Optical Theorem

Scattering theory owns exactly one equation that is true for every potential, every energy, every number of open channels, with no approximation anywhere in sight: $\sigma_{\mathrm{tot}}=(4\pi/k)\,\mathrm{Im}\,f(0)$ . On the left sits the total cross section — everything a target does to a beam, elastic ricochets and swallowed flux alike, integrated over all $4\pi$ of solid angle. On the right sits a single number read at a single angle: the imaginary part of the scattering amplitude in the exact forward direction. A detector parked at $\theta=0$ , where the unscattered beam is blinding, somehow knows the whole budget. Scattering Amplitude found the mechanism — the beam–ripple interference is exiled to a forward cone shrinking as $(kr)^{-1/2}$ — and promised that its surviving value is this article's. We collect the debt twice: by flux bookkeeping (one closed sphere, three currents, a two-endpoint integral whose backward member we compute rather than wave away, and show dies as $1/kr$ ), and by partial-wave assembly, where the theorem is a one-line identity between two appearances of one sum — machine-verified to $10^{-12}$ , including through a resonance saturating the unitarity ceiling. Along the way we repair two defects in our source text (a factor of 2 and a garbled bound, both machine-proven), exhibit why the Born approximation must violate the theorem and measure the $V^3$ rate at which second order repays the debt, and put real numbers on both ends of physics: thermal neutrons on protons, and protons on protons at 13 TeV, where this seventy-year-old identity is still how the LHC measures its total cross section.

Key takeaways

The optical theorem: $\sigma_{\mathrm{tot}}=\frac{4\pi}{k}\,\mathrm{Im}\,f(0)$ — exact, potential-independent, valid with absorption. Derived twice (flux ledger; partial-wave assembly), machine-verified to $10^{-12}$ including through a resonance.
The forward endpoint alone gives $-\sigma_{\mathrm{tot}}$ — at large $r$ the flux sits on it with no averaging needed; the backward endpoint $e^{2ikr}f(\pi)$ survives as a residual oscillation suppressed by $1/(kr)$ that vanishes as $r\to\infty$ — the honesty step most texts skip, machine-measured to fall off with $r$ .
In partial waves both sides equal $\frac{1}{k}\sum_\ell(2\ell+1)\sin^2\delta_\ell$ (times $4\pi/k$ for $\sigma$ ): the theorem is unitarity, channel by channel.
First Born order cannot obey it: $f_B$ is real, yet $\sigma_B>0$ . The repair $\mathrm{Im}\,f^{(2)}(0)=\frac{k}{4\pi}\int|f^{(1)}|^2\mathrm{d}\Omega$ is Born Series and Higher-Order Corrections' theorem; we measure the residual gap closing as $V^{3.0}$ .
Real numbers: thermal n–p ( $\sigma=20.4$ b) $\Rightarrow\mathrm{Im}\,f(0)=2.8\times10^{-3}$ fm (0.02% of $|f|$ ); pp at 13 TeV ( $\sigma=110.6$ mb) $\Rightarrow\mathrm{Im}\,f(0)\approx2.9\times10^{4}$ fm and optical point $\mathrm{d}\sigma/\mathrm{d}t|_0\approx630$ mb/GeV² — how TOTEM measures $\sigma_{\mathrm{tot}}$ .
Key idea: probability conservation forces the forward-scattered wave to interfere destructively with the beam by exactly the amount removed from it, so the total cross section is written, in full, in the imaginary part of one forward number.

The core insight

Probability conservation forces the forward-scattered wave to interfere destructively with the incident beam; the strength of that interference, $\mathrm{Im}\,f(0)$ , must equal the total flux removed from the beam, giving $\sigma_{\mathrm{tot}}=(4\pi/k)\,\mathrm{Im}\,f(0)$ .

Where this lives today

Forward physics at the LHC is an optical-theorem industry. TOTEM and ATLAS-ALFA park silicon detectors in Roman pots millimetres from the beam to measure elastic $\mathrm{d}\sigma/\mathrm{d}t$ at $|t|$ down to $10^{-3}\ \mathrm{GeV^2}$ , then extract $\sigma_{\mathrm{tot}}$ from the optical point, Eq. (7). The 2019 measurement of $\rho$ at 13 TeV strained Pomeron fits and revived the "odderon" (C-odd gluonic exchange) — hunted through the real part of a forward amplitude. The absorptive face runs the nuclear data pipeline: optical-model fits of neutron total cross sections (whence "optical model") anchor reactor and shielding calculations, and the same identity governs how an engineered nanoresonator dims a beam. Every "unitarity bound" on a new-physics amplitude is the channel-diagonal version assembled here.

