Approximation Checklist

This checklist is for deriving, reading, or debugging an open-system master equation. It is most useful when a proposed equation claims to describe weak coupling to a bath, a measurement apparatus, a reservoir, or an engineered noise source.

The goal is not to force every problem into a Lindblad equation. The goal is to make the assumptions visible: what is ignored, which time scales are separated, whether the map is completely positive, and what physical limits the equation should reproduce.

For formulas used below, see the Formula Sheet. For the Hamiltonian starting point, see System–Bath Hamiltonians. For terminology, see Baths, Reservoirs, and Environments. For the weak-coupling factorization step, see Born Approximation; for the short-memory step, see Markov Approximation; for Bohr-frequency averaging, see Secular Approximation. For the canonical distinction between exact reduced dynamics and approximate generators, see Reduced Dynamics and Lindblad–GKSL Equation.

One-Minute Audit

Before trusting a master equation, ask:

• What is the system, what is the environment, and where is the boundary?
• Is the initial state factorized, correlated, thermal, stationary, or prepared by a prior measurement?
• What small parameter controls the approximation?
• What is the bath correlation time?
• What are the relevant system time scales and Bohr-frequency splittings?
• Is the equation trace preserving and Hermiticity preserving?
• Is it positive or completely positive on the stated domain?
• Does it have the expected steady state?
• Does it reduce to the correct limits when the coupling, temperature, or drive is turned off?

If any answer is “not specified,” treat the equation as a model ansatz until the missing assumption is supplied.

Step 1: Define the Boundary

Start with an explicit split:

$$\mathcal H_{\text{tot}} = \mathcal H_S\otimes\mathcal H_B.$$

Write the Hamiltonian as

$$H = H_S+H_B+H_I,$$

with interaction

$$H_I = \lambda \sum_\alpha A_\alpha\otimes B_\alpha.$$

The coupling parameter $\lambda$ may be literal, such as a weak dipole coupling, or bookkeeping, such as the scale of matrix elements in $H_I$. The split is not unique: moving a mean bath force into $H_S$ can change the apparent interaction.

If

$$\operatorname{Tr}_B(B_\alpha\rho_B)\ne 0,$$

usually shift

$$B_\alpha \mapsto B_\alpha - \operatorname{Tr}_B(B_\alpha\rho_B)I_B$$

and absorb the corresponding mean-field term into $H_S$. Otherwise a first-order Hamiltonian correction may be hidden inside a “dissipative” derivation.

Step 2: State the Initial Condition

The simplest weak-coupling derivations assume a factorized initial state:

$$\rho_{SB}(0) = \rho_S(0)\otimes\rho_B.$$

They also usually assume the bath state is stationary:

$$[H_B,\rho_B]=0.$$

If the initial state is correlated,

$$\rho_{SB}(0)\ne\rho_S(0)\otimes\rho_B,$$

then the reduced dynamics for arbitrary system inputs may fail to be a completely positive map. This does not mean the total evolution is unphysical; it means the reduced description has a restricted preparation domain. See Initial Correlations for assignment maps and initial-slip caveats.

Record whether the state is:

• prepared independently from the bath,
• prepared by a prior measurement,
• already equilibrated with the bath,
• driven into a nonequilibrium steady state,
• conditioned on a measurement record.

Those cases lead to different reduced descriptions even when the same symbols $H_S$, $H_B$, and $H_I$ appear.

Step 3: Identify the Prediction Target

Do not derive a master equation before deciding what it predicts.

An unconditional state obeys an equation for the ensemble state $\rho(t)$. A conditional state depends on a measurement record. A selected subensemble may evolve under a trace-nonincreasing operation. These are different objects:

$$\rho(t), \qquad \rho(t\mid \text{record}), \qquad \mathcal I_x(\rho).$$

Confusing these is one of the fastest ways to misread measurement backaction. See Selective and Nonselective Measurements and Quantum Instruments for the operational distinction.

Step 4: Compute Bath Correlations

In the interaction picture, bath correlation functions have the form

$$C_{\alpha\beta}(t) = \operatorname{Tr}_B \!\left[ B_\alpha(t)B_\beta(0)\rho_B \right],$$

where

$$B_\alpha(t) = e^{iH_Bt/\hbar} B_\alpha e^{-iH_Bt/\hbar}.$$

Estimate a correlation time $\tau_B$ from the decay of $C_{\alpha\beta}(t)$. If the bath is structured, finite, nearly resonant, glassy, critical, or low dimensional, this decay may be slow or oscillatory rather than short and featureless.

