Noether's Theorem for Continuous Symmetries

One theorem in this chapter outranks all the others, because everything the remaining articles compute — stress-energy tensors, angular-momentum currents, internal charges, the generators the Poisson-bracket article turns loose on phase space — is an output of a single machine. Noether's theorem for fields converts the geometric statement "the action does not change under a continuous transformation" into the dynamical statement "a definite current obeys a continuity equation." You met the particle version in the Lagrangian chapter: one symmetry, one conserved number. The field version is strictly stronger and strictly more physical: one symmetry, one conserved current $j^\mu$ — a density plus a flux, locally accounted at every point of spacetime. Charge cannot vanish here and reappear there; it must flow through the boundary in between. Relativity, it turns out, demands exactly this local form, and the conserved-currents article makes that demand precise. The proof is a masterpiece of economy — one chain rule, one integration by parts, one appeal to the equations of motion — and this article walks it in full, then cranks the machine on the two great symmetry families: spacetime translations (out comes the stress-energy tensor) and internal phase rotations (out comes the prototype of electric charge). The deep-dive anatomy of those outputs belongs to the four articles that follow; here you learn the machine itself, including its two famous fine-print clauses — the total-divergence allowance $K^\mu$ that trips even professionals, and the improvement ambiguity that the Belinfante-Rosenfeld article turns from nuisance into tool.

Key takeaways

Every continuous symmetry of the field action yields a locally conserved current: $\partial_\mu j^\mu = 0$ — the field-theory upgrade of the particle theorem's conserved number.
The current is built from the field's response to the symmetry: $j^\mu = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\Delta\phi_a - K^\mu$ , where $K^\mu$ compensates any total-divergence shift of $\mathcal{L}$ .
Integrating the time component gives a constant charge, $Q = \int\mathrm{d}^3x\, j^0$ , provided currents fall off at spatial infinity.
Spacetime translations produce the canonical stress-energy tensor $T^{\mu\nu}$ ; Lorentz transformations produce the angular-momentum current; internal symmetries produce charges like electric charge.
Only invariance of the action is required — $\mathcal{L}$ may shift by $\partial_\mu K^\mu$ ; forgetting to subtract $K^\mu$ yields a "current" that is not conserved.
Conservation is on-shell: it holds when the fields obey the Euler-Lagrange equations, not identically.
The current is unique only up to identically conserved improvement terms $\partial_\nu B^{[\nu\mu]}$ — same charge, different density; the Belinfante-Rosenfeld construction lives in this freedom.
Key idea: symmetry of the action + equations of motion = a local continuity equation, one for every continuous parameter.

The core insight

Vary the action under an infinitesimal symmetry; the Euler-Lagrange equations kill the bulk variation, leaving a pure boundary term whose vanishing is exactly the local conservation law $\partial_\mu j^\mu = 0$ .

Where this lives today

Noether's theorem is the organizing syntax of fundamental physics: the Standard Model is specified by declaring its symmetries, and the theorem converts that declaration into the conserved quantities experiments actually track — electric charge, color, baryon and lepton number (the last two "accidental" symmetries of the renormalizable Lagrangian, and their possible tiny violation is why proton-decay detectors like Super-Kamiokande watch 50,000 tonnes of water for a single flash). The quantum version of the theorem — Ward-Takahashi and Slavnov-Taylor identities — is load-bearing machinery: it enforces the cancellations that make gauge-theory calculations finite, and every LHC cross-section prediction passes through it. Where the theorem's hypotheses fail, modern physics finds its most interesting structure: spontaneously broken symmetries still deliver currents but realize them as Goldstone bosons (the pion's existence and low-energy interactions are Noether bookkeeping for broken chiral symmetry), while anomalies — symmetries of the action that quantization cannot preserve — turn the on-shell fine print into measurable physics like the $\pi^0 \to \gamma\gamma$ decay rate. Condensed-matter theory runs the theorem daily in its hydrodynamic form: the conserved densities of a material (energy, momentum, particle number, spin) are exactly the slow variables of transport theory, so even the Navier-Stokes equations of the continuum chapter are downstream of symmetry counting. Noether's second theorem, meanwhile, quietly governs every gauge theory and general relativity itself — the constraint structure the Hamiltonian article flagged for gauge fields is its fingerprint. Within the chapter: the next four articles are this theorem's outputs catalogued — currents and charges in general, the internal-symmetry charge, the stress-energy tensor, and the angular-momentum current — before the Poisson-bracket machinery turns every one of those charges into a generator.