The physical picture

Stand behind a scatterer: there is a shadow, less beam than the source sent. But a stationary state absorbs no probability locally (for a real potential, none at all), so the shadow is not a deletion — it is interference. Directly behind the target the forward-scattered wave $f(0)\,e^{ikr}/r$ rides in lockstep with $e^{ikz}$ — same direction, same wavenumber, phase-locked forever — and partly cancels it. Every missing particle reappears elsewhere, in the ring of detectors reading $|f(\theta)|^2$ or in a reaction channel. The optical theorem is that closed ledger: cancelled forward flux on one page, scattered-plus-absorbed flux on the other, and they must match. Why only the imaginary part? Cancellation is a phase statement. Downstream the two waves share $e^{ikr}$ , so their relative phase is set by $f(0)$ ; the angular integration over the forward cone supplies one stationary-phase factor of $i$ (optics' Gouy shift). That $i$ rotates $\mathrm{Re}$ into $\mathrm{Im}$ : $\mathrm{Re}f(0)$ only slides the fringes in $r$ , while $\mathrm{Im}f(0)$ alone changes the count. A scatterer that merely phase-shifts the forward wave ( $f(0)$ real) dents nothing on average — and the theorem forces $\sigma_{\mathrm{tot}}=0$ . The partial-wave view says the same channel by channel. Each $f_\ell$ lives on The S-Matrix and Unitarity in Scattering's unitarity circle (radius $1/2k$ , centre $i/2k$ ), so $\mathrm{Im}\,f_\ell=k|f_\ell|^2$ — the optical theorem for one channel. Summing with the $(2\ell+1)$ weights assembles the whole. Unitarity is not an ingredient of the theorem; it is the theorem, diagonalized.

Figure

The interference flux vs r: forward endpoint fixes −σ_tot, backward wobble dies as 1/kr

The theorem and its three faces

The statement itself is one line. Every quantity in it was built earlier in this chapter: the amplitude and the far-field form $\psi\to e^{ikz}+f(\theta,\phi)\,e^{ikr}/r$ are Scattering Amplitude's and Asymptotic Form of the Scattering Wave Function's; the cross section and its counting definition are Differential and Total Cross Section's. New here is the lock between them.

\sigma_{\mathrm{tot}}=\frac{4\pi}{k}\,\mathrm{Im}\,f(\theta=0) \qquad (1)

where: $\sigma_{\mathrm{tot}}$ is the total cross section (elastic plus every open reaction channel), $k=\sqrt{2mE}/\hbar$ the incident wavenumber, $f(0)$ the elastic amplitude in the exact forward direction. The lock is dimensionally tight: $f$ and $4\pi/k$ are both lengths, their product the area the beam loses. Note what is not in it — no potential, no truncation, no weak coupling: it holds for a hard sphere at $ka=10^{-3}$ and a proton at 13 TeV. The theorem wears three faces: the flux ledger (interference deficit at $\theta=0$ equals outgoing $\sigma$ , Derivation A); the partial-wave sum ( $\mathrm{Im}\,f(0)$ and $\frac{k}{4\pi}\sigma$ are one $\sum_\ell(2\ell+1)\sin^2\delta_\ell$ , Derivation B); and channel unitarity ( $\mathrm{Im}\,f_\ell=k\lvert f_\ell\rvert^2$ from $\lvert S_\ell\rvert=1$ , The S-Matrix and Unitarity in Scattering's, assembled here).

Derivation AThe flux ledger: the whole theorem from one closed sphere

We want the exact forward-interference value Scattering Amplitude located but did not evaluate. The strategy is an audit: surround the target with a sphere of radius r in the far field, demand that the net probability current through it vanish (stationary state, real potential), and split the current into beam, ripple, and interference pieces. The delicate step is the angular integral of the interference piece — a two-endpoint affair in which we keep the oscillating backward term most treatments discard silently, because seeing what kills it is the difference between believing the theorem and understanding it.

