Rayleigh Scattering

"Why is the sky blue?" has a short answer — short wavelengths scatter more — and this article exists for the long one. The strong-UG article Electric Dipole Radiation already showed the exponent: dipole power carries $\omega^4$ , so with the dipole amplitude fixed, blue beats red by $(\lambda_{\rm red}/\lambda_{\rm blue})^4$ . What it could not answer is the load-bearing question hiding in "amplitude fixed": why does an air molecule's induced dipole not care about frequency? A free electron's would (that is Thomson Scattering, flat in $\omega$ ); a resonant one's would care catastrophically. The answer is the bound electron's spring: drive it far below resonance and its displacement locks to the field with a stiff, frequency-independent polarizability — air's ultraviolet resonances sit near 100 nm, the visible sits far downhill, and the $\omega^4$ of re-radiation survives untouched. This article derives the whole response function honestly: one driven-oscillator formula, $\sigma(\omega) = \sigma_T\,\omega^4/[(\omega_0^2-\omega^2)^2 + \gamma^2\omega^2]$ , containing Rayleigh's law, Thomson's plateau, and resonance fluorescence as its three floors. Then it does the sky quantitatively — molecular cross section from the measured refractivity of air, vertical optical depths, sunset extinction ratios, the polarization at $90°$ , and the honest fine print (why not violet; why the exponent is measurably a bit above 4). Last comes the derivation's guilty secret: $10^7$ molecules sit inside every cubic wavelength, so why do they scatter as independents rather than cancel like a dense medium? The resolution — density fluctuations, priced by the coherence theorem of Larmor Formula from the Liénard–Wiechert Fields — is why the sky is bright, why liquids are dim, and how Rayleigh weighed the molecule itself. The bare slug's full optics story lives in the Optics track; this is the electrodynamics engine underneath it.

Key takeaways

Rayleigh scattering = elastic re-radiation by a polarizable object much smaller than $\lambda$ , driven far below its resonances: induced dipole $\mathbf{p} = 4\pi\varepsilon_0\alpha\mathbf{E}$ with $\alpha$ frequency-stiff, so the dipole $\omega^4$ passes straight through to the cross section: $\sigma = \dfrac{8\pi}{3}k^4|\alpha|^2 \propto \lambda^{-4}$ .
One bound-electron formula holds all regimes: $\sigma(\omega) = \sigma_T\,\dfrac{\omega^4}{(\omega_0^2-\omega^2)^2 + \gamma^2\omega^2}$ — Rayleigh floor ( $\omega\ll\omega_0$ ), resonance tower ( $\sigma_{\max} = 6\pi\bar{\lambda}_0^2$ , $\sim10^{15}\,\sigma_T$ for an optical line), Thomson plateau ( $\omega\gg\omega_0$ ).
The same oscillator drives Absorption and Dispersion's $\varepsilon(\omega)$ — one response function, spent there on propagation, here on scattering; air's refractivity $n-1 = 2.79\times10^{-4}$ hands us the molecular cross section with no atomic theory: $\sigma(550\,\mathrm{nm}) \approx 4.5\times10^{-31}\ \mathrm{m}^2$ .
Sky, by the numbers (computed): vertical optical depth $\tau = 0.22 / 0.098 / 0.050$ at 450/550/650 nm — zenith transmission 80/91/95%; at a $\sim$ 38-airmass sunset path the blue survives at $2.5\times10^{-4}$ while the red keeps 15% — blue sky, red sunset, one exponent.
Skylight at $90°$ from the Sun is strongly linearly polarized — Thomson Scattering's $\Pi(\theta)$ transfer, capped near 94% by molecular anisotropy and lower in practice by multiple scattering.
The independent-scatterer sum is justified by fluctuations: ideal-gas density fluctuations $\langle\Delta N^2\rangle = \bar{N}$ make the incoherent result exact (Smoluchowski–Einstein); dense liquids scatter $\sim$ 16× less per molecule (computed for water).
Key idea: a small polarizable particle in an oscillating field becomes a tiny radiating dipole; since dipole radiated power $\propto \omega^4|\mathbf{p}|^2$ and $\mathbf{p} = \alpha\mathbf{E}$ is frequency-independent at low $\omega$ , the scattering cross section scales as $\omega^4 \propto \lambda^{-4}$ .

Where this lives today

The $\lambda^{-4}$ law is infrastructure. Optical fibers are quota-limited by it: frozen-in density fluctuations of silica (Derivation C's physics, vitrified) set a Rayleigh floor of $\sim$ 0.13 dB/km near 1550 nm — why telecom lives at 1.55 μm, and the wall submarine cables engineer against. Atmospheric science runs the law daily: Rayleigh lidars read stratospheric density and temperature from the known molecular return; satellite radiometers subtract the computable Rayleigh signal to retrieve aerosols. Astronomy reads the exponent at range: exoplanet transmission spectra sloping up toward the blue betray clear hydrogen skies (measured on HD 189733b and dozens since), while flat spectra mean clouds — the Rayleigh/Mie dichotomy as a remote cloud sensor. At the bench, the resonant floor became a technology: near-unitary $3\lambda^2/2\pi$ cross sections underlie laser cooling and single-atom imaging — and the $90°$ polarization transfer recurs in insect skylight compasses and the CMB story next door in Thomson Scattering.

