The Four-Potential

At the end of the radiation chapter, electrodynamics stood in a curious pose. Lorenz Gauge had decoupled the potential equations — one wave equation for the scalar potential $V$ sourced by charge, three for the vector potential $\mathbf{A}$ sourced by current — and Inhomogeneous Wave Equations for Potentials solved them. Four equations, identical in form, differing only in which piece of the source sits on the right. When a theory produces four copies of the same equation, it is telling you that you have been counting one object as four. Relativity supplies the missing grammar. The previous article fused charge and current densities into the four-current $J^\mu=(c\rho,\mathbf{J})$ ; this one performs the twin fusion on the potentials. The quadruple $A^\mu=(V/c,\mathbf{A})$ is a four-vector — the scalar potential is the time shadow and the vector potential the space shadow of a single spacetime object — and the four wave equations collapse into one covariant statement, $\Box A^\mu=\mu_0 J^\mu$ , with the Lorenz condition revealed as the Lorentz-scalar equation $\partial_\mu A^\mu=0$ . A boost mixes $V$ and $\mathbf{A}$ into each other exactly as it mixes $ct$ and $x$ , which is why a humming current-carrying wire, viewed from a passing frame, sprouts a scalar potential and an electric field it never had at rest. The four-potential is also where a genuinely modern idea enters: gauge freedom — many $A^\mu$ , one physics — stated covariantly as $A^\mu\to A^\mu-\partial^\mu\chi$ . What looks like bookkeeping slack is the structural template for the Standard Model. One article from now, a single antisymmetric derivative of $A^\mu$ builds the field tensor and completes the chapter's destination.

Key takeaways

The four-potential $A^\mu=(V/c,\,\mathbf{A})$ packages the scalar and vector potentials into a single four-vector; the $1/c$ gives all four slots the same units (V·s/m), the precondition for mixing under boosts.
The radiation chapter's twin Lorenz-gauge results — $\Box^2V=-\rho/\varepsilon_0$ , $\Box^2\mathbf{A}=-\mu_0\mathbf{J}$ — assemble into one covariant equation: $\Box A^\mu=\mu_0 J^\mu$ , with $\Box=\partial_\mu\partial^\mu=-\Box^2$ .
The Lorenz condition is the scalar statement $\partial_\mu A^\mu=0$ : impose it in one inertial frame and every frame inherits it. The Coulomb condition $\nabla\cdot\mathbf{A}=0$ enjoys no such privilege — it must be re-chosen frame by frame.
Boosts mix the potentials like coordinates: $V'/c=\gamma(V/c-\beta A_x)$ , $A'_x=\gamma(A_x-\beta V/c)$ . A pure Coulomb potential boosts into the constant-velocity Liénard–Wiechert potentials — retardation algebra reproduced by geometry alone.
Gauge freedom is covariant: $A^\mu\to A^\mu-\partial^\mu\chi$ changes the potentials, never the fields; within the Lorenz family the residual freedom is $\Box\chi=0$ .
Honesty clause: " $A^\mu$ is a four-vector" is a property of the Lorenz-gauge family, not a gauge-invariant fact.
Key idea: bundle $(V/c,\mathbf{A})$ into one four-vector $A^\mu$ — then the potentials' wave equations, their frame mixing, and their gauge redundancy all become single covariant statements, and the entire electromagnetic field waits one derivative away.

The core insight

Bundle $(V/c,\mathbf{A})$ into one four-vector $A^\mu$ ; then the entire electromagnetic field, both homogeneous Maxwell equations, and the wave-equation dynamics all follow from a single covariant object plus its gauge freedom.

Where this lives today

Open any laboratory Hamiltonian written since 1930 and the four-potential is sitting in it: the minimal-coupling rule $p_\mu\to p_\mu-qA_\mu$ , stated here as a classical Lagrangian fact, is how a superconducting flux qubit feels the bias line threading it and how every Standard Model term coupling matter to force was written down — $A^\mu$ generalized to the strong and weak gauge fields, the grammar unchanged. Weyl's 1929 realization that demanding local phase freedom forces a field with exactly $A^\mu$ 's transformation law inverted this article's logic: gauge redundancy stopped being a quirk of potentials and became the generating principle of interactions. The experimental capstone is that the potential's reality is now routinely measured. Aharonov–Bohm interferometry — electrons acquiring a phase $\frac{q}{\hbar}\oint A_\mu dx^\mu$ while flying through field-free regions — has migrated from thought experiment to workhorse: solid-state AB rings map mesoscopic coherence, and electron holography reads magnetic textures through the loop integral of $\mathbf{A}$ . The fine print matters: the measured object is the closed-loop integral, gauge-invariant by construction, never a local value of $A^\mu$ . Meanwhile lattice gauge theory simulates $A^\mu$ (as link variables) and derives fields afterward — the computational world long ago voted for the potentials. Within this chapter the vote happens next article: one antisymmetric derivative of $A^\mu$ , and the field tensor finally exists.

