Optical Theorem

An accounting miracle: to know the total power a scatterer removes from a beam — summed over every scattering angle, plus everything absorbed as heat — you do not need the field everywhere. You need one complex number: the amplitude scattered into the exact forward direction. The optical theorem, $\sigma_{\rm tot} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0)$ , is energy conservation compressed to a single line, and it holds for light, sound, and quantum particles alike, exactly and non-perturbatively. The physics is the shadow. Behind an obstacle the beam is dimmer — but nothing "blocks" a wave; the only mechanism a linear theory owns is interference. The scattered wave, in the forward direction alone, stays phase-coherent with the incident beam over arbitrarily long distances, and their cross term destructively drains exactly the removed power. Every shadow you have ever seen is forward scattering doing arithmetic. This article derives the theorem from Poynting bookkeeping on a giant sphere, with the stationary-phase step executed honestly. Then it runs the audit that makes this the electrodynamics telling of the story: a lossless driven oscillator naively has a real polarizability — the theorem then predicts zero extinction while the dipole plainly radiates. The books balance only because of the radiation-reaction force of the E9 self-force articles: the tiny $\omega^3$ damping term hands $\alpha$ exactly the imaginary part the theorem demands, verified here symbolically and to twelve digits numerically. Unitarity forces the self-force. The bare slug for this topic belongs to the Optics track; what lives here is the electromagnetic engine underneath it.

Key takeaways

The optical theorem: $\sigma_{\rm tot} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0)$ — the total cross section (scattering into all angles plus absorption) is fixed by the imaginary part of the forward scattering amplitude alone.
It is pure flux conservation: on a far sphere, the incident–scattered interference term survives only near $\theta = 0$ (stationary phase), and it drains from the forward beam exactly what scattering and absorption remove.
A real $f(0)$ means a phase-shifted but unattenuated beam — no extinction. Only $\mathrm{Im}\,f(0)$ dims light; that is why absorption and scattering both feed it.
The radiation-reaction audit: a lossless dipole scatterer with real polarizability contradicts the theorem. The self-force of Radiation Reaction Force supplies $\mathrm{Im}\,\alpha = \frac{k^3}{6\pi}|\alpha|^2$ , and then $\frac{4\pi}{k}\mathrm{Im}\,f(0) = \sigma_{\rm scat}$ holds exactly — verified symbolically and to $10^{-12}$ numerically in this article.
Forward scattering also builds the refractive index: $n = 1 + \frac{2\pi N}{k^2}f(0)$ — extinction and dispersion are the per-scatterer and per-volume readings of the same amplitude (the bulk story is Absorption and Dispersion).
The extinction paradox delivered: for a large obstacle the forward diffraction lobe carries $f(0) = ika^2/2$ , so $\sigma_{\rm ext} = 2\pi a^2$ — twice the geometric area, the factor the sphere article deferred here.
In the Born approximation $f^{(1)}(0)$ is real: the theorem is satisfied only at second order — a built-in unitarity test for any approximation scheme.
Key idea: whatever a target removes from a beam must reappear as destructive interference between the incident wave and the forward-scattered wave — so one number, $\mathrm{Im}\,f(0)$ , carries the whole energy budget.

The core insight

Conservation of flux forces the depletion of the forward beam to equal the interference between the incident wave and the forward-scattered wave; hence the entire total cross section is fixed by $\mathrm{Im}\,f(0)$ alone.

Where this lives today

Particle physics runs the theorem as a measurement tool: at the LHC, the TOTEM and ATLAS-ALFA experiments extract the total proton–proton cross section from the elastic forward slope — Roman-pot detectors creep to within millimeters of the beam to measure scattering at microradian angles, extrapolate to $t = 0$ , and read $\sigma_{\rm tot}$ (about 111 mb at 13 TeV) from the S-matrix edition of this article's formula. No detector counts "all interactions"; the forward amplitude counts them by conservation. In light scattering the theorem is baked into the daily numerics. The discrete-dipole approximation — the workhorse for interstellar dust, aerosols, and nanoparticle optics — assigns each dipole the radiatively corrected polarizability of Derivation B ( $\mathrm{Im}\,\alpha = \frac{k^3}{6\pi}|\alpha|^2$ , the Draine correction), precisely so that computed extinction respects unitarity; omit it and absorption cross sections of lossless particles come out negative. Atmospheric lidar, satellite aerosol retrievals, and climate radiative-transfer codes all use extinction efficiencies computed through the forward-amplitude sum of Mie theory (the $Q_{\rm ext}$ ledger of Scattering by Conducting and Dielectric Spheres), and the extinction paradox's factor two — with its acceptance-angle fine print — is a standing calibration item for every transmissometer. When the same identity appears in quantum mechanics as the unitarity of the S-matrix, nothing new happens: it is this Poynting audit wearing probability current.

