Time Evolution of Phase-Space Functions

The Heisenberg picture — the quantum face of this article's master equation — has become the workhorse of quantum simulation. When experimentalists benchmark a quantum processor or a tensor-network code simulates one, the efficient question is rarely "what is the state?" but "what is the operator?": evolve $\hat A(t)=e^{i\hat Ht/\hbar}\hat A e^{-i\hat Ht/\hbar}$ and watch how far it spreads. Operator-spreading and out-of-time-order-correlator diagnostics, central to 2026-era studies of quantum chaos and information scrambling on superconducting and trapped-ion hardware, are direct descendants of equation (10) — and their classical limits are statements about iterated Poisson brackets exactly like the Lie series (8). On the classical side, the master equation is silently executed trillions of times a day. Every molecular-dynamics engine — the codes behind protein-folding studies and battery-electrolyte screening — reports temperature, pressure, and stress as phase functions of the instantaneous $(q,p)$, propagated by symplectic integrators that are truncated Lie-series exponentials of the bracket. And Wigner–Moyal phase-space methods, which treat quantum dynamics as Poisson evolution plus $O(\hbar^2)$ corrections, are the standard tool for semiclassical chemistry and for reading quantum optics experiments, keeping Poisson's 1809 algebra in daily scientific production.

Picture a phase-space function the way a hiker pictures altitude: $F(q,p)$ paints a landscape of contour lines over the phase plane. The system, meanwhile, is a dot gliding along its trajectory, indifferent to that landscape. The value $F(t)$ you observe is simply the altitude read off along the path — and it changes for exactly one of two reasons. Either the trajectory carries the dot across contour lines of $F$, or the landscape itself is shifting under the dot because $F$ depends explicitly on time. The Poisson bracket $\{F,H\}$ is precisely the first effect: the directional derivative of $F$ along the Hamiltonian velocity field. The partial $\partial F/\partial t$ is the second.

In the figure the orbit is the harmonic-oscillator circle and the landscape belongs to $F=qp$. Where the trajectory runs parallel to a contour, $F$ is momentarily stationary; where it cuts contours steeply, $F$ changes fastest. Every wiggle of the right-hand graph is readable off the left-hand geometry — that is the entire content of $\dot F = \{F,H\}$, drawn rather than derived. Now change the viewpoint from geometry to algebra. Think of $H$ as a generator: just as momentum generates spatial translations and angular momentum generates rotations, the Hamiltonian generates translations in time. The bracket is the machine that applies the generator. Feed it any observable and $H$; it returns the observable's rate of change. Feed it the coordinates themselves and it prints Hamilton's canonical equations. Feed it $H$ and antisymmetry forces the output to vanish — energy conservation without a single line of calculus.

Conservation, in this picture, stops being a differential-equation problem and becomes an algebra problem. If $F$ has no internal clock ($\partial_t F=0$) and the machine returns zero ($\{F,H\}=0$), then $F$ is frozen along every trajectory: its contour surfaces are invariant sheets the motion can never leave. Hunting conserved quantities means hunting functions that bracket to zero with $H$ — pure differentiation, no equations solved. And because the machine is built from the symplectic structure of phase space rather than from any particular Hamiltonian, everything it produces survives the passage to quantum mechanics almost untouched.

Let a system with $n$ degrees of freedom evolve under Hamilton's canonical equations, and let $F(q_1,\dots,q_n,p_1,\dots,p_n,t)$ be any smooth quantity built from the phase-space coordinates — energy, angular momentum, anything. Along an actual trajectory the coordinates become functions of time, so $F$ does too, and its total rate of change compresses into the master evolution equation of Hamiltonian mechanics,

where: $F(q,p,t)$ — any smooth dynamical quantity on phase space; $\partial F/\partial t$ — its explicit (intrinsic) time dependence at a frozen phase point; $\{F,H\}$ — the Poisson bracket of $F$ with the Hamiltonian $H$. Read it as a division of labor: the bracket term is the change because the flow carries the point, the partial term is the change because the function itself ticks. For the overwhelming majority of observables — anything defined once and for all by a formula in $q$ and $p$ — the partial term is zero and the bracket does all the work. The object doing that work deserves its own definition. For any two phase-space functions $A$ and $B$,

