The Schrodinger Picture

Here is a question that sounds philosophical but is actually bookkeeping: when a quantum system evolves, what is doing the moving? Every prediction quantum mechanics makes is a sandwich — a state on each side, an operator in the middle, $\langle\psi|\hat A|\psi\rangle$ — and only the whole sandwich is measurable. So the theory gets to choose where inside the sandwich the time dependence lives. The Schrödinger picture makes the choice you have been using all along without naming it: the state carries all the time dependence, the operators stand still. The wavefunction spreads and sloshes; the position operator just sits there being multiplication by $x$ . This article makes the default explicit — because the chapter is about to offer two rival bookkeeping schemes (the Heisenberg picture, where operators move and states freeze, and the interaction picture, which splits the difference), and you cannot appreciate the alternatives until the familiar choice is stated as a choice. We formalize the moving state against the static scaffolding of observable eigenkets, derive the two-state Bohr oscillation, package the practical craft into the expand–rotate–resum workflow of Spectral Decomposition of Time Evolution, and — most importantly — establish which numbers are picture-independent, so nothing physical can ever depend on the bookkeeping. Mastering this costs little (it is The Time-Evolution Operator applied once more) and buys the entire pictures trilogy: the next article inverts this one wholesale, and the equivalence proof waits two serials ahead.

Key takeaways

In the Schrödinger picture the state $|\psi(t)\rangle$ carries all dynamical time dependence: $i\hbar\,\frac{d}{dt}|\psi(t)\rangle = \hat H|\psi(t)\rangle$ , solved by $|\psi(t)\rangle = \hat U(t,t_0)|\psi(t_0)\rangle$ .
Operators representing observables are static scaffolding — $\hat x$ , $\hat p$ , $\hat H$ do not move; their eigenkets form fixed axes against which the rotating state is read.
"Static" means not dragged by the dynamics: an operator may still carry explicit time dependence built in by hand, like a driven field $\hat V(t)$ .
A stationary state only spins its phase, $e^{-iE_nt/\hbar}$ ; all observable time dependence comes from relative phases between energy components, beating at Bohr frequencies $\omega_{mn} = (E_m - E_n)/\hbar$ .
Expectation values, probabilities, and amplitudes are picture-independent; the Schrödinger picture is a choice of bookkeeping, not a claim about nature.
Practical workflow: expand in energy eigenstates → rotate each phase → resum. Wave mechanics is this picture projected onto $\langle x|$ .
Key idea: put all the time dependence on the state and none on the observables; the moving state read against fixed operator axes reproduces every measurable number.

The core insight

Put all the time dependence on the state and none on the observables: $|\psi(t)\rangle = \hat U(t,t_0)|\psi(t_0)\rangle$ with $i\hbar\,\partial_t\hat U = \hat H\hat U$ — expectation values then come from sandwiching fixed operators between evolving states.

Where this lives today

Every numerical quantum-dynamics code that propagates a state — split-operator solvers sweeping femtosecond chemistry, Krylov-subspace exponentiators, tensor-network TDVP evolving a thousand-site spin chain — is a Schrödinger-picture engine: one state vector updated timestep by timestep against fixed, sparse operators. That is the decision table's third row industrialized: a vector is cheap to store and update; a full operator matrix is not. The same logic governs quantum hardware: a circuit on a superconducting or trapped-ion processor is literally prepare–evolve–measure — the qubit worked example running at scale, its 5 GHz splitting and 200 ps precession being honest laboratory numbers. The modern twist is that the rival picture powers the classical simulation of those same devices, tracking evolving operators instead of the exponentially large state (next article's today-band). Knowing when to hop pictures is now a working skill, not a formalism footnote.

The physical picture

Picture the state vector as a unit-length arrow in Hilbert space and the eigenkets of your observables as a rigid frame of axes bolted to the laboratory. In the Schrödinger picture the arrow rotates — rigidly, unitarily, never changing length — while the frame never moves. Measurement statistics ask one geometric question: where does the rotating arrow point, relative to the fixed axes, right now? Its component along $|a_n\rangle$ is the amplitude whose squared magnitude the Born rule turns into the probability of outcome $a_n$ .

