Conservation Laws and Energy-Momentum Flow

The last article built the ledger; this one makes it flow. The stress-energy tensor $\theta^{\mu\nu}$ holds the field's energy, momentum, and stress in sixteen slots — but a ledger becomes a conservation law only when you can prove what leaves one account enters another. The proof is a single covariant line,

\partial_\mu\theta^{\mu\nu}=-F^{\nu\lambda}J_\lambda,

and its derivation is the chapter's most satisfying piece of teamwork: the two lines of the covariant Maxwell system work in concert — the inhomogeneous equation inserts the source, the Bianchi identity annihilates everything else — and the right-hand side lands, with no room for negotiation, on the negative of the Lorentz-force density. The field loses four-momentum at exactly the rate matter gains it. E7 discovered the pieces of this statement one at a time: Poynting's theorem is its time slot, the momentum balance law its space slots. Here the four theorems become one equation with one free index, and E7's closing article — which stated this equation as a filing system and deferred the covariant machinery to E10 — gets its receipt. Two payoffs follow. Globally: for a closed system of fields plus charges, the total four-momentum $P^\nu_{\text{field}}+P^\nu_{\text{matter}}$ is exactly conserved — proven by combining this article's balance law with the particle equation of motion, and sharpened by Laue's warning that only closed totals transform as four-vectors. And rotationally: because $\theta^{\mu\nu}$ is symmetric, the angular-momentum current $x^\mu\theta^{\nu\lambda}-x^\nu\theta^{\mu\lambda}$ is conserved for free — the payoff the symmetry was purchased for. A copper wire's Joule heat and self-pinch, and a photon rocket's mass budget, put numbers on the flow.

Key takeaways

The covariant balance law is $\partial_\mu\theta^{\mu\nu}=-F^{\nu\lambda}J_\lambda$ : derived by inserting Maxwell's inhomogeneous equation and killing the remainder with the Bianchi identity — both lines of the covariant system working together.
The right-hand side is forced: its time slot is $-\mathbf{E}\cdot\mathbf{J}/c$ (Joule transfer) and its space slots are $-(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})$ (Lorentz-force density) — the field's loss is matter's gain, component by component.
$\nu=0$ is Poynting's theorem, $\nu=i$ is E7's momentum balance: four 3+1 theorems, one tensor equation, one free index.
In vacuum ( $J^\mu=0$ ) the field's four-momentum is separately conserved; with sources, only the total — field plus particles — is, proven via the covariant Lorentz force closing the loop.
Laue's closure doctrine: the integrated $P^\nu$ of an open subsystem (a charge's own field) is not a four-vector; totals of closed systems are — the 4/3 problem's diagnosis, stated once and deferred to E11 for surgery.
Symmetry of $\theta$ cashes out here: $M^{\mu\nu\lambda}=x^\mu\theta^{\nu\lambda}-x^\nu\theta^{\mu\lambda}$ is conserved because $\theta^{\nu\mu}=\theta^{\mu\nu}$ — field angular momentum (E7's) joins the covariant family.
Key idea: the two-line Maxwell system plus the symmetric stress-energy tensor yield one balance law, $\partial_\mu\theta^{\mu\nu}=-F^{\nu\lambda}J_\lambda$ — all of electromagnetic energy-momentum accounting in a single covariant sentence.

The core insight

The divergence of the electromagnetic stress-energy tensor equals minus the Lorentz-force density on the sources, $\partial_\mu\theta^{\mu\nu}=-F^{\nu\lambda}J_\lambda$ ; in vacuum the field's four-momentum is conserved outright, and with matter included the total four-momentum and angular momentum of field plus particles are exact constants.

Where this lives today

This balance law is the daily bread of every field that moves energy with light or plasma. Inertial-confinement fusion targets are imploded by the space slots (radiation pressure and ablation recoil) while the time slot prices the drive energy; the National Ignition Facility's gain milestones are audits of equation (1) at kilojoule scale. Plasma astrophysics runs on the same accounting at its extremes: pulsar wind nebulae convert spin-down four-momentum into field and particle ledgers whose exchange the Crab Nebula displays in X-ray movies; black-hole jets are giant Poynting-flux conduits, their power estimated precisely as $\int\theta^{0i}dA$ . Gravitational-wave astronomy borrowed the whole framework — the energy a binary radiates is computed from the gravitational analogue of this article's flux slots, and LIGO's "3 solar masses radiated" headline for GW150914 is a Derivation-B statement about a closed total. Even the photon rocket is no longer pure fiction at small scale: laser-sail proposals (Starshot) are engineering studies of Example 2's slots, with the sail's momentum credit purchased at the beam's $1/c$ exchange rate. Wherever you meet "energy flux," "radiation pressure," or "field momentum" in the literature, you are reading one slot of the sixteen, and equation (1) is the guarantee they never contradict each other.

