Relativistic Elastic Collisions

Roll a billiard ball into an identical one at rest. If the collision is elastic and off-centre, the two balls leave at exactly ninety degrees to each other — always, whatever the impact parameter, whatever the speed. Every pool player knows it in their hands, and it falls out of Newton in two lines: momentum gives $\vec p_1 = \vec p_3 + \vec p_4$ , kinetic energy gives $p_1^2 = p_3^2 + p_4^2$ , and together they are Pythagoras. In Newtonian mechanics the right angle is not an approximation — it is exact. It is also false. Fire the same collision at $1\ \mathrm{GeV}$ of kinetic energy — a proton on a proton in a bubble chamber — and the tracks come out at $\mathbf{77.9^\circ}$ . At $\gamma = 5$ the answer is exactly $\mathbf{60^\circ}$ ; at $\gamma = 100$ the pair emerges inside a $16^\circ$ cone. The opening angle is always less than $90^\circ$ , and it closes toward zero as the beam energy climbs. What survives is better than the rule it replaces. The sum of the lab angles is not fixed — it depends on how the collision was struck. But the product of their tangents is, and it is fixed by nothing but the beam's Lorentz factor:

\tan\theta_1 \, \tan\theta_2 \;=\; \frac{2}{\gamma + 1}

Set $\gamma = 1$ and this reads $\tan\theta_1\tan\theta_2 = 1$ , which says $\theta_1 = 90^\circ - \theta_2$ : Newton's right angle, recovered exactly. Set $\gamma > 1$ and the product drops below $1$ , which forces $\theta_1 + \theta_2 < 90^\circ$ — with no exceptions, at any scattering angle. Champion photographed it in a cloud chamber in 1932.

Key takeaways

Signature $(+,-,-,-)$ : $p\cdot p = (E/c)^2 - |\vec p\,|^2 = +\,m^2c^2$ on shell. $s$ is in energy² units, so $\sqrt{s}$ is an energy (Centre-of-Momentum Frame locks this).
Elastic means every $m_i$ is unchanged. Four-momentum is conserved in every collision; what is additionally true here is that the rest masses survive — and therefore so does the total kinetic energy.
In the CM frame an elastic collision is almost nothing: the momenta stay back-to-back and $|\vec p^{\;*}|$ is unchanged. The collision merely rotates the pair through the CM angle $\theta^{*}$ .
Owned result. Boosting that rotation to the lab gives $\tan\theta_1 = \tan(\theta^{*}/2)/\gamma_{\mathrm{cm}}$ and $\tan\theta_2 = \cot(\theta^{*}/2)/\gamma_{\mathrm{cm}}$ , so the product is $1/\gamma_{\mathrm{cm}}^2 = 2/(\gamma+1)$ — independent of $\theta^{*}$ . Below $1$ for any $\gamma > 1$ , so the opening angle is strictly under $90^\circ$ , always.
The minimum opening (equal energy sharing) is $\theta_{\min} = 2\arctan\sqrt{2/(\gamma+1)}$ , with $\cos\theta_{\min} = (\gamma-1)/(\gamma+3)$ — exactly $60^\circ$ at $\gamma = 5$ .
Key idea: the Newtonian right angle is a Pythagoras theorem about kinetic energy, and relativity breaks it because $E^2 = (pc)^2 + (mc^2)^2$ is a DIFFERENT Pythagoras. What replaces it is not a bigger formula but a sharper one — a product of tangents that knows only $\gamma$ .

The core insight

Newton's right angle rests on $p_1^2 = p_3^2 + p_4^2$ — conservation of kinetic energy written in momentum variables — and it works only because $T = p^2/2m$ makes kinetic energy quadratic in momentum. Relativistically $T = \sqrt{(pc)^2 + (mc^2)^2} - mc^2$ is not quadratic in $p$ , the Pythagorean identity fails, and the right angle goes with it. The relativistic collision keeps a hidden simplicity, but you must boost to the CM frame to see it: there, an elastic collision does nothing but turn a rigid back-to-back pair of arrows through an angle. Every lab-frame complication — the closing opening angle, the forward collimation, the uneven energy sharing — is that one rotation, seen from a moving train.

Where this lives today

Elastic scattering is how the structure of matter gets read. Rutherford found the nucleus by watching alpha particles bounce off gold elastically; the shape of the angular distribution, not the fact of scattering, is what carries the information. Every later generation ran the same play at higher energy — elastic electron–proton scattering at SLAC mapped the proton's charge distribution and then, past a threshold, stopped being elastic at all and revealed quarks. The kinematics here is the frame within which the dynamics is measured: the cross-section supplies the physics, the opening angle supplies the ruler. At a collider the elastic channel is a specialist's business: two protons in, two protons out, deflected by microradians, so detecting them needs silicon detectors inside the beam pipe hundreds of metres downstream — which is what TOTEM and ATLAS/ALFA do. The pay-off is the total cross-section, extracted from the elastic rate by the optical theorem, and one of the few LHC numbers that cannot be predicted from first principles. The opening angle itself remains the cleanest demonstration that relativity is a replacement, not a correction: a few-GeV proton beam in a hydrogen bubble chamber makes elastic events by the thousand, every one a pair of tracks a Newtonian would call impossible. Compton Scattering runs the same conservation laws with a massless projectile, where the invariant statement is a wavelength shift rather than an angle.

