Symmetries and Conservation Laws in Hamiltonian Form

Noether's theorem is usually told as a construction job: spot a continuous symmetry of the action, then follow a recipe through the Euler–Lagrange machinery until a conserved charge falls out. The recipe works — Noether's Theorem walks it end to end on the Lagrangian side — but it runs one way. Symmetry is the input, conservation the output, and the two live in different mathematical houses: transformations act on paths, charges are functions of the state. Hamiltonian mechanics collapses the two houses into one room. On phase space, any observable $G$ is simultaneously a quantity you can measure and the generator of a flow — turn the crank by $\epsilon$ and every observable shifts by $\epsilon\{f,G\}$ . Ask how the Hamiltonian responds to $G$ 's flow and you get $\{H,G\}$ ; ask how $G$ responds to time and you get $\{G,H\}$ . These are the same bracket with its arguments swapped, and the Poisson bracket is antisymmetric. So " $G$ leaves $H$ unchanged" and "time leaves $G$ unchanged" are one equation read in two directions. Symmetry does not imply conservation here; symmetry is conservation, seen from the other slot of the bracket. This page develops that reciprocity into a working toolkit: the two-way theorem itself, the dictionary of standard generators, conserved charges that carry explicit clocks (the Galilean boost, with the mass appearing as a central charge), symmetry algebras rather than isolated laws, and — just as useful — the bracket as a quantitative meter for how badly a broken symmetry fails.

Key takeaways

Any observable $G$ generates a canonical flow $\delta f=\epsilon\{f,G\}$ ; the Hamiltonian's response is $\delta H=\epsilon\{H,G\}$ .
Antisymmetry converts that response into $-\epsilon\{G,H\}=-\epsilon\,\dot G$ (for time-independent $G$ ): the flow of $G$ preserves $H$ iff the flow of $H$ preserves $G$ .
The general criterion is $\dot G=\{G,H\}+\partial_t G=0$ — explicitly time-dependent generators like the boost $G=pt-mx$ can be conserved even though $\{G,H\}\neq0$ .
Conserved charges close under the bracket (Poisson's theorem), so symmetries arrive as Lie algebras — $\mathfrak{so}(3)$ for rotations, the Galilei algebra with its mass central charge $\{K_i,P_j\}$ for free particles.
When the symmetry is broken, the same bracket quantifies the damage: $\dot G=\{G,H\}$ is the leak rate — for $L_z$ it is literally the torque.
Key idea: symmetry and conservation are one antisymmetric bracket equation, $\delta H/\epsilon=\{H,G\}=-\{G,H\}=-\dot G$ , read from its two slots — the Hamiltonian face of Noether's theorem.

The core insight

On phase space every observable is also a generator, and the antisymmetry of the Poisson bracket makes symmetry and conservation the same statement: $G$ 's flow leaves $H$ invariant exactly when $H$ 's flow (time evolution) leaves $G$ invariant.

Where this lives today

The reciprocity theorem is production machinery wherever equations are solved by exploiting structure. Lattice gauge theory codes verify their discretizations by checking that charge conservation holds as a bracket identity — Gauss-law constraints commuting with the discrete Hamiltonian — before a single physics run; when the check fails, the symmetry was broken by the discretization, and the theorem says exactly which charge will drift. The same logic drives the current generation of machine-learned interatomic potentials and "Hamiltonian neural networks": build translation and rotation invariance into the learned $H$ and the theorem guarantees momentum and angular-momentum conservation of the simulated dynamics, for free, forever — enforcing the symmetry is strictly cheaper than penalizing the leak. At the other end of the energy scale, every LHC analysis is organized by the theorem's quantum image: conserved charges are operators commuting with $\hat H$ , so events are classified by the eigenvalues — energy, momentum, angular momentum, and the internal charges of the Standard Model's symmetry algebra. And when experimentalists quantify symmetry violation (CP asymmetries, baryon-number searches), they are reading precisely the meter of equation (5): the measured rate at which a would-be charge leaks is the measured size of the symmetry-breaking term in the Hamiltonian.

The physical picture

Picture phase space with two families of level curves drawn on it: the energy surfaces of $H$ , and the level sets of some observable $G$ . Each function generates a flow that slides phase points along its own level curves — that is the machinery of Infinitesimal Canonical Transformations, viewed as standing equipment. Now the whole theorem is a picture: if $G$ 's flow happens to slide points along the energy surfaces (never across), then $H$ is invariant under $G$ — a symmetry. But "the flows of $G$ and $H$ each stay on the other's level sets" is a single geometric relationship between the two families of curves. Time evolution is just $H$ 's flow, so the same relationship, read the other way, says the dynamics never carries the system across a level set of $G$ : conservation. One picture, two captions.

