Quantum Scars, MBL, Chaos Bound

The Berry connection on \(S^{2}\) with monopole is:

By Stokes' theorem:

where \(\Sigma\) is the surface bounded by \(\gamma\). \(\blacksquare\)

□

Step 1: In the semiclassical approximation, the wavefunction near a periodic orbit is a sum over repetitions:

where \(r\) is the number of traversals.

Step 2: The sum converges due to orbit instability (each traversal contributes a factor \(\sim e^{-\lambda_{\gamma}T_{\gamma}}\)):

Step 3: When the resonance condition is satisfied, all terms in the sum have the same phase, giving constructive interference:

Step 4: Off-resonance, the phases are random and the sum is suppressed by \(O(1/\sqrt{r_{\max}})\) due to destructive interference.

Step 5: The amplitude \(A_{\gamma}^{(1)} \sim 1/(2j+1)^{1/2}\) from WKB normalization, giving the scar enhancement \(\sim 1/\sqrt{2j+1}\). \(\blacksquare\)

□

Step 1: The resonance condition \(\phi_{\gamma} = 2\pi n\) defines a codimension-1 surface in the space of periodic orbits.

Step 2: The width of the near-resonance region (where scars are visible) is \(\Delta\phi \sim 1/\sqrt{j}\) (the quantum uncertainty).

Step 3: For orbits with action \(S_{\gamma} \sim E T_{\gamma}\), the number of orbits with period \(\lesssim T\) is \(N(T) \sim e^{hT}/(hT)\).

Step 4: The number of resonant orbits is:

Step 5: At fixed energy, \(T_{\max} \sim \ln(j)/\lambda_{L}\) (from semiclassical density of states). Therefore:

Step 6: The scar fraction is:

where \(S = \ln(2j+1)\) is the entropy. \(\blacksquare\)

□

Step 1: Define the scar strength as the excess probability along the orbit:

where the tube has width \(\sim 1/\sqrt{j}\).

Step 2: For resonant orbits, the semiclassical amplitude squared is:

The \(\sinh\) factor comes from summing the geometric series of amplitudes.

Step 3: The amplitude \(A_{\gamma}\) depends on the WKB prefactor, which varies randomly between eigenstates. For Gaussian random wave eigenstates, \(|A_{\gamma}|^{2}\) follows an exponential distribution.

Step 4: The exponential distribution of \(|A_{\gamma}|^{2}\) implies exponential distribution of \(\Xi_{\gamma}\):

with mean \(\bar{\Xi} \propto 1/[\sqrt{2j+1}\sinh(\lambda_{\gamma}T_{\gamma}/2)]\). \(\blacksquare\)

□

Step 1 (Thermal phase): In the absence of strong disorder, the energy shell \(\Sigma_{E}\) is connected. Ergodicity implies that almost all trajectories explore the entire shell. The microcanonical measure covers extensive volume \(\sim e^{NS}\).

Step 2 (Disorder creates barriers): Strong disorder \(W\) modifies the energy function. Local minima and saddle points appear in \((S^{2})^{N}\).

Step 3 (Barrier heights): The effective barrier height between configuration regions scales as:

where \(\beta\) depends on the disorder correlation structure.

Step 4 (Confinement criterion): Classical trajectories at energy \(E\) cannot overcome barriers with \(\Delta E > E - E_{\min}\). When barrier heights exceed available kinetic energy, the trajectory is confined.

Step 5 (MBL from confinement): Confined trajectories explore only a local region of \((S^{2})^{N}\). The Husimi distribution of eigenstates localizes to this region:

This breaks ETH and produces MBL. \(\blacksquare\)

□

Step 1 (Percolation in high dimensions): Percolation on \((S^{2})^{N}\) is effectively percolation in dimension \(2N\).

Step 2 (High-dimensional percolation): In \(d \gg 1\) dimensions, the percolation threshold approaches the mean-field value:

Step 3 (Translation to MBL): The percolation probability \(p\) corresponds to the fraction of passable barriers. This fraction decreases with disorder:

Step 4 (Critical condition): MBL occurs when \(p(W_{c}) = p_{c}(2N)\):

Step 5 (Dimension dependence): In low physical dimensions, correlations reduce the effective configuration space dimension, modifying this scaling. \(\blacksquare\)

□

We provide three complementary derivations.

