Scattering by Conducting and Dielectric Spheres

Shine a plane wave on a sphere and ask where the light goes. This is the one scattering problem classical electrodynamics solves exactly — and the solution has organized a century of physics, from why the sky is blue to why radar engineers calibrate on metal balls to why a gold colloid glows ruby red. Everything is controlled by a single dimensionless dial, the size parameter $x = ka = 2\pi a/\lambda$ : turn it small and the sphere collapses to a pair of point dipoles whose cross section climbs as $\omega^4$ ; turn it to order one and the sphere becomes a resonator with a spectrum of sharp Mie modes; turn it large and the cross section settles onto twice the geometric area — a paradox this article states and its neighbor resolves. The small- $x$ story is richer than the folklore version. A dielectric sphere scatters as one electric dipole, inherited intact from the electrostatics chapter. But a conducting sphere carries a second, secret radiator — a magnetic dipole from expelled flux, half the strength of the electric one and opposite in sign — and the interference of the two tilts the pattern hard toward backscatter, nine to one, with a direction of perfect polarization at exactly $60°$ . Miss the magnetic dipole and every one of those numbers comes out wrong. The exact machinery is the multipole expansion wearing boundary conditions: coefficients $a_\ell, b_\ell$ as Riccati–Bessel ratios, whose $\ell = 1$ terms reproduce the dipole pair — certified here numerically to four digits. One problem, one dial, every regime.

Key takeaways

One dial rules everything: the size parameter $x = ka = 2\pi a/\lambda$ divides sphere scattering into Rayleigh ( $x \ll 1$ ), Mie resonant ( $x \sim 1$ ), and geometric ( $x \gg 1$ ) regimes.
Small dielectric sphere: one induced electric dipole with the electrostatic polarizability $\mathbf{p} = 4\pi\varepsilon_0 a^3\frac{\varepsilon_r - 1}{\varepsilon_r + 2}\mathbf{E}_0$ (the factor owned by Dielectric Sphere in a Uniform Field), radiating $\sigma = \frac{8\pi}{3}k^4a^6\left(\frac{\varepsilon_r-1}{\varepsilon_r+2}\right)^2$ — the $a^6/\lambda^4$ law.
Small conducting sphere: two moments, $\mathbf{p} = 4\pi\varepsilon_0 a^3\mathbf{E}_0$ and $\mathbf{m} = -2\pi a^3\mathbf{H}_0$ ; their interference gives $\frac{d\sigma}{d\Omega} = k^4a^6\left[\frac{5}{8}(1 + \cos^2\theta) - \cos\theta\right]$ — backscatter beats forward $9{:}1$ , total $\sigma = \frac{10\pi}{3}k^4a^6$ , perfect polarization at $\theta = 60°$ .
The exact solution expands all fields in the electric/magnetic multipole channels; boundary matching yields the Mie coefficients $a_\ell, b_\ell$ as Riccati–Bessel ratios, with the dipole pair as their $\ell = 1$ , small- $x$ limit (verified numerically here to $10^{-4}$ ).
Sharp Mie resonances appear where the coefficient denominators nearly vanish; for $x \gg 1$ the extinction efficiency tends to $Q_{\rm ext} \to 2$ — twice geometric, the extinction paradox owned by Optical Theorem.
Why clouds are white, delivered: droplets with $x \sim 100$ have left the $k^4a^6$ regime entirely — all colors scatter alike, and a 16-meter photon mean free path makes the cloud opaque.
Key idea: all sphere scattering is one boundary-value problem read at different $x = ka$ — a dipole pair when small, a resonator ladder when comparable, a shadow-casting disk when large.

The core insight

Sphere scattering is controlled entirely by $x = ka$ : small $x$ gives dipole scattering with $\sigma \propto a^6/\lambda^4$ (plus a magnetic dipole and $9{:}1$ backscatter for conductors), intermediate $x$ requires the full Mie partial-wave sum with its resonances, and large $x$ tends to the geometric-shadow limit $\sigma_{\rm ext} \to 2\pi a^2$ .

Where this lives today

Mie's coefficients are production code across half of applied physics. Climate science runs the $a_\ell, b_\ell$ sums billions of times daily inside radiative-transfer models — every satellite retrieval of aerosol load or droplet size is a fit to computed Mie curves, and the $Q_{\rm ext} \to 2$ plateau is baked into every cloud albedo estimate. Radar and lidar live on the same ladder from the other end: the conducting-sphere curve computed above is the calibration standard, and the $9{:}1$ backscatter bias sets detection budgets from insect-scale drones to space debris. Nanophotonics has made the small- $x$ story fashionable again under new names. Gold and silver nanoparticles sit on the Fröhlich pole $\varepsilon_r \to -2$ — the localized surface plasmon coloring stained glass and cancer-therapy colloids alike. High-index dielectric nanospheres (silicon, $\varepsilon_r \approx 16$ ) hunt the magnetic channel instead: the $b_1$ resonance gives a designer magnetic dipole without metal losses, and balancing it against $a_1$ — Derivation A's interference algebra tuned to cancel backscatter rather than enhance it — is the Kerker condition behind reflectionless Huygens metasurfaces. The century-old interference term has become a design dial.