Figure

The Schrödinger picture — a rotating state read against fixed operator axes

Two features of this cartoon deserve promotion to physics. First, the rotation is rigid: $\hat U(t,t_0)$ is unitary (proved in The Time-Evolution Operator), so overlaps and probabilities are preserved and norms never drift. Second, the axes are special only when they are energy eigenkets: a state pointing exactly along $|E_n\rangle$ never tips toward any other axis — it just twirls its internal complex phase, which no single measurement can see. That is the geometric soul of a stationary state, and it means all visible motion lives in the beats between energy components.

Figure

Only relative phase moves the needle — two energy components beating at the Bohr frequency

Why make the default explicit at all? Because the sandwich $\langle\psi(t)|\hat A|\psi(t)\rangle$ can be regrouped into a frozen state sandwiching a moving operator — the entire Heisenberg picture, one serial ahead. Nothing measurable can tell the two apart; the pictures differ only in which factor of the sandwich holds the clock. This article works the left-hand side and flags what is invariant, so the handoff is clean.

The law of motion and its solution

The dynamical postulate, stated in ket language in Time-Dependent Schrodinger Equation and integrated in The Time-Evolution Operator, is a first-order linear equation for the moving arrow — so the initial state alone fixes the entire future and past:

i\hbar\,\frac{d}{dt}|\psi(t)\rangle = \hat H\,|\psi(t)\rangle \qquad (1)

where: $\hat H$ is the Hamiltonian, the hermitian generator of time translation, and the state ket does all the moving. Linearity means superpositions evolve component by component — evolution never scrambles the principle of superposition. The equation's solution map is the unitary operator this chapter opened with:

|\psi(t)\rangle = \hat U(t,t_0)\,|\psi(t_0)\rangle, \qquad i\hbar\,\frac{\partial}{\partial t}\hat U(t,t_0) = \hat H(t)\,\hat U(t,t_0) \qquad (2)

where: $\hat U(t_0,t_0) = \hat 1$ , and unitarity $\hat U^\dagger\hat U = \hat 1$ — equivalent to the hermiticity of $\hat H$ , as The Time-Evolution Operator proved — keeps $\langle\psi(t)|\psi(t)\rangle = 1$ forever. Probability is conserved because the rotation is rigid. For the workhorse case of a time-independent Hamiltonian the solution exponentiates, $\hat U = e^{-i\hat H(t-t_0)/\hbar}$ ; the general time-dependent case needs the time-ordered machinery of Time-Ordered Exponential and the Dyson Series and is quoted, never re-derived, here. Sandwiching a fixed operator between moving states gives every prediction of the theory:

\langle\hat A\rangle_t = \langle\psi(t)|\,\hat A\,|\psi(t)\rangle \qquad (3)

where: $\hat A$ carries no dynamical time dependence — position today is the same operator as position tomorrow. An operator may still depend on time explicitly, because you built the time in by hand: a laser field $\hat V(t) = -e\hat x\,\mathcal E_0\cos\omega t$ is a Schrödinger-picture operator with a clock in its coefficient, not an operator dragged by the dynamics. The distinction matters enormously one article ahead, where a second, dynamical time dependence gets layered on top. Which equation to reach for, when:

you have	reach for	because
$\hat H$ time-independent, spectrum known	expand–rotate–resum (eq. 5)	phases are the whole story
$\hat H$ time-independent, spectrum unknown	$e^{-i\hat H t/\hbar}$ numerically	one matrix exponential
$\hat H(t)$ , all times commuting	$\exp[-\tfrac{i}{\hbar}\int\hat H\,dt']$	ordering is free
$\hat H(t)$ non-commuting	Dyson series (its own article)	ordering is the physics

The spectral solution: expand, rotate, resum

For time-independent $\hat H$ with eigenkets $\hat H|n\rangle = E_n|n\rangle$ , the picture's practical craft is three moves long. Expand the initial state on the energy basis, spin each component's phase clock, and resum:

|\psi(t)\rangle = \sum_n c_n\,e^{-iE_n(t-t_0)/\hbar}\,|n\rangle, \qquad c_n = \langle n|\psi(t_0)\rangle \qquad (4)

where: the $c_n$ are fixed at $t_0$ once and for all — dynamics never transfers amplitude between energy eigenstates, only rotates phases. The algorithm is Spectral Decomposition of Time Evolution's; this article's contribution is the reading: in the energy basis the picture's motion is as simple as possible — $N$ independent phase clocks, nothing else. (Verifier: recipe versus direct matrix exponential on a random three-level system, $10^{-13}$ .) A single eigenstate evolves as $e^{-iE_mt/\hbar}|m\rangle$ : a pure phase, the same ray forever — the stationary states of Stationary States and Energy Eigenstates. All observable dynamics therefore lives in cross terms. Take two components and sandwich an observable:

