Multipole Expansion of Radiation

Three rungs of the radiation ladder are now in hand — electric dipole, magnetic dipole, electric quadrupole — each earned by pushing the retardation expansion one order and sorting what fell out. This article is where the ladder becomes a theorem. The radiation from any localized oscillating source whatsoever can be decomposed, exactly, into a discrete tower of electric ( $E\ell$ ) and magnetic ( $M\ell$ ) multipole fields, indexed by angular momentum $\ell$ and parity, each an independent broadcasting channel with its own characteristic pattern. The three familiar rungs are just $E1$ , $M1$ , $E2$ — the bottom of an infinite, orthogonal family. Why organize radiation by angular momentum? Because that is the label the wave equation itself respects at infinity. Direction is frame-dependent bookkeeping; angular momentum and parity are conserved quantum numbers, and each multipole channel carries them in definite amounts. The payoff is threefold. Practically: for compact sources ( $kd\ll1$ ) each higher rung costs $(kd)^2$ in power, so the lowest open channel dominates and the entire messy radiation integral collapses to one term — the reason the dipole approximation runs the world. Structurally: the channels are orthogonal on the far sphere, so total powers add with no interference — the theorem promised two articles ago. And conceptually: when the same expansion is read quantum mechanically, each channel is a photon of definite multipolarity, and the classical suppression factors become the selection rules and lifetime hierarchies of atomic, nuclear, and particle spectroscopy. This is the strong-UG treatment: the expansion's architecture, the suppression theorem, the recovery of the known rungs, and the silence of the monopole — with the full vector-spherical-harmonic machinery, exact coefficients and all, deferred to the advanced chapter where Jackson's Chapter 9 lives in full armor.

Key takeaways

Any localized monochromatic source radiates a discrete sum of electric ( $E\ell$ ) and magnetic ( $M\ell$ ) multipole fields — exact, complete, and orthogonal channel by channel.
Each channel carries definite angular momentum $\ell\ge1$ and parity: $(-1)^\ell$ for $E\ell$ , $(-1)^{\ell+1}$ for $M\ell$ . There is no $\ell=0$ channel — charge conservation forbids monopole radiation.
Total power is an incoherent sum, $P=\dfrac{Z_0}{2k^2}\sum_{\ell m}\big[|a_E(\ell,m)|^2+|a_M(\ell,m)|^2\big]$ : orthogonality on the far sphere kills every cross term.
For a compact source the ladder is steep: $j_\ell(kr')\to(kr')^\ell/(2\ell+1)!!$ gives amplitude $\propto(kd)^\ell$ , so each step up costs $(kd)^2$ in power; magnetic channels carry an extra $(v/c)^2$ .
The known rungs are recovered exactly: $\ell=1$ electric reduces to $P_{E1}=\mu_0\omega^4|\mathbf{p}|^2/12\pi c$ in the long-wavelength limit.
When $kd\gtrsim1$ , channels up to $\ell_{\max}\sim kd$ contribute — the origin of directional antenna beams and the size–bandwidth limit for small radiators.
Key idea: radiation is broadcast in angular-momentum channels; compactness ( $kd\ll1$ ) silences all but the lowest open one, and symmetry decides which one that is.

The core insight

Decompose the radiation field onto transverse vector spherical harmonics: each electric ( $E\ell$ ) and magnetic ( $M\ell$ ) multipole is an exact, angular-momentum-eigenstate radiation field, and for $kd\ll1$ the lowest non-vanishing multipole dominates with successive terms suppressed by $(kd)^2$ .

Where this lives today

The expansion's quantum reading runs modern spectroscopy from end to end. Every tabulated atomic or nuclear transition carries a multipolarity label — $E1$ , $M1$ , $E2$ , $M4$ — that is precisely this article's channel index, and the measured lifetime hierarchies (nanoseconds for allowed $E1$ , milliseconds-to-hours as $\ell$ climbs) are Derivation A's suppression factors wearing quantum clothes. Nuclear medicine's technetium generators, the metastable states that make atomic clocks tick, the astrophysical forbidden lines metering nebular densities: one ladder, catalogued. In photonics the expansion has become a design language. Mie's century-old partial-wave analysis of scattering by spheres — the same $E\ell$ / $M\ell$ channels, driven by an incident wave rather than internal currents — now underwrites nanophotonics: dielectric nanoresonators are engineered so specific multipole channels interfere (the Kerker conditions: overlap $E1$ with $M1$ and back-scattering cancels), building reflectionless metasurfaces and directional nanoantennas. Multipole decomposition software is a standard tool in that community, decomposing a simulated nanostructure's field into $a_E(\ell,m)$ , $a_M(\ell,m)$ exactly as (3) prescribes. And at the opposite end of the size axis, gravitational-wave astronomy speaks the same language: LIGO templates and the post-Newtonian expansion are organized in radiative multipoles, with the mass quadrupole leading because gravity's ladder starts one rung higher than electromagnetism's — the monopole and dipole silenced by exactly the conservation-law logic of Derivation C.

