Coulomb Gauge

Back in magnetostatics, the Coulomb gauge was a modest housekeeping rule: impose $\nabla\cdot\mathbf{A}=0$ , and the vector potential's equation collapses to three tidy Poisson problems. That article — Gauge Freedom and the Coulomb Gauge — proved the condition attainable, priced its leftover freedom, and closed with a promissory note: in time-dependent theory the same gauge produces a scalar potential that updates instantaneously across the universe, an apparent causality scandal whose acquittal was deferred "to the radiation chapter." This is the radiation chapter, and this article pays the note. Impose $\nabla\cdot\mathbf{A}=0$ when the sources move, and the equations split with striking asymmetry. The scalar potential obeys the electrostatic Poisson equation at every instant — charge wiggles in a distant galaxy and $V$ here rearranges now, with no light-travel delay. The vector potential, meanwhile, obeys a genuine wave equation, but sourced by only part of the current: the transverse (divergence-free) part, with the longitudinal part surgically cancelled by the very term the instantaneous $V$ generates. Everything acausal is quarantined in $V$ ; everything radiative travels at $c$ inside $\mathbf{A}$ . The scandal dissolves once you watch the fields rather than the potentials: outside the light cone of any source change, the transverse field conspires to cancel the updated Coulomb piece exactly, and the measurable $\mathbf{E}$ still carries yesterday's news. Prosecuting that cancellation honestly — and understanding why the transverse/longitudinal split is nonlocal enough to pull it off — is this article's core business. The reward for tolerating the weirdness is the gauge modern physics quantizes light in: two transverse polarizations, an explicit Coulomb interaction, and the working coordinates of atomic physics and QED.

Key takeaways

The Coulomb gauge imposes $\nabla\cdot\mathbf{A}=0$ in full electrodynamics. Gauss's law then forces $\nabla^2 V=-\rho/\varepsilon_0$ — Poisson's equation with the charge density at the same instant: $V$ is the electrostatic potential of the frozen present configuration, recomputed at every $t$ .
The vector potential obeys a true wave equation, $\Box^2\mathbf{A}=-\mu_0\mathbf{J}_\perp$ , sourced only by the transverse (divergence-free) part of the current; the longitudinal part is cancelled by $\mu_0\varepsilon_0\nabla(\partial V/\partial t)$ , courtesy of charge conservation.
The instantaneous $V$ violates nothing: potentials are gauge coordinates, not observables. Outside the light cone of a source change, the transverse piece of $\mathbf{E}$ cancels the updated Coulomb piece exactly; the total field is causal in every gauge.
The Helmholtz longitudinal/transverse split is nonlocal — that is how a "transverse" field can carry the instantaneous antidote — and it is frame-dependent, which is the covariance price of this gauge.
Far from charges one may set $V=0$ outright: the radiation gauge, where the two transverse components of $\mathbf{A}$ are the only dynamical variables — exactly the photon's two polarizations.
Division of labor: the magnetostatic twin owns attainability, the harmonic residual, and the static toolkit; this article owns the dynamical structure and the causality acquittal.
Key idea: with $\nabla\cdot\mathbf{A}=0$ , the scalar potential is the instantaneous Coulomb integral and the vector potential is a causal wave driven by the transverse current alone — and the observable fields remain strictly causal.

The core insight

Imposing $\nabla\cdot\mathbf{A}=0$ makes the scalar potential obey an instantaneous Poisson equation while the vector potential is driven only by the transverse current; the observable fields nonetheless remain causal.

Where this lives today

The Coulomb gauge is where light gets quantized. Set $\rho = 0$ , take $V = 0$ and $\nabla\cdot\mathbf{A} = 0$ , and the field's only dynamical variables are the two transverse components of $\mathbf{A}$ — each Fourier mode a harmonic oscillator, each quantum a photon with exactly two polarization states, no ghost bookkeeping required. That radiation-gauge quantization, essentially in Fermi's 1932 formulation, still underlies quantum optics: cavity QED, circuit QED with superconducting qubits, and the input–output theory behind every quantum-limited amplifier are written in it. Atomic, molecular, and condensed-matter physics stay in Coulomb gauge for a different reason: the instantaneous $1/r$ Coulomb interaction sits explicitly in the Hamiltonian — the term that makes atoms bind — while the transverse field enters as the $\mathbf{p}\cdot\mathbf{A}$ coupling that drives absorption and emission. The gauge's org chart is the structure of quantum chemistry. The frame-dependence that embarrasses the gauge relativistically is priced and paid daily: high-energy theory uses covariant gauges for loop integrals, then checks gauge-invariant answers, while bound-state QED (Lamb shifts, positronium, precision hydrogen) still often runs in Coulomb gauge because binding is a rest-frame concept. And on the computational side, "solve a Poisson equation every step, project the current transverse" is the literal inner loop of electromagnetic particle-in-cell plasma codes and Gummel-cycle device simulators — the two clerks, implemented as a fast elliptic solve plus a hyperbolic update, integrating the same split this article derived.

