Rapidity

Board a rocket and boost to half the speed of light. From that rocket, launch a second rocket at half the speed of light relative to the first. Then a third, from the second. Do it ten times, each boost identical, each one a perfectly ordinary half-light-speed shove relative to the ship you just left. Arithmetic says you are now doing five times the speed of light. Arithmetic is wrong. You are doing $0.9999661c$ — and the exact fraction, if you want it, is $29524/29525$ . Ten enormous boosts have bought you a hundredth of a percent, and no number of further boosts will ever buy you the last sliver. Something has clearly failed to add. But something did add, perfectly and linearly, all ten times: a quantity called the rapidity, $\varphi = \operatorname{artanh}\beta$ . Each half-light-speed boost adds exactly $\operatorname{artanh}(0.5) = 0.549306$ to it, so after ten you have $\varphi = 5.493061$ , on the nose, with no correction terms and no relativistic funny business. Velocity is what a speedometer reads. Rapidity is what a boost actually does. And the reason relativistic velocities behave so strangely is simply that we have been keeping score with the wrong variable.

Key takeaways

Signature $(+,-,-,-)$ : $\Delta s^2 = c^2\Delta t^2 - \Delta x^2$ , timelike $> 0$ . The identity $\gamma^2 - (\gamma\beta)^2 = 1$ is exactly $\cosh^2\varphi - \sinh^2\varphi = 1$ , so a boost is a hyperbolic rotation.
Rapidity $\varphi = \operatorname{artanh}\beta = \tfrac12\ln\frac{1+\beta}{1-\beta}$ , and then $\beta = \tanh\varphi$ , $\gamma = \cosh\varphi$ , $\gamma\beta = \sinh\varphi$ .
Collinear boosts add their rapidities: $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1+\varphi_2)$ , exactly. Einstein's velocity-composition law is the $\tanh$ addition formula, in disguise.
Owned result: velocity is not a group parameter; rapidity is. The $1{+}1$ D boost group is $(\mathbb{R},+)$ — and $\operatorname{artanh}$ is, up to an overall scale, the only function of $\beta$ that adds. $c$ is unreachable because $\operatorname{artanh}$ has a pole at $\beta = 1$ : reaching light speed costs infinite rapidity.
Exact anchor: $\beta = 0.6 \iff \varphi = \ln 2$ , giving $\gamma = 5/4$ and $\gamma\beta = 3/4$ .
Key idea: boosts compose by addition, always — just not addition of velocity. Change variables from $\beta$ to $\varphi$ and every awkward relativistic formula collapses into ordinary arithmetic.

The core insight

Relativity did not break addition; we were adding the wrong thing. A boost is a rotation through an imaginary angle — a hyperbolic rotation — and rotations compose by adding angles. The hyperbolic angle is the rapidity. Once you parametrise a boost by $\varphi$ instead of $\beta$ , the velocity-addition law, the $\gamma$ factor, the impossibility of reaching $c$ , and the whole apparatus of hyperbolic identities become one fact restated four ways.

Where this lives today

Rapidity is not a pedagogical convenience. It is the working variable of experimental particle physics. At a hadron collider the interesting collisions happen between partons carrying unknown, unequal fractions of the beam momentum, so the centre of momentum of any given event is moving along the beam axis by an unknown amount — and every measured quantity that depends on velocity along that axis is therefore contaminated. Rapidity solves this exactly: because a boost along the beam simply adds a constant to the rapidity of every particle, differences of rapidity are invariant under the unknown longitudinal boost, and so are shapes of rapidity distributions. Detectors are built with their granularity uniform in rapidity for precisely this reason. The numbers are worth seeing. A $6.5\ \mathrm{TeV}$ proton in the LHC has $\gamma = E/mc^2 = 6500/0.938272 = 6927.6$ , hence a rapidity of $\varphi = \operatorname{arcosh}\gamma = 9.536$ . Its $\beta$ is $0.99999999$ — a number that conveys nothing and cannot usefully be plotted. Its rapidity, $9.54$ , is a perfectly ordinary quantity that sits comfortably on a linear axis alongside the rapidities of the particles it produces. And notice the scaling: doubling the beam energy adds only $\ln 2 = 0.69$ to the rapidity. The entire history of accelerator physics, in one number. Rapidity also gives the cleanest statement of what constant acceleration means. A rocket whose crew feels a constant $g$ underfoot is not one whose velocity increases linearly — that would take it past $c$ within a year. It is one whose rapidity increases linearly: $\varphi = a\tau/c$ , where $\tau$ is the proper time aboard and $a$ the proper acceleration. One Julian year of $1g$ gives $\varphi = (9.80665)(3.15576\times10^7)/(2.99792458\times10^8) = 1.0323$ , hence $\beta = \tanh(1.0323) = 0.7748$ and $\gamma = 1.5818$ . The Galilean bookkeeping would have read $g\tau = 3.09\times10^8\ \mathrm{m/s}$ , which exceeds $c$ — the failure is not subtle. (The resulting worldline is a hyperbola, and its geometry and its horizon belong to the articles on Relativistic Acceleration and on Rindler Coordinates; they are not treated here.)

