Larmor Formula

The previous article ended with a verdict: charges radiate if and only if they accelerate, and the part of the field that does the radiating — the acceleration field — falls as $1/r$ , stays transverse, and reads the acceleration at the retarded time. What it deliberately did not do is present the bill. This article computes it: integrate the acceleration field's Poynting flux over a distant sphere and out comes the single most quoted result in radiation theory, $P=\mu_0q^2a^2/6\pi c$ — Larmor's formula, the wattage of a wiggled charge. Its structure repays a slow read. The power depends on the charge and the acceleration squared — not on the velocity at all, in this nonrelativistic limit — and the numbers it produces span nature's whole dynamic range: an electron accelerated at $9.8$ m/s² radiates $10^{-51}$ watts (gravity will never light a bulb), while the same electron in an atomic-scale electric field radiates enough to collapse classical hydrogen in sixteen picoseconds. That last number is not a fringe application; it is the formula's most consequential output ever — the quantitative proof that classical physics cannot hold at the atomic scale, the crisis that demanded quantum mechanics. Two derivations share the work. The flagship integrates the Poynting flux honestly, inheriting the acceleration field from the point-charge article. The second is a jewel of a shortcut: J.J. Thomson's geometric construction, which extracts the same $1/r$ transverse field from nothing but a kinked field line and similar triangles — the previous article's picture, made quantitative. The formula's own limits are marked honestly too: it is the $v\ll c$ anchor, and its relativistic completion — where $\gamma^4$ and $\gamma^6$ enhancements change accelerator design — is the next article's whole subject.

Key takeaways

Only the acceleration ( $1/r$ ) part of a point charge's field carries energy to infinity; integrating its Poynting flux gives the Larmor formula $P=\dfrac{\mu_0q^2a^2}{6\pi c}=\dfrac{q^2a^2}{6\pi\varepsilon_0c^3}$ .
The angular distribution is $dP/d\Omega\propto\sin^2\theta$ about the acceleration direction: zero along $\mathbf{a}$ , maximal broadside — the universal dipole donut.
Velocity does not appear: for $v\ll c$ a charge's radiated power depends only on how hard it is being shoved, not how fast it is moving.
Thomson's kink construction derives the same radiation field geometrically: the transverse-to-radial field ratio in the expanding kink shell is $a\,r\sin\theta/c^2$ , forcing $E_t\propto1/r$ .
Flagship numbers: classical hydrogen collapses by radiation in $1.6\times10^{-11}$ s — the formula that broke classical physics; an electron cyclotroning in 1 T cools radiatively in $2.6$ s.
The formula is the nonrelativistic anchor: the exact-any-speed generalization (Liénard's $\gamma^6$ result) is the next article, and connecting to it goes through the invariance of $P$ .
Key idea: radiated power equals $\mu_0q^2a^2/6\pi c$ — acceleration squared, velocity ignored, direction pattern $\sin^2\theta$ — because only the $1/r$ acceleration field survives the trip to infinity.

The core insight

Only the part of a point charge's field that falls off as $1/r$ — the acceleration field — carries energy to infinity; its Poynting flux integrated over a sphere gives $P=\mu_0q^2a^2/6\pi c$ .

Where this lives today

Every machine that makes X-rays runs on this formula's ledger. In an X-ray tube, electrons slam into tungsten and decelerate violently — bremsstrahlung, "braking radiation," is Larmor applied to $a\sim10^{23}$ m/s² stopping events, and the tube's continuous spectrum is the chirp of a hundred million kinks per second. Medical linacs, airport scanners, synchrotron light sources (with the relativistic corrections of the next article), and the free-electron lasers that now film chemistry in femtoseconds are all engineering exercises in arranging large $a^2$ on demand. In plasmas the formula runs the other way: thermal bremsstrahlung — electrons Larmor-radiating as they swerve past ions — is a dominant energy-loss channel in fusion devices (where high- $Z$ impurities are scrubbed because the loss scales with the ion charge squared) and the X-ray glow by which astronomers weigh galaxy-cluster gas. The formula also patrols a frontier: what it does not predict keeps mattering. The equivalence-principle puzzle — does a charge sitting in a gravitational field radiate? does one falling freely? — turns on exactly which acceleration Larmor should read, and its resolution (radiation is not a local, frame-free fact; the answer depends on the observer's horizon structure) feeds directly into Unruh radiation and black-hole thermodynamics. And the energy Larmor exports must, by the conservation chapter's books, be debited from the particle's motion: the back-force that does the debiting — radiation reaction, with its infamous pathologies — is two articles ahead, with the invariant completion of the formula itself next.