Step 1. Split the current. With $\psi=e^{ikz}+f(\theta)e^{ikr}/r$ , the current $\mathbf j=(\hbar/m)\,\mathrm{Im}(\psi^*\nabla\psi)$ splits into beam, ripple, and interference pieces. The pure beam piece is a uniform $\hat z$ current: over any closed surface it integrates to zero — what enters upstream leaves downstream.

Step 2. The ripple carries σ. The radial $\mathbf j_{\mathrm{sc}}$ is $(\hbar k/m)|f(\theta)|^2/r^2$ to leading order, so its outward flux is $(\hbar k/m)\int|f|^2\mathrm{d}\Omega=(\hbar k/m)\,\sigma$ — the area's $r^2$ cancelling the intensity's $1/r^2$ , the signature of a genuine radiating wave.

Step 3. The interference piece is a two-endpoint integral. The cross current carries the relative phase $e^{ikr(1-\cos\theta)}$ , and its surface integral reduces to $2\pi r\int_{-1}^{1}e^{ikr(1-u)}g(u)\,\mathrm{d}u$ -type integrals ( $u=\cos\theta$ , $g$ slowly varying in $f,u$ ). The bare exponential integrates exactly to

\int_{-1}^{1}e^{ikr(1-u)}\,\mathrm{d}u=\frac{e^{2ikr}-1}{ikr} \qquad (2)

where: the $-1$ term is the forward endpoint $u=1$ — $r$ -independent in phase, permanent; the $e^{2ikr}$ term is the backward endpoint $u=-1$ , spinning twice per wavelength of $r$ . For slowly varying $g$ , stationary phase pins each endpoint to $g(\pm1)$ , i.e. $f(0)$ and $f(\pi)$ ; corrections enter at next order in $1/kr$ , the forward cone having the $\theta_1\simeq2\sqrt{\pi/kr}$ width Scattering Amplitude measured.

Step 4. Evaluate, and keep the honest wobble. Assembling the geometry (ripple's $1/r$ , area's $r^2$ , the $1/ikr$ of Eq. 2), the interference flux is a permanent forward deficit plus a backward endpoint riding $e^{2ikr}$ :

\Phi_{\mathrm{int}}(r)=-\frac{\hbar}{m}\,4\pi\,\mathrm{Im}\,f(0)\;+\;\frac{\hbar}{m}\,\frac{4\pi}{kr}\,\mathrm{Im}\!\left[e^{2ikr}f(\pi)\right]+\cdots \qquad (3\ast)

The forward term is $r$ -independent: the flux sits on it with no averaging needed. The backward endpoint is a real oscillation, but suppressed by $1/(kr)$ — the machine run (hard sphere, $ka=2$ ) measures it at only 0.07% of $\sigma$ by $kr\approx100$ , decaying to $\sim10^{-6}$ of $\sigma$ by $kr\approx3000$ , where the flux equals $-\sigma_{\mathrm{tot}}$ to $7\times10^{-4}$ from the forward endpoint alone. It dies twice over: geometrically as $r\to\infty$ (the limit the theorem is stated in), and physically through any beam's energy spread, which spins $e^{2ikr}$ through full turns. The forward term never fades — its phase is locked at every $r$ .

Step 5. Close the books. A stationary state with a real potential holds no probability permanently: $\oint\mathbf j\cdot\mathrm{d}\mathbf A=0$ . Beam gives zero, ripple $+(\hbar k/m)\sigma$ , interference $-(\hbar k/m)(4\pi/k)\,\mathrm{Im}f(0)$ (forward endpoint, at large $r$ ). The sum vanishes only if $\sigma=(4\pi/k)\,\mathrm{Im}f(0)$ — Eq. (1). For an absorbing target ( $\oint\mathbf j\cdot\mathrm{d}\mathbf A=-(\hbar k/m)\sigma_{\mathrm{abs}}$ ) the same ledger gives $\sigma_{\mathrm{el}}+\sigma_{\mathrm{abs}}=(4\pi/k)\,\mathrm{Im}f(0)$ : the forward dent counts flux it never sees again. Verified: Eq. (2) exact in sympy; the full interference flux computed numerically on a 200,000-point angular grid — at large r it equals −σ_tot to 7×10⁻⁴ from the forward endpoint alone; the backward-endpoint oscillation is nonzero at moderate r (0.07% of σ at kr≈100) and decays with r, negligible (10⁻⁶) by kr≈3000 (this article's verification script).

