Growth and collapse of subsystem complexity under… — Plain-Language Explanation

✨

This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a giant, chaotic kitchen where a team of chefs (the quantum circuit) is constantly mixing ingredients (quantum states) together. You start with a very simple, organized meal: a plate where every ingredient is separate and untouched (a "product state").

As time passes, the chefs throw random spices and sauces onto the food, stirring it up wildly. This paper asks a specific question: How hard is it to recreate a specific portion of this chaotic meal, just by looking at that portion?

In the world of quantum physics, "complexity" is a measure of how many simple steps (or "local quantum channels") you need to build a specific state from scratch. If a state is simple, you need few steps. If it's a tangled mess of chaos, you need millions of steps.

Here is what the authors, Jeongwan Haah and Douglas Stanford, discovered about how this "complexity" grows and then collapses over time, using a model called a random brickwork circuit (think of it like a wall of bricks where each layer is randomly shuffled).

The Two Types of Plates: Small vs. Large

The researchers looked at two different sizes of plates (subsystems) taken from the giant kitchen:

The Small Plate: Less than half the size of the whole kitchen.
The Large Plate: More than half the size of the whole kitchen.

They found that these two plates behave very differently as the chefs keep mixing.

1. The Large Plate: The Never-Ending Puzzle

If you take a plate that is larger than half the total system, the complexity keeps growing linearly with time.

The Analogy: Imagine trying to describe a massive, swirling storm. As time goes on, the storm gets more and more intricate. To recreate this storm from scratch, you need more and more instructions.
The Result: For a long, long time (exponentially long), the number of steps needed to recreate this large portion grows steadily. It never stops getting harder to describe.

2. The Small Plate: The Rise and Sudden Fall

If you take a small plate (less than half the system), the story is more dramatic.

The Rise: At first, as the chefs mix, the small plate becomes more complex. It's like watching a simple salad get tossed with more and more unique dressings. The complexity grows linearly with time.
The Sudden Crash: However, once the time reaches a specific point (roughly when the time equals half the length of the plate), something strange happens. The complexity abruptly drops to zero.
The Analogy: Imagine you are trying to memorize a specific pattern of noise in a crowded room. At first, the pattern is unique and hard to replicate. But eventually, the room gets so loud and chaotic that the noise becomes a uniform "white noise" (static). Once it's just static, it's incredibly simple to describe: "It's just random noise." You don't need a million steps to recreate static; you just need to say "turn on the static."
The Result: The small plate "thermalizes." It forgets its specific history and becomes a generic, boring, high-temperature soup. Because it's so generic, it has almost zero complexity.

The "Memory" of the Circuit

One of the most fascinating parts of the paper is the question: Does the small plate remember the specific recipe the chefs used?

Early Times: Yes. If you change just one spice in the recipe used on the small plate, the final taste (the quantum state) changes completely. The plate "remembers" every single step the chefs took. This is why the complexity is high; there are so many different possible outcomes that you need a huge instruction manual to distinguish them.
Late Times (After the crash): No. Once the plate becomes "thermal" (just static), it stops remembering the specific spices. Whether the chefs added salt first or pepper first, the final result looks the same. The specific history is lost. This is why the complexity crashes: there is no unique history to reconstruct anymore.

The Holographic Connection (The "Black Hole" View)

The authors also looked at this through the lens of holography (a theory that connects our 3D world to a 2D surface, like a black hole's event horizon).

In this view, the "complexity" is like the volume of a hidden room behind a black hole's horizon.
For the small plate, this hidden room grows larger and larger as time passes.
But at the critical moment ( $T = \ell/2$ ), the geometry of this room suddenly shifts. The "door" to the hidden room closes, and the volume instantly shrinks to zero.
This supports the idea that the complexity doesn't fade away slowly; it snaps shut like a trapdoor.

Summary of the Findings

Big Systems: Keep getting more complex forever (until the heat death of the universe).
Small Systems: Get complex for a while, but then suddenly become simple and forget their past.
The Transition: The moment a small system becomes simple is sharp and sudden, not a slow fade. It's like a light switch flipping off.
Why it matters: This helps us understand how information is stored and lost in chaotic quantum systems. It shows that while a small part of a chaotic system can hold a lot of information for a while, it eventually gives up and becomes a generic, uninformative mess.

The paper uses rigorous math to prove these behaviors, showing that for small systems, the "memory" of the specific quantum operations is lost exactly when the system becomes indistinguishable from random noise.

1. Problem Statement

The paper investigates the quantum state complexity of reduced density matrices (subsystems) evolving under chaotic dynamics modeled by random unitary circuits in $1+1$ dimensions.

Context: While the complexity of a global pure state in a chaotic system is known to grow linearly with time until exponential times, the behavior of subsystem complexity (which corresponds to mixed states) is less understood.
The Core Question: How does the complexity of a subsystem of length $\ell$ $ℓ$ evolve as a function of time $T$ $T$ ? Specifically, how does this behavior differ between:
- Large subsystems ( $\ell > L/2$ ): Expected to retain memory of the global state.
- Small subsystems ( $\ell < L/2$ ): Expected to thermalize to a maximally mixed state (infinite temperature), implying low complexity at late times.
The Puzzle: How does the complexity of a small subsystem transition from linear growth (early times) to zero (late times)? Does this transition happen continuously or discontinuously (abruptly)?

2. Methodology

The authors employ a combination of rigorous mathematical proofs, holographic arguments, and random matrix theory.

