Poisson Brackets for Fields

The Hamiltonian formulation for fields ended with a promissory note: it exhibited the equal-time bracket $\{\phi(\mathbf{x}), \pi(\mathbf{y})\} = \delta^3(\mathbf{x}-\mathbf{y})$ as a summary of the canonical structure and deferred the full algebra. This article pays the note — and the algebra turns out to be the single most load-bearing structure classical field theory hands to quantum field theory. The Poisson bracket of the Hamiltonian chapters, generalized from $2N$ coordinates to two functions' worth, becomes a machine with three outputs: it is the equations of motion ( $\dot F = \{F, H\}$ regenerates Klein-Gordon in three lines), it is the symmetry apparatus (every Noether charge $Q$ acts through $\{Q, \cdot\}$ as the generator of its own transformation), and it is — after one substitution — the operator algebra of QFT, because Dirac's rule $\{,\} \to [,]/i\hbar$ maps this article's every equation onto the equal-time commutators that open every quantum field theory course. The construction costs almost nothing new. The discrete bracket $\{q_i, p_j\} = \delta_{ij}$ passes through the chapter's dictionary — index $i \to$ label $\mathbf{x}$ , sum $\to$ integral, partial derivative $\to$ functional derivative, Kronecker $\to$ Dirac delta — and everything the Poisson-bracket articles of the Hamiltonian chapters established (antisymmetry, Leibniz, Jacobi, charges closing on Lie algebras) survives the passage intact. What is genuinely new is the payoff scale: the charges built from the stress-energy tensor close under this bracket on the Poincaré algebra, so an entire spacetime symmetry group lives inside the bracket structure of one scalar field. Master this article and the capstone's "quantization is one substitution" claim stops being a slogan and becomes an inventory you can check line by line.

Key takeaways

The discrete canonical relations $\{q_i, p_j\} = \delta_{ij}$ become the equal-time field brackets $\{\phi(\mathbf{x}), \pi(\mathbf{y})\} = \delta^3(\mathbf{x}-\mathbf{y})$ , $\{\phi,\phi\} = \{\pi,\pi\} = 0$ — one independent canonical pair per spatial point.
For functionals: $\{F, G\} = \int \mathrm{d}^3x\, \big(\tfrac{\delta F}{\delta\phi}\tfrac{\delta G}{\delta\pi} - \tfrac{\delta F}{\delta\pi}\tfrac{\delta G}{\delta\phi}\big)$ — the discrete bracket with $\sum_i \to \int\mathrm{d}^3x$ and partials $\to$ functional derivatives.
Dynamics is algebra: $\dot F = \{F, H\} + \partial_t F$ ; taking $F = \phi, \pi$ regenerates Hamilton's field equations and hence Klein-Gordon.
The bracket is antisymmetric, bilinear, Leibniz, and satisfies Jacobi — a genuine Poisson algebra on the infinite-dimensional phase space, guaranteed because the fundamental brackets are c-numbers.
Noether charges are generators: $\{Q, \phi\}$ produces the infinitesimal symmetry transformation; $P^i$ generates translations, $H$ time evolution, $Q^A$ internal rotations.
Charge brackets close on the symmetry's Lie algebra: $\{Q^A, Q^B\} = f^{ABC}Q^C$ ; the $P^\mu, J^{\mu\nu}$ of the stress-energy articles close on the Poincaré algebra.
Everything is equal-time: the bracket lives on a spatial slice; Lorentz covariance is carried by the charge algebra, not by the bracket itself.
Key idea: one bracket, three jobs — equations of motion, symmetry generation, and the classical template of quantum commutators.

The core insight

Replace the discrete canonical pair by the field-momentum pair with $\{\phi(\mathbf{x}), \pi(\mathbf{y})\} = \delta^3(\mathbf{x}-\mathbf{y})$ at equal time; then $\dot F = \{F, H\}$ governs all dynamics, and Noether charges generate their symmetries through the bracket — the complete algebraic skeleton that canonical quantization promotes to quantum field theory.

Where this lives today

The equal-time bracket is the exact classical template of the commutation relations that define quantum field theory: every QFT course opens by postulating $[\hat\phi(\mathbf{x}), \hat\pi(\mathbf{y})] = i\hbar\,\delta^3(\mathbf{x}-\mathbf{y})$ , which is equation (1) processed through Dirac's substitution, and the charge algebras of this article become the operator algebras — Poincaré generators on Fock space, current algebras constraining scattering amplitudes through Ward identities — that organize the Standard Model's calculations. Where the naive bracket fails, the modern machinery is still Dirac's repair: constrained Hamiltonian analysis and Dirac brackets underlie the BRST quantization of gauge theories, the ADM formulation that numerical relativity uses to evolve black-hole mergers (its constraints are monitored at every timestep of every LIGO-template simulation), and loop quantum gravity's canonical program. And the bracket keeps earning its keep on the classical frontier: geometers study Poisson manifolds as a subject in their own right, deformation quantization constructs quantum theories directly as $\hbar$ -deformations of Poisson algebras (Kontsevich's formality theorem, one of the results behind his 1998 Fields Medal), and plasma physics and magnetohydrodynamics run stable numerical schemes built to preserve noncanonical Poisson structures exactly — bracket-preserving integrators being the field-theory descendants of the symplectic integrators that keep solar-system simulations honest for gigayears. Within the chapter, this article arms the three that follow: Noether's theorem for fields supplies the charges the bracket turns into generators, the stress-energy and angular-momentum articles assemble the Poincaré charges whose algebra (7) certifies covariance, and the capstone performs the substitution this article's every equation was staged for.

