Rapidity and the Hyperbolic Form of Boosts

Velocities do not add. That is the first genuinely disorienting thing a student of relativity learns, and it never quite stops being disorienting, because addition is the one operation we trust. Run at $0.5c$ inside a rocket that is already doing $0.5c$ and you are not doing $c$ ; you are doing $0.8c$ . Do it again and you get $0.928\,571c$ . Again: $0.975\,610c$ . The numbers crawl toward the light barrier and never arrive, and the rule that produces them — $(\beta_1+\beta_2)/(1+\beta_1\beta_2)$ — looks like an arbitrary tax levied on arithmetic by an unsympathetic universe. It is not arbitrary. There is a variable in which boosts do add, plainly and exactly, and it is not the velocity. It is the rapidity, $\varphi = \operatorname{arctanh}\beta$ . Boost by $\varphi_1$ , then by $\varphi_2$ , and you have boosted by $\varphi_1 + \varphi_2$ — no correction term, no denominator, no tax. And the velocity-addition law turns out to be the identity $\tanh(\varphi_1+\varphi_2) = (\tanh\varphi_1 + \tanh\varphi_2)/(1 + \tanh\varphi_1\tanh\varphi_2)$ : a formula that appears in tables of hyperbolic functions and has nothing to do with physics. Relativity did not invent a strange way of adding velocities. It told us that velocity was the wrong coordinate, and handed us the right one. And once you are in the right coordinate, the light barrier stops being mysterious. Rapidity runs over the whole real line: $\varphi$ can be $1$ , or $10$ , or $10^{100}$ , and every one of those is a physically realisable boost. But $\tanh$ maps all of $\mathbb{R}$ into the open interval $(-1, 1)$ — it approaches $\pm 1$ and never touches. So no finite sum of rapidities can ever produce $\beta = 1$ , not because something intervenes to stop you, but because the answer is not in the codomain. The cosmic speed limit is the range of a function.

Key takeaways

Define the rapidity $\varphi$ by $\beta = \tanh\varphi$ , equivalently $\varphi = \operatorname{arctanh}\beta = \tfrac12\ln\!\frac{1+\beta}{1-\beta}$ . Then $\gamma = \cosh\varphi$ and $\gamma\beta = \sinh\varphi$ .
The additivity theorem (this article's core): $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1 + \varphi_2)$ . Rapidities add. Exactly, with no correction term, for collinear boosts.
The velocity-composition law is the $\tanh$ addition formula. $\beta_1 \oplus \beta_2 = \tanh(\varphi_1+\varphi_2)$ — not a rule to memorise, an identity in disguise.
The ceiling is the range of a function. $\tanh:\mathbb{R}\to(-1,1)$ maps onto the open interval, so any finite $\varphi$ gives $|\beta| < 1$ strictly. $N$ boosts of $\beta_1 = 0.5$ give exactly $\beta_N = \dfrac{3^N-1}{3^N+1}$ — $0.5$ , $0.8$ , $0.928571$ , $0.975610$ , … — with $1 - \beta_N = 2/(3^N+1)$ : exponentially close to $c$ , never equal to it.
Light sits at $\varphi = \infty$ , and a finite boost adds a finite rapidity, so the invariance of $c$ is the statement that infinity is a fixed point of addition.
Key idea: relativity is linear in the right variable. Rapidity is that variable — it is the additive parameter of the boost group, velocity is merely its hyperbolic tangent, and every "strange" feature of relativistic kinematics is the strangeness of that one function.

The core insight

Rapidity is the angle of a boost: the quantity that adds when you compose two boosts, exactly as ordinary angles add when you compose two rotations. Velocity does not add because velocity is not the angle — it is $\tanh$ of the angle, and $\tanh$ is not linear. The speed limit is then the elementary fact that $\tanh$ never reaches $1$ .

Where this lives today

Rapidity is not a pedagogical curiosity; it is the working coordinate of experimental particle physics. Every LHC analysis bins its data in rapidity $y$ rather than in velocity or angle, for the reason established above: a longitudinal boost — and the colliding partons always deliver one, since they carry unequal momentum fractions — merely translates $y$ , so distribution shapes and rapidity differences are boost-invariant while velocities and angles are not. Detector acceptance is quoted as a range of pseudorapidity ( $|\eta| < 2.5$ for the ATLAS inner tracker) for the same reason. The $7\ \mathrm{TeV}$ beam's rapidity, $9.61$ , is a more useful description of the machine than $\beta = 0.999999991$ . The additive structure also reaches places velocity cannot. Motion under constant proper acceleration $a$ is simply linear in rapidity: $\varphi(\tau) = a\tau/c$ , so $\beta(\tau) = \tanh(a\tau/c)$ and $\gamma(\tau) = \cosh(a\tau/c)$ . The hyperbolic worldline that underlies the Rindler horizon and the Unruh effect is exactly the statement that rapidity accumulates at a constant rate on the traveller's own clock while velocity saturates. The rocket's engine never stops working, the rapidity climbs forever, and the velocity does not reach $c$ .

