Markovian open-system dynamics from first principles.
When can the evolution of an open quantum system be generated by a fixed rule, and what form must that rule take? Lindblad, and independently Gorini, Kossakowski and Sudarshan, answered in 1976. You will prove the structure of their generator, derive it from a microscopic system–bath model, and recover T1, T2 and detailed balance from first principles.
What you’ll learn
State the GKSL theorem and explain why complete positivity, not mere positivity, is required.
Derive the Lindblad form from a short-time Kraus expansion, including the gauge freedom of H and the jump operators.
Carry out the Born–Markov–secular derivation and state the regime of validity of each approximation.
Derive the master equation of a qubit in a thermal bosonic bath, with detailed balance.
Unravel a master equation into quantum-jump trajectories.
Quantum dynamical semigroups
The semigroup property is the mathematical form of Markovianity: the future depends only on the present state, not on the history of system–environment correlations. Complete positivity is non-negotiable, because the system may be entangled with an ancilla that does not evolve, and Λt⊗id must still map states to states. The transpose is the classic map that is positive but not completely positive: applied to half of a Bell pair it produces an operator with eigenvalue −21.
DerivationWhy the generator must look like this
K0=I+(K−iH)dt,Kk=dtLk(k≥1),K=K†,H=H†
Write Λdt in Kraus form. To first order one Kraus operator stays close to the identity and the others are O(dt):
I+2Kdt+dtk≥1∑Lk†Lk+O(dt2)=I⟹K=−21k≥1∑Lk†Lk
Trace preservation, ∑kKk†Kk=I, fixes the Hermitian part K:
Λdt(ρ)=ρ+dt(−i[H,ρ]+{K,ρ}+k≥1∑LkρLk†)+O(dt2)
Expand Λdt(ρ)=∑kKkρKk† to first order:
L(ρ)=−i[H,ρ]+k≥1∑(LkρLk†−21{Lk†Lk,ρ})
Substituting K gives the GKSL form, with the rates absorbed into the Lk. The rigorous proof replaces this scaling ansatz by an analysis of the generator's Choi matrix: L generates a CP semigroup iff it preserves Hermiticity and its Choi matrix is positive on the orthogonal complement of the maximally entangled vector (conditional complete positivity).
From a microscopic model: Born, Markov, secular
Take H=HS+HB+HI with HI=∑αAα⊗Bα (Aα, Bα Hermitian), an initial product state ρS(0)⊗ρB, a stationary bath state ([HB,ρB]=0) and Tr(BαρB)=0. Tildes denote the interaction picture.
DerivationThe Born–Markov–secular derivation
dtdρ~S=−∫0tdsTrB[H~I(t),[H~I(s),ρ~(s)]]
Integrate the interaction-picture von Neumann equation once, substitute it back, and trace out the bath (the first-order term vanishes since ⟨Bα⟩=0):
ρ~(s)≈ρ~S(s)⊗ρB
Born approximation (weak coupling): the bath is barely disturbed and stays uncorrelated with the system.
Markov approximation: bath correlations decay on a time τB much shorter than the system's relaxation time τR. Replace ρ~S(s)→ρ~S(t), substitute s→t−s and extend the integral to infinity. (Stopping here gives the Redfield equation, which need not be completely positive.)
Decompose the coupling into eigenoperators of HS: Aα(ω)=∑ϵ′−ϵ=ωΠ(ϵ)AαΠ(ϵ′), so that A~α(ω,t)=e−iωtAα(ω). With Γαβ(ω)=∫0∞dseiωs⟨B~α†(s)Bβ(0)⟩, terms oscillating as ei(ω′−ω)t appear:
Secular approximation: if τR≫1/∣ω′−ω∣ for all distinct Bohr frequencies, the oscillating terms average out; keep only ω′=ω. Split Γ into its dissipative and dispersive parts:
The result is a GKSL generator. Each matrix γ(ω) is the Fourier transform of a positive-definite correlation function, hence positive semidefinite (Bochner's theorem): exactly the Kossakowski condition. The Lamb-shift Hamiltonian HLS=∑ω∑αβSαβ(ω)Aα†(ω)Aβ(ω) commutes with HS.
nearly degenerate levels (use partial-secular or coarse-grained equations)
The three approximations of the weak-coupling (Davies) limit.
