Retarded Potentials and Liénard-Wiechert Fields

LEP, the electron–positron collider that filled CERN's 27-kilometre ring through the 1990s, lost 2.33 GeV of beam energy every single lap at 95 GeV per beam. Not to collisions — to light. The electrons, bent by the ring's magnets, radiated it away, and the RF cavities spent their lives pouring it back in. Push to 104.5 GeV and the loss climbs to roughly three per cent of the beam's entire energy, every lap. That is why LEP stopped there, and why the same tunnel now holds protons. The equation that predicted this is Liénard's, from 1898 — seven years before relativity:

P = \frac{q^2}{6\pi\varepsilon_0 c}\,\gamma^6\Big[\dot{\vec\beta}^{\,2} - \big(\vec\beta\times\dot{\vec\beta}\big)^2\Big].

Behind it sits the retarded potential of a point charge, and behind that the deepest statement in the subject: the field at a point knows nothing about where the charge is now. It knows only where the charge was, one light-travel-time ago. Causality is not imposed on this solution — it is built into the Green function, which is supported exactly on the light cone, a Lorentz-invariant surface. The construction is relativistic before you do anything to it.

Key takeaways

Signature $(+,-,-,-)$ , $p\cdot p = +m^2c^2$ ; $m$ is the invariant rest mass. $u^\mu = \gamma(c,\vec v)$ , $J^\mu = (c\rho,\vec J)$ .
The covariant potential: $\boxed{A^\mu(x) = \dfrac{\mu_0 c}{4\pi}\,\dfrac{q\,u^\mu}{u_\nu(x - w)^\nu}\bigg|_{\text{ret}}}$ — evaluated at the retarded point on the worldline. The three-vector fields are owned by `aq-em-lienard-wiechert-potentials`; this page owns the covariant form and what it explains.
Causality is in the Green function. $\Box G_{\text{ret}} = \delta^4$ is solved by $G_{\text{ret}} \propto \theta(\Delta t)\,\delta\big((\Delta x)^2\big)$ — supported on the light cone, both factors Lorentz-invariant. Manifest causality, no extra postulate.
The $(1 - \hat R\cdot\vec\beta)$ factor is a JACOBIAN, not time dilation. It comes from the retarded-time condition, not from $\gamma$ — the most-misattributed factor in electrodynamics.
Radiation is the $1/R$ piece. The velocity field falls as $1/R^2$ and carries no energy to infinity; the acceleration field falls as $1/R$ , and $(1/R)^2 \times 4\pi R^2$ survives.
Key idea: $\gamma^6$ is the LINAC; $\gamma^4$ is the RING — and the published LEP data proves it. $\gamma^6[\dot{\vec\beta}^2 - (\vec\beta\times\dot{\vec\beta})^2] = \gamma^4\dot{\vec\beta}^2$ when $\vec a \perp \vec v$ . LEP lost 140 MeV/turn at 47 GeV and 2330 MeV/turn at 95 GeV. If the law is $\gamma^4$ the ratio must be $(95/47)^4 = \mathbf{16.69}$ ; measured, $2330/140 = \mathbf{16.64}$ . Agreement to 0.3%.

The core insight

Two ideas, and they are one idea. First: the retarded Green function $G_{\text{ret}} \propto \theta(\Delta t)\,\delta((\Delta x)^2)$ is built from two Lorentz invariants — the interval, and the sign of a time separation (invariant for causally connected events). So the solution cannot violate causality: the field at $x$ sums over exactly the source points on $x$ 's past light cone, and every observer agrees which points those are. Second: the resulting $A^\mu \propto q\,u^\mu/u_\nu(x-w)^\nu$ is manifestly a four-vector divided by a scalar — its transformation law is visible in the symbols. Retarded electrodynamics did not have to be made relativistic. It was relativistic, because the light cone was in the Green function from the beginning.

Where this lives today

Every antenna is Liénard–Wiechert: the $1/R$ acceleration field is the radiated field, the $1/R^2$ velocity field is the reactive near field, and antenna design is the art of arranging accelerating charges so the far fields add where you want them and cancel where you do not. The $\gamma^4$ of a bending magnet defines the architecture of high-energy physics. It is why an electron machine at the energy frontier must be linear (SLAC, and the proposed ILC and CLIC) while a hadron machine can be circular; why LEP's tunnel was handed to the LHC; and why the proposed Future Circular Collider must be very large, since $\Delta E \propto E^4/\rho$ can only be tamed by making $\rho$ enormous. Harvested deliberately, the same loss is the world's brightest X-ray source: the ESRF, Diamond, APS and SPring-8 are machines built to radiate. It also dominates the non-thermal radio sky. When a radio telescope maps the Crab Nebula it is looking, quite literally, at Liénard's 1898 formula evaluated for $\gamma \sim 10^6$ .

