The Heisenberg Equation of Motion
The previous article built the Heisenberg picture and left it at the threshold of its own law: state kets frozen at , every operator carrying the full burden of motion, the free particle solved as an appetizer. This article writes the law itself. Differentiate the conjugation once, and out falls the single most classical-looking equation in quantum mechanics:
Read it next to Hamilton's and the resemblance is not decorative — it is the deepest structural fact about quantization. The commutator divided by sits exactly where the Poisson bracket sat in 1809, when celestial mechanics invented the bracket to track how orbital elements drift. Dirac saw the correspondence in a flash of an October Sunday in 1925 and quantum mechanics has never let go of it. The payoffs are immediate. Conservation laws become one-line statements: commute with and you are frozen — as an operator, not merely on average. Linear systems integrate in closed form: this article solves the oscillator's operators exactly, watching and rotate into each other like a phase-space clock while the operator algebra rides along undamaged. Spin precession drops out of a two-line commutator evaluation. And sandwiched between fixed states, the equation becomes Ehrenfest's theorem — the expectation-value engine two articles on. Master this one derivation and the rest of the chapter is applications.
- Heisenberg-picture operators obey — the operator law of motion, obtained by differentiating .
- For a time-independent Hamiltonian, so : the same generates the motion in both pictures.
- The structure mirrors Hamilton's equation with — Dirac's 1925 correspondence, the grammar of canonical quantization.
- and make forever: a constant of the motion at the full operator level.
- The oscillator solves exactly: , , or in one line .
- Equal-time commutators survive the flow: at every — unitary conjugation preserves the algebra.
- Key idea: the Hamiltonian rotates every observable through operator space at a rate set by their commutator — Hamilton's equations, upgraded to operators, with the Poisson bracket replaced by .
Time evolution acts on observables by unitary conjugation ; differentiating gives — the operator form of Hamilton's equations, with the Poisson bracket replaced by times the commutator.
Where this lives today
Field theory is written almost entirely in this article's grammar: in QED the states are timeless while field operators evolve by Heisenberg equations, with equal-time commutators — preserved by the flow exactly as in Derivation B's Step 5 — as axioms. Closer to hardware, the input–output formalism of quantum optics propagates cavity and qubit operators (not states) through amplifiers and lines: LIGO's squeezer analysis and every superconducting-qubit readout chain are Heisenberg-picture calculations, with as the zeroth-order skeleton and commutator preservation enforcing the quantum noise floor. The equation also anchors modern many-body physics: "operator spreading" — how a simple grows non-local under conjugation — is today's working definition of quantum-information scrambling, measured on quantum processors through out-of-time-order correlators built from .
The physical picture
The Heisenberg Picture set the stage with a change of frame: instead of a state rotating through a fixed Hilbert space, hold the state fixed and let the measuring apparatus — the operators — rotate the opposite way. What that article did not yet supply is the rate law: given , how fast, and along what direction in operator space, does each observable turn? The answer is the commutator. measures precisely the failure of to fit the energy's eigenframe, and that misfit is the torque. An observable that commutes with is aligned with the flow and rides it unchanged — conservation. An observable that fails to commute gets churned, at a rate proportional to the failure. Energy is the generator of time translation exactly as momentum generates displacements (the translation family of Canonical Commutation Relation) and rotations will be generated by angular momentum in the next chapter.
The second picture to hold onto is the oscillator, because there the flow is literally a rotation. Under , the pair turns rigidly at angular rate , forever, as operators. Nothing spreads, nothing damps — spreading is a property of states, and the states are standing still in this picture. The whole quantum content lives in the fact that the two hands of the clock refuse to commute, and the rotation is exactly rigid because it preserves that refusal: at every instant.
