Fading ergodicity and quantum dynamics in random matrix ensembles

This paper demonstrates that the Rosenzweig-Porter and ultrametric random matrix ensembles belong to the same universality class of ergodicity breaking when embedded in many-body spin systems, unifying them under the "fading ergodicity" regime where local observables thermalize on timescales shorter than the Heisenberg time.

The Gist

Imagine you have a giant, chaotic dance floor filled with thousands of dancers (the quantum particles). In a perfectly "ergodic" system, everyone eventually mixes together, forgets who they started with, and the whole floor looks like a uniform, happy crowd. This is how most normal materials behave: if you heat one corner, the heat spreads until the whole room is the same temperature.

But what happens if the dance floor starts to get weird? What if some dancers get stuck in corners, or the music changes so that people only dance with their immediate neighbors? This is called ergodicity breaking, and physicists are trying to understand exactly how and when this happens.

This paper is about a "middle ground" phase called Fading Ergodicity. It's like the dance floor isn't fully mixed yet, but it's not completely frozen either. The dancers are starting to get stuck, but they haven't given up on dancing entirely.

Here is the story of the paper, broken down into simple concepts:

1. The Two Different Dance Floors (The Models)

The researchers studied two very different mathematical models to see how this "stuckness" happens:

• The Rosenzweig-Porter (RP) Model: Imagine a dance floor where the music is mostly random, but there's a subtle rule that makes it harder for dancers to jump across the room. As you tweak the music (a parameter called $\gamma$), the dancers slowly stop mixing.
• The Ultrametric (UM) Model: Imagine a dance floor organized like a pyramid or a family tree. Dancers at the top of the tree can talk to everyone, but dancers at the bottom can only talk to their immediate family. As you tweak the rules (a parameter called $\alpha$), the "family" groups get too isolated to mix with the rest of the room.

The Big Discovery: Even though these two models look completely different on paper, the researchers found that when they are tuned to the point where the system is just about to stop mixing, they behave exactly the same way. They belong to the same "universality class." It's like realizing that a traffic jam caused by a red light and a traffic jam caused by a broken bridge follow the exact same flow patterns right before cars stop moving.

2. The "Fading" Effect

The paper introduces the idea of Fading Ergodicity.

• Normal Mixing: In a healthy system, if you look at the energy of a specific dancer, it fluctuates wildly and quickly.
• Fading Ergodicity: As the system approaches the "stuck" phase, those fluctuations don't disappear instantly. Instead, they fade away slowly. The dancers are still moving, but their movements are becoming predictable and repetitive. They are "forgetting" how to be chaotic.

The researchers showed that in this fading phase, the system still manages to "thermalize" (reach a steady temperature) within a reasonable amount of time, but it takes longer than usual. It's like a party that is slowly winding down; people are still talking, but the energy is dropping, and eventually, everyone goes home.

3. The "Thouless Time" (The Stopwatch)

To compare these two different models, the researchers needed a common stopwatch. They used something called the Thouless time.

• Think of this as the time it takes for a rumor to spread from one side of the dance floor to the other.
• In a normal party, the rumor spreads fast.
• In the "fading" phase, the rumor spreads slower and slower.
• The researchers calibrated their two models so that the "rumor spreading time" was identical. Once they did this, they found that the behavior of the dancers (the quantum particles) was identical in both models.

4. The "Quantum Quench" (The Surprise Party)

To test this, they performed a "Quantum Quench." Imagine the dance floor is quiet, and suddenly, the DJ blasts a new song (a sudden change in energy).

• They watched how the dancers reacted.
• The Result: Even in the "fading" phase, the dancers eventually settled into a new rhythm that matched the average of the whole room. They didn't get stuck forever; they just took a little longer to get there.
• Crucially, they found that the system reaches this steady state before the "Heisenberg time" (the absolute maximum time limit for a quantum system to do anything interesting). This proves that the system is still "alive" and thermalizing, even though it's on the edge of breaking down.

5. The Noise and the Spectrum

The researchers also listened to the "noise" of the system (the power spectrum).