Derivation 1: Resolution Argument

Step 1 (Quantum resolution): The minimum resolvable separation on \(S^{2}\) is set by the uncertainty principle:

where \(L \sim \hbar j\) is the characteristic angular momentum.

Step 2 (Thermal resolution): The thermal smearing \((\Delta\Omega)_{T}\) provides a maximum meaningful separation. Beyond this, differences are masked by thermal fluctuations.

Step 3 (Lyapunov growth): Chaos causes exponential separation of initially close trajectories:

Step 4 (Effective chaos): Meaningful Lyapunov growth can only occur while:

This gives the scrambling time:

Step 5 (Consistency with thermal time): The scrambling time cannot be shorter than the thermal correlation time \(\tau_{T}\), since thermal equilibration is required to define chaos:

Step 6 (Bound derivation): Combining Steps 4–5, the scrambling time satisfies \(t_* \geq \tau_T\), and since \(t_* = S/\lambda_L\):

This resolution argument gives \(\lambda_L \lesssim S k_B T/\hbar\), which is weaker than the MSS bound by a factor of \(S\). The tighter bound \(\lambda_L \leq 2\pi k_BT/\hbar\) (independent of \(S\)) requires the analyticity argument of Derivation 2.

Derivation 2: Imaginary Time Periodicity

Step 1 (Thermal field theory): At finite temperature, time becomes periodic in the imaginary direction:

The imaginary time circle has circumference \(\beta\hbar = \hbar/(k_{B}T)\).

Step 2 (OTOC analyticity): The out-of-time-ordered correlator \(C(t)\) can be analytically continued to the complex time plane. The KMS condition (thermal periodicity) constrains \(C(t - i\tau)\) to be analytic in the strip \(0 \leq \tau \leq \beta\hbar/2\) (not the full thermal circle, because the OTOC involves operators at shifted imaginary times \(\beta\hbar/4\)).

Step 3 (Analyticity strip constraint): Following Maldacena, Shenker, and Stanford (2016): the exponential growth \(1 - C(t) \sim e^{\lambda_{L}t}\), when continued to \(t - i\tau\), must remain bounded in the strip \(|\tau| \leq \beta\hbar/4\). The maximum growth rate compatible with analyticity in this strip is:

Step 4 (Origin of \(2\pi\)): The factor \(2\pi\) arises because the analytic continuation must respect the half-period \(\beta\hbar/2\) of the thermal boundary condition. The growth rate \(e^{\lambda_L t}\), analytically continued over a half-strip of width \(\beta\hbar/4\), gives \(e^{\lambda_L \beta\hbar/4} \lesssim O(1)\) at the boundary, yielding \(\lambda_L \lesssim 4/(\beta\hbar)\). The precise \(2\pi\) follows from a careful Phragmén–Lindelöf argument on the strip (see MSS, Appendix A).

This is the MSS bound.

Derivation 3: \(S^{2}\) Coherent State Evolution

Step 1 (Coherent state): Consider a coherent state \(|\Omega_{0}\rangle\) on \(S^{2}\) with width \(\sim 1/\sqrt{j}\).

Step 2 (Thermal spreading): In a thermal bath at temperature \(T\), the coherent state width grows due to thermal fluctuations. The equilibrium width is:

where \(E_{0} \sim m R^{2}\omega^{2}\) is the characteristic energy.

Step 3 (Chaos-induced spreading): Classical chaos spreads the coherent state exponentially:

Step 4 (Thermal saturation): Spreading saturates when \((\Delta\Omega)(t) \sim (\Delta\Omega)_{\text{eq}}\). This occurs at the thermalization time:

Step 5 (Quantum bound): Quantum mechanics requires \(t_{\text{th}} \geq \hbar/(k_{B}T)\) (time-energy uncertainty). For \((\Delta\Omega)_{\text{eq}}/(\Delta\Omega)_{0} \sim \sqrt{j}\):

This gives \(\lambda_L \lesssim k_BT\ln\sqrt{j}/\hbar\) for any \(j\). Derivations 1 and 3 both yield bounds that depend on system size; only Derivation 2 (the MSS analyticity strip argument) produces the universal, \(S\)-independent bound \(\lambda_L \leq 2\pi k_BT/\hbar\). Derivations 1 and 3 provide complementary physical intuition for why such a bound must exist. \(\blacksquare\)

□