Origins · 1 minJames Clerk Maxwell and how this was discovered

James Clerk Maxwell

1831–1879 · Edinburgh, Cambridge, London — whose own equations lived in the Coulomb gauge

It is one of history's neat ironies that the man who discovered electromagnetic waves never wrote the gauge that makes them look simplest. Throughout the Treatise on Electricity and Magnetism (1873), Maxwell worked with the potentials as primary objects and systematically imposed $\nabla\cdot\mathbf{A}=0$ — the Coulomb gauge, decades before the name existed. His scalar potential was therefore the instantaneous Coulomb integral, and the strange split this article prosecutes — omniscient $V$ , causal transverse $\mathbf{A}$ — is native to the original formulation of electrodynamics, not a modern curiosity. FitzGerald and Heaviside later grumbled that Maxwell's potential-first, Coulomb-gauge presentation obscured the propagation physics (Heaviside campaigned to "murder the potentials" outright), and the fields-first rewrite became the textbook standard. The pendulum swung back: quantum mechanics restored the potentials to primacy, and Fermi's Coulomb-gauge QED put Maxwell's coordinates at the center of the quantum theory of light. Maxwell had, characteristically, chosen the gauge that the physics of bound charges and photons would eventually demand.

History

1824

— Poisson's equation for the electrostatic potential: the instantaneous ledger's mathematics, a half-century early

1867

— Ludvig Lorenz publishes retarded potentials with the rival condition — causal potentials, decades before covariance made them fashionable

1873

— Maxwell's Treatise develops electrodynamics with potentials primary and ∇·A = 0 imposed throughout

1888

— Hertz generates electromagnetic waves; the fields-first rewrite of Maxwell by Heaviside and Hertz sidelines the potentials

1932

— Fermi's Reviews of Modern Physics article fixes Coulomb-gauge quantization as the working formulation of QED for bound systems

1951

— Gupta and Bleuler tame covariant (Lorenz-gauge) quantization with indefinite metric — the price the Coulomb gauge never pays

1967

— Brill and Goodman publish the explicit causality reconciliation: the light-cone cancellation between longitudinal and transverse pieces

The physical picture

Picture the gauge as a firm with two clerks. The first clerk — the scalar potential — maintains the Coulomb ledger: at every tick of the clock, he looks at where all the charge is right now and writes down the electrostatic potential of that frozen arrangement, as if the world were a statics problem refreshed infinitely fast. He is omniscient and instantaneous and completely unobservable. The second clerk — the transverse vector potential — handles everything that actually propagates: her entries obey a wave equation, travel at $c$ , and carry the radiation. The gauge condition $\nabla\cdot\mathbf{A}=0$ is the firm's org chart: it assigns all divergence (all charge-tracking) to the first clerk and all curl (all circulation and radiation) to the second, with no overlap. The current splits along the same line. Any current density decomposes — the Helmholtz decomposition of the vector-calculus chapter — into a curl-free longitudinal piece $\mathbf{J}_\parallel$ , which shuffles charge from place to place and is fully dictated by $\partial\rho/\partial t$ through continuity, and a divergence-free transverse piece $\mathbf{J}_\perp$ , which circulates without accumulating charge anywhere. The longitudinal piece is the first clerk's problem: it only updates the Coulomb ledger. The transverse piece is the only thing that radiates. This is not a gauge artifact dressed as insight — it is the reason antenna engineers speak of effective currents and why the dipole radiation pattern three articles ahead carries a $\sin\theta$ : that factor is the transverse projection.