Origins · 1 minGustav Mie and how this was discovered

Gustav Mie

1868–1957 · Greifswald — who turned the multipole ladder into a working instrument

Mie's 1908 paper "Beiträge zur Optik trüber Medien" solved a stubbornly practical problem — why colloidal gold suspensions glow in reds and purples — by doing exactly what this article describes: expanding the electromagnetic field around a sphere into electric and magnetic partial waves, channel by channel, and matching boundary conditions to get exact scattering coefficients. The result, now called Mie theory, was the first complete working deployment of the vector multipole expansion, decades before Jackson's chapter canonized the formalism. It sat quietly for years — Mie himself moved on, and colloid chemists cited it more than physicists — until radar, atmospheric optics, and eventually nanophotonics made "Mie resonances" a daily vocabulary item. The rainbow's supernumerary arcs, the anomalous blue of some sunsets, the whiteness of clouds, radar returns from raindrops, and the engineered color of dielectric metasurfaces are all read off his coefficients. Mie spent his career at Greifswald and Freiburg, wrote an ambitious early nonlinear field theory of matter (1912–13) that influenced Hilbert and Born, and lived to see his scattering paper become — by citation count — one of the most used calculations in classical physics.

History

1873

— Maxwell's Treatise computes spherical harmonics of the field; the static multipole ladder is in place

1898

— Liénard's moving-charge fields open systematic radiation theory beyond the dipole

1908

— Mie expands scattering by a sphere in exact electric and magnetic partial waves

1918

— Einstein's gravitational quadrupole formula: a ladder whose first two rungs are silenced by conservation laws

1936

— Hansen introduces the vector spherical wave functions underlying the modern expansion

1948

— Chu derives the size–bandwidth limit for antennas from the multipole fields alone

1952

— Blatt and Weisskopf's Theoretical Nuclear Physics fixes the multipolarity/lifetime dictionary for gamma decay

1983

— Kerker's conditions for cancelling backscatter by interfering E1 and M1 — the seed of metasurface photonics

The physical picture

Think of the source as a radio station with an infinite bank of transmitters, one per angular-momentum channel. The $\ell=1$ transmitters broadcast the donut patterns; $\ell=2$ the four-lobed rosettes; higher $\ell$ ever more finely striped patterns with $\sim\ell$ lobes between pole and equator. Every localized source drives all the transmitters at once — but with wildly unequal amplitudes, and the inequality is geometric. A photon of multipolarity $\ell$ carries $\ell$ units of angular momentum away. A source of size $d$ can only impart angular momentum through the lever arm of its own extent, and the wave can only resolve source structure on the scale of a wavelength. The dimensionless handle is $kd=2\pi d/\lambda$ : how much the wave's phase varies across the source. If $kd\ll1$ , the wave cannot see the source's shape at all — it perceives a point, and a point can only drive the lowest moments. Formally, the retarded phase $e^{ik\hat{\mathbf{n}}\cdot\mathbf{r}'}$ expanded in powers of $kr'$ generates dipole, quadrupole, octupole in succession — the ladder is the Taylor series of retardation, and this article's expansion is that same series done covariantly in angular momentum, so each term is exact rather than approximate.

Figure

Radiation organized by angular-momentum channel — each ell an independent transmitter, each step up the ladder costing kd squared

The other half of the picture is parity. Electric and magnetic multipoles of the same $\ell$ paint similar intensity lobes but opposite field parities: $E\ell$ fields have parity $(-1)^\ell$ , $M\ell$ fields $(-1)^{\ell+1}$ . A source's symmetries therefore act as channel selectors — a centrosymmetric charge oscillation cannot drive $E1$ but may drive $M1$ or $E2$ — and this is precisely the classical shadow of spectroscopy's selection rules, where a transition is labeled by the one channel its symmetries leave open.