The physical picture

Two protons, one moving and one at rest, scatter elastically: two protons out, same masses, nothing created. Four-momentum is conserved, as in every collision (Conservation of Four-Momentum owns that). What makes it elastic is the extra condition that no rest mass changed — and since $E_i = T_i + m_ic^2$ with every $m_i$ fixed, conservation of total energy then forces conservation of total kinetic energy too. That is the whole of the extra input. Now go to the centre-of-momentum frame (Centre-of-Momentum Frame owns it: $\vec V_{\mathrm{CM}} = c^2\vec P_{\mathrm{tot}}/E_{\mathrm{tot}}$ , and $\sqrt{s}$ is its energy). Both protons approach with equal and opposite momenta $\pm\vec p^{\;*}$ . Afterwards they must still be equal and opposite, because the total is still zero — and because the masses are unchanged and $\sqrt{s}$ is conserved, each still has $E^{*} = \sqrt{s}/2$ and therefore the same $|\vec p^{\;*}|$ . So in the CM frame an elastic collision is a rotation, and nothing more. A rigid back-to-back pair of arrows turns through some angle $\theta^{*}$ — which angle is a question about forces and impact parameter that kinematics cannot answer and does not need to. Everything the lab measures is that rotation, seen from a frame moving at $\vec V_{\mathrm{CM}}$ .

Figure

PUNCHLINE — the equal-mass elastic opening angle collapses below Newton's 90°. Marked with real numbers: a 1 GeV proton on a stationary proton (γ = 2.066) opens to 77.85°, not 90°. At γ = 5 the answer is EXACTLY 60°. At γ = 100 the pair emerges inside a 16° cone. The dashed line at 90° is the Newtonian prediction — exact in Newton, and never attained relativistically

The core equations

Elastic means the masses survive, which is what makes total kinetic energy a conserved quantity:

\sum_{\mathrm{in}} p_i^\mu = \sum_{\mathrm{out}} p_f^\mu \quad\text{(always)}, \qquad \{m_i\}_{\mathrm{in}} = \{m_f\}_{\mathrm{out}} \quad\text{(elastic)} \;\Longrightarrow\; \sum_{\mathrm{in}} T_i = \sum_{\mathrm{out}} T_f \qquad (1)

where: $T = E - mc^2 = (\gamma-1)mc^2$ . Physical reading: four-momentum conservation is not a property of elastic collisions — it holds in every collision, including the wholly inelastic ones of the next article. The elastic condition is the one on the right, and kinetic-energy conservation is its consequence, not its definition. In the CM frame, equal masses give both particles the same energy, and the collision is a pure rotation:

E^{*} = \frac{\sqrt{s}}{2}, \qquad |\vec p^{\;*}| = \frac{1}{c}\sqrt{E^{*2} - m^2c^4} \;=\; mc\sqrt{\frac{\gamma - 1}{2}} \quad\text{(unchanged by the collision)} \qquad (2)

where: $\gamma$ — the beam's Lorentz factor in the lab, so $E_1^{\mathrm{lab}} = \gamma mc^2$ , and $s = 2m^2c^4(1+\gamma)$ . Physical reading: the only thing the collision does in the CM frame is choose $\theta^{*}$ . Both energies and both momentum magnitudes are fixed before it happens. The boost back to the lab is controlled by two numbers, and for equal masses they are startlingly clean:

\beta_{\mathrm{cm}} = \sqrt{\frac{\gamma-1}{\gamma+1}}\,, \qquad \gamma_{\mathrm{cm}} = \sqrt{\frac{\gamma+1}{2}}\,, \qquad\text{and}\qquad \beta_{\mathrm{cm}}\,\frac{E^{*}}{c} = |\vec p^{\;*}| \qquad (3)

where: the last identity says the CM frame moves at exactly the speed each particle has within it — obvious once seen, since in the CM frame the target recedes at $\beta_{\mathrm{cm}}$ and it is one of the two particles. Physical reading: this coincidence is special to equal masses, and it is what makes the next equation so simple. And now the owned result. Boosting the CM rotation back into the lab gives the two lab angles, and their product forgets $\theta^{*}$ entirely:

\boxed{\;\tan\theta_1 = \frac{\tan(\theta^{*}/2)}{\gamma_{\mathrm{cm}}}, \qquad \tan\theta_2 = \frac{\cot(\theta^{*}/2)}{\gamma_{\mathrm{cm}}} \qquad\Longrightarrow\qquad \tan\theta_1\,\tan\theta_2 \;=\; \frac{1}{\gamma_{\mathrm{cm}}^2} \;=\; \frac{2}{\gamma + 1}\;} \qquad (4)

where: $\theta_1$ , $\theta_2$ — the lab angles of the scattered and recoil particles, on opposite sides of the beam. Physical reading: the sum is not fixed; the product of the tangents is. That is the correct relativistic generalisation of Newton's right angle, and the two coincide at $\gamma = 1$ , where the product is $1$ and $\theta_1 + \theta_2 = 90^\circ$ exactly. The consequence is immediate: for $\gamma > 1$ the product is below $1$ , so $\tan\theta_1 < \cot\theta_2 = \tan(90^\circ - \theta_2)$ , and $\tan$ is increasing on $(0^\circ, 90^\circ)$ , so

\gamma > 1 \qquad\Longrightarrow\qquad \boxed{\;\theta_1 + \theta_2 \;<\; 90^\circ\;} \qquad\text{for every } \theta^{*} \qquad (5)

Deriving the opening angle

Derivation AIn the centre-of-momentum frame an elastic collision is nothing but a rotation

Before any angle can be computed the collision must be reduced to its simplest description, and for an elastic collision that description is almost embarrassingly small: in the CM frame the two particles keep their energies, keep their momentum magnitudes, stay back-to-back, and differ before and after by a single rotation angle. Everything the lab sees is that rotation, distorted by a boost.

Step 1. Impose the CM condition. By definition $\vec p_1^{\;*} + \vec p_2^{\;*} = \vec 0$ before the collision, so $\vec p_2^{\;*} = -\vec p_1^{\;*}$ . Four-momentum conservation makes the total spatial momentum zero after as well, so $\vec p_4^{\;*} = -\vec p_3^{\;*}$ . Why: back-to-back in, back-to-back out, and this needs nothing but momentum conservation — it is true for inelastic collisions too.

Step 2. Fix the energies from the invariant. Both particles have mass $m$ , so $E_3^{*2} - E_4^{*2} = (m^2 - m^2)c^4 = 0$ , and with $E_3^{*} + E_4^{*} = \sqrt{s}$ this forces $E_3^{*} = E_4^{*} = \sqrt{s}/2$ . The same argument before the collision gives $E_1^{*} = E_2^{*} = \sqrt{s}/2$ . Why: this is Centre-of-Momentum Frame's two-body split $E_1^{*} = (s + (m_1^2-m_2^2)c^4)/(2\sqrt{s})$ , with the mass-difference term vanishing because the masses are equal. Equal masses share the CM energy equally — always, and regardless of how the collision went.

Step 3. Conclude the momentum magnitude is unchanged. Each particle is on its mass shell, so $|\vec p^{\;*}|c = \sqrt{E^{*2} - m^2c^4}$ , and since $E^{*} = \sqrt{s}/2$ both before and after, $|\vec p^{\;*}|$ takes the same value before and after. Using $s = 2m^2c^4(1+\gamma)$ this is $\boxed{|\vec p^{\;*}| = mc\sqrt{(\gamma-1)/2}}$ . Why: THIS is what "elastic" buys, and it buys it only because the masses were unchanged. Had the collision created a new particle, $\sqrt{s}$ would have been shared differently and the magnitude would have dropped.

Step 4. Read off the geometry. Four vectors of equal length, two in and two out, each pair back-to-back. The out-pair is the in-pair rotated through an angle $\theta^{*}$ , and that single number is the only freedom the collision has. Why: kinematics cannot predict $\theta^{*}$ — that requires the dynamics (the force law, the impact parameter, the cross-section). But it does not need to. Every kinematic statement below is true for whatever $\theta^{*}$ nature chooses, which is exactly why the result that follows is so strong. *Verified in verify_aq-sr-relativistic-elastic-collisions.py: $E^{*} = \sqrt{s}/2$ is derived symbolically from the two-body split with $m_1 = m_2$ ; $|\vec p^{\;*}| = mc\sqrt{(\gamma-1)/2}$ is checked against the mass shell; and $s = 2m^2c^4(1+\gamma)$ is re-derived from the fixed-target formula.*

Go deeper · Derivation B · 2 minThe product of the tangents is two over gamma plus one

Derivation BThe product of the tangents is two over gamma plus one

This is the article's owned result. We take the CM rotation of the previous card and boost it into the lab, using the standard transformation of transverse and longitudinal momentum. The half-angle identity does all the work, and the payoff is an expression in which the CM scattering angle — the one thing kinematics cannot predict — cancels identically between the two particles.