Figure

The two-way street — the flow of G slides along energy surfaces while the flow of H slides along level sets of G, and both statements are the same geometric fact

Contrast this with the Lagrangian side. There, a symmetry is a transformation of trajectories and a charge is a function built from a recipe — two different kinds of object, related by a theorem with an actual construction in the middle. Here both are the single object $G$ . The charge does not merely follow from the symmetry; the charge generates the symmetry, and the symmetry condition is the charge's own conservation law. That is why physicists say quantum mechanics inherited its symmetry machinery from Hamiltonian, not Lagrangian, mechanics: replace the bracket by a commutator and this entire page transcribes line for line. The reciprocity also pays when the symmetry is imperfect. Since $\dot G=\{G,H\}$ for clock-free $G$ , the bracket is a gauge needle: zero means exact symmetry, and any nonzero value is the measured leak rate of the would-be charge. For angular momentum the needle reads in newton-metres — it is the torque — and you can watch it swing to zero as a perturbation is switched off.

Figure

A helical channel — neither rotation alone nor translation alone leaves the potential invariant, but the locked screw combination does, and one combined charge survives

The reciprocity theorem

Everything runs on one recalled fact from Infinitesimal Canonical Transformations: the flow generated by $G$ with parameter $\epsilon$ shifts any observable by

\delta f=\epsilon\,\{f,G\} \qquad (1)

where: $f$ — any phase-space function; $G$ — the generator, itself an ordinary observable; $\epsilon$ — the infinitesimal flow parameter (an angle for $G=L_z$ , a displacement for $G=p$ , a time for $G=H$ ). The F2 construction behind (1) lives in the companion article and is not repeated here. Set $f=H$ and the Hamiltonian's response to the transformation is a bracket:

\delta H=\epsilon\,\{H,G\} \qquad (2)

where: $\delta H$ — the first-order change in the energy function when every phase point is pushed one $\epsilon$ -step along $G$ 's flow. The transformation is a symmetry of the Hamiltonian when this vanishes. So far this is bookkeeping; the physics enters through antisymmetry.

\frac{\delta H}{\epsilon}=\{H,G\}=-\{G,H\}=-\frac{\mathrm{d}G}{\mathrm{d}t}\qquad(\partial_t G=0) \qquad (3)

where: the last equality is the evolution equation $\dot G=\{G,H\}+\partial_t G$ of Poisson Brackets: Definition, specialized to a clock-free $G$ . Equation (3) is the whole theorem: the number that measures how much $G$ 's flow tilts the energy landscape is the negative of the number that measures how fast time drains the charge. Either both vanish or neither does.

Derivation ASymmetry equals conservation from bracket antisymmetry

We are proving the reciprocity theorem (3): a time-independent observable $G$ generates a symmetry of $H$ if and only if $G$ is a constant of the motion. The strategy is to compute the change of $H$ under $G$ 's flow by the chain rule, recognize the result as a Poisson bracket, and then let antisymmetry swap the roles of the two functions. Watch how short the "proof" of a famous theorem becomes once the generator formalism is standing equipment — and note precisely where time-independence of $G$ is used, because that is the loophole the boost will exploit later.

Step 1. Vary $H$ along the flow. Under $\delta q_i=\epsilon\{q_i,G\}=\epsilon\,\partial G/\partial p_i$ and $\delta p_i=\epsilon\{p_i,G\}=-\epsilon\,\partial G/\partial q_i$ , the chain rule gives

\delta H=\sum_i\left(\frac{\partial H}{\partial q_i}\,\delta q_i+\frac{\partial H}{\partial p_i}\,\delta p_i\right)=\epsilon\sum_i\left(\frac{\partial H}{\partial q_i}\frac{\partial G}{\partial p_i}-\frac{\partial H}{\partial p_i}\frac{\partial G}{\partial q_i}\right)

Step 2. Recognize the bracket: the parenthesis is exactly $\{H,G\}$ , so $\delta H=\epsilon\{H,G\}$ — equation (2). Why this needed no new input: the same chain-rule collapse defines the bracket itself; (2) is just the statement "the bracket measures rates along flows" with $G$ driving and $H$ riding, the mirror image of time evolution.

