Noether's Theorem

Ask a physicist why energy is conserved and the honest answer is not "because of forces" but "because the laws do not care what time it is." Noether's theorem is the exact statement of that idea: every continuous symmetry of an action functional is matched, one-for-one, by a quantity that does not change along the motion. Time-translation symmetry buys energy; space-translation buys momentum; rotational symmetry buys angular momentum; an internal phase symmetry buys electric charge. What had been a loose collection of empirical conservation laws becomes a single theorem with a constructive proof — hand it a symmetry and it hands back the conserved charge, built explicitly from the Lagrangian and the generators of the transformation. Emmy Noether proved this in 1918, in two theorems. The first, treated here, governs global continuous symmetries and produces ordinary conservation laws; the second governs local (gauge) symmetries and produces identities among the equations of motion. The setting is Hamilton's principle of stationary action: physical trajectories extremize $S=\int L\,\mathrm{d}t$ , equivalently they solve the Euler-Lagrange equation. A symmetry is a transformation of coordinates and fields that leaves the action invariant, or invariant up to a boundary term. The proof is a two-line accounting trick — compute the change in the action once using the symmetry and once using the equations of motion, then equate — and out drops a total time derivative that must vanish, whose interior is the Noether charge. In field theory the same logic upgrades the conserved number to a conserved current $j^\mu$ obeying $\partial_\mu j^\mu=0$ , the backbone of every relativistic field theory and the bridge from classical mechanics to the quantum world.

Key takeaways

Every continuous symmetry of an action — one that leaves $S$ invariant, or invariant up to a boundary term — corresponds to a quantity conserved along solutions of the Euler-Lagrange equations.
The correspondence is explicit: from generators $(\tau,\eta^i)$ the charge is $Q=p_i\eta^i-(p_i\dot q^i-L)\tau-K$ , built from the canonical momentum and the energy.
Time translation gives energy, space translation gives momentum, rotation gives angular momentum, internal phase symmetry gives electric charge — all one theorem.
Conservation is on-shell: $\mathrm{d}Q/\mathrm{d}t=0$ uses the equations of motion, so it is a property of solutions, not a kinematic identity.
Quasi-symmetries that change $L$ by a total derivative $\mathrm{d}K/\mathrm{d}t$ still give a charge, once the boundary term $K$ is subtracted — this is how Galilean boosts conserve the center of mass.
In field theory the statement is local: $\partial_\mu j^\mu=0$ , a continuity equation whose spatial integral $Q=\int j^0\,\mathrm{d}^3x$ is the conserved charge.
Key idea: A continuous symmetry of the action $\delta S=0$ (or $\delta S$ equal to a boundary term), combined with the Euler-Lagrange equations, forces a total time derivative to vanish — and the quantity inside that derivative is the conserved Noether charge.

Where this lives today

Noether's first theorem is the load-bearing wall of modern physics. The entire Standard Model is organized by it: the electric charge conserved in every particle collision is the Noether charge of the $U(1)$ phase symmetry of quantum electrodynamics, color charge belongs to the $SU(3)$ of the strong force, weak isospin to the $SU(2)$ of the electroweak sector — the same current construction as the complex-scalar example above. Searches for proton decay test whether baryon number — a Noether charge of an approximate symmetry — is exactly conserved; the rule that "approximate symmetry means $\partial_\mu j^\mu$ equals the small breaking term" is Noether's, and it underlies the partially conserved axial current of pion physics. In general relativity the stress-energy tensor $T^{\mu\nu}$ is the Noether current of spacetime translations, and each Killing vector supplies a conserved quantity along geodesics — which is why the energy and angular momentum of orbits around a black hole are well defined, and why LIGO can bookkeep the energy gravitational waves carry away from a binary inspiral. On the computational side, symplectic integrators preserve the discrete analog of the Noether charges exactly, so a solar-system simulation conserves energy and angular momentum over billions of timesteps. From the Ward identities that make quantum field theory renormalizable to the topological charges of condensed matter, the pattern Noether found — symmetry buys conservation — remains the single most reliable organizing principle in physics.

