Accurate Lindblad-form master equation for weakly damped quantum systems across all regimes

The paper addresses the lack of a completely positive, time-independent Markovian master equation for weakly damped quantum systems in the near-degenerate regime, where transition frequencies are distinct but separated by less than a few linewidths. Existing Lindblad master equations cover only the degenerate and far non-degenerate limits, forcing reliance on the Bloch-Redfield (B-R) equation in the intermediate regime despite its known potential to violate positivity. The authors aim to derive a single Lindblad-form master equation valid across degenerate, near-degenerate, and non-degenerate regimes under weak damping, and to determine the physical conditions (notably, slow variation of the bath spectral density) under which such a Markovian description is accurate. The work further investigates when Markovian descriptions break down, and provides an adiabatic extension for time-dependent transitions, including cases where levels cross.

Two distinct Lindblad master equations are known: one for degenerate transitions and one for non-degenerate transitions obtained via the secular (rotating-wave) approximation applied to the Bloch-Redfield (B-R) equation. However, no Lindblad formulation existed for the near-degenerate regime. The B-R equation, widely used in some fields (e.g., photochemistry), has been controversial due to non-positivity. Recent results (e.g., Eastham et al. on coupled oscillators; Jeske et al. noting reduction to degenerate Lindblad when transitions sample the same spectral density) suggest B-R can be accurate and nearly positive when dynamics are close to Markovian and the spectral density varies slowly. Other works posited the near-degenerate regime is intrinsically non-Markovian due to the absence of a Lindblad form. The authors build on this background to assess conditions under which a positive, Lindblad-form equation can be obtained without the secular approximation, unifying prior degenerate and non-degenerate results.

• Modeling framework: Standard oscillator-bath model of thermal damping with system Hamiltonian H_sys coupled linearly to a continuum of harmonic oscillators with spectral density J(ω) and high cutoff Ω. System-bath interaction decomposed into lowering operators A = ∑_j g_j σ_j and an optional diagonal dephasing term D (treated in Methods; it contributes standard Lindblad dephasing and does not affect near-degenerate analysis). Weak-damping regime is assumed, enabling a first rotating-wave approximation (RWA) in the interaction Hamiltonian.
• Step 1: System identification (SID) on exact simulations:
  • Perform exact many-body simulations (matrix-product-state methods) of a V system coupled to an Ohmic bath (J(ω) ∝ ω, sharp cutoff). Explore detunings Δω spanning degenerate through non-degenerate regimes while keeping linewidths fixed.
  • Apply subspace system identification to the reduced system trajectories obtained from multiple initial states to infer the minimal linear time-invariant model generating the dynamics of the observable subspace (upper-level populations and coherences). Extract the effective dynamical matrix D = ln[M(t)]/t from the learned propagator M(t).
  • Finding: Dynamics are effectively 4-dimensional, time-independent, and Markovian; the inferred master equation matches the structure of the degenerate Lindblad equation but with rates and Lamb shifts evaluated at the actual (possibly distinct) transition frequencies.
• Step 2: Analytical derivation from Bloch-Redfield with a Slowly Varying Spectrum (SVS) condition:
  • Start from the RWA Born-Markov derivation to obtain the Bloch-Redfield equation with frequency-dependent coefficients Γ_j = R_j + i I_j, where γ_j ∝ J(ω_j) and Lamb shifts Δ_j arise from principal-value integrals over J(ω).
  • Observe that cross terms in B-R oscillate at detunings Δω_jk and vanish for large detunings (recovering non-degenerate Lindblad). To cover near-degenerate pairs without losing positivity, impose a slow-variation condition on J(ω): J(ω_j + Δ_j) ≈ J(ω_j) over frequency scales of the Lamb shifts/linewidths. This implies Γ_j ≈ Γ_k for transitions with small detuning, allowing replacements R_j ≈ √(R_j R_k) and I_j ≈ √(I_j I_k) in cross terms and enabling a refactorization into a Lindblad form.
  • Resulting zero-temperature Lindblad master equation (Eq. 37): ρ̇ = −i[H_0 − ħ D†D, ρ] + 𝒟[Σ]ρ, with jump operator Σ = ∑_j √γ_j e^{iφ_j} σ_j and Lamb-shift operator D = ∑_j √Δ_j e^{iφ_j} σ_j, yielding bath-induced coherent couplings between upper levels when transitions are not strictly degenerate.
  • Finite-temperature extension (Eq. 50): add temperature-dependent jump operators Θ(T) and Y(T) with thermal occupations n_T(ω_j), and corresponding Hamiltonian terms built from B and C operators involving Δ_j and additional temperature-dependent Lamb shifts Δ′_j (given by principal-value integrals weighted by n_T).
• Numerical validation and scope:
  • Benchmarks at T = 0 for Ohmic bath: V system across detunings; a trident system (three transitions); two co-located qubits. Also evaluate an adiabatic, time-dependent extension where system parameters γ(ω), Δ(ω), σ_j are updated instantaneously along a slow schedule (changing detunings, including level crossings à la Landau-Zener).
  • Breakdown analysis: Employ a piecewise-linear spectral density J_r(ω) with tunable slope r in a central segment to test SVS and Markovianity via SID-inferred effective dimension. Compare the new Lindblad equation with the B-R equation and with exact simulations.