Origins · 1 minHenri Poincaré and how this was discovered

Henri Poincaré

1854–1912 · Paris

Poincaré appears twice in this chapter's genealogy, wearing different hats. In 1900 — five years before special relativity — he confronted a scandal in Lorentz's electron theory: radiation from an emitter seemed to violate Newton's action-reaction, momentum appearing from nowhere. His repair was radical: treat the electromagnetic field as a momentum-bearing fluid with density $\mathbf{S}/c^2$ , so that the recoil of an emitter is balanced by momentum genuinely stored in flight. In the same analysis he noted that a pulse of energy $E$ behaves as if it had mass $E/c^2$ — the inertia-of-energy relation, in print before Einstein, though Poincaré read it as a property of the electromagnetic fluid rather than of energy as such. His second appearance is structural: the "Poincaré stresses" — the non-electromagnetic glue his 1905–06 electron model needed to hold itself against its own field pressure — are exactly the closure terms Laue's theorem demands before the self-field's four-momentum makes sense, the resolution this article's warning band points toward. The mathematician who named the Lorentz group thus also drafted both the field-momentum ledger and the fine print about when its totals may be trusted.

History

1884

— Poynting's theorem: the energy row of the balance law, in 3+1 dress

1900

— Poincaré endows the field with momentum density

\mathbf{S}/c^2

to save action-reaction, and prices energy's inertia at

E/c^2

1908

— Minkowski packages field mechanics covariantly; the four-divergence form appears

1911

— Laue's closure theorem: only closed systems' integrated momenta are four-vectors

1918

— Noether grounds the whole ledger in spacetime symmetry

1972

— Compact-object astrophysics begins auditing Poynting-dominated flows at stellar scale — the balance law becomes an observational tool

The physical picture

Think of spacetime as threaded by two circulatory systems. One carries the mechanical four-momentum of matter — every charge's $p^\mu$ , evolving by the covariant Lorentz force of Relativistic Dynamics of a Charged Particle. The other carries the field's four-momentum — the $\theta^{0\nu}$ densities of the last article, transported by its flux slots. The two systems meet wherever charge meets field, and there is exactly one teller window between them: the contraction $F^{\nu\lambda}J_\lambda$ . Every joule a resistor dissipates, every newton a motor delivers, every recoil a laser imparts passes through that window, and this article's theorem says the window is honest — debits equal credits, at every event, in every frame. The 3+1 shadow of this picture is E7 territory, worth recalling by name because every slot below lands on it. Poynting's Theorem was the energy row: $\partial_t u+\nabla\cdot\mathbf{S}=-\mathbf{E}\cdot\mathbf{J}$ . Conservation of Linear Momentum in Electrodynamics was the momentum rows: $\partial_t\mathbf{g}+$ (stress divergence) $=-(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})$ , every coefficient forced. What E7 could not see is that its two theorems are one object's time and space faces — and that the proof, done covariantly, is four lines rather than two pages of vector identities, because the Bianchi identity silently performs every curl-and-cancel that made the 3+1 derivations sweat.

Figure

Two circulatory systems, one teller window — the exchange term that keeps both ledgers honest

One conceptual upgrade over E7 deserves emphasis before the derivation. In 3+1 language, "energy conservation" and "momentum conservation" look like four separate facts. Covariantly they cannot be separated: a boost mixes the time slot into the space slots — the Covariant Form of Maxwell's Equations article showed the laws themselves blending under boosts, and the same happens here. A moving observer's "Poynting theorem" is a $\gamma$ -weighted blend of the lab's energy and momentum balances. Conservation of energy without conservation of momentum is not a relativistically consistent option; the four-index package is the minimum sellable unit.

