Properties and Time Evolution of the Density Matrix

The density operator was built to carry a state; this article makes it move. For a closed system every ensemble member obeys the Schrödinger equation, so the whole density operator obeys $\rho(t) = U(t)\,\rho(0)\,U^\dagger(t)$ with $U = e^{-i\hat Ht/\hbar}$ . Differentiating gives the von Neumann equation $i\hbar\,\partial_t\rho = [\hat H, \rho]$ — the equation of motion for a state written as an operator, and the workhorse of open-system physics, magnetic resonance, and quantum optics. Two structural facts, owned here, do most of the conceptual work. First, the sign: the von Neumann equation has the opposite sign to the Heisenberg equation of motion $\dot{\hat A}_H = \tfrac{i}{\hbar}[\hat H, \hat A_H]$ for an observable. This is not a typo waiting to bite you but a consequence of what $\rho$ is — a state, transforming by conjugation $U(\cdot)U^\dagger$ , where an observable transforms the other way. We derive both side by side and check that $\langle\hat A\rangle(t)$ comes out identical whether you evolve the state or the observable, so the two signs are two bookkeeping conventions for one physics. Second, the invariants: under unitary flow the eigenvalues of $\rho$ never move, so $\mathrm{Tr}\,\rho$ , the purity $\mathrm{Tr}\,\rho^2$ , and every $\mathrm{Tr}\,\rho^n$ — hence the entropy — are conserved. A closed quantum system therefore cannot purify or mix: a pure state stays pure, a mixed state stays exactly as mixed. The punchline is the launch point for the rest of the track: mixing requires an environment. The full open-system theory — the Lindblad master equation, dephasing, decoherence — is developed in Quantum Decoherence and the Density Matrix by name; this article owns the closed-system law and the reason mixedness demands a partner.

Key takeaways

Closed evolution: $\rho(t) = U(t)\rho(0)U^\dagger(t)$ , $U = e^{-i\hat Ht/\hbar}$ ; differential form the von Neumann equation $i\hbar\,\partial_t\rho = [\hat H, \rho]$ (derived).
Sign contrast (owned): $i\hbar\,\partial_t\rho = [\hat H,\rho]$ for the state versus $\dot{\hat A}_H = \tfrac{i}{\hbar}[\hat H,\hat A_H]$ for an observable — opposite signs because $\rho$ transforms by $U(\cdot)U^\dagger$ ; $\langle\hat A\rangle(t)$ is identical either way (checked).
Invariants of unitary flow (owned): eigenvalues frozen $\Rightarrow$ $\mathrm{Tr}\,\rho^n$ constant for all $n$ ; purity and entropy conserved. A closed system never purifies or mixes.
Punchline: mixing requires an environment. The open-system Lindblad form is stated; its derivation and decoherence are owned by the decoherence article by name.
Coherence lab: for $\hat H = \tfrac12\hbar\omega\,\sigma_z$ , $\rho_{01}(t) = \rho_{01}(0)\,e^{-i\omega t}$ rotates while populations freeze; an open dephasing term would decay it with a $T_2$ time (stated).
Worked: $|+x\rangle$ precesses, $\langle\sigma_x\rangle(t) = \cos\omega t$ , purity $= 1$ always; a mixed Bloch vector precesses with $|\mathbf r|$ constant.
Key idea: a closed system evolves $\rho$ by unitary conjugation, which freezes its spectrum — so purity and entropy are conserved, and any change in mixedness is proof of an environment.

The core insight

Closed-system evolution rotates $\rho$ by $U(\cdot)U^\dagger$ , whose differential form is the von Neumann equation $i\hbar\dot\rho = [\hat H,\rho]$ — sign opposite to the Heisenberg observable equation because $\rho$ is a state. Unitary conjugation cannot move eigenvalues, so purity, entropy, and every $\mathrm{Tr}\,\rho^n$ are conserved: mixing a state requires an environment, not a Hamiltonian.