The physical picture

Picture a marble rolling in a perfectly circular bowl. Rotate the whole bowl about its vertical axis and nothing looks different — the landscape is indifferent to that rotation. Set the marble moving and its angular momentum about the axis never changes, because nothing in the surroundings can exert a tangential torque. A continuous symmetry is exactly this "direction of indifference" made precise: a one-parameter family of transformations you can dial smoothly toward zero, and if the action does not notice the dialing, the quantity conjugate to its generator is frozen. Noether's genius was to turn that intuition into an algorithm — feed in the symmetry, read out the charge — and the first figure is that algorithm as a dictionary.

Figure

The Noether dictionary: each continuous symmetry of the action maps to exactly one conserved charge through the master formula Q = p η − H τ − K, so time translation gives energy H, space translation gives momentum p, rotation gives angular momentum Lz, and an internal phase rotation gives electric charge

The proof underneath is a bookkeeping trick. Compute the change in the action two ways: the symmetry makes it at most a boundary term, while the general variational identity makes it the Euler-Lagrange expression times the field variation plus a total derivative — and on a physical path the Euler-Lagrange piece vanishes. Equate the two and a total time derivative is pinned to zero; its interior is the charge. Nowhere is this clearer than in a planet's orbit: rotational symmetry of the Sun's central potential conserves angular momentum, and that single conserved number is Kepler's second law — the radius vector sweeps equal areas in equal times. The second figure makes it concrete, with the numbers from the first worked example.

Figure

Kepler's equal-area law is angular-momentum conservation in disguise: for a particle of mass 2.0 kg in a central potential, the line from the Sun sweeps equal areas in equal times because Lz = m(x ẏ − y ẋ) = 24.0 kg·m²∕s stays constant, while time-translation symmetry pins the energy at E = ½ m v² − k∕r = 2.67 J along the whole orbit

The core equations

Everything starts from the action — a single number assigned to each candidate path by integrating the Lagrangian along it. The whole theorem is a statement about how this number responds when the path and the clock are nudged.

S[q]=\int_{t_1}^{t_2} L\big(q^i(t),\dot q^i(t),t\big)\,\mathrm{d}t \qquad (1)

where: $S$ — the action, a functional of the path; $L$ — the Lagrangian; $q^i,\dot q^i$ — generalized coordinates and velocities, $i=1,\dots,n$ ; $t$ — time. Reading: the theorem asks which nudges of $q^i$ and $t$ leave this integral fixed. Physical paths are the stationary points of the action, and stationarity under fixed-endpoint variations is the Euler-Lagrange equation — the only place the equations of motion enter the proof, the "on-shell" condition that makes conservation more than an identity.

\frac{\partial L}{\partial q^i}-\frac{\mathrm{d}}{\mathrm{d}t}\frac{\partial L}{\partial \dot q^i}=0 \qquad (2)

where: $\partial L/\partial q^i$ — the generalized force; $p_i\equiv\partial L/\partial\dot q^i$ — the canonical momentum conjugate to $q^i$ . Reading: only on solutions does the Noether charge stop changing. A symmetry acts on both the clock and the coordinates: the most general continuous transformation displaces time by $\varepsilon\tau$ and the coordinates by $\varepsilon\eta^i$ , with $\varepsilon$ the smooth dial and $(\tau,\eta^i)$ the generators encoding which symmetry it is.

t\to t'=t+\varepsilon\,\tau,\qquad q^i(t)\to q'^i(t')=q^i(t)+\varepsilon\,\eta^i \qquad (3)

where: $\varepsilon$ — infinitesimal parameter; $\tau$ — generator of the time shift; $\eta^i$ — generator of the coordinate shift. Reading: choosing $(\tau,\eta^i)$ selects the symmetry, and the master formula converts it into a charge. The centerpiece is the charge itself: when the transformation is a symmetry (up to the boundary term $K$ ) and the path is a solution, this combination of momentum, energy, and boundary function is constant in time.