The physical picture

Start from the one equation everyone already knows and nobody reads carefully:

\gamma = \frac{1}{\sqrt{1-\beta^2}} \qquad\Longrightarrow\qquad \gamma^2 - (\gamma\beta)^2 = 1.

Look at what that says. The pair $(\gamma,\ \gamma\beta)$ — which is exactly the pair of numbers appearing in the boost matrix — lies on the unit hyperbola $X^2 - Y^2 = 1$ , for every $\beta$ whatsoever. It cannot lie anywhere else. And a point on the unit hyperbola has a standard parametrisation, one that has been sitting in the textbooks since the 1750s: $X = \cosh\varphi$ , $Y = \sinh\varphi$ . So there is a number $\varphi$ such that $\gamma = \cosh\varphi$ and $\gamma\beta = \sinh\varphi$ . Divide the second by the first and you get $\beta = \tanh\varphi$ . Invert, and $\varphi = \operatorname{artanh}\beta$ . That is the entire definition of rapidity, and it was forced on us: nothing was chosen, we simply named the parameter that the geometry already had. The payoff is immediate and enormous. In Euclidean space, the transformation that mixes $x$ and $y$ is a rotation, and two rotations compose by adding angles — nobody would dream of asking how two rotation angles "combine" beyond adding them. In Minkowski space, the transformation that mixes $ct$ and $x$ is a boost, and two boosts compose by adding rapidities. Same statement, one sign changed. What made relativistic velocity addition look bizarre is that $\beta = \tanh\varphi$ is a nonlinear function of the thing that actually adds — like insisting on adding the sines of two angles and then being astonished that $\sin(\theta_1 + \theta_2) \ne \sin\theta_1 + \sin\theta_2$ .

Interactive Simulation

Minkowski Boost ↔ Euclidean Rotation — a boost is the hyperbolic analogue of a rotation; the rapidity φ = atanh β is the hyperbolic angle. Compare the two side by side.

Figure

PUNCHLINE — N successive boosts of 0.5c each, relative to the ship you just left. GALILEO says the speeds add: 5c after ten boosts (the dashed line, straight off the top of the frame). RELATIVITY says β saturates: β_N = (3^N − 1)/(3^N + 1) exactly, reaching 0.9999661c at N = 10 and never reaching 1. RAPIDITY says φ adds, exactly and forever: φ_N = N × 0.549306, a straight line with no ceiling. Galileo was right about the shape of the law — he was wrong about which variable obeys it

The core equations

Rapidity is defined by inverting the hyperbolic tangent, and everything else follows.

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

where: $\beta = v/c$ . Physical reading: $\varphi$ is the hyperbolic angle of the boost. It runs over the whole real line, $-\infty < \varphi < \infty$ , while $\beta$ is confined to $(-1, 1)$ — and the map between them is a bijection. Every attainable velocity has a rapidity; no rapidity, however large, corresponds to $\beta = 1$ . The two factors that dominate every relativistic formula are now just $\cosh$ and $\sinh$ :

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

where: $\gamma = (1-\beta^2)^{-1/2}$ . Physical reading: the famous identity $\gamma^2(1-\beta^2) = 1$ is the hyperbolic Pythagoras. It was never anything else. This is why $\gamma$ and $\gamma\beta$ are always the two legs of a hyperbola and never of a circle. The boost matrix, written in $\varphi$ , stops looking like a formula and starts looking like a rotation:

\Lambda(\varphi) = \begin{pmatrix}\cosh\varphi & -\sinh\varphi \\ -\sinh\varphi & \cosh\varphi\end{pmatrix} \quad\text{acting on}\quad \begin{pmatrix}ct \\ x\end{pmatrix} \qquad (3)

where: $\Lambda$ is the standard boost along $+x$ . Physical reading: compare the Euclidean rotation $\begin{pmatrix}\cos\theta & -\sin\theta \\ \sin\theta & \cos\theta\end{pmatrix}$ . The trigonometric functions have become hyperbolic and one sign has moved; that is the whole difference between turning your axes and changing your speed. It also makes $\det\Lambda = \cosh^2 - \sinh^2 = 1$ obvious at a glance. And the composition law is now the simplest equation in relativity:

\Lambda(\varphi_1)\,\Lambda(\varphi_2) = \Lambda(\varphi_1 + \varphi_2), \qquad\qquad \tanh(\varphi_1+\varphi_2) = \frac{\tanh\varphi_1 + \tanh\varphi_2}{1 + \tanh\varphi_1\tanh\varphi_2} \qquad (4)

where: the second identity is the standard $\tanh$ addition formula. Physical reading: read it right to left. Written in $\beta$ , it is $\beta_{\text{tot}} = (\beta_1+\beta_2)/(1+\beta_1\beta_2)$ — Einstein's velocity-composition law, the one that looks like an ad-hoc relativistic correction. Written in $\varphi$ , it is $\varphi_{\text{tot}} = \varphi_1 + \varphi_2$ . They are the same equation. The "correction" was an artefact of the coordinate we chose for speed.

Where rapidity comes from

Derivation AThe boost is a rotation through a hyperbolic angle

The claim that a Lorentz boost "is" a hyperbolic rotation is usually presented as an elegant re-writing, an optional flourish for the mathematically inclined. It is not optional. The hyperbolic form is forced by a single algebraic fact about $\gamma$ , and once you see the forcing there is no other parametrisation available. What follows is that derivation, done in the $(+,-,-,-)$ signature, with no appeal to any prior knowledge of hyperbolic functions beyond their definitions.

Step 1. Write the boost and extract its two entries. The standard boost along $+x$ sends $ct \mapsto \gamma(ct - \beta x)$ and $x \mapsto \gamma(x - \beta\, ct)$ , so its matrix is built entirely out of the two numbers $\gamma$ and $\gamma\beta$ . Why: whatever a boost is, it is determined by this pair. Understanding boosts means understanding the pair.

Step 2. Discover the constraint the pair obeys. From $\gamma = (1-\beta^2)^{-1/2}$ we get $\gamma^2(1-\beta^2) = 1$ , that is, $\gamma^2 - (\gamma\beta)^2 = 1$ . So the point $(X, Y) = (\gamma,\ \gamma\beta)$ lies on the unit hyperbola $X^2 - Y^2 = 1$ — for every $\beta$ , without exception. Why: this is the geometric content of the Lorentz factor, and it is exactly the metric's minus sign showing through. In a Euclidean world the constraint would read $X^2 + Y^2 = 1$ and we would be on a circle.

Step 3. Parametrise the hyperbola. Any point on the branch $X \ge 1$ of $X^2 - Y^2 = 1$ can be written $X = \cosh\varphi$ , $Y = \sinh\varphi$ for exactly one real $\varphi$ , because $\cosh^2 - \sinh^2 = 1$ identically and $\sinh$ is a bijection from $\mathbb{R}$ onto $\mathbb{R}$ . Therefore $\gamma = \cosh\varphi$ and $\gamma\beta = \sinh\varphi$ . Why: this is not an ansatz; it is the standard parametrisation of the curve we have just been forced onto, and it is unique.

Step 4. Read off $\beta$ and invert. Dividing, $\beta = \sinh\varphi/\cosh\varphi = \tanh\varphi$ , so $\boxed{\varphi = \operatorname{artanh}\beta = \tfrac12\ln\frac{1+\beta}{1-\beta}}$ . Substituting into the matrix gives $\Lambda(\varphi)$ with entries $\cosh\varphi$ and $-\sinh\varphi$ . Why: the logarithmic form is worth memorising — it is what makes $\varphi$ computable, and it is where the pole at $\beta = 1$ will come from.

Step 5. Check the anchor. Put $\beta = 3/5$ . Then $\varphi = \tfrac12\ln\frac{8/5}{2/5} = \tfrac12\ln 4 = \ln 2$ , exactly. And indeed $\cosh(\ln 2) = \tfrac12(2 + \tfrac12) = 5/4 = \gamma$ and $\sinh(\ln 2) = \tfrac12(2 - \tfrac12) = 3/4 = \gamma\beta$ , with $(5/4)^2 - (3/4)^2 = 1$ . Why: $\beta = 0.6$ is the one velocity at which everything in relativity comes out as a clean fraction, and the reason is precisely that its rapidity is $\ln 2$ , so $e^{\varphi} = 2$ and $e^{-\varphi} = \tfrac12$ . *Verified symbolically in verify_aq-sr-rapidity.py: $\Lambda^{\mathsf T}\eta\Lambda = \eta$ and $\det\Lambda = 1$ in both the $\beta$ -form and the $\varphi$ -form; the two forms are identical at $\beta = \tanh\varphi$ ; and $\operatorname{artanh}(3/5) = \ln 2$ , $\cosh(\ln 2) = 5/4$ , $\sinh(\ln 2) = 3/4$ exactly.*

Go deeper · Derivation B · 2 minVelocity is not a group parameter. Rapidity is.

