This lecture explores the statistical properties of quantum chaotic systems, establishing deep connections with random matrix theory (RMT). It begins by classifying quantum states based on unitary and anti-unitary symmetries, particularly time-reversal symmetry, which dictates whether eigenstates are generically real or complex. The statistical distribution of these "random states" is then quantified using the inverse participation ratio, illustrating how chaos leads to delocalization, with distinct behaviors for time-reversal symmetric and broken systems. The talk then delves into eigenvalue correlations, demonstrating the phenomenon of level repulsion in chaotic systems through nearest-neighbor spacing distributions (Wigner-Dyson vs. Poisson) and the spectral form factor. Transitioning to many-body physics, the lecture discusses entanglement as a basis-independent measure, introducing the Page value for typical random states and its emergence in chaotic coupled systems. It also examines many-body localization, revealing how entanglement patterns change across phases. Finally, the discussion touches upon Out-of-Time-Order Correlators (OTOCs) as a quantum analogue for detecting scrambling and chaos, highlighting the interdisciplinary nature and ongoing challenges in this field.
Exploring Quantum Chaos: From Random States to Entanglement and Scrambling
0:00 - Introduction to Quantum Chaos Statistics: The lecture begins by reviewing quantum chaotic eigenstates in single-particle systems (e.g., standard map, baker's map), noting they can appear as random states or be scarred by periodic orbits. The primary focus will be on the statistical aspects of these "random states."
Unitary vs. Anti-unitary Symmetries: Unitary symmetries include geometric transformations (parity, rotation), while anti-unitary symmetries are typically associated with time-reversal.
Time-Reversal Symmetry (TRS): In quantum mechanics, TRS involves an anti-linear operator that leaves position states unchanged but reverses momentum.
Generically Real Eigenstates (Bosons): For bosonic systems, if the flow operator possesses TRS (operator squared = +1), it implies symmetric unitary flow, leading to eigenstates that are generically real in a time-adapted basis.
6:05 - Characterizing Real Random States with IPR:
Definition: For systems with TRS, eigenstates are treated as "real random states" – uniformly distributed on an N-sphere, subject only to normalization.
Gaussian Amplitude Distribution: The probability distribution for a single amplitude component is Gaussian.
Inverse Participation Ratio (IPR): This measure quantifies the spread of an eigenfunction. For large dimensions (N) and real random states, the average IPR is 3/N, indicating that roughly 1/3 of the components are significant. This is presented as a "quantum effect."
11:37 - IPR in the Standard Map: The lecture demonstrates that as the chaos parameter (K) in the standard map increases, the IPR transitions from showing high localization (integrable regime) to fluctuating around the 1/3 value (chaotic regime), confirming the random state model. The initial basis dependence disappears in the chaotic limit.
15:04 - Complex Random States (Broken TRS): When TRS is broken (e.g., by a magnetic flux), eigenstates become complex. In this case, the average IPR for complex random states is 2/N, implying more delocalization (about 1/2 of components are significant) compared to real states. This is also demonstrated using the standard map.
Unitary Symmetries: Introduce symmetry sectors (e.g., parity) that block-diagonalize the Hamiltonian, allowing for separate analysis of eigenvalue spectra.
Level Crossings vs. Avoided Crossings: In the integrable regime (small K), eigenvalue plots show many crossings. In the chaotic regime (large K), these turn into "avoided crossings," a phenomenon known as level repulsion.
Level Velocities: The slopes of "straight line" features in eigenvalue plots (vs. K) correspond to highly localized states influenced by stable orbits, with these slopes being important in metrology as "level velocities."
28:41 - Statistical Properties of Energy Level Spacings:
Poisson vs. Wigner-Dyson: Integrable systems (e.g., circle billiard) typically exhibit Poisson statistics (exponential decay) for nearest-neighbor spacing (NNS), indicating uncorrelated levels. Chaotic systems (e.g., stadium billiard) follow Wigner-Dyson statistics, characterized by level repulsion (NNS vanishing at small separations).
Universality in Chaos: The ratio of successive spacings provides a universal measure to distinguish integrable (0.39), time-reversal symmetric chaotic (0.52), and time-reversal broken chaotic (~0.6) systems.