\langle\hat A\rangle_t = |c_1|^2A_{11} + |c_2|^2A_{22} + 2\,\mathrm{Re}\!\left[c_1^*c_2\,A_{12}\,e^{-i\omega_{21}t}\right], \qquad \omega_{21} = \frac{E_2 - E_1}{\hbar} \qquad (5)

where: $A_{jk} = \langle j|\hat A|k\rangle$ and hermiticity supplies $A_{21} = A_{12}^*$ , which is what folds the two cross terms into one real oscillation. The diagonal terms are frozen; the interference term beats at the Bohr frequency $\omega_{21}$ — the only rate the system can broadcast. Shift every energy by a constant and $\omega_{21}$ doesn't move: only energy differences are observable, a theme Time-Dependent Schrodinger Equation proves as the global-phase theorem and this equation displays in action.

Derivation ATwo energy components produce one Bohr oscillation

We derive equation (5), the workhorse formula of the picture: a two-component superposition makes any observable's expectation value oscillate sinusoidally at exactly the Bohr frequency $(E_2 - E_1)/\hbar$ . The strategy is direct substitution of the spectral solution into the sandwich (3), watching which phases survive. The step to watch is the pairing of cross terms: hermiticity of $\hat A$ is what makes the two off-diagonal contributions complex conjugates of each other, so they add up to something real — as an expectation value must be. This is the calculation the base provenance asks every verifier to recompute, and ours did, numerically, on a random hermitian $\hat A$ across a full grid of times.

Step 1. Write the evolving state with both phase clocks running, taking $t_0 = 0$ :

|\psi(t)\rangle = c_1 e^{-iE_1t/\hbar}|1\rangle + c_2 e^{-iE_2t/\hbar}|2\rangle

Why the $c_n$ are constants: dynamics only rotates each component's phase — the expansion coefficients' magnitudes are set once, at preparation.

Step 2. Substitute into $\langle\hat A\rangle_t = \langle\psi(t)|\hat A|\psi(t)\rangle$ and expand the four terms. The diagonal pieces carry phase factors $e^{+iE_nt/\hbar}e^{-iE_nt/\hbar} = 1$ :

\langle\hat A\rangle_t = |c_1|^2A_{11} + |c_2|^2A_{22} + c_1^*c_2\,A_{12}\,e^{-i(E_2-E_1)t/\hbar} + c_2^*c_1\,A_{21}\,e^{+i(E_2-E_1)t/\hbar}

Why the diagonal terms freeze: a bra and a ket of the same energy spin their clocks in opposite directions at the same rate — the phases cancel identically, for every observable. Frozen diagonals are the stationary-state theorem in miniature.

Step 3. Use hermiticity, $A_{21} = A_{12}^*$ . The two cross terms become a number plus its own complex conjugate, and $z + z^* = 2\,\mathrm{Re}\,z$ :

\langle\hat A\rangle_t = |c_1|^2A_{11} + |c_2|^2A_{22} + 2\,\mathrm{Re}\!\left[c_1^*c_2\,A_{12}\,e^{-i\omega_{21}t}\right]

Step 4. Read the anatomy. A constant plateau (the weighted average of the two frozen diagonals) plus one cosine of amplitude $2|c_1c_2A_{12}|$ and frequency $\omega_{21}$ , phase-offset by the arguments of $c_1^*c_2$ and $A_{12}$ . Maximal contrast needs an even mix ( $|c_1| = |c_2| = 1/\sqrt2$ ) and an observable that connects the levels ( $A_{12} \neq 0$ ): an operator diagonal in the energy basis shows no beat at all — it cannot tell the clocks apart. Verified: numerically — the boxed form matches the direct sandwich on a random hermitian operator at 25 times across several periods to 10⁻¹³, and $A_{21} = A_{12}^*$ holds to machine precision (verify script, sections 5 and 8).

