Scalar and Vector Potentials

The wave chapter just closed having computed everything about electromagnetic waves except where they come from. Radiation — the subject this new chapter opens — means solving Maxwell's equations with the sources kept, and the chapter opener of E8 already confessed the field equations' weakness there: six coupled components, drives built from awkward derivatives of $\rho$ and $\mathbf{J}$ , constraints to police at every step. It also promised the fix by name: the potential formulation would get its own opening. This is it. The idea is a trade this track has made twice before, now executed with time dependence switched on. In electrostatics, $V$ turned three constrained field components into one function; in magnetostatics, $\mathbf{A}$ did the same for $\mathbf{B}$ . Electrodynamics threatens to wreck both — Faraday's law makes $\mathbf{E}$ non-conservative, so $\mathbf{E} = -\nabla V$ dies — but the wreck is precise and repairable: $\nabla\cdot\mathbf{B} = 0$ survives untouched (so $\mathbf{B} = \nabla\times\mathbf{A}$ stands), and feeding that into Faraday's law shows that not $\mathbf{E}$ but the combination $\mathbf{E} + \partial_t\mathbf{A}$ is curl-free. One new equation, $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ , and the two source-free Maxwell equations become identities: four potential components now do the work of six field components, with the two dynamical Maxwell equations turning into two second-order equations for $(V, \mathbf{A})$ . Those two equations arrive coupled and ugly — and hiding inside them, exhibited here in a box, sits one combination, $\nabla\cdot\mathbf{A} + \mu_0\varepsilon_0\,\partial_t V$ , begging to be set to zero. Doing so, and understanding why one is allowed to, is the gauge story of the next three articles. This article builds the objects, proves they always exist, prices their redundancy honestly, and shows a plane wave wearing potentials — with the Aharonov–Bohm effect waiting at the end to argue that the potentials were never mere bookkeeping.

Key takeaways

The two source-free Maxwell equations are integrability conditions: $\nabla\cdot\mathbf{B} = 0$ grants $\mathbf{B} = \nabla\times\mathbf{A}$ (the E0 vector-potentials theorem, recalled), and Faraday's law then makes $\mathbf{E} + \partial_t\mathbf{A}$ curl-free, granting $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ (the E0 scalar-potentials theorem, applied to the right object).
Statics is the $\partial_t \to 0$ audit: $V$ collapses to the electrostatic potential (E1), $\mathbf{A}$ to the magnetostatic one (E4) — the twins own those stories; this article owns their marriage under time dependence.
The two dynamical Maxwell equations become two coupled second-order potential equations; written suggestively, each contains the d'Alembertian $\Box$ plus one obstructing term built from the single combination $\nabla\cdot\mathbf{A} + \mu_0\varepsilon_0\,\partial_t V$ .
The potentials are redundant: $\mathbf{A} \to \mathbf{A} + \nabla\lambda$ , $V \to V - \partial_t\lambda$ leaves both fields untouched for any $\lambda(\mathbf{r},t)$ — time-dependent gauge freedom, generalizing the statics versions (E0/E4, recalled). Spending it is the next three articles' business (Coulomb and Lorenz gauges, deferred by name).
A plane wave in potential clothing: $V = 0$ , $\mathbf{A} = (E_0/\omega)\sin(kz - \omega t)\,\hat{\mathbf{x}}$ reproduces the E8 triad exactly — and a 1 mW laser's $A_0 \sim 3\times10^{-13}$ T·m sets the scale of the canonical momentum $q\mathbf{A}$ that quantum mechanics runs on.
Aharonov–Bohm: electrons circling a confined flux pick up the gauge-invariant phase $q\Phi_B/\hbar$ where $\mathbf{B} = 0$ on their entire path — potentials carry physical content the local fields do not.
KEY IDEA: the homogeneous half of Maxwell is solved identically by $\mathbf{B} = \nabla\times\mathbf{A}$ , $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ ; the inhomogeneous half becomes two coupled equations whose obstruction is one gauge-adjustable combination — the doorway to gauge choices, retarded potentials, and all of radiation theory.

The core insight

The two source-free Maxwell equations are solved identically by $\mathbf{B} = \nabla\times\mathbf{A}$ and $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ ; the potentials $(V, \mathbf{A})$ are determined only up to the gauge transformation $\mathbf{A} \to \mathbf{A} + \nabla\lambda$ , $V \to V - \partial_t\lambda$ , and the remaining two Maxwell equations become coupled second-order equations whose obstruction is the single combination $\nabla\cdot\mathbf{A} + \mu_0\varepsilon_0\,\partial_t V$ .

Where this lives today

The potential formulation is the working language of every field this chapter feeds. Radiation engineering (antennas to synchrotrons) computes from the gauged potential equations — the retarded- $\mathbf{A}$ pipeline this chapter is about to build; numerical electromagnetics splits between field solvers (FDTD, the E8 opener's territory) and potential-based formulations prized where geometry is multiply connected or coupling to quantum matter is the point. In quantum technology the potentials are simply primary: superconducting circuits are engineered in terms of flux $\oint\mathbf{A}\cdot d\boldsymbol{\ell}$ through loops (the flux qubit's variable), Josephson physics runs on gauge-invariant phase differences, and Aharonov–Bohm interferometry has graduated into a tool — solid-state AB rings measure decoherence, and AB phases from engineered gauge fields let neutral cold atoms impersonate charged particles in magnetic fields. The deepest modern payload is the one Derivation B's consolidation hinted at: promoting the potentials to the fundamental variables and demanding local gauge invariance generates interactions. The electromagnetic coupling itself, and its non-abelian siblings governing the weak and strong forces, follow from exactly the redundancy pattern (2) — the "defect" of the classical potentials turned out to be the organizing principle of the Standard Model. When the E10 chapter assembles $(V/c, \mathbf{A})$ into the four-potential and the E11 frontier gestures at field theory, they are cashing the check this article writes: the objects introduced here as calculational conveniences are, by current best understanding, closer to nature's own bookkeeping than the fields they generate.

