and function in random systems, revealing that the path to freedom for quantum particles is longer and more complex than previously thought.

In the end, the paper tells a story of separation. In the landscape of a regular tree, the moment the road opens and the moment the car can drive down it are not the same. There is a stretch of road that is open but impassable. The researchers have proven that this stretch exists, defined its boundaries, and explained why it is impassable. Their work stands as a definitive answer to a question that had lingered in the field, separating the fact of existence from the fact of flow, and showing that in the quantum world, being connected is only the first step toward being free.

Technical Summary: Quantum Percolation Conjecture on Regular Trees

Problem Statement
The paper addresses the quantum percolation problem on infinite $(k+1)$ -regular trees ( $T_{k+1}$ ) with $k \ge 2$ . In Bernoulli bond percolation, edges are retained independently with probability $p$ . Above the percolation threshold $p_c = 1/k$ , infinite connected components (clusters) appear almost surely. The central question is whether these infinite clusters support extended quantum states, characterized by the presence of an absolutely continuous (a.c.) component in the spectrum of the adjacency operator $A_\omega$ .

The authors investigate the relationship between two critical thresholds:

$p_c$ : The threshold for the emergence of infinite clusters.
$p_{ac}$ : The threshold for the emergence of an absolutely continuous spectral component on the infinite clusters.

While it is known that $p_c = 1/k$ and that $p_{ac} < 1$ (due to Bordenave's theorem), the strict inequality $p_c < p_{ac}$ had not been proven for regular trees. The paper aims to establish that there exists an explicit supercritical interval $(p_c, p_c + \varepsilon_k)$ where infinite clusters exist, yet the adjacency operator on every such cluster almost surely possesses no absolutely continuous spectrum.

Methodology
The proof relies on a spectral analysis of the diagonal resolvent (Green's function) $G_x(z) = \langle \delta_x, (A_\omega - z)^{-1} \delta_x \rangle$ . The absence of absolutely continuous spectrum is equivalent to the condition that the imaginary part of the boundary value $G_x(E + i0)$ vanishes for almost every energy $E$ .

The methodology proceeds through the following steps:

Resolvent Recursion and Conditioning:
The authors utilize the recursive structure of the tree. By conditioning on the survival of the cluster (i.e., the cluster is infinite), they derive a fixed-point equation for the law $\nu$ of the forward resolvent in the upper half-plane $\mathbb{H}$ . The recursion distinguishes between branches that are finite and those that are infinite.
The law $\nu$ satisfies an equation of the form:
$\nu = (1-q)P\nu + qQ\nu$
where $q$ is the probability of having at least two infinite child branches given survival, $P$ corresponds to the case of exactly one infinite child branch, and $Q$ corresponds to two or more.
Contraction Estimates in $L^2(\mathbb{H})$ :
To prove that no probability solution $\nu$ exists on $\mathbb{H}$ (implying the resolvent must be real almost everywhere), the authors analyze the operators $P$ and $Q$ acting on $L^2(\mathbb{H})$ with respect to the hyperbolic measure.
- The Operator $P^2$ : The core of the argument is showing that the operator $P^2$ is a strict contraction. This is achieved by considering two independent realizations of the finite side branches. The authors identify two specific finite configurations: one with no finite side branches and one with a single-vertex side branch. These configurations induce distinct Möbius transformations.
- Free Group Gap: The relative action of these transformations generates a free group structure. By invoking Kesten's spectral gap estimate for free groups, the authors establish a uniform contraction bound $\|P^2 f\|_2 \le \rho_k \|f\|_2$ with $\rho_k < 1$ .
- The Operator $Q$ : While $Q$ is not necessarily contractive, its weight $q$ is proportional to $kp - 1$. In the regime where $p$ is close to $p_c = 1/k$ , $q$ becomes arbitrarily small.
Fixed-Point Exclusion:
Using a quantitative fixed-point exclusion criterion, the authors show that if the contraction of $P^2$ and the smallness of $q$ satisfy a specific inequality (specifically $\rho_k + 2qB_k < 1$ ), no stationary probability measure $\nu$ supported on $\mathbb{H}$ can satisfy the fixed-point equation. Consequently, the resolvent boundary values must be real almost everywhere.
Handling Singular Measures:
Since the Bernoulli distribution has no density, standard averaging arguments requiring smoothness are not directly applicable. The authors employ a regularization technique using hyperbolic ball averages and convolution bounds for singular measures to extend the contraction estimates to arbitrary probability laws.

Key Contributions and Results

Theorem 1.1: For every $k \ge 2$ , there exists an explicit constant $\varepsilon_k > 0$ such that for $p \in (1/k, 1/k + \varepsilon_k]$ , the absolutely continuous subspace of the adjacency operator $A_\omega$ is trivial ( $H_{ac}(A_\omega) = \{0\}$ ) almost surely on $T_{k+1}$ .
Corollary 1.2: This result establishes the strict separation of thresholds:
$p_c(T_{k+1}) = \frac{1}{k} < \frac{1}{k} + \varepsilon_k \le p_{ac}(T_{k+1}) < 1$
This confirms that the appearance of infinite clusters strictly precedes the appearance of absolutely continuous spectrum.
Explicit Constants: The paper provides explicit, though not optimized, formulas for $\varepsilon_k$ derived from the contraction constants and the probability of multiple infinite branches.

Significance and Claims
The paper claims to resolve the lower bound of the quantum percolation conjecture on regular trees by proving that the spectral transition is distinct from the geometric percolation transition.

Spectral vs. Dynamical: The authors clarify that the absence of absolutely continuous spectrum does not imply pure-point spectrum or dynamical localization; it is a purely spectral statement.
Regime Specificity: Unlike previous works that studied trees close to regular trees or assumed smooth potentials, this work addresses the regime immediately above the first appearance of infinite clusters, where the randomness of finite branches is the dominant factor.
Mechanism: The work highlights that the randomness of finite side branches, even when the infinite backbone exists, is sufficient to suppress the absolutely continuous component in the low-supercritical regime.
AI Usage: The authors explicitly acknowledge using AI (ChatGPT) to identify the specific "two-step contraction mechanism" involving the free-group spectral gap and to assist in deriving the quantitative proposition from classical ideas (Shubin–Vakilian–Wolff and Kesten), though the final mathematical arguments and responsibility remain with the authors.

The paper does not claim to determine the exact value of $p_{ac}$ or to classify the spectral type in the intermediate region between $p_c + \varepsilon_k$ and Bordenave's threshold $p_B(k)$ . It strictly establishes the existence of a gap where infinite clusters are localized in the spectral sense.

More like this