What no picture can change

Before handing the sandwich to the Heisenberg picture, we record what is not up for negotiation. Regroup the expectation value using $|\psi(t)\rangle = \hat U|\psi(0)\rangle$ :

\langle\phi(t)|\,\hat A\,|\psi(t)\rangle \;=\; \langle\phi(0)|\,\underbrace{\hat U^\dagger\hat A\,\hat U}_{\text{a moved operator}}\,|\psi(0)\rangle \qquad (6)

where: the left side is Schrödinger bookkeeping (moving states, fixed $\hat A$ ), the right side is Heisenberg bookkeeping (fixed states, moved operator) — the same number. Amplitudes, probabilities, expectation values, and spectra are all built from such sandwiches, so all are picture-independent. The systematic statement — that all pictures agree on every physical prediction, with the full dictionary — is the theorem of Equivalence of the Dynamical Pictures; this article derives only the seed identity.

Go deeper · Derivation B · 1 minOne sandwich, two bookkeepings — the regrouping identity

Derivation BOne sandwich, two bookkeepings — the regrouping identity

We derive equation (6), the identity from which the whole pictures trilogy grows: any amplitude computed with evolved states equals the same amplitude computed with frozen states and a conjugated operator. The strategy is embarrassingly short — insert the propagation law and re-associate the product — which is precisely the point: the freedom to choose a picture costs nothing because it is associativity. Watch where the dagger comes from: the bra evolves with $\hat U^\dagger$ , the antilinear flip of Hermitian Conjugate and Adjoint Operators on duty.

Step 1. Evolve both states from $t_0 = 0$ : $|\psi(t)\rangle = \hat U(t)|\psi(0)\rangle$ , and correspondingly for the bra, $\langle\phi(t)| = \langle\phi(0)|\hat U^\dagger(t)$ . Why the dagger: the ket-to-bra correspondence conjugates operators — $\hat U|\phi\rangle$ has bra $\langle\phi|\hat U^\dagger$ , the dual dictionary of the adjoint article.

Step 2. Substitute into the sandwich and re-associate:

\langle\phi(t)|\hat A|\psi(t)\rangle = \langle\phi(0)|\,\hat U^\dagger\hat A\,\hat U\,|\psi(0)\rangle = \langle\phi(0)|\,\hat A_H(t)\,|\psi(0)\rangle

Step 3. Name the middle object: $\hat A_H(t) \equiv \hat U^\dagger\hat A\hat U$ , with $\hat A_H(0) = \hat A$ since $\hat U(0) = \hat 1$ . That object — the operator wearing the time dependence the states gave up — is the Heisenberg-picture operator, constructed properly in the next article. Nothing measurable changed; only the label saying who holds the clock. Verified: numerically — the identity holds on ten random three-level systems (random hermitian $\hat H$ , $\hat A$ , random state pairs, random times) to 10⁻¹³, and $\hat A_H(0) = \hat A$ exactly (verify script, section 1).

Two invariance facts worth naming now: the spectrum of an observable never moves (a similarity transform cannot change eigenvalues), and the algebra never moves (commutators conjugate cleanly). The Heisenberg Picture carries the full who-carries-time ledger; we do not duplicate it here.

Choosing a picture: the decision table

The choice is pure convenience, and the conveniences are real. This table is the article's practical takeaway — the Heisenberg article recalls it rather than re-printing it:

situation	picture of choice	why
wavefunction problems, wells, barriers	Schrödinger	wave mechanics is this picture in the position basis
state preparation → evolve → measure (circuits, pulses)	Schrödinger	you track exactly the object the protocol manipulates
numerical propagation on a grid or in Krylov space	Schrödinger	one vector to update; operators stay sparse and fixed
operator algebra, conservation laws, correspondence with classical mechanics	Heisenberg	operators obey classical-looking equations (next article)
two-time correlation functions $\langle\hat A(t)\hat B(t')\rangle$	Heisenberg	needs operators wearing two different time labels
quantum field theory	Heisenberg	fields are operators labeled by spacetime points
perturbation theory with $\hat H = \hat H_0 + \hat V(t)$	interaction	splits the labor: fast known part on operators, slow part on states