where: $q_i,p_i$ — canonically conjugate coordinate–momentum pairs; $n$ — the number of degrees of freedom. The bracket pairs the $q$-slope of one function with the $p$-slope of the other, antisymmetrically — the algebraic shadow of the symplectic area form from the phase-space article. Note what it is not: it contains no dynamics. Equation (2) works between any two functions whatever; dynamics enters only when one slot is filled with $H$. Evaluating (2) on the coordinates themselves gives the multiplication table of the whole formalism, the fundamental brackets

where: $\delta_{ij}$ — the Kronecker delta. These three statements characterize canonical coordinates completely: a change of variables is canonical precisely when it preserves (3). They are also the exact classical ancestors of $[\hat q_i,\hat p_j]=i\hbar\,\delta_{ij}$ — the resemblance is not decorative, and we will make it precise below.

The bracket's power comes from three structural properties, all provable directly from definition (2). Antisymmetry and the Jacobi identity read

where: $A,B,C$ — arbitrary smooth phase-space functions. Antisymmetry is visible by inspection of (2); the Jacobi identity is a half-page of second derivatives cancelling in pairs, painful to verify and priceless to own — it is what makes the set of phase-space functions a Lie algebra under the bracket, the same species of structure as commutators of matrices or of quantum operators. The third property ties the bracket to ordinary multiplication. For any three functions,

where: $BC$ — the pointwise product of functions. This is the Leibniz (product) rule: bracketing with $A$ acts like a derivative. Together, (4) and (5) say each function $A$ defines a differential operator $\{A,\cdot\}$ on phase space — the Hamiltonian vector field of $A$. For $A=H$ that operator is time evolution itself; for $A=p_x$ it generates spatial translations; for $A=L_z$, rotations about the $z$-axis. The article on conservation laws in the Hamiltonian formalism develops this generator picture systematically; here we need only the case $A=H$. With no explicit time dependence, equation (1) collapses to $\dot F=\{F,H\}$, and conservation becomes the algebraic criterion

where: $F$ — a candidate constant of motion with $\partial_t F=0$. This is the practical payoff: to test whether angular momentum survives in your potential you differentiate twice and add — no equations of motion are solved. For the central-force Hamiltonian $H=p_r^2/2m+p_\theta^2/2mr^2+V(r)$, the bracket $\{p_\theta,H\}=-\partial H/\partial\theta$ vanishes because $\theta$ is absent from $H$: angular momentum is conserved, Noether's theorem in one line of algebra. Constants of motion breed. If $F$ and $G$ are both conserved, Poisson's theorem asserts their bracket is conserved too,

where: $F,G$ — two time-independent constants of motion; $\{F,G\}$ — the new candidate. The proof, and the obvious question it raises — why can't this bootstrap infinitely many constants and trivialize every system? — are the business of the next derivation.

Equation (1) gives the rate of change. To get the finite-time change, iterate. If $\dot F=\{F,H\}$, then differentiating again gives $\ddot F=\{\{F,H\},H\}$, and every higher derivative is one more nested bracket. Assembling the Taylor series of $F(t)$ around $t=0$ therefore produces a purely algebraic solution of the dynamics,

where: $\mathcal L_H$ — the Liouvillian operator $\mathcal L_H F=\{F,H\}$; all brackets are evaluated at the initial phase point. This Lie series is time evolution written as an exponential map: the Hamiltonian, applied through the bracket over and over, is the flow. The companion article on Hamiltonian flow develops the same exponential as a map of phase space onto itself; equation (8) is that map read through the eyes of an observable. For Hamiltonians quadratic in $q$ and $p$ the series converges everywhere and resums in closed form — Derivation C does it for the oscillator, and Worked Example 2 tests the truncations numerically. One nearby result deserves careful sign-reading. The phase-space density $\rho(q,p,t)$ of Liouville's theorem evolves by

where: $\rho$ — a probability density carried by the phase fluid. The sign looks flipped relative to (1) — it is a change of viewpoint, not a contradiction. Equation (1) is Lagrangian: it rides one moving phase point. Liouville's theorem states that a density is constant along the ride, $\mathrm{d}\rho/\mathrm{d}t=0$; feeding that into (1) leaves $\partial\rho/\partial t=-\{\rho,H\}$, an Eulerian statement — the density as seen from one fixed spot while the fluid streams past. Same master equation, one extra physical input, opposite-looking sign. The Liouville article builds its whole incompressible-fluid picture on (9).

