Power-law banded random matrix ensemble as a model for quantum many-body Hamiltonians

This paper investigates the interpretation of the power-law banded random matrix ensemble as one-dimensional quantum many-body Hamiltonians by comparing labeling schemes and demonstrating how its single-particle phases correspond to distinct entanglement transitions, including the identification of an intermediate phase with volume-law scaling that deviates from the Page value.

The Gist

Imagine you have a giant, chaotic dance floor representing a quantum system. In this dance floor, every dancer is a particle, and their movements are governed by a set of rules called a Hamiltonian (essentially the system's energy blueprint).

Physicists usually try to understand these complex dance floors by comparing them to a "perfectly random" dance floor, where every dancer is equally likely to bump into any other dancer, no matter how far apart they are. This is the standard model (called GOE or GUE in physics jargon).

However, real physical systems are different. In the real world, dancers mostly bump into their neighbors. It's rare for someone in the corner to suddenly grab hands with someone on the opposite side of the room.

This paper explores a new, more realistic model called the Power-Law Banded Random Matrix (PLBRM). Think of this as a "structured" dance floor where the likelihood of two dancers interacting drops off like a power law:

• Close neighbors: Very likely to interact.
• Far neighbors: Unlikely to interact, but not impossible.
• Very far neighbors: Almost never interact.

The authors ask: If we treat this structured dance floor as a model for a real quantum system, what happens to the "entanglement" (a fancy way of saying how much the dancers are "linked" or "entangled" with each other)?

Here is the breakdown of their findings using simple analogies:

1. The Three "Moods" of the Dance Floor

The behavior of the system changes based on a single knob they can turn, called $\alpha$ (alpha). This knob controls how quickly the interaction strength drops off with distance.

• The "Party" Mode (Ergodic Phase, $\alpha < 0.5$):
  The interactions are so strong that everyone mixes perfectly. It's like a mosh pit where everyone is jostling with everyone else. In this state, the dancers are maximally entangled. If you look at half the room, the information about the other half is completely scrambled. This is the "Volume Law" state—chaos and connection everywhere.

• The "Isolated" Mode (Localized Phase, $\alpha > 1$):
  The interactions are so weak that dancers only stick to their immediate neighbors. The room freezes into small, isolated groups. No one talks to the other side of the room. This is the "Area Law" state—entanglement is low and confined to the edges of small groups.

• The "Strange Middle" Mode (Weakly Ergodic Phase, $0.5 < \alpha < 1$):
  This is the most interesting part of the paper. It's a mix. The dancers in the middle of the energy spectrum (the "bulk") are still having a wild party (Volume Law). But the dancers at the very edges of the energy spectrum (the "ground state" and "highest energy state") are acting like they are in the Isolated Mode (Area Law).

  • The Analogy: Imagine a concert. The people in the middle of the crowd are jumping and singing along (chaotic/entangled). But the people at the very front and very back of the venue are standing quietly in their own little bubbles (localized).

2. The "Labeling" Problem (The Seating Chart)

To make this math work for a real quantum system (like a chain of spins), the authors had to decide how to assign the "dance floor seats" to the "dancers."

• The Binary Scheme: They tried assigning seats like counting numbers (000, 001, 010...). This turned out to be a bad idea because it created a bias: the "left" side of the room behaved differently than the "right" side, just because of how they numbered the seats. It was like having a rule where the left side of the room had better speakers than the right.
• The Gray Code Scheme: They found a smarter way to number the seats (Gray code) where changing one number only changes one seat slightly. This made the room behave more fairly and uniformly.
• Site Randomization: Even with the better numbering, they had to shuffle the physical locations of the seats randomly to ensure the "physics" didn't depend on which side of the room you were sitting on.

3. The "Rainbow" Discovery

The biggest finding is what happens in that "Strange Middle" mode.
In standard random models, the entanglement looks the same everywhere. But in this new model, the entanglement forms a Rainbow:

• Top of the rainbow (High Energy): Low entanglement (Area Law).
• Bottom of the rainbow (Low Energy): Low entanglement (Area Law).
• Middle of the rainbow (Mid Energy): High entanglement (Volume Law).

This perfectly mimics real-world quantum systems, which the old "perfectly random" models failed to do.

4. The "Zombie" Zone (Intermediate States)

The authors dug even deeper into that "Strange Middle" mode. They found a specific group of dancers right on the border between the "Party" and the "Isolated" zones.