Limiting cases

case	form	meaning
time-independent $\hat H$	$\|\psi(t)\rangle = e^{-i\hat H(t-t_0)/\hbar}\|\psi(t_0)\rangle$	one matrix exponential; phases do everything
energy eigenstate	$\|\psi(t)\rangle = e^{-iE_n(t-t_0)/\hbar}\|n\rangle$	stationary: same ray forever, all statistics frozen
commuting family $[\hat H(t_1),\hat H(t_2)] = 0$	$\hat U = \exp[-\tfrac{i}{\hbar}\int_{t_0}^{t}\hat H(t')dt']$	ordering costs nothing; area under $\hat H$ is all that counts
general $\hat H(t)$	$\hat U = \mathcal T\exp[-\tfrac{i}{\hbar}\int_{t_0}^{t}\hat H(t')dt']$	time-ordering required — the Dyson series article owns this
$\hat H \to \hat H + E_0\hat 1$	$\|\psi(t)\rangle \to e^{-iE_0t/\hbar}\|\psi(t)\rangle$	global phase only; energy zero is a gauge choice

Common trapOperators are time-independent in the Schrödinger picture — so nothing about them can ever involve time

Two traps hide in that sentence, and the base article's own pitfall flags both. First: "time-independent" means not dragged by the dynamics — an operator can still carry explicit, built-in-by-hand time dependence, like a driven potential $\hat V(x,t)$ or a rotating-field Hamiltonian. The Schrödinger picture forbids dynamical motion of operators, not clocks in coefficients. Second, subtler: a stationary state's phase $e^{-iE_nt/\hbar}$ looks like pure decoration, and in isolation it is — but the relative phases between different-energy components are the entire observable dynamics of the theory. Every beat, every Rabi flop, every oscillating dipole that makes an atom radiate is relative phase at work (equation 5). Dismiss phases as unphysical and you have dismissed motion itself; the correct statement is that only global phase is unphysical.

Worked examples

Example 1 — the two-level beat, end to end (confidence builder). A system has two energy levels $E_1 = 1.0$ eV and $E_2 = 1.5$ eV, prepared at $t = 0$ in the even mix $|\psi(0)\rangle = (|1\rangle + |2\rangle)/\sqrt2$ . An observable $\hat A$ has matrix elements $A_{11} = A_{22} = 0$ , $A_{12} = 1$ in the energy basis. Find $\langle\hat A\rangle_t$ and the beat period.

Step 1. Coefficients: $c_1 = c_2 = 1/\sqrt2$ , so $|c_1|^2 = |c_2|^2 = 1/2$ and $c_1^*c_2 = 1/2$ .

Step 2. Apply the boxed form (5). Diagonals contribute zero; the cross term gives $\langle\hat A\rangle_t = 2\,\mathrm{Re}[\tfrac12 e^{-i\omega_{21}t}] = \cos\omega_{21}t$ .

Step 3. The Bohr frequency: $\hbar\omega_{21} = E_2 - E_1 = 0.5$ eV. The period is $T = 2\pi/\omega_{21} = h/(E_2 - E_1) = 4.1357\times10^{-15}\,\text{eV}\,\cdot\,\text{s}/0.5\,\text{eV} = \boxed{8.27\ \text{fs}}$ .

Step 4. Audit the quarter period: at $t = T/4$ the phase is $\pi/2$ , so $\langle\hat A\rangle = 0$ — machine-checked. An observable diagonal in the energy basis ( $A_{12} = 0$ ) would read a flat line instead: beats need an operator that connects the levels.

Example 2 — a 5 GHz qubit on the Bloch equator (exam level). A superconducting qubit has level splitting $E_+ - E_- = \hbar\omega$ with $\omega/2\pi = 5.0$ GHz; take $\hat H = \tfrac{\hbar\omega}{2}\hat\sigma_z$ . It is prepared in $|{+}x\rangle = (|{+}z\rangle + |{-}z\rangle)/\sqrt2$ . Find the splitting in eV, the precession trajectory, and the state of the Bloch vector at $t = 50$ ps.

Step 1. The splitting: $\hbar\omega = hf = 4.1357\times10^{-15} \times 5.0\times10^{9} = \boxed{20.7\ \mu\text{eV}}$ — sub-millikelvin in temperature units, which is why these devices live in dilution refrigerators.