Origins · 1 minGustav Mie and how this was discovered

Gustav Mie

1868–1957 · Greifswald, 1908 — the exact solution, written to explain the colors of gold colloids

Mie was not hunting a mathematical trophy. Colloid chemists of the 1900s could make gold suspensions in ruby red, violet, and blue from the same metal, and nobody could say why. Mie's 1908 paper in Annalen der Physik solved Maxwell's equations exactly for a sphere of arbitrary size and complex permittivity, summed the partial waves, and computed spectra for gold particles of increasing radius — matching the observed color march. The solution had been found independently by Ludvig Lorenz (1890, in Danish, largely unread) and was refound by Peter Debye (1909, for radiation pressure); "Lorenz–Mie–Debye theory" honors all three, but it was Mie's version — worked numerically against real colloid data — that made the machinery an instrument. He computed his Bessel-function sums by hand to $\ell = 3$ ; the same sums now run inside every climate model and radar simulator on Earth. His later career went to an early rival of general relativity (a different "Mie theory"), but the sphere is the immortality: the standard proof that exactly solvable and endlessly useful can be the same problem.

History

1871

— Rayleigh derives the

\lambda^{-4}

law for small scatterers: the blue sky quantified

1890

— Lorenz publishes the exact sphere solution in Danish; it goes unnoticed

1908

— Mie solves the sphere exactly to explain gold-colloid colors: the partial-wave ladder is born

1909

— Debye rederives the solution computing radiation pressure on cosmic dust

1944

— the resonance region enters radar practice; conducting-sphere curves become calibration standards

1957

— van de Hulst's Light Scattering by Small Particles systematizes the field

1983

— Bohren and Huffman publish the reference text and the Mie codes still in use

2013

— Kerker-condition experiments: dielectric nanospheres cancel backscatter by balancing the dipole pair

The physical picture

A plane wave arrives and shakes the sphere's charges in step. The sphere answers by re-radiating — and how it answers depends only on how many wavelengths fit across it. When the sphere is tiny compared to $\lambda$ , the incident field is effectively uniform and slowly breathing over the sphere's volume at any instant. The electrostatics chapter already solved this problem: a dielectric sphere in a uniform field develops the dipole moment with the $(\varepsilon_r - 1)/(\varepsilon_r + 2)$ factor of Dielectric Sphere in a Uniform Field. The single new ingredient is that the moment now oscillates, and an oscillating dipole radiates with the donut pattern and the $\omega^4$ power law of the radiation chapter. Volume gives $p \propto a^3$ ; radiated power squares it; hence $\sigma \propto a^6$ , with $k^4$ from the $\omega^4$ — short wavelengths scatter overwhelmingly more. A conducting sphere does something a dielectric cannot. It expels not just the electric field (surface charges) but also the oscillating magnetic field: induced eddy currents make it a perfect diamagnet, and a perfect diamagnet in a uniform $\mathbf{H}_0$ carries $\mathbf{m} = -2\pi a^3\mathbf{H}_0$ — exactly half the electric polarizability, with a minus sign. Two co-located radiators, phase-locked by the same incident wave, interfere. The cross term goes as $-\cos\theta$ : it suppresses the forward direction and reinforces the backward one. A small metal ball is a backscatter machine — the pattern below shows the $9{:}1$ contrast — which is precisely why it makes an honest radar target. Grow the sphere toward $x \sim 1$ and the quasi-static picture dies: the field varies across the sphere, higher multipoles wake up, and each partial wave $\ell$ behaves as a resonator with its own natural frequencies. The exact solution — the Mie ladder — takes over, and its coefficients ring. Grow further still and ray optics emerges, but with a diffraction surcharge: the total power removed tends to twice the geometric shadow.

Figure

Two computed small-sphere patterns, beam arriving from the left — a dielectric sphere radiates the symmetric dipole peanut, while a conducting sphere's electric-magnetic interference throws nine times more power backward than forward

The one dial

Every statement in this article is organized by one dimensionless number and one material ratio. Before any field is computed, agree on:

x \equiv ka = \frac{2\pi a}{\lambda}, \qquad m = \frac{n_{\rm sphere}}{n_{\rm medium}}. \qquad (1)

where: $a$ — sphere radius; $k$ — wavenumber in the surrounding medium; $m$ — relative refractive index ( $m^2 = \varepsilon_r$ for nonmagnetic spheres). The size parameter $x$ counts how many radians of phase the wave accumulates across one radius: $x \ll 1$ means the sphere never notices the wave's spatial structure (quasi-statics applies), $x \sim 1$ means retardation across the sphere is order unity (all multipoles couple), $x \gg 1$ means rays and shadows. A 20 nm aerosol particle in green light has $x \approx 0.23$ ; a 10 μm cloud droplet has $x \approx 114$ . Same material, same light — different physics, because different $x$ .

Derivation ASmall conducting sphere — the two induced moments and their interference pattern

The goal is the complete Rayleigh-limit answer for a perfectly conducting sphere: both induced moments, the polarized differential cross sections, the unpolarized average $\frac{d\sigma}{d\Omega} = k^4a^6[\frac{5}{8}(1 + \cos^2\theta) - \cos\theta]$ , and the total $\frac{10\pi}{3}k^4a^6$ . The strategy is quasi-static: at $x \ll 1$ the incident fields are uniform over the sphere at each instant, so the electrostatic and magnetostatic boundary-value problems — both solved in earlier chapters and recalled here by name — hand us the moments, and the radiation chapter's combined dipole far field does the rest. Watch the relative sign and the factor $\tfrac12$ : they carry all the angular structure.

