Electric Quadrupole Radiation

The dipole approximation was built on a deliberate act of blindness: treat the whole source as one point, so that every part of it broadcasts from the same retarded time. That is the zeroth-order reading of retardation, and it delivered the electric dipole formula. This article opens its eyes one notch. Across a source of size $d$ , the retarded time varies by $\sim d/c$ — the far side reports slightly staler news than the near side — and expanding the current to first order in that skew produces the entire next generation of radiation. The correction term splits, by pure algebra, into an antisymmetric piece and a symmetric piece. The antisymmetric piece is the magnetic dipole of the previous article, arriving from a second direction. The symmetric piece is new: electric quadrupole radiation, sourced not by any vector moment but by a rank-2 tensor, the traceless quadrupole moment $Q_{ij}$ . The tensor character changes everything visible. The radiated power rides the third time derivative, $\dddot{Q}_{ij}$ , so the power law steepens from $\omega^4$ to $\omega^6$ . The angular pattern is no longer the single donut but a four-lobed rosette, $\sin^2\theta\cos^2\theta$ for the simplest case, silent both along the axis and on the equator. And the price of the extra moment is another factor of $(kd)^2$ in power — which is why quadrupole transitions in atoms are "forbidden": not impossible, just tens of thousands of times slower, visible only where collisions cannot beat them to it. One more reason to master this rung, and it is a large one: gravity radiates starting here. Mass has no negative charge, so gravitational radiation has no dipole term at all — the leading gravitational-wave channel is the quadrupole, structurally the same $\dddot{Q}$ formula. The mathematics you earn in this article is the classical skeleton of LIGO's chirps.

Key takeaways

First-order retardation across the source — the correction $(\hat{\mathbf{n}}\cdot\mathbf{x}'/c)\,\dot{\mathbf{J}}$ inside the radiation integral — generates both magnetic dipole and electric quadrupole radiation: same order in $d/\lambda$ , split by symmetry.
The quadrupole channel is governed by the traceless tensor $Q_{ij}=\int(3x_i'x_j'-r'^2\delta_{ij})\rho\,d^3x'$ , through its third time derivative $\dddot{Q}_{ij}$ .
Radiation fields: $\mathbf{B}_{\text{rad}}=\dfrac{\mu_0}{24\pi c^2 r}\,\hat{\mathbf{n}}\times\dddot{\mathbf{Q}}(t_0)$ with $[\dddot{\mathbf{Q}}]_i=\dddot{Q}_{ij}\hat{n}_j$ , and $\mathbf{E}_{\text{rad}}=c\mathbf{B}_{\text{rad}}\times\hat{\mathbf{n}}$ .
Total power: $P_{E2}=\dfrac{\mu_0}{1440\pi c^3}\sum_{ij}\langle\dddot{Q}_{ij}^2\rangle$ — for harmonic sources $\propto\omega^6|Q|^2$ , two powers of frequency above the dipole.
Suppression: $P_{E2}/P_{E1}\sim(kd)^2=(2\pi d/\lambda)^2$ ; quadrupole radiation leads only when the dipole (and often $M1$ ) is silenced by symmetry.
The simplest oscillating quadrupole radiates a four-lobed $\sin^2\theta\cos^2\theta$ pattern: nulls on the axis and the equator, maxima at $45°$ .
Key idea: quadrupole radiation is the first retardation correction made audible — a tensor moment, an $\omega^6$ law, a four-lobed pattern, and the structural template for gravitational waves.

The core insight

Quadrupole radiation is the first-order retardation correction to the dipole approximation: expanding the retarded current to next order in $\hat{\mathbf{n}}\cdot\mathbf{x}'/c$ splits into an antisymmetric piece (magnetic dipole) and a symmetric piece (electric quadrupole), with radiated power $\propto\omega^6$ and a multi-lobed angular pattern set by the traceless tensor $\dddot{Q}_{ij}$ .

