Compton Scattering

Point a gamma-ray source — a lump of caesium-137, say — at a scintillator, and plot the energies it records. You expect a single spike at $662\ \mathrm{keV}$ , the energy of the gamma the caesium emits. You get the spike. But you also get a broad smear of counts below it, and the smear does not fade away gradually. It ends. At $\mathbf{477.7\ keV}$ there is a cliff — counts, counts, counts, then almost nothing until the smear resumes at $184.3\ \mathrm{keV}$ as a second little bump. Nothing about the detector explains those two numbers. They are not calibration artefacts, not electronics, not the crystal. They are kinematics — the exact places where a photon bouncing off a free electron is forced to stop by conservation of four-momentum, and you can compute both of them from the electron's mass and nothing else. That is the strange power of this collision. The wavelength a scattered photon comes back with obeys

\Delta\lambda = \lambda_C\,(1 - \cos\theta), \qquad \lambda_C = \frac{h}{m_ec} = 2.42631\ \mathrm{pm}

and look at what is not in that formula. Not the incoming wavelength. Not the material. Not the intensity. The shift depends on the angle and on one universal constant, and on nothing else in the universe. That rigidity is why the effect settled an argument: a wave scattering off an electron would come back with the frequency it went in with, always, at every angle. Compton's did not. And there is a second reading, which is the one special relativity is for. Ride alongside a fast electron and the same equation, unchanged, says the photon gains energy — enormously. A microwave photon from the cosmic background, bounced off an electron in an astrophysical jet, comes back as an X-ray. It is not a different process. It is this process, seen from a different frame.

Key takeaways

Signature $(+,-,-,-)$ ; $P^\mu = (E/c,\ \vec p\,)$ with $P\cdot P = +m^2c^2$ . A photon is the massless case: $P_\gamma \cdot P_\gamma = 0$ .
The Compton shift, from four-momentum conservation alone:

\boxed{\;\Delta\lambda = \lambda' - \lambda = \lambda_C\,(1-\cos\theta),\qquad \lambda_C \equiv \frac{h}{m_ec} = 2.42631\ \mathrm{pm}\;}

The derivation is one line, and the trick is the whole of SR. Isolate the unobserved electron, $P_e' = P_\gamma + P_e - P_\gamma'$ , and square it. Its norm is $m_e^2c^2$ whatever it is doing, so the recoil angle never enters the algebra.
The shift is independent of $\lambda$ : an absolute shift of a few picometres, so the fractional shift is what varies — $5\times10^{-6}$ for visible light (hopeless), $3.4\%$ for the X-rays Compton used. That is why the effect needed X-rays.
In energy form, usually more useful: $\dfrac{1}{E'} - \dfrac{1}{E} = \dfrac{1-\cos\theta}{m_ec^2}$ .
The recoil electron takes $T = E - E'$ , maximal at backscatter. For Cs-137's $662\ \mathrm{keV}$ : the Compton edge at $477.7\ \mathrm{keV}$ and the backscatter peak at $184.3\ \mathrm{keV}$ — both visible in every real gamma spectrum.
The photon energy saturates: however hard you throw it, a backscattered photon returns with at most $m_ec^2/2 = 255.5\ \mathrm{keV}$ .
Inverse Compton is the same equation in a boosted frame. Head-on against a $\gamma$ -factor electron the photon's energy is multiplied by about $4\gamma^2$ : a CMB microwave at $6.35\times10^{-4}\ \mathrm{eV}$ off a $\gamma = 1000$ electron comes back at $\mathbf{2.5\ keV}$ — an X-ray.
Key idea: squaring a four-vector you do not care about is how relativity deletes the variables you did not measure. The Compton shift is what is left when the electron's recoil angle has been algebraically thrown away.

The core insight

Nobody asked about the electron's recoil direction — and there is a way to erase it. Put the electron's final four-momentum alone on one side, $P_e' = P_\gamma + P_e - P_\gamma'$ , and take the invariant norm of both sides. The left side is $m_e^2c^2$ — constant, known, and utterly indifferent to where the electron went. The right side is built from things you did measure. One contraction, and an entire unwanted variable has vanished. This is not a trick special to Compton scattering; it is the standard move of relativistic kinematics, and it works precisely because the norm of a four-vector is the same in every frame.