The physical picture

Start from what the bracket meant in particle mechanics, because the meaning survives the continuum limit unchanged. The Poisson-bracket articles of the Hamiltonian chapters established the geometry: phase space carries a symplectic pairing, and $\{A, B\}$ measures how the observable $A$ changes when the system is dragged along the flow that $B$ generates. The bracket table $\{q_i, p_j\} = \delta_{ij}$ is the statement that each coordinate is paired with exactly one momentum — its own — and is blind to all the others.

Figure

From Kronecker to Dirac — the canonical pairing passes to the continuum with its locality intact

For fields, the phase space is the one the Hamiltonian article built — a canonical pair $(\phi(\mathbf{x}), \pi(\mathbf{x}))$ at every point of the slice — and the bracket's meaning is identical: $\{A, B\}$ still reads "the rate of change of $A$ along the flow generated by $B$ ." The delta function in the fundamental bracket is not a technicality but a physical statement of locality: perturb the momentum density in a small region around $\mathbf{y}$ and the conjugate response of the field is confined to exactly that region — no spread, no leakage, one independent pair per point, precisely as the bead chain promised before the continuum limit was taken. The picture to carry through the whole article is the bracket as a universal flow machine. Feed it the Hamiltonian and it cranks out time evolution. Feed it the total field momentum $P^i$ and it drags every field configuration sideways through space. Feed it an internal charge and it rotates the field multiplet in its internal plane, touching no point of spacetime at all. Conservation and generation become two sides of one identity: $\{Q, H\} = 0$ says simultaneously that $H$ 's flow preserves $Q$ (the charge is conserved) and that $Q$ 's flow preserves $H$ (the transformation is a symmetry). That reciprocity — proved for particles in the Hamiltonian chapters and inherited here wholesale — is what makes the bracket the natural home of Noether's theorem on the canonical side.

The bracket, promoted to functionals

Everything begins with the fundamental equal-time relations, the dictionary image of the discrete bracket table. On one fixed time slice,

\{\phi(\mathbf{x}, t), \pi(\mathbf{y}, t)\} = \delta^3(\mathbf{x} - \mathbf{y}), \qquad \{\phi(\mathbf{x}), \phi(\mathbf{y})\} = \{\pi(\mathbf{x}), \pi(\mathbf{y})\} = 0 \qquad (1)

where: $\pi = \partial\mathcal{L}/\partial\dot\phi$ is the momentum density conjugate to the field, and $\delta^3$ is the three-dimensional Dirac delta — the continuum Kronecker delta, with dimensions of inverse volume, which is exactly the factor that converts the discrete $\delta_{ij}$ into a per-unit-volume pairing. Both arguments sit at the same time $t$ : the bracket is a structure on one slice of the phase space, not on spacetime. Observables of a field theory are functionals — energy, momentum, charges, smeared field values — so the bracket must eat functionals. The general definition transcribes the discrete $\{F, G\} = \sum_i (\partial_{q_i}F\,\partial_{p_i}G - \partial_{p_i}F\,\partial_{q_i}G)$ term for term:

\{F, G\} = \int \mathrm{d}^3x \left( \frac{\delta F}{\delta\phi(\mathbf{x})} \frac{\delta G}{\delta\pi(\mathbf{x})} - \frac{\delta F}{\delta\pi(\mathbf{x})} \frac{\delta G}{\delta\phi(\mathbf{x})} \right) \qquad (2)

where: the functional derivatives are the objects constructed in the functional-derivatives article, and the result of the integral is again a functional — the algebra closes. Consistency check, worth doing once in your life: take $F = \phi(\mathbf{y})$ , $G = \pi(\mathbf{z})$ . Then $\delta F/\delta\phi(\mathbf{x}) = \delta^3(\mathbf{x}-\mathbf{y})$ and $\delta G/\delta\pi(\mathbf{x}) = \delta^3(\mathbf{x}-\mathbf{z})$ , so (2) collapses two deltas against each other: $\{\phi(\mathbf{y}), \pi(\mathbf{z})\} = \int\mathrm{d}^3x\,\delta^3(\mathbf{x}-\mathbf{y})\,\delta^3(\mathbf{x}-\mathbf{z}) = \delta^3(\mathbf{y}-\mathbf{z})$ — definition (2) reproduces (1), as it must. From (2) the bracket inherits the full algebraic kit of its finite-dimensional ancestor:

\{A, B\} = -\{B, A\}, \qquad \{A, BC\} = \{A, B\}\,C + B\,\{A, C\}, \qquad \{A, \{B, C\}\} + \text{cyc} = 0 \qquad (3)

where: antisymmetry and bilinearity are visible by inspection of (2); the Leibniz rule follows from the product rule for functional derivatives; and the Jacobi identity ("cyc" = the two cyclic permutations of $A, B, C$ ) is the non-trivial one — it is what makes the bracket a genuine Poisson structure rather than a mere antisymmetric pairing, and it is proved below from the c-number nature of (1). Jacobi is also the property with the longest quantum shadow: it is the classical counterpart of the associativity of operator products. Dynamics enters through the master evolution equation, unchanged from particle mechanics:

\dot F = \{F, H\} + \frac{\partial F}{\partial t} \qquad (4)

where: $H = \int\mathcal{H}\,\mathrm{d}^3x$ is the Hamiltonian functional of the previous article and $\partial F/\partial t$ collects only explicit time dependence (a time-dependent smearing function, say) — for the observables of a closed field theory it vanishes. Setting $F = \phi(\mathbf{x})$ and $F = \pi(\mathbf{x})$ in (4) yields $\dot\phi = \delta H/\delta\pi$ and $\dot\pi = -\delta H/\delta\phi$ : Hamilton's field equations are the two most elementary instances of the bracket's dynamics, and the first derivation below runs them all the way to Klein-Gordon.

