🔢 mathematics

Iterated Graph Systems (III): Alexander-Orbach conjecture on hierarchical lattices

This paper constructs scaling limits of critical percolation clusters on hierarchical lattices to demonstrate that the Alexander-Orbach conjecture generally fails for these systems, with their associated spectral dimensions forming a dense set within the interval $(1,2)$ .

Original authors: Ziyu Neroli

Published 2026-10-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ziyu Neroli

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 a vast, tangled network of roads where some paths are open and others are blocked, a landscape that looks the same no matter how closely you zoom in. This is the world of critical percolation, a mathematical model used to understand how fluids move through porous rock, how electricity flows through random materials, or how diseases spread through a population. At a specific tipping point, known as the critical point, these networks form a giant, infinite cluster that connects one side of the world to the other. For decades, physicists have been fascinated by how a random walker, like a person wandering aimlessly, would move across this giant, jagged structure. In 1982, two researchers proposed a bold idea: that no matter how complex the underlying network is, the way this walker spreads out over time follows a universal rule, characterized by a specific number, four-thirds. This idea, known as the Alexander–Orbach conjecture, has held up in high-dimensional spaces but has faced skepticism in lower dimensions, where the geometry is more intricate and the path is more winding.

A team of mathematicians has now taken a deep dive into this question by constructing a special class of mathematical models called hierarchical lattices. These are not the flat, grid-like streets of a city, but rather self-similar structures built by repeatedly replacing a single line with a more complex pattern, creating a fractal shape that repeats itself infinitely. By studying the movement of a random walker on these specific, infinitely detailed structures, the researchers were able to calculate exactly how the walker spreads. They found that the universal rule proposed by Alexander and Orbach fails for these lattices outside a narrow region of possible shapes. Instead, the speed at which the walker diffuses depends entirely on the specific shape of the lattice being used. While the researchers could find examples where the spreading rate comes very close to the predicted value, they proved that for almost every lattice they examined, outside a specific narrow region, the actual rate is different.

The researchers began by building these fractal networks step by step. They started with a simple line and replaced it with a small, branching pattern, then replaced every line in that pattern with the same branching shape again, and again. This process creates a structure that is infinitely complex yet perfectly ordered. At the critical point, where the network is just barely connected enough to span the whole system, they identified the "chemical distance," which is the actual number of steps a walker must take to get from one end to the other, rather than the straight-line distance. They then studied how the electrical resistance of the network grows as it gets larger. In these fractal worlds, resistance does not grow in a simple, predictable way; it grows at a specific rate that depends on the geometry of the branches.

By combining the growth of the distance with the growth of the resistance, the team calculated the "spectral dimension," a number that describes how the random walker explores the space. If the Alexander–Orbach conjecture were true, this number would always be exactly four-thirds, regardless of the lattice's shape. However, the calculations revealed a different reality. The spectral dimension varies widely depending on the specific rules used to build the lattice. The researchers proved that for the vast majority of these structures, specifically those outside a narrow region of dimension space, the dimension is not four-thirds. It can be higher or lower, but it is almost never that specific value.

There is a very narrow range of possibilities where the dimension could theoretically equal four-thirds, but the researchers could not find a single example of a lattice that actually achieves it. They showed that while one can construct a sequence of lattices whose dimensions get closer and closer to four-thirds, no finite, simple rule produces a lattice that hits the number exactly. This suggests that the conjecture, while a useful approximation in some high-dimensional settings, is likely incorrect for these types of self-similar structures, though proving it fails within that narrow region remains difficult. The work provides a rigorous mathematical proof that the geometry of the underlying network dictates the behavior of the random walker, and that there is no single, universal number that describes this behavior across all hierarchical lattices outside that narrow exception.

The study also confirmed that these mathematical objects behave in a very consistent way. The researchers proved that the growth rates for both distance and resistance are stable and predictable, allowing them to define a precise "spectral dimension" for almost every point on the infinite structure. They found that these dimensions are dense in the range between one and two, meaning that by tweaking the rules of the lattice, one can create a structure with a dimension almost anywhere in that interval. This flexibility highlights the richness of these fractal systems and shows that the behavior of random walks is far more sensitive to the details of the geometry than the old conjecture suggested.

In the end, the paper does not just disprove a famous idea; it maps out the true landscape of how diffusion works on these complex, self-repeating networks. It shows that the universe of these mathematical lattices is diverse, with each shape having its own unique signature for how a random walker moves through it. The Alexander–Orbach value of four-thirds appears to be a rare coincidence rather than a fundamental law of nature for these systems. The researchers have left the door open for the possibility that a very specific, complex rule might one day be found that produces the exact value within the narrow remaining region, but they have shown that such a case would be an exception rather than the rule. For now, the conjecture stands as a beautiful but ultimately likely incorrect guess about the behavior of random walks on hierarchical lattices, with the proof of its failure established for all but a narrow set of cases.

Technical Summary: Iterated Graph Systems (III): Alexander-Orbach conjecture on hierarchical lattices

Problem Statement
The paper addresses the validity of the Alexander–Orbach (AO) conjecture on hierarchical lattices. The conjecture, proposed in 1982, posits that the spectral dimension of the simple random walk on the incipient infinite cluster of critical Bernoulli percolation on $\mathbb{Z}^d$ is exactly $4/3$ for all $d \ge 2$ . While established for high-dimensional mean-field regimes ( $d \ge 11$ ), the conjecture remains mathematically open for $2 \le d \le 10$ and is suspected to fail in low dimensions. The author investigates this problem within the context of Edge Iterated Graph Systems (EIGS), specifically focusing on classical EIGS hierarchical lattices. These are self-similar, fractal graphs generated by deterministic substitution rules, which allow for exact renormalization and rigorous analysis of critical phenomena. The central question is whether the natural limiting diffusion on the critical percolation cluster of such a lattice possesses a spectral dimension of $4/3$ .