Step 2. Evolve. With this sign convention the verifier pins the trajectory: $\langle\hat\sigma_x\rangle_t = \cos\omega t$ , $\langle\hat\sigma_y\rangle_t = +\sin\omega t$ , $\langle\hat\sigma_z\rangle_t = 0$ . The Bloch vector precesses in the equatorial plane, from $+x$ toward $+y$ , at exactly the Bohr frequency — this is equation (5) three times over, once per Pauli component.

Step 3. Period and snapshot: one full revolution takes $T = 1/f = \boxed{200\ \text{ps}}$ . At $t = 50$ ps $= T/4$ : $\langle\hat\sigma_x\rangle = \cos(\pi/2) = 0$ , $\langle\hat\sigma_y\rangle = \sin(\pi/2) = \boxed{1}$ — the Bloch vector points along $+y$ (machine-checked to $10^{-10}$ ).

Step 4. Read it as bookkeeping: the state moved; $\hat\sigma_x$ , $\hat\sigma_y$ never did. The next article re-runs this identical qubit with the state frozen and the operators precessing — same measurable numbers, the cleanest side-by-side the pictures allow.

The physicist: dynamics as a principle

Origins · 1 minJulian Schwinger and how this was discovered

Julian Schwinger

1918–1994 · Harvard, then UCLA

Schwinger was the counter-Feynman: where Feynman drew pictures, Schwinger wrote operator equations of such compressed elegance that colleagues joked his papers were "written for God." A New York prodigy who published at sixteen and took his Columbia PhD at twenty-one, he became Harvard's oracle of formalism, supervising over seventy doctoral students — four future Nobel laureates among them, alongside his own prize. His signature contribution to this article's subject is the quantum action principle (1951): where the Schrödinger picture says "states evolve, here is the differential equation," Schwinger recast dynamics as a variational statement about transformation amplitudes $\langle\phi(t)|\psi(t_0)\rangle$ — one principle from which the equation of motion, the commutation relations, and the conservation laws all flow. It is the sharpest expression of the idea this article opens with: dynamics is a statement about amplitudes, and the split into moving states versus moving operators is ours to choose. Legend has him lecturing without notes, ambidextrously writing two equations at once — one hand per blackboard.

History

1925

— Matrix mechanics arrives first: time dependence carried by observables, not states

1926

— Schrödinger's wave mechanics puts the time dependence on a evolving wavefunction — the picture named after him

1927

— Dirac's transformation theory reveals the two formulations as one dynamics in different clothes

1951

— Schwinger's quantum action principle recasts evolution as a variational statement about amplitudes

1982

— Feit and Fleck's split-operator method makes numerically propagating the Schrödinger-picture state an industrial tool

2023

— Utility-scale quantum processors run prepare–evolve–measure circuits: the Schrödinger picture as hardware protocol

Going further

References. Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §2.1–2.2 · Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., Ch. 2 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 4 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. I, Ch. III · Townsend, A Modern Approach to Quantum Mechanics, 2nd ed., Ch. 4.

Provenance. Equations and structure follow Sakurai & Napolitano (Modern QM, 2nd ed., §2.1–2.2), Griffiths & Schroeter (3rd ed., Ch. 2), Shankar (Ch. 4), Cohen-Tannoudji (Vol. I, Ch. III), and Townsend (Ch. 4). A verifier should recompute: (1) the cancellation establishing $\partial_t(U^\dagger U)=0$ ; (2) the two-state expectation value reducing to the boxed Bohr-oscillation form with $A_{21}=A_{12}^*$ ; (3) the operator regrouping $A_H=U^\dagger A U$ with $A_H(t_0)=A$ . The numerical value of $\hbar$ is the CODATA standard $1.054571817\times10^{-34}\,\mathrm{J\,s}$ . For this edition all three provenance items plus every numeric claim were machine-verified (numpy/sympy, 25 checks): the regrouping identity on ten random three-level systems ( $10^{-13}$ ), the symbolic and numeric unitarity cancellations, the pinned Bloch-precession convention ( $\hat H = +\tfrac{\hbar\omega}{2}\hat\sigma_z$ , $|{+}x\rangle$ : $\cos\omega t$ , $+\sin\omega t$ ), the expand–rotate–resum recipe against direct exponentiation, the boxed two-state form on a random hermitian observable, the stationary-state and global-phase theorems, and both worked examples (8.27 fs beat; 20.7 μeV, 200 ps, quarter-period snapshot). Per the chapter ownership map: the construction and properties of $\hat U$ are owned by The Time-Evolution Operator; the expansion algorithm by Spectral Decomposition of Time Evolution; the equivalence theorem by Equivalence of the Dynamical Pictures; $d\langle\hat A\rangle/dt$ in commutator form by the Quantum Dynamics chapter's Ehrenfest's Theorem; this article owns the picture's contract, the decision table, and the two-state Bohr-oscillation derivation.