Figure

The Coulomb-gauge division of labor — an instantaneous Coulomb clerk for V and a causal wave clerk for the transverse A

The equations to be earned are, from Gauss's law and the Ampère–Maxwell law with the potential definitions of the chapter opener inserted:

\nabla^2 V = -\frac{\rho}{\varepsilon_0}, \qquad\qquad \Box^2\mathbf{A} \equiv \nabla^2\mathbf{A} - \mu_0\varepsilon_0\frac{\partial^2\mathbf{A}}{\partial t^2} = -\mu_0\mathbf{J}_\perp, \qquad (1)

where: $\rho(\mathbf{r},t)$ is the present charge density — no retardation — and $\mathbf{J}_\perp$ is the transverse part of the current in the Helmholtz sense, $\nabla\cdot\mathbf{J}_\perp = 0$ . Read the asymmetry: the first equation has no time derivatives at all (elliptic, instantaneous, boundary-value character), while the second is hyperbolic and causal. One gauge, two completely different kinds of physics problem — this split is the gauge's entire personality, and both its power and its scandal flow from it. The scalar equation solves exactly as in electrostatics, instant by instant:

V(\mathbf{r},t) = \frac{1}{4\pi\varepsilon_0}\int \frac{\rho(\mathbf{r}',t)}{|\mathbf{r}-\mathbf{r}'|}\,d^3r'. \qquad (2)

where: the source is evaluated at the same $t$ as the field point — the "instantaneous Coulomb integral." Note what this is not: it is not a claim that anything physical moves instantaneously; it is the definition of one gauge coordinate. The retarded integral that answers the corresponding question causally is two articles ahead, and comparing the two side by side is the fastest way to see what a gauge choice does and does not commit you to.

Derivation AThe Coulomb-gauge equations — instantaneous Poisson for V, transverse-sourced wave for A

The goal is the full system (1): impose $\nabla\cdot\mathbf{A}=0$ on the general coupled potential equations and reduce them, tracking exactly where the longitudinal current disappears — the cancellation is the theorem here, not a bookkeeping afterthought. The strategy: Gauss's law first (one line, delivering the instantaneous Poisson equation), then the Ampère–Maxwell law via the curl-of-curl identity, and finally the identification of the leftover $\nabla(\partial_t V)$ term as exactly $\mu_0\mathbf{J}_\parallel$ using continuity plus the Helmholtz uniqueness argument recalled from the vector-calculus chapter. Watch for the moment charge conservation enters — without it the gauge would not close.

Step 1. The scalar equation. Gauss's law with $\mathbf{E}=-\nabla V - \partial_t\mathbf{A}$ :

\nabla\cdot\mathbf{E} = -\nabla^2 V - \frac{\partial}{\partial t}(\nabla\cdot\mathbf{A}) = \frac{\rho}{\varepsilon_0} \qquad\xrightarrow{\ \nabla\cdot\mathbf{A}=0\ }\qquad \nabla^2 V = -\frac{\rho}{\varepsilon_0}.

Why this is the gauge's signature move: the only time derivative in Gauss's law rides on $\nabla\cdot\mathbf{A}$ , and the condition erases it. What remains is elliptic — no wave operator, no retardation, the same equation solved throughout electrostatics, now with a time-stamped source.

Step 2. The vector equation, raw. Ampère–Maxwell with both potential definitions inserted and the curl-of-curl identity $\nabla\times(\nabla\times\mathbf{A}) = \nabla(\nabla\cdot\mathbf{A}) - \nabla^2\mathbf{A}$ :

\nabla(\underbrace{\nabla\cdot\mathbf{A}}_{=0}) - \nabla^2\mathbf{A} = \mu_0\mathbf{J} - \mu_0\varepsilon_0\,\nabla\frac{\partial V}{\partial t} - \mu_0\varepsilon_0\frac{\partial^2\mathbf{A}}{\partial t^2},

so that

\Box^2\mathbf{A} = -\mu_0\mathbf{J} + \mu_0\varepsilon_0\,\nabla\frac{\partial V}{\partial t}.

Why the leftover term is a feature: the gauge has killed the $\nabla(\nabla\cdot\mathbf{A})$ ugliness, but the scalar potential still contaminates the vector equation through $\nabla\partial_t V$ . The next two steps show this contamination is precisely engineered — it subtracts the non-radiating part of the current.

Step 3. Identify the contamination as the longitudinal current. Apply $\partial_t$ to Step 1's Poisson equation and use continuity, $\partial_t\rho = -\nabla\cdot\mathbf{J}$ :

\nabla^2\frac{\partial V}{\partial t} = -\frac{1}{\varepsilon_0}\frac{\partial\rho}{\partial t} = \frac{1}{\varepsilon_0}\nabla\cdot\mathbf{J} \qquad\Longrightarrow\qquad \nabla\cdot\!\left(\varepsilon_0\nabla\frac{\partial V}{\partial t}\right) = \nabla\cdot\mathbf{J}_\parallel,

where the last equality holds because the transverse part of $\mathbf{J}$ is divergence-free by definition. Now compare the two fields $\varepsilon_0\nabla\partial_t V$ and $\mathbf{J}_\parallel$ : both are curl-free (one is a gradient; the other by the definition of the Helmholtz split), they have equal divergences everywhere, and both vanish at infinity for localized sources. By the Helmholtz uniqueness theorem of the vector-calculus chapter,

\mathbf{J}_\parallel = \varepsilon_0\,\nabla\frac{\partial V}{\partial t}.