The equation and its anatomy
Recall the two facts this article differentiates. The Time-Evolution Operator built obeying ; The Heisenberg Picture defined moving observables by conjugation, , and showed by matrix-element matching that this is the unique assignment agreeing with the Schrödinger picture on all physics. One derivative later (Derivation A below), the law of motion reads:
where: is the moving observable; is the Hamiltonian transported to the same picture; and collects any time dependence the operator carries by construction — a ramped field, a rotating trap axis — before dynamics touches it. The first term is universal and structural; the second is optional bookkeeping that vanishes for the static observables (, , , parity…) that fill most problems. One subtlety deserves its own sentence. When is time-independent, is a function of alone, so and
where: the subscript can simply be dropped — the same operator generates the motion in both pictures, which is why equation (1) is usually quoted with a bare . When depends on time the evolution operator becomes a time-ordered exponential ([H(t),H(t′)]≠0 in general), the two Hamiltonians part company, and the subscript earns its keep — Time-Ordered Exponential and the Dyson Series owns that regime. The third piece of anatomy is the classical dictionary. In Hamiltonian mechanics an observable drifts along the flow as , with the Poisson bracket. Equation (1) is that sentence with one substitution:
where: the correspondence runs deep — both brackets are antisymmetric, both satisfy the Jacobi identity and the product rule (The Commutator and Compatible Observables proved the algebra pack), so the two dynamics share their entire formal skeleton. This is Dirac's 1925 insight, and it is why so many Heisenberg equations read word-for-word like their classical ancestors: the free particle inherits uniform motion, the oscillator inherits rotation, spin inherits Larmor's precession. What changes is the cast: the letters now refuse to commute, and sets the exchange rate.
We derive equation (1) by differentiating the conjugation directly. The strategy is the product rule plus the one dynamical input available — the operator Schrödinger equation for and its adjoint — followed by a re-dressing step that turns every surviving Schrödinger operator into its Heisenberg counterpart. The manoeuvre to watch is the insertion of between and : that single move is what makes the final equation close among Heisenberg-picture objects instead of leaving a stranded mixed expression. We keep general so the explicit-time term appears honestly rather than being bolted on.
Step 1. Differentiate the definition with the product rule — three terms, one per moving factor:
Step 2. Feed in the evolution law and its adjoint (hermitian ):
Why the signs land this way: carries 's minus sign, its adjoint the plus — the two dynamical terms arrive with opposite signs, already shaped like a commutator-to-be.
Step 3. Insert between each pair to re-dress both factors:
Step 4. Collect. The two dynamical terms assemble into the commutator, the third term is by definition :
Why nothing was assumed about 's time dependence: the derivation used only , which holds for time-dependent Hamiltonians too (with the time-ordered exponential). The time-independent case then adds the bonus of equation (2). *Verified: the identity machine-checked symbolically on a generic hermitian 2×2 Hamiltonian (sympy, exact zero residual); signs of both -equations cross-checked against Sakurai §2.2.*
The commutator craft — and the operator Newton laws
Using equation (1) is a two-move routine: evaluate one commutator, read off one differential equation. The evaluation move was banked in the formalism chapter — Canonical Commutation Relation proved the corollary ladder and — and for the standard Hamiltonian it delivers, in two lines each:
where: ; only the kinetic term sees (giving ) and only the potential sees . Feeding (4) into (1) produces the operator Newton laws:
where: the right-hand sides are evaluated at the Heisenberg operators — conjugation by passes through functions, . Equation (5) is with hats. For the free particle () it integrates to — but that solution, its terminating BCH derivation, and the startling non-equal-time commutator are owned by The Heisenberg Picture; we quote and move on to the potential that (5) solves exactly and nontrivially.
Go deeper · Derivation B · 1 minSolving the harmonic oscillator at the operator level
We integrate the coupled operator equations (5) for in closed form. The strategy exploits linearity: the equations couple and linearly with c-number coefficients, so they decouple and solve exactly like their classical counterparts, with the initial operators standing where initial values would stand classically. The care point is ordering — we never need to commute past , precisely because the coefficients are c-numbers; that is why the classical solution survives verbatim. We finish with the ladder-operator form, one line long, and the consistency check that the flow preserves the canonical algebra.
Step 1. Specialize (5): and . Differentiate the first and insert the second:
Why this is legitimate operator calculus: each Heisenberg operator is a fixed linear combination of and with time-dependent scalar coefficients, so differentiation acts on the coefficients only — no ordering ambiguity ever arises.
Step 2. Solve the simple-harmonic operator equation with operator initial conditions , :
Step 3. Differentiate and multiply by to get the momentum partner:
Step 4. Repackage with the ladder operator . Substituting Steps 2–3:
Why the ladder form is the cleanest statement: makes equation (1) read — a single first-order equation whose solution is one exponential. All oscillator dynamics is this one rotating phase.