The physical picture

Two pictures, one for each half of the theorem's content. The first is the flat direction. Think of the action as a cost assigned to every conceivable field history, with physical histories sitting at stationary points. A continuous symmetry is a direction in history space along which the cost is exactly flat — slide any history along it and the action does not budge. Sliding a stationary history along a flat direction lands on another stationary history, so symmetries map solutions to solutions; and the theorem's computation shows that the flatness, evaluated on a solution, condenses into a quantity whose total time derivative is a spatial divergence — something that cannot change except by flowing.

Figure

The two halves of the theorem — a flat direction of the action becomes a locally accounted flow in spacetime

The second picture is the ledger. The continuity equation $\partial_t j^0 + \nabla\cdot\mathbf{j} = 0$ is bookkeeping with teeth: integrate it over any region and the charge inside changes exactly by the flux through the walls. This locality is the genuinely new content compared with the particle theorem — a particle's conserved energy is one global number, but a field's energy has a location, and Noether's theorem tells you not just that the total is constant but how every joule moves from place to place. That is why the theorem's output is a four-component object $j^\mu = (j^0, \mathbf{j})$ : the density and its flux are two faces of one spacetime vector, and only together do they make a frame-independent statement. One organizational fact tames the zoo of examples: the current always splits into a "field response" piece $\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi)}\Delta\phi$ and, for symmetries that move spacetime itself, a "dragged density" piece $K^\mu$ . Internal symmetries (phase rotations) have only the first; translations and Lorentz transformations carry both. Every conserved current in physics you will ever meet is an instance of this one two-term template.

The master identity

The setting is the chapter's standard one: fields $\phi_a(x)$ (the index spans multiplet components), Lagrangian density $\mathcal{L}(\phi_a, \partial_\mu\phi_a)$ , action $S = \int\mathcal{L}\,\mathrm{d}^4x$ . A continuous transformation with parameter $\epsilon$ changes the fields by $\delta\phi_a = \epsilon\,\Delta\phi_a$ , where $\Delta\phi_a$ may depend on the fields and their derivatives. The transformation is a symmetry when it leaves the action invariant — which, at the level of the density, allows a total divergence:

\delta\mathcal{L} = \epsilon\,\partial_\mu K^\mu \qquad (1)

where: $K^\mu$ is a definite four-vector determined by the transformation (zero for a strict invariance like a phase rotation, nonzero for translations, boosts, and Galilean shifts). This allowance is not a loophole but a necessity: a total divergence integrates to a boundary term and cannot affect the equations of motion, so demanding $\delta\mathcal{L} = 0$ outright would blind the theorem to some of the most important symmetries in physics. The theorem's output is the current

j^\mu = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\Delta\phi_a - K^\mu, \qquad \partial_\mu j^\mu = 0 \quad \text{(on-shell)} \qquad (2)

where: the first term is the canonical momentum flowing in the $\mu$ direction times the field's symmetry response (summed over $a$ ), and the subtraction of $K^\mu$ is mandatory whenever (1) has a right-hand side. "On-shell" means the conservation holds for fields obeying their Euler-Lagrange equations — the theorem says nothing about arbitrary field configurations. The derivation is three moves long.

Derivation AThe master Noether identity for internal symmetries

We derive the conservation law (2) for transformations that act on the fields alone, coordinates fixed — the cleanest setting, and the template every other case reduces to. The strategy: expand $\delta\mathcal{L}$ by the chain rule, massage the derivative term into "total divergence plus Euler-Lagrange bracket" by the product rule, then set the fields on-shell so the bracket dies. What remains equates two total divergences, and their difference is the conserved current. Watch where the equations of motion enter — that step is what makes the result on-shell only.

Step 1. Chain rule. Under $\phi_a \to \phi_a + \delta\phi_a$ with $\delta(\partial_\mu\phi_a) = \partial_\mu(\delta\phi_a)$ (variation and derivative commute for coordinate-fixed transformations):

\delta\mathcal{L} = \frac{\partial\mathcal{L}}{\partial\phi_a}\,\delta\phi_a + \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\partial_\mu(\delta\phi_a)

Step 2. Product rule on the second term, at the level of the integrand:

\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\partial_\mu(\delta\phi_a) = \partial_\mu\!\left(\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a\right) - \partial_\mu\!\left(\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\right)\delta\phi_a

Substituting back regroups the variation into Euler-Lagrange bracket plus divergence:

\delta\mathcal{L} = \underbrace{\left[\frac{\partial\mathcal{L}}{\partial\phi_a} - \partial_\mu\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\right]}_{\text{EL}_a}\delta\phi_a + \partial_\mu\!\left(\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\delta\phi_a\right)

Why this regrouping is the heart of the proof: it is the same integration-by-parts that derived the Euler-Lagrange equations in the field-equations article, recycled — but now the variation is a symmetry direction rather than an arbitrary one, so instead of discarding the boundary term and keeping the bracket, we discard the bracket and keep the boundary term.

