Noether's theorem — symmetries and conservation laws

Intuition [Beginner]

You know conservation laws: energy is conserved, momentum is conserved, angular momentum is conserved. You were probably told these are experimental facts, or that they follow from Newton's laws. They are deeper than that.

Noether's theorem says: every continuous symmetry of the laws of physics produces a conserved quantity, and conversely, every conserved quantity comes from a symmetry. A symmetry is a transformation you can apply to a system that leaves the physics unchanged. "Continuous" means you can dial the transformation by small amounts — shift everything one millimetre, or half a millimetre, or a thousandth.

The three classic examples fall out immediately. If the laws of physics do not change with time — an experiment performed today gives the same result as tomorrow — then energy is conserved. If the laws do not depend on where you are in space — shifting your entire lab one metre to the left changes nothing — then momentum is conserved. If the laws do not depend on which direction you face — rotating the lab by any angle changes nothing — then angular momentum is conserved.

This is why conservation laws are universal. They come from the geometry of space and time, not from the details of any particular force. Gravity, electromagnetism, the strong and weak nuclear forces — all of them conserve energy, momentum, and angular momentum, because all of them live in a spacetime with time-translation, space-translation, and rotation symmetry.

Consider a marble rolling inside a perfectly smooth, round bowl. The bowl has rotational symmetry: turn the bowl by any angle and it looks the same. Noether's theorem says angular momentum about the centre is conserved. If the marble starts moving in a circle, it stays in that circle forever. Now put the marble on a bumpy, irregular surface — no rotational symmetry. Angular momentum is no longer conserved. The marble wobbles, speeds up, slows down. The conservation law was not a fact about the marble; it was a fact about the shape of the surface.

The Lagrangian — kinetic energy minus potential energy, $L=T-V$ — encodes the physics of the system. A symmetry of the system is a transformation that leaves the Lagrangian unchanged (or changes it only by a total derivative, which does not affect the equations of motion). Noether's theorem gives you a recipe: hand it a symmetry, get back a conserved quantity. The recipe is mechanical — you do not need physical intuition for each case.

This theorem was proved by Emmy Noether in 1918. It is one of the most important results in all of theoretical physics. It unifies conservation laws under a single principle, and it extends far beyond classical mechanics — into general relativity, quantum field theory, and gauge theories. The Standard Model of particle physics is built on it.

Visual [Beginner]

Figure: A top panel shows three symmetry transformations applied to a mechanical system: (1) a clock running, with a double arrow labelled "time translation" — the Lagrangian before and after the shift is the same, and the text reads "energy conserved"; (2) a particle trajectory shifted sideways in space, with an arrow labelled "spatial translation" — the Lagrangian is unchanged, and the text reads "momentum conserved"; (3) a particle trajectory rotated about an axis, with a curved arrow labelled "rotation" — the Lagrangian is unchanged, and the text reads "angular momentum conserved." A bottom panel contrasts this with a bumpy potential landscape labelled "no symmetry" — the trajectory wobbles irregularly, and the text reads "no conserved quantity from this direction."

Worked example [Beginner]

A planet of mass $m$ orbits the Sun (mass $M$, treated as fixed) in a central gravitational potential $V(r)=-GMm/r$ where $r$ is the distance from the Sun. The Lagrangian in polar coordinates $(r,\theta )$ is

$$L=\frac{1}{2}m({\dot{r}}^2+r^2{\dot{\theta }}^2)+\frac{GMm}{r}.$$

Does this Lagrangian change when you rotate the system by an angle $\alpha$? In polar coordinates, a rotation replaces $\theta$ with $\theta +\alpha$. The Lagrangian depends on $\dot{\theta }$ but not on $\theta$ itself. So replacing $\theta$ by $\theta +\alpha$ leaves $L$ exactly as it was. The Lagrangian is symmetric under rotation.

Noether's theorem says: this symmetry produces a conserved quantity. The recipe gives us the angular momentum

$$p_{\theta }=m{r}^2\dot{\theta }.$$

Because $\theta$ does not appear in $L$, the Euler-Lagrange equation for $\theta$ is $\frac{d}{dt}(m{r}^2\dot{\theta })=0$ — angular momentum is conserved.

For the Earth-Sun system, $GM\approx 1.327\times 10^{20}$ m${}^3$/s${}^2$, $m\approx 5.97\times 10^{24}$ kg, mean orbital radius $r\approx 1.496\times 10^{11}$ m, orbital period $T\approx 3.156\times 10^7$ s. The angular velocity is $\dot{\theta }=2\pi /T\approx 1.99\times 10^{-7}$ rad/s. The conserved angular momentum is $p_{\theta }=m{r}^2\dot{\theta }\approx 2.66\times 10^{40}$ kg m${}^2$/s. This number is constant to high precision — one of the best-tested conservation laws in physics, and it comes from the rotational symmetry of empty space, not from the gravitational force law itself.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $L(q,\dot{q},t)$ be the Lagrangian of a system with $n$ degrees of freedom $q=(q^1,\dots ,q^n)$. Consider a one-parameter family of transformations

$$q^i\mapsto Q^i(q,\dot{q},t;ϵ)=q^i+ϵ{\xi }^i(q,t)+O({ϵ}^2),$$

where $ϵ$ is a real parameter and ${\xi }^i$ are smooth functions called the infinitesimal generators of the transformation. The transformation is a variational symmetry of $L$ if

$$L(Q,\dot{Q},t)=L(q,\dot{q},t)+ϵ\frac{dF}{dt}(q,t)+O({ϵ}^2)$$

for some function $F(q,t)$ (which may be zero). That is, the Lagrangian changes at most by a total time derivative — which does not affect the Euler-Lagrange equations.