Where this lives today

The quadrupole formula's biggest modern stage is not electromagnetic at all. Gravitational radiation begins at the quadrupole — conservation of mass kills the monopole channel and conservation of momentum kills the dipole, precisely the way charge conservation killed the electromagnetic monopole here — and the gravitational luminosity formula is the direct structural analogue of Derivation B's result, with the mass quadrupole in place of $Q_{ij}$ and $G/c^5$ setting the scale. Every LIGO/Virgo chirp is a $\dddot{Q}$ calculation: as a binary tightens, $\omega$ climbs and the steep frequency dependence (the gravitational counterpart of $\omega^6$ ) drives the runaway sweep in amplitude and pitch that ends in merger. The angular integrals you did in Step 2 and Step 3 are, symbol for symbol, the ones in the gravitational-wave literature. On the electromagnetic side, quadrupole physics runs quietly through modern spectroscopy and quantum optics. Clock transitions in trapped-ion optical clocks — the electric-quadrupole $S\to D$ lines of Sr⁺, Ca⁺, Yb⁺ — are prized because they are $E2$ -forbidden: millisecond-to-second lifetimes mean millihertz linewidths, the narrowness that makes eighteen-digit timekeeping possible. Nebular diagnostics still lean on forbidden-line ratios to meter temperatures and densities of interstellar gas. And in nanophotonics the dipole approximation itself is failing by design: plasmonic nanostructures and "metatoms" are engineered with field gradients steep enough that quadrupole transitions light up at useful rates — first-order retardation, weaponized.

Origins · 1 minLev Davidovich Landau and how this was discovered

Lev Davidovich Landau

1908–1968 · Kharkov and Moscow — whose Classical Theory of Fields made the quadrupole formula canonical

Landau's §71 — quadrupole and magnetic dipole radiation, in the volume of the Course of Theoretical Physics written with Evgeny Lifshitz — is where generations of physicists learned the calculation this article performs, compressed to a page and a half of inevitabilities. The book's fingerprints are on the formalism itself: the systematic expansion in retardation, the traceless moment as the natural radiating object, and the immediate leap to the application that mattered — Landau and Lifshitz apply the quadrupole formula to gravitating binaries, deriving the orbital decay of a two-body system by gravitational radiation decades before pulsar timing made it measurable. (The Hulse–Taylor binary's spin-down, matched to that very formula, won the 1993 Nobel Prize.) Landau's career was itself a study in extremes: the Landau levels, Landau damping, superfluidity theory that earned his own Nobel in 1962; arrest in 1938 during the purges, released after Kapitsa's personal intervention; and the 1962 car crash that ended his working life at 54. The Course outlived it all — the quadrupole section essentially unchanged, because there was nothing to change.

History

1908

— Mie's sphere-scattering paper organizes fields into partial waves — multipoles beyond the dipole become working tools

1918

— Einstein's quadrupole formula for gravitational radiation: the no-dipole theorem makes ℓ=2 the leading gravitational channel

1927

— Bowen identifies the mysterious "nebulium" lines as forbidden M1/E2 transitions of ionized oxygen and nitrogen

1930

s — Condon and Shortley systematize E1/M1/E2 selection rules for atomic spectra

1951

— the 21 cm hyperfine line detected, opening forbidden-transition astronomy at radio wavelengths

1974

— Hulse and Taylor discover the binary pulsar; its orbital decay matches the quadrupole radiation formula

2015

— LIGO records GW150914: the quadrupole chirp of two merging black holes, ending the multipole ladder's purely electromagnetic career

The physical picture

Build the simplest source with no dipole moment: two identical dipoles, back to back. Put $+q$ at $z=\pm d$ and $-2q$ at the origin, and let the outer charges oscillate in step. At every instant the upper half is a dipole pointing up and the lower half a dipole pointing down; the vector sum is zero, always. In the dipole approximation — every part broadcasting from one shared retarded time — the two halves' radiation fields cancel identically, and the source is silent. But they do not broadcast from the same retarded time. An observer above the source hears the upper dipole slightly fresher than the lower one — the news is offset by the light-crossing time of the separation, a phase skew of order $kd$ . Two waves of equal strength, opposite sign, and a small relative delay do not cancel; they leave a residue proportional to the delay. That residue is quadrupole radiation: the amplitude carries the extra factor $kd$ , the power its square. Everything in the formalism — the extra moment of the charge distribution, the extra time derivative, the extra powers of frequency — is this one picture in equations.

Figure

Two dipoles back to back — perfect cancellation at zeroth order, and the retardation skew that survives it

The object that catalogs "shape sloshing" is the traceless quadrupole tensor — the same second moment that governed the $1/r^3$ term of the static multipole ladder in the electrostatics chapter, now promoted to a function of time:

Q_{ij}(t)=\int\big(3x_i'x_j'-r'^2\delta_{ij}\big)\,\rho(\mathbf{x}',t)\,d^3x'. \qquad (1)

where: $Q_{ij}$ is symmetric and traceless ( $Q_{ii}=0$ ) by construction — five independent numbers describing the ellipsoidal distortion of the charge cloud. Subtracting the trace is not cosmetic: the trace part is a spherically symmetric breathing of the source, and spherical breathing, by the same charge-conservation argument that silences the monopole, cannot radiate. Only the shape-changing part of the second moment couples to transverse waves at infinity, and the standard coefficients below assume the traceless form — feed the raw moment $\int x_i'x_j'\rho\,d^3x'$ into them and the prefactors come out wrong.