Step 3. Go on-shell and invoke the symmetry. On solutions, $\text{EL}_a = 0$ . The symmetry condition (1) fixes the left side. With $\delta\phi_a = \epsilon\,\Delta\phi_a$ and $\epsilon$ constant:

\epsilon\,\partial_\mu K^\mu = \epsilon\,\partial_\mu\!\left(\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\Delta\phi_a\right) \quad\Longrightarrow\quad \partial_\mu\!\left(\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\Delta\phi_a - K^\mu\right) = 0

which is (2). One conserved current per continuous parameter: a symmetry group with $n$ parameters delivers $n$ currents. Why on-shell matters: Step 3 used $\text{EL}_a = 0$ . Off-shell, the identity instead reads $\partial_\mu j^\mu = -\text{EL}_a\,\Delta\phi_a$ — a useful formula in its own right (it quantifies how badly the current leaks when the equations of motion are violated), but conservation it is not. *Verified: matches Peskin & Schroeder §2.2 and Goldstein §13.7; setting fields to depend on $t$ only reduces every step to the particle-mechanics proof of the Noether article in the Lagrangian chapter, with $j^0 \to p\,\Delta q - K$ .*

The conserved charge follows by integrating the time component over a spatial slice:

Q = \int_{\mathbb{R}^3} \mathrm{d}^3x\; j^0, \qquad \frac{\mathrm{d}Q}{\mathrm{d}t} = -\oint_{\infty} \mathbf{j}\cdot\mathrm{d}\mathbf{A} = 0 \qquad (3)

where: the surface term vanishes for localized field configurations — the standing fine print on every "constant charge" claim. The charge-current anatomy, including what happens when the fine print fails, is the next article's whole subject; here we only note that (3) is where the field theorem hands back the particle theorem's kind of statement: one number, constant in time.

Cranking the machine: the two great families

Spacetime translations. Under $x^\mu \to x^\mu + a^\mu$ , the field at a fixed coordinate changes because the configuration is dragged past it: $\Delta\phi_a = -a^\nu\partial_\nu\phi_a$ . The density is dragged too, so this is the archetype of a symmetry with $K^\mu \neq 0$ .

Go deeper · Derivation B · 1 minThe stress-energy tensor from translation invariance

Derivation BThe stress-energy tensor from translation invariance

We apply the master identity to spacetime translations and obtain the canonical stress-energy tensor — four conserved currents at once, one for each direction of translation, packaging energy and momentum conservation into a single object. The strategy is pure substitution: identify $\Delta\phi$ and $K^\mu$ for a translation, insert into (2), and strip off the arbitrary parameter. The step needing care is $K^\mu$ : the Lagrangian density is not invariant under translations — it is transported — and that transport is exactly a total divergence.

Step 1. Identify the response. For $\phi_a(x) \to \phi_a(x - a)$ : $\Delta\phi_a = -a^\nu\,\partial_\nu\phi_a$ per unit $a$ . The density, having no explicit $x$ -dependence (a closed system), is transported the same way:

\delta\mathcal{L} = -a^\nu\partial_\nu\mathcal{L} = \partial_\mu(-a^\mu\mathcal{L}) \quad\Longrightarrow\quad K^\mu = -a^\mu\mathcal{L}

Step 2. Insert into the master current (2):

j^\mu = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,(-a^\nu\partial_\nu\phi_a) - (-a^\mu\mathcal{L}) = -a^\nu\left[\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\partial_\nu\phi_a - \delta^\mu_\nu\,\mathcal{L}\right]

Step 3. Strip the parameter. The four components $a^\nu$ are independent constants, so each bracket is separately conserved. Raising the index with $\eta$ :

T^{\mu\nu} = \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}\,\partial^\nu\phi_a - \eta^{\mu\nu}\mathcal{L}, \qquad \partial_\mu T^{\mu\nu} = 0 \qquad (4)

where: $\nu$ labels which translation (which conserved current), $\mu$ labels the flow direction. $\nu = 0$ is energy conservation; $\nu = i$ is momentum conservation.