Origins · 1 minHendrik van de Hulst and how this was discovered

Hendrik van de Hulst

1918–2000 · Leiden — who made extinction a precision science, and predicted the 21-cm line on the side

Van de Hulst wrote the book — literally: Light Scattering by Small Particles (1957) remains the field's reference, and its treatment of extinction, the optical theorem, and the paradox's acceptance-angle fine print turned scattered folklore into an engineering discipline. His anomalous-diffraction approximation is still the fastest honest estimate of $Q_{\rm ext}$ for soft particles, and his accounting of what a real detector actually measures (the forward lobe rides into every finite aperture) resolved decades of factor-of-two confusion in transmission experiments. The same man, as a 26-year-old student in occupied Leiden in 1944, answered Oort's challenge to find a radio line from interstellar gas by predicting the 21-cm hyperfine transition of hydrogen — detected in 1951, it became radio astronomy's master tool and mapped the Galaxy's spiral arms through the dust his scattering theory described. Extinction by day, the transparent universe by night: few careers have owned both sides of the light-and-obstacle ledger so completely.

History

1871

— Rayleigh's scattering papers connect molecular scattering to the sky's blue and to attenuation

1908

— Mie's sphere solution carries the extinction sum — the forward-amplitude ledger in partial waves

1915

— Ewald and Oseen's extinction theorem: the medium's molecules rebuild the wave, index from forward scattering

1939

— Bohr, Peierls and Placzek state the general optical theorem for quantum scattering; the name sticks

1946

— van de Hulst's anomalous-diffraction work systematizes extinction between the regimes

1957

— Light Scattering by Small Particles codifies the theorem, the paradox, and the detector fine print

1988

— Draine's radiative-reaction polarizability makes discrete-dipole codes unitarity-safe

2011

— TOTEM at the LHC measures the pp total cross section through the forward elastic amplitude

The physical picture

Put a detector far behind a scatterer, on axis. The field there is a superposition: the incident plane wave, plus the scattered spherical wave $f(\theta)\,e^{ikr}/r$ . At any angle off the forward direction, the two accumulate relative phase $kr(1 - \cos\theta)$ — enormous and wildly oscillating at large $r$ , so their interference averages away and only intensities add. But exactly forward, the spherical wave rides along with the beam, phase-locked forever. There, and only there, amplitudes add before squaring. That surviving cross term is the shadow's entire mechanism. If $f(0)$ has an imaginary part with the right sign, the interference is destructive: the forward beam emerges dimmer, and the missing power is precisely what went sideways ( $\sigma_{\rm scat}$ ) or into heat ( $\sigma_{\rm abs}$ ). A purely real $f(0)$ only shifts the phase of the forward beam — a lens does that — without dimming it at all. So attenuation lives entirely in $\mathrm{Im}\,f(0)$ : the same division of labor as a refractive index, whose real part bends and whose imaginary part absorbs. Indeed the two statements are one — a dilute slab of $N$ scatterers per volume has $n - 1 = \frac{2\pi N}{k^2}f(0)$ , and taking its imaginary part converts the optical theorem into Beer's attenuation law. The theorem's generality is its glory and its warning. It needs only a conserved flux and the asymptotic form "plane wave + outgoing spherical wave" — nothing about the target. But it pays out only the total: $f(0)$ does not determine where the scattered power goes, and a first-order (Born) amplitude, being real, satisfies the theorem trivially at first order — the cross section it implies appears one order later. Unitarity is a constraint between orders, which is exactly how this article's centerpiece audit uses it: demanding the books balance for a simple dipole scatterer derives the radiation-reaction correction to its polarizability.

Figure

The flux ledger on a giant sphere — scattered power flows out everywhere, but the incident-scattered interference survives only in the forward stripe, where it drains the beam by exactly the total cross section

The objects on the table

Far from any finite scatterer, a monochromatic field settles into the universal asymptotic form

\psi(\mathbf{r}) \xrightarrow[r\to\infty]{} e^{ikz} + f(\theta,\phi)\,\frac{e^{ikr}}{r}, \qquad (1)

where: $\psi$ stands for the relevant amplitude — one polarization component of $\mathbf{E}$ here, a quantum wavefunction in the QM edition — and $f(\theta,\phi)$ , with dimensions of length, is the scattering amplitude: the target's complete far-field business card. Its square is the differential cross section, $d\sigma/d\Omega = |f|^2$ , and its angular integral is $\sigma_{\rm scat}$ . For electromagnetic scattering the honest vector statement replaces $f(0)$ by the polarization-projected forward amplitude, $\sigma_{\rm ext} = \frac{4\pi}{k}\mathrm{Im}\big[\hat{\boldsymbol{\epsilon}}_0^*\cdot\mathbf{f}(0)\big]$ — same theorem, with bookkeeping for the two transverse channels. The total cross section splits by fate:

\sigma_{\rm tot} = \sigma_{\rm scat} + \sigma_{\rm abs}, \qquad (2)

where: $\sigma_{\rm abs}$ counts flux that enters the target and never re-emerges (Ohmic heat, excitation). The theorem's astonishing claim is that the sum (2) — one term needing an all-angle integral, the other a look inside the target — is read off one number at one angle.