Origins · 1 minGustav Kirchhoff and how this was discovered

Gustav Kirchhoff

1824–1887 · Königsberg, Heidelberg, Berlin

Kirchhoff wrote this article's central equation first: his 1857 analysis of signals in telegraph wires expressed the electric field as $-\nabla V - \partial_t\mathbf{A}$ — the earliest appearance of the full electrodynamic potential decomposition — and noticed, in the same work, that the signal velocity emerging from his equations was suspiciously close to the measured speed of light, four years before Maxwell made that coincidence the center of physics (Kirchhoff, cautious to a fault, declined to speculate). It was characteristic: he had already, as a 21-year-old student, given electrical engineering its permanent foundation (the circuit laws bearing his name), and would go on to co-found spectroscopy with Bunsen — identifying cesium and rubidium, explaining the Fraunhofer lines as solar absorption, and stating the radiation law (emissivity equals absorptivity, function of temperature alone) whose pursuit led Planck, his successor in Berlin, to the quantum. His blackbody concept, his diffraction theory (the rigorous basis of the optics this chapter's Fresnel articles touched), his thermochemistry law — Kirchhoff's fingerprints are on an implausible fraction of physics' load-bearing walls, applied with a rigor so austere that Boltzmann eulogized his lectures as models of clarity from which all warmth had been carefully removed. He worked much of his life on crutches after an accident, declining sympathy as firmly as speculation.

History

1845

— Franz Neumann builds induction theory around a vector function of currents: the vector potential enters physics

1857

— Kirchhoff writes E as minus grad V minus dA/dt in telegraph theory and notes the light-speed coincidence

1865

— Maxwell makes A ("electromagnetic momentum") central to the dynamical theory of the field

1867

— Ludvig Lorenz publishes retarded potentials and the condition later misattributed to Lorentz

1880

s — Heaviside and Hertz campaign to "murder the potentials," recasting Maxwell in fields alone

1926

— Schrödinger's equation couples matter to light through q A in the canonical momentum: the potentials return as primary

1959

— Aharonov and Bohm predict a measurable phase from potentials in field-free regions

1986

— Tonomura's shielded-toroid experiment confirms the AB phase beyond objection

The physical picture

Maxwell's four equations are not four of a kind. Two of them — $\nabla\cdot\mathbf{B} = 0$ and Faraday's law — mention no charge and no current; they are structural statements, true for every electromagnetic field in every circumstance, sources or not. The other two carry $\rho$ and $\mathbf{J}$ and do the dynamics. The potential formulation is the observation that the structural pair can be solved once and for all: they are precisely the integrability conditions under which fields can be written as derivatives of simpler objects. Solve them identically, and only the dynamical pair remains to fight — with four unknowns instead of six, and no constraints left to police. This is why the E8 chapter's completion routine (check the divergences, build the partner field) has no analog here: a field assembled from potentials cannot violate the homogeneous equations, no matter how carelessly the potentials are chosen. Half of Maxwell becomes automatic. The physical reading of the new equation deserves a slow look. $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ says the electric field has two parents. The first term is the old electrostatic story: charges pile up, potential hills form, field lines run downhill. The second is Faraday's law wearing new clothes: a time-varying vector potential — think of $\mathbf{A}$ as the "flux field," since its loop integral is the magnetic flux (the E0 twins' central identity, recalled) — pushes charges around closed loops, the induced electric field of every generator and transformer in the E6 chapter. What gauge freedom repackages is exactly the split between the parents: $\mathbf{A} \to \mathbf{A} + \nabla\lambda$ , $V \to V - \partial_t\lambda$ moves contributions from one ledger column to the other while every measurable field — and every loop integral $\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \Phi_B$ — holds still. The freedom is larger than statics': $\lambda$ may now depend on time, so even the "zero of potential" can be re-chosen instant by instant, and $V$ can be gauged away entirely (the plane-wave construction below does exactly that). Why bother, beyond economy? The E8 opener said it and this chapter will spend it: the potential equations, once gauged, take the form $\Box(\text{potential}) = -(\text{source, undifferentiated})$ — the cleanest inhomogeneous wave equations in the theory, with $\rho$ and $\mathbf{J}$ entering raw rather than as gradients and time-derivatives. Green's-function technology applies directly, causality can be built in by hand (the retarded potentials, two articles ahead by name), and every radiation formula of this chapter — dipole fields, Larmor's power, the Liénard–Wiechert potentials of a moving charge — falls out of that pipeline. The potentials are the price of admission to radiation theory; the gauge articles ahead are the ticket booth.