Step 4. Sanity check on the free scalar, $\mathcal{L} = \tfrac12(\partial\phi)^2 - \tfrac12 m^2\phi^2$ : the $00$ component evaluates to $T^{00} = \tfrac12\dot\phi^2 + \tfrac12(\nabla\phi)^2 + \tfrac12 m^2\phi^2$ — precisely the Hamiltonian density of the Hamiltonian-formulation article, positive and correct. The theorem has derived what that article defined. Verified: matches Goldstein §13.7 and Landau-Lifshitz Classical Theory of Fields §32; $T^{00}$ cross-checked against the Legendre-transform result of the Hamiltonian article. The full anatomy of $T^{\mu\nu}$ — component meanings, asymmetry disease, examples — is owned by the stress-energy article; here it stands as the theorem's premier output.

Internal symmetries. The complementary family touches no coordinate: the transformation rotates field components among themselves, $K^\mu = 0$ , and the current is the bare response term. The cleanest example carries the rest of the chapter's charge story.

Go deeper · Derivation C · 1 minThe U(1) current of a complex scalar field

Derivation CThe U(1) current of a complex scalar field

We crank the machine on the complex scalar $\mathcal{L} = \partial_\mu\phi^*\partial^\mu\phi - m^2\phi^*\phi$ , invariant under the global phase rotation $\phi \to e^{i\alpha}\phi$ — the prototype of every internal symmetry in the Standard Model. Strategy: treat $\phi$ and $\phi^*$ as independent fields (the standard trick, justified because their real and imaginary parts are independent), read off $\Delta\phi$ for each, and sum the response terms. Since the density is strictly invariant, $K^\mu = 0$ and the current is two terms with no subtraction. We then verify conservation explicitly against the Klein-Gordon equation — a habit worth keeping.

Step 1. First-order responses. From $\phi \to e^{i\alpha}\phi \approx \phi + i\alpha\phi$ :

\Delta\phi = i\phi, \qquad \Delta\phi^* = -i\phi^*

Step 2. Canonical momenta: $\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi)} = \partial^\mu\phi^*$ and $\frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi^*)} = \partial^\mu\phi$ . Sum the two response terms of (2):

j^\mu = \partial^\mu\phi^*\,(i\phi) + \partial^\mu\phi\,(-i\phi^*) = i\big(\phi\,\partial^\mu\phi^* - \phi^*\partial^\mu\phi\big)

(Overall sign and normalization are convention; the physics is in the relative sign between the two terms, which makes the current vanish identically for a real field — a real scalar carries no charge.)

Step 3. Verify conservation on-shell. Take the divergence; the $\partial^\mu\phi^*\partial_\mu\phi$ cross terms cancel between the two products:

\partial_\mu j^\mu = i\big(\phi\,\Box\phi^* - \phi^*\Box\phi\big) = i\big(\phi(-m^2\phi^*) - \phi^*(-m^2\phi)\big) = 0

using the Klein-Gordon equations $\Box\phi = -m^2\phi$ and its conjugate — the on-shell input, arriving exactly where the master derivation said it would. Verified: matches the standard Klein-Gordon charge current (Peskin & Schroeder eq. 2.15 up to sign convention); the cancellation pattern recomputed term by term. The physical reading of this current — antiparticles, why $j^0$ is not positive-definite, gauging — is owned by the internal-symmetries article.

Figure

The Noether harvest — one machine, every conservation law of the chapter, with the article that owns each output

The remaining family — Lorentz rotations and boosts, with their orbital-plus-spin current $M^{\mu\nu\rho} = x^\nu T^{\mu\rho} - x^\rho T^{\mu\nu} + \frac{\partial\mathcal{L}}{\partial(\partial_\mu\phi_a)}(\Sigma^{\nu\rho})_{ab}\phi_b$ — runs on the same rails with one new ingredient (the spin matrix $\Sigma$ ), and has its own article. So does the theorem's fine print about uniqueness: the replacement

j'^\mu = j^\mu + \partial_\nu B^{\nu\mu}, \qquad B^{\nu\mu} = -B^{\mu\nu} \qquad (5)

where: $B$ is any antisymmetric tensor built from the fields — leaves $\partial_\mu j'^\mu = \partial_\mu j^\mu$ (antisymmetry kills the double divergence) and leaves the charge unchanged (the added $j'^0 - j^0 = \partial_i B^{i0}$ is a pure spatial divergence integrating to a boundary term). Same conservation law, same charge, different local density. This is the improvement ambiguity, and it is the entire subject of the Belinfante-Rosenfeld article.