Derivation AThe theorem from Poynting bookkeeping on a far sphere

The goal is $\sigma_{\rm tot} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0)$ , by auditing energy flux through a sphere of radius $r \to \infty$ around the scatterer. The route: insert the asymptotic field (1) into the flux, split into incident, scattered, and interference pieces, and execute the stationary-phase evaluation of the interference — the step where "only forward survives" stops being a slogan and becomes a Gaussian integral. We work with the scalar amplitude (one fixed polarization; the flux normalization divides out), writing the cycle-averaged radial flux as $j_r \propto \mathrm{Im}(\psi^*\partial_r\psi)$ — the Poynting bookkeeping of the conservation chapter in amplitude form.

Step 1. The conservation statement. In steady state, the net outward flux through the far sphere equals minus the power absorbed inside:

\oint_{S_R} \mathbf{j}\cdot d\mathbf{A} = -\,v\,\sigma_{\rm abs}\,|A|^2,

normalizing the incident beam to unit amplitude ( $|A|^2 = 1$ ) and flux velocity $v$ ( $= c$ for light). With no absorber the integral vanishes; with absorption it is negative by exactly the absorbed power.

Step 2. Split the flux. Inserting $\psi = e^{ikz} + f e^{ikr}/r$ gives three pieces: $\mathbf{j}_{\rm inc}$ , $\mathbf{j}_{\rm scat}$ , and the cross term $\mathbf{j}_{\rm int}$ . The incident piece is a uniform beam: its net flux through any closed surface is zero. The scattered piece is radial with magnitude $v|f|^2/r^2$ , integrating to $v\,\sigma_{\rm scat}$ . Everything interesting hides in the cross term.

Step 3. The interference term, honestly. The cross term carries the relative phase $e^{ik(r - z)} = e^{ikr(1 - \cos\theta)}$ . Integrate over the sphere with $u = \cos\theta$ :

\oint \mathbf{j}_{\rm int}\cdot d\mathbf{A} \;\propto\; \mathrm{Re}\left[\,2\pi r\int_{-1}^{1} \frac{1+u}{2}\,f(\theta)\,e^{ikr(1-u)}\,du\right]\!,

with $\frac{1+u}{2}$ the slowly varying obliquity of the cross term's radial projection; the integrand oscillates in $u$ with enormous frequency $kr$ . Integrating by parts (or Gaussian stationary phase at the endpoint $u = 1$ ), the integral is dominated by the forward endpoint, contributing $\frac{i}{kr}\,f(0)\,\times$ (slowly varying factors), while the backward endpoint's contribution oscillates as $e^{2ikr}$ and vanishes upon the infinitesimal frequency-averaging any physical beam performs. Assembling the constants — the $2\pi$ from azimuth, the $1/kr$ from the endpoint, the $r$ from the area element against $1/r$ of the scattered wave — the interference flux converges to a finite, $r$ -independent deficit:

\oint \mathbf{j}_{\rm int}\cdot d\mathbf{A} = -\,v\,\frac{4\pi}{k}\,\mathrm{Im}\,f(0).

Why the imaginary part appears: the endpoint evaluation multiplies $f(0)$ by $i$ ; taking the real part of the flux then selects $\mathrm{Im}\,f(0)$ . A real amplitude shifts the forward phase — the interference term then merely displaces the beam's wavefronts (that is refraction, Derivation C) — while only the imaginary part produces a net energy deficit.

Step 4. Assemble. Conservation (Step 1) with the three pieces:

0 + v\,\sigma_{\rm scat} - v\,\frac{4\pi}{k}\,\mathrm{Im}\,f(0) = -\,v\,\sigma_{\rm abs} \qquad\Longrightarrow\qquad \boxed{\;\sigma_{\rm tot} = \sigma_{\rm scat} + \sigma_{\rm abs} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0).\;}

Nothing about the target entered: any scatterer, any strength, any internal physics — if flux is conserved and the far field looks like (1), the theorem holds exactly. Verified: against Jackson §10.11 and Newton's Scattering Theory of Waves and Particles Ch. 1; the endpoint/stationary-phase constant $4\pi/k$ rechecked by the disk audit in Derivation C, where $f(0) = ika^2/2$ reproduces $\sigma_{\rm ext} = 2\pi a^2$ exactly; dimensions: $[f] =$ length, so $[\mathrm{Im} f/k] =$ area.

Go deeper · Derivation B · 2 minThe radiation-reaction audit — unitarity forces the self-force

Derivation BThe radiation-reaction audit — unitarity forces the self-force

The goal is the article's centerpiece: show that for the simplest electromagnetic scatterer — an induced point dipole $\mathbf{p} = \varepsilon_0\alpha\,\mathbf{E}_0$ — the optical theorem fails unless the polarizability carries the imaginary part supplied by radiation reaction, and that with it the theorem holds exactly, at every frequency. The self-force itself is owned by the E9 pair Radiation Reaction Force and Abraham–Lorentz Equation and Its Pathologies — recalled by name, never re-derived; what this derivation owns is the consistency theorem that makes their $\omega^3$ term non-optional.