The physical picture

The wave equation $\Box A^\mu = \mu_0 J^\mu$ (`aq-sr-wave-equation-four-potential`) is linear, so it is solved by a Green function. For a point charge the four-current is concentrated on a worldline $w^\mu(\tau)$ , and the superposition collapses to a single term — the one point on that worldline lying on the observation point's past light cone. That is the retarded position, and the whole of Liénard–Wiechert is the statement that the field at $x$ is fixed by the charge's position, velocity and acceleration there. Two consequences deserve separating, because they are usually run together.

The kinematic one. Because the charge is moving while the light cone sweeps past it, the amount of charge that contributes is not $q$ but $q$ divided by $(1 - \hat R\cdot\vec\beta)$ . This factor is a Jacobian — it is $|\partial t_{\text{ret}}/\partial t|^{-1}$ — and it has nothing whatever to do with time dilation. It survives at $\gamma \to 1$ ; a charge crawling at $\beta = 0.01$ has it.

The dynamical one. Differentiating $A^\mu$ to get $F^{\mu\nu}$ produces two terms with different powers of distance: a velocity field $\propto 1/R^2$ (a boosted Coulomb field — `aq-sr-field-uniformly-moving-charge`) and an acceleration field $\propto 1/R$ . Only the second radiates. The Poynting flux goes as (field) $^2$ and a sphere's area grows as $R^2$ , so the $1/R$ piece delivers a constant power through every sphere, all the way to infinity, while the $1/R^2$ piece reaches nowhere. Radiation exists because of an exponent.

Figure

WHY RADIATION EXISTS, IN ONE EXPONENT. Differentiating the Liénard-Wiechert potential gives TWO field terms. The VELOCITY field falls as 1/R² — the boosted Coulomb field, bound to the charge, carrying its energy around with it. The ACCELERATION field falls as 1/R, and exists only when the charge accelerates. Now integrate the Poynting flux over a sphere of radius R. The flux goes as the SQUARE of the field, and the sphere's area goes as 4πR². For the velocity field: (1/R²)² × R² = 1/R² → the power through the sphere VANISHES as R → ∞. For the acceleration field: (1/R)² × R² = CONSTANT → the power through every sphere is the SAME, all the way to infinity. That surviving constant IS the radiated power, and it is Liénard's formula. The entire existence of radiation — of light, of antennas, of synchrotrons — turns on the difference between the exponent 1 and the exponent 2.

The core equations

The manifestly covariant retarded solution — the equation this page owns:

\boxed{\;A^\mu(x) \;=\; \frac{\mu_0 c}{4\pi}\;\frac{q\,u^\mu}{u_\nu\,(x-w)^\nu}\;\bigg|_{\text{ret}}\;} \qquad (1)

where: $w^\mu(\tau)$ is the charge's worldline, $u^\mu = \gamma(c,\vec v)$ its four-velocity, and "ret" means evaluation at the unique proper time at which $w^\mu(\tau)$ crosses the past light cone of $x$ . Physical reading: a four-vector divided by a Lorentz scalar — $u_\nu(x-w)^\nu$ is a contraction of two four-vectors — so $A^\mu$ transforms as a four-vector by inspection. The three-vector fields are derived in `aq-em-lienard-wiechert-potentials`, which prints exactly this potential as $A^\mu = \tfrac{\mu_0 qc}{4\pi}\,u^\mu/(R_\nu u^\nu)$ . The Green function that puts causality in:

\Box\,G_{\text{ret}}(x - x') = \delta^4(x - x'), \qquad G_{\text{ret}}(x-x') = \frac{1}{2\pi}\,\theta\big(t - t'\big)\;\delta\big((x-x')^2\big) \qquad (2)

where: $(x-x')^2 = c^2(t-t')^2 - |\vec x - \vec x\,'|^2$ is the invariant interval. Physical reading: the Green function is supported exactly on the light cone, and both factors are Lorentz-invariant — $\delta((x-x')^2)$ obviously, and $\theta(t-t')$ because the time-ordering of causally connected events is invariant (`aq-sr-causal-structure`). So the retarded solution is manifestly causal and manifestly covariant, and neither had to be arranged. The $\delta$ supplies both cones; the $\theta$ chooses the past one, and that choice encodes the arrow of time, not the geometry. The Jacobian, which is not time dilation:

\kappa \;\equiv\; 1 - \hat R\cdot\vec\beta, \qquad \frac{\partial t_{\text{ret}}}{\partial t} = \frac{1}{\kappa} \qquad (3)

where: $\hat R$ points from the retarded position to the field point. Physical reading: $\kappa$ is a Jacobian of the retarded-time map, and it is not $\gamma$ — first order in $\beta$ , and it survives at $\gamma = 1$ . Liénard's radiated power — a Lorentz-invariant statement:

\boxed{\;P = \frac{q^2}{6\pi\varepsilon_0 c}\;\gamma^6\Big[\dot{\vec\beta}^{\,2} - \big(\vec\beta\times\dot{\vec\beta}\big)^2\Big] \;=\; \frac{\mu_0 q^2 c}{6\pi}\,\gamma^6\Big[\dot{\vec\beta}^{\,2} - \big(\vec\beta\times\dot{\vec\beta}\big)^2\Big]\;} \qquad (4)

where: $\vec\beta = \vec v/c$ and $\dot{\vec\beta} = \mathrm{d}\vec\beta/\mathrm{d}t$ , a rate with units $\mathrm{s^{-1}}$ — not an acceleration. Physical reading: $P$ is a Lorentz scalar — every inertial observer agrees on it. The covariant form makes that visible: $P = -\tfrac{\mu_0 q^2}{6\pi c}\,a^\mu a_\mu$ with $a^\mu = \mathrm{d}u^\mu/\mathrm{d}\tau$ , a scalar built from a four-vector (`aq-em-relativistic-larmor-formula`). ⚠ The prefactor in (4) carries $c$ in the numerator, not the denominator — see the warning card.

