Poisson brackets — definition, properties, connection to commutators: Worked examples

Before anything, one reminder of the only formula we need, so no symbol is unearned:

We will lean on two facts throughout, so let us frame them upfront as named theorems before any example uses them.

One more piece of notation is used only in Cell F, so we define it right where it is earned:

The scenario matrix

The grid below is drawn as Figure 1 — a visual map with the nine cells A–I laid out as coloured tiles, so you can see the whole territory at a glance before diving in. The table restates the same nine cells in words. Every worked example is tagged with the cell it fills; together they cover the entire grid.

Figure 1 — The scenario matrix (nine cells A–I). deepdives/dd-physics-2.1.15-d3-s01.png

What the figure shows. Nine rounded tiles sit in a 3×3 grid. The top row (mint) holds the elementary cases: A ${q, p}$ , B ${q, H}$ equations of motion, C degenerate/zero. The middle row (butter) holds the structural cases that lean on a property arrow: D sign-swap, E product rule, F vector components. The bottom row (lavender) holds the advanced cases: G explicit time, H a real-world word problem, I the Jacobi exam twist. A small coral legend strip along the bottom names the three difficulty bands. Use it as your checklist — each tile lights up as its example is worked.

Cell	Case class	What can trip you	Example
A	Both are simple coordinates	order & sign of ${q, p}$	Ex 1
B	One is a coordinate, one is $H$	recovering equations of motion	Ex 2
C	Zero / degenerate: bracket with itself or a constant	"is it just $0$ ?"	Ex 3
D	Sign trap: swap the order	antisymmetry ${f, g} = - {g, f}$	Ex 4
E	Product of quantities	Leibniz product rule	Ex 5
F	Vector components (all cyclic 3D indices)	angular-momentum $ϵ_{ij k}$	Ex 6
G	Explicit time dependence (limiting/boundary in $t$ )	the $\partial f / \partial t$ term	Ex 7
H	Real-world word problem	translating physics into a bracket	Ex 8
I	Exam twist: manufacture a new conservation law	Jacobi identity / Poisson's theorem	Ex 9

We use one degree of freedom ( $n = 1$ , variables $q, p$ ) unless the cell needs 3D, in which case the coordinates are $x, y, z$ with partner momenta $p_{x}, p_{y}, p_{z}$ , and the shortcut rules ${x, p_{x}} = 1$ , ${x, p_{y}} = 0$ , and so on.

Cell A — the simplest bracket

Worked example Ex 1 ·

${q, p}$ (Cell A) Compute ${q, p}$ for one degree of freedom.

Forecast: guess the number before reading. Is it $0$ , $1$ , or $- 1$ ?

Write $f = q$ , $g = p$ . Plug into the recipe: ${q, p} = \frac{\partial q}{\partial q} \frac{\partial p}{\partial p} - \frac{\partial q}{\partial p} \frac{\partial p}{\partial q}$ . Why this step? The definition is the only starting point; everything is just careful bookkeeping of four partial derivatives.
Evaluate each slope. $\frac{\partial q}{\partial q} = 1$ ( $q$ changes fully with itself), $\frac{\partial p}{\partial p} = 1$ , $\frac{\partial q}{\partial p} = 0$ ( $q$ doesn't care about $p$ ), $\frac{\partial p}{\partial q} = 0$ . Why this step? $q$ and $p$ are independent axes of phase space; nudging one never moves the other, so cross-slopes vanish.
Combine: ${q, p} = (1) (1) - (0) (0) = 1$ . Why this step? We substitute the four evaluated slopes back into the recipe and do the arithmetic — the first product gives $1$ , the second product is $0$ , and " $1$ minus $0$ " is the final number. Nothing else can contribute, so the bracket is fully determined.

Verify: This is the canonical bracket. It is the classical seed of $[\overset{q}{^}, \overset{p}{^}] = i ℏ$ (multiply by $i ℏ$ ). Units: $q$ has length, $p$ has momentum, so ${q, p}$ has units of $\frac{length \cdot momentum}{length \cdot momentum} = 1$ , dimensionless — consistent with it equalling the pure number $1$ . ✓

Cell B — bracket with the Hamiltonian recovers motion

Worked example Ex 2 ·

${q, H}$ and ${p, H}$ for a harmonic oscillator (Cell B) Take $H = \frac{p ^{2}}{2 m} + \frac{1}{2} m ω^{2} q^{2}$ . Find $\overset{q}{˙}$ and $\overset{p}{˙}$ .

Forecast: you already know Newton's answer for a spring — momentum drives position, and position pulls momentum back. Guess the two brackets before computing.