• These dancers are technically "entangled" (Volume Law), but they aren't fully chaotic like the ones in the middle. They are "almost" there, but with a slight deviation.
• It's like a group of people who are dancing, but they are slightly out of sync with the main crowd. They aren't frozen, but they aren't fully wild either.

Why Does This Matter?

This paper provides a better "toy model" for physicists to study Quantum Chaos and Many-Body Localization.

• Realism: It captures the fact that real quantum systems have a "bulk" (chaotic) and "edges" (ordered), which standard models miss.
• Predictions: It helps predict how quantum computers might behave. If a quantum computer gets too "localized" (frozen), it stops working. If it's too "chaotic," it loses information. This model helps find the sweet spot and understand the transition between the two.

In a nutshell: The authors built a better, more realistic simulation of a quantum system. They discovered that these systems aren't just "all chaotic" or "all frozen"; they have a complex structure where the middle is wild, the edges are calm, and there's a weird, in-between zone that behaves like a "half-party." This helps us understand how real quantum matter holds together.

Technical Summary

1. Problem Statement

Random Matrix Theory (RMT), specifically Gaussian Orthogonal (GOE) and Unitary (GUE) ensembles, is the standard model for chaotic quantum many-body systems. However, these ensembles possess a significant limitation: their eigenvectors exhibit identical statistical properties (e.g., entanglement entropy) throughout the entire energy spectrum, from the spectral edges to the bulk.

In contrast, physical many-body Hamiltonians with local interactions display a distinct "edge-bulk" dichotomy:

• Spectral Edges (Low/High Energy): Eigenstates typically follow an area law (low entanglement).
• Spectral Bulk (Mid-Energy): Eigenstates follow a volume law (high entanglement, approaching the Page value).
• Result: A "rainbow" or arch-shaped plot of entanglement entropy versus energy.

The Power-Law Banded Random Matrix (PLBRM) ensemble is known to exhibit different localization phases based on a power-law exponent $\alpha$. While previously studied as single-particle models for Anderson localization, this paper investigates whether PLBRMs can serve as improved models for quantum many-body Hamiltonians that naturally reproduce the edge-bulk entanglement distinction absent in standard GOE/GUE models.

2. Methodology

A. Model Definition

The authors utilize the PLBRM ensemble defined by matrix elements:
$$H_{ij} = G_{ij} a(|i-j|)$$
where $G_{ij}$ are GOE random variables, and the envelope function $a(r)$ decays as a power law:
$$a(r) = \frac{1}{1 + (r/\beta)^\alpha}$$
The system size is $N = 2^L$ for an $L$-site spin-$1/2$ chain. The exponent $\alpha$ controls the phase:

• $\alpha < 1/2$: Fully ergodic.
• $1/2 < \alpha < 1$: Weakly ergodic (multifractal).
• $\alpha > 1$: Localized.

B. Basis State Labeling Schemes

A critical methodological step is mapping random matrix indices ($1 \dots N$) to many-body basis states (bit strings of length $L$). The authors propose and compare three schemes to minimize "badness" ($B$), defined as the average Hamming distance between consecutive basis states:

1. Random Labeling: High badness ($B \sim L/2$); interactions are highly non-local.
2. Binary Labeling: Low badness ($B \sim 2$); index $i$ is the binary representation of the spin configuration.
3. Gray Code Labeling: Optimal badness ($B = 1$); consecutive indices differ by exactly one spin flip. This is proposed as the superior scheme for modeling local interactions.

C. Site Randomization

The authors identify that fixed labeling schemes (Binary/Gray) introduce spatial inhomogeneity in the ensemble-averaged Hamiltonian. To correct this, they implement "site randomization", where the mapping between the binary digits of the basis states and the physical lattice sites is randomized for each realization. This ensures the resulting Hamiltonian is spatially homogeneous.

D. Analysis

The primary observable is the half-system bipartite entanglement entropy ($S_{ent}$) of eigenstates. The authors perform finite-size scaling analyses on:

• Spectral bulk eigenstates (mid-spectrum).
• Spectral edge eigenstates (ground and highest excited states).
• Intermediate eigenstates to identify boundaries between different scaling behaviors.