Derivation BVelocity is not a group parameter. Rapidity is.

Here is the article's owned result, and it is a statement about group theory, not about kinematics. The collinear boosts form a group; a one-dimensional continuous group has a parameter in which composition is simply addition; and the question is which function of the velocity that parameter is. The answer is $\operatorname{artanh}\beta$ — and, crucially, nothing else. Rapidity is not merely a convenient change of variable. It is the unique additive coordinate on the boost group, and velocity is provably not one.

Step 1. Show the rapidities add. Multiply two boosts in the form (3). The entries of the product are $\cosh\varphi_1\cosh\varphi_2 + \sinh\varphi_1\sinh\varphi_2 = \cosh(\varphi_1+\varphi_2)$ and $-(\sinh\varphi_1\cosh\varphi_2 + \cosh\varphi_1\sinh\varphi_2) = -\sinh(\varphi_1+\varphi_2)$ , by the hyperbolic addition formulas. So $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1+\varphi_2)$ , with no approximation and no restriction. Why: this is a one-parameter group, and $\varphi$ is its parameter. Composition of boosts is addition of reals — the simplest group there is.

Step 2. Collect the group axioms and name the group. $\Lambda(0) = \mathbb{1}$ (identity), $\Lambda(\varphi)^{-1} = \Lambda(-\varphi)$ (inverses), and composition is $+$ on $\mathbb{R}$ , which is associative and commutative. Hence $\boxed{\text{the }1{+}1\text{D boost group is }(\mathbb{R},\,+)}$ . Why: boosts along a line are, as an abstract group, nothing more exotic than the real numbers under addition. All the apparent strangeness lives in the map $\varphi \mapsto \beta$ , not in the group.

Step 3. Recover Einstein's law as a corollary. Apply $\tanh$ to Step 1: $\beta_{\text{tot}} = \tanh(\varphi_1 + \varphi_2) = \dfrac{\tanh\varphi_1 + \tanh\varphi_2}{1+\tanh\varphi_1\tanh\varphi_2} = \dfrac{\beta_1 + \beta_2}{1 + \beta_1\beta_2}$ . Why: the velocity-composition law is not an extra postulate, a correction, or a coincidence. It is the $\tanh$ addition formula wearing a physicist's hat, and it has been a theorem of hyperbolic trigonometry since Riccati and Saladini wrote it down in the 1760s — a century and a half before anyone needed it.

Step 4. Show velocity fails. Compose $\beta = 0.6$ with itself: $(0.6+0.6)/(1+0.36) = 1.2/1.36 = 15/17 = 0.882353\ldots$ , not $1.2$ . And $\gamma$ fails too: $\gamma(15/17) = 17/8 = 2.125$ , whereas $\gamma(0.6) + \gamma(0.6) = 2.5$ . But the rapidities: $\ln 2 + \ln 2 = \ln 4$ , and $\tanh(\ln 4) = (4 - \tfrac14)/(4 + \tfrac14) = 15/17$ — dead on, with $\cosh(\ln 4) = 17/8$ and $\sinh(\ln 4) = 15/8$ . Why: neither $\beta$ nor $\gamma$ is additive, and no amount of wishing makes them so. Only $\varphi$ is.

Step 5. Prove the uniqueness — the part that makes this a theorem rather than a nice observation. Suppose some function $f(\beta)$ is additive: $f(\beta_1 \oplus \beta_2) = f(\beta_1) + f(\beta_2)$ , where $\oplus$ is the composition law. Compose an infinitesimal $\epsilon$ onto $\beta$ : expanding $(\beta+\epsilon)/(1+\beta\epsilon)$ to first order gives $\beta \oplus \epsilon = \beta + \epsilon(1-\beta^2) + O(\epsilon^2)$ . Additivity then requires $f(\beta) + f'(\beta)\,\epsilon(1-\beta^2) = f(\beta) + f'(0)\,\epsilon$ , so $f'(\beta) = f'(0)/(1-\beta^2)$ . Integrating, and using $f(0) = 0$ (the identity boost has zero parameter), $\boxed{f(\beta) = f'(0)\cdot\operatorname{artanh}\beta}$ . Why: this closes the case. Any additive parametrisation of the boost group is rapidity, up to a choice of units. There is no second candidate hiding somewhere — the differential equation has one solution.

