Poisson brackets, and the structure quantum mechanics inherits

Derivation · the atom

Poisson brackets, and the structure quantum mechanics inherits

D-005 Home PU-201 Threads symmetry Depends on D-004 verified

Statement

The time evolution of any phase-space function is given by a single bracket operation with the Hamiltonian, and that operation has the same algebraic structure as the quantum commutator.

Why it matters

Once evolution is written as a bracket with H, conservation laws become a computation rather than an insight: a quantity with no explicit time dependence is conserved exactly when its bracket with H vanishes. That is already worth the formalism. The deeper payoff is structural — the fundamental brackets of classical mechanics map term-for-term onto the quantum commutators of PU-202, which is what makes canonical quantisation a motivated procedure rather than a blind guess.

Assumptions

Hamilton's equations hold.The bracket formulation is equivalent to them, not independent of them — everything below rests on D-004. Coordinates are canonical.{q_i, p_j} = δ_ij. Under a non-canonical change of variables the bracket changes form and the elegance is lost.

Derivation

{f, g} ≡ Σ_i ( ∂f/∂q_i · ∂g/∂p_i − ∂f/∂p_i · ∂g/∂q_i )

Definition. Note the antisymmetry: {f,g} = −{g,f}. It is inherited directly from the antisymmetry of Hamilton's equations. A

df/dt = Σ_i ( ∂f/∂q_i · q̇_i + ∂f/∂p_i · ṗ_i ) + ∂f/∂t

Chain rule for a function of phase-space coordinates that may also depend explicitly on time. A

df/dt = Σ_i ( ∂f/∂q_i · ∂H/∂p_i − ∂f/∂p_i · ∂H/∂q_i ) + ∂f/∂t

Substitute Hamilton's equations from D-004 (q̇_i = ∂H/∂p_i, ṗ_i = −∂H/∂q_i). The two terms now match the bracket definition exactly, with g = H. B

{q_i, q_j} = 0, {p_i, p_j} = 0, {q_i, p_j} = δ_ij

Taking f = q_i and f = p_i recovers Hamilton's equations, so nothing was lost. These fundamental brackets encode the entire canonical structure. C

[x̂_i, x̂_j] = 0, [p̂_i, p̂_j] = 0, [x̂_i, p̂_j] = iħδ_ij

The quantum commutators of PU-202 have the identical form under {·,·} → [·,·]/iħ. Heisenberg's equation dÂ/dt = [Â,Ĥ]/iħ + ∂Â/∂t is the boxed result with that single substitution. Not a coincidence and not a proof — canonical quantisation is a procedure, motivated by this match, justified because it works. C

Result

df/dt = {f, H} + ∂f/∂t

Reading. The Hamiltonian generates time translations. Any quantity with no explicit time dependence is conserved if and only if its bracket with H vanishes. Conservation laws are now a computation, not an insight — which is what makes the formalism worth learning.

Units check. With canonical coordinates, ∂H/∂p_i = q̇_i carries units of [q]/[t] and ∂f/∂q_i carries [f]/[q], so each bracket term is [f]/[t] — matching df/dt and ∂f/∂t. The fundamental bracket {q,p} = δ_ij is dimensionless, so its quantum image iħδ_ij correctly carries the units of action.

Limiting cases

Coordinates themselves. f = q_i gives q̇_i = {q_i, H} = ∂H/∂p_i, and f = p_i gives ṗ_i = {p_i, H} = −∂H/∂q_i. The result collapses back to Hamilton's equations — a consistency check that nothing was added or lost.
Conserved quantity. If ∂f/∂t = 0 and {f, H} = 0, then f is a constant of motion; in particular {H, H} = 0 gives energy conservation for a time-independent Hamiltonian.
Classical limit of the quantum law. As ħ → 0 the commutator image [·,·]/iħ reduces to {·,·}, and Heisenberg's equation returns to df/dt = {f, H} + ∂f/∂t.

Breaks when

The correspondence is pushed too far. Groenewold's theorem shows no consistent map exists from all classical observables to quantum operators preserving brackets exactly. Ordering ambiguities are real and unavoidable.
The coordinates are not canonical — the bracket must then be written with the general symplectic form.