Where this lives today

The von Neumann equation is the base layer of every quantum-dynamics simulation and every spin-resonance experiment. Magnetic resonance (NMR, MRI, EPR) is Bloch precession read on hardware: the magnetization vector is $\mathbf r$ , precessing at the Larmor frequency (equation 4) and — crucially — relaxing, with $T_1$ and $T_2$ times measuring exactly the non-unitary decay this article flags as environmental. Quantum control and gate design solve the von Neumann equation (or its open-system Lindblad extension) to find pulse sequences that steer $\rho$ ; the closed-system invariants tell a designer that a unitary gate can rotate but never purify, so purification requires measurement or dissipation-engineering. In open quantum systems and quantum thermodynamics, the split this article draws — reversible commutator term versus dissipative Lindblad term — is the organizing dichotomy: work is the unitary part, heat and entropy production the dissipative part, and the second law appears precisely because the environment moves eigenvalues the Hamiltonian cannot.

Figure

The Bloch picture: closed evolution is a rotation, open evolution spirals inward

The physical picture

A closed quantum system is a rigid dance. The Hamiltonian rotates the state, and because that rotation is unitary — length-preserving, invertible — it can rearrange amplitudes and shuffle coherences, but it cannot create or destroy the fundamental "spread" of the state. Picture $\rho$ by its eigenvalues: a pure state is all weight on one eigenvector, a mixed state has weight spread across several. Unitary evolution spins the eigenvectors around Hilbert space but leaves the eigenvalues exactly where they were. So the shape of the state — how pure, how mixed, how much entropy — is frozen, and only its orientation moves. This is why decoherence is such a different animal. Watching a coherent superposition "turn into" a classical mixture — a live qubit going dead — is watching eigenvalues move toward each other, and no Hamiltonian on the system alone can do that. The move requires coupling to something you do not track: the system entangles with an environment, you trace the environment out, and that non-unitary step redistributes the eigenvalues. The von Neumann equation is the law of the reversible dance; its failure is the signature of an open system, and the failure is where the rest of this chapter lives.

Figure

Unitary flow rotates eigenvectors but freezes eigenvalues — coherence spins, purity holds

The equation of motion

Every member of an ensemble is a ket, and every ket obeys the Schrödinger equation, so the density operator's motion follows by averaging. The unitary evolution operator $U(t) = e^{-i\hat Ht/\hbar}$ and the time-dependent Schrödinger equation are owned by the dynamics chapter and recalled here by name.

Derivation AThe von Neumann equation and its sign — state versus observable

We derive $\rho(t) = U\rho(0)U^\dagger$ and its differential form, then place it beside the Heisenberg equation for an observable to expose the sign difference and prove the two pictures agree. The point: $\rho$ is a state, evolving by conjugation, while an observable in the Heisenberg picture evolves by the reverse conjugation — the opposite sign is forced, not chosen.

Step 1. Evolve the ensemble. Each ket runs as $|\psi_i(t)\rangle = U(t)|\psi_i(0)\rangle$ , so

\rho(t) = \sum_i p_i\,|\psi_i(t)\rangle\langle\psi_i(t)| = U(t)\Bigl[\sum_i p_i|\psi_i(0)\rangle\langle\psi_i(0)|\Bigr]U^\dagger(t) = U(t)\,\rho(0)\,U^\dagger(t) \qquad (1)

Why the weights ride along untouched: the classical probabilities $p_i$ do not evolve — only the quantum states do — so the whole operator transforms by the single conjugation $U(\cdot)U^\dagger$ . Note the $U$ on the left and $U^\dagger$ on the right: a state sandwiched between the evolution and its inverse.

Step 2. Differentiate. Using $\dot U = -\tfrac{i}{\hbar}\hat HU$ and $\dot U^\dagger = +\tfrac{i}{\hbar}U^\dagger\hat H$ ,

\dot\rho = \dot U\rho_0U^\dagger + U\rho_0\dot U^\dagger = -\tfrac{i}{\hbar}\hat H\rho + \tfrac{i}{\hbar}\rho\hat H = -\tfrac{i}{\hbar}[\hat H,\rho], \quad\text{i.e.}\quad i\hbar\,\partial_t\rho = [\hat H,\rho] \qquad (2)

the von Neumann equation, the density-operator equation of motion. Why the commutator: $\rho$ transforms by conjugation, and the generator of conjugation is the commutator with $\hat H$ — the same structure as the classical Liouville equation $\partial_t\rho = \{H,\rho\}_{\rm PB}$ with the Poisson bracket replaced by $[\cdot,\cdot]/i\hbar$ .