The balance law

\partial_\mu\theta^{\mu\nu}=-F^{\nu\lambda}J_\lambda, \qquad (1)

where: $\theta^{\mu\nu}$ is the last article's tensor, $F^{\nu\lambda}$ the field tensor, $J_\lambda$ the (lowered) four-current, and the free index $\nu$ makes this four scalar balance laws in one envelope — the same packaging discipline as the two-line system itself. The left side is a four-divergence: "rate of change of density plus outflow of flux," slot by slot. The right side, unpacked with the track dictionary (verified symbolically for this edition):

-F^{\nu\lambda}J_\lambda=\left(-\frac{\mathbf{E}\cdot\mathbf{J}}{c},\;-(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})\right),

where: the time slot is (minus) the rate at which the field does work on charges per unit volume — Joule transfer — and the space slots are (minus) the Lorentz-force density. Nothing about this pairing was chosen: Derivation A shows both slots are outputs of the Maxwell system. Equation (1) with $J=0$ says the free field conserves its own four-momentum; with sources it says every deficit is a transfer through the teller window, never a leak.

Derivation AThe divergence — Maxwell's two lines working in concert

We compute $\partial_\mu\theta^{\mu\nu}$ from the definition of $\theta$ and show it collapses to $-F^{\nu\lambda}J_\lambda$ . The inputs are exactly the two lines of Covariant Form of Maxwell's Equations: the inhomogeneous equation $\partial_\mu F^{\mu\alpha}=\mu_0J^\alpha$ supplies the source term, and the Bianchi identity $\partial_\lambda F_{\mu\nu}+\partial_\mu F_{\nu\lambda}+\partial_\nu F_{\lambda\mu}=0$ (the homogeneous line, in The Dual Field Tensor's equivalent form) annihilates the rest. The strategy is product-rule bookkeeping followed by a symmetrization trick; the one danger is index hygiene on the cross term, so every renaming is displayed. Watch the division of labor: dynamics inserts $J$ , geometry cancels the remainder — neither line alone could produce the balance law.

Step 1. Differentiate the definition $\theta^{\mu\nu}=\tfrac{1}{\mu_0}[F^{\mu\alpha}F_\alpha{}^\nu+\tfrac14\eta^{\mu\nu}F_{\alpha\beta}F^{\alpha\beta}]$ by the product rule:

\mu_0\,\partial_\mu\theta^{\mu\nu} =(\partial_\mu F^{\mu\alpha})F_\alpha{}^{\nu} +F^{\mu\alpha}\,\partial_\mu F_\alpha{}^{\nu} +\tfrac12 F_{\alpha\beta}\,\partial^{\nu}F^{\alpha\beta}.

Step 2. The first term is the dynamics. Insert the inhomogeneous line, $\partial_\mu F^{\mu\alpha}=\mu_0 J^\alpha$ :

(\partial_\mu F^{\mu\alpha})F_\alpha{}^{\nu}=\mu_0 J^\alpha F_\alpha{}^{\nu}=-\mu_0F^{\nu\lambda}J_\lambda,

using antisymmetry to swap the index order at the cost of one sign. The right-hand side of (1) has already arrived; everything else must vanish.

Step 3. Organize the remainder. Raise/lower to write both leftover terms with all-up field tensors: $F^{\mu\alpha}\partial_\mu F_\alpha{}^\nu=F_{\mu\alpha}\partial^\mu F^{\alpha\nu}$ , and relabel dummies so the common factor is $F_{\alpha\beta}$ :

\text{remainder}=F_{\alpha\beta}\left(\partial^{\alpha}F^{\beta\nu}+\tfrac12\,\partial^{\nu}F^{\alpha\beta}\right).

Why the first piece could be symmetrized: $F_{\alpha\beta}$ is antisymmetric under $\alpha\leftrightarrow\beta$ , so only the antisymmetric part of its cofactor survives the contraction — we may freely replace $\partial^\alpha F^{\beta\nu}$ by $\tfrac12(\partial^\alpha F^{\beta\nu}-\partial^\beta F^{\alpha\nu})$ .

Step 4. Kill it with geometry. The Bianchi identity, written with the indices at hand, reads $\partial^\alpha F^{\beta\nu}-\partial^\beta F^{\alpha\nu}=-\,\partial^\nu F^{\alpha\beta}$ (cyclic identity rearranged; this is where the homogeneous line enters). Substituting into the symmetrized remainder:

F_{\alpha\beta}\left(-\tfrac12\partial^{\nu}F^{\alpha\beta}+\tfrac12\partial^{\nu}F^{\alpha\beta}\right)=0.

