Logarithmic Scrambling in Quantum Systems

• Logarithmic scrambling time is defined as the minimal timescale for a quantum system to delocalize local information, scaling as the logarithm of its entropy.
• It is diagnosed via entanglement entropy growth, decay of out-of-time-ordered correlators, and mutual information, establishing universal lower bounds on scrambling dynamics.
• Various models—from black holes and holographic theories to quantum circuits—demonstrate the ln(N) scaling, informing quantum chaos, error correction, and experimental design.

Logarithmic scrambling time is the minimal timescale required for a quantum many-body system to delocalize initially local quantum information, distributing it irretrievably over its full Hilbert space in a time that scales as the logarithm of the system’s entropy or degrees of freedom. This phenomenon is central to the fast scrambling conjecture, which asserts that black holes and certain quantum systems saturate a universal lower bound on information scrambling, characterized by timescales of the form $t_* \sim \beta \ln S$, where $S$ is entropy and $\beta$ is the inverse temperature. Scrambling is typically diagnosed by the growth of entanglement entropy, the decay of out-of-time-ordered correlators (OTOCs), or mutual information. Fast scrambling at the logarithmic bound plays a pivotal role in quantum information theory, quantum gravity, and the phenomenology of many-body quantum chaos.

1. Precise Definitions and Operational Criteria

Two principal operational definitions of scrambling time appear recurrently:

• Entropic definition: $t_*$ is the minimal time such that, under the system dynamics, the reduced density matrix of every subsystem of size $m<N/2$ achieves nearly maximal ("Page") entanglement entropy, i.e., $S_m(t_*) \geq m \ln 2 - O(1)$ (0808.2096, Lashkari et al., 2011).
• Dynamical (OTOC) definition: Using out-of-time-ordered correlators,

$C(t) = -\langle [V(0), W(t)]^2 \rangle$

one defines $t_*$ as the time for $C(t)$ to reach $O(1)$, meaning that local operators have spread over the entire system in a nonlocal fashion (Belyansky et al., 2020, 0808.2096).

Further, the tripartite mutual information $S$0 provides a multipartite diagnostic, with $S$1 defined as the first time $S$2, indicating the emergence of nonlocal quantum correlations (Kuriyattil et al., 2023).

2. Universal Lower Bounds and Saturation Mechanisms

The fast scrambling conjecture posits a universal lower bound: $S$3 where $S$4 is the characteristic local interaction rate. All-to-all and sufficiently nonlocal models can saturate this bound, with explicit constructions yielding (0808.2096, Vikram et al., 2024, Belyansky et al., 2020, Brown et al., 2012): $S$5 where $S$6 denotes system size or Hilbert space dimension, and $S$7 is the local inverse temperature. Holographic models and matrix quantum mechanics realize this scaling exactly, with black holes conjectured to be the fastest scramblers in nature (0808.2096, Barbon et al., 2011).

Significantly, a recent universal speed limit applies to all quantum Hamiltonian dynamics, regardless of locality or interaction structure: $S$8 where $S$9 is subsystem Rényi-2 entropy, affirming the logarithmic lower bound as a property of quantum mechanics itself (Vikram et al., 2024).

3. Scrambling in Quantum Circuit, Hamiltonian, and Geometric Models

Logarithmic scrambling times arise in a diverse range of physical models:

• Random quantum circuits and Brownian circuits: Circuits with parallel two-qubit gates on complete graphs, or Brownian time-dependent Hamiltonians, scramble in $\beta$0 layers or steps (Lashkari et al., 2011, Brown et al., 2012, Kuriyattil et al., 2023, Hashizume et al., 2021).
• Sparse and long-range graphs: Models on expander graphs, random regular graphs, and hypercube graphs (realized in neutral atom arrays) achieve $\beta$1 scrambling time provided the interaction diameter is $\beta$2 (Bentsen et al., 2018, Hashizume et al., 2021, Kuriyattil et al., 2023).
• Minimal fast-scrambling Hamiltonians: All-to-all Ising interactions normalized as $\beta$3 plus local chaos suffice for $\beta$4 scaling (Belyansky et al., 2020).
• Ultrametric and geometrically motivated processes: Classical diffusion on ultrametric trees with logarithmic barrier heights directly reproduces $\beta$5 for horizon degrees of freedom in black holes (Barbon et al., 2013). Similarly, kinetic models of null geodesics in optical AdS near-horizon regions yield the same scaling via Lyapunov exponents for classical chaos on hyperbolic billiards (Barbon et al., 2011).

These models underpin the claim that logarithmic scrambling can be achieved with a minimal set of ingredients: sufficiently nonlocal coupling and some mechanism for rapid entanglement generation.

4. Scrambling Across Physical and Mathematical Settings

Black holes and holography: Multiple derivations, including the AdS/CFT correspondence, shock-wave analysis in TFD states, and membrane-paradigm calculations, show that black hole horizons scramble information in time $\beta$6 (0808.2096, Barbon et al., 2011, Wang et al., 2022, Shor, 2018). This is a crucial element in the black hole information problem—the logarithmic bound ensures that Hawking radiation can carry away information without violating no-cloning or relativistic causality (Shor, 2018). For 2D large-$\beta$7 CFTs, the scrambling time from OTOCs at finite temperature is $\beta$8 when the identity block dominates (Liu et al., 2018).