Origins · 1 minJoseph John Larmor and how this was discovered

Joseph John Larmor

1857–1942 · Cambridge — Lucasian professor who priced the wiggling electron

Larmor derived his power formula in 1897 — the same year J.J. Thomson (no relation between the surnames' owners beyond a shared Cambridge) identified the electron — as part of his program Aether and Matter, an attempt to build all of physics from singularities in the ether. The formula appeared en route to explaining how atomic "vibrators" could emit light, and alongside it Larmor gave the precession of orbiting charges in magnetic fields (the Larmor precession of every MRI machine) and, remarkably, wrote down the time transformation $t'=t-vx/c^2$ with the full Lorentz factor before Lorentz's 1904 paper — one of relativity's several half-credited anticipations. He held the Lucasian chair (Newton's, later Dirac's and Hawking's) for twenty-nine years, edited the collected works of Stokes and Kelvin, and sat in Parliament for Cambridge University. History's little irony: the man who computed how violently classical orbits must radiate lived to see that very number — sixteen picoseconds for hydrogen — become Exhibit A for the quantum mechanics that retired his beloved ether.

History

1897

— Larmor publishes the radiated-power formula for an accelerated charge; Thomson identifies the electron the same year

1898

— Liénard generalizes the result to arbitrary velocity (the next article's subject)

1904

— Thomson's kink construction gives the radiation field geometrically, without potentials

1909

— Rutherford's scattering experiments make the nuclear atom — and its classical collapse problem — unavoidable

1913

— Bohr postulates stationary states: Larmor radiation is switched off by fiat where quantum rules hold

1929

— X-ray tube spectra fully understood as bremsstrahlung: Larmor's kinks, industrialized

1947

— General Electric team observes visible synchrotron light, opening the relativistic chapter of the formula

1985

— single trapped electrons observed radiatively cooling at the Larmor rate — the formula verified one particle at a time

The physical picture

Start from the inventory the point-charge article left us. The full field of a moving charge splits into a velocity field — the boosted Coulomb field, $1/r^2$ , pinned to the charge, carrying no net energy away — and an acceleration field: transverse, proportional to the acceleration at the retarded time, falling as $1/r$ . The $1/r$ is everything. Energy flux goes as the field squared, $1/r^2$ ; the area of a distant sphere grows as $r^2$ ; their product is constant. Whatever power crosses the sphere at one radius crosses every larger sphere too — the energy is not spreading and thinning into nothing, it is leaving. Radiated power is well defined precisely because, and only because, some part of the field falls as slowly as $1/r$ . Why acceleration squared? The radiation field is linear in $a$ — twice the shove, twice the transverse field — and power is quadratic in field. Why no velocity? Uniform motion is relativity's freebie: a charge coasting at constant $\mathbf{v}$ is a charge at rest viewed from a moving frame, and a static Coulomb field radiates nothing in any frame. Only the change of velocity is news, and the news is billed quadratically. The angular pattern completes the picture. The radiation field is the acceleration projected on the sky — the component of $\mathbf{a}$ transverse to the line of sight — so it vanishes along $\mathbf{a}$ ( $\sin\theta=0$ : looking down the barrel of the acceleration, there is no transverse component to see) and peaks broadside. The intensity donut $\sin^2\theta$ wraps the acceleration axis, and — worth engraving — it is oriented by $\mathbf{a}$ , not by $\mathbf{v}$ : a charge moving along $x$ while shoved along $y$ radiates a donut around $y$ .