The remainder dies identically — not for special fields, but as an identity of the potential formulation. Assembling Steps 2–4:

\partial_\mu\theta^{\mu\nu}=-F^{\nu\lambda}J_\lambda.

Step 5. Audit the slots against E7. $\nu=0$ with the dictionary: left side $=\tfrac1c(\partial_tu+\nabla\cdot\mathbf{S})$ , right side $=-\mathbf{E}\cdot\mathbf{J}/c$ — Poynting's Theorem, exactly, coefficient for coefficient. $\nu=i$ : left side $=\partial_t g_i+\partial_j(-\sigma_{ji})$ (momentum density rate plus stress outflow), right side $=-(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})_i$ — E7's Conservation of Linear Momentum in Electrodynamics, whose every coefficient that article derived "forced" from the 3+1 side. Two chapters of vector calculus have become one contraction and one identity. Verified: the RHS unpacking ( $-\mathbf{E}\cdot\mathbf{J}/c$ ; $-(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})$ ) recomputed symbolically with the track dictionary (sympy, all four slots); the Bianchi rearrangement checked component-wise; slot audit matched by name to the E7 statements. Matches Jackson §12.10 and Landau–Lifshitz §33.

Go deeper · Derivation B · 2 minGlobal conservation — the closed total, and Laue's warning about open pieces

Derivation BGlobal conservation — the closed total, and Laue's warning about open pieces

We integrate the balance law over space and combine it with the particle equation of motion to prove that the total four-momentum of fields plus charges is constant, then state precisely when the integrated field momentum is a legitimate four-vector. The strategy is the covariant version of every conservation argument in the track: convert a divergence to a boundary flux (which dies for localized systems), and let the source term hand its content to the matter ledger. Watch the matter side arrive from outside electromagnetism: the covariant Lorentz force is what turns the field's deficit into the particles' credit — the two articles' results interlock like gears.

Step 1. Define the field's candidate four-momentum by integrating the density column over a constant-time slice:

P^\nu_{\text{field}}(t)=\frac{1}{c}\int\theta^{0\nu}(t,\mathbf{x})\,d^3x .

Integrate the $\mu$ -sum of (1) over all space: the spatial-divergence part $\partial_i\theta^{i\nu}$ becomes a surface flux at infinity, which vanishes for localized field configurations. What survives:

\frac{dP^\nu_{\text{field}}}{dt}=-\int F^{\nu\lambda}J_\lambda\,d^3x .

Step 2. Evaluate the right side for point charges: $J^\mu=\sum_a q_a\mathbf{v}_a$ -supported delta functions, so the integral collapses onto the worldlines, giving $-\sum_a q_aF^{\nu\lambda}(x_a)u_{a\lambda}/\gamma_a$ . By the covariant Lorentz force (Relativistic Dynamics of a Charged Particle), each term is $-dp_a^\nu/dt$ (the $1/\gamma$ converting the proper-time rate to lab time). Hence

\frac{d}{dt}\left[P^\nu_{\text{field}}+\sum_a p_a^\nu\right]=0 .

All four components, every frame: the total is exact. Energy conservation and momentum conservation are the $\nu$ -slots of one statement, as promised.

Step 3. Laue's fine print, stated as a theorem because it prevents famous errors. The integral defining $P^\nu_{\text{field}}$ is taken over a simultaneity slice, and different observers slice differently; the integrated object transforms as a four-vector iff the integrand is divergence-free everywhere ( $\partial_\mu\theta^{\mu\nu}=0$ ) — i.e. iff the subsystem is closed. Free radiation pulses qualify: their $P^\nu$ is null-directed and boosts like a photon's. A charge's own Coulomb field does not: it exchanges momentum with the charge (and with whatever binds the charge), the slice integrals of the field alone disagree between frames, and the notorious "4/3" mismatch between the field's energy-mass and momentum-mass is precisely this open-system artifact. The diagnosis lives here; the surgery — Poincaré stresses, self-energy renormalization — is E11's Classical Electron Theory and Self-Energy, by name.