Why continuity is the load-bearing wall: without $\partial_t\rho = -\nabla\cdot\mathbf{J}$ the divergences would not match and the identification would fail. The instantaneous $V$ "knows" how charge is about to rearrange only because the longitudinal current is enslaved to $\rho$ by charge conservation — the gauge is spending the continuity equation, not luck.

Step 4. The cancellation. Substitute Step 3 into Step 2:

\Box^2\mathbf{A} = -\mu_0\left(\mathbf{J} - \mathbf{J}_\parallel\right) = -\mu_0\mathbf{J}_\perp.

Why this earns the gauge its aliases: the wave equation for $\mathbf{A}$ — the only causal, propagating object in the formalism — is driven by the transverse current alone. Longitudinal charge-shuffling never radiates; it only refreshes the Coulomb ledger. Hence "transverse gauge" and, in the source-free case, "radiation gauge." Consistency check: taking the divergence of both sides gives $\Box^2(\nabla\cdot\mathbf{A}) = -\mu_0\nabla\cdot\mathbf{J}_\perp = 0$ — the gauge condition, once imposed, propagates itself. *As a consistency check: the static limit of both equations recovers the electrostatic Poisson equation and the magnetostatic $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ of the twin article, where $\mathbf{J}=\mathbf{J}_\perp$ automatically since steady currents are divergence-free.*

The scandal, prosecuted

Equation (2) says that if a charge on Proxima Centauri twitches at noon, the Coulomb-gauge $V$ in this room changes at noon — four years before any light arrives. Before dissolving the paradox, sharpen it: this is not a small effect, not a far-field subtlety, and not an approximation artifact. It is the exact, intended behavior of the gauge. The resolution has to be structural, and it is: the acausal content of $V$ is exactly cancelled, outside the light cone, by the transverse field — and the cancellation works because the Helmholtz split is nonlocal. That last point deserves its own paragraph, because it is where intuition usually breaks. "Transverse" in the Helmholtz sense does not mean "locally perpendicular to something"; it means divergence-free, a condition diagnosed by integrals over all space. The transverse projection of a current localized in a wire is not localized in the wire — it has tails filling all space (the projection kernel falls off as a power law). So $\mathbf{E}_\perp = -\partial_t\mathbf{A}$ , despite obeying a causal-looking wave equation sourced by $\mathbf{J}_\perp$ , receives contributions from transverse-current tails that switched on the instant the distant charge moved. The transverse field is causal with respect to $\mathbf{J}_\perp$ — but $\mathbf{J}_\perp$ itself is an instantaneously nonlocal functional of the physical current. Two instantaneous wrongs, one causal right.

Figure

The light-cone cancellation — a source rearranges at t = 0; outside r = ct the updated Coulomb piece and the transverse antidote sum to the old field

Go deeper · Derivation B · 2 minCausality of the fields — the light-cone cancellation, exhibited

Derivation BCausality of the fields — the light-cone cancellation, exhibited

The goal is the acquittal: show that the total electric field in Coulomb gauge is strictly causal — outside the light cone of any source change, $\mathbf{E}$ still reports the old configuration — and exhibit the mechanism, namely that the transverse piece $-\partial_t\mathbf{A}$ develops an instantaneous component equal and opposite to the updated Coulomb field. The strategy is a bookkeeping argument built on two facts: (i) the total fields obey gauge-invariant, sourced wave equations — recalled from the wave chapter's opener, which derived $\Box^2\mathbf{E} = \tfrac{1}{\varepsilon_0}\nabla\rho + \mu_0\,\partial_t\mathbf{J}$ directly from Maxwell — whose solutions with quiescent past are retarded and causal (the uniqueness and domain-of-dependence argument is the next article's Derivation A; the explicit retarded field formulas are Jefimenko's Equations, which serve as this derivation's receipts); and (ii) the Coulomb-gauge split of that causal total into $\mathbf{E}_\parallel + \mathbf{E}_\perp$ is exact and unambiguous. Squeezing (ii) against (i) forces the cancellation.

Step 1. Split the field along the gauge's org chart. In Coulomb gauge $\mathbf{A}$ is divergence-free, so $-\partial_t\mathbf{A}$ is transverse; and $-\nabla V$ is a gradient, hence longitudinal. The split of $\mathbf{E}$ is therefore aligned with the potentials:

\mathbf{E}_\parallel = -\nabla V, \qquad \mathbf{E}_\perp = -\frac{\partial\mathbf{A}}{\partial t}.