\frac{\mathrm{d}Q}{\mathrm{d}t}=0,\qquad Q=\frac{\partial L}{\partial\dot q^i}\,\eta^i-\Big(\frac{\partial L}{\partial\dot q^i}\dot q^i-L\Big)\tau-K \qquad (4)

where: $Q$ — the conserved Noether charge; $K$ — the boundary function of a quasi-symmetry, defined by $\delta L=\varepsilon\,\mathrm{d}K/\mathrm{d}t$ ; $H=p_i\dot q^i-L$ — the energy that multiplies $\tau$ . Reading: read off $Q$ by plugging the generators $(\tau,\eta^i,K)$ into this one line. In a field theory the time integral becomes a spacetime integral and the conserved number becomes a conserved current, obeying the continuity equation $\partial_\mu j^\mu=0$ — the local face of conservation, where charge cannot appear or vanish, only flow.

\partial_\mu j^\mu=0,\qquad j^\mu=\frac{\partial \mathcal L}{\partial(\partial_\mu\phi^a)}\,\bar\delta\phi^a-T^{\mu}{}_{\nu}\,\delta x^\nu - K^\mu \qquad (5)

where: $\mathcal L$ — the Lagrangian density, $S=\int\mathcal L\,\mathrm{d}^4x$ ; $\phi^a$ — the fields; $j^\mu$ — the conserved four-current; $T^{\mu}{}_{\nu}$ — the canonical stress-energy tensor; $\bar\delta\phi^a$ — the fixed-point field variation. Reading: internal symmetries ( $\delta x=0$ ) give matter currents; spacetime symmetries give $T^{\mu}{}_{\nu}$ . Integrating the time component over all space collapses the local law back to a global one — provided the current falls off fast enough that nothing leaks across the boundary.

Q=\int_{\mathbb R^3} j^0\,\mathrm{d}^3x,\qquad \frac{\mathrm{d}Q}{\mathrm{d}t}=0 \qquad (6)

where: $j^0$ — the charge density; $Q$ — the total conserved charge, assuming $j^i\to0$ at infinity. Reading: mechanics is field theory in $0+1$ dimensions, where (6) reduces to (4).

Derivation AMaster derivation of the Noether charge in mechanics

The goal is to carry a general continuous transformation of time and coordinates through the action and prove, without assuming any particular symmetry, that invariance up to a boundary term forces a specific combination of momentum and energy to be conserved along every solution. We compute $\delta S$ once, split it into an Euler-Lagrange bulk term and a total derivative, then spend the two hypotheses — symmetry and the equations of motion — once each. The one subtlety is that the transformation moves the integration limits as well as the integrand, and tracking that measure change is what produces the energy term $L\tau$ .

Step 1. Transform the action to $S'=\int_{t_1'}^{t_2'} L(q'^i(t'),\dot q'^i(t'),t')\,\mathrm{d}t'$ , changing both the integrand through the new fields and the limits through $t'=t+\varepsilon\tau$ . Why: a symmetry must leave the whole integral invariant, so integrand and domain are transported together.

Step 2. Change variables back to $t$ using $\mathrm{d}t'=(1+\varepsilon\dot\tau)\,\mathrm{d}t$ . Why: this Jacobian is the piece a fixed-time variation would miss, and it is the origin of the $L\tau$ term in the charge.

Step 3. Expand to first order in $\varepsilon$ . Writing the fixed-time variation $\bar\delta q^i=\eta^i-\dot q^i\tau$ , the measure factor and the explicit-time pieces combine into $\delta S=\varepsilon\int_{t_1}^{t_2}\big[\tfrac{\partial L}{\partial q^i}\bar\delta q^i+\tfrac{\partial L}{\partial\dot q^i}\dot{\overline{\delta q}}{}^i+\tfrac{\mathrm{d}}{\mathrm{d}t}(L\tau)\big]\mathrm{d}t$ . Why: separating the "vertical" field change $\bar\delta q^i$ from the time shift lets the Euler-Lagrange operator appear cleanly next.

Step 4. Integrate the velocity term by parts and collect: $\delta S=\varepsilon\int\big[\mathrm{EL}_i\,\bar\delta q^i+\tfrac{\mathrm{d}}{\mathrm{d}t}(p_i\bar\delta q^i+L\tau)\big]\mathrm{d}t$ , with $\mathrm{EL}_i=\tfrac{\partial L}{\partial q^i}-\tfrac{\mathrm{d}}{\mathrm{d}t}p_i$ . Why: every variational proof funnels through this split — a bulk term proportional to the equations of motion plus a total derivative that becomes the charge.