Methodology
The author builds upon the percolation theory and random walk convergence results established in their previous works (IGS I and IGS II). The methodology involves a multi-stage construction of scaling limits and the analysis of geometric and probabilistic exponents:

Chemical Geometry Construction: The author constructs the Gromov–Hausdorff–Prokhorov (GHP) scaling limit of the full terminal-connected critical cluster using the chemical metric (intrinsic graph distance along open edges). This replaces the ambient scale of the rule graph with an intrinsic growth factor, $\ell^{\text{ch}}_C$ . They identify the Minkowski dimension of this limit, denoted $\text{dim}_B(C, d_{\text{ch}})$ , with the chemical mass exponent derived from the spectral radius of the mass-transfer matrix, $\rho(M_C)$ .
Resistance Exponent Analysis: The paper proves the existence of a deterministic resistance growth exponent, $\lambda_C > 1$ , such that the effective resistance between terminals scales as $\lambda_C^n$ . This is achieved by analyzing the electrical functional on the substitution networks and establishing the convergence of normalized resistances to a stationary random field.
Diffusion Limit and Spectral Dimension: By combining the chemical metric, the limiting mass measure $\mu$ , and the limiting resistance form, the author constructs a diffusion process as the scaling limit of simple random walks. They derive the pointwise spectral dimension, $\text{dim}^{(N)}_S(C:x)$ , in terms of the chemical Minkowski dimension and the resistance dimension ( $\text{dim}_R(C, d_{\text{ch}}) = \log \lambda_C / \log \ell^{\text{ch}}_C$ ).
Inequality Proofs: The core of the disproof involves establishing strict inequalities between the geometric growth factors ( $\ell^{\text{ch}}_C$ , $\lambda_C$ , and $\rho(M_C)$ ) that prevent the spectral dimension from equalling $4/3$ except in a narrow, unproven region.

Key Contributions and Results

Existence of Scaling Limits: The paper rigorously constructs the GHP scaling limit of the critical percolation cluster for every classical EIGS hierarchical lattice. It proves that the chemical Minkowski dimension is almost surely given by $\frac{\log \rho(M_C)}{\log \ell^{\text{ch}}_C}$ .
Resistance Growth: It resolves the resistance-growth question for all classical EIGS lattices (extending previous results on the Diamond Hierarchical Lattice), proving the almost-sure existence of the limit $\lim_{n \to \infty} \frac{1}{n} \log R_{C_n}(v_+, v_-) = \log \lambda_C$ .
Spectral Dimension Formula: The author derives an explicit formula for the typical spectral dimension of the limiting diffusion:
$\text{dim}^{(N)}_S(C) = \frac{2 \text{dim}_B(C, d_{\text{ch}})}{\text{dim}_B(C, d_{\text{ch}}) + \text{dim}_R(C, d_{\text{ch}})}$
This confirms that the spectral dimension is determined by the interplay between the mass growth and resistance growth.
Failure of the Alexander–Orbach Conjecture: The main theorem (Theorem 5.4) states that for every classical EIGS rule, if the pair of dimensions $(\text{dim}_B, \text{dim}_R)$ $(dim_{B}, dim_{R})$ falls outside the region $(4/3, 2) \times (2/3, 1)$ $(4/3, 2) \times (2/3, 1)$ , the spectral dimension is strictly not $4/3$ $4/3$ .
- The author proves that $\text{dim}_R < 1$ and $\text{dim}_B + \text{dim}_R > 2$ for all classical rules.
- These inequalities imply that the equality $\text{dim}^{(N)}_S(C) = 4/3$ (which requires $\text{dim}_B = 2\text{dim}_R$ ) can only potentially hold if $\text{dim}_B \in (4/3, 2)$ and $\text{dim}_R \in (2/3, 1)$ .
Density of Spectral Dimensions: The paper demonstrates that the set of typical spectral dimensions achievable by classical EIGS lattices is dense in the interval $(1, 2)$ . This implies that while values arbitrarily close to $4/3$ can be approximated, no finite rule has been found that yields exactly $4/3$ .
Formalization: The proofs for Sections 2–5 and the Appendices have been formally verified using Lean 4 and mathlib, ensuring logical rigor without additional axioms beyond standard set theory.

Significance and Claims
The paper claims to provide a rigorous disproof of the Alexander–Orbach conjecture for the broad class of classical EIGS hierarchical lattices, except for a narrow, unverified region in the parameter space. The author explicitly states that despite the density of spectral dimensions in $(1, 2)$ , they have not found a single example where the dimension is exactly $4/3$ . Consequently, they propose Conjecture 5.7: that the Alexander–Orbach conjecture fails on every classical EIGS hierarchical lattice.

The work is significant because it moves beyond numerical estimates and heuristic scaling arguments to provide exact, rigorous bounds on the spectral dimension for a wide class of fractal structures. It clarifies the relationship between resistance, mass, and diffusion on critical percolation clusters, showing that the "universality" of the $4/3$ value does not hold in these hierarchical settings. The paper maintains a modest tone regarding the remaining open region, acknowledging that excluding the equality within the narrow interval $(4/3, 2) \times (2/3, 1)$ remains difficult, but asserts that the conjecture fails outside this specific window.

Technical Summary: Iterated Graph Systems (III): Alexander-Orbach conjecture on hierarchical lattices

More like this