• In the RP model, the noise looked like a smooth, predictable wave (a Lorentzian shape).
• In the UM model, the noise had long, jagged tails (power-law decay).
• The Surprise: Even though the shape of the noise was different, the timing of the noise (how fast it faded) was the same. This confirmed that the underlying mechanism of "fading ergodicity" is the same, even if the surface details look different.

The Takeaway

This paper is like finding a universal rule for how things break down.

• Before: Physicists thought different types of systems (like the RP model and the UM model) might break down in totally different ways.
• Now: We know that there is a specific "fading" phase where they all behave the same. It's a bridge between a perfectly mixed system and a completely frozen one.

In simple terms: The universe has a "slow-motion" mode right before it freezes. Whether you are looking at a random mess of particles or a structured hierarchy, if you slow them down just right, they all start to dance to the same beat. This helps scientists understand how quantum computers might fail or how new materials might behave in extreme conditions.

Technical Summary

1. Problem Statement

The paper addresses the challenge of characterizing the boundaries of ergodicity in isolated quantum systems and identifying universal mechanisms for ergodicity breaking. While the Eigenstate Thermalization Hypothesis (ETH) describes thermalization in chaotic systems and Many-Body Localization (MBL) describes non-ergodic phases, the intermediate regime where ergodicity "fades" remains a subject of intense study.

Specifically, the authors aim to bridge the gap between:

1. Physical models with few-body interactions (e.g., the Quantum Sun Model) that exhibit "fading ergodicity."
2. Structured random matrix models (specifically the Rosenzweig–Porter and Ultrametric models) which are often used as theoretical proxies for localization transitions but are defined in dense Hilbert spaces with unphysical multi-body interactions.

The central question is whether these distinct classes of models share a universal mechanism for ergodicity breaking when embedded in a many-body Hilbert space of spins-1/2.

2. Methodology

The authors employ a comparative numerical and analytical approach involving two paradigmatic random matrix models embedded in a Hilbert space of $L$ coupled spin-1/2 particles (dimension $D=2^L$).

• Models Studied:

  • Rosenzweig–Porter (RP) Model: Defined by a Hamiltonian $\hat{H} = \hat{H}_0 + D^{-\gamma/2}\hat{M}$. The parameter $\gamma$ controls the transition: $\gamma < 1$ (ergodic/GOE), $1 < \gamma < 2$ (fractal phase/fading ergodicity), and $\gamma > 2$ (localized/Poisson).
  • Ultrametric (UM) Model: A hierarchical model describing central impurities interacting with remaining qubits via spatially decaying coupling, controlled by parameter $\alpha$. The transition occurs at $\alpha_c = 1/\sqrt{2}$.

• Calibration Strategy:
  To compare the models, the authors calibrate their parameters ($\gamma$ and $\alpha$) by aligning their Thouless times ($t_{Th}$), or equivalently, their Thouless energies ($\Gamma$). They derive a scaling relation (Eq. 9) linking $\gamma$ and $\alpha$ in the thermodynamic limit:
  $$\gamma = 1 + \frac{\ln \alpha}{\ln \alpha_c}$$
  This ensures that both models are compared at equivalent distances from their respective ergodicity-breaking transitions.

• Observables and Dynamics:

  • Observable: The local spin operator $\hat{O} = \hat{S}^z_L$ (at the site most weakly coupled in the UM model).
  • Analysis:
    1. Matrix Elements: Statistical properties of $\langle n|\hat{O}|m\rangle$ to extract the fluctuation exponent $\eta$.
    2. Spectral Functions: Fourier transforms of autocorrelation functions to extract $\Gamma$.
    3. Quantum Quench: Evolution from initial product states $|\psi_0\rangle$ to study thermalization, long-time averages, and temporal fluctuations.
    4. Power Spectra: Analysis of noise in the steady state and survival probabilities.

3. Key Contributions

1. Universality of Fading Ergodicity: The paper establishes that the RP and UM models, despite their structural differences (dense vs. hierarchical), belong to the same universality class of ergodicity breaking when mapped to the fading ergodicity regime.
2. Unified Framework: It provides a unified theoretical framework connecting the "fractal phase" of the RP model with the "ergodic phase" of the UM model, showing both are precursors to ergodicity breaking characterized by fading ergodicity.
3. Decomposition of Power Spectra: The authors demonstrate that the power spectrum of observable fluctuations can be approximately decomposed into the product of the observable's spectral function and the local density of states (LDoS) spectrum of the initial state.
4. Thermalization Timescales: They prove that in the fading ergodicity regime, local observables thermalize on timescales shorter than the Heisenberg time ($t_H$), distinguishing this regime from full localization.