Step 1. Boost each outgoing momentum. Under a boost of speed $\beta_{\mathrm{cm}}$ along the beam, transverse momentum is unchanged and longitudinal momentum mixes with energy: $p_{\perp} = p^{*}\sin\theta^{*}$ and $p_{\parallel} = \gamma_{\mathrm{cm}}\big(p^{*}\cos\theta^{*} + \beta_{\mathrm{cm}}E^{*}/c\big)$ . The lab angle is $\tan\theta_1 = p_{\perp}/p_{\parallel}$ . Why: the transverse component is the one a boost along the beam cannot touch, so all the angular distortion comes from the longitudinal component growing.

Step 2. Use the equal-mass coincidence. For equal masses $\beta_{\mathrm{cm}}E^{*}/c = p^{*}$ exactly — the CM frame moves at precisely the speed the particles have within it. Substituting, the whole expression collapses:

\tan\theta_1 = \frac{p^{*}\sin\theta^{*}}{\gamma_{\mathrm{cm}}\big(p^{*}\cos\theta^{*} + p^{*}\big)} = \frac{1}{\gamma_{\mathrm{cm}}}\cdot\frac{\sin\theta^{*}}{1 + \cos\theta^{*}} = \frac{\tan(\theta^{*}/2)}{\gamma_{\mathrm{cm}}}

Why: $p^{*}$ cancels completely, and the half-angle identity $\sin\theta/(1+\cos\theta) = \tan(\theta/2)$ turns the mess into one tangent. That cancellation is the whole trick, and it is a gift of the equal masses.

Step 3. Do the recoil particle. It leaves the CM frame at $\pi - \theta^{*}$ , so $\cos\theta^{*} \to -\cos\theta^{*}$ while $\sin\theta^{*}$ is unchanged. The same algebra gives $\tan\theta_2 = \dfrac{1}{\gamma_{\mathrm{cm}}}\cdot\dfrac{\sin\theta^{*}}{1 - \cos\theta^{*}} = \dfrac{\cot(\theta^{*}/2)}{\gamma_{\mathrm{cm}}}$ . Why: the half-angle identity for $1 - \cos\theta$ produces a cotangent instead of a tangent — and a tangent times its own cotangent is $1$ . The result is already visible.

Step 4. Multiply. $\tan\theta_1\tan\theta_2 = \dfrac{\tan(\theta^{*}/2)\cot(\theta^{*}/2)}{\gamma_{\mathrm{cm}}^2} = \boxed{\dfrac{1}{\gamma_{\mathrm{cm}}^2}}$ . The CM scattering angle has cancelled identically. Why: this is the payoff. The individual angles depend on how the collision was struck; the product does not, and so it is a prediction that can be tested on every event in a photograph without knowing anything about the dynamics.

Step 5. Put in $\gamma_{\mathrm{cm}}$ . For a beam of Lorentz factor $\gamma$ on an equal mass at rest, $\gamma_{\mathrm{cm}} = E_{\mathrm{tot}}/\sqrt{s} = (\gamma+1)mc^2/\big(mc^2\sqrt{2(1+\gamma)}\big) = \sqrt{(\gamma+1)/2}$ , so $\gamma_{\mathrm{cm}}^2 = (\gamma+1)/2$ and $\boxed{\tan\theta_1\tan\theta_2 = 2/(\gamma+1)}$ . Why: everything is now expressed in the one quantity an experimenter actually sets — the beam energy. No CM quantities survive in the final statement. *Verified in verify_aq-sr-relativistic-elastic-collisions.py: the full CM-to-lab boost is carried out SYMBOLICALLY for an arbitrary $\theta^{*}$ , and sympy reduces $\tan\theta_1\tan\theta_2 - 2/(\gamma+1)$ to exactly $0$ — the cancellation of $\theta^{*}$ is proved, not sampled. The identity $\beta_{\mathrm{cm}}E^{*}/c = p^{*}$ and $\gamma_{\mathrm{cm}}^2 = (\gamma+1)/2$ are proved separately.*

Go deeper · Derivation C · 2 minWhy the opening angle is always under ninety degrees, and where the minimum sits

Derivation CWhy the opening angle is always under ninety degrees, and where the minimum sits

The tangent identity is compact but its meaning is not yet visible. Here we extract the two statements an experimenter can check on a photograph: that the opening angle is strictly below $90^\circ$ for any relativistic beam and any scattering angle, and that it is smallest when the two particles share the energy equally. Along the way we recover Newton exactly, and we can say precisely where the base article goes wrong.

Step 1. Recover Newton. At $\gamma \to 1$ the product $2/(\gamma+1) \to 1$ , so $\tan\theta_1 = 1/\tan\theta_2 = \cot\theta_2 = \tan(90^\circ - \theta_2)$ , hence $\theta_1 = 90^\circ - \theta_2$ and $\boxed{\theta_1 + \theta_2 = 90^\circ}$ — for every $\theta^{*}$ . Why: the non-relativistic limit is not a coincidence to be checked afterwards. The right angle IS the statement $\tan\theta_1\tan\theta_2 = 1$ , and Newton is the $\gamma = 1$ case of a one-parameter family.