Origins · 1 minLord Rayleigh and how this was discovered

Lord Rayleigh

1842–1919 · Terling Place and the Cavendish — the sky explained, then weighed

John William Strutt, third Baron Rayleigh, solved the sky twice. In 1871 — before Maxwell's theory was orthodoxy — he derived the $\lambda^{-4}$ law from dimensional analysis and correctly located the scatterers in the air molecules themselves, against the era's consensus of suspended dust (Tyndall's aerosols had misled everyone). His 1881 and 1899 revisits translated the law into electromagnetic language and executed Derivation C's astonishing inversion: from atmospheric transparency plus air's refractivity, a count of the molecules in a cubic meter — the sky as a molecular scale. This is the same Rayleigh whose resolution criterion Antenna Arrays and Directivity borrows (that band belongs there); scattering was one province of an empire spanning the theory of sound, surface waves, and the discovery of argon (Nobel Prize, 1904) — some 450 papers from a private laboratory at his family estate, none careless.

History

1869

— Tyndall scatters white light off laboratory aerosols: blue sideways, polarized — but credits dust

1871

— Rayleigh derives the λ⁻⁴ law and pins the sky on the molecules of air itself

1899

— Rayleigh inverts sky transparency plus refractivity into a count of the molecules

1908

— Smoluchowski frames scattering by density fluctuations; Mie solves the large sphere

1910

— Einstein quantifies fluctuation scattering; critical opalescence becomes molecular evidence

1914

— Fowle's Mount Wilson transmission data yield Avogadro's number from the sky to a few percent

1928

— Raman finds the inelastic sibling: scattered light with shifted frequency

1966

— Kao identifies Rayleigh scattering as the fundamental loss floor of optical fibers

2008

— Rayleigh slope measured in an exoplanet's transmission spectrum (HD 189733b)

The physical picture

Every scattering problem in this chapter is one machine with a different spring: wave shakes charge, charge re-radiates the dipole donut, cross section = re-radiated power over incident flux. Thomson Scattering ran it springless — cross section flat. Here the electron is bound, and everything turns on where the drive sits relative to the resonance. Far below resonance — visible light against air's far-ultraviolet lines — the spring wins the tug-of-war with inertia. The electron's displacement faithfully tracks the instantaneous field: $x = qE/m\omega_0^2$ , no $\omega$ in sight. The molecule is a dipole of fixed, frequency-blind amplitude, gently waved at whatever rate the light demands. But radiation is paid in acceleration, and differentiating a fixed-amplitude oscillation twice costs $\omega^2$ — squared in the power, that is the $\omega^4$ . Blue light does not grip air molecules any harder than red; it merely waves the same little dipole faster, and the Larmor meter charges quartically for speed. The geometry rides along from the dipole pattern. Skylight reaching you at $90°$ from the Sun comes from dipoles driven transverse to your line of sight — one driving axis invisible (the donut's null), the other fully visible — so that patch is strongly linearly polarized, Thomson Scattering's transfer function verbatim. Rotate polarized sunglasses against it and the sky dims and brightens; bees navigate by the same pattern. The one dishonesty in the naive picture is the word "independent." A cubic wavelength of air holds ten million molecules; forward, their re-radiations add coherently and build the refractive index (bookkeeping owned by Optical Theorem); sideways, a perfectly uniform medium would cancel to blackness. The sky is visible only because air fluctuates — and pricing that fluctuation, in Derivation C, is how the molecule was first counted.

Figure

The whole response of one bound electron computed from the driven-oscillator cross section — Rayleigh's omega-fourth ramp, the resonance tower, and Thomson's plateau on one log-log curve

The response function

The scatterer's entire personality is its polarizability. We use the volume convention $\mathbf{p} = 4\pi\varepsilon_0\,\alpha\,\mathbf{E}$ ( $\alpha$ in m³; the dictionary to Induced Dipoles and Atomic Polarizability's convention is $\alpha_{\rm SI} = 4\pi\varepsilon_0\alpha$ ). The machine then has two moving parts: what $\alpha(\omega)$ is, and what a radiating dipole costs. The second is settled upstream — the $\omega^4$ Poynting ledger of Electric Dipole Radiation, exact edition in Electric Dipole Radiation: Exact Fields and Zone Structure — so everything new lives in $\alpha(\omega)$ : the same Lorentz oscillator whose real part builds the refractive index in Absorption and Dispersion. One response function, two customers. That identity is not a coincidence to admire but a tool we will spend twice — first to get the ladder (Derivation A), then to extract air's cross section from its measured refractivity without solving any molecule (Derivation B).

