Logarithmic growth of operator entanglement in a clean non-integrable circuit

This paper demonstrates that in a clean, non-integrable semi-ergodic dual-unitary circuit, the operator entanglement of a single-site operator grows at most logarithmically over time, exhibiting intermediate dynamics between chaotic and free systems despite the absence of disorder.

The Gist

Imagine you are watching a massive, chaotic dance floor where thousands of dancers (quantum particles) are interacting. In physics, we usually expect two extreme outcomes for how this dance evolves:

1. The "Perfectly Ordered" Dance (Integrable): The dancers follow a strict, predictable script. They never get truly tangled up with each other. This is easy to predict, but boring.
2. The "Total Chaos" Dance (Chaotic): The dancers move randomly, bumping into everyone, and quickly become a giant, inseparable knot of entanglement. This is incredibly hard to predict and impossible for a normal computer to simulate once the knot gets too big.

This paper introduces a fascinating third option: a "Semi-Ergodic" dance. It's a Goldilocks scenario where the system is complex enough to be interesting, but not so chaotic that it becomes impossible to track.

Here is a breakdown of what the researchers found, using simple analogies:

1. The Setup: A Special Kind of Circuit

The scientists built a specific type of quantum circuit (a "brickwork" pattern of gates) that acts like a conveyor belt.

• The Two Light Rays: Imagine the dance floor has two directions of movement.
  • Direction A (The Chaotic Ray): If you look at how dancers interact moving this way, they mix up completely and forget their past. This is "ergodic" (normal chaos).
  • Direction B (The Quiet Ray): If you look at the other direction, the dancers are surprisingly stubborn. They keep some of their original identity and don't mix up as much. This is "non-ergodic."

By combining these two directions, they created a system that is half-chaotic and half-stable.

2. The Main Character: The Qutrit vs. The Qubits

To understand what happens, the researchers simplified the math. They realized that if you start with a single "dancer" (a specific quantum operator) on the floor, its evolution can be imagined as a special traveler (a "qutrit," which is like a 3-sided coin) walking through a crowd of regular people (qubits, or 2-sided coins).

• The Scenario: The 3-sided coin (the qutrit) walks down a hallway. Every step, it bumps into one regular person (a qubit) and they swap places or change states.
• The Twist: The regular people (qubits) are actually "solitons"—they are like ghostly waves that just keep moving forward without changing, unless they hit the special traveler.
• The Result: The traveler (qutrit) is constantly scattering off these ghosts, but because of the special rules of this "semi-ergodic" dance, the traveler never gets lost in a total mess. It stays in a restricted, manageable zone.

3. The Big Surprise: Logarithmic Growth

In a normal chaotic system, the "entanglement" (how tangled the system gets) grows linearly. Imagine a rope being tied into knots; in chaos, you tie a knot every second, and the knot pile grows fast and tall.

The Discovery: In this semi-ergodic system, the entanglement grows logarithmically.

• The Analogy: Imagine you are tying knots, but every time you tie one, the rope gets slightly harder to tie the next one. You tie a knot, then wait a bit, then tie another, then wait longer. The pile of knots grows, but it grows very slowly.
• Why it matters: Usually, if a system isn't perfectly ordered (integrable), we expect it to be chaotic and hard to simulate. The fact that this system is not perfectly ordered but still grows entanglement so slowly is a huge surprise. It means this system is "hard" enough to be interesting, but "easy" enough that we might still be able to simulate it on a computer for a long time.

4. The "Bimodal" Size Distribution

The researchers also looked at the "size" of the quantum information.

• In Chaos: The information spreads out everywhere, becoming a giant, complex blob.
• In Order: The information stays small and simple.
• In this System: The information becomes bimodal (two-peaked). Sometimes it stays small and simple (like a single dancer), and sometimes it spreads out into a large, complex blob. It's like the system is constantly flipping between being "lazy" and being "busy," but never fully committing to total chaos.

5. Why Should We Care?

This paper is a "proof of concept" for a new kind of quantum behavior.

• For Quantum Computers: It shows there is a middle ground between "too simple to be useful" and "too complex to simulate."
• For Understanding Nature: It helps us understand how complexity arises. It suggests that you don't need total chaos to get interesting dynamics; you just need a little bit of "stubbornness" (non-ergodicity) mixed in.

In a nutshell: The authors found a quantum system that behaves like a traveler walking through a crowd. The traveler gets bumped around, but because the crowd has a special rule (one side is chaotic, one side is stubborn), the traveler never gets completely lost. This keeps the system's complexity growing at a snail's pace, defying the usual rules of quantum chaos.

Technical Summary

1. Problem Statement

The paper addresses the challenge of simulating quantum dynamics on classical computers, particularly the difficulty posed by quantum chaotic systems.

• The Dichotomy: Integrable systems are easy to simulate but have slow-decaying correlations. Chaotic systems are hard to simulate due to rapid entanglement generation (linear growth of operator entanglement) and exponential decay of correlations.
• The Gap: There is a need to explore intermediate regimes between chaos and integrability to understand the complexity classification of quantum dynamics and potentially identify systems that are "hard" for classical simulation but retain "survivable" correlations.
• The Specific Model: The authors investigate semi-ergodic dual-unitary circuits. These are brickwork circuits where two-point correlation functions exhibit ergodic behavior along one light ray (decaying to zero) and non-ergodic behavior along the other (surviving at late times). While infinite-width limits of such circuits are analytically tractable, finite-width circuits are generally non-integrable and expected to exhibit linear operator entanglement growth (a signature of classical simulation hardness).