Deriving the covariant solution

Derivation AThe invariant Green function — causality is not imposed, it is inherited from the light cone

The goal is to solve $\Box A^\mu = \mu_0 J^\mu$ for an arbitrary source and to notice, on the way, that the solution could not have violated causality even if we had wanted it to. What belongs here, rather than in the electromagnetism track, is the observation that the Green function is built entirely out of Lorentz invariants — which is why the whole construction is relativistic before any relativity is done to it.

Step 1. Ask for the Green function. Linearity gives $A^\mu(x) = \mu_0\int G(x-x')\,J^\mu(x')\,\mathrm{d}^4x'$ provided $\Box G = \delta^4$ . Why: since $\Box$ is a scalar operator and $\delta^4$ is invariant, $G$ can only depend on $x - x'$ through the invariant interval $(x-x')^2$ — and on the sign of $t - t'$ .

Step 2. Solve it. The unique solution supported in the past is G_{\text{ret}}(x-x') = \frac{1}{2\pi}\,\theta(t-t')\,\delta\big((x-x')^2\big). Why: the delta puts the support exactly on the light cone — not inside, not outside, on it. The field at $x$ is influenced only by source points from which a light signal arrives precisely now.

Step 3. Check that both factors are invariant. $(x-x')^2$ is invariant by construction. And $\theta(t-t')$ is invariant on the support of the delta, because there the separation is null, and no Lorentz transformation can reverse the time-ordering of null- or timelike-separated events. Why: this is the whole point. The step function looks like it picks out a preferred time direction, which would break covariance. It does not, because it is only ever evaluated on the light cone, where all observers agree on the order.

Step 4. Reduce the delta, then specialise to a point charge. Its roots $t' = t \mp R/c$ are cut down by $\theta(t-t')$ to the retarded one alone, giving $A^\mu = \tfrac{\mu_0}{4\pi}\int R^{-1}J^\mu(\vec x\,', t - R/c)\,\mathrm{d}^3x'$ — the retarded potentials, as a four-vector equation. Now put $J^\mu(x) = qc\int u^\mu(\tau)\,\delta^4(x - w(\tau))\,\mathrm{d}\tau$ ; the $\tau$ integral against the light-cone delta collapses everything onto a single retarded proper time: \boxed{\;A^\mu(x) = \frac{\mu_0 c}{4\pi}\,\frac{q\,u^\mu}{u_\nu(x-w)^\nu}\bigg|_{\text{ret}}\;} Why: the denominator is the Jacobian of that last delta, and it is a Lorentz scalar. Expanding it gives $u_\nu(x-w)^\nu = \gamma c R(1 - \hat R\cdot\vec\beta) = \gamma c R\,\kappa$ , so the $\gamma$ in $u^\mu$ cancels against the $\gamma$ here — and what remains in the three-vector potentials is $\kappa$ , not $\gamma$ . That cancellation is the cleanest available proof that $\kappa$ is not a time-dilation factor. *Verified in verify_aq-sr-lienard-wiechert-potentials.py: $u_\nu(x-w)^\nu$ is expanded in components and shown to equal $\gamma cR(1 - \hat R\cdot\vec\beta)$ exactly, so the $\gamma$ 's cancel; and $A^\mu$ is confirmed to reduce to the Coulomb potential for a charge at rest.*

Go deeper · Derivation B · 2 minOwned result: Lienard's power, and the gamma-6 versus gamma-4 split that killed LEP

Derivation BOwned result: Lienard's power, and the gamma-6 versus gamma-4 split that killed LEP

Liénard's 1898 formula gives the total power radiated by a charge in arbitrary motion. Two things about it are distinctively relativistic and belong here rather than in the electromagnetism track: it is a Lorentz invariant — all observers agree on it — and the way its $\gamma$ -dependence collapses depends entirely on whether the acceleration is along the velocity or across it. That collapse is two lines of algebra, and it is the single most consequential fact in the design of particle accelerators.