The physical picture

Consider what a boost actually does to the pair $(ct, x)$ . It mixes them, linearly, in a way that leaves $s^2 = (ct)^2 - x^2$ untouched — that invariance belongs to The Invariant Spacetime Interval, and the geometric account of the boost as an isometry of $s^2$ belongs to Lorentz Boosts as Hyperbolic Rotations. Take both as given. What concerns us here is a question of parametrisation: which number should we use to label a boost? The obvious candidate is $\beta$ . It is the thing we measure. But it is a bad label, and the tell is that it composes badly: label boosts by $\beta$ and the composition of two of them is that ugly quotient. A good label composes by addition. Rotations have one — the angle. Do boosts? They do, and finding it takes one line. The two numbers $\gamma$ and $\gamma\beta$ obey $\gamma^2 - (\gamma\beta)^2 = \gamma^2(1-\beta^2) = 1$ identically — which is the hyperbolic Pythagorean identity $\cosh^2\varphi - \sinh^2\varphi = 1$ staring back at us. So there is exactly one real $\varphi$ with $\gamma = \cosh\varphi$ and $\gamma\beta = \sinh\varphi$ , and dividing, $\beta = \tanh\varphi$ . That number is the rapidity, and the rest of the article is the discovery that it is precisely the additive label we were looking for.

Figure

Two boosts of β = 0.5 composed. Multiply the matrices and the cosh/sinh ADDITION FORMULAE fall out: the product of Λ(φ₁) with itself is Λ(2φ₁), with cosh 2φ₁ = cosh²+sinh² = 1.3333 + 0.3333 = 1.6667 and sinh 2φ₁ = 2 sinh cosh = 1.3333. Read the composed velocity off the bottom: β = sinh/cosh = 1.3333/1.6667 = 0.8 — exactly what the velocity-addition law gives, because it IS the tanh addition formula

The core equations

Signature $(+,-,-,-)$ : $s^2 = c^2\Delta t^2 - \Delta x^2 - \Delta y^2 - \Delta z^2$ , timelike-positive. The Lorentz transformation itself is derived in The Lorentz Transformation Equations and is taken as given; what follows is a change of variable inside it.

\varphi \;\equiv\; \operatorname{arctanh}\beta \;=\; \frac{1}{2}\ln\!\left(\frac{1+\beta}{1-\beta}\right), \qquad \beta = \tanh\varphi \qquad (1)

where: $\varphi$ — the rapidity, dimensionless, running over all of $\mathbb{R}$ ; $\beta = v/c \in (-1, 1)$ . Physical reading: a relabelling of the velocity, and a faithful one — the map is a bijection, so nothing is lost and nothing is added. What changes is how boosts compose. Note the logarithmic form: as $\beta \to 1$ the argument diverges, so $\varphi \to \infty$ . The light barrier sits at infinite rapidity.

\gamma = \cosh\varphi, \qquad \gamma\beta = \sinh\varphi, \qquad \cosh^2\varphi - \sinh^2\varphi = 1 \qquad (2)

where: the rapidity dictionary. Physical reading: the identity on the right is not an analogy with $\gamma^2(1-\beta^2) = 1$ — it is that identity, rewritten. That is what licenses the substitution: the pair $(\gamma, \gamma\beta)$ lies on the unit hyperbola, and every point of its forward branch is $(\cosh\varphi, \sinh\varphi)$ for exactly one real $\varphi$ .

\Lambda(\varphi) = \begin{pmatrix} \cosh\varphi & -\sinh\varphi \\ -\sinh\varphi & \cosh\varphi \end{pmatrix} \qquad (3)

where: the boost matrix, acting on $(ct, x)^{\mathsf T}$ , in the rapidity variable. Physical reading: substitute (2) into the usual $\gamma\begin{pmatrix}1 & -\beta\\ -\beta & 1\end{pmatrix}$ and this is what you get — the same matrix, better labelled. Its resemblance to a rotation matrix with $\cos \to \cosh$ , $\sin \to \sinh$ is exact and not a coincidence; the geometry behind that correspondence is developed in Lorentz Boosts as Hyperbolic Rotations and is not re-derived here.

\Lambda(\varphi_1)\,\Lambda(\varphi_2) = \Lambda(\varphi_1 + \varphi_2) \qquad (4)

where: the additivity theorem — the result this article owns. Physical reading: collinear boosts compose by adding their rapidities, exactly as rotations about a common axis add their angles. That is why rapidity, not velocity, is the natural parameter of the boost group.

\beta_1 \oplus \beta_2 \;=\; \tanh(\varphi_1 + \varphi_2) \;=\; \frac{\tanh\varphi_1 + \tanh\varphi_2}{1 + \tanh\varphi_1 \tanh\varphi_2} \;=\; \frac{\beta_1 + \beta_2}{1 + \beta_1\beta_2} \qquad (5)

where: the velocity-composition law, whose derivation belongs to Relativistic Velocity Addition. Physical reading: the physics is on the far left, the formula everyone memorises is on the far right, and the middle step is a hyperbolic identity known to mathematicians long before relativity existed. The composition law is not a law. It is $\tanh$ of a sum.

