Skip to content

Quantum Dynamical Semigroups on the Trace Class

A conservative quantum dynamical semigroup is a continuous evolution on normal states whose maps compose according to elapsed time and preserve complete positivity and trace. The canonical Quantum Dynamical Semigroups article develops its physical interpretation. This extension fixes the Banach space and continuity convention and explains how a master equation remains meaningful when its generator is unbounded.

Required background. Quantum Channels in Infinite Dimensions defines the state maps. Helpful background. Strong Continuity distinguishes pointwise state continuity from operator-norm continuity.

Conservative semigroups on the trace class

Section titled “Conservative semigroups on the trace class”

On a separable Hilbert space H\mathcal H, let Φt:S1(H)→S1(H)\Phi_t:\mathcal S_1(\mathcal H)\to\mathcal S_1(\mathcal H), t≥0t\ge0, satisfy:

  • Φ0=id\Phi_0=\mathrm{id} and Φt+s=ΦtΦs\Phi_{t+s}=\Phi_t\Phi_s;
  • each Φt\Phi_t is CP and trace preserving;
  • ∥Φt(X)−X∥1→0\|\Phi_t(X)-X\|_1\to0 as t↓0t\downarrow0 for every trace-class XX.

These are the conventions used here. Some literature allows trace-decreasing maps under the name “quantum dynamical semigroup” and adds conservative or Markov for the trace-preserving case.

Each map is a contraction in trace norm. The adjoints Tt=Φt†T_t=\Phi_t^\dagger form a semigroup of normal unital CP maps on B(H)B(\mathcal H), continuous pointwise in the ultraweak topology. The duality is

Tr⁡[Φt(X)A]=Tr⁡[XTt(A)].\operatorname{Tr}[\Phi_t(X)A]=\operatorname{Tr}[XT_t(A)].

Normality is continuity with respect to monotone operator limits at a fixed time. Time continuity is a separate condition on the family. Neither should be silently substituted for the other.

A one-parameter semigroup is time homogeneous. A two-time propagator Φ(t,s)\Phi(t,s) can instead obey Φ(t,r)Φ(r,s)=Φ(t,s)\Phi(t,r)\Phi(r,s)=\Phi(t,s) while depending on both endpoints. The mathematical language of Markovianity separates these structures.

Differentiable trajectories start in the generator domain

Section titled “Differentiable trajectories start in the generator domain”

Define

D(L)={X:lim⁡h↓0Φh(X)−Xh exists in S1},LX=lim⁡h↓0Φh(X)−Xh.D(\mathcal L)=\left\{X: \lim_{h\downarrow0}\frac{\Phi_h(X)-X}{h} \text{ exists in }\mathcal S_1\right\}, \qquad \mathcal L X=\lim_{h\downarrow0}\frac{\Phi_h(X)-X}{h}.

This is a dense linear domain, but it need not contain every density operator. If X∈D(L)X\in D(\mathcal L), the semigroup law gives

Φh(ΦtX)−ΦtXh=ΦtΦhX−Xh.\frac{\Phi_h(\Phi_tX)-\Phi_tX}{h} =\Phi_t\frac{\Phi_hX-X}{h}.

Boundedness of Φt\Phi_t proves that ΦtX∈D(L)\Phi_tX\in D(\mathcal L) and LΦtX=ΦtLX\mathcal L\Phi_tX=\Phi_t\mathcal LX. The trajectory is differentiable and solves

ddtΦtX=LΦtX,Φ0X=X.\frac{d}{dt}\Phi_tX=\mathcal L\Phi_tX,\qquad \Phi_0X=X.

For X∉D(L)X\notin D(\mathcal L) the trajectory still exists and is trace-norm continuous. It is a mild solution, represented by the integrated relation

∫0tΦsX ds∈D(L),ΦtX−X=L∫0tΦsX ds.\int_0^t\Phi_sX\,ds\in D(\mathcal L),\qquad \Phi_tX-X=\mathcal L\int_0^t\Phi_sX\,ds.

The integral is a trace-class Bochner integral. The domain and resolvent proof is given in Generators of CP Semigroups. Writing ρ˙=Lρ\dot\rho=\mathcal L\rho for every initial state without this qualification is stronger than the semigroup theorem warrants.