$\overset{q}{˙} = {q, H} = \frac{\partial q}{\partial q} \frac{\partial H}{\partial p} - \frac{\partial q}{\partial p} \frac{\partial H}{\partial q} = 1 \cdot \frac{\partial H}{\partial p} - 0 = \frac{p}{m}$ . Why this step? By Theorem T1 with $f = q$ (no explicit time), $\overset{q}{˙} = {q, H}$ . Only the first term survives because $\partial q / \partial p = 0$ ; then $\partial H / \partial p = \partial (p^{2} /2 m) / \partial p = p / m$ .
$\overset{p}{˙} = {p, H} = \frac{\partial p}{\partial q} \frac{\partial H}{\partial p} - \frac{\partial p}{\partial p} \frac{\partial H}{\partial q} = 0 - 1 \cdot \frac{\partial H}{\partial q} = - m ω^{2} q$ . Why this step? Now only the second term survives ( $\partial p / \partial q = 0$ ), and $\partial H / \partial q = m ω^{2} q$ .
Read the physics: $\overset{q}{˙} = p / m$ is "velocity = momentum over mass," and $\overset{p}{˙} = - m ω^{2} q$ is "force = $- k q$ " with $k = m ω^{2}$ . These are Hamilton's equations.

Verify: Differentiate $\overset{q}{˙} = p / m$ : $\overset{q}{¨} = \overset{p}{˙} / m = - ω^{2} q$ — the simple-harmonic equation, exactly what a spring gives. ✓ Units of $\overset{p}{˙}$ : $m ω^{2} q$ has $kg \cdot s^{- 2} \cdot m = N$ , a force. ✓

Cell C — the degenerate / zero cases

Worked example Ex 3 ·

${f, f}$ and ${f, c}$ (Cell C) Compute the bracket of any quantity with itself, and with a constant $c$ (say $c = 7$ ).

Forecast: one of these might not be zero — which do you trust?

${f, f} = \sum_{i} (\frac{\partial f}{\partial q _{i}} \frac{\partial f}{\partial p _{i}} - \frac{\partial f}{\partial p _{i}} \frac{\partial f}{\partial q _{i}})$ . The two terms are literally identical, so they cancel: ${f, f} = 0$ . Why this step? Multiplication commutes ( $ab = ba$ ), so the " $q p$ " and " $pq$ " pieces are the same number subtracted from itself.
${f, c}$ : every partial derivative of a constant is zero, $\frac{\partial c}{\partial q _{i}} = \frac{\partial c}{\partial p _{i}} = 0$ . Both terms vanish: ${f, c} = 0$ . Why this step? A constant has no slope anywhere — it's a flat sheet over phase space.

Verify: Sanity via antisymmetry: ${f, f} = - {f, f}$ forces $2 {f, f} = 0$ , so ${f, f} = 0$ independently. ✓ And ${f, c} = 0$ says a constant is trivially conserved — obviously true, since $\overset{c}{˙} = {c, H} = 0$ . ✓

Cell D — the sign trap

Worked example Ex 4 ·

${p, q}$ vs ${q, p}$ (Cell D) Compute ${p, q}$ and compare with Ex 1.

Forecast: many students say ${p, q} = 1$ "because brackets are symmetric like the dot product." Is that right?

${p, q} = \frac{\partial p}{\partial q} \frac{\partial q}{\partial p} - \frac{\partial p}{\partial p} \frac{\partial q}{\partial q} = 0 - 1 \cdot 1 = - 1$ . Why this step? We just apply the recipe with $f = p, g = q$ : the first product uses cross-slopes ( $0$ ), the second uses self-slopes ( $1 \cdot 1$ ), and " $0$ minus $1$ " is $- 1$ .
So ${p, q} = - {q, p} = - 1$ . Why this step? We invoke the antisymmetry arrow from the property wheel (Figure 2): swapping $f$ and $g$ literally exchanges the two terms of the recipe, so the whole bracket flips sign. This gives a second, independent route to $- 1$ — and confirms ${q, p} = 1$ from Ex 1 must have the opposite sign of ${p, q}$ .