Derivation ABracket dynamics regenerates the Klein-Gordon equation

We show that the equal-time bracket, fed the free scalar Hamiltonian, reproduces the Klein-Gordon equation $\Box\phi + m^2\phi = 0$ — closing the loop between the algebraic formulation and the Euler-Lagrange dynamics of the Lagrangian articles. The strategy is the direct one: compute $\dot\phi = \{\phi, H\}$ and $\dot\pi = \{\pi, H\}$ using the functional bracket (2), then combine the two first-order equations into one second-order equation. The step to watch is the gradient term, where an integration by parts inside the functional derivative supplies the Laplacian and a sign.

Step 1. The free scalar Hamiltonian, from the Legendre transform of $\mathcal{L} = \tfrac12(\partial_\mu\phi)(\partial^\mu\phi) - \tfrac12 m^2\phi^2$ with $\pi = \dot\phi$ :

H = \int \mathrm{d}^3x \left[ \tfrac{1}{2}\pi^2 + \tfrac{1}{2}(\nabla\phi)^2 + \tfrac{1}{2} m^2\phi^2 \right]

Step 2. Velocity equation. In (2) with $F = \phi(\mathbf{y})$ , only $\delta H/\delta\pi$ survives, and only the $\tfrac12\pi^2$ term of $H$ depends on $\pi$ :

\dot\phi(\mathbf{y}) = \{\phi(\mathbf{y}), H\} = \frac{\delta H}{\delta\pi(\mathbf{y})} = \pi(\mathbf{y})

Why this is a result and not a restatement: in the phase-space formulation $\phi$ and $\pi$ are independent coordinates; the Legendre relation $\pi = \dot\phi$ is recovered here as the first equation of motion, exactly as in the Hamiltonian article's variational treatment.

Step 3. Momentum equation. Now $F = \pi(\mathbf{y})$ picks out $-\delta H/\delta\phi$ . The mass term gives $m^2\phi$ directly; the gradient term needs one integration by parts, $\delta\!\int\tfrac12(\nabla\phi)^2 = \int \nabla\phi\cdot\nabla\delta\phi = -\int (\nabla^2\phi)\,\delta\phi$ , with the boundary term dropped for fields decaying at spatial infinity:

\dot\pi(\mathbf{y}) = \{\pi(\mathbf{y}), H\} = -\frac{\delta H}{\delta\phi(\mathbf{y})} = \nabla^2\phi(\mathbf{y}) - m^2\phi(\mathbf{y})

Step 4. Combine. Differentiate Step 2 in time and substitute Step 3:

\ddot\phi = \dot\pi = \nabla^2\phi - m^2\phi \quad\Longrightarrow\quad \Box\phi + m^2\phi = 0, \qquad \Box = \partial_t^2 - \nabla^2

The bracket formalism and the Lagrangian formalism are the same dynamics in different clothes — three lines of algebra, no variational calculus in sight. Verified: matches the covariant Euler-Lagrange derivation of the field-equations article and Goldstein ch. 13; the integration-by-parts sign checked against the functional-derivatives article's component formula $\delta H/\delta\phi = \partial\mathcal{H}/\partial\phi - \nabla\cdot\partial\mathcal{H}/\partial(\nabla\phi)$ .

Go deeper · Derivation B · 2 minThe Jacobi identity from the fundamental brackets

Derivation BThe Jacobi identity from the fundamental brackets

We prove that the functional bracket (2) satisfies the Jacobi identity, certifying that field phase space carries a genuine Poisson algebra — the property that guarantees flows compose consistently, that time evolution is a one-parameter family of canonical transformations, and that the charge algebras of the second half of this article close properly. The strategy exploits a structural gift: it suffices to verify Jacobi on the elementary fields themselves, because bilinearity and the Leibniz rule then extend it to all functionals built from them. The thing to notice is why the elementary case is trivial — the fundamental brackets are c-numbers.

Step 1. Observe the key structural fact: every fundamental bracket in (1) is either zero or $\delta^3(\mathbf{x}-\mathbf{y})$ — a fixed distribution with no dependence on the fields. A bracket of anything with a c-number vanishes, since all functional derivatives of a constant are zero.

Step 2. Take an elementary triple, say $A = \phi(\mathbf{x})$ , $B = \pi(\mathbf{y})$ , $C = \phi(\mathbf{z})$ . The inner brackets are $\{B, C\} = -\delta^3(\mathbf{y}-\mathbf{z})$ , $\{C, A\} = 0$ , $\{A, B\} = \delta^3(\mathbf{x}-\mathbf{y})$ — all c-numbers.

