Operator space fragmentation in perturbed Floquet-Clifford circuits

This paper demonstrates that random Floquet-Clifford circuits perturbed by non-Clifford gates exhibit robust operator localization and emergent local integrals of motion for all perturbation probabilities $p < 1$, driven by the fragmentation of operator space into disjoint sectors separated by wall configurations, which creates entanglement bottlenecks and delays the onset of quantum chaos.

The Gist

Imagine a quantum computer not as a single, giant brain, but as a long line of tiny, dancing particles (qubits) passing information back and forth. In a perfectly chaotic system, if you drop a pebble (an "operator" or piece of information) into one end of the line, ripples would spread out instantly, mixing with everything else until the whole line is a jumbled, entangled soup. This is called ergodicity or chaos.

However, this paper explores a special kind of quantum circuit that acts more like a dammed river. Even when the water is pushed and stirred, the flow gets stuck in specific pools, unable to mix with the rest of the river.

Here is a breakdown of their findings using simple analogies:

1. The Setup: The "Brickwork" Dance

The researchers built a model using a "brickwork" pattern. Imagine a wall being built where bricks are laid in alternating rows.

• The Dancers: The "bricks" are quantum gates (operations) that make two qubits interact.
• The Perfect Dancers (Clifford): In their base model, the dancers follow strict, predictable rules (Clifford gates). These are easy to simulate on a regular computer.
• The Wild Dancers (Perturbations): To make it realistic, they introduced "wild" dancers (random unitary gates) that don't follow the strict rules. They sprinkle these in with a probability $p$. If $p=0$, everyone follows the rules. If $p=1$, everyone is wild.

2. The Discovery: "Walls" that Stop the Flow

The team discovered that even with these wild dancers, the system doesn't always turn into a chaotic soup. Instead, it forms invisible walls.

• The Analogy: Imagine a hallway where people are trying to walk from left to right. Suddenly, a specific arrangement of furniture (a "k-wall") appears. No matter how hard a person pushes from the left, they cannot pass through this furniture arrangement to get to the right side.
• The Mechanism: In the quantum world, these "walls" are specific sequences of gates. When an operator (a piece of information) hits this wall, it gets trapped. It can wiggle around inside the wall, but it cannot spread out to the rest of the system.
• The Result: The long line of qubits breaks up into isolated "fragments" or pools. Information stays trapped in its own pool.

3. The "Magic" of Stability

You might think that adding random "wild" dancers ($p > 0$) would destroy these walls. The paper shows that the walls are surprisingly sturdy.

• The Analogy: Think of a wall made of dominoes. If you shake the floor a little bit (add some random noise), the wall might wobble, but it doesn't fall. It takes a lot of shaking (a very high probability of wild dancers) to break the wall completely.
• The Finding: As long as the probability of wild dancers ($p$) is less than 100%, these walls keep forming. The system remains "localized," meaning information stays stuck in small regions rather than spreading everywhere.

4. The "Bottleneck" of Entanglement

In quantum physics, "entanglement" is like a deep connection between two particles. Usually, in a chaotic system, everything gets deeply connected to everything else.

• The Analogy: Imagine two rooms separated by a narrow, locked door (the wall). People in Room A can talk to each other, and people in Room B can talk to each other, but they can't really talk to the other room.
• The Finding: Because of these walls, the quantum system creates an "entanglement bottleneck." The two sides of the wall remain only weakly connected. They don't become a single, giant, entangled mess. The paper calculates that the amount of connection across these walls is very small and bounded—it never grows infinitely large.

5. The "Chaos" Inside the Pools

Even though the walls stop information from leaving the pools, what happens inside the pools?

• The Analogy: Inside each isolated room, the people are still dancing wildly and chaotically. They are mixing up perfectly within their own room.
• The Finding: The researchers found that inside these trapped fragments, the system behaves like a random, chaotic system (similar to a "Circular Unitary Ensemble" in physics). So, you have orderly isolation (the walls) containing chaotic islands (the fragments).

