Effective dynamics of qubit networks via phase-covariant quantum ensembles

Technical Summary: Effective Dynamics of Qubit Networks via Phase-Covariant Quantum Ensembles

Problem Statement
Most quantum systems used for measurement and computation are open rather than isolated. While ensembles of noisy circuits provide an approximate description of complex closed systems, there is a need for a constructive procedure to generate ensembles of open-system dynamics that are rigorously constrained by the properties of the underlying closed system. Specifically, the authors seek to characterize ensembles where the individual maps may or may not possess specific symmetries (here, phase-covariance), but where the ensemble-averaged map respects these symmetries. Existing approaches often rely on random Lindbladian master equations (restricting dynamics to be time-homogeneous and Markovian) or random non-Hermitian Hamiltonians, which are generally disconnected from the internal dynamics of the specific closed system being modeled. This work aims to bridge that gap by deriving ensembles of phase-covariant maps directly from the partitioning of closed spin networks.

Methodology
The authors develop a constructive procedure to generate ensembles of phase-covariant dynamical maps associated with individual spins in closed quantum systems. The methodology proceeds in three main stages:

1. Exact Derivation in Small Systems: The authors explicitly compute single-spin dynamical maps for small XXZ networks (both ring topologies with nearest-neighbor interactions and completely connected networks) with $N$ qubits. They specialize to a class of initial states (totally uncorrelated and diagonal in the computational basis) that guarantees phase-covariant dynamics for each spin, satisfying specific thermodynamic constraints.
2. Time-Averaging and Extraction of Stationary Channels: Recognizing that dynamics in finite closed systems exhibit oscillatory features dependent on system size, the authors define a long-time averaging procedure to extract time-homogeneous (stationary) dynamics. They define the average map $\Lambda^\infty_N$ as the limit of the time-average of the ensemble average of individual maps:
   $$\Lambda^\infty_N \equiv \lim_{t \to \infty} \frac{1}{t} \int_0^t \left[ \frac{1}{N} \sum_{i=1}^N \Lambda_i(t') \right] dt'$$
   They analyze the dependence of these average maps on the initial state parameters (Bloch vector components $z_i$) and the symmetries of the Hamiltonian.
3. Distributional Construction: Using the exact results from small systems, the authors motivate the form of distributional functions for the parameters of the phase-covariant map ($\lambda_1, \lambda_3, \theta, \tau_3$). They propose using truncated Gaussian distributions where the mean parameters are set by the long-time average values derived from the XXZ networks, and the variance (scale) parameters are constrained by the observed fluctuations around these averages. This allows for the generation of arbitrary-sized ensembles of channels that mimic the statistical properties of the finite-size spin networks.

The paper also extends this framework to cases where individual maps are not phase-covariant (e.g., due to disordered Hamiltonians), demonstrating that the ensemble average can still recover phase-covariance if the disorder distribution satisfies specific symmetry constraints (e.g., even distributions of mixing angles).

Key Contributions and Results

• Derivation of Average Maps: The authors derive explicit expressions for the long-time averaged phase-covariant channels ($\Lambda^\infty_N$) for XXZ networks with $N=3$ through $N=6$. They find that the average map depends only on the initial state and the symmetries of the Hamiltonian, not on the specific coupling strengths (provided they are not commensurate in specific ways).
  • For completely connected networks, the average map is invariant under the full permutation group of the initial state.
  • For ring networks, the invariance is generally restricted to cyclic permutations, though full permutation symmetry is recovered for specific $N$ (e.g., $N=5$).
• Characterization of Fluctuations: The study quantifies the fluctuations of dynamical map parameters around their steady values. They establish that the normalized fluctuations of non-unitary parameters ($\lambda_3, \tau_3$) are bounded by $|\Delta \zeta| \leq C_\zeta t_J / (Nt)$, where $t_J$ is the dynamical timescale. This provides a physical basis for setting the variance in the proposed distributional models.
• Disordered Hamiltonians: The paper demonstrates that phase-covariance can emerge on average even when individual maps are not phase-covariant. By considering ensembles of qubit pairs with disordered Hamiltonian parameters (specifically magnetic fields and coupling strengths), they show that if the disorder distribution is symmetric (e.g., Gaussian with zero mean for mixing angles), the non-phase-covariant terms vanish in the ensemble average.
• Constructive Procedure for Arbitrary Ensembles: The authors propose a specific algorithm to generate ensembles of arbitrary size using truncated Gaussian distributions. The means of these distributions are fixed by the derived average maps, and the scales are determined by the fluctuation bounds. This procedure allows for the simulation of open-system evolution that respects the partitioning of a larger closed system.
• Comparison with General Channels: The paper compares the eigenvalue distributions of phase-covariant ensembles against more general non-phase-covariant ensembles. It highlights that breaking phase-covariance introduces additional structures in the complex eigenvalue plane, such as a lower density of points off the real axis and a higher likelihood of real eigenvalues.

Significance and Claims
The paper claims to provide a "qualitatively new approach" to ensembles of open systems. Rather than imposing loose restrictions (like the Lindblad form) disconnected from the system's internal dynamics, this method constructs ensembles that are physically motivated by the partitioning of a specific closed system.

The significance lies in the ability to:

1. Realize Random Families of Open-System Dynamics: The procedure generates ensembles constrained by global properties (initial state, Hamiltonian symmetries) that approximate a partition of a closed system.
2. Study Thermalization and Dephasing: The resulting time-independent phase-covariant channels generate a notion of thermalization where single-spin states are driven toward a specific Gibbs state (defined by an effective inverse temperature $\beta^*$), which is distinct from the maximally mixed state. This is particularly relevant for studying thermalization beyond the Markovian regime.
3. Efficient Simulation: The authors suggest that these results could open a route to the efficient simulation of a broader class of open systems, potentially leveraging the exponential advantage of quantum simulators for open-system evolution.

The authors remain modest regarding the generality of their findings, noting that while they conjecture the form of map parameters for specific topologies, a general proof for the convergence of these averages in the limit of infinitely many spins (requiring absolute convergence proofs via Fubini's Theorem) is beyond the scope of the current work. Furthermore, they observe that in their examples, the phase-covariant dynamics of 1-qubit subsystems did not appear to differ based on whether the underlying spin-chain Hamiltonian was integrable or non-integrable, suggesting a need for further study into how closed-system properties manifest in open-system ensembles.