Conserved Currents and Charges

Noether's theorem hands you a four-vector $j^\mu$ with $\partial_\mu j^\mu = 0$ and moves on. This article stays behind and asks what that object actually is — because the distance between "a current is conserved" and "a charge is conserved" is where a surprising amount of physics lives. The continuity equation is a local statement: stuff is neither created nor destroyed at any point, it only flows. The constant charge $Q = \int j^0\,\mathrm{d}^3x$ is a global statement, and it follows from the local one only through a boundary condition — the spatial current must die off fast enough that nothing escapes through the sphere at infinity. Radiating systems violate exactly this fine print, which is why a "conservation law" can coexist with an antenna that visibly pumps charge-weighted field out of every finite region. The deeper question this article answers is why fundamental physics insists on the local form. Special relativity is the enforcer: a charge that vanished here and reappeared there — globally conserved, locally lawless — would appear in a boosted frame to exist twice at one moment and not at all a moment later, because moving observers slice spacetime differently. Local continuity is the only version of conservation that all frames can agree on, and as a bonus it makes the total charge of a localized system a genuine Lorentz scalar: every observer counts the same $Q$ , a fact we prove rather than assert. Along the way the current's built-in ambiguity — the freedom to add divergence-free circulation without touching the charge — gets its precise statement, ready for the Belinfante-Rosenfeld article to exploit.

Key takeaways

Noether currents obey the continuity equation $\partial_\mu j^\mu = 0$ , i.e. $\partial_t j^0 + \nabla\cdot\mathbf{j} = 0$ : local bookkeeping — density changes only by flux.
The global charge $Q = \int\mathrm{d}^3x\,j^0$ is constant only if $\mathbf{j}$ falls off faster than $1/r^2$ : the divergence theorem converts $\dot Q$ into a surface flux at infinity.
Relativity demands the local form: "action at a distance" conservation is inconsistent between frames, because simultaneity is frame-dependent.
For localized configurations, $Q$ is the same on every time slice — flat or tilted — making the charge of a system a Lorentz-invariant number.
The current is defined only up to an improvement term $j^\mu \to j^\mu + \partial_\nu\Sigma^{\mu\nu}$ ( $\Sigma$ antisymmetric): identically conserved, zero added charge, different local flow pattern.
Static configurations have divergence-free spatial currents; radiation is the standard way the fine print fails.
In gauge theories, $\partial_\mu \to D_\mu$ : covariant conservation, with charge bookkeeping subtleties that classical field theory already foreshadows.
Key idea: conservation in field theory is local first, global second — the charge is constant because, and only because, nothing crosses the boundary.

The core insight

A continuous symmetry yields an on-shell divergence-free current $\partial_\mu j^\mu = 0$ ; its time component integrated over space is a conserved charge $Q = \int\mathrm{d}^3x\,j^0$ , constant provided the spatial current carries no charge to infinity.

Where this lives today

The continuity equation is arguably the most-verified equation in physics, and its modern frontiers run in three directions. In precision fundamental physics, charge conservation is tested by hunting its failure: searches for electron decay ( $e^- \to \nu\gamma$ would violate charge conservation) bound the electron lifetime beyond $10^{28}$ years — twenty orders of magnitude past the age of the universe — and "milli-charged" dark-sector particles are constrained by exactly the frame-consistency logic of this article. In gravitational physics, the leak is the signal: the binary-pulsar and LIGO measurements of orbital decay are precision audits of Step 2's ledger, energy-momentum flux through a large sphere balancing the system's loss, with general relativity's prediction confirmed at the sub-percent level. And in condensed matter and hydrodynamics, conserved densities are the definition of the subject's slow variables: transport theory, from the Navier-Stokes equations of the continuum chapter to the exotic "hydrodynamics of electrons" observed in graphene, is the systematic expansion of currents in gradients of their conserved densities. The quantum frontier sharpened rather than blurred the classical story: anomalies make some classical currents unconservable (with physical consequences — the neutral pion's decay rate is an anomaly coefficient), while the generalized-symmetry program of the last decade reads every topological statement about currents as a symmetry, extending Noether bookkeeping to objects — strings, surfaces, defects — the classical theorem never dreamed of. Within the chapter: the internal-symmetry article builds the most important single instance of this machinery; the stress-energy and angular-momentum articles apply the same ledger to energy, momentum, and spin; and the Poisson-bracket article has already shown what the charges do — generate their own symmetries — once you have them.