Verify: This is exactly antisymmetry: ${f, g} = - {g, f}$ . The bracket is not like a dot product (which is symmetric); it behaves like a cross product — order matters, and swapping flips sign. ✓

The properties we keep reaching for are collected in Figure 2, the property wheel:

Figure 2 — The Poisson-bracket property wheel. deepdives/dd-physics-2.1.15-d3-s02.png

What the figure shows. A soft central hub labelled ${f, g}$ — the bracket itself. Six coloured arrows fan out, each to one property the bracket obeys: a coral arrow to Antisymmetry ${f, g} = - {g, f}$ (swap flips sign, used in Ex 4); a mint arrow to Bilinearity ${a f + b g, h} = a {f, h} + b {g, h}$ (split sums, pull out constants, used in Ex 8); a butter arrow to Leibniz ${f g, h} = f {g, h} + {f, h} g$ (products split like ordinary derivatives, used in Ex 5); a lavender arrow to Jacobi (the cyclic sum is zero, used in Ex 9); a second mint arrow to Constants ${f, c} = 0$ (Ex 3); and a lavender arrow to Canonical ${q, p} = 1$ (Ex 1). Every calculation below is just "which arrow am I using right now?"

Cell E — the product rule

Cell F — vector components, all index cases

Worked example Ex 6 · The angular-momentum algebra

${L_{i}, L_{j}}$ (Cell F) With $L_{x} = y p_{z} - z p_{y}$ , $L_{y} = z p_{x} - x p_{z}$ , $L_{z} = x p_{y} - y p_{x}$ , compute ${L_{x}, L_{y}}$ .

Forecast: the answer is one of the three components. Which one, and with what sign?

The cyclic sign rule is drawn in Figure 3, the angular-momentum ring:

Figure 3 — The cyclic ring for ${L_{i}, L_{j}}$ . deepdives/dd-physics-2.1.15-d3-s03.png

What the figure shows. Three pastel circles at the corners of a triangle: $L_{x}$ (lavender, top), $L_{y}$ (mint, lower-left), $L_{z}$ (butter, lower-right). Curved lavender arrows sweep the ring in the positive cyclic direction $x \to y \to z \to x$ : bracket two components in that forward order and the answer is the next component with a $+$ sign, so ${L_{x}, L_{y}} = + L_{z}$ . A dashed coral arrow runs backward, $y \to x$ , reminding you that reading against the arrows (anti-cyclic) picks up a minus sign, e.g. ${L_{y}, L_{x}} = - L_{z}$ . That is the $ϵ_{ij k}$ rule made visual: forward $= + 1$ , backward $= - 1$ .

Expand using bilinearity: ${L_{x}, L_{y}} = {y p_{z} - z p_{y}, z p_{x} - x p_{z}}$ splits into four brackets. Only pairs sharing a canonical partner survive. Why this step? A bracket like ${y p_{z}, z p_{x}}$ is nonzero only when a coordinate meets its own momentum, because ${x, p_{x}} = 1$ but ${x, p_{y}} = 0$ .
The surviving contributions come from the $z$ and $p_{z}$ terms. Carefully: ${y p_{z}, z p_{x}} = y p_{x} {p_{z}, z} = y p_{x} (- 1) = - y p_{x}$ , and ${- z p_{y}, - x p_{z}} = x p_{y} {z, p_{z}} = x p_{y} (1) = x p_{y}$ ; the other two of the four brackets ( ${y p_{z}, - x p_{z}}$ and ${- z p_{y}, z p_{x}}$ ) have no matching coordinate–momentum pair and vanish. Why this step? Treat spectator factors as constants (Leibniz), pulling out $y p_{x}$ and $x p_{y}$ , leaving the elementary brackets ${p_{z}, z} = - 1$ and ${z, p_{z}} = 1$ .
Sum the two survivors: ${L_{x}, L_{y}} = - y p_{x} + x p_{y} = x p_{y} - y p_{x} = L_{z}$ . Why this step? We add the two nonzero pieces from step 2 and reorder. The combination $- y p_{x} + x p_{y}$ is, letter-for-letter, the definition of $L_{z} = x p_{y} - y p_{x}$ — so the messy four-term expansion collapses into a single named quantity with no leftover terms. The final answer is therefore exactly $L_{z}$ , positive sign, matching the forward direction on the ring in Figure 3.

Verify: This is the $ϵ_{ij k}$ rule ${L_{i}, L_{j}} = ϵ_{ij k} L_{k}$ with $(i, j, k) = (x, y, z)$ : since $(x, y, z)$ is cyclic, $ϵ_{x y z} = + 1$ , so ${L_{x}, L_{y}} = + L_{z}$ . ✓ Quantizing gives $[\hat{L}_{x}, \hat{L}_{y}] = i ℏ \hat{L}_{z}$ . ✓

Cell G — explicit time dependence (boundary case)

Worked example Ex 7 · A quantity carrying its own clock (Cell G)

A particle feels constant force $F$ , so $H = \frac{p ^{2}}{2 m} - F q$ . Show the quantity $f = p - F t$ is conserved.