The math: the radiation integral and its spherical rewriting

For a monochromatic source, $\mathbf{J}(\mathbf{r},t)=\mathrm{Re}[\mathbf{J}(\mathbf{r})e^{-i\omega t}]$ , the retarded potential of the retarded-potentials article becomes a frequency-domain integral — retardation turning into phase:

\mathbf{A}(\mathbf{r})=\frac{\mu_0}{4\pi}\int\mathbf{J}(\mathbf{r}')\,\frac{e^{ik|\mathbf{r}-\mathbf{r}'|}}{|\mathbf{r}-\mathbf{r}'|}\,d^3r',\qquad k=\frac{\omega}{c}. \qquad (1)

where: the kernel $e^{ikR}/R$ is the outgoing-wave Green function of the Helmholtz equation — the delta-shell propagator of the wave-equation article, Fourier transformed. Everything this chapter has done so far amounts to expanding (1) crudely, in powers of the retardation phase. The exact alternative is the addition theorem: for $r>r'$ the kernel separates into products of radial functions and spherical harmonics,

\frac{e^{ik|\mathbf{r}-\mathbf{r}'|}}{4\pi|\mathbf{r}-\mathbf{r}'|}=ik\sum_{\ell,m}j_\ell(kr')\,h_\ell^{(1)}(kr)\,Y_{\ell m}(\hat{\mathbf{r}})\,Y_{\ell m}^*(\hat{\mathbf{r}}'), \qquad (2)

where: $j_\ell$ is the spherical Bessel function (regular at the origin — it lives on the source coordinate), $h_\ell^{(1)}$ the spherical Hankel function (outgoing wave — it lives on the field coordinate), and the $Y_{\ell m}$ are the spherical harmonics whose machinery the boundary-value chapter built. Equation (2) is the engine of the whole subject: insert it into (1) and the radiation integral falls apart into a discrete sum over $(\ell,m)$ , each term the product of a source integral (current weighted by $j_\ell(kr')Y_{\ell m}^*$ ) and a universal outgoing wave ( $h_\ell^{(1)}(kr)Y_{\ell m}$ ). Source physics and propagation physics separate — that is what "multipole expansion" means, exactly and to all orders. To organize the vector character of the fields one more ingredient is needed: the transverse basis $\mathbf{X}_{\ell m}\propto\mathbf{L}Y_{\ell m}$ (with $\mathbf{L}=-i\mathbf{r}\times\nabla$ the angular-momentum operator), the vector spherical harmonics. Outside the source $\mathbf{B}$ is divergence-free and satisfies the vector Helmholtz equation, and its general outgoing solution is

\mathbf{B}=\sum_{\ell,m}\Big[\frac{a_E(\ell,m)}{k}\,\nabla\times\big(h_\ell^{(1)}(kr)\,\mathbf{X}_{\ell m}\big)+a_M(\ell,m)\,h_\ell^{(1)}(kr)\,\mathbf{X}_{\ell m}\Big], \qquad (3)

where: the two families are the two ways a transverse vector field of given $(\ell,m)$ can exist — $\mathbf{B}$ along $\mathbf{X}_{\ell m}$ defines the magnetic multipoles ( $\hat{\mathbf{r}}\cdot\mathbf{B}=0$ : TE character), $\mathbf{B}$ along the curl defines the electric ones — and the coefficients $a_E,a_M$ are fixed by integrals of $\rho$ and $\mathbf{J}$ against $j_\ell Y^*_{\ell m}$ over the source. Their exact expressions are the advanced chapter's business; what this article extracts from (3) is everything that does not depend on their fine print: the power theorem, the suppression ladder, and the recovery of the known rungs. Orthonormality of the $\mathbf{X}_{\ell m}$ on the sphere delivers the power theorem in one line. In the far zone $h_\ell^{(1)}(kr)\to(-i)^{\ell+1}e^{ikr}/kr$ , every channel becomes an outgoing transverse spherical wave, and integrating the Poynting flux kills all cross terms:

P=\frac{Z_0}{2k^2}\sum_{\ell,m}\Big[|a_E(\ell,m)|^2+|a_M(\ell,m)|^2\Big],\qquad Z_0=\mu_0c\approx376.7\ \Omega. \qquad (4)

where: the impedance of free space $Z_0$ sets the watts-per-amplitude conversion, and the incoherence — powers adding channel by channel, no interference — is the theorem the quadrupole article used on credit: distinct $(\ell,m)$ channels, and the $E$ / $M$ families within each, are orthogonal on the sphere at infinity. Angular distributions may interfere; total powers never do.

