Linearized Density-Matrix Master Eq

• Linearized density-matrix master equations are linear differential forms that describe open quantum system dynamics using approximations like Born, Markov, and rotating-wave.
• These formulations encompass methods such as Redfield, GKLS/Lindblad, and non-Markovian approaches to capture both coherent and dissipative evolution.
• Efficient numerical strategies, including vectorization and real-valued expansions, optimize computations for applications in quantum transport and terahertz laser models.

A linearized density-matrix master equation is an evolution equation for a density matrix in which the dynamics is cast as a linear differential problem, typically after a perturbative truncation in the system–environment coupling, a Markov or Born approximation, or a basis transformation that converts operator evolution into linear equations for expansion coefficients. In open-system practice, a canonical form is

$$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$

with $\mathcal{R}_t$ a dissipative superoperator, while in basis-expanded or vectorized formulations the same dynamics becomes a linear system for a coefficient vector or Bloch vector (Becker et al., 2022, Liniov et al., 2018, Kunold, 2023). The term therefore refers less to a single equation than to a class of reduced-density-matrix descriptions spanning Redfield, GKLS/Lindblad, non-Markovian kernel equations, and real-valued coefficient systems.

1. Canonical form and the meaning of linearization

In weak-coupling open-system theory, the standard linearized structure is the second-order quantum master equation

$$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$

where the coherent part is generated by the system Hamiltonian and the dissipative part is encoded in a Redfield-type superoperator (Becker et al., 2022). In field-theoretic treatments, the reduced density matrix can also be written in an exact operator form before approximation. Within thermo field dynamics and the Schwinger–Keldysh formalism, the exact reduced evolution obeys

$$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$

and the “linearized” equation arises after expanding the influence action perturbatively, typically to second order in the coupling (Käding et al., 11 Mar 2025).

The same term is also used for algebraic embeddings of density-matrix dynamics. In a Lie-algebra basis, one writes

$$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$

which yields

$$i\hbar \frac{d}{dt}\boldsymbol{\rho}(t)=\mathcal{H}(t)\boldsymbol{\rho}(t).$$

In the standard column-stacking representation this becomes

$$i\hbar\frac{d}{dt}|\boldsymbol{\rho}\rangle = [I\otimes H(t)-H^\top(t)\otimes I]\,|\boldsymbol{\rho}\rangle,$$

so that the von Neumann equation is “linearized” into an ordinary linear system on a larger vector space (Kunold, 2023). Taken together, these formulations use linearization in two closely related senses: perturbative truncation of reduced dynamics, and linear embedding of operator evolution.

2. Perturbative derivations from reduced dynamics

A large part of the modern literature derives linearized density-matrix equations from an exact system–environment problem by expanding in the interaction strength. In the TFD–Schwinger–Keldysh construction, the influence action operator is expanded as a cumulant-like series, and inserting this into the exact reduced evolution produces a perturbative master equation with connected correlator terms. In this framework the resulting equation is local in time, and the non-Markovian character is carried by time-dependent coefficients rather than by explicit time integrals over past density matrices (Käding et al., 11 Mar 2025).

An inflationary example follows the same logic. For a factorized initial state,

$$\rho_I(\eta_0)=\rho_{I\chi}(\eta_0)\otimes \rho_{I\psi}(\eta_0),$$

the full interaction-picture density matrix is expanded iteratively, the environment is traced out, and a quantum master equation is obtained up to $\mathcal{O}(\lambda^2)$. The derivation assumes weak coupling and uses Born and Markov approximations, with the reduced density matrix $\rho_r(\eta')$ replaced by $\mathcal{R}_t$0 inside the memory integral (Boyanovsky, 2015).

A distinct route starts from a naive perturbative solution of the von Neumann equation and applies coarse graining plus renormalization-group resummation. For a total Hamiltonian

$\mathcal{R}_t$1

weak coupling, a stationary bath, and short bath correlation time $\mathcal{R}_t$2 lead to a coarse-grained discrete evolution. The renormalization-group step removes explicit dependence on the initial time and yields a Markovian semigroup master equation in GKLS form (Nambu et al., 2015). A related field-theoretical program formulates reduced dynamics through a transmission matrix satisfying a Dyson equation with an irreducible kernel. In Born approximation, followed by a condensed-matter pole or quasiparticle-type approximation and then the rotating-wave approximation, the general non-Markovian equation reduces first to a Markov master equation and then to Lindblad form (Fogedby, 2022).