Step 3. Each nested bracket therefore vanishes individually:

\{A, \{B, C\}\} = \{B, \{C, A\}\} = \{C, \{A, B\}\} = 0

and the cyclic sum is zero. The same argument disposes of every other choice of elementary triple — the inner bracket is always a c-number.

Step 4. Extend to arbitrary functionals. Any $F[\phi, \pi]$ is built from the elementary fields by sums, products, and limits; the bracket acts on such combinations through bilinearity and Leibniz (3), and a direct check shows these operations propagate the Jacobi identity: if Jacobi holds for the generators of an algebra and the bracket differentiates products, it holds on everything they generate. Why the general theorem lurks here: in symplectic language, Jacobi for a bracket $\{A,B\} = \omega^{ab}\partial_a A\,\partial_b B$ holds automatically when the symplectic components $\omega^{ab}$ are constant — and (1) says precisely that the field-theory $\omega$ is the constant off-diagonal block $\delta^3$ . Non-constant brackets (curved phase spaces, Dirac brackets) must earn Jacobi case by case. *Verified: matches José & Saletan ch. 9's theorem that constant-coefficient symplectic structures satisfy Jacobi; consistent with the finite-dimensional proof in the Poisson-bracket-algebra article, of which this is the verbatim continuum transcription.*

Charges as generators

Here is the payoff that elevates the bracket from a compact notation to the organizing structure of field theory. Noether's theorem (next article, for the field-theory version; the particle version is the W3 article) assigns each continuous symmetry a conserved charge $Q = \int\mathrm{d}^3x\, j^0$ . The bracket assigns each charge a flow. The two assignments are inverse to each other:

\{Q, \phi(\mathbf{x})\} = -\,\delta\phi(\mathbf{x}) \qquad (5)

where: $\delta\phi$ is the infinitesimal transformation of the field under the symmetry that $Q$ came from (sign conventions vary by author; what is universal is that the bracket with $Q$ is the transformation). Read (5) forward: the conserved quantity is not an inert number but an operator on phase space, dragging configurations along its symmetry direction. Read it backward: if you know a charge, you can reconstruct the symmetry it belongs to by computing one bracket. The cleanest concrete case is spatial translations. The total field momentum, from the stress-energy article's $T^{0i}$ , is $P^i = -\int\mathrm{d}^3x\,\pi\,\partial^i\phi$ . Bracket it with the field:

\{\phi(\mathbf{y}), P^i\} = -\int\mathrm{d}^3x\,\{\phi(\mathbf{y}), \pi(\mathbf{x})\}\,\partial^i\phi(\mathbf{x}) = -\,\partial^i\phi(\mathbf{y}) \qquad (6)

where: the delta from (1) collapsed the integral, and $-\partial^i\phi$ is exactly the first-order change of $\phi$ under the shift $\mathbf{x} \to \mathbf{x} + \boldsymbol{\epsilon}$ : the momentum charge generates translations, just as $H$ generates time evolution via (4). One machine, every symmetry: rotations from the angular-momentum charges, boosts from the boost moments, internal rotations from internal charges.

Figure

The bracket as a flow machine — each conserved charge drags the field configuration along its own symmetry direction

The final structural fact is that the generators do not merely act on fields — they act on each other, and their mutual brackets reproduce the Lie algebra of the symmetry group. For the spacetime charges assembled from the stress-energy tensor,

\{P^\mu, P^\nu\} = 0, \qquad \{J^{\mu\nu}, P^\rho\} = \eta^{\nu\rho}P^\mu - \eta^{\mu\rho}P^\nu \qquad (7)

where: $P^\mu$ is the energy-momentum four-vector of charges, $J^{\mu\nu}$ the angular-momentum/boost charges of the spin-current article, and $\eta = \mathrm{diag}(+,-,-,-)$ the Minkowski metric. Together with $\{J, J\}$ (which closes on $J$ 's) these are the Poincaré algebra: the complete relativistic symmetry group of flat spacetime, living inside the equal-time bracket structure of a field theory. This is how covariance survives the slicing: the bracket itself picked a frame, but the algebra of the charges knows the full spacetime group — the theme the Lorentz-covariance article develops. For internal symmetries the same closure reads $\{Q^A, Q^B\} = f^{ABC}Q^C$ , and the derivation below earns it honestly.

Go deeper · Derivation C · 1 minClosure of the internal charge algebra

Derivation CClosure of the internal charge algebra

We derive the hallmark result that the Noether charges of an internal symmetry close, under the field Poisson bracket, on the Lie algebra of the symmetry group: $\{Q^A, Q^B\} = f^{ABC}Q^C$ . This is the classical origin of "current algebra" and the exact structure that quantization turns into charge commutators. The setting: a multiplet $\phi_a$ transforming as $\delta\phi_a = -i\epsilon^A (T^A)_{ab}\phi_b$ , with generator matrices obeying $[T^A, T^B] = if^{ABC}T^C$ . The strategy is two bracket computations: first show each charge generates its rotation on the fields, then apply Leibniz to the bracket of two charges and watch the generator commutator assemble itself.