Step 5. Spend the two hypotheses. Symmetry means $\delta S=\varepsilon\int(\mathrm{d}K/\mathrm{d}t)\,\mathrm{d}t$ (a pure boundary term, with $K=0$ for a strict symmetry); the equations of motion mean $\mathrm{EL}_i=0$ on the path. Both bulk contributions vanish, leaving $\tfrac{\mathrm{d}}{\mathrm{d}t}(p_i\bar\delta q^i+L\tau-K)=0$ . Why: this is the on-shell nature of the theorem made explicit — kill the Euler-Lagrange term with the dynamics, kill the rest with the symmetry, and only a conserved total derivative survives.

Step 6. Substitute $\bar\delta q^i=\eta^i-\dot q^i\tau$ to read off the charge $Q=p_i\eta^i-(p_i\dot q^i-L)\tau-K$ , whose vanishing time derivative is equation (4). Why: momentum multiplies the coordinate generator, energy multiplies the time generator, the boundary term is subtracted — the complete Noether charge. Verification: for $\tau=1,\eta=0,K=0$ the charge is $Q=-(p_i\dot q^i-L)=-H$ , so energy is conserved whenever $\partial L/\partial t=0$ — exactly the Beltrami identity, a consistency check against Goldstein §13.7.

Go deeper · Derivation B · 1 minRotational invariance and the conservation of angular momentum

Derivation BRotational invariance and the conservation of angular momentum

The goal is to run the master formula on the most important mechanical symmetry — rotation of a central-force system — and watch it manufacture the conservation of angular momentum from nothing but the invariance of the kinetic energy and a radial potential. We write the infinitesimal rotation as its generator, confirm the action is strictly invariant so no boundary term appears, and substitute into $Q=p_i\eta^i$ . The payoff is that the abstract charge becomes the cross product $\mathbf r\times\mathbf p$ , and its constancy is Kepler's second law.

Step 1. Take $L=\tfrac12 m\,\dot{\mathbf r}^2-V(|\mathbf r|)$ and the rotation about the $z$ -axis, generated by $\tau=0$ and $\boldsymbol\eta=\hat{\mathbf z}\times\mathbf r=(-y,\,x,\,0)$ , so $\delta x=-\varepsilon y$ and $\delta y=\varepsilon x$ . Why: a rotation shifts the coordinates but not the clock, so $\tau=0$ and only the coordinate generator $\eta$ is active.

Step 2. Check invariance: $\delta L=m\dot{\mathbf r}\cdot\delta\dot{\mathbf r}-V'\,\hat{\mathbf r}\cdot\delta\mathbf r=0$ , because both $|\dot{\mathbf r}|^2$ and $|\mathbf r|$ are unchanged by rotation. Why: a strict symmetry ( $\delta L=0$ ) means the boundary function $K=0$ , so the charge is just $p_i\eta^i$ with no correction.

Step 3. Substitute the momenta $p_i=m\dot x_i$ into $Q=p_i\eta^i=m\dot x(-y)+m\dot y(x)=m(x\dot y-y\dot x)$ , which is the $z$ -component of $\mathbf L=\mathbf r\times\mathbf p$ . Why: the master charge with $\tau=0,K=0$ is exactly the projection of the angular momentum on the rotation axis.

Step 4. Conclude $\mathrm{d}L_z/\mathrm{d}t=0$ by Noether; repeating for the $x$ - and $y$ -axes gives the full vector $\mathbf L$ conserved for any central potential. Why: rotational symmetry about every axis holds for spherically symmetric $V$ , so all three components are conserved together. Verification: directly, $\dot L_z=m(x\ddot y-y\ddot x)$ , which vanishes because $m\ddot{\mathbf r}=-V'\hat{\mathbf r}\parallel\mathbf r$ — consistent with the central-force result in Goldstein §1.2.

Limiting cases

Specializing the generators $(\tau,\eta^i,K)$ in the master charge $Q=p_i\eta^i-(p_i\dot q^i-L)\tau-K$ recovers each elementary conservation law as one line of arithmetic.