Derivation AThe universal ladder — one oscillator from Rayleigh through resonance to Thomson

The goal is the full frequency dependence of scattering by one classically bound electron: the formula $\sigma(\omega) = \sigma_T\,\omega^4/[(\omega_0^2-\omega^2)^2+\gamma^2\omega^2]$ , its three regimes, and the resonance ceiling. The route: solve the driven-damped oscillator for the induced dipole, feed its amplitude to the dipole-radiation meter, divide by the flux. The Larmor/dipole power and the plane-wave flux are imports, recalled by name; what is new is watching one denominator manufacture three different physical laws.

Step 1. The driven electron. Displacement $x$ of a bound electron (charge $-e$ , resonance $\omega_0$ , damping $\gamma$ ) under $E_0e^{-i\omega t}$ :

m\ddot{x} + m\gamma\dot{x} + m\omega_0^2x = -eE_0e^{-i\omega t} \qquad\Longrightarrow\qquad x_0 = \frac{-eE_0/m}{\omega_0^2 - \omega^2 - i\gamma\omega}.

Why this oscillator is legitimate for a quantum atom: below and near a discrete line the atom's linear response is exactly of this form with an oscillator strength as prefactor — the same license Absorption and Dispersion invokes for $\varepsilon(\omega)$ ; we are re-spending it, not re-arguing it.

Step 2. The polarizability. With $p_0 = -ex_0 \equiv 4\pi\varepsilon_0\alpha(\omega)E_0$ :

\alpha(\omega) = \frac{e^2/4\pi\varepsilon_0 m}{\omega_0^2 - \omega^2 - i\gamma\omega} = \frac{r_ec^2}{\omega_0^2 - \omega^2 - i\gamma\omega},

where: $r_ec^2 = e^2/4\pi\varepsilon_0m$ makes the classical electron radius' appearance inevitable: it is the strength of one electron's charge-to-inertia response. Static limit: $\alpha(0) = r_ec^2/\omega_0^2$ — real, positive, frequency-blind: the stiff spring of the physical picture.

Step 3. Radiate and divide. The time-averaged power of the oscillating dipole is the imported meter, $\langle P\rangle = \omega^4|p_0|^2/12\pi\varepsilon_0c^3$ ; the incident flux $\langle S\rangle = \tfrac12\varepsilon_0cE_0^2$ . Their ratio (the $E_0^2$ cancels, as always):

\sigma(\omega) = \frac{8\pi}{3}\,\frac{\omega^4}{c^4}\,|\alpha(\omega)|^2 = \boxed{\;\sigma_T\,\frac{\omega^4}{(\omega_0^2 - \omega^2)^2 + \gamma^2\omega^2}\;} \qquad (1)

using $\sigma_T = \tfrac{8\pi}{3}r_e^2$ from Thomson Scattering. One formula; now read its floors.

Step 4. The three regimes. (i) $\omega \ll \omega_0$ : denominator $\to \omega_0^4$ , so $\sigma \approx \sigma_T(\omega/\omega_0)^4$ — Rayleigh's law, the $\omega^4$ passed through undamaged because $\alpha \to \alpha(0)$ . (ii) $\omega \gg \omega_0$ : denominator $\to \omega^4$ , so $\sigma \to \sigma_T$ — the spring forgotten, the free-electron plateau; this is why X-rays see all electrons as free. (iii) $\omega = \omega_0$ : $\sigma_{\max} = \sigma_T\,\omega_0^2/\gamma^2$ — the resonance tower.

Step 5. The resonance ceiling. How tall is the tower? For an isolated atom the damping is radiative: $\gamma_{\rm rad}(\omega_0) = \omega_0^2\tau$ with $\tau = 2r_e/3c = 6.27\times10^{-24}$ s (the radiation-reaction timescale — the same constant that runs Abraham-Lorentz Radiation Reaction Force). Substituting:

\sigma_{\max} = \frac{\sigma_T}{\omega_0^2\tau^2} = 6\pi\left(\frac{c}{\omega_0}\right)^2 = 6\pi\bar{\lambda}_0^2 = \frac{3\lambda_0^2}{2\pi},

— the electron's charge and mass have cancelled out: on resonance the cross section is set by the photon's wavelength alone. For the sodium D line (589 nm): $\sigma_{\max} = 1.66\times10^{-13}\ \mathrm{m}^2$ , fifteen orders of magnitude above $\sigma_T$ and nine above the atom's geometric size — an atom on resonance is a barn door the size of the light itself. (Real lines are Doppler- and collision-broadened, so laboratory peaks fall short of this unitary ceiling; the quantum edition $3\lambda_0^2/2\pi$ survives as the two-level limit, and the line shape is owned downstream by Radiative Frequency Shift and Line Width.) Verified: (1) reduces correctly in all three regimes (computed: $\sigma/\sigma_T = 1.0\times10^{-4}$ at $\omega = 0.1\omega_0$ ✓ fourth power; $\to 1$ at $\omega = 30\omega_0$ ✓); the $6\pi\bar{\lambda}_0^2$ ceiling computed for 589 nm: $6\pi(93.7\,\mathrm{nm})^2 = 1.656\times10^{-13}\ \mathrm{m}^2$ ✓, ratio to $\sigma_T = 2.5\times10^{15}$ ; matches Jackson §14.7/§17.8 conventions and Griffiths §11.1.4's limit structure.