Step 1. The Noether charge for parameter $A$ (from $j^{0A} = \pi_a\,\delta\phi_a/\delta\epsilon^A$ , anticipating the next two articles' construction):

Q^A = -\int\mathrm{d}^3x\;\pi_a\,(iT^A)_{ab}\,\phi_b

Step 2. The charge generates the rotation. Bracket with a field component, using $\{\phi_c(\mathbf{y}), \pi_a(\mathbf{x})\} = \delta_{ca}\,\delta^3(\mathbf{x}-\mathbf{y})$ to collapse the integral:

\{Q^A, \phi_c(\mathbf{y})\} = -i\,(T^A)_{cb}\,\phi_b(\mathbf{y})

which is exactly $\delta\phi_c/\epsilon^A$ — equation (5) verified in this case. The matching computation on $\pi$ gives $\{Q^A, \pi_c\} = +i\,\pi_b(T^A)_{bc}$ : the momentum rotates with the transposed (inverse) matrix, as conjugate variables must.

Step 3. Bracket two charges. Write $Q^B = -\int\mathrm{d}^3y\,\pi_c\,(iT^B)_{cd}\,\phi_d$ and apply the Leibniz rule, letting $Q^A$ act on the $\pi_c$ and $\phi_d$ inside via Step 2:

\{Q^A, Q^B\} = -\int\mathrm{d}^3y\;\pi\,\big(iT^A\, iT^B - iT^B\, iT^A\big)\phi \cdot(-1) = -\int\mathrm{d}^3y\;\pi_a\,\big(i\,[T^A, T^B]\big)_{ab}\,\phi_b

Why only the commutator survives: the rotation of $\pi$ contributes $T^A$ acting from the left of $T^B$ , the rotation of $\phi$ contributes it from the right, with opposite signs from antisymmetry — precisely the commutator, the same mechanism that makes matrix Lie algebras close.

Step 4. Substitute $[T^A, T^B] = if^{ABC}T^C$ and recognize the result:

\{Q^A, Q^B\} = f^{ABC}\left(-\int\mathrm{d}^3y\;\pi_a\,(iT^C)_{ab}\,\phi_b\right) = f^{ABC}\,Q^C

The charges form a copy of the symmetry's Lie algebra inside the Poisson algebra of observables. Quantize ( $\{,\}\to[,]/i\hbar$ ) and this becomes $[\hat Q^A, \hat Q^B] = i f^{ABC}\hat Q^C$ — the charge algebra of quantum field theory, derived before quantum mechanics enters. Verified: closure matches the standard QFT result (e.g. any treatment of current algebra); Step 2 recomputed by collapsing the delta explicitly; signs consistent with the antisymmetry conventions of (1)-(2).

Limiting cases

Case	Result	Meaning
single mode kept	$\{\phi_k, \pi_k\} = 1$	bracket contracts to one particle-mechanics pair
discrete lattice, spacing $a$	$\{\phi_i, \pi_j\} = \delta_{ij}/a^3$	Dirac delta regularized: the bead-chain bracket
$F, G$ both functionals of $\phi$ only	$\{F, G\} = 0$	configuration observables commute — no symplectic pairing without $\pi$
$Q$ conserved ( $\{Q,H\}=0$ )	$Q$ 's flow is a symmetry of $H$	conservation ⇔ generation, both read from one bracket
unequal times	bracket undefined	must solve dynamics first; only equal-time structure is canonical
gauge theories ( $\pi^0 \equiv 0$ )	naive bracket inconsistent	constraints; repair = Dirac bracket
quantization	$\{,\} \to [,]/i\hbar$	(1) becomes $[\hat\phi, \hat\pi] = i\hbar\,\delta^3$ : QFT's opening axiom

Common trapThe bracket lives at equal time — and π is a density, not a number

Two standing errors, both fatal. First: writing $\{\phi(x), \pi(y)\}$ for spacetime points at different times is meaningless. The Poisson bracket is a structure on one phase-space slice; relating fields at different times is not algebra but dynamics — you must solve the equations of motion first (the objects that do carry unequal-time information, Green functions and their quantum descendants like the Feynman propagator, are built from solutions, not from the canonical structure alone). Second: $\pi(\mathbf{x})$ is a momentum density and the bracket (1) is a distribution. Brackets of unsmeared fields produce deltas, and manipulating them as if they were smooth functions invites divergences that are bookkeeping artifacts: $\{\pi(\mathbf{x}), H\}$ is fine (the integral in $H$ eats the delta), but expressions like $\{\phi(\mathbf{x}), \pi(\mathbf{x})\}$ at coincident points are $\delta^3(0)$ — the infinite-volume-density symptom, not physics. The clean practice, inherited from the functional-derivatives article: smear with test functions, $\phi[f] = \int f\phi\,\mathrm{d}^3x$ , whenever an expression looks singular. Then $\{\phi[f], \pi[g]\} = \int f g\,\mathrm{d}^3x$ — finite, and everything downstream is honest.

Worked examples

1 · One mode of the bank, evolved by the bracket (confidence builder)

Keep a single real Fourier mode of the free scalar field in a box: canonical pair $(\phi_k, \pi_k)$ with $\{\phi_k, \pi_k\} = 1$ and mode Hamiltonian $H_k = \tfrac12\pi_k^2 + \tfrac12\omega_k^2\phi_k^2$ . Take $m = 0.50$ , $k = 2\pi/L$ with $L = 10.0$ (natural units), and initial data $\phi_k(0) = 0.40$ , $\pi_k(0) = 0$ . Find $\omega_k$ , evolve with the bracket, and give $\phi_k$ at $t = 2.0$ .

Step 1. The wavevector: $k = 2\pi/10.0 = 0.6283$ . The mode frequency from the dispersion relation: $\omega_k^2 = k^2 + m^2 = 0.3948 + 0.2500 = 0.6448$ , so $\omega_k = \mathbf{0.8030}$ .