Step 3. Swap the slots. Antisymmetry gives $\{H,G\}=-\{G,H\}$ , and for $G$ with no explicit time dependence the evolution equation reads $\dot G=\{G,H\}$ . Therefore

\delta H=-\epsilon\,\dot G

Step 4. Read it both ways. If the transformation is a symmetry, $\delta H=0$ , then $\dot G=0$ : the generator is conserved. If $G$ is conserved, $\dot G=0$ , then $\delta H=0$ : the flow it generates never changes the energy. One vanishing bracket, both theorems. Why the Lagrangian version cannot be this symmetric: on configuration space a symmetry is a map of paths and a charge is a function of $(q,\dot q)$ — different types, so the implication needs Noether's construction as a bridge. On phase space both are the single function $G$ ; the bridge has contracted to the antisymmetry of one bilinear form. Verified: matches Goldstein §9.6 and Landau–Lifshitz §40; consistency check — setting $G=H$ gives $\delta H=\epsilon\{H,H\}=0$ , time-translation symmetry, and simultaneously $\dot H=0$ , energy conservation for autonomous systems.

The standard conservation laws are now dictionary entries, each one a choice of $G$ whose flow you already know from Momentum and Angular Momentum as Generators and Hamiltonian as Generator of Time Evolution:

Generator $G$	Flow it generates	$\{H,G\}=0$ means	Conserved charge
$P_x=\sum_a p_{x,a}$	rigid translation of all particles	$H$ built from coordinate differences only	total momentum
$L_z$	rigid rotation of positions and momenta about $z$	$H$ rotationally invariant about $z$	angular momentum
$H$	time evolution itself	$\partial_t H=0$	energy
$K=mx-pt$ (clocked)	Galilean boost, velocity shift	full criterion $\{K,H\}+\partial_t K=0$	center-of-mass launch point
$p_j$ with $q_j$ cyclic	translation of the cyclic coordinate	$\partial H/\partial q_j=0$	the memorized freshman rule

The last row is worth a sentence: "cyclic coordinate $\Rightarrow$ conserved momentum" from Cyclic Coordinates and Conserved Momenta is the diagonal special case of this page — the symmetry flow is translation of $q_j$ itself, and $\partial H/\partial q_j=0$ is $\{H,p_j\}=0$ written in coordinates. The bracket version is what survives when the symmetry does not line up with any single coordinate axis, as in the screw example below.

Charges that carry a clock

Equation (3) used $\partial_t G=0$ in its final step. The honest, fully general criterion — straight from the evolution equation — is

\frac{\mathrm{d}G}{\mathrm{d}t}=\{G,H\}+\frac{\partial G}{\partial t}=0 \qquad (4)

where: $\partial G/\partial t$ — the explicit clock dependence of $G$ 's formula, over and above the motion of the phase point. A generator can fail the naive test $\{G,H\}=0$ and still be exactly conserved because its built-in clock ticks at precisely the compensating rate. This is not a curiosity: the most physical charge of a free particle — where it started — is of this clocked type.

Go deeper · Derivation B · 2 minThe Galilean boost charge and its mass central charge

Derivation BThe Galilean boost charge and its mass central charge

We are showing that the boost generator $G=pt-mx$ of a free particle $H=p^2/2m$ is conserved even though $\{G,H\}\neq0$ , and then extracting the strangest feature of the Galilei algebra: the bracket of a boost with a translation is not another symmetry generator but the mass itself. The strategy is direct computation of both terms of criterion (4), then one more bracket for the central charge. Keep track of signs — two conventions for the boost circulate, and we fix ours explicitly.

Step 1. The bracket term. Only the $-mx$ piece sees $H$ :

\{G,H\}=\{pt-mx,\ \tfrac{p^2}{2m}\}=-m\,\{x,\tfrac{p^2}{2m}\}=-m\cdot\frac{p}{m}=-p

using $\{x,p^2\}=2p\{x,p\}=2p$ by the Leibniz rule. Nonzero: the boost flow genuinely tilts the free-particle Hamiltonian at any frozen instant.

Step 2. The clock term. The formula $G=pt-mx$ contains $t$ explicitly, so $\partial G/\partial t=p$ .

Step 3. Total: $\dot G=\{G,H\}+\partial_t G=-p+p=0$ . Exactly conserved. Physically, $G=pt-mx=-m\left(x-\tfrac{p}{m}t\right)=-m\,x_0$ : minus the mass times the launch position. Inertial motion never forgets where it started, and that memory is the boost charge. Why the cancellation is not fine-tuning: the boost is a symmetry of the dynamics (Galilean relativity) but not of the instantaneous energy — boosting changes kinetic energy. The clocked charge is precisely how Hamiltonian mechanics accommodates symmetries that shuffle energy surfaces instead of preserving them; the Lagrangian-side counterpart is the total-derivative freedom in Noether's Theorem, whose boost charge this $G$ reproduces.