Step 4. The vacuum corollary, for the record: with $J=0$ , equation (1) is $\partial_\mu\theta^{\mu\nu}=0$ and the field's four-momentum is separately conserved — light in flight keeps its own books. This is the covariant license behind every "photon budget" argument in the worked examples and beyond. Verified: the delta-function collapse and $\gamma$ bookkeeping in Step 2 rechecked against the RPART reduction; the slice-dependence statement checked against the plane-wave pulse (closed, null $P^\nu$ ) and the static Coulomb field (open; momentum integral frame-mismatched by the classic 4/3) — Laue's theorem per Jackson §16.4–16.5 and Zangwill §22.7.

Angular momentum — the symmetry's payoff

The last article bought symmetry ( $\theta^{\mu\nu}=\theta^{\nu\mu}$ ) and promised it would fund rotation. Payment is three lines. Define the angular-momentum current

M^{\mu\nu\lambda}=x^{\mu}\theta^{\nu\lambda}-x^{\nu}\theta^{\mu\lambda}, \qquad (2)

where: the $(\mu\nu)$ pair labels which rotation (or boost) plane, and $\lambda$ is the flow index. Its divergence, using the product rule and the free-field balance law $\partial_\lambda\theta^{\nu\lambda}=0$ :

\partial_\lambda M^{\mu\nu\lambda}=\theta^{\nu\mu}-\theta^{\mu\nu}=0,

because and only because $\theta$ is symmetric — an asymmetric tensor would leave a torque-out-of-nowhere residue, which is the physical content of the symmetry requirement. The space-space components integrate to E7's field angular momentum $\int\mathbf{x}\times\mathbf{g}\,d^3x$ (Angular Momentum of the Electromagnetic Field, including its paradox-dissolving duties in the Feynman-disk affair — recalled, not reopened). The time-space components are the boost charges, whose conservation is the center-of-energy theorem — the ledger Conservation Laws and Spacetime Symmetries derived and E7's hidden-momentum article spent; with sources, the same computation routes torque through the teller window, and the total angular momentum of field plus matter is conserved. Noether's accounting closes: translations fund $P^\nu$ , rotations and boosts fund $M^{\mu\nu}$ , and the field pays its share of both through the one tensor.

question	reach for	why
local energy audit	$\nu=0$ slot: $\partial_tu+\nabla\cdot\mathbf{S}=-\mathbf{E}\cdot\mathbf{J}$	Poynting's theorem, covariant edition
local force on matter	$\nu=i$ slots: RHS $=+(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})_i$ into matter	momentum balance
total budget, closed system	$P^\nu_{\text{field}}+\sum p^\nu=$ const	Derivation B
is my integrated field momentum a four-vector	closure test $\partial_\mu\theta^{\mu\nu}=0$	Laue; else expect 4/3-type artifacts
torque bookkeeping	$M^{\mu\nu\lambda}$ , eq. (2)	symmetry of $\theta$ does the work
radiation escaping to infinity	flux slots at large $r$	the E9 power integrals, covariantly licensed

Limiting cases

case	result	meaning
vacuum	$\partial_\mu\theta^{\mu\nu}=0$	free fields keep their own four-momentum
statics	$\nu=0$ : $\nabla\cdot\mathbf{S}=-\mathbf{E}\cdot\mathbf{J}$	steady Joule flow — E7's battery-to-resistor Poynting picture
single charge in external field	Derivation B collapses to RPART's law	the two articles are consistent by construction
no charges, one wave pulse	$P^\nu$ null four-vector	light's budget boosts like a photon
self-field of a charge	open system — no four-vector $P^\nu$	4/3 diagnosis; E11 surgery
media instead of vacuum	$\theta$ must be renegotiated	Abraham–Minkowski; E7's verdict + next article's tensors

Common trapThe right-hand side is a transfer, not a violation — and never write the vacuum law when sources are present

Two mirror-image misreadings of equation (1) cause most downstream errors. First: seeing $\partial_\mu\theta^{\mu\nu}\neq0$ and concluding field energy-momentum "is not conserved." It is — jointly. The nonzero divergence is the signature of an open subsystem doing its job: transferring through the teller window at exactly the Lorentz-force rate. Conservation statements attach to the closed total (Derivation B), never to one ledger. Second, the converse: writing $\partial_\mu\theta^{\mu\nu}=0$ inside matter "because energy is conserved." Inside a resistor, an antenna, a plasma — anywhere $J\neq0$ — the field ledger alone is not conserved, and pretending otherwise double-counts the energy that Joule heating already moved to matter. The practical protocol: draw the system boundary first; if charges are inside, use (1) with its right side; if you want a conserved quantity, add the matter ledger or shrink the boundary until nothing charged is inside. And before quoting any integrated $P^\nu$ across frames, run Laue's closure test — open pieces (self-fields, "the field in this half of the box") do not boost like four-vectors, no matter how carefully you computed them in one frame.