Step 1. The electric moment. A conductor in a uniform field $\mathbf{E}_0$ is the $\varepsilon_r \to \infty$ limit of the dielectric sphere: the polarizability factor $(\varepsilon_r - 1)/(\varepsilon_r + 2) \to 1$ , so

\mathbf{p} = 4\pi\varepsilon_0 a^3\,\mathbf{E}_0.

Why the limit is legitimate: the grounded-conductor surface condition (equipotential, field expelled) is exactly what $\varepsilon_r \to \infty$ enforces in the E3 solution — the same uniform-field images construction that Method of Images and Dielectric Sphere in a Uniform Field own; nothing new is solved here.

Step 2. The magnetic moment. The oscillating $\mathbf{H}_0$ cannot penetrate a perfect conductor (eddy currents cancel it inside — the $\delta \to 0$ limit of Skin Depth and Reflection from Conductors), so the sphere responds as a perfect diamagnet, $\mu \to 0$ . The magnetizable-sphere solution of the E5 chapter with $\mu \to 0$ gives the polarizability factor $(\mu - \mu_0)/(\mu + 2\mu_0) \to -\tfrac12$ :

\mathbf{m} = 4\pi a^3\left(\frac{\mu - \mu_0}{\mu + 2\mu_0}\right)\mathbf{H}_0 \;\longrightarrow\; -2\pi a^3\,\mathbf{H}_0.

Half the electric strength, opposite sign — both facts are about to become geometry.

Step 3. The combined far field. Both moments oscillate at $\omega$ and radiate; the combined radiation field (electric dipole from Electric Dipole Radiation: Exact Fields and Zone Structure, magnetic dipole by the duality of Magnetic Dipole and Electric Quadrupole Radiation, both recalled) is

\mathbf{E}_{\rm sc} = \frac{k^2 e^{ikr}}{4\pi\varepsilon_0 r}\Big[(\hat{\mathbf{n}}\times\mathbf{p})\times\hat{\mathbf{n}} - \frac{1}{c}\,\hat{\mathbf{n}}\times\mathbf{m}\Big].

For the incident plane wave $E_0 = c\mu_0 H_0$ , so the magnetic term's amplitude is $|\mathbf{m}|/c = 2\pi a^3 H_0/c = 2\pi\varepsilon_0 a^3 E_0$ — exactly half of $|\mathbf{p}| = 4\pi\varepsilon_0 a^3 E_0$ .

Step 4. Polarized cross sections. Take incidence along $\hat{\mathbf{z}}$ , scattering direction at angle $\theta$ in a plane; resolve the incident polarization in and out of that plane. Projecting the two dipole patterns and their relative orientation (the algebra is two cross products per case) gives

\frac{d\sigma_\parallel}{d\Omega} = k^4a^6\left(\cos\theta - \tfrac12\right)^2, \qquad \frac{d\sigma_\perp}{d\Omega} = k^4a^6\left(1 - \tfrac12\cos\theta\right)^2.

Why one channel can die: at $\cos\theta = \tfrac12$ — that is, $\theta = 60°$ — the parallel amplitude vanishes identically: the electric and magnetic contributions cancel exactly in that channel. Light scattered at $60°$ from a small metal sphere is perfectly polarized perpendicular to the scattering plane, whatever the incident polarization.

Step 5. Unpolarized average and total. Averaging the two channels (numerically verified as an identity to $4\times10^{-16}$ across all angles):

\frac{d\sigma}{d\Omega} = \frac{1}{2}\left[\frac{d\sigma_\parallel}{d\Omega} + \frac{d\sigma_\perp}{d\Omega}\right] = k^4a^6\left[\frac{5}{8}(1 + \cos^2\theta) - \cos\theta\right].

Forward ( $\theta = 0$ ): $\tfrac14$ . Backward ( $\theta = \pi$ ): $\tfrac94$ . Backscatter wins, nine to one — the $-\cos\theta$ interference term suppresses forward and feeds backward. Integrating (the $\cos\theta$ term integrates to zero; $\int(1 + \cos^2\theta)\,d\Omega = 16\pi/3$ ):

\sigma_{\rm sc} = \frac{10\pi}{3}k^4a^6 = \frac{8\pi}{3}k^4a^6\left(1 + \tfrac14\right),

the electric dipole's total plus one quarter from $|\mathbf{m}/c|^2 = \tfrac14|\mathbf{p}|^2$ — the cross term redistributes angle but adds no total. Verified: matches Jackson Eqs. 10.13–10.14; the sphere integral recomputed numerically ( $10.4716$ vs $10\pi/3 = 10.4720$ ); forward/backward $= 1/4$ vs $9/4$ checked directly; the $\theta = 60°$ zero exact to machine precision. The base article's intuition text called this pattern "forward-peaked" — the computation says the opposite, and Jackson agrees: for the conducting sphere the interference favors backscatter (see the provenance addendum).