3. Key Contributions

1. Systematic Labeling Framework: The paper establishes that the Gray code labeling scheme combined with site randomization is the most physically consistent method for interpreting PLBRMs as many-body Hamiltonians, minimizing artificial non-locality and spatial inhomogeneity.
2. Quantification of Edge-Bulk Distinction: It provides the first detailed quantitative mapping of how the PLBRM phases translate to many-body entanglement transitions, specifically confirming that the weakly ergodic phase ($1/2 < \alpha < 1$) reproduces the "rainbow" entanglement profile seen in physical chaotic systems.
3. Discovery of Intermediate Scaling Regimes: The authors identify a previously uncharacterized set of eigenstates in the weakly ergodic phase that exhibit volume-law scaling but with a non-vanishing deviation from the Page value ($S_{Page}$).
4. Critical Exponents for Boundaries: The paper defines and numerically extracts two critical exponents ($\delta_1$ and $\delta_2$) that demarcate the boundaries between different entanglement behaviors in the spectrum.

4. Key Results

A. Phase Diagram and Entanglement Scaling

• Fully Ergodic Phase ($\alpha < 1/2$): All eigenstates (bulk and edge) exhibit volume-law scaling, with $S_{ent} \to S_{Page}$ as $L \to \infty$.
• Localized Phase ($\alpha > 1$): All eigenstates exhibit area-law scaling ($S_{ent} \sim L^0$).
• Weakly Ergodic Phase ($1/2 < \alpha < 1$):
  • Bulk: Volume law ($S_{ent} \approx S_{Page}$).
  • Edges: Area law ($S_{ent} \ll S_{Page}$).
  • Transition: The transition from volume to area law occurs at $\alpha \approx 1/2$ for the edges and $\alpha \approx 1$ for the bulk.

B. Identification of Three Entanglement Regimes in the Weakly Ergodic Phase

In the range $1/2 < \alpha < 1$, the spectrum is partitioned into three distinct regions based on the eigenstate index $n$ (scaled by $N$):

1. Deep Edge ($n \sim N^{\delta_2}$): Eigenstates follow an area law.
   • The exponent $\delta_2$ increases from 0 to 1 as $\alpha$ goes from $1/2$ to $1$.
2. Intermediate Edge ($N^{\delta_2} < n < N^{\delta_1}$): Eigenstates follow a volume law but with a non-vanishing deviation from the Page value ($S_{Page} - S_{ent} > 0$).
   • The exponent is found to be $\delta_1 = \alpha$.
   • As $\alpha \to 1/2^+$, the number of these states scales as $N^{1/2}$ (diverging number, vanishing fraction).
3. Bulk ($n > N^{\delta_1}$): Eigenstates follow a volume law with $S_{ent} \to S_{Page}$ (maximally ergodic).

C. Scaling Analysis

• The transition point for bulk eigenstates (volume $\to$ area) is confirmed at $\alpha_c \approx 1$.
• The transition point for edge eigenstates is confirmed at $\alpha_c \approx 1/2$.
• The relationship $\delta_1 = \alpha$ is a striking, simple analytic result derived numerically, suggesting a deep connection between the power-law decay of matrix elements and the spectral fraction of non-ergodic states.

5. Significance

• Improved Modeling of Chaos: The PLBRM ensemble, when interpreted via the Gray code and site randomization, serves as a superior minimal model for chaotic many-body systems compared to GOE/GUE. It captures the crucial physical feature of spectral edge-bulk differentiation (the "entanglement rainbow").
• Insight into Ergodicity Breaking: The work clarifies the nature of the "weakly ergodic" phase, showing it is not a uniform state but contains a complex hierarchy of eigenstates ranging from area-law to perfect volume-law.
• New Critical Phenomena: The identification of the intermediate set of states (volume law with deviation) and the precise scaling $\delta_1 = \alpha$ offers new targets for analytical theory. It suggests a "quantum Zeno-like" transition where the number of non-ergodic states diverges even as their fraction vanishes at the critical point $\alpha = 1/2$.
• Future Directions: The paper opens avenues for studying open/driven systems, exploring the dependence on labeling schemes, and applying these findings to real disordered interacting systems (e.g., Many-Body Localization transitions).

In summary, this paper successfully bridges the gap between single-particle random matrix theory and many-body quantum chaos, providing a rigorous framework and quantitative results for how power-law correlations in Hamiltonian matrices manifest as entanglement phase transitions in physical systems.