Noether's theorem. If a one-parameter family of transformations is a variational symmetry of $L$, then the quantity

$$\boxed{J=\frac{\partial L}{\partial {\dot{q}}^i}{\xi }^i-F}$$

is conserved along any solution of the Euler-Lagrange equations: $dJ/dt=0$.

The quantity $J$ is called the Noether charge (or Noether conserved quantity). The functions $p_i=\partial L/\partial {\dot{q}}^i$ are the conjugate momenta. The generators ${\xi }^i$ specify how the coordinates change under the symmetry.

The three classical cases.

1. Time translation. The transformation is $t\mapsto t+ϵ$, with ${\xi }^i={\dot{q}}^i$ (the coordinates shift along the trajectory) and $F=L$. The Noether charge becomes $J=p_i{\dot{q}}^i-L=H$, the Hamiltonian. Time-translation invariance produces energy conservation.

2. Spatial translation. The transformation is $q^i\mapsto q^i+ϵ{\delta }_k^i$ for a fixed direction $k$, with ${\xi }^i={\delta }_k^i$ and $F=0$. The Noether charge is $J=p_k$, the conjugate momentum in the $k$-th direction. Translation invariance in direction $k$ produces conservation of the $k$-th component of momentum.

3. Rotation. For rotation about the $z$-axis in three dimensions: $x\mapsto x-ϵy$, $y\mapsto y+ϵx$, $z\mapsto z$. The generators are ${\xi }^x=-y$, ${\xi }^y=x$, ${\xi }^z=0$. The Noether charge is $J=p_x(-y)+p_y(x)=x{p}_y-y{p}_x=L_z$, the $z$-component of angular momentum. Rotational symmetry about an axis produces conservation of the corresponding angular momentum component.

Counterexamples to common slips

• Noether's theorem requires the Euler-Lagrange equations. The Noether charge $J$ is conserved along solutions of the equations of motion. An arbitrary curve $(q(t),\dot{q}(t))$ that does not satisfy the Euler-Lagrange equations will not generally conserve $J$, even if the symmetry is present.

• A total-derivative shift in $L$ matters. A symmetry that changes $L$ by more than a total derivative does not produce a conserved quantity via Noether's theorem. The transformation must be a variational symmetry, not merely leave the equations of motion invariant in some weaker sense.

• Not every conserved quantity comes from an obvious geometric symmetry. The Laplace-Runge-Lenz vector in the Kepler problem is conserved but comes from a hidden $\mathrm{SO}(4)$ symmetry that is not visible from the geometry of space alone. Noether's theorem still applies — the symmetry is there, just not geometrically obvious.

• Noether's theorem gives one conserved quantity per one-parameter symmetry. A discrete symmetry (reflection, parity) does not generate a Noether charge. A finite symmetry group with $k$ generators gives $k$ conserved quantities, one per generator.

Key theorem with proof [Intermediate+]

Theorem (Noether). Let $L(q,\dot{q},t)$ be a smooth Lagrangian on $TQ$. Suppose there exists a one-parameter family of transformations $q^i\mapsto q^i+ϵ{\xi }^i(q,t)+O({ϵ}^2)$ such that

$$L(q+ϵ\xi ,\dot{q}+ϵ\dot{\xi },t)=L(q,\dot{q},t)+ϵ\frac{dF(q,t)}{dt}+O({ϵ}^2).$$

Then the quantity $J=p_i{\xi }^i-F$ (where $p_i=\partial L/\partial {\dot{q}}^i$) satisfies $dJ/dt=0$ along every solution of the Euler-Lagrange equations.

Proof. Expand the left side to first order in $ϵ$:

$$L(q+ϵ\xi ,\dot{q}+ϵ\dot{\xi },t)=L(q,\dot{q},t)+ϵ\left(\frac{\partial L}{\partial q^i}{\xi }^i+\frac{\partial L}{\partial {\dot{q}}^i}{\dot{\xi }}^i\right)+O({ϵ}^2).$$

The symmetry condition equates the $O(ϵ)$ terms with $dF/dt$:

$$\frac{\partial L}{\partial q^i}{\xi }^i+\frac{\partial L}{\partial {\dot{q}}^i}{\dot{\xi }}^i=\frac{dF}{dt}.$$

Now compute $dJ/dt$ along a trajectory, where $p_i=\partial L/\partial {\dot{q}}^i$:

$$\frac{dJ}{dt}=\frac{d}{dt}(p_i{\xi }^i)-\frac{dF}{dt}={\dot{p}}_i{\xi }^i+p_i{\dot{\xi }}^i-\frac{dF}{dt}.$$