The sky, quantitatively

Go deeper · Derivation B · 3 minFrom air's refractive index to the blue sky — cross section, optical depths, sunset

Derivation BFrom air's refractive index to the blue sky — cross section, optical depths, sunset

The goal is the sky with real numbers and no molecular theory: express the molecular cross section through the measured refractivity of air, then integrate through the atmosphere. This is Rayleigh's own 1899 maneuver, and its power is that $\omega_0$ , oscillator strengths, and quantum mechanics all hide inside one tabulated number, $n - 1$ . Inputs: $n - 1 = 2.79\times10^{-4}$ (dry air, optical), $N = 2.55\times10^{25}\ \mathrm{m^{-3}}$ (15°C, 1 atm), vertical column $N_{\rm col} = 2.15\times10^{29}\ \mathrm{m^{-2}}$ ( $= P/m_{\rm air}g$ , exact).

Step 1. Polarizability from refractivity. A dilute gas of dipoles has $n^2 = 1 + N\alpha_{\rm SI}/\varepsilon_0 = 1 + 4\pi N\alpha$ , so for $n - 1 \ll 1$ :

\alpha = \frac{n-1}{2\pi N} = 1.74\times10^{-30}\ \mathrm{m}^3.

Why the dilute (non-Clausius–Mossotti) form suffices: the local-field correction enters at order $N\alpha \sim 10^{-4}$ — the E3 Lorentz-cavity machinery is quantitatively idle in a gas, one of the few places it can be skipped with a clear conscience.

Step 2. Cross section from polarizability. Insert into Derivation A's $\sigma = \tfrac{8\pi}{3}k^4\alpha^2$ and eliminate $\alpha$ :

\boxed{\;\sigma(\lambda) = \frac{32\pi^3}{3}\,\frac{(n-1)^2}{N^2\lambda^4}\;} \qquad (2)

where: the density appears downstairs squared while the coherent refractivity appears upstairs squared — a fingerprint of Derivation C's fluctuation logic. Evaluating at 550 nm: $\sigma = 4.34\times10^{-31}\ \mathrm{m}^2$ ; multiplying by the King factor 1.049 (N₂/O₂ anisotropy depolarization — quoted, not derived) gives $4.55\times10^{-31}\ \mathrm{m}^2$ , within a few percent of the standard tabulated value.

Step 3. Optical depths. $\tau(\lambda) = N_{\rm col}\,\sigma(\lambda)$ :

$\lambda$	$\sigma$ (m²)	$\tau$ vertical	zenith transmission	38-airmass sunset
450 nm (blue)	$1.02\times10^{-30}$	0.218	0.80	$2.5\times10^{-4}$
550 nm (green)	$4.55\times10^{-31}$	0.098	0.91	$2.4\times10^{-2}$
650 nm (red)	$2.33\times10^{-31}$	0.050	0.95	0.15

Overhead, one photon in five of blue is scattered out of the direct beam and becomes sky; the measured $\tau(550) \approx 0.098$ for a clean dry atmosphere is reproduced to the percent. At sunset the horizon path is $\sim$ 38 airmasses (spherical-shell geometry caps the flat-Earth $\sec\theta$ divergence): the blue direct beam is extinguished six hundred times more thoroughly than the red — the transmitted disk reddens, and the scattered blue has lit everyone else's daytime sky on the way.

Step 4. The two honesty clauses. Why not violet? $\sigma$ keeps rising toward 400 nm, but the sky's color multiplies three spectra: solar output (already falling into the violet), Rayleigh's $\lambda^{-4}$ , and the eye's photopic response (nearly blind at 410 nm). The product peaks as saturated blue. Is the exponent exactly 4? No — $\alpha(\omega)$ itself rises toward the UV resonances (Step 2 of Derivation A), steepening the law; with air's effective single-oscillator $\lambda_0 \approx 156$ nm (computed from $\alpha(0)$ and $r_e c^2$ ), the local exponent at 550 nm is $4 + 4\omega^2/(\omega_0^2 - \omega^2) \approx 4.35$ in this one-line model, while the multi-line reality tabulated for air is $\approx 4.09$ . The deviation from 4 is dispersion talking — the same physics, audited.

Step 5. The polarization dividend. Each scattering is a dipole event, so the transfer function $\Pi(\theta) = (1-\cos^2\theta)/(1+\cos^2\theta)$ of Thomson Scattering applies verbatim: skylight at $90°$ from the Sun should be fully polarized. Molecular anisotropy (the King factor's other face) caps it near 94%, and multiple scattering plus ground albedo bring real zenith-sky maxima to 70–85% — still strong enough to navigate by, and the polarization pattern (a great circle of maxima $90°$ from the Sun, neutral points near it) is a daily whole-sky confirmation of the dipole donut. Verified: every number in the table computed for this edition from (2) with CODATA constants ( $N$ from $P/k_BT$ at 288 K; column from $P/m_{\rm air}g = 2.15\times10^{29}\ \mathrm{m^{-2}}$ ); $\tau(550) = 0.098$ vs the standard clean-atmosphere value 0.097–0.10 ✓; $(650/450)^4 = 4.35$ ✓; sunset factors are $e^{-38\tau}$ ; effective $\lambda_0$ : $\omega_0 = \sqrt{r_ec^2/\alpha(0)} = 1.21\times10^{16}$ rad/s → 156 nm ✓ consistent with air's far-UV absorption bands.