Figure

The energy audit at infinity — velocity field stays home, acceleration field pays out at a rate independent of the sphere's radius

The math: from acceleration field to wattage

The input is inherited, by name, from Fields of a Moving Point Charge: in the nonrelativistic limit ( $\beta\to0$ , so the Doppler-like $(1-\hat{\mathbf{n}}\cdot\boldsymbol\beta)$ factors go to one) the acceleration field is

\mathbf{E}_{\text{rad}}=\frac{q}{4\pi\varepsilon_0c^2}\,\frac{\hat{\mathbf{n}}\times(\hat{\mathbf{n}}\times\mathbf{a})}{r}\Big|_{t_r},\qquad \mathbf{B}_{\text{rad}}=\frac{1}{c}\,\hat{\mathbf{n}}\times\mathbf{E}_{\text{rad}}. \qquad (1)

where: $\hat{\mathbf{n}}$ is the line of sight from the (retarded) charge position, and the double cross product is a projector — it strips $\mathbf{a}$ of its longitudinal component and leaves $-\mathbf{a}_\perp$ , the acceleration as seen on the sky, magnitude $a\sin\theta$ . The triad $(\mathbf{E},\mathbf{B},\hat{\mathbf{n}})$ is right-handed with $E=cB$ : locally, the radiation field is a plane wave headed out. Everything in this article follows from (1) plus the Poynting machinery of the conservation chapter.

Derivation AThe Larmor formula from the Poynting flux of the acceleration field

The goal is the flagship result: total power $P=\mu_0q^2a^2/6\pi c$ crossing any large sphere, from the fields (1). The strategy is the chapter's standard energy audit — form the Poynting vector, express the flux per solid angle, watch the $r$ 's cancel (the signature of genuine radiation), then do the one angular integral. Nothing is subtle except keeping track of which constants combine into $\mu_0$ ; the physics was all in earning (1).

Step 1. Magnitude of the radiation field. The projector gives $|\hat{\mathbf{n}}\times(\hat{\mathbf{n}}\times\mathbf{a})|=a\sin\theta$ with $\theta$ measured from $\mathbf{a}$ :

E_{\text{rad}}=\frac{q\,a\sin\theta}{4\pi\varepsilon_0c^2\,r}.

Step 2. Poynting flux. For the plane-wave-like triad, $\mathbf{S}=\tfrac{1}{\mu_0}\mathbf{E}\times\mathbf{B}$ points along $\hat{\mathbf{n}}$ with magnitude $E^2/\mu_0c$ ; using $1/\varepsilon_0^2c^4\cdot(1/\mu_0 c)=\mu_0/c$ (via $\mu_0\varepsilon_0=1/c^2$ , the constants' one identity worth memorizing):

S=\frac{E_{\text{rad}}^2}{\mu_0c}=\frac{\mu_0\,q^2a^2\sin^2\theta}{16\pi^2c\,r^2}.

Why checking the $r$ -dependence here matters: $S\propto1/r^2$ exactly, so $r^2S$ is $r$ -independent — the power through a sphere does not care which sphere. Any leftover $r$ would mean a bookkeeping error or a non-radiative term smuggled in.

Step 3. Power per solid angle, and the donut:

\frac{dP}{d\Omega}=r^2S=\frac{\mu_0q^2a^2}{16\pi^2c}\,\sin^2\theta.

Step 4. The angular integral. With $\oint\sin^2\theta\,d\Omega=2\pi\int_0^\pi\sin^3\theta\,d\theta=2\pi\cdot\tfrac43=\tfrac{8\pi}{3}$ :

P=\frac{\mu_0q^2a^2}{16\pi^2c}\cdot\frac{8\pi}{3}=\boxed{\frac{\mu_0q^2a^2}{6\pi c}}=\frac{q^2a^2}{6\pi\varepsilon_0c^3}.

Why the two SI spellings: $\mu_0/6\pi c$ and $1/6\pi\varepsilon_0c^3$ are the same coefficient — worth recognizing both, since references split evenly. In Gaussian units the same physics reads $P=2q^2a^2/3c^3$ . *Verified: matches Griffiths Eq. 11.70 and Jackson Eq. 14.22; dimensional audit $[\mu_0q^2/c]=$ kg·s and $[a^2]=$ m²/s⁴ gives watts; the $\sin^2\theta$ pattern and $8\pi/3$ integral match every dipole calculation in the chapter — as they must, since a wiggling charge is an infinitesimal dipole antenna ( $\ddot{\mathbf{p}}=q\mathbf{a}$ , and substituting into the dipole power $\mu_0\langle\ddot p^2\rangle/6\pi c$ reproduces Larmor verbatim).*

Go deeper · Derivation B · 2 minThomson's kink construction — the radiation field from similar triangles

Derivation BThomson's kink construction — the radiation field from similar triangles

The goal is the same $1/r$ transverse field, derived with no Liénard–Wiechert machinery at all: J.J. Thomson's 1904 construction, which turns the previous article's kink picture into a two-line quantitative argument. The setup: a charge at rest is given a brief kick — acceleration $a$ for a short time $\tau$ — reaching the tiny velocity $v=a\tau$ , then coasts. Look at the field a long time $t$ later. The strategy: field lines must be continuous; inside the sphere $r=ct$ the news has arrived and lines point away from the moved charge, outside they still point away from the original position; the kink connecting them, a shell of thickness $c\tau$ , contains a transverse field whose size pure geometry fixes.