On solutions of the Euler-Lagrange equations, ${\dot{p}}_i=\partial L/\partial q^i$. Substituting:

$$\frac{dJ}{dt}=\frac{\partial L}{\partial q^i}{\xi }^i+\frac{\partial L}{\partial {\dot{q}}^i}{\dot{\xi }}^i-\frac{dF}{dt}.$$

The right side is zero by the symmetry condition. Therefore $dJ/dt=0$. $\square$

Bridge. Noether's theorem is the foundational reason that conservation laws are not independent physical postulates but consequences of the geometry of the action functional. The proof identifies the Noether charge $J=p_i{\xi }^i-F$ as exactly the obstruction to the symmetry transforming the Lagrangian — the bridge is between the infinitesimal invariance of $L$ and the constancy of $J$ on-shell. This result appears again in 09.04.01 pending via the Legendre transform, where $J$ becomes the Hamiltonian generator of the symmetry; it generalises in 09.05.01 pending to the moment-map framework of symplectic geometry, which packages all Noether charges for a Lie-group action into a single ${\mathfrak{g}}^{\ast }$-valued function; and it builds toward 10.09.01 pending where gauge invariance of the electromagnetic Lagrangian produces charge conservation via the field-theoretic version of the theorem.

Example: the Laplace-Runge-Lenz vector and hidden SO(4) symmetry

The Kepler/Coulomb problem with $L=\frac{1}{2}m|\dot{\mathbf{q}}{|}^2+k/|\mathbf{q}|$ has, in addition to the conserved angular momentum $\mathbf{L}=\mathbf{q}\times \mathbf{p}$ (from rotation symmetry) and energy $H$ (from time-translation symmetry), a fourth independent conserved quantity: the Laplace-Runge-Lenz vector

$$\mathbf{A}=\mathbf{p}\times \mathbf{L}-mk\hat{\mathbf{r}},$$

where $\hat{\mathbf{r}}=\mathbf{q}/|\mathbf{q}|$. This vector points from the force centre to the perihelion, and its conservation means the orbit does not precess — the ellipse is fixed in space. The conservation of $\mathbf{A}$ reflects a hidden symmetry: the bound-state Kepler problem is invariant under a larger group $\mathrm{SO}(4)$ (not just $\mathrm{SO}(3)$). The six generators of $\mathrm{SO}(4)$ — three components of $\mathbf{L}$ and a rescaled $\mathbf{A}$ — close into an $\mathfrak{so}(4)$ Lie algebra under the Poisson bracket. This makes the Kepler problem superintegrable: five independent conserved quantities ($H$, three components of $\mathbf{L}$ minus one for the constraint $\mathbf{A}\cdot \mathbf{L}=0$, plus two components of $\mathbf{A}$) on a six-dimensional phase space. The $\mathrm{SO}(4)$ symmetry is "hidden" in the sense that it is not a geometric symmetry of three-dimensional space; it mixes position and momentum in a non-obvious way. Noether's theorem still captures it — the symmetry is a canonical transformation of phase space that leaves the action invariant.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib has Lie group actions and infinitesimal generators in fragments (Geometry.Manifold.Algebra, Topology.ContinuousMap), the calculus of variations for real-valued functionals, and smooth manifold theory sufficient for tangent bundles. It does not formalise Noether's theorem as a named result connecting one-parameter variational symmetries of a Lagrangian to conserved Noether charges along Euler-Lagrange trajectories. The moment-map framework (symplectic geometry) is also absent. The proof requires assembling: the EL equations, the infinitesimal symmetry condition, differentiation of the Noether charge along a trajectory, and the cancellation step. Each piece exists in isolation; the synthesis does not. lean_status: none.

Advanced results [Master]

Theorem 1 (Noether on manifolds). Let $Q$ be a smooth $n$-manifold and $L:TQ\times \mathbb{R}\to \mathbb{R}$ a smooth Lagrangian. A one-parameter group of diffeomorphisms ${\varphi }_{ϵ}:Q\to Q$ (with ${\varphi }_0=\mathrm{id}$) induces a lifted transformation ${\hat{\varphi }}_{ϵ}:TQ\to TQ$ by differentiation. The infinitesimal generator is the vector field $X_{\xi }$ on $Q$ with local components ${\xi }^i$. If ${\varphi }_{ϵ}$ is a variational symmetry — that is, $L\circ {\hat{\varphi }}_{ϵ}=L+ϵdF/dt+O({ϵ}^2)$ for some $F$ — then the charge $J=p_i{\xi }^i-F$ is constant along every Euler-Lagrange trajectory on $TQ$.

The proof on manifolds is identical to the coordinate proof given in the Intermediate tier; the manifold setting ensures the statement is coordinate-free. Arnold (1989) §20 gives the formulation on cotangent bundles $T^{\ast }Q$ ^{[Arnold 1989 §20]}.