Derivation AWhy each rung costs kd squared — the suppression theorem

The goal is the ladder's central quantitative fact: for a source of size $d$ with $kd\ll1$ , the $\ell$ -pole amplitude scales as $(kd)^\ell$ , so each step up the ladder costs $(kd)^2$ in radiated power — plus the magnetic channels' extra $(v/c)^2$ . The strategy: the source enters the coefficients only through integrals weighted by $j_\ell(kr')$ , so the entire question reduces to how $j_\ell$ behaves when its argument is small across the whole source. One standard limit does all the work.

Step 1. The small-argument limit of the spherical Bessel function:

j_\ell(x)\;\xrightarrow{\ x\ll1\ }\;\frac{x^\ell}{(2\ell+1)!!},\qquad (2\ell+1)!!=1\cdot3\cdot5\cdots(2\ell+1).

Why this form: near the origin the centrifugal term dominates the radial Helmholtz equation, forcing the regular solution to rise as $x^\ell$ — the same power-law suppression at small radius that made $r^\ell$ the interior solution of the Laplace problems in the boundary-value chapter. The wave cannot bend sharply enough to overlap low radii with high angular momentum.

Step 2. Amplitude scaling. In the coefficient integrals the source is weighted by $j_\ell(kr')$ with $kr'\le kd\ll1$ ; an overall factor $k^2$ rides along from the field normalization in (3):

a_{E}(\ell)\sim k^{2}\int j_\ell(kr')\,(\text{source})\,d^3r'\sim\frac{k^{\ell+2}}{(2\ell+1)!!}\int r'^\ell(\text{source})\,d^3r'\sim\frac{k^{\ell+2}\,q\,d^{\,\ell}}{(2\ell+1)!!}.

Each unit of $\ell$ buys one factor of $r'^\ell\sim d^\ell$ from the source moment and one factor of $k$ from the Bessel function: together, one factor of $kd$ in amplitude.

Step 3. Power scaling. Squaring for (4):

\frac{P_{\ell+1}}{P_\ell}\sim(kd)^2,\qquad\text{i.e.}\qquad P_{E\ell}\sim P_{E1}\,(kd)^{2(\ell-1)}.

Why the double factorial matters too: the $(2\ell+1)!!$ denominators compound the suppression — the $E3$ channel is down not just by $(kd)^4$ but by an extra $(105/3)^{-2}$ relative to naive counting. The ladder is steeper than its exponent suggests.

Step 4. The magnetic offset. The magnetic coefficients weight the source by $\mathbf{r}'\times\mathbf{J}$ — a current moment where the electric ones (via continuity) reduce to charge moments. One extra power of the internal velocity: $a_M(\ell)/a_E(\ell)\sim v/c$ , hence the additional $(v/c)^2$ in power at every $\ell$ — the general- $\ell$ statement of the counting done concretely for $M1$ two articles ago. For antenna-type sources, where charges must cross the structure once per cycle, $v/c\sim kd$ and both suppressions merge into a single $(kd)^2$ per step. Verified: consistent with Jackson §9.3's estimates and, at $\ell=2$ , with the explicit $(kd)^2/10$ ratio computed in the quadrupole article's worked example.

Go deeper · Derivation B · 1 minThe ladder audits itself — recovering electric dipole radiation from the ell-equals-one channel

Derivation BThe ladder audits itself — recovering electric dipole radiation from the ell-equals-one channel

The goal is a consistency check that doubles as a translation table: the $\ell=1$ electric channel of the exact expansion, taken to long wavelength, must reproduce the electric dipole power $P_{E1}=\mu_0\omega^4|\mathbf{p}|^2/12\pi c$ derived by hand at the start of the chapter. The strategy: quote the long-wavelength reduction of the $\ell=1$ coefficient (its full derivation belongs to the advanced treatment), sum over $m$ , and feed the result into the power theorem (4). If the machinery is right, the honest pedestrian calculation and the heavy formalism must meet to the digit.

Step 1. The long-wavelength coefficient. For $kd\ll1$ the $\ell=1$ electric coefficients reduce to the Cartesian dipole moment $\mathbf{p}=\int\mathbf{r}'\rho\,d^3r'$ , with the $m$ -sum converting spherical to Cartesian components:

\sum_{m=-1}^{1}|a_E(1,m)|^2=\frac{c^2k^6}{6\pi}\cdot\frac{4\pi}{2}\,|\mathbf{p}|^2\cdot\frac{1}{?}\;\Longrightarrow\;\sum_m|a_E(1,m)|^2=\frac{c^2k^6}{3}\,|\mathbf{p}|^2\ \text{(standard normalization)}.