Step 1. The geometry of the mismatch. At time $t$ , the charge has moved $\approx vt$ (the kick displacement $\tfrac12a\tau^2$ is negligible for $\tau\ll t$ ). A field line leaving the charge at angle $\theta$ from the eventual velocity must, across the kink shell at radius $r=ct$ , jog sideways by the transverse projection of the displacement: $vt\sin\theta$ .

Step 2. Similar triangles in the shell. Within the shell of radial thickness $c\tau$ the line runs at a slant; the ratio of its transverse to radial field components equals the ratio of the jog to the thickness:

\frac{E_t}{E_r}=\frac{v\,t\sin\theta}{c\,\tau}=\frac{a\,t\sin\theta}{c}=\frac{a\,r\sin\theta}{c^2},

using $v=a\tau$ (the $\tau$ 's cancel — the kick's duration drops out, as it must for a clean point result) and $t=r/c$ .

Step 3. Feed in the radial field. Gauss's law fixes $E_r=q/4\pi\varepsilon_0r^2$ through the shell (the un-updated Coulomb value — the shell hasn't heard otherwise). Multiply:

E_t=\frac{a\,r\sin\theta}{c^2}\cdot\frac{q}{4\pi\varepsilon_0r^2}=\frac{q\,a\sin\theta}{4\pi\varepsilon_0c^2\,r}.

One factor of $r$ from the growing mismatch cancels one of Coulomb's two — and the field of equation (1) appears, coefficient and all, from triangles.

Step 4. Read the construction physically. The transverse field falls as $1/r$ because the kink must bridge a mismatch that grows linearly with distance while its thickness stays fixed at $c\tau$ — the slant gets steeper forever. The $\sin\theta$ is the projection of the displacement on the sky. And squaring $E_t$ , multiplying by the shell volume $4\pi r^2\cdot c\tau\cdot(\text{angular average})$ and dividing by the kick duration reproduces the Larmor power — energy $\tfrac{\mu_0q^2a^2\tau}{6\pi c}$ deposited in the shell per kick, i.e. $P=\mu_0q^2a^2/6\pi c$ while the kick lasts. Verified: reproduces Purcell & Morin Appendix H (the classic presentation of Thomson's construction); coefficient identical to Derivation A's Step 1 — two independent routes, one field.

Figure

Thomson's construction — the kinked field line, and the similar triangles that price the radiation

Limiting cases

case	result	meaning
$a=0$ (any constant $\mathbf{v}$ )	$P=0$	uniform motion is a boosted static field; no news, no bill
$v\ll c$ , any $\mathbf{a}$	$P=\mu_0q^2a^2/6\pi c$ , donut about $\mathbf{a}$	this article: velocity-blind, acceleration-squared
harmonic wiggle $x_0\cos\omega t$	$\langle P\rangle=\mu_0q^2x_0^2\omega^4/12\pi c$	$a=\omega^2x_0$ : the $\omega^4$ of dipole radiation, recovered per-charge
collection of slow charges	$P=\mu_0\langle\ddot p^2\rangle/6\pi c$ , $\mathbf{p}=\sum q_i\mathbf{r}_i$	Larmor is dipole radiation when charges cooperate
relativistic, $\mathbf{a}\parallel\mathbf{v}$	$P\to\gamma^6\times$ Larmor form	next article: Liénard's completion
relativistic, $\mathbf{a}\perp\mathbf{v}$	$P\to\gamma^4\times$ Larmor form	synchrotron regime — geometry decides the exponent
Gaussian units	$P=2q^2a^2/3c^3$	same physics, different clothing; conversion via $\mu_0/4\pi\to1/c^2$

Common trapThe donut wraps the acceleration, not the velocity — and Larmor with lab-frame a is wrong the moment v approaches c