Many-body localization: In disordered, interacting spin chains, scrambling is drastically slowed—OTOCs decay logarithmically in time, $\beta$9, so the arrival time at distance $t_*$0 is $t_*$1: a "logarithmic light cone" (Chen, 2016, Kim et al., 2023, Sahu et al., 2024). The logarithmic rate $t_*$2 analogizes a Lyapunov exponent.

Floquet time crystals: Scrambling emerges only after exponentially long frozen phase evolution, with subsequent logarithmic-in-time OTOC and entanglement growth, controlled by disorder strength (Sahu et al., 2024).

RG flows and model dependence: While fast scrambling typically accelerates in the IR as degrees of freedom are integrated out, certain holographic models with additional IR matter content display slower scrambling, as additional IR degrees increase the effective Hilbert space (Kundu, 2020).

5. Information-Theoretic and Quantum Coding Consequences

Logarithmic-depth circuits suffice to:

• Generate Page-scrambled states and achieve locally indistinguishable reduced marginals on all subsystems up to size $t_*$3 (Brown et al., 2012, Hashizume et al., 2021).
• Realize good quantum error correction and entanglement-assisted erasure codes (encoding $t_*$4 logical into $t_*$5 physical qubits) with only $t_*$6 circuit depth (Brown et al., 2012).
• Implement deterministic, efficiently programmable fast scramblers in neutral atom arrays, using only $t_*$7 interaction layers (Hashizume et al., 2021).

The link between entropy growth, decoupling, and error correction capacity is codified in circuit and Hamiltonian models respect these optimal timescales.

6. Hierarchies of Scrambling: Designs, Rényi Entropy, and Max-Scrambling

A refined hierarchy distinguishes between different strengths of "scrambling" by associating the order-$t_*$8 Rényi entanglement entropy with unitary $t_*$9-designs:

• Partial scrambling ($m<N/2$0): 2-designs achieve near-maximal second Rényi entropy—relevant for OTOC decay and operator growth.
• Max-scrambling ($m<N/2$1): Only $m<N/2$2-designs suffice to maximize the min-entropy, making all reduced marginals indistinguishable from Haar (Liu et al., 2017). Max-scrambling (maximal min-entropy) thus constitutes a strong form of "Haar-randomness" achieved in minimal time for appropriately constructed circuits.

This framework generalizes the fast scrambling conjecture to "max-scrambling in time roughy linear in the number of degrees of freedom," enlarging the set of meaningful scrambling measures.

7. Limitations, Universality, and Open Problems

• Universality: The logarithmic lower bound is protected by quantum unitarity and analyticity of the spectral form factor—for any quantum Hamiltonian, regardless of locality or randomness (Vikram et al., 2024). Graph diameter and Cheeger constant control whether sparse models saturate the bound (Bentsen et al., 2018).
• Limitations: Operationally, some definitions (e.g., strong "two-hemisphere" entanglement) appear incompatible with black hole physics unless nonstandard physics acts outside the event horizon (Shor, 2018).
• Many-body localization: Logarithmic scrambling does not mean "fast" scrambling—rather, it is the slowest possible delocalization consistent with exponentially localized integrals of motion (Chen, 2016, Kim et al., 2023).
• Experimental realizability: Neutral atom and cavity QED systems now approach the threshold for direct observation of logarithmic scrambling times (Hashizume et al., 2021, Kuriyattil et al., 2023), including table-top analogues of black hole scrambling (Wang et al., 2022).
• Open Questions: The precise relation between scrambling, thermalization, late-stage chaos, and the role of conserved quantities remains under investigation, especially in systems outside the holographic class or with disorder (Kaikov, 2022).

---

Table: Summary of Key Models and Scrambling Time Scaling

Physical Model	Scrambling Time $m<N/2$3	Reference
Random circuit (complete graph)	$m<N/2$4	(Brown et al., 2012)
All-to-all Ising + local chaos	$m<N/2$5	(Belyansky et al., 2020)
Matrix quantum mechanics (black hole)	$m<N/2$6	(0808.2096)
Holographic CFT/AdS black hole	$m<N/2$7	(Barbon et al., 2011)
Sparse expander graph	$m<N/2$8	(Bentsen et al., 2018)
MBL spin chain	$m<N/2$9	(Chen, 2016)
Quantum time crystals (FTC-MBL)	$S_m(t_*) \geq m \ln 2 - O(1)$0	(Sahu et al., 2024)
Neutral atom hypercube circuits	$S_m(t_*) \geq m \ln 2 - O(1)$1 layers	(Hashizume et al., 2021)

---

Logarithmic scrambling time thus frames a universal speed limit for quantum delocalization processes, with implications for black hole physics, quantum information, and experimental realizations of quantum chaos. Whether via OTOCs, entanglement entropy, quantum codes, or explicit dynamical models, the emergence of the $S_m(t_*) \geq m \ln 2 - O(1)$2 law signifies the maximal efficiency with which physical systems can entangle and hide quantum information.