The math: one order deeper into retardation

Everything begins from the far-zone radiation integral, the Lorenz-gauge retarded vector potential of the retarded-potentials article with the far-field replacement $|\mathbf{r}-\mathbf{x}'|\approx r-\hat{\mathbf{n}}\cdot\mathbf{x}'$ in the time argument (and just $r$ in the amplitude):

\mathbf{A}(\mathbf{r},t)=\frac{\mu_0}{4\pi r}\int\mathbf{J}\!\left(\mathbf{x}',\,t_0+\frac{\hat{\mathbf{n}}\cdot\mathbf{x}'}{c}\right)d^3x',\qquad t_0=t-\frac{r}{c}. \qquad (2)

where: $\hat{\mathbf{n}}$ points from the source's center to the observer, and the term $\hat{\mathbf{n}}\cdot\mathbf{x}'/c$ inside the time argument is the per-element freshness correction — the near side of the source ( $\hat{\mathbf{n}}\cdot\mathbf{x}'>0$ ) reports later, fresher news. The dipole approximation drops this term entirely; every multipole beyond $E1$ lives inside it. Taylor-expanding the current in that small time shift is the single move that generates the ladder, and this article executes the first step of it.

Derivation AThe first retardation correction splits into magnetic dipole plus electric quadrupole

The goal is the structural theorem of the article: expanding (2) to first order in $\hat{\mathbf{n}}\cdot\mathbf{x}'/c$ yields exactly two new radiation channels — the magnetic dipole and the electric quadrupole — and identifies the quadrupole's source as $\ddot{Q}_{ij}$ . The strategy: Taylor-expand the current, split the resulting first-moment integral into antisymmetric and symmetric parts (a pure algebra move), recognize the antisymmetric part as the magnetic moment, and convert the symmetric part from current moments to charge moments using the continuity equation — the same conversion trick the electric dipole derivation used one order down. Watch the symmetry bookkeeping: it is what guarantees $M1$ and $E2$ arrive together, at the same order, always.

Step 1. Taylor-expand the current in the freshness correction:

\mathbf{J}\!\left(\mathbf{x}',t_0+\tfrac{\hat{\mathbf{n}}\cdot\mathbf{x}'}{c}\right)\approx\mathbf{J}(\mathbf{x}',t_0)+\frac{\hat{\mathbf{n}}\cdot\mathbf{x}'}{c}\,\dot{\mathbf{J}}(\mathbf{x}',t_0).

Why stopping at first order is legitimate: each further term costs another power of $kd$ — the expansion parameter of the whole chapter. The zeroth term integrates to $\dot{\mathbf{p}}$ (the electric dipole channel, owned by that article); everything new is in the first-order piece.

Step 2. Split the first-order integral by symmetry. Writing components, the object is $\int x_j'\,\dot{J}_i\,d^3x'$ , and any such two-index integral splits identically:

\int x_j'\dot{J}_i\,d^3x'=\underbrace{\tfrac12\int\big(x_j'\dot{J}_i-x_i'\dot{J}_j\big)d^3x'}_{\text{antisymmetric}}+\underbrace{\tfrac12\int\big(x_j'\dot{J}_i+x_i'\dot{J}_j\big)d^3x'}_{\text{symmetric}}.

The antisymmetric part is, by the definition $\mathbf{m}=\tfrac12\int\mathbf{x}'\times\mathbf{J}\,d^3x'$ , precisely $\dot{\mathbf{m}}$ contracted with $\hat{\mathbf{n}}$ — the magnetic dipole channel of the previous article, rediscovered from the systematic side. Why this split is physics and not just algebra: antisymmetric = circulation = magnetic moment; symmetric = stretching = shape change. The one first-order correction contains exactly one of each, which is the theorem that $M1$ and $E2$ share an order in $kd$ — no calculation that keeps one may drop the other on grounds of size alone.

Step 3. Convert the symmetric part to charge moments via continuity. The identity (integrate $\nabla'\cdot\mathbf{J}=-\dot\rho$ against $x_i'x_j'$ and integrate by parts, localized source, surface term dead):

\int\big(x_i'J_j+x_j'J_i\big)\,d^3x'=\frac{d}{dt}\int x_i'x_j'\,\rho\,d^3x'.