Worked example: a qubit in a thermal bosonic bath
Let HS=−2ω0Z, so that ∣0⟩ is the ground state, and couple A=X=σ++σ− to a bath of oscillators, B=∑k(gkbk+gk∗bk†), at temperature T. Here σ−=∣0⟩⟨1∣ lowers the energy, so the eigenoperators are A(ω0)=σ− and A(−ω0)=σ+.
DerivationAmplitude damping from first principles
⟨B~(s)B(0)⟩=k∑∣gk∣2[(nˉk+1)e−iωks+nˉkeiωks]
Bath correlation function, with nˉ(ω)=1/(eℏω/kBT−1):
Read off populations and coherences (ω0′ is the Lamb-shifted frequency):
ρ00ssρ11ss=γ↓γ↑=nˉ+1nˉ=e−ℏω0/kBT
Hence 1/T1=γ↓+γ↑=2πJ(ω0)coth(ℏω0/2kBT) and T2=2T1: this coupling causes no pure dephasing. The steady state obeys detailed balance, the KMS condition γ(−ω)/γ(ω)=e−ℏω/kBT:
At T=0 the solution is ρ11(t)=e−γ↓tρ11(0) and ∣ρ01(t)∣=e−γ↓t/2∣ρ01(0)∣: exactly the amplitude-damping channel with γ(t)=1−e−t/T1. Pure dephasing needs a coupling with a zero-frequency component, such as A=Z. It gives γφD[Z] with γφ=γ(0), the bath spectrum at zero frequency, and then 1/T2=1/(2T1)+2γφ. The lesson: relaxation probes the noise spectrum at the qubit frequency; dephasing probes it near zero frequency.
Predict before you look
A qubit's bath has nˉ(ω0)=0.1 at the qubit frequency. What is the steady-state excited population ρ11ss, in %?
Show the answer
8.3 %
ρ11ss=γ↑/(γ↑+γ↓)=nˉ/(2nˉ+1)=0.1/1.2≈8.3%. For comparison, a 5 GHz qubit in equilibrium with a 50 mK bath has nˉ≈0.008, i.e. an excited population below 1%.
Quantum trajectories and the limits of Markovianity
Rewrite the generator as L(ρ)=−i(Heffρ−ρHeff†)+∑kLkρLk† with the non-Hermitian Heff=H−2i∑kLk†Lk. This suggests a stochastic unraveling (Dalibard, Castin & Mølmer 1992; Carmichael 1993):
In each step dt, jump k occurs with probability dpk=dt⟨ψLk†Lkψ⟩, and then ∣ψ⟩↦Lk∣ψ⟩/∥Lk∣ψ⟩∥.
Otherwise the state evolves under e−iHeffdt and is renormalized.
Averaging ∣ψ⟩⟨ψ∣ over trajectories reproduces ρ(t) exactly.
For amplitude damping this is literally the photodetection record: each click is a jump σ−. Trajectories are how large open systems are simulated, and they describe continuously monitored qubits.
Check your understanding
Which generator produces a CPTP semigroup (rates γ,κ>0)?
L(ρ)=−γD[σ−]ρ
L(ρ)=−i(Hρ−ρH†) with H=H0−iγ∣1⟩⟨1∣
L(ρ)=γ(σ−ρσ+−σ+σ−ρ)
L(ρ)=γD[σ−]ρ+κ(ZρZ−ρ)
Show the answer
D.L(ρ)=γD[σ−]ρ+κ(ZρZ−ρ)
Both terms are GKSL dissipators with positive rates: κ(ZρZ−ρ)=κD[Z]ρ because Z†Z=I. Together they describe relaxation plus pure dephasing, with 1/T1=γ and 1/T2=γ/2+2κ.