Step 2. Prove the inequality. For $\gamma > 1$ we have $2/(\gamma+1) < 1$ , so $\tan\theta_1\tan\theta_2 < 1$ , i.e. $\tan\theta_1 < \cot\theta_2 = \tan(90^\circ - \theta_2)$ . Both $\theta_1$ and $90^\circ - \theta_2$ lie in $(0^\circ, 90^\circ)$ , where $\tan$ is strictly increasing, so $\theta_1 < 90^\circ - \theta_2$ , which is $\boxed{\theta_1 + \theta_2 < 90^\circ}$ . Why: no approximation, no limit, no assumption about $\theta^{*}$ — this is a strict inequality holding event by event. It is the single most testable statement in relativistic collision kinematics.

Step 3. Find the minimum. With $t \equiv \tan\theta_1$ and the product fixed at $k = 2/(\gamma+1)$ , the opening angle is $f(t) = \arctan t + \arctan(k/t)$ . Then $f'(t) = 1/(1+t^2) - k/(t^2 + k^2)$ , which vanishes at $t = \sqrt{k}$ — the symmetric configuration $\theta_1 = \theta_2$ , reached when $\theta^{*} = 90^\circ$ and the two particles share the energy equally. Why: and it is a MINIMUM, not a maximum — as $t \to 0$ or $t \to \infty$ , $f \to 90^\circ$ . So the opening angle ranges over $[\theta_{\min},\, 90^\circ)$ , approaching the Newtonian value only in the grazing limit where one particle is barely deflected.

Step 4. Evaluate it. At $t = \sqrt{k}$ , $\theta_{\min} = 2\arctan\sqrt{k} = \boxed{2\arctan\sqrt{2/(\gamma+1)}}$ , and using $\cos(2\arctan x) = (1-x^2)/(1+x^2)$ ,

\cos\theta_{\min} \;=\; \frac{1 - k}{1 + k} \;=\; \frac{1 - \tfrac{2}{\gamma+1}}{1 + \tfrac{2}{\gamma+1}} \;=\; \boxed{\frac{\gamma - 1}{\gamma + 3}}

Why: at $\gamma = 5$ this is $4/8 = 1/2$ , so $\theta_{\min} = 60^\circ$ exactly — a clean anchor to remember and an easy one to check. At $\gamma = 1$ it is $0$ , giving $90^\circ$ , as it must.

Step 5. Locate the base article's error, because it is instructive. The AllQFT base asserts $\cos(\theta_1 + \theta_2) = (\gamma-1)/(\gamma+1)$ as a flat equality. Expand the cosine: $\cos(\theta_1+\theta_2) = \cos\theta_1\cos\theta_2\big(1 - \tan\theta_1\tan\theta_2\big) = \cos\theta_1\cos\theta_2\cdot\dfrac{\gamma-1}{\gamma+1}$ . The base dropped the factor $\cos\theta_1\cos\theta_2$ — which is not $1$ , and is not even constant. Its formula is therefore not the symmetric value ( $(\gamma-1)/(\gamma+3)$ ), nor an identity, nor a bound: at $\gamma = 2.066$ it returns $69.7^\circ$ , whereas the true minimum is $77.9^\circ$ and the true range is $[77.9^\circ,\, 90^\circ)$ . Why: the opening angle is simply NOT a function of $\gamma$ alone — it depends on $\theta^{*}$ as well. Only the product of tangents is $\theta^{*}$ -free, and that is exactly why the product, not the sum, is the right thing to write down. *Verified in verify_aq-sr-relativistic-elastic-collisions.py: the inequality $\theta_1 + \theta_2 < 90^\circ$ is checked over a dense grid of $(\gamma, \theta^{*})$ ; the minimum is confirmed at $\theta^{*} = 90^\circ$ by direct minimisation; $\cos\theta_{\min} = (\gamma-1)/(\gamma+3)$ is proved symbolically; and the base's $(\gamma-1)/(\gamma+1)$ is shown to disagree with the true opening angle at every $\theta^{*}$ .*

Limiting cases

Case	Result	Meaning
$\gamma \to 1$ (Newtonian)	$\tan\theta_1\tan\theta_2 \to 1$ , $\theta_1 + \theta_2 \to 90^\circ$	the billiard-ball right angle, recovered exactly
$\gamma > 1$ , any $\theta^{*}$	$\theta_1 + \theta_2 < 90^\circ$	strict, event by event — the marquee result
$\theta^{*} = 90^\circ$ (equal sharing)	$\theta_1 = \theta_2 = \arctan\sqrt{2/(\gamma+1)}$	the minimum opening angle; $\cos\theta_{\min} = (\gamma-1)/(\gamma+3)$
$\theta^{*} \to 0$ or $\pi$ (grazing)	$\theta_1 + \theta_2 \to 90^\circ$	the Newtonian corner — approached, never reached
$\gamma \gg 1$	$\theta_{\min} \approx 2\sqrt{2/\gamma}$	forward collimation into a narrow cone
$\gamma = 5$	$\theta_{\min} = 60^\circ$ exactly	$\cos\theta_{\min} = 4/8 = 1/2$ — the anchor to memorise
$m_2 \gg m_1$ (heavy target)	$\theta_1$ can exceed $90^\circ$	the equal-mass rule does not generalise

