Markovian Master Equations

Markovian master equations describe reduced quantum dynamics when relevant environmental memory can be neglected on the timescale of interest. Their central structural form is the Lindblad–Gorini–Kossakowski–Sudarshan equation, usually shortened to Lindblad or GKSL form.

In one common convention,

$$\begin{aligned} \frac{d\rho}{dt} ={}& -\frac{i}{\hbar}[H,\rho] \\ &+ \sum_k\gamma_k \left( L_k\rho L_k^\dagger - \frac12 \{L_k^\dagger L_k,\rho\} \right), \end{aligned}$$

with nonnegative rates $\gamma_k$. The Hamiltonian term generates coherent motion. Each dissipator describes one representation of noise, damping, pumping, dephasing, or engineered dissipation.

The form guarantees a physical completely positive, trace-preserving evolution under its mathematical assumptions. It does not, by itself, prove that a chosen $H$, $L_k$, and $\gamma_k$ accurately model a particular apparatus, bath, or timescale.

Four Layers to Keep Separate

The phrase “Lindblad equation” is used for several related claims.

Layer	Claim	What it does not establish
structural theorem	a continuous CPTP semigroup generator has GKSL form	which generator a physical device has
working master equation	specified $H$, $L_k$, and rates define a CPTP model	microscopic validity of those choices
weak-coupling derivation	Born–Markov–secular assumptions produce rates and Lamb shifts	validity outside the stated regime
trajectory unraveling	a monitoring scheme resolves the generator into stochastic records	a unique underlying sequence of jumps

Confusing these layers leads to many common errors. A phenomenological GKSL equation can be useful without a microscopic derivation. A microscopic derivation can fail even though the resulting equation is in GKSL form. One unconditional generator can admit many different trajectory unravelings.

Reading Path

Read this page	Use it for
Lindblad–GKSL Equation	Learning the working equation, trace preservation, and standard examples.
Lindblad Operators	Interpreting decay, dephasing, pumping, collective noise, and representation freedom.
Lindblad Theorem	Understanding the finite-dimensional semigroup theorem and its assumptions.
Quantum Dynamical Semigroups	Working with time-homogeneous composition, generators, and relaxation modes.
Detailed Balance	Relating upward and downward rates to thermal equilibrium and KMS spectra.
Thermal Master Equations	Constructing weak-coupling generators with Gibbs steady states.
Quantum Optical Master Equation	Modeling atoms, cavities, and light-matter systems coupled to broadband fields.
Optical Bloch Equations	Solving a driven two-level system with relaxation, detuning, and dephasing.
Pauli Rate Equations	Reducing decoupled population dynamics to a classical master equation.
Pure Dephasing Master Equation	Modeling phase-coherence loss without population transfer.
Amplitude Damping Master Equation	Modeling spontaneous emission, qubit $T_1$ decay, and oscillator loss.
Steady States and Relaxation	Finding fixed states, Liouvillian modes, gaps, dark states, and metastability.

A compact first route is

$$\begin{gathered} \text{GKSL equation} \longrightarrow \text{Lindblad operators}, \\ \text{semigroups} \longrightarrow \text{steady states}, \\ \text{dephasing and damping examples}. \end{gathered}$$

Then choose thermal, optical, or rate-equation pages according to the application.

Dissipator Notation

It is convenient to define

$$\mathcal D[L](\rho) = L\rho L^\dagger - \frac12 \{L^\dagger L,\rho\}.$$

The generator is then

$$\mathcal L(\rho) = -\frac{i}{\hbar}[H,\rho] + \sum_k\gamma_k\mathcal D[L_k](\rho).$$

Some authors absorb $\sqrt{\gamma_k}$ into $L_k$. Others place factors of two in the definition of $\mathcal D$. Rate comparisons are meaningless until this convention is fixed.

The anticommutator term is essential. It balances the gain term $L\rho L^\dagger$ so that trace is preserved:

$$\operatorname{Tr}\mathcal D[L](\rho)=0.$$

Complete positivity is not apparent from an arbitrary differential equation, but the GKSL structure ensures that the finite-time maps generated by a time-independent $\mathcal L$ are CPTP.