Origins · 1 minHermann Weyl and how this was discovered

Hermann Weyl

1885–1955 · Zürich, Göttingen, Princeton — who read the four-potential's redundancy as the organizing principle of physics

Weyl took this article's most awkward feature — many $A^\mu$ , one physics — and declared it the point. In 1918, fresh from writing the first great textbook of general relativity (Raum, Zeit, Materie), he proposed unifying gravity and electromagnetism by letting the scale ("gauge") of length vary from point to point, with $A_\mu$ as the compensating field. Einstein's objection was lethal — atomic spectra would betray a rider's path history — but the mathematics refused to die. In 1929, with quantum mechanics in hand, Weyl relocated the freedom from length scale to the phase of the wavefunction: demand that physics survive $\psi\to e^{iq\chi/\hbar}\psi$ with $\chi$ arbitrary at each point, and a field must exist that transforms exactly as $A_\mu\to A_\mu+\partial_\mu\chi$ — the four-potential, derived rather than postulated. That inversion, gauge symmetry as the source of interactions, became the template for Yang–Mills theory and the Standard Model. The name "gauge" survives as a fossil of the failed 1918 version — physics kept the idea and the word, and quietly swapped the meaning.

History

1867

— Ludvig Lorenz publishes retarded potentials and the condition now bearing his name, decades before its covariant meaning could be stated

1905

— Einstein's electrodynamics paper transforms fields and coordinates; the potentials still live as separate

V

and

\mathbf{A}

1908

— Minkowski's spacetime program: four-dimensional objects become the natural containers for electromagnetic quantities

1910

— Sommerfeld's four-dimensional vector algebra;

(V/c,\mathbf{A})

takes its place beside

x^\mu

and

J^\mu

in the covariant toolkit

1918

— Weyl's gauge idea is born as a failed scale-invariance theory of gravity plus electromagnetism

1929

— Weyl recasts gauge freedom as local phase symmetry:

A_\mu

becomes the prototype gauge field of quantum theory

1954

— Yang and Mills generalize the four-potential's transformation law to non-abelian symmetry, the future language of the strong and weak forces

1959

— Aharonov and Bohm predict a measurable phase from potentials in field-free regions; Chambers confirms it within a year

The physical picture

Take a charge $q$ sitting at rest. Its electromagnetic personality could not be simpler: a Coulomb scalar potential $V=q/4\pi\varepsilon_0 r$ , and no vector potential at all. Now run past it. In your frame the charge moves — it is a current — and currents make vector potentials. Nothing happened to the charge; you changed, and part of its $V$ reappeared as $\mathbf{A}$ . This should feel familiar. A clock at rest presents a pure time interval between its ticks; viewed from a moving frame, that interval becomes a mixture of time and space. The pair $(V/c,\mathbf{A})$ behaves under boosts exactly as the pair $(ct,\mathbf{x})$ does, and Four-Vectors and the Metric Tensor taught us to read that behavior as a birth certificate: the potentials are the time and space components of a single four-vector, and which part you call "scalar" and which "vector" is frame gossip. The second half of the picture concerns what the potentials are for. The fields are derivatives of the potentials — $\mathbf{E}=-\nabla V-\partial_t\mathbf{A}$ , $\mathbf{B}=\nabla\times\mathbf{A}$ , the working definitions since magnetostatics and Gauge Transformations — the way a force is the gradient of an energy landscape. And just as shifting the landscape by a constant leaves every force intact, shifting $A^\mu$ by a four-gradient leaves every field intact. The redundancy is not a defect; it is declared bookkeeping slack, and choosing how to spend it (a gauge) is what turned four coupled potential equations into the radiation chapter's clean wave equations in the first place. What relativity adds is economy with teeth. Once $A^\mu$ is certified a four-vector, frame-independence is auditable at a glance: $\partial_\mu A^\mu=0$ is a scalar equation, so the Lorenz gauge travels across frames; $\Box A^\mu=\mu_0 J^\mu$ equates two four-vectors, so it holds everywhere if it holds anywhere. The four separate wave equations were never four facts — they were one fact, written by an observer who had not yet been told that $V$ and $\mathbf{A}$ are the same object's shadows.

Figure

One spacetime object, two shadows — the rest frame sees a pure scalar potential, the lab sees scalar and vector mixed by the boost

The math: one four-vector, one wave equation, one scalar condition

Start with the definition and its covariant partner, index conventions as in Four-Vectors and the Metric Tensor:

A^\mu=\left(\frac{V}{c},\,\mathbf{A}\right),\qquad A_\mu=\eta_{\mu\nu}A^\nu=\left(\frac{V}{c},\,-\mathbf{A}\right). \qquad (1)

where: $V$ is the scalar potential (volts), $\mathbf{A}$ the vector potential (T·m = V·s/m), signature the track's locked $(+,-,-,-)$ . The $1/c$ is load-bearing: it gives all four slots the units of $\mathbf{A}$ , and components can only mix under boosts if commensurable — the same reason $x^0$ is $ct$ , not $t$ . The upper index in (1) is, for now, a promissory note; Derivation A pays it. Because $A^\mu$ transforms like $x^\mu$ , the standard boost along $x$ at speed $v=\beta c$ (the matrix of Lorentz Transformations) mixes the potentials componentwise:

\frac{V'}{c}=\gamma\left(\frac{V}{c}-\beta A_x\right),\qquad A'_x=\gamma\left(A_x-\beta\frac{V}{c}\right),\qquad A'_y=A_y,\quad A'_z=A_z. \qquad (2)

where: $\gamma=(1-\beta^2)^{-1/2}$ ; transverse components ride along untouched, exactly as $y$ and $z$ do. Read it physically, both directions. A charge distribution at rest (pure $V$ ) acquires, in a frame where it streams past, a vector potential $A'_x=-\gamma\beta V/c$ : moving charge is current, and current makes $\mathbf{A}$ . Conversely a neutral current-carrying wire (pure $\mathbf{A}$ ) acquires a scalar potential in any frame gliding along it — worked example 1 computes the electric field that appears, and Magnetism as a Relativistic Effect builds its whole argument on the charge-density version of the same fact. Next, the condition that made the radiation chapter's potentials tractable: Lorenz Gauge imposed $\nabla\cdot\mathbf{A}=-\mu_0\varepsilon_0\,\partial V/\partial t$ , proving it attainable and self-maintaining for conserved sources. Compute the four-divergence of (1) with the four-gradient $\partial_\mu=(\partial_t/c,\nabla)$ of Four-Vectors and the Metric Tensor:

\partial_\mu A^\mu=\frac{1}{c^2}\frac{\partial V}{\partial t}+\nabla\cdot\mathbf{A}=0. \qquad (3)

where: $\mu_0\varepsilon_0=1/c^2$ converts the E9 form into this one term for term — the same equation. But the packaging upgrades its status: a four-divergence of a four-vector is a Lorentz scalar (the second derivation of Four-Vectors and the Metric Tensor, which promised on exactly this line that the Lorenz condition would one day be "invariantly imposed"). Zero in one frame means zero in all: the Lorenz gauge is frame-proof. Contrast the Coulomb condition $\nabla\cdot\mathbf{A}=0$ of Coulomb Gauge: three spatial derivatives with no time partner form nobody's scalar, so a boosted observer finds it violated and must re-gauge from scratch. Nothing is wrong with Coulomb gauge in any single frame — its causality scandal was honestly acquitted in that article — but it is a per-frame outfit, while Lorenz gauge is a spacetime one. With the condition in scalar form, the chapter's twin wave equations are ready to fuse.

Derivation AAssembling the covariant wave equation from the twin E9 results

The goal is the single equation $\Box A^\mu=\mu_0 J^\mu$ , obtained by packaging the Lorenz-gauge wave equations of Lorenz Gauge and Inhomogeneous Wave Equations for Potentials — quoted, not re-derived — into four-vector grammar, and then to establish honestly in what sense $A^\mu$ is a four-vector at all. The strategy is deliberate bookkeeping: reconcile the two sign conventions for the wave operator (the radiation chapter used Griffiths's $\Box^2$ ; the covariant d'Alembertian is its negative), convert each slot, and watch the factor $c^2=1/\mu_0\varepsilon_0$ do the one nontrivial step. The watch-out is the four-vector claim itself: it leans on the gauge condition, and the derivation ends by saying so exactly.

Step 1. Quote the inputs. In Lorenz gauge, with all sources conserved, the radiation chapter established

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

together with the condition $\nabla\cdot\mathbf{A}=-\mu_0\varepsilon_0\,\partial V/\partial t$ . These are Lorenz Gauge's boxed results, solved in Inhomogeneous Wave Equations for Potentials; nothing here re-derives them. Why quote rather than start from Maxwell: the covariant route — inserting the field tensor into $\partial_\nu F^{\nu\mu}=\mu_0J^\mu$ — is shorter but circular at this point in the chapter, since $F^{\mu\nu}$ is built from $A^\mu$ one article ahead. Chapter order is honest order: the 3+1 results exist, so we package them.

Step 2. Reconcile the operators. The covariant d'Alembertian, proven a scalar operator in Four-Vectors and the Metric Tensor, is

\Box\equiv\partial_\mu\partial^\mu=\frac{1}{c^2}\frac{\partial^2}{\partial t^2}-\nabla^2=-\,\Box^2,

so the quoted results read $\Box V=\rho/\varepsilon_0$ and $\Box\mathbf{A}=\mu_0\mathbf{J}$ : same physics, opposite dialect. (Griffiths's $\Box^2$ symbol weights space positively; the index-built $\Box$ inherits the metric's time-positive signature. Every sign in this article is in the covariant dialect from here on.)

