Compton scattering

Compton scattering, $\gamma + e^- \to \gamma + e^-$ , is the process where quantum electrodynamics stops being a formalism and starts being a laboratory. A photon strikes a free electron, recoils with a longer wavelength, and the electron carries off the difference — a result that only makes sense if light delivers its energy and momentum in particle-like quanta. Arthur Compton measured the wavelength shift in 1923 and won a Nobel Prize for it, because the effect could not be reconciled with any wave-only picture: a classical wave should re-radiate at the frequency it drives, with no shift at all. There are two calculations hiding inside this one process, and this article does both. The first is pure relativistic kinematics: treat the collision as two billiard balls, one massless, conserve four-momentum, and the wavelength shift $\lambda' - \lambda = (h/m_e c)(1-\cos\theta)$ drops out with no field theory whatsoever — the Compton wavelength $h/m_e c = 2.426\ \mathrm{pm}$ appearing as a pure kinematic constant. The second calculation is dynamical — how likely is each scattering angle? — and this is where QED earns its keep. From the Dirac spinors for the electron and the two leading-order Feynman diagrams one computes the Klein–Nishina cross section, the first fully relativistic scattering rate ever derived from a quantum field theory. This article recalls the Dirac machinery by name from its prerequisites, sets up the amplitude, quotes the spin-averaged square (whose trace evaluation is owned by the trace-techniques article that follows), and reads off the physics.

Key takeaways

Compton scattering is the QED process $\gamma + e^- \to \gamma + e^-$ : a photon scatters off a free electron and emerges softer and redder.
Pure four-momentum conservation gives the Compton shift $\lambda' - \lambda = \dfrac{h}{m_e c}(1-\cos\theta)$ , independent of any field theory; $h/m_e c = 2.426\ \mathrm{pm}$ .
The QED amplitude is the sum of two tree diagrams — an $s$ -channel and a $u$ -channel graph sharing a virtual electron propagator; neither is gauge-invariant alone, but their sum is.
The spin-averaged rate is the Klein–Nishina formula, $\dfrac{d\sigma}{d\Omega}=\dfrac{\alpha^2}{2m^2}\Big(\dfrac{\omega'}{\omega}\Big)^{2}\Big[\dfrac{\omega}{\omega'}+\dfrac{\omega'}{\omega}-\sin^2\theta\Big]$ .
In the low-energy (Thomson) limit it reduces to $\dfrac{r_e^2}{2}(1+\cos^2\theta)$ with total cross section $\sigma_T = \dfrac{8\pi\alpha^2}{3m^2}=0.665\ \mathrm{barn}$ .
Key idea: Compton scattering is two calculations in one — relativistic kinematics fixes how much the photon reddens, and QED (two Feynman diagrams built from Dirac spinors) fixes how often it scatters into each angle, giving the Klein–Nishina cross section that reduces to classical Thomson scattering at low energy.

The core insight

Compton scattering is two calculations in one: relativistic four-momentum conservation fixes the wavelength shift $\lambda'-\lambda=(h/m_e c)(1-\cos\theta)$ with no field theory, while QED — two tree diagrams built from Dirac spinors — fixes the angular rate, giving the Klein–Nishina cross section that collapses to classical Thomson scattering when the photon is soft.

Where this lives today

Compton scattering is a daily working tool across physics and medicine. In gamma-ray astronomy, Compton telescopes (COMPTEL on NASA's Compton Gamma Ray Observatory, and the modern COSI mission) reconstruct the direction of MeV photons from the recoil-electron energy and scattered-photon angle, using exactly equation (2). In medical imaging and radiotherapy, Compton scattering dominates photon interactions in soft tissue between roughly $100\ \mathrm{keV}$ and $10\ \mathrm{MeV}$ , so the Klein–Nishina cross section sets radiation dose. Inverse Compton scattering — a relativistic electron up-scattering a soft photon — powers the X-ray glow of astrophysical jets and black-hole coronae, and drives laser-electron sources of tunable gamma beams on Earth. And in fundamental physics, high-precision Compton polarimetry monitors electron-beam polarization at facilities like Jefferson Lab, because the spin-dependent cross section is calculable to sub-percent accuracy from QED — the same theory Klein and Nishina wrote down in 1929.