Step 2. Bracket dynamics: $\dot\phi_k = \{\phi_k, H_k\} = \pi_k$ and $\dot\pi_k = \{\pi_k, H_k\} = -\omega_k^2\phi_k$ — the harmonic oscillator, as the Hamiltonian article's diagonalization promised. With the given initial data: $\phi_k(t) = 0.40\cos(\omega_k t)$ .

Step 3. At $t = 2.0$ : $\omega_k t = 1.6059$ rad, $\cos(1.6059) = -0.0352$ , so $\phi_k(2.0) = 0.40 \times (-0.0352) = \mathbf{-0.0141}$ . The mode has just passed its first zero crossing (quarter period $T/4 = \pi/2\omega_k = 1.956$ ). Energy check: $H_k = \tfrac12\omega_k^2(0.40)^2 = \tfrac12(0.6448)(0.16) = \mathbf{0.0516}$ , conserved because $\{H_k, H_k\} = 0$ — antisymmetry alone guarantees it.

2 · The momentum charge generates translations — checked on a wavepacket, with numbers (exam level)

A one-dimensional scalar field carries a rightward-moving Gaussian packet: $\phi(x) = A\,e^{-(x-x_0)^2/2\sigma^2}$ with $\pi(x) = -c\,\partial_x\phi(x)$ (the profile translating at speed $c$ ). Take $A = 0.20$ , $\sigma = 1.00$ , $c = 0.80$ . (i) Compute the field momentum $P = -\int\pi\,\partial_x\phi\,\mathrm{d}x$ . (ii) Verify that $\{\phi, P\}$ translates the packet, and use it to predict $\phi$ at the point $x = x_0$ after the flow parameter $\epsilon = 0.05$ .

Step 1. Substitute $\pi = -c\,\partial_x\phi$ into the charge: $P = c\int(\partial_x\phi)^2\,\mathrm{d}x$ . For the Gaussian, $\partial_x\phi = -A\,\frac{(x-x_0)}{\sigma^2}\,e^{-(x-x_0)^2/2\sigma^2}$ , so $(\partial_x\phi)^2 = \frac{A^2 u^2}{\sigma^4}e^{-u^2/\sigma^2}$ with $u = x - x_0$ , and $\int u^2 e^{-u^2/\sigma^2}\mathrm{d}u = \tfrac{\sqrt{\pi}}{2}\sigma^3$ .

Step 2. Therefore $P = c\,\frac{A^2}{\sigma^4}\cdot\frac{\sqrt{\pi}}{2}\sigma^3 = \frac{\sqrt{\pi}}{2}\,\frac{cA^2}{\sigma}$ . Numbers: $\frac{\sqrt{\pi}}{2} = 0.8862$ , so $P = 0.8862 \times 0.80 \times 0.04 / 1.00 = \mathbf{0.0284}$ — a positive momentum for a rightward-moving packet, sign consistent with (6).

Step 3. The generator statement (6): the flow of $P$ with parameter $\epsilon$ changes the field by $\delta\phi = \epsilon\{\phi, P\} = -\epsilon\,\partial_x\phi$ , i.e. $\phi(x) \to \phi(x - \epsilon)$ to first order — a rigid shift rightward by $\epsilon$ , shape untouched. No re-solving of dynamics, no equations of motion: the charge itself moves the configuration.

Step 4. Predict the new value at $x = x_0$ . Exactly: $\phi(x_0 - \epsilon) = A\,e^{-\epsilon^2/2\sigma^2} = 0.20\,e^{-0.00125} = \mathbf{0.19975}$ . First-order flow: $\delta\phi(x_0) = -\epsilon\,\partial_x\phi|_{x_0} = 0$ (the peak is a stationary point), so the generator predicts no first-order change at the crest — and the exact answer confirms it: the deficit $0.00025 = 0.125\%$ is second order, $O(\epsilon^2)$ , exactly as a first-order generator statement permits. At a flank point $x = x_0 + \sigma$ , by contrast, the first-order shift is $-\epsilon\,\partial_x\phi|_{x_0+\sigma} = 0.05 \times 0.20\,e^{-1/2}/1.0 = \mathbf{+0.00607}$ — the packet's leading edge rises as the lump advances. The bracket turned a conserved number, $P = 0.0284$ , into a verified motion of the whole configuration.

Origins · 1 minSiméon Denis Poisson and how this was discovered

Siméon Denis Poisson

1781–1840 · École Polytechnique, Paris

Poisson introduced his bracket in 1809 as a computational device for planetary perturbation theory — a shortcut for producing new constants of motion from old ones — and would be astonished by what it became. The son of a provincial administrator, steered to Polytechnique at seventeen, he succeeded Fourier as professor at twenty-five and produced across four decades a staggering range: the Poisson equation of gravitation and electrostatics, the Poisson distribution of probability, the Poisson ratio of elasticity — and the bracket. His theorem that the bracket of two conserved quantities is again conserved (a one-line consequence of the Jacobi identity, though Jacobi's clean formulation came a generation later) is precisely the closure mechanism behind this article's charge algebras. The bracket's promotion from trick to foundation happened twice after his death: Jacobi and Lie uncovered its geometric depth in the nineteenth century, and in 1925 Dirac — reportedly during a Sunday walk, anxious until the library opened Monday morning so he could check Poisson's definition — recognized Heisenberg's mysterious quantum products as the bracket's operator image, $[,] = i\hbar\{,\}$ . Every equal-time commutator in quantum field theory is Poisson's 1809 bookkeeping device, promoted mode for mode. He is one of the seventy-two names engraved on the Eiffel Tower.