Step 3. Convert the time slot. Divide $\Box V=\rho/\varepsilon_0$ by $c$ and use $1/\varepsilon_0=\mu_0c^2$ :

\Box\left(\frac{V}{c}\right)=\frac{\rho}{c\,\varepsilon_0}=\mu_0\,c\rho=\mu_0 J^0,

where $J^0=c\rho$ is the time component of the four-current, certified in The Four-Current Density. The three spatial slots already read $\Box A^i=\mu_0 J^i$ . Four equations, one index:

\boxed{\;\Box A^\mu=\mu_0 J^\mu\;}\qquad\text{together with}\qquad \partial_\mu A^\mu=0.

Why the factor of $c$ landed so neatly: the $1/c$ in $A^0=V/c$ and the $c$ in $J^0=c\rho$ are the same unit-matching convention applied to both four-vectors — SI's $\mu_0$ – $\varepsilon_0$ split hides the symmetry that Gaussian units display openly, and the identity $\mu_0\varepsilon_0 c^2=1$ is where SI pays it back.

Step 4. Now the four-vector claim, honestly. On the right of the boxed equation sits $\mu_0 J^\mu$ — a certified four-vector times an invariant constant. On the left acts $\Box$ — a scalar operator. The gauge condition $\partial_\mu A^\mu=0$ is a scalar equation. So the system is form-invariant: if $A^\mu$ solves it in frame $S$ with sources $J^\mu$ , then the reshuffled components $A'^\mu=\Lambda^\mu{}_\nu A^\nu$ , viewed as functions of the primed coordinates, solve the identical system with the correctly transformed sources $J'^\mu$ , Lorenz condition included. Retarded boundary conditions select the same solution in both frames — "retarded" is a frame-proof specification because the past light cone is one, by the invariant-causal-order theorem of Spacetime and Minkowski Geometry. Conclusion: within the Lorenz-gauge family, the potentials of the two frames are related by $\Lambda$ , which is precisely the statement that $A^\mu$ is a four-vector.

Step 5. State the fine print. Step 4 used the gauge condition twice — to write the wave equation and to certify that the primed potentials remain in the family. In a gauge whose defining condition is not a scalar, the argument dies: boost a Coulomb-gauge solution and the primed potentials fail $\nabla'\cdot\mathbf{A}'=0$ ; the transformation law becomes boost plus gauge patch — not a four-vector law. " $A^\mu$ transforms as a four-vector" is thus a true, useful, and gauge-dependent statement: a property of the Lorenz family, not of the potentials as such. The systematic theory of covariant conditions belongs to Gauge Transformations and Covariant Gauge Conditions; the manifestly covariant retarded solution to Covariant Wave Equation and Retarded Solutions. Why the $\mu=0$ component is self-consistent: $\Box A^0=\mu_0 J^0$ with $A^0=V/c$ and $J^0=c\rho$ gives $\Box V=\mu_0c^2\rho=\rho/\varepsilon_0$ , recovering the Lorenz-gauge scalar wave equation; in the static limit the time derivative vanishes and this reduces to Poisson's equation $\nabla^2V=-\rho/\varepsilon_0$ , as required. The identity $\mu_0\varepsilon_0c^2=1$ underpins the conversion throughout.

The boxed pair is relativistic electrodynamics in Lorenz gauge, complete: one four-vector unknown, one four-vector source, one scalar operator, one scalar constraint — the entire content of Maxwell's equations for the potentials, frame-independence readable by index grammar alone. What the equation does not yet exhibit is gauge freedom, and that too covariantizes in one line. The three-vector shifts of Gauge Transformations — $\mathbf{A}\to\mathbf{A}+\nabla\chi$ , $V\to V-\partial\chi/\partial t$ — are exactly the components of

A^\mu\;\longrightarrow\;A^\mu-\partial^\mu\chi, \qquad (4)

where: $\chi(x)$ is any differentiable scalar field, and $\partial^\mu=(\partial_t/c,-\nabla)$ supplies the relative sign between the two 3+1 rules automatically — one four-gradient replaces two memorized formulas. The fields, being the antisymmetric derivative pattern of $A^\mu$ (next paragraph), never notice. Within the Lorenz family the surviving slack is the set of $\chi$ with $\Box\chi=0$ : gauge waves riding the light cone without touching physics. Two structural remarks. First, applying (4) inside the Lorenz condition shifts it by $\Box\chi$ — so any potential can be brought into the family by solving $\Box\chi=-\partial_\mu A^\mu$ , a wave equation with known source: the gauge choice is itself covariant machinery. Second, gauge transformations and Lorentz transformations answer different questions — (4) re-describes the same physics in the same frame; $\Lambda$ re-describes it in a different frame. The pitfall box below keeps them apart. One derivative now separates the potentials from the fields. The antisymmetric combination $\partial^\mu A^\nu-\partial^\nu A^\mu$ is unchanged by (4) term by term, packages $\mathbf{E}=-\nabla V-\partial_t\mathbf{A}$ and $\mathbf{B}=\nabla\times\mathbf{A}$ into a single rank-2 tensor, and renders the two source-free Maxwell equations automatic identities. Executing that construction is the whole business of The Electromagnetic Field Tensor; the identity theorem behind "automatic" belongs to The Dual Field Tensor. Here the arrow is enough: the four-potential is the object the field tensor is made of. The interaction side gets its statement too: a charge couples through the invariant $u_\mu A^\mu$ built with its four-velocity — in Lagrangian form $L_{\text{int}}=q(\mathbf{v}\cdot\mathbf{A}-V)=-q\,u_\mu A^\mu/\gamma$ — the minimal coupling that will drive the covariant force law of Relativistic Dynamics of a Charged Particle and, as $p_\mu\to p_\mu-qA_\mu$ , every gauge interaction physics has since written down.