The physical picture

A classical light wave of frequency $f$ falls on a free electron. The wave's field shakes the electron at frequency $f$ ; an oscillating charge radiates at the frequency it is being shaken at. So the scattered light comes back at $f$ — every angle, every material, every intensity. This is Thomson scattering, and nineteenth-century physics predicted it with total confidence. In 1923 Arthur Compton fired molybdenum $\mathrm{K}\alpha$ X-rays ( $\lambda = 71.1\ \mathrm{pm}$ ) at graphite and measured, with a Bragg spectrometer, what came out at various angles. The scattered X-rays had a longer wavelength, and the excess grew with angle. No classical mechanism can do that. The quantum picture does it immediately. Treat the light as a particle carrying energy $E = hf$ and momentum $p = E/c = h/\lambda$ . Now it is a billiard-ball collision: the photon strikes a stationary electron, the electron recoils, and the photon — which cannot slow down, because photons only travel at $c$ — must pay for that by reddening. The harder the bounce, the longer. So how much longer? The answer,

\Delta\lambda = \lambda_C(1-\cos\theta),

says the photon always stretches by the same absolute amount at a given angle — $2.43\ \mathrm{pm}$ at $90°$ — whether it arrived as a soft X-ray or a hard gamma. Two and a half picometres is a colossal fractional change for a gamma ray and a rounding error for visible light. The physics is scale-free; the visibility is not. Compton could see the effect because his X-rays had wavelengths within a factor of thirty of $\lambda_C$ . Had he used a sodium lamp, the shift would have been five parts in a million. The last move belongs to relativity rather than to quantum mechanics. Nothing says the electron has to be at rest — that was a convenience. Boost into a frame where the electron is charging at the photon and the same conserved four-momenta say the photon comes back blueshifted, hugely. Astrophysicists call this inverse Compton scattering and it lights up half the high-energy sky. It is the same collision. We are just standing somewhere else.

The core equations

The two four-momenta, with $P \cdot P = +m^2c^2$ in the locked signature. The photon is the massless case:

P_\gamma = \left(\frac{E}{c},\ \frac{E}{c}\hat{\mathbf n}\right),\quad P_\gamma\cdot P_\gamma = 0; \qquad\qquad P_e = (m_ec,\ \vec 0\,),\quad P_e \cdot P_e = m_e^2c^2 \qquad (1)

where: $\hat{\mathbf n}$ is the photon's direction and $E = hf = hc/\lambda$ . Physical reading: the photon's four-momentum has zero norm — that is what "massless" means, and it is what makes the algebra below collapse. Massless Particles owns the statement; here we use it. Conservation, and the only equation in this article:

P_\gamma + P_e = P_\gamma' + P_e' \qquad (2)

where: primes are after the collision. Physical reading: one four-vector equation is four scalar equations — energy plus three momentum components — and it is the entire physical content of the problem. The result, the Compton shift:

\boxed{\;\Delta\lambda \;=\; \lambda' - \lambda \;=\; \lambda_C\,(1 - \cos\theta), \qquad \lambda_C \equiv \frac{h}{m_ec} \;=\; 2.42631\ \mathrm{pm}\;} \qquad (3)

where: $\theta$ is the photon's scattering angle. Physical reading: look at what is absent. No $\lambda$ , no material, no intensity, no charge — only the angle and the electron's mass. $\Delta\lambda$ ranges from $0$ (forward) to $2\lambda_C = 4.85\ \mathrm{pm}$ (backscatter), and that is the entire range available to it, at any energy, forever. The same statement in energies, which is what you actually use:

\boxed{\;\frac{1}{E'} - \frac{1}{E} \;=\; \frac{1-\cos\theta}{m_ec^2}\;} \qquad\Longleftrightarrow\qquad E' = \frac{E}{1 + \dfrac{E}{m_ec^2}(1-\cos\theta)} \qquad (4)

where: $m_ec^2 = 511.0\ \mathrm{keV}$ is the scale that decides everything. Physical reading: the shift is linear in $1/E$ , not in $E$ — i.e. $\Delta\lambda$ , not $\Delta E$ , is the natural variable. Both limits are visible: for $E \ll m_ec^2$ the correction dies and $E' \to E$ (Thomson); for $E \gg m_ec^2$ the $1$ dies and $E' \to m_ec^2/(1-\cos\theta)$ — the scattered energy stops depending on the incident energy at all. And the recoil electron, which is what a detector usually sees:

T = E - E' = E\,\frac{\dfrac{E}{m_ec^2}(1-\cos\theta)}{1 + \dfrac{E}{m_ec^2}(1-\cos\theta)} \qquad\Longrightarrow\qquad \boxed{\;T_{\max} = \frac{2E^2}{m_ec^2 + 2E}\;} \qquad (5)

where: $T_{\max}$ is at $\theta = 180°$ . Physical reading: the electron cannot take all the energy. There is a hard ceiling — the Compton edge — and for Cs-137's $662\ \mathrm{keV}$ gammas it is $477.7\ \mathrm{keV}$ , which no free electron in the detector may exceed. That cliff is the most recognisable feature of any real gamma spectrum.

Deriving the shift

Derivation AThe Compton shift from four-momentum conservation, in one contraction

The goal is $\Delta\lambda = \lambda_C(1-\cos\theta)$ , exactly, with no approximation and — this is the point — without ever introducing the electron's recoil angle. That angle exists, it is perfectly well defined, and nobody measured it. A good relativistic calculation deletes it before it can cause trouble, and the tool that deletes it is the invariant norm of a four-vector.