Worked example: resetting to a normal state

Section titled “Worked example: resetting to a normal state”

Fix any density operator τ\tau on H\mathcal H and γ>0\gamma>0. The reset channel Rτ(X)=Tr⁡(X)τR_\tau(X)=\operatorname{Tr}(X)\tau satisfies Rτ2=RτR_\tau^2=R_\tau. Hence

Φt=e−γtid+(1−e−γt)Rτ\Phi_t=e^{-\gamma t}\mathrm{id} +(1-e^{-\gamma t})R_\tau

is a channel for each tt. Decompose X=RτX+(I−Rτ)XX=R_\tau X+(I-R_\tau)X. The first component is fixed and the second is multiplied by e−γte^{-\gamma t}, proving the semigroup law directly.

The generator is the bounded operator

L=γ(Rτ−id),∥L∥1→1≤2γ.\mathcal L=\gamma(R_\tau-\mathrm{id}),\qquad \|\mathcal L\|_{1\to1}\le2\gamma.

For every state,

∥Φt(ρ)−τ∥1=e−γt∥ρ−τ∥1.\|\Phi_t(\rho)-\tau\|_1 =e^{-\gamma t}\|\rho-\tau\|_1.

Thus τ\tau is the unique stationary density operator and attraction is uniform over states. This construction works in infinite dimension because τ\tau is a specified trace-class state. Replacing it by a purported normalized identity would be invalid when Tr⁡I=∞\operatorname{Tr}I=\infty.

An explicit countable Kraus representation for RτR_\tau follows if τ=∑mqm∣m⟩⟨m∣\tau=\sum_mq_m|m\rangle\langle m|: use Kmn=qm∣m⟩⟨n∣K_{mn}=\sqrt{q_m}|m\rangle\langle n|. The strong sum of Kmn†KmnK_{mn}^\dagger K_{mn} is II. This exhibits a perfectly bounded generator even though its natural list of jump operators is infinite.

Semigroup structure does not establish microscopic validity

Section titled “Semigroup structure does not establish microscopic validity”

The abstract axioms prove positivity, normalization, composition, and continuity within a reduced model. They do not prove that a given system–environment Hamiltonian produces that model over a chosen time window.

In particular, tracing out a fixed environment after joint unitary evolution generally does not give the semigroup law. Reusing the environment retains correlations, whereas composing a reduced channel often corresponds to preparing a fresh environment for the next interval. The mathematical open-system overview works through this distinction explicitly.

Also, a semigroup of one-time reduced maps does not by itself specify every probability for experiments with interventions at intermediate times. That stronger statement requires a multitime process model.

Fixed point and generator kernel. Show that a state is stationary for all tt if and only if it belongs to D(L)D(\mathcal L) and Lρ=0\mathcal L\rho=0.

Solution

A stationary orbit has zero difference quotient, so it is in the domain with zero generator. Conversely, domain invariance gives dΦt(ρ)/dt=Φt(Lρ)=0d\Phi_t(\rho)/dt=\Phi_t(\mathcal L\rho)=0. The initial condition then fixes the orbit to ρ\rho.

A reset map at a random time. Give a probabilistic construction of the reset semigroup and explain why the reset state must be the same at every event.

Solution

Let reset events occur at Poisson rate γ\gamma. With probability e−γte^{-\gamma t} no event occurs, so the input survives. With complementary probability at least one occurs, and every later reset leaves the same state τ\tau. Averaging gives Φt\Phi_t. If the reset target changes with absolute time, the resulting propagator generally depends on both time endpoints.

  • Engel, Klaus-Jochen, and Rainer Nagel. One-Parameter Semigroups for Linear Evolution Equations. Graduate Texts in Mathematics 194, Springer, 2000. doi:10.1007/b97696. Generator domains, classical solutions, and integrated trajectories.
  • Holevo, Alexander S. Quantum Systems, Channels, Information: A Mathematical Introduction, 2nd ed. De Gruyter, 2019. doi:10.1515/9783110642490. Normal channels and quantum dynamical semigroups.
  • Lindblad, Göran. “On the generators of quantum dynamical semigroups.” Communications in Mathematical Physics 48, 119–130 (1976). doi:10.1007/BF01608499. Norm-continuous quantum semigroups.