Figure

The assembly — two E9 wave equations and a gauge condition fuse into one covariant statement

Limiting cases

case	result	meaning
static charges	$A^\mu=(V/c,\mathbf{0})$ , $\nabla^2V=-\rho/\varepsilon_0$	wave equation degenerates to Poisson; electrostatics is the rest frame of the four-potential
steady currents	$A^\mu=(0,\mathbf{A})$ (neutral source), $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$	the magnetostatic vector-Poisson equation; Biot–Savart follows as in the magnetostatics chapter
source-free region	$\Box A^\mu=0$	each component a free wave at $c$ ; plane-wave solutions $A^\mu=\epsilon^\mu e^{-ik_\nu x^\nu}$ with $k_\mu k^\mu=0$ — the classical skeleton of the photon
$\beta\to0$	$V'\approx V-vA_x$ , $A'_x\approx A_x-vV/c^2$	first-order mixing survives even at Galilean speeds — the seed of worked example 1's field
gauge shift $\chi$	$A^\mu\to A^\mu-\partial^\mu\chi$ , fields unchanged	the redundancy; with $\Box\chi=0$ it preserves the Lorenz family
invariant norm	$A_\mu A^\mu=(V/c)^2-\lvert\mathbf{A}\rvert^2$	frame-proof but gauge-dependent — physical conclusions may not be built on it

Common trapGauge invariance is not frame invariance — and neither one makes the potentials themselves physical

Three distinct statements are routinely mashed into one. A gauge transformation $A^\mu\to A^\mu-\partial^\mu\chi$ happens within a single inertial frame: it changes the potentials, touches no field, and signals redundancy. A Lorentz transformation $A'^\mu=\Lambda^\mu{}_\nu A^\nu$ changes the observer: it mixes $V$ with $\mathbf{A}$ (and, downstream, $\mathbf{E}$ with $\mathbf{B}$ ), changing what each frame measures. Confuse them and you will "prove" absurdities — e.g. that a boost can be undone by a clever $\chi$ (it cannot: no gauge shift alters a single field value). The third mash-up is subtler: the norm $A_\mu A^\mu$ is a Lorentz scalar (in Lorenz gauge) but not gauge-invariant, so it carries no physics by itself. Frame-proof and gauge-proof are independent filters; only quantities passing both — field-tensor contractions, $\oint A_\mu dx^\mu$ around a closed loop — describe the world. And a units trap for good measure: the time slot is $V/c$ , not $V$ ; omit the $c$ and equation (2) manufactures dimensional nonsense that no later step will flag.

Worked examples

1 · The wire that grows an electric field (confidence builder)

Setup. A long neutral wire on the $x$ -axis carries current $I=10\,$ A in the $+x$ direction; outside it, in the wire frame, $V=0$ and (choosing the standard magnetostatic gauge with reference radius $\rho_0=10\,$ cm) $A_x(\rho)=\dfrac{\mu_0 I}{2\pi}\ln\dfrac{\rho_0}{\rho}$ . An observer drifts along the wire at $\beta=0.6$ . Find the scalar potential and the radial electric field she measures at $\rho=1\,$ cm.

Step 1. Potentials in the wire frame at $\rho=1$ cm: $A_x=(2\times10^{-7}\ \mathrm{T\,m/A})(10\ \mathrm{A})\ln 10=4.61\times10^{-6}$ T·m, and $V=0$ everywhere.

Step 2. Boost with equation (2), $\gamma=1.25$ : the moving observer finds

V'=\gamma\left(V-\beta c\,A_x\right)=-\gamma\beta c\,A_x=-(1.25)(0.6)(2.998\times10^8)(4.61\times10^{-6})\ \mathrm{V}=-1.04\times10^{3}\ \mathrm{V}.

A kilovolt-scale scalar potential, conjured from a wire that has none — the reference-radius constant rides along inside the logarithm, but differences and derivatives of $V'$ are reference-free.