Step 1. Write conservation and immediately isolate the thing you do not want. From $P_\gamma + P_e = P_\gamma' + P_e'$ ,

P_e' = P_\gamma + P_e - P_\gamma'

Why: the electron's final four-momentum is the only unknown we do not care about — it carries the recoil angle. So put it alone on one side, where we can hit it with something that does not care either.

Step 2. Square both sides — that is, contract each with itself. The left is free:

P_e' \cdot P_e' = m_e^2c^2

whatever the electron did. Why: this is the entire derivation. The norm of the electron's four-momentum is its mass, always, in every frame, at every recoil angle. By taking a norm we have thrown the recoil direction away without ever writing it down. Everything below is expansion.

Step 3. Expand the right-hand side and watch most of it die:

m_e^2c^2 = \underbrace{P_\gamma^2}_{0} + \underbrace{P_e^2}_{m_e^2c^2} + \underbrace{P_\gamma'^2}_{0} + 2P_\gamma\!\cdot\!P_e - 2P_\gamma\!\cdot\!P_\gamma' - 2P_e\!\cdot\!P_\gamma'

The photon norms vanish (massless), and $m_e^2c^2$ appears on both sides and cancels, leaving a single relation among dot products:

P_\gamma\cdot P_e - P_e\cdot P_\gamma' = P_\gamma\cdot P_\gamma'

Why: two of the six terms die because $P_\gamma\cdot P_\gamma = 0$ , and a third cancels against the left. The masslessness of the photon is not decoration — it is what makes this a one-line problem. And read what survives: the left is the energy the photon lost, the right is the momentum it transferred. The whole effect is that these are equal.

Step 4. Evaluate the three dot products in the lab, where the electron starts at rest, $P_e = (m_ec, \vec 0\,)$ . With the target at rest only the time components contract:

P_\gamma\cdot P_e = \frac{E}{c}\,m_ec = m_eE, \qquad P_e\cdot P_\gamma' = m_eE'

and for the two photons, whose spatial momenta have magnitudes $E/c$ and $E'/c$ and an angle $\theta$ between them,

P_\gamma\cdot P_\gamma' = \frac{EE'}{c^2} - \frac{EE'}{c^2}\cos\theta = \frac{EE'}{c^2}(1-\cos\theta)

Why: here, and only here, does $\theta$ enter — through the dot product of the two photon momenta. The cosine is not a modelling assumption; it is what a dot product of two vectors is.

Step 5. Assemble and rearrange:

m_e(E - E') = \frac{EE'}{c^2}(1-\cos\theta) \qquad\Longrightarrow\qquad \frac{E-E'}{EE'} = \frac{1-\cos\theta}{m_ec^2} \qquad\Longrightarrow\qquad \boxed{\frac{1}{E'} - \frac{1}{E} = \frac{1-\cos\theta}{m_ec^2}}

Why: dividing by $EE'$ makes the structure visible — the natural variable is $1/E$ , i.e. $\lambda$ , not $E$ .

Step 6. Convert to wavelength with $E = hc/\lambda$ , so $1/E = \lambda/hc$ :

\frac{\lambda' - \lambda}{hc} = \frac{1-\cos\theta}{m_ec^2} \qquad\Longrightarrow\qquad \boxed{\Delta\lambda = \frac{h}{m_ec}(1-\cos\theta) = \lambda_C(1-\cos\theta)}

Why: and there is the punchline. $\lambda$ has cancelled. It appeared on both sides and left. The shift cannot depend on the incident wavelength, because the algebra gave it no way to. *Verified in verify_aq-sr-compton-scattering.py: Steps 1–6 are reproduced symbolically in sympy from explicit four-momenta — squaring $P_e' = P_\gamma + P_e - P_\gamma'$ and imposing $P_e'\cdot P_e' = m_e^2c^2$ yields $1/E' - 1/E = (1-\cos\theta)/m_ec^2$ identically, with the electron's recoil angle never appearing; $\lambda_C = h/m_ec = 2.42631$ pm is computed from CODATA; and $\Delta\lambda$ is confirmed independent of $\lambda$ by evaluating it at three incident wavelengths spanning six decades.*

Go deeper · Derivation B · 2 minThe recoil electron, the Compton edge, and the cliff in every gamma spectrum

Derivation BThe recoil electron, the Compton edge, and the cliff in every gamma spectrum

A scintillator does not see the scattered photon — that usually escapes. It sees the recoil electron, which stops in the crystal and deposits its kinetic energy. So the observable spectrum is the spectrum of $T = E - E'$ , and equation (4) already contains it. What comes out is a sharp, computable maximum that no electron may exceed, and it is visible as a cliff in real data.