Figure

The architecture of the trade — two Maxwell equations solved identically by potentials, two turned into dynamics, with one gauge-adjustable obstruction in the way

The construction and the coupled equations

The objects first (Derivation A), then what the dynamics does to them (Derivation B). The identical solution of the homogeneous pair:

\mathbf{B} = \nabla\times\mathbf{A}, \qquad \mathbf{E} = -\nabla V - \frac{\partial\mathbf{A}}{\partial t}, \qquad (1)

where: the first equation is unchanged from magnetostatics — $\nabla\cdot\mathbf{B} = 0$ holds in electrodynamics exactly as before, so the E0 existence theorem (Vector Potentials, recalled by name) applies verbatim — while the second is this article's one genuinely new object: the $-\partial_t\mathbf{A}$ term is the induced electric field of Faraday's law, and dropping it (the classic blunder) silently reasserts electrostatics. Audit the statics limit: $\partial_t \to 0$ collapses (1) to $\mathbf{E} = -\nabla V$ and $\mathbf{B} = \nabla\times\mathbf{A}$ — the E1 and E4 stories, owned by those chapters and their E0 twins, recovered exactly. The redundancy, stated precisely:

\mathbf{A} \to \mathbf{A} + \nabla\lambda, \qquad V \to V - \frac{\partial\lambda}{\partial t} \qquad\Longrightarrow\qquad \mathbf{E}, \mathbf{B}\ \text{unchanged}, \qquad (2)

where: $\lambda(\mathbf{r}, t)$ is any smooth scalar function of space and time. The check is two lines — the curl kills $\nabla\lambda$ in $\mathbf{B}$ ; in $\mathbf{E}$ the two new pieces are $-\nabla(-\partial_t\lambda) - \partial_t(\nabla\lambda) = 0$ by equality of mixed partials — but note the pairing: the shifts of $\mathbf{A}$ and $V$ must be made together, cross-wired through $\lambda$ , which is why time-dependent gauge freedom is richer than the statics versions (a constant offset of $V$ , a gradient added to static $\mathbf{A}$ — both special cases, recalled from the E0/E4 articles). That this is the entire freedom (any two potential pairs giving the same fields differ by exactly a $\lambda$ ) is proved in the next article, Gauge Transformations, deferred by name. Feeding (1) into the two sourced Maxwell equations (Derivation B) yields the dynamics:

\nabla^2 V + \frac{\partial}{\partial t}\left(\nabla\cdot\mathbf{A}\right) = -\frac{\rho}{\varepsilon_0}, \qquad (3)

\left(\nabla^2\mathbf{A} - \mu_0\varepsilon_0\frac{\partial^2\mathbf{A}}{\partial t^2}\right) - \nabla\left(\nabla\cdot\mathbf{A} + \mu_0\varepsilon_0\frac{\partial V}{\partial t}\right) = -\mu_0\mathbf{J}, \qquad (4)

where: all of electrodynamics is now these two equations — Griffiths calls the pair "ugly," and the ugliness is informative. Equation (4) is one gradient away from beauty: its first bracket is exactly $\Box\mathbf{A}$ (the E8 chapter's d'Alembertian, in the track's sign convention), and the spoiler is the gradient of the single combination

\mathcal{L} \equiv \nabla\cdot\mathbf{A} + \mu_0\varepsilon_0\frac{\partial V}{\partial t}, \qquad (5)

where: the same $\mathcal{L}$ obstructs (3) as well — add and subtract $\mu_0\varepsilon_0\,\partial_t^2 V$ in (3) and it reads $\Box V + \partial_t\mathcal{L} = -\rho/\varepsilon_0$ . One scalar function blocks both equations from being clean wave equations. And (2) is precisely enough freedom to dictate $\mathcal{L}$ : under a gauge transformation, $\mathcal{L} \to \mathcal{L} + \Box\lambda$ , so choosing $\lambda$ means choosing $\mathcal{L}$ . Set $\mathcal{L} = 0$ and both equations decouple into $\Box V = -\rho/\varepsilon_0$ , $\Box\mathbf{A} = -\mu_0\mathbf{J}$ — the Lorenz gauge, stated here as the visible punchline and derived, justified, and spent two articles ahead (by name; the Coulomb gauge $\nabla\cdot\mathbf{A} = 0$ , which instead cleans (3) at (4)'s expense, gets the article between). This division is deliberate: the present article's job is to make the obstruction visible; the gauge articles own the surgery.

Figure

One electric field, two bookkeepers — gauge freedom repackages E between its gradient parent and its induced parent while every measurable holds still

Derivation AThe potentials exist — solving half of Maxwell identically

The goal is (1): prove that every electromagnetic field, however wild its time dependence, can be written in terms of a scalar and a vector potential — and that this construction uses up exactly the two source-free Maxwell equations. The strategy is to harvest the two E0 existence theorems (recalled by name, not re-proved) in the right order: divergence-free first, then a curl-free combination that Faraday's law hands us. Watch which object each theorem is applied to — the entire trick of the derivation is applying the scalar-potential theorem not to $\mathbf{E}$ but to a repaired version of it.

Step 1. The magnetic potential survives electrodynamics untouched. $\nabla\cdot\mathbf{B} = 0$ holds always (no monopoles — the one Maxwell equation nothing in this track has ever modified), and the E0 theorem (Vector Potentials) states: a divergence-free field on all of space is a curl. So $\mathbf{B} = \nabla\times\mathbf{A}$ for some $\mathbf{A}(\mathbf{r}, t)$ , at every instant, with the time dependence riding along parametrically.

Step 2. Faraday's law repairs the electric field. In electrodynamics $\nabla\times\mathbf{E} = -\partial_t\mathbf{B} \neq 0$ : $\mathbf{E}$ itself is not curl-free and admits no scalar potential — the electrostatic construction is dead. But substitute Step 1 and move everything left:

\nabla\times\mathbf{E} = -\frac{\partial}{\partial t}\left(\nabla\times\mathbf{A}\right) \quad\Longrightarrow\quad \nabla\times\left(\mathbf{E} + \frac{\partial\mathbf{A}}{\partial t}\right) = 0.

Why the swap of derivatives is legal: space and time derivatives commute on smooth fields; the curl and $\partial_t$ pass through each other, letting Faraday's law be read as a statement that a particular combination is irrotational.