Go deeper · Derivation B · 1 minThe partial-wave assembly: one sum appears on both sides

Derivation BThe partial-wave assembly: one sum appears on both sides

We now re-derive Eq. (1) with no surface, no stationary phase, no averaging — three lines on machinery this chapter owns — and repair a factor-of-two defect in the source text. Partial-Wave Expansion gave the ladder f(θ)=Σ(2ℓ+1)f_ℓ P_ℓ(cosθ), f_ℓ=e^{iδ_ℓ}sinδ_ℓ/k, and the sum σ=(4π/k²)Σ(2ℓ+1)sin²δ_ℓ; The S-Matrix and Unitarity in Scattering gave each channel's circle identity Im f_ℓ=k|f_ℓ|². Our job is only the assembly — but the assembly IS the theorem.

Step 1. Forward value. At $\theta=0$ every Legendre polynomial equals one, $P_\ell(1)=1$ , so

\mathrm{Im}\,f(0)=\frac{1}{k}\sum_\ell(2\ell+1)\,\mathrm{Im}\!\left[e^{i\delta_\ell}\sin\delta_\ell\right]=\frac{1}{k}\sum_\ell(2\ell+1)\sin^2\delta_\ell \qquad (4)

where: the only trigonometry is $\mathrm{Im}(e^{i\delta}\sin\delta)=\sin^2\delta$ — each channel contributes its $\sin^2\delta_\ell$ , the same quantity measuring how strongly it scatters. This is the circle identity: $\mathrm{Im}f_\ell=\sin^2\delta_\ell/k=k|f_\ell|^2$ .

Step 2. Compare with the cross-section sum. Partial-Wave Expansion's $\sigma=(4\pi/k^2)\sum_\ell(2\ell+1)\sin^2\delta_\ell$ is the same sum times $4\pi/k^2$ . Divide: $\sigma/\mathrm{Im}f(0)=4\pi/k$ term by term — Eq. (1), now a channel-by-channel conservation statement, not a global accident. The machine check runs both sides independently (quadrature of $\int|f|^2\mathrm{d}\Omega$ vs the forward imaginary part) for hard spheres at $ka=0.5,2,5$ and a δ-shell resonance model swept through its peak, where $\sigma_0$ saturates the $4\pi/k^2$ ceiling: agreement to $10^{-12}$ everywhere.

Step 3. Repair the source's factor of two. Our base text states the unitarity route as " $2\,\mathrm{Im}f(0)=(k/4\pi)\int|f|^2\mathrm{d}\Omega$ " — off by exactly 2 (the machine evaluates both sides for generic phase shifts; the ratio is 2, identically). The slip is instructive: the operator relation $-i(T-T^\dagger)=T^\dagger T$ has left side $2\,\mathrm{Im}\,T$ , and it is tempting to carry that 2 into the amplitude statement. But the normalization connecting $\langle\mathbf k'|T|\mathbf k\rangle$ to $f$ absorbs it: with $\mathrm{d}\sigma/\mathrm{d}\Omega=|f|^2$ , the on-shell relation is $\mathrm{Im}f(0)=(k/4\pi)\int|f|^2\mathrm{d}\Omega$ , no 2. Eq. (4) — three lines of trigonometry — is the arbiter. Verified: symbolic two-route equality for generic δ₀…δ₃ (sympy, exact); the base's factor-2 line reproduced and shown ≡ 2× the true relation; numeric two-route equality ≤10⁻¹² for hard sphere ka=0.5/2/5 and through the δ-shell resonance; ceiling saturation at the resonance confirmed to 0.2% (verification script).