Step 1. The forward amplitude of a dipole. An induced dipole radiates the field of Electric Dipole Radiation: Exact Fields and Zone Structure; reading off its far-zone amplitude along the incident polarization in the forward direction gives

f(0) = \frac{k^2\alpha}{4\pi},

with $\alpha$ the (volume-dimensioned) polarizability. The theorem then demands

\sigma_{\rm tot} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0) = k\,\mathrm{Im}\,\alpha.

Step 2. The contradiction. Take a lossless bound electron — the driven oscillator of Absorption and Dispersion, no collisions, no resistance: naively $\alpha(\omega) = \frac{e^2/m\varepsilon_0}{\omega_0^2 - \omega^2}$ , purely real. The theorem then says $\sigma_{\rm tot} = 0$ . But the dipole demonstrably radiates: the Rayleigh cross section $\sigma_{\rm scat} = \frac{k^4|\alpha|^2}{6\pi}$ is finite. Energy books refuse to balance — something in the lossless oscillator is missing.

Step 3. The rescue, by name. What is missing is the force the E9 chapter derived from energy bookkeeping itself: radiation reaction. Including the self-force term (the $\ddot{a}$ force of the Abraham–Lorentz equation, acting on the driven oscillator as a damping $\gamma_{\rm rad}(\omega) = \omega^2\tau$ with $\tau = \frac{e^2}{6\pi\varepsilon_0 m c^3}$ — the characteristic time of the E9 articles) gives

\alpha(\omega) = \frac{e^2/m\varepsilon_0}{\omega_0^2 - \omega^2 - i\,\omega^3\tau}.

Step 4. The identity, exactly. With that single change, compute both sides. The claim: $k\,\mathrm{Im}\,\alpha = \frac{k^4}{6\pi}|\alpha|^2$ identically in $\omega$ — equivalently $\mathrm{Im}\,\alpha = \frac{k^3}{6\pi}|\alpha|^2$ . Proof in one line: for $\alpha = A/D$ with $D = \omega_0^2 - \omega^2 - i\omega^3\tau$ , $\mathrm{Im}\,\alpha = A\,\omega^3\tau/|D|^2$ , while $\frac{k^3}{6\pi}|\alpha|^2 = \frac{\omega^3}{6\pi c^3}A^2/|D|^2$ — equal iff $\tau = \frac{A}{6\pi c^3} = \frac{e^2}{6\pi\varepsilon_0 mc^3}$ , which is precisely the Larmor damping time. The optical theorem does not merely tolerate the radiation-reaction coefficient — it pins it: any other value of $\tau$ breaks energy conservation. Verified symbolically (computer algebra: the difference simplifies to zero) and numerically at $\lambda = 550$ , $200$ , and $91$ nm for a bound electron with $\omega_0$ in the far-UV: $\sigma_{\rm tot}/\sigma_{\rm scat} = 1.000000000000$ each time, while the damping-free $\alpha$ gives $\mathrm{Im}\,f(0) = 0$ against a finite scattered power — the contradiction, exhibited.

Step 5. The Thomson audit. Set $\omega_0 = 0$ (free electron): $\alpha = -\frac{e^2/m\varepsilon_0}{\omega^2(1 + i\omega\tau)}$ , and $k\,\mathrm{Im}\,\alpha$ evaluates to $\frac{(e^2/m\varepsilon_0 c^2)^2}{6\pi}(1 + \omega^2\tau^2)^{-1} \to \frac{8\pi}{3}r_e^2 = \sigma_T$ — the Thomson cross section of Thomson Scattering, recovered from the forward amplitude alone to eight digits numerically. The theorem, the self-force, and the free-electron cross section form one closed, consistent triangle. Verified: the identity $\mathrm{Im}\,\alpha = \frac{k^3}{6\pi}|\alpha|^2$ against the standard radiative-correction result (Draine 1988; Bohren & Huffman); $\tau = 6.2664\times10^{-24}$ s $= 2r_e/3c$ exactly; all numerical checks recomputed for this edition. The pathologies of the underlying self-force (runaways, preacceleration) stay with the E9 owners and the E11 self-force set — this audit needs only the damping term's value.