Step 1. State the formula and check its dimensions, because the base does not. P = \frac{q^2}{6\pi\varepsilon_0 c}\,\gamma^6\Big[\dot{\vec\beta}^{\,2} - \big(\vec\beta\times\dot{\vec\beta}\big)^2\Big] = \frac{\mu_0 q^2 c}{6\pi}\,\gamma^6\Big[\dot{\vec\beta}^{\,2} - \big(\vec\beta\times\dot{\vec\beta}\big)^2\Big]. Why: $\dot{\vec\beta}$ is a rate, $\mathrm{s^{-1}}$ , not an acceleration, so the prefactor must carry $\mathrm{[J\,s]}$ to make watts — and $q^2/(\varepsilon_0 c) = \mathrm{N\,m\,s}$ does ✓. The base's $\mu_0q^2/(6\pi c)$ gives $\mathrm{N\,s\,m^{-1}}$ : not watts, short by exactly $c^2$ .

Step 2. Confirm the non-relativistic limit. Set $\gamma\to1$ , $\vec a \parallel \vec v$ , so the cross product vanishes and $\dot{\vec\beta} = \vec a/c$ : P \to \frac{q^2}{6\pi\varepsilon_0 c}\cdot\frac{a^2}{c^2} = \frac{q^2 a^2}{6\pi\varepsilon_0 c^3} . Why: exactly Larmor's 1897 result, with coefficient exactly 1. This is the dimensional check with teeth: the broken prefactor misses it by $c^2 \approx 10^{17}$ . A formula that misses Larmor by seventeen orders of magnitude is not a convention dispute.

Step 3. Now the geometry, and it is the whole story. Linear acceleration ( $\vec a\parallel\vec v$ , a linac): the cross product is zero, nothing cancels, P_{\parallel} \propto \gamma^6\,\dot{\vec\beta}^{\,2} . Circular motion ( $\vec a\perp\vec v$ , a storage ring): now $|\vec\beta\times\dot{\vec\beta}| = \beta|\dot{\vec\beta}|$ , so the bracket becomes $\dot\beta^2(1 - \beta^2) = \dot\beta^2/\gamma^2$ , and two powers of $\gamma$ are eaten: P_{\perp} \propto \gamma^6\cdot\frac{\dot\beta^2}{\gamma^2} = \gamma^4\,\dot{\vec\beta}^{\,2} . Why: $\gamma^6$ is the LINAC and $\gamma^4$ is the RING, and it is easy to get backwards. A ring bends the beam, so its acceleration is perpendicular and it suffers $\gamma^4$ — not $\gamma^6$ . Anyone who "corrects" a ring's $\gamma^4$ to $\gamma^6$ has broken it.

Step 4. Convert to the engineer's form. For a ring of bending radius $\rho$ , the energy radiated per turn is U_0 = \frac{q^2\beta^3\gamma^4}{3\varepsilon_0\rho} \qquad\Longrightarrow\qquad \boxed{\;\Delta E\,[\mathrm{keV}] = 88.46\,\frac{E\,[\mathrm{GeV}]^4}{\rho\,[\mathrm{m}]}\;} for electrons at $\beta\approx1$ . Why: every synchrotron designer knows this by heart, and its fourth power is the $\gamma^4$ of Step 3, since $E = \gamma m_ec^2$ .

Step 5. Test it against LEP. The sharpest test needs no formula at all — only the published losses and the claim that the exponent is 4: \frac{2330}{140} = \mathbf{16.64} \qquad\text{versus}\qquad \left(\frac{95}{47}\right)^{\!4} = \mathbf{16.69} . Why: they agree to 0.3%, and nothing but the exponent went into the comparison. Had the ring's law been $\gamma^6$ , the ratio would have been $(95/47)^6 = 68.2$ — off by a factor of four. LEP's own logbook confirms $\gamma^4$ , and the worked examples below check the absolute numbers too. *Verified in verify_aq-sr-lienard-wiechert-potentials.py: $\gamma^6[\dot{\vec\beta}^2 - (\vec\beta\times\dot{\vec\beta})^2] = \gamma^4\dot{\vec\beta}^2$ for $\vec a\perp\vec v$ is proved symbolically; the $\gamma\to1$ limit is shown to be exactly Larmor; the base's prefactor is shown to be short by $c^2$ ; and all three LEP rows plus the LHC row are recomputed in `mpmath` at 50 digits against the article's printed digits.*

Go deeper · Derivation C · 1 minOwned result: the radiation field carries BOTH invariants equal to zero

Derivation COwned result: the radiation field carries BOTH invariants equal to zero

"Radiation" has an invariant definition, and it is sharper than "the $1/R$ bit". This derivation connects the acceleration field to the invariant classification of `aq-sr-electromagnetic-field-invariants` and shows they say the same thing — so that "does this radiate?" becomes a question with a frame-independent answer.

Step 1. Take the radiation field. Keeping only the $1/R$ terms gives \vec E_{\text{rad}} = \frac{q}{4\pi\varepsilon_0 c}\,\frac{\hat R\times\big[(\hat R - \vec\beta)\times\dot{\vec\beta}\big]}{\kappa^3 R}, \qquad \vec B_{\text{rad}} = \frac{\hat R\times\vec E_{\text{rad}}}{c}, so $\vec E_{\text{rad}} \perp \vec B_{\text{rad}}$ and $|\vec E_{\text{rad}}| = c|\vec B_{\text{rad}}|$ . Why: these are the standard acceleration fields, derived in full in `aq-em-lienard-wiechert-potentials` — we import them. The two relations say the far field is a locally plane wave travelling radially outward, which is what "reaches infinity" has to mean.