6. The Tipping Point

The paper identifies a clear tipping point:

• If $p < 1$ (Even a tiny bit of order remains): The walls exist. The system is fragmented. Information is localized. It is not chaotic across the whole system.
• If $p = 1$ (Everything is wild): The walls eventually break down. The "river" flows freely again, and the whole system becomes a chaotic, fully mixed soup.

Summary

This paper describes a quantum system that acts like a fractured mirror. Even when you shake it, the pieces (fragments) stay separate because of specific structural "walls" that form naturally. Inside each piece, things are chaotic, but the pieces themselves never fully merge into one big chaotic whole. This provides a new way to understand how quantum systems can stay "localized" (stuck) rather than becoming "ergodic" (fully mixed), which is a key concept in understanding quantum memory and stability.

Technical Summary

Technical Summary: Operator Space Fragmentation in Perturbed Floquet-Clifford Circuits

Problem Statement

The dynamics of quantum many-body systems far from equilibrium are central to understanding thermalization, chaos, and localization. While random quantum circuits (specifically brickwork architectures) are rigorously established to converge to Haar-random ensembles and exhibit ergodicity, the introduction of disorder can lead to Many-Body Localization (MBL). In MBL, systems fail to thermalize due to the emergence of quasi-local integrals of motion (l-bits). However, constructing and analyzing these l-bits in generic Hamiltonian systems is analytically and numerically intractable.

This work investigates the stability of operator localization and the emergence of chaos in random Floquet-Clifford circuits subjected to unitary perturbations that drive them away from the Clifford limit. The authors aim to understand how the interplay between Clifford dynamics (which are classically simulable) and generic non-Clifford perturbations affects operator spreading, entanglement growth, and spectral statistics. Specifically, they address whether localization persists when the system is perturbed by single-qubit unitaries sampled from the Haar distribution with probability $p$.

Methodology

The authors employ a combination of analytical group-theoretic methods and exact numerical simulations on finite-size systems.

1. Model Construction:

   • The system is a one-dimensional chain of qubits evolving under a brickwork Floquet circuit.
   • The unitary evolution consists of layers of random two-qubit Clifford gates (excluding product gates to ensure connectivity) and stochastic single-qubit perturbations ($R_i$).
   • Perturbations are sampled from the Haar distribution on $U(2)$ with probability $p$, and are identity with probability $1-p$.
   • The limit $p=0$ corresponds to a pure Clifford circuit, while $p=1$ represents a fully perturbed, generic unitary circuit.

2. Analytical Framework:

   • Clifford Formalism: The authors utilize the isomorphism between the Clifford group modulo Paulis and the symplectic group $Sp(2n, \mathbb{Z}_2)$. This allows for the representation of operator spreading as linear transformations in a discrete phase space.
   • k-Wall Definition: A central concept is the "k-wall," a subsystem of $k$ sites that arrests the spreading of operators from the left or right environment. The authors define these via invariant subspaces in the symplectic vector space.
   • Fragmentation: They analyze how the operator space fragments into disjoint sectors (invariant subspaces) bounded by these walls, leading to emergent local conserved quantities.

3. Numerical Probes:

   • Entanglement Entropy: Calculated via exact diagonalization for small systems ($n \approx 10$) to study the growth of Von Neumann entropy across fragment boundaries.
   • Spectral Form Factor (SFF): Used as a probe for chaos and ergodicity. The authors compute the SFF and its sample-to-sample fluctuations to distinguish between localized (non-ergodic) and chaotic (ergodic) regimes, comparing results against the Circular Unitary Ensemble (CUE).

Key Contributions

1. Formalization of Operator Localization via k-Walls

The paper rigorously defines k-walls as configurations of gates that stop the spreading of arbitrary operators.

• Structure: The authors prove that irreducible k-walls in shallow Clifford circuits must begin and end with gates from the CZ-equivalence class.
• Conserved Charges: They demonstrate that these walls host emergent local integrals of motion. Specifically, for 1-walls, the internal subspace corresponds to a single traceless Pauli matrix (e.g., $Z$) that is conserved within the wall.
• Probability: The authors calculate the sampling probability of forming these walls in a random Clifford ensemble. For 1-walls, the probability is approximately $0.025$; for 2-walls, it is $\approx 0.013$. Higher-order walls are exponentially unlikely.

