🔢 mathematics

Disorder on the hyperbolic square lattice I: Anderson delocalization and absolutely continuous spectrum

This paper establishes the existence of absolutely continuous spectrum and the absence of singular spectrum for the Anderson model on a specific hyperbolic square lattice at weak disorder and under certain bounded density conditions, demonstrating Anderson delocalization in this geometric setting.

Original authors: Simon Becker, Izak Oltman

Published 2026-09-18

📖 7 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

In the study of how electricity moves through materials, scientists often look at what happens when the path is not perfectly smooth. Imagine a crystal where atoms are arranged in a perfect grid; electrons can flow through it easily, like water in a clear pipe. But in the real world, materials are messy. Atoms might be slightly heavier or lighter than their neighbors, or impurities might be scattered randomly throughout the structure. This disorder acts like rocks in that pipe, potentially blocking the flow entirely. This phenomenon, known as Anderson localization, suggests that if the disorder is strong enough, electrons get trapped in one spot and cannot move, turning a conductor into an insulator. For decades, researchers have known that on a simple, flat grid, this trapping happens no matter how weak the disorder is. However, the rules change when the geometry of the space itself becomes more complex.

A new study by Simon Becker and Izak Oltman investigates this question on a specific, unusual shape: a hyperbolic square lattice. Unlike the flat grid of a standard computer chip, this structure is built like a saddle, curving away from itself in every direction. In this world, the space expands exponentially as you move outward; there are simply more places to go the further you travel. The researchers combined this hyperbolic shape with a flat, grid-like extension to create a hybrid model. They asked a precise question: if you introduce a small amount of randomness to this specific, curved structure, do the electrons get stuck, or can they still flow freely? Their work provides a definitive answer for weak disorder regarding the spectral properties, proving that on this particular lattice, if the disorder is sufficiently small, it is not enough to stop the flow in the spectral sense. Instead, the electrons remain in a state where they possess an absolutely continuous spectrum, indicating a lack of singular, trapped states, though the study does not establish dynamical transport.

The researchers focused on a mathematical model where the "grid" is formed by connecting squares in a way that five of them meet at every single corner. On a standard flat surface, only four squares can meet at a point without overlapping. By forcing five to meet, the structure must curve, creating the hyperbolic geometry. The team simulated electrons moving across this lattice, adding a random variable to each point to represent the disorder. They found that when the disorder is weak, specifically below certain strict thresholds, the electrons do not get trapped. Instead, they found a specific range of energies where the electrons move freely, and the probability of finding them at any given spot is smooth and continuous, rather than jagged or concentrated in isolated points. This is a significant departure from what happens on a flat grid, where even the slightest disorder eventually stops the flow.

The study goes further by proving that this free-flowing state is not just a possibility but a certainty for a specific set of conditions. The authors demonstrated that for a specific type of random distribution, the electrons are not only free to move but that the system is free of "singular" behavior, which would indicate trapping or chaotic localization, but only on a specific deterministic set of energy levels with positive measure. They showed that this holds true even when the lattice is combined with a standard flat grid in multiple dimensions. The proof relies on a clever counting method. Because the hyperbolic space grows so rapidly, there are an enormous number of different paths an electron can take to get from point A to point B. The researchers showed that this sheer abundance of routes compensates for the random obstacles. Even if one path is blocked, the sheer number of alternative routes ensures that the electron can still find a way through, maintaining its ability to conduct in the spectral sense.

The findings are robust and apply to a variety of scenarios, but strictly within the bounds of weak disorder. The team proved that as long as the disorder is weak enough, there is a guaranteed set of energy levels where the material exhibits an absolutely continuous spectrum. They also showed that this result is stable; even if the random disorder is slightly changed or perturbed, the free-flowing state remains intact. This suggests that the geometry of the material itself is the dominant factor in determining whether electricity can flow. If the space expands fast enough, as it does in this hyperbolic lattice, the system resists the tendency to localize. The work does not just suggest this behavior; it provides a rigorous mathematical proof that such a state exists and is stable under the conditions they described.