So the symmetric channel is driven by the second time derivative of the second charge moment — and adding the (non-radiating) trace subtraction to form (1), by $\ddot{Q}_{ij}/3$ up to that trace. Why continuity keeps appearing at every rung: current moments are what enter $\mathbf{A}$ , but charge moments are what we catalog sources by; the continuity equation is the dictionary, converting one current moment into one time derivative of a charge moment — at this order exactly as it converted $\int\mathbf{J}\,d^3x'=\dot{\mathbf{p}}$ for the dipole.

Step 4. Assemble the quadrupole potential and differentiate. The symmetric part of the first-order potential reads

\mathbf{A}_{E2}=\frac{\mu_0}{24\pi c\,r}\,\ddot{\mathbf{Q}}(t_0),\qquad \big[\ddot{\mathbf{Q}}\big]_i\equiv\ddot{Q}_{ij}\,\hat{n}_j,

and one radiation-zone curl (each derivative $\to-(\hat{\mathbf{n}}/c)\partial_t$ , exactly as in the $M1$ derivation) delivers the fields:

\mathbf{B}_{\text{rad}}=\frac{\mu_0}{24\pi c^2 r}\,\hat{\mathbf{n}}\times\dddot{\mathbf{Q}}(t_0),\qquad \mathbf{E}_{\text{rad}}=c\,\mathbf{B}_{\text{rad}}\times\hat{\mathbf{n}}.

Why a THIRD derivative: one dot per conversion. Continuity turned the current moment into $\ddot{Q}$ inside the potential; the curl's action on the retarded time added the third. Radiation always costs one more derivative than the potential carries — the dipole's $\dot{\mathbf{p}}$ in $\mathbf{A}$ became $\ddot{\mathbf{p}}$ in the fields the same way. *Verified: the prefactor $\mu_0/24\pi c^2$ matches Jackson §9.3 (SI) and the structure matches Griffiths §11.1.4's order counting; setting the symmetric moment to a point dipole's reproduces no term — the channels are cleanly separated.*

The contraction $[\dddot{\mathbf{Q}}]_i=\dddot{Q}_{ij}\hat{n}_j$ is worth reading twice: the tensor is a property of the source alone, but the vector that actually radiates toward you depends on your direction $\hat{\mathbf{n}}$ . A rank-2 source shows a different face to every observer — that is where the multi-lobed patterns come from, and it is the structural difference between this rung and everything below it.

Go deeper · Derivation B · 1 minTotal quadrupole power — the rank-4 angular integral and the omega-sixth law

Derivation BTotal quadrupole power — the rank-4 angular integral and the omega-sixth law

The goal is the Larmor-analogue for the quadrupole channel, $P_{E2}=\mu_0\sum_{ij}\langle\dddot{Q}_{ij}^2\rangle/1440\pi c^3$ , with every angular factor earned. Strategy: Poynting flux of the fields from Derivation A, then integrate $|\hat{\mathbf{n}}\times\dddot{\mathbf{Q}}|^2$ over the sphere. The work is two standard sphere averages — a rank-2 and a rank-4 — plus the tracelessness of $Q$ doing one crucial kill. This is the first genuinely tensorial angular integral of the track; the same machinery returns for the full multipole expansion and (in another theory) for the gravitational quadrupole formula.

Step 1. Flux per solid angle. With $|\mathbf{E}|=c|\mathbf{B}|$ and $S=|\mathbf{E}||\mathbf{B}|/\mu_0=c|\mathbf{B}|^2/\mu_0$ , and the cross-product identity $|\hat{\mathbf{n}}\times\mathbf{V}|^2=V_iV_i-(\hat n_iV_i)^2$ applied to $V_i=\dddot{Q}_{ij}\hat n_j$ :

\frac{dP}{d\Omega}=r^2S=\frac{\mu_0}{1152\pi^2c^3}\Big[\dddot{Q}_{ij}\dddot{Q}_{ik}\,\hat n_j\hat n_k-\dddot{Q}_{ij}\dddot{Q}_{kl}\,\hat n_i\hat n_j\hat n_k\hat n_l\Big].

Step 2. The two sphere averages, quoted from the standard table and verifiable by symmetry alone:

\oint\hat n_j\hat n_k\,d\Omega=\frac{4\pi}{3}\delta_{jk},\qquad \oint\hat n_i\hat n_j\hat n_k\hat n_l\,d\Omega=\frac{4\pi}{15}\big(\delta_{ij}\delta_{kl}+\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk}\big).

Why no calculation is really needed: each integral must be an isotropic tensor of its rank built from Kronecker deltas, and the overall constants follow from taking traces ( $\hat n_j\hat n_j=1$ fixes the first; a double trace fixes the second).