Rapidity, and why it is the right variable

Derivation AThe definition: the pair (γ, γβ) lives on a hyperbola, so it has an angle

Why arctanh, of all things? Not by inspiration, and not by analogy — it is forced. The hyperbolic functions are not imported from outside; they are discovered inside the Lorentz factor, and the derivation uses nothing but the definition of $\gamma$ .

Step 1. Start from $\gamma = 1/\sqrt{1-\beta^2}$ and square-and-rearrange: $\gamma^2(1-\beta^2) = 1$ , i.e. $\gamma^2 - (\gamma\beta)^2 = 1$ . Why: this is the single most useful identity in special relativity, and here it is doing something new — it says the point $(\gamma,\ \gamma\beta)$ is constrained to the unit hyperbola $X^2 - Y^2 = 1$ , for every possible velocity.

Step 2. Compare with $\cosh^2\varphi - \sinh^2\varphi = 1$ , which holds for every real $\varphi$ . Since $\gamma \ge 1 > 0$ , our point is on the forward branch, and the map $\varphi \mapsto (\cosh\varphi, \sinh\varphi)$ is a bijection from $\mathbb{R}$ onto that branch. Hence there is exactly one real $\varphi$ with $\gamma = \cosh\varphi$ and $\gamma\beta = \sinh\varphi$ . Why: uniqueness matters. The rapidity is not a convenient substitution one might or might not make; it is the unique coordinate on the set of boosts, delivered by the identity itself.

Step 3. Divide the two: $\dfrac{\sinh\varphi}{\cosh\varphi} = \dfrac{\gamma\beta}{\gamma} = \beta$ , so $\beta = \tanh\varphi$ and $\varphi = \operatorname{arctanh}\beta$ . Writing $\operatorname{arctanh}$ out, $\varphi = \tfrac12\ln\frac{1+\beta}{1-\beta}$ . Why: and now the ceiling is already visible. As $\beta \uparrow 1$ the argument of the log diverges, so $\varphi \to +\infty$ — a light pulse has infinite rapidity, which is a much more informative statement than " $\gamma$ blows up".

Step 4. Substitute the dictionary into the boost matrix: $\gamma\begin{pmatrix}1 & -\beta \\ -\beta & 1\end{pmatrix} = \begin{pmatrix}\gamma & -\gamma\beta \\ -\gamma\beta & \gamma\end{pmatrix} = \begin{pmatrix}\cosh\varphi & -\sinh\varphi \\ -\sinh\varphi & \cosh\varphi\end{pmatrix}$ , which is equation (3). Why: nothing has been assumed and nothing has been changed — the same boost, in a different label. Whether this matrix "is" a rotation through an imaginary angle, and what geometry that reflects, is a separate question, and its owner is Lorentz Boosts as Hyperbolic Rotations.

Step 5. Sanity-check the small- $\varphi$ limit: $\tanh\varphi = \varphi - \varphi^3/3 + \dots$ , so for $\varphi \ll 1$ , $\beta \approx \varphi$ — rapidity is velocity to first order, and at $\varphi = 0.1$ , $\beta = 0.099668$ . Why: this is why nobody needed rapidity before 1905. In the Galilean regime the two labels agree, velocities add, and the distinction is invisible. *Verified in verify_aq-sr-rapidity-hyperbolic-boosts.py: $\gamma^2 - (\gamma\beta)^2 = 1$ symbolically; $\cosh(\operatorname{arctanh}\beta) = \gamma$ and $\sinh(\operatorname{arctanh}\beta) = \gamma\beta$ symbolically; the matrix substitution of Step 4 entrywise; strict monotonicity ( $\operatorname{sech}^2\varphi > 0$ ); and $\tanh(0.1) = 0.099668$ .*

Go deeper · Derivation B · 2 minTHE ADDITIVITY THEOREM — and the discovery that velocity addition is a trigonometric identity

Derivation BTHE ADDITIVITY THEOREM — and the discovery that velocity addition is a trigonometric identity

This is the result the article owns, and it deserves to be stated as a theorem rather than an observation: for collinear boosts, $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1+\varphi_2)$ . The rapidity is the additive parameter of the one-dimensional boost group — the exact counterpart of the angle for rotations about a fixed axis. The payoff is not aesthetic: the velocity-addition law then falls out in one line, and stops being an independent piece of physics you have to remember.