Step 3. In her frame the configuration is again steady, so $\mathbf{E}'=-\nabla' V'$ with the transverse coordinate unchanged ( $\rho'=\rho$ ):

E'_\rho=-\frac{\partial V'}{\partial\rho}=-\gamma\beta c\,\frac{\mu_0 I}{2\pi\rho}=-\gamma\beta c\,B(\rho)=-(1.25)(0.6)(2.998\times10^8)(2.00\times10^{-4})\ \mathrm{V/m},

giving $E'_\rho=\mathbf{-4.50\times10^{4}\ V/m}$ at $\rho=1$ cm — pointing radially inward: the wire she chases looks negatively charged.

Step 4. Two independent audits. Charge bookkeeping: the four-current gives $\lambda'=\gamma(\lambda-\beta I/c)=-25.0$ nC/m, and that line charge gives $E'_\rho=\lambda'/2\pi\varepsilon_0\rho=-4.50\times10^4$ V/m — identical. Field bookkeeping: the rules of Transformation of Electric and Magnetic Fields send $(E=0,\ B=\mu_0I/2\pi\rho)$ to $E'_\perp$ of magnitude $\gamma\beta cB$ — identical again. The lesson: potentials, sources, and fields transform as one covariant package; and the "electrified wire" seen by a moving observer, read off the potentials in three lines, is the phenomenon Magnetism as a Relativistic Effect explains microscopically.

2 · Boosted Coulomb potential, by the numbers (exam level)

Setup. A charge $q=1$ nC moves at $\beta=0.9$ along $+x$ . Using Derivation B's result, evaluate $V$ and $A_x$ at present-position distance $R=1$ cm broadside ( $\theta=90^\circ$ ) and dead ahead ( $\theta=0$ ), and verify the invariant $A_\mu A^\mu$ against the rest frame.

Step 1. Rest-frame anchor: $V_0=q/4\pi\varepsilon_0 R=(8.99\times10^9)(10^{-9})/(0.01)=898.8$ V; $\gamma=(1-0.81)^{-1/2}=2.294$ .

Step 2. Broadside, Derivation B's formula with $\sin\theta=1$ : $V=V_0/\sqrt{1-\beta^2}=\gamma V_0=\mathbf{2062\ V}$ , and $A_x=\beta V/c=(0.9)(2062)/(2.998\times10^8)=\mathbf{6.19\times10^{-6}}$ T·m, directed along the motion.

Step 3. Dead ahead, $\sin\theta=0$ : $V=V_0/\sqrt{1-0}=\mathbf{898.8\ V}$ — exactly Coulomb, enhancement factor 1.000, matching the fore/aft audit in Liénard–Wiechert Potentials. (At $\theta=45^\circ$ the factor is $(1-0.81/2)^{-1/2}=1.296$ — the equipotential pancake fills in smoothly between the extremes.)

Step 4. Invariant audit at the broadside event. Lab: $(V/c)^2-A_x^2=(2062/c)^2-(6.19\times10^{-6})^2=8.988\times10^{-12}$ (SI). Rest frame, same event ( $x'=\gamma(x-vt)=0$ , $y'=R$ , so $r'=R$ ): $(V_0/c)^2=(898.8/c)^2=8.988\times10^{-12}$ . Equal to machine precision, as the norm of a four-vector must be. The lesson: the boosted Coulomb potential is anisotropic in components and isotropic in no frame but the charge's own — but its invariant is blind to the boost, and checking it costs one line once the object is a certified four-vector.

Go deeper · Derivation B · 2 minBoosting the Coulomb potential into the Lienard-Wiechert form

Derivation BBoosting the Coulomb potential into the Lienard-Wiechert form

The goal is the potential of a charge in uniform motion, derived with no retardation analysis whatsoever: boost the electrostatic Coulomb solution using the four-vector law (2), and land exactly on the constant-velocity form that Liénard–Wiechert Potentials extracted from the retarded integral. The strategy showcases the four-potential's power — a hard light-cone bookkeeping problem becomes three lines of special relativity — and the cross-check against the retarded route is the point, not a bonus: two utterly different formalisms, one answer. Watch one subtlety: the boost delivers the potentials at the same event, so the rest-frame radius must be rewritten in lab coordinates before the result can be read.

Step 1. Solve the trivial problem. In the charge's rest frame $S'$ , electrostatics gives $A'^\mu=(V'/c,\,\mathbf{0})$ with $V'=q/4\pi\varepsilon_0 r'$ — the Coulomb potential, no vector potential, no retardation (nothing moves).

Step 2. Boost to the lab $S$ , where the charge travels at $v\hat{\mathbf{x}}$ . Inverting (2) (equivalently, boosting with $-\beta$ ):

\frac{V}{c}=\gamma\left(\frac{V'}{c}+\beta A'_x\right)=\gamma\frac{V'}{c},\qquad A_x=\gamma\left(A'_x+\beta\frac{V'}{c}\right)=\frac{\gamma\beta V'}{c},

so $V=\gamma V'$ and $\mathbf{A}=\mathbf{v}V/c^2$ — the vector potential is the scalar potential times $\mathbf{v}/c^2$ , at every point, automatically. Why $\mathbf{A}=\mathbf{v}V/c^2$ was inevitable: in the rest frame the four-potential is "at rest" (pure time component), so in the lab it must point along the source's four-velocity — $A^\mu\propto u^\mu$ for uniform motion, the same alignment the covariant Liénard–Wiechert package displays.