This research bridges a gap between two very different worlds of physics. On one side are flat, Euclidean grids where disorder always wins eventually. On the other are tree-like structures where disorder can be overcome easily. The hyperbolic square lattice sits in between, possessing the loops and cycles of a flat grid but the rapid expansion of a tree. The study confirms that this specific combination allows for a unique behavior: the electrons can delocalize, or spread out, despite the presence of randomness. The authors did not rely on computer simulations to guess the outcome; they constructed a logical argument that holds up under strict mathematical scrutiny. They identified the exact conditions under which the electrons remain free and showed that these conditions are met for a specific range of weak disorder strengths.

The implications of this work extend beyond just this specific lattice. It offers a new perspective on how the shape of a material influences its electrical properties. By understanding that geometry can protect against localization, scientists might one day design materials that remain conductive even when they are imperfect or disordered. The study also clarifies the boundary between order and chaos in quantum systems. It shows that the transition from a conductor to an insulator is not just about how much disorder is present, but also about the underlying structure of the space the particles inhabit. In this curved, expanding world, the rules of localization are rewritten, allowing for a state of matter that is both disordered and capable of supporting an absolutely continuous spectrum.

The authors also explored how these results hold up when the randomness is not perfectly uniform. They found that even if the disorder is slightly different from the ideal case, or if it is cut off at certain limits, the free-flowing state persists. This stability is crucial for real-world applications, as perfect conditions are rare. The proof covers cases where the disorder is bounded and follows specific statistical patterns, ensuring that the findings are not just theoretical curiosities but describe a physically realizable state. The work stands as a complete and verified explanation of why, in this specific geometric setting with weak disorder, the electrons refuse to be trapped.

In the end, the paper delivers a clear message about the power of geometry in physics. It shows that by changing the way space is connected, one can fundamentally alter how particles move through it. The hyperbolic square lattice acts as a shield against the trapping effects of disorder, allowing electrons to maintain their freedom in terms of spectral properties. This discovery adds a new chapter to our understanding of quantum transport, proving that in the right shape, even a messy world can support a smooth, uninterrupted flow.

Technical Summary: Disorder on the Hyperbolic Square Lattice I: Anderson Delocalization and Absolutely Continuous Spectrum

1. Problem Statement and Context
The paper investigates the spectral properties of the Anderson model $H_{\lambda, \omega} = A + \lambda \text{diag}(\omega_x)$ on the graph $X_d = \mathbb{Z}^d \square G_{4,5}$ . Here, $A$ is the unit-weight adjacency operator, $\lambda > 0$ is the disorder strength, and $\omega_x$ are independent and identically distributed (i.i.d.) random variables with a bounded probability density $\rho$ . The graph $G_{4,5}$ is the regular hyperbolic square lattice where five squares meet at each vertex (contrasted with the Euclidean $\mathbb{Z}^2 = G_{4,4}$ where four squares meet).

The central question is how the geometry of the underlying graph influences Anderson localization. It is established that on $\mathbb{Z}$ , localization occurs at all disorder strengths, while on regular trees (Bethe lattices), absolutely continuous (a.c.) spectrum exists at weak disorder. The status of $\mathbb{Z}^2$ remains open. Hyperbolic lattices like $G_{4,5}$ combine the exponential volume growth of trees with the polygonal loops of Euclidean lattices. The authors aim to determine whether the presence of loops in $G_{4,5}$ destroys the delocalization observed on trees or if a.c. spectrum persists at weak disorder.

2. Methodology
The proof strategy builds upon the resonance mechanism developed by Aizenman and Warzel for trees, adapted to handle the non-trivial loops present in $G_{4,5}$ .