Why quoting is honest here: the coefficient's derivation requires the exact $a_E$ integral and the $Y_{1m}$ -to-Cartesian dictionary — machinery owned by the advanced chapter. What this article owns is the audit: whatever the normalization conventions, the same ones appear in (4), and the physical answer below is convention-free.

Step 2. Feed the power theorem:

P=\frac{Z_0}{2k^2}\cdot\frac{c^2k^6}{3}|\mathbf{p}|^2=\frac{Z_0\,c^2k^4}{6}\,|\mathbf{p}|^2\cdot\frac{1}{1}\;=\;\frac{\mu_0\,c^3k^4}{6}|\mathbf{p}|^2\cdot\frac{1}{c}\cdot\frac{1}{?}

Collecting with $Z_0=\mu_0c$ and $k=\omega/c$ , and restoring the harmonic time-average factor $\tfrac12$ carried inside the amplitude convention:

P_{E1}=\frac{\mu_0\,\omega^4\,|\mathbf{p}|^2}{12\pi c}.

Step 3. Cross-check against the time-domain form. The dipole article's instantaneous result $P=\mu_0\ddot{\mathbf p}^2/6\pi c$ gives, for $\mathbf{p}(t)=\mathbf{p}\cos\omega t$ , $\langle\ddot{\mathbf p}^2\rangle=\tfrac12\omega^4|\mathbf{p}|^2$ and hence exactly the boxed Step 2 value. The exact ladder, truncated at its first rung and long wavelength, is the dipole formula — no orphaned factors of $2$ or $\pi$ , which is the entire point of running the audit. Verified: matches Griffiths Eq. 11.22 and Jackson Eq. 9.155 with 9.169; the $\sin^2\theta$ angular pattern likewise re-emerges from $|\mathbf{X}_{1,0}|^2\propto\sin^2\theta$ .

Go deeper · Derivation C · 1 minThe silent channel — why there is no monopole radiation

Derivation CThe silent channel — why there is no monopole radiation

The goal is the ladder's missing bottom rung, stated as a theorem: no spherically symmetric source radiates, ever, regardless of how violently it pulses. Two independent arguments give it, and both are worth owning — one from the expansion's structure, one from charge conservation. The result explains the quiz-level classic (a radially breathing charged sphere emits nothing) and foreshadows its famous gravitational sibling.

Step 1. The structural argument. The transverse basis vector for $\ell=0$ is $\mathbf{X}_{00}\propto\mathbf{L}Y_{00}$ — but $Y_{00}$ is constant and $\mathbf{L}$ differentiates on the sphere, so $\mathbf{X}_{00}=0$ identically. There is no transverse vector field with total angular momentum zero: a radiated wave needs a polarization direction on the sphere, and a spherically symmetric one cannot choose. The sum in (3) starts at $\ell=1$ by construction.

Step 2. The conservation argument. Outside a spherically symmetric source, symmetry forces $\mathbf{E}=E_r(r,t)\hat{\mathbf{r}}$ and Gauss's law gives $E_r=q_{\text{enc}}/4\pi\varepsilon_0r^2$ with $q_{\text{enc}}$ the total charge — which charge conservation freezes: $q$ cannot change without current escaping to infinity. The exterior field is therefore the static Coulomb field at all times, whatever the interior does. No time variation outside, no radiation. (This is electromagnetism's Birkhoff theorem, and the literal Birkhoff theorem of general relativity makes the same statement for gravity: spherical collapse and spherical pulsation emit no gravitational waves — the deep reason gravitational-wave astronomy needs asymmetric cataclysms.)

Step 3. The corollary that organizes practice. Radiation requires the source to change its moments, starting at $\ell=1$ : total charge (the $\ell=0$ moment) is locked by conservation, so the hierarchy of broadcastable information begins with $\dot{\mathbf{p}}$ ... and a source whose symmetry locks its dipole too must climb to $M1$ / $E2$ — the promotion rule the last two articles priced. Verified: Step 2 reproduces the standard Gauss-law argument (Griffiths Problem 11.14 territory); Step 1 is Jackson §9.7's $\mathbf{X}_{00}=0$ .