The physical picture

A photon of energy $\omega$ meets an electron sitting at rest. It cannot simply be absorbed — a free electron cannot swallow a photon and conserve both energy and momentum — so it must scatter. The electron recoils, and because it carries off some energy, the outgoing photon is softer: lower energy, lower frequency, longer wavelength. The harder the deflection (the larger the angle $\theta$ ), the more recoil the electron takes and the redder the photon returns. At $180^\circ$ backscatter the photon loses the most. That single sentence — "scattered light is redshifted by an amount that grows with angle" — is the entire Compton effect, and it is invisible to classical wave theory, which predicts no shift. The first figure shows why QED needs two diagrams: the virtual electron between the two photon vertices can be in either time order.

Figure

The two leading-order QED diagrams for γ + e⁻ → γ + e⁻. Left, the s-channel: the electron absorbs the incoming photon k, propagates as a virtual electron of momentum p+k, then emits k′. Right, the u-channel: the electron emits k′ first, propagates as p−k′, then absorbs k. The propagator denominators are s − m² = 2p·k and u − m² = −2p·k′. Only the sum of both graphs is gauge invariant.

The kinematics fixes the redshift; the dynamics fixes the rate. As the photon energy climbs, the recoiling electron increasingly steals forward momentum, and the scattering swings from the gentle fore-aft symmetry of classical Thomson scattering toward a sharp forward spike. The second figure is the quantity a detector physicist actually reaches for: the Klein–Nishina angular distribution, plotted in units of the classical electron radius squared, at three energies. At low energy it is the symmetric $1+\cos^2\theta$ dumbbell whose integral is exactly the Thomson cross section $\sigma_T = 0.665\ \mathrm{barn}$ ; at high energy almost everything scatters forward.

Figure

Klein–Nishina differential cross section dσ/dΩ in units of the classical electron radius squared r_e², versus scattering angle θ. The Thomson curve (green, photon energy → 0) is the symmetric 1+cos²θ shape whose integral is σ_T = 0.665 barn. At 511 keV (blue) and 10 MeV (amber) the distribution tips strongly forward as the recoiling electron carries off forward momentum.

The core equations

The kinematic heart of the effect is a single line. Model the collision as a photon (four-momentum $k$ ) striking an electron at rest (four-momentum $p=(m,\mathbf 0)$ ), scattering to $k'$ while the electron recoils to $p'=p+k-k'$ . Squaring $p'$ and using $k^2=k'^2=0$ , $p^2=m^2$ gives the shift directly.

\lambda' - \lambda = \frac{h}{m_e c}\,(1-\cos\theta) \qquad (1)

where: $\lambda,\lambda'$ — incoming and outgoing photon wavelengths; $\theta$ — photon scattering angle; $h/m_e c \equiv \lambda_C = 2.426\ \mathrm{pm}$ is the electron's Compton wavelength. The shift depends only on the angle, never on the wavelength — a scale-free signature that no classical wave picture can reproduce. Reading it: forward scatter ( $\theta=0$ ) gives no shift; backscatter ( $\theta=\pi$ ) gives the maximum, $2\lambda_C$ . The same relation written in photon energies (with $\hbar=c=1$ , so energy $\omega=$ frequency) is more convenient for the QED calculation, because it isolates the two invariants the amplitude depends on, $p\cdot k = m\omega$ and $p\cdot k' = m\omega'$ .

\frac{1}{\omega'}-\frac{1}{\omega}=\frac{1}{m}\,(1-\cos\theta) \quad\Longleftrightarrow\quad \omega'=\frac{\omega}{1+\dfrac{\omega}{m}\,(1-\cos\theta)} \qquad (2)

where: $\omega,\omega'$ — incoming and outgoing photon energies; $m=m_e$ . When $\omega\ll m$ the denominator is $\approx 1$ and $\omega'\approx\omega$ (no recoil, the classical limit); when $\omega\gg m$ the photon dumps almost all its energy into the electron at large angles. The dynamics comes from the two Feynman diagrams. Applying the QED Feynman rules — a vertex $-ie\gamma^\mu$ , external polarizations $\varepsilon_\mu(k)$ and $\varepsilon'^{*}_\nu(k')$ , spinors $u(p)$ and $\bar u(p')$ , and the electron propagator $i(\slashed q+m)/(q^2-m^2)$ — gives the invariant amplitude as a sum over the two time orderings.