Why this alignment is special to this gauge: in a general gauge both potentials contribute to both Helmholtz components. Coulomb gauge is exactly the choice that makes the split of $\mathbf{E}$ read off the potentials — which is why the scandal is so visible here.

Step 2. The longitudinal field is the instantaneous Coulomb field. From (2),

\mathbf{E}_\parallel(\mathbf{r},t) = \frac{1}{4\pi\varepsilon_0}\int \rho(\mathbf{r}',t)\,\frac{\mathbf{r}-\mathbf{r}'}{|\mathbf{r}-\mathbf{r}'|^3}\,d^3r'

— the electrostatic field of the present configuration, updated with zero delay everywhere. Taken alone, this piece is flagrantly acausal. Why we do not panic: $\mathbf{E}_\parallel$ alone is not measurable. No detector projects onto Helmholtz components — the projection requires knowledge of the field everywhere at one instant. Detectors read the total $\mathbf{E}$ .

Step 3. The total field is causal. The total $\mathbf{E}$ obeys the sourced wave equation recalled above, with sources $\nabla\rho$ and $\partial_t\mathbf{J}$ that are physical, local, and causal. With a quiescent (or previously settled) past, the unique solution is the retarded one: $\mathbf{E}(\mathbf{r},t)$ depends on sources only on the past light cone. Concretely — the scenario to hold in mind — let a source sit static until $t=0$ , then rearrange. For $|\mathbf{r}| > ct$ , the total field is still the old static field, unchanged.

Step 4. Force the books to balance. Outside the light cone ( $|\mathbf{r}| > ct$ ):

\underbrace{\mathbf{E}}_{\text{old field (Step 3)}} = \underbrace{\mathbf{E}_\parallel}_{\text{NEW Coulomb field (Step 2)}} + \;\mathbf{E}_\perp \qquad\Longrightarrow\qquad \mathbf{E}_\perp = \mathbf{E}^{\text{old}} - \mathbf{E}_\parallel^{\text{new}} \neq 0 .

The transverse field outside the light cone is forced to be nonzero: it must supply the old field while subtracting the updated Coulomb field. The instantaneous piece of $\mathbf{E}_\perp$ — sourced by the nonlocal tails of $\mathbf{J}_\perp$ discussed above — is exactly the antidote, point by point, to the instantaneous update of $\mathbf{E}_\parallel$ . Inside the light cone, both pieces are "new" and their sum is the genuine retarded field. Why the conspiracy is guaranteed rather than miraculous: both statements — the causal total and the instantaneous split — are theorems about the same solution of Maxwell's equations. The cancellation is their only possible reconciliation. What looks like fine-tuning is bookkeeping: the gauge put the acausality into $V$ by hand, so it must hide the antidote in $\mathbf{A}$ by necessity. *This is the classic analysis of Brill & Goodman (Am. J. Phys. 35, 832, 1967). The explicit retarded field formulas that serve as Step 3's receipts are Jefimenko's equations, three articles ahead, which contain no reference to any gauge — as befits gauge-invariant observables.*

Residual freedom, and what pins it

The magnetostatic twin showed that $\nabla\cdot\mathbf{A}=0$ survives further gauge shifts with harmonic gauge functions, $\nabla^2\lambda = 0$ . Dynamics adds a wrinkle: the shift $V \to V - \partial_t\lambda$ must also preserve the instantaneous Poisson character of $V$ , and since $\nabla^2(\partial_t\lambda) = \partial_t(\nabla^2\lambda) = 0$ , any time-dependent harmonic $\lambda(\mathbf{r},t)$ is formally admissible. What kills the residue in practice is decay: a harmonic function that vanishes at infinity is identically zero (the mean-value property of the boundary-value chapter), so demanding that both potentials die at infinity at every instant leaves $\lambda =$ constant — no residual freedom at all. For localized sources, Coulomb gauge plus decay conditions is a complete gauge fixing: equations (1)–(2) have exactly one solution pair. This is tidier than the Lorenz gauge, whose residual freedom (next article) is a genuine family of gauge waves.