Step 1. Get $T$ from equation (4). With $\kappa \equiv (E/m_ec^2)(1-\cos\theta)$ we have $E' = E/(1+\kappa)$ , so

T = E - E' = E\,\frac{\kappa}{1+\kappa}

Why: $T$ is monotonic in $\kappa$ , and $\kappa$ is monotonic in $\theta$ on $[0,\pi]$ . So the hardest bounce gives the hottest electron, with no turning points to worry about.

Step 2. Find the ends. At $\theta = 0$ , $\kappa = 0$ and $T = 0$ — a photon that misses transfers nothing. At $\theta = 180°$ , $1-\cos\theta = 2$ , so $\kappa_{\max} = 2E/m_ec^2$ and

\boxed{\,T_{\max} = E\,\frac{2E/m_ec^2}{1 + 2E/m_ec^2} = \frac{2E^2}{m_ec^2 + 2E}\,}

Why: this is the Compton edge, and it is a hard kinematic ceiling. Not a probability, not a distribution — an electron with more energy than this cannot have come from Compton scattering of a photon of energy $E$ , full stop.

Step 3. Put Cs-137 through it: $E = 662\ \mathrm{keV}$ , $m_ec^2 = 511.0\ \mathrm{keV}$ .

T_{\max} = \frac{2(662)^2}{511.0 + 1324} \ \mathrm{keV} = \frac{876{,}488}{1835.0}\ \mathrm{keV} = \mathbf{477.7\ keV}

Why: that is the cliff in the overview, computed from the electron's mass and the source's energy and nothing else. Every gamma spectroscopy lab in the world sees it, and it is the standard calibration check.

Step 4. And the photon that got away. At $\theta = 180°$ it returns with $E' = E - T_{\max} = 662 - 477.7 = \mathbf{184.3\ keV}$ . Why: this is the backscatter peak. Gammas that leave the source, bounce off the shielding, and come back into the detector arrive at exactly this energy — so the little bump at $184\ \mathrm{keV}$ is not the source, it is the room. The edge and the bump are one event counted from both ends: $477.7 + 184.3 = 662.0$ .

Step 5. Take the ultra-relativistic limit and find something genuinely strange. As $E \to \infty$ , equation (4) at $\theta = 180°$ gives

E' \;\longrightarrow\; \frac{m_ec^2}{1 - \cos 180°} = \frac{m_ec^2}{2} = \mathbf{255.5\ keV}

Why: the backscattered photon's energy saturates. Throw a $1\ \mathrm{TeV}$ photon at an electron and what comes straight back is a $255\ \mathrm{keV}$ photon — the same as if you had thrown a $10\ \mathrm{MeV}$ one. All the extra energy goes to the electron. There is a ceiling on how hard a stationary electron can throw a photon backwards, and it is half the electron's rest energy. *Verified in verify_aq-sr-compton-scattering.py: $T_{\max} = 2E^2/(m_ec^2+2E)$ is derived symbolically from equation (4); the Cs-137 numbers 477.7 keV (edge), 184.3 keV (backscatter peak), 288.4 keV and 373.6 keV (the 90° pair) are recomputed from CODATA and asserted against the literals printed here; the ledger $477.7 + 184.3 = 662$ is checked to close; and the $E \to \infty$ saturation to $m_ec^2/2 = 255.5$ keV is confirmed as a limit.*

Figure

THE SHIFT IS ABSOLUTE, SO THE VISIBILITY IS NOT — this is why Compton needed X-rays. The curve is Δλ = λ_C (1 − cos θ), the SAME curve for every incident photon in the universe: zero at θ = 0, one Compton wavelength (2.43 pm) at 90 degrees, two (4.85 pm) at backscatter. But a fixed absolute shift is a wildly different FRACTIONAL shift depending on what you came in with. Visible light at 500 nm: 0.0005 percent, utterly unmeasurable. Compton's molybdenum X-rays at 71.1 pm: 3.4 percent, right at the edge of a 1923 Bragg spectrometer. A caesium-137 gamma at 1.87 pm: 130 percent, the photon loses most of its energy. Same physics, three verdicts