i\mathcal M = -ie^2\,\bar u(p')\!\left[\frac{\slashed\varepsilon'^{*}(\slashed p+\slashed k+m)\slashed\varepsilon}{2\,p\cdot k}-\frac{\slashed\varepsilon(\slashed p-\slashed k'+m)\slashed\varepsilon'^{*}}{2\,p\cdot k'}\right]u(p) \qquad (3)

where: $\slashed a\equiv\gamma^\mu a_\mu$ ; the first term is the $s$ -channel graph (denominator $s-m^2=2p\cdot k$ ), the second the $u$ -channel graph (denominator $u-m^2=-2p\cdot k'$ ). Reading it: the two diagrams differ only by swapping the roles of the two photons, $(\varepsilon,k)\leftrightarrow(\varepsilon'^{*},-k')$ — a crossing symmetry that will also relate Compton scattering to pair annihilation. Squaring the amplitude, summing over final spins and averaging over initial spins turns the bispinor sandwich into traces of gamma matrices. That trace evaluation is the subject of the article this one leads to; the result is compact and famous.

\frac14\sum_{\text{spins}}|\mathcal M|^2 = 2e^4\!\left[\frac{p\cdot k'}{p\cdot k}+\frac{p\cdot k}{p\cdot k'}+2m^2\!\left(\frac{1}{p\cdot k}-\frac{1}{p\cdot k'}\right)+m^4\!\left(\frac{1}{p\cdot k}-\frac{1}{p\cdot k'}\right)^{\!2}\right] \qquad (4)

where: the factor $\tfrac14$ averages over the two initial electron spins and two photon polarizations. Every appearance of $p\cdot k=m\omega$ and $p\cdot k'=m\omega'$ is fixed by kinematics, so this expression is a pure function of $\omega,\omega',\theta$ once frame is chosen. Folding equation (4) into the two-body phase space in the lab frame (electron initially at rest) gives the Klein–Nishina differential cross section — the observable rate.

\frac{d\sigma}{d\Omega}=\frac{\alpha^2}{2m^2}\left(\frac{\omega'}{\omega}\right)^{2}\!\left[\frac{\omega}{\omega'}+\frac{\omega'}{\omega}-\sin^2\theta\right] \qquad (5)

where: $\alpha=e^2/4\pi\approx1/137$ ; $\omega'/\omega$ is fixed by equation (2); $\alpha^2/2m^2=\tfrac12 r_e^2$ with $r_e=\alpha/m$ the classical electron radius. Reading it: the $(\omega'/\omega)^2$ prefactor is the phase-space and flux effect that suppresses back-scattering at high energy; the bracket is the polarization-summed dynamics. Integrating equation (5) at low energy ( $\omega'\to\omega$ ) collapses it to the classical Thomson result — energy-independent, and the benchmark every X-ray and $\gamma$ -ray cross section is measured against.

\sigma_T=\frac{8\pi}{3}\,\frac{\alpha^2}{m^2}=\frac{8\pi}{3}r_e^2 = 6.65\times10^{-29}\ \mathrm{m^2}=0.665\ \mathrm{barn} \qquad (6)

where: $r_e=2.818\times10^{-15}\ \mathrm{m}$ . This is the effective cross-sectional "size" a free electron presents to a soft photon; the Klein–Nishina formula shows this size shrinks as the photon hardens.

Derivation AThe Compton wavelength shift from four-momentum conservation

The goal is to obtain equation (1) — the redshift of a scattered photon — using nothing but relativistic kinematics, no field theory at all, so that we cleanly separate what the Compton effect proves (light carries particle-like momentum) from what QED later adds (the rate). We conserve total four-momentum, exploit the masslessness of the photon and the on-shell condition of the electron, and eliminate the unmeasured recoil electron entirely by squaring. The only physical input is that a free electron cannot absorb a photon, so it must scatter it.

Step 1. Write four-momentum conservation as $p+k=p'+k'$ and isolate the electron recoil, $p'=p+k-k'$ . Square both sides: $p'^2=p^2+k^2+k'^2+2p\cdot k-2p\cdot k'-2k\cdot k'$ . Why: squaring is the standard trick to eliminate the recoil electron, whose momentum we never measure; the invariant $p'^2$ is fixed by the mass shell regardless of direction.