Figure

The audit computed — total cross section from the forward amplitude versus integrated scattered power for a radiatively damped oscillator across the resonance; the two curves coincide to machine precision, while dropping the self-force sends the theorem's prediction to zero

Go deeper · Derivation C · 2 minWhat forward scattering builds — the refractive index, and the extinction paradox delivered

Derivation CWhat forward scattering builds — the refractive index, and the extinction paradox delivered

The goal is the theorem's two payoffs in bulk matter and in the large-obstacle limit: first, $n = 1 + \frac{2\pi N}{k^2}f(0)$ — the refractive index is coherent forward scattering — which converts the optical theorem into the attenuation coefficient of a medium; second, the extinction paradox $\sigma_{\rm ext} \to 2\pi a^2$ that Scattering by Conducting and Dielectric Spheres stated and deferred here. Both are exercises in reading $f(0)$ .

Step 1. The thin-slab phase audit. Send the unit-amplitude beam through a slab of thickness $dz$ containing $N$ scatterers per volume. Downstream on axis, each scatterer contributes its forward-scattered wavelet; summing them across the slab (the same transverse Fresnel-zone integral as the theorem's stationary phase, executed with the flat-slab geometry) gives the on-axis field

\psi(z + dz) = \psi(z)\left[1 + \frac{2\pi i N}{k^2}\,f(0)\,k\,dz\right].

Exponentiating: $\psi \propto e^{iknz}$ with

n = 1 + \frac{2\pi N}{k^2}\,f(0).

Why this is profound rather than cute: the refractive index of glass, water, air — the whole of linear optics — is the accumulated forward-scattering phase of its molecules. For the dipole scatterer of Derivation B, $f(0) = k^2\alpha/4\pi$ gives $n = 1 + N\alpha/2$ : exactly the dilute limit of the Clausius–Mossotti chain, tying this formula to the E3 machinery and to the oscillator dispersion of Absorption and Dispersion, by name.

Step 2. Attenuation = the theorem, per volume. Take the imaginary part: $\mathrm{Im}\,n = \frac{2\pi N}{k^2}\mathrm{Im}\,f(0) = \frac{N\sigma_{\rm tot}}{2k}$ . The intensity $\propto e^{-2k\,\mathrm{Im}(n)\,z} = e^{-N\sigma_{\rm tot} z}$ — Beer's law with exactly the theorem's cross section. One amplitude, three readings: its real part is dispersion, its imaginary part per scatterer is extinction, and the Kramers–Kronig relations linking them (owned by Absorption and Dispersion) are the frequency-domain shadow of this same causal structure.

Step 3. The extinction paradox. A large opaque disk of radius $a$ ( $ka \gg 1$ ) blocks its geometric area — and also diffracts. Babinet's reading of the forward field (the aperture machinery of Scalar and Vector Diffraction Theory): removing a disk-shaped piece of wavefront produces a forward diffraction amplitude equal to minus the Fraunhofer amplitude of the missing disk, whose on-axis value is

f(0) = \frac{ik}{2\pi}\int_{\rm disk} da' = \frac{ika^2}{2}.

Purely imaginary — pure extinction, no phase shift — and the theorem pays out

\sigma_{\rm ext} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0) = \frac{4\pi}{k}\cdot\frac{ka^2}{2} = 2\pi a^2.

Twice the geometric area, exactly (numerical audit for a 1 μm droplet at 550 nm: both routes give $6.2832\times10^{-12}$ m²). The accounting: $\pi a^2$ of flux is geometrically intercepted (absorbed or reflected — the "blocking"), and another $\pi a^2$ is scattered into the narrow forward diffraction lobe of angular width $\sim 1/ka$ . A detector with any finite acceptance angle swallows that lobe along with the beam — which is why the paradox hid for so long: measuring the true $2\pi a^2$ requires angular resolution finer than $1/ka$ , a real constraint on real instruments (rain on a windshield "disappears" for exactly this reason). Verified: the slab formula against the standard multiple-scattering result (Jackson §10.11, van de Hulst §4.3); $n = 1 + N\alpha/2$ against the dilute Clausius–Mossotti limit; the disk amplitude against the Fraunhofer transform at zero angle; air-molecule audit in worked example 1.

Limiting cases

case	result	meaning
lossless dipole, no self-force	$\mathrm{Im}\,f(0) = 0$ but $\sigma_{\rm scat} > 0$	contradiction — unitarity demands radiation reaction
dipole + radiation damping	$\mathrm{Im}\,\alpha = \frac{k^3}{6\pi}\|\alpha\|^2$ , theorem exact at all $\omega$	Derivation B's audit
free electron ( $\omega_0 \to 0$ )	$\frac{4\pi}{k}\mathrm{Im}\,f(0) \to \sigma_T = 0.665$ b	Thomson recovered from the forward amplitude
first Born approximation	$f^{(1)}(0)$ real $\Rightarrow$ theorem trivial at first order; $\mathrm{Im}\,f^{(2)}(0)$ carries $\|f^{(1)}\|^2$	unitarity as an order-by-order consistency test
large opaque obstacle	$f(0) = ika^2/2 \Rightarrow \sigma_{\rm ext} = 2\pi a^2$	extinction paradox: geometry + forward diffraction lobe
dilute medium	$n = 1 + \frac{2\pi N}{k^2}f(0)$ ; $I \propto e^{-N\sigma_{\rm tot}z}$	dispersion and Beer's law from one amplitude