Step 6. Read off the impossibility of $c$ . Since $\varphi = \operatorname{artanh}\beta$ and $\operatorname{artanh}$ has a pole at $\beta = 1$ , reaching light speed requires infinite rapidity: $\varphi \to \infty$ as $\beta \to 1^-$ . Equivalently, $\tanh\varphi < 1$ for every finite $\varphi$ , however large. Why: this is the sharpest statement of why $c$ is a barrier. It is not that a massive body needs infinite energy as a brute fact — it is that in the variable in which boosting is additive, light speed sits at infinity. You cannot get there by adding finitely many finite amounts. *Verified symbolically in verify_aq-sr-rapidity.py: $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1+\varphi_2)$ exactly; $\Lambda(0) = \mathbb{1}$ ; $\Lambda(\varphi)^{-1} = \Lambda(-\varphi)$ ; $\tanh(\varphi_1+\varphi_2)$ equals the Einstein composition; $0.6 \oplus 0.6 = 15/17$ with $\gamma = 17/8$ (and $\gamma$ is confirmed non-additive); $\beta \oplus \epsilon = \beta + \epsilon(1-\beta^2) + O(\epsilon^2)$ ; the ODE $f' = k/(1-\beta^2)$ with $f(0) = 0$ has the unique solution $k\operatorname{artanh}\beta$ ; and $\operatorname{artanh}\beta \to \infty$ as $\beta \to 1^-$ .*

Go deeper · Derivation C · 2 minThe ladder of ten boosts, in closed form

Derivation CThe ladder of ten boosts, in closed form

The rocket in the overview is not a thought experiment that needs numerical integration. Because rapidities add, $N$ identical boosts is just $N$ times one rapidity, and the whole ladder collapses to a single exact formula that you can evaluate by hand for any $N$ . This is what a good change of variables buys you: a problem that is intractable in $\beta$ (compose the Einstein law with itself ten times, and try to keep the fractions straight) becomes a one-line multiplication in $\varphi$ .

Step 1. Compute the rung. One boost of $\beta = 1/2$ has rapidity $\varphi_1 = \operatorname{artanh}\tfrac12 = \tfrac12\ln\frac{1 + 1/2}{1 - 1/2} = \tfrac12\ln 3 = 0.549306\ldots$ Why: this is the only number in the whole problem. Everything else is arithmetic on it.

Step 2. Climb. After $N$ boosts, $\varphi_N = N\varphi_1 = \tfrac{N}{2}\ln 3$ , so $e^{\varphi_N} = 3^{N/2}$ . Then

\beta_N = \tanh\varphi_N = \frac{e^{\varphi_N} - e^{-\varphi_N}}{e^{\varphi_N} + e^{-\varphi_N}} = \frac{3^{N/2} - 3^{-N/2}}{3^{N/2} + 3^{-N/2}} = \boxed{\frac{3^N - 1}{3^N + 1}}

after multiplying top and bottom by $3^{N/2}$ . Why: an exact rational, for every $N$ , with no square roots and no floating point anywhere. The awkwardness of the Einstein law has completely evaporated.

Step 3. Check the low rungs against the composition law directly. $N=1$ : $(3-1)/(3+1) = 1/2$ . ✓ $N=2$ : $(9-1)/(9+1) = 4/5 = 0.8$ — and Einstein's law gives $(\tfrac12+\tfrac12)/(1+\tfrac14) = 1/1.25 = 0.8$ . ✓ $N=3$ : $(27-1)/(27+1) = 13/14 = 0.928571\ldots$ Why: the closed form is not a trick; it agrees with composing the velocities one at a time, as it must.

Step 4. Reach the top. $N = 10$ : $\varphi = 5\ln 3 = \ln 243 = 5.493061\ldots$ , and

\beta_{10} = \frac{3^{10}-1}{3^{10}+1} = \frac{59048}{59050} = \frac{29524}{29525} = 0.9999661304\ldots

with $\gamma_{10} = \cosh(5\ln 3) = \tfrac12\!\left(243 + \tfrac{1}{243}\right) = \frac{59050}{486} = 121.502058\ldots$ Why: ten half-light-speed boosts, and you are still $0.0000339c$ short of light. The Galilean bookkeeping would have said $5c$ — off not by a correction but by a category error.