Step 3. Rewrite the rest-frame radius at the common event. The Lorentz transformation gives $x'=\gamma(x-vt)$ , $y'=y$ , $z'=z$ , so with $\rho^2\equiv y^2+z^2$ ,

r'=\sqrt{\gamma^2(x-vt)^2+\rho^2}=\gamma\sqrt{(x-vt)^2+(1-\beta^2)\,\rho^2},

and the $\gamma$ from Step 2 cancels the $\gamma$ pulled out of the root:

V(\mathbf{r},t)=\frac{q}{4\pi\varepsilon_0}\,\frac{1}{\sqrt{(x-vt)^2+(1-\beta^2)(y^2+z^2)}}.

Step 4. Recognize the geometry. Let $\mathbf{R}=(x-vt,\,y,\,z)$ be the vector from the charge's present position, with $\sin\theta=\rho/R$ . Then $(x-vt)^2+(1-\beta^2)\rho^2=R^2(1-\beta^2\sin^2\theta)$ , and

\boxed{\;V=\frac{q}{4\pi\varepsilon_0 R}\,\frac{1}{\sqrt{1-\beta^2\sin^2\theta}},\qquad \mathbf{A}=\frac{\mathbf{v}}{c^2}\,V\;}

— term for term the constant-velocity present-position form of Liénard–Wiechert Potentials, where it emerged from the delta-function Jacobian of the retarded integral. Here it emerged from a rotation of spacetime axes. The two machineries must agree for uniform motion, and do, exactly; where they part company — accelerated charges, where no rest frame persists — only the retarded route survives, covariantly executed in Covariant Wave Equation and Retarded Solutions. Two independent routes agree at $\beta=0.9$ , $q=1$ nC, $R=1$ cm: the boost formula gives broadside $V=\gamma V_0=2.294\times898.8=2061.9$ V, and the boxed present-position formula gives $898.8/\sqrt{0.19}=2061.9$ V — identical. The fore/aft enhancement factor is exactly 1 (pure Coulomb), and the invariant $(V/c)^2-A^2=(V'/c)^2$ holds numerically to machine precision at the broadside event.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §12.3.4–§12.3.5 · Jackson, Classical Electrodynamics, 3rd ed., §11.9, §6.3 · Zangwill, Modern Electrodynamics, §22.3, §24.1 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 5–6 The scalar potential is written $V$ throughout, matching E1–E9 usage. The field-tensor component identities $F^{0i}=-E^i/c$ and $F^{ij}=-\varepsilon^{ijk}B_k$ , the covariant gauge-invariance argument (which follows from commuting partials), and the reduction from $\partial_\nu F^{\nu\mu}=\mu_0J^\mu$ to the wave equation are treated in their respective owner articles (The Electromagnetic Field Tensor, The Dual Field Tensor, Covariant Form of Maxwell's Equations). Derivation A here assembles $\Box A^\mu=\mu_0J^\mu$ from the E9 Lorenz-gauge results — the chapter-order-honest route — and adds the gauge-dependence fine print on four-vector status; Derivation B derives the boost of the Coulomb potential to the constant-velocity Liénard–Wiechert form. The worked numbers use no numerical constants beyond standard $\mu_0,\varepsilon_0,c$ : wire 10 A, $\beta=0.6$ , 1 cm gives $V'=-1.04$ kV, $E'_\rho=-4.50\times10^4$ V/m, dual-audited via $\lambda'=-25.0$ nC/m and $\gamma\beta cB$ ; boosted Coulomb at $\beta=0.9$ gives broadside 2062 V, $A_x=6.19\times10^{-6}$ T·m, fore/aft exactly Coulomb, invariant norm matched.

Builds on: Four-Vectors and the Metric Tensor · Lorentz Transformations · The Four-Current Density · Spacetime and Minkowski Geometry Leads to: The Electromagnetic Field Tensor · Transformation of Electric and Magnetic Fields · Lorentz Invariants of the Field · The Dual Field Tensor

Are the potentials physically real, or just calculational scaffolding for the fields?

The classical answer was "scaffolding": only fields push charges, and gauge freedom means $A^\mu$ is not even unique. The modern answer is sharper. Quantum mechanically, a charge acquires a phase $\frac{q}{\hbar}\oint A_\mu dx^\mu$ around a closed loop, and the Aharonov–Bohm experiment measures that phase even when the loop never touches nonzero fields — the potentials act where the fields are absent. But notice what is measured: a closed-loop integral, gauge-invariant even though $A^\mu$ is not. The defensible statement: the gauge-equivalence class of $A^\mu$ carries physically real, sometimes nonlocal information (the holonomy of a connection), while any particular representative remains partly conventional. The potentials are more real than the nineteenth century thought and less real than their kilovolt values suggest — worked example 1's $V'=-1.04$ kV contains an arbitrary reference constant; its derivative, the 45 kV/m field, does not.