Step 1. Multiply the matrices of equation (3):

\Lambda(\varphi_1)\Lambda(\varphi_2) = \begin{pmatrix} \cosh\varphi_1\cosh\varphi_2 + \sinh\varphi_1\sinh\varphi_2 & -(\sinh\varphi_1\cosh\varphi_2 + \cosh\varphi_1\sinh\varphi_2) \\ -(\sinh\varphi_1\cosh\varphi_2 + \cosh\varphi_1\sinh\varphi_2) & \cosh\varphi_1\cosh\varphi_2 + \sinh\varphi_1\sinh\varphi_2\end{pmatrix}.

Why: no cleverness — just the product, written out. Now look at the entries.

Step 2. Recognise them. The hyperbolic addition formulae are $\cosh(\varphi_1+\varphi_2) = \cosh\varphi_1\cosh\varphi_2 + \sinh\varphi_1\sinh\varphi_2$ and $\sinh(\varphi_1+\varphi_2) = \sinh\varphi_1\cosh\varphi_2 + \cosh\varphi_1\sinh\varphi_2$ . Those are exactly the entries of Step 1. Hence the product is $\begin{pmatrix}\cosh(\varphi_1+\varphi_2) & -\sinh(\varphi_1+\varphi_2) \\ -\sinh(\varphi_1+\varphi_2) & \cosh(\varphi_1+\varphi_2)\end{pmatrix} = \Lambda(\varphi_1+\varphi_2)$ . ∎ Why: that is the theorem, and note how little it cost. The composed boost is a boost (closure), and its parameter is the sum. The awkward $\gamma_1\gamma_2(1+\beta_1\beta_2)$ prefactor that appears when you do this in the velocity variable has simply evaporated — it was never anything but $\cosh(\varphi_1+\varphi_2)$ in disguise.

Step 3. Extract the velocity-composition law. The composed boost has $\beta_3 = \tanh\varphi_3 = \tanh(\varphi_1+\varphi_2)$ . Apply the $\tanh$ addition formula:

\beta_3 = \frac{\tanh\varphi_1 + \tanh\varphi_2}{1 + \tanh\varphi_1\tanh\varphi_2} = \frac{\beta_1+\beta_2}{1+\beta_1\beta_2}.

Why: this is the punchline of the whole subject compressed into one substitution. The relativistic velocity-addition law — whose derivation from the transformation belongs to Relativistic Velocity Addition — is the tanh addition formula. It is not a physical law that happens to resemble a mathematical identity; it is the identity, evaluated on the physical parameter.

Step 4. Check it on real numbers. $\beta_1 = \beta_2 = 0.5$ gives $\varphi_1 = \operatorname{arctanh} 0.5 = 0.549306$ , so $\varphi_3 = 1.098612$ and $\beta_3 = \tanh(1.098612) = 0.8$ . The velocity law agrees: $(0.5+0.5)/(1+0.25) = 1/1.25 = 0.8$ ✓. And $\gamma_3 = \cosh(1.098612) = 1.6667 = 1/\sqrt{1-0.64}$ ✓. Why: two independent routes, one number. This is the arithmetic drawn in the first figure.

Step 5. The group structure falls out for free: $\Lambda(0) = I$ , $\Lambda(\varphi)\Lambda(-\varphi) = I$ (the reciprocity of The Inverse Lorentz Transformation and Reciprocity), and associativity from that of $+$ . In the rapidity label the boost group is just $(\mathbb{R}, +)$ . Why: this is the strongest statement of the article. The collinear boosts are isomorphic to the additive real numbers, and all the apparent nonlinearity of relativistic kinematics was an artefact of a bad coordinate. *Verified in verify_aq-sr-rapidity-hyperbolic-boosts.py: $\Lambda(\varphi_1)\Lambda(\varphi_2) - \Lambda(\varphi_1+\varphi_2)$ simplifies to the zero matrix symbolically; the equivalence of $\tanh(\varphi_1+\varphi_2)$ with $(\beta_1+\beta_2)/(1+\beta_1\beta_2)$ is proved symbolically; and both are checked numerically against each other on a $21\times21$ grid of velocities.*

Go deeper · Derivation C · 2 minTHE CEILING — why no finite number of boosts reaches c

Derivation CTHE CEILING — why no finite number of boosts reaches c

Here is the second owned result. The usual account of the speed limit invokes energy: accelerating a massive particle to $c$ would take infinite energy, so it cannot be done. True — but it is a dynamical answer to a kinematical question, and it leaves the impression that $c$ is almost reachable and merely expensive. The rapidity picture says something sharper. Boosting is free; you may apply as many boosts as you like, each adding its rapidity. The barrier is not that the sum is hard to make large. The barrier is that $\tanh$ never returns $1$ .

Step 1. State the fact about the function. $\tanh: \mathbb{R} \to (-1, 1)$ is a bijection onto the open interval: it is strictly increasing ( $\mathrm{d}\tanh/\mathrm{d}\varphi = \operatorname{sech}^2\varphi > 0$ everywhere), and $\lim_{\varphi\to\pm\infty}\tanh\varphi = \pm 1$ , a supremum that is never attained. Why: everything else follows from this one line of elementary analysis. Note where the physics is not: there is no force, no energy, no mass in this statement.