Limiting cases

Case	Result	Meaning
fields depend on $t$ only	$\frac{\mathrm{d}}{\mathrm{d}t}(p\,\Delta q - K) = 0$	the particle Noether theorem of the Lagrangian chapter, recovered exactly
time translation	$\partial_\mu T^{\mu 0} = 0$ : $E = \int T^{00}\mathrm{d}^3x$ const	energy conservation; $T^{00} = \mathcal{H}$
space translation	$\partial_\mu T^{\mu i} = 0$ : $P^i = \int T^{0i}\mathrm{d}^3x$ const	field momentum conservation
internal $U(1)$	$j^\mu = i(\phi\,\partial^\mu\phi^* - \phi^*\partial^\mu\phi)$	prototype of electric charge
real scalar field	$U(1)$ current $\equiv 0$	no internal symmetry, no charge — chargeless field
strict invariance	$K^\mu = 0$ : $j^\mu$ = response term only	internal symmetries; the simple half of the template
explicit $x$ -dependence in $\mathcal{L}$	translation symmetry broken	$\partial_\mu T^{\mu\nu} \neq 0$ : external backgrounds pump energy-momentum
discrete symmetry (parity, $\phi \to -\phi$ )	no current	theorem needs a continuous parameter to differentiate along

Common trapInvariance of the action is not invariance of the Lagrangian — forget K-mu and your current leaks

The single most common Noether error: demanding $\delta\mathcal{L} = 0$ and building the current from the response term alone. Many bona fide symmetries — spacetime translations above, boosts, Galilean boosts in nonrelativistic theory — shift the density by a total divergence $\delta\mathcal{L} = \epsilon\,\partial_\mu K^\mu$ , which leaves the action invariant (boundary term) and the physics symmetric. For these, the conserved current is response minus $K^\mu$ ; omit the subtraction and you will compute a "current" whose divergence is $-\partial_\mu K^\mu \neq 0$ and spend an afternoon hunting a sign error that is actually a concept error. The translation derivation above is the canonical exhibit: without $K^\mu = -a^\mu\mathcal{L}$ there is no $-\eta^{\mu\nu}\mathcal{L}$ term in $T^{\mu\nu}$ , and the "energy density" comes out as $\dot\phi\,\pi$ instead of the correct $\tfrac12\dot\phi^2 + \tfrac12(\nabla\phi)^2 + \tfrac12 m^2\phi^2$ — wrong by exactly the Legendre-transform structure. And remember the second clause of the fine print: conservation is on-shell. Noether's identity off-shell reads $\partial_\mu j^\mu = -\text{EL}_a\Delta\phi_a$ ; the current of a field configuration that does not solve the equations of motion is under no obligation to be conserved.

Worked examples

1 · Charge and current of a plane-wave complex scalar (confidence builder)

A complex scalar plane wave $\phi = A\,e^{-i(\omega t - kx)}$ with $A = 0.10$ , $m = 1.00$ , $k = 0.75$ (natural units) solves Klein-Gordon. Compute $\omega$ , the charge density $j^0$ , the current density $j^x$ , and check the transport speed $j^x/j^0$ .

Step 1. Dispersion: $\omega = \sqrt{k^2 + m^2} = \sqrt{0.5625 + 1.0000} = \sqrt{1.5625} = \mathbf{1.250}$ — exact, by design of the numbers.

Step 2. Charge density. With the convention $j^\mu = i(\phi^*\partial^\mu\phi - \phi\,\partial^\mu\phi^*)$ (opposite overall sign to the derivation's — both legal, fixed once per book): $\partial_t\phi = -i\omega\phi$ , so $\phi^*\partial_t\phi - \phi\,\partial_t\phi^* = -2i\omega|A|^2$ and $j^0 = 2\omega A^2 = 2(1.250)(0.0100) = \mathbf{0.0250}$ per unit volume — constant in space and time, as a single plane wave's must be.

Step 3. Spatial current: $\partial_x\phi = ik\phi$ gives $j^x = 2kA^2 = 2(0.75)(0.0100) = \mathbf{0.0150}$ .

Step 4. Transport speed: $j^x/j^0 = k/\omega = 0.75/1.250 = \mathbf{0.600}$ . Compare the wave's group velocity $v_g = \mathrm{d}\omega/\mathrm{d}k = k/\omega = 0.600$ : charge rides at the group velocity, not the phase velocity $\omega/k = 1.667$ (which exceeds 1 and transports nothing). Continuity is trivially satisfied here — both densities are constants — but the ratio just computed is the honest, frame-consistent content: this wave carries charge at $0.6c$ .

2 · Energy conservation from the theorem, checked to the digit (exam level)

A real scalar standing wave in a box of length $L = 10.0$ with periodic boundaries: $\phi = A\cos(kx)\cos(\omega t)$ , $A = 0.20$ , $m = 0.60$ , $k = 2\pi/L$ . Verify that the Noether energy $E = \int_0^L T^{00}\,\mathrm{d}x$ is time-independent even though every term of the integrand oscillates.