Case	Result	Meaning
Time translation	$\tau=1,\ \eta^i=0\ \Rightarrow\ Q=-(p_i\dot q^i-L)=-H$	If $L$ has no explicit time dependence, the energy $H=p_i\dot q^i-L$ is conserved (the Beltrami identity in mechanics).
Space translation	$\tau=0,\ \eta^i=a^i\ \Rightarrow\ Q=p_i a^i$	If $L$ is invariant under shifting a coordinate ( $q^i\to q^i+a^i$ , a cyclic coordinate), the conjugate momentum $p_i$ is conserved.
Rotation	$\tau=0,\ \eta^i=(\boldsymbol\omega\times\mathbf r)^i\ \Rightarrow\ Q=\boldsymbol\omega\cdot(\mathbf r\times\mathbf p)$	Rotational invariance gives conservation of the angular-momentum component $\boldsymbol\omega\cdot\mathbf L$ along every symmetry axis.
Galilean boost	$\eta^i=v^i t,\ \delta L=\tfrac{\mathrm{d}}{\mathrm{d}t}(m v^i x_i)\ \Rightarrow\ Q=m x_i-p_i t$	A quasi-symmetry: $L$ changes by a total derivative, so $K=m v_i x^i$ is subtracted; the charge encodes uniform center-of-mass motion.
Internal phase	$\delta x=0,\ \delta\phi=i\alpha\phi\ \Rightarrow\ j^\mu\propto i(\phi^\partial^\mu\phi-\phi\,\partial^\mu\phi^)$	A global $U(1)$ symmetry of a complex field gives a conserved current whose charge is electric charge (or particle number).

Common trapA symmetry of the equations of motion is not the same as a symmetry of the action

Two traps recur, and both give the wrong charge. First, conservation is on-shell, not a kinematic identity: $\mathrm{d}Q/\mathrm{d}t=0$ is proven only after imposing the Euler-Lagrange equations, so it is a property of solutions, not of arbitrary paths. Second, and more subtly, the theorem requires invariance of the action, not merely of the equations of motion — genuinely different things. A scaling transformation can leave the equations of motion unchanged while rescaling the action by a constant, and it then produces no Noether charge. Only invariance of the action, up to a boundary term, is guaranteed to deliver a conserved quantity. The related error is dropping the quasi-symmetry term: when $L$ changes by a total derivative $\mathrm{d}K/\mathrm{d}t$ (as under a Galilean boost), forgetting to subtract $K$ from $Q$ yields a quantity that is not conserved. Scan for the boundary term before trusting the charge.

Worked examples

1 · The two Noether charges of a central-force orbit (fully worked)

A particle of mass $m=2.0\ \mathrm{kg}$ moves in the attractive central potential $V(r)=-k/r$ with $k=40\ \mathrm{J\cdot m}$ . At one instant it is at $\mathbf r=(3.0,\,0,\,0)\ \mathrm{m}$ with velocity $\mathbf v=(0,\,4.0,\,0)\ \mathrm{m/s}$ . Find the two conserved Noether charges — angular momentum and energy — and say which symmetry each comes from.

Step 1. The potential depends only on $|\mathbf r|$ , so the Lagrangian $L=\tfrac12 m\dot{\mathbf r}^2+k/r$ is invariant under rotation about any axis and under time translation. Noether hands us two charges: $\mathbf L=\mathbf r\times\mathbf p$ from rotation, and $H$ from time translation.

Step 2. The momentum is $\mathbf p=m\mathbf v=(0,\,8.0,\,0)\ \mathrm{kg\,m/s}$ , so the angular momentum is $\mathbf L=\mathbf r\times\mathbf p=(3.0,0,0)\times(0,8.0,0)=(0,\,0,\,24.0)$ , giving the rotation charge $L_z=m(x\dot y-y\dot x)=2.0(3.0\times4.0-0)=\mathbf{24.0\ kg\,m^2/s}$ .

Step 3. The energy comes from time-translation symmetry: $E=\tfrac12 m v^2-\dfrac{k}{r}=\tfrac12(2.0)(4.0)^2-\dfrac{40}{3.0}=16.0-13.33=\mathbf{2.67\ J}$ . Here $E>0$ , so the orbit is an unbound hyperbola — but $E$ and $L_z$ keep these values at every instant.