Semigroup Setting

A quantum dynamical semigroup is a family of channels satisfying

$$\Phi_0=\operatorname{id}, \qquad \Phi_{t+s}=\Phi_t\Phi_s, \qquad t,s\ge0.$$

For a time-independent generator,

$$\Phi_t=e^{t\mathcal L}.$$

The semigroup law expresses time homogeneity and memoryless composition: evolution over the next interval depends only on its duration and the present state. The maps need not have physical CPTP inverses, so this is a semigroup rather than a group.

The Lindblad Theorem states, in finite dimensions and with the appropriate continuity assumptions, that a CPTP semigroup generator has GKSL form and that a GKSL generator produces a CPTP semigroup.

The theorem is structural. It does not derive weak coupling, a flat spectrum, a temperature, or a particular decay rate.

Time-Dependent GKSL Form

A time-local equation can also have time-dependent coefficients:

$$\frac{d\rho}{dt} = \mathcal L_t(\rho).$$

If $\mathcal L_t$ has instantaneous GKSL form with nonnegative rates for every time, the evolution is CP-divisible under standard regularity conditions. It is generally not a time-homogeneous semigroup because the generator changes with time.

If an instantaneous rate becomes negative, the evolution is not CP-divisible over that interval in the ordinary canonical representation. The finite-time map from the initial time can still be completely positive. Thus

$$\begin{gathered} \text{negative instantaneous rate} \\ \text{does not automatically mean} \\ \text{unphysical finite-time evolution}. \end{gathered}$$

Time-local, semigroup, CP-divisible, and Markovian are related but not interchangeable terms.

Origin from Microscopic Models

A standard weak-coupling route begins with

$$H = H_S+H_E + \sum_\alpha S_\alpha\otimes B_\alpha.$$

The derivation then uses distinct steps:

1. The Born approximation keeps the bath near a reference state to the retained perturbative order.
2. The Markov approximation uses short bath memory relative to reduced evolution.
3. System operators are decomposed into Bohr-frequency components.
4. Secular or controlled coarse-graining removes rapidly rotating cross-frequency terms.
5. Positive rate matrices are diagonalized to obtain Lindblad operators and rates.
6. Principal-value terms contribute a Lamb-shift Hamiltonian.

Each step has its own validity condition. Lindblad form at the end does not retroactively validate weak coupling, bath stationarity, or secularization. Near degeneracies, strong coupling, structured reservoirs, low-frequency noise, and correlated initial states require special care.

Use Open Quantum Systems for the derivation architecture and the Approximation Checklist for practical audits.

Lindblad Operators Are Not Unique Events

The operators $L_k$ identify how the environment couples to the system in one representation. Typical choices include

Operator	Common model role
$\sigma_z$	qubit dephasing
$\sigma_-$	qubit relaxation or photon emission
$\sigma_+$	incoherent excitation or thermal pumping
$a$	bosonic-mode loss
$a^\dagger$	bosonic-mode gain
$n=a^\dagger a$	oscillator number dephasing
$\sum_j\sigma_-^{(j)}$	collective emission

But the representation is not unique. A unitary mixing

$$L_j' = \sum_k u_{jk}L_k$$

leaves the total dissipator unchanged when the rates have been absorbed into the operators. Shifts by identity can also be compensated by a Hamiltonian change.

Therefore a Lindblad operator is not automatically an observable, a unique detector event, or a unique microscopic bath mode. Physical event labels require a specified monitoring scheme or dilation.

Jump and No-Jump View

One possible quantum-jump unraveling introduces an effective non-Hermitian Hamiltonian

$$H_{\mathrm{eff}} = H - \frac{i\hbar}{2} \sum_k\gamma_kL_k^\dagger L_k.$$

Over a short interval $dt$, a normalized pure state has jump probability

$$dp_k = \gamma_k \langle L_k^\dagger L_k\rangle dt.$$

Conditioned on a jump of type $k$, the state is updated by $L_k$ and renormalized. Conditioned on no detected jump, it evolves under $H_{\mathrm{eff}}$ and is renormalized. Averaging the trajectories recovers the unconditional GKSL equation.

