Level Statistics — Quantum Chaos

Integrable systems:

• Harmonic oscillator: \(H = \frac{p^2}{2m} + \frac{m\omega^2 x^2}{2}\)
• Kepler problem: \(H = \frac{p^2}{2m} - \frac{k}{r}\)
• Free particle on \(S^2\): \(H = \frac{L^2}{2I}\) (only \(L_z\) is conserved in absence of perturbation)

Chaotic systems:

• Double pendulum (large amplitudes)
• Three-body gravitational problem
• Kicked top (See §subsec:kicked-top)

The Schrödinger equation is linear:

Linearity implies:

• The superposition principle holds
• Evolution is unitary (distances preserved in Hilbert space)
• No exponential divergence of “nearby” quantum states can occur

The Question: How does classical chaos manifest in quantum mechanics?

The Answer: Through spectral statistics and eigenstate properties. Classical chaos influences the statistical correlations among energy levels and the structure of eigenstates.

Regular systems (Poisson statistics):

For integrable systems, energy levels behave like independent random points on a line:

Key features:

• Levels are statistically uncorrelated
• \(P(0) = 1\): level clustering is allowed
• No level repulsion
• \(\langle s \rangle = 1\) (by definition of normalization)

Chaotic systems (Wigner-Dyson statistics):

For chaotic systems, the level spacing distribution follows:

where the Dyson index \(\beta\) depends on the symmetry class:

• \(\beta = 1\) for GOE
• \(\beta = 2\) for GUE
• \(\beta = 4\) for GSE

Key features:

• \(P(0) = 0\): strong level repulsion near \(s = 0\)
• Levels “avoid” each other with \(P(s) \propto s^\beta\) for small \(s\)
• Long-range correlations present (spectral rigidity)
• Gaussian decay for large \(s\)

Level repulsion arises from avoided level crossings. Consider a \(2 \times 2\) matrix:

The eigenvalues are:

Key insight: For \(V \neq 0\), the levels cannot cross; they “repel” each other with minimum separation \(\sim 2|V|\).

Symmetry dependence: The probability distribution of off-diagonal elements \(V\) depends on the symmetry class:

• In GOE: real \(V\), matrix is symmetric
• In GUE: complex \(V\), matrix is Hermitian
• In GSE: quaternionic \(V\), matrix is self-dual

This symmetry dependence leads to the different \(\beta\) exponents and spacing distributions.

In TMT, the connection between geometry and spectral statistics is direct:

• The \(S^2\) interface has isometry group SO(3) \(\cong\) SU(2)/\(\mathbb{Z}_2\)
• Quantum states on \(S^2\) are represented by spinors transforming under SU(2)
• The monopole (Part 5, Chapter 54) breaks time-reversal symmetry
• Therefore, chaotic dynamics on \(S^2\) generically gives GUE statistics

This connection shows that quantum chaos is not imposed externally but emerges from TMT's fundamental \(S^2\) geometry.

In spherical coordinates, the monopole field strength is:

This \(\sin\theta\) factor might suggest the time-reversal breaking varies across \(S^2\)—strong at the equator, vanishing at the poles. But this is a coordinate illusion. In polar field coordinates:

The monopole's time-reversal breaking is uniform: every patch of \(S^2\) contributes equally to symmetry violation. The \(\sin\theta\) factor in the spherical expression is entirely a Jacobian artifact from the coordinate transformation \(du = -\sin\theta\,d\theta\).

The GUE classification follows from two polar properties:

(1) Constant field strength: \(F_{u\phi} = 1/2\) means the magnetic flux through any polar rectangle \([u_1, u_2] \times [\phi_1, \phi_2]\) is proportional to its area:

(2) Flat integration measure: The ensemble average in the Gaussian weight (eq:gaussian-ensemble-measure) involves \(\mathrm{Tr}(H^2)\). On \(S^2\), trace operations reduce to integrals over \(d\Omega = du\,d\phi\)—a flat measure with no \(\sin\theta\) weighting. Every region of \(S^2\) contributes equally to the matrix element statistics.

Together: uniform field \(+\) flat measure \(=\) every degree of freedom experiences the same time-reversal breaking. No GOE subregions survive. The result is pure GUE.

Poisson (regular systems):

Linear growth indicates no long-range correlations—levels are essentially independent.