Common trapA real forward amplitude does not mean weak scattering — and the theorem never tells you where the power goes

Trap 1 — the Born zero. In first-order perturbation theory $f(0)$ comes out real, and a hasty reading of the theorem concludes $\sigma_{\rm tot} = 0$ . Wrong order: the theorem is exact, and it relates $\mathrm{Im}\,f$ at order $2n$ to $|f|^2$ at order $n$ — the first-order cross section lives in the second-order imaginary part. Never use a lowest-order amplitude inside the theorem and expect the full cross section back; instead, use the mismatch as a diagnostic (any simulation or approximation violating $\frac{4\pi}{k}\mathrm{Im} f(0) = \sigma_{\rm scat} + \sigma_{\rm abs}$ is leaking energy — Derivation B shows even the humble lossless dipole needs its self-force term to pass). Trap 2 — total means total only. $\mathrm{Im}\,f(0)$ fixes the energy budget, not its angular distribution: two targets with identical $f(0)$ can throw their power into completely different patterns. Detector design, radiative-transfer phase functions, backscatter radar budgets — all need $|f(\theta)|^2$ angle by angle. The theorem is a conservation law, not a substitute for solving the scattering problem.

Worked examples

1 · Measuring a molecule's forward amplitude with a refractometer (confidence builder)

Setup. Dry air at 15 °C: refractivity $n - 1 = 2.77\times10^{-4}$ at $\lambda = 589$ nm, number density $N = 2.55\times10^{25}$ m⁻³. Find the forward scattering amplitude per molecule.

Step 1. Invert the slab formula. $\mathrm{Re}\,f(0) = \frac{(n-1)k^2}{2\pi N}$ with $k = 2\pi/\lambda = 1.067\times10^7$ m⁻¹.

Step 2. The number. $\mathrm{Re}\,f(0) = \frac{2.77\times10^{-4}\times(1.067\times10^7)^2}{2\pi\times2.55\times10^{25}} = \mathbf{2.0\times10^{-16}}$ m $= 0.20$ fm — a refractometer is a forward-scattering-amplitude meter, and every air molecule scatters visible light with an amplitude a fifth of a femtometer long.

Step 3. The imaginary part, by the theorem. Rayleigh's cross section for air at 589 nm is $\sigma \approx 4\times10^{-31}$ m², so $\mathrm{Im}\,f(0) = k\sigma/4\pi \approx \mathbf{3\times10^{-25}}$ m — nine orders below the real part. Light in air is phase-shifted mightily and dimmed barely: precisely the ratio that makes the atmosphere transparent yet refractive, and the sky's slow blue extinction (the $\lambda^{-4}$ story of Rayleigh Scattering) is this tiny imaginary part accumulated over kilometers.

The lesson: dispersion and extinction are the real and imaginary parts of one measurable number — and for air, you have measured its real part every time you trusted a star's apparent position.

2 · The audit at three wavelengths, and the Thomson closure (exam level)

Setup. A bound electron with resonance at $\lambda_0 = 90$ nm (far-UV, the air-like oscillator of Absorption and Dispersion), damped only by radiation reaction, $\tau = 6.266\times10^{-24}$ s. Check the optical theorem at $\lambda = 550$ nm, then close the loop on the free electron.

Step 1. The polarizability. At $\lambda = 550$ nm: $\alpha = \frac{e^2/m\varepsilon_0}{\omega_0^2 - \omega^2 - i\omega^3\tau}$ ; the two routes to the cross section are $\sigma_{\rm tot} = k\,\mathrm{Im}\,\alpha$ (theorem) and $\sigma_{\rm scat} = \frac{k^4|\alpha|^2}{6\pi}$ (Larmor integral).

Step 2. The numbers. $\sigma_{\rm tot} = \mathbf{5.035894\times10^{-32}}$ m² and $\sigma_{\rm scat} = \mathbf{5.035894\times10^{-32}}$ m² — ratio $1.000000000000$ : the books balance to machine precision, and repeat at 200 nm ( $4.29\times10^{-30}$ m²) and at resonance-adjacent 91 nm ( $1.33\times10^{-25}$ m²). Drop the $i\omega^3\tau$ term and $\mathrm{Im}\,\alpha = 0$ : the theorem predicts zero against a finite $\sigma_{\rm scat}$ — energy conservation broken by exactly the amount the self-force restores.

Step 3. Thomson closure. Free the electron ( $\omega_0 = 0$ ) at any optical frequency: $k\,\mathrm{Im}\,\alpha \to \frac{8\pi}{3}r_e^2 = \mathbf{0.6652}$ barn $= 6.652\times10^{-29}$ m², matching $\sigma_T$ of Thomson Scattering to eight digits. The classical electron radius $r_e$ entered through $\tau = 2r_e/3c$ : the self-force timescale, the Thomson cross section, and the optical theorem are one consistent structure.