Two standing errors. First, orientation: students reflexively draw the radiation pattern around the direction of motion. Wrong axis — equation (1) knows only $\mathbf{a}$ . A charge orbiting in a magnetic field has $\mathbf{a}$ pointing at the center, so at each instant the (nonrelativistic) donut points its null at the orbit's center, not along the track. Second, jurisdiction: Larmor is the $v\ll c$ limit, full stop. Feed it the lab-frame acceleration of a relativistic particle and you undercount catastrophically — circular motion at Lorentz factor $\gamma$ radiates $\gamma^4$ times more than the naive Larmor estimate, a factor of $10^{16}$ for LEP's electrons. The honest chain: use Larmor in the particle's instantaneous rest frame, then let the invariance argument of the next article export it to the lab. The formula is an anchor, not a universal law.

Worked examples

1 · An electron in a 1 tesla field — radiative cooling you can time with a stopwatch (confidence builder)

Setup. An electron moves at $v=0.01c=3.0\times10^6$ m/s perpendicular to a uniform $B=1$ T field (a Penning-trap or storage-ring scale field). Find the radiated power and the timescale for the motion to decay.

Step 1. The acceleration. The magnetic force supplies $a=\dfrac{evB}{m}=\dfrac{(1.60\times10^{-19})(3.0\times10^6)(1)}{9.11\times10^{-31}}=5.3\times10^{17}\ \text{m/s}^2$ — seventeen orders of magnitude beyond gravity, routine for an electron.

Step 2. The power:

P=\frac{\mu_0e^2a^2}{6\pi c}=\frac{(1.26\times10^{-6})(1.60\times10^{-19})^2(5.3\times10^{17})^2}{6\pi(3.0\times10^8)}=\mathbf{1.6\times10^{-18}\ W}.

Step 3. The cooling time. Kinetic energy $K=\tfrac12mv^2=4.1\times10^{-18}$ J, so $\tau=K/P\approx\mathbf{2.6\ s}$ . Better: since $a\propto v$ , $P\propto v^2\propto K$ , so $\dot K=-K/\tau$ with constant $\tau$ — the energy decays exponentially, $\tau=3\pi\varepsilon_0m^3c^3/e^4B^2=2.6$ s at 1 T, independent of the starting speed.

Step 4. Reality check. This is measurable, mundane physics: single electrons in Penning traps are watched radiating their cyclotron energy away on exactly this timescale (at 5 T, sixteen times faster), and the exponential decay of trapped-electron cyclotron motion is a standard calibration. Larmor is not an abstraction; it is a lab clock.

The lesson: compute $a$ from the force, square it, and the formula prices the energy leak — with the exponential-decay shortcut whenever $P\propto K$ .

2 · The sixteen-picosecond catastrophe — classical hydrogen radiates itself to death (exam level)

Setup. In the classical planetary atom, an electron circles the proton at radius $r$ . Show the orbit decays by radiation and compute the time to spiral from $r_0=a_0=5.29\times10^{-11}$ m into the nucleus. This is the calculation that ended classical physics' jurisdiction over the atom.

Step 1. Orbit energetics. Coulomb attraction supplies the centripetal force, giving the virial bookkeeping $K=-E$ and total energy $E(r)=-\dfrac{e^2}{8\pi\varepsilon_0r}$ ; the inward acceleration is $a=\dfrac{e^2}{4\pi\varepsilon_0mr^2}$ .

Step 2. Radiated power at radius $r$ . Feeding $a$ into Larmor:

P(r)=\frac{\mu_0e^2}{6\pi c}\left(\frac{e^2}{4\pi\varepsilon_0mr^2}\right)^2.

At $r=a_0$ : $a=9.0\times10^{22}$ m/s², $P=4.7\times10^{-8}$ W — colossal against the $2.2\times10^{-18}$ J binding scale. The instantaneous ratio $|E|/P\approx4.7\times10^{-11}$ s already announces the disaster's timescale.

Step 3. The spiral. Energy balance $dE/dt=-P$ with $E(r)$ from Step 1 converts to a radius equation; the slow-spiral (adiabatic) approximation — orbits fast, decay slow, marginally honest here and conservative — gives

\frac{dr}{dt}=-\frac{4}{3}\frac{r_e^2c}{r^2}\cdot 3\ \Rightarrow\ r^2\dot r=-4r_e^2c \quad\Longrightarrow\quad t_{\text{collapse}}=\frac{a_0^3}{4r_e^2c},

with $r_e=e^2/4\pi\varepsilon_0mc^2=2.82\times10^{-15}$ m the classical electron radius. Numerically:

t_{\text{collapse}}=\frac{(5.29\times10^{-11})^3}{4(2.82\times10^{-15})^2(3.0\times10^8)}=\mathbf{1.6\times10^{-11}\ s}.