RMT (chaotic systems):

Logarithmic growth indicates strong long-range correlations. Energy levels are “rigid” like a crystal lattice, not random like a gas. Each new eigenstate is “aware” of the spacing of all nearby states.

Physical interpretation: In chaotic systems, one cannot add arbitrary levels without violating RMT constraints. Levels are correlated over energy scales of order \(\hbar\omega\), where \(\omega\) is the smallest characteristic frequency.

In the classical limit (\(j \to \infty\)), the kicked top describes a unit vector \(\vec{n} = (n_x, n_y, n_z)\) on \(S^2\) evolving via the map:

Kick phase: \(\phi \to \phi + k n_z\)

Rotation phase: \(\vec{n} \to R_y(p) \vec{n}\) (rotation around \(y\)-axis)

where \(\phi\) is the azimuthal angle conjugate to \(L_z\).

Dynamical behavior:

• For small \(k \ll 1\): Regular motion. The kick is a small perturbation to the precession. Trajectories oscillate regularly around a precessing axis.
• For intermediate \(k \sim 1\): Mixed behavior. Some orbits are regular, others chaotic.
• For large \(k \gtrsim 3\): Chaotic motion. Sensitive dependence on initial conditions. The \(S^2\) surface is densely filled by ergodic trajectories.

The critical transition occurs around \(k \approx 3\).

The classical kicked top map involves the \(z\)-component \(n_z = \cos\theta\). In polar field coordinates, \(n_z = u\), so the kick phase becomes:

This is a linear shear in the \((u, \phi)\) rectangle \([-1,+1] \times [0, 2\pi)\). What appeared as a nonlinear transformation \(\phi \to \phi + k\cos\theta\) in spherical coordinates is revealed as the simplest possible area-preserving map: a linear shear in the flat polar plane.

The full one-period map in polar coordinates:

Kick: \(\phi \to \phi + ku\) \quad (linear shear in \(u\))

Rotation: \(\vec{n} \to R_y(p)\vec{n}\) \quad (mixes \(u\) and \(\phi\))

The kicked top is a Hamiltonian system, so it preserves the symplectic form on \(S^2\). In polar coordinates, this form is:

which is the standard flat symplectic form on a rectangle. Area preservation in the \((u, \phi)\) plane is manifest: the Jacobian of the kick map is

The transition to chaos at \(k \approx 3\) corresponds to the shear becoming strong enough that the stretching-and-folding mechanism operates across the full \(u\)-range \([-1,+1]\).

Numerical studies (using exact diagonalization of the Floquet matrix for \(j \lesssim 100\)) confirm:

• For \(k = 0.1\): \(\chi^2\) test strongly rejects GUE, accepts Poisson
• For \(k = 1\): Intermediate statistics (mixture of Poisson and GUE)
• For \(k = 5\): \(\chi^2\) test strongly rejects Poisson, accepts GUE
• For \(k = 10\): Excellent agreement with GUE predictions (Wigner surmise)

The transition width decreases as \(j \to \infty\) (semiclassical limit). This validates the BGS conjecture for the kicked top and demonstrates that quantum chaos on \(S^2\) follows universal random matrix predictions.

The monopole field on \(S^2\) (Chapter 54, Berry connection) plays a crucial role in quantum chaos:

• Time-reversal breaking: The monopole field \(\vec{A}(\theta, \phi) \propto \hat{n}(\theta, \phi)\) is odd under \(T\), breaking \(T\)-symmetry \(\to\) GUE rather than GOE
• Berry phase as a tracer: The geometric phase \(\gamma_B = \frac{qg_m}{2} \Omega\) (where \(\Omega\) is the solid angle enclosed) tracks the trajectory of the \(S^2\) configuration. Chaotic trajectories explore more of \(S^2\), accumulating more sensitive Berry phase.
• Constructive/destructive interference: Different \(S^2\) paths to the same final state can interfere, with phases determined by their Berry phase difference. Chaotic paths explore more paths, enhancing quantum complexity.

Key insight: The same \(S^2\) geometry that produces quantum mechanics (via the electromagnetic duality, Part 5) also produces quantum chaos. These are two aspects of the same structure.

A powerful heuristic is Berry's random wave conjecture: In the semiclassical limit, chaotic eigenstates behave like random superpositions of plane waves:

with random phases \(a_{\vec{k}} \approx e^{i\phi_{\vec{k}}}\) where \(\phi_{\vec{k}}\) are drawn from a uniform distribution.

