Scattering Theory (an introduction)

Almost everything we know about the subatomic world we learned by throwing particles at things and watching how they bounce. Rutherford fired alpha particles at a sheet of gold foil and, from the surprising few that ricocheted straight back, deduced that the atom hides a tiny dense nucleus. Neutron beams diffract off crystals to reveal where the atoms sit; electrons scattered off protons showed the proton is not a point but a fuzzy cloud of quarks; and every collider from SLAC to the LHC is, at heart, the same experiment scaled up — smash things together, catch the debris, and read the target's structure off the pattern of where the pieces fly.

Scattering theory is the language that turns "how often does it bounce, and which way?" into "what does the target — the force V(r) — actually look like?" It is the bridge between a number you can measure in a detector and the potential you cannot see. This page builds that bridge in four planks: the scattering amplitude that carries all the physics, the cross-section that a detector actually counts, the Born approximation that ties the amplitude to the shape of the force, and partial waves with their phase shifts, the natural language when the force is central.

The setup: a plane wave meets a potential

Idealise a beam as a plane wave of definite momentum p = \hbar k travelling along the z-axis, e^{ikz}. It flies in from far away, hits a localised potential V(r) (one that dies off outside some small region — a nucleus, an atom, a short-range force), and scatters. Far downstream, the wavefunction settles into a clean and universal shape: the original plane wave, still ploughing ahead, plus an outgoing spherical ripple spreading from the target.

\psi(\mathbf{r}) \;\xrightarrow{\;r \to \infty\;}\; \underbrace{e^{ikz}}_{\text{incoming beam}} \;+\; \underbrace{f(\theta)\,\frac{e^{ikr}}{r}}_{\text{scattered ripple}}.

The e^{ikr}/r is just an expanding sphere of probability — its 1/r falloff is exactly what keeps the total probability crossing each shell constant (surface area grows as r^2, so |{\cdot}|^2 must fall as 1/r^2). Everything specific to this target — how much scatters, and in which directions — is bundled into the single function out front, f(\theta), the scattering amplitude. It has units of length, it generally depends on the scattering angle \theta (and energy), and it is the one object the whole theory is built to compute. Know f(\theta) and you know the experiment.

Cross-sections: what a detector actually counts

A detector does not see amplitudes; it sees counts. Park a detector subtending a small solid angle d\Omega in the direction (\theta,\phi) and ask: what fraction of the incoming flux lands in it? That ratio, per unit solid angle, is the differential cross-section, and it is simply the squared amplitude:

\frac{d\sigma}{d\Omega} = |f(\theta)|^2 .

The name is geometric. A cross-section has units of area — since f is a length, |f|^2 is an area — and it behaves like the effective "size" of the target as the beam sees it. A big d\sigma/d\Omega in some direction means the target presents a large effective face there, deflecting many particles that way. Integrate over all directions and you get the total cross-section, the whole effective target area:

\sigma = \int |f(\theta)|^2\, d\Omega = \int_0^{2\pi}\!\!\int_0^{\pi} |f(\theta)|^2 \sin\theta\, d\theta\, d\phi .

Nuclear cross-sections are so small that they are measured in barns, 1\,\text{barn} = 10^{-28}\,\text{m}^2 — roughly the geometric cross-section of a heavy nucleus. (The name is a physicists' joke: to a beam a uranium nucleus is, by subatomic standards, "as big as a barn.") A large cross-section means a fat, easy-to-hit target; a small one means the beam mostly sails through untouched.

The classic slip. Because d\sigma/d\Omega = |f(\theta)|^2 looks exactly like the Born-rule "amplitude squared = probability" you have met everywhere else in quantum mechanics, it is tempting to read the cross-section as a probability, a pure number between 0 and 1. It is not. f(\theta) has units of length, so |f|^2 is an area (m², or barns), and the total \sigma can be any positive size at all.

The way to keep it straight: a cross-section answers "how big a target does the interaction present?", not "how likely is one event?". You only get an actual count rate by multiplying the cross-section by the beam's luminosity (particles per area per second). Amplitude squared, yes — but the amplitude here is a length, so its square is an area.

The Born approximation: the amplitude is the Fourier transform of the force

So how do we get f(\theta) from a given potential? Exactly is hard; but if the potential is weak (or the energy high), first-order perturbation theory gives a beautifully clean answer — the Born approximation. Treat the scattered wave as small and keep only the leading term, and the amplitude becomes an integral over the potential:

Read that slowly, because it is the punchline of the whole subject: measuring the angular distribution of scattered particles measures the Fourier transform of the force. Different angles \theta sample different spatial frequencies q of V(\mathbf{r}); sweep the detector around and you build up the full transform, then invert it to reconstruct the potential you could never touch. This is precisely how form factors and the internal structure of nuclei and nucleons were mapped — the charge distribution of the proton was read straight off the way electrons scattered from it. Scattering is a microscope whose lens is a Fourier transform.

Because it is only the first term of a series. The Born approximation is first-order perturbation theory: it assumes the scattered wave is a small correction to the incoming plane wave, so it is trustworthy only for a weak potential or high energy, where the particle barely notices the target on its way through. Turn the potential up — a deep well, a strong resonance, a slow particle that lingers — and higher-order terms (the particle scattering twice, three times, …) matter, and the neat single Fourier transform is no longer enough.