2. Stability of Localization against Perturbations

A primary contribution is the demonstration that operator localization is stable for any $0 \le p < 1$.

• Mechanism: In the thermodynamic limit, the probability of a wall being "broken" by a perturbation on all its constituent qubits decreases exponentially with the wall width. Consequently, unperturbed walls persist with high probability, fragmenting the system.
• Fragmentation: The circuit decomposes into spatially localized fragments. Within each fragment, operators evolve under an approximate Haar ensemble, but the fragments themselves are weakly coupled.
• Contrast with MBL: Unlike conventional MBL where l-bits are strictly local and the system is non-ergodic globally, this model features large ergodic regions (fragments) separated by "bottlenecks" (walls). The conserved quantities are strictly local to the walls, not the bulk.

3. Entanglement Dynamics and the "Entanglement Bottleneck"

The authors show that operator localization leads to a bound on entanglement growth across fragment boundaries.

• Bounded Entropy: For an unperturbed wall, the entanglement entropy across the cut is bounded by a constant (specifically $S \le 1$ for 1-walls in the stabilizer formalism).
• Weak Entanglement: Even with perturbations ($p < 1$), the system exhibits "weakly entangled" fragments. The average entanglement across a wall converges to a value determined by the probability of the conserved Pauli being preserved (analytically estimated at $\le 2/3$).
• Scaling: This results in an area-law-like scaling for the average entanglement in the localized phase, tunable by the perturbation probability $p$.

4. Spectral Signatures of Non-Ergodicity

The study of the Spectral Form Factor (SFF) reveals distinct signatures of the fragmented phase.

• Quadratic Ramp: In the presence of walls, the SFF exhibits a $t^2$ ramp (quadratic) at early times, deviating from the linear ramp characteristic of the CUE. This reflects the product structure of the fragmented operator space.
• Non-Ergodicity: For $p < 1$, the system fails to converge to the CUE ensemble even at late times, indicating a non-ergodic quantum phase. The fluctuations in the SFF remain large, signaling the lack of phase space chaos typical of fully ergodic systems.
• Transition: At $p=1$, the walls are destroyed, and the system eventually approaches CUE statistics, though a "fragmentation time scale" is observed before random matrix theory sets in.

Results

• Localization Length: The typical localization length (distance between unperturbed walls) is derived as $\mu(p) \sim \frac{44}{1-p}$. As $p \to 1$, the localization length diverges, signaling a transition to a delocalized phase.
• Entanglement Bounds: Numerical simulations confirm that for $p < 1$, entanglement across walls is bounded and fluctuates significantly, consistent with the analytical prediction of a "bottleneck" effect.
• SFF Behavior: The SFF for $p=0.5$ shows a quadratic ramp and large fluctuations, distinct from the linear ramp and smooth behavior of the CUE. For $p=1$, the system eventually thermalizes, but the approach is governed by the interplay of dual-unitary classes and the specific circuit geometry.

Significance and Claims

The authors claim that their work provides an explicit, analytically tractable description of quantum phases in operator dynamics that can be realized on current NISQ devices.

• Toy Model for MBL: The model serves as a "toy (yet not semi-classical) model of many-body localization" where local conservation laws (k-walls) occur without fine-tuning of circuit elements. It bridges the gap between integrable Clifford circuits and generic chaotic systems.
• Stability: The paper establishes that localization in this context is robust against generic, disordered unitary perturbations, contrasting with earlier results that relied on finely-tuned constructions.
• Fragmentation: It highlights the role of operator space fragmentation in unitary circuit dynamics, showing how disjoint sectors emerge naturally from the competition between dual-unitary (ballistic) and CZ-class (localizing) gates.
• Generalizability: The authors suggest that the concept of k-walls and fragmentation can be extended to higher-dimensional lattices (where it maps to a percolation problem) and higher local Hilbert space dimensions (qudits), potentially engineering non-ergodic phases in more complex geometries.

The work concludes that while the model does not exhibit the conventional l-bits of the MBL phase (due to the ergodic nature of the fragments), it successfully demonstrates a regime of operator localization and non-ergodicity driven by the structural constraints of the circuit gate set.