This is an unraveling, not a unique hidden history. Direct, homodyne, and heterodyne monitoring of the same output field can generate jump-like or diffusive records with the same unconditional master equation. The canonical stochastic treatment belongs to Continuous Measurement and Quantum Trajectories.

Pure Dephasing

A common qubit pure-dephasing equation is

$$\frac{d\rho}{dt} = -\frac{i}{\hbar}[H,\rho] + \frac{\gamma_\phi}{2} (\sigma_z\rho\sigma_z-\rho).$$

When $H$ commutes with $\sigma_z$, the populations are constant and

$$\rho_{01}(t) = e^{-\gamma_\phi t} e^{-i\omega_0t} \rho_{01}(0)$$

in the displayed convention. The basis and factor-of-two convention must be stated.

This generator does not thermalize the populations. It is the Markovian limit of more general phase-noise models that can have nonexponential envelopes or revivals.

Amplitude Damping

Zero-temperature qubit relaxation is generated by

$$\frac{d\rho}{dt} = -\frac{i}{\hbar}[H,\rho] + \gamma_1\mathcal D[\sigma_-](\rho).$$

The excited population and coherence obey

$$\begin{aligned} \rho_{ee}(t) &=e^{-\gamma_1t}\rho_{ee}(0), \\ \rho_{eg}(t) &=e^{-\gamma_1t/2} e^{-i\omega_0t} \rho_{eg}(0). \end{aligned}$$

Population relaxation therefore contributes half its rate to coherence decay. With independent pure dephasing,

$$\frac1{T_2} = \frac1{2T_1} + \frac1{T_\phi}.$$

At finite temperature, upward and downward operators are both required. Zero-temperature amplitude damping is not a thermal model at arbitrary temperature.

Thermal Generators and Detailed Balance

For a two-level transition of frequency $\omega_0>0$, equilibrium detailed balance requires

$$\frac{\Gamma_\uparrow} {\Gamma_\downarrow} = e^{-\beta\hbar\omega_0}.$$

The stationary population ratio then agrees with the Gibbs state. For a bosonic mode coupled to a thermal bath,

$$\begin{aligned} \frac{d\rho}{dt} ={}& -\frac{i}{\hbar}[H,\rho] \\ &+ \kappa(\bar n+1)\mathcal D[a](\rho) \\ &+ \kappa\bar n\mathcal D[a^\dagger](\rho). \end{aligned}$$

Detailed balance is stronger than merely having a steady state. Driven systems, multiple reservoirs, feedback, and engineered dissipation can have stationary states that are not Gibbs states and do not satisfy equilibrium detailed balance.

The global-versus-local master-equation distinction also matters for coupled subsystems. Local dissipators written in a bare basis can violate thermodynamic consistency when internal coupling is not negligible relative to level splittings and bath resolution.

Quantum Optical and Optical Bloch Equations

The quantum optical master equation applies GKSL dynamics to atoms, cavities, and light-matter systems coupled to broadband electromagnetic reservoirs. Standard terms include spontaneous emission, cavity loss, thermal excitation, collective decay, and coherent drive.

For a driven qubit in a rotating frame, the Hamiltonian can be written

$$H_{mathrm{rot}} = -\frac{\hbar\Delta}{2}\sigma_z + \frac{\hbar\Omega}{2}\sigma_x,$$

with detuning $\Delta$ and Rabi frequency $\Omega$. Adding amplitude damping and pure dephasing yields the optical Bloch equations for the Bloch components. They describe saturation, power broadening, transient Rabi oscillations, and steady-state spectroscopy.

The equations are a reduced unconditional model. A single photon-counting record follows a stochastic trajectory, not the smooth ensemble-averaged Bloch vector.

Pauli Rate Equations

When populations decouple from coherences, the diagonal probabilities can obey

$$\dot p_n = \sum_{m\ne n} \left( W_{n\leftarrow m}p_m - W_{m\leftarrow n}p_n \right).$$

This is a classical continuous-time Markov equation embedded in the quantum dynamics. It can arise after secularization or when only populations are relevant.