Limiting cases

Case	Result	Meaning
$\theta \to 0$	$\Delta\lambda \to 0$ , $E' \to E$	no deflection, no transfer — the beam sails through
$\theta = 90°$	$\Delta\lambda = \lambda_C = 2.43\ \mathrm{pm}$	exactly one Compton wavelength — the definition made visible
$\theta = 180°$	$\Delta\lambda = 2\lambda_C = 4.85\ \mathrm{pm}$	the maximum shift there is, at any energy, ever
$E \ll m_ec^2$ (Thomson)	$\Delta\lambda/\lambda \to 0$ , $E' \to E$	the classical limit: wavelength unchanged, as a wave would have it
$E \gg m_ec^2$	$E' \to \dfrac{m_ec^2}{1-\cos\theta}$ ; at $180°$ , $m_ec^2/2 = 255.5$ keV	the outgoing energy forgets the incoming energy — a hard ceiling
recoil, $\theta = 180°$	$T_{\max} = \dfrac{2E^2}{m_ec^2+2E}$	the Compton edge: 477.7 keV for Cs-137
target a proton, not an electron	$\lambda_C \to h/m_pc = 1.32\ \mathrm{fm}$	1836× smaller shift — hence the unshifted line in real data
electron moving at you ( $\gamma \gg 1$ )	$E' \approx 4\gamma^2 E$	inverse Compton — the same equation, another frame

Common trapThe shift is absolute, not fractional — and that is the whole reason the effect hid for a century

The commonest misreading of $\Delta\lambda = \lambda_C(1-\cos\theta)$ is to treat it as a fractional change. It is not. It is a fixed number of picometres, the same for a radio wave and a gamma ray. At $90°$ every photon in the universe gains exactly $2.43\ \mathrm{pm}$ . For a $500\ \mathrm{nm}$ photon from a sodium lamp that is a fractional change of $\mathbf{4.9\times10^{-6}}$ — far below what any 1920s spectrometer could resolve, and swamped by line width. For Compton's $71.1\ \mathrm{pm}$ X-rays it is $\mathbf{3.4\%}$ — small, but visible on a Bragg spectrometer and unmistakably growing with angle. For a $662\ \mathrm{keV}$ gamma ( $\lambda = 1.87\ \mathrm{pm}$ ) it is $\mathbf{130\%}$ .

So "why did nobody see this before 1923?" has a precise answer: nobody had a bright, monochromatic beam whose wavelength was within a few orders of magnitude of $2.4\ \mathrm{pm}$ . Compton did not have a better idea than his predecessors; he had X-rays. A related trap: real spectra show an unshifted line beside the shifted one. That is not a failure of the formula — it is the formula applied to the right mass. Photons scattering off tightly bound inner electrons recoil against the whole atom, so $m_e \to M_{\mathrm{atom}}$ and $\lambda_C$ collapses by orders of magnitude. The shift is still there; it is just too small to see.

Worked examples

1 · Compton's own experiment, 1923

Molybdenum $\mathrm{K}\alpha$ X-rays, $\lambda = 71.1\ \mathrm{pm}$ , scattered from graphite. What is the scattered wavelength at $90°$ , and is it measurable?

Step 1 — identify. Free-electron Compton scattering; equation (3) applies with $\theta = 90°$ , so $1 - \cos 90° = 1$ exactly.

Step 2 — apply. $\Delta\lambda = \lambda_C \times 1 = \mathbf{2.4263\ pm}$ . So $\lambda' = 71.1 + 2.43 = \mathbf{73.53\ pm}$ .

Step 3 — the boundary number. The fractional shift is $2.4263/71.1 = \mathbf{3.41\%}$ . A Bragg spectrometer of the period could resolve a few tenths of a percent, so a $3.4\%$ displacement of a sharp $\mathrm{K}\alpha$ line is glaring, not marginal. And at $135°$ the shift grows to $\lambda_C(1+\tfrac{1}{\sqrt2}) = \mathbf{4.14\ pm}$ , i.e. $5.8\%$ : the effect grows with angle exactly as a cosine says it should, and that angular march — not the single number — is what made the result decisive.

2 · Reading a caesium-137 gamma spectrum

A Cs-137 source emits $662\ \mathrm{keV}$ gammas into a scintillator. Predict the two kinematic features of the recorded spectrum.

Step 1 — the Compton edge. Equation (5) with $E = 662$ , $m_ec^2 = 511.0\ \mathrm{keV}$ :

T_{\max} = \frac{2(662)^2}{511.0 + 2(662)} = \frac{876{,}488}{1835.0} = \mathbf{477.7\ keV}

Electrons deposit anything up to this and nothing above it — hence a continuum with a cliff.

Step 2 — the backscatter peak. The photon that delivered that maximal kick left with $E' = 662 - 477.7 = \mathbf{184.3\ keV}$ . Gammas backscattering off the shielding outside the detector and re-entering deposit exactly this: a small bump at $184\ \mathrm{keV}$ .

Step 3 — the boundary number, and the calibration check. $477.7 + 184.3 = 662.0\ \mathrm{keV}$ : the two features are one event counted from both ends, so they must sum to the photopeak. If your measured edge and backscatter peak do not add up to 662, your calibration is wrong — and that is exactly how the check is used in practice. For contrast, $90°$ scattering gives $E' = \mathbf{288.4\ keV}$ and $T = \mathbf{373.6\ keV}$ : even a right-angle bounce costs the photon more than half its energy.