Step 5. Consistency: compute the equal-time commutator of the solutions,
Why it had to work: is an algebra homomorphism — unitary conjugation preserves every commutator, so the canonical structure is carried rigidly by the flow. Getting is the internal audit that Steps 2–3 really are a unitary orbit. *Verified: both closed forms machine-checked — symbolically (the coupled ODEs and the all- commutator, sympy with noncommuting symbols) and numerically (conjugation by in a 60-level Fock truncation matches and on the interior 40×40 block to at three times; the truncation-corner caveat of Canonical Commutation Relation applies to the last rows only).*
Conservation in one line
Set and in equation (1) and the right side dies:
where: the conclusion is operator-valued — not just frozen, but the entire observable is carried unchanged by the flow, eigenvalues, eigenprojectors and all. Energy itself is the first citizen (); the oscillator's number operator is another (, machine-checked exactly in the verifier). The full theory of what conservation buys — frozen outcome distributions, good quantum numbers, the algebra of constants — is the next article's business, Constants of the Motion and Conservation Laws; here we only note that the Heisenberg equation makes the criterion visible at a glance. Sandwich equation (1) between fixed Heisenberg states and the expectation-value law appears: — picture-independent and directly measurable. That statement is owned three serials on by Ehrenfest's Theorem. Between the operator law and its expectation shadow sits The Interaction (Dirac) Picture (next), where splits and the two terms of (1) go to different masters.
Limiting cases
| case | result | meaning |
|---|---|---|
| the workhorse form — all dynamics generated by the commutator | ||
| , | constant of the motion, frozen as an operator (theory → next article) | |
| time-independent | one Hamiltonian for both pictures; subscript freely dropped | |
| correspondence | equation (1) → Hamilton's | |
| free-particle drift — solved in The Heisenberg Picture | ||
| rigid rotation of | Derivation B — exact, no spreading at operator level |
Worked examples
Example 1 — a CO molecule's vibrating bond, solved as operators. The CO stretch vibrates at wavenumber . Model it as a QHO in the reduced coordinate and find for the superposition , using the Heisenberg solution — no wavefunctions.
Step 1. Convert the spectroscopic frequency: , so one vibrational period is .
Step 2. Reduced mass: .
Step 3. In the Heisenberg picture the state never moves, so with . Evaluate both static matrix elements once: and (the – momentum matrix element is purely imaginary and cancels in this real superposition).
Step 4. The oscillator length gives . Therefore
— the bond's charge centroid swings sinusoidally with amplitude 3.39 pm, about 3% of the 113 pm bond length, at period 15.6 fs. One operator solution served every observable question; a different initial state would only change the two static matrix elements in Step 3.
Example 2 — spin precession from a two-line commutator. An electron sits in along , so with . Find the motion of and the precession frequency.
Step 1. Evaluate the rate: .
Step 2. The spin algebra gives , so equation (1) reads , and likewise , (the sign pattern machine-audited in the verifier).
Step 3. The coupled pair is again a rigid rotation: — the spin operator precesses about while is conserved ().
Step 4. Frequency: 14.01 GHz, period 71.4 ps — exactly half the 28.02 GHz that The Time-Evolution Operator computed at 1 T, as linearity in demands. Every initial spin state inherits this motion by sandwiching; the operator solution is the whole problem.
Origins · 1 minSiméon Denis Poisson and how this was discovered
The bracket in Hamilton's — the exact classical shadow of this article's commutator — was invented in 1809 by a man who never suspected mechanics had a quantum future. Poisson, the École Polytechnique's prodigy-turned-professor ("life is good for only two things: doing mathematics and teaching it"), was hunting the slow drift of planetary orbital elements when he built as a bookkeeping device for perturbed celestial mechanics. The device turned out to be load-bearing: antisymmetric, product-rule-obeying, Jacobi-closing — the complete grammar of how observables change along a Hamiltonian flow. In October 1925, Dirac took his Sunday walk stewing over Heisenberg's strange non-commuting arrays and was struck by a resemblance he could not check until Monday, because no library open on Sunday could remind him what a Poisson bracket was. The match was exact: obeys every algebraic law Poisson's bracket obeys, and the Heisenberg equation is Poisson's dynamics with hats on. His own treatises mostly aged into footnotes; his bracket became the sentence in which quantum dynamics is still written.