Theorem 2 (Moment map — existence). Let $(M,\omega )$ be a symplectic manifold and $G$ a Lie group acting on $M$ by symplectomorphisms. If the action is Hamiltonian, there exists a smooth map $\mathbf{J} : M \to \mathfrak{g}^$satisfying$d\langle \mathbf{J}, \xi \rangle = i_{\xi_M}\omega$forall$\xi \in \mathfrak{g}$,where$\xi_M$istheinfinitesimalgenerator.Thecomponent$J_\xi = \langle \mathbf{J}, \xi \rangle$istheNoethercharge:${J_\xi, H} = 0$whenever$H$is$G$-invariant.*

The moment map simultaneously encodes all Noether charges for the $G$-action in a single ${\mathfrak{g}}^{\ast }$-valued function. Marsden-Weinstein reduction 09.05.01 pending constructs the reduced phase space ${\mathbf{J}}^{-1}(\mu )/G_{\mu }$ as a symplectic manifold of dimension $\dim M-2\dim G_{\mu }$ — the effective phase space after modding out the conserved quantities.

Theorem 3 (Coadjoint-orbit symplectic structure). The image $\mathbf{J}(M) \subset \mathfrak{g}^$isaunionofcoadjointorbitsofthe$\mathrm{Ad}^_g$-actionof$G$on$\mathfrak{g}^$.Eachcoadjointorbit$\mathcal{O}\mu$carriestheKirillov-Kostant-Souriausymplecticform$\omega^+\mu(\mathrm{ad}^_\xi \mu, \mathrm{ad}^_\eta \mu) = \langle \mu, [\xi, \eta]\rangle$,makingitahomogeneoussymplectic$G$-space.*

This is the symplectic-geometric refinement of "conserved quantities live on orbits of the symmetry group." Kostant 1970 and Souriau 1970 established the orbit method independently ^{[Kostant 1970]}; Kirillov's 1962 classification of irreducible unitary representations of nilpotent Lie groups was the prototype ^{[Kirillov 1962 *Russ. Math. Surveys* 17]}.

Theorem 4 (Noether for fields — conserved currents). For a field theory with Lagrangian density $\mathcal{L}(\varphi ,{\partial }_{\mu }\varphi )$ on spacetime, a continuous symmetry $\varphi \to \varphi +ϵ\delta \varphi$ of the action produces a conserved current $j^{\mu }$ satisfying the continuity equation ${\partial }_{\mu }{j}^{\mu }=0$ on solutions of the Euler-Lagrange field equations. The spatial integral $Q=\int j^0{d}^{3}x$ is the conserved charge.

Translation symmetry gives the energy-momentum tensor $T^{\mu \nu }$ (four conserved quantities: energy and three momentum components). Lorentz symmetry gives the angular momentum tensor $M^{\mu \nu \lambda }$. Internal symmetries (phase rotations, isospin, colour in QCD) give conserved currents for electric charge, baryon number, and lepton number. The Standard Model gauge symmetry $\mathrm{SU}(3)\times \mathrm{SU}(2)\times \mathrm{U}(1)$ produces a conserved current for each generator ^{[Peskin-Schroeder 1995 Ch. 2]}.

Theorem 5 (Moment map as Lie-algebra homomorphism). If the moment map $\mathbf{J}$ is $G$-equivariant, then the map $\xi \mapsto J_{\xi }=⟨\mathbf{J},\xi ⟩$ is a Lie-algebra homomorphism from $\mathfrak{g}$ to the Poisson algebra of smooth functions on $M$: $\{J_{\xi },J_{\eta }\}=J_{[\xi ,\eta ]}$ for all $\xi ,\eta \in \mathfrak{g}$.

This identifies the Noether charges with the symmetry generators themselves: $J_{\xi }$ is conserved under the Hamiltonian flow of $H$ (because $\{J_{\xi },H\}=0$), and simultaneously generates the symmetry via its own Hamiltonian flow. The equivariance condition excludes cocycle corrections that arise for projective representations (e.g., the magnetic translation group).

Theorem 6 (Noether charge as symmetry generator). The Hamiltonian flow of $J_{\xi }$ on $(M,\omega )$ is the one-parameter group of canonical transformations induced by the symmetry $\exp (t\xi )\subset G$. The conservation law $\{J_{\xi },H\}=0$ and the generation law $\{J_{\xi },\cdot \}={\mathcal{L}}_{{\xi }_M}$ are dual faces of the same symplectic-geometric fact.

Putting these together: the Hamiltonian version of Noether's theorem states that the symmetry group $G$ acts on phase space by the Hamiltonian flows of the moment-map components, and conservation is the statement $\{J_{\xi },H\}=0$ for all $\xi$.

Synthesis. The manifold, moment-map, coadjoint-orbit, and field-theoretic formulations are not separate theorems but facets of a single geometric principle. The foundational reason is that the action functional $S[q]=\int Ldt$ is a function on the infinite-dimensional space of paths, and a symmetry of $S$ produces a level set of the Noether charge along stationary paths. The central insight is that passing from $TQ$ to $T^{\ast }Q$ via the Legendre transform converts the Lagrangian Noether charge into the Hamiltonian moment map, and this is exactly the structure that Marsden-Weinstein reduction exploits to quotient phase space by the symmetry group. The bridge is between the variational calculus of the action and the symplectic geometry of the reduced phase space: the pattern generalises from finite-dimensional configuration spaces to field theories on spacetime, where conserved currents replace conserved charges, and appears again in 10.09.01 pending for gauge theories and 10.05.01 pending for general relativity, where Noether's second theorem governs the interplay between diffeomorphism invariance and gravitational energy localisation.

