Galilean Relativity

In the 1650s Christiaan Huygens drew a picture that belongs on the first page of every relativity book. A boat glides along a canal; a man standing in it swings two balls on strings together, and they collide and rebound, while on the towpath an assistant watches the same collision. The two observers write down different velocities for the same two balls, and they are obliged to agree on what the balls do. Huygens saw that this obligation is not a nuisance but a tool: shift to the frame in which the collision is symmetric, solve it there by inspection, shift back. Out of that trick came the correct laws of elastic impact, half a century before anyone had the modern word for momentum. The machinery of the shift is not ours to build. The mechanics article Relative Motion and Galilean Transformations (`aq-mech-relative-motion-galilean-transformations`) owns it: the coordinate dictionary $\mathbf{r}' = \mathbf{r} - \mathbf{V}t$ with $t' = t$ , the velocity rule $\mathbf{v}' = \mathbf{v} - \mathbf{V}$ , the invariance of acceleration $\mathbf{a}' = \mathbf{a}$ , the covariance $\mathbf{F}' = m\mathbf{a}' = m\mathbf{a} = \mathbf{F}$ , the ten-parameter family of boosts, rotations and translations, and the frame-hopping recipe that boosts by $\mathbf{V}_{\mathrm{cm}} = \mathbf{P}/M$ into the zero-momentum frame. We recall that by name and put it to work; we do not rebuild it. What this article owns is the machinery read as a claim about the world. Galilean relativity says the laws of physics are blind to uniform motion: no experiment sealed inside a smoothly moving laboratory can reveal its own velocity. We prove the claim where it holds — for Huygens' collision law, identical in every inertial frame — then walk it up to the one moving object that breaks it. Feed the same innocent addition rule a beam of light and it hands back $c \pm v$ : a speed of light that depends on who is looking. The rest of special relativity is the repair.

Key takeaways

Galilean relativity: the laws of mechanics take the same form in every inertial frame, so uniform motion is undetectable from inside a closed laboratory — Galileo's ship, 1632.
Its engine is Galilean velocity addition, $u' = u - v$ : a walker at $5\ \mathrm{m/s}$ on a train at $30\ \mathrm{m/s}$ moves at exactly $35\ \mathrm{m/s}$ over the ground.
Differentiate once more and the frame velocity dies: $a' = a$ . Acceleration is a fact all inertial observers share, so $\mathbf{F} = m\mathbf{a}$ survives the boost word for word — galilean invariance.
Huygens' law of elastic collision is therefore frame-independent: the relative velocity reverses, and conservation of momentum and of kinetic energy hold in every inertial frame at once — though the values of $p$ and $K$ differ between frames.
Equal masses, head-on, elastic: the velocities exchange. A ball at $4\ \mathrm{m/s}$ striking an identical stationary ball stops dead and the target leaves at $4\ \mathrm{m/s}$ — the Newton's-cradle click.
Key idea: a Galilean boost subtracts a constant from every velocity and leaves every acceleration, force and conservation law untouched — so mechanics cannot tell you how fast you are moving. Hand the same rule a light beam and it returns $c \pm v$ , a frame-dependent speed of light: the one place where the principle and the arithmetic cannot both survive.

The core insight

A boost to a frame moving at constant velocity $v$ subtracts $v$ from every velocity, leaves every acceleration alone, and therefore leaves Newton's laws — and every conservation law built on them — reading exactly the same. Uniform motion has no mechanical signature. But the same arithmetic predicts a light beam travelling at $c + v$ for an observer running toward it — and that prediction is false.

Where this lives today

Every collider on Earth runs on Huygens' trick. When two proton beams meet, what matters is not the energy in the laboratory but the energy available in the zero-momentum frame — the frame you reach by boosting, exactly as in worked example 1 — and every event recorded at CERN is analysed by transforming into and out of it. The same manoeuvre is routine in molecular-beam chemistry and nuclear scattering; accident reconstruction is its low-speed cousin, and the collision-avoidance systems that keep aircraft apart compute the invariant of worked example 2. The deeper legacy is methodological: "choose the frame where the physics is simplest, solve it there, transform back" survives the revolution. Relativity swaps the Galilean dictionary for the Lorentz one, but the strategy becomes more central, not less — Einstein's 1905 paper is that strategy applied with a straight face to a light beam, precisely where equation (5) ran out of road.