3. Markovian structure, steady states, and canonical consistency

The most familiar linearized density-matrix master equation is the Lindblad or GKLS equation,

$\mathcal{R}_t$3

whose semigroup structure preserves trace and complete positivity (Gu et al., 2024, Nambu et al., 2015). In diagrammatic derivations, this form is obtained only after Born, Markov, timescale-separation, and rotating-wave approximations; without the rotating-wave step, one generally remains at a Markov master equation that is not yet in Lindblad form (Fogedby, 2022).

The limitations of standard linearized weak-coupling equations are most explicit in the comparison between Redfield, Lindblad, and canonically corrected equations. The Redfield equation can violate positivity and yields incorrect steady states at nonzero coupling, whereas the secular Lindblad equation produces a steady-state density matrix that is independent of the system–bath coupling. The canonically consistent quantum master equation modifies the Redfield dissipator to

$\mathcal{R}_t$4

where $\mathcal{R}_t$5 is determined by the second-order correction to the reduced mean force Gibbs state. This construction makes the mean force Gibbs state stationary up to $\mathcal{R}_t$6, improves the transient dynamics, and suppresses positivity violations, although it does not guarantee complete positivity. In the harmonic-oscillator benchmark the method matches the exact solution extremely well, reduces the trace distance across a large parameter range, and remains accurate provided the second-order expansion is valid, with benchmarks quoted for $\mathcal{R}_t$7 (Becker et al., 2022).

A separate consistency issue appears for PT-symmetric non-Hermitian Hamiltonians. There the linear Lindblad equation is preserved only if one evolves the generalized density matrix

$\mathcal{R}_t$8

and imposes pseudo-Hermiticity on the Lindblad operators,

$\mathcal{R}_t$9

Using the normalized density matrix

$$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$0

instead introduces nonlinear terms and breaks the linearity requirement. The generalized density matrix is therefore the correct carrier of linear master-equation dynamics in that setting (Ohlsson et al., 2020).

4. Non-Markovian formulations and memory

Linearized density-matrix master equations are not synonymous with Markovianity. In the TFD–Schwinger–Keldysh treatment of open systems, the dynamical map is not divisible in a general basis, which identifies the dynamics as non-Markovian, yet the perturbative master equation remains local in time and contains no time integrals over past density matrices. Memory is encoded in the coefficients through connected correlation functions (Käding et al., 11 Mar 2025). This directly contradicts the common assumption that a local-in-time master equation must be Markovian.

In the field-theoretical transmission-matrix formulation, non-Markovianity appears through the kernel in the Dyson equation,

$$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$1

where the irreducible kernel is built diagrammatically. The approach is explicitly designed to incorporate secular effects and to remain independent of the initial preparation, which distinguishes it from standard projection-operator derivations (Fogedby, 2022).

Quantum transport of fermionic carriers provides a concrete single-particle-density-matrix realization. After formally integrating out reservoir–system coherence, the reduced SPDM satisfies a non-Markovian equation with a memory kernel. In the Markov approximation the SPDM equation is solvable analytically, but in the Born approximation without Markov reduction the problem becomes a Redfield-form algebraic equation for the stationary density matrix. The non-Markovian equation yields resonant transport similar to Landauer’s conductance (Maksimov et al., 2022).

Cosmological perturbations offer another non-Lindbladian example. An explicit interaction between system and environment produces a reduced-density-matrix equation with fluctuation and dissipation terms analogous to quantum Brownian motion. The equation is not in Lindblad form, and positivity can be violated on sub-horizon scales, but typical solutions become positive on super-horizon scales. In that regime the density matrix admits a physically meaningful stochastic interpretation, and a Langevin equation with Gaussian white noise can be written for the emergent classical trajectories (Hollowood et al., 2017).

5. Real-valued expansions, vectorization, and integration algorithms

One major computational strategy is to unfold the master equation into a real linear system. Expanding the density operator over generators of $$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$2,

$$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$3

turns the Lindblad master equation into a non-homogeneous system of $$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$4 real-valued linear differential equations,

$$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$5

The matrix $$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$6 is real and skew-symmetric, and the construction preserves normalization and Hermiticity. The paper emphasizes that a straightforward implementation scales as $$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$7, but that complexity can be reduced when the number of dissipative operators is independent of $$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$8; the reported algorithm handles a model with $$\partial_t \varrho(t)= -i[H_S,\varrho(t)] + \mathcal{R}_t[\varrho(t)],$$9 states on a single node of a computer cluster (Liniov et al., 2018).