Gaussian Orthogonal Ensemble (GOE): RMT models quantum chaotic Hamiltonians as ensembles of random matrices. For time-reversal symmetric systems, real symmetric matrices (GOE) are used, with Gaussian-distributed elements.
Invariant Ensemble: The joint probability distribution of matrix elements is invariant under orthogonal transformations, making it basis-independent.
Eigenvalue Repulsion: The determinant of the Jacobian transformation from matrix elements to eigenvalues introduces a Vandermonde factor, which causes eigenvalues to repel each other.
Wigner Surmise: For a 2x2 GOE matrix, the spacing distribution is linear in separation (s * e^(-s²)), providing the basis for Wigner's surmise for larger matrices.
Wigner Semicircle Law: The density of states for large random matrices follows a semicircle law, but this is generally not universal for physical Hamiltonians; universality lies in spectral fluctuations (like spacing), not the overall density.
GUE: For time-reversal breaking systems, Hermitian matrices (GUE) are used, resulting in even stronger level repulsion (s² * e^(-s²)).
Circular Ensembles: These model flow operators (unitary matrices). Their eigenangle spacing distributions also follow Wigner-Dyson statistics, demonstrating the broad universality of RMT.
Two-Point Correlations: The SFF measures correlations between all pairs of eigenvalues, calculated from the trace of a time-evolved operator.
Distinct Behavior: SFF exhibits characteristic behaviors for different ensembles (e.g., linear ramp for GUE, a 'dip' for GOE, flat for Poisson).
Non-Self-Averaging: Unlike NNS, SFF for individual quantum chaotic systems is not self-averaging; it requires ensemble averaging (e.g., over a range of K values for the standard map) to match RMT predictions.
Basis-Independent Measures: In many-body systems, entanglement is a crucial measure for quantum chaos, as it is independent of the single-particle basis.
Definition of Entanglement: States that cannot be written as a product of subsystem states are entangled.
Von Neumann Entropy: The Von Neumann entropy of the reduced density matrix (obtained by tracing out part of a bipartite system) quantifies entanglement.
1:20:29 - Average Entanglement and the Page Value:
Nearly Maximally Entangled: For typical "random states" (e.g., coefficients drawn from a Gaussian distribution), the average entanglement (measured by Von Neumann entropy) approaches the maximum possible value (log N) minus a small factor (e.g., log N - 1/2 for time-reversal breaking systems), known as the Page value.
Chaos Creates Entanglement: Coupled standard maps demonstrate that entanglement grows with increased coupling and classical chaos, eventually saturating at the Page value. Both initially unentangled and maximally entangled states evolve towards this typical Page value over time, indicating quantum thermalization.
1:25:30 - Monogamy of Entanglement and Many-Body Localization (MBL):
Monogamy: A key quantum property is that if two subsystems are maximally entangled, they cannot be entangled with other parts of the system.
Multipartite Entanglement: In chaotic many-body systems, due to monogamy, entanglement tends to be "multipartite" (spread across many particles), resulting in low entanglement between small subsystems (e.g., individual spins).
MBL Transition: The Many-Body Localization (MBL) model shows a transition from an integrable phase (Poisson NNS, high local entanglement) to a chaotic phase (Wigner NNS, low local entanglement but high multipartite) and then to a localized phase (Poisson NNS again, but with persistent local entanglement).
Quantum Lyapunov Exponent: OTOCs, measuring the squared commutator of operators at different times, serve as a quantum analogue to classical Poisson brackets and Lyapunov exponents.
Exponential Growth: In some quantum chaotic models (e.g., kicked top) with large-N or semiclassical limits, the logarithm of the OTOC grows linearly in time, providing a "quantum Lyapunov exponent."
Ongoing Research: However, for purely quantum systems (e.g., spin-1/2 models without classical limits), OTOCs often show only polynomial, not exponential, growth, indicating that the definition of quantum chaos and scrambling in purely quantum regimes is still an active research area.
1:49:20 - Discussion and Key Takeaways: The lecture concludes by reiterating that RMT provides a robust statistical framework for understanding quantum chaos, particularly regarding spectral fluctuations and entanglement patterns. It emphasizes the interdisciplinary nature of the field and the ongoing quest to understand phenomena like scrambling and thermalization in many-body quantum systems, even in the absence of a classical limit. The universality in quantum chaos lies in fluctuations, not necessarily in the overall density of states.