Limiting cases

case	result	meaning
statics	$\nabla^2V=-\rho/\varepsilon_0$ , $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$	both gauges' equations coincide ( $\partial_t V=0$ makes Lorenz ≡ Coulomb); the twin article's home territory
source-free region	$V=0$ , $\Box^2\mathbf{A}=0$ , $\nabla\cdot\mathbf{A}=0$	the radiation gauge: two transverse wave amplitudes, the classical skeleton of photon quantization
slow sources ( $v\ll c$ , sizes $\ll\lambda$ )	fields ≈ instantaneous Coulomb + small transverse correction	the quasistatic regime: Coulomb gauge makes the hierarchy explicit — the leading physics is the first clerk's ledger
single Fourier mode $e^{i\mathbf{k}\cdot\mathbf{r}}$	$\mathbf{A}\perp\mathbf{k}$ ; $\mathbf{J}_\perp = \mathbf{J}-\hat{\mathbf{k}}(\hat{\mathbf{k}}\cdot\mathbf{J})$	in $k$ -space the Helmholtz split is a local projector — the nonlocality in real space is the Fourier transform's spreading
relativistic boost	$\nabla\cdot\mathbf{A}=0$ fails in the new frame	the condition is not Lorentz-invariant: each frame has its own Coulomb gauge — the covariance price

Common trapThe instantaneous potential is not a faster-than-light signal — and the transverse field is not locally transverse

Two traps, one root. First: no experiment reads $V$ ; the measurable fields are causal in every gauge, with the light-cone cancellation of Derivation B doing the work. Anyone proposing to signal superluminally via the Coulomb-gauge potential has confused a coordinate with an observable — the same error as reading simultaneity conventions as physics. Second, subtler: "transverse" is a global (Helmholtz) property, not a local one. $\mathbf{J}_\perp$ of a current confined to a wire fills all space; $\mathbf{E}_\perp$ outside the light cone is nonzero precisely where naive intuition says nothing has arrived. If you find yourself Helmholtz-splitting fields inside an argument about locality or signal speed, stop: the split itself is instantaneously nonlocal and frame-dependent. Causality claims belong to the total fields.

Worked examples

1 · The blinking dipole and the one-millisecond ledger (confidence builder)

Setup. Two small spheres 1 m apart exchange charge through a wire at $f = 1$ kHz: $q(t) = q_0\cos\omega t$ on one sphere, $-q(t)$ on the other, with $q_0 = 1\ \mu$ C — an oscillating dipole with $p_0 = q_0 d = 10^{-6}$ C·m. An observer sits at $r = 300$ km, on the dipole axis. How do the Coulomb-gauge $V$ and the physical field each respond? (Chosen distance is no accident: $\lambda = c/f = 300$ km, so the observer sits right at the radiation-zone boundary, and the light-travel time $r/c = 1.0$ ms is one full period.)

Step 1. The instantaneous ledger. The Coulomb-gauge $V$ at the observer is the present dipole potential, $V(t) = \frac{1}{4\pi\varepsilon_0}\frac{p_0\cos\omega t}{r^2}$ on axis. Amplitude: $V_0 = (8.99\times10^9)(10^{-6})/(3\times10^5)^2 = \mathbf{1.0\times10^{-7}\ V}$ . Its phase tracks the source with zero delay: when the laboratory clock reads $q = q_0$ , the potential 300 km away already reads its maximum.

Step 2. The physical field. The measurable $\mathbf{E}$ at 300 km is built from the source one light-time earlier — a full period back. In this radiation zone the field is dominated not by the Coulomb-ledger $1/r^2$ piece but by the $1/r$ radiation term (delivered in Electric Dipole Radiation), and every crest of it left the source 1.0 ms before arriving. A fast field mill at 300 km and an identical one at 3 km would see the same waveform shifted by $\Delta t = (300-3)\ \text{km}/c = 0.99$ ms — the fields march outward at $c$ , indifferent to the ledger's omniscience.

Step 3. Where the books balance. Between the two clerks: the observer's $\mathbf{E}_\parallel$ (instantaneous, amplitude set by $V_0/r$ -type scales) is cancelled outside each light cone by the transverse clerk's antidote piece, and what survives is the retarded total. Nothing about the choice of 1 kHz or 300 km matters to the principle — scale it to Proxima Centauri and four years, and the arithmetic of the cancellation is identical.

The lesson: the gauge's instantaneous entries are enormous in extent and zero in consequence. Causality lives in the totals.

2 · Transverse projection is the radiation pattern (exam level)

Setup. A short center-fed antenna carries a uniform oscillating current $\mathbf{I} = I_0\cos\omega t\,\hat{\mathbf{z}}$ , $I_0 = 1$ A. Consider the far field propagating toward a direction $\hat{\mathbf{k}}$ at angle $\theta$ from the antenna axis. Using the $k$ -space form of the Helmholtz split (the limiting-cases table), find the transverse current fraction driving $\mathbf{A}$ in that direction, and connect it to the dipole radiation pattern.