Step 3. Harvest the scalar theorem. The combination $\mathbf{E} + \partial_t\mathbf{A}$ is curl-free everywhere, so the E0 theorem (Scalar Potentials) grants a scalar function — call it $-V$ for continuity with statics — with

\mathbf{E} + \frac{\partial\mathbf{A}}{\partial t} = -\nabla V \quad\Longrightarrow\quad \mathbf{E} = -\nabla V - \frac{\partial\mathbf{A}}{\partial t}.

Step 4. Confirm the books balance. Two Maxwell equations were spent building (1); check they are now automatic: $\nabla\cdot\mathbf{B} = \nabla\cdot(\nabla\times\mathbf{A}) = 0$ identically (div of curl), and $\nabla\times\mathbf{E} = -\nabla\times\nabla V - \partial_t\nabla\times\mathbf{A} = -\partial_t\mathbf{B}$ identically (curl of gradient dies). Any $(V, \mathbf{A})$ whatsoever — smooth but otherwise arbitrary — generates fields satisfying the homogeneous pair. This is the promised inversion of the E8 completion routine: there, candidate fields had to be policed against the constraints; here, the constraints are woven into the parametrization and cannot be violated. Why this is a real theorem and not notation: the count matters. Six field components minus two scalar constraint equations leaves four honest degrees of freedom — and four potential components (one $V$ , three $\mathbf{A}$ ) is exactly that count, with the gauge redundancy (2) accounting for the fact that the constraints were differential rather than algebraic. The deeper accounting (four potentials, minus gauge, = two physical polarizations) belongs to the gauge articles.

Step 5. Statics audit and inheritance. $\partial_t = 0$ : Step 3 collapses to $\mathbf{E} = -\nabla V$ (E1's object), Step 1 to magnetostatics' $\mathbf{A}$ (E4's object, with its Coulomb-gauge machinery recalled from Gauge Freedom and the Coulomb Gauge). Everything those chapters built — Poisson solutions, boundary-value technology, the flux reading $\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \Phi_B$ — is inherited unchanged as the low-frequency limit of this construction. Verified: reproduces Griffiths §10.1.1 (Eqs. 10.2–10.4) and Jackson §6.2; the E0 theorems applied are exactly those of the landed Scalar Potentials and Vector Potentials articles (simply-connected all-space case); identity audit (div-curl, curl-grad) per the E0 second-derivatives article; statics limits match E1/E4.

Go deeper · Derivation B · 2 minThe coupled dynamics — and the one combination that obstructs both equations

Derivation BThe coupled dynamics — and the one combination that obstructs both equations

The goal is (3)–(5): substitute the potentials into the two sourced Maxwell equations, organize the wreckage honestly, and exhibit — without yet spending — the single gauge-adjustable combination that stands between the coupled mess and two clean wave equations. The strategy is direct substitution plus one vector identity, with the bookkeeping arranged to make the d'Alembertian structure impossible to miss.

Step 1. Gauss's law. Substitute $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ into $\nabla\cdot\mathbf{E} = \rho/\varepsilon_0$ :

\nabla^2 V + \frac{\partial}{\partial t}(\nabla\cdot\mathbf{A}) = -\frac{\rho}{\varepsilon_0},

which is (3). Why it is not yet Poisson's equation: the new term couples $V$ to $\mathbf{A}$ 's divergence — the two potentials cannot be solved independently, and $V$ 's equation is first-order in time through its partner. Statics ( $\partial_t \to 0$ ) recovers Poisson exactly.

Step 2. Ampère–Maxwell. Substitute both parts of (1) into $\nabla\times\mathbf{B} = \mu_0\mathbf{J} + \mu_0\varepsilon_0\,\partial_t\mathbf{E}$ ; the left side needs the E0 double-curl identity $\nabla\times(\nabla\times\mathbf{A}) = \nabla(\nabla\cdot\mathbf{A}) - \nabla^2\mathbf{A}$ (recalled by name — owned by the del-operator article):

\nabla(\nabla\cdot\mathbf{A}) - \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}.

Step 3. Organize for visibility. Collect the wave-operator pieces on the left and force the remaining terms into a single gradient:

\left(\nabla^2\mathbf{A} - \mu_0\varepsilon_0\frac{\partial^2\mathbf{A}}{\partial t^2}\right) - \nabla\left(\nabla\cdot\mathbf{A} + \mu_0\varepsilon_0\frac{\partial V}{\partial t}\right) = -\mu_0\mathbf{J},

which is (4). Why this grouping and not another: the first bracket is $\Box\mathbf{A}$ — the operator every article of E8 ran on — so the equation reads "wave equation for $\mathbf{A}$ , sourced by the raw current, spoiled by one gradient." The spoiler's argument is the combination $\mathcal{L}$ of (5). The same $\mathcal{L}$ hides in Step 1's equation: adding and subtracting $\mu_0\varepsilon_0\,\partial_t^2V$ rewrites (3) as $\Box V + \partial_t\mathcal{L} = -\rho/\varepsilon_0$ . One scalar function obstructs both equations — a remarkable consolidation, and no accident: relativistically, $(V, \mathbf{A})$ assemble into a four-vector and $\mathcal{L}$ is its four-divergence, a single invariant scalar. The E10 chapter cashes that reading; here it explains the coincidence.

Step 4. Measure the freedom against the obstruction. Under (2), compute how $\mathcal{L}$ moves: $\nabla\cdot\mathbf{A} \to \nabla\cdot\mathbf{A} + \nabla^2\lambda$ and $\mu_0\varepsilon_0\,\partial_tV \to \mu_0\varepsilon_0\,\partial_tV - \mu_0\varepsilon_0\,\partial_t^2\lambda$ , so

\mathcal{L} \to \mathcal{L} + \Box\lambda.

