Ergodic Schrodinger operators on the Bethe lattice and a modified Thouless formula

This paper establishes a modified Thouless formula for ergodic Schrödinger operators on the Bethe lattice that relates the density of states to the Lyapunov exponent via a nontrivial remainder term for connectivity $\kappa \geq 2$, while also clarifying the role of the lattice's automorphism group and the multiparameter noncommutative ergodic theorem in evaluating Green's function limits.

The Gist

The Big Picture: A New Rule for a Weird World

Imagine you are trying to understand how electricity (or a wave) moves through a material. In the real world, we often model materials as a grid, like a checkerboard or a city block. In math, this is called a lattice.

For a long time, physicists have had a perfect "rulebook" (called the Thouless Formula) for predicting how waves behave on a simple, straight line (like a 1D wire). This rulebook connects two things:

1. The Density of States: How many "parking spots" are available for the waves to sit in.
2. The Lyapunov Exponent: How quickly the waves die out or get lost as they travel.

The Problem: This paper asks: What happens if the material isn't a straight line, but a giant, branching tree?

In math, this tree is called the Bethe Lattice. It's an infinite tree where every branch splits into more branches, forever. It looks like a coral reef or a fractal fern. The authors, Peter Hislop and Christoph Marx, discovered that the old rulebook doesn't work perfectly here. They had to write a Modified Rulebook with a new "correction term" to make the math work.

---

The Setting: The Infinite Tree (The Bethe Lattice)

Think of the Bethe Lattice as a massive, infinite family tree.

• The Root: There is one starting point (the origin).
• The Branches: From the root, there are $\kappa + 1$ branches. From every other point on the tree, there are $\kappa$ new branches going forward.
• The Geometry: This is where it gets weird. On a flat grid (like a city), if you walk 100 blocks, the "surface" of your walk is small compared to the total area. But on this tree, if you walk 100 steps, the number of new branches you could have taken at every step is huge. The "surface" grows exponentially.

The Analogy: Imagine walking down a hallway (the straight line). You only have two walls touching you. Now, imagine walking down a hallway where, at every step, the hallway splits into 10 new hallways, and every one of those splits again. You are surrounded by possibilities on every side, not just the front and back. This is the "hyperbolic geometry" of the Bethe Lattice.

The Discovery: The "Remainder Term"

The authors proved that the relationship between the "parking spots" (Density of States) and the "decay rate" (Lyapunov Exponent) is almost the same as the old rulebook, but with a twist.

The Old Rule (Straight Line):
$$\text{Decay Rate} = \text{Average Parking Spots}$$
(It's a perfect balance.)

The New Rule (The Tree):
$$\text{Decay Rate} = \text{Average Parking Spots} + \mathbf{R}$$
(There is an extra term, R, called the Remainder Term.)

Why does this extra term exist?

This is the most creative part of the paper.

• On a Straight Line: When you walk down a path, you are only connected to the "outside world" at the very beginning and the very end of your path. It's like walking through a tunnel with two doors. As your path gets longer, the influence of those two doors becomes tiny compared to the length of the tunnel. The extra term vanishes.
• On the Tree: When you walk down a path on the tree, you are connected to the "outside world" at every single step. At every vertex (branching point), there are other trees growing off to the side.
  • The Metaphor: Imagine walking down a path in a forest. On a straight path, the forest is just on your left and right. On the Bethe Lattice, every time you take a step, a whole new forest sprouts from your shoulder.
  • Because the path is constantly "leaking" energy into these side forests, the math doesn't balance out perfectly. The "leakage" from the interior of the path adds up. This accumulated leakage is the Remainder Term ($R$).

The "Non-Commutative" Puzzle

To prove this, the authors had to deal with a tricky math problem involving shifts.

• On a straight line, if you move "Right" then "Up," it's the same as moving "Up" then "Right." (They commute).
• On the Tree, the order matters! If you move "Right" then "Up," you end up in a different spot than if you move "Up" then "Right."