That is exactly when the partial-wave picture below takes over: it makes no weak-potential assumption and is the tool of choice at low energy and strong coupling. The two approaches are complementary — Born for weak-and-fast, partial waves for strong-and-slow.

Partial waves and phase shifts

When the potential is central — V(r) depends only on the distance r, not the direction — angular momentum is conserved, and a different decomposition is natural. Break the incoming plane wave into components of definite orbital angular momentum \ell = 0, 1, 2, \dots (the familiar s, p, d, … waves). A central force cannot change \ell, so each partial wave scatters completely independently — and, being elastic, all a partial wave can do is get shifted in phase relative to the free wave it would have been. That shift is the phase shift \delta_\ell, and it holds all the physics of the \ell-th channel.

The phase shift reads like a story. A positive \delta_\ell means the potential pulled the wave inward — the mark of an attractive force; a negative \delta_\ell means the wave was pushed out, a repulsive force. When a phase shift races rapidly up through \pi/2, its \sin^2\delta_\ell hits its maximum value of 1 and that channel scatters as hard as quantum mechanics allows — a resonance, the fingerprint of a temporary bound state (a short-lived particle forming and decaying). And crucially, at low energy the higher partial waves are frozen out (a slow particle carries little angular momentum, so it cannot reach past the target's edge for \ell > 0): only \ell = 0, the s-wave, survives, and scattering becomes beautifully simple and isotropic.

Watch the pattern form: s-wave and p-wave interference

You do not need the full infinite sum to feel how partial waves shape the angular pattern. Keep just the first two — the isotropic s-wave (a constant, since P_0 = 1) and the p-wave (P_1(\cos\theta) = \cos\theta) — so the amplitude is f(\theta) \approx f_0 + f_1\cos\theta and the differential cross-section is

\frac{d\sigma}{d\Omega} = \big|\,f_0 + f_1\cos\theta\,\big|^2 .

Drag the two weights below. With only the s-wave (f_1 = 0) the pattern is a flat circle — scattering is equally likely in every direction, the hallmark of low-energy s-wave physics. Turn up the p-wave and the constant and the \cos\theta begin to interfere: they add in the forward direction (\theta = 0, \cos\theta = 1) and fight in the backward direction (\theta = 180^\circ, \cos\theta = -1), tilting the whole distribution forward. Flip the sign of f_1 and the peak swings to the back. The angular pattern is a direct readout of which partial waves are in play and how they interfere.

This forward–backward asymmetry is not a curiosity; it is data. An experimenter who measures the angular distribution and fits it to a sum of Legendre polynomials reads off the individual phase shifts \delta_\ell — and from those, the shape and strength of the force.

Worked examples

Example 1 — isotropic (s-wave) total cross-section. At low energy only \ell = 0 survives, so f(\theta) = f_0 is a constant and d\sigma/d\Omega = |f_0|^2 is the same in every direction. The total cross-section is then that constant times the whole solid angle \int d\Omega = 4\pi:

\sigma = \int |f_0|^2\, d\Omega = 4\pi\,|f_0|^2 .

If, say, |f_0| = 2\,\text{fm}, then \sigma = 4\pi(2)^2 = 16\pi \approx 50\,\text{fm}^2 = 0.5\,\text{barn}. An isotropic amplitude of a few femtometres already gives a cross-section of order a barn.

Example 2 — momentum transfer at a given angle. A beam with wavenumber k = 3\,\text{fm}^{-1} scatters through \theta = 60^\circ. The momentum transfer is

q = 2k\sin\!\frac{\theta}{2} = 2(3)\sin 30^\circ = 6 \times \tfrac12 = 3\,\text{fm}^{-1}.

At exact back-scattering, \theta = 180^\circ, \sin(\theta/2) = 1, the transfer is maximal: q = 2k. That is why the sharpest structural detail lives in the back-scattered particles — they carry the largest q, the highest spatial frequency of the potential.

Example 3 — an s-wave cross-section from its phase shift. With only the s-wave active, the total cross-section is \sigma_0 = \dfrac{4\pi}{k^2}\sin^2\delta_0. Take k = 1\,\text{fm}^{-1} and a phase shift racing through resonance, \delta_0 = 90^\circ, so \sin^2\delta_0 = 1:

\sigma_0 = \frac{4\pi}{(1)^2}\times 1 = 4\pi \approx 12.6\,\text{fm}^2 .

This is the unitarity limit — with \sin^2\delta_0 = 1 the s-wave scatters as strongly as it possibly can. A phase shift passing through 90^\circ is the very signature of a resonance.

A gorgeous piece of bookkeeping called the optical theorem: \sigma_{\text{total}} = \frac{4\pi}{k}\,\text{Im}\,f(0). The total amount of scattering in all directions is fixed by the imaginary part of the forward amplitude alone. It sounds like magic, but the logic is pure conservation: every particle knocked out of the forward beam must go somewhere, so the beam is slightly dimmed straight ahead by exactly the amount scattered everywhere else — and that shadowing shows up as the imaginary (absorptive) part of f(0). Probability in equals probability out; the optical theorem is that ledger, written in one line.