Figure

The photon rocket's closed budget — null exhaust makes the gentlest possible mass ratio

Worked examples

Example 1 — reading both slots in a copper wire. A copper wire of radius $a=1.0$ mm carries a steady current density $J=1.0\times10^{7}$ A/m² (a stiff but realistic loading; $I=J\pi a^2=31.4$ A). Copper's resistivity: $1.68\times10^{-8}$ Ω·m.

Step 1. The driving field: $E=\rho_{\text{res}}J=0.168$ V/m along the wire. Time slot of (1): the field ledger pays matter $\mathbf{E}\cdot\mathbf{J}=1.68\times10^{6}$ W/m³ — about 5.3 W per centimeter of wire, arriving (per E7's Poynting Vector audit) sideways through the surface, not down the copper.

Step 2. Space slots: at the surface the wire's own field is $B=\mu_0I/2\pi a=6.28$ mT, so the momentum transfer density is $|\mathbf{J}\times\mathbf{B}|=6.3\times10^{4}$ N/m³, directed inward — the self-pinch. Integrated over the cross-section this squeezes the conductor (harmless at 31 A; at the mega-ampere currents of a Z-pinch machine the same slot implodes the load — that is the point of the machine).

Step 3. The covariant reading: one equation supplied both numbers — the watts and the newtons are the $\nu=0$ and $\nu=r$ slots of a single four-balance, and a moving observer would see them blend.

Example 2 — the photon rocket's exact mass budget. The cleanest closed-total problem in relativity: a rocket of initial rest mass $m_i$ converts mass to radiation and beams it backward perfectly. What mass $m_f$ remains when it reaches $\beta=0.5$ ?

Step 1. Conservation applies to the total: (rocket $p^\nu$ ) + (radiation $P^\nu_{\text{field}}$ ), the radiation's budget being a legitimate null four-vector by Derivation B's vacuum corollary. In the launch frame: initially $(m_ic,\,0)$ ; finally the rocket carries $(\gamma m_fc,\,\gamma m_f\beta c)$ and the beamed light carries $(E_L/c,\,-E_L/c)$ (null: energy and momentum locked, the best exhaust ratio physics permits).

Step 2. Momentum slot: $\gamma m_f\beta c=E_L/c$ . Energy slot: $\gamma m_fc+E_L/c=m_ic$ . Eliminate $E_L$ :

m_i=\gamma m_f(1+\beta) \qquad\Longrightarrow\qquad \frac{m_i}{m_f}=\sqrt{\frac{1+\beta}{1-\beta}} .

Step 3. Numbers: at $\beta=0.5$ , $m_i/m_f=\sqrt{3}=1.73$ — 42% of the ship's rest mass leaves as light to reach half lightspeed. (For perspective: the same $\Delta v$ with 30 km/s chemical exhaust would demand a mass ratio of $e^{5000}$ — the covariant budget is why interstellar daydreams always end in photon or antimatter drives.) A sanity slot: the emitted light's budget $(E_L/c)(1,-1,0,0)$ with $E_L=\gamma m_f\beta c^2=0.5\,\gamma m_fc^2$ is exactly null, as radiated four-momentum must be.

Going further

Reading. Jackson, Classical Electrodynamics, 3rd ed., §12.10 and §16.4–16.5 — the covariant conservation law and the closure subtleties. Landau & Lifshitz, Classical Theory of Fields, §§32–33. Zangwill, Modern Electrodynamics, §22.7. Griffiths, 4th ed., §8.1–8.2 for the 3+1 anchors.

