Strong-disorder expansion of the root-averaged density of states for the Anderson model on the Bethe lattice

arXiv:2605.00478 · math-ph, math.MP · Submitted 2026-05-01 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematical Physics. The day in mathematical physics: rigorous models of physical systems.

Kai: We're looking at the paper: "Strong-disorder expansion of the root-averaged density of states for the Anderson model on the Bethe lattice".

Mira: Here's the gist of the paper. We study the root-averaged density of states for the Anderson model on the Bethe lattice in the strong-disorder regime.

Kai: This paper derives an explicit finite-order strong-disorder expansion of this density of states measure, which is significant because it provides a combinatorial description of the spectral properties in this challenging physical regime.

Mira: The Core Methodology The study relies on combining the random-walk expansion on the tree with a complex-analytic argument for single-site Stieltjes transforms.

Kai: The authors assume that the singlesite distribution has compact support and has a locally analytic density on an interval I containing a given interval I♯.

Title and authors: Mira: So, we’re looking at this paper titled "Strong-disorder expansion of the root-averaged density of states for the Anderson model on the Bethe lattice." It sounds really technical, but essentially they are taking a complex problem—the spectral properties of a disordered system—and finding an explicit way to describe it when disorder is very strong.

Kai: That’s right, Mira. And what’s interesting about this specific paper is that they use the Bethe lattice, which is a tree structure. This structure allows them to do something combinatorial with the random walks on the tree to get this expansion for the density of states measure.

Lev: From my side, when I hear "strong-disorder expansion," I immediately think about how much overhead this kind of calculation would actually take on real hardware. If we were trying to simulate this, we'd be looking at a lot of these terms multiplying together, and the complexity really lands somewhere in the error correction part.

Mira: Exactly, Lev. The point they are making is that they have a finite-order expansion for how the density of states behaves in that strong disorder regime. It provides a combinatorial description of what’s happening spectrally without needing to solve the full Hamiltonian directly for every single configuration of disorder.

Kai: So, instead of just getting some general statement about the spectrum, they are giving us a specific formula—a series in terms of lambda—that tells us how the density of states changes as we scale our energy window up by lambda. It’s about finding an explicit coefficient structure for this spectral behavior.

Lev: And that's where it gets tricky for simulation. If you want to run this on actual hardware, you need those coefficients, right? You need to know exactly what the leading terms are before you can even think about the corrections.

The paper's summary: Kai: Okay, so looking at the summary of "Strong-disorder expansion of the root-averaged density of states for the Anderson model on the Bethe lattice," they explain their main technique: they combine a random-walk expansion on trees with a complex analytic argument for single-site Stieltjes transforms.

Mira: The key assumption they make is that the single-site distribution—that’s how disorder is distributed at one site—has compact support and has a locally analytic density on some interval I, which must contain another interval I sharp. That local analyticity assumption is what lets them do the holomorphic continuation across that whole interval.

Lev: From an engineering standpoint, that assumption about the single-site distribution being locally analytic is a big hurdle for me. In real materials, especially at strong disorder, we often deal with distributions that aren't perfectly analytic in the way they need to be for these complex analysis tools to work smoothly across the whole support I.

Kai: They state that if you assume that local analyticity holds, then for a sufficiently large lambda, the scaled averaged diagonal resolvent has a holomorphic continuation into a complex neighborhood of I. This means we can analytically continue this important quantity even when we are dealing with very strong disorder effects.

Mira: And they give us the structure of that continuation, which is an expansion in terms of lambda and some coefficient functions Mn. These coefficient functions are defined by sums over closed walks starting and ending at the root of the tree.

Lev: Closed walks on a Bethe lattice are always even length because it’s bipartite, so all those odd-length contributions just vanish. That simplifies things a lot for the combinatorial part, but we still have to compute these sums over all possible subtrees that could form those walks.

Kai: So, the actual density of states measure is derived from this resolvent using the Stieltjes inversion formula. They say that when you look at the scaled energy window E equals lambda times xi, the density of states measure becomes absolutely continuous on that window and has a finite-order strong-disorder expansion there.

Mira: And they go as far as explicitly computing the leading coefficient, which they identify with the local density of the single-site distribution itself. This is a really neat result because it tells us that in this strong disorder limit, the diagonal random potential dominates everything else in determining the spectral shape.