What first order owes, and how the ledger collects

Here is the theorem's most useful day job: policing approximations. Born Approximation gives a real $f_B$ for any real $V$ (a symmetric Fourier transform), so $\mathrm{Im}f_B(0)=0$ while $\sigma_B=\int|f_B|^2\mathrm{d}\Omega>0$ : first order violates the theorem by construction, scattering flux it never debits from the forward beam. Nothing is wrong — the theorem relates order $n$ amplitudes to order $2n$ cross sections, so the $O(V^2)$ deficit in $\sigma$ is repaid by the $O(V^2)$ amplitude, second Born order. That repayment is exact, Born Series and Higher-Order Corrections' structural theorem: $\mathrm{Im}f^{(2)}(0)=\frac{k}{4\pi}\int|f^{(1)}|^2\mathrm{d}\Omega$ , the on-shell ( $i\varepsilon$ ) piece of the second-order propagator supplying the missing imaginary part. This article adds the measurement. For an attractive square well ( $\hbar=m=1$ , $a=1$ , $k=1$ ) we tracked the deficit $\big|\mathrm{Im}f_{\mathrm{exact}}(0)-\tfrac{k}{4\pi}\sigma_B\big|$ as the depth was halved: it fell by $2^{3.0}$ — the residual closes as $V^{3}$ , exactly as an identity satisfied through $O(V^2)$ demands. Symbolically too: expanding $\delta_\ell=\lambda e_\ell+O(\lambda^2)$ , the $O(\lambda^2)$ imaginary part of $f(0)$ equals $(k/4\pi)\int|f^{(1)}|^2\mathrm{d}\Omega$ term for term, for arbitrary channel coefficients.

Figure

One number, both routes — and where Born sits

The forward floor: what the theorem forces on measurement

Because $|f(0)|^2=[\mathrm{Re}f(0)]^2+[\mathrm{Im}f(0)]^2$ , the theorem sets a hard floor under the forward differential cross section. Our source text shipped this bound garbled (a literal "?" for the inequality's left side — an export corruption, repaired here and flagged in the cheat sheet); the correct statement is one line, worth deriving because a billion-dollar measurement program rests on it.

Go deeper · Derivation C · 1 minThe forward floor and the optical point

Derivation CThe forward floor and the optical point

We want the minimum forward peak a scatterer must show given its total cross section, and the dictionary turning this floor into the collider observable dσ/dt at t=0 — the "optical point". The payoff is the actual method by which total hadronic cross sections are measured: σ_tot is not directly countable at a collider (you cannot instrument 4π at millidegree angles), so experiments measure the forward elastic peak and let this floor-turned-equality do the rest.

Step 1. The floor. Split $f(0)$ into real and imaginary parts and use Eq. (1):

\frac{\mathrm{d}\sigma}{\mathrm{d}\Omega}\bigg|_{0}=|f(0)|^2\;\ge\;[\mathrm{Im}\,f(0)]^2=\left(\frac{k\,\sigma_{\mathrm{tot}}}{4\pi}\right)^{\!2} \qquad (5)

where: equality holds when $\mathrm{Re}f(0)=0$ (purely absorptive). The floor is quadratic in $\sigma_{\mathrm{tot}}$ and grows with $k^2$ : fast, strongly interacting systems are forced to be forward-peaked. A measured peak below this floor is a normalization error, not new physics.

Step 2. Promote to an equality with one parameter. Define $\rho\equiv\mathrm{Re}f(0)/\mathrm{Im}f(0)$ . Then exactly

\frac{\mathrm{d}\sigma}{\mathrm{d}\Omega}\bigg|_{0}=(1+\rho^2)\left(\frac{k\,\sigma_{\mathrm{tot}}}{4\pi}\right)^{\!2} \qquad (6)

where: $\rho$ is small when absorption dominates (pp at LHC: $\rho\approx0.10$ , a 1% $\rho^2$ correction). Forward peak plus $\rho$ gives $\sigma_{\mathrm{tot}}$ — no $4\pi$ detector needed.