The physical picture

Think of $j^0$ as a stuff-density and $\mathbf{j}$ as its flow — water in a network of pipes, heat in a metal, probability in quantum mechanics, energy and momentum in the stress-energy article. The continuity equation is the statement that the stuff is honest: the amount in any region changes only because stuff crosses the walls. Nothing pops into existence; nothing quietly evaporates. Draw any closed surface anywhere in the system, at any time, and the ledger balances: rate of change inside equals minus the flux through the boundary.

Figure

The worldtube of a conserved charge — every slice through it, flat or tilted, counts the same total

The distinction this article guards is between that local ledger and the single global number $Q$ . Field theory naturally produces the stronger, local statement — Noether's output is a current, not a number — and the global charge is recovered by integration, inheriting a boundary condition in the process. Picture the charge distribution's history as a worldtube in spacetime: the continuity equation says the tube's contents are conserved along its length, so any cross-section — any way of slicing "now" through the tube — counts the same total. Two consequences follow immediately, and both are theorems below. First, $Q$ is constant in time (slice horizontally at $t_1$ and $t_2$ : same count). Second, and less obviously, $Q$ is the same for all inertial observers (slice at an angle, the boosted observer's "now": still the same count) — the charge of an electron is not "the charge in some frame"; it is a scalar. The third piece of the picture is the ambiguity. The current pins down where the charge is and how it moves; but you can superpose a divergence-free circulation — closed loops of flow that carry no net stuff across any closed surface — and the ledger still balances with the same totals. That gauge-like freedom in the local description, made precise below, is not a defect: it is the working room that lets the Belinfante-Rosenfeld article reshape the stress-energy tensor into its symmetric form without touching a single conserved quantity.

From local law to global charge

The continuity equation, written covariantly and in 3+1 form:

\partial_\mu j^\mu = 0 \qquad\Longleftrightarrow\qquad \frac{\partial j^0}{\partial t} + \nabla\cdot\mathbf{j} = 0 \qquad (1)

where: $j^0$ is the charge density and $\mathbf{j}$ the charge flux (charge per unit area per unit time); the covariant form makes manifest that density and flux are one four-vector's components, mixing under boosts the way $(ct, \mathbf{x})$ do. This is an on-shell statement inherited from Noether's theorem — the current of a non-solution need not balance its books.

Derivation ACharge conservation from current conservation — and its fine print

We derive the global conservation law $\mathrm{d}Q/\mathrm{d}t = 0$ from the local law (1), tracking exactly where the boundary condition enters — because the resulting fine print (currents must decay faster than $1/r^2$ ) is physics, not pedantry, and radiating systems live on its wrong side. The strategy: differentiate the charge in a fixed volume, trade the time derivative for a divergence via (1), convert to a surface flux with the divergence theorem, then push the surface to infinity and see what is required of $\mathbf{j}$ .

Step 1. Charge in a fixed volume $V$ , differentiated under the integral (the volume does not move):

\frac{\mathrm{d}Q_V}{\mathrm{d}t} = \frac{\mathrm{d}}{\mathrm{d}t}\int_V \mathrm{d}^3x\; j^0 = \int_V \mathrm{d}^3x\;\partial_t j^0

Step 2. Use the local law: $\partial_t j^0 = -\nabla\cdot\mathbf{j}$ from (1). Then apply the divergence theorem:

\frac{\mathrm{d}Q_V}{\mathrm{d}t} = -\int_V \mathrm{d}^3x\;\nabla\cdot\mathbf{j} = -\oint_{\partial V} \mathbf{j}\cdot\mathrm{d}\mathbf{S}

Why this intermediate form is the physically strongest statement: it holds for every finite region — the charge in your laboratory changes exactly by the current through its walls. No limit has been taken and no condition assumed; this is the form experiments verify.