Step 3. Contrast with the observable. In the Heisenberg picture the state is fixed and the observable carries the time dependence, $\hat A_H(t) = U^\dagger\hat A\,U$ (note the order reversed from a state), giving

\dot{\hat A}_H = +\tfrac{i}{\hbar}[\hat H,\hat A_H] \qquad (3)

— the opposite sign to (2). Why they must differ: a state is $U\rho U^\dagger$ and an observable is $U^\dagger\hat AU$ ; the conjugations are inverses of each other, so their generators are negatives. There is no physical asymmetry — the sign is entirely which object you decided to move.

Step 4. Prove consistency. The observable's mean must not depend on which picture you chose. In the Schrödinger picture $\langle\hat A\rangle(t) = \mathrm{Tr}(\rho(t)\hat A) = \mathrm{Tr}(U\rho_0U^\dagger\hat A)$ ; cyclicity of the trace rewrites this as $\mathrm{Tr}(\rho_0\,U^\dagger\hat AU) = \mathrm{Tr}(\rho_0\hat A_H(t))$ , the Heisenberg answer. The two signs give one number. Verified numerically: for $\hat H = \tfrac12\hbar\omega\sigma_z$ and $\rho_0 = |+x\rangle\langle+x|$ , equation (2) holds (finite-difference $i\hbar\dot\rho = [\hat H,\rho]$ to $10^{-4}$ ), the Heisenberg equation (3) holds with the opposite sign, and $\mathrm{Tr}(\rho(t)\sigma_x) = \mathrm{Tr}(\rho_0\,\sigma_{x,H}(t))$ exactly.

The invariants: why a closed system can never mix

Go deeper · Derivation B · 2 minUnitary flow freezes the spectrum — purity and entropy are conserved

Derivation BUnitary flow freezes the spectrum — purity and entropy are conserved

We prove that under $\rho(t) = U\rho_0U^\dagger$ every trace power $\mathrm{Tr}\,\rho^n$ is constant, hence purity and the von Neumann entropy are conserved, and conclude that a closed system cannot change how mixed it is. The engine is the cyclicity of the trace together with the eigenvalue-preserving nature of conjugation.

Step 1. Conjugation preserves eigenvalues. $\rho(t) = U\rho_0U^\dagger$ is a similarity transformation by a unitary, so $\rho(t)$ and $\rho_0$ have the same spectrum $\{\lambda_k\}$ — the eigenvectors rotate to $U|k\rangle$ , the eigenvalues stay put. Anything depending only on $\{\lambda_k\}$ is therefore a constant of the motion.

Step 2. Trace powers are invariant. Directly, $\mathrm{Tr}\,\rho(t)^n = \mathrm{Tr}\bigl((U\rho_0U^\dagger)^n\bigr) = \mathrm{Tr}\bigl(U\rho_0^nU^\dagger\bigr) = \mathrm{Tr}\,\rho_0^n$ for every $n$ , where the $U^\dagger U$ pairs between factors cancel and cyclicity removes the outer $U$ . Why this matters: $n = 1$ is conservation of probability; $n = 2$ is conservation of purity; and $-\mathrm{Tr}(\rho\ln\rho)$ , a power series in $\rho$ , is conserved too — the entropy is a constant of the unitary motion.

Step 3. Draw the conclusion. Since purity and entropy are frozen, a closed system that starts pure stays pure ( $P = 1$ for all $t$ ) and one that starts mixed keeps exactly its initial mixedness. On the Bloch ball this is visible: unitary evolution is a rotation about some axis, which moves the Bloch vector on a sphere of fixed radius $|\mathbf r|$ — never inward toward the mixed interior. Why "mixing requires an environment": to lower purity you must move eigenvalues toward equality, and no Hamiltonian on the system alone does that; the only route is to entangle with an untracked partner and trace it out — a non-unitary CPTP step.