Step 2. Impose the on-shell and massless conditions $p'^2=p^2=m^2$ and $k^2=k'^2=0$ . Everything collapses to $0=2p\cdot k-2p\cdot k'-2k\cdot k'$ , i.e. $p\cdot k-p\cdot k'=k\cdot k'$ . Why: this single invariant equation already contains the whole shift — it says the energy the electron gains equals a term built purely from the two photon momenta.

Step 3. Evaluate the dot products in the lab frame with the electron at rest, $p=(m,\mathbf 0)$ : then $p\cdot k=m\omega$ , $p\cdot k'=m\omega'$ , and $k\cdot k'=\omega\omega'(1-\cos\theta)$ , where $\theta$ is the angle between the incoming and outgoing photon. Substituting gives $m(\omega-\omega')=\omega\omega'(1-\cos\theta)$ . Why: choosing the rest frame turns abstract invariants into the actual measured photon energies and the lab scattering angle.

Step 4. Divide through by $m\omega\omega'$ to get $1/\omega'-1/\omega=(1-\cos\theta)/m$ , then restore $\hbar,c$ and convert energy to wavelength via $\omega=hc/\lambda$ to obtain $\lambda'-\lambda=(h/m_e c)(1-\cos\theta)$ . Why: the reciprocal-energy form is what the QED calculation wants, while the wavelength form is what Compton's spectrometer measured — same physics, two languages. *Verified: for $\theta=90^\circ$ , $\Delta\lambda=\lambda_C=2.426\ \mathrm{pm}$ ; a $511\ \mathrm{keV}$ photon at $90^\circ$ leaves at $255.5\ \mathrm{keV}$ ; backscatter of $511\ \mathrm{keV}$ gives the Compton edge at $340.7\ \mathrm{keV}$ .*

Go deeper · Derivation B · 2 minAssembling the QED amplitude and its gauge invariance

Derivation BAssembling the QED amplitude and its gauge invariance

The aim is to build the invariant amplitude of equation (3) from the QED Feynman rules and then demonstrate the non-negotiable consistency check every QED amplitude must pass: gauge invariance, expressed as the Ward identity. This matters because neither diagram is separately gauge invariant — replace a photon polarization by its momentum and each graph gives a nonzero answer — yet the physical sum must vanish, and seeing that cancellation is the surest way to know the two diagrams belong together and that no third diagram is missing.

Step 1. Assign the external factors: incoming electron $u(p)$ , outgoing $\bar u(p')$ , incoming photon polarization $\varepsilon_\mu(k)$ , outgoing $\varepsilon'^{*}_\nu(k')$ , each vertex $-ie\gamma^\mu$ , and the internal electron propagator $i(\slashed q+m)/(q^2-m^2)$ . Why: these are read straight off the QED Feynman rules established in the prerequisite article; Compton scattering is the simplest process that uses the fermion propagator on an internal line.

Step 2. For the $s$ -channel graph the virtual electron carries $q=p+k$ , so $q^2-m^2=(p+k)^2-m^2=2p\cdot k$ ; for the $u$ -channel graph $q=p-k'$ , so $q^2-m^2=(p-k')^2-m^2=-2p\cdot k'$ . Assembling both with a relative sign fixed by the shared external electron line gives equation (3). Why: the two graphs are the only two ways to order the two photon vertices along the electron line, and both are the same order in $e$ , so both must be kept.

Step 3. Write $\mathcal M=\varepsilon_\mu\varepsilon'^{*}_\nu\mathcal M^{\mu\nu}$ and test the Ward identity by the replacement $\varepsilon_\mu\to k_\mu$ . The $s$ -channel piece produces $\slashed k(\slashed p+\slashed k+m)/(2p\cdot k)$ ; using $\slashed k\,\slashed k=k^2=0$ and the Dirac equation $\slashed p\,u(p)=m\,u(p)$ , the numerator simplifies to $2p\cdot k$ times a vertex, cancelling the denominator. Why: the on-shell conditions are exactly what make the longitudinal piece of one diagram a pure vertex term.