The paper's improvements: Kai: Now, let’s talk about what they suggest as improvements or what their results actually change for how we think about these systems. They focus heavily on deriving an explicit finite-order strong-disorder expansion of this density of states measure.

Mira: The improvement is really in the explicitness. Instead of just knowing that an expansion exists, they provide a concrete formula, specifically for the uniform single-site distribution case. For that case, they get a two-term strong-disorder expansion for the density of states in terms of lambda.

Lev: A two-term expansion is helpful because it gives you something you can actually calculate and check against simulations. But I have to ask, how reliable is this approximation when we move beyond that uniform distribution to more realistic disorder distributions?

Kai: The authors say that while the two-term formula works for the uniform distribution, it exhibits singularities at the endpoints of that interval I, specifically at plus a and minus a. This means you can’t just use this formula everywhere without being careful about those boundaries.

Mira: The leading term they find for the uniform case is one over two times a lambda to the power of one <ref:2605.00478#pg3>. They also get a first correction term, which involves s two and s one, where s ones and s twos are related to some properties of the single-site distribution’s density.

Lev: If we were trying to implement this on hardware, those s terms would require very precise measurements of the disorder statistics. We wouldn't be able to just guess those values; they would dictate the fidelity of our entire prediction for that spectral region.

Kai: So, the practical implication is that this method gives us a way to calculate the averaged local density of states at a typical vertex in strong disorder regimes, using this explicit two-term expansion derived from the uniform case as a baseline.

Conclusion: Mira: To wrap up, what we’ve seen here with the paper "Strong-disorder expansion of the root-averaged density of states for the Anderson model on the Bethe lattice" is that they have successfully derived an explicit coefficient-level strong-disorder expansion.

Kai: The main implication is that because they found that the leading coefficient equals the local density of the single-site distribution, it confirms that in strong disorder, we really should focus our attention on how that single-site law behaves spectrally. The hopping term seems to only give lower order corrections when you scale things up by lambda.

Lev: For me, it means if we want to build error correction codes for these disordered systems, we don't need to worry about the full complexity immediately; we can focus our resources on precisely characterizing that leading single-site density term first.

Mira: And for the uniform case specifically, they gave us this formula: n lambda times xi equals one over two a lambda to the power of minus one minus q plus one over two a times a squared minus xi squared times lambda to the power of minus three plus Olambda to the power of minus five as lambda goes to infinity.

Kai: That formula shows how fast the density of states scales with that energy window size, and it confirms those singularities at plus a and minus a, which we have to be careful about.

Lev: It’s a solid result for understanding the scaling behavior under extreme disorder, but we still need to figure out how to map these continuous analytical results onto the discrete structure of physical quantum hardware.

Mira: And that’s where it leaves us. We’ve got this explicit expansion, and now we look forward to seeing how this combinatorial approach helps us predict spectral behavior in even more complex, realistic disorder landscapes.

Department of Information Technology, Faculty of Engineering, Tohoku Gakuin University

math-ph, math.MP

Submitted: 2026-05-01

Updated: 2026-11-04

Comments: 19 pages, 1 figure

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: We study the root-averaged density of states for the Anderson model on the Bethe lattice in the strong-disorder regime.

Key concepts

Anderson Model on Bethe Lattice
This is a mathematical model used to study how electrons behave in materials with random potential (disorder). The Bethe lattice is a specific, highly connected structure of points where these electrons move.
Strong-Disorder Regime
This regime occurs when the randomness in the material's potential is very high. In this situation, the disorder dominates the system's physics, meaning the random potential has a much larger effect than normal hopping terms.
Stieltjes Transform
This is a mathematical tool used to represent spectral data (like density of states) in a way that allows for analytic continuation. It helps map complex spectral information onto functions that can be manipulated using calculus.

Terminology

Summary

We study the root-averaged density of states for the Anderson model on the Bethe lattice in the strong-disorder regime. p. 0

Where the paper says it

Strong-disorder expansion of the root-averaged density of states for the Anderson model on the Bethe lattice Masahiro Kaminaga∗ Department of Information Technology, Faculty of Engineering, Tohoku Gakuin University, Sendai, Miyagi 984-8588, Japan Abstract We study the root-averaged density of states for the Anderson model on the Bethe lattice in the strong-disorder regime. Here the density of states means the root-averaged spectral measure, not a finite-volume eigenvalue counting limit.