Step 4. Take stock. Sixteen picoseconds — about $10^5$ ever-faster orbits, the emitted "light" sweeping continuously up in frequency as the spiral tightens: a smooth chirp, nothing like the discrete lines atoms actually emit. Both predictions are catastrophically wrong: matter is stable for eons and spectra are sharp. The failure is not in the electrodynamics — Larmor is impeccable — but in applying continuous classical trajectories at scales where action is of order $\hbar$ . Bohr's 1913 postulates were engineered precisely to switch this formula off in stationary states; quantum mechanics then rebuilt radiation as transition rates. Larmor's most important number is the one nature refused.

The lesson: a formula's greatest service can be a wrong prediction delivered with certainty — the classical atom's sixteen picoseconds forced the twentieth century's hand.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §11.2.1 · Jackson, Classical Electrodynamics, 3rd ed., §14.2 · Purcell & Morin, Electricity and Magnetism, 3rd ed., App. H · Zangwill, Modern Electrodynamics, §23.3

Provenance. Base derivation verified against Griffiths Eq. 11.70 (route: Poynting flux of the acceleration field, angular integral $\int\sin^2\theta\,d\Omega=8\pi/3$ ) and Jackson Eq. 14.22 (Gaussian form $2q^2a^2/3c^3$ ). A verifier should recompute: (1) the field magnitude $qa\sin\theta/4\pi\varepsilon_0c^2r$ from the $\beta\to0$ Liénard–Wiechert acceleration field; (2) the constants chain $1/(\varepsilon_0^2c^4\mu_0c)=\mu_0/c$ ; (3) the collapse time $a_0^3/4r_e^2c\approx1.6\times10^{-11}$ s. Addendum for this edition: per the chapter's division of labor, the acceleration field (1) is quoted by name from Fields of a Moving Point Charge — its derivation, the velocity/acceleration split, and the qualitative kink picture live there; this edition owns the flagship power integration (Derivation A) and the quantitative Thomson kink construction (Derivation B, Purcell App. H route), which sharpens that article's picture into the field formula. The relativistic generalization is stated only as a limiting-cases row and owned by Relativistic Larmor Formula (Liénard's Result); radiation reaction is deferred to its own pair of articles; the covariant treatment is the E11 twin's. Worked numbers for this edition (cyclotron at 1 T: $a=5.3\times10^{17}$ m/s², $P=1.6\times10^{-18}$ W, $\tau=2.6$ s two routes — ratio and closed form $3\pi\varepsilon_0m^3c^3/e^4B^2$ ; hydrogen collapse $1.6\times10^{-11}$ s $\approx10^5$ orbits; $P(a=g)=5.5\times10^{-52}$ W) recomputed and verified for this edition.

Builds on: Liénard–Wiechert Potentials · Fields of a Moving Point Charge · Retarded Potentials Leads to: Electric Dipole Radiation · Magnetic Dipole Radiation · Relativistic Larmor Formula (Liénard's Result) · Radiation Reaction Force

If radiated power is velocity-independent, why do faster particles in accelerators radiate so much more?

Because they are relativistic, and Larmor's velocity-blindness holds only at $v\ll c$ . Two things change as $v\to c$ : the transformation between the particle's rest frame (where Larmor applies exactly, instantaneously) and the lab multiplies the power by $\gamma$ factors, and the acceleration for a given force is itself modified by relativistic inertia. The exact bookkeeping — Liénard's formula, next article — gives $P\propto\gamma^6$ for acceleration along the motion and $\gamma^4$ for circular motion. Within the strictly nonrelativistic regime, though, the statement is exact and worth internalizing: a charge dawdling at 1 m/s and one cruising at $10^6$ m/s, given the same shove, send the same bill. Radiation reads force, not speed.

Where does the radiated energy actually come from — the charge slows down, but which force does the accounting?