Step 2. Compute the invariants and compare. With $E = cB$ , $\vec E\perp\vec B$ : F_{\alpha\beta}F^{\alpha\beta} = 2\!\left(B^2 - \frac{E^2}{c^2}\right) = 0, \qquad F_{\alpha\beta}\tilde F^{\alpha\beta} = -\frac{4}{c}\,\vec E\cdot\vec B = 0 . The $1/R^2$ velocity field, by contrast, is a boosted Coulomb field with $F_{\alpha\beta}F^{\alpha\beta} < 0$ and $F_{\alpha\beta}\tilde F^{\alpha\beta} = 0$ . Why: both invariants vanish for radiation — a null field — and that is frame-independent. No boost turns a radiation field into a pure electric or pure magnetic one. But the velocity field can always be boosted to a pure electrostatic field: it is a Coulomb field in disguise. That is the invariant statement of "this is not radiation". Verified in verify_aq-sr-lienard-wiechert-potentials.py: for $\vec E\perp\vec B$ with $E = cB$ both invariants vanish identically; for a boosted Coulomb field the first is strictly negative and the second exactly zero.*

Limiting cases

Case	Result	What it says
Charge at rest	$A^\mu \to (\phi/c, \vec 0)$ , $\phi = q/4\pi\varepsilon_0 R$	(1) reduces to Coulomb — the retardation is invisible
Uniform velocity	velocity field only, $\propto 1/R^2$	the boosted Coulomb field (`aq-sr-field-uniformly-moving-charge`). No radiation
$\vec a \parallel \vec v$ (linac)	$P \propto \gamma^6\,\dot\beta^2$	the cross product vanishes; nothing cancels
$\vec a \perp \vec v$ (ring)	$P \propto \gamma^4\,\dot\beta^2$	the bracket eats $\gamma^2$ . This is what killed LEP
$\gamma\to1$ , $\vec a\parallel\vec v$	$P = q^2a^2/6\pi\varepsilon_0c^3$	Larmor, coefficient exactly 1
Ultrarelativistic	emission within $\theta \sim 1/\gamma$	the searchlight cone: $\kappa^{-3}$ beams it forward
Far zone, $1/R$ term	both invariants $= 0$	a null field — the invariant signature of radiation
Protons, not electrons	$P$ down by $(m_p/m_e)^4 = 1.14\times10^{13}$	why the LEP tunnel now holds the LHC

Worked examples

1 · LEP — reading the exponent straight off the logbook

LEP lost 140 MeV per turn at 47 GeV per beam and 2330 MeV per turn at 95 GeV. Determine the exponent of $\gamma$ in the synchrotron loss law using nothing but these two numbers, then check against the formula.

Step 1 — the model-free test. If $\Delta E \propto E^n$ then n = \frac{\ln(2330/140)}{\ln(95/47)} = \frac{2.8121}{0.70371} = \mathbf{3.996} . The exponent is 4, to one part in a thousand, and no theory was used to get it. Equivalently: the $\gamma^4$ law demands $(95/47)^4 = \mathbf{16.69}$ and the logbook says $2330/140 = \mathbf{16.64}$ — 0.3%. A $\gamma^6$ ring would have demanded $\mathbf{68.2}$ .

Step 2 — the absolute numbers. With $\Delta E[\mathrm{keV}] = 88.46\,E[\mathrm{GeV}]^4/\rho[\mathrm{m}]$ and $\rho \approx 3096\ \mathrm{m}$ :

$E$ (GeV/beam)	predicted	published
47	$\mathbf{139.4\ MeV}$	140 MeV
95	$\mathbf{2327\ MeV}$	2330 MeV
104.5	$\mathbf{3407\ MeV}$	$\approx$ 3.4 GeV

Three rows, one formula, no free parameters. At 104.5 GeV the loss is about 3% of the beam energy — every lap, and the RF must replace all of it at a cost scaling as $E^4$ . That, not any wall of principle, is why LEP stopped where it did.

Step 3 — why protons. The loss goes as $\gamma^4 = (E/mc^2)^4$ , so at equal energy the proton/electron ratio is $(m_e/m_p)^4 = 1/\mathbf{1.14\times10^{13}}$ . Check it: the LHC runs 7 TeV protons in essentially the same tunnel ( $\rho = 2803.95\ \mathrm{m}$ ), and $\gamma_p = 7000/0.938 = 7461$ , giving U_0 = \frac{q^2\gamma^4\beta^3}{3\varepsilon_0\rho} = \mathbf{6.66\ keV\ per\ turn} \quad (\text{published:}\ \approx6.7\ \mathrm{keV}). Six kilo-electronvolts, against LEP's two and a third giga-electronvolts — same tunnel, same bending radius, heavier particle. The LHC exists because of the fourth power of a mass ratio.

