RELATIVISTIC VELOCITY ADDITION

In Galilean kinematics, velocities add. Stand on a flatbed truck moving at 30 m/s and roll a ball forward at 5 m/s. The road observer sees 35 m/s. Two centuries of mechanics rest on this arithmetic fact, encoded in the Galilean invariance of Newton's laws.

§02.4 is where that arithmetic breaks. The replacement, due to in 1905 and equivalently to in the 1904 algebra Einstein had read, is

Two velocities combine through this rational function rather than through addition. The denominator is the entire story. At everyday speeds, $u' v / c^2$ is negligible — for two cars at 30 m/s the correction is one part in $10^{16}$ — and the formula collapses back to $u = u' + v$. At cosmic speeds it does the work all of §02 has been pointing toward: it caps the universe.

Three facts come welded together. First, the formula reduces to Galileo at low speed — special relativity does not contradict ordinary kinematics, it contains it. Second, two sub-luminal velocities never compose to a super-luminal one. Third — and this is the postulate, encoded as algebra — $c + v = c$ for every $v$. Light, uniquely, refuses to add.

The canonical demonstration is two rockets. Earth watches Rocket A pull away at 0.6 c. Rocket A then fires Rocket B forward at 0.6 c in its own frame. Galilean kinematics demands Earth see Rocket B at 1.2 c. Einstein's formula gives $(0.6 + 0.6) / (1 + 0.36) = 1.2 / 1.36 \approx 0.882$. Earth sees Rocket B at 0.882 c. Half a c of arithmetic difference, and only one answer is what nature delivers.

Push the slider with both β's toward 0.99: Galileo's ghost runs off the canvas, Einstein's answer creeps closer to the c-wall but never reaches it. That asymptotic approach is the kinematic content of "no matter how hard you push, you cannot make a massive object cross c". The function $(u' + v) / (1 + u' v / c^2)$ maps the open box $(-c, c)^2$ smoothly onto $(-c, c)$. The boundary is approached, never crossed.

What if one of the inputs is $c$? Plug $u' = c$. The numerator is $c + v$. The denominator is $1 + v/c = (c + v)/c$. The ratio is exactly $c$. The formula has just enforced Einstein's second postulate. Light moves at $c$ in every inertial frame — and the postulate falls out of the algebra once the formula is in hand. Not an extra constraint: a corollary.

A sharper way to feel that. Put a flashlight on a spaceship. The ship flies past Earth at velocity $v$ rightward, the flashlight points forward, at $t = 0$ the captain switches it on. Galilean kinematics demands that the photons — which left the bulb at $c$ in the ship's frame — are seen on Earth at $c + v$. For a ship at 0.6 c that would be 1.6 c. Galilean velocity addition, applied to light, would give faster-than-light photons emitted by every moving lamp. The Michelson-Morley null already said, in 1887, that this is not what happens.

The cyan photon front always reaches the right-edge mark at the same lab-time $L/c$, regardless of how fast the ship is moving. The amber Galilean ghost reaches it at $L/(c+v)$, sooner — at high $v$ the ghost is already off the canvas while the actual photon is still mid-flight. The c-wall is not a property of light, it is a property of the spacetime. Einstein elevated that fact in 1905 from "what Michelson and Morley measured" to a postulate of nature: the laws of physics, including the value of $c$, are the same in every inertial frame.

The historical hint that classical addition was wrong sat on a lab bench fifty-four years before special relativity. in 1851 measured the speed of light in moving water. The setup was an interferometer with two arms; one passed light along a flowing water column, the other against. Recombining the beams produced fringes; by rotating which arm went with the flow, Fizeau shifted the fringes and extracted, from the shift, the velocity at which light propagated through the moving water.

The question was whether the water "drags" the light. Three answers were on the table. (i) Full drag — the water carries its luminiferous medium with it, $u = c/n + v_{\text{water}}$ (Stokes). (ii) No drag — the aether is fixed, $u = c/n$ regardless. (iii) Partial drag — Fresnel's 1818 ad-hoc proposal that light picks up a fraction $1 - 1/n^2$ of the medium's velocity, on the grounds that the aether was "partially blocked" by optical density.

Fizeau measured option (iii). The drag coefficient came out to $1 - 1/n^2$ within 1%. For water ($n = 1.33$) the coefficient is $\approx 0.43$. Fresnel's ad-hoc factor was vindicated empirically. Nineteenth-century optics treated this as a fine-tuned property of the aether.

The relativistic interpretation is structural. Feed the velocity-addition formula $u' = c/n$ (light in the water's rest frame) and $v = v_{\text{water}}$, and expand to first order:

The Fresnel coefficient falls straight out. It was never a quirk of how the aether interacted with optical media — it was the v/c-order term of relativistic kinematics. Fresnel guessed it from the wrong physics. Fizeau measured it from a moving water tube. Einstein derived it, fifty-four years later, as a corollary of the postulate that there is no aether at all.

There is a structural argument for why velocity addition cannot be addition. Suppose two boosts compose by some function $f(u', v)$ mapping $(-c, c)^2 \to (-c, c)$. Three constraints fix its shape: (i) Galilean limit, $f(u', v) \approx u' + v$ for $|u'|, |v| \ll c$; (ii) light is invariant, $f(c, v) = c$ for every sub-luminal $v$; (iii) boosts form a group, $f(f(u', v), w) = f(u', f(v, w))$ — composition is associative.

Up to overall scale, the unique smooth function satisfying all three is $f(u', v) = (u' + v)/(1 + u' v / c^2)$. The constraints over-determine it. The Galilean limit fixes linear-in-v behaviour; c-invariance fixes the denominator's pole; the group law fixes the denominator to be exactly $1 + u' v / c^2$. The formula is forced — and it is forced by the postulates, not by empirical fit.

This is the same argument-from-symmetry that produces the Lorentz transformation. Velocity addition is what that transformation says about how velocities, rather than coordinates, transform — the same group law in different costumes. (For the mathematician: the formula is the rapidity-additive structure of the boost subgroup, transported back to velocity coordinates via $u = c \tanh \phi$. The bounded range of $\tanh$ is geometrically why velocities live in the open ball $|u| < c$.)

$c + c = c$ is the universe's hard ceiling. It is hard because it is a property of the metric, not of any apparatus. Build a faster engine, push a faster particle, fire a brighter laser — the formula refuses to return a number bigger than $c$. The 0.882 c answer in the two-rocket scene is not a friction loss or fuel limit. It is the kinematic content of being alive in 4-dimensional Minkowski space.

§02.5 observes how the same algebra reshapes frequencies. A source moving toward you at 0.5 c does not Doppler-shift by the classical $(1 + \beta)$ — it shifts by $\sqrt{(1+\beta)/(1-\beta)}$, and the difference is exactly the relativistic time-dilation factor that has been propagating through every section since §02.1. By the close of §02 we will have seen four shadows of the same 4D rotation: time, length, velocity, frequency.