Step 3. Extend $V \to \mathbb{R}^3$ . The boundary becomes a sphere of radius $r \to \infty$ , whose area grows as $4\pi r^2$ . The flux vanishes if and only if

|\mathbf{j}| \to 0 \;\text{ faster than } \frac{1}{r^2} \quad\Longrightarrow\quad \frac{\mathrm{d}Q}{\mathrm{d}t} = -\oint_{S_\infty}\mathbf{j}\cdot\mathrm{d}\mathbf{S} = 0

Step 4. Read the failure mode. Localized lumps and any field with compact support qualify trivially. Radiation does not: an outgoing wave carries current falling as exactly $1/r^2$ (the amplitude falls as $1/r$ , currents are quadratic), so the flux through the sphere at infinity tends to a finite constant — the radiated power of the charge in question. For such systems $Q$ inside any finite region genuinely decreases, and the honest statement is Step 2's, not Step 3's. Verified: matches Landau-Lifshitz §29; the $1/r^2$ threshold recomputed from area $\times$ current scaling; the radiating counterexample is the standard antenna/wave-zone analysis.

The output is the conserved charge and its conditions of validity:

Q = \int_{\mathbb{R}^3}\mathrm{d}^3x\; j^0, \qquad \frac{\mathrm{d}Q}{\mathrm{d}t} = 0 \quad\text{(localized configurations)} \qquad (2)

where: "localized" is shorthand for the Step-3 decay condition. Note the asymmetry of roles: $j^0$ defines the charge, $\mathbf{j}$ polices it. Both are needed; neither alone is a conservation law.

Why relativity insists on the local form

Newtonian intuition tolerates a weaker notion: total charge constant, mechanism unspecified — charge could teleport from here to Andromeda as long as the books balance globally. Relativity does not tolerate it, and the reason is the relativity of simultaneity.

Go deeper · Derivation B · 2 minTeleportation is frame-inconsistent, and the charge on any slice is the same

Derivation BTeleportation is frame-inconsistent, and the charge on any slice is the same

We prove two statements with one construction. First, the negative: globally conserved but nonlocal "conservation" contradicts Lorentz invariance. Second, the positive: for a genuinely conserved current, the charge counted on any spacelike slice — any observer's "now" — is identical, making $Q$ a Lorentz scalar. The tool for both is the four-dimensional divergence theorem applied to the spacetime region between two slices, and the negative statement falls out of the positive one's failure conditions. Watch how completely the argument runs on geometry: nothing about the specific field theory enters.

Step 1. The negative, by simultaneity. Suppose charge $q$ vanishes at event $A$ and reappears at spacelike-separated event $B$ , "simultaneously" in frame $S$ , with no current in between. For spacelike separation there exist frames $S'$ in which $B$ occurs before $A$ (the temporal order of spacelike-separated events is frame-dependent — the SR-chapter result). In $S'$ , for a stretch of time, the charge exists at both locations: total charge $2q$ , then $q$ again. Another frame sees a period with the charge nowhere. "Globally conserved" in one frame is not even globally conserved in another. The only escape is a current connecting the disappearance to the reappearance — locality.

Step 2. The positive, by the 4D divergence theorem. Take a conserved current, $\partial_\mu j^\mu = 0$ , with localized support. Consider the spacetime region $\Omega$ bounded below by slice $\Sigma_1$ (constant $t_1$ in frame $S$ ), above by slice $\Sigma_2$ (constant $t'_2$ in a boosted frame $S'$ — a tilted plane in $S$ ), and at the sides by a large timelike cylinder. Integrate the identity over $\Omega$ :

0 = \int_\Omega \partial_\mu j^\mu\;\mathrm{d}^4x = \oint_{\partial\Omega} j^\mu\,\mathrm{d}\Sigma_\mu = \int_{\Sigma_2} j^\mu\,\mathrm{d}\Sigma_\mu - \int_{\Sigma_1} j^\mu\,\mathrm{d}\Sigma_\mu + (\text{side flux})

where $\mathrm{d}\Sigma_\mu$ is the outward-oriented surface element of each boundary piece (for a constant- $t$ slice, $\mathrm{d}\Sigma_\mu = (\mathrm{d}^3x, \mathbf{0})$ , so the slice integral is exactly $\int j^0\,\mathrm{d}^3x = Q$ ; for the tilted slice it is the boosted observer's $\int j'^0\,\mathrm{d}^3x' = Q'$ ).