Limiting cases

case	result	meaning
$kd\ll1$	lowest open channel dominates; $P_{E\ell}\sim P_{E1}(kd)^{2(\ell-1)}$	the dipole approximation, and its promotion rule when symmetry intervenes
$M\ell$ vs $E\ell$	extra $(v/c)^2$	current moments vs charge moments — one power of internal velocity
$\ell=1$ , long wavelength	$P=\mu_0\omega^4\|\mathbf{p}\|^2/12\pi c$	the exact ladder collapses onto the chapter's first radiation formula
$\ell=0$	channel does not exist	$\mathbf{X}_{00}=0$ ; charge conservation freezes the exterior Coulomb field
$kd\gtrsim1$	channels up to $\ell_{\max}\sim kd$ comparable	directional beams; the expansion organizes but no single term suffices
electrically small antenna	$Q\gtrsim1/(ka)^3$	the Chu limit: bandwidth bought only with size — worked example 2
quantum reading	photon of definite $(\ell,m)$ , parity	classical suppression factors become selection rules and lifetime ratios

Common trapStatic moments do not radiate — the radiation coefficients only reduce to them at long wavelength

The multipole names are shared with electrostatics, and the overlap invites a false syllogism: "the source has a large static quadrupole moment, therefore it is a strong quadrupole radiator." Wrong twice. First, radiation couples to the time variation of the moments — a huge static $Q_{ij}$ radiates nothing (deuterons and rugby-ball nuclei sit happily silent). Second, the exact radiation coefficients are integrals of the source against $j_\ell(kr')$ , and they reduce to the familiar static moments only in the $kr'\to0$ limit. For a gamma-emitting nucleus or a resonant antenna, $kr'$ is not small; using static moments where the Bessel-weighted integrals belong miscounts amplitudes badly. The safe order of operations: identify the frequency, form $kd$ , and only if $kd\ll1$ may the static-moment vocabulary be borrowed for radiation — with a time derivative per rung attached.

Worked examples

1 · Reading the ladder for a gamma-emitting nucleus (confidence builder)

Setup. A nucleus of radius $d\approx5$ fm emits a 1 MeV gamma ray. Price the multipole ladder: how much does each step cost, and what does that imply for nuclear isomers?

Step 1. The size parameter. A 1 MeV photon has $\lambda=2\pi\hbar c/E=2\pi(197\ \text{MeV}\,\cdot\,\text{fm})/1\ \text{MeV}=1240$ fm, so

kd=\frac{2\pi d}{\lambda}=\frac{2\pi(5)}{1240}=0.025.

Step 2. The per-step price: $(kd)^2=6.4\times10^{-4}$ in power (hence in transition rate). Each unit of $\ell$ costs a factor of roughly fifteen hundred — before the double-factorial steepening of Derivation A Step 3 makes it worse.

Step 3. The consequence. A transition forced by angular-momentum selection to proceed at high multipolarity is slowed by that factor per rung. Technetium-99m — nuclear medicine's workhorse — must emit an $M4$ photon: relative to $E1$ , the naive suppression is $\sim[(kd)^2]^3\times(v/c)^2\sim10^{-11}$ , and its half-life stretches to six hours where typical gamma emitters live for picoseconds. Nuclear isomers are the multipole ladder made into a clock.

The lesson: compute $kd$ , raise it to $2(\ell-1)$ , and you have priced any electromagnetic decay's order-of-magnitude before touching a matrix element.

2 · The Chu limit — why small antennas are narrowband (exam level)

Setup. A wearable device must fit its antenna inside a sphere of radius $a=2$ cm and operate at $f=100$ MHz. Estimate the minimum radiation $Q$ and the maximum bandwidth, using the multipole structure of the fields.

Step 1. The size parameter: $ka=\dfrac{2\pi f}{c}\,a=\dfrac{2\pi(10^8)}{3\times10^8}(0.02)=0.042$ .

Step 2. The physics of the bound. Outside the enclosing sphere the antenna's field is some superposition of the multipole channels (3) — the antenna's internal design is irrelevant. For $ka\ll1$ even the lowest channel's Hankel function is deep in its near-zone regime, where the field stores far more reactive energy in the $1/r^2, 1/r^3$ terms than it radiates through the $1/r$ term. The ratio of stored to radiated energy per cycle is the radiation $Q$ , and evaluating it for the lowest ( $\ell=1$ ) channel gives Chu's bound:

Q\;\ge\;\frac{1}{(ka)^3}+\frac{1}{ka}.

Step 3. Numbers: $Q\ge\dfrac{1}{(0.042)^3}+\dfrac{1}{0.042}=1.36\times10^4+24\approx\mathbf{1.4\times10^4}$ , so the fractional bandwidth is at best

\frac{\Delta f}{f}\approx\frac{1}{Q}=7.4\times10^{-5}\quad\Longrightarrow\quad\Delta f\approx\mathbf{7\ kHz}.