The moment map and symplectic Noether theorem [Master]

The Hamiltonian reformulation of Noether's theorem provides the sharpest geometric statement. Let $(M,\omega )$ be a symplectic manifold — for a mechanical system, $M=T^{\ast }Q$ with the canonical symplectic form $\omega =d{p}_i\wedge d{q}^i$. A Lie group $G$ acts on $M$ by symplectomorphisms: ${\Phi }_g^{\ast }\omega =\omega$ for all $g\in G$. For each $\xi \in \mathfrak{g}$, the infinitesimal generator ${\xi }_M$ is a symplectic vector field: ${\mathcal{L}}_{{\xi }_M}\omega =0$. By Cartan's magic formula, this is equivalent to $d(i_{{\xi }_M}\omega )=0$, so the one-form $i_{{\xi }_M}\omega$ is closed. If $i_{{\xi }_M}\omega$ is in fact exact — that is, there exists a function $J_{\xi }$ with $d{J}_{\xi }=i_{{\xi }_M}\omega$ — then the action is called Hamiltonian and the assignment $\xi \mapsto J_{\xi }$ assembles into a map $\mathbf{J}:M\to {\mathfrak{g}}^{\ast }$ called the moment map.

The defining equation $d{J}_{\xi }=i_{{\xi }_M}\omega$ has a direct physical reading. In canonical coordinates on $T^{\ast }Q$, the symplectic form is $\omega =d{p}_i\wedge d{q}^i$ and ${\xi }_M$ has components $({\xi }^i,{\partial }^{2}L/\partial q^i\partial {\dot{q}}^j{\xi }^j)$. The moment-map equation integrates to $J_{\xi }=p_i{\xi }^i-F$, recovering the Lagrangian Noether charge. The moment map is thus the Hamiltonian incarnation of the Noether machinery.

The conservation theorem is immediate. Along the Hamiltonian flow ${\varphi }_t^H$ of $H$:

$$\frac{d}{dt}(J_{\xi }\circ {\varphi }_t^H)=\{J_{\xi },H\}\circ {\varphi }_t^H.$$

If $H$ is $G$-invariant — equivalently $\{H,J_{\xi }\}=0$ for all $\xi$, since $\{H,J_{\xi }\}=-{\mathcal{L}}_{{\xi }_M}H$ — then $J_{\xi }$ is conserved. The proof is two lines: $\{J_{\xi },H\}=\omega (X_{J_{\xi }},X_H)=\omega ({\xi }_M,X_H)=-dH({\xi }_M)=-{\mathcal{L}}_{{\xi }_M}H=0$.

Example: angular momentum as moment map. The rotation group $\mathrm{SO}(3)$ acts on $T^{\ast }{\mathbb{R}}^3$ by ${\Phi }_g(\mathbf{q},\mathbf{p})=(g\mathbf{q},g\mathbf{p})$. The moment map $\mathbf{J}:T^{\ast }{\mathbb{R}}^3\to \mathfrak{so}(3{)}^{\ast }\cong {\mathbb{R}}^3$ is $\mathbf{J}(\mathbf{q},\mathbf{p})=\mathbf{q}\times \mathbf{p}=\mathbf{L}$, the angular momentum vector. The three components $L_i=(\mathbf{q}\times \mathbf{p}{)}_i$ satisfy $\{L_i,L_j\}={ϵ}_{ijk}{L}_k$, which is the Lie-algebra homomorphism property $\{J_{\xi },J_{\eta }\}=J_{[\xi ,\eta ]}$ for $\mathfrak{so}(3)$. Conservation of angular momentum under a rotation-invariant $H$ is $\{L_i,H\}=0$.

Example: linear momentum as moment map. The translation group ${\mathbb{R}}^3$ acts on $T^{\ast }{\mathbb{R}}^3$ by ${\Phi }_{\mathbf{a}}(\mathbf{q},\mathbf{p})=(\mathbf{q}+\mathbf{a},\mathbf{p})$. The moment map is $\mathbf{J}(\mathbf{q},\mathbf{p})=\mathbf{p}$, the linear momentum. The Poisson brackets $\{p_i,p_j\}=0$ reflect the Abelian Lie algebra ${\mathbb{R}}^3$. Translation-invariant $H$ gives $\{p_i,H\}=0$ — conservation of linear momentum.

Noether's second theorem and gauge symmetries [Master]

Noether's 1918 paper contains a second theorem, less well known but equally fundamental, that deals with infinite-dimensional symmetry groups — families of transformations parametrised by arbitrary functions rather than by a finite set of parameters. The paradigmatic example is the diffeomorphism group of a manifold.