Step 2. Apply additivity. Compose any finite number of boosts: $\varphi_{\text{total}} = \varphi_1 + \varphi_2 + \dots + \varphi_N$ , a finite real number whenever each $\varphi_i$ is finite. The resulting velocity is $\beta_{\text{total}} = \tanh(\varphi_{\text{total}})$ , and by Step 1 this satisfies $|\beta_{\text{total}}| < 1$ strictly. Why: there is the theorem. Not "you get close to $c$ "; not "you would need infinite energy". You cannot reach $\beta = 1$ because $1$ is not in the range of the function, and a finite sum of finite rapidities is finite.

Step 3. Make it quantitative for equal boosts. Let each boost have velocity $\beta_1$ , rapidity $\varphi_1$ , and set $k \equiv e^{2\varphi_1} = \dfrac{1+\beta_1}{1-\beta_1}$ . Then $N$ boosts give $\varphi_N = N\varphi_1$ and

\beta_N = \tanh(N\varphi_1) = \frac{e^{2N\varphi_1} - 1}{e^{2N\varphi_1} + 1} = \frac{k^N - 1}{k^N + 1}, \qquad\text{so}\qquad 1 - \beta_N = \frac{2}{k^N + 1}.

Why: a closed form, exact, for any number of boosts — the kind of thing that is invisible in the velocity variable, where you would have to iterate the composition law $N$ times and would never see the pattern.

Step 4. Put $\beta_1 = 0.5$ in it. Then $k = 1.5/0.5 = 3$ exactly, so $\beta_N = \dfrac{3^N-1}{3^N+1}$ — and the numbers are rational: $\tfrac12,\ \tfrac{8}{10},\ \tfrac{26}{28} = 0.928571,\ \tfrac{80}{82} = 0.975610,\ \tfrac{242}{244} = 0.991803, \dots$ and $1-\beta_N = 2/(3^N+1)$ , which shrinks by a factor of $3$ with every boost but is never zero. At $N = 10$ : $\beta = 59048/59050 = 0.9999661$ . Why: this is the ladder drawn in the punchline figure — equal steps in $\varphi$ , crowding in $\beta$ . The approach to $c$ is exponential and endless.

Step 5. Finally, light. A light pulse has $\beta = 1$ , hence $\varphi = \operatorname{arctanh}(1) = \infty$ . Boost it by any finite $\varphi_0$ : the new rapidity is $\infty + \varphi_0 = \infty$ , so the new velocity is still $\tanh(\infty) = 1$ . Why: the invariance of the speed of light is the statement that infinity is a fixed point of addition. The second postulate, which is normally an axiom one swallows, becomes in this variable an arithmetical triviality — and that is about as strong a sign as one could ask for that rapidity is the correct coordinate. *Verified in verify_aq-sr-rapidity-hyperbolic-boosts.py: strict monotonicity and the open-range property of $\tanh$ ; the closed form $\beta_N = (k^N-1)/(k^N+1)$ checked against $N$ -fold iteration of the velocity-composition law for $N = 1\ldots12$ at three different $\beta_1$ ; the exact rationals $13/14$ , $40/41$ , $121/122$ , $59048/59050$ ; $1-\beta_N = 2/(3^N+1)$ ; and $\beta \oplus 1 = 1$ for a sweep of $\beta$ .*

Figure

PUNCHLINE — the speed limit is the range of a function. β = tanh φ, with the rapidity ticks of ten successive β = 0.5 boosts: EQUALLY spaced in φ (steps of φ₁ = 0.549306), crowding hopelessly into the ceiling in β. The exact values are (3ᴺ−1)/(3ᴺ+1): 0.5, 0.8, 0.928571, 0.975610, 0.991803, … 0.9999661 at N = 10. The gap 1 − β = 2/(3ᴺ+1) shrinks threefold per boost and never closes

Limiting cases

Case	Result	Meaning
$\varphi \to 0$	$\beta \approx \varphi - \varphi^3/3$ , $\gamma \approx 1 + \varphi^2/2$	rapidity is velocity in the Galilean regime; at $\varphi = 0.1$ , $\beta = 0.099668$
$\varphi \to \infty$	$\beta \to 1^-$ , $\gamma \approx \tfrac12 e^{\varphi}$	the light barrier is at infinite rapidity; $\gamma$ grows exponentially in $\varphi$
$\varphi_2 = -\varphi_1$	$\Lambda(\varphi_1)\Lambda(-\varphi_1) = I$	reciprocity, in one symbol
$N$ equal boosts	$\beta_N = \dfrac{k^N-1}{k^N+1}$ , $k = \dfrac{1+\beta_1}{1-\beta_1}$	exponential approach; $1-\beta_N = 2/(k^N+1) > 0$ always
light, $\beta = 1$	$\varphi = \infty$	$\infty + \text{finite} = \infty$ : the second postulate as arithmetic
$\beta_1 = 0.5$	$k = 3$ , $\beta_N = \dfrac{3^N-1}{3^N+1}$	the ladder: $0.5,\ 0.8,\ 0.928571,\ 0.975610,\ \dots$
non-collinear boosts	$\varphi$ does not simply add	a Thomas–Wigner rotation appears; additivity holds along one axis only