Step 3. Contract, and let tracelessness kill the trace term. The rank-2 average gives $\tfrac{4\pi}{3}\dddot{Q}_{ij}\dddot{Q}_{ij}$ . In the rank-4 contraction, the $\delta_{ij}\delta_{kl}$ term produces $\dddot{Q}_{ii}\dddot{Q}_{kk}=0$ — the trace subtraction of (1) earning its keep — while the other two each produce $\dddot{Q}_{ij}\dddot{Q}_{ij}$ :

P=\frac{\mu_0}{1152\pi^2c^3}\left[\frac{4\pi}{3}-\frac{8\pi}{15}\right]\dddot{Q}_{ij}\dddot{Q}_{ij}=\frac{\mu_0}{1152\pi^2c^3}\cdot4\pi\left(\frac13-\frac{2}{15}\right)\dddot{Q}_{ij}\dddot{Q}_{ij}.

Step 4. Collect: $\tfrac13-\tfrac{2}{15}=\tfrac15$ and $\tfrac{4\pi}{1152\pi^2}\cdot\tfrac15=\tfrac{1}{1440\pi}$ , hence

P_{E2}=\frac{\mu_0}{1440\pi c^3}\sum_{i,j}\big\langle\dddot{Q}_{ij}^{\,2}\big\rangle \;\xrightarrow{\ Q_{ij}(t)=Q_{ij}\cos\omega t\ }\; \frac{\mu_0\,\omega^6}{2880\pi c^3}\sum_{i,j}|Q_{ij}|^2,

the harmonic form using $\langle\dddot{Q}^2\rangle=\tfrac12\omega^6|Q|^2$ — three dots, six powers of $\omega$ , the promised steepening from the dipole's $\omega^4$ . Verified: coefficient $1/1440\pi$ reproduces Landau–Lifshitz §71 (after Gaussian→SI conversion) and Jackson §9.3; the numeric cross-check in worked example 1 confirms the assembled formula against the $(kd)^2$ suppression estimate.

For the simplest source — the axial quadrupole with $Q_{zz}=Q_0$ , $Q_{xx}=Q_{yy}=-Q_0/2$ (the back-to-back dipoles of the picture section) — the general contraction of Step 1 collapses to

\frac{dP}{d\Omega}=\frac{\mu_0\,\langle\dddot{Q}_0^2\rangle}{512\pi^2c^3}\,\sin^2\theta\cos^2\theta, \qquad (3)

where: the $\tfrac94\sin^2\theta\cos^2\theta$ shape (the $9/4$ absorbed into the prefactor here) has two null directions — along the axis, where no transverse component survives, and on the equator, where the two half-dipoles are heard perfectly in phase and cancel exactly. Maxima sit at $\theta=45°$ and $135°$ : a four-lobed rosette, the visible signature of a rank-2 source, and the first pattern in the track that a single dipole of any orientation cannot mimic.

Figure

Dipole donut vs quadrupole rosette — the angular pattern is the multipole's fingerprint

Limiting cases

case	result	meaning
dipole-allowed source	$P_{E2}/P_{E1}\sim(kd)^2\ll1$	quadrupole invisible whenever the dipole speaks
$\mathbf{p}=0$ by symmetry	$P\approx P_{M1}+P_{E2}$ , same order	the two first-correction channels share the stage
$\dddot{Q}_{ij}=0$	$P_{E2}=0$	static or uniformly translating quadrupoles are silent
trace part of the moment	contributes nothing	spherical breathing cannot radiate — the $\delta_{ij}$ subtraction in (1)
axial quadrupole	$dP/d\Omega\propto\sin^2\theta\cos^2\theta$	four lobes, nulls on axis and equator, maxima at $45°$
$kd\to0$ strictly	only $E1$ survives	the point-source limit; the ladder collapses to its first rung
gravity	no monopole, no dipole $\Rightarrow$ quadrupole leads	the same $\dddot{Q}$ structure, with mass for charge

Common trapKeep the magnetic dipole, and keep the trace subtraction — the two omissions that silently corrupt every quadrupole calculation

Both errors produce answers that look plausible and are wrong. First: computing "the next correction to dipole radiation" and evaluating only the quadrupole. Derivation A shows $M1$ and $E2$ are two halves of one integral — the antisymmetric and symmetric parts of the same first-order term. Unless a symmetry kills one of them (a perfect loop has no $E2$ ; back-to-back dipoles have no $M1$ ), both are present at the same order, they interfere in the angular distribution, and dropping either is not an approximation but a mistake. Second: feeding the raw second moment $\int x_i'x_j'\rho\,d^3x'$ into the power formula. The coefficient $1/1440\pi$ was derived using $Q_{ii}=0$ — the trace term was killed by tracelessness in Step 3. With a traceful moment the rank-4 contraction leaves an extra piece and every prefactor shifts. The trace itself never radiates (spherical breathing has no transverse projection), so the physics is indifferent — but the bookkeeping is not. Use (1), always.