Figure

THE CLIFF IS KINEMATICS, NOT ELECTRONICS. A real caesium-137 gamma spectrum has three features and two of them are pure four-momentum conservation. The photopeak at 662 keV is the source. The Compton continuum is the recoil-electron spectrum, and it STOPS DEAD at T_max = 2E²/(m_e c² + 2E) = 477.7 keV — the Compton edge, the hardest kick a 662 keV photon can give a free electron, reached at 180 degree backscatter. The photon that delivered that kick leaves with 662 − 477.7 = 184.3 keV, and if it bounces off the shielding back INTO the detector it makes the backscatter peak. Edge plus backscatter peak equals the photopeak: 477.7 + 184.3 = 662. That sum is the standard calibration check

Physics of the portrait

The constant in the formula is not a fitted parameter. It is a length built out of the electron, and it is the length at which a photon starts to matter to it.

Origins · 2 minCompton wavelength and how this was discovered

Compton wavelength

1923 · the length at which light starts to hurt · h/mc · and the only constant in the formula

$\lambda_C = h/m_ec = 2.42631\ \mathrm{pm}$ is what you get when you ask: at what wavelength does a photon carry as much momentum as the electron's own mass-scale, $m_ec$ ? Set $h/\lambda = m_ec$ and the answer falls out. Above that wavelength a photon is a feather against a bowling ball and bounces off with its energy intact — that is Thomson scattering, and it is why light does not redden crossing a room. Below it, the photon hits hard enough to move what it hit. The Compton wavelength is the frontier between those two worlds, and the shift formula measures your distance from it. It is not the size of the electron: the classical electron radius is $r_e = 2.818\ \mathrm{fm}$ , smaller by the fine-structure constant, and the Bohr radius is larger by the same factor — so $r_e : \bar{\lambda}_C : a_0 = \alpha^2 : \alpha : 1$ arranges the three fundamental lengths of atomic physics on one ladder whose rungs are $\alpha \approx 1/137$ . Arthur Compton, at Washington University in St. Louis, scattered molybdenum X-rays off graphite and found the shift his 1923 Physical Review paper had predicted from exactly this kinematics; Peter Debye, in Zürich, derived the same result independently and got into print first. Compton took the 1927 Nobel Prize, shared with C. T. R. Wilson, whose cloud chamber had made the recoil electrons visible.

History

1905

— Einstein proposes the light quantum, giving the photon an energy hf; its momentum h/lambda follows from relativity.

1923

— Compton publishes the shift and its derivation in Physical Review 21, 483, having submitted it in December 1922; he scattered molybdenum X-rays off graphite.

1923

— Debye publishes the same result independently in Physikalische Zeitschrift 24, 161, and reaches print before Compton does.

1923

— C. T. R. Wilson photographs the recoil electrons in a cloud chamber, making the struck particle visible.

1925

— Bothe and Geiger show by coincidence counting that the scattered photon and the recoil electron appear together, event by event, refuting the Bohr-Kramers-Slater proposal that energy is conserved only statistically.

1927

— Compton shares the Nobel Prize in Physics with C. T. R. Wilson.

1929

— Klein and Nishina compute the scattering cross-section from the Dirac equation, supplying the intensity that the kinematics cannot.

Where this lives today

Compton scattering is what a gamma ray does to matter between about $100\ \mathrm{keV}$ and a few $\mathrm{MeV}$ — the band covering most of nuclear medicine and radiotherapy. It is the dominant photon interaction in soft tissue across the diagnostic range, so it governs where the dose lands in a treatment plan and it is why a raw X-ray image is fogged by scatter. Every gamma spectrometer's continuum is the recoil-electron spectrum computed above. But the reading that belongs to relativity runs the other way. Nothing forced the electron to be at rest — that was our choice of frame, not a fact about the world. Let it be ultra-relativistic and charging at the photon, and the identical conservation law says the photon comes back blueshifted, by roughly $4\gamma^2$ . This is inverse Compton scattering, an engine of high-energy astrophysics: electrons in the jets of active galaxies kick CMB photons up into X-rays, and the resulting distortion of the CMB through a galaxy cluster (the Sunyaev–Zel'dovich effect) is now used to find clusters.

Go deeper · Derivation C · 2 minInverse Compton: the same equation, read from a moving frame

Derivation CInverse Compton: the same equation, read from a moving frame

The claim in the overview was that inverse Compton scattering is not a different process — it is this one, viewed from elsewhere. That is worth proving, because it is the most relativistic thing in this article: nothing changes but the observer, and a microwave becomes an X-ray. We work in the Thomson regime, where the algebra is transparent and the physics unobscured.