Population rates do not determine the full quantum generator. Many different dephasing and coherence-transfer models can share the same Pauli equation. Near degeneracies, discarding coherences can remove important quantum transport effects.

Steady States and Relaxation

A steady state satisfies

$$\mathcal L(\rho_{\mathrm{ss}})=0.$$

If right eigenoperators obey

$$\mathcal L(R_j)=\lambda_jR_j,$$

then modes with $\operatorname{Re}\lambda_j<0$ decay. The zero-eigenvalue subspace contains stationary operators. A Liouvillian gap is set by the nonzero eigenvalue with real part closest to zero and often controls the slowest asymptotic relaxation.

Important distinctions include:

• stationary versus attracting;
• unique versus multiple steady states;
• dark states versus states reached from a given initial condition;
• equilibrium versus nonequilibrium steady states;
• true degeneracy versus long-lived metastability;
• physical convergence versus truncation artifacts in oscillator models.

Conserved quantities and symmetries can produce several stationary sectors. Engineered Lindblad operators can make a target state dark, but dark-state status alone does not prove that every initial state converges to it.

Validating a Markovian Model

GKSL form handles mathematical consistency, not empirical adequacy. For a proposed model, ask:

1. Are the Hilbert space and truncation physically adequate?
2. Which bath correlations and spectral densities generate the rates?
3. Is the bath correlation time short relative to system relaxation?
4. Are the Born and secular approximations justified?
5. Are nearly degenerate transitions grouped correctly?
6. Is temperature represented by both upward and downward processes?
7. Are Lamb shifts or drive-frame transformations included consistently?
8. Are collective and correlated decay channels required?
9. Does the generator reproduce known symmetry, steady-state, and limiting behavior?
10. Are fitted rates stable across observables and experimental protocols?

Numerically, verify trace, Hermiticity, positivity, cutoff convergence, steady-state residuals, and agreement between direct integration and independent benchmarks.

Canonical Boundaries

This chapter owns Markovian generators, their structural theorem, and standard finite-dimensional rate models.

• Open Quantum Systems owns microscopic Hamiltonians and the Born–Markov–secular derivation chain.
• Quantum Channels and Noise owns finite-time CPTP maps and channel representations.
• Non-Markovian Dynamics owns divisibility failures, information backflow, memory measures, and structured-environment methods.
• Continuous Measurement and Quantum Trajectories owns conditional stochastic states and detector records.
• Quantum Noise, Dissipation, and Baths owns correlation spectra, Langevin equations, input-output theory, and canonical bath models.
• Quantum Control owns reservoir engineering and dissipative state preparation as design tasks.
• Quantum Thermodynamics owns heat, work, entropy production, and fluctuation relations.

Use these boundaries to avoid presenting one Lindblad model as a complete microscopic or thermodynamic account.

Common Mistakes

• Calling every time-local equation Markovian.
• Treating GKSL form as proof that a microscopic derivation is valid.
• Omitting the anticommutator term or using inconsistent rate conventions.
• Interpreting Lindblad operators as unique observables or events.
• Assuming one unconditional generator has one trajectory unraveling.
• Forgetting upward thermal transitions.
• Calling pure dephasing thermalization.
• Equating a steady state with a Gibbs state or an attracting state.
• Secularizing across nearly degenerate transitions without a scale check.
• Using local thermal dissipators outside their regime.
• Ignoring collective decay and correlated noise.
• Trusting a truncated bosonic steady state without convergence tests.
• Inferring the full quantum generator from population rates alone.

Exercises

Trace preservation of one dissipator

Show that $\operatorname{Tr}\mathcal D[L](\rho)=0$ for every operator $L$ and trace-class $\rho$.

Solution

Using cyclicity of the trace,

$$\begin{aligned} \operatorname{Tr}\mathcal D[L](\rho) ={}& \operatorname{Tr}(L\rho L^\dagger) \\ &- \frac12 \operatorname{Tr}(L^\dagger L\rho) \\ &- \frac12 \operatorname{Tr}(\rho L^\dagger L) \\ ={}&0. \end{aligned}$$

The Hamiltonian commutator also has zero trace, so the full GKSL generator preserves trace.