Common trapRapidity is a hyperbolic angle, not an angle — and cosh is not cos

The parallel with rotations is exact enough to be useful and seductive enough to be dangerous. A rotation matrix has $\cos$ and $\sin$ ; a boost has $\cosh$ and $\sinh$ ; angles add and rapidities add. But the trigonometric functions are bounded and periodic, and the hyperbolic ones are neither. Rotating by $2\pi$ returns you to where you started; boosting by rapidity $2\pi$ does not — it lands you at $\beta = \tanh(2\pi) = 0.999993$ , with $\gamma = 267.7$ . There is no periodicity, no "full turn", and no upper limit on $\varphi$ . The consequences: $\gamma = \cosh\varphi$ grows without bound, where $\cos$ never exceeds $1$ ; and $\beta = \tanh\varphi$ is confined to $(-1,1)$ , where $\tan$ ranges over all of $\mathbb{R}$ . The unboundedness of $\gamma$ and the boundedness of $\beta$ are the same fact, seen through two functions of one unbounded parameter. And do not let the analogy tempt you into writing $\cosh^2 + \sinh^2 = 1$ : the sign is a minus, and that minus is the metric signature $(+,-,-,-)$ making itself felt.

Common trapThe collider's "rapidity" y and its cousin "pseudorapidity" η are not always this φ

Particle physicists use the word constantly, and it means something slightly narrower than the boost parameter of this article. The collider rapidity of a particle is defined with respect to the beam axis $z$ :

y = \frac{1}{2}\ln\!\left(\frac{E + p_z c}{E - p_z c}\right).

For a particle moving along the beam axis this is exactly $\operatorname{arctanh}(\beta_z) = \varphi$ — the same object. But for a particle with transverse momentum it is not $\operatorname{arctanh}|\vec\beta\,|$ : it is the rapidity of the longitudinal motion only, and $|\vec\beta| > \tanh y$ in general. Physicists use $y$ rather than the polar angle for precisely the reason this article gives: under a boost along the beam — which is what happens whenever the colliding partons carry unequal momentum fractions — $y$ shifts by a constant, so differences in $y$ , and the shapes of $\mathrm{d}N/\mathrm{d}y$ distributions, are boost-invariant. Sharper still, pseudorapidity $\eta = -\ln\tan(\theta/2)$ is a different quantity again: a function of the polar angle alone, ignoring the mass, equal to $y$ only when $p \gg mc$ . Detector geometry is quoted in $\eta$ ; physics is done in $y$ .

Worked examples

1 · How many half-light-speed boosts does it take to beat the LHC?

A $7\ \mathrm{TeV}$ proton at the LHC. Find its rapidity. Then find how many successive boosts of $\beta_1 = 0.5$ — each one a perfectly ordinary boost — it would take to exceed that speed, and check the answer exactly.

Step 1 — get $\gamma$ . For a proton, $\gamma = E/(m_pc^2) = 7000/0.938272 = \mathbf{7460.5}$ .

Step 2 — get the rapidity. $\varphi = \operatorname{arccosh}\gamma = \ln\!\left(\gamma + \sqrt{\gamma^2-1}\right) = \ln(7460.5232 + 7460.5231) = \ln(14921.046) = \mathbf{9.6105}$ . (The square root equals $\gamma$ to seven significant figures: the proton is so ultrarelativistic that its momentum and energy have all but converged.) That is the whole of the LHC's kinematic reach, expressed as a single number of order ten. (The corresponding $\beta$ is $0.999\,999\,991$ — nine nines, and utterly uninformative by comparison. This is the practical argument for rapidity: it is the variable in which relativistic speeds have human-sized numbers.)

Step 3 — one boost's worth of rapidity. $\varphi_1 = \operatorname{arctanh}(0.5) = \tfrac12\ln 3 = \mathbf{0.549306}$ .

Step 4 — divide. $N = \varphi_{\text{LHC}}/\varphi_1 = 9.6105/0.549306 = \mathbf{17.50}$ . So seventeen boosts is not enough and eighteen is more than enough.

Step 5 — check it exactly. With $k = 3$ , the rapidity ladder gives $\gamma_N = \cosh(N\varphi_1) = \tfrac12(3^{N/2} + 3^{-N/2})$ :

\gamma_{17} = 5682.0 \quad (<7460.5), \qquad \gamma_{18} = 9841.5 \quad (>7460.5). \;\checkmark

Eighteen boosts of half the speed of light overshoot the LHC's $7\ \mathrm{TeV}$ proton. And the corresponding velocity, $\beta_{18} = (3^{18}-1)/(3^{18}+1) = 0.999\,999\,995$ , is still strictly less than $1$ — as it must be, and as it would remain after eighteen thousand boosts.