Step 4. That $L_z=24.0\ \mathrm{kg\,m^2/s}$ is constant is Kepler's second law — the radius vector sweeps area at the fixed rate $\mathrm{d}A/\mathrm{d}t=L_z/2m=6.0\ \mathrm{m^2/s}$ — while constant $E$ is what lets the orbit be solved by quadrature. Two symmetries, two charges, one Lagrangian. L_z = 24.0 kg·m²/s (rotation), E = 2.67 J (time translation).

2 · A Galilean boost and the conserved center of mass (quasi-symmetry)

A free particle of mass $m=1.0\ \mathrm{kg}$ starts at $x=0$ at $t=0$ and moves at $u=5.0\ \mathrm{m/s}$ . Under a Galilean boost the free Lagrangian is not invariant. Show that the Noether charge $Q=mx-pt$ is nonetheless conserved, and check it at two times.

Step 1. The boost is $x\to x+\varepsilon v t$ , so $\eta=vt$ and $\tau=0$ . The free Lagrangian $L=\tfrac12 m\dot x^2$ changes by $\delta L=m\dot x\,(v)=\dfrac{\mathrm{d}}{\mathrm{d}t}(mvx)$ , a total time derivative. This is a quasi-symmetry with boundary function $K=mvx$ .

Step 2. The master charge is $Q=p\,\eta-K=p(vt)-mvx=-v(mx-pt)$ . Dropping the arbitrary boost speed $v$ , the conserved quantity is $Q=mx-pt$ — the initial position of the center of mass, run backward from the current state.

Step 3. Check it. At $t=0$ : $x=0$ , $p=mu=5.0\ \mathrm{kg\,m/s}$ , so $Q=1.0(0)-5.0(0)=\mathbf{0}$ . At $t=2.0\ \mathrm{s}$ : $x=ut=10.0\ \mathrm{m}$ , still $p=5.0$ , so $Q=1.0(10.0)-5.0(2.0)=\mathbf{0}$ . The charge holds.

Step 4. Had we forgotten $K$ and written $Q=p\eta=pvt$ , the "charge" would grow linearly in time. The boundary term is the whole difference between a symmetry and a quasi-symmetry. Q = mx − pt = 0 at both t = 0 and t = 2 s; omitting K would have given a growing, non-conserved charge.

3 · The conserved current of a complex scalar field

A free complex scalar field has Lagrangian density $\mathcal L=\partial_\mu\phi^*\,\partial^\mu\phi-m^2\phi^*\phi$ , invariant under the global phase rotation $\phi\to e^{i\alpha}\phi$ . Construct the Noether current and confirm it is conserved on-shell.

Step 1. The transformation is $\delta\phi=i\alpha\phi$ , $\delta\phi^*=-i\alpha\phi^*$ , with no change of spacetime point ( $\delta x^\nu=0$ ). The current is the internal-symmetry piece of equation (5): $j^\mu=\dfrac{\partial\mathcal L}{\partial(\partial_\mu\phi)}\dfrac{\delta\phi}{\alpha}+\dfrac{\partial\mathcal L}{\partial(\partial_\mu\phi^*)}\dfrac{\delta\phi^*}{\alpha}$ .

Step 2. With $\partial\mathcal L/\partial(\partial_\mu\phi)=\partial^\mu\phi^*$ and $\partial\mathcal L/\partial(\partial_\mu\phi^*)=\partial^\mu\phi$ , substitution gives the conserved four-current $j^\mu=\partial^\mu\phi^*(i\phi)+\partial^\mu\phi(-i\phi^*)=i\big(\phi\,\partial^\mu\phi^*-\phi^*\partial^\mu\phi\big)$ .

Step 3. Its divergence is $\partial_\mu j^\mu=i\big(\phi\,\Box\phi^*-\phi^*\Box\phi\big)$ ; the kinetic cross-terms cancel identically.