Step 4. The moral, quantitatively. At 550 nm the radiative width is $\omega^3\tau/\omega_0\omega \sim 10^{-8}$ of the resonance frequency — utterly negligible for the real part of $\alpha$ (dispersion doesn't notice) yet carrying all of the extinction. Unitarity lives entirely in terms you would round away.

The lesson: the optical theorem is an equality between a second-order quantity ( $|\alpha|^2$ ) and a "small correction" ( $\mathrm{Im}\,\alpha$ ) — which is why it is the sharpest consistency test a scattering calculation can face.

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., §10.11 · Newton, Scattering Theory of Waves and Particles, 2nd ed., Ch. 1 · Bohren & Huffman, Absorption and Scattering of Light by Small Particles, Ch. 3–4 · van de Hulst, Light Scattering by Small Particles · Zangwill, Modern Electrodynamics, Ch. 21 · Draine, ApJ 333, 848 (1988)

Provenance. A verifier should reproduce (1) the master relation $\sigma_{\mathrm{tot}}=(4\pi/k)\mathrm{Im}\,f(0)$ via the flux/stationary-phase derivation in Jackson §10.11 and Newton's Scattering Theory; (2) the partial-wave identity $\sigma_{\mathrm{tot}}=(2\pi/k^2)\sum(2\ell+1)\mathrm{Re}(1-S_\ell)$ ; (3) the Rayleigh dipole results $\sigma_{\mathrm{scatt}}=k^4|\alpha|^2/6\pi$ and the radiative correction $\mathrm{Im}\,\alpha=k^3|\alpha|^2/6\pi$ (Bohren & Huffman; Draine 1988); and (4) the large-object extinction limit $\sigma_{\mathrm{ext}}\to2\pi a^2$ . Numbers quoted are standard textbook results, not fabricated. Addendum for this edition: item (3) was upgraded from a quoted correction to the article's centerpiece consistency theorem — the radiatively damped oscillator was run through both sides of the theorem symbolically (the difference vanishes identically iff $\tau = e^2/6\pi\varepsilon_0 mc^3$ , the Larmor time of the E9 self-force articles) and numerically at three wavelengths (ratio $= 1.000000000000$ ), with the Thomson closure $\sigma_T$ recovered to eight digits from the forward amplitude; the audit SVG is traced from these computed curves. The partial-wave/S-matrix edition (2) is stated here with its unitarity reading but its quantum machinery belongs to the QM track; the Mie-coefficient extinction sum is cross-referenced to Scattering by Conducting and Dielectric Spheres, whose deferred extinction paradox is delivered in Derivation C (both routes to $2\pi a^2$ agreeing exactly, with the acceptance-angle fine print). Division of labor: the self-force is owned by Radiation Reaction Force and Abraham–Lorentz Equation and Its Pathologies (recalled by name, never re-derived — this article pins its coefficient by unitarity); bulk dispersion and Kramers–Kronig by Absorption and Dispersion; $\sigma_T$ by Thomson Scattering; the sky's numbers by Rayleigh Scattering. The bare slug `optical-theorem` is owned by the Optics track — instrument-level extinction practice, radiative transfer, and coherence live there (the full story in the Optics track); this article owns the electromagnetic derivation, the radiation-reaction audit, and the index/paradox payoffs. Worked numbers (air $f(0) = 0.20$ fm; audit $5.035894\times10^{-32}$ m² both routes; $\sigma_T = 0.6652$ b) recomputed for this edition.

Builds on: Multipole Expansion of Radiation Fields · Electric Dipole Radiation: Exact Fields and Zone Structure · Larmor Formula from the Liénard–Wiechert Fields · Vector Spherical Harmonics Leads to: Magnetic Dipole and Electric Quadrupole Radiation · Antenna Arrays and Directivity

How can one number at one angle know about power scattered into every other angle?

Because conservation is a global constraint and the forward direction is where the ledger is kept. The beam's energy can only decrease by interference with something phase-locked to it, and the only surviving phase-locked partner is the forward-scattered wave — everything else oscillates itself to zero in the flux integral. So whatever the target does internally (scatter sideways, absorb), the sum of the losses is debited from the forward beam, and the debit mechanism is $\mathrm{Im}\,f(0)$ . The theorem is blind to the distribution by construction: the cashier's till knows the total, not the shopping list.

Is the optical theorem exact even for strong scatterers, resonances, and absorbing targets?

Yes — that is its entire value. The derivation used only two ingredients: a conserved flux (Poynting's theorem, owned by the E7 conservation chapter) and the asymptotic form "plane wave plus outgoing spherical wave." Nothing assumed weakness, smallness, or transparency. It holds on a plasmon resonance where perturbation theory diverges, for a black absorber, for a Mie sphere at any size parameter — the partial-wave form $\sigma_{\rm tot} = \frac{2\pi}{k^2}\sum(2\ell+1)\mathrm{Re}(1 - S_\ell)$ makes it term-by-term explicit, with $|S_\ell| < 1$ signaling absorption. What is not exact is any approximate amplitude you insert into it — and the mismatch between the theorem's two sides is then the sharpest available measure of how badly your approximation leaks energy.