Theorem (Noether's second theorem). If the action functional $S[\varphi ]=\int \mathcal{L}(\varphi ,\partial \varphi )d^{n}x$ is invariant under a symmetry group parametrised by $r$ arbitrary smooth functions ${ϵ}^{\alpha }(x)$ ($\alpha =1,\dots ,r$) vanishing on the boundary, then the Euler-Lagrange equations satisfy $r$ differential identities. These identities relate the field equations without imposing further constraints — they are off-shell identities among the variational derivatives.

The consequence is that the field equations are not all independent: $r$ of them are determined by the others via the identities. In general relativity, the diffeomorphism invariance of the Einstein-Hilbert action $S_{\mathrm{EH}}=\int (R/2\kappa )\sqrt{-g}{d}^{4}x$ produces four Bianchi identities ${\nabla }_{\mu }{G}^{\mu \nu }=0$ among the ten Einstein field equations $G^{\mu \nu }=\kappa T^{\mu \nu }$. These four identities reduce the independent equations from ten to six — matching the four free functions available in any coordinate choice (the gauge freedom of general covariance).

The interplay between the first and second theorems in general relativity resolves the long-standing puzzle about energy conservation in curved spacetime. The first theorem, applied to the time-translation isometry of a specific spacetime (e.g., Schwarzschild), produces a conserved energy. The second theorem, applied to the full diffeomorphism group, produces the Bianchi identities, which show that gravitational energy-momentum cannot be localised as a tensor — it is a pseudotensor that depends on the coordinate choice. This is not a deficiency of the theory but a structural consequence of gauge invariance: a generally covariant theory cannot have a local, tensorial energy-momentum density for the gravitational field itself.

The gauge-theoretic reading. In Yang-Mills theory with gauge group $G$ and Lagrangian density ${\mathcal{L}}_{\mathrm{YM}}=-\frac{1}{4}{F}_{\mu \nu }^a{F}^{a\mu \nu }$, the gauge invariance $A_{\mu }\to A_{\mu }+D_{\mu }ϵ(x)$ (with $ϵ:\text{spacetime}\to \mathfrak{g}$ arbitrary) is an infinite-dimensional symmetry. The second theorem produces the covariant conservation law $D_{\mu }{J}^{a\mu }=0$ where $J^{a\mu }$ is the matter current — not a Noether conservation law of the first type (which would give ${\partial }_{\mu }{j}^{\mu }=0$), but a covariant version compatible with the gauge structure. The first theorem applies to global (rigid) symmetries — constant gauge transformations $ϵ=\text{const}$ — and produces the genuinely conserved charges (total electric charge, total colour charge) that are measurable in experiments.

The distinction between the two theorems is the distinction between global symmetries (which produce conserved quantities) and local/gauge symmetries (which produce identities among the equations of motion and constrain the structure of the theory). The Standard Model of particle physics is built on this distinction: $\mathrm{SU}(3)\times \mathrm{SU}(2)\times \mathrm{U}(1)$ is a local gauge symmetry (second theorem), while baryon number and lepton number conservation come from global $\mathrm{U}(1)$ symmetries (first theorem, modulo anomalies).

The mathematical distinction is sharp. A finite-dimensional Lie group $G$ with $\dim G<\infty$ yields the first theorem: $\dim G$ independent conserved charges. An infinite-dimensional gauge group (functions on spacetime valued in a Lie algebra) yields the second theorem: differential identities among the field equations, not new conserved charges. Brading and Brown (2000) give a systematic comparison ^{[Brading-Brown 2000 *arXiv:hep-th/0009058*]}; Kosmann-Schwarzbach (2011) provides the historical account of how the two theorems were understood and misunderstood across the twentieth century ^{[Kosmann-Schwarzbach 2011]}.

Full proof set [Master]

Proposition 1 (Moment map satisfies the homomorphism property). Let $G$ act on $(M,\omega )$ in a Hamiltonian fashion with equivariant moment map $\mathbf{J}$. Then $\{J_{\xi },J_{\eta }\}=J_{[\xi ,\eta ]}$ for all $\xi ,\eta \in \mathfrak{g}$.

Proof. For any smooth function $f$ on $M$, the Hamiltonian vector field $X_f$ is defined by $i_{X_f}\omega =-df$. The Poisson bracket is $\{f,g\}=\omega (X_f,X_g)$. The infinitesimal generator ${\xi }_M$ of the $G$-action equals $X_{J_{\xi }}$ (by the moment-map equation $d{J}_{\xi }=i_{{\xi }_M}\omega$, noting $i_{X_{J_{\xi }}}\omega =-d{J}_{\xi }=-i_{{\xi }_M}\omega$, so $X_{J_{\xi }}={\xi }_M$ up to sign convention).