Since $\Box\lambda = f$ is solvable for any reasonable $f$ (it is the E8 chapter's inhomogeneous wave equation — solutions exist; the retarded construction two articles ahead even builds one explicitly), the gauge freedom is exactly strong enough to assign $\mathcal{L}$ any value one likes. Set it to zero: both equations decouple into $\Box V = -\rho/\varepsilon_0$ and $\Box\mathbf{A} = -\mu_0\mathbf{J}$ — the Lorenz-gauge punchline, stated here so the destination is visible, with the derivation, the residual-freedom analysis, and the covariance story owned by Lorenz Gauge (and the alternative $\nabla\cdot\mathbf{A} = 0$ surgery by Coulomb Gauge), both deferred by name. Chapter architecture in one sentence: this article built the objects and found the knob; the next three turn it. Verified: (3)–(4) reproduce Griffiths Eqs. 10.4–10.5 and Jackson §6.2's coupled pair; the double-curl identity per the E0 article; the $\mathcal{L} \to \mathcal{L} + \Box\lambda$ shift matches Griffiths §10.1.3's gauge accounting; statics limits — (3) → Poisson, (4) → the E4 vector-Poisson equation in Coulomb gauge — check against E1/E4.

Go deeper · Derivation C · 2 minA plane wave in potential clothing — and the scale of A in a laser beam

Derivation CA plane wave in potential clothing — and the scale of A in a laser beam

The goal is a working demonstration that the potentials carry all of electrodynamics, radiation included: dress the E8 chapter's plane wave in $(V, \mathbf{A})$ , verify it reproduces the full field triad, read off which gauges the construction happens to satisfy, and put numbers on $\mathbf{A}$ for a real laser — the quantity that quantum mechanics' canonical momentum $q\mathbf{A}$ runs on. The strategy: guess the simplest potentials imaginable and let (1) do the work — an exercise in how much simpler fields-from-potentials is than fields-with-constraints.

Step 1. The ansatz. Try zero scalar potential and a single-component vector potential oscillating with the wave:

V = 0, \qquad \mathbf{A} = \frac{E_0}{\omega}\,\sin(kz - \omega t)\,\hat{\mathbf{x}}, \qquad \omega = ck.

Step 2. Generate the fields from (1). Electric: $\mathbf{E} = -\nabla V - \partial_t\mathbf{A} = E_0\cos(kz - \omega t)\,\hat{\mathbf{x}}$ ✓ — the E8 plane wave, amplitude and phase exact. Magnetic: $\mathbf{B} = \nabla\times\mathbf{A} = \partial_z A_x\,\hat{\mathbf{y}} = \frac{E_0 k}{\omega}\cos(kz - \omega t)\,\hat{\mathbf{y}} = \frac{E_0}{c}\cos(kz - \omega t)\,\hat{\mathbf{y}}$ ✓ — transverse partner, correct ratio $E/B = c$ , correct direction ( $\mathbf{E}\times\mathbf{B}$ forward). The entire completion-routine checklist of the E8 opener — transversality, the $E = cB$ lock, the phase lock — emerged automatically, because Derivation A made the homogeneous equations unbreakable. Why no constraint-checking was needed: that is the formulation's entire selling point, displayed live.

Step 3. Audit the gauges this ansatz satisfies. $\nabla\cdot\mathbf{A} = \partial_x A_x = 0$ (the wave depends only on $z$ ): Coulomb gauge ✓. With $V = 0$ as well, $\mathcal{L} = 0$ : Lorenz gauge ✓ simultaneously. Why both at once is possible here but not generally: with $\rho = 0$ (no sources) the two conditions are compatible; the moment charges enter, $V$ cannot be globally zeroed and the gauges part company — the dividing line the two gauge articles will draw carefully. The choice $V = 0$ itself (the "temporal gauge") is popular in quantum optics for exactly the reason on display: one vector function carries the whole radiation field.

Step 4. Numbers — the scale of $\mathbf{A}$ . A 1 mW HeNe laser beam focused to 1 mm² carries intensity $I = 10^3$ W/m², hence $E_0 = \sqrt{2I/\varepsilon_0 c} = 868$ V/m (the E8 intensity machinery, recalled), and at $\omega = 2.98\times10^{15}$ rad/s:

A_0 = \frac{E_0}{\omega} = 2.9\times10^{-13}\ \text{T}\,\cdot\,\text{m}.

Tiny — but watch it in the canonical momentum $\mathbf{p} = m\mathbf{v} + q\mathbf{A}$ (the charged-particle Lagrangian's momentum, and the operator quantum mechanics quantizes): for an electron, $qA_0 = 4.7\times10^{-32}$ kg·m/s. Compare the photon momentum $\hbar k = 1.0\times10^{-27}$ kg·m/s: the classical field's momentum handle on a single electron per cycle is five orders below a single photon kick — the honest scale of "weak-field" light–matter coupling, and the small parameter of perturbative quantum optics. Crank the same formula at a petawatt laser's $E_0 \sim 10^{14}$ V/m and $qA_0$ exceeds $m_ec$ — the relativistic-optics threshold, defined in practice by exactly this article's $A$ . Verified: fields from Step 2 match the E8 plane-wave triad (wave-equation article's completion-routine output) identically; gauge audits by direct differentiation; laser numbers recomputed — $E_0 = \sqrt{2\times10^3/(8.854\times10^{-12}\times2.998\times10^8)} = 868$ V/m, $A_0 = E_0/\omega = 2.9\times10^{-13}$ T·m, $eA_0 = 4.7\times10^{-32}$ kg·m/s vs $\hbar k = 1.05\times10^{-27}$ kg·m/s; the $a_0 = eA_0/m_ec \sim 1$ relativistic threshold is the standard laser-physics parameter.