Worked examples

1 · The linear quadrupole, priced in watts (confidence builder)

Setup. Charges $+q$ at $z=\pm d$ and $-2q$ at the origin, with $q=1$ nC, $d=1$ cm, oscillating at $f=100$ MHz ( $\omega=6.28\times10^8$ s⁻¹, $kd=\omega d/c=0.0210$ ). Find the radiated power and compare with a single dipole of the same scale.

Step 1. The moment. Point charges on the axis have $r'^2=z'^2$ , so each outer charge contributes $q(3d^2-d^2)=2qd^2$ to $Q_{zz}$ and the central charge nothing:

Q_0\equiv Q_{zz}=4qd^2=4(10^{-9})(10^{-2})^2=4.0\times10^{-13}\ \text{C}\,\cdot\,\text{m}^2,

with $Q_{xx}=Q_{yy}=-Q_0/2$ enforced by tracelessness (each outer charge contributes $-qd^2$ to $Q_{xx}$ : check, $2\times(-qd^2)=-2qd^2=-Q_0/2$ ✓).

Step 2. The power. $\sum Q_{ij}^2=Q_0^2(1+\tfrac14+\tfrac14)=\tfrac32Q_0^2$ , so the harmonic formula gives

P_{E2}=\frac{\mu_0\omega^6}{2880\pi c^3}\cdot\frac32 Q_0^2=\frac{(1.257\times10^{-6})(6.15\times10^{52})}{2880\pi(2.70\times10^{25})}\cdot\frac32(1.6\times10^{-25})=\mathbf{7.6\times10^{-8}\ W}.

Step 3. The comparison. A single dipole $p_0=qd=10^{-11}$ C·m at the same frequency radiates $P_{E1}=\mu_0p_0^2\omega^4/12\pi c=1.73\times10^{-3}$ W. The ratio:

\frac{P_{E2}}{P_{E1}}=4.4\times10^{-5}=\frac{(kd)^2}{10}.

Cancelling the dipoles cost a factor of twenty-three thousand — and the exact ratio lands at $(kd)^2/10$ , the suppression estimate with its honest $O(1)$ coefficient computed rather than waved at.

The lesson: the $(kd)^2$ law is quantitative. Before any quadrupole calculation, compute $kd$ ; its square, within a factor of order ten, is your suppression.

2 · Forbidden lines — why quadrupole transitions take milliseconds, and why the sky glows green (exam level)

Setup. Estimate the ratio of $E2$ to $E1$ transition rates for an atom radiating visible light ( $\lambda=500$ nm), and read off the lifetime scale of quadrupole "forbidden" lines given typical allowed lifetimes of order nanoseconds.

Step 1. The atom's size parameter. The radiating structure has the scale of the Bohr radius $a_0=5.29\times10^{-11}$ m, and $k=2\pi/\lambda$ :

ka_0=\frac{2\pi(5.29\times10^{-11})}{5.0\times10^{-7}}=6.6\times10^{-4}.

Step 2. The rate suppression. Transition rates follow the classical power ratio:

\frac{\Gamma_{E2}}{\Gamma_{E1}}\sim(ka_0)^2=4.4\times10^{-7}.

Step 3. The lifetime. With allowed lifetimes $\tau_{E1}\sim1.6$ ns (hydrogen $2p\to1s$ as the anchor):

\tau_{E2}\sim\frac{1.6\times10^{-9}}{4.4\times10^{-7}}\approx\mathbf{3.6\ ms}

— six orders of magnitude slower, and real $E2$ lines (with their specific matrix elements) range from milliseconds to seconds. The famous green auroral and nebular oxygen lines come from exactly such metastable states, with lifetimes near a second.

Step 4. The density condition that makes them visible. A state that waits $\sim1$ s to radiate only gets to do so if nothing collides with it first: the collision time must exceed the radiative lifetime. In laboratory gas at atmospheric density (collision time $\sim10^{-9}$ s) the state is quenched a billion times over — forbidden lines are laboratory-invisible, which is why nebular spectra baffled astronomers into inventing the element "nebulium." In an emission nebula at $\sim10^8$ particles/m³ ( $\sim10^{-17}$ of atmospheric), collisions arrive hours apart and the slow channels win. The forbidden lines are not exotic emission — they are ordinary $M1$ / $E2$ radiation in an environment dilute enough to wait for it.