Figure

The sky computed from the optical-depth table — what the vertical atmosphere and the sunset path each do to blue, green and red, and where the scattered blue goes

The fine print that counts the molecules

Go deeper · Derivation C · 2 minCoherence, fluctuations, and why the sky is bright at all

Derivation CCoherence, fluctuations, and why the sky is bright at all

The goal is to justify — and then exploit — the step every naive treatment glosses: we computed one molecule's $\sigma$ and multiplied by $N$ . But radiation superposes as fields, and the coherence theorem of Larmor Formula from the Liénard–Wiechert Fields warns that $N$ in-phase scatterers give $N^2$ , while a perfectly ordered medium gives (sideways) zero. Air holds $\sim10^7$ molecules per cubic optical wavelength: which regime is the sky? The resolution is statistical, it is quantitative, and it once weighed the molecule.

Step 1. The dense-uniform limit is dark. Partition a strictly uniform medium into cells much smaller than $\lambda$ . For any off-forward direction, for every cell there exists a partner half a projected wavelength away whose re-radiation arrives in antiphase; a uniform medium pairs off perfectly and the side-scattered field cancels exactly (only the forward direction, where all phases agree regardless of position, survives — that coherent remainder is the refracted beam itself, the business of Optical Theorem). A continuum does not scatter; if air were a continuum the sky would be black at noon.

Step 2. Fluctuations break the pairing. Real cells hold $N_c \pm \Delta N_c$ molecules. Write each cell's amplitude as (mean + fluctuation): the means cancel by Step 1; the fluctuations are uncorrelated between cells in an ideal gas, so their fields add in quadrature — incoherently. The scattered intensity per cell is then $\langle\Delta N_c^2\rangle \times |{\rm one\ molecule}|^2$ , and summing cells:

I_{\rm scattered} \propto \sum_{\rm cells}\langle\Delta N_c^2\rangle = \langle\Delta N^2\rangle_{\rm total}.

Step 3. The ideal-gas miracle. Poisson statistics for independent molecules gives $\langle\Delta N_c^2\rangle = \bar{N}_c$ exactly, so the fluctuation sum equals the number of molecules:

I \propto \bar{N} \times |{\rm one\ molecule}|^2,

— precisely the independent-scatterer answer we used in Derivation B. The naive multiplication was right, but for a statistical reason: the sky's brightness is not "molecules each scattering" so much as thermal density fluctuations diffracting sunlight, which happen, for an ideal gas, to tally one unit per molecule. (Smoluchowski 1908 and Einstein 1910 built exactly this theory; the general result scales as $\langle\Delta N^2\rangle/\bar{N} = Nk_BT\kappa_T$ , the compressibility form.)

Step 4. The liquid audit. Water at 20°C: $N = 3.34\times10^{28}\ \mathrm{m^{-3}}$ , $\kappa_T = 4.5\times10^{-10}\ \mathrm{Pa^{-1}}$ , so $Nk_BT\kappa_T = 0.061$ — a liquid's molecules scatter 16× less per molecule than the same dipoles as dilute gas (computed): stiff media fluctuate weakly and cancellation nearly completes. Hence you can see across a swimming pool but not across 100 km of horizontal air. Push toward a critical point, where $\kappa_T \to \infty$ , and the same formula diverges: critical opalescence, the fluctuation theory's spectacular confirmation.

Step 5. Counting the molecules. Read Derivation B's equation (2) backwards: $\sigma \propto (n-1)^2/N^2\lambda^4$ , so the measured sky extinction plus the measured refractivity determine $N$ — the number density of air, hence Avogadro's number. Rayleigh (1899) ran exactly this inversion from visibility on clean days; refined by Kelvin and by Fowle's Mount Wilson transmission data (1914), it gave $N_A$ within a few percent of modern values. The blueness of the sky, quantified, was among the earliest weighings of the molecule — a fact Einstein and Smoluchowski advertised deliberately against the era's molecular skeptics. Verified: Step 3 reproduces the standard Einstein–Smoluchowski result (Landau–Lifshitz Electrodynamics of Continuous Media §120); water suppression factor computed: $3.34\times10^{28} \times 1.381\times10^{-23} \times 293 \times 4.5\times10^{-10} = 0.061$ → 16.4× ✓; the Poisson clause $\langle\Delta N^2\rangle = \bar{N}$ is exact for an ideal gas; the forward-coherent remainder is deferred by name to Optical Theorem (index building) — no double counting between the two ledgers.