Step 5. Contemplate the last sliver. To halve the gap $1 - \beta$ you must add roughly $\tfrac12\ln 2 = 0.347$ to $\varphi$ , since $1 - \tanh\varphi \approx 2e^{-2\varphi}$ for large $\varphi$ . Every further boost of $0.5c$ adds $0.549$ and so shrinks the gap by a factor of $3$ — forever, without ever closing it. Why: this is the geometry of the pole. Light speed is not a wall you can reach and be stopped at; it is a point at infinity in the only variable that composes. *Verified symbolically in verify_aq-sr-rapidity.py: $\operatorname{artanh}(1/2) = \tfrac12\ln 3$ ; the closed form $\beta_N = (3^N-1)/(3^N+1)$ is confirmed against $\tanh(N\operatorname{artanh}\tfrac12)$ for $N = 1\ldots12$ ; $\beta_{10} = 29524/29525$ and $\gamma_{10} = 59050/486$ exactly, with $\gamma_{10}^2 - (\gamma\beta)_{10}^2 = 1$ ; and the companion $\beta = 0.6$ ladder $\beta_N = (4^N-1)/(4^N+1)$ is confirmed for $N = 1\ldots8$ .*

Figure

The rapidity ruler. Equal boosts are equal steps in φ — the top scale is uniform, forever. The SAME boosts, read in β on the bottom scale, pile up against the wall at β = 1 and never touch it. Marked: the exact anchor β = 0.6 at φ = ln 2 = 0.693, the two-boost point 15/17 at φ = ln 4, the three-boost point 63/65 at φ = ln 8 — and, far off to the right, a 6.5 TeV LHC proton at φ = 9.54, which in β is 0.99999999

Limiting cases

Case	Result	Meaning
$\varphi \to 0$	$\beta \approx \varphi$ , $\gamma \approx 1 + \varphi^2/2$	rapidity is $v/c$ at low speed — which is why Galilean velocity addition works so well
$\beta$ small	$\varphi = \beta + \beta^3/3 + \cdots$	the first correction is third order: at $\beta = 0.1$ , $\varphi$ and $\beta$ differ by $0.03\%$
$\varphi$ large	$\gamma \approx e^{\varphi}/2$	$\gamma$ grows exponentially in the additive variable — one more boost multiplies it
$\varphi$ large	$1 - \beta \approx 2e^{-2\varphi}$	the gap to light speed shrinks geometrically and never closes
$\beta \to 1^-$	$\varphi \to +\infty$	the pole. Light speed costs infinite rapidity
Photon	$\varphi$ undefined (infinite)	light has no rest frame, and correspondingly no finite rapidity
$\varphi < 0$	boost in $-x$	rapidity is signed, and $\Lambda(-\varphi) = \Lambda(\varphi)^{-1}$
Non-collinear boosts	rapidities do not simply add	additivity is a collinear statement; see Hyperbolic Geometry of Boosts

Common trapRapidity adds only for collinear boosts

Everything in this article is one-dimensional: all the boosts point along the same axis. The additivity $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1+\varphi_2)$ is a statement about the one-parameter subgroup generated by a single direction, and it is exact there. Boosts in different directions do not commute, and their composition is not a boost of the summed rapidity — it is a boost combined with a rotation. Adding rapidities as if they were vectors is a real and common error. The composition rule for non-collinear boosts, and the rotation it drags in, belong to Hyperbolic Geometry of Boosts and to the article on Thomas Precession; nothing about them is derived here.

Common trapTwo symbols, one quantity — and a signature to check

This corpus writes rapidity as $\varphi$ throughout. Other sources write $\theta$ , $w$ , $\zeta$ , $\eta$ (unfortunate, given the metric) or — following Robb's original — $\omega$ ; particle physicists call the same quantity $y$ and define it as $y = \tfrac12\ln\frac{E + p_z c}{E - p_z c}$ , which for a particle moving along $z$ is exactly $\operatorname{artanh}\beta_z$ . The AllQFT base for this slug uses $\theta$ in its treatment of accelerated motion while using no symbol elsewhere; that inconsistency is repaired here in favour of $\varphi$ . The signature is $(+,-,-,-)$ : $\Delta s^2 = c^2\Delta t^2 - \Delta x^2$ , timelike positive. In $(-,+,+,+)$ the hyperbolic functions are unchanged — rapidity is signature-blind, because $\gamma^2 - (\gamma\beta)^2 = 1$ holds in both — but the interval it acts on flips sign.

Worked examples

1 · Ten boosts of half light speed

A rocket boosts to $0.5c$ . From it, a second rocket boosts to $0.5c$ relative to the first. Ten stages. What is the final speed relative to the original frame?

Step 1 — identify. Each stage is a collinear boost, so the rapidities add and nothing else does. The velocity-composition law is the awkward route; the rapidity route is one multiplication. Compute the rapidity of one stage: $\varphi_1 = \operatorname{artanh}(0.5) = \tfrac12\ln 3 = 0.5493061$ .

Step 2 — apply. Ten identical stages: $\varphi_{10} = 10 \times 0.5493061 = 5.4930614$ . Exactly, with no correction terms. (In closed form, $\varphi_{10} = 5\ln 3 = \ln 243$ .)