Pure-dephasing coherence

For $H=0$ and

$$\dot\rho = \frac{\gamma}{2} (\sigma_z\rho\sigma_z-\rho),$$

derive the equations for $\rho_{00}$ and $\rho_{01}$.

Solution

Conjugation by $\sigma_z$ leaves the diagonal entries unchanged and changes the sign of the off-diagonal entries. Therefore

$$\dot\rho_{00}=0, \qquad \dot\rho_{01}=-\gamma\rho_{01}.$$

Hence

$$\rho_{01}(t) = e^{-\gamma t}\rho_{01}(0),$$

while both populations remain constant.

Amplitude-damping fixed state

For $\dot\rho=\gamma\mathcal D[\sigma_-](\rho)$, identify the unique stationary density operator and explain why the maximally mixed state is not stationary.

Solution

The ground state is annihilated by $\sigma_-$, so

$$\mathcal D[\sigma_-] (|g\rangle\langle g|)=0.$$

Every excited-state population decays toward the ground state, making $|g\rangle\langle g|$ the unique stationary density operator for the two-level zero-temperature model.

The channel is not unital. Acting on $I/2$ transfers excited population to the ground state, so $I/2$ is not stationary.

Thermal two-level state

For upward and downward rates $\Gamma_\uparrow$ and $\Gamma_\downarrow$, solve the stationary population ratio and impose detailed balance.

Solution

The population equation is

$$\dot p_e = \Gamma_\uparrow p_g - \Gamma_\downarrow p_e.$$

At stationarity,

$$\frac{p_e}{p_g} = \frac{\Gamma_\uparrow} {\Gamma_\downarrow}.$$

Thermal detailed balance gives

$$\frac{p_e}{p_g} = e^{-\beta\hbar\omega_0},$$

which is the Gibbs ratio for the two levels.

Semigroup decay law

A dephasing channel has coherence factor $f(t)$ and satisfies the time-homogeneous semigroup law. Assuming continuity, $f(0)=1$, and nonzero $f(t)$, determine its form.

Solution

Composition requires

$$f(t+s)=f(t)f(s).$$

The continuous nonzero solutions are exponential:

$$f(t)=e^{ct}.$$

Complete positivity requires the coherence magnitude not to grow beyond one for $t\ge0$, so $\operatorname{Re}c\le0$. Writing $c=-\gamma-i\omega$ gives exponential dephasing with a Hamiltonian phase:

$$f(t)=e^{-\gamma t}e^{-i\omega t}.$$

Cross-Links

• Applications and Experimental Platforms
• Computational Notebooks
• Reference
• Lindblad–GKSL Equation
• Lindblad Operators
• Lindblad Theorem
• Quantum Dynamical Semigroups
• Detailed Balance
• Thermal Master Equations
• Quantum Optical Master Equation
• Optical Bloch Equations
• Pauli Rate Equations
• Pure Dephasing Master Equation
• Amplitude Damping Master Equation
• Steady States and Relaxation
• Open Quantum Systems
• Non-Markovian Dynamics
• Quantum Jump Trajectories
• Solving Lindblad Equations
• Formula Sheet

References

• G. Lindblad, “On the generators of quantum dynamical semigroups,” Communications in Mathematical Physics 48, 119–130 (1976).
• V. Gorini, A. Kossakowski, and E. C. G. Sudarshan, “Completely positive dynamical semigroups of N-level systems,” Journal of Mathematical Physics 17, 821–825 (1976).
• E. B. Davies, Quantum Theory of Open Systems, Academic Press (1976).
• R. Alicki and K. Lendi, Quantum Dynamical Semigroups and Applications, 2nd ed., Springer (2007).
• H.-P. Breuer and F. Petruccione, The Theory of Open Quantum Systems, Oxford University Press (2002).
• Á. Rivas and S. F. Huelga, Open Quantum Systems: An Introduction, Springer (2012).
• C. W. Gardiner and P. Zoller, Quantum Noise, 3rd ed., Springer (2004).
• H. J. Carmichael, An Open Systems Approach to Quantum Optics, Springer (1993).