Limiting cases

Case	Result	Meaning
$\partial_t \to 0$ , charges only	$\mathbf{E} = -\nabla V$ , $\nabla^2V = -\rho/\varepsilon_0$	E1 electrostatics recovered; twins own the story
$\partial_t \to 0$ , currents only	$\mathbf{B} = \nabla\times\mathbf{A}$ , $\nabla^2\mathbf{A} = -\mu_0\mathbf{J}$	E4 magnetostatics (Coulomb gauge) recovered
forget $-\partial_t\mathbf{A}$	wrong $\mathbf{E}$ in any dynamic problem	the classic blunder — warning box
$\lambda$ time-independent	statics gauge freedom	E0/E4 special case of (2)
$\mathcal{L} = 0$ imposed	$\Box V = -\rho/\varepsilon_0$ , $\Box\mathbf{A} = -\mu_0\mathbf{J}$	Lorenz gauge: stated here, owned two articles on
$\nabla\cdot\mathbf{A} = 0$ imposed	$V$ instantaneous-Poisson	Coulomb gauge: next-next article
source-free, $V = 0$	$\mathbf{A}$ alone carries radiation	Derivation C's plane wave; temporal gauge
loop integral	$\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \Phi_B$ , gauge-invariant	the E0 flux identity survives dynamics intact

Common trapIn dynamics, E is never minus grad V — and no particular potential is physically real

Two errors, one per potential. First, the amputation: carrying $\mathbf{E} = -\nabla V$ into any time-dependent problem quietly deletes Faraday's law — the $-\partial_t\mathbf{A}$ term is the induced electric field, and without it every generator, transformer, and radiated wave in the theory vanishes (a plane wave in the $V = 0$ gauge is one hundred percent $-\partial_t\mathbf{A}$ ; amputate it and light goes dark). The tell that you need the full (1): any $\mathbf{B}$ changing anywhere. Second, the reification: no individual $(V, \mathbf{A})$ is measurable — the same fields wear infinitely many potential outfits, connected by (2), and any quantity that changes under a gauge transformation is bookkeeping, not physics. Classical measurables are the fields and gauge-invariant integrals like $\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \Phi_B$ ; statements like "the scalar potential propagates instantaneously" (true in Coulomb gauge!) describe the outfit, not the body, and panic about them dissolves in the gauge articles ahead. Quantum mechanics sharpens rather than breaks this rule — the Aharonov–Bohm phase is built from exactly the gauge-invariant loop integral, which is why it is physics and not artifact. Between the two errors lies the discipline: compute with whatever gauge is convenient, trust only what survives (2).

Worked examples

1 · The two-parent electric field of a ramping solenoid (confidence builder)

A long solenoid (radius $R = 2$ cm, $n = 10^4$ turns/m) carries current ramping at $dI/dt = 100$ A/s. Using potentials, find the electric field at radius $s = 5$ cm outside the solenoid, where $\mathbf{B} = 0$ at all times.

Step 1. The vector potential outside, from the flux identity (E0/E4, recalled): $\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \Phi_B$ gives $2\pi s\,A_\phi = \pi R^2 B_{\rm in}$ with $B_{\rm in} = \mu_0 n I$ , so $A_\phi = \mu_0 n I R^2/2s$ — circulating, nonzero where the magnetic field is exactly zero, its loop integral storing the enclosed flux.

Step 2. No charges anywhere: $V = 0$ throughout. The electric field has one parent only:

E_\phi = -\frac{\partial A_\phi}{\partial t} = -\frac{\mu_0 n R^2}{2s}\frac{dI}{dt}.

Step 3. Numbers: $E_\phi = -\frac{4\pi\times10^{-7}\times10^4\times(0.02)^2}{2\times0.05}\times100 = \mathbf{-5.0\times10^{-3}\ V/m}$ — five millivolts per meter of induced field, circling the solenoid, in a region where $\mathbf{B}$ never appears. (Sanity check against Faraday's integral law: $\oint\mathbf{E}\cdot d\boldsymbol{\ell} = -d\Phi/dt = -\pi R^2\mu_0 n\,dI/dt = -1.58$ mV; divided by $2\pi s = 0.314$ m gives $-5.0$ mV/m ✓.)

Step 4. The reading: the potential formulation localizes the bookkeeping — $E$ at the point follows from $A$ at the point — where the field-only description must reach across space to the flux through a distant loop. Classically the two are equivalent accounts. But hold the picture of this example: a charge outside the solenoid responding to $\mathbf{A}$ where $\mathbf{B} = 0$ is the classical antechamber of example 2, where quantum mechanics makes the antechamber the whole house.

2 · Aharonov–Bohm: how much flux moves the fringes one period? (exam level)

In an electron interferometer, the two paths enclose a micrometer-scale solenoid (confined flux $\Phi_B$ ; $\mathbf{B} = 0$ on both paths). (a) State the phase difference. (b) Find the flux that shifts the interference pattern by one full fringe, and the corresponding solenoid field for a 0.5 μm-radius solenoid. (c) Why does no choice of gauge make the effect disappear?

Step 1. The phase. Each electron's wavefunction accumulates $\varphi = (q/\hbar)\int\mathbf{A}\cdot d\boldsymbol{\ell}$ along its path (the canonical-momentum coupling $\mathbf{p} - q\mathbf{A}$ of quantum mechanics); the difference between the two arms closes a loop:

\Delta\varphi = \frac{q}{\hbar}\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \frac{q\,\Phi_B}{\hbar},

by the flux identity — gauge-invariant, because loop integrals of $\mathbf{A}$ are (the warning box's discipline paying off).