Common trapThe base article's opening-angle formula is wrong

The AllQFT base prints $\cos(\theta_1+\theta_2) = (\gamma-1)/(\gamma+1)$ as a flat equality. It is not one. The true identity is

\cos(\theta_1+\theta_2) = \cos\theta_1\cos\theta_2\big(1 - \tan\theta_1\tan\theta_2\big) = \cos\theta_1\cos\theta_2\cdot\frac{\gamma-1}{\gamma+1}

and the dropped factor $\cos\theta_1\cos\theta_2$ is neither $1$ nor constant. At $\gamma = 2.066$ the base returns $69.7^\circ$ , which is outside the physical range $[77.9^\circ,\ 90^\circ)$ entirely. The deeper point: an equal-mass elastic collision has one free parameter ( $\theta^{*}$ ), so no function of $\gamma$ alone can give the opening angle. Test any candidate at $\gamma = 1$ — the true minimum-angle formula returns $90^\circ$ ; the base's returns $0^\circ$ , which would mean a Newtonian collision fires both particles straight down the beam.

Worked examples

1 · A 1 GeV proton in a hydrogen bubble chamber

A proton of kinetic energy $T = 1.000\ \mathrm{GeV}$ scatters elastically off a proton at rest. Predict the smallest opening angle you could see, and the opening angle for an event in which the scattered proton comes off at $\theta_1 = 30^\circ$ . Take $m_pc^2 = 0.938272\ \mathrm{GeV}$ .

Step 1 — identify. Everything is set by the beam's Lorentz factor: $\gamma = 1 + T/m_pc^2 = 1 + 1.000/0.938272 = \mathbf{2.0658}$ .

Step 2 — the invariant. $\tan\theta_1\tan\theta_2 = 2/(\gamma+1) = 2/3.0658 = \mathbf{0.6524}$ — already below $1$ , so this collision cannot open to $90^\circ$ however it is struck.

Step 3 — the minimum opening. Symmetric sharing gives $\theta_{\min} = 2\arctan\sqrt{0.6524} = 2\arctan(0.8077) = 2\times 38.93^\circ = \mathbf{77.85^\circ}$ . Cross-check: $\cos\theta_{\min} = (\gamma-1)/(\gamma+3) = 1.0658/5.0658 = 0.2104$ , and $\arccos(0.2104) = 77.85^\circ$ . Two routes, same answer.

Step 4 — an asymmetric event. If $\theta_1 = 30^\circ$ then $\tan\theta_2 = 0.6524/\tan 30^\circ = 0.6524/0.5774 = 1.1299$ , so $\theta_2 = \mathbf{48.49^\circ}$ and the opening is $\mathbf{78.49^\circ}$ . Different event, different sum — still below $90^\circ$ , and only $0.6^\circ$ above the minimum.

Step 5 — check the CM picture. $\sqrt{s} = m_pc^2\sqrt{2(1+\gamma)} = 0.938272\times\sqrt{6.1316} = \mathbf{2.3234\ GeV}$ , so $E^{*} = \sqrt{s}/2 = 1.1617\ \mathrm{GeV}$ and $p^{*}c = \sqrt{1.1617^2 - 0.8804} = \mathbf{0.6849\ GeV}$ . The frame moves at $\beta_{\mathrm{cm}} = \sqrt{(\gamma-1)/(\gamma+1)} = \mathbf{0.5896}$ , giving $\gamma_{\mathrm{cm}} = \sqrt{(\gamma+1)/2} = \mathbf{1.2381}$ — and $1/\gamma_{\mathrm{cm}}^2 = 1/1.5329 = 0.6524$ , the product from Step 2.

Step 6 — the boundary number. The deficit from Newton is $90^\circ - 77.85^\circ = \mathbf{12.15^\circ}$ . On a bubble-chamber photograph, where tracks are measurable to a fraction of a degree, that is not a subtlety but a gulf — and it was measurable in Champion's 1932 cloud chamber too.

2 · How much energy can the beam give up?

In the same collision, what is the largest kinetic energy the recoiling target proton can take, and at what angle?

Step 1 — identify and apply. The recoil takes most energy thrown straight forward, $\theta_2 = 0$ , which is $\theta^{*} = \pi$ : a head-on hit in the CM frame, where the two particles simply exchange momenta. The outgoing CM pair is then the incoming pair reversed, so in the lab the projectile stops dead and the target carries the entire beam kinetic energy: $T_2^{\max} = T = \mathbf{1.000\ GeV}$ at $\theta_2 = \mathbf{0^\circ}$ .