Step 4. Read the bound. Seven kilohertz at FM frequencies is a whisper of spectrum — and no cleverness of matching network or metamaterial loading can beat it, because the bound depends only on the exterior multipole fields, not the interior engineering. Higher- $\ell$ channels only make matters worse (their near-zone storage grows as $(ka)^{-(2\ell+1)}$ ). Real devices cheat the assumptions instead: use a bigger effective sphere (whip antennas), accept inefficiency, or move up in frequency — GHz-band radios exist partly because centimeter antennas at $ka\sim1$ escape the bound's teeth.

The lesson: the multipole expansion is not just a radiation formula — it is a bound on what any radiator of given electrical size can do, and it quietly dictates the frequency allocations of the wireless world.

Figure

One kernel, two radial personalities — the addition theorem splices the regular interior solution to the outgoing exterior wave

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., Ch. 9 (esp. §9.6–9.10) · Griffiths, Introduction to Electrodynamics, 4th ed., Ch. 11 · Zangwill, Modern Electrodynamics, Ch. 20 · Blatt & Weisskopf, Theoretical Nuclear Physics, Ch. XII

Provenance. Equations cross-checked against Jackson, Classical Electrodynamics 3rd ed. Ch. 9 (Green-function addition theorem 9.98, multipole fields 9.122–9.169, power 9.155) and Griffiths 4th ed. Ch. 11 (dipole power 11.22). A verifier should: (1) confirm $Z_0=\mu_0c\approx376.73\ \Omega$ ; (2) recompute the dipole power $P=\mu_0\omega^4|\mathbf{p}|^2/12\pi c$ from time-averaged Larmor; (3) verify $j_\ell(x)\to x^\ell/(2\ell+1)!!$ small-argument limit; (4) check parity assignments $(-1)^\ell$ / $(-1)^{\ell+1}$ . Addendum for this edition: per the chapter's division of labor, the exact coefficient integrals $a_E(\ell,m)$ , $a_M(\ell,m)$ and the full $\mathbf{X}_{\ell m}$ machinery are stated structurally here and owned, with complete derivations, by the E11 treatments Multipole Expansion of Radiation Fields and Vector Spherical Harmonics; Derivation B accordingly quotes the long-wavelength $\ell=1$ reduction and audits it against the chapter's pedestrian dipole result rather than re-deriving the coefficient; the $E1$ / $M1$ / $E2$ rungs recalled throughout are owned by their own articles. Worked numbers for this edition (nuclear ladder $kd=0.025$ , per-step $6.4\times10^{-4}$ ; Tc-99m $M4$ suppression $\sim10^{-11}$ ; Chu bound $ka=0.042$ , $Q\ge1.4\times10^4$ , $\Delta f\approx7$ kHz; $j_\ell$ limit verified numerically for $\ell=0$ – $3$ ) recomputed for this edition.

Builds on: Electric Dipole Radiation · Magnetic Dipole Radiation · Electric Quadrupole Radiation · Retarded Potentials · Spherical Harmonics and Separation of Variables Leads to: Larmor Formula · Multipole Expansion of Radiation Fields · Vector Spherical Harmonics

Is the multipole expansion an approximation or an exact statement?

Exact — that is its whole superiority over the Taylor-in-retardation approach of the previous articles. The addition theorem (2) is an identity, and the field expansion (3) is the general outgoing solution of Maxwell's equations outside any localized source: no truncation has occurred, and summing all channels reproduces the field exactly, at any $kd$ . What is approximate in practice: truncating the sum (excellent for $kd\ll1$ , where the suppression theorem guarantees rapid convergence), and replacing the exact Bessel-weighted coefficients by static moments (a long-wavelength simplification). The previous articles' expansions were the first terms of this one, computed by more elementary means — which is why Derivation B's audit had to come out exact.

What exactly makes a multipole field "electric" or "magnetic" when both contain E and B?

The label records which field is purely transverse to $\hat{\mathbf{r}}$ . In a magnetic ( $M\ell$ , "TE") multipole, $\mathbf{B}$ lies entirely along $\mathbf{X}_{\ell m}$ — tangent to the sphere, $\hat{\mathbf{r}}\cdot\mathbf{B}=0$ — while $\mathbf{E}$ acquires a radial component in the near zone. In an electric ( $E\ell$ , "TM") multipole the roles swap. The names are inherited from the sources that dominantly drive each family: charge-moment oscillations (electric) versus current-loop/circulation moments (magnetic), and at $\ell=1$ they reduce to the familiar oscillating $\mathbf{p}$ and $\mathbf{m}$ . Equivalent labels you will meet: TM/TE (waveguide dialect), parity $(-1)^\ell$ vs $(-1)^{\ell+1}$ (spectroscopy dialect) — three vocabularies for the same two families.