The corresponding two-sided spectrum is

$$\Gamma_{\alpha\beta}(\omega) = \int_{-\infty}^{\infty} dt\, e^{i\omega t} C_{\alpha\beta}(t).$$

In a thermal bath, detailed balance is encoded in the KMS relation. For a single bath operator with common spectrum conventions,

$$S_{BB}(-\omega) = e^{-\beta\hbar\omega}S_{BB}(\omega).$$

If the derived rates do not satisfy the expected temperature relation, check the Fourier-transform convention and the sign convention for Bohr frequencies before concluding that the physics is wrong.

Step 5: Identify System Frequencies and Time Scales

Diagonalize the unperturbed system Hamiltonian if the derivation uses energy eigenoperators:

$$H_S = \sum_\epsilon \epsilon\,\Pi(\epsilon).$$

The Bohr-frequency components of a system coupling operator are

$$A_\alpha(\omega) = \sum_{\epsilon'-\epsilon=\hbar\omega} \Pi(\epsilon)A_\alpha\Pi(\epsilon').$$

They satisfy

$$[H_S,A_\alpha(\omega)] = -\hbar\omega A_\alpha(\omega).$$

List the relevant scales:

• $\tau_B$: bath correlation time,
• $\tau_S$: time scale for resolved system evolution,
• $\Gamma^{-1}$: relaxation or dephasing time,
• $|\omega-\omega'|^{-1}$: beat time between distinct transition frequencies,
• $\Omega_R^{-1}$: drive or Rabi time if the system is driven.

The approximation is easiest to trust when the hierarchy is clear. It becomes delicate near degeneracies, exceptional points, strong drives, band edges, and critical baths.

Step 6: Born Approximation

The Born Approximation assumes weak coupling and weak buildup of system–bath correlations. In a common form, it replaces the evolving total interaction-picture state by

$$\rho_{SB}^{I}(t) \approx \rho_S^{I}(t)\otimes\rho_B.$$

This is not the claim that the true state never entangles with the bath. It is the claim that, to the retained order in the coupling, the bath remains effectively stationary and system–bath correlations do not need their own dynamical variables.

Practical checks:

• The dimensionless coupling is small compared with the relevant system and bath scales.
• The bath is large enough or strongly mixing enough not to be appreciably depleted.
• Relaxation and dephasing rates are slow compared with microscopic bath dynamics.
• Backaction on the bath does not accumulate over the time interval of interest.

Warning signs:

• small environments,
• strong coupling,
• narrow resonant modes,
• repeated collisions with the same ancilla,
• long-lived system–bath correlations,
• low-temperature structured reservoirs.

Step 7: Markov Approximation

The Markov Approximation replaces memory-dependent evolution by time-local evolution. Schematically, a memory integral of the form

$$\int_0^t ds\, C(s)\,\rho_S(t-s)$$

is approximated by

$$\rho_S(t) \int_0^\infty ds\, C(s),$$

when bath correlations decay on a time $\tau_B$ much shorter than the system state changes.

Useful checks:

$$\tau_B\ll \Gamma^{-1}, \qquad \tau_B\ll \tau_S.$$

If the system is driven, the drive can add a new slow or fast time scale. A Markov approximation in the lab frame may not be equivalent to one in a rotating frame.

Warning signs:

• algebraic correlation tails,
• photonic band gaps or band edges,
• finite-size recurrences,
• low-frequency $1/f$ noise,
• strong non-equilibrium reservoirs,
• coarse graining comparable to the bath memory time.

Step 8: Secular Approximation

After decomposing system operators into Bohr frequencies, the Secular Approximation addresses terms oscillating as

$$e^{i(\omega'-\omega)t}.$$

The secular approximation drops terms with $\omega\ne\omega'$ when their oscillations average out on the dissipative time scale:

$$|\omega-\omega'|\gg \Gamma.$$

After full secularization, a weak-coupling thermal-bath generator commonly has the structure

$$\frac{d\rho}{dt} = -\frac{i}{\hbar}[H_S+H_{\text{LS}},\rho] + \sum_{\omega,\alpha,\beta} \Gamma_{\alpha\beta}(\omega) \left( A_\beta(\omega)\rho A_\alpha^\dagger(\omega) - \frac{1}{2} \{A_\alpha^\dagger(\omega)A_\beta(\omega),\rho\} \right).$$

The rate matrix $\Gamma_{\alpha\beta}(\omega)$ should be positive semidefinite for each $\omega$ in the usual weak-coupling construction. This positivity is what allows diagonalization into Lindblad operators.

If the secular equation decouples populations from coherences and only populations are being modeled, the reduced description is a Pauli Rate Equation.

Use caution when:

• transitions are degenerate or nearly degenerate,
• the secular approximation destroys physically important coherences,
• the system is driven and the relevant frequencies are quasienergies,
• a partial-secular approximation is used.

A nonsecular Redfield Equation can be accurate for short and intermediate times while failing complete positivity outside its validity regime. A secular Lindblad equation can be completely positive while losing important near-degenerate coherence physics. The choice is a physics judgment, not just an algebraic preference.