Step 1. The projector. For the Fourier mode traveling along $\hat{\mathbf{k}}$ , the transverse projection of the source is $\mathbf{J}_\perp = \mathbf{J} - \hat{\mathbf{k}}(\hat{\mathbf{k}}\cdot\mathbf{J})$ . With $\mathbf{J} \parallel \hat{\mathbf{z}}$ and $\hat{\mathbf{k}}\cdot\hat{\mathbf{z}} = \cos\theta$ :

|\mathbf{J}_\perp| = |\mathbf{J}|\sqrt{1-\cos^2\theta} = |\mathbf{J}|\sin\theta.

Step 2. Numbers. Toward $\theta = 90°$ (broadside): $|\mathbf{J}_\perp|/|\mathbf{J}| = 1.000$ — the whole current radiates. At $\theta = 60°$ : $0.866$ . At $45°$ : $0.707$ . At $30°$ : $0.500$ . Along the axis ( $\theta = 0$ ): zero — the current is purely longitudinal as seen by that direction, and nothing radiates along the axis of a linear antenna.

Step 3. Power. Radiated power per solid angle rides on $|\mathbf{J}_\perp|^2$ , i.e. on $\sin^2\theta$ : $1.00$ broadside, $0.75$ at $60°$ , $0.50$ at $45°$ , $0.25$ at $30°$ , $0$ on axis. This is exactly the $\sin^2\theta$ donut of Electric Dipole Radiation, obtained three articles early from pure gauge structure: the dipole pattern is the transverse projector made visible. The $\Box^2\mathbf{A} = -\mu_0\mathbf{J}_\perp$ equation was never solved — only its source examined — yet the angular shape of the answer is already fixed.

The lesson: "only the transverse current radiates" is quantitative engineering, not slogan. Antenna nulls are directions in which the current looks longitudinal.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §10.1.3 · Jackson, Classical Electrodynamics, 3rd ed., §6.3, §6.5 · Zangwill, Modern Electrodynamics, §15.3 · Brill & Goodman, "Causality in the Coulomb gauge," Am. J. Phys. 35, 832 (1967) · Fermi, "Quantum theory of radiation," Rev. Mod. Phys. 4, 87 (1932) The three key algebraic checks underlying this article are: (1) substituting the field forms into Gauss's law to obtain $\nabla^2 V=-\rho/\epsilon_0$ under $\nabla\cdot\mathbf{A}=0$ ; (2) the curl-of-curl reduction of Ampère–Maxwell to $\Box^2\mathbf{A}=-\mu_0\mathbf{J}_\perp$ ; (3) using continuity and the time-differentiated Poisson equation to establish $\mathbf{J}_\parallel=\epsilon_0\nabla\partial_t V$ . Constants $\epsilon_0$ and $\mu_0$ are standard vacuum values. Consistency with Jefimenko's equations confirms field causality. The magnetostatic twin Gauge Freedom and the Coulomb Gauge owns the attainability/repair construction, the harmonic-residual analysis, and the static toolkit, all recalled here by name and not re-derived; this article owns the dynamical system (Derivation A, with the $\mathbf{J}_\parallel$ identification as Step 3), the light-cone causality acquittal (Derivation B, after Brill & Goodman), the nonlocality analysis of the Helmholtz split, and the dynamic residual-freedom closure. Worked examples — the $\lambda = 300$ km blinking dipole and the $\sin\theta$ transverse-projection table — have all numbers recomputed.

Builds on: Scalar and Vector Potentials · Gauge Transformations · Gauge Freedom and the Coulomb Gauge (magnetostatics twin) · Helmholtz Decomposition Leads to: Lorenz Gauge · Retarded Potentials · Jefimenko's Equations · Electric Dipole Radiation

What exactly does this article add to the magnetostatics Coulomb-gauge article?

Division of labor, mandated by the chapter plan. The magnetostatic twin owns the gauge machinery: the proof that curl-free differences are gradients, the attainability of $\nabla\cdot\mathbf{A}=0$ by solving a Poisson repair problem, the harmonic residual freedom, the uniform-field menagerie, and the doctrine of gauge-invariance audits — all with static sources, where $V$ never enters the story. This article owns everything that time does to the gauge: the instantaneous Poisson equation for $V$ , the transverse-sourced wave equation for $\mathbf{A}$ with the continuity-powered $\mathbf{J}_\parallel$ cancellation, the causality scandal and its light-cone acquittal (the twin's closing FAQ promised exactly this), the nonlocality of the Helmholtz split, and the radiation-gauge reduction that quantum optics builds on. Read the twin for how to reach and manipulate the gauge; read this for what the gauge does when sources move.

