Skip to content

Poisson Brackets

The Poisson bracket is the canonical antisymmetric product on classical phase-space observables. In canonical coordinates (qi,pi)(q^i,p_i), it is defined by

{f,g}=∑i(∂f∂qi∂g∂pi−∂f∂pi∂g∂qi),\{f,g\} = \sum_i \left( \frac{\partial f}{\partial q^i} \frac{\partial g}{\partial p_i} - \frac{\partial f}{\partial p_i} \frac{\partial g}{\partial q^i} \right),

where f(q,p,t)f(q,p,t) and g(q,p,t)g(q,p,t) are smooth functions on phase space. It packages Hamilton’s equations, constants of motion, infinitesimal canonical transformations, and the classical side of the commutator correspondence.

Poisson brackets matter in quantum mechanics because they are the classical algebraic structure that canonical commutators deform.

They appear when one:

  • passes from a classical Hamiltonian system to canonical quantization;
  • compares Heisenberg equations with classical Hamiltonian flow;
  • derives Ehrenfest and semiclassical correspondence formulas;
  • studies canonical transformations before their quantum unitary analogues;
  • writes phase-space formulations such as Wigner and Moyal brackets;
  • checks whether a proposed “quantization rule” is only a heuristic or a controlled construction.

The bracket is not a quantum commutator with ℏ\hbar removed. It is a classical operation on functions. The relation to quantum mechanics is powerful, but it has ordering, domain, and representation caveats.

For one degree of freedom with coordinates (q,p)(q,p),

{f,g}=∂f∂q∂g∂p−∂f∂p∂g∂q.\{f,g\} = \frac{\partial f}{\partial q} \frac{\partial g}{\partial p} - \frac{\partial f}{\partial p} \frac{\partial g}{\partial q}.

For nn degrees of freedom, the definition is the sum over conjugate pairs:

{f,g}=∑i=1n(∂f∂qi∂g∂pi−∂f∂pi∂g∂qi).\{f,g\} = \sum_{i=1}^n \left( \frac{\partial f}{\partial q^i} \frac{\partial g}{\partial p_i} - \frac{\partial f}{\partial p_i} \frac{\partial g}{\partial q^i} \right).

The bracket takes two classical observables and returns another classical observable. Its value at a phase-space point depends on how the two functions vary in the canonical position and momentum directions.

The fundamental coordinate brackets are

{qi,qj}=0,{pi,pj}=0,{qi,pj}=δji.\{q^i,q^j\}=0, \qquad \{p_i,p_j\}=0, \qquad \{q^i,p_j\}=\delta^i_j.

These are the classical ancestors of the canonical commutation relations.

The Poisson bracket is bilinear:

{αf+βg,h}=α{f,h}+β{g,h},\{\alpha f+\beta g,h\} = \alpha\{f,h\} + \beta\{g,h\},

and similarly in the second slot.

It is antisymmetric:

{f,g}=−{g,f},{f,f}=0.\{f,g\} = -\{g,f\}, \qquad \{f,f\}=0.

It satisfies a product rule in each slot:

{fg,h}=f{g,h}+g{f,h},\{fg,h\} = f\{g,h\} + g\{f,h\},

and

{f,gh}={f,g}h+g{f,h}.\{f,gh\} = \{f,g\}h + g\{f,h\}.

It also satisfies the Jacobi identity:

{f,{g,h}}+{g,{h,f}}+{h,{f,g}}=0.\{f,\{g,h\}\} + \{g,\{h,f\}\} + \{h,\{f,g\}\} = 0.

These identities make smooth phase-space functions into a Lie algebra under the Poisson bracket. The ordinary product of functions is commutative, but the Poisson bracket records the symplectic geometry that makes Hamiltonian mechanics nontrivial.

Hamilton’s equations are equivalent to a single bracket formula. If the Hamiltonian is H(q,p,t)H(q,p,t), then

q˙i={qi,H},p˙i={pi,H}.\dot q^i = \{q^i,H\}, \qquad \dot p_i = \{p_i,H\}.

For any observable f(q,p,t)f(q,p,t) along a Hamiltonian trajectory,

dfdt={f,H}+∂f∂t.\frac{df}{dt} = \{f,H\} + \frac{\partial f}{\partial t}.

Thus the Hamiltonian is the generator of time evolution with respect to the Poisson bracket. This is the classical counterpart of the Heisenberg-picture equation

dAHdt=iℏ[H,AH]+(∂AS∂t)H.\frac{dA_H}{dt} = \frac{i}{\hbar}[H,A_H] + \left(\frac{\partial A_S}{\partial t}\right)_H.

The signs agree when the commutator correspondence is written as

{f,H}⟷1iℏ[f^,H^],\{f,H\} \quad \longleftrightarrow \quad \frac{1}{i\hbar}[\hat f,\hat H],

or equivalently

1iℏ[f^,H^]=iℏ[H^,f^].\frac{1}{i\hbar}[\hat f,\hat H] = \frac{i}{\hbar}[\hat H,\hat f].