On \(S^2\), this becomes a random superposition of spherical harmonics:

with random amplitudes. The Husimi distribution then exhibits random Gaussian fluctuations around the mean value \(1/(4\pi)\).

This explains the eigenstate thermalization hypothesis (Chapter 72): if eigenstates look “random” (like a thermal ensemble), they thermalize locally.

The statement “eigenstates are uniformly delocalized on \(S^2\)” (Theorem thm:eigenstate-delocalization) has a subtlety in spherical coordinates: the “uniform” Husimi distribution \(Q \approx 1/(4\pi)\) must be weighted by the non-uniform measure \(\sin\theta\,d\theta\,d\phi\) when computing expectation values. Near the poles (\(\theta \to 0, \pi\)), the solid angle element shrinks, so “uniform on \(S^2\)” is not the same as “uniform in coordinates.”

In polar field coordinates, this subtlety vanishes. The integration measure is flat:

A uniform Husimi distribution \(Q(u, \phi) = 1/(4\pi)\) combined with the flat measure gives:

This is the unweighted flat average over the polar rectangle. No Jacobian corrections are needed.

In polar coordinates, the random wave ansatz (eq:random-spherical-harmonic) factorizes into AROUND (\(\phi\)) and THROUGH (\(u\)) contributions. For a chaotic eigenstate:

AROUND (gauge) sector: The \(\phi\)-dependent part of each spherical harmonic is \(e^{im\phi}\). Random superposition over \(m\) values produces uniform \(\phi\)-distribution. This is the gauge sector: \(L_z = -i\hbar\partial_\phi\) has zero mean.

THROUGH (mass) sector: The \(u\)-dependent part involves Legendre polynomials \(P_\ell^m(u)\), which are polynomials in \(u\). Random superposition produces uniform \(u\)-distribution in the flat measure. This is the mass sector: the particle explores the full THROUGH range \(u \in [-1,+1]\).

Delocalization criterion in polar form:

The transition from localized (regular) to delocalized (chaotic) is a transition from a curve in the \((u, \phi)\) rectangle to filling the entire rectangle.

Quantum scars were discovered by E.J. Heller in numerical studies of the stadium billiard—a classically chaotic billiard ball problem. Despite the classical system being fully chaotic with no periodic orbits of measure zero:

• Heller computed eigenstate Husimi distributions
• Several eigenstates showed clear scars along short unstable periodic orbits
• The scarred states were rare but undeniable deviations from random matrix predictions

This discovery was surprising: random matrix theory predicted uniform delocalization, yet here were states “remembering” classical structures. The implication is that ergodic theory (all trajectories explore the full phase space) doesn't fully capture the quantum mechanics.

The kicked top on \(S^2\) exhibits scars along multiple periodic orbits:

• Fixed points: For certain parameters, the north and south poles are fixed points (period-1 orbits). They enclose solid angle \(\Omega = 0\) or \(4\pi\).
• Period-2 orbits: Equatorial bouncing patterns (rotate by \(\pi\), return by rotation)
• Higher-period orbits: Multiply-winding paths around \(S^2\)

The monopole field contributes \(\gamma_B = \frac{q g_m}{2} \Omega\) to the resonance condition. For topological reasons (Chapter 54), the product \(q g_m\) is quantized:

This quantization affects which orbits support scars: - For \(n=1\) (minimal monopole), certain orbits resonate; others don't - The scar pattern is a signature of the monopole charge

The Berry phase in the scar resonance condition (eq:berry-phase-orbit) involves the solid angle \(\Omega_\gamma\) enclosed by the periodic orbit. In polar coordinates, the solid angle is simply the area of the enclosed region in the flat \((u, \phi)\) rectangle:

For a periodic orbit that traces a closed curve in the \((u, \phi)\) plane, this is the ordinary planar area—no metric corrections, no \(\sin\theta\) weights.

Example: A great circle orbit at colatitude \(\theta_0\) (fixed \(u_0 = \cos\theta_0\), \(\phi\) runs \(0\) to \(2\pi\)) encloses:

This is the rectangular area below \(u = u_0\) in the polar strip: width \(2\pi\), height \((1 - u_0)\) measured from the south pole \(u = -1\). The scar resonance condition becomes:

where we used \(qg_m = 2\pi n\). The Berry phase contribution is linear in \(u_0\)—the polar coordinate of the orbit.