Provenance. Built from the AllQFT base article for this topic (balance law, global conservation, angular-momentum corollary retained), conventions locked to this track's $(+,-,-,-)$ dictionary and to the stress-energy form sign-audited by The Electromagnetic Stress-Energy Tensor. A verifier should: (1) rerun Derivation A including the Bianchi rearrangement and the symbolic RHS unpacking ( $-\mathbf{E}\cdot\mathbf{J}/c$ ; $-(\rho\mathbf{E}+\mathbf{J}\times\mathbf{B})$ , sympy-checked for this edition); (2) confirm the slot audit against Poynting's theorem and the momentum balance; (3) recompute Example 1 ( $E=0.168$ V/m, $\mathbf{E}\cdot\mathbf{J}=1.7\times10^6$ W/m³, $B_{\text{surf}}=6.28$ mT, pinch $6.3\times10^4$ N/m³ at $I=31.4$ A) and Example 2 ( $m_i/m_f=\sqrt{(1+\beta)/(1-\beta)}=\sqrt3$ at $\beta=0.5$ ). Division of labor for this edition: the 3+1 slot owners are E7's — Poynting's Theorem (energy), Conservation of Linear Momentum in Electrodynamics (momentum, coefficients forced), Angular Momentum of the Electromagnetic Field (rotation), Hidden Momentum and Conservation Laws and Spacetime Symmetries (boost charge and center-of-energy; the filing-cabinet deferral delivered here) — recalled by name, never re-derived; the tensor object by The Electromagnetic Stress-Energy Tensor (previous); the two Maxwell lines by Covariant Form of Maxwell's Equations and The Dual Field Tensor; the particle-side gear by Relativistic Dynamics of a Charged Particle; the 4/3 surgery by E11's Classical Electron Theory and Self-Energy; media bookkeeping by E7's media article and Tensor Formulation of Fields in Media (ahead). This edition owns the covariant derivation, the closure theorem's statement and use, the angular-momentum corollary, and the two audits.

Builds on: The Electromagnetic Stress-Energy Tensor · Covariant Form of Maxwell's Equations · The Dual Field Tensor · Relativistic Dynamics of a Charged Particle · The Four-Current Density

Leads to: Covariant Wave Equation and Retarded Solutions · Tensor Formulation of Fields in Media · Relativistic Doppler Effect and Aberration

Why is the right-hand side exactly the negative of the Lorentz-force density — could there have been extra terms?

No, and the derivation shows why with unusual finality: after the inhomogeneous equation inserts $\mu_0J$ , everything else cancels by the Bianchi identity — an identity, not an equation, so no field configuration evades it. Physically this is the statement that the field has exactly one handle by which matter can grab it: the minimal coupling $-J^\mu A_\mu$ of the action article. Vary that coupling one way and you get the Lorentz force on particles; vary it the other way and you get the current source in Maxwell's equations; the balance law's RHS is those two variations meeting. A different RHS would require a different coupling — dipole terms, say, for particles with structure — and then both the force law and the transfer term would change together, keeping the books balanced. The teller window is honest because both ledgers were built from the same action.

Is Poynting's theorem literally the time component of this law, or just analogous to it?

Literally, coefficient for coefficient. Set $\nu=0$ in equation (1): the left side is $\partial_\mu\theta^{\mu0}=\tfrac1c\partial_t\theta^{00}+\partial_i\theta^{i0}=\tfrac1c(\partial_tu+\nabla\cdot\mathbf{S})$ , using the slot identifications audited in the stress-energy article; the right side is $-\mathbf{E}\cdot\mathbf{J}/c$ (sympy-verified for this edition). Multiply by $c$ and you hold E7's Poynting theorem, unchanged. What the covariant form adds is the companionship: a boost mixes this slot with the momentum slots, so the moving observer's energy theorem contains the lab's force balance. E7 derived four theorems with four vector-calculus campaigns; the tensor derivation runs one campaign and gets the other three by relabeling $\nu$ — the compression is the physics, not just the notation.

If field energy alone is not conserved near charges, what stops energy from silently leaking between frames?

The transfer term's own covariance. The teller window $-F^{\nu\lambda}J_\lambda$ is a four-vector density: what one observer books as pure Joule heating, a boosted observer books as part heating, part force-times-velocity — but both agree it left the field ledger and entered the matter ledger at the same events. There is no frame in which a transfer disappears or a leak opens, because the balance law is a tensor equation (zero residual in one frame, zero in all — the same argument that made Maxwell's system frame-proof). The only genuine hazard is the one Laue flagged: integrated single-ledger totals compare different simultaneity slices across frames, and for open subsystems those totals are honestly different numbers — not a leak, but an accounting mismatch from cutting the system open along different seams.