Step 3. Kill the sides. The side cylinder sits at large spatial distance, where the localized current vanishes (the same fine print as before). Hence

Q' \;=\; \int_{\Sigma_2} j^\mu\,\mathrm{d}\Sigma_\mu \;=\; \int_{\Sigma_1} j^\mu\,\mathrm{d}\Sigma_\mu \;=\; Q

The charge is slice-independent: constant in time (take both slices flat in $S$ ) and frame-independent (take one flat, one tilted). One number, owned by the system.

Step 4. Close the loop on Step 1. The teleporting scenario has $\partial_\mu j^\mu \neq 0$ at $A$ and $B$ (a sink and a source), so Step 2's identity acquires bulk terms whose values depend on how much of $A$ and $B$ the slice has passed — which is exactly what frame-dependence of the count means. Local conservation is not an aesthetic preference; it is the unique way to make "conserved" a relativistically meaningful word. Verified: the 4D Gauss argument is standard (Weinberg QFT Vol. 1 §7.4 context; Landau-Lifshitz Classical Theory of Fields); the simultaneity reversal for spacelike pairs is the SR-chapter result on temporal order; slice integrals recomputed for flat and tilted normals.

Figure

Why globally balanced books are not enough — a teleporting charge counted by two observers

The improvement freedom, stated precisely

Noether's construction produces a current, not the current. The full freedom compatible with the same physics:

j^\mu \;\longrightarrow\; j'^\mu = j^\mu + \partial_\nu \Sigma^{\mu\nu}, \qquad \Sigma^{\mu\nu} = -\Sigma^{\nu\mu} \qquad (3)

where: $\Sigma$ is any antisymmetric tensor built from the fields, decaying at infinity. Two properties make this a genuine redundancy rather than a physical change, and both are one-line checks. Identical conservation: $\partial_\mu\partial_\nu\Sigma^{\mu\nu} = 0$ because a symmetric pair of derivatives contracts an antisymmetric index pair — no equations of motion needed, the added piece is conserved off-shell. Zero added charge: the added density is $j'^0 - j^0 = \partial_\nu\Sigma^{0\nu} = \partial_i\Sigma^{0i}$ (antisymmetry kills $\nu = 0$ ), a pure spatial divergence whose integral is a surface term at infinity — zero for decaying $\Sigma$ . The improvement is the mathematical version of superposing closed circulation loops on a flow: the local pattern changes, every honest total does not. What it is for — trading the canonical stress tensor's defects for the symmetric Belinfante form — is the business of two articles ahead; what matters here is that the freedom exists, is exactly this large, and costs nothing observable within the field theory itself.

Limiting cases

Case	Result	Meaning
static configuration	$\partial_t j^0 = 0 \Rightarrow \nabla\cdot\mathbf{j} = 0$	spatial current field-lines close or run to infinity
compact support / fast decay	$\oint_{S_\infty}\mathbf{j}\cdot\mathrm{d}\mathbf{S} = 0$	the standard case: $Q$ strictly constant
radiation ( $j \sim 1/r^2$ exactly)	finite flux at infinity	$Q$ in any finite region decays; conservation is the Step-2 ledger only
finite region $V$	$\dot Q_V = -\oint_{\partial V}\mathbf{j}\cdot\mathrm{d}\mathbf{S}$	the always-true, experimentally tested form
improvement term	$j^\mu + \partial_\nu\Sigma^{[\mu\nu]}$	conserved off-shell, adds zero charge — pure repackaging
tilted (boosted) slice	same $Q$	charge is a Lorentz scalar for localized systems
gauge theories	$\partial_\mu \to D_\mu$ : $D_\mu j^\mu = 0$	covariant conservation; naive $\int j^0$ no longer frame-portable (see FAQ)
fields on a lattice / bounded box	flux through walls explicit	boundary conditions decide conservation — periodic boxes conserve exactly