The physical picture

Walk forward along the aisle of a train. To the other passengers you are ambling; to a farmer in a field you are a blur. Neither is wrong — the farmer's number is your number plus the train's, and that is the whole of Galilean velocity addition. It took two centuries and a beam of light to notice that this most natural of additions was an assumption at all.

Figure

Galilean velocity addition — a walker at u′ = 5 m/s on a train at v = 30 m/s reads u = u′ + v = 35 m/s over the ground. The two contributions stack end-to-end on the number line, with no correction term of any kind.

Now seal the walker in a windowless carriage and let her play catch, pour coffee, bounce two billiard balls off each other. Every experiment comes out as it would in a station car park — not because the train is slow, but because the train's speed cancels. Carriage positions differ from ground positions by $vt$ , velocities by $v$ , accelerations by nothing at all: subtract a constant, differentiate twice, and the constant is gone. That is why Huygens could treat a moving boat as a laboratory — and why the punchline figure repays a stare. Drawn as a graph of ground speed against frame speed, the addition rule is an exactly straight line: slope one, forever. Push it far enough and it calmly predicts that a stroll on a fast enough train carries you past the speed of light.

Figure

PUNCHLINE — Galilean addition is exactly linear. Left: the everyday case u = v + 5, with 30 + 5 = 35 m/s dead on the line. Right: the same straight rule for a walker at u′ = 0.5c — it sails clean through the speed of light and predicts u = 1.5c. The dashed curve is the rule that actually holds up there: it bends over and stops at c.

The core equations

Two inertial frames: $S$ (the ground, the towpath, the lab) and $S'$ (the train, the boat), with $S'$ sliding at constant velocity $v$ along the shared $x$ -axis, origins together at $t = 0$ . Classical physics adds one silent clause — a single clock serves them both.

x' = x - vt, \qquad y' = y, \qquad z' = z, \qquad t' = t \qquad (1)

where: $(x,y,z,t)$ — an event's coordinates in $S$ ; $(x',y',z',t')$ — the same event's coordinates in $S'$ ; $v$ — the constant speed of $S'$ relative to $S$ . Physical reading: by time $t$ the moving frame has carried its origin $vt$ down the track, so its position reading is the ground's minus that head start, while $t' = t$ asserts absolute time. The dictionary belongs to the mechanics article named above — here it is the input, not the output. Differentiate the first line with respect to the shared time and the boost passes to velocities. Differentiate once more and the character of the statement changes completely, because the frame velocity is a constant and a constant has no derivative left to give.

u' = u - v \qquad\Longleftrightarrow\qquad u = u' + v, \qquad\qquad a' = a \qquad (2)

where: $u = \mathrm{d}x/\mathrm{d}t$ and $a = \mathrm{d}^2x/\mathrm{d}t^2$ — a particle's velocity and acceleration in $S$ ; $u', a'$ — the same in $S'$ . Physical reading: ground velocity is velocity-in-the-frame plus velocity-of-the-frame ( $5 + 30 = 35$ for the walker), while acceleration is identical in the two frames. Velocity is an opinion; acceleration is a fact — which is what a law of motion needs, since forces prescribe accelerations, not velocities. Classical forces depend on relative positions and relative velocities, and the boost leaves every such difference unchanged, because the $vt$ and $v$ shifts cancel between the two bodies; mass is a scalar. So the second law crosses the translation intact:

F' = ma' = ma = F \qquad (3)

where: $F$ — the net force, computed from the same physics in either frame; $m$ — mass, invariant in classical mechanics. Physical reading: galilean invariance in one line, and Galileo's ship argument in equation form. A physicist sealed in either frame writes identical equations, with no lever to decide whose laboratory is "really" moving. Now the collision law Huygens extracted from that freedom. Two bodies of masses $m_1, m_2$ approach along a line with velocities $v_1, v_2$ and leave with $w_1, w_2$ , elastically, so kinetic energy as well as momentum is conserved. The outcome is forced, and it takes a strikingly frame-friendly shape.