Step 2 — the boundary number. This is the one degenerate configuration: the scattered projectile has no direction, because it has no momentum, and $\tan\theta_1\tan\theta_2 = 2/(\gamma+1)$ survives only as the limit $0\times\infty$ . For equal masses a relativistic elastic collision can transfer 100 % of the beam's kinetic energy — exactly as a Newtonian one can, and for the same reason: the CM-frame exchange is a symmetry of the equal-mass problem at any speed.

Figure

The same physics, drawn twice. LEFT — in the CM frame an elastic collision is a pure rotation: two equal, opposite arrows of length p = 0.685 GeV/c turn through θ, and nothing else happens. RIGHT — boosting back to the lab at β_cm = 0.5896 rakes both arrows forward. The Newtonian 90° corner (dashed) is never reached: the pair closes to 77.85° when the energy is shared equally, and to 78.49° for the θ₁ = 30° event

Physics of the portrait

The elastic collision is the oldest solved problem in mechanics and it keeps being the newest. Huygens got the rules right in 1669, more than a decade before the Principia; three centuries later the same collision, done fast enough, became a sharp test of relativity.

Origins · 2 minElastic collision and how this was discovered

Elastic collision

1669–2011 · Huygens' rules · the right-angle corollary · Champion's cloud chamber · TOTEM at the LHC

Christiaan Huygens communicated his rules of impact to the Royal Society in 1669, having worked them out years earlier and — characteristically — declined to publish. His insight was that in a perfectly elastic collision two quantities survive: the total momentum, and the sum of $mv^2$ , which would not be named kinetic energy for another century and a half. From those two rules the equal-mass right angle follows in a line. Its status as a theorem rather than an observation is what made it dangerous: the proof leans entirely on $T = p^2/2m$ , and when Einstein replaced that with $E^2 = (pc)^2 + (mc^2)^2$ the theorem lost its foundation without anyone immediately noticing — nothing on Earth moved fast enough to tell. The check came from radioactivity. In 1932 F. C. Champion photographed close collisions between fast $\beta$ -particles and atomic electrons in a Wilson cloud chamber — equal masses, relativistic speeds, tracks measurable with a ruler — expressly as a direct test of the conservation laws and of relativistic mechanics against the Newtonian rule (Proc. Roy. Soc. A 136, 630). The opening angle was the discriminating observable, and it remains so: elastic proton–proton scattering is still measured at the frontier, by TOTEM at the LHC at $\sqrt{s} = 7\ \mathrm{TeV}$ in 2011.

History

1669

— Huygens communicates his rules of impact to the Royal Society: in a perfectly elastic collision both the total momentum and the total of mv² are conserved. The equal-mass right angle follows from them.

1687

— Newton's Principia systematises collision mechanics; the equal-mass 90° rule becomes standard textbook physics for two centuries.

1905

— Einstein's relativistic dynamics replaces T = p²/2m with E² = (pc)² + (mc²)², quietly removing the hinge on which the right-angle proof rested.

1925

— Bothe and Geiger, and independently Compton and Simon, show energy and momentum are conserved in individual scattering events, not merely on average.

1932

— F. C. Champion photographs close β-particle–electron collisions in a Wilson cloud chamber as a direct test of the conservation laws and of relativistic mechanics; the equal-mass opening angle is the observable that separates them (Proc. Roy. Soc. A 136, 630).

2011

— TOTEM measures elastic proton–proton scattering at the LHC at √s = 7 TeV, where the same kinematics is still the tool for extracting the total cross-section.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §31 · Griffiths, Introduction to Elementary Particles, 2nd ed., §3.4 · Morin, Introduction to Classical Mechanics, ch. 13 · Hagedorn, Relativistic Kinematics (1963), ch. 5 · Champion, Proc. Roy. Soc. A 136, 630 (1932).

Provenance. Signature $(+,-,-,-)$ ; $p\cdot p = +m^2c^2$ ; $s$ in energy² units. Base defect corrected: the AllQFT base asserts $\cos(\theta_1+\theta_2) = (\gamma-1)/(\gamma+1)$ , which is wrong — it drops a factor $\cos\theta_1\cos\theta_2$ , and the opening angle is not a function of $\gamma$ alone. The correct invariant is $\tan\theta_1\tan\theta_2 = 2/(\gamma+1)$ , proved symbolically for arbitrary $\theta^{*}$ in verify_aq-sr-relativistic-elastic-collisions.py — which also parses this article and re-derives every number it prints: $\gamma = 2.0658$ , $0.6524$ , $\theta_{\min} = 77.85^\circ$ , $\cos\theta_{\min} = 0.2104$ , the $30^\circ \Rightarrow 48.49^\circ$ event, $\sqrt{s} = 2.3234$ , $E^{*} = 1.1617$ , $p^{*}c = 0.6849$ , $\beta_{\mathrm{cm}} = 0.5896$ , $\gamma_{\mathrm{cm}} = 1.2381$ , the $12.15^\circ$ deficit, and the table $90^\circ / 78.46^\circ / 60.00^\circ / 46.19^\circ / 16.02^\circ$ .