This paper derives an explicit finite-order strong-disorder expansion of this density of states measure, which is significant because it provides a combinatorial description of the spectral properties in this challenging physical regime. p. 1
Where the paper says it

Instead, we derive an explicit finite-order strong-disorder expansion of the rootaveraged density of states in the scaled energy window E = λξ. The coefficients are finite combinatorial sums over closed walks on the Bethe lattice, or equivalently over their occupation profiles.

The Core Methodology

The study relies on combining the random-walk expansion on the tree with a complex-analytic argument for single-site Stieltjes transforms. The authors assume that the singlesite distribution has compact support and has a locally analytic density on an interval I containing a given interval I. p. 0

Where the paper says it

Here the density of states means the root-averaged spectral measure, not a finite-volume eigenvalue counting limit. We assume that the singlesite distribution has compact support and has a locally analytic density on an interval I ♯ containing a given interval I.

This local analyticity assumption is crucial for obtaining a holomorphic continuation of the single-site Stieltjes transforms across the interval I. p. 8
Where the paper says it

, mn+1(γ)) proves the assertion. 4 Holomorphic continuation of the single-site transforms The key point is that the local analyticity of ρ enables us to continue each sk across the interval I.

The Expansion Structure

The main result states that for sufficiently large λ, the scaled averaged diagonal resolvent has a holomorphic continuation to a complex neighborhood of I. This continuation is given by an expansion of the form mλ(z) = X∞ n=0 λ−n−1Mn(z/λ) + RN,λ(z), z/λ ∈ omegaδ(I), where Mn are holomorphic on omegaδ(I). p. 2

Where the paper says it

Our main result says that, under the above local analyticity assumption on ρ, the averaged diagonal resolvent mλ(z) := Ehδ0,(Hλ,ω − z) −1 δ0i, Im z > 0, admits, after the scaling z = λζ, a holomorphic continuation in ζ to Ωδ(I) for all sufficiently large λ. More precisely, for each N ≥ 0 we obtain an expansion of the form mλ(z) = X N n=0 λ −n−1Mn(z/λ) + RN,λ(z), z/λ ∈ Ωδ(I), where the coefficient functions Mn are holomorphic on Ωδ(I) and are given by finite sums over closed walks of length n starting and ending at the root.

The coefficient functions Mn are given by finite sums over closed walks of length n starting and ending at the root. p. 2
Where the paper says it

Our main result says that, under the above local analyticity assumption on ρ, the averaged diagonal resolvent mλ(z) := Ehδ0,(Hλ,ω − z) −1 δ0i, Im z > 0, admits, after the scaling z = λζ, a holomorphic continuation in ζ to Ωδ(I) for all sufficiently large λ. More precisely, for each N ≥ 0 we obtain an expansion of the form mλ(z) = X N n=0 λ −n−1Mn(z/λ) + RN,λ(z), z/λ ∈ Ωδ(I), where the coefficient functions Mn are holomorphic on Ωδ(I) and are given by finite sums over closed walks of length n starting and ending at the root.

Furthermore, all odd coefficients vanish because the Bethe lattice is bipartite, meaning every closed walk has even length. p. 2
Where the paper says it

Since Tq is bipartite, every closed walk starting and ending at the root has even length, and hence all odd coefficients vanish. Moreover, because the graph is a tree, the contribution of a closed walk is encoded by the finite subtree it visits together with the occupation numbers of its vertices.

Density of States Derivation

The density of states measure nλ(E) is derived from the Stieltjes transform mλ(z) via the formula dNλ is absolutely continuous on λI and its density is given by nλ(E) = 1/π Im λ(E), E ∈ λI. The leading coefficient a0(ξ) is equal to the local density of the single-site distribution ρ(ξ). p. 0

Where the paper says it

By the Stieltjes inversion formula, the root-averaged density of states measure is absolutely continuous on the scaled energy window λI, and its density is real analytic and has a finite-order strong-disorder expansion there. In the scaled form E = λξ, the leading coefficient is the local density of the single-site distribution.