Step 4. The central charge. Define the conventional boost $K\equiv-G=mx-pt$ (so that $K/m=x_0$ is the launch point with a plus sign) and bracket it with the translation generator $p$ :

\{K,p\}=\{mx-pt,\ p\}=m\,\{x,p\}=m

A pure number — not zero, and not another generator. In $n$ particles and three dimensions the same computation gives $\{K_i,P_j\}=M\delta_{ij}$ with $M$ the total mass. Boosts and translations almost commute; their failure to do so is the system's total inertia. Why this matters beyond classical mechanics: an algebra that closes only "up to a constant" is a centrally extended Lie algebra. Quantized, this central term is why non-relativistic quantum mechanics superselects mass (Bargmann's rule: no coherent superpositions of different masses) — a deep quantum fact whose entire content is already sitting in $\{K,p\}=m$ . Verified: Arnold §40 and Landau–Lifshitz §8 (Galilei invariants); sign convention cross-check — with $K=mx-pt$ , $\dot K=\{K,H\}+\partial_t K=p-p=0$ , matching Step 3 with the overall sign flipped, and the provenance form $G=pt-mx$ gives $\{G,p\}=-m$ : same physics, opposite sign, so state your convention before comparing with any text.

Symmetries come in algebras

A single conserved charge is rarely alone. If $G_1$ and $G_2$ both commute with $H$ , then Poisson's theorem — proved via the Jacobi identity in Algebra of Poisson Brackets, and only cited here — says $\{G_1,G_2\}$ commutes with $H$ too. The set of all symmetry generators is therefore closed under the bracket: symmetries form a Lie algebra, with the bracket as its product. Rotations are the canonical example: the closure $\{L_i,L_j\}=\epsilon_{ijk}L_k$ (derived from fundamentals in the algebra article) says the bracket of two conserved rotation generators is the third — the $\mathfrak{so}(3)$ structure constants realized on phase-space functions. For the free particle, assembling $H$ , $\mathbf P$ , $\mathbf L$ , $\mathbf K$ and computing all mutual brackets yields the full Galilei algebra, complete with the mass central charge of the derivation above; the relativistic particle yields the Poincaré algebra the same way. Group theory has quietly entered mechanics through the bracket's back door — this is the structure Generators of Continuous Transformations develops for general flows, now restricted to flows that happen to be symmetries. Two practical corollaries. First, the algebra sometimes manufactures charges: knowing $L_x$ and $L_y$ are conserved forces $L_z$ 's conservation for free. Second, the algebra polices how many charges you can use at once: since $\{L_x,L_y\}\neq0$ , their flows are incompatible, and only mutually commuting ("in involution") charges can be simultaneously turned into momenta of one canonical chart — the counting that decides integrability, taken up in Constants of Motion via Poisson Brackets and consummated in Hamilton–Jacobi theory.

When the symmetry is broken, the bracket is the meter

The reciprocity theorem has a quantitative converse that is used daily and stated rarely. For clock-free $G$ ,

\frac{\mathrm{d}G}{\mathrm{d}t}=\{G,H\} \qquad (5)

where: the right side, evaluated at the current phase point, is the instantaneous leak rate of the almost-conserved quantity. Nothing requires it to vanish; whatever it equals is how fast the charge drains. For $G=L_z$ this number is the torque in disguise — Poisson Brackets: Definition computed it as $-3kxy$ for a $1{:}4$ anisotropic trap — and the worked example below runs the same logic on a different symmetry-breaking term with the cross-check made explicit. The mindset shift is the point: do not sort Hamiltonians into "symmetric" and "not symmetric"; compute the bracket and read the needle.

Go deeper · Derivation C · 1 minThe screw charge of a helical potential

Derivation CThe screw charge of a helical potential

We are constructing the conserved charge of a screw symmetry — a potential invariant under neither rotations nor translations separately, but exactly invariant under a locked combination of both, like a particle threading an idealized helical channel of pitch $h$ . The goal is to verify $\{G,H\}=0$ for $G=p_\phi+\tfrac{h}{2\pi}\,p_z$ by honest partial derivatives, showing how the bracket formalism handles symmetries that no single coordinate makes cyclic. The strategy: write $H$ in cylindrical coordinates, encode the screw invariance as a functional dependence of $V$ , and let the chain rule do the rest.

Step 1. Set up. In cylindrical coordinates $(r,\phi,z)$ with conjugate momenta $(p_r,p_\phi,p_z)$ ,

H=\frac{p_r^{2}}{2m}+\frac{p_\phi^{2}}{2mr^{2}}+\frac{p_z^{2}}{2m}+V(r,\phi,z),\qquad V=U\!\left(r,\ \phi-\frac{2\pi}{h}z\right)

The single combined argument $s\equiv\phi-\tfrac{2\pi}{h}z$ is the screw symmetry: advance $\phi$ by $\mathrm{d}\phi$ while sliding $z$ by $\mathrm{d}z=\tfrac{h}{2\pi}\mathrm{d}\phi$ and $s$ — hence $V$ — is unchanged. One full turn advances $z$ by the pitch $h$ .