2 · The searchlight, and why a synchrotron is a light source

An electron at $\gamma = 1000$ (about 511 MeV) is bent by a magnet. Into what cone does it radiate?

Step 1 — the beaming angle. The angular distribution of the acceleration field carries $\kappa^{-3}$ , so emission is sharply peaked forward, with cone half-angle \theta \simeq \frac{1}{\gamma} = \frac{1}{1000} = \mathbf{1.0\ mrad} = 0.057^\circ . All of the radiation comes out in a cone a milliradian wide, tangent to the orbit.

Step 2 — where it comes from. $\kappa = 1 - \beta\cos\theta \approx \tfrac12(1/\gamma^2 + \theta^2)$ at $\beta\to1$ : tiny for $\theta \lesssim 1/\gamma$ , growing fast beyond. $\kappa$ cubed in the denominator is a very sharp searchlight — and note once more that $\kappa$ is the retarded-time Jacobian, not $\gamma$ : the beaming is a retardation effect, a source chasing its own emitted wavefront.

Step 3 — read it as an instrument. A charge going in a circle sweeps this searchlight round like a lighthouse, and a detector on a tangent sees a brief flash each turn. A short pulse in time means a broad spectrum in frequency — which is why synchrotrons, built as an unwanted energy loss, became the brightest laboratory X-ray sources on Earth (`aq-sr-synchrotron-radiation`).

:::svg[γ⁶ IS THE LINAC. γ⁴ IS THE RING. AND THE LEP LOGBOOK PROVES IT. Liénard's bracket is [β̇² − (β × β̇)²]. When the acceleration is PARALLEL to the velocity — a linear accelerator — the cross product is zero, nothing cancels, and the power goes as γ⁶. When the acceleration is PERPENDICULAR — a storage ring bending its beam — the cross product is β·β̇, the bracket collapses to β̇²(1 − β²) = β̇²/γ², and TWO POWERS OF γ ARE EATEN: the power goes as γ⁴. The published LEP numbers settle it with no theory at all: fit ΔE ∝ Eⁿ to the two measured points (140 MeV at 47 GeV, 2330 MeV at 95 GeV) and you get n = 3.996. A γ⁶ ring would have predicted a ratio of 68.2 where 16.64 was measured.]

Common trapThis article's base gets Liénard's power DIMENSIONALLY WRONG — and dimensional analysis catches it in five seconds.

The AllQFT base for this article prints

P = \frac{\mu_0 q^2}{6\pi c}\,\gamma^6\Big[\dot{\vec\beta}^{\,2} - \big(\vec\beta\times\dot{\vec\beta}\big)^2\Big] \qquad \textbf{— WRONG, by a factor of } c^2 .

The error is that $\dot{\vec\beta}$ is a rate, not an acceleration. $\dot{\vec\beta} = \mathrm{d}(\vec v/c)/\mathrm{d}t$ has units $\mathrm{s^{-1}}$ , so the prefactor must supply $\mathrm{J\,s}$ to make watts — and $\mu_0q^2/(6\pi c)$ supplies $\mathrm{N\,s\,m^{-1}}$ instead. It is not a power. The correct prefactor puts $c$ in the numerator:

\frac{\mu_0 q^2 c}{6\pi} \;=\; \frac{q^2}{6\pi\varepsilon_0 c} \qquad\Longrightarrow\qquad \gamma\to1,\ \vec a\parallel\vec v:\quad P = \frac{q^2a^2}{6\pi\varepsilon_0c^3} \quad\text{✓ exactly Larmor.}

Two independent checks — dimensions, and the $\gamma\to1$ limit — kill the broken form instantly, and neither was run. ⚠ Sweep it to the source. The same $c^2$ slip is in the served text of the live `aq-em-relativistic-larmor-formula`, which prints $P = \tfrac{\mu_0q^2\gamma^6}{6\pi c}[\dot{\vec\beta}^2 - (\vec\beta\times\dot{\vec\beta})^2]$ — while its own covariant form $P = -\tfrac{\mu_0q^2}{6\pi c}a^\mu a_\mu$ (dimensionally right, since $a^\mu$ is a four-acceleration) with its own $a^\mu a_\mu = -\gamma^6c^2[\dot{\vec\beta}^2 - (\vec\beta\times\dot{\vec\beta})^2]$ implies the correct $\mu_0q^2c/6\pi$ . That article contradicts itself by $c^2$ , exactly as this base does. Flagged for repair; this page prints the correct form.

Common trapThe (1 − R̂·β) factor is a JACOBIAN. It is not time dilation, and calling it that will make you wrong.

The factor $\kappa = 1 - \hat R\cdot\vec\beta$ that sits, cubed, in the denominator of every Liénard–Wiechert field is the most-misattributed quantity in electrodynamics. It is not $\gamma$ and not a time-dilation factor. It is a Jacobian: $\partial t_{\text{ret}}/\partial t = 1/\kappa$ . As the observer's clock ticks forward by $\mathrm{d}t$ , the retarded time — the instant on the worldline the observer is currently "seeing" — advances by only $\mathrm{d}t/\kappa$ , because the charge moved during the light-travel interval and the cone must catch up with it. A chasing effect, first order in $\beta$ .