The dielectric sphere — electrostatics that radiates

The dielectric case needs no new boundary-value work at all: the E3 chapter's Dielectric Sphere in a Uniform Field already earned the interior-field factor $3/(\varepsilon_r + 2)$ and with it the induced moment, which we now simply let oscillate:

\mathbf{p}(t) = 4\pi\varepsilon_0 a^3\,\frac{\varepsilon_r - 1}{\varepsilon_r + 2}\,\mathbf{E}_0\,e^{-i\omega t}. \qquad (2)

where: $\varepsilon_r$ — the sphere's relative permittivity at the driving frequency $\omega$ (for optical light, $\varepsilon_r = n^2$ of the material — water droplets carry $\varepsilon_r \approx 1.77$ , not the static $80$ ). The factor $(\varepsilon_r - 1)/(\varepsilon_r + 2)$ is a dial with three famous settings: it vanishes at $\varepsilon_r = 1$ (index-matched sphere — invisible), it flips sign for $\varepsilon_r < 1$ (a bubble in water scatters with inverted contrast but the same cross section — only the square enters), and it diverges at $\varepsilon_r = -2$ , the Fröhlich condition where metals with negative permittivity host the localized surface-plasmon resonance that colors gold colloids.

Go deeper · Derivation B · 1 minRayleigh cross section of a small dielectric sphere

Derivation BRayleigh cross section of a small dielectric sphere

The goal is $\sigma_{\rm sc} = \frac{8\pi}{3}k^4a^6\left(\frac{\varepsilon_r - 1}{\varepsilon_r + 2}\right)^2$ — the quantitative $a^6/\lambda^4$ law — by the shortest honest route: take the electrostatic moment (2) as given by the E3 owner, price its radiation with the harmonic-dipole power owned by the radiation chapter, and divide by the incident flux. Every ingredient is a recall; the article's contribution is the assembly into a cross section, the object scattering theory trades in.

Step 1. Radiated power of the oscillating moment. For a harmonic dipole of amplitude $p_0$ , the cycle-averaged radiated power (the $P_{E1}$ result of Electric Dipole Radiation: Exact Fields and Zone Structure, recalled not re-derived) is

\langle P \rangle = \frac{\omega^4 p_0^2}{12\pi\varepsilon_0 c^3} = \frac{c\,k^4 p_0^2}{12\pi\varepsilon_0},

using $\omega = ck$ to trade frequency for wavenumber — the move that makes the $\lambda^{-4}$ visible.

Step 2. Incident flux. The time-averaged Poynting flux of the incident plane wave is $\langle S \rangle = \tfrac12\varepsilon_0 c E_0^2$ .

Step 3. The cross section. By definition, $\sigma$ is the power removed per unit incident intensity:

\sigma_{\rm sc} = \frac{\langle P \rangle}{\langle S \rangle} = \frac{k^4 p_0^2}{6\pi\varepsilon_0^2 E_0^2}.

Why this is the right definition: a cross section is the fictitious area that, intercepting the incident beam, would capture exactly the scattered power — it converts a field computation into the geometric currency detectors and mean-free-path arguments use.

Step 4. Insert the moment. With $p_0 = 4\pi\varepsilon_0 a^3\frac{\varepsilon_r-1}{\varepsilon_r+2}E_0$ from (2):

\sigma_{\rm sc} = \frac{8\pi}{3}\,k^4 a^6\left(\frac{\varepsilon_r - 1}{\varepsilon_r + 2}\right)^2.

The anatomy: $a^6$ because the moment is volumetric and power squares it; $k^4$ because dipole radiation prices acceleration twice; the dielectric factor squared because only the contrast with the surroundings scatters. The angular pattern is the unpolarized dipole peanut $\tfrac12(1 + \cos^2\theta)$ — symmetric fore and aft, since there is no second moment to interfere with. Verified: matches Jackson Eq. 10.11; the $\varepsilon_r \to \infty$ limit gives $\frac{8\pi}{3}k^4a^6$ , which is the conducting sphere's electric-dipole part alone — consistent with Derivation A's total $\frac{10\pi}{3} = \frac{8\pi}{3}(1 + \frac14)$ ; the small- $x$ Mie coefficient reproduces this to $0.03\%$ at $x = 0.05$ (Derivation C).

The promised delivery: why clouds are white. Rayleigh Scattering explains the blue sky with molecules deep in the $x \ll 1$ regime and defers the cloud question here, by name. A 10 μm droplet in green light has $x \approx 114$ : it has left the $k^4a^6$ regime entirely, its extinction saturated at $Q_{\rm ext} \approx 2$ — wavelength-independent — so red, green, and blue scatter alike: white. The numbers also make clouds opaque: $\sigma \approx 2\pi a^2 = 6.3\times10^{-10}$ m² per droplet at $10^8$ droplets/m³ gives a 16 m photon mean free path. Sky and cloud are the same physics at two settings of $x$ .

Go deeper · Derivation C · 2 minThe Mie frame — exact partial waves, and the dipole pair certified at ell equals one

Derivation CThe Mie frame — exact partial waves, and the dipole pair certified at ell equals one

The goal is the structure of the exact solution: the scattered field as a sum over electric and magnetic multipole channels with coefficients $a_\ell, b_\ell$ fixed by boundary matching, the perfect-conductor limit, and the numerical certification that the $\ell = 1$ terms at small $x$ reproduce Derivations A and B. The channel machinery itself is owned elsewhere — the exact $E\ell/M\ell$ decomposition by Multipole Expansion of Radiation Fields, the transverse harmonics $\mathbf{X}_{\ell m}$ by Vector Spherical Harmonics, both recalled by name — so this derivation's work is the matching, which is where the sphere's physics enters.