Key takeaways
Markovian (semigroup) dynamics of a finite-dimensional system is generated by L(ρ)=−i[H,ρ]+∑kγk(LkρLk†−21{Lk†Lk,ρ}) with γk≥0 (GKSL).
Complete positivity forces the Kossakowski matrix to be positive semidefinite; H and the Lk are defined only up to unitary mixing and shifts.
The Born (weak coupling), Markov (τB≪τR) and secular (τR≫1/Δω) approximations yield GKSL form with rates set by bath spectra: T1 probes the spectrum at ω0, pure dephasing near ω=0.
Detailed balance γ↑/γ↓=e−ℏω0/kBT fixes the thermal steady state, and quantum-jump unravelings reproduce ρ(t) as an average over pure trajectories.
Check yourself
In the interactive lesson these are scored with your confidence, and what you miss comes back for review.
Question 1
Why must the maps etL be completely positive rather than merely positive?
Because positivity alone would not preserve the trace
Because the system may be entangled with an untouched ancilla, and etL⊗id must still map states to states
Because complete positivity guarantees that the evolution is unitary
Because positive maps cannot be written as exponentials of generators
Show the answer
B. Because the system may be entangled with an untouched ancilla, and etL⊗id must still map states to states
The physical system can be half of an entangled state. A positive but not completely positive map, like the transpose, can then produce a negative eigenvalue on the joint state. Complete positivity is precisely the requirement that Λ⊗id be positive for every ancilla.
Question 2
In the Born–Markov–secular derivation, which condition justifies the secular approximation?
The bath correlation time is much shorter than the relaxation time
The system–bath coupling is weak
The relaxation time is long compared with the inverse differences of the system's Bohr frequencies
The bath is at zero temperature
Show the answer
C. The relaxation time is long compared with the inverse differences of the system's Bohr frequencies
Terms oscillating as ei(ω′−ω)t average out only if the dissipative dynamics is slow compared with 1/∣ω′−ω∣. Nearly degenerate transitions violate this.
Question 3
For a qubit coupled through X to a thermal bosonic bath, what is the ratio γ↑/γ↓?
e−ℏω0/kBT
1
nˉ(ω0)
e+ℏω0/kBT
Show the answer
A.e−ℏω0/kBT
γ↑/γ↓=nˉ/(nˉ+1)=e−ℏω0/kBT, the KMS (detailed-balance) condition. It drives the qubit to the Gibbs state.
Question 4
In the quantum-jump unraveling of ρ˙=γD[σ−]ρ, a qubit starts in ∣+⟩. If no jump is recorded for a long time, the conditional state
stays ∣+⟩, since nothing happened
becomes the maximally mixed state
rotates toward ∣1⟩
is continuously pulled toward ∣0⟩
Show the answer
D. is continuously pulled toward ∣0⟩
Under Heff=−2iγ∣1⟩⟨1∣ (in the rotating frame) the ∣1⟩ amplitude decays as e−γt/2 while the ∣0⟩ amplitude does not; after renormalization the state drifts toward ∣0⟩. Not seeing a photon is information too.
Further reading
G. Lindblad, "On the generators of quantum dynamical semigroups", Communications in Mathematical Physics 48, 119 (1976).
V. Gorini, A. Kossakowski & E. C. G. Sudarshan, "Completely positive dynamical semigroups of N-level systems", Journal of Mathematical Physics 17, 821 (1976).
E. B. Davies, "Markovian master equations", Communications in Mathematical Physics 39, 91 (1974).
H.-P. Breuer & F. Petruccione, The Theory of Open Quantum Systems, Ch. 3 (Oxford, 2002).
D. A. Lidar, "Lecture notes on the theory of open quantum systems", arXiv:1902.00967 (2019).