Everything above was classical. The reason this article carries so much weight in a physics education is that its central equation survives quantization essentially verbatim — and the dictionary between the two worlds is worth displaying before the quantum equation itself.

Dirac's observation of 1925 is that the commutator of quantum operators, divided by $i\hbar$, has every algebraic property of the Poisson bracket: antisymmetry, Leibniz rule, Jacobi identity. Making the substitution $\{A,B\}\to[\hat A,\hat B]/i\hbar$ in the master equation (1) yields the Heisenberg equation of motion,

where: $\hat A,\hat H$ — operators replacing the phase functions $A$ and $H$; $[\hat A,\hat B]=\hat A\hat B-\hat B\hat A$ — the commutator; $\hbar$ — the reduced Planck constant. Every structural statement made in this article transfers: operators commuting with $\hat H$ are conserved, the commutator of two conserved operators is conserved (quantum Poisson theorem), and the finite-time solution exponentiates to $\hat A(t)=e^{i\hat Ht/\hbar}\hat A\,e^{-i\hat Ht/\hbar}$ — the quantum image of the Lie series (8). The fine print matters at second order: the exact quantum bracket on phase space is the Moyal bracket, which equals the Poisson bracket plus corrections of order $\hbar^2$ built from third derivatives. For Hamiltonians quadratic in $q$ and $p$ those corrections vanish identically, which is why the quantum harmonic oscillator's Heisenberg operators obey exactly the classical equations of Derivation C.

1 · The bracket versus the stopwatch (confidence builder)

A harmonic oscillator has $m=0.5\ \mathrm{kg}$, $\omega=2.0\ \mathrm{rad/s}$, released from $q_0=1.0\ \mathrm{m}$, $p_0=0$. For the phase function $F=qp$ (the one-particle virial), evaluate $\mathrm{d}F/\mathrm{d}t$ at $t=0.4\ \mathrm{s}$ twice — once from the Poisson bracket, once by differentiating the explicit solution — and check they agree.

Step 1. Compute the bracket once and for all. With $H=p^2/2m+\tfrac12 m\omega^2q^2$ and definition (2): $\{qp,H\}=\dfrac{\partial(qp)}{\partial q}\dfrac{\partial H}{\partial p}-\dfrac{\partial(qp)}{\partial p}\dfrac{\partial H}{\partial q}=p\cdot\dfrac{p}{m}-q\cdot m\omega^2 q=\dfrac{p^2}{m}-m\omega^2q^2$. This is twice the kinetic minus twice the potential energy — the virial combination, whose vanishing time-average is the virial theorem.

Step 2. Locate the phase point at $t=0.4\ \mathrm{s}$. The solution is $q(t)=q_0\cos\omega t$, $p(t)=-m\omega q_0\sin\omega t$, so $q=\cos(0.8)=0.6967\ \mathrm{m}$ and $p=-(0.5)(2.0)(1.0)\sin(0.8)=-0.7174\ \mathrm{kg\,m/s}$.

Step 3. Evaluate the bracket there. $p^2/m=(-0.7174)^2/0.5=1.0292\ \mathrm{J}$ and $m\omega^2q^2=(0.5)(4.0)(0.6967)^2=0.9708\ \mathrm{J}$, so $\mathrm{d}F/\mathrm{d}t=\{F,H\}=1.0292-0.9708=\boxed{+0.0584\ \mathrm{J}}$ (units: $F$ is an action, J·s, so its rate is joules).

Step 4. Now the stopwatch route. Along the solution, $F(t)=q(t)p(t)=-m\omega q_0^2\cos\omega t\sin\omega t=-\tfrac12 m\omega q_0^2\sin 2\omega t=-\tfrac12\sin(4t)\ \mathrm{J\,s}$. Differentiating directly: $\mathrm{d}F/\mathrm{d}t=-2\cos(4t)$, and at $t=0.4$: $-2\cos(1.6)=-2(-0.02920)=\boxed{+0.0584\ \mathrm{J}}$. The two routes agree to every computed digit ($0.058399$ both ways) — the bracket knows the derivative without ever seeing the solution.

Step 5. Read the physics. $F=qp$ oscillates at $2\omega$ (twice per orbit, as the hero figure shows: the ellipse crosses each hyperbolic contour of $qp$ four times per cycle), and $\langle\{qp,H\}\rangle=0$ over a period forces $\langle p^2/m\rangle=\langle m\omega^2q^2\rangle$ — equipartition of the oscillator's energy, extracted from one bracket.