Step 9: Trace and Hermiticity Checks

A legitimate density-operator equation must preserve trace and Hermiticity on its intended domain.

For a time-local equation

$$\frac{d\rho}{dt} = \mathcal L(\rho),$$

trace preservation means

$$\operatorname{Tr}\mathcal L(\rho)=0$$

for all allowed $\rho$.

Hermiticity preservation means

$$\mathcal L(\rho^\dagger) = \mathcal L(\rho)^\dagger.$$

For Lindblad form, both properties are automatic if $H=H^\dagger$. For an equation assembled from rates and matrices by hand, check them explicitly.

Step 10: Positivity and Complete Positivity

Positivity means $\rho(t)\ge 0$ whenever $\rho(0)\ge 0$. Complete positivity means positivity remains true when the system is entangled with an arbitrary spectator:

$$(\Phi_t\otimes\mathrm{id}_R)(X)\ge 0 \quad \text{whenever} \quad X\ge 0.$$

For a time-independent Lindblad–GKSL generator, complete positivity of the semigroup follows from the Lindblad theorem. For a finite-time map, the Kraus representation or Choi matrix is the cleanest test.

If a time-local equation is written as

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

then $\gamma_k(t)\ge 0$ for all $k,t$ is sufficient for CP-divisible Markovian evolution. Temporarily negative rates can occur in legitimate non-Markovian descriptions, but then the equation should not be advertised as a standard Markovian Lindblad equation.

For general approximate equations, test positivity on:

• pure states,
• eigenstates of $H_S$,
• coherent superpositions of near-degenerate states,
• maximally entangled states with a reference system,
• states near the boundary of the density-operator set.

See Completely Positive Maps for why positivity on the system alone is not enough.

Step 11: Steady State and Detailed Balance

Solve

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

Then check whether the answer matches the physical setting.

For a single thermal bath and weak coupling to a nondegenerate system, the expected steady state of a thermal master equation is often close to the Gibbs state

$$\rho_\beta = \frac{e^{-\beta H_S}}{\operatorname{Tr}(e^{-\beta H_S})},$$

possibly with Hamiltonian renormalization or weak-coupling corrections. For a driven system, multiple baths, measurement feedback, or nonthermal reservoir, a nonequilibrium steady state is expected instead.

Population rates should obey the relevant detailed-balance condition. For a two-level system with transition frequency $\omega_0>0$,

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

for a single thermal bath in equilibrium.

If the steady state has negative populations, grows without bound in a bounded Hilbert space, or heats at zero temperature without a drive or inverted bath, the model needs review.

Step 12: Limit Checks

Good equations behave correctly in simple limits.

Set the coupling to zero:

$$H_I\to 0.$$

The equation should reduce to closed-system unitary dynamics:

$$\frac{d\rho}{dt} = -\frac{i}{\hbar}[H_S,\rho].$$

Set temperature to zero. Upward thermal rates should vanish for an ordinary passive bath:

$$\Gamma_\uparrow\to 0.$$

Set the bath occupation high. For oscillator damping, rates should approach the expected classical-noise scaling with $\bar n$.

Remove the measurement record. A conditional stochastic equation should average to the corresponding nonselective master equation.

Check short-time behavior. A Markovian exponential decay may not reproduce exact quadratic survival probability at extremely short times; that mismatch can be acceptable only if the Markov model is not being used at microscopic times.

Step 13: Units, Frames, and Conventions

Many apparent disagreements are convention errors.

Check:

• whether $\hbar=1$ is being used,
• whether rates are angular-frequency or cycle-frequency quantities,
• whether spectra include factors of $2\pi$,
• whether $\sigma_z$ has eigenvalues $\pm1$ or $\pm\hbar/2$ is being used as spin,
• whether $\sigma_-=\lvert g\rangle\langle e\rvert$ or the opposite convention is used,
• whether the equation is in the Schrödinger, interaction, rotating, or toggling frame,
• whether a Lamb shift has been included in $H_S$ or reported separately.

When comparing two sources, translate the conventions before comparing numerical rates.

Step 14: Degeneracies and Near Degeneracies

Degenerate subspaces require special care. If several transitions share the same Bohr frequency, secularization should keep the coherent structure inside that frequency block rather than treating every matrix element independently.

Near degeneracies are harder. If

$$|\omega-\omega'|\lesssim \Gamma,$$

full secularization can erase dynamics that are experimentally visible. A partial-secular, coarse-grained, or Redfield-type equation may be more accurate, but it must be checked for positivity in the regime used.

The checklist question is:

Is the resolution time of the experiment long enough to average away the beat frequency?

If not, do not drop the term merely because it is nonsecular in a formal derivation.