Step 1. Ingredients: $k = 2\pi/10.0 = 0.6283$ , $\omega^2 = k^2 + m^2 = 0.3948 + 0.3600 = 0.7548$ , $\omega = \mathbf{0.8688}$ . Derivatives: $\dot\phi = -A\omega\cos(kx)\sin(\omega t)$ , $\partial_x\phi = -Ak\sin(kx)\cos(\omega t)$ .

Step 2. Integrate each piece of $T^{00} = \tfrac12\dot\phi^2 + \tfrac12(\partial_x\phi)^2 + \tfrac12 m^2\phi^2$ using $\int_0^L\cos^2(kx)\,\mathrm{d}x = \int_0^L\sin^2(kx)\,\mathrm{d}x = L/2$ :

E(t) = \frac{LA^2}{4}\Big[\omega^2\sin^2(\omega t) + k^2\cos^2(\omega t) + m^2\cos^2(\omega t)\Big]

Step 3. The dispersion relation is the conservation mechanism: $k^2 + m^2 = \omega^2$ , so the two $\cos^2$ terms fuse and $E = \frac{LA^2\omega^2}{4}\big[\sin^2 + \cos^2\big] = \frac{LA^2\omega^2}{4}$ — all time dependence cancels identically. Numerically: $E = \frac{10.0 \times 0.0400 \times 0.7548}{4} = \mathbf{0.0755}$ .

Step 4. Spot-check two instants. At $\omega t = 0$ : potential-type terms only, $E = \frac{LA^2}{4}(k^2 + m^2) = \frac{0.4}{4}(0.7548) = 0.0755$ ✓. At $\omega t = \pi/2$ : kinetic only, $E = \frac{LA^2\omega^2}{4} = 0.0755$ ✓. The energy sloshes between kinetic and (gradient + mass) reservoirs at $2\omega$ while the Noether total holds to the last digit — and it holds because $\omega^2 = k^2 + m^2$ : on-shell conservation, visible in arithmetic.

Origins · 1 minEmmy Noether and how this was discovered

Emmy Noether

1882–1935 · Göttingen and Bryn Mawr

Noether proved her theorem in 1918 to settle a crisis in general relativity: Hilbert and Klein had found that energy conservation in Einstein's brand-new theory seemed to hold only vacuously, and they recruited Göttingen's finest invariant-theorist to sort out why. Her paper "Invariante Variationsprobleme" answered with two theorems — the first (this article's) tying each finite-parameter symmetry to a conserved current, the second explaining the relativity puzzle itself: when the symmetry group is infinite-dimensional (spacetime-dependent parameters, as in GR's coordinate freedom), the conservation laws become identities among the field equations rather than independent statements. That she did this as an unpaid, untitled lecturer — Göttingen's philosophical faculty having refused to habilitate a woman, with Hilbert famously retorting that the senate was not a bathhouse — belongs to the theorem's history as much as the mathematics. Her deeper fame among mathematicians rests elsewhere: she rebuilt abstract algebra (rings, ideals, the ascending chain condition — "Noetherian" is an adjective in every algebra text) and taught the structural, axiomatic style that defines the subject. Dismissed by the Nazi race laws in 1933, she emigrated to Bryn Mawr College and died two years later at 53. Einstein's obituary called her the most significant creative mathematical genius yet produced since the higher education of women began; physics remembers her differently and just as absolutely — as the person who told us what conservation laws are.

History

1842

— Jacobi connects symmetries to first integrals in point mechanics — conserved quantities from cyclic coordinates, the theorem's ancestor.

1915

— Hilbert and Klein, wrestling with energy conservation in general relativity, bring the problem to Noether.

1918

— "Invariante Variationsprobleme": both Noether theorems, proved in full generality for arbitrary variational systems.

1954

— Yang and Mills gauge an internal symmetry; Noether currents become the source terms of non-abelian force fields.

1971

— 't Hooft's renormalization proof leans on the Ward-Takahashi identities — quantum Noether — sealing the theorem's role in the Standard Model.