Step 4. The same replacement in the $u$ -channel produces an equal and opposite vertex term, so $k_\mu\mathcal M^{\mu\nu}=0$ : the sum is gauge invariant even though each graph is not. Why: this is the diagnostic that the amplitude is complete and correct — a QED result that fails the Ward identity has a dropped diagram or a sign error, so this check is done before any trace algebra. *Verified: each diagram's denominator equals the Mandelstam combination $s-m^2=2p\cdot k$ and $u-m^2=-2p\cdot k'$ ; the longitudinal replacement cancels between diagrams (Peskin & Schroeder §5.5).*

Go deeper · Derivation C · 1 minThe Thomson limit and the total cross section

Derivation CThe Thomson limit and the total cross section

The goal is to show that the Klein–Nishina formula (5) reduces, when the photon is soft ( $\omega\ll m$ ), to the classical Thomson differential cross section $\tfrac12 r_e^2(1+\cos^2\theta)$ , and to integrate it to the total $\sigma_T=8\pi\alpha^2/3m^2$ . This limit is the bridge between QED and classical electromagnetism — Thomson scattering is just an oscillating charge re-radiating a driving wave — so recovering it is both a sanity check on the relativistic calculation and the origin of the classical electron radius as a cross-section scale.

Step 1. In the soft limit equation (2) gives $\omega'/\omega\to1$ , because the recoil term $(\omega/m)(1-\cos\theta)\to0$ . So the prefactor $(\omega'/\omega)^2\to1$ and the bracket $\omega/\omega'+\omega'/\omega-\sin^2\theta\to 2-\sin^2\theta=1+\cos^2\theta$ . Why: at low energy the electron barely recoils, the photon keeps its energy, and only the polarization dynamics — the $1+\cos^2\theta$ shape of dipole re-radiation — survives.

Step 2. Equation (5) therefore becomes $d\sigma/d\Omega=\tfrac{\alpha^2}{2m^2}(1+\cos^2\theta)=\tfrac12 r_e^2(1+\cos^2\theta)$ , the Thomson differential cross section, symmetric under $\theta\to\pi-\theta$ . Why: the fore-aft symmetry is the fingerprint of the classical limit; the Klein–Nishina forward-peaking is a purely relativistic, recoil-driven effect that switches off here.

Step 3. Integrate over solid angle: $\sigma_T=\int\tfrac12 r_e^2(1+\cos^2\theta)\,d\Omega=\pi r_e^2\int_{-1}^{1}(1+x^2)\,dx=\pi r_e^2\cdot\tfrac83=\tfrac{8\pi}{3}r_e^2$ . Why: the $\int(1+\cos^2\theta)\,d(\cos\theta)$ integral gives the characteristic factor $8/3$ that turns the microscopic length $r_e$ into the famous $0.665\ \mathrm{barn}$ .

Step 4. Numerically $r_e=2.818\times10^{-15}\ \mathrm m$ gives $\sigma_T=6.65\times10^{-29}\ \mathrm m^2=0.665\ \mathrm{barn}$ , and at high energy the full Klein–Nishina integral falls as $\sigma\approx(2\pi\alpha^2/m\omega)[\ln(2\omega/m)+\tfrac12]$ . Why: the electron looks its full classical size only to soft photons; hard photons see a shrinking target, which is why the Compton cross section drops with energy — the physics behind $\gamma$ -ray penetration. *Verified: $\sigma_T=8\pi r_e^2/3=0.665\ \mathrm{barn}$ ; the $1+\cos^2\theta$ shape integrates to the factor $8/3$ ; high-energy fall-off $\sim\ln\omega/\omega$ .*

Limiting cases

Case	Result	Meaning
Soft photon, $\omega\ll m$	$\dfrac{d\sigma}{d\Omega}\to\tfrac12 r_e^2(1+\cos^2\theta)$ , $\sigma\to\sigma_T$	classical Thomson scattering; no recoil, fore-aft symmetric, energy-independent
Forward scatter, $\theta\to0$	$\omega'\to\omega$ , $\dfrac{d\sigma}{d\Omega}\to r_e^2$	no wavelength shift; all three energy curves coincide
Back-scatter, $\theta=\pi$	$\lambda'-\lambda=2\lambda_C$ ; $\omega'=\dfrac{\omega}{1+2\omega/m}$	maximum energy loss; the Compton edge in a detector spectrum
Hard photon, $\omega\gg m$	$\sigma\approx\dfrac{2\pi\alpha^2}{m\omega}\big[\ln\tfrac{2\omega}{m}+\tfrac12\big]$	cross section falls as $\ln\omega/\omega$ ; strongly forward-peaked
$511\ \mathrm{keV}$ , $\theta=90^\circ$	$\omega'=255.5\ \mathrm{keV}$	annihilation photon loses half its energy; a standard calibration line