w_2 - w_1 = -(v_2 - v_1), \qquad\text{and for } m_1 = m_2: \quad w_1 = v_2, \quad w_2 = v_1 \qquad (4)

where: $v_2 - v_1$ and $w_2 - w_1$ — the relative velocity of the bodies before and after impact. Physical reading: an elastic collision reverses the relative velocity, whatever the masses, and equal masses simply exchange velocities. Notice what the law contains — only a difference of velocities. Every observer subtracts the same $v$ from both bodies, so every observer computes the same relative velocity and agrees on the law before any calculation is done. That is the structural result this article owns. Finally, the equation that ends the classical world. Apply the velocity rule not to a ball but to a light beam, $u = c$ , and it returns a speed of light that depends on the observer.

u = c \quad\Longrightarrow\quad u' = c - v \qquad (5)

where: $c = 2.998\times10^{8}\ \mathrm{m/s}$ — the speed Maxwell's electrodynamics assigns an electromagnetic wave, built from two laboratory constants in $c = 1/\sqrt{\mu_0\varepsilon_0}$ and naming no frame at all. Physical reading: a chasing observer should measure $c - v$ , an approaching one $c + v$ , so at most one inertial frame could see light obey Maxwell's law in every direction. That frame — the ether — would be detectable, contradicting the very principle equation (3) established. Either the principle fails for electromagnetism, or the addition rule fails for light. The Michelson–Morley Experiment was built to find out; the arc runs on through The Constancy of the Speed of Light and Einstein's Two Postulates.

Where the results come from

Derivation AVelocity addition and the invariance of acceleration

The goal is to extract the two workhorses of this article — the addition rule $u = u' + v$ and the invariance $a' = a$ — from the coordinate dictionary (1) and nothing else, while watching where each classical assumption is spent. Absolute time is consumed in the first step, before any calculus happens; the constancy of $v$ is consumed in the third. Knowing which assumption pays for which conclusion is what lets us predict, later, exactly which conclusion collapses when absolute time turns out to be false.

Step 1. Write the boost for a particle whose ground trajectory is $x(t)$ : in $S'$ it is $x'(t') = x(t) - vt$ , with $t' = t$ . Why: the clause $t' = t$ is not decoration — it lets both observers differentiate with respect to the same variable. Without it, $\mathrm{d}/\mathrm{d}t'$ and $\mathrm{d}/\mathrm{d}t$ are different operators and nothing below survives.

Step 2. Differentiate once with respect to that shared time:

u' = \frac{\mathrm{d}x'}{\mathrm{d}t'} = \frac{\mathrm{d}}{\mathrm{d}t}\big(x - vt\big) = u - v

Why: the linear term $vt$ differentiates to the constant $v$ , so the frames disagree about every velocity by exactly the frame velocity — identically for every particle in the world. That universality is what makes the disagreement harmless.

Step 3. Differentiate again, noting that $v$ is a constant of the problem:

a' = \frac{\mathrm{d}u'}{\mathrm{d}t} = \frac{\mathrm{d}u}{\mathrm{d}t} - \frac{\mathrm{d}v}{\mathrm{d}t} = a - 0 = a

Why: a constant cannot survive a derivative. This is the engine of the whole principle, and where the constancy of $v$ is spent — let $v$ depend on time and an extra term $-\dot v$ appears, the fictitious-force acceleration of a non-inertial frame.

Step 4. Close the loop on dynamics. Under the boost $x_i' - x_j' = x_i - x_j$ and $u_i' - u_j' = u_i - u_j$ , so classical forces — which depend only on such differences — satisfy $F' = F$ ; with $m' = m$ and $a' = a$ , the law $F = ma$ becomes $F' = ma'$ . Why this step is not optional: $a' = a$ alone proves nothing about dynamics — a force law depending on absolute velocity would break the invariance anyway. Form-invariance needs both an invariant acceleration and an invariant force, and it is the second that nature has to grant us. *Verified symbolically in verify_aq-sr-galilean-relativity.py: sympy differentiates $x' = x - vt$ to give $u' = u - v$ and $a' = a$ , confirms relative separations and velocities are boost-invariant, recovers $a' = a - \dot v$ , and reproduces $5 + 30 = 35\ \mathrm{m/s}$ .*

Go deeper · Derivation B · 2 minHuygens: the elastic-collision law is the same in every inertial frame

Derivation BHuygens: the elastic-collision law is the same in every inertial frame

This is the result the article owns. We show that the outcome of a one-dimensional elastic collision computed in the ground frame is exactly what you get by boosting into another inertial frame, solving there, and boosting back — and that momentum and kinetic energy are conserved in both frames at once, although their values differ in each. The proof is Huygens' own: rather than grind through simultaneous quadratics, choose the frame in which the collision is symmetric, where the answer is forced by inspection, then boost back. That frame is the one in which the total momentum vanishes.