Forecast: $p$ alone grows in time (the force pushes it). Can subtracting $F t$ freeze it?

Full law (Theorem T1): $\frac{df}{d t} = {f, H} + \frac{\partial f}{\partial t}$ . We must NOT drop the last term because $f$ mentions $t$ explicitly. Why this step? The bracket only sees how $f$ rides the phase-space flow; an explicit $t$ inside $f$ is extra motion the bracket cannot know about, so T1's second term is exactly what accounts for it.
Bracket part: ${p - F t, H} = {p, H}$ (the $- F t$ piece is a constant in phase space, bracket $0$ ). And ${p, H} = - \frac{\partial H}{\partial q} = - (- F) = F$ . Why this step? ${p, H} = - \partial H / \partial q$ from the recipe; here $\partial H / \partial q = - F$ .
Explicit term: $\frac{\partial f}{\partial t} = \frac{\partial ( p - F t )}{\partial t} = - F$ . Why this step? Freeze $q, p$ and differentiate $f$ in $t$ only.
Add: $\frac{df}{d t} = F + (- F) = 0$ . So $f = p - F t$ is conserved. Why this step? We combine the two contributions from T1. The flow term (step 2) says the current pushes $p$ upward at rate $+ F$ ; the explicit-clock term (step 3) subtracts $F t$ at rate $- F$ . These are equal and opposite by construction, so they cancel exactly and the total change is zero — that is what "conserved" means.

Verify: From Newton, $\overset{p}{˙} = F$ so $p = p_{0} + F t$ , giving $p - F t = p_{0} =$ constant. ✓ The explicit term was essential — dropping it would have (wrongly) given $\dot{f} = F \neq = 0$ .

Cell H — a real-world word problem

Worked example Ex 8 · Is the total momentum of two beads on a frictionless track conserved? (Cell H)

Two beads on a wire interact only through each other via a potential $V (q_{1} - q_{2})$ (it depends only on their separation). $H = \frac{p _{1}^{2}}{2 m} + \frac{p _{2}^{2}}{2 m} + V (q_{1} - q_{2})$ . Take total momentum $P = p_{1} + p_{2}$ . Is it conserved?

Forecast: intuitively, internal forces are equal and opposite (Newton's third law), so total momentum should not change. Let the bracket confirm it.

Test conservation via ${P, H}$ (no explicit $t$ , so by Theorem T1 $\dot{P} = {P, H}$ ). Why this step? A quantity with ${f, H} = 0$ and no explicit time is conserved — that's the master test T1 with the last term gone.
${P, H} = {p_{1} + p_{2}, H} = {p_{1}, H} + {p_{2}, H}$ . Why this step? We use the bilinearity arrow from Figure 2: the bracket is additive in its first slot, so a bracket of a sum splits into a sum of brackets. This lets us handle each bead's momentum separately.
${p_{1}, H} = - \frac{\partial H}{\partial q _{1}} = - V^{'} (q_{1} - q_{2})$ and ${p_{2}, H} = - \frac{\partial H}{\partial q _{2}} = - V^{'} (q_{1} - q_{2}) \cdot (- 1) = + V^{'} (q_{1} - q_{2})$ . Why this step? Each ${p_{i}, H} = - \partial H / \partial q_{i}$ from the recipe. Now the chain rule is essential: $V$ is a function of the single combination $u = q_{1} - q_{2}$ , so $\partial V / \partial q_{1} = V^{'} (u) \cdot (+ 1)$ while $\partial V / \partial q_{2} = V^{'} (u) \cdot (- 1)$ . The opposite inner derivatives $+ 1$ and $- 1$ are the whole story.
Add: ${P, H} = - V^{'} + V^{'} = 0$ . Total momentum is conserved. Why this step? We sum the two contributions from step 3. Because the chain rule handed us $- V^{'}$ from bead 1 and exactly $+ V^{'}$ from bead 2, the terms are equal in size and opposite in sign, so they annihilate — leaving ${P, H} = 0$ with nothing left over.

Verify: The cancellation is Newton's third law in disguise: the force on bead 1 is $- V^{'}$ , on bead 2 is $+ V^{'}$ — equal and opposite. Sum of forces $= 0 \Rightarrow \dot{P} = 0$ . ✓ This is also Noether's Theorem & Conservation Laws: $H$ is unchanged if you shift both beads by the same amount (translation symmetry) $\Rightarrow$ total momentum conserved.