Common trapCompton scattering is not the same as the photoelectric effect or Thomson scattering

Three "photon-hits-electron" processes are constantly confused. In the photoelectric effect the photon is absorbed whole and the electron is ejected — this requires a bound electron so the atom can absorb the recoil momentum, and it dominates at low photon energy. In Thomson scattering the photon scatters elastically off a free electron with no wavelength change — it is the $\omega\ll m$ classical limit of Compton, valid only when recoil is negligible. Compton scattering is the inelastic scattering off a free (or quasi-free) electron with a wavelength shift, dominating at intermediate $\gamma$ -ray energies. The tell is the wavelength: unchanged means Thomson, shifted means Compton, gone means photoelectric. And note that a genuinely free electron cannot photo-absorb at all — energy and momentum cannot both be conserved — which is exactly why the photon must scatter rather than be swallowed.

Worked examples

1 · Compton's original experiment — molybdenum X-rays off graphite (fully worked)

Compton scattered Mo K $\alpha$ X-rays ( $\lambda=71.1\ \mathrm{pm}$ , about $17.4\ \mathrm{keV}$ ) off the loosely bound electrons in graphite. Find the wavelength of the scattered line at $45^\circ$ , $90^\circ$ , and $135^\circ$ , and confirm the shift is independent of the incident wavelength.

Step 1. The shift depends only on angle through equation (1), with $\lambda_C=h/m_e c=2.426\ \mathrm{pm}$ : $\Delta\lambda=\lambda_C(1-\cos\theta)$ . At $45^\circ$ : $\Delta\lambda=2.426(1-0.707)=\mathbf{0.711\ pm}$ . At $90^\circ$ : $\Delta\lambda=2.426(1-0)=\mathbf{2.426\ pm}$ . At $135^\circ$ : $\Delta\lambda=2.426(1+0.707)=\mathbf{4.142\ pm}$ .

Step 2. Add to the incident $71.1\ \mathrm{pm}$ : the scattered lines sit at $\mathbf{71.81\ pm}$ , $\mathbf{73.53\ pm}$ , and $\mathbf{75.24\ pm}$ . Compton saw exactly this — a shifted line moving to longer wavelength with angle, alongside an unshifted line from tightly bound (whole-atom recoil) electrons.

Step 3. The decisive point: $\Delta\lambda$ contains no factor of $\lambda$ . Repeat the experiment with a different X-ray tube and the shifts in picometres are identical. A shift that is set by the electron mass alone, not by the incident wavelength, is the signature that light delivers momentum in quanta $p=h/\lambda$ — the result classical wave theory cannot produce.

2 · The Compton edge of a 511 keV annihilation photon (fully worked)

A $\gamma$ -ray detector records $511\ \mathrm{keV}$ photons from positron annihilation. Compton scattering inside the crystal deposits the recoil-electron energy. Find the maximum energy the electron can receive (the Compton edge) and the energy of the fully back-scattered photon.

Step 1. Maximum energy transfer occurs at back-scatter, $\theta=180^\circ$ , where equation (2) gives $\omega'=\dfrac{\omega}{1+2\omega/m}$ . With $\omega=511\ \mathrm{keV}$ and $m=511\ \mathrm{keV}$ : $\omega'=\dfrac{511}{1+2}=\mathbf{170.3\ keV}$ — the "back-scatter peak" seen in real spectra.

Step 2. The electron takes the rest: $T_{\max}=\omega-\omega'=511-170.3=\mathbf{340.7\ keV}$ . This is the Compton edge — a sharp upper cutoff in the deposited-energy spectrum, below the $511\ \mathrm{keV}$ full-absorption photopeak.

Step 3. Check the $90^\circ$ case: $\omega'=511/(1+1)=\mathbf{255.5\ keV}$ , so the electron gets $255.5\ \mathrm{keV}$ . The spread from $0$ up to the $340.7\ \mathrm{keV}$ edge is the continuous Compton shoulder every gamma spectroscopist learns to read. The edge sits at $340.7\ \mathrm{keV}$ and the back-scatter peak at $170.3\ \mathrm{keV}$ , separated from the photopeak by exactly the recoil kinematics.

3 · The Thomson cross section as an interaction rate

Estimate the mean free path of a soft photon travelling through a fully ionized hydrogen plasma of electron density $n_e=10^{29}\ \mathrm{m^{-3}}$ (roughly a stellar interior scale), using the Thomson cross section.