Step 1. Cast all three fields in the channel basis. Outside the sphere live the incident wave and the scattered wave; inside lives the transmitted field. The plane wave expands in regular (spherical-Bessel $j_\ell$ ) multipoles with known coefficients — one electric and one magnetic channel per $\ell$ , with $m = \pm1$ only, since a plane wave carries unit angular momentum about its propagation axis per photon. The scattered field must be outgoing (Hankel $h_\ell^{(1)}$ , the radiation condition), the interior field regular ( $j_\ell$ of argument $mkr$ ):

\mathbf{E}_{\rm sc} \propto \sum_\ell \big[i a_\ell\,\mathbf{N}^{(3)}_{\ell} - b_\ell\,\mathbf{M}^{(3)}_{\ell}\big], \qquad \mathbf{E}_{\rm int} \propto \sum_\ell \big[c_\ell\,\mathbf{M}^{(1)}_{\ell} - i d_\ell\,\mathbf{N}^{(1)}_{\ell}\big],

with $\mathbf{M} \propto j_\ell(kr)\,\mathbf{X}_{\ell m}$ the magnetic (TE) and $\mathbf{N} \propto \nabla\times\mathbf{M}/k$ the electric (TM) channel fields.

Step 2. Match at the surface. Continuity of tangential $\mathbf{E}$ and tangential $\mathbf{H}$ at $r = a$ — the electrodynamic boundary conditions of the E6 chapter — gives, channel by channel (the channels never mix: orthogonality of the $\mathbf{X}_{\ell m}$ , recalled), four linear equations per $\ell$ . Eliminating the interior amplitudes leaves the scattering coefficients as ratios of Riccati–Bessel functions $\psi_\ell(z) = z j_\ell(z)$ , $\xi_\ell(z) = z h_\ell^{(1)}(z)$ :

a_\ell = \frac{m\psi_\ell(mx)\psi'_\ell(x) - \psi_\ell(x)\psi'_\ell(mx)}{m\psi_\ell(mx)\xi'_\ell(x) - \xi_\ell(x)\psi'_\ell(mx)}, \qquad b_\ell = \frac{\psi_\ell(mx)\psi'_\ell(x) - m\psi_\ell(x)\psi'_\ell(mx)}{\psi_\ell(mx)\xi'_\ell(x) - m\xi_\ell(x)\psi'_\ell(mx)}.

Why the denominators matter most: $\xi_\ell$ is complex, so the denominators never quite vanish for real $m$ — but they pass near zero, and each near-miss is a Mie resonance: the sphere ringing as an open cavity in channel $\ell$ . For large dielectric spheres these are the whispering-gallery modes; for metals, the plasmon ladder.

Step 3. Cross sections from orthogonality. Because the channels are orthogonal on the sphere at infinity, powers add with no interference in the totals:

Q_{\rm sc} = \frac{\sigma_{\rm sc}}{\pi a^2} = \frac{2}{x^2}\sum_{\ell=1}^{\infty}(2\ell + 1)\big(|a_\ell|^2 + |b_\ell|^2\big), \qquad Q_{\rm ext} = \frac{2}{x^2}\sum_{\ell=1}^{\infty}(2\ell + 1)\,\mathrm{Re}(a_\ell + b_\ell).

The extinction sum — linear, not quadratic, in the coefficients — is the sphere's edition of the forward-amplitude bookkeeping owned by Optical Theorem, where its meaning and its famous factor-two limit are prosecuted.

Step 4. The perfect-conductor limit and the $\ell = 1$ certification. For a PEC sphere the interior field vanishes and the conditions collapse to $E_{\rm tan} = 0$ :

a_\ell = \frac{\psi'_\ell(x)}{\xi'_\ell(x)}, \qquad b_\ell = \frac{\psi_\ell(x)}{\xi_\ell(x)}.

Numerically, at $x = 0.01$ : $a_1 = -\tfrac{2i}{3}x^3$ to one part in $10^4$ , $b_1 = +\tfrac{i}{3}x^3$ , ratio $b_1/a_1 = -0.5000$ — the magnetic dipole at exactly half strength and opposite sign, and $Q_{\rm sc} = \frac{2}{x^2}\cdot3\left(\tfrac{4}{9} + \tfrac{1}{9}\right)x^6 = \tfrac{10}{3}x^4$ , reproducing Derivation A's total to four digits. For the dielectric sphere at $m = 1.33$ , $a_1$ matches $-\tfrac{2i}{3}x^3\frac{m^2-1}{m^2+2}$ to $0.01\%$ at $x = 0.05$ — Derivation B recovered, with $\varepsilon_r = m^2$ . Verified: coefficients against Bohren & Huffman Eqs. 4.56–4.57 (and the PEC forms against Jackson §10.4); small- $x$ certifications run numerically at $x = 0.01, 0.05, 0.1$ with the stated agreement; the exact PEC sum at $x = 0.21$ exceeds the Rayleigh formula by $1\%$ — the first whisper of the resonance region.