The test that settles it. At $\beta = 0.01$ , $\gamma = 1.00005$ — utterly negligible. But $\kappa$ ranges from $0.99$ to $1.01$ depending on direction: a 1% effect, two hundred times larger. If $\kappa$ were a time-dilation factor it would be direction-independent and second order in $\beta$ . It is neither. And the cleanest proof is in the covariant form: $u_\nu(x-w)^\nu = \gamma cR\,\kappa$ , so the $\gamma$ from $u^\mu$ cancels exactly against the $\gamma$ here. What survives into the three-vector potentials is $\kappa$ ; the $\gamma$ 's have gone.

Physics of the portrait

The portrait is the formula that turned an inconvenience into an industry.

Origins · 2 minRelativistic Larmor formula and how this was discovered

Relativistic Larmor formula

1897–1898 · the power an accelerating charge radiates · a Lorentz invariant · the γ⁴ that stopped LEP and the γ⁶ that did not

Joseph Larmor obtained the non-relativistic radiated power in 1897 — Philosophical Magazine, Series 5, volume 44, number 271, pages 503–512, with the result on p. 512 — for a charge whose speed is small: $P = q^2a^2/6\pi\varepsilon_0c^3$ in modern SI, or $\tfrac23 e^2a^2/c^3$ in the Gaussian units he used. One year later Alfred-Marie Liénard, in a three-part paper in L'Éclairage Électrique (1898, pp. 5–14, 53–59, 106–112), generalised it to arbitrary speed and obtained the $\gamma^6$ bracket — seven years before special relativity existed. Liénard did not know he was writing a relativistic formula; he simply solved Maxwell's equations for a point charge honestly, and Maxwell's equations were already relativistic. Emil Wiechert reached the potentials in 1900 (Archives Néerlandaises; also Annalen der Physik, 1901), on the standard account independently of Liénard — though that independence is historical convention rather than established fact. What makes the formula relativistic is that the total radiated power is a Lorentz invariant, visible only when it is written covariantly, $P = -\tfrac{\mu_0q^2}{6\pi c}\,a^\mu a_\mu$ . The practical consequence took fifty more years to bite. In April 1947, technicians at General Electric in Schenectady, running a 70 MeV electron synchrotron behind a transparent shield, saw a brilliant arc of blue-white light coming off the beam: the loss Liénard had predicted, seen for the first time. To particle physicists it was a catastrophe — it is precisely what forced LEP to stop at 104.5 GeV and hand its tunnel to protons. To everyone else it was a gift. That unwanted blue glow is now the brightest laboratory X-ray source on Earth.

History

1897

— Larmor gives the radiated power of a slowly moving accelerated charge (Phil. Mag. Ser. 5, 44 (271), 503–512, p. 512).

1898

— Liénard generalises Larmor to arbitrary speed in a three-part paper (L'Éclairage Électrique, pp. 5–14, 53–59, 106–112), obtaining the γ⁶ bracket seven years before relativity.

1900

— Wiechert publishes the retarded potentials of a point charge (Archives Néerlandaises; also Ann. Phys. 1901); on the standard account independently of Liénard.

1908

— Minkowski's four-vector formalism makes the retarded solution manifestly covariant and the radiated power manifestly invariant.

1947

— Synchrotron radiation is seen for the first time, as a visible arc of light from the 70 MeV electron synchrotron at General Electric, Schenectady.

1989

— LEP begins operation; by 2000 it reaches about 104.5 GeV per beam, where synchrotron loss consumes roughly 3% of the beam energy every turn.

Going further

Texts. Griffiths, Introduction to Electrodynamics, 4th ed., §10.3, §11.2 · Jackson, Classical Electrodynamics, 3rd ed., §14.1–14.3 · Landau & Lifshitz, The Classical Theory of Fields, 4th ed., §63–74.

Provenance. Signature $(+,-,-,-)$ , hence $p\cdot p = +m^2c^2$ ; $m$ is the invariant rest mass. $u^\mu = \gamma(c,\vec v)$ , $a^\mu = \mathrm{d}u^\mu/\mathrm{d}\tau$ . This article's base prints Liénard's power with the prefactor $\mu_0q^2/(6\pi c)$ , dimensionally wrong by $c^2$ — it yields $\mathrm{N\,s\,m^{-1}}$ , not watts. The corrected $\mu_0q^2c/6\pi = q^2/(6\pi\varepsilon_0c)$ is derived, dimensionally checked and machine-asserted here. ⚠ The identical $c^2$ slip is present in the served text of the live `aq-em-relativistic-larmor-formula`, whose own covariant form and own $a^\mu a_\mu$ contradict its own printed 3-vector Liénard formula. Flagged for repair at source. No claim is made on this page that Larmor's original paper contained an arithmetic error — that story is an undocumented myth, and the primary scan shows the correct coefficient. All collider numbers are recomputed in `mpmath` at 50 digits against this article's printed digits.