Step 1. For soft photons the relevant cross section is $\sigma_T=0.665\ \mathrm{barn}=6.65\times10^{-29}\ \mathrm{m^2}$ (equation 6), energy-independent.

Step 2. The mean free path is $\ell=1/(n_e\sigma_T)=1/(10^{29}\times6.65\times10^{-29})\approx\mathbf{0.15\ m}$ . A photon scatters every $\sim15\ \mathrm{cm}$ — which is why radiation diffuses out of stellar interiors over millennia rather than escaping at light speed. Thomson scattering off free electrons is the dominant opacity source in hot ionized matter, and $\sigma_T$ sets its scale.

Physics of the portrait

Compton scattering is where the photon stopped being a bookkeeping device and became a particle you could collide. The man who did the colliding built the definitive argument against the wave-only view of light.

Origins · 1 minArthur Compton and how this was discovered

Arthur Compton

1892–1962 · X-ray scattering and the particle nature of light

Arthur Holly Compton, at Washington University in St. Louis, performed in 1922–1923 the experiment that forced physicists to accept that light carries momentum in discrete quanta. He directed monochromatic molybdenum K $\alpha$ X-rays at a graphite target and measured, with a Bragg crystal spectrometer, the wavelength of the radiation scattered at various angles. Classical electromagnetism was unambiguous: an oscillating charge driven by an incident wave should re-radiate at the same frequency, so the scattered wavelength should not change. Compton found instead a second, longer-wavelength line whose shift grew with angle exactly as $\Delta\lambda=(h/m_e c)(1-\cos\theta)$ — the formula he derived by treating the process as an elastic collision between a photon of momentum $h/\lambda$ and a free electron. Peter Debye published the same analysis independently and almost simultaneously, but Compton had the data. The measurement was decisive because the shift depends only on the electron mass, not on the incident wavelength or the target material — precisely what a particle-collision picture predicts and a wave picture forbids. It earned Compton the 1927 Nobel Prize and cemented wave–particle duality. Six years later Oskar Klein and Yoshio Nishina used Dirac's brand-new relativistic electron theory to compute the full angular cross section — one of the earliest triumphs of relativistic quantum mechanics, and a prediction still among the most-checked in all of physics.

History

1923

— Compton measures the angle-dependent wavelength shift of X-rays scattered by graphite

1923

— Debye independently derives the same photon-collision shift formula

1928

— Dirac's relativistic wave equation supplies the spin-½ electron and its bispinor solutions

1929

— Klein and Nishina compute the full relativistic cross section from Dirac theory

1949

— Feynman's diagrammatic QED recasts Compton scattering as the sum of two tree graphs

Going further

Texts. Peskin & Schroeder, An Introduction to Quantum Field Theory, §5.5 (Compton scattering, the Klein–Nishina formula, and the Ward-identity check) · Griffiths, Introduction to Elementary Particles, 2nd ed., §7 and §8.3 · Halzen & Martin, Quarks and Leptons, §6 · Thomson, Modern Particle Physics, §11 · Schwartz, Quantum Field Theory and the Standard Model, §13.

Provenance. The Dirac equation, its gamma-matrix algebra $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}$ , and the plane-wave bispinors $u(p),v(p)$ used as external legs are recalled by name from the prerequisite articles aq-pp-dirac-equation, aq-pp-gamma-matrices-clifford-algebra, and aq-pp-dirac-solutions; the vertex $-ie\gamma^\mu$ and the propagator come from aq-pp-feynman-rules-qed. The spin sum in equation (4) and its reduction to the Klein–Nishina result via trace theorems are owned by the articles this one leads to, aq-pp-matrix-element-spin-averaging and aq-pp-trace-techniques (forward references, quoted not re-derived here). A verifier should confirm: (1) $\lambda'-\lambda=(h/m_ec)(1-\cos\theta)$ with $\lambda_C=2.426\ \mathrm{pm}$ ; (2) $\omega'=\omega/[1+(\omega/m)(1-\cos\theta)]$ giving $255.5\ \mathrm{keV}$ at $90^\circ$ and $170.3\ \mathrm{keV}$ at $180^\circ$ for a $511\ \mathrm{keV}$ photon; (3) the Ward identity $k_\mu\mathcal M^{\mu\nu}=0$ ; (4) $\sigma_T=8\pi\alpha^2/3m^2=0.665\ \mathrm{barn}$ .