History

1809

— Poisson introduces the bracket in celestial perturbation theory; his theorem generates new invariants from old.

1842

— Jacobi proves the identity bearing his name and builds the bracket into the core of analytical dynamics.

1890

— Lie's transformation-group theory reveals brackets and generators as two faces of one structure — the modern reading of equation (5).

1925

— Dirac identifies the quantum commutator as

i\hbar

times the Poisson bracket: canonical quantization in one line.

1949

— Dirac generalizes to constrained systems, repairing the bracket for gauge fields — the Dirac bracket that quantizes electromagnetism and, later, gravity programs.

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 13 (Continuous Systems and Fields) · Landau & Lifshitz, Mechanics, Vol. 1, §40–42 and the continuum extension · Morin, Introduction to Classical Mechanics, Hamiltonian and Lagrangian chapters · Taylor, Classical Mechanics, Ch. 13 (Hamiltonian Mechanics) · José & Saletan, Classical Dynamics: A Contemporary Approach, Ch. 9 (field-theoretic Poisson structure)

Provenance. A verifier should recompute: (1) the collapse of two deltas giving $\{\phi(\mathbf{y}),\pi(\mathbf{z})\}=\delta^3(\mathbf{y}-\mathbf{z})$ ; (2) the Klein–Gordon derivation $\dot\phi=\pi,\ \dot\pi=\nabla^2\phi-m^2\phi\Rightarrow\Box\phi+m^2\phi=0$ ; (3) Jacobi vanishing for elementary fields; (4) the closure $\{Q^A,Q^B\}=f^{ABC}Q^C$ via $[T^A,T^B]=if^{ABC}T^C$ . Cross-references: Goldstein, Poole & Safko Ch. 13; Landau–Lifshitz Vol. 1 §40–42; José & Saletan Ch. 9 for the symplectic/Dirac-bracket structure. All numerical factors (the $\tfrac12$ coefficients in $\mathcal{H}$ , signs of integration-by-parts terms) are standard textbook values. Additionally for this upgrade: worked-example numbers recomputed directly ( $k = 0.6283$ , $\omega_k = 0.8030$ , $\phi_k(2.0) = -0.0141$ , $H_k = 0.0516$ ; Gaussian-packet charge $P = \tfrac{\sqrt{\pi}}{2}cA^2/\sigma = 0.0284$ , flank shift $+0.00607$ , crest deficit $0.125\% = O(\epsilon^2)$ ). Scope note: the construction of Noether currents and charges is owned by the next three articles; the Poincaré algebra's full derivation by the angular-momentum and Lorentz-covariance articles; Dirac brackets are flagged, not developed.

Builds on: Hamiltonian Formulation for Fields · Functional Derivatives and Field Variations · From Discrete Systems to Continuous Fields · Covariant Euler-Lagrange Equations for Fields Leads to: Noether's Theorem for Continuous Symmetries · Conserved Currents and Charges · Internal Symmetries and Conserved Charge · Lorentz Covariance of Classical Field Theory

Is the Dirac delta in the fundamental bracket an approximation that gets smoothed out in real physics?

No — it is the exact statement of canonical locality, and its singular character is meaningful. The delta says the symplectic pairing has zero range: the momentum at $\mathbf{y}$ is conjugate to the field at $\mathbf{y}$ and to nothing else. Any smoothing would say a momentum kick here changes the field value over there, which would wreck the independence of the canonical pairs and, after quantization, violate microcausality (field operators at spacelike separation must commute). What is true is that physically meaningful observables are smeared — $\phi[f] = \int f\phi$ for a localized test function $f$ — and brackets of smeared observables are perfectly finite. The delta is the kernel of the structure; the smearing is how you read finite numbers out of it. The lattice regularization makes this concrete: on a grid of spacing $a$ the bracket is $\delta_{ij}/a^3$ , a large but finite number per cell, and $\delta^3$ is its $a \to 0$ limit — same locality, sharper language.

The bracket picks a time slice. Does a boosted observer, slicing spacetime differently, build a different theory?

They build a different-looking presentation of the same theory, and the charge algebra is the certificate. Each observer erects canonical coordinates on their own simultaneity slices, with their own equal-time deltas; the two phase spaces are related by a canonical transformation whose generator is built from the boost charges $J^{0i}$ . The mutual brackets of all ten Poincaré charges close on the Poincaré algebra (7) — and that closure is precisely the statement that translations, rotations, boosts and time evolution knit together consistently, so any invariant question (scattering outcomes, charge values, energy of a bound lump measured in its rest frame) receives the same answer on every slicing. What genuinely differs between observers is the bookkeeping: which part of the energy is "kinetic," which pairs of points are "simultaneous," what the instantaneous configuration looks like. The situation is exactly the Hamiltonian-versus-Lagrangian trade throughout this chapter: the bracket buys algebraic power at the cost of manifest covariance, and the Lorentz-covariance article audits what survives.

What breaks when I try to run this machinery on electromagnetism, and how bad is the damage?