The lesson: "forbidden" means suppressed by the multipole ladder, and the ladder's small parameter $(ka_0)^2$ converts directly into the lifetimes and the astrophysical conditions where such lines dominate.

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., §9.2–9.3, §9.10 · Griffiths, Introduction to Electrodynamics, 4th ed., §11.1.2–11.1.4 · Landau & Lifshitz, The Classical Theory of Fields, §71 · Zangwill, Modern Electrodynamics, Ch. 20

Provenance. Equations follow the standard SI multipole-radiation treatment in Jackson (§9.2–9.3) and Landau–Lifshitz (§71). A verifier should recompute: (1) the angular integrals $\oint\hat n_j\hat n_k\,d\Omega=\tfrac{4\pi}{3}\delta_{jk}$ and the rank-4 average; (2) the algebra $\tfrac13-\tfrac{2}{15}=\tfrac15$ yielding the coefficient $1/(1440\pi)$ ; (3) the $\omega^6$ scaling from $\dddot Q\sim\omega^3Q$ . Cross-check the field prefactor $\mu_0/(24\pi c^2r)$ against Jackson and the power formula against Landau–Lifshitz (after Gaussian→SI conversion). Addendum for this edition: per the chapter's division of labor, the zeroth-order (dipole) channel and the radiation-zone approximations are owned by Electric Dipole Radiation; the antisymmetric partner discovered in Derivation A Step 2 is owned by Magnetic Dipole Radiation and recalled, not re-derived; the systematic all- $\ell$ machinery (vector spherical harmonics, exact coefficients) is deferred to Multipole Expansion of Radiation and, at Jackson register, the E11 treatments. Worked numbers for this edition (linear quadrupole $Q_0=4qd^2=4.0\times10^{-13}$ C·m², $P_{E2}=7.6\times10^{-8}$ W, ratio $(kd)^2/10=4.4\times10^{-5}$ at $kd=0.021$ ; atomic estimate $ka_0=6.6\times10^{-4}$ , $\tau_{E2}\sim3.6$ ms; the axial pattern coefficient $\tfrac94\sin^2\theta\cos^2\theta$ and $\sum Q_{ij}^2=\tfrac32Q_0^2$ ) recomputed and sympy-verified for this edition.

Builds on: Electric Dipole Radiation · Magnetic Dipole Radiation · Retarded Potentials · Scalar and Vector Potentials · Lorenz Gauge Leads to: Multipole Expansion of Radiation · Jefimenko's Equations

Why does the quadrupole need a third time derivative when the dipole only needed two?

Count the conversions. The radiation field always costs one time derivative acting through the retarded time — that is universal, one dot in every channel. What differs is how many dots the potential already carries. The dipole potential carries $\dot{\mathbf{p}}$ (continuity converting $\int\mathbf{J}=\dot{\mathbf{p}}$ : one dot), so its field rides $\ddot{\mathbf{p}}$ . The quadrupole potential carries $\ddot{Q}_{ij}$ — the symmetric current moment converts via continuity to the time derivative of the second charge moment, and the expansion itself contributed one more derivative through the $\dot{\mathbf{J}}$ in the first-order term: two dots in the potential, three in the field. Each rung of the ladder adds one moment of $\mathbf{x}'$ and one time derivative together, which is exactly why each rung costs one factor of $kd\sim\omega d/c$ in amplitude: the $\omega$ from the dot and the $d$ from the moment arrive as a pair.

Is the four-lobed pattern universal for quadrupole radiation, or specific to the linear case?

Specific. A general $Q_{ij}$ has five independent components, and the radiated pattern is the direction-dependent contraction $\dddot Q_{ij}\dddot Q_{ik}\hat n_j\hat n_k-(\dddot Q_{ij}\hat n_i\hat n_j)^2$ — a family of shapes, not one shape. The $\sin^2\theta\cos^2\theta$ rosette belongs to the axially symmetric (diagonal, $Q_{zz}=-2Q_{xx}=-2Q_{yy}$ ) tensor: back-to-back dipoles, symmetric stretching modes. A rotating quadrupole — two masses or charges orbiting, the binary-star configuration — has two components ninety degrees out of phase and radiates a different pattern, strongest along the rotation axis (where the radiation is circularly polarized) and weakest, but not zero, in the plane. What is universal: an $\ell=2$ pattern always has more angular structure than any dipole can produce, its lobes count and placement encode the tensor's principal axes, and the total power is always the invariant $\sum\dddot Q_{ij}^2$ contraction, independent of orientation.

In the back-to-back dipole picture, the fields cancel at zeroth order. Why does the surviving term come with kd rather than some other measure of the imperfection?