Step 3 — convert back. $\beta_{10} = \tanh(5\ln 3)$ . Since $e^{\varphi_{10}} = 243$ , \beta_{10} = \frac{243 - 1/243}{243 + 1/243} = \frac{59048}{59050} = \frac{29524}{29525} = 0.9999661304.

Step 4 — cross-check on the low rungs. Two stages by Einstein's law: $(0.5+0.5)/(1+0.25) = 0.8$ ; by the closed form, $(3^2-1)/(3^2+1) = 8/10 = 0.8$ . ✓ Three stages: $13/14 = 0.928571$ . ✓ The two routes agree, as they must — they are the same law.

Step 5 — the boundary number. The Galilean answer is $10 \times 0.5c = 5c$ . The true answer is $0.99997c$ . The discrepancy is not a small correction to be added on; it is the difference between adding the right variable and the wrong one. Meanwhile $\gamma_{10} = \cosh(5\ln 3) = \tfrac12(243 + \tfrac{1}{243}) = 121.502$ — so although the speed has barely moved since stage five, the ship's energy and momentum have gone up by a factor of over a hundred. That is the physical content of the saturation: the boosts are still doing real work; they are simply no longer buying speed.

Step 6 — read the structure. Each further boost multiplies $e^{\varphi}$ by $\sqrt{3}$ , hence multiplies $\gamma$ by (asymptotically) $\sqrt{3}$ and divides the gap $1 - \beta$ by $3$ . You can do this forever. You will never arrive.

2 · The exact anchor, and three boosts of 0.6c

Show that $\beta = 0.6$ has rapidity exactly $\ln 2$ , and use it to compose three such boosts without ever touching the velocity-addition law.

Step 1 — identify. Use the logarithmic form, $\varphi = \tfrac12\ln\frac{1+\beta}{1-\beta}$ , which turns a rapidity into a logarithm of a rational number.

Step 2 — the anchor. $\varphi = \tfrac12\ln\dfrac{1 + 3/5}{1 - 3/5} = \tfrac12\ln\dfrac{8/5}{2/5} = \tfrac12\ln 4 = \ln 2 = 0.693147$ . So $e^{\varphi} = 2$ and $e^{-\varphi} = \tfrac12$ , exactly, and hence $\gamma = \cosh\varphi = \tfrac12(2 + \tfrac12) = \tfrac54$ and $\gamma\beta = \sinh\varphi = \tfrac12(2 - \tfrac12) = \tfrac34$ . Check: $(5/4)^2 - (3/4)^2 = 25/16 - 9/16 = 1$ . ✓

Step 3 — apply, three times. Three collinear boosts of $0.6c$ : $\varphi_{\text{tot}} = 3\ln 2 = \ln 8$ . So $e^{\varphi_{\text{tot}}} = 8$ , and immediately \beta_{\text{tot}} = \frac{8 - 1/8}{8 + 1/8} = \frac{63}{65} = 0.969231, \qquad \gamma_{\text{tot}} = \frac{8 + 1/8}{2} = \frac{65}{16} = 4.0625, \qquad \gamma\beta = \frac{63}{16}. Check: $(65/16)^2 - (63/16)^2 = (65^2 - 63^2)/256 = 256/256 = 1$ . ✓

Step 4 — cross-check the hard way. Compose two: $(0.6+0.6)/(1+0.36) = 15/17$ . Compose the third: $(15/17 + 3/5)/(1 + 9/17) = 63/65$ . ✓ Identical — but note how much more work it was, and that the intermediate $15/17$ had to be carried exactly to avoid rounding.

Step 5 — the boundary number. Galileo would report $1.8c$ . Relativity reports $0.969c$ , and it reports it as the exact rational $63/65$ , with $\gamma = 65/16$ exactly. The general rule for this ladder is $\beta_N = (4^N - 1)/(4^N + 1)$ , since $e^{\varphi_N} = 2^N$ — which is why $\beta = 0.6$ is the velocity that makes every relativity problem come out in clean fractions. Its rapidity is a logarithm of $2$ .

Physics of the portrait

Rapidity has no discoverer to put in a portrait, because the mathematics was finished long before the physics needed it. The portrait therefore belongs to the functions themselves.