The very first step fails: the momentum conjugate to the scalar potential, $\pi^0 = \partial\mathcal{L}/\partial\dot A_0$ , is identically zero for $\mathcal{L} = -\tfrac14 F_{\mu\nu}F^{\mu\nu}$ — the Legendre transform cannot be inverted, and the phase space is born carrying the constraint $\pi^0 = 0$ . Demanding the constraint be preserved in time generates a second one, and it is a famous equation: $\nabla\cdot\mathbf{E} = \rho$ , Gauss's law, appearing not as dynamics but as a restriction on allowed initial data. With constraints present, the naive bracket (2) is inconsistent — it generates flows that leave the constraint surface. Dirac's repair subtracts the offending directions: $\{A, B\}_D = \{A, B\} - \{A, \chi_a\}C^{ab}\{\chi_b, B\}$ , where the $\chi$ 's are the (second-class) constraints and $C$ is the inverse of their bracket matrix. The damage, then, is real but organized: gauge freedom means some canonical directions are unphysical, and the constrained formalism is the systematic accounting. For this chapter's scalar field none of this arises — every momentum inverts — which is why the scalar is the pedagogical vehicle and gauge theory is the graduate course that follows.

Poisson's theorem says the bracket of two conserved quantities is conserved. Does that mean I can generate endless new conservation laws?

The theorem is true — Jacobi's identity applied to $\{Q_1, Q_2\}$ with $\{Q_1, H\} = \{Q_2, H\} = 0$ gives $\{\{Q_1, Q_2\}, H\} = 0$ in one line — but the harvest is usually finite, for a structural reason this article makes visible: the charges close on a Lie algebra. Bracketing two Poincaré charges returns a linear combination of Poincaré charges (7); bracketing two internal charges returns $f^{ABC}Q^C$ — combinations of charges you already had. New, independent invariants appear only when the algebra has not yet closed, and that is exactly how it happens in practice: for the Kepler problem, bracketing angular-momentum components with Runge-Lenz components generates the full hidden $SO(4)$ algebra from a smaller seed. So the right reading of Poisson's theorem is not "free conservation laws forever" but "conserved charges organize themselves into a closed algebra" — and finding that algebra is finding the symmetry group of the dynamics, hidden pieces included.

Why is the bracket antisymmetric — what physics would break with a symmetric pairing?

Antisymmetry encodes two facts you cannot do without. First, conservation of the generator itself: $\dot H = \{H, H\} = 0$ is automatic only because any antisymmetric pairing of a thing with itself vanishes — a symmetric "bracket" would let the energy that generates time evolution drift under its own flow. Second, the action-reaction structure of generation: $\{A, B\} = -\{B, A\}$ says the rate at which $B$ 's flow changes $A$ is minus the rate at which $A$ 's flow changes $B$ . That reciprocity is the algebraic heart of Noether's theorem read canonically: $\{Q, H\} = 0$ means both "time evolution preserves $Q$ " (conservation) and " $Q$ 's flow preserves $H$ " (symmetry) — one equation, two theorems, exchangeable only because of the sign flip. Geometrically, antisymmetry is what makes the pairing a symplectic form (areas, not lengths), and Liouville's theorem, canonical invariance, and ultimately the unitarity of the quantized theory all flow from that. The symmetric analogue of the bracket exists in mathematics — it just does not do mechanics.

Does every conserved quantity generate a symmetry, or only Noether charges?

In the Hamiltonian formalism the implication runs both ways, and this is a genuine strength over the Lagrangian picture. If $\{Q, H\} = 0$ , then by antisymmetry $Q$ 's flow preserves $H$ — so every conserved phase-space functional generates a canonical transformation that is a symmetry of the dynamics. No Lagrangian pedigree required: the Runge-Lenz vector, which has no simple point-transformation origin, generates perfectly good canonical flows (they deform orbits into other orbits of the same energy). The asymmetry with Noether's theorem is one of scope, not validity: Noether's field-theory version (next article) produces not just the charge but a local current $j^\mu$ with $\partial_\mu j^\mu = 0$ — a statement about local bookkeeping in spacetime that the canonical formalism, working on one slice, does not by itself provide. So the honest division of labor: the bracket certifies charge-flow duality on phase space; Noether supplies the local current that makes the charge's conservation a statement about flux through surfaces — and the conserved-currents article shows why that locality is physically non-negotiable.

How literally does the classical charge algebra survive quantization?

Remarkably literally in structure — with celebrated exceptions that became physics in their own right. Dirac's substitution maps $\{Q^A, Q^B\} = f^{ABC}Q^C$ to $[\hat Q^A, \hat Q^B] = i f^{ABC}\hat Q^C$ , and for the Poincaré algebra and simple internal symmetries this holds exactly: the quantum theory represents the same Lie algebra by operators, which is why "particles are irreducible representations of Poincaré" is a sentence that makes sense. The exceptions are anomalies: cases where the quantum theory cannot preserve all classical brackets simultaneously, and central or non-closing terms appear — Schwinger terms in current-current commutators, the central charge of the Virasoro algebra in two dimensions, the chiral anomaly that lets the neutral pion decay to two photons at a rate the classical symmetry would forbid. Far from defects, these controlled failures are among the deepest probes in physics: anomaly cancellation fixes the particle content of the Standard Model generation by generation. The classical algebra of this article is thus the reference structure — what survives is the skeleton; where it bends, precisely and computably, is where the quantum theory says something classical physics cannot.