Step 1. Take equal masses $m$ , body 1 approaching at $v_1$ , body 2 at rest. Boost into the frame moving at $V = \tfrac12(v_1 + v_2)$ ; for equal masses this is exactly the zero-momentum frame, since $P' = m(v_1 - V) + m(v_2 - V) = 0$ . In it the bodies approach with equal and opposite velocities $\pm\tfrac12(v_1 - v_2)$ . Why: the trick is choosing the frame in which the problem has a symmetry it did not visibly have before. Nothing was done to the bodies — only the bookkeeping moved.

Step 2. In that frame invoke Huygens' one postulate: two equal bodies meeting head-on at equal and opposite speeds rebound at equal and opposite speeds, so each velocity reverses. Boost back by adding $V$ and body 1 leaves with $w_1 = -\tfrac12(v_1 - v_2) + V = v_2$ , body 2 with $w_2 = v_1$ — the velocities have been exchanged, which is equation (4). Why the postulate is forced rather than assumed: momentum must stay zero, so the final velocities are $\pm w$ , and kinetic energy must be unchanged, which fixes $|w|$ at the incoming speed. The exchange then follows because reflecting two numbers about their mean swaps them.

Step 3. Now show the law itself is frame-blind. Apply (2) to each body and subtract: the relative velocity transforms as $(v_2 - v) - (v_1 - v) = v_2 - v_1$ — a Galilean invariant. So "the relative velocity reverses" reads identically in every inertial frame, for any masses. Why this is the deep point: a law written purely in relative velocities cannot betray the frame it was written in. Huygens never had to check his result on the towpath; the form of the law guaranteed it.

Step 4. Audit the conservation laws in two frames. Momentum in the boosted frame is $P' = \sum_i m_i(v_i - v) = P - Mv$ , and mass is conserved in the collision, so that shift is the same before and after: $\Delta P' = \Delta P = 0$ . Kinetic energy is $K' = \sum_i \tfrac12 m_i (v_i - v)^2 = K - Pv + \tfrac12 Mv^2$ ; again $P$ , $M$ and $v$ are untouched by the collision, so $\Delta K' = \Delta K = 0$ . Why this settles a famous confusion: the two observers write down different momenta and different kinetic energies, and both are right. They share not the values but the verdicts — an elastic collision is elastic for everybody. *Verified in verify_aq-sr-galilean-relativity.py: sympy solves momentum and kinetic-energy conservation for general masses and returns $w_2 - w_1 = -(v_2 - v_1)$ and, for $m_1 = m_2$ , the exchange $w_1 = v_2$ , $w_2 = v_1$ ; the numbers $4.0 \to 0$ and $0 \to 4.0\ \mathrm{m/s}$ are checked for conservation of $p$ and $K$ in the ground frame and in the frame boosted at $2.0\ \mathrm{m/s}$ , with the shift identity $K' = K - Pv + \tfrac12 Mv^2$ .*

Limiting cases

Case	Result	Meaning
Low speeds, $\beta = v/c \ll 1$	addition is exact to the precision of any mechanical measurement	on Earth $\beta \lesssim 10^{-4}$ , so corrections enter at $\beta^2 \lesssim 10^{-8}$
The formal limit $c \to \infty$	absolute time and $u = u' + v$ become exactly true	the world you would inhabit if light were infinitely fast — a limit, not a blunder
Equal masses, elastic, head-on	$w_1 = v_2$ , $w_2 = v_1$	the Newton's-cradle click; a special case of the relative-velocity reversal
A light beam, $u = c$	$u' = c - v$ , frame-dependent	the rule breaks — see The Constancy of the Speed of Light
Frame accelerates, $v = v(t)$	$a' = a - \dot v$	the frame is no longer inertial; fictitious forces appear