Over an accelerating stretch, the agent pushing the charge pays: part of its work goes into kinetic energy, part into the field. But the local bookkeeping is genuinely subtler than "power in equals Larmor out at each instant." The energy flowing through a distant sphere now was emitted at the retarded time; the fields near the charge also hold a reservoir (the Schott, or acceleration, energy) that can lend and reabsorb energy within a cycle. Instantaneous Larmor power equals the instantaneous drain on the particle only after averaging over a period or over a complete event (kick, orbit, wave packet). The force that mediates the drain — the field of the charge acting back on the charge itself — is the radiation-reaction force, and making its accounting precise (including exactly what the Schott term defers) is the entire business of the next-but-one article.

Does a charge held at rest in Earth's gravity radiate? What about one falling freely?

This is the famous equivalence-principle trap, and the honest short answer is: "radiates" is not a local, observer-free property. By the equivalence principle a supported charge is locally indistinguishable from one accelerating at $g$ in flat space — suggesting it radiates — while a freely falling charge is locally inertial — suggesting it does not. The resolution, worked out over decades (Boulware's analysis is the standard landmark), is that the radiation a detector reports depends on the detector's own state of motion and horizon structure: a co-supported detector next to the supported charge sees no radiation (the would-be radiation lies beyond its horizon), while a freely falling detector watching the supported charge does. Nothing paradoxical survives careful bookkeeping, but the lesson is real and deep: the Larmor formula's clean $P$ is an inertial-frame, asymptotic-observer statement. Locality, gravity, and radiation mix only with general-relativistic care — the same care that later prices Unruh and Hawking radiation.

Why does the formula have exactly one factor of the classical electron radius hidden in it?

Write Larmor's energy loss per unit time against the natural atomic scales and $r_e=e^2/4\pi\varepsilon_0mc^2$ keeps surfacing — the collapse time is $a_0^3/4r_e^2c$ , the cyclotron cooling time is $\tfrac{3}{2}c/(r_e\omega_c^2)\cdot$ (order one), the Thomson cross-section (light scattering off free electrons, built from Larmor) is $\tfrac{8\pi}{3}r_e^2$ . The reason: $r_e$ is the length at which a charge's field energy is comparable to its rest energy, so it is the built-in yardstick for "how strongly does this particle's own field react on it." Any process where radiation competes with inertia gets priced in powers of $r_e$ . That same yardstick, pushed to its limit, marks where classical electrodynamics stops being self-consistent — the subject of the radiation-reaction articles, where $r_e/c$ appears as the pathological timescale $\tau$ .

How is Larmor's single-charge formula related to the dipole radiation formula — which is more fundamental?

They are the same statement at two levels of aggregation, and which is "fundamental" depends on direction of travel. For a single slow charge, $\ddot{\mathbf{p}}=q\mathbf{a}$ makes Larmor and the point-dipole formula $P=\mu_0\ddot p^2/6\pi c$ literally identical. For a system of charges, the dipole formula with $\mathbf{p}=\sum q_i\mathbf{r}_i$ is the coherent sum: at wavelengths long compared to the system, the fields of all charges add before squaring, so a neutral pair shaking in step radiates via its net $\ddot{\mathbf{p}}$ (which may vanish — then the multipole ladder takes over), not via the sum of individual Larmor powers. Incoherent Larmor-per-charge is instead correct when the charges wiggle with random phases (thermal bremsstrahlung). So: Larmor is the microscopic brick; dipole and multipole radiation are the architecture of how bricks interfere; and choosing coherent versus incoherent addition — squaring the sum versus summing the squares — is the physical judgment call the wavelength decides.

The derivation integrated flux at one instant. What happens if the acceleration is changing rapidly — is P still meaningful?

The formula remains exact as a statement about flux at the sphere: the power crossing radius $r$ at time $t$ is $\mu_0q^2a(t_r)^2/6\pi c$ with $t_r=t-r/c$ — each shell of radiation carries the imprint of the acceleration at its own emission moment, and nothing requires $a$ to be slowly varying. What becomes delicate for rapidly changing $a$ is the attribution at the source: identifying the instantaneous flux with an instantaneous energy drain on the particle runs into the Schott-energy lending business (see the energy FAQ above), and for motions with structure on timescales approaching $r_e/c$ the whole point-charge framework begins to creak. Practical rule: for any macroscopic or atomic-scale motion, use $P(t)=\mu_0q^2a(t)^2/6\pi c$ freely, retarded when mapping to a distant detector; reserve the fine print for the radiation-reaction articles where it becomes the story.