arXiv:2610.40808
Liouvillian spectral snapshots: A reduced-subspace approach for time-dependent and noisy open quantum systems
Hans Christiansen, André Melo, Jens Paaske, Fabrizio Minganti
Abstract
Driven-dissipative quantum devices are routinely operated with time-dependent drives and exposed to temporally correlated classical noise. When these perturbations vary on timescales comparable to the system's intrinsic dynamics, rotating-wave and Markovian approximations can fail, making effective static descriptions unreliable. As a result, accurately modeling systems in this regime often relies on simulating the full time-dependent master equation. This is particularly demanding for processes requiring long integration times, such as rare switching between metastable states, or for stochastic perturbations requiring averaging over many noise realizations. To address these challenges, we introduce Liouvillian spectral snapshots (LSS), a reduced-basis method that enables efficient simulation of such systems by projecting the master equation onto a low-dimensional subspace capturing their coarse-grained dynamics. LSS constructs this subspace by computing the slowly decaying, low-frequency eigenoperators of the Liouvillian at a few adaptively sampled perturbation values and compressing them into a common basis. This basis is built once and can be reused across noise realizations and perturbation parameters, reducing the cost of ensemble averages and parameter sweeps.
AI Summary
Simulating driven open systems with temporally correlated classical noise can require long integrations and many realizations, particularly when the quantities of interest are rare switching events or logical relaxation rates. This paper introduces Liouvillian spectral snapshots (LSS), a projection-based reduction targeting these slow dynamics without replacing the perturbation by a static or Markovian model. LSS computes slow, low-frequency eigenoperators of instantaneous Liouvillians at adaptively sampled perturbation values, combines them using singular-value decomposition, and propagates the master equation in the resulting fixed basis. A greedy residual test controls enrichment across the sampled parameter range. The basis can be reused across noise realizations and spectra; identity inclusion and conjugation symmetry enforce trace and Hermiticity preservation. Benchmarks compare LSS with full-space evolution. For a bistable Kerr resonator, bases of roughly $100$ operators replace a $1024$-dimensional Liouville space. Floquet switching rates agree within $1.5\%$ at three tested detunings. Lorentzian detuning noise resonant with a long-lived oscillating mode enhances switching by about an order of magnitude; one resonant benchmark has a $3.4\%$ rate discrepancy. For an eleven-qubit continuous repetition-code model, a $37$-dimensional basis replaces $4^{11}\approx4.2\times10^6$ dimensions and closely reproduces tested logical trajectories. Accuracy depends on spectral separation and perturbation-induced leakage into discarded modes. Tightening the residual tolerance does not remove spectral-window truncation errors, and the test can miss slow modes with negligible basis overlap. The paper derives a trace-distance bound from the integrated projection residual and a perturbative leakage estimate. The projected generator is not constrained to Lindblad form.
Error found: The paper extracts a long-time switching rate from one finite-time noise-averaged propagator without establishing that the extracted rate has converged with integration time, although colored-noise averages generally lack the semigroup property. The paper's general argument that discarded-mode dissipation limits long-time propagation errors overlooks feedback that can shift retained slow rates even when fast-sector population stays small; its first-order leakage estimate does not establish saturation of the total error.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 22 pages, 12 figures