Limiting cases

case	result	meaning
$\omega \ll \omega_0$	$\sigma = \sigma_T(\omega/\omega_0)^4$	Rayleigh's law — visible light on air, the sky
$\omega \gg \omega_0$	$\sigma \to \sigma_T$	Thomson Scattering recovered: X-rays see free electrons
$\omega = \omega_0$ , radiative damping	$\sigma_{\max} = 6\pi\bar{\lambda}_0^2$	resonance fluorescence: wavelength-sized, mass-blind
small dielectric sphere	$\alpha = a^3\dfrac{\varepsilon_r-1}{\varepsilon_r+2}$ , $\sigma \propto k^4a^6$	the particle edition — owned next door by Scattering by Conducting and Dielectric Spheres
$a \gtrsim \lambda$	$\lambda^{-4}$ fails; Mie regime	why clouds are white — deferred by name to the spheres article
uniform dense medium	side-scattering $\to 0$	continuum limit: only fluctuations scatter (Derivation C)

Common trapThe omega-fourth law has a postcode — outside the far-below-resonance, far-smaller-than-lambda, dilute-or-fluctuating regime, every one of its factors fails separately

Three separate abuses to refuse. Regime abuse: $\sigma \propto \omega^4$ holds only for $\omega \ll \omega_0$ ; near a line the same oscillator delivers $10^{15}\sigma_T$ (sodium-vapor "Rayleigh" estimates fail by that much), and above the resonances it flattens to Thomson — one formula, three floors, check which floor you're on. Size abuse: the point-dipole step needs $a \ll \lambda$ so the particle is driven in one phase; cloud droplets ( $a \sim 10\ \mu$ m) break it completely, higher multipoles enter, the color-blindness of Mie scattering results — white clouds are not "strong Rayleigh," they are its failure, priced in Scattering by Conducting and Dielectric Spheres. Statistics abuse: multiplying one-molecule $\sigma$ by $N$ presumes fluctuation statistics (Derivation C); in a liquid it overestimates by $\sim$ 16×, in a crystal by far more, and in any medium the forward direction is coherent and belongs to the refractive index, not to extinction-per-molecule. Also resist the folk version "blue is absorbed" — nothing is absorbed; the blue is redirected, which is why the lost light of one observer's sunset is another's blue noon.

Worked examples

1 · The sky, end to end (confidence builder)

Setup. Using only $n - 1 = 2.79\times10^{-4}$ , $N = 2.55\times10^{25}\ \mathrm{m^{-3}}$ , and the vertical column $2.15\times10^{29}\ \mathrm{m^{-2}}$ , predict: the blue/red scattering ratio, the fraction of overhead blue sunlight turned into skylight, and the sunset color budget.

Step 1. The ratio. From $\sigma \propto \lambda^{-4}$ : $(650/450)^4 = \mathbf{4.35}$ — each blue photon is 4.35× more likely to be redirected than a red one. (The strong-UG Electric Dipole Radiation quotes 9.4 for its 400-vs-700 endpoints — same law, wider lever.)

Step 2. Cross section and optical depth. Equation (2) at 450 nm with the King factor: $\sigma = 1.02\times10^{-30}\ \mathrm{m}^2$ ; $\tau_{450} = 2.15\times10^{29} \times 1.02\times10^{-30} = \mathbf{0.218}$ . So overhead, $1 - e^{-0.218} = \mathbf{20\%}$ of blue is scattered at least once — that fifth of the blue solar beam, redistributed over $4\pi$ with the $(1+\cos^2\theta)$ pattern, is the daytime sky. For red, only 5% — the sky is blue by budget, not by magic.

Step 3. Sunset. At 38 airmasses: blue survives $e^{-38\times0.218} = \mathbf{2.5\times10^{-4}}$ , red $e^{-38\times0.050} = \mathbf{0.15}$ — a 600:1 relative massacre of blue. Add aerosols (real sunsets are dustier than pure Rayleigh) and the transmitted disk runs orange-to-crimson. The same photons deleted from your sunset beam were somebody's zenith sky an hour earlier: global bookkeeping closes.

The lesson: two tabulated numbers about air — its refractivity and its pressure — predict the color of the planet's sky to the percent level, with no molecular physics anywhere in sight. That economy is Derivation B's whole point.

2 · The resonance tower — sodium vapor and the unitary ceiling (exam level)

Setup. Sodium's D₂ line sits at $\lambda_0 = 589$ nm with radiative lifetime 16.2 ns. Find the on-resonance cross section, compare with $\sigma_T$ and with the atom's geometric size, and check the damping bookkeeping.

Step 1. The ceiling. From Derivation A Step 5: $\sigma_{\max} = 3\lambda_0^2/2\pi = 6\pi\bar{\lambda}_0^2$ with $\bar{\lambda}_0 = 93.7$ nm: $\sigma_{\max} = \mathbf{1.66\times10^{-13}\ m^2}$ .

Step 2. The comparisons. Against Thomson: $\sigma_{\max}/\sigma_T = \mathbf{2.5\times10^{15}}$ . Against geometry: a sodium atom's radius is $\sim$ 0.2 nm, area $\sim10^{-19}\ \mathrm{m}^2$ — the resonant atom scatters as if it were a million times its own area. Nothing paradoxical: a cross section measures response, not silhouette, and on resonance the response is bounded only by re-radiation itself.