Because the only thing distinguishing the two half-sources, as seen from far away, is when their news left. Their amplitudes are equal and opposite by construction; their directions to the observer agree to $O(d/r)$ , which vanishes in the far zone; what cannot be transformed away is the emission-time offset $\Delta t_r\approx(2d\cos\theta)/c$ between the two halves. For harmonic sources a time offset is a phase offset $\omega\Delta t_r\sim kd\cos\theta$ , and the sum of two opposite signals with relative phase $\varphi$ has residual amplitude $2\sin(\varphi/2)\approx\varphi$ . So the residue is linear in $kd$ , with the $\cos\theta$ factor already foreshadowing the equatorial null: at $\theta=90°$ the two halves are equidistant, the phase offset vanishes, and the cancellation is exact. The formal expansion of Derivation A is this argument with its bookkeeping done properly.

Do M1 and E2 interfere with each other when both are present?

In the angular distribution, yes; in the total power, no. At each direction $\hat{\mathbf{n}}$ the two channels' fields add coherently, and since $M1$ radiates $\mathbf{E}\parallel\hat{\boldsymbol\phi}$ while $E2$ 's polarization is direction-dependent, the cross term generally survives and skews the lobes — measured angular asymmetries in atomic and nuclear emission are a standard way to extract mixing ratios between multipoles. But integrate over the whole sphere and the cross terms die: the multipole fields are orthogonal on the sphere at infinity (the general theorem lives in the next article, where the expansion is organized in vector spherical harmonics), so total powers add incoherently, $P=P_{M1}+P_{E2}$ . The pattern remembers the interference; the wattmeter does not.

The oxygen forbidden lines are visible in nebulae but not in the lab. Could we not just build a better vacuum?

We can, and we do — but the numbers are brutal and instructive. Visibility requires the radiative lifetime ( $\sim1$ s for the [O III] lines) to beat the collisional de-excitation time, and it also requires enough atoms along the line of sight to produce detectable photons. A nebula wins on both: densities near $10^8$ – $10^9$ m⁻³ make collisions hours apart, and the emitting column is light-years deep. A laboratory chamber at ultra-high vacuum can beat the collision requirement, but its centimeters of column length emit immeasurably few photons from a transition that fires once per second per excited atom. The modern resolution is not a better vacuum but a better geometry: trap a single ion, excite the metastable state deliberately, and watch it — trapped-ion spectroscopy routinely drives exactly these $E2$ lines, using their second-long lifetimes as the frequency references inside optical clocks. The nebula and the ion trap are the same physics at opposite extremes of quantity.

How literal is the analogy to gravitational waves?

Structurally very close, with two honest asterisks. Close: gravity has no monopole radiation (mass conservation) and no dipole radiation (momentum conservation — the mass dipole's second derivative is $\dot{\mathbf{P}}_{\text{total}}=0$ ), precisely mirroring charge conservation's kill of the EM monopole; the leading gravitational luminosity is a traceless-quadrupole formula, $P\propto G\langle\dddot{Q}_{ij}^2\rangle/c^5$ with the mass quadrupole, and the sphere-average machinery is identical to Derivation B's. The asterisks: first, the graviton's spin-2 character means the radiation field is itself a rank-2 tensor (two polarizations, $+$ and $\times$ ), not a vector field contracted with a tensor source — the angular patterns differ in detail from the EM quadrupole's. Second, the numerical prefactor ( $32/5$ in the standard form for a binary) comes from that tensor structure and from general relativity's field equations, not from Maxwell's. The multipole logic — conservation laws silencing low $\ell$ , $(kd)$ -type suppression, third-derivative sources — transfers exactly; the polarization structure and coefficients do not.

When the source is NOT small compared to the wavelength, which formula takes over?

None on this page — the systematic expansion of the next article does. Everything here assumed $kd\ll1$ twice over: once to truncate the Taylor expansion of the retarded time at first order, and once to use the static-form moments as the coefficients. When $kd\sim1$ (resonant antennas, nuclei emitting gamma rays with $\lambda\sim$ nuclear size, large phased arrays), three things happen: higher multipoles contribute at comparable strength; the exact multipole coefficients — integrals of the source against spherical Bessel functions $j_\ell(kr')$ , which reduce to the static moments only for $kr'\to0$ — must replace the moments used here; and the radiation pattern sharpens into directional lobes as terms up to $\ell_{\max}\sim kd$ pile up coherently. The vector-spherical-harmonic machinery that organizes all of this exactly, to every order, is the subject of the next article, with the full Jackson-register treatment waiting in the advanced chapter.