Step 2. One fringe = $\Delta\varphi = 2\pi$ : $\Phi_B = 2\pi\hbar/e = h/e = \mathbf{4.14\times10^{-15}\ Wb}$ — the (single-charge) flux quantum, a universal constant containing no property of the electron beam, the geometry, or the material. For a solenoid of radius 0.5 μm (area $7.9\times10^{-13}$ m²): $B = \Phi_B/\pi r^2 = \mathbf{5.3\ mT}$ — a modest laboratory field in a micron-scale core marches the fringes by a full period, which is exactly how Tonomura's 1986 experiment (toroidal magnets, superconducting shielding to guarantee zero field on the paths) settled the question.

Step 3. Why gauge cannot kill it: to remove $\mathbf{A}$ along both arms one needs $\mathbf{A} = \nabla\lambda$ in the whole doubly-connected region around the solenoid — possible locally, impossible globally when $\oint\mathbf{A}\cdot d\boldsymbol{\ell} = \Phi_B \neq 0$ (a gradient's loop integral vanishes; the region's hole is load-bearing — the same simply-connected fine print the E0 scalar-potentials article flagged). The physical content is the gauge-invariant flux, and the lesson is permanent: in quantum mechanics the potentials couple locally to matter and produce effects the local fields cannot account for — while everything measurable remains gauge-invariant. Fields underdescribe; potentials overdescribe; the loop integral is exactly right.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §10.1 · Jackson, Classical Electrodynamics, 3rd ed., §6.2–6.3 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 6 · Zangwill, Modern Electrodynamics, §15.1–15.3

Provenance. All equations are standard and can be verified against Griffiths 4th ed. §10.1 (Eqs. 10.2–10.16), Jackson 3rd ed. §6.2–6.3, and Zangwill §15.1–15.3. A verifier should recompute: (1) the curl-of-curl reduction giving the coupled potential equations; (2) the Lorenz-gauge decoupling into $\Box V=\rho/\epsilon_0$ , $\Box\mathbf{A}=\mu_0\mathbf{J}$ ; (3) the curls of the symmetric and Landau-gauge vector potentials both yielding $B\hat{\mathbf{z}}$ and their gauge function $\lambda=\tfrac12 Bxy$ ; (4) the Aharonov–Bohm phase $q\Phi_B/\hbar$ . No empirical constants beyond $c=1/\sqrt{\mu_0\epsilon_0}$ are asserted. Addendum for this edition: scope split executed — the E0 twins (Scalar Potentials, Vector Potentials) own the existence theorems and are recalled by name; the magnetostatic $\mathbf{A}$ machinery (including the uniform-field symmetric/Landau construction of the base's third derivation) is owned by the E4 articles and recalled, not re-derived; the Lorenz-gauge decoupling is stated as the visible punchline with its derivation deferred by name to the Lorenz Gauge article (base's second derivation rescoped accordingly), and the Coulomb-gauge surgery to its own article. This article owns: the electrodynamic construction (Derivation A), the coupled equations with the obstruction $\mathcal{L}$ exhibited and its gauge shift $\mathcal{L} \to \mathcal{L} + \Box\lambda$ computed (Derivation B), and the plane-wave-in-potentials construction with laser scales (Derivation C). Worked numbers recomputed: ramping solenoid $E_\phi = 5.0$ mV/m at $s = 5$ cm (cross-checked via Faraday's integral law); flux quantum $h/e = 4.14\times10^{-15}$ Wb, 5.3 mT in a 0.5 μm solenoid per fringe; laser scales $E_0 = 868$ V/m, $A_0 = 2.9\times10^{-13}$ T·m, $eA_0 = 4.7\times10^{-32}$ kg·m/s vs $\hbar k = 1.0\times10^{-27}$ kg·m/s.

Builds on: Maxwell's Equations in Vacuum · Vector Calculus Identities · Faraday's Law of Induction · Helmholtz Theorem

Leads to: Gauge Transformations · Coulomb Gauge · Lorenz Gauge · Inhomogeneous Wave Equations for Potentials · Retarded Potentials · Liénard–Wiechert Potentials

How does this article relate to the four potential articles the track already has?

By division of era, not duplication. The two E0 articles (Scalar Potentials, Vector Potentials) are mathematics: existence theorems — curl-free fields have scalar potentials, divergence-free fields have vector potentials — proved as vector calculus with the gauge-as-gradient observation attached. The two E4 articles (Magnetic Vector Potential, Gauge Freedom and the Coulomb Gauge) apply the second theorem to magnetostatics: constructing $\mathbf{A}$ from steady currents, spending the static gauge freedom. This article is the electrodynamic installment: time dependence breaks $\mathbf{E} = -\nabla V$ , Faraday's law repairs it into $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ , the gauge freedom becomes time-dependent and cross-wired between the two potentials, and the sourced Maxwell equations become the coupled pair with the $\mathcal{L}$ obstruction. Every earlier result is recalled and inherited — the existence theorems in Derivation A, the flux identity in both worked examples — and everything after (gauges, retarded potentials, radiation) builds on this article's objects. One construction, told in three eras, each owning its own theorem.

If potentials are redundant and fields are not, why did physics keep the potentials at all — didn't Heaviside and Hertz try to eliminate them?