Step 2. Bracket the candidate charge $G=p_\phi+\tfrac{h}{2\pi}p_z$ with $H$ . Since $G$ is built purely from momenta, only the $\partial V/\partial q$ terms of the bracket survive:

\{G,H\}=\{p_\phi,H\}+\frac{h}{2\pi}\{p_z,H\}=-\frac{\partial V}{\partial \phi}-\frac{h}{2\pi}\frac{\partial V}{\partial z}

Why the kinetic terms drop out: $p_\phi$ and $p_z$ bracket to zero with every momentum and with $r$ -dependent factors, because $\phi$ and $z$ appear nowhere in the kinetic energy — the centrifugal factor $1/r^2$ involves only $r$ .

Step 3. Chain rule through the screw variable. With $U'\equiv\partial U/\partial s$ : $\ \partial V/\partial\phi=U'\cdot1$ and $\partial V/\partial z=U'\cdot\left(-\tfrac{2\pi}{h}\right)$ . Substitute:

\{G,H\}=-U'+\frac{h}{2\pi}\cdot\frac{2\pi}{h}\,U'=-U'+U'=0

Step 4. Interpret. Neither $p_\phi$ ( $=L_z$ ) nor $p_z$ is conserved alone — each has $\dot p=\mp$ a nonzero force component — but the pitch-weighted sum rides through untouched. The helix trades angular momentum for linear momentum at the fixed exchange rate $\tfrac{2\pi}{h}$ , and $G$ is the invariant balance of that account. Verified: limiting checks — $h\to\infty$ (channel straightens into a groove along $z$ ): the screw variable becomes $\phi$ alone, $G\to p_\phi$ , plain rotational symmetry; $h\to0$ with $U$ smooth forces $U'\to0$ contributions to rearrange into pure $z$ -translation symmetry, $G$ dominated by $p_z$ . Both match intuition; algebra matches Goldstein's treatment of symmetry-adapted momenta (§9.6).

Limiting cases

Case	Result	Meaning
$G=H$ , autonomous	$\delta H=\epsilon\{H,H\}=0$ always	time translation is automatically a symmetry; energy conservation is reflexive
$G=p_j$ , $q_j$ cyclic	$\{H,p_j\}=-\partial H/\partial q_j=0$	the freshman rule is the diagonal entry of the theorem
Central potential	$\{H,L_i\}=0$ for all three $i$	full $\mathfrak{so}(3)$ of charges, only $L^2$ and one $L_i$ usable at once
Clocked generator	$\{G,H\}\neq0$ yet $\dot G=0$	boost-type charges; the clock term pays the bracket's bill
Broken symmetry	$\dot G=\{G,H\}\neq0$	the bracket reads the leak rate — for $L_z$ , literally the torque
$\hbar$ -side image	$\{G,H\}\to[\hat G,\hat H]/i\hbar$	conserved charges become operators commuting with $\hat H$ — good quantum numbers

Common trapIf the bracket with H vanishes it is conserved — not when the charge carries a clock, and not the only route to conservation

The two failure modes of the slogan are mirror images. First: $\{G,H\}=0$ certifies conservation only for $G$ with no explicit time dependence — for a clocked observable the criterion is the full $\{G,H\}+\partial_t G=0$ , and quoting the bracket alone gets the boost charge wrong in both directions ( $\{K,H\}=p\neq0$ , yet $K$ is exactly conserved). Second, the converse trap: finding $\{G,H\}\neq0$ does not mean "no symmetry here" — it means this $G$ is not the generator; a clocked completion (boost) or a combination with other momenta (the screw charge $p_\phi+\tfrac{h}{2\pi}p_z$ ) may still commute with the dynamics. The reliable procedure is always the same: write the candidate with any explicit $t$ it needs, compute both terms of (4), and let the algebra — not the slogan — deliver the verdict.

Worked examples

1 · Reading the symmetry-breaking meter (confidence builder)

A particle of mass $m$ moves in the plane in $V=\tfrac{k}{2}\left(x^{2}+y^{2}\right)+b\,xy$ with $k=5.0\ \mathrm{N/m}$ and $b=0.80\ \mathrm{N/m}$ . At the point $x=0.30\ \mathrm{m}$ , $y=0.50\ \mathrm{m}$ (any momenta), find the instantaneous leak rate of $L_z$ , cross-check it against the torque, and identify the limit in which conservation is restored.