If V updates instantly, could two experimenters use it to communicate faster than light?

No, and it is worth seeing exactly where the scheme dies. To "read" $V$ , the receiving experimenter must measure something, and everything measurable — forces on charges, induced currents, interference phases — is built from $\mathbf{E}$ and $\mathbf{B}$ , which Derivation B shows are unchanged outside the sender's light cone: the updated Coulomb piece and the transverse antidote cancel to whatever precision you compute them. The instantaneous update is real as a statement about the gauge coordinate and empty as a channel, in precisely the way a change of map projection can move every latitude line on Earth "instantly" without moving any city. The same acquittal covers all gauges at once, since gauge transformations never touch the fields.

Is the longitudinal and transverse decomposition of E something I could measure separately?

Not with local apparatus. The split is defined by conditions on divergence and curl — properties that require field data over all space at one instant — and the projection kernels are long-ranged. A detector at a point reads the total field; assigning its reading to $\mathbf{E}_\parallel$ versus $\mathbf{E}_\perp$ requires already knowing the global field configuration. Worse, the split is frame-dependent: a boosted observer's Helmholtz components mix yours. The decomposition is a superbly useful calculational coordinate system — it diagonalizes the gauge's structure and, in $k$ -space, the radiation problem — but no invariant meaning attaches to the pieces separately. (Free radiation far from all sources is the exception where the transverse piece is the whole field.)

Why does atomic physics insist on the Coulomb gauge rather than the manifestly covariant Lorenz gauge?

Because atoms are bound by the Coulomb interaction, and this gauge hands it over explicitly. The instantaneous $V$ produces the $1/r$ electron–nucleus and electron–electron potentials directly in the Hamiltonian — the hydrogen problem, the helium problem, all of quantum chemistry — while the transverse $\mathbf{A}$ enters separately as the radiation field that drives transitions. In Lorenz gauge, the same physics arrives disguised: the binding interaction hides inside retarded photon exchange, and extracting a Schrödinger-like problem requires undoing the covariance you paid for. For systems with a preferred frame (the atom's rest frame), frame-dependence is no defect. Relativistic corrections come in as systematic expansions (Breit interaction, retardation corrections) on top of the Coulomb-gauge skeleton.

Does the transverse current really extend outside the wire that carries the actual current?

Yes — and computing one case makes it believable. The transverse projection subtracts from $\mathbf{J}$ the gradient field $\varepsilon_0\nabla\partial_t V$ , and $V$ is the Coulomb integral of the charge the current is rearranging, which fills all space. For a short current element, $\mathbf{J}_\perp$ outside the element is the (time-derivative of the) dipole-pattern gradient field — falling off as a power law, present everywhere, switching on simultaneously with the current. That is not physical charge flow at those points; it is a mathematical component of a nonlocal decomposition. The physical current is confined to the wire; its transverse part is not, and the wave equation for $\mathbf{A}$ is sourced by the latter. This is the mechanism behind the instantaneous antidote in Derivation B.

In the radiation gauge, where did the scalar potential go — is setting V to zero always allowed?

Only where there is no net charge density. With $\rho=0$ , the instantaneous Poisson equation plus decay at infinity forces $V=0$ with no further choice — the radiation gauge is the Coulomb gauge meeting empty space, not an extra condition. With $\rho\neq0$ you cannot zero $V$ while keeping $\nabla\cdot\mathbf{A}=0$ : the two conditions together would overdetermine the gauge function (temporal gauge $V=0$ is attainable alone, at the price of a nonzero $\nabla\cdot\mathbf{A}$ ). In practice the radiation gauge is the natural habitat for free-field problems — quantizing light, cavity modes, scattering states — while the full Coulomb gauge handles matter and field together, with $V$ carrying the binding interactions.

The equation for A is a clean wave equation. Why do we not just solve it with retarded integrals like the Lorenz-gauge potentials?

You can — with one catch: the source is $\mathbf{J}_\perp$ , not $\mathbf{J}$ . The retarded solution of $\Box^2\mathbf{A}=-\mu_0\mathbf{J}_\perp$ is the retarded integral over the transverse current, which (previous FAQ) fills all space even for a compact physical current. So the "retarded" Coulomb-gauge $\mathbf{A}$ integrates over an unbounded effective source, which is computationally awkward and conceptually the very origin of its instantaneous content. The Lorenz gauge's selling point, delivered over the next articles, is exactly that its retarded integrals run over the physical, compact $\rho$ and $\mathbf{J}$ . Rule of thumb: Coulomb gauge for Hamiltonians and mode counting; Lorenz gauge for propagation and radiation integrals.