Common trap“The kinetic energy has a different value in the boosted frame, so it can’t be conserved there as well”

This is the commonest stumble in frame-hopping, and it confuses a value with a verdict. In the worked example below the ground observer measures a total kinetic energy of $4.0\ \mathrm{J}$ before the collision and $4.0\ \mathrm{J}$ after; the observer boosted at $2.0\ \mathrm{m/s}$ measures $2.0\ \mathrm{J}$ before and $2.0\ \mathrm{J}$ after. The totals differ by a factor of two — and both observers see energy conserved. Step 4 of the Huygens derivation says why: a boost shifts the kinetic energy by $-Pv + \tfrac12 Mv^2$ , built entirely from constants of the process, so it is the same before as after and cancels out of every change. Values are frame-dependent bookkeeping; conservation is frame-independent physics — compare before with after within each frame, never energies across frames.

Worked examples

1 · The Newton's-cradle click, computed the way Huygens computed it

Two identical steel balls, $m = 0.50\ \mathrm{kg}$ each. Ball 1 rolls at $v_1 = 4.0\ \mathrm{m/s}$ into ball 2, at rest, head-on and elastically. Find the final velocities by a frame boost, then confirm that momentum and kinetic energy are conserved in the ground frame and in a frame moving at $2.0\ \mathrm{m/s}$ .

Step 1. Choose the frame that makes the problem symmetric. The total momentum is $P = (0.50)(4.0) + (0.50)(0) = 2.0\ \mathrm{kg\,m/s}$ and the total mass is $M = 1.0\ \mathrm{kg}$ , so the zero-momentum frame moves at $V = P/M = \boxed{2.0\ \mathrm{m/s}}$ — for equal masses, the mean of the two velocities.

Step 2. Boost into it with equation (2): $v_1' = 4.0 - 2.0 = +2.0\ \mathrm{m/s}$ and $v_2' = 0 - 2.0 = -2.0\ \mathrm{m/s}$ . The balls approach at equal and opposite speeds — Huygens' symmetric collision — and indeed $P' = (0.50)(+2.0) + (0.50)(-2.0) = 0$ .

Step 3. Collide, then boost back. In the symmetric frame the only outcome consistent with zero momentum and unchanged kinetic energy is a clean reversal, $w_1' = -2.0$ and $w_2' = +2.0\ \mathrm{m/s}$ ; adding $V = 2.0\ \mathrm{m/s}$ returns $w_1 = \boxed{0}$ and $w_2 = \boxed{4.0\ \mathrm{m/s}}$ . The striker stops dead and the target leaves with its whole velocity — the exchange promised by equation (4), with no algebra anywhere.

Step 4. Audit in the ground frame. Momentum: $2.0\ \mathrm{kg\,m/s}$ before, $(0.50)(0) + (0.50)(4.0) = \boxed{2.0\ \mathrm{kg\,m/s}}$ after. Kinetic energy: $\tfrac12(0.50)(4.0)^2 = \boxed{4.0\ \mathrm{J}}$ before, the same after. Both conserved.

Step 5. Audit again in the boosted frame, where every number is different. Momentum: $\boxed{0}$ before and after. Kinetic energy: $2 \times \tfrac12(0.50)(2.0)^2 = \boxed{2.0\ \mathrm{J}}$ before and after — half the ground-frame total, and just as perfectly conserved. Different values, identical verdicts; and the relative velocity flips from $+4.0$ to $-4.0\ \mathrm{m/s}$ in both frames, because it is a Galilean invariant.

2 · Closing speed on a two-lane road

One car travels east at $25\ \mathrm{m/s}$ , another west at $20\ \mathrm{m/s}$ , both relative to the road, and they are $200\ \mathrm{m}$ apart. How fast is one approaching the other, and how long until they meet?

Step 1. Ride in the eastbound car, so $S'$ is that car and $v = +25\ \mathrm{m/s}$ with east positive; the westbound car has ground velocity $u = -20\ \mathrm{m/s}$ . Apply equation (2): $u' = u - v = (-20) - (+25) = \boxed{-45\ \mathrm{m/s}}$ . The other car rushes at him at $45\ \mathrm{m/s}$ — the closing speed, the same subtraction that gave $35\ \mathrm{m/s}$ for the walker, with different signs. Sitting in the westbound car instead gives $(+25) - (-20) = +45\ \mathrm{m/s}$ : the same magnitude, because relative velocity is a Galilean invariant.