Step 1. Split the bracket. $\dot L_z=\{L_z,H\}$ with $H=T+\tfrac{k}{2}r^2+bxy$ . The kinetic term is rotationally invariant, so $\{L_z,T\}=0$ ; the isotropic $\tfrac{k}{2}(x^2+y^2)=\tfrac{k}{2}r^2$ term is too, so $\{L_z,\tfrac{k}{2}r^2\}=0$ . Only the cross term can torque.

Step 2. Bracket the perturbation by the Leibniz rule, using $\{L_z,x\}=+y$ and $\{L_z,y\}=-x$ (the rotation flow of Momentum and Angular Momentum as Generators, signs flipped through antisymmetry):

\{L_z,\,bxy\}=b\left(\{L_z,x\}\,y+x\,\{L_z,y\}\right)=b\left(y^{2}-x^{2}\right)

Step 3. Numbers: $\dot L_z=0.80\times\left(0.50^{2}-0.30^{2}\right)=0.80\times(0.25-0.09)=0.80\times0.16=\mathbf{+0.128\ N\,m}$ . Note $k$ has dropped out entirely — the isotropic part of the trap, however stiff, contributes zero leak.

Step 4. Cross-check with elementary mechanics. The force is $\mathbf F=-\nabla V=(-kx-by,\ -ky-bx)$ , and the torque about the origin is

\tau_z=xF_y-yF_x=x(-ky-bx)-y(-kx-by)=-bx^{2}+by^{2}=b\left(y^{2}-x^{2}\right)=+0.128\ \mathrm{N\,m}\ \checkmark

(the $-kxy$ and $+kxy$ pieces cancel identically). The bracket computed the torque with no cross product in sight, because for $G=L_z$ equation (5) is the torque equation.

Step 5. Restore the symmetry: as $b\to0$ the meter winds down to zero and $L_z$ is conserved — the reciprocity theorem watched continuously, not just at the symmetric point. On the diagonal lines $y=\pm x$ the leak also vanishes instantaneously even with $b\neq0$ : the perturbing force there points radially. A symmetry argument alone would never hand you that local detail.

2 · Banking on the screw charge (exam level)

A bead of mass $m=0.50\ \mathrm{kg}$ moves through a helical guide field of pitch $h=0.20\ \mathrm{m}$ whose potential has the screw form $V=U\!\left(r,\ \phi-\tfrac{2\pi}{h}z\right)$ . At time $t_1$ its angular momentum is $p_\phi=L_z=0.40\ \mathrm{J\,s}$ and its linear momentum along the axis is $p_z=1.2\ \mathrm{kg\,m/s}$ . Later, at $t_2$ , interaction with the channel wall has reduced $L_z$ to $0.25\ \mathrm{J\,s}$ . Find $p_z$ at $t_2$ , and check the bookkeeping.

Step 1. Identify the conserved object. Neither $L_z$ nor $p_z$ commutes with this $H$ , but the screw charge does (Derivation C): $G=L_z+\dfrac{h}{2\pi}\,p_z$ , with exchange rate $\dfrac{h}{2\pi}=\dfrac{0.20}{2\pi}=0.031831\ \mathrm{m}$ .

Step 2. Evaluate the charge at $t_1$ :

G=0.40+0.031831\times1.2=0.40+0.038197=\mathbf{0.43820\ J\,s}

Step 3. Conservation at $t_2$ : $0.43820=0.25+0.031831\,p_z(t_2)$ , so

p_z(t_2)=\frac{0.43820-0.25}{0.031831}=\frac{0.18820}{0.031831}=\mathbf{5.912\ kg\,m/s}

Step 4. Sanity checks. Direction: the bead lost angular momentum, so the screw converts it into forward momentum — $p_z$ jumped from $1.2$ to $5.9\ \mathrm{kg\,m/s}$ , a factor set entirely by the steep exchange rate $2\pi/h=31.42\ \mathrm{m^{-1}}$ (a tight pitch is a low gear: small angular change, large linear payout). Units: $\mathrm{J\,s}=\mathrm{kg\,m^2/s}$ , divided by metres gives $\mathrm{kg\,m/s}$ ✓. Limits: $h\to\infty$ makes the exchange rate vanish — the charge degenerates to $L_z$ alone and no conversion is possible, matching the straight-groove limit of Derivation C.

Step 5. The moral for exams: when a problem's geometry couples two standard symmetries, do not hunt for conserved $L_z$ or $p_z$ separately — build the combination the bracket certifies, and treat the pitch (or its analogue) as a fixed exchange rate between the two accounts.