Step 3. The collider dictionary. Experiments bin in $t=-q^2=-2k^2(1-\cos\theta)$ , so $\mathrm{d}\sigma/\mathrm{d}t=(\pi/k^2)\,\mathrm{d}\sigma/\mathrm{d}\Omega$ . At $t=0$ , Eq. (6) becomes the optical-point formula

\frac{\mathrm{d}\sigma}{\mathrm{d}t}\bigg|_{0}=\frac{(1+\rho^2)\,\sigma_{\mathrm{tot}}^{2}}{16\pi} \qquad (7)

where: every factor of $k$ cancels — the optical point depends only on $\sigma_{\mathrm{tot}}$ and $\rho$ , a clean experimental target. (Units: with $\sigma$ in GeV⁻² via $(\hbar c)^2=0.3894$ mb·GeV², Eq. (7) returns GeV⁻⁴; × $(\hbar c)^2$ again for mb/GeV².) Verified: Eqs. (5)–(7) checked numerically on the np and pp numbers of the worked examples; the np floor is satisfied by 7 orders of magnitude of headroom, and the pp optical point evaluates to 631 mb/GeV² (verification script).

Limiting cases

case	result	meaning
real potential, purely elastic	$\sigma_{\mathrm{el}}=(4\pi/k)\,\mathrm{Im}f(0)$	every dented beam particle reappears in $\lvert f\rvert^2$
absorbing target ( $\lvert S_\ell\rvert<1$ )	$\sigma_{\mathrm{el}}+\sigma_{\mathrm{abs}}=(4\pi/k)\,\mathrm{Im}f(0)$	forward dent counts undetected channels; anatomy in Inelastic Scattering and Multichannel Theory
on a resonance, $\delta_\ell=\pi/2$	both sides saturate the $4\pi(2\ell+1)/k^2$ ceiling	verified here to $10^{-12}$ at the δ-shell peak; lineshape is Resonances and the Breit-Wigner Formula's
$k\to0$ (s-wave world)	$\mathrm{Im}f(0)=ka^2/(1+k^2a^2)=k\sigma/4\pi$ exactly	scattering-length regime: both sides vanish linearly in $k$ ; the amplitude turns real at threshold, owned by Low-Energy Scattering and the Scattering Length
first Born order	$\mathrm{Im}f_B(0)=0$ while $\sigma_B>0$	the theorem fails at $O(V^2)$ and is repaid exactly at second order — Born Series and Higher-Order Corrections
black disk at high energy	$\sigma_{\mathrm{el}}=\sigma_{\mathrm{abs}}=\sigma_{\mathrm{tot}}/2$	maximal absorption still satisfies the theorem — half the "total" is the shadow's own diffraction

Worked examples

Example 1 — thermal neutrons on protons: how small an exact statement can be. The thermal n–p total cross section is $\sigma=20.4$ b — Differential and Total Cross Section's ladder anchor. What does the theorem force on the forward amplitude?

Step 1. Kinematics in the CM. A neutron at $E_{\mathrm{lab}}=25.3$ meV on a nearly free proton: $k_{\mathrm{lab}}=\sqrt{2m_nE}/\hbar c=3.49\times10^{-5}\ \mathrm{fm}^{-1}$ , and with $\mu\approx m_n/2$ the relative wavenumber is $k=1.75\times10^{-5}\ \mathrm{fm}^{-1}$ .

Step 2. Apply the theorem. $\mathrm{Im}\,f(0)=k\sigma/4\pi=(1.75\times10^{-5}\times2044\ \mathrm{fm})/4\pi$ $=2.8\times10^{-3}\ \mathrm{fm}$ .

Step 3. Compare sizes. S-wave dominance gives $|f|\simeq\sqrt{\sigma/4\pi}=12.8$ fm, so $\mathrm{Im}f(0)/|f(0)|\approx2.2\times10^{-4}$ : the forward amplitude is 99.98% real. No contradiction — the theorem never says $\mathrm{Im}f$ is large, only exactly $k\sigma/4\pi$ , and at threshold both sides vanish linearly in $k$ . Indeed $f=-a/(1+ika)$ gives $\mathrm{Im}f(0)=ka^2/(1+k^2a^2)=k\sigma/4\pi$ identically (sympy-verified). Answer: $\mathrm{Im}f(0)=2.8\times10^{-3}$ fm on a 12.8 fm nearly-real amplitude.

Example 2 — pp at 13 TeV: measuring σ_tot with the theorem. TOTEM's measured $\sigma_{\mathrm{tot}}=110.6$ mb at $\sqrt s=13$ TeV. Run the machine forward, then explain how the experiment actually ran it backward.