Resonance and Tunneling: The core idea is that the exponential volume growth provides a sufficient number of distant sites to compensate for the low probability of any single site resonating with a fixed origin. On trees, deleting a vertex separates branches, allowing exact recursive identities for the resolvent. On $G_{4,5}$ , square loops connect these branches. The authors control the interference between paths by analyzing the "tunneling amplitude" $\tau_{xy}(z)$ , which measures the coupling between two sites $x$ and $y$ in the presence of the rest of the graph.
Logarithmic Estimates: The authors utilize complex analysis techniques, specifically mapping the spectral complement to a disk and applying Jensen's formula. This allows them to bound the energy average of $\log |\tau_{xy}|$ from below by the number of shortest paths ( $N_{xy}$ ) between vertices, minus a term related to the distance. This establishes that the tunneling decay is slower than the volume growth.
Resampling and Decoupling: To handle correlations between resonance events at different sites, the authors employ a finite resampling technique. They approximate exterior events by cylinder events and resample the potentials on a finite block. Using the rank-one resolvent formula, they show that this resampling changes the logarithmic tunneling amplitude by a sub-linear amount ( $o(R)$ ) with high probability, effectively decoupling the resonance probabilities.
Poincaré Inequality and Second Moments: A Poincaré inequality on the probability space is used to bound the variance of the logarithmic response. If the boundary imaginary part of the Green's function vanished (implying singular spectrum), the existence of distant resonances would force a fluctuation in the root response that contradicts the uniform logarithmic bounds derived from the Poincaré inequality.
Spectral Averaging: To rule out singular spectrum on the identified energy sets, the authors condition on all potentials except one and apply spectral averaging. This ensures that the local spectral measure has no singular mass on the deterministic energy sets where the averaged density is positive.
Cauchy Identity and Interval Results: For stronger results (purely a.c. spectrum on intervals), the authors utilize the Cauchy distribution's property that averaging over the potential is equivalent to adding a fixed imaginary potential. This allows for strict finite-block inequalities comparing the logarithmic propagation cost with endpoint entropy, which persists under perturbations of the distribution.

3. Key Contributions and Results

Theorem 1.1 (Weak Disorder Delocalization): For the graph $X_d = \mathbb{Z}^d \square G_{4,5}$ , there exists a deterministic Borel set $D_{d,\lambda}$ of positive Lebesgue measure within the spectrum of the free operator such that, for sufficiently weak disorder ( $\lambda \le 10^{-4}$ for $d=0$ and $\lambda \le 1/5$ for $d \ge 1$ ), the local spectral measures have a positive absolutely continuous density almost everywhere on $D_{d,\lambda}$ and no singular part. This proves the existence of a non-zero absolutely continuous spectral subspace almost surely.
Theorem 1.2 (Larger Disorder Range): If the single-site density $\rho$ is bounded below by a positive constant on the entire interval $[-1, 1]$ , the result holds for a larger range of disorder, specifically $0 < \lambda \le 2$ in every finite dimension.
Theorem 1.3 (Interval Results for Truncated Cauchy): For truncated Cauchy distributions, the authors prove that the spectrum is purely absolutely continuous on specific intervals (e.g., $[4.75, 4.9] \cup [-4.9, -4.75]$ for $d=0$ ) that lie outside the spectrum of the free operator. The local spectral densities are positive almost everywhere on these intervals.
Theorem 1.4 (Stability): The interval results are stable under perturbations of the distribution. If the density is sufficiently close in total variation to a truncated Cauchy distribution and bounded below on a necessary interval, the purely absolutely continuous spectrum persists on the same intervals.

4. Significance and Claims
The paper claims to establish the first rigorous proof of Anderson delocalization (in the spectral sense) on a regular hyperbolic lattice containing non-trivial loops. While numerical studies have suggested Anderson transitions on hyperbolic lattices, this work provides a mathematical proof that the interplay between exponential volume growth and loop structure allows for the existence of extended states (a.c. spectrum) at weak disorder.

The authors explicitly state that their results establish delocalization in the spectral sense (positive a.c. density and absence of singular spectrum) but do not establish dynamical transport. The work bridges the gap between tree-like structures (where delocalization is known) and Euclidean lattices (where it is conjectured but unproven), demonstrating that the "hyperbolic" nature of $G_{4,5}$ is sufficient to support a.c. spectrum despite the presence of square loops.

The paper also highlights the role of artificial intelligence (ChatGPT) in the development process, noting that LLM assistance helped identify a route to sharpen the argument from the more hyperbolic $G_{8,8}$ to the geometrically more Euclidean $G_{4,5}$ , and suggested potential extensions to other non-amenable geometries, though these are not pursued in the current work. The authors take full responsibility for the final mathematical arguments and verification.

More like this