Step 2. Collapse the gap in whichever frame you like. In the eastbound car's frame $200\ \mathrm{m}$ shrinks at $45\ \mathrm{m/s}$ , so $t = 200/45 = \boxed{4.44\ \mathrm{s}}$ ; in the road frame it closes at $25 + 20 = 45\ \mathrm{m/s}$ , giving the same $4.44\ \mathrm{s}$ — as it must, since time intervals are Galilean invariants too. Two frames, one answer: the arithmetic behind every airborne collision-avoidance alert.

Physics of the portrait

Galileo gave physics the principle; Huygens gave it the first serious proof of concept. In the decades after Galileo's death the theory of impact was a shambles — Descartes had published seven rules of collision, most of them wrong. Huygens dismantled them with one idea: a collision seen from a moving boat must agree with the same collision seen from the bank. It is the earliest instance in physics of a symmetry argument replacing a calculation — and still how a particle physicist attacks a collision three centuries later.

Origins · 1 minChristiaan Huygens and how this was discovered

Christiaan Huygens

1629–1695 · The Hague · Paris

Huygens was the most formidable physicist between Galileo and Newton, and the first to turn the relativity of motion into a working method. Between 1652 and 1656 he worked out the correct laws of elastic impact in a manuscript, De Motu Corporum ex Percussione, using the relativity principle as his engine: he imagined the collision performed by a man on a canal boat while an assistant on the bank watched, and demanded that the two accounts agree. From the one symmetric case he could solve outright — equal balls meeting at equal and opposite speeds, which simply rebound — the boost-and-return trick generated the general law, and with it two results a century ahead of their time: that the relative velocity of the bodies reverses, and that $\sum m v^2$ is conserved in an elastic collision (Leibniz's vis viva, our kinetic energy). Cautious to a fault, he did not publish; his rules reached the world in 1669 through the Royal Society's competition on impact, and the treatise itself only in 1703. The irony is exquisite: the man who first exploited the relativity of motion also handed the nineteenth century the wave theory of light, whose speed that same relativity could not accommodate.

History

1632

— Galileo's Dialogo argues from a ship's sealed cabin that shared uniform motion is undetectable — the relativity principle.

1656

— Huygens completes De Motu Corporum ex Percussione, deriving the elastic-collision laws by shifting between a moving boat and the canal bank; he leaves it unpublished.

1668

— The Royal Society sets the laws of impact as a problem; Wallis, Wren and Huygens each send in solutions.

1669

— Huygens' collision rules appear in the Journal des Sçavans and in the Philosophical Transactions.

1687

— Newton's Principia, Corollary V: the motions of bodies among themselves are the same whether their space is at rest or moves uniformly.

1703

— De Motu Corporum ex Percussione is printed posthumously in Huygens' Opuscula postuma.

Going further

Texts. Taylor and Wheeler, Spacetime Physics, 2nd ed., ch. 1 and 3 · Rindler, Introduction to Special Relativity, 2nd ed., ch. I · Morin, Introduction to Classical Mechanics, ch. 5 · Landau and Lifshitz, Mechanics, Vol. 1, §3 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.1.

Provenance. Everything here follows from the boost $x' = x - vt$ , $t' = t$ and is re-derived in `verify_aq-sr-galilean-relativity.py`, which reproduces $u' = u - v$ , $a' = a$ and $a' = a - \dot v$ for a non-inertial frame (Rindler ch. I; Landau and Lifshitz §3); the elastic-collision solution, the reversal $w_2 - w_1 = -(v_2 - v_1)$ and the equal-mass exchange (Huygens 1656/1669; Morin ch. 5); and every printed number — $5 + 30 = 35\ \mathrm{m/s}$ ; the collision $4.0 \to 0$ , $0 \to 4.0\ \mathrm{m/s}$ with $p = 2.0\ \mathrm{kg\,m/s}$ and $K = 4.0\ \mathrm{J}$ conserved in the ground frame while $p' = 0$ and $K' = 2.0\ \mathrm{J}$ are conserved in the frame boosted at $2.0\ \mathrm{m/s}$ ; the closing speed $45\ \mathrm{m/s}$ with $200/45 = 4.44\ \mathrm{s}$ ; and Galilean addition applied to light returning $c - v$ . Constants: $c = 2.998\times10^{8}\ \mathrm{m/s}$ , Earth's orbital $\beta \approx 9.94\times10^{-5}$ .