Common trapCurrent conserved is not charge conserved — the boundary term is physics

The continuity equation $\partial_\mu j^\mu = 0$ is a local identity on solutions; the constancy of $Q$ is a global theorem with a hypothesis. Conflating them produces real errors in both directions. Direction one: concluding that any Noether current implies a constant total — false for radiating or infinite non-decaying configurations, where charge-weighted flux genuinely escapes (compute $\dot Q_V = -\oint\mathbf{j}\cdot\mathrm{d}\mathbf{S}$ and it is honestly nonzero for every finite $V$ ; the energy of a radiating source really does leave). Direction two, subtler: treating $\mathrm{d}Q/\mathrm{d}t = 0$ as the full content of the symmetry and discarding the current — which throws away exactly the local structure that relativity requires and that the worldtube argument uses. A conserved number with no current behind it cannot be made frame-consistent (the teleportation derivation), cannot be localized in an experiment, and cannot couple to a gauge field. When you audit a conservation claim, always ask the two questions in order: does a local current exist ( $\partial_\mu j^\mu = 0$ )? and do the boundary conditions kill the flux (decay faster than $1/r^2$ )? "Yes, yes" is conservation; "yes, no" is a leak with a ledger; "no" is not conservation at all.

Worked examples

1 · Auditing a leaking sphere (confidence builder)

A charge distribution decays in place: inside a sphere of radius $R = 0.50$ the density is uniform, $j^0(t) = \rho_0\,e^{-t/\tau}$ with $\rho_0 = 2.00$ and $\tau = 4.00$ , and the loss is carried by a radial surface current $j_r(R, t) = \frac{\rho_0 R}{3\tau}e^{-t/\tau}$ (natural units). Verify the continuity ledger at $t = 0$ : does the flux through the sphere account exactly for the charge loss?

Step 1. Total charge at $t = 0$ : $Q_0 = \tfrac{4}{3}\pi R^3 \rho_0 = \tfrac{4}{3}\pi (0.125)(2.00) = \mathbf{1.047}$ .

Step 2. Loss rate from the density: $\dot Q = \tfrac{4}{3}\pi R^3\,\dot\rho = -\tfrac{4}{3}\pi R^3 \frac{\rho_0}{\tau} = -\frac{1.047}{4.00} = \mathbf{-0.2618}$ .

Step 3. Flux through the surface: $\oint\mathbf{j}\cdot\mathrm{d}\mathbf{S} = 4\pi R^2\, j_r(R,0) = 4\pi(0.25)\times\frac{2.00 \times 0.50}{3 \times 4.00} = 3.1416 \times 0.08333 = \mathbf{+0.2618}$ .

Step 4. The ledger balances to four digits: $\dot Q = -\oint\mathbf{j}\cdot\mathrm{d}\mathbf{S}$ ✓. Also check the local law at an interior point: inside a uniformly decaying ball the consistent interior current is $j_r(r) = \frac{\rho_0 r}{3\tau}e^{-t/\tau}$ (linear in $r$ ), whose divergence is $\nabla\cdot\mathbf{j} = \frac{1}{r^2}\partial_r(r^2 j_r) = \frac{\rho_0}{\tau}e^{-t/\tau}$ — exactly $-\partial_t j^0$ ✓ pointwise, not just in total. This is Step 2 of the first derivation, exercised with numbers.

2 · Charge transport and the two-observer count (exam level)

(i) The complex-scalar plane wave of the Noether article ( $A = 0.10$ , $k = 0.75$ , $m = 1.00$ , so $\omega = 1.25$ , $j^0 = 0.0250$ , $j^x = 0.0150$ ) flows through a fixed surface of area $\mathcal{A} = 1.00$ for a time $T = 10.0$ . How much charge crosses? Cross-check by a density sweep. (ii) A localized lump at rest carries $Q = 0.800$ . An observer boosts past at $v = 0.60$ ( $\gamma = 1.25$ ). Reconcile the transformed density and volume, and state what each observer measures.