How can the four-vector character of the potentials depend on the gauge — either something transforms covariantly or it doesn't?

The subtlety is that "the potential" is not one function but an equivalence class, and a gauge condition is a rule for picking representatives. A boost maps the class covariantly — that much is gauge-proof, because the fields transform tensorially. The question is whether the boost maps your chosen representative in frame $S$ to the representative your rule picks in $S'$ . For the Lorenz rule, yes: $\partial_\mu A^\mu=0$ is a scalar statement, so $\Lambda A$ already satisfies it, and the four-vector law closes. For the Coulomb rule, no: $\nabla\cdot\mathbf{A}=0$ singles out a preferred time direction, $\Lambda A$ violates the primed condition, and the honest transformation is "boost, then re-gauge" — not a linear componentwise law. Covariance here is a property of the (object, selection rule) pair. The systematic account of which conditions are covariant is the business of Gauge Transformations and Covariant Gauge Conditions later in this chapter.

The radiation chapter wrote the wave operator as del-squared minus time derivatives, but this article's d'Alembertian has the opposite sign. Which is right?

Both — they are dialects, and the reconciliation is one minus sign: Griffiths's $\Box^2\equiv\nabla^2-\frac{1}{c^2}\partial_t^2$ (used across the E9 articles) versus the covariant $\Box\equiv\partial_\mu\partial^\mu=\frac{1}{c^2}\partial_t^2-\nabla^2=-\Box^2$ , whose sign is dictated by the $(+,-,-,-)$ metric with no room for choice. Hence the paired appearances: $\Box^2V=-\rho/\varepsilon_0$ and $\Box V=+\rho/\varepsilon_0$ are the same equation. When consulting references, check the operator's definition before comparing any signs — Jackson's $\Box$ matches ours; texts in the $(-,+,+,+)$ signature flip it again. The physics test is the static limit: every convention must degenerate to Poisson's equation $\nabla^2V=-\rho/\varepsilon_0$ with the minus sign that makes a positive charge a potential hill.

Does V equals gamma V-prime contradict the invariance of charge — the source got stronger without more charge?

No, because $V$ was never a scalar and never claimed to be: it is the time component of a four-vector, and components grow under boosts exactly as $ct$ does. Charge invariance lives one four-vector over, in $J^\mu$ — the total charge $\int\rho\,dV$ is frame-proof because $\gamma\rho$ integrates over a $1/\gamma$ -contracted volume (the bookkeeping of The Four-Current Density). The broadside potential enhancement $V=\gamma V_0$ coexists with an exactly Coulomb potential fore and aft, so the angular average is reshuffled, not inflated; Gauss's law audits of the boosted fields close on the same $q/\varepsilon_0$ in every frame. The moral generalizes: never read physics off a lone component — energies ( $p^0$ ), charge densities ( $J^0$ ), and potentials ( $A^0$ ) all dilate under boosts, and all for the same geometric reason.

What happens to the leftover gauge freedom after imposing the Lorenz condition — is the potential finally unique?

Not yet, and the leftover slack is physically eloquent. Within the Lorenz family you may still shift $A^\mu\to A^\mu-\partial^\mu\chi$ with $\Box\chi=0$ : gauge functions solving the free wave equation, rippling along the light cone while changing no field. Uniqueness then comes from boundary conditions — retarded behavior and decay at infinity nail the physical solution. The residual freedom is a counting device, not a nuisance: a photon mode $\epsilon^\mu e^{-ik_\nu x^\nu}$ starts with four polarization components, loses one to the Lorenz condition ( $k_\mu\epsilon^\mu=0$ ) and one to residual gauge ( $\epsilon^\mu\sim\epsilon^\mu+\alpha k^\mu$ ), leaving the two transverse polarizations light actually exhibits — the counting Coulomb Gauge displayed by construction, recovered covariantly. The full treatment belongs to Gauge Transformations and Covariant Gauge Conditions.

Is the invariant norm of the four-potential ever useful, given that it is gauge-dependent?

Rarely as physics, often as a diagnostic — and knowing the difference is the skill. Because $A_\mu A^\mu$ changes under gauge transformations, no measurable quantity may depend on it; any "prediction" built from it is a gauge artifact. But within a fixed, covariantly specified gauge it is a legitimate Lorentz scalar, and checking its frame-independence — as worked example 2 does for the boosted Coulomb potential — is a cheap, sharp audit of covariance: if your boosted potentials fail the norm check, you have made an algebra error, not discovered physics. The genuinely physical scalars are built one storey up, from the field tensor — $F_{\mu\nu}F^{\mu\nu}$ and its dual partner — and Lorentz Invariants of the Field is devoted to exactly what they say.