Does the radiation-reaction audit mean the pathological Abraham-Lorentz force is actually correct?

It means the damping coefficient is not optional — a subtler and safer statement. The audit pins one number: in the driven, steady-state, small-amplitude regime, the oscillator's effective damping must be $\gamma_{\rm rad} = \omega^2\tau$ with $\tau = e^2/6\pi\varepsilon_0 mc^3$ , or energy conservation fails by exactly the deficit computed in Derivation B. That steady-state coefficient is common to the Abraham–Lorentz equation and to its pathology-free Landau–Lifshitz reduction — the runaway and preacceleration diseases live in the transient, self-driven sector that a steadily driven oscillator never visits (their prosecution is Abraham–Lorentz Equation and Its Pathologies' business, with the systematic treatment in this chapter's Abraham-Lorentz Radiation Reaction Force). So the optical theorem certifies the part of the self-force that is physically unambiguous — and stays diplomatically silent on the part that is not.

If the imaginary part of alpha is a factor of ten to the minus eight, why can it not just be neglected like every other tiny correction?

Because it is not a correction to something — it is the something. For dispersion (the real part of $n$ ), $\mathrm{Im}\,\alpha$ contributes at second order and is genuinely negligible: no refractometer notices. But extinction has no zeroth-order term: the entire attenuation of a lossless scattering medium is $k\,\mathrm{Im}\,\alpha$ per molecule. Neglecting it does not make a small error in the answer — it makes the answer identically zero, qualitatively wrong (a sky that never dims the sun, a fog with no opacity). This is the general shape of unitarity constraints: they equate a "small" quantity ( $\mathrm{Im}\,f$ ) to the square of a leading one ( $|f|^2$ ), so the small term is precisely as important as the leading term's consequences. Perturbative habits — "drop what is small" — need the refinement "drop what is small compared to something else doing the same job."

What exactly happens in the shadow region behind a large obstacle?

Superposition, not blockage. The total field behind the disk is (incident) + (scattered), and in the geometric shadow the scattered wave has amplitude nearly equal and phase nearly opposite to the incident — their sum is small. The missing energy is redistributed: half the extinction budget went to reflection/absorption at the obstacle, half rides the forward diffraction lobe just outside the shadow boundary. Walk downstream and the shadow heals — the lobe spreads and refills the axis beyond $z \sim a^2/\lambda$ (the Fresnel-number boundary of Scalar and Vector Diffraction Theory; the Poisson–Arago spot is the first symptom). The theorem is indifferent to the drama: it audits the sphere at infinity, where the books have long since balanced.

How does the electromagnetic version differ from the scalar theorem — does polarization change anything?

Only the bookkeeping, not the structure. Light carries two transverse channels, so the amplitude becomes a vector $\mathbf{f}(\hat{\mathbf{n}})$ and the theorem projects it onto the incident polarization: $\sigma_{\rm ext} = \frac{4\pi}{k}\mathrm{Im}[\hat{\boldsymbol{\epsilon}}_0^*\cdot\mathbf{f}(0)]$ . The projection matters: only the forward-scattered component in the same polarization state stays coherent with the beam and can interfere. Consequences worth knowing: extinction can be polarization-dependent (dichroism — aligned interstellar grains polarize starlight by differential extinction, which is how galactic magnetic fields were first mapped); for anisotropic scatterers $\mathbf{f}(0)$ need not be parallel to $\hat{\boldsymbol{\epsilon}}_0$ (birefringence of the dilute medium follows); and in the Mie edition the projection is what makes $Q_{\rm ext} = \frac{2}{x^2}\sum(2\ell+1)\mathrm{Re}(a_\ell + b_\ell)$ a linear sum over channels while $Q_{\rm sc}$ is quadratic — extinction is interference, scattering is power.

Where does the optical theorem sit relative to Kramers-Kronig and causality?

They are two constraints on the same amplitude, from two different sanctities. The optical theorem is unitarity at fixed frequency: energy conservation ties $\mathrm{Im}\,f(0,\omega)$ to $\sigma_{\rm tot}$ at that same $\omega$ . Kramers–Kronig is causality across frequencies: $\mathrm{Re}\,f(0,\omega)$ is an integral over $\mathrm{Im}\,f(0,\omega')$ at all others — the machinery Absorption and Dispersion builds for $\varepsilon(\omega)$ applies verbatim as forward dispersion relations. Combined they are extraordinarily restrictive: measure total cross sections at all energies and causality reconstructs the real forward amplitude — used in earnest in hadron physics to cross-check the very $\sigma_{\rm tot}$ the theorem extracts from elastic slopes. One amplitude, two sacred principles, almost no freedom left.