Step 1. Direct flux count: $\Delta Q = j^x\,\mathcal{A}\,T = 0.0150 \times 1.00 \times 10.0 = \mathbf{0.150}$ .

Step 2. Density-sweep cross-check: the charge crossing the surface in time $T$ is the charge initially filling a slab of length $v_{\text{transport}}T$ upstream, where $v_{\text{transport}} = j^x/j^0 = 0.600$ . Slab charge: $j^0 \times (0.600 \times 10.0) \times 1.00 = 0.0250 \times 6.00 = \mathbf{0.150}$ ✓. Flux and density bookkeeping agree because — and only because — the current is conserved.

Step 3. The boosted lump. The density is the time component of a four-vector: for the observer moving at $v = 0.60$ , the lump's charge density rises to $j'^0 = \gamma\,j^0_{\text{rest}}$ (for a lump with no rest-frame current, the transformation is the clean $\gamma$ factor). Simultaneously the lump's volume Lorentz-contracts: $V' = V/\gamma$ .

Step 4. Multiply: $Q' = j'^0 V' = (\gamma j^0)(V/\gamma) = j^0 V = \mathbf{0.800}$ — the $\gamma$ 's cancel identically. Both observers measure charge $0.800$ ; they disagree about density ( $\times 1.25$ ) and volume ( $\div 1.25$ ), never about the total. This is the worldtube theorem in its most economical costume: $j^0$ and $\mathrm{d}^3x$ are each frame-dependent, but $j^\mu\mathrm{d}\Sigma_\mu$ is a scalar — the count belongs to the tube, not the slice.

Origins · 1 minJulian Schwinger and how this was discovered

Julian Schwinger

1918–1994 · Harvard and UCLA

No physicist wrung more physics out of conserved currents than Schwinger. A prodigy who published at sixteen and earned his Columbia PhD at twenty-one, he shared the 1965 Nobel Prize with Feynman and Tomonaga for renormalized quantum electrodynamics — his formulation built precisely on the field-theoretic action principle and the local currents of this chapter, developed with a Greek-mathematics insistence on structure over pictures (he famously let students use Feynman diagrams only after they could do without them). His quantum action principle made the Noether machinery an operator statement, and the Ward-Takahashi-Schwinger identities — quantum continuity equations — became the load-bearing constraints of gauge theory. Two of his discoveries live exactly at this article's fault lines: Schwinger terms (1959) showed that quantum current-current commutators acquire extra pieces the classical bracket algebra lacks, and his 2D model of electrodynamics (the Schwinger model, 1962) remains the cleanest laboratory for anomalies — currents conserved classically whose quantum conservation fails, computably. He supervised over seventy PhD students, four of them future Nobel laureates, lecturing in legendarily polished improvisations; his summary of the field-theory creed could headline this chapter: the local field, with its local currents, is how nature reconciles quantum mechanics with relativity.

History

1747

— Continuity equations enter physics with Euler's fluid dynamics: mass conservation as a local flow law.

1864

— Maxwell builds displacement current into his equations precisely so that charge obeys

\partial_t\rho + \nabla\cdot\mathbf{j} = 0

exactly.

1918

— Noether derives currents for every continuous symmetry: local conservation becomes a theorem, not an observation.

1950

— Ward's identity makes current conservation a quantum operator statement, taming QED's divergences.

1969

— Adler, Bell and Jackiw compute the anomaly: a classically conserved current whose quantum divergence is nonzero — and measurable in

\pi^0 \to \gamma\gamma

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §13.7 · Landau & Lifshitz, The Classical Theory of Fields, §29 and §32–33 · Peskin & Schroeder, An Introduction to QFT, §2.2 · Weinberg, The Quantum Theory of Fields, Vol. 1, §7.4 (currents, improvements, and charge algebra context)