Scope. ⚠ The electromagnetism track KEEPS the Liénard–Wiechert derivation. The three-vector potentials and fields, the $\kappa^{-3}$ angular structure and the velocity/acceleration decomposition are owned by `aq-em-lienard-wiechert-potentials` (EMAG.RAD.LW.0610), the covariant Larmor invariant by `aq-em-relativistic-larmor-formula` (EMAG.RAD.RLARM.0617). This page owns what is distinctively relativistic: the invariant Green function; the covariant four-potential (1); $\kappa$ as a Jacobian; the invariant character of the radiation field; and Liénard's power as a Lorentz-invariant statement with the collider numbers. The spectrum is `aq-sr-synchrotron-radiation`.

Builds on. `aq-sr-wave-equation-four-potential` · `aq-sr-four-potential` · `aq-sr-four-current-density` · `aq-sr-gauge-transformations-lorenz-gauge` · `aq-sr-charge-invariance-continuity`. Leads to. `aq-sr-field-uniformly-moving-charge` · `aq-sr-electromagnetic-field-invariants` · `aq-sr-covariant-lorentz-force`.

Why is the retarded solution chosen over the advanced one? Is that not an extra assumption?

It is, and it is worth being honest about it. $\Box G = \delta^4$ has both a retarded and an advanced solution, and the wave equation prefers neither. Choosing $\theta(t - t')$ encodes the arrow of time, not the geometry of spacetime; Feynman and Wheeler famously built a consistent electrodynamics on half retarded plus half advanced. The wave equation gives you a cone; it does not tell you which way to go round it. What relativity does guarantee is that whichever you pick, every observer agrees — because the time-order of null-separated events is invariant.

Is the factor κ = 1 − R̂·β really not a relativistic effect at all?

It is a retardation effect, which is a different thing from a relativistic one, though both involve the finite speed of light. $\kappa$ appears at first order in $\beta$ and is direction-dependent; time dilation appears at second order and is not. You would get $\kappa$ in a Galilean world with a finite signal speed — it is the geometry behind the ordinary Doppler shift of a moving siren. Retardation gives you $\kappa$ ; relativity tells you what to do with it — namely, that the resulting potential assembles into a four-vector and the radiated power is an invariant.

Why does a uniformly moving charge not radiate, if its field is changing at every point in space?

Because "the field is changing here" is not what radiation means. Radiation means energy irreversibly delivered to infinity, and that requires the Poynting flux through a sphere to survive $R\to\infty$ . The velocity field falls as $1/R^2$ , so its flux falls as $1/R^4$ against an area growing as $R^2$ : the power through the sphere goes to zero. Its energy is carried along with the charge, like the bow wave of a boat — sloshing, never leaving. It takes an acceleration to produce a $1/R$ field, and only a $1/R$ field can escape.

Is the total radiated power really the same in every frame? That seems to contradict the Doppler effect.

It does not. $P = \mathrm{d}W/\mathrm{d}t$ is a Lorentz invariant because $W$ and $t$ transform the same way — the radiated energy is the time component of a four-momentum, divided by the time component of a four-vector, so the ratio is a scalar. Covariantly this is obvious: $P = -\tfrac{\mu_0q^2}{6\pi c}a^\mu a_\mu$ . What is not invariant is the angular distribution and the frequency spectrum — which is exactly the Doppler effect and the searchlight beaming. They agree on the total; they disagree on everything else about it.

Why does a storage ring suffer γ⁴ and a linac γ⁶? Which is worse?

The exponents come from the geometry, not the machine: $\gamma^6$ when $\vec a\parallel\vec v$ (a linac), $\gamma^4$ when $\vec a\perp\vec v$ (a ring). $\gamma^6$ is the larger exponent, so it sounds worse — but it is not, and this is the point most often garbled. In a linac the achievable gradient is so small that $\dot{\vec\beta}$ is minuscule and the radiated power is negligible at any energy anyone can build. In a ring the bending is unavoidable — you must turn the beam through $2\pi$ every lap — so $\dot{\vec\beta}$ is large. The $\gamma^4$ term with a big $\dot\beta$ beats the $\gamma^6$ term with a tiny one, by a mile. Linear colliders exist to escape the ring's $\gamma^4$ bending loss, not because $\gamma^6$ is tamer.

How do I get the Coulomb field back from the covariant potential?

Put the charge at rest: $u^\mu = (c, \vec 0)$ , so on the light cone $u_\nu(x-w)^\nu = cR$ and A^\mu = \frac{\mu_0c}{4\pi}\cdot\frac{q\,(c,\vec 0)}{cR} = \left(\frac{q}{4\pi\varepsilon_0 cR},\ \vec 0\right), which is $A^\mu = (\phi/c, \vec 0)$ with $\phi = q/4\pi\varepsilon_0 R$ — Coulomb, exactly, using $\mu_0c^2 = 1/\varepsilon_0$ . The retardation is invisible because a static charge was in the same place at the retarded time as it is now.