The authors had to invent a new way to navigate this tree, creating a specific set of "moves" (automorphisms) to map the tree onto itself. They showed that even though the order of moves matters, you can still find a consistent path to calculate the average behavior of the waves. This was necessary to prove that the "Remainder Term" is real and not just a math error.

The Conclusion: Why It Matters

The authors didn't just find a formula; they proved that the "Remainder Term" is non-zero whenever the tree has branches ($\kappa \ge 2$).

• If $\kappa = 1$: The tree is just a straight line. The remainder is zero. The old rule works.
• If $\kappa \ge 2$: The tree branches. The remainder is not zero.

The Takeaway:
This paper tells us that the geometry of a material fundamentally changes how waves behave. You cannot simply take the rules for a straight wire and apply them to a branching structure. The "side branches" of the tree constantly interact with the path, creating a permanent, measurable effect that must be accounted for.

In short: On a tree, you can't ignore the neighbors you pass along the way; they change the destination.

Technical Summary

1. Problem Statement

The paper addresses a fundamental gap in the spectral theory of random (ergodic) Schrödinger operators on the Bethe lattice (infinite Cayley tree) compared to the standard one-dimensional lattice ($\mathbb{Z}$).

• Context: For ergodic Schrödinger operators on $\mathbb{Z}$, the Thouless formula relates the Lyapunov exponent ($L(z)$), which characterizes the exponential decay of Green's functions, to the integrated density of states (IDS) via the logarithmic potential of the density of states measure ($dnv$).
• The Challenge: The Bethe lattice possesses hyperbolic geometry. Unlike $\mathbb{Z}$, where the surface-to-volume ratio vanishes as the system size grows, the Bethe lattice has an exponential growth rate where the "surface" (boundary of a ball) is of the same order as the volume.
• The Question: Does the standard Thouless formula hold for the Bethe lattice? Previous work suggested the existence of a Lyapunov exponent, but the precise relationship between this exponent and the density of states remained unclear, particularly regarding whether a "remainder term" exists due to the lattice's geometry.

2. Methodology

The authors employ a combination of geometric group theory, functional analysis, and probabilistic ergodic theory.

A. Geometric Structure and Automorphisms

• Labeling and Labeling: The authors introduce a specific labeling scheme for vertices on the Bethe lattice with connectivity $\kappa \ge 2$. A vertex is represented by a tuple $(0, a_1, \dots, a_\ell)$, where $a_1 \in \{0, \dots, \kappa\}$ and $a_j \in \{0, \dots, \kappa-1\}$ for $j > 1$.
• Non-Commuting Shifts: They define two elementary graph automorphisms, $\tau_1$ (generalized translation/level shift) and $\tau_2$ (generalized rotation). Crucially, these maps do not commute.
• Generalized Shifts: Any vertex $x$ can be reached from the root $(0)$ via a non-commutative product of these generators. This allows the definition of a generalized shift $\tau_x$ mapping the root to $x$.
• Correction of Literature: The authors correct a previous result by Acosta and Klein regarding the representation of vertices as products of these automorphisms, highlighting the non-commutative nature which complicates the application of standard ergodic theorems.

B. Green Function Expansions

• Random Walk (RW) vs. Self-Avoiding Walk (SAW): The paper utilizes expansions of the resolvent (Green's function).
  • The RW-expansion sums over all walks.
  • The SAW-expansion (Feenberg expansion) sums only over self-avoiding walks (paths). On a tree, there is a unique path between any two vertices, simplifying the SAW expansion to a single term.
• Finite Volume Approximation: A key technical step involves proving that the Green's function of the infinite-volume operator can be approximated by finite-volume restrictions. Due to the hyperbolic geometry, the authors prove that one must enlarge the finite volume (e.g., from radius $L$ to $L + e(L)$ where $e(L) \to \infty$) to compensate for "surface-to-volume effects."