Builds on: Time-Dependent Schrodinger Equation · The Time-Evolution Operator · Stationary States and Energy Eigenstates

Leads to: The Heisenberg Picture · The Interaction (Dirac) Picture · Equivalence of the Dynamical Pictures · Ehrenfest's Theorem · Spectral Decomposition of Time Evolution

If the pictures are all equivalent, why does this one get to be the default?

Partly history (wave mechanics conquered the textbooks early), partly ergonomics. Most first problems are wavefunction problems, and $\psi(x,t) = \langle x|\psi(t)\rangle$ is precisely the Schrödinger-picture state in the position basis. A moving vector is also the cheapest object to compute with: codes update one state, not a full matrix of operators. The Heisenberg picture wins in different terrain — operator algebra, conservation laws, field theory — which is why the chapter teaches both and gives you the decision table.

Is the wavefunction ψ(x,t) the same thing as the state ket |ψ(t)⟩?

The wavefunction is the state's shadow in one particular basis: $\psi(x,t) = \langle x|\psi(t)\rangle$ , the component of the moving ket along the fixed position eigenket $|x\rangle$ . All of wave mechanics is the Schrödinger picture projected onto position eigenkets — the position representation article of the Formalism chapter built that dictionary. The ket is basis-free; the wavefunction is one column of its coordinates.

The phase e^(−iEt/ħ) of a stationary state is unobservable — so why does the article keep insisting phases matter?

Because global and relative are different words. One overall phase on the entire state drops out of every sandwich — that is the global-phase theorem. But a superposition of two energies carries a phase difference that advances at the Bohr frequency, and that difference sits inside the cross term of equation (5), producing every observable oscillation in quantum physics: quantum beats, Rabi flopping, the radiating dipole of an excited atom. The phase of one clock is invisible; the angle between two clocks is dynamics itself.

Why does an observable that is diagonal in the energy basis never show any time dependence?

The cross term in equation (5) is proportional to $A_{12}$ ; diagonal-in-energy observables have $A_{12} = 0$ , so only the frozen diagonal terms survive — a flat line for any initial state. Such an observable cannot see the relative phase of the energy components; it commutes with $\hat H$ and is a constant of the motion (developed in Constants of the Motion and Conservation Laws). To see a beat you need an operator that connects the beating levels.

Does the state really "rotate" in Hilbert space, or is that just a metaphor?

Nearly literal. Unitary evolution preserves inner products the way a rigid rotation preserves angles and lengths, so the state traces a path on the unit sphere of Hilbert space, never stretching. Two cautions: the space is complex (the "rotation" mixes real and imaginary parts — that is where phases live), and along an energy eigenket the motion is purely internal, a phase spin that moves the vector but not the ray. The qubit's Bloch sphere is the one case where it becomes an exactly visualizable real rotation.

When the Hamiltonian itself depends on time, is the Schrödinger picture still usable?

Fully — the state still obeys equation (1) and $\hat U$ still exists and is unitary; only the convenient formulas break. The exponential $e^{-i\hat Ht/\hbar}$ and the expand–rotate–resum recipe assume a time-independent $\hat H$ . For driven systems you integrate (2) numerically, or — when the drive is weak — use the Dyson series and the interaction picture, the next articles' entire business.

How literally should I take "operators never move" when the lab apparatus clearly moves — magnets rotate, lasers switch on?

Moving apparatus means explicit time dependence in a coupling operator — a clock in a coefficient, $\hat V(x,t)$ — perfectly at home in this picture (the warning's first trap). What operators never acquire here is dynamical time dependence: the conjugation by $\hat U$ the Heisenberg picture performs. The two kinds appear as two separate terms in the Heisenberg equation of motion, one article ahead.