4. Key Results

A. Spectral Properties and Calibration

• Thouless Energy Extraction: By analyzing the spectral function $\tilde{O}(\omega)$, the authors extract the Thouless energy $\Gamma$. They confirm that $\Gamma$ scales exponentially with system size $L$ in the fading regime.
• Parameter Mapping: The numerical data confirms the analytical scaling relation between $\gamma$ and $\alpha$ (Eq. 9) in the thermodynamic limit, with finite-size corrections scaling as $1/L$.

B. Matrix Element Fluctuations

• Exponent $\eta$: The variance of off-diagonal matrix elements scales as $D^{-2/\eta}$.
• Universal Behavior: Both models exhibit the same exponent $\eta$ when parameters are matched via Eq. 9.
  • In the fading regime, $2 < \eta < \infty$.
  • At the critical point, $\eta$ diverges.
  • The relation $\eta = 2/(2-\gamma)$ (for RP) holds for both models, linking matrix element fluctuations directly to the fractal dimension of eigenstates ($d_2^{(eig)} = 2-\gamma$).

C. Quantum Quench Dynamics

• Thermalization: Following a quench from product states, the expectation values of $\hat{S}^z_L$ relax to the microcanonical ensemble (ME) predictions.
• Timescales: The relaxation occurs on the Thouless timescale ($t_{Th} \sim 1/\Gamma$), which is much shorter than the Heisenberg time ($t_H \sim D$).
• Long-time Averages: The deviation between the diagonal ensemble (DE) and ME predictions, $\Delta Q$, vanishes exponentially with system size $L$ in the fading ergodicity regime, confirming thermalization.

D. Temporal Fluctuations and Survival Probability

• Fluctuation Exponent ($\eta_t$): The variance of temporal fluctuations scales as $\sigma_t^2 \propto D^{-2/\eta_t}$.
  • RP Model: $\eta_t$ matches the matrix element exponent $\eta$. At the critical point $\gamma_c=2$, $\eta_t \to \infty$ (fluctuations do not vanish), indicating a lack of equilibration.
  • UM Model: $\eta_t$ remains finite at the critical point $\alpha_c$, suggesting that equilibration persists even at the transition, a distinct difference from the RP model.
• Power Spectra:
  • RP Model: The power spectrum $S(\omega)$ follows a Lorentzian decay ($\propto \omega^{-4}$) due to the product of two Lorentzian functions (spectral function and LDoS).
  • UM Model: Near the critical point, the LDoS exhibits power-law tails ($\propto \omega^{-1+d_2^{(0)}}$), leading to a power spectrum decay of $\propto \omega^{-3+d_2^{(0)}}$. This encodes the fractal dimension of the initial state.

5. Significance

This work is significant because it:

• Validates the "Fading Ergodicity" Concept: It demonstrates that fading ergodicity is not an artifact of specific toy models (like the Quantum Sun Model) but a robust phenomenon observable in structured random matrix ensembles.
• Bridges Theory and Physics: It reconciles the study of dense random matrices (often criticized for unphysical interactions) with physical many-body systems, showing they share universal critical exponents and dynamical behaviors near the ergodicity-breaking transition.
• Clarifies Thermalization: It refines the understanding of thermalization in non-ergodic phases, showing that "fading" ergodicity allows for thermalization on intermediate timescales ($t_{Th} < t < t_H$) before the system eventually freezes or localizes at very long times.
• Provides Diagnostic Tools: The decomposition of power spectra and the scaling of fluctuation exponents offer concrete numerical diagnostics for identifying the nature of ergodicity breaking in complex quantum systems.

In conclusion, the paper establishes that the Rosenzweig–Porter and Ultrametric models, when properly calibrated, exhibit identical fading ergodicity signatures, providing a unified framework for understanding the transition from chaos to localization in quantum many-body systems.