2. Methodology

The authors employ a combination of analytical mapping and exact numerical simulations to study the Heisenberg evolution of a single-site traceless operator (e.g., a Pauli matrix) in a finite, periodic brickwork circuit.

• Model Setup:

  • The system consists of $L$ qubits with periodic boundary conditions.
  • The evolution is driven by a Floquet operator $U_F = U_e U_o$, where $U_o$ and $U_e$ are alternating layers of dual-unitary gates.
  • The gates are parametrized such that the quantum channel $M_+$ (right-moving light ray) is non-ergodic (eigenvalue 1 exists for non-trivial operators), while $M_-$ (left-moving light ray) is ergodic.

• Analytical Mapping (The Qutrit Scattering Picture):

  • The authors prove that the Heisenberg evolution of a single-site operator initially on an odd site is confined to a restricted operator subspace.
  • This subspace is spanned by Pauli strings where the operator acts as a qutrit (encoding $\sigma_x, \sigma_y, \sigma_z$) moving along the ergodic light ray, interacting sequentially with a "bath" of qubits (encoding $I$ or $\sigma_z$) moving along the non-ergodic light ray.
  • The dynamics are mapped to a single qutrit scattering sequentially with $L/2$ qubits arranged in a helix.
  • By exploiting a hidden $U(1)$ symmetry in the scattering matrix, the evolution is rewritten in terms of products of $SO(3)$ matrices. The operator state is represented as a superposition of a qutrit and qubits, where the qutrit's state vector is rotated by $SO(3)$ matrices at each scattering event.

• Numerical Approach:

  • Due to the reduction to a $6 \times 6$ matrix multiplication per time step (describing the qutrit-qubit scattering), the authors perform exact numerical calculations for large system sizes ($L \approx 60$), which is significantly larger than what is typically possible for generic non-integrable circuits using tensor networks.
  • They compute:
    1. Auto-correlation functions.
    2. Operator entanglement entropy.
    3. Operator size distribution.

3. Key Contributions

1. Discovery of Logarithmic Entanglement Growth: Contrary to the expectation that non-integrable, clean circuits should exhibit linear operator entanglement growth (indicating high complexity), the authors demonstrate that this semi-ergodic model exhibits at most logarithmic growth in time.
2. Mapping to Soliton Scattering: They provide a rigorous mapping of the complex many-body evolution to a simple single-particle scattering problem (qutrit vs. solitons/qubits), revealing a hidden structure that prevents rapid scrambling.
3. Random Matrix Theory (RMT) Prediction: They show that the late-time auto-correlation functions can be described by a sum of products of $SO(3)$ matrices. In the "semi-ergodic" parameter regime, these products converge to a Haar average, predicting a late-time auto-correlation value of $1/3$.
4. Bimodal Operator Size Distribution: The study reveals a unique dynamical feature where the operator size distribution becomes bimodal at specific times (multiples of $L/2$). This indicates the coexistence of simple operators (small size) and complex operators (large size), a behavior intermediate between chaotic and integrable systems.

4. Key Results

• Operator Entanglement:

  • For both semi-ergodic and non-semi-ergodic parameter choices, the operator entanglement entropy grows logarithmically with time, eventually saturating at the logarithm of the subsystem dimension.
  • This is a stark contrast to chaotic systems (linear growth) and resembles the behavior of Many-Body Localized (MBL) systems, despite the system being clean (no disorder) and non-integrable.
  • The semi-ergodic regime shows an initial saturation at $\ln 3$ (maximal entanglement of a qutrit) before the slow logarithmic growth begins.

• Auto-Correlation Functions:

  • The auto-correlation function $C_{XX}(t)$ is analyzed.
  • In the semi-ergodic regime (specific parameter values), the time-averaged correlation converges to the RMT prediction of $1/3$ at times $t = k \cdot L/2$.
  • In the non-semi-ergodic regime, the correlation does not converge to this value, indicating a breakdown of the random matrix behavior.

• Operator Size Distribution:

  • The distribution of operator sizes ($S_l$) is bimodal at times $t = k \cdot L/2$.
  • There is a sharp peak at small sizes ($l=1$), corresponding to the survival of the initial operator structure (due to the non-ergodic channel), and a broad peak at large sizes, representing the chaotic spreading.
  • This bimodality is a signature of the "semi-ergodic" nature, where information is neither fully scrambled nor fully preserved.

5. Significance and Implications

• Complexity Classification: The work identifies a new class of quantum dynamics that is non-integrable yet possesses slow entanglement growth. This challenges the conventional wisdom that non-integrability necessarily implies linear entanglement growth and classical simulation hardness.
• Quantum Advantage: While the specific model studied is not complex enough to demonstrate quantum advantage (as it is classically simulatable due to the logarithmic entanglement), it provides a blueprint for searching for "semi-ergodic" dynamics in more complex systems (e.g., higher dimensions or triunitary circuits) that might be hard to simulate classically while retaining measurable late-time signals.
• Physical Insight: The results deepen the understanding of the interplay between chaos and integrability. The "semi-ergodic" regime acts as a bridge, exhibiting features of both: chaotic spreading (large operator size peak) and integrable survival (small operator size peak and slow entanglement growth).
• Future Directions: The authors suggest extending this framework to higher-dimensional dual-unitary circuits or triunitary circuits, where having one non-ergodic direction and multiple ergodic directions might tilt the balance toward chaotic behavior while preserving non-trivial late-time correlations.

In summary, the paper demonstrates that by carefully engineering the ergodicity of light rays in dual-unitary circuits, one can create a clean, non-integrable system that defies the typical chaotic scaling of entanglement, offering a novel perspective on the complexity of quantum dynamics.