Step 1. Convert: $\sigma=110.6/0.3894=284\ \mathrm{GeV^{-2}}$ ; CM momentum $k\simeq\sqrt s/2=6500$ GeV.

Step 2. Forward amplitude: $\mathrm{Im}f(0)=k\sigma/4\pi=6500\times284/4\pi=1.47\times10^{5}\ \mathrm{GeV^{-1}}$ , i.e. $\times\,\hbar c=0.1973\ \mathrm{GeV\,fm}$ : $\mathbf{2.9\times10^{4}}$ fm — thirty nanometers of forward amplitude from a femtometer-sized proton. Forward peaking at high energy is not an option; it is Eq. (5).

Step 3. Optical point, Eq. (7): $\mathrm{d}\sigma/\mathrm{d}t|_0=(1+0.10^2)\times284^2/16\pi=\mathbf{631\ mb/GeV^2}$ . In practice the logic inverts: TOTEM measures the elastic peak, extrapolates to $t=0$ , takes $\rho$ from Coulomb–nuclear interference, and solves Eq. (7) for $\sigma_{\mathrm{tot}}$ — the optical theorem is the measurement, not a check. Answer: $\mathrm{Im}f(0)\approx1.5\times10^{5}$ GeV⁻¹; optical point $\approx630$ mb/GeV².

Origins · 1 minOptical Theorem and how this was discovered

Optical Theorem

1871–1955 · Cambridge, Copenhagen

The theorem is older than quantum mechanics. In the 1870s optics already knew — through Sellmeier's dispersion work and Rayleigh's scattered-light analysis — that a beam's attenuation through a medium is fixed by the forward response of its scatterers, the optics "extinction theorem." Quantum mechanics inherited the ledger: in 1932 Eugene Feenberg proved the relation for matter waves (electron–atom scattering). Its full generality — total cross section including all reaction channels, from the forward elastic amplitude alone — crystallized in a 1939 nuclear-dispersion analysis by Bohr, Peierls and Placzek that was never published (the war intervened; physics cites the manuscript anyway). Postwar, van de Hulst made the extinction form standard and the "optical" name migrated into particle physics, where the theorem now converts LHC forward-peak measurements into total cross sections no detector could integrate directly.

History

1871

— Rayleigh's scattering analysis and Sellmeier's dispersion theory tie a medium's attenuation to the forward response of its scatterers — the optics ancestry of the theorem

1932

— Eugene Feenberg proves the quantum relation between total cross section and forward amplitude for electron scattering (Phys. Rev. 40, 40)

1939

— Bohr, Peierls and Placzek generalize it to all reaction channels in a nuclear-dispersion manuscript famously never published, yet cited for decades

1957

— van de Hulst's "Light Scattering by Small Particles" canonizes the extinction theorem in optics; the "optical theorem" name settles into scattering theory

2019

— TOTEM at the LHC measures ρ = Re f(0)/Im f(0) ≈ 0.09–0.10 at 13 TeV via Coulomb-nuclear interference — precision forward physics built on Eq. (1)

Going further

References. Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §6.3 · Griffiths, Introduction to Quantum Mechanics, 3rd ed., Ch. 11 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 19 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. II, Ch. VIII · van de Hulst, Light Scattering by Small Particles (1957) · R. G. Newton, "Optical theorem and beyond," Am. J. Phys. 44, 639 (1976)

Provenance. Verifier should recompute: (1) the flux-integral proof, confirming the stationary-phase cross term gives $-(4\pi/k)\mathrm{Im}\,f(0)$ (Sakurai §6.3, Cohen-Tannoudji Vol. II Ch. VIII); (2) the partial-wave identity that both $\mathrm{Im}\,f(0)$ and $\sigma_{\mathrm{tot}}$ equal $\frac{4\pi}{k^2}\sum(2\ell+1)\sin^2\delta_\ell$ up to the $4\pi/k$ factor, using $P_\ell(1)=1$ and Legendre orthogonality; (3) the unitarity-limit value $\frac{4\pi}{k^2}(2\ell+1)$ at $\delta_\ell=\pi/2$ ; (4) the second-Born relation $\mathrm{Im}\,f^{(2)}(0)=\frac{k}{4\pi}\int|f^{(1)}|^2\mathrm{d}\Omega$ . All formulas are standard (Sakurai, Griffiths Ch. 11, Shankar Ch. 19). (Lane note: items (1)–(3) are executed in this article's verification script, item (4) symbolically at order λ²; the systematic second-Born derivation is chapter property of Born Series and Higher-Order Corrections. Two base display defects repaired, machine-proven: the "generalized optical theorem" line carried a spurious factor 2, and the forward-bound display shipped with a literal "?" — both corrected in Derivation B and Derivation C respectively.)