Step 4. The Klein-Gordon equations $\Box\phi+m^2\phi=0$ and $\Box\phi^*+m^2\phi^*=0$ then give $\partial_\mu j^\mu=i\big(\phi(-m^2\phi^*)-\phi^*(-m^2\phi)\big)=0$ — conserved, but only on-shell. The charge $Q=\int j^0\,\mathrm{d}^3x$ is the electric charge (or net particle-minus-antiparticle number). j^μ = i(φ ∂^μφ − φ ∂^μφ), with ∂_μ j^μ = 0 by Klein-Gordon.

Physics of the portrait

Noether's theorem is a rare result whose name, statement, and proof are all due to one person — who had to fight for the right to lecture at all.

Origins · 1 minEmmy Noether and how this was discovered

Emmy Noether

1882-1935 · invariant theory and the algebra of conservation laws

Amalie Emmy Noether was the greatest algebraist of her generation and, by Einstein's assessment in his 1935 obituary, "the most significant creative mathematical genius thus far produced since the higher education of women began." Born in Erlangen to the mathematician Max Noether, she took her doctorate in 1907 on invariant theory under Paul Gordan, then worked for years without pay or title because German universities did not admit women to the faculty. In 1915 David Hilbert and Felix Klein brought her to Göttingen because they were stuck: general relativity seemed to threaten the conservation of energy, and Hilbert believed her mastery of invariants could resolve it. It could. Her 1918 paper Invariante Variationsprobleme contains the two theorems that now bear her name: the first, the subject of this article, showed that the conservation laws of physics are the shadows of continuous symmetries, and the second explained the peculiar identity-like form energy conservation takes in general relativity. She went on to found modern abstract algebra — her work on ideals and chain conditions gave us the word "Noetherian" — before the Nazi regime expelled her as a Jew in 1933. She emigrated to Bryn Mawr College and died there in 1935, at 53. Physicists who never open an algebra textbook use her theorem every day.

History

1834

— Hamilton casts mechanics as the stationary action, the setting the theorem needs

1888

— Lie's theory of continuous transformation groups gives symmetry its algebra

1915

— Hilbert and Klein bring Noether to Göttingen to rescue energy conservation in general relativity

1918

— Noether publishes Invariante Variationsprobleme, proving both her theorems

1954

— Yang and Mills build non-abelian gauge field theory on conserved Noether currents

Going further

Texts. Stone & Goldbart, Mathematics for Physics, Ch. 1 (calculus of variations and Noether's theorem) · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., Ch. 22 · Hassani, Mathematical Physics, Ch. 36 (Lagrangian dynamics and symmetries) · Nakahara, Geometry, Topology and Physics, Ch. 1 · Byron & Fuller, Mathematics of Classical and Quantum Physics, Ch. 2 · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §13.7 · Noether, E. (1918), Invariante Variationsprobleme, Nachr. Ges. Wiss. Göttingen. The master identity follows from the $\varepsilon$ -expansion of the action and integration by parts. The time-translation case gives $Q=-H$ , the Beltrami identity. The orbit numbers $L_z=24.0\ \mathrm{kg\,m^2/s}$ and $E=2.67\ \mathrm{J}$ follow from $m=2.0\ \mathrm{kg}$ , $\mathbf r=(3,0,0)\ \mathrm{m}$ , $\mathbf v=(0,4,0)\ \mathrm{m/s}$ , $k=40\ \mathrm{J\,m}$ . The Galilean charge $Q=mx-pt=0$ at both $t=0$ and $t=2\ \mathrm{s}$ for $m=1.0\ \mathrm{kg}$ , $u=5.0\ \mathrm{m/s}$ . The current $j^\mu=i(\phi\,\partial^\mu\phi^*-\phi^*\partial^\mu\phi)$ satisfies $\partial_\mu j^\mu=0$ via the Klein-Gordon equations, and $\mathrm{d}Q/\mathrm{d}t=0$ then follows from $\partial_\mu j^\mu=0$ by the divergence theorem. All formulas are standard; no nonstandard constants are introduced.

Prerequisites. Functionals and Function Spaces · First Variation and Stationarity · Euler-Lagrange Equation · Hamilton's Principle of Stationary Action · Beltrami Identity and First Integrals. Leads to. Constrained Variation and Lagrange Multipliers · Geodesics as Variational Problems · Functionals of Several Functions and Higher Derivatives.