Going further

References. Goldstein, Poole & Safko, Classical Mechanics (3rd ed.), §13.7 (Noether's theorem) · Landau & Lifshitz, Mechanics (Vol. 1), §6–9 and Classical Theory of Fields §32 · Morin, Introduction to Classical Mechanics, Ch. 5–6 (symmetries and conservation) · Taylor, Classical Mechanics, Ch. 7 (Lagrangian mechanics and symmetry) · Peskin & Schroeder, An Introduction to QFT, §2.2 (Noether's theorem for fields)

Provenance. Derivations follow the standard treatments in Goldstein §13.7, Landau-Lifshitz Mechanics §6 and Classical Theory of Fields §32, and Peskin & Schroeder §2.2. A verifier should recompute: (1) the master Noether identity by integration by parts using Euler-Lagrange; (2) the canonical $T^{\mu\nu}$ for the free real scalar and confirm $T^{00}=\tfrac12\dot\phi^2+\tfrac12(\nabla\phi)^2+\tfrac12 m^2\phi^2$ ; (3) the complex-scalar $U(1)$ current $j^\mu=i(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*)$ and verify $\partial_\mu j^\mu=0$ via Klein-Gordon. All numbers and tensor expressions are standard textbook values. Additionally for this upgrade: worked-example numbers recomputed directly ( $\omega = 1.250$ from $k = 0.75$ , $m = 1$ ; $j^0 = 0.0250$ , $j^x = 0.0150$ , transport speed $0.600 = v_g$ ; standing-wave energy $E = LA^2\omega^2/4 = 0.0755$ with $\omega^2 = 0.7548$ , verified at two phases). Scope note: stress-energy anatomy owned by the stress-energy article; charge-current physics by the conserved-currents article; the $U(1)$ deep dive by the internal-symmetries article; improvement terms by the Belinfante-Rosenfeld article; the particle-mechanics version by the Noether article of the Lagrangian chapter.

Builds on: The Lagrangian Density and Field Action · Covariant Euler-Lagrange Equations for Fields · Functional Derivatives and Field Variations · From Discrete Systems to Continuous Fields Leads to: Conserved Currents and Charges · The Canonical Stress-Energy Tensor · Internal Symmetries and Conserved Charge · Angular Momentum and Spin Current in Fields · Symmetrization and the Belinfante-Rosenfeld Tensor

Does every conservation law come from a symmetry — is the converse of Noether's theorem true?

Within the Hamiltonian framework, essentially yes, and the Poisson-bracket article states the mechanism: any conserved $Q$ satisfies $\{Q, H\} = 0$ , which by antisymmetry means $Q$ generates a canonical transformation preserving $H$ — a symmetry. So on phase space, conservation and symmetry are strictly dual. The subtleties live in the Lagrangian translation. Some conserved quantities generate symmetries that are not point transformations of the fields (they mix fields and momenta — the Runge-Lenz vector is the classic case), so they fall outside the theorem's hypotheses while confirming its spirit. And some conservation-like statements are not Noetherian at all: topological charges (winding numbers, soliton counts) are conserved because configurations cannot be continuously deformed across sectors — no symmetry involved, no current from the theorem's formula, and their conservation is exact even off-shell. The clean summary: continuous symmetry ⇒ Noether current (this article); conserved phase-space function ⇔ canonical symmetry (bracket article); topological conservation is a third thing entirely.

What exactly is the difference between the first and second Noether theorems?

The 1918 paper proved two results, and the difference is whether the symmetry parameters are constants or functions. The first theorem — this article — handles global symmetries: finitely many constant parameters, each yielding a genuine conserved current. The second handles gauge symmetries: parameters that are arbitrary functions of spacetime, as in electromagnetism ( $\lambda(x)$ ) or general relativity (coordinate freedom). There the theorem's output changes character: instead of new conservation laws you get identities satisfied by the field equations themselves (for electromagnetism, the statement that $\partial_\mu\partial_\nu F^{\mu\nu} \equiv 0$ forces charge conservation of the source; for GR, the contracted Bianchi identities). This is why gauge symmetry gives constraints rather than conserved charges — the connection to the vanishing momentum $\pi^0$ and the constraint analysis flagged in the Hamiltonian article is exact. Noether proved the second theorem precisely to answer Hilbert's GR energy puzzle: in a generally covariant theory, "energy conservation" is an identity, not an independent law — which is why defining gravitational energy density is famously subtle to this day.

If the symmetry is only approximate, is everything lost?

No — the theorem degrades gracefully, and that grace is a working tool. If the action is invariant up to a small explicit breaking term, $\delta\mathcal{L} = \epsilon\,\partial_\mu K^\mu + \epsilon\,\mathcal{B}$ , then running the same derivation gives $\partial_\mu j^\mu = \mathcal{B}$ : the current's divergence equals the breaking term, so the charge changes at a computable rate rather than being constant. This "partially conserved current" logic is quantitatively predictive: isospin in nuclear physics is broken at the percent level by quark mass differences and electromagnetism, and Noether bookkeeping tells you which observables inherit that percent. The chiral symmetry of light quarks is broken by small quark masses, and the resulting divergence equation (PCAC) successfully predicted pion properties before QCD existed. The pattern to remember: symmetry ⇒ conservation; near-symmetry ⇒ near-conservation with the leak rate set by the breaking — often the most informative regime, because the leak is measurable.