A closely related approach uses Lie-algebraic vectorization. The standard stacked representation,

$$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$0

is identified as a special case of a more general algebra-dependent linearization. Hermitian algebras, especially the Pauli-string algebra, yield real density-matrix coefficients and substantially simplify tomography. On that basis a quantum algorithm was proposed that avoids Trotterization for the Pauli-string case and was demonstrated for two toy Hamiltonians using the IBM noisy quantum circuit simulator (Kunold, 2023).

For Markovian problems, an integration method based directly on the Lindbladian exponential has also been proposed: $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$1 Truncating this Taylor series yields a practical propagator that is mathematically equivalent to vectorization but keeps all operations at the level of $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$2 matrices. The stated numerical complexity is $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$3, compared with $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$4 for vectorization, and the method integrates naturally with tensor networks. Its validity was illustrated for damped Rabi oscillations in a two-level system and for a driven dissipative Heisenberg chain (Gu et al., 2024).

An alternative to direct density-matrix propagation is the non-stochastic open system Schrödinger equation. There one evolves a matrix of auxiliary wave-functions $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$5 such that

$$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$6

with the compatibility condition

$$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$7

This reformulation guarantees positive semi-definiteness by construction and applies to both Markovian and non-Markovian master equations (Joubert-Doriol et al., 2014).

6. Applications, domain-specific forms, and limits of linearization

The density-matrix Lindblad framework has been adapted to a quantum free-electron laser by modeling the electron as a quantum two-level system with spontaneous emission included through a dissipator. In the full reduced system, the populations $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$8, $$\partial_t \hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle = i\,\partial_t[\widehat{S}_\mathcal{S}(t)+\widehat{S}_{IF}(t)]\,\hat{\rho}_\mathcal{S}^+(t)|1_\mathcal{S}\rangle,$$9, the coherence $$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$0, and the scaled field amplitude $$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$1 satisfy coupled equations. In the linear regime, with $$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$2, $$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$3, and $$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$4, the field obeys the second-order linear equation

$$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$5

The model is stated to remain valid for $$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$6, beyond which spontaneous emission dominates and the two-level approximation becomes inaccurate (Fares et al., 2018).

In Coulomb-coupled quantum-dot arrays, the microscopic density-matrix formulation reduces to a linear quantum master equation in the sequential tunneling regime. The diagonal density-matrix elements obey population equations, while off-diagonal terms decay rapidly and are adiabatically eliminated. The resulting state probabilities satisfy a coupled system of linear first-order differential equations,

$$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$7

with rates that preserve Coulomb blockade and state-dependent energy shifts (Singha, 2020).

For terahertz quantum cascade lasers, a Markovian master equation for the single-electron density matrix was derived that preserves positivity, retains coherences, resolves in-plane dynamics, and includes phonon and impurity scattering. The formal structure is linear in the density matrix,

$$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$8

and simulations show very good agreement with both experiment and nonequilibrium Green’s functions at optimal lasing bias. The reported results also show that the magnitude of coherences can be a significant fraction of the diagonal matrix elements and that the in-plane energy distribution can deviate far from a heated Maxwellian distribution (Jonasson et al., 2016).

Not every density-matrix kinetic equation is linear. For Bogoliubov–BCS quasiparticles, the general master equation

$$\rho(t)=\sum_i \rho_i(t) g_i,\qquad H(t)=\sum_i a_i(t) g_i,$$9

is described as not linear in general, and the BdG structure is preserved only if gain and loss operators satisfy a specific symmetry constraint (Huang et al., 2010). For identical particles in mean-field approximation, the reduced single-particle equation contains explicit $$i\hbar \frac{d}{dt}\boldsymbol{\rho}(t)=\mathcal{H}(t)\boldsymbol{\rho}(t).$$0 factors,

$$i\hbar \frac{d}{dt}\boldsymbol{\rho}(t)=\mathcal{H}(t)\boldsymbol{\rho}(t).$$1

so the dynamics is nonlinear because of Pauli blocking (Bondarev, 2013). This suggests that linearization is often an approximation scheme or a representation choice, not an intrinsic property of density-matrix dynamics itself.