2 · The muon's rapidity, and a boost split in half

In the Frisch–Smith experiment the muon travels at $\beta = 0.995$ . Find its rapidity. Then split the boost into two equal halves and verify, by composing them with the velocity-addition law, that they reconstruct the original velocity — a direct test of additivity on a real measurement.

Step 1 — the rapidity. $\varphi = \operatorname{arctanh}(0.995) = \tfrac12\ln\!\left(\dfrac{1.995}{0.005}\right) = \tfrac12\ln(399) = \mathbf{2.9945}$ .

Step 2 — cross-check against $\gamma$ . $\cosh(2.9945) = \mathbf{10.0125}$ , and directly $\gamma = 1/\sqrt{1-0.995^2} = 1/\sqrt{0.009975} = 10.0125$ ✓. The dictionary holds on a real experimental number.

Step 3 — halve the rapidity. $\varphi/2 = 1.49724$ , and $\beta_{1/2} = \tanh(1.49724) = \mathbf{0.904648}$ . Note what this is not: it is not $0.995/2 = 0.4975$ . Halving a boost does not halve its velocity, because velocity is not the boost parameter.

Step 4 — compose the two halves with the velocity law. Using $\oplus$ and nothing else:

0.904648 \oplus 0.904648 = \frac{1.809297}{1 + 0.818389} = \frac{1.809297}{1.818389} = \mathbf{0.995000}. \;\checkmark

Exactly the muon's speed, recovered. The velocity-addition law and rapidity additivity are the same statement, and here they agree to six decimal places on a number that came out of a mountain in New Hampshire.

Step 5 — read the physics off. One parameter runs everything: $\cosh\varphi$ is the time-dilation factor, so the muon's $1907\ \mathrm{m}$ flight (the detector-to-detector drop; the summit's $1917\ \mathrm{m}$ elevation is a different quantity), which takes $\Delta t = 6.393\ \mathrm{\mu s}$ on the mountain, takes $6.393/\cosh(2.9945) = \mathbf{0.6385\ \mu s}$ on the muon's own clock; $\sinh\varphi$ is $\gamma\beta$ , the quantity in the momentum $p = mc\sinh\varphi$ ; and $\tanh\varphi$ is the velocity.

Origins · 1 minRapidity and how this was discovered

Rapidity

1908–1924 · Minkowski's hyperbolic angle · Varićak's Lobachevskian velocity space · Robb gives it its name · Silberstein and Eddington carry it into the textbooks

The parameter came before the word. Minkowski, in 1908, already wrote the boost as a rotation through an imaginary angle — the same object, seen in a mirror. But the recognition that the natural variable of relativistic velocity is a hyperbolic arc length belongs to two men working independently in 1910: Vladimir Varićak (1865–1942), a Croatian mathematician at Zagreb, and Edmund Taylor Whittaker, in the first edition of his History of the Theories of Aether and Electricity. Varićak's insight was the deeper: he showed that the space of relativistic velocities is not Euclidean but hyperbolic — Lobachevskian — and that the awkward velocity-addition law is nothing but the addition of arcs in that geometry. Where classical physics draws a Euclidean triangle of velocities, relativity draws a Lobachevskian triangle of rapidities, whose angular defect is the Thomas–Wigner rotation. The name is Alfred Robb's: in Optical Geometry of Motion (Cambridge, 1911) he called the parameter the rapidity, and the word stuck — Varićak adopted it in 1912, Silberstein used it in The Theory of Relativity (1914), and by Eddington's Mathematical Theory of Relativity (1924) it was simply what the quantity was called.

History

1908

— Minkowski, in "Raum und Zeit", presents the boost as a rotation through an imaginary angle

\psi

with

\tan(i\psi) = i\beta

— the hyperbolic parameter in Euclidean disguise.

1910

— Varićak shows that the space of relativistic velocities is hyperbolic (Lobachevskian), and that velocity addition is the addition of arcs in that geometry; Whittaker independently introduces the same parameter in his History of the Theories of Aether and Electricity.

1911

— Robb, in Optical Geometry of Motion (Cambridge), gives the parameter its name: rapidity.

1912

— Varićak adopts Robb's term.

1914

— Silberstein's The Theory of Relativity puts the rapidity into a textbook; Eddington's Mathematical Theory of Relativity (1924) settles the usage for good.

Going further

Texts. Taylor & Wheeler, Spacetime Physics, 2nd ed., Ch. 8 ("the rapidity", and the boost-as-hyperbolic-rotation table) · Rindler, Introduction to Special Relativity, 2nd ed., §§2.9, 3.4 · Griffiths, Introduction to Electrodynamics, 4th ed., Problem 12.11. On the hyperbolic geometry of velocity space: V. Varićak, Physikalische Zeitschrift 11, 93 (1910). On the name: A. A. Robb, Optical Geometry of Motion (Cambridge, 1911). For the collider variable: Particle Data Group, Review of Particle Physics, "Kinematics" §48.