Step 15: Model Diagnostics

For an analytic or numerical master equation, run at least the following diagnostics:

• Trace remains $1$ to numerical precision.
• Eigenvalues of $\rho(t)$ stay nonnegative within the expected approximation regime.
• The steady state is positive and normalized.
• Entropy and energy flows have the expected signs for simple thermal limits.
• Turning off a bath removes its dissipator.
• Increasing coupling changes rates in the expected perturbative order.
• Results are stable under a smaller time step or larger Hilbert-space cutoff.
• The equation reproduces a known exactly solvable limit when available.

For channels, check the finite-time map. For generators, check both the instantaneous generator and the integrated dynamics.

Checklist Template

Use this compact template when reviewing a derivation:

• System Hilbert space and bath Hilbert space are explicitly defined.
• Hamiltonian split $H_S+H_B+H_I$ is stated.
• Mean bath forces are removed or included in $H_S$.
• Initial system–bath correlations are specified.
• Bath state is stationary or its time dependence is modeled.
• Coupling strength and perturbative order are identified.
• Bath correlation functions and spectra are defined with conventions.
• System Bohr frequencies and degeneracies are listed.
• Born approximation is justified or explicitly avoided.
• Markov approximation is justified or explicitly avoided.
• Secular or partial-secular approximation is justified.
• Trace preservation is verified.
• Positivity or complete positivity is verified on the intended domain.
• Lamb shift and Hamiltonian renormalization are accounted for.
• Steady state is found and interpreted.
• Zero-coupling, zero-temperature, high-temperature, and no-drive limits are checked where relevant.
• Numerical cutoffs and time steps are tested if the equation is simulated.

Common Failure Modes

Correct form, wrong regime

A Lindblad equation can be mathematically well formed but physically inappropriate if its operators or rates come from invalid assumptions. Complete positivity is necessary for a standard Markovian channel, but it does not by itself certify the model.

Redfield equation used too long

Redfield-type equations can capture nonsecular coherence dynamics better than a fully secular equation, but they may produce negative populations when pushed outside their regime. If negativity appears only beyond the perturbative time window, the model may still be useful; if it appears immediately for reasonable states, the derivation should be revisited.

Thermal bath without thermal steady state

If a single equilibrium bath coupled weakly to an undriven system does not drive the system toward the expected thermal state, check detailed balance, sign conventions, and whether the secular approximation was applied consistently.

Dephasing confused with relaxation

Pure dephasing damps coherences in a preferred basis without changing populations in that basis. Relaxation changes populations and often contributes to coherence decay. See Dephasing vs Dissipation for the canonical distinction and Pure Dephasing Master Equation for the Markovian generator.

Measurement record mixed with ensemble state

A stochastic conditioned state and the ensemble state answer different questions. Averaging over records should recover the nonselective state, but an individual trajectory should not be interpreted as the unconditional density operator.

Audit Exercises

Factorized initial state

A derivation assumes $\rho_{SB}(0)=\rho_S(0)\otimes\rho_B$, but the physical preparation is “let the system equilibrate with the bath, then begin observing.” What should be checked?

Solution

Equilibration generally creates system–bath correlations. The factorized initial-state assumption may be inconsistent with the preparation. One should check whether the subsequent dynamics is being applied only to a restricted family of prepared system states, whether initial-slip corrections are needed, and whether the bath can still be treated as stationary. If the equation is used as a phenomenological model after equilibration, its steady state and short-time behavior should be tested rather than assumed from the factorized derivation.

Secular approximation near a doublet

Two transition frequencies satisfy $|\omega-\omega'|\sim \Gamma$. Should the cross terms always be dropped?

Solution

No. The secular approximation requires the beat frequency $|\omega-\omega'|$ to be large compared with the dissipative rates and with the inverse coarse-graining time. If $|\omega-\omega'|\sim\Gamma$, the cross terms can affect observable coherence and population transfer. A partial-secular, coarse-grained, or nonsecular treatment may be necessary, followed by explicit positivity and steady-state checks.

Negative rate in a time-local equation

A time-local equation is written as $\dot\rho=\gamma(t)\mathcal D[\sigma_-]\rho$, and $\gamma(t)$ becomes negative over a short interval. What conclusion is justified?

Solution

A negative instantaneous rate means the evolution is not CP-divisible during that interval, so it should not be described as a standard Markovian Lindblad semigroup. It does not automatically prove that the finite-time map is unphysical. One must check the integrated map, for example through a Kraus or Choi test, and verify positivity on the intended time interval.

References

• 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, Springer (2004).
• H. J. Carmichael, An Open Systems Approach to Quantum Optics, Springer (1993).
• H. M. Wiseman and G. J. Milburn, Quantum Measurement and Control, Cambridge University Press (2010).
• 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).