Builds on. Inertial Frames of Reference · The Principle of Relativity. Leads to. Galilean Transformations · Symmetries of Galilean Spacetime · Absolute Space and Absolute Time · Electromagnetism and the Failure of Galilean Invariance · Einstein's Two Postulates.

If motion is relative, how does the collision "know" to exchange the velocities rather than do something else?

It knows nothing — the exchange is forced by two conservation laws that hold in every frame. Frame-hopping just makes it feel less arbitrary. Boost into the frame where the balls approach at equal and opposite speeds, and the only outcome keeping momentum at zero and kinetic energy unchanged is a clean reversal; boosting back turns that reversal into an exchange, because reflecting two numbers about their mean swaps them. What is relative is the description — a stop and a start on the ground, a symmetric bounce in the boosted frame — while the physics, the reversal of the relative velocity, is identical in both.

The kinetic energy is 4.0 J in one frame and 2.0 J in another. Which is the real one?

Both, and there is no further answer. Kinetic energy is a property not of a body but of a body and a frame. A boost shifts the total by $-Pv + \tfrac12 Mv^2$ , and since $P$ , $M$ and $v$ are constants of a closed process, that shift is identical before and after the collision — so every observer's books balance, at different totals. If you feel the pull of "one must be the true energy", notice that the same pull would demand a true velocity, hence a true rest frame — exactly what Galilean relativity forbids. The invariant content is the change, and every frame agrees it is zero.

Is Galilean relativity simply wrong, then?

No — it is exact in a limit and superbly accurate outside it. Every mechanical result on this page is right to the last digit any experiment can measure; no laboratory has caught $a' = a$ or the collision law failing. What fails is the extension of velocity addition to light, and the failure hides at ordinary speeds because it is governed by $\beta = v/c$ , about $10^{-4}$ even for the Earth in orbit, with corrections at $\beta^2 \approx 10^{-8}$ . Read Galilean relativity as the $c \to \infty$ theory — the physics of a universe where signals could be infinitely fast, so a universal "now" would make sense. Ours very nearly is that universe.

Exactly what does the principle of relativity forbid?

It forbids any experiment whose result depends on the uniform velocity of the laboratory. It does not forbid detecting acceleration or rotation: an accelerometer works with the blinds drawn, a Foucault pendulum reveals the Earth's spin, coffee slides across the dashboard when you brake. The line runs exactly where equation (2) runs — a constant $v$ dies under the second derivative, a time-dependent $v$ does not, and the surviving $-\dot v$ is the fictitious force. Velocity is undetectable from inside; acceleration announces itself. Any claim to have measured "our absolute speed", through the ether or otherwise, is a claim to have broken the principle.

Why is the collision law credited to Huygens rather than to Galileo or Newton?

Galileo supplied the principle but never applied it to impact, and Newton's Principia came three decades after Huygens had the answer. Huygens' contribution was to see that the relativity of motion could be used as a computational instrument — that a collision you cannot solve becomes one you can solve if you view it from the right boat. From that idea alone he derived the reversal of relative velocity and the conservation of $\sum m v^2$ , in a manuscript completed in 1656, then sat on it for over a decade, publishing a summary only in 1669 when the Royal Society made a competition of the problem.

If the addition rule is what breaks for light, why does everyone say relativity is about time?

Because the addition rule is not an independent assumption — it is a consequence, and the derivation above shows which assumption it rests on. Step 1 spends the clause $t' = t$ : only because the observers share a clock can they differentiate along the same variable and conclude $u' = u - v$ . A velocity rule that is wrong for light is the symptom; the disease is absolute time. That is why the repair is not a patched-up velocity formula but a new coordinate dictionary in which $t'$ depends on position as well as on $t$ — so observers in relative motion stop agreeing about which events are simultaneous.