If ff has no explicit time dependence, then

dfdt={f,H}.\frac{df}{dt} = \{f,H\}.

So ff is conserved along Hamiltonian flow if

{f,H}=0.\{f,H\}=0.

This is the classical version of the quantum conservation test [A,H]=0[A,H]=0 for an observable with no explicit time dependence.

Energy conservation is the simplest example. Since {H,H}=0\{H,H\}=0,

dHdt=∂H∂t.\frac{dH}{dt} = \frac{\partial H}{\partial t}.

If HH has no explicit time dependence, the Hamiltonian is conserved.

For a free particle in one dimension,

H(x,p)=p22m.H(x,p) = \frac{p^2}{2m}.

The coordinate brackets give

x˙={x,H}=∂H∂p=pm,\dot x = \{x,H\} = \frac{\partial H}{\partial p} = \frac{p}{m},

and

p˙={p,H}=−∂H∂x=0.\dot p = \{p,H\} = - \frac{\partial H}{\partial x} = 0.

Thus the Poisson bracket reproduces the horizontal phase-space trajectories of the free particle.

For the harmonic oscillator,

H(x,p)=p22m+12mω2x2.H(x,p) = \frac{p^2}{2m} + \frac12m\omega^2x^2.

The brackets are

x˙={x,H}=pm,p˙={p,H}=−mω2x.\dot x = \{x,H\} = \frac{p}{m}, \qquad \dot p = \{p,H\} = -m\omega^2x.

Combining them gives

x¨+ω2x=0.\ddot x+\omega^2x=0.

In this example the bracket is not an extra assumption. It is a compact way to express the same phase-space flow already encoded in Hamilton’s equations.

Generators of Infinitesimal Transformations

Section titled “Generators of Infinitesimal Transformations”

A phase-space function G(q,p)G(q,p) generates an infinitesimal transformation by

δf=ϵ{f,G}\delta f = \epsilon\{f,G\}

for small parameter ϵ\epsilon. Applying this to coordinates gives

δqi=ϵ{qi,G}=ϵ∂G∂pi,\delta q^i = \epsilon\{q^i,G\} = \epsilon \frac{\partial G}{\partial p_i},

and

δpi=ϵ{pi,G}=−ϵ∂G∂qi.\delta p_i = \epsilon\{p_i,G\} = - \epsilon \frac{\partial G}{\partial q^i}.

This is why the Hamiltonian generates time translations, momentum generates spatial translations, and angular momentum generates rotations.

For example, in two spatial dimensions let

Lz=xpy−ypx.L_z = xp_y-yp_x.

Then

{x,Lz}=−y,{y,Lz}=x,\{x,L_z\} = -y, \qquad \{y,L_z\} = x,

and

{px,Lz}=−py,{py,Lz}=px.\{p_x,L_z\} = -p_y, \qquad \{p_y,L_z\} = p_x.

Up to the sign convention for the transformation parameter, this is the infinitesimal rotation of positions and momenta in the plane.

Canonical quantization often starts from the correspondence

{f,g}⟶1iℏ[f^,g^].\{f,g\} \quad \longrightarrow \quad \frac{1}{i\hbar}[\hat f,\hat g].

The coordinate brackets become

{qi,pj}=δji⇝[Qi,Pj]=iℏδjiI.\{q^i,p_j\} = \delta^i_j \quad \leadsto \quad [Q^i,P_j] = i\hbar\delta^i_j I.

This is the algebraic bridge to the Canonical Commutation Relations and the Heisenberg Group.

The caveat is essential. Classical functions commute under ordinary multiplication, while quantum operators need not commute. A classical product such as q2p2q^2p^2 does not uniquely determine whether the quantum operator should be Q2P2Q^2P^2, PQ2PP Q^2 P, a symmetrized expression, or something else. Unbounded operators also require domains. For this reason, Poisson brackets guide quantization but do not make it automatic.

From Phase Space to Canonical Quantization extends this correspondence to functional field brackets, equal-time commutators, and constrained systems.

The inverse direction is also subtle. In semiclassical analysis, commutators of well-behaved quantum observables often have leading classical behavior

1iℏ[f^,g^]∼{f,g}^\frac{1}{i\hbar}[\hat f,\hat g] \sim \widehat{\{f,g\}}

as ℏ\hbar becomes small relative to the relevant action scale, but higher-order terms can matter. Phase-space formulations make this precise through the Moyal bracket in appropriate settings.

In canonical coordinates, the Poisson bracket is built from the standard symplectic form

ω=∑idqi∧dpi.\omega = \sum_i dq^i\wedge dp_i.