What is the difference between a symmetry being broken explicitly and spontaneously, from the current's point of view?

Everything. Explicit breaking (a term in $\mathcal{L}$ that is not invariant) means the hypothesis fails: $\partial_\mu j^\mu \neq 0$ , the current leaks at a rate set by the breaking term, as in the previous answer. Spontaneous breaking is subtler and more interesting: the action is exactly symmetric, so Noether's theorem applies in full force and $\partial_\mu j^\mu = 0$ holds exactly — but the lowest-energy configuration (the vacuum) is not invariant under the symmetry. The current is conserved; the charge becomes problematic: acting on the asymmetric vacuum, it does not annihilate it but tries to rotate it into a different, degenerate vacuum, and in infinite volume the integral $Q = \int j^0$ fails to converge on such states. The physical residue is spectacular rather than null: Goldstone's theorem says the exactly conserved current must create gapless excitations — the Goldstone bosons (phonons of a crystal breaking translations; magnons of a ferromagnet breaking rotations; pions, approximately, for chiral symmetry). So: explicit breaking kills the current; spontaneous breaking keeps the current and converts the symmetry into massless particles. Distinguishing the two cases is a standing exam trap and a working particle-physics skill.

Why do translations give a rank-2 tensor while a phase rotation gives only a vector current?

Count parameters — the theorem is strictly one current per parameter. A $U(1)$ phase rotation has one parameter, hence one four-vector current $j^\mu$ . Spacetime translations have four independent parameters $a^\nu$ (one per direction), hence four currents — and four four-vector currents stack into a two-index object $T^{\mu\nu}$ : the $\nu$ index labels which current (which translation), the $\mu$ index is each current's ordinary flow index. Lorentz symmetry has six parameters (three rotations, three boosts), so the angular-momentum article's $M^{\mu\nu\rho}$ carries a flow index plus an antisymmetric pair labeling the six. An internal $SU(2)$ has three parameters and yields a triplet of currents $j^{\mu A}$ . Nothing mysterious: the tensor rank of the conserved object is bookkeeping for the dimension of the symmetry group, and reading indices this way ("which current" versus "flow direction") prevents a whole class of index errors in the articles ahead.

Noether's theorem needs a Lagrangian. What about systems that do not have one?

The theorem is genuinely a statement about variational systems — its proof lives inside the action principle, and for dynamics with no Lagrangian formulation (dissipative systems being the everyday case) the symmetry-conservation link weakens or fails. A damped oscillator's equations are time-translation invariant, yet its energy demonstrably decays: no action invariance, no theorem, no contradiction. This is not a defect so much as a diagnosis — the theorem quietly explains why fundamental physics is dissipation-free at bottom (dissipation is always emergent, energy leaking into untracked microscopic modes; include the modes and the Lagrangian, with its Noether energy, returns). For systems where a Lagrangian exists but is not obvious, the Hamiltonian route often rescues the connection: conservation ⇔ symmetry holds on phase space via the bracket whenever a Hamiltonian exists. And modern formulations (Schwinger-Keldysh actions for open systems) recover Noether-like statements even for dissipative field theories, at the cost of doubling the fields. For this chapter's closed, conservative field theories, the hypothesis is always satisfied and the theorem always fires.

Is the Noether current physically measurable, given the improvement ambiguity?

The honest answer is layered. The charge is unambiguous: improvement terms change $j^0$ by a total spatial divergence, which integrates away, so $Q$ is measurable full stop. The local density is convention-dependent in a pure field theory — two physicists using canonical and improved currents report different $j^0(\mathbf{x})$ maps, identical charges, and no experiment internal to the theory distinguishes them. The ambiguity is resolved the moment the current couples to something: whatever the coupling probes is then the physical current by definition. For stress-energy the arbiter is gravity — spacetime curvature responds to a specific tensor (the symmetric, metric-derived one), which is why the Belinfante-Rosenfeld article can speak of the "right" $T^{\mu\nu}$ ; for the electromagnetic current the arbiter is the photon field, which couples to the specific gauge-invariant $j^\mu$ . There is a satisfying experimental coda: differences between candidate stress tensors are total derivatives whose moments can differ, and precision tests of how light bends and how binary pulsars radiate confirm gravity sources on the Belinfante-improved tensor — the improvement ambiguity settled by observation, not by taste.