Model: A brickwork random circuit on a 1D chain of $L$ qudits (dimension $q$ ). The system starts in a product pure state. At each time step, layers of random two-site unitary gates (drawn uniformly from $U(q^2)$ ) are applied.
Complexity Definition: The complexity $C(\sigma)$ of a mixed state $\sigma$ is defined as the minimum number of local quantum channels (elementary operations) required to generate $\sigma$ from a product state within a constant trace-distance error $\alpha$ .
Key Techniques:
- Unitary $k$ -designs: Using the property that random circuits approximate the Haar measure for moments up to $k \approx T/\text{poly}(L)$ .
- Counting Arguments: Estimating the volume of the space of states reachable by $G$ gates versus the volume of states generated by the circuit.
- Mutual Information Bounds: Deriving lower bounds on complexity using the monotonicity of mutual information under local channels.
- Holographic Correspondence: Analyzing the "Complexity = Volume" (CV) conjecture in AdS/CFT to predict the behavior of the entanglement wedge.
- Replica Trick & Random Matrix Theory: Analyzing the fidelity between states generated by slightly different circuits to determine memory capacity.

3. Key Contributions and Results

A. Complexity of Large Subsystems ( $\ell > L/2$ )

Result: The complexity grows linearly with time ( $C \propto T$ ) up to exponentially late times ( $T \sim e^L$ ).
Proof: The authors prove that random circuits generate a vast number of nearly orthogonal states on the subsystem. By constructing an $\alpha$ -covering net for the space of states reachable by $G$ gates, they show that $G$ must scale linearly with $T$ to distinguish the typical states produced by the circuit.
Significance: This rigorously confirms that large subsystems retain the "memory" of the global chaotic evolution for an exponentially long time.

B. Complexity of Small Subsystems ( $\ell < L/2$ )

The paper establishes a dichotomy between early and late times:

Late-Time Thermalization: For $T \gtrsim \ell/2$ , the subsystem thermalizes to the maximally mixed state (infinite temperature). Consequently, the complexity drops to zero (or a small constant).
Early-Time Growth: For $T < \ell/2$ $T < ℓ /2$ , the complexity grows linearly.
- Using mutual information, the authors derive a lower bound showing that complexity grows at least linearly up to $T \approx \ell/4$ .
- The mutual information profile $I(A:B)$ across a cut in the subsystem forms a trapezoid that peaks at $T = \ell/4$ and vanishes at $T = \ell/2$ .

C. The Nature of the Collapse (Continuous vs. Discontinuous)

The paper addresses the critical question of how the complexity drops to zero.

Rigorous Bounds: The rigorous lower bounds derived from mutual information (Section 4) are consistent with a continuous decay but do not rule out a sharp transition.
Holographic Evidence (Section 5): Using the Complexity=Volume conjecture, the authors show that the entanglement wedge of a small subsystem undergoes a phase transition at $T = \ell/2$ $T = ℓ /2$ .
- For $T < \ell/2$ , the wedge volume (and thus complexity) grows as $\ell T$ .
- For $T > \ell/2$ , the minimal surface jumps to a configuration that does not probe the interior of the black hole (the "thermal" region), causing the complexity to abruptly collapse to zero.
Memory Argument (Section 6): Using the replica trick on a modified circuit (with alternating dimensions $q$ $q$ and $Q \gg q$ $Q ≫ q$ ), the authors show that the subsystem "remembers" the specific gates applied to it only while the entanglement wedge contains them.
- Before $T = \ell/2$ , the density matrix distinguishes different circuit histories.
- At $T = \ell/2$ , the entanglement wedge no longer covers the entire history of the gates applied to the subsystem, and the density matrix becomes indistinguishable from the maximally mixed state regardless of the specific gates. This supports the discontinuous scenario.

D. Counting Distinguishable States (Section 7)

The authors analyze the maximum number of pairwise orthogonal density matrices with rank $r$ and dimension $d$ .

Finding: Even for states with entropy close to maximal (rank $r \approx d$ ), there exists an exponentially large number of nearly orthogonal states.
Implication: This suggests that the "phase space" available to the subsystem just before thermalization is large enough to support the high complexity predicted by the holographic model, even though the state is nearly thermal.

4. Significance and Implications

Resolution of Subsystem Complexity Dynamics: The paper provides the first rigorous lower bounds for subsystem complexity in random circuits, confirming linear growth for large subsystems and establishing the time scale ( $T \sim \ell/2$ ) for the collapse of small subsystems.
Support for Discontinuous Transitions: While rigorous proofs of the sharp transition are difficult in lattice models, the combination of holographic duality and memory analysis strongly suggests that subsystem complexity does not decay smoothly but rather collapses abruptly when the subsystem thermalizes. This contrasts with the continuous saturation often assumed in simpler models.
Distinction from Shallow Designs: The authors highlight a crucial difference between standard brickwork circuits and "shallow" unitary designs (where gates are sparse). In shallow designs, complexity can continue to grow even for small subsystems because the rank of the reduced density matrix remains low. In contrast, standard brickwork circuits drive the rank to the maximum, forcing thermalization and zero complexity.
Holographic Validation: The results provide strong non-rigorous evidence that holographic complexity proposals (specifically the CV conjecture) correctly capture the phenomenology of quantum information scrambling and thermalization in 1D systems.

Summary Conclusion

Haah and Stanford demonstrate that in chaotic 1D quantum systems, the complexity of a subsystem is governed by the size of its entanglement wedge. Large subsystems ( $\ell > L/2$ ) maintain linear complexity growth indefinitely. Small subsystems ( $\ell < L/2$ ) exhibit linear growth until $T \approx \ell/2$ , at which point they undergo a sharp, discontinuous collapse to zero complexity as they thermalize. This collapse is driven by the loss of "memory" of the specific unitary gates applied to the system, a phenomenon rigorously supported by counting arguments and holographic duality.

Growth and collapse of subsystem complexity under random unitary circuits