Failure modes

Reading the correspondence as a derivation of quantum mechanics. It is an analogy strong enough to guide a guess and no stronger.
Sign errors from the antisymmetry. Fix a convention for {q,p} once and enforce it site-wide; this is a schema-level decision, not a per-page one.

Worked number

Take the harmonic oscillator H = p²/2m + ½mω²q² with m = 1 kg, ω = 2 rad/s, at the instant q = 0.5 m, p = 3 kg·m/s.

q̇ = {q, H} = ∂H/∂p = p/m = 3 m/s

ṗ = {p, H} = −∂H/∂q = −mω²q = −(1)(4)(0.5) = −2 N

Combining gives q̈ = −ω²q, the correct oscillator equation, and {H, H} = 0 confirms energy is conserved. Run the check →

Discussion

The boxed result df/dt = {f, H} + ∂f/∂t says something sharper than "the Hamiltonian gives the energy." It says the Hamiltonian is the generator of the flow of time through phase space: to advance any observable by an instant, bracket it with H. Nothing about the observable matters except how it varies across phase space, encoded in its gradients ∂f/∂q and ∂f/∂p. Every dynamical question — what moves, what is conserved, how fast — collapses to a single bilinear operation, and the physics has been recoded as algebra.

The particular shape of that operation is not decorative. The bracket is bilinear, antisymmetric ({f,g} = −{g,f}), obeys the Leibniz product rule {f,gh} = g{f,h} + {f,g}h, and satisfies the Jacobi identity {f,{g,h}} + {g,{h,f}} + {h,{f,g}} = 0. The first three make the smooth functions on phase space a Lie algebra that also respects multiplication — a Poisson algebra. The Jacobi identity is the load-bearing one: it guarantees that bracketing-with-H behaves as a derivation, so that the bracket of two conserved quantities is itself conserved (Poisson's theorem). This is why conservation laws breed more conservation laws rather than existing in isolation.

Read one level deeper, the antisymmetric second-derivative structure is a coordinate report of an object that does not need coordinates: the symplectic 2-form ω = ∑_i dp_i ∧ dq_i. The bracket {f,g} is just ω evaluated on the Hamiltonian vector fields of f and g, which is exactly why the assumption "coordinates are canonical" matters — a non-canonical change of variables leaves ω untouched but scrambles its component expression, so the tidy formula in step 1 is a statement about a chosen chart, while the geometry underneath is invariant. Every conserved quantity is then the generator of a symplectomorphism (a canonical transformation): H generates time translation, linear momentum generates spatial translation, angular momentum generates rotation. Noether's theorem, viewed from here, is the observation that {Q,H} = 0 is symmetric in its two slots — if Q is unchanged by the flow of H, then H is unchanged by the flow of Q.

The correspondence {·,·} → [·,·]/iℏ is the reason this page sits upstream of quantum mechanics, but the match is a structural one, not an identity. Both algebras are Lie algebras with the same fundamental relations; canonical quantisation guesses that the quantum bracket should be the classical one, rescaled by iℏ. Groenewold and van Hove proved this guess cannot be completed consistently: there is no map from all classical observables to operators that preserves every bracket, because the classical algebra is commutative under ordinary multiplication while the quantum one is not, and the two constraints collide at cubic order and above. The modern resolution — deformation quantisation — keeps the Poisson bracket as the leading term in an ℏ-expansion of the Moyal bracket, so the classical structure is literally the ℏ → 0 shadow of the quantum one.

Common misconceptions. (i) "{f,H} = 0 means f is constant." It means f is conserved only if f has no explicit time dependence; the ∂f/∂t term is not optional (see the free-particle boost generator in the problems). (ii) The correspondence "derives" quantum mechanics — it does not; it motivates a procedure whose warrant is experimental. (iii) Sign carelessness: the whole formalism rests on a fixed convention for {q,p} = +1; flip it and every equation of motion flips, silently.