Provenance. All equations are standard Noether/field-theory results. A verifier should recompute: (1) the Noether current derivation via Euler-Lagrange (Goldstein §13.7, Peskin-Schroeder §2.2); (2) charge conservation by divergence theorem (Landau-Lifshitz §29); (3) the U(1) scalar current sign/normalization; (4) the symmetry condition $T^{\nu\rho}-T^{\rho\nu}=-\partial_\mu S^{\mu\nu\rho}$ (Belinfante-Rosenfeld; Weinberg QFT Vol 1 §7.4). No numerical constants are claimed. Additionally for this upgrade: the 4D-divergence slice-independence argument follows the standard treatment (Weinberg §7.4 context); worked-example numbers recomputed directly ( $Q_0 = 1.047$ , $\dot Q = -0.2618$ balancing $4\pi R^2 j_r = 0.2618$ ; interior divergence check $\nabla\cdot\mathbf{j} = \rho_0 e^{-t/\tau}/\tau$ ; transport $\Delta Q = 0.150$ both ways; boosted-lump cancellation $\gamma/\gamma = 1$ ). Scope note: the master Noether identity is owned by the Noether article; the $T^{\mu\nu}$ -symmetry condition and the use of improvements are owned by the angular-momentum and Belinfante-Rosenfeld articles; the $U(1)$ current's deep dive by the internal-symmetries article.

Builds on: Noether's Theorem for Continuous Symmetries · Covariant Euler-Lagrange Equations for Fields · The Lagrangian Density and Field Action · From Discrete Systems to Continuous Fields Leads to: The Canonical Stress-Energy Tensor · Internal Symmetries and Conserved Charge · Symmetrization and the Belinfante-Rosenfeld Tensor · Poisson Brackets for Fields · Angular Momentum and Spin Current in Fields

Is the charge density j⁰ measurable, given that improvements can change it?

Within a pure, uncoupled field theory the local density is convention-laden — improvements shuffle it — but the moment the current couples to anything, the ambiguity collapses. Electromagnetic charge density is measured by the fields it sources; the photon couples to one specific $j^\mu$ (the gauge-invariant one), so probing the field around a system maps the physical density with no residual freedom. The same resolution applies to energy density via gravity, as the Belinfante-Rosenfeld article develops. There is also an internal consistency point worth keeping: improvements change $j^0(\mathbf{x})$ but not the flux through the boundary of any region where $\Sigma$ vanishes — so measurements confined to charge-free surroundings of a lump agree between conventions even before couplings are invoked. In practice: totals and externally probed densities are physical; unprobed local repackagings are the theorist's gauge.

Radiation makes the charge in every finite region decay. Is anything still conserved then?

Yes — the accounting simply needs a bigger book. For a radiating system, the correct conserved object is charge-in-the-region plus charge-in-flight: define $Q_V(t) + \int_{t_0}^{t}\oint_{\partial V}\mathbf{j}\cdot\mathrm{d}\mathbf{S}\,\mathrm{d}t'$ and it is exactly constant, by the finite-region ledger. Physically, the escaping flux is not lost; it is in transit, and an observer with a larger sphere catches it later (the retardation is visible in the flux's arrival time). This is precisely how energy bookkeeping works for the binary pulsar: the orbit's mechanical energy decreases, the wave zone's energy content increases, and the sum tracks to the precision of the ephemeris. The formal statement is that $\mathrm{d}Q/\mathrm{d}t = 0$ holds for the charge on a complete spacelike slice, radiation included; it fails only for the charge assigned to a spatially bounded subsystem — and the failure rate, the flux, is itself the measurable of interest (luminosity, radiated power, emission rate). Conservation is never violated; it is delocalized.

In gauge theories the conservation law reads D-mu j-mu = 0. Why is covariant conservation weaker, and what happens to the charge?

Because the covariant divergence contains the gauge field itself: $D_\mu j^\mu = \partial_\mu j^\mu + [A_\mu, j^\mu] = 0$ means the matter current alone satisfies an ordinary continuity equation only where the commutator vanishes. Physically, a non-abelian gauge field carries the charge it couples to (gluons are colored, unlike the photon which is neutral), so charge flows between matter and field, and only the total — matter plus field contribution — obeys $\partial_\mu J^\mu = 0$ with an ordinary divergence. The price of that repair is that the total current is not gauge-covariant, and the associated total charge is gauge-dependent except when evaluated with boundary conditions that freeze the gauge at infinity. This tangle, already visible classically, is one reason color charge is not measured the way electric charge is (no Gauss-law readout of "total color in a region") and connects to confinement lore in the quantum theory. The abelian case dodges every subtlety: the photon is chargeless, $D_\mu = \partial_\mu$ on the current, and electric charge enjoys the full classical story of this article — which is exactly why electromagnetism is the safe first example and Yang-Mills is a later course.