Step 3. The damping audit. Classical radiative width: $\gamma_{\rm rad} = \omega_0^2\tau$ with $\omega_0 = 3.20\times10^{15}$ rad/s: $\gamma_{\rm rad} = (3.20\times10^{15})^2 \times 6.27\times10^{-24} = 6.4\times10^{7}\ \mathrm{s^{-1}}$ → lifetime $1/\gamma = \mathbf{16\ ns}$ — the classical oscillator predicts sodium's measured 16.2 ns to a few percent (the D-line's oscillator strength is near 1, the classical electron's exactly 1). This spooky success of $\tau = 2r_e/3c$ is the same constant that powers the radiation-reaction articles, cashed as spectroscopy.

Step 4. The honesty. Room-temperature vapor is Doppler-broadened to $\sim$ 50× the natural width, diluting the peak accordingly; the full ceiling is recovered only for cold atoms — exactly how laser-cooling experiments make single trapped atoms camera-visible by resonance fluorescence.

The lesson: the ladder formula is not three separate facts but one; walking it from the sky's $10^{-31}\ \mathrm{m}^2$ to a cold atom's $10^{-13}\ \mathrm{m}^2$ spans eighteen orders of magnitude with a single denominator.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §11.1.4 · Jackson, Classical Electrodynamics, 3rd ed., §10.1–10.2 and §14.7 · Zangwill, Modern Electrodynamics, §21.5–21.6 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 9 · Landau & Lifshitz, Electrodynamics of Continuous Media, §120 (fluctuation scattering) · Bohren & Huffman, Absorption and Scattering of Light by Small Particles, Ch. 5. The full optics story — radiative transfer, whole-sky polarization maps, twilight and atmospheric optics — lives in the Optics track's Rayleigh Scattering article; this edition is the electrodynamics engine beneath it.

Provenance. Equations are standard results from Griffiths §11.1.4, Jackson §10.1–10.2 and §14.7, and Zangwill §21.5–21.6. A verifier should: (1) recompute the total cross section $\sigma=(8\pi/3)(\omega^4/c^4)|\alpha|^2$ from dipole Larmor power over incident flux; (2) check the $1+\cos^2\theta$ angular average over polarizations; (3) verify the Lorentz-oscillator $\sigma$ reduces to $\sigma_T=(8\pi/3)r_e^2$ for $\omega\gg\omega_0$ and to $\sigma_T(\omega/\omega_0)^4$ for $\omega\ll\omega_0$ ; (4) confirm the wavelength-ratio numbers ( $2^4=16$ , $(700/450)^4\approx5.7$ ). Gaussian units used throughout for the radiation formulae. Addendum for this edition: converted throughout to SI with the volume-polarizability convention $\mathbf{p} = 4\pi\varepsilon_0\alpha\mathbf{E}$ (dictionary to E3's $\alpha_{\rm SI} = 4\pi\varepsilon_0\alpha$ stated in-text); the base's $(700/450)^4 \approx 5.7$ evaluates to 5.86 and is superseded here by the 450/650 pair (4.35) with the E9 article's 400/700 (9.4) cross-referenced. Division of labor: the $(1+\cos^2\theta)/2$ average and $\Pi(\theta)$ are recalled by name from Thomson Scattering (checked against item (2), not re-derived); the oscillator's $\varepsilon(\omega)$ /KK side belongs to Absorption and Dispersion; the dielectric-sphere $\alpha = a^3(\varepsilon_r-1)/(\varepsilon_r+2)$ and all $a \sim \lambda$ physics to Scattering by Conducting and Dielectric Spheres; line shape to Radiative Frequency Shift and Line Width; the forward-coherent index bookkeeping to Optical Theorem. This edition owns: the three-regime ladder with the $6\pi\bar{\lambda}_0^2$ ceiling (sodium: $1.66\times10^{-13}\ \mathrm{m}^2$ , lifetime audit 16 ns vs measured 16.2 ns), the refractivity route $\sigma = 32\pi^3(n-1)^2/3N^2\lambda^4$ with King factor 1.049 and the computed sky table ( $\tau$ = 0.218/0.098/0.050 at 450/550/650 nm; zenith and 38-airmass transmissions; effective $\lambda_0 = 156$ nm and local exponent 4.35 vs tabulated 4.09 honesty), and the fluctuation ledger (Poisson $\langle\Delta N^2\rangle = \bar N$ ; water suppression $Nk_BT\kappa_T = 0.061 \to 16\times$ , computed; Rayleigh–Fowle molecule count) — all numbers recomputed for this edition; figure curves and bars traced from computed data.

Builds on: Electric Dipole Radiation · Larmor Formula · Multipole Expansion of Radiation Fields Leads to: Magnetic Dipole and Electric Quadrupole Radiation · Angular Distribution of Radiation from Accelerated Charges · Mie scattering

If a cubic wavelength of air contains ten million molecules, why don't they scatter coherently with an N-squared boost — or cancel entirely?