Worked examples

Example 1 — The oscillator, run entirely through brackets. A mass on a spring has H = p²/2m + ½mω²q² with m = 0.20 kg and spring constant k = mω² = 8.0 N/m. Find the equations of motion, confirm energy is conserved, and evaluate the rate of change of the virial G = qp at the instant q = 0.050 m, p = 0.30 kg·m/s.

dq/dt = {q, H} = ∂H/∂p = p/m

Only the ∂q/∂q = 1 term survives in the bracket, picking out ∂H/∂p.

dp/dt = {p, H} = −∂H/∂q = −mω²q = −kq

Now only the ∂p/∂p = 1 term survives, with its leading minus sign. These two lines are Hamilton's equations, recovered.

dH/dt = {H, H} + ∂H/∂t = 0 + 0 = 0

Antisymmetry forces {H,H} = 0; the potential is static so ∂H/∂t = 0. Energy is conserved — a one-line proof.

dG/dt = {qp, H} = q{p,H} + p{q,H} = p²/m − kq² = 2T − 2V

Leibniz rule, then substitute steps 1–2. This is the virial relation in bracket form: the instantaneous imbalance of kinetic and potential energy.

dG/dt = (0.30)²/0.20 − (8.0)(0.050)² = 0.45 − 0.020 = 0.43 J

Numbers in only at the end. (Check: T = p²/2m = 0.225 J, V = ½kq² = 0.010 J, so 2T − 2V = 0.430 J.)

dq/dt = p/m = 1.5 m/s, dp/dt = −kq = −0.40 N, dG/dt = 0.43 J

Answer. Bracketing with H reproduces both equations of motion, proves dH/dt = 0 without integration, and gives the virial rate 0.43 J at the stated point. The energy itself is E = T + V = 0.235 J, fixed for all time.

Example 2 — Angular momentum, and why it is conserved in a central field. For a particle in three dimensions, verify the fundamental angular-momentum bracket {L_x, L_y} = L_z, then show L_z is conserved for any central Hamiltonian H = p²/2m + V(r), and evaluate L_z for r = (0.30, 0.10, 0) m, p = (0, 0.50, 0) kg·m/s.

L_x = y p_z − z p_y, L_y = z p_x − x p_z

Components of L = r × p. Each depends on two coordinates and two momenta, so the bracket has few surviving terms.

{L_x, L_y} = ∑_i ( ∂L_x/∂q_i · ∂L_y/∂p_i − ∂L_x/∂p_i · ∂L_y/∂q_i )

Write out the definition summing over i = x, y, z.

only the i = z term is nonzero: (−p_y)(−x) − (y)(p_x) = x p_y − y p_x = L_z

The x and y slots cancel; the z slot uses ∂L_x/∂z = −p_y, ∂L_y/∂p_z = −x, etc. The so(3) Lie algebra is exact.

{L_z, r²} = −∑_i ∂L_z/∂p_i · ∂r²/∂q_i = −[ (−y)(2x) + (x)(2y) ] = 0

With r² = x²+y²+z². By the same cancellation {L_z, p²} = 0, so L_z brackets to zero with any function of r and of p².

dL_z/dt = {L_z, H} = {L_z, p²/2m} + {L_z, V(r)} = 0

Rotational invariance of H about the z-axis, now a computation rather than a picture.

L_z = x p_y − y p_x = (0.30)(0.50) − (0.10)(0) = 0.15 kg·m²/s

Evaluate the conserved value at the given phase point.

{L_x, L_y} = L_z, {L_z, H} = 0, L_z = 0.15 kg·m²/s

Answer. The components of angular momentum close into the rotation algebra, and any central Hamiltonian commutes (in the Poisson sense) with each component about its axis, so L_z = 0.15 kg·m²/s is a constant of the motion. This is the classical parent of the quantum relation [L̂_x, L̂_y] = iℏL̂_z.

Problems

Fundamental brackets, warm-up. Using the definition and {q,p} = 1, evaluate (a) {q, p²}, (b) {q³, p}, (c) {q², p²}.
Solution

With one degree of freedom {f,g} = ∂f/∂q · ∂g/∂p − ∂f/∂p · ∂g/∂q.

(a) {q, p²} = (1)(2p) − (0)(0) = 2p.

(b) {q³, p} = (3q²)(1) − (0)(0) = 3q².

(c) {q², p²} = (2q)(2p) − (0)(0) = 4qp. Note {qⁿ, p} = n qⁿ⁻¹: the bracket with p acts as ∂/∂q, foreshadowing p̂ = −iℏ∂/∂x.
Complex amplitude of the oscillator. For H = p²/2m + ½mω²q², define α = p − imωq. Show dα/dt = −iωα, hence α(t) = α₀e^−iωt, and identify the conserved quantity |α|². Evaluate it for m = 0.20 kg, ω² = 40 s⁻², q = 0.050 m, p = 0.30 kg·m/s.
Solution

Because the bracket is bilinear, dα/dt = {α,H} = {p,H} − imω{q,H}. From Example 1, {p,H} = −mω²q and {q,H} = p/m, so

dα/dt = −mω²q − iωp = −iω(p − imωq) = −iωα.