Step 1. Set the scene. A photon of lab energy $E$ meets an electron of Lorentz factor $\gamma$ head-on. Go to the electron's rest frame — which we may, because it exists — and call quantities there starred. In that frame the electron is at rest, and everything derived above applies verbatim. Why: the derivation assumed a stationary electron, so find the frame where the electron is stationary. That frame always exists for a massive particle, and this is what it is for.

Step 2. Boost the photon in. A head-on collision means the photon and electron approach, so the relativistic Doppler shift blueshifts the photon into the electron's frame by $\gamma(1+\beta)$ :

E^* = \gamma(1+\beta)\,E \;\approx\; 2\gamma E \qquad (\beta \to 1)

Why: this is the longitudinal Doppler factor, owned by Relativistic Doppler Effect. For $\gamma \gg 1$ , $\gamma(1+\beta) \to 2\gamma$ . The electron has not done anything yet. It has merely been looked at from its own frame, and the photon is already $2\gamma$ times more energetic.

Step 3. Scatter, and check we may neglect the shift. In the electron frame this is ordinary Compton scattering, and if $E^* \ll m_ec^2$ we are in the Thomson limit, where equation (4) gives $E'^* \approx E^*$ . Why: the condition $2\gamma E \ll m_ec^2$ is very easily met. For a CMB photon ( $E = 6.35\times10^{-4}\ \mathrm{eV}$ ) and $\gamma = 1000$ : $2\gamma E = 1.27\ \mathrm{eV}$ , against $511{,}000\ \mathrm{eV}$ . Deep in the Thomson regime by five orders of magnitude — the electron's rest frame sees a completely elastic bounce.

Step 4. Boost back out. The photon that scatters straight back in the electron frame is travelling along the electron's motion in the lab, so coming home it is Doppler-boosted again by the same factor:

E' \;\approx\; \gamma(1+\beta)\,E'^* \;\approx\; 2\gamma \cdot 2\gamma E \;=\; \boxed{4\gamma^2 E}

Why: two boosts, one in and one out — and the electron never gave up any energy in its own frame at all. The entire gain is kinematic, the price of changing frames twice. That is why the answer goes as $\gamma^2$ and not as $\gamma$ .

Step 5. Put the number in. A cosmic-microwave-background photon, $E = 6.35\times10^{-4}\ \mathrm{eV}$ (a microwave), off an electron with $\gamma = 1000$ (a $511\ \mathrm{MeV}$ electron, unremarkable in an astrophysical jet):

E' \approx 4(1000)^2 \times 6.35\times10^{-4}\ \mathrm{eV} = 4\times10^6 \times 6.35\times10^{-4}\ \mathrm{eV} = 2540\ \mathrm{eV} = \mathbf{2.5\ keV}

Why: a microwave photon has come back as an X-ray, its energy multiplied by four million, and the equation that did it is the same $\Delta\lambda = \lambda_C(1-\cos\theta)$ that reddens X-rays in a graphite block. The only thing that changed is where we were standing. That is what it means for a law to be relativistic. *Verified in verify_aq-sr-compton-scattering.py: the $4\gamma^2$ result is confirmed against an EXACT relativistic calculation — the photon's four-momentum is boosted into the electron rest frame with a real Lorentz matrix, scattered through 180° with the exact equation (4) (not the Thomson approximation), and boosted back — and the exact answer agrees with $4\gamma^2E$ to better than 0.1% at $\gamma = 1000$ ; the Thomson-regime condition $2\gamma E \ll m_ec^2$ is asserted numerically; and the printed 2.5 keV is reproduced.*

Going further

Texts. Griffiths, Introduction to Elementary Particles, 2nd ed., §3.5 · Rindler, Introduction to Special Relativity, 2nd ed., §35 · Rybicki & Lightman, Radiative Processes in Astrophysics, ch. 7 (inverse Compton) · A. H. Compton, "A Quantum Theory of the Scattering of X-rays by Light Elements", Physical Review 21, 483 (1923) · P. Debye, Physikalische Zeitschrift 24, 161 (1923).

Provenance. Signature $(+,-,-,-)$ ; $P\cdot P = +m^2c^2$ ; photons massless. The AllQFT base was checked and found clean — its recoil-energy card was independently re-derived and is correct. verify_aq-sr-compton-scattering.py re-derives every number printed above from CODATA: $\lambda_C = 2.42631$ pm, $m_ec^2 = 511.0$ keV, the Mo K $\alpha$ numbers, the Cs-137 numbers (edge 477.7 keV, backscatter peak 184.3 keV, ledger 477.7 + 184.3 = 662), the saturation $m_ec^2/2 = 255.5$ keV, the length ladder $r_e : \bar{\lambda}_C : a_0 = \alpha^2 : \alpha : 1$ , and the inverse-Compton 2.5 keV checked against an exact two-boost Lorentz calculation.