Origins · 1 minEmmy Noether and how this was discovered

Emmy Noether

1882–1935 · Göttingen and Bryn Mawr

Noether proved her symmetry–conservation theorems in 1918, answering a question Hilbert and Klein could not settle: why does energy conservation misbehave in general relativity? Her paper sorted the matter with two theorems of shocking generality — and she regarded it as a minor detour from the abstract algebra that was her life's work. The detour outgrew the century. Denied a paid position at Göttingen for years (Hilbert famously fumed that a university senate was "not a bathhouse"), she habilitated only in 1919, built modern abstract algebra through the 1920s while her "Noether boys" spread it worldwide, and was expelled as a Jew in 1933, finishing her career at Bryn Mawr before dying suddenly at 53. She worked on the Lagrangian side; the Hamiltonian reciprocity of this page is her theorem reborn inside the bracket formalism, where the charge she taught physicists to construct turns out to be the symmetry it certifies. Einstein's obituary called her "the most significant creative mathematical genius thus far produced since the higher education of women began" — a sentence that manages to be both true and an indictment of its era.

History

1788

— Lagrange's Mécanique analytique ties cyclic coordinates to conserved momenta, the theorem's diagonal case

1834

— Hamilton recasts dynamics on phase space, making observables and transformations the same kind of object

1842

— Jacobi's lectures develop canonical transformations and note conservation laws arriving in related families

1918

— Noether proves the general symmetry–conservation theorems on the Lagrangian side, resolving the GR energy puzzle

1925

— Dirac's bracket-commutator dictionary carries the reciprocity theorem verbatim into quantum mechanics

1974

— Marsden and Weinstein's momentum-map reduction industrializes symmetry: each conserved charge removes a full canonical pair

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §9.4–9.6 (infinitesimal canonical transformations and generators) · Landau & Lifshitz, Mechanics, 3rd ed., §40–42 (Poisson brackets and conservation) · Morin, Introduction to Classical Mechanics, Ch. 15 (Hamiltonian formalism) · Taylor, Classical Mechanics, Ch. 13 · Arnold, Mathematical Methods of Classical Mechanics, §40 (Hamiltonian symmetry and momentum maps)

Provenance. All Poisson brackets computed from $\{q_i,p_j\}=\delta_{ij}$ and the canonical algebra (antisymmetry, Leibniz, Jacobi). A verifier should recompute: the Type-2 expansion giving $\delta q=\epsilon\{q,G\}$ (Goldstein §9.4); the equivalence derivation via antisymmetry (Landau–Lifshitz §40, Goldstein §9.6); the $\mathfrak{so}(3)$ closure $\{L_x,L_y\}=L_z$ ; the Galilei boost cancellation $\dot G=-p+p=0$ for $G=pt-mx$ , $H=p^2/2m$ (Arnold §40). Structure constants $\epsilon_{ijk}$ and the central charge $\{K_i,P_j\}=m\delta_{ij}$ are standard. No numerical constants beyond integer structure constants are used. The numeric examples (leak rate $b(y^2-x^2)=0.128\ \mathrm{N\,m}$ at $b=0.80$ , $x=0.30$ , $y=0.50$ , cross-checked against the torque; screw charge $G=0.43820\ \mathrm{J\,s}$ and $p_z=5.912\ \mathrm{kg\,m/s}$ at $h=0.20\ \mathrm{m}$ ) were computed for this article, each verified by an independent route.

Builds on: Poisson-Bracket Formulation of Dynamics · Algebra of Poisson Brackets · Definition of Canonical Transformations · Type-2 Generating Function F2(q,P,t) · Fundamental Poisson Brackets Leads to: Symplectic Group and Phase-Space Geometry · Liouville's Theorem · Hamilton–Jacobi Theory · Quantization and Symmetry

Noether's theorem already gives me conserved charges. What does the Hamiltonian version actually add?

Three things. Reciprocity: the Lagrangian theorem runs symmetry → charge; the bracket version is a genuine iff, so you can also run charge → symmetry — hand me any conserved $G$ and I hand you the invariance of $H$ it generates, no action variation needed. Algebra: charges close under the bracket (Poisson's theorem), so you get the symmetry group's Lie algebra realized on observables, with central charges like $\{K,p\}=m$ that the naive Lagrangian recipe hides in total-derivative bookkeeping. And quantization: $\{G,H\}=0\to[\hat G,\hat H]=0$ transcribes this page directly into quantum mechanics, which is why symmetry arguments there are Hamiltonian-flavored. The physics content coincides where both apply; the packaging does different work.

Is every conserved quantity the generator of some symmetry, or only the famous ones?

Every one — that is the theorem's cleanest punchline. Conservation means $\{G,H\}+\partial_t G=0$ ; for clock-free $G$ this is exactly $\delta H=0$ under $G$ 's flow, so the "symmetry associated with $G$ " always exists: it is the canonical flow $G$ itself generates. It may look unglamorous. The Laplace–Runge–Lenz charge of Constants of Motion via Poisson Brackets generates a peculiar squashing of the orbit that no one would call obvious, and a generic integrable system's charges generate flows with no household names. "Symmetry" in this formalism just means "canonical flow preserving $H$ " — famous or not.