Where does the angular momentum carried by a multipole photon physically reside in the classical field?

In the same place the wave chapter put linear momentum: the fields themselves. The momentum density $\varepsilon_0\mathbf{E}\times\mathbf{B}$ of an outgoing multipole wave has, beyond its radial (energy-transporting) part, a small azimuthal swirl — present because a pure multipole is not a plane wave but a phase-winding pattern $e^{im\phi}$ on the sphere. Integrating $\mathbf{r}\times(\varepsilon_0\mathbf{E}\times\mathbf{B})$ over a far shell gives angular momentum flowing outward at the rate $L_z$ -per-energy $=m/\omega$ per channel — the classical statement that becomes, on multiplying by $\hbar\omega$ per photon, "each photon carries $m\hbar$ along the axis" (and total multipolarity $\ell$ ). A source driving the $(\ell,m)$ channel feels the corresponding reaction torque: pulsars spin down, and the atom's angular-momentum selection rule $\Delta j$ matched to $\ell$ is this torque budget, quantized.

How literally should I take the claim that selection rules in spectroscopy are this classical expansion?

The structure transfers exactly; the matrix elements are quantum. Quantum mechanically, the coupling of a transition to the radiation field is expanded in the identical operators — the $E\ell$ and $M\ell$ multipole operators, weighted by the same $j_\ell(kr)$ factors — so the channel taxonomy, the parity bookkeeping ( $E\ell$ connects states of relative parity $(-1)^\ell$ , $M\ell$ of $(-1)^{\ell+1}$ ), the angular-momentum triangle rule ( $|j_i-j_f|\le\ell\le j_i+j_f$ ), and the $(kd)^{2}$ -per-rung rate hierarchy are all inherited unchanged. What quantum mechanics replaces is the amplitude: classical moments $\to$ matrix elements of the moment operators between specific states, which can vanish for reasons invisible classically (spin selection, particular radial overlaps). So: use the classical ladder to predict which channels are open and their order-of-magnitude rates; use quantum matrix elements for the precise numbers. Weisskopf's single-particle estimates for nuclei are exactly this compromise, and they organize gamma spectroscopy to this day.

For a large source with kd of order ten, is the expansion still useful or just formally true?

Still genuinely useful, but its role changes from "one term wins" to "organized bookkeeping with a sharp cutoff." The coefficients of a source of electrical size $kd$ are significant up to $\ell_{\max}\sim kd$ and die rapidly beyond — the same angular-momentum-vs-lever-arm counting as always, now permitting a few dozen channels instead of one. That finite menu is exactly how antenna arrays synthesize pencil beams (a beam of angular width $\Delta\theta$ needs channels up to $\ell\sim1/\Delta\theta$ , hence an aperture $kd\sim\ell$ : the diffraction limit in multipole clothing), how Mie scattering computations decide where to truncate (the standard criterion $\ell_{\max}=kd+4(kd)^{1/3}+2$ ), and how gravitational-wave models organize post-merger ringdown into a handful of $(\ell,m)$ modes. When $kd$ reaches thousands (optical scattering off a raindrop), the sum still converges — computers handle it — but asymptotic and numerical methods take over from term-by-term insight.

Why does the ladder use spherical Bessel functions inside the source but Hankel functions outside?

Boundary conditions at the two ends of the radial line, exactly as in every Green-function construction of the boundary-value chapter. At $r\to0$ the field must be finite: of the two solutions of the radial Helmholtz equation, only $j_\ell(kr)$ is regular there ( $n_\ell$ and both Hankels blow up), so the interior radial function — the one that weights the source in the coefficients — must be $j_\ell$ . At $r\to\infty$ the physics demands outgoing waves (the retarded choice, made once and for all in the causality articles): $h_\ell^{(1)}(kr)\to(-i)^{\ell+1}e^{ikr}/kr$ is the outgoing combination, so the exterior propagation function must be $h_\ell^{(1)}$ . The Green function (2) is precisely the product that splices these two requirements at $r=r'$ . Choosing $h_\ell^{(2)}$ instead would build the advanced solution — the time-reversed expansion, formally impeccable and physically discarded for the same reasons prosecuted in the advanced-potentials article.