Step 4. State the open-system generalization. When the system is open, unitarity fails and the evolution acquires dissipative terms; the general Markovian form is the Lindblad master equation $\dot\rho = -\tfrac{i}{\hbar}[\hat H,\rho] + \sum_k\bigl(L_k\rho L_k^\dagger - \tfrac12\{L_k^\dagger L_k,\rho\}\bigr)$ , which does move eigenvalues and lower purity. Its derivation, the dephasing that gives a $T_2$ time, and decoherence are owned by Quantum Decoherence and the Density Matrix by name; here we only note that the von Neumann equation is the $L_k = 0$ special case. Verified numerically: for the precessing $|+x\rangle$ , $\mathrm{Tr}\,\rho^n$ is constant for $n = 1,2,3,4$ across a full period, and the purity stays exactly $1$ .

The coherence lab

Put $\hat H = \tfrac12\hbar\omega\,\sigma_z$ — a spin in a field along $z$ — and watch a single off-diagonal element. In the energy eigenbasis the populations are stationary (energy eigenstates have time-independent occupation), while the coherence carries the phase difference of the two energies:

\rho_{01}(t) = \rho_{01}(0)\,e^{-i(E_0 - E_1)t/\hbar} = \rho_{01}(0)\,e^{-i\omega t} \qquad (4)

where: $\rho_{01}$ is the amplitude linking the up and down branches, $E_0 - E_1 = \hbar\omega$ is their energy gap, and the coherence rotates in the complex plane at the Larmor frequency $\omega$ while its magnitude $|\rho_{01}|$ stays fixed. This is the reversible dance: the two-level system's Bloch vector precesses about $z$ , tracing a circle of constant radius. Contrast the open-system fate: a dephasing (Lindblad) term with rate $\Gamma$ multiplies (4) by $e^{-\Gamma t}$ , so $|\rho_{01}(t)| = |\rho_{01}(0)|\,e^{-t/T_2}$ with $T_2 = 1/\Gamma$ — the coherence decays, purity falls, and the qubit dies. The magnitude decay is the environment's fingerprint; the phase rotation is the Hamiltonian's. The decay term and its microscopic origin are owned by the decoherence article by name; here we own the reversible rotation and the statement that only the open case touches the magnitude.

Limiting cases

case	evolution	reading
$[\hat H, \rho_0] = 0$	$\rho(t) = \rho_0$ , stationary	$\rho$ diagonal in the energy basis is a fixed point
$\rho_0$ pure	stays pure, $P = 1$ for all $t$	unitary preserves purity
$\rho_0 = \mathbb 1/d$	stays $\mathbb 1/d$	the maximally mixed state is invariant under any unitary
$\hat H = \tfrac12\hbar\omega\sigma_z$	Bloch precession about $z$ , radius fixed	coherence rotates, populations freeze
add Lindblad dephasing $\Gamma$	$\|\rho_{01}\|\to e^{-t/T_2}\|\rho_{01}(0)\|$	open system; purity falls (owned downstream)
classical limit	$\partial_t\rho = \{H,\rho\}_{\rm PB}$ (Liouville)	$[\cdot,\cdot]/i\hbar\to\{\cdot,\cdot\}$

Common trapThe von Neumann sign is opposite to the Heisenberg sign — and both are right