Prerequisites. Dirac Equation · Klein–Gordon Equation · Gamma Matrices and the Clifford Algebra · Dirac Plane-Wave Solutions · QED Feynman Rules. Leads to. Matrix Element and Spin Averaging · Trace Techniques · Introduction to QED.

Ask the tutor: Ask me to walk through the gamma-matrix trace that turns the two-diagram amplitude into the Klein–Nishina formula, or to derive the inverse-Compton blueshift that lights up astrophysical jets.

Why does the scattered X-ray get a longer wavelength, not a shorter one?

Because the photon can only lose energy to the electron, never gain it, when the electron starts at rest. Conservation of energy and momentum in the collision forces the recoiling electron to carry away kinetic energy, and that energy comes out of the photon. Less photon energy means lower frequency and, since $\lambda=hc/E$ , longer wavelength. The loss grows with scattering angle because a larger deflection requires transferring more momentum to the electron. Only if the electron were already moving toward the photon (inverse Compton scattering) could the photon come out bluer — and that is exactly how astrophysical jets turn starlight into X-rays.

What is the difference between Compton scattering and the photoelectric effect?

Both start with a photon and an electron, but the outcomes differ. In the photoelectric effect the photon is completely absorbed and a bound electron is ejected; this needs an atom to absorb the leftover momentum, so it cannot happen with a truly free electron, and it dominates at low photon energies. In Compton scattering the photon survives — it scatters off a free or loosely bound electron and comes away with less energy and a longer wavelength. Experimentally you tell them apart by looking for the scattered photon: in Compton it is still there (reddened); in the photoelectric effect it is gone. Compton dominates the middle range of gamma energies, between photoelectric absorption at low energy and pair production above $1.022\ \mathrm{MeV}$ .

Why can't a free electron simply absorb the photon?

Try to conserve energy and momentum at once. In the electron's rest frame the photon brings momentum $\omega$ and energy $\omega$ . If absorbed, the electron would have momentum $\omega$ and hence energy $\sqrt{m^2+\omega^2}$ , but energy conservation demands $m+\omega$ . These match only if $\sqrt{m^2+\omega^2}=m+\omega$ , i.e. $\omega=0$ — no photon at all. So a free electron cannot absorb a real photon; it can only scatter it, which is why Compton scattering exists and why the photoelectric effect needs a third body (the nucleus) to soak up the momentum mismatch.

Why does the QED calculation need two diagrams instead of one?

The electron interacts with a photon at a vertex, and Compton scattering has two photons — one absorbed, one emitted. There are two distinct time orderings: the electron can absorb the incoming photon first and emit the outgoing one later (the $s$ -channel diagram, with a virtual electron of momentum $p+k$ ), or it can emit the outgoing photon first and absorb the incoming one later (the $u$ -channel diagram, virtual momentum $p-k'$ ). Both are the same order in the coupling $e$ , so quantum mechanics says you must add their amplitudes. Crucially, neither diagram is gauge invariant by itself — the check that they belong together is the Ward identity, which only the sum satisfies. A single-diagram "answer" would violate charge conservation.

What is the Compton wavelength, physically?

The Compton wavelength $\lambda_C=h/m_ec=2.426\ \mathrm{pm}$ is the natural length set by the electron's mass, and here it is the size of the maximum wavelength shift. It marks the scale at which quantum field theory becomes unavoidable: to localize an electron within its Compton wavelength you would need photons energetic enough to create electron–positron pairs, so the single-particle picture breaks down. A neat coincidence makes it tangible — a photon whose energy equals the electron rest energy, $511\ \mathrm{keV}$ , has wavelength exactly $\lambda_C$ , which is why annihilation photons and the Compton shift are numerically intertwined.

Why does the Compton cross section fall off at high energy?

At low energy a photon sees the full Thomson cross section $\sigma_T=0.665\ \mathrm{barn}$ , independent of energy. As the photon energy climbs past the electron rest mass, two effects suppress the rate: recoil steals phase space (the $(\omega'/\omega)^2$ factor), and the scattering funnels sharply forward, so the total falls roughly as $\ln(\omega)/\omega$ . This is why high-energy gamma rays penetrate matter far more easily than the electron's "size" suggests — a hard photon effectively sees a smaller electron — and why you need much more lead to stop MeV gammas than keV X-rays.