Figure

The computed exact scattering efficiency of a conducting sphere across three decades of size parameter — the Rayleigh fourth-power rise, the resonance overshoot near x of one, and the settling toward the geometric-optics value of two

Limiting cases

case	result	meaning
dielectric, $x \ll 1$	$\sigma = \frac{8\pi}{3}k^4a^6\left(\frac{\varepsilon_r-1}{\varepsilon_r+2}\right)^2$ , pattern $\propto \tfrac12(1+\cos^2\theta)$	one electric dipole; the $a^6/\lambda^4$ law
conductor, $x \ll 1$	$\sigma = \frac{10\pi}{3}k^4a^6$ , pattern $\frac58(1+\cos^2\theta) - \cos\theta$	$\mathbf{p}$ + $\mathbf{m}$ interference: $9{:}1$ backscatter, polarization zero at $60°$
$\varepsilon_r \to 1$	$\sigma \to 0$	index-matched sphere is invisible — only contrast scatters
$\varepsilon_r \to -2$	polarizability diverges (lossless limit)	Fröhlich condition: localized surface plasmon of metal nanoparticles
$x \sim 1$	full Mie sums; $Q_{\rm sc}$ rings, first PEC peak $Q \approx 2.3$ at $x \approx 1.2$	sphere as open resonator; whispering-gallery / plasmon ladders
$x \gg 1$	$Q_{\rm ext} \to 2$	extinction paradox — stated here, owned by Optical Theorem

Common trapThe forgotten magnetic dipole — and the direction the pattern actually leans

Two errors, one root. Error 1: treating a small metal sphere like a small dielectric — a single electric dipole. That loses a quarter of the total cross section ( $\frac{10\pi}{3}$ vs $\frac{8\pi}{3}k^4a^6$ ), the entire forward–backward asymmetry, and the perfectly polarized direction at $60°$ ; every one of those is the magnetic dipole's fingerprint. A perfect conductor must expel the oscillating $\mathbf{H}$ , and the resulting $\mathbf{m} = -2\pi a^3\mathbf{H}_0$ is only a factor of $2$ below the electric moment — never negligible. Error 2: remembering the interference but guessing its direction. The cross term is $-\cos\theta$ : it starves the forward direction and feeds backscatter, $9{:}1$ — the reverse of the beamed-forward intuition imported from large-particle (Mie, $x \gtrsim 1$ ) scattering, where diffraction really does throw a forward lobe. Small conducting spheres shine backward (radar's good fortune); large spheres of any kind shine forward (the diffraction half of the extinction paradox). Check which regime of $x$ you are in before trusting either instinct.

Worked examples

1 · The radar calibration sphere (confidence builder)

Setup. A perfectly conducting sphere of radius $a = 1$ cm in a 1 GHz radar beam. Find the size parameter, the total scattering cross section, the backscatter (radar) cross section, and check the Rayleigh formulas against the exact Mie sum.

Step 1. The dial. $\lambda = c/f = 30.0$ cm, so $x = ka = 2\pi a/\lambda = \mathbf{0.210}$ — comfortably Rayleigh, and $k^4a^6 = 1.93\times10^{-7}$ m².

Step 2. Total cross section. $\sigma_{\rm sc} = \frac{10\pi}{3}k^4a^6 = \mathbf{2.02\times10^{-6}}$ m² — an efficiency $Q_{\rm sc} = \sigma/\pi a^2 = 6.4\times10^{-3}$ : at a fifth of a radian of size parameter, the sphere intercepts less than $1\%$ of its geometric share. The exact Mie sum at this $x$ gives $Q = 6.50\times10^{-3}$ , $1\%$ above Rayleigh — the resonance region announcing itself.

Step 3. Radar cross section. Radar convention quotes $\sigma_{\rm RCS} = 4\pi\,(d\sigma/d\Omega)|_{\theta=\pi}$ — the backscatter direction, where the pattern is at its $\tfrac94 k^4a^6$ maximum: $\sigma_{\rm RCS} = 9\pi k^4 a^6 = \mathbf{5.46\times10^{-6}}$ m², a factor $2.7$ above the total $\sigma_{\rm sc}$ . The $9{:}1$ interference bias is not a curiosity — it is the radar engineer's signal budget.

Step 4. The $\lambda^{-4}$ leverage. Doubling the frequency multiplies every cross section above by $16$ — why small debris hides from long-wavelength radar, and why calibration spheres are sized to the band.

The lesson: one measured sphere at one frequency pins the whole curve — the exact Mie ladder is the only calibration standard radar needs.

2 · One material, two skies — haze droplet vs cloud droplet (exam level)

Setup. Water at 550 nm ( $\varepsilon_r = n^2 = 1.77$ ). Compare a 20 nm haze droplet with a 10 μm cloud droplet: cross sections, spectral behavior, and the photon mean free path in a cloud ( $n_d = 10^8$ droplets/m³).