Both happen, in different directions, and the sky is the leftover. Forward, all molecules radiate in phase regardless of position, so the forward amplitude is coherent and $N$ -fold — but that wave doesn't remove light from the beam; it is the beam, phase-shifted: the refractive index (the bookkeeping Optical Theorem formalizes). Sideways, a uniform medium cancels pairwise to exact zero; what survives is the fluctuation part, and Poisson statistics ( $\langle\Delta N^2\rangle = \bar N$ ) makes the incoherent one-per-molecule sum exact for an ideal gas — Derivation C. The correct slogan is not "each molecule scatters" but "the gas's thermal roughness scatters, at exactly one molecule's worth per molecule." In liquids the roughness is 16× smaller; in a perfect zero-temperature crystal only defects and phonons would scatter.

Why is the sky blue rather than violet, since violet has an even shorter wavelength?

Because your eye and the Sun both vote alongside the $\lambda^{-4}$ law. The cross section does keep rising into the violet (~40% higher at 410 nm than at 450 nm), but the solar spectrum is already falling there, ozone trims the near-UV, and the eye's photopic sensitivity collapses below ~430 nm. Multiply solar output × $\lambda^{-4}$ × visual response and the hue lands on saturated blue. A flat-response detector does report a violet-peaked sky; the blueness is partly air, partly retina.

Is the blue of the sky the same physics as the blue of Rayleigh scattering in optical fibers?

Yes — with the temperature frozen in. Air's scattering roughness is live thermal fluctuation; silica's is the density-fluctuation pattern present at the glass transition (~1500 K), frozen in as the melt fell out of equilibrium. A fiber scatters like a snapshot of very hot gas statistics — which is why lowering the fictive temperature by slow annealing is the current route to record-low-loss submarine fiber (~0.14 dB/km). The $\lambda^{-4}$ signature survives intact: loss falls as $\lambda^{-4}$ until the infrared multiphonon edge turns it back up, and the 1.55 μm telecom window is the valley between — the sky's exponent deciding where the internet lives.

How does one formula give both the faint sky-blue scattering and the enormous resonance fluorescence of a sodium lamp?

By the denominator's mood. Far below resonance the denominator is the huge constant $\omega_0^4$ : air at 550 nm sits a factor $(156/550)^4 \approx 6\times10^{-3}$ further suppressed below the already-tiny $\sigma_T$ . On resonance the real part vanishes and only $\gamma^2\omega_0^2$ survives downstairs: the cross section rockets to $\sigma_T(\omega_0/\gamma)^2$ — with radiative damping, $6\pi\bar\lambda_0^2 \sim 10^{15}\sigma_T$ , the unitary ceiling. Eighteen orders of magnitude, one formula: just how close the drive sits to the spring's note. The quantum footnotes — oscillator strengths, saturation, the line shape — belong to Radiative Frequency Shift and Line Width and the quantum track.

Does multiple scattering matter for the sky, or is the single-scattering picture complete?

Single scattering carries the structure; multiple scattering softens it. With zenith $\tau \sim 0.1$ –0.2 most detected skylight has scattered once, so the single-scattering predictions — the $90°$ polarization maximum, the $(1+\cos^2\theta)$ pattern, the color — are recognizably correct. The twice-scattered tail matters at the edges: it dilutes the polarization maximum to the observed 70–85%, creates the neutral points near Sun and anti-Sun, whitens the horizon, and keeps twilight luminous. Iterating to convergence is the Optics track's radiative-transfer territory; the input — cross section, pattern, polarization transfer — is what this article computed.

Rayleigh versus Raman scattering — what is the difference?

Rayleigh is elastic: the scattered photon keeps the incident frequency because the molecule's linear response returns it to its initial state. Raman is inelastic: the molecule exits in a different vibrational or rotational state and the photon shifts down (Stokes) or up (anti-Stokes) by the transition frequency — classically, the polarizability modulated by the vibration, producing sidebands; quantum-mechanically, an exchange with internal structure. Raman lines run $\sim10^{-3}$ of the already-faint Rayleigh line. The division of labor: Rayleigh light measures how much (density — hence Rayleigh lidar); Raman light measures what (a vibrational fingerprint — hence Raman spectroscopy). Every Raman spectrometer's first job is filtering out the towering elastic peak this article computes.

This article and the Optics track both have a Rayleigh Scattering page — which does what?

Deliberate division of labor. The Optics track owns the phenomenology and transport: whole-sky polarization maps, twilight, visibility and haze, radiative transfer. This electrodynamics edition owns the engine: where the $\omega^4$ comes from (Derivation A), how air's cross section follows from its refractivity with zero atomic modeling (Derivation B), and why the independent-scatterer sum is statistically legal at all (Derivation C). To predict an actual sky pixel, go there; to know why the exponent is 4, why the sky isn't black, or why one electron gives $10^{-31}\ \mathrm{m}^2$ in air and $10^{-13}\ \mathrm{m}^2$ on resonance, stay here.