Scope. The base ships Compton scattering as two equation cards and a derivation; that is Compton Scattering's payload and is not derived here. Mandelstam $t$ is not introduced — Mandelstam Invariants owns it. Cross-sections are dynamics: this article predicts where the particles go, never how often.

Builds on. Centre-of-Momentum Frame · Conservation of Four-Momentum · Relativistic Four-Momentum · Relativistic Kinetic Energy. Leads to. Relativistic Inelastic Collisions · Threshold Energy · Compton Scattering · Mandelstam Invariants.

Is the opening angle a fixed number for a given beam energy?

No — and this is the commonest misreading of the result. For a given $\gamma$ the opening angle can be anything in $[\theta_{\min},\ 90^\circ)$ , with $\theta_{\min} = 2\arctan\sqrt{2/(\gamma+1)}$ . Which value you get depends on the CM scattering angle $\theta^{*}$ , which kinematics cannot predict — that is set by the impact parameter and the force law. What is fixed by $\gamma$ alone is the product of the tangents. That is why the product, not the sum, is the thing worth writing down: it is the only combination the dynamics cannot touch.

Where exactly does the Newtonian 90° proof break?

At one line. Momentum conservation gives $p_1^2 = p_3^2 + p_4^2 + 2\vec p_3\cdot\vec p_4$ ; kinetic-energy conservation with $T = p^2/2m$ gives $p_1^2 = p_3^2 + p_4^2$ ; subtracting, $\vec p_3\cdot\vec p_4 = 0$ . The second equation is the culprit. Relativistically $T = \sqrt{(pc)^2+(mc^2)^2} - mc^2$ , so conserving kinetic energy does not say the squared momenta add, and the dot product does not vanish. Note that momentum conservation was never in doubt — what failed was the energy relation. That is the recurring shape of these repairs: relativity rarely breaks a conservation law, it breaks the algebraic relation between the conserved quantities.

Does the sub-90° rule hold for unequal masses?

No, and the derivation shows exactly why not. Two things were used: that both particles have CM energy $\sqrt{s}/2$ (true only for equal masses), and that $\beta_{\mathrm{cm}}E^{*}/c = p^{*}$ (again equal masses only — it says the CM frame moves at exactly the speed the particles have within it). Break either and the tangents no longer multiply to a $\theta^{*}$ -free constant. A light projectile on a heavy target can backscatter, with $\theta_1$ exceeding $90^\circ$ on its own; a heavy projectile on a light target is confined to a forward cone. The equal-mass case is special — which is precisely what makes it such a clean test.

Can an elastic collision transfer all of the beam's kinetic energy?

Yes, for equal masses, and in the configuration Newton would have named. Take $\theta^{*} = \pi$ , a head-on hit in the CM frame: the two particles exchange momenta, and in the lab the projectile is left at rest while the target carries off the whole kinetic energy $T$ . This is the relativistic stop shot, and it survives untouched at any speed because it follows from the symmetry of the equal-mass CM problem rather than from any low-speed approximation. It is also the degenerate corner of the tangent identity, where $\theta_1$ is undefined because the scattered particle has no momentum at all.

Is total kinetic energy really conserved here? I thought only four-momentum was.

Both are true, and the order of the logic matters. Four-momentum conservation holds in every collision and is the fundamental statement. Total kinetic energy is conserved only when total rest energy is unchanged — because $\sum E = \sum T + \sum mc^2$ , so if $\sum mc^2$ is the same before and after, conservation of $\sum E$ forces conservation of $\sum T$ . That mass condition is exactly what "elastic" means. In an inelastic collision the masses change, $\sum mc^2$ rises, and $\sum T$ falls by the same amount — which is the next article, and the reason a system's invariant mass is not the sum of its parts.

The opening angle closes as gamma grows. Do the particles go forward together?

Increasingly, yes — and the cause is the boost, not the collision. In the CM frame the outgoing pair is back-to-back however large $\gamma$ is. What changes is $\gamma_{\mathrm{cm}} = \sqrt{(\gamma+1)/2}$ , which grows without bound, and a large boost rakes everything forward, compressing the whole angular range into a narrow cone. At $\gamma \gg 1$ the minimum opening scales as $\theta_{\min} \approx 2\sqrt{2/\gamma}$ . This forward collimation is why high-energy detectors are long forward-pointing cylinders rather than spheres: at collider energies almost nothing comes out sideways.