Scope — the split, stated accurately. `aq-qm-compton-scattering` (QUANT.WAVE.COMPT.1108) is the quantum-mechanics home of this effect: it presents Compton scattering as the decisive evidence for the photon, and reaches the shift by squaring the three-vector momentum equation. This article is the relativistic home: it derives the same shift covariantly, in one four-vector contraction with no components and no recoil angle, and extends to what kinematics alone can still deliver — the Compton edge, the high-energy saturation, and inverse Compton scattering as a change of frame, which the QM article names but does not derive. Neither article computes the scattering intensity: the angular distribution is the Klein–Nishina cross-section (1929), it is quantum electrodynamics, and a future QFT track owns it. Both articles cross-link the other.

Builds on. Relativistic Four-Momentum · Energy-Momentum Relation · Massless Particles · Invariant Mass · Conservation of Four-Momentum · Relativistic Doppler Effect. Leads to. Relativistic Energy · Mass-Energy Equivalence · Relativistic Momentum · Mandelstam Invariants.

Why is the shift independent of the incident wavelength? That still feels like a coincidence.

It is not — you can watch it happen in Step 6. The relation the four-momentum algebra gives is $1/E' - 1/E = (1-\cos\theta)/m_ec^2$ , and $1/E$ is proportional to $\lambda$ . So the equation is already a statement about a difference of wavelengths, and a difference has no memory of where it started. The deeper reason: squaring the four-momenta left exactly one energy scale in the problem, $m_ec^2$ — and a shift must be measured in units of something. The only length available is $h/m_ec$ . There was nothing else it could have been.

The electron's recoil angle never appears. Where did it go?

It was deliberately destroyed, in Step 2, and that is the most important line in the article. The recoil angle lives in the direction of $P_e'$ ; the norm $P_e'\cdot P_e' = m_e^2c^2$ depends only on that four-vector's length, which is the electron's mass, the same no matter where it went. Contracting $P_e'$ with itself projects out exactly the information we did not have and did not want. This is the standard opening move of relativistic kinematics: whenever a final-state particle goes unobserved, isolate it, square it, and the mass shell will eat it. It works for decays, for missing-energy searches, and for every $2\to2$ collision you will ever meet.

If the photon loses energy, does it slow down?

No — it cannot, ever; it always travels at exactly $c$ . This is where classical intuition fails hardest. A billiard ball that loses energy slows down, because for a massive particle energy is tied to speed. A photon is massless, so its energy is tied to its frequency ( $E = hf$ ), and it pays for the collision by reddening rather than by dawdling. The energy goes into the electron, entirely, as kinetic energy $T = E - E'$ . The photon that comes out of a Compton scatter is a different colour, not a slower photon.

Can a photon just be absorbed by a free electron instead of scattering?

No — it is kinematically forbidden, and the four-momentum machinery shows it instantly. If a free electron absorbed a photon, the final state would be a single electron with four-momentum $P_e + P_\gamma$ . Take its norm: $(P_e+P_\gamma)^2 = m_e^2c^2 + 0 + 2P_e\cdot P_\gamma = m_e^2c^2 + 2m_eE > m_e^2c^2$ for any $E > 0$ . But a lone electron must have norm exactly $m_e^2c^2$ . Contradiction. So a free electron can only ever scatter a photon, never swallow it. Absorption — the photoelectric effect — needs a third body, a bound nucleus, to soak up the excess. `aq-qm-compton-scattering` makes the same point.

Real spectra show an unshifted peak next to the shifted one. Does that break the formula?

No — it confirms it, applied honestly. The formula's $m_e$ is the mass of whatever recoils. A photon that scatters off a loosely bound outer electron effectively meets a free electron, and you get the full $\lambda_C = 2.43\ \mathrm{pm}$ shift. But a photon that scatters off a tightly bound inner electron cannot dislodge it, so the recoil is taken up by the entire atom — and then $m_e \to M_{\mathrm{atom}}$ , thousands of times larger, and the Compton wavelength shrinks by the same factor. For carbon the shift becomes about $10^{-4}\ \mathrm{pm}$ : still there, utterly unresolvable. The "unshifted" line is Compton scattering, off a target 22,000 times heavier.

Is inverse Compton scattering really the same thing, or is that just a slogan?

It is literally the same thing, and the derivation above is the proof: we never wrote a new equation. We took the standard result — which assumed a stationary electron — went to the frame where the electron is stationary (which always exists), applied it unchanged, and came back. The gain of roughly $4\gamma^2$ came entirely from Doppler shifting twice: once in, once out. In its own frame the electron barely notices, handing back a photon of almost exactly the energy it received. Nothing physical distinguishes the two processes — only the observer's velocity does.

Compton Scattering

The physical picture

The core equations

Deriving the shift

Limiting cases

Worked examples

1 · Compton's own experiment, 1923

2 · Reading a caesium-137 gamma spectrum

Physics of the portrait

Where this lives today

Going further

Before this

Where this leads