Ask the tutor: Name any symmetry — a rotation, a boost, a phase, a Killing vector — and I will run Noether's master formula to hand you the conserved charge, with the numbers checked.

Is Noether's theorem the reason energy is conserved, or just a restatement of it?

It is a genuine explanation. Before Noether, conservation of energy was an empirical law abstracted from experiment. She showed it is a consequence of a deeper fact — that the laws of physics do not depend explicitly on time. Feed the theorem the generator $\tau=1,\eta=0$ and it derives the conserved quantity $-H=L-p_i\dot q^i$ , the energy. The explanatory power is real: it predicts which quantity is conserved from the symmetry alone, and tells you exactly when energy is not conserved — when $L$ depends explicitly on time, as for a driven oscillator, where $\mathrm{d}H/\mathrm{d}t=-\partial L/\partial t\neq0$ .

Why must the equations of motion be used — isn't the symmetry enough?

The symmetry alone is not enough, and seeing why is the heart of the theorem. Invariance of the action tells you $\delta S$ equals a pure boundary term for arbitrary nearby paths. But the general variational identity writes $\delta S$ as (Euler-Lagrange expression) × (variation) plus a total derivative. To conclude that the total derivative alone vanishes — the conservation law — you must first kill the Euler-Lagrange term, and it vanishes only on solutions. Hence conservation is "on-shell": the symmetry buys the form of the charge; the dynamics buys its constancy.

What is a quasi-symmetry, and why does it still give a conserved charge?

A quasi-symmetry changes the Lagrangian by a total time derivative, $\delta L=\varepsilon\,\mathrm{d}K/\mathrm{d}t$ , rather than leaving it strictly invariant. The action then changes only by boundary terms, which do not affect the equations of motion, so the variational problem is untouched. The theorem still applies, provided you subtract $K$ from the charge: $Q=p_i\eta^i-(p_i\dot q^i-L)\tau-K$ . The canonical example is the Galilean boost of a free particle, where $L$ shifts by $\mathrm{d}(mvx)/\mathrm{d}t$ ; the corrected charge $Q=mx-pt$ encodes uniform center-of-mass motion. Forgetting the $K$ term is the most common way to get the wrong answer.

How does the mechanical charge Q relate to the field-theory current j^μ?

Mechanics is field theory in $0+1$ dimensions — one time coordinate, no space. There the index $\mu$ takes only the time value, the continuity equation $\partial_\mu j^\mu=0$ reduces to $\mathrm{d}j^0/\mathrm{d}t=0$ , and $j^0$ is the mechanical charge $Q$ . In a genuine field theory the density $j^0(\mathbf x,t)$ lives at every point and the spatial components $j^i$ describe how it flows. The integral $Q=\int j^0\,\mathrm{d}^3x$ recovers a single conserved number, but the local statement is strictly stronger: it forbids charge from teleporting between regions, insisting it flow through the boundary.

Does invariance of the equations of motion alone guarantee a conserved charge?

No, and this is a genuine subtlety. There exist transformations that map solutions of the equations of motion to other solutions — symmetries of the dynamics — that are nevertheless not symmetries of the action. A scaling transformation is the classic case: it can rescale the action by an overall constant, leaving the Euler-Lagrange equations unchanged yet $S$ non-invariant, and it produces no Noether charge. Only invariance of the action, up to a boundary term, guarantees a conserved quantity. So when hunting for conservation laws, always check the action, not just the equations of motion.

What is Noether's second theorem, and how is it different?

Noether's 1918 paper contains two theorems. The first, treated here, concerns global continuous symmetries — those with finitely many constant parameters, like a single rotation angle — and produces genuine conservation laws. The second concerns local or gauge symmetries, where the parameter is an arbitrary function of spacetime, like the position-dependent phase of electromagnetism. Rather than a conservation law, it produces identities among the equations of motion (the Bianchi identities and gauge constraints), reflecting a redundancy in the description. This is why "energy conservation" in general relativity looks strange: gravity's diffeomorphism symmetry is a second-theorem symmetry, so the conserved energy is not a simple integral but a subtle, frame-dependent quantity.