The Jacobi identity for the Poisson bracket gives $\{J_{\xi },\{J_{\eta },f\}\}-\{J_{\eta },\{J_{\xi },f\}\}=\{\{J_{\xi },J_{\eta }\},f\}$ for all $f$. The left side is $\{J_{\xi },{\mathcal{L}}_{{\eta }_M}f\}-\{J_{\eta },{\mathcal{L}}_{{\xi }_M}f\}={\mathcal{L}}_{{\xi }_M}({\mathcal{L}}_{{\eta }_M}f)-{\mathcal{L}}_{{\eta }_M}({\mathcal{L}}_{{\xi }_M}f)={\mathcal{L}}_{[{\xi }_M,{\eta }_M]}f$.

The bracket of the fundamental vector fields satisfies $[{\xi }_M,{\eta }_M]=-[\xi ,\eta {]}_M$ (the minus sign comes from the Lie-algebra convention). Hence ${\mathcal{L}}_{[{\xi }_M,{\eta }_M]}f=-{\mathcal{L}}_{[\xi ,\eta {]}_M}f=-\{J_{[\xi ,\eta ]},f\}$.

Comparing: $\{\{J_{\xi },J_{\eta }\},f\}=-\{J_{[\xi ,\eta ]},f\}$ for all $f$. If the moment map is equivariant, the coboundary term vanishes and $\{J_{\xi },J_{\eta }\}=J_{[\xi ,\eta ]}$. $\square$

Proposition 2 (Conserved current from spacetime translation symmetry). Let $\mathcal{L}(\varphi ,{\partial }_{\mu }\varphi )$ be a Lagrangian density on Minkowski spacetime that does not depend explicitly on the coordinates $x^{\mu }$. Then the energy-momentum tensor $T^{\mu }{}_{\nu }=\frac{\partial \mathcal{L}}{\partial ({\partial }_{\mu }\varphi )}{\partial }_{\nu }\varphi -{\delta }_{\nu }^{\mu }\mathcal{L}$ satisfies ${\partial }_{\mu }{T}^{\mu \nu }=0$ on solutions of the Euler-Lagrange equations.

Proof. Under an infinitesimal translation $x^{\mu }\to x^{\mu }+ϵa^{\mu }$ (with $a^{\mu }$ constant), the field changes by $\delta \varphi =a^{\mu }{\partial }_{\mu }\varphi$ and the Lagrangian density changes by $\delta \mathcal{L}=a^{\mu }{\partial }_{\mu }\mathcal{L}={\partial }_{\mu }(a^{\mu }\mathcal{L})$. By the field-theoretic Noether formula, the conserved current is

$$j^{\mu }=\frac{\partial \mathcal{L}}{\partial ({\partial }_{\mu }\varphi )}\delta \varphi -a^{\mu }\mathcal{L}=\frac{\partial \mathcal{L}}{\partial ({\partial }_{\mu }\varphi )}{a}^{\nu }{\partial }_{\nu }\varphi -a^{\mu }\mathcal{L}=a_{\nu }{T}^{\mu \nu }.$$

Since $a^{\nu }$ is arbitrary, ${\partial }_{\mu }{T}^{\mu \nu }=0$. The conserved charges are $P^{\nu }=\int T^{0\nu }{d}^{3}x$: energy ($\nu =0$) and momentum ($\nu =1,2,3$). $\square$

Proposition 3 (Reduction of degrees of freedom by symmetries). Let a compact Lie group $G$ of dimension $k$ act freely and in a Hamiltonian fashion on a $2n$-dimensional symplectic manifold $(M,\omega )$ with moment map $\mathbf{J}$. For a regular value $\mu \in \mathfrak{g}^$,theMarsden-Weinsteinquotient$M_\mu = \mathbf{J}^{-1}(\mu)/G_\mu$isasymplecticmanifoldofdimension$2(n - k)$.*

The dimension count encodes the physical fact that each independent symmetry removes two degrees of freedom from the effective phase space: one from the conserved quantity (the moment-map constraint $\mathbf{J}=\mu$) and one from the quotient by the symmetry group orbit. For a free particle in ${\mathbb{R}}^3$ with rotational symmetry ($k=3$), the unreduced phase space has $2n=6$ dimensions and the reduced space has $2(3-3)=0$ dimensions — meaning the motion is entirely determined by the three conserved components of angular momentum and the three free parameters of the rotation group. For the Kepler problem, the $\mathrm{SO}(4)$ symmetry ($k=6$) on a six-dimensional phase space leaves the reduced space zero-dimensional, explaining why the Kepler orbit is a unique closed ellipse determined by its conserved quantities.

Connections [Master]

• Action principle 09.02.01 pending — Noether's theorem is proved by varying the action under a symmetry transformation; the action principle is the substrate on which the symmetry-conservation link is established.

• Euler-Lagrange equations 09.02.02 pending — the Noether charge is conserved along EL trajectories; the EL equations are used in the proof to replace ${\dot{p}}_i$ by $\partial L/\partial q^i$.

• Conservation laws (Newtonian) 09.01.03 — the Newtonian conservation laws (energy, momentum, angular momentum) are the three special cases of Noether's theorem for the Galilean symmetries of Newtonian spacetime.

• Hamiltonian formalism 09.04.02 pending — the Noether charge becomes the Hamiltonian generator of the symmetry; $\{J,H\}=0$ is the Hamiltonian form of conservation, and the moment map $\mathbf{J}:T^{\ast }Q\to {\mathfrak{g}}^{\ast }$ is the coordinate-free packaging.