Step 1. The haze droplet. $x = 2\pi(20\,\text{nm})/550\,\text{nm} = 0.23$ — Rayleigh. The dielectric factor: $(\varepsilon_r - 1)/(\varepsilon_r + 2) = 0.77/3.77 = 0.204$ . Then $\sigma = \frac{8\pi}{3}k^4a^6(0.204)^2 = \mathbf{3.8\times10^{-19}}$ m² with $k = 1.14\times10^7$ m⁻¹. Blue-to-red contrast: $(650/450)^4 = 4.4$ — this droplet still scatters selectively, tinting hazes.

Step 2. The cloud droplet. $x = 2\pi(10\,\mu\text{m})/550\,\text{nm} = \mathbf{114}$ — deep geometric regime, $Q_{\rm ext} \to 2$ : $\sigma_{\rm ext} = 2\pi a^2 = \mathbf{6.3\times10^{-10}}$ m², independent of wavelength. Nine orders of magnitude above the haze droplet — and flat across the visible, hence white.

Step 3. The mean free path. $\ell_{\rm mfp} = 1/(n_d\,\sigma_{\rm ext}) = 1/(10^8 \times 6.3\times10^{-10}) = \mathbf{16}$ m. A photon entering a 300 m thick cloud scatters $\sim20$ times: direction randomized, spectrum untouched — a white, opaque ceiling.

Step 4. The crossover reading. Between 20 nm and 10 μm the scaling walks from $\sigma \propto a^6$ (Rayleigh, volume-squared) to $\sigma \propto a^2$ (geometric, area): growth by $500\times$ in radius buys $10^9$ in cross section, not $(500)^6 \approx 10^{16}$ — the Mie region in between absorbs the difference in its ringing ladder.

The lesson: same water, same sunlight — the size parameter alone decides between a blue-tinting haze and a white opaque cloud.

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., Ch. 10 (Scattering and Diffraction) · Griffiths, Introduction to Electrodynamics, 4th ed., Ch. 9, 11 · Bohren & Huffman, Absorption and Scattering of Light by Small Particles · Zangwill, Modern Electrodynamics, Ch. 21 · van de Hulst, Light Scattering by Small Particles

Provenance. Rayleigh cross section: derive $\sigma=\frac{8\pi}{3}k^4a^6\left(\frac{\varepsilon-\varepsilon_0}{\varepsilon+2\varepsilon_0}\right)^2$ from the static dipole and Larmor formula — verify against Jackson 3rd ed. Eq. (10.11) and Griffiths Ch. 9. Conducting-sphere moments $\mathbf{p}=4\pi\varepsilon_0a^3\mathbf{E}_0$ , $\mathbf{m}=-2\pi a^3\mathbf{H}_0$ and unpolarized $\frac{\mathrm{d}\sigma}{\mathrm{d}\Omega}=k^4a^6[\frac58(1+\cos^2\theta)-\cos\theta]$ with total $\frac{10\pi}{3}k^4a^6$ : verify against Jackson Eqs. (10.13)–(10.14). Note sign/convention of the forward vs backward peak (the relative phase of $\mathbf{m}$ ) differs across texts; a verifier should recompute the polarized cross sections $\sigma_{\parallel,\perp}$ to confirm the fully-polarized angle at $\cos\theta=\tfrac12$ . Mie coefficients $a_\ell,b_\ell$ in Riccati–Bessel form: cross-check Bohren & Huffman Eqs. (4.56)–(4.57) and the efficiency sums $Q_{\rm sc},Q_{\rm ext}$ . The $\lambda^{-4}$ scaling and extinction paradox $Q_{\rm ext}\to2$ are standard (van de Hulst). No fabricated numerical constants used. Addendum for this edition: the flagged convention question was settled by direct computation — the polarized channels recomputed from the combined dipole far field give forward $\tfrac14$ vs backward $\tfrac94$ ( $9{:}1$ backscatter) with the $60°$ zero exact; the base's intuition text calling the pattern "forward-peaked" contradicted its own eqCards and is corrected throughout (base error fixed). All small- $x$ limits were certified against exact Mie coefficients ( $b_1/a_1 = -0.5000$ at $x = 0.01$ ; Rayleigh recovery to $0.03\%$ ; exact PEC sum $+1\%$ at $x = 0.21$ ); the $Q_{\rm sc}(x)$ SVG is traced from the computed exact sum (peak $Q = 2.29$ at $x \approx 1.2$ ). Division of labor: the polarizability factor is owned by Dielectric Sphere in a Uniform Field, the channel machinery by Multipole Expansion of Radiation Fields and Vector Spherical Harmonics, the dipole radiation formulas by the E11 radiation set — recalled by name; the extinction paradox is deferred to Optical Theorem, the molecular sky story to Rayleigh Scattering, whose cloud question is delivered here. All worked numbers recomputed for this edition.

Builds on: Multipole Expansion of Radiation Fields · Vector Spherical Harmonics · Electric Dipole Radiation: Exact Fields and Zone Structure · Magnetic Dipole and Electric Quadrupole Radiation Leads to: Antenna Arrays and Directivity · Synchrotron Radiation

Why does a perfect conductor acquire a magnetic dipole moment when a dielectric does not?