Origins · 1 minHyperbolic functions and how this was discovered

Hyperbolic functions

1757–1911 · Bologna · Mulhouse · Zagreb · Cambridge

The functions were introduced by Vincenzo Riccati (1707–1775) in Bologna, who defined $\sinh$ and $\cosh$ in 1757 in the first volume of his Opusculorum ad res physicas et mathematicas pertinentium by way of the geometry of the unit hyperbola $X^2 - Y^2 = 1$ — deliberately in parallel with the circular functions on $X^2 + Y^2 = 1$ . In the Institutiones Analyticae (1765–1767), written with his student Girolamo Saladini, he set down the addition formulas, including the one that would become Einstein's velocity-composition law. Johann Heinrich Lambert (1728–1777) systematised them in the following decade, credited Riccati for the names, and fixed the abbreviations still in use. For a century and a half they were a curiosity of analysis. Then Minkowski's 1908 geometry made $\cosh$ and $\sinh$ the natural coordinates of a boost; Vladimir Varičak (1865–1942), a Croatian mathematician who had spent 1903–1908 working on Bolyai–Lobachevsky geometry, published in 1910 the reading of special relativity as hyperbolic geometry, in which velocities compose by the hyperbolic triangle rule; and in 1911 Alfred Arthur Robb (1873–1936), in Optical Geometry of Motion, gave the hyperbolic angle its physical name: rapidity. A formula written for the pleasure of it in 1765 turned out, in 1911, to be a law of nature.

History

1757

— Vincenzo Riccati defines the hyperbolic sine and cosine from the geometry of the unit hyperbola.

1765

— Riccati and Girolamo Saladini publish the hyperbolic addition formulas in the Institutiones Analyticae.

1770

— Johann Heinrich Lambert systematises the functions and fixes the modern notation.

1908

— Minkowski's Cologne lecture makes spacetime a geometry with an indefinite metric, in which boosts are hyperbolic rotations.

1910

— Vladimir Varičak reads special relativity as Bolyai-Lobachevsky geometry: velocities compose by the hyperbolic triangle rule.

1911

— Alfred A. Robb, in Optical Geometry of Motion, names the hyperbolic angle of a boost the rapidity.

1912

— Varičak's non-Euclidean interpretation of relativity is published in full.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§2, 9 · Taylor & Wheeler, Spacetime Physics, 2nd ed., ch. 8 ("velocity parameter") · Landau & Lifshitz, The Classical Theory of Fields, §4 · Robb, Optical Geometry of Motion (Heffer, Cambridge, 1911) · Varičak, "Die Relativtheorie und die Lobatschefskijsche Geometrie", Physikalische Zeitschrift 11, 93 (1910).

Provenance. Signature $(+,-,-,-)$ . All numbers recomputed in verify_aq-sr-rapidity.py (92 checks): the anchor $\operatorname{artanh}(3/5) = \ln 2$ with $\gamma = 5/4$ , $\gamma\beta = 3/4$ ; the group law $\Lambda(\varphi_1)\Lambda(\varphi_2) = \Lambda(\varphi_1+\varphi_2)$ ; the uniqueness of $\operatorname{artanh}$ as the additive parametrisation; the ladders $\beta_N = (3^N-1)/(3^N+1)$ and $(4^N-1)/(4^N+1)$ ; $\beta_{10} = 29524/29525 = 0.9999661304$ with $\gamma_{10} = 59050/486$ ; the LHC proton at $\varphi = 9.536$ ; and $1g$ for one Julian year giving $\varphi = 1.0323$ , $\beta = 0.7748$ .

Scope. Non-collinear composition, the rotation it induces, and the Rindler treatment of constant proper acceleration are not derived here — see Hyperbolic Geometry of Boosts, Thomas Precession, and Relativistic Acceleration. See also: aq-sr-rapidity-hyperbolic-boosts.

Builds on. Hyperbolic Geometry of Boosts · The Spacetime Interval · Four-Vectors. Leads to. Lorentz Invariance · Causal Structure of Spacetime · Spacetime Diagrams.

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 you can measure is "the rapidity" in the way a speedometer measures speed — but the group of boosts is a physical structure, and rapidity is its canonical coordinate, the one in which composition is addition. Compare angle: no instrument reads "angle" directly either, yet nobody doubts that rotating by $30°$ and then $40°$ is rotating by $70°$ . Rapidity is the angle of a boost, and it earns its reality the same way.

If rapidity is so much better, why is every textbook formula written in terms of γ and β?

Historical accident, mostly, plus the fact that $\beta$ is what you measure. But the moment a problem involves composing boosts — chained frames, successive accelerations, collider kinematics, the relativistic rocket — the $\beta$ formulation becomes a swamp and the $\varphi$ formulation becomes trivial. A useful working rule: if you are transforming between two frames, use $\gamma$ and $\beta$ ; if you are composing transformations, switch to $\varphi$ first and switch back at the end.

Does rapidity add for boosts in different directions?

No, and this is the one trap in the subject. Rapidity is additive along a single axis; that is a statement about a one-parameter subgroup. Compose a boost along $x$ with a boost along $y$ and the result is not a pure boost at all — it is a boost together with a rotation, whose angle is not zero