• Legendre transform 09.04.01 pending — converts the Lagrangian Noether charge $J=p_i{\xi }^i-F$ into the Hamiltonian moment map.

• Canonical transformations 09.05.01 pending — a symmetry of the Lagrangian induces a canonical transformation on $T^{\ast }Q$; Marsden-Weinstein reduction uses the moment map to quotient phase space by the symmetry group.

• EM Lagrangian and gauge invariance 10.09.01 pending — gauge invariance of the electromagnetic Lagrangian $\mathcal{L}=-\frac{1}{4}{F}_{\mu \nu }{F}^{\mu \nu }+j^{\mu }{A}_{\mu }$ produces electric-charge conservation via the field-theoretic Noether theorem.

• Quantum mechanics: symmetries and representations 12.02.01 pending — the Wigner-Eckart theorem and representation theory of symmetry groups on Hilbert space are the quantum analogues of the classical Noether framework; conserved quantities become operators that commute with $\hat{H}$.

• Thermodynamics: first law 11.01.01 pending — energy conservation in thermodynamics is the Noether-theorem consequence of time-translation symmetry; the first law is a special case of the symmetry-conservation connection.

• General relativity 10.05.01 pending — diffeomorphism invariance of the Einstein-Hilbert action produces the contracted Bianchi identities (a weakened form of energy-momentum conservation in curved spacetime); Noether's second theorem (for infinite-dimensional symmetry groups) governs this case.

Historical & philosophical context [Master]

Emmy Noether published "Invariante Variationsprobleme" in 1918 in the Nachrichten von der Koniglichen Gesellschaft der Wissenschaften zu Gottingen. The paper was solicited by Felix Klein and David Hilbert, who were grappling with energy conservation in general relativity. Einstein's theory had raised a puzzle: in a generally covariant theory, the action is invariant under arbitrary coordinate transformations, and the naive application of energy-momentum conservation seemed to break down. Hilbert suspected the problem was not with general relativity but with the understanding of the relationship between symmetries and conservation laws. He asked Noether to investigate.

Noether's paper delivered two theorems. The first (the one taught in physics courses) deals with finite-dimensional Lie groups: a one-parameter variational symmetry produces a single conserved quantity. The second deals with infinite-dimensional symmetry groups (like the diffeomorphism group of a manifold): such symmetries produce not conserved quantities but differential identities among the field equations (the Bianchi identities in GR are the paradigmatic example). Both theorems apply simultaneously in general relativity, and the interplay between them resolves the energy-conservation puzzle — or rather, reframes it: gravitational energy cannot be localised in a generally covariant theory, and Noether's second theorem explains why.

The significance of Noether's theorem extends far beyond the problem that motivated it. The first theorem unifies all of the known conservation laws of classical mechanics under a single principle. The second theorem underlies the structure of gauge theories. Together, they establish that the conservation laws of physics are not independent facts but are logical consequences of the symmetry structure of the action.

Noether was 36 when she published this work. She had been excluded from the Gottingen faculty for seven years because of her gender, lecturing under Hilbert's name. The paper is 23 pages long. It is, by any measure, one of the most consequential papers in the history of mathematical physics.

Bibliography [Master]

• Noether, E., "Invariante Variationsprobleme," Nachr. d. Konig. Gesellsch. d. Wiss. zu Gottingen, Math.-Phys. Klasse (1918), pp. 235–257.
• Arnold, V. I., Mathematical Methods of Classical Mechanics, 2nd ed. (Springer GTM 60, 1989), §20.
• Goldstein, H., Poole, C. P. & Safko, J., Classical Mechanics, 3rd ed. (Pearson, 2002), Ch. 2.6.
• Taylor, J. R., Classical Mechanics (University Science Books, 2005), Ch. 7.5.
• Marsden, J. E. & Ratiu, T. S., Introduction to Mechanics and Symmetry, 2nd ed. (Springer TAM 17, 1999), Ch. 11.
• Susskind, L. & Hrabovsky, G., The Theoretical Minimum: Classical Mechanics (Basic Books, 2014), Lecture 7.
• Tong, D., Classical Dynamics (DAMTP Cambridge lecture notes), §2.
• Landau, L. D. & Lifshitz, E. M., Mechanics, 3rd ed. (Course of Theoretical Physics Vol. 1, Pergamon, 1976), §2.
• Abraham, R. & Marsden, J. E., Foundations of Mechanics, 2nd ed. (Addison-Wesley, 1978), §3.3.
• Olver, P. J., Applications of Lie Groups to Differential Equations, 2nd ed. (Springer GTM 107, 1993), Ch. 4–5.
• Brading, K. & Brown, H. R., "Noether's theorems and gauge symmetries," arXiv/0009058 (2000).
• Kosmann-Schwarzbach, Y., The Noether Theorems: Invariance and Conservation Laws in the Twentieth Century (Springer, 2011).
• Byers, N., "E. Noether's discovery of the deep connection between symmetries and conservation laws," arXiv/9807044 (1998).