Because a perfect conductor excludes time-varying magnetic flux. The oscillating $\mathbf{H}_0$ drives surface eddy currents that cancel the interior field — the sphere behaves as a perfect diamagnet ( $\mu \to 0$ ), and the magnetostatic sphere solution assigns it $\mathbf{m} = -2\pi a^3\mathbf{H}_0$ . A nonmagnetic dielectric has no mobile charge to run those currents; its molecular diamagnetism is negligible at $x \ll 1$ . The $-\tfrac12$ relative to the electric polarizability is pure geometry — the $(\mu-\mu_0)/(\mu+2\mu_0)$ dial at $\mu = 0$ against the $(\varepsilon_r-1)/(\varepsilon_r+2)$ dial at $\varepsilon_r = \infty$ — which is why it survives every convention change.

Is the conducting-sphere pattern forward-peaked or backward-peaked? Different books seem to disagree.

Backward-peaked — decisively: $\tfrac94$ backward against $\tfrac14$ forward, from the $-\cos\theta$ interference term. The confusion has two honest sources: some texts measure $\theta$ from the backward direction, which flips the visual; and intuition trained on large particles — where diffraction genuinely throws a forward lobe — gets imported into the Rayleigh regime. This article's convention is standard (forward $= \theta = 0$ ), the polarized channels were recomputed from the combined far field, and the numbers verified against Jackson Eq. 10.14. Rederive with $\mathbf{m}$ 's sign flipped and you get forward-peaking with a polarization zero at $120°$ — both wrong: the $60°$ zero is the experimental tiebreaker.

Where exactly does the quasi-static approximation break, and how would I know?

Its first casualty is visible in this article's own numbers: at $x = 0.21$ the exact Mie sum already exceeds the Rayleigh formula by $1\%$ . The approximation assumes a uniform incident field and instantaneous response; both fail at relative order $x^2$ (retardation across the sphere) — expect percent-level errors near $x \approx 0.1$ and qualitative failure by $x \approx 0.5$ , where the $\ell = 2$ channels wake up. The practical test costs nothing: compute $a_1, b_1$ exactly (two Bessel ratios) and compare with $-\tfrac{2i}{3}x^3$ times the polarizability factor. When they diverge, you have left Rayleigh.

What physically happens at a Mie resonance?

The sphere becomes a ringing cavity for one partial wave. Each channel $\ell$ must match an interior radial structure to an exterior outgoing wave at the surface; at special combinations of $x$ and $m$ the match is nearly perfect for a free oscillation — the denominator of $a_\ell$ or $b_\ell$ passes close to zero, and the driven response spikes. For high-index dielectric spheres these are whispering-gallery modes (quality factors up to $10^8$ in optical microspheres — microsphere lasers, single-virus biosensors); for metals with $\varepsilon_r < 0$ the same denominators host the plasmon ladder, with the $\ell = 1$ Fröhlich mode at $\varepsilon_r = -2$ coloring gold colloids. The overshoot in this article's $Q_{\rm sc}$ curve ( $Q = 2.29$ at $x \approx 1.2$ ) is the same phenomenon in its mildest form.

Why do the totals add as a-squared plus b-squared with no cross terms, when the angular pattern clearly shows interference?

Because the channels are orthogonal on the far sphere while directions are not. At a given angle all multipole amplitudes superpose coherently — exactly where Derivation A's $-\cos\theta$ cross term lives. Integrating over solid angle kills every cross term by the orthogonality of the vector spherical harmonics (owned by Vector Spherical Harmonics), electric against magnetic included. So totals are incoherent sums — the same theorem behind the clean power formula of Multipole Expansion of Radiation Fields — while patterns are coherent. Both statements are exact; confusing their domains is a standard exam trap.

Does absorption change the Mie framework?

No — it is already inside. Let the sphere's permittivity be complex ( $m = n + i\kappa$ ) and the same boundary matching yields the same $a_\ell, b_\ell$ formulas with complex argument $mx$ ; nothing structural changes. What changes is the bookkeeping: $Q_{\rm sc}$ (the quadratic sum) now counts only re-radiated power, while $Q_{\rm ext}$ (the linear, forward-amplitude sum) counts everything removed from the beam, and the difference $Q_{\rm abs} = Q_{\rm ext} - Q_{\rm sc} \geq 0$ is the Ohmic deposit. For gold nanoparticles near the plasmon resonance, absorption dominates below $\sim40$ nm radius (the principle of photothermal therapy) while larger particles mostly scatter — why imaging and therapy colloids are sized differently. The energy audit behind the $Q_{\rm ext}$ sum is the subject of Optical Theorem.

Is Rayleigh scattering by air molecules just this article's dielectric-sphere formula at tiny x?

Structurally yes, with one substitution and one subtlety. A molecule is not a continuum sphere, so the sphere polarizability $4\pi\varepsilon_0 a^3\frac{\varepsilon_r-1}{\varepsilon_r+2}$ is replaced by the molecular polarizability $\alpha(\omega)$ — the object Clausius–Mossotti Relation connects to the bulk index — and the $k^4|\alpha|^2$ structure survives untouched. The subtlety is coherence: $N$ molecules per cubic wavelength should naively interfere; ideal-gas density fluctuations restore the independent-molecule answer. That whole story — the $\omega^4$ derivation from the driven oscillator, the sky's numbers, the fluctuation argument — is owned by Rayleigh Scattering; this article supplies the sphere limit it calibrates against, and delivers the question it defers: clouds, where $x$ is large and whiteness is the answer.