The single most common slip in density-matrix dynamics is writing $\dot\rho = +\tfrac{i}{\hbar}[\hat H,\rho]$ by pattern-matching the Heisenberg equation $\dot{\hat A}_H = +\tfrac{i}{\hbar}[\hat H,\hat A_H]$ . The state equation is $i\hbar\dot\rho = [\hat H,\rho]$ , i.e. $\dot\rho = -\tfrac{i}{\hbar}[\hat H,\rho]$ — the opposite sign. The reason is structural: a state transforms by $U(\cdot)U^\dagger$ and an observable (Heisenberg) by $U^\dagger(\cdot)U$ , inverse conjugations with negated generators. The reliable check is consistency of $\langle\hat A\rangle(t)$ : compute it both ways and they must agree ( $\mathrm{Tr}(\rho(t)\hat A) = \mathrm{Tr}(\rho_0\hat A_H(t))$ by cyclicity). If your two pictures disagree, you flipped a sign — and it is almost always the $\rho$ one. Physically: with the correct sign, a coherence goes as $e^{-i(E_0-E_1)t/\hbar}$ ; the wrong sign predicts the coherence rotating the wrong way, which a Ramsey-fringe experiment would catch immediately.

Worked examples

Example 1 — a precessing pure spin (confidence builder). Take $\hat H = \tfrac12\hbar\omega\,\sigma_z$ and the pure initial state $|+x\rangle$ , $\rho_0 = \tfrac12(\mathbb 1 + \sigma_x)$ . Find $\langle\sigma_x\rangle(t)$ and the purity.

Step 1. Evolve. $U(t) = e^{-i\omega t\sigma_z/2}$ rotates the Bloch vector about $z$ by angle $\omega t$ . The initial Bloch vector $\mathbf r(0) = \hat x$ precesses to $\mathbf r(t) = (\cos\omega t,\,-\sin\omega t,\,0)$ — a circle in the equatorial plane at radius $1$ .

Step 2. Read the observable. $\langle\sigma_x\rangle(t) = r_x(t) = \boxed{\cos\omega t}$ (machine-verified across a period): the spin's $x$ -component oscillates sinusoidally, the classic Larmor precession seen from the $x$ -axis.

Step 3. Read the purity. $|\mathbf r(t)| = 1$ for all $t$ , so $P = \tfrac12(1 + 1) = \boxed{1}$ always — the state stays pure, precessing on the surface of the Bloch ball and never dipping inside. A closed evolution cannot decohere it; that would require the environment this article deliberately excludes.

Example 2 — a partially mixed spin precesses without mixing further (exam level). Now start mixed: $\mathbf r(0) = (0.6, 0, 0)$ , a state of purity $P = \tfrac12(1 + 0.36) = 0.68$ . Evolve under the same $\hat H$ and confirm the mixedness is conserved.

Step 1. Precess. The Bloch vector rotates about $z$ exactly as before: $\mathbf r(t) = (0.6\cos\omega t,\,-0.6\sin\omega t,\,0)$ , a circle of radius $0.6$ . At $t = 0.5/\omega$ : $r_x = 0.6\cos(0.5) = 0.5265$ , $r_y = -0.6\sin(0.5) = -0.2877$ (machine-verified).

Step 2. Check the invariant. $|\mathbf r(t)| = \sqrt{r_x^2 + r_y^2} = 0.6$ for all $t$ (machine-verified around the circle), so the purity holds at $P = 0.68$ throughout: the state precesses inside the ball on a fixed-radius circle, never approaching either the pure surface or the mixed center.

Step 3. Bank the lesson. Unitary evolution moves the orientation of the state (the coherence phase, the Bloch direction) but never its shape (purity, entropy). To make this mixed state purer or more mixed you would have to break the closed-system assumption — measure it, or couple it to a bath. That is the door the decoherence article opens; this article closes it firmly.