Builds on: Scattering Amplitude · Asymptotic Form of the Scattering Wave Function · Differential and Total Cross Section · Partial-Wave Expansion

Leads to: The S-Matrix and Unitarity in Scattering · Phase Shifts · Resonances and the Breit-Wigner Formula · Born Series and Higher-Order Corrections

If the interference term only survives at θ = 0, why does anything else matter to the theorem?

Because the value the forward direction retains is an integral over everything else. Stationary phase pins the surviving interference to $f(0)$ , but conservation equates it to $\int|f|^2\mathrm{d}\Omega$ plus all absorption — that is the theorem. The forward direction is not where the physics happens; it is where the accounting is filed. Which is why one measured point (plus $\rho$ ) fixes $\sigma_{\mathrm{tot}}$ at a collider.

Is Im f(0) > 0 guaranteed?

$\mathrm{Im}f(0)=k\sigma_{\mathrm{tot}}/4\pi\ge0$ always, with equality only for no scattering. That fixed sign and size powers dispersion relations: the energy dependence of $\mathrm{Im}f(0)\propto\sigma_{\mathrm{tot}}(E)$ constrains $\mathrm{Re}f(0,E)$ at each energy — the LHC's measured $\rho$ parameter is exactly such a dispersion-relation output, which is why its 2019 value carried theoretical weight.

Does the theorem hold at second order in Born, or only in the exact theory?

Order by order, interlocking: $\mathrm{Im}f^{(2)}(0)=(k/4\pi)\int|f^{(1)}|^2\mathrm{d}\Omega$ , then $\mathrm{Im}f^{(3)}$ pairs with the $f^{(1)*}f^{(2)}$ cross terms, and so on — each order's imaginary part pays the previous orders' bill. The exact amplitude satisfies it identically; truncate and you owe the highest order's debt, which is why our square-well deficit closes as $V^3$ .

The Coulomb potential scatters plenty — why does the theorem break there?

Because $\sigma_{\mathrm{tot}}$ diverges (the $1/\sin^4(\theta/2)$ forward divergence integrates to infinity) and the asymptotic state never becomes a clean plane wave — logarithmic phase distortions persist at all distances, as Asymptotic Form of the Scattering Wave Function details. Both sides of Eq. (1) are infinite; the finite statement lives one article over: screen the potential (Scattering by a Screened Coulomb Potential) and the theorem holds with the screening length setting the forward scale.

How can the LHC measure a total cross section it cannot see? Detectors cover almost no forward angle at all.

That is precisely the theorem's job. Elastic protons at milliradian angles are tracked in Roman-pot detectors far downstream; the measured $\mathrm{d}\sigma/\mathrm{d}t$ is extrapolated to $t=0$ ; Coulomb–nuclear interference at ultra-small $|t|$ fixes $\rho$ ; and Eq. (7) converts the optical point into $\sigma_{\mathrm{tot}}$ . The luminosity-independent variant compares elastic and inelastic rates so that even the beam normalization drops out. Every quoted "total pp cross section" above ISR energies is an optical-theorem number.

Is there a version for identical particles or for classical waves?

Yes to both. For identical particles the theorem applies to the properly (anti)symmetrized $f(0)$ — the bookkeeping shifts but conservation does not; Identical-Particle Scattering handles it. And the identity predates quantum mechanics: any linear wave with a conserved quadratic flux — light, sound, water waves around a piling — obeys an extinction theorem of identical structure, which is where the "optical" name came from and why antenna engineers use Eq. (1) with $f$ a classical field amplitude.

Optical Theorem

Where this lives today

The physical picture

The theorem and its three faces

What first order owes, and how the ledger collects

The forward floor: what the theorem forces on measurement

Limiting cases

Worked examples

Going further

Before this

Where this leads