How does angular momentum conservation depend on the symmetry of the stress-energy tensor?

Compute the divergence of the candidate current $M^{\mu\nu\lambda}=x^\mu\theta^{\nu\lambda}-x^\nu\theta^{\mu\lambda}$ : the product rule gives $\delta^\mu_\lambda\theta^{\nu\lambda}-\delta^\nu_\lambda\theta^{\mu\lambda}=\theta^{\nu\mu}-\theta^{\mu\nu}$ plus terms proportional to $\partial_\lambda\theta^{\nu\lambda}$ (zero for free fields, transfer terms with sources). The antisymmetric residue $\theta^{\nu\mu}-\theta^{\mu\nu}$ would be a torque density with no source — angular momentum appearing from bookkeeping — unless the tensor is symmetric. This is why the Belinfante healing (mechanics track) matters beyond aesthetics: the canonical tensor's asymmetry is exactly compensated by a spin current, and the healed, symmetric $\theta$ absorbs that spin bookkeeping into the orbital-looking current (2). For electromagnetism the physical checks are E7's: circularly polarized beams really torque absorbers, and the field angular momentum in crossed static fields really balances the Feynman disk — all of it consistent because the symmetric tensor exists.

Why is light the ideal rocket exhaust in the covariant budget?

Because the exhaust's four-momentum is null: for radiation, momentum equals energy over $c$ — no exhaust can buy more momentum per joule spent, since any massive exhaust at speed $v<c$ delivers $p=Ev/c^2\cdot(\ldots)<E/c$ of momentum for the same energy outlay. Example 2's algebra makes the optimality visible: the null exhaust saturates the constraint between the two conservation slots, giving the gentlest possible mass ratio $\sqrt{(1+\beta)/(1-\beta)}$ — and even that is brutal (42% of the ship gone by $0.5c$ ). The same null-budget logic runs the other direction in laser-sail schemes: leave the power plant at home, beam the null momentum at the ship, and the ship's ledger credits up to $2E/c$ per reflected joule (the factor 2 from the mirror's double exchange). Both designs are corollaries of Derivation B plus the stress-energy article's null dyad.

Does the balance law survive inside a material medium?

The logic survives; the objects must be renegotiated, and the chapter has been honest about this twice. In vacuum-with-charges, $\theta^{\mu\nu}$ is unambiguous and equation (1) is exact with the microscopic fields and currents — that always remains true if you resolve matter into its charges. The dispute begins when you average into macroscopic fields ( $\mathbf{D}$ , $\mathbf{H}$ ) and try to split the total stress-energy into "field in the medium" plus "medium": Minkowski's and Abraham's splits assign different momenta to light in glass, and E7's Electromagnetic Stress, Momentum, and Energy in Media owns the resolution (kinetic vs canonical momentum; both right about different questions). The covariant machinery for the macroscopic split — the tensors $G^{\mu\nu}$ and $M^{\mu\nu}$ in which the Minkowski form is naturally written — is precisely the business of the next-but-two article, Tensor Formulation of Fields in Media. Until then: microscopic version, always safe; macroscopic version, declare your split.

Where does this conservation law come from in the Noether picture?

From spacetime translation symmetry of the action built two articles ago. Noether's theorem (mechanics track, and the field-theoretic edition in its field-theory chapter) assigns a conserved current to every continuous symmetry; invariance under the four translations $x^\mu\to x^\mu+a^\mu$ yields the four currents packaged as the stress-energy tensor, and Conservation Laws and Spacetime Symmetries told this story for electromagnetism at the structural level, deferring the covariant machinery here. The honest fine print from that lineage: Noether's recipe outputs the canonical tensor, which needs the Belinfante healing before it matches the physical $\theta$ used in this article — surgery owned by the mechanics track and cited by the previous article. With sources included, translation symmetry of the total action (field + particles + coupling) is what guarantees the total conservation of Derivation B; the field-only ledger fails translation symmetry exactly where charges sit, which is one more way of reading the teller window.