Can a current be conserved without any symmetry behind it?

Yes — and the exceptions map the boundary of Noether's theorem. The cleanest classical family is topological currents: in one spatial dimension, $j^\mu = \epsilon^{\mu\nu}\partial_\nu\phi$ satisfies $\partial_\mu j^\mu = \epsilon^{\mu\nu}\partial_\mu\partial_\nu\phi = 0$ identically — off-shell, for any field configuration, symmetric or not — because the antisymmetric epsilon annihilates the symmetric pair of derivatives (the same mechanism as the improvement term's conservation). Its charge, $Q = \int\partial_x\phi\,\mathrm{d}x = \phi(+\infty) - \phi(-\infty)$ , counts the field's boundary winding: a soliton number, conserved because configurations in different sectors cannot be continuously deformed into one another, not because anything is invariant. Such conservation laws need no equations of motion, survive arbitrary symmetry breaking, and protect objects (kinks, vortices, skyrmions, magnetic monopoles in suitable theories) whose stability is geometric. The taxonomy worth memorizing: Noether currents are conserved on-shell because of symmetry; topological currents are conserved identically because of index antisymmetry; and the modern "generalized symmetries" program deliberately reunifies the two by treating every identically conserved current as the signature of a (possibly higher-form) symmetry.

Why is the charge the integral of j⁰ specifically, and not some other component?

Because a charge is a count on a slice of constant time, and $j^0$ is the component of the current normal to such a slice. The covariant statement makes this transparent: for any spacelike hypersurface $\Sigma$ with oriented element $\mathrm{d}\Sigma_\mu$ , the charge is $Q = \int_\Sigma j^\mu\,\mathrm{d}\Sigma_\mu$ — a scalar built from a four-vector and a surface element. For a constant- $t$ slice in a given frame, $\mathrm{d}\Sigma_\mu = (\mathrm{d}^3x, \mathbf{0})$ and the general formula collapses to the familiar $\int j^0\,\mathrm{d}^3x$ ; for a boosted observer's tilted slice, their $\mathrm{d}\Sigma'_\mu$ picks out their $j'^0$ — and the worldtube derivation shows all such counts agree. So "integrate the time component" is not a rule about the number zero; it is the coordinate expression of "count what crosses your now." The same logic explains the flux: $\mathbf{j}$ is what crosses a timelike surface (a detector sitting at fixed position integrates $j^x$ over its area and its operating time), which is exactly how the worked example's two counts — slab of density versus flux through a plane — came out equal.

Does the improvement freedom mean the current of electromagnetism is ambiguous too?

The electromagnetic source current is the instructive case where every abstract freedom gets fixed by a physical handle. Formally, yes: given any conserved $j^\mu$ you may add $\partial_\nu\Sigma^{\mu\nu}$ . But electromagnetism couples through the interaction $A_\mu j^\mu$ , and demanding the coupling be gauge-invariant and reproduce the observed Lorentz force selects the specific matter current — for the complex scalar, the (gauge-covariantized) $U(1)$ current of the Noether article. Meanwhile, Maxwell's equations themselves enforce the conservation: $\partial_\nu F^{\mu\nu} = j^\mu$ implies $\partial_\mu j^\mu = \partial_\mu\partial_\nu F^{\mu\nu} \equiv 0$ by the antisymmetry of $F$ — the field equations are inconsistent with a non-conserved source, which is why Maxwell needed the displacement current in 1864 before anyone knew why. Notice the pleasing inversion: in this article, symmetry produced conservation; in electromagnetism, the consistency of the force law demands it. The two arguments meet in Noether's second theorem — gauge symmetry is exactly the structure that makes the source's conservation an identity — closing a loop the FAQ on gauge theories opened.