• Main theoretical result: A single Lindblad-form master equation (Eqs. 37 and 50) accurately models weakly damped quantum systems across degenerate, near-degenerate, and non-degenerate regimes, provided the bath spectral density varies slowly on frequency scales set by Lamb shifts and linewidths (SVS condition), and the first RWA/weak-coupling conditions hold. It unifies and generalizes prior degenerate and non-degenerate Lindblad equations without invoking the secular approximation across all pairs.
• When SVS is satisfied, the new Lindblad equation and the B-R equation agree to very high accuracy; the Lindblad form guarantees complete positivity and enables efficient quantum trajectory simulations.
• Numerical accuracy (Ohmic bath, T = 0):
  • V system: average error in populations and coherences is below 8×10⁻⁴ over explored detunings, including Δω = 100γ₁; captures small oscillatory features missed by the standard non-degenerate Lindblad equation.
  • Trident system: maximum population error < 2×10⁻³ over simulated time.
  • Two co-located qubits: with HRWA simulated for practical Q factors, maximum error < 5×10⁻³ over the shown duration.
• Breakdown under steep spectral slopes: Increasing the central-segment slope r of a piecewise-linear J_r(ω) distorts decay from exponential, signaling non-Markovianity. System identification shows the effective dynamical dimension increases beyond the system’s, indicating that all time-independent Markovian master equations (including B-R) fail before the Lindblad and B-R equations deviate from each other. For a two-level system, Lindblad and B-R coincide and both deviate increasingly with r; even best-fit parameter tuning cannot prevent rapid error growth. For the V system, Lindblad and B-R errors remain almost indistinguishable while both diverge from exact dynamics as r grows; B-R maintains near-positivity up to r ≈ 10 in tested cases (most negative eigenvalue magnitude < 10⁻¹²).
• Time-dependent (adiabatic) extension: For a V system with time-varying detuning, adiabatic master equation errors are 3.4×10⁻³ (gradual ramp F₁) and 1.5×10⁻² (faster-varying F₂). For a generalized Landau-Zener crossing in a 4-level system, maximum error < 5.4×10⁻³, demonstrating reliable handling of level crossings when parameter changes are sufficiently slow.
• Corollaries: (1) The secular approximation is not necessary for positivity; weak damping, high cutoff, and SVS suffice. (2) The controversy over B-R is resolved for thermal damping: under SVS, B-R is close to a Lindblad equation and approximately preserves positivity; outside SVS, B-R (and any time-independent Markovian equation) is invalid.

The work demonstrates that near-degenerate weakly damped systems can remain effectively Markovian and be captured by a time-independent Lindblad generator when the bath spectrum varies slowly on relevant frequency scales. The derived master equation reduces to the known degenerate or non-degenerate Lindblad limits as appropriate and reproduces B-R predictions where B-R is valid, while ensuring complete positivity. This addresses the central question of whether a positive, efficient master equation exists for the near-degenerate regime and clarifies the physical conditions (SVS) underpinning Markovianity in oscillator-bath models. The equation’s simpler structure provides insight into bath-induced coherent couplings between transitions and supports efficient Monte Carlo simulations. When SVS is violated (steep spectral gradients), exact simulations and SID show increased effective dynamical dimension and non-exponential decay, indicating non-Markovian behavior and the breakdown of all time-independent Markovian equations, including both B-R and the new Lindblad equation. The adiabatic extension broadens applicability to slowly time-varying Hamiltonians and crossing transitions, relevant to tasks like control of super/sub-radiance and Landau-Zener processes.

The authors present a unified Lindblad-form master equation valid across degenerate, near-degenerate, and non-degenerate regimes for weakly damped systems with slowly varying bath spectra. It replaces the Bloch-Redfield equation for thermal damping in its regime of validity, guarantees complete positivity, enables trajectory-based simulation methods, and clarifies when Markovian descriptions hold. Exact simulations verify high accuracy for Ohmic baths and show that when the SVS condition fails, all time-independent Markovian approaches break down. The adiabatic extension accurately treats systems with slowly time-dependent transition frequencies, including level crossings. Future directions include quantifying the adiabatic extension’s accuracy versus rate of parameter change, exploring applicability to other system-bath coupling models beyond oscillator quadratures, and developing improved estimates for frequency-dependent coefficients to extend accuracy toward steeper spectral variations.

• Assumes the standard oscillator-bath model with system coupling to oscillator quadratures; extension to other bath couplings remains open.
• Requires weak damping and a sufficiently high but not excessively large cutoff (to keep Lamb shifts ≪ transition frequencies); relies on an initial RWA in the interaction Hamiltonian.
• Central SVS assumption: the bath spectral density must vary slowly on scales set by Lamb shifts/linewidths near relevant transition frequencies; accuracy degrades as spectral slopes increase, leading to non-Markovian dynamics where no time-independent Markovian equation is valid.
• The adiabatic time-dependent extension is accurate only for sufficiently slow parameter changes; quantitative bounds on allowable rates are not fully characterized.
• Some numerical validations (e.g., two co-located qubits) used HRWA due to computational constraints and may require higher Q factors for the full model.