The Hamiltonian vector field XfX_f associated with a function ff is the vector field whose flow is generated by ff. In canonical coordinates,

Xf=∑i(∂f∂pi∂∂qi−∂f∂qi∂∂pi).X_f = \sum_i \left( \frac{\partial f}{\partial p_i} \frac{\partial}{\partial q^i} - \frac{\partial f}{\partial q^i} \frac{\partial}{\partial p_i} \right).

Then the bracket can be read as

{g,f}=Xf(g).\{g,f\} = X_f(g).

Canonical Transformations, Symplectic Vector Spaces, and Symplectic Manifolds, First Look develop this structure-preserving viewpoint. For most quantum-mechanics calculations, the canonical-coordinate formula is the workhorse.

  • Confusing the Poisson bracket {f,g}\{f,g\} with the quantum anticommutator {A,B}=AB+BA\{A,B\}=AB+BA.
  • Treating the rule {f,g}↦[f^,g^]/(iℏ)\{f,g\}\mapsto [\hat f,\hat g]/(i\hbar) as an exact quantization algorithm.
  • Forgetting the minus sign in the pp equation, p˙i={pi,H}\dot p_i=\{p_i,H\}.
  • Computing brackets as if qiq^i and pip_i were independent labels but not conjugate variables.
  • Assuming a change of variables is canonical just because it is invertible.
  • Ignoring explicit time dependence in df/dt={f,H}+∂f/∂tdf/dt=\{f,H\}+\partial f/\partial t.
  • Forgetting that Poisson brackets are classical functions, not operators.
  • H. Goldstein, C. Poole, and J. Safko, Classical Mechanics, 3rd ed., Addison-Wesley, 2002.
  • V. I. Arnold, Mathematical Methods of Classical Mechanics, 2nd ed., Springer, 1989.
  • R. Abraham and J. E. Marsden, Foundations of Mechanics, 2nd ed., AMS Chelsea, 2008.
  • L. D. Landau and E. M. Lifshitz, Mechanics, 3rd ed., Butterworth-Heinemann, 1976.
  • P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford University Press, 1958.
  • M. de Gosson, Symplectic Geometry and Quantum Mechanics, Birkhäuser, 2006.
  • B. C. Hall, Quantum Theory for Mathematicians, Springer, 2013.
  1. Compute the fundamental Poisson brackets {q,p}\{q,p\}, {q,q}\{q,q\}, and {p,p}\{p,p\} for one degree of freedom.
Solution

Using

{f,g}=∂f∂q∂g∂p−∂f∂p∂g∂q,\{f,g\} = \frac{\partial f}{\partial q} \frac{\partial g}{\partial p} - \frac{\partial f}{\partial p} \frac{\partial g}{\partial q},

one has

{q,p}=1,{q,q}=0,{p,p}=0.\{q,p\}=1, \qquad \{q,q\}=0, \qquad \{p,p\}=0.
  1. Let H=p2/(2m)+V(q)H=p^2/(2m)+V(q). Use Poisson brackets to derive Hamilton’s equations.
Solution

For the coordinate,

q˙={q,H}=∂H∂p=pm.\dot q = \{q,H\} = \frac{\partial H}{\partial p} = \frac{p}{m}.

For the momentum,

p˙={p,H}=−∂H∂q=−V′(q).\dot p = \{p,H\} = - \frac{\partial H}{\partial q} = - V'(q).

These are Hamilton’s equations for the one-dimensional potential problem.

  1. Show that if ff and HH have no explicit time dependence and {f,H}=0\{f,H\}=0, then ff is constant along the Hamiltonian trajectory.
Solution

The total derivative along the trajectory is

dfdt={f,H}+∂f∂t.\frac{df}{dt} = \{f,H\} + \frac{\partial f}{\partial t}.

Both terms vanish by assumption, so df/dt=0df/dt=0. Therefore ff is conserved along the motion.

  1. In two dimensions, use Lz=xpy−ypxL_z=xp_y-yp_x to compute {x,Lz}\{x,L_z\} and {py,Lz}\{p_y,L_z\}.
Solution

Using canonical coordinates (x,y,px,py)(x,y,p_x,p_y),

{x,Lz}=∂Lz∂px=−y.\{x,L_z\} = \frac{\partial L_z}{\partial p_x} = -y.

Also,

{py,Lz}=−∂Lz∂y=px.\{p_y,L_z\} = - \frac{\partial L_z}{\partial y} = p_x.

These are two components of the infinitesimal rotation generated by LzL_z, up to the sign convention for the small rotation parameter.

  1. Why does {q,p}=1\{q,p\}=1 not by itself determine a unique quantum theory?
Solution

The bracket suggests the canonical commutation relation [Q,P]=iℏI[Q,P]=i\hbar I, but a quantum theory still needs a Hilbert space, a representation of QQ and PP, domains for unbounded operators, a self-adjoint Hamiltonian, boundary conditions, and choices for ordering noncommuting products. Different choices can agree on the leading classical bracket while differing in quantum predictions.