Liouville's flow, made quantum

Origins · 1 mindensity operator and how this was discovered

density operator

1838–1932 · Paris, Göttingen

The equation of motion for a state has a classical ancestor and a quantum author. In Paris in 1838, Joseph Liouville proved that a classical probability density flows through phase space without compression, $\partial_t\rho = \{H, \rho\}_{\rm PB}$ — the density is carried rigidly along the Hamiltonian flow, its "shape" (and hence its entropy) conserved. A century later, as quantum mechanics acquired its statistical operator, the same structure reappeared with the Poisson bracket replaced by the commutator over $i\hbar$ : John von Neumann, in the 1932 Mathematische Grundlagen, wrote $i\hbar\,\partial_t\rho = [\hat H, \rho]$ , the exact quantum image of Liouville's law, and proved its conservation properties — that the entropy $-\mathrm{Tr}(\rho\ln\rho)$ is a constant of the closed-system motion, the quantum echo of Liouville's incompressibility. The correspondence is exact and deep: $\{\cdot,\cdot\}\to[\cdot,\cdot]/i\hbar$ is one of the pillars of quantization, and the fact that both the classical and quantum flows freeze the entropy is why the origin of the second law had to be sought outside the closed dynamics — in coarse-graining classically, in tracing out an environment quantum-mechanically. The dissipative extension came only in 1976, when Lindblad (and, independently, Gorini–Kossakowski–Sudarshan) found the most general form a non-unitary Markovian evolution can take.

History

1838

— Liouville: a classical phase-space density flows incompressibly, ∂ₜρ = {H, ρ}

1927

— von Neumann's statistical operator; the commutator equation of motion

1932

— Mathematische Grundlagen: entropy conserved under closed evolution

1946

— Bloch equations add phenomenological T₁, T₂ relaxation to the precession

1976

— Lindblad and Gorini–Kossakowski–Sudarshan: the general open-system master equation

Going further

References. Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §3.4 · Cohen-Tannoudji, Diu & Laloë, Complement $E_{III}$ · Nielsen & Chuang, §2.4, §8 · Breuer & Petruccione, The Theory of Open Quantum Systems, Ch. 3 (Lindblad) · Blum, Density Matrix Theory and Applications · Liouville, J. Math. Pures Appl. 3, 342 (1838) (classical phase-flow ancestor); von Neumann (1932).

Provenance. All results are standard. The unitary evolution operator $U = e^{-i\hat Ht/\hbar}$ , the time-dependent Schrödinger equation, and the density-operator credentials are owned by Time-Dependent Schrödinger Equation and The Density Operator and Mixed States, recalled by name. Owned and derived here: (1) the von Neumann equation $i\hbar\dot\rho = [\hat H,\rho]$ from $\rho(t) = U\rho_0U^\dagger$ , machine-verified by finite difference for $\hat H = \tfrac12\hbar\omega\sigma_z$ ; (2) the sign contrast with the Heisenberg equation $\dot{\hat A}_H = +\tfrac{i}{\hbar}[\hat H,\hat A_H]$ , with the consistency $\mathrm{Tr}(\rho(t)\hat A) = \mathrm{Tr}(\rho_0\hat A_H(t))$ checked; (3) the invariants of unitary flow — $\mathrm{Tr}\,\rho^n$ constant for $n = 1,2,3,4$ (checked over a period), hence purity and entropy conserved; and (4) the coherence law $\rho_{01}(t) = \rho_{01}(0)e^{-i\omega t}$ with frozen populations (checked). Worked numbers recomputed: $\langle\sigma_x\rangle(t) = \cos\omega t$ with purity $1$ (Example 1); a mixed $|\mathbf r| = 0.6$ Bloch vector precesses with $|\mathbf r|$ constant, $r_x(0.5/\omega) = 0.5265$ (Example 2). The Lindblad master equation is stated as the open-system generalization ( $L_k = 0$ recovering the von Neumann equation); its derivation, dephasing, the $T_2$ time, and decoherence are owned by Quantum Decoherence and the Density Matrix by name; entropy production and its meaning by Entanglement Entropy and Measures of Entanglement; the non-unitary measurement jump by Measurement and Wavefunction Collapse and The Quantum Measurement Problem — all by name.

Builds on: The Density Operator and Mixed States · Pure versus Mixed States and the Purity Criterion · Time-Dependent Schrödinger Equation · The Commutator and Compatible Observables

Leads to: The Reduced Density Matrix and Partial Trace · Entanglement Entropy and Measures of Entanglement · Quantum Entanglement and Non-Separability

If the entropy of a closed system is conserved, how does the universe ever increase in entropy?