Why does antisymmetry of the bracket carry the entire theorem? What would happen with a symmetric bracket?

Because the theorem is literally the statement $\{H,G\}=-\{G,H\}$ dressed up. The left side is the symmetry meter (change of $H$ along $G$ 's flow), the right is the conservation meter (rate of $G$ along $H$ 's flow); antisymmetry pins them to be negatives, so one vanishes iff the other does. With a symmetric pairing the two meters would be independent numbers — you could have flows preserving $H$ whose generators drift, and conserved quantities generating flows that pump energy. Antisymmetry is also what makes $\{H,H\}=0$ automatic, i.e. energy conservation for autonomous systems; a symmetric bracket would not even grant you that.

How do I find the conserved combination when no single coordinate is cyclic, like the helix?

Work backwards from the invariance you can see. Write the finite transformation that leaves the potential unchanged — for the helix, rotate by $\mathrm{d}\phi$ while translating by $\mathrm{d}z=\tfrac{h}{2\pi}\mathrm{d}\phi$ — and read off the generator as the momentum-weighted sum of the displaced coordinates: $G=p_\phi\cdot1+p_z\cdot\tfrac{h}{2\pi}$ . Each coordinate's displacement per unit parameter multiplies its conjugate momentum. Then verify $\{G,H\}=0$ by the chain rule as in Derivation C, because eyeballed symmetries are sometimes only approximate. The same recipe produces the boost ( $\delta x=t$ per unit velocity, so $G\supset pt$ , plus the clock piece) and, in crystals, the quasi-translation charges of helical dislocations.

What exactly is a central charge, and why should the mass show up as one?

Close the brackets of your symmetry generators and sometimes you get, instead of another generator, a pure number that brackets to zero with everything — a central element. For Galilean boosts and translations, $\{K,p\}=m$ : performing an infinitesimal boost and translation in the two orders differs by a constant, and that constant is the inertia. It appears because boosts do not preserve $H$ instant-by-instant (kinetic energy changes) — they are clocked symmetries, and the clock's bookkeeping deposits the mass into the algebra. The consequence survives quantization intact: the extended algebra forces mass superselection (Bargmann), and in relativistic physics the analogous central extensions control anomalies. All of that machinery is visible classically in one two-line bracket.

If the bracket of L_z with H is the torque, is that a coincidence or a general pattern?

Fully general pattern. Equation (5) says $\dot G=\{G,H\}$ , so whatever force-like quantity drives $G$ 's drift is what the bracket computes: for $G=p_x$ it returns $-\partial H/\partial x$ , the force component; for $G=L_z$ , the torque; for $G=H$ in a driven system, the input power $\partial H/\partial t$ appears through the clock term instead. The bracket with $H$ is the universal "equation of motion generator," and each classical rate law — Newton's second law, the torque equation, the work-energy theorem — is one choice of first slot. That is the practical content of Poisson-Bracket Formulation of Dynamics, and it is why checking conservation and computing dynamics are the same act in this formalism.

Can a symmetry of the equations of motion fail to be a symmetry of H?

Yes, and the boost is the standard example: Galilean relativity is a perfect symmetry of free-particle dynamics, yet boosting changes the energy, so no clock-free $G$ with $\{G,H\}=0$ represents it. The formalism absorbs this through explicitly time-dependent generators satisfying the full criterion $\{G,H\}+\partial_t G=0$ — the flow then maps solutions to solutions while sliding along energy surfaces rather than staying on one. (Scaling symmetries of the Kepler problem behind Kepler's third law are similar: they map solutions to solutions while rescaling $H$ , and correspondingly yield no conserved charge at all.) The sharp statement of this page's theorem concerns invariance of $H$ ; symmetry of the dynamics is the slightly larger tent, and the clock term is the flap that lets the extra cases in.

Which conserved charges can I actually use simultaneously to simplify a problem?

Only charges in involution — pairwise vanishing brackets — can be made simultaneous momenta of a single canonical chart, each rendering its conjugate coordinate cyclic and cutting the problem by one degree of freedom. The rotation algebra shows the obstruction: $L_x,L_y,L_z$ are all conserved in a central field, but $\{L_i,L_j\}=\epsilon_{ijk}L_k\neq0$ , so you may use $L^2$ and one component — precisely the $(\ell,m)$ bookkeeping quantum mechanics inherits. Given $n$ independent charges in involution on $2n$ -dimensional phase space, the system is Liouville-integrable and Hamilton–Jacobi theory finishes the job; count them with the tools of Constants of Motion via Poisson Brackets.