2 · Evolving q with brackets alone (exam level)

An oscillator has $m=1.0\ \mathrm{kg}$, $\omega=2.0\ \mathrm{rad/s}$, initial data $q_0=0.30\ \mathrm{m}$, $p_0=0.80\ \mathrm{kg\,m/s}$. Using the Lie series (8) for $F=q$, compute $q$ at $t=0.25\ \mathrm{s}$ (so $\omega t=0.5$) keeping (a) two terms, (b) four terms, (c) six terms, and compare with the exact resummed answer.

Step 1. Assemble the iterated brackets from Derivation C: $\mathcal L_Hq=p/m$, $\mathcal L_H^2q=-\omega^2q$, $\mathcal L_H^3q=-\omega^2p/m$, $\mathcal L_H^4q=\omega^4q$, $\mathcal L_H^5q=\omega^4p/m$. At $t=0$ these evaluate on the initial data $q_0=0.30$, $p_0/m=0.80\ \mathrm{m/s}$.

Step 2. Two terms (through order $t$): $q\approx q_0+t\,p_0/m=0.30+(0.25)(0.80)=\boxed{0.5000\ \mathrm{m}}$. This is straight-line inertial extrapolation — the tangent to the true motion.

Step 3. Four terms (through $t^3$): add $-\omega^2q_0t^2/2=-\,(4)(0.30)(0.0625)/2=-0.0375$ and $-\omega^2(p_0/m)t^3/6=-(4)(0.80)(0.015625)/6=-0.008333$, giving $q\approx0.5000-0.0375-0.008333=\boxed{0.4542\ \mathrm{m}}$.

Step 4. Six terms (through $t^5$): add $+\omega^4q_0t^4/24=(16)(0.30)(0.00390625)/24=+0.000781$ and $+\omega^4(p_0/m)t^5/120=(16)(0.80)(0.0009765625)/120=+0.000104$, giving $q\approx0.4542+0.000885=\boxed{0.45505\ \mathrm{m}}$.

Step 5. Exact answer from the resummed series: $q=q_0\cos(0.5)+\dfrac{p_0}{m\omega}\sin(0.5)=(0.30)(0.877583)+(0.40)(0.479426)=0.263275+0.191770=\boxed{0.455045\ \mathrm{m}}$.

Step 6. Convergence scorecard. Two terms: $0.5000$, error $+0.0450\ \mathrm{m}$ ($+9.9\%$). Four terms: $0.4542$, error $-0.00088\ \mathrm{m}$ ($-0.19\%$). Six terms: $0.45505$, error $+7\times10^{-6}\ \mathrm{m}$ ($+0.0016\%$). Each pair of bracket orders buys roughly a factor of fifty at $\omega t=0.5$ — $\cos\omega t$ and $\sin\omega t$ built term by term out of pure algebra, which is exactly how a symplectic integrator's truncated Lie series propagates a molecular-dynamics step.

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §9.5–9.6 (Poisson brackets) · Landau & Lifshitz, Mechanics, 3rd ed., §42 (Poisson brackets) · Morin, Introduction to Classical Mechanics, Ch. 15 (Hamiltonian methods) · Taylor, Classical Mechanics, Ch. 13 · Arnold, Mathematical Methods of Classical Mechanics, §40

Provenance. Master formula and Poisson-bracket definition verified against Goldstein §9.5–9.6 and Landau-Lifshitz §42. A verifier should re-derive $\mathrm{d}F/\mathrm{d}t=\partial_t F+\{F,H\}$ via the chain rule plus Hamilton's equations, confirm $\{H,H\}=0$ from antisymmetry, recompute the harmonic-oscillator brackets ($\dot q=p/m$, $\dot p=-m\omega^2 q$, $\ddot q=-\omega^2 q$), and check the canonical brackets $\{q_i,p_j\}=\delta_{ij}$ and quantization rule $[\hat q,\hat p]=i\hbar$ against any standard quantum text. Jacobi identity and Poisson's theorem cross-checked with Arnold §40.

Builds on: Hamilton's Canonical Equations · The Hamiltonian Function · Phase Space · Generalized Momenta Leads to: Hamiltonian Flow and the Phase Fluid · Fixed Points and Stability in Phase Space · The Harmonic Oscillator in Phase Space