This linear equation integrates to α(t) = α₀e^−iωt. Its modulus is constant: |α|² = p² + m²ω²q² = 2mH, i.e. the energy up to the factor 2m. Numerically |α|² = (0.30)² + (0.20)²(40)(0.050)² = 0.090 + 0.0040 = 0.094 (kg·m/s)², giving E = |α|²/2m = 0.094/0.40 = 0.235 J, matching Example 1. The single complex number α carries both amplitude and phase; α → â is the ladder operator.
A conserved quantity that depends on time. A free particle has H = p²/2m. Show that G = q − (p/m)t satisfies dG/dt = 0 even though {G,H} ≠ 0. Interpret G. If m = 0.50 kg and p = 2.0 kg·m/s with the particle at q = 3.0 m when t = 1.0 s, give G.
Solution

Here G depends explicitly on time, so both terms of the master equation are needed:

{G,H} = {q,H} − (t/m){p,H} = p/m − 0 = p/m, and ∂G/∂t = −p/m.

Therefore dG/dt = {G,H} + ∂G/∂t = p/m − p/m = 0. So G is conserved. Physically G = q(t) − v t is the initial position q₀; it is the generator of Galilean boosts. This is the standard trap the master equation resolves: {G,H} alone is nonzero, yet the full time derivative vanishes. Numerically G = 3.0 − (2.0/0.50)(1.0) = 3.0 − 4.0 = −1.0 m (the position at t = 0).
Closing the rotation algebra. Show {L_z, L_x} = L_y, confirming the cyclic pattern {L_i, L_j} = ε_ijkL_k. What does this tell you about measuring two components of angular momentum at once, once the correspondence to commutators is applied?
Solution

L_z = x p_y − y p_x, L_x = y p_z − z p_y. The two share only the coordinate y and the momentum p_y, so just the i = y term of the sum survives. The needed derivatives are ∂L_z/∂y = −p_x, ∂L_x/∂p_y = −z, ∂L_z/∂p_y = 0, ∂L_x/∂y = p_z. Hence

{L_z, L_x} = (−p_x)(−z) − (0)(p_z) = z p_x. Adding the symmetric p_x-slot contribution −x p_z from ∂L_x/∂p_x and ∂L_z/∂x gives z p_x − x p_z = L_y.

So the brackets are cyclic, {L_i, L_j} = ε_ijkL_k — the Lie algebra so(3). Under {·,·} → [·,·]/iℏ this becomes [L̂_i, L̂_j] = iℏε_ijkL̂_k ≠ 0: the components do not commute, so no two Cartesian components of angular momentum have simultaneous eigenstates and cannot be sharply measured together. Only |L|² and one component can.
Poisson's theorem (hard). Assume f and g are conserved and have no explicit time dependence, so {f,H} = {g,H} = 0. Prove that {f,g} is also conserved, and explain why this makes the conservation of L_z automatic once L_x and L_y are known to be conserved. Where could the argument still fail to produce a new conservation law?
Solution

The time derivative of a bracket is d{f,g}/dt = {{f,g}, H} (no explicit time dependence anywhere). The Jacobi identity, rearranged, says that bracketing-with-H is a derivation over the bracket:

{{f,g}, H} = {{f,H}, g} + {f, {g,H}}.

Both right-hand terms contain {f,H} = 0 or {g,H} = 0, so d{f,g}/dt = 0. Hence {f,g} is conserved — conserved quantities form a closed Lie subalgebra. Applied to angular momentum: if L_x and L_y are conserved (central field), then {L_x,L_y} = L_z must be conserved too, with no extra calculation.

The catch: the theorem is not a conservation-law factory. {f,g} may turn out to be a constant, a function of f and g already known, or identically zero — in which case nothing new is learned. It generates a genuinely independent integral of motion only when {f,g} is functionally independent of the quantities in hand, and a system with more than n independent commuting integrals would be over-determined (this is the ceiling that makes Liouville integrability — exactly n integrals in involution — special).