The kicked top's periodic orbits, when mapped to the \((u, \phi)\) rectangle, have a simple geometric interpretation:

Fixed points: The poles \(u = \pm 1\) are points in the rectangle. They enclose \(\Omega = 0\) (north pole, \(u = +1\)) or \(\Omega = 4\pi\) (south pole, \(u = -1\)).

Period-2 orbits: Equatorial bouncing patterns appear as horizontal lines at \(u = 0\), enclosing \(\Omega = 2\pi\) (half the sphere = half the rectangle).

Higher-period orbits: General periodic orbits trace closed curves in the \((u, \phi)\) rectangle. The scar strength depends on the enclosed rectangular area through the Berry phase.

Key insight: The scar resonance condition selects orbits whose enclosed rectangular area \(\Omega_\gamma\) satisfies \(\pi n \Omega_\gamma/(2\pi) + S_\gamma/\hbar = 2\pi m\). The geometry of scarring is rectangle geometry in the flat polar plane.

The Eigenstate Thermalization Hypothesis (ETH, developed in Chapter 72) predicts that for chaotic systems:

For any local observable \(O\), the expectation value in a highly excited eigenstate equals the thermal (microcanonical) expectation value:

Scars violate this expectation:

• A scarred eigenstate concentrated along a periodic orbit will have different expectation values for observables sensitive to that orbit
• For example, \(\langle \psi_{\text{scar}} | L_z | \psi_{\text{scar}} \rangle\) may differ significantly from \(\langle L_z \rangle_{\text{thermal}}\) if the orbit preferentially samples certain \(L_z\) values
• These scarred states are rare (measure zero in the spectrum), but non-negligible
• They are important for dynamics: a system starting in a scarred eigenstate will exhibit slower thermalization (return time to initial state is longer)

TMT perspective: Scars occur when the \(S^2\) geometry creates phase-matching conditions (Theorem thm:scars-s2-resonance). They are not violations of quantum mechanics but manifestations of the delicate interplay between Berry phase and classical periodic orbits. The fine structure of scars encodes information about the monopole charge and the underlying \(S^2\) geometry.

A major recent development (2018+) is the discovery of many-body quantum scars in interacting spin systems:

• The PXP (Rydberg atom array) model exhibits many-body scars
• Certain initial states avoid thermalization, showing persistent oscillations (oscillations don't decay)
• Scars are connected to “towers” of scarred eigenstates that form a separate structure in the spectrum

TMT extension: Each spin in a many-body system has its own \(S^2\) (Bloch sphere). Many-body scars may arise from collective Berry phase resonances across coupled \(S^2\) interfaces:

When multiple spins have phases that collectively resonate (e.g., due to conservation laws or symmetries), the system can avoid thermalization.

Status: This is an active research frontier. A complete TMT treatment of many-body scars would require extension of the formalism to systems with many \(S^2\) components coupled through interactions.

Beyond quantum scars, other localization phenomena exist and merit mention:

Anderson localization (single-particle): Disorder in position space can localize eigenstates, preventing spreading and thermalization. The eigenstates decay exponentially away from the disorder, with localization length \(\xi(\hbar)\) depending on disorder strength.

Many-body localization (MBL): In interacting disordered systems, localization can persist even at high energy density, preventing thermalization. MBL breaks ergodicity: the system retains memory of initial conditions indefinitely.

TMT perspective: Localization in TMT would occur when the \(S^2\) configuration becomes “trapped” in a region, unable to explore the full interface. Possible mechanisms:

• Disorder barriers: If disorder on \(S^2\) creates energy barriers (e.g., random potential \(V(\theta, \phi)\)), trajectories cannot escape, localizing states
• Berry phase interference: Berry phase could create destructive interference that suppresses spreading. Paths leading away from a region accumulate incompatible phases, canceling the wavefunction.
• Emergent conservation laws: In certain systems, additional conserved quantities beyond energy emerge. These confine the particle on lower-dimensional regions of \(S^2\).

Status: A full TMT treatment of localization phenomena requires further development. The existing framework (Chapters 54-70) provides the \(S^2\) foundation; the many-body extension to handle localization is still being formulated.