They did — "the potentials should be murdered" is (paraphrased) Heaviside's actual position, and the field-only formulation of Maxwell that this track teaches is their victory: for classical field bookkeeping, $(\mathbf{E}, \mathbf{B})$ suffice. The potentials survived on three counts that only grew. Practicality: radiation problems are dramatically easier in potentials (undifferentiated sources, Green's functions, four components with gauge freedom versus six with constraints) — the reason this chapter switches formulations at its door. Mechanics: the Lorentz force has no potential-energy function, but it does have a Lagrangian — $L = \tfrac{1}{2}mv^2 - qV + q\mathbf{v}\cdot\mathbf{A}$ — so the entire canonical apparatus (least action, Hamiltonians, conservation via symmetry) needs $(V, \mathbf{A})$ directly; there is no field-only Lagrangian for a charge. Quantum mechanics: the Schrödinger equation couples matter to electromagnetism through the canonical momentum $\mathbf{p} - q\mathbf{A}$ — the potentials, not the fields, appear in the fundamental equation — and Aharonov–Bohm made the consequence observable. The modern verdict inverts Heaviside: the potentials (as a gauge field) are the fundamental object; the fields are their gauge-invariant shadow.

Is the vector potential just a mathematical trick, or does it have a physical picture?

It has at least two honest pictures, both inherited from earlier articles and worth keeping. The flux picture (E0/E4): $\mathbf{A}$ 's loop integral equals the magnetic flux through the loop, making $\mathbf{A}$ a "flux density in transit" — the ramping-solenoid example runs on it, and superconducting-circuit engineers think in it daily. The momentum picture (Maxwell's own — he called $\mathbf{A}$ the electromagnetic momentum): $q\mathbf{A}$ is the field's contribution to a charge's canonical momentum, the electromagnetic IOU a charge holds — switch the field off and the IOU is paid as a mechanical impulse ( $m\Delta\mathbf{v} = q\mathbf{A}_{\rm before}$ , for a suddenly-killed field, is a nice one-line check of the Lagrangian coupling). Where caution enters: the pictures attach to gauge-invariant aspects (loop integrals, differences, sudden changes); the pointwise value of $\mathbf{A}$ remains outfit, not body. The E7 hidden-momentum article's careful distinctions between kinds of momentum are the same discipline applied one chapter earlier — and the quantum chapter of that story, where $q\mathbf{A}/\hbar$ is a phase gradient, is precisely Aharonov–Bohm.

The coupled equations look worse than the six field equations they replace — where is the promised economy?

Fair audit — and the answer has three installments. Immediately: the potentials never need constraint-checking (half of Maxwell is identically satisfied — the completion routine of E8 is abolished), and four unknowns replace six. One gauge choice later: the coupled pair decouples into $\Box V = -\rho/\varepsilon_0$ , $\Box\mathbf{A} = -\mu_0\mathbf{J}$ — compare the field wave equations of the E8 opener, whose sources were $\nabla\rho$ and $\partial_t\mathbf{J}$ tangles: the potential equations take their sources raw, which is the difference between textbook Green's-function integrals and derivative-juggling. And structurally: the "ugliness" of (3)–(4) is concentrated in one scalar combination $\mathcal{L}$ — a consolidation that is itself information (relativistically, $\mathcal{L}$ is the four-divergence of the four-potential; the mess was one Lorentz scalar all along). The honest ledger: the potential formulation costs one gauge decision and pays with the entire machinery of radiation theory. Every article left in this chapter is the dividend.

In Coulomb gauge the scalar potential responds instantaneously to distant charges — doesn't that violate relativity?

It violates nothing, and the puzzle is a wholesome vaccination against reifying potentials. In Coulomb gauge ( $\nabla\cdot\mathbf{A} = 0$ , next-next article), equation (3) becomes an instantaneous Poisson equation: wiggle a charge here and $V$ everywhere updates now. But $V$ is not measurable — the fields are, and the field $\mathbf{E} = -\nabla V - \partial_t\mathbf{A}$ assembles from both potentials, whose superluminal pieces cancel exactly: the $-\partial_t\mathbf{A}$ term in this gauge contains a compensating acausal-looking piece, and the physical fields propagate at exactly $c$ (guaranteed in advance, since the fields satisfy the manifestly causal field equations regardless of gauge). The general moral: causality constrains gauge-invariant quantities only; gauge-dependent bookkeeping may do anything it likes en route. The Lorenz gauge is popular precisely because it keeps even the bookkeeping causal (retarded potentials, two articles ahead), which is aesthetically soothing — but Coulomb gauge's "violation" costs nothing and its clean transverse $\mathbf{A}$ pays handsomely in quantum optics. Choose outfits by convenience; trust only the body.

Does the Aharonov–Bohm effect really prove the potentials are physical — could it not be reinterpreted as nonlocal action of the fields?

This is a genuine, live foundational question with a sharp consensus core. The unarguable part: the AB phase is measured (Tonomura's shielded toroids closed the loopholes — flux fully confined, paths field-free, fringes shifted by exactly $q\Phi/\hbar$ ), and no local account in terms of fields at the electron's location exists — the local fields there are zero. The interpretive fork: either (a) potentials act locally on quantum matter (the standard reading — the Schrödinger equation's $q\mathbf{A}$ coupling, with gauge invariance ensuring only $\oint\mathbf{A}\cdot d\boldsymbol{\ell}$ matters), or (b) fields act nonlocally — the electron responds to flux it never touches. Formally these are equivalent; the choice is which principle to sacrifice, locality or gauge-freedom-as-pure-redundancy. Modern field theory takes a refined third path: the fundamental object is the holonomy — the set of loop integrals $e^{iq\oint\mathbf{A}\cdot d\boldsymbol{\ell}/\hbar}$ — which is both local-enough (built from potentials along paths) and exactly gauge-invariant; potentials overdescribe it, fields underdescribe it. That "Wilson loop" language, born from this effect, is now the working vocabulary of lattice QCD and topological quantum matter — a considerable career for a phase shift of interference fringes.