Cell I — the exam twist: manufacture a conservation law

Worked example Ex 9 · Two conserved quantities breed a third (Cell I)

A system conserves $L_{x}$ and $L_{y}$ (both have ${L_{x}, H} = {L_{y}, H} = 0$ ). Prove $L_{z}$ is automatically conserved too, without touching $H$ 's details.

Forecast: you might think you must recompute ${L_{z}, H}$ from scratch. The Jacobi identity gives it for free — guess how.

Recall from Ex 6 that ${L_{x}, L_{y}} = L_{z}$ . So proving " $L_{z}$ is conserved," i.e. ${L_{z}, H} = 0$ , is the same as proving ${{L_{x}, L_{y}}, H} = 0$ . Why this step? We rewrite the unknown $L_{z}$ as a bracket of two things we already know are conserved. Only then can Jacobi — which is a statement about brackets-of-brackets — get a grip.
Write Theorem T2 (Jacobi) with the three quantities $f = L_{x}$ , $g = L_{y}$ , $h = H$ : ${L_{x}, {L_{y}, H}} + {L_{y}, {H, L_{x}}} + {H, {L_{x}, L_{y}}} = 0.$ Why this step? Jacobi is the one algebraic law relating three quantities through nested brackets, so it is exactly the tool for "does the new object ${L_{x}, L_{y}}$ also commute with $H$ ?" We just slot our three quantities into its cyclic pattern $f \to g \to h \to f$ .
Kill the first two terms using the assumptions. The inner bracket of the first term is ${L_{y}, H} = 0$ (given), so ${L_{x}, 0} = 0$ . The inner bracket of the second term is ${H, L_{x}} = - {L_{x}, H} = - 0 = 0$ (given ${L_{x}, H} = 0$ , then antisymmetry), so ${L_{y}, 0} = 0$ . Why this step? Both $L_{x}$ and $L_{y}$ are conserved, so any bracket built on ${L_{x}, H}$ or ${L_{y}, H}$ collapses to a bracket with $0$ , which is $0$ by the Constants arrow. Antisymmetry is what turns ${H, L_{x}}$ into $- {L_{x}, H}$ so we can use the given fact.
What remains from T2 is therefore ${H, {L_{x}, L_{y}}} = 0$ . Flip it by antisymmetry: ${{L_{x}, L_{y}}, H} = - {H, {L_{x}, L_{y}}} = 0$ . Since ${L_{x}, L_{y}} = L_{z}$ , this reads ${L_{z}, H} = 0$ . Why this step? With two of the three Jacobi terms gone, the identity " $0 + 0 + (third) = 0$ " forces the third term to vanish. A last antisymmetry flip puts it in the standard " ${f, H}$ " order, and substituting ${L_{x}, L_{y}} = L_{z}$ delivers the conclusion.
By Theorem T1 (no explicit time in $L_{z}$ ), $\dot{L}_{z} = {L_{z}, H} = 0$ , so $L_{z}$ is conserved. ∎ Why this step? Conservation is defined through T1; having shown the bracket with $H$ vanishes, the time-evolution law immediately certifies $L_{z}$ as constant.

Verify: This is Poisson's theorem: the bracket of two constants of motion is itself a constant of motion. Physically, if two components of angular momentum are conserved (rotational symmetry about two axes), full rotational symmetry follows and all of $L$ is conserved. ✓ Quantizing echoes it: simultaneous symmetry generators close into a Lie algebra — see Angular Momentum Algebra.

Recall Quick self-test

${p, q} =$ ? ::: $- 1$ (antisymmetry flips ${q, p} = 1$ ). ${q^{2}, p} =$ ? ::: $2 q$ (power rule / Leibniz). ${L_{x}, L_{y}} =$ ? ::: $+ L_{z}$ (cyclic $x \to y \to z$ ). If $f = p - F t$ under force $F$ , is $f$ conserved? ::: Yes — the bracket gives $+ F$ , the explicit term gives $- F$ , sum $0$ . Two conserved $A, B$ : what is ${A, B}$ ? ::: Also conserved (Poisson's theorem, via Jacobi).

Connections

Parent topic (Hinglish) — the core definition and properties.
Hamiltonian Mechanics — supplies the $H$ and Hamilton's equations used in Cells B, G, H.
Noether's Theorem & Conservation Laws — the symmetry story behind Cell H.
Angular Momentum Algebra — Cells F and I live here.
Commutators in Quantum Mechanics — every classical bracket above has a $\frac{1}{i ℏ} [\cdot]$ twin.
Canonical Transformations — why brackets like ${q, p} = 1$ are preserved.
Liouville's Theorem — the flow whose "current" the brackets measure.