Provenance. The definition and dictionary follow Rindler §2.9. The additivity theorem and its corollary — that the velocity-composition law is the $\tanh$ addition formula — and the ceiling as a range statement, with the exact closed form $\beta_N = (k^N-1)/(k^N+1)$ , are derived and machine-checked above; they are the article's owned results. The geometry of the boost as an isometry of $s^2$ (the generator, $\Lambda = \exp(\varphi K)$ , the invariant hyperbolae) is not re-derived here — it is owned by Lorentz Boosts as Hyperbolic Rotations. Every number is recomputed in verify_aq-sr-rapidity-hyperbolic-boosts.py, including the exact rationals of the ladder, the LHC proton and the muon.

Builds on. The Lorentz Transformation Equations · The Lorentz Factor Gamma · The Invariant Spacetime Interval · Deriving the Lorentz Transformation. Leads to. Lorentz Boosts as Hyperbolic Rotations · Relativistic Velocity Addition · Proper Time · Minkowski Diagrams and Skewed Axes.

Is rapidity a real physical quantity, or just a change of variable?

It is a change of variable — and that is exactly what makes it physical. Nothing about the world changes when you relabel a boost by $\varphi$ instead of $\beta$ ; the two carry identical information, since $\tanh$ is a bijection. What changes is which structure is visible. In the $\beta$ label, the boosts are a set with a bizarre composition rule. In the $\varphi$ label, they are the additive group of real numbers. The physics was always the group; the velocity was hiding it — the same sense in which a logarithm is "just" a change of variable, and the reason slide rules worked.

If rapidities add, why do particle physicists still bother with velocity at all?

Because velocity is what you measure. A time-of-flight detector reports $\beta$ ; a tracker reports momentum, which gives $\gamma\beta = \sinh\varphi$ . Rapidity is not measured directly — it is computed. Its value is economy: it makes boost composition trivial, makes constant proper acceleration linear ( $\varphi = a\tau/c$ ), turns the velocity ceiling into a statement about a function's range, and — the clincher at a collider — makes a longitudinal boost a mere translation. Velocity is the observable; rapidity is the coordinate you should compute in.

Does rapidity still add for boosts in different directions?

No, and this is the one place the beautiful picture breaks. Additivity, equation (4), is proved for boosts along a common axis. Two boosts along different axes do not compose into a pure boost at all — the product is a boost plus a rotation, the Thomas–Wigner rotation, so there is no single rapidity to be the sum. In Varićak's picture this is natural: velocity space is hyperbolic, and the failure of rapidities to add around a closed circuit is the angular defect of a hyperbolic triangle. The Thomas precession of an electron's spin is that defect, made observable in a spectrum.

What is the rapidity of light, and why is that not just a cheap way of writing "infinity"?

$\varphi = \operatorname{arctanh}(1) = \infty$ , and it is not cheap — it does real work. In the velocity variable, the invariance of $c$ is a postulate: you assert that $\beta = 1$ is unchanged by boosts, then verify it survives the composition law. In the rapidity variable it is a triviality with content: boosting adds a finite rapidity, and $\infty + \varphi_0 = \infty$ , so a pulse at $\varphi = \infty$ stays at $\beta = 1$ in every frame. The reason light is a fixed point while everything else moves is that light alone sits at the unreachable end of the parameter.

Why exactly is the tanh addition formula the same as the velocity-addition law? Isn't that a coincidence?

It cannot be a coincidence, because it is a derivation. The boost matrix in the rapidity variable is built from $\cosh$ and $\sinh$ (equation 3, which follows from $\gamma^2 - (\gamma\beta)^2 = 1$ and nothing else). Multiplying two such matrices produces, entry by entry, precisely the $\cosh$ and $\sinh$ addition formulae — not an interpretation, but the arithmetic of the product. So the composed boost carries rapidity $\varphi_1 + \varphi_2$ , and its velocity is $\tanh(\varphi_1+\varphi_2)$ , which expands into $(\beta_1+\beta_2)/(1+\beta_1\beta_2)$ . Every step is forced. The "coincidence" is a statement about which group the boosts form, and the hyperbolic functions are simply the functions that parametrise it.

If I keep firing my rocket forever, my rapidity grows without bound. What does that feel like?

Entirely ordinary — and this is the most physical thing rapidity has to say. Under constant proper acceleration $a$ , the rapidity grows linearly on your own clock: $\varphi(\tau) = a\tau/c$ . You feel a constant push, your accelerometer reads a constant $a$ , and nothing in the cabin ever suggests a limit. Meanwhile $\beta = \tanh(a\tau/c)$ saturates, and an observer back home watches you creep toward $c$ and never arrive. Both descriptions are correct. You are not being throttled by anything: the parameter you actually accumulate is unbounded. It is only its hyperbolic tangent — the thing home calls your speed — that has a ceiling.