The first nonzero correction term is explicitly computed for the uniform single-site distribution, yielding a second term involving s2 and s1. p. 0
Where the paper says it

All odd coefficients vanish, and the higher coefficients are finite sums determined by occupation profiles of short closed walks on the tree. For the uniform single-site distribution, we compute the first nonzero correction term explicitly.

Explicit Results in the Uniform Case

For the uniform distribution on an interval, the authors obtain an explicit two-term strong-disorder expansion of the density of states in terms of λ. The resulting formula is nλ(λξ) = 1/2a λ−1 − (q + 1)/2a(a squared − ξ 2) λ−3 + O(λ−5) as λ → ∞. p. 0

Where the paper says it

By the Stieltjes inversion formula, the root-averaged density of states measure is absolutely continuous on the scaled energy window λI, and its density is real analytic and has a finite-order strong-disorder expansion there. In the scaled form E = λξ, the leading coefficient is the local density of the single-site distribution.

This expansion is uniform on fixed closed subintervals of I but exhibits singularities at the endpoints ±a. p. 2
Where the paper says it

More precisely, for each N ≥ 0 we obtain an expansion of the form mλ(z) = X N n=0 λ −n−1Mn(z/λ) + RN,λ(z), z/λ ∈ Ωδ(I), where the coefficient functions Mn are holomorphic on Ωδ(I) and are given by finite sums over closed walks of length n starting and ending at the root. The remainder satisfies the uniform bound |RN,λ(z)| ≤ CN,δλ −N−2 .

The leading term a0(ξ) is found to be 1/2a. p. 0
Where the paper says it

By the Stieltjes inversion formula, the root-averaged density of states measure is absolutely continuous on the scaled energy window λI, and its density is real analytic and has a finite-order strong-disorder expansion there. In the scaled form E = λξ, the leading coefficient is the local density of the single-site distribution.

Conclusion and Interpretation

The paper concludes by showing that the leading term of the root-averaged density of states on the scale E = λξ is simply the density of the single-site law. This confirms that for strong disorder, the diagonal random potential is dominant and hopping term gives only lower order corrections. p. 5

Where the paper says it

The interval λI moves linearly with the disorder strength. This is natural in the strong-disorder region, because the diagonal part of Hλ,ω is of size λ, while the hopping term is of size 1.

The method successfully provides an explicit coefficient-level strong-disorder expansion which has not appeared in the literature before. p. 1
Where the paper says it

The coefficients are finite combinatorial sums over closed walks on the Bethe lattice, or equivalently over their occupation profiles. To the best of our knowledge, no explicit coefficient-level strong-disorder expansion for the root-averaged density of states on the Bethe lattice has appeared in the literature.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Strong disorder expansion for Anderson model spectral measures allows AI to predict density of states in highly disordered systems by using a finite-order strong-disorder expansion derived from random walk expansions on trees.

  2. The explicit two term strong disorder expansion for the uniform distribution provides a method for AI to calculate the averaged local density of states at a typical vertex, which is useful for simulating material properties under extreme disorder.

  3. The result that the leading coefficient is the local density of the single-site distribution enables AI systems to quickly estimate spectral behavior in strong disorder regimes by focusing on the single-site law's density.

Abstract

We study the root-averaged density of states for the Anderson model on the Bethe lattice in the strong-disorder regime. Here the density of states means the root-averaged spectral measure, not a finite-volume eigenvalue counting limit. We assume that the single-site distribution has compact support and has a locally analytic density on an interval I containing a given interval I. Combining the random-walk expansion on the tree with a complex-analytic argument for the single-site Stieltjes transforms, we prove that the scaled averaged diagonal resolvent has a holomorphic continuation to a complex neighborhood of I for all sufficiently large λ. By the Stieltjes inversion formula, the root-averaged density of states measure is absolutely continuous on the scaled energy window λI, and its density is real analytic and has a finite-order strong-disorder expansion there. In the scaled form E=λξ, the leading coefficient is the local density of the single-site distribution. All odd coefficients vanish, and the higher coefficients are finite sums determined by occupation profiles of short closed walks on the tree. For the uniform single-site distribution, we compute the first nonzero correction term explicitly.

Related papers