The conservation is of the fine-grained (von Neumann) entropy of the whole closed system — and the whole universe, if closed, indeed keeps its total fine-grained entropy fixed. Entropy increase is a statement about subsystems: trace out the parts you do not track (the environment, the unmeasured degrees of freedom) and the reduced state's entropy grows, because the traced-out correlations hide information. So the second law is not a violation of the von Neumann equation but its shadow after a partial trace — exactly the non-unitary step this article identifies as the only way to move eigenvalues. The mechanism is owned by The Reduced Density Matrix and Partial Trace and the decoherence article by name.

Is the von Neumann equation just the Schrödinger equation rewritten?

For a closed system, yes — it is the Schrödinger equation lifted to operators, and it carries exactly the same information for a pure state ( $\rho = |\psi\rangle\langle\psi|$ , and (2) reduces to the Schrödinger equation on $|\psi\rangle$ up to an irrelevant global phase). Its value is twofold: it applies to mixed states, which have no wavefunction, and it is the natural starting point for the open-system generalization (the Lindblad equation), where the Schrödinger equation has no analog because there is no single ket to evolve. So the von Neumann equation is the more general home, even though it says nothing new in the closed pure-state case.

Why the opposite sign from the Heisenberg equation — how do I never get it wrong?

Anchor on what moves. In the Schrödinger picture the state moves: $\rho(t) = U\rho_0U^\dagger$ , and differentiating a conjugation-by- $U$ gives $-\tfrac{i}{\hbar}[\hat H,\rho]$ . In the Heisenberg picture the observable moves the other way: $\hat A_H = U^\dagger\hat AU$ , giving $+\tfrac{i}{\hbar}[\hat H,\hat A_H]$ . The mnemonic: the object with $U$ on the left (the state) gets the minus. The foolproof check is to compute $\langle\hat A\rangle(t)$ both ways — they must agree — so a disagreement flags a flipped sign, and it is almost always on $\rho$ .

Does a time-dependent Hamiltonian change any of this?

No — the invariants still hold. For a time-dependent $\hat H(t)$ the solution is $\rho(t) = U(t)\rho_0U^\dagger(t)$ with $U$ a time-ordered exponential, but $U$ is still unitary, so the eigenvalues of $\rho$ are still frozen and purity and entropy are still conserved. The von Neumann equation (2) holds instantaneously with $\hat H(t)$ . A time-dependent Hamiltonian can steer the state anywhere on its constant-purity manifold (that is what quantum gates do), but it cannot leave that manifold — purification and mixing remain the environment's job.

What exactly does the Lindblad term do that the commutator can't?

The commutator term $-\tfrac{i}{\hbar}[\hat H,\rho]$ generates unitary rotation, which preserves eigenvalues. The Lindblad (dissipator) terms $L_k\rho L_k^\dagger - \tfrac12\{L_k^\dagger L_k,\rho\}$ are not of conjugation form, so they move the eigenvalues — flattening the spectrum, lowering purity, raising entropy — and they break time-reversal. A pure dephasing Lindblad operator $L = \sqrt{\Gamma}\,\sigma_z$ , for instance, decays the coherence $|\rho_{01}|\to e^{-t/T_2}|\rho_{01}(0)|$ while leaving populations fixed. The full theory (which $L_k$ arise, how, and the resulting decoherence) is owned by Quantum Decoherence and the Density Matrix by name.

Is there a Bloch-equation picture for the full open-system dynamics too?

Yes — the optical/magnetic Bloch equations are the von Neumann equation plus Lindblad relaxation written for the Bloch vector: $\dot{\mathbf r} = \boldsymbol\omega\times\mathbf r$ (the precession this article owns) plus relaxation terms $-r_x/T_2,\,-r_y/T_2,\,-(r_z - r_{\rm eq})/T_1$ that shrink the vector toward equilibrium. The precession is the reversible, purity-preserving part; the $T_1$ / $T_2$ terms are the environment pulling the vector into the ball's interior. NMR, laser cooling, and cavity QED all run on this combined equation; the relaxation half is owned downstream by name.