Going further
References. Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §2.2 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 4 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. I, Complement G · Dirac, The Principles of Quantum Mechanics, 4th ed., §28 · Merzbacher, Quantum Mechanics, 3rd ed., Ch. 14.
Provenance. Heisenberg EOM derivation verified against Sakurai & Napolitano §2.2 and Shankar Ch. 4. Oscillator solution and operator Newton's laws cross-checked with Cohen-Tannoudji Vol. I Complement G. Larmor precession standard. Verifier should reconfirm: (1) the product-rule derivation yielding ; (2) , for ; (3) preservation of under unitary conjugation; (4) . No fabricated constants used. For this edition all four were machine-executed (sympy/numpy), plus: the QHO closed forms symbolically and in a 60-level Fock truncation (interior block, ); numerically; frozen exactly; the CO numbers ( rad/s, kg, amplitude 3.39 pm) and the spin numbers (0.5 T → 14.01 GHz, 71.4 ps) from CODATA constants. Per the chapter ownership map: the conjugation construction, free-particle solve and non-equal-time commutator belong to The Heisenberg Picture (recalled, never re-derived); the commutator corollary ladder to Canonical Commutation Relation (consumed with citation); the expectation-value form to Ehrenfest's Theorem (stated with owner named); conservation theory to Constants of the Motion and Conservation Laws (teased); the time-dependent-Hamiltonian regime to Time-Ordered Exponential and the Dyson Series.
Builds on: The Time-Evolution Operator · The Schrodinger Picture · The Heisenberg Picture · Time-Dependent Schrodinger Equation
Leads to: Ehrenfest's Theorem · Constants of the Motion and Conservation Laws · Symmetries and Conservation in Quantum Dynamics · The Interaction (Dirac) Picture · Equivalence of the Dynamical Pictures
Is the Heisenberg equation more fundamental than the Schrödinger equation?
Neither outranks the other — they are the same dynamical postulate in different clothes, connected by the unitary re-draping the equivalence article proves. Historically the Heisenberg form came first (matrix mechanics, 1925), and in relativistic field theory it is the more natural dress because it keeps operators local in spacetime while states stay global. For a hydrogen spectrum you will still prefer Schrödinger's PDE. Fundamentality is not the question; convenience per problem is.
Why does the commutator, of all things, control the rate of change?
Because generates time translation, and a generator moves exactly what fails to commute with it. The one-parameter family conjugates operators; expanding to first order in , the change of is proportional to — commuting parts cancel between and , non-commuting parts survive. It is the same logic by which momentum (translation generator) fails to commute with position, and by which anything failing to commute with changes under displacement.
Does the Heisenberg picture make wave packets stop spreading?
No — it relocates the spreading. The state is frozen, but ask a physical question (say the variance of position at time ) and you must evaluate , and the growing operator content of (free case) delivers exactly the same growing variance the Schrödinger picture shows. Pictures cannot change physics; they change which symbol carries it.
What exactly is the difference between dÂ_H/dt and (∂Â/∂t)_H?
The total derivative is everything — the actual rate of change of the moving observable. The partial term is only the built-in time dependence the Schrödinger operator had before dynamics acted: a field you are ramping, an axis you are rotating by hand. For , , (time-independent case), parity, and most textbook observables the partial term is zero and the commutator is the whole story. The split is the same one classical mechanics makes in .
Why does the oscillator solution look exactly classical while the ISW dynamics looks nothing like a bouncing ball?
Because the Heisenberg equations for the oscillator are linear with c-number coefficients, so operators obey literally the classical equations, and expectations follow suit for every state. The square well's walls are violently nonlinear ( is not linear in ), so and expectation values do not follow classical trajectories — the carpet-weaving interference of Spectral Decomposition of Time Evolution takes over instead. Linearity of the operator equations, not smallness of , is what makes a quantum system "look classical" here; Ehrenfest's Theorem sharpens this.
Is â_H(t) = â e^{−iωt} saying the oscillator loses energy, since â annihilates quanta?
No quanta go anywhere — the number operator is exactly conserved (; the verifier checks to machine precision). The phase is unitary bookkeeping, not decay: is not hermitian, so it is not an observable and its "value" is never measured directly. What the rotating phase encodes is the joint motion of the two quadratures — and swapping into each other each quarter period. Damping requires coupling to an environment, which adds non-Hamiltonian terms well outside this article.