C. Ergodic Theory on Non-Commuting Groups

• Ergodic Family: They define ergodic Schrödinger operators using a family of measure-preserving transformations $T_x$ on a probability space, induced by the geometric shifts $\tau_x$.
• Macroscopically Representative (MRP) Paths: Because the shifts are non-commuting, standard multi-parameter ergodic theorems (like Zygmund's) do not directly apply to arbitrary paths. The authors define MRP paths as those along which the ergodic average of a function converges to its expectation. They show that paths with specific asymptotic behaviors (eventually constant exponents) are MRP.

3. Key Contributions and Results

A. The Modified Thouless Formula

The central result is Theorem 4.1, which establishes a modified Thouless formula for the Bethe lattice:
$$L(z) = \int_{\Sigma} \log |z - E| \, dn\nu(E) + R(z)$$
Where:

• $L(z)$ is the Lyapunov exponent.
• The integral term is the standard "Thouless-like" term involving the density of states measure $dn\nu$.
• $R(z)$ is a non-trivial remainder term that depends on the connectivity $\kappa$.

B. Properties of the Remainder Term $R(z)$

• Vanishing for $\kappa=1$: When $\kappa=1$ (the lattice $\mathbb{Z}$), the remainder term $R(z)$ vanishes, recovering the classical Thouless formula.
• Non-Vanishing for $\kappa \ge 2$: For the Bethe lattice, $R(z)$ is generally non-zero.
• Physical Interpretation: The remainder term arises from the coupling of the path to its surroundings.
  • On $\mathbb{Z}$, a path segment is coupled to the rest of the lattice only through its two endpoints. As the path length $L \to \infty$, the average contribution of these boundary terms decays to zero.
  • On the Bethe lattice ($\kappa \ge 2$), every interior vertex of a path is connected to $\kappa-1$ additional rooted trees. The number of these "side branches" grows linearly with the path length. Upon averaging, these contributions do not decay, resulting in a persistent remainder term.
• Bound: The authors prove the bound $|R(z)| \le C(z)(\kappa - 1)$.

C. Explicit Calculation for the Free Laplacian

In Section 4.2, the authors explicitly compute the remainder term for the free Laplacian (zero potential).

• They show that for the free Laplacian, the derivative of the Thouless-like term is non-zero for $\kappa \ge 2$.
• Since the Lyapunov exponent for the free Laplacian is constant ($\frac{1}{2}\log \kappa$) within the spectrum, the non-constant nature of the integral term necessitates a non-zero $R(z)$ to balance the equation.
• The explicit form derived is:
  $$R(z) = \frac{1}{2}\log \kappa + \frac{\kappa - 1}{4} \log[(\kappa + 1)^2 - z^2] + C_0(\kappa)$$

4. Significance

• Theoretical Breakthrough: This work resolves the long-standing question of the relationship between the Lyapunov exponent and the density of states on tree graphs. It demonstrates that the hyperbolic geometry fundamentally alters the spectral relations found in Euclidean lattices.
• Localization vs. Delocalization: The existence of a non-trivial remainder term provides a rigorous mathematical underpinning for the known phenomenon that random Schrödinger operators on the Bethe lattice exhibit absolutely continuous spectrum (delocalization) for weak disorder, unlike the Anderson localization seen on $\mathbb{Z}$. The "extra" degrees of freedom (the side branches) prevent the exponential decay required for localization.
• Methodological Rigor: The paper provides a rigorous framework for handling ergodic averages on non-commuting groups (the automorphism group of the tree), correcting previous literature and establishing the necessary conditions (MRP paths) for the existence of limits.
• Correction of Prior Work: By correcting the representation of vertices in terms of automorphisms (Proposition 2.1), the paper ensures the mathematical foundation for future studies on random operators on trees is sound.

In summary, Hislop and Marx prove that the standard Thouless formula is insufficient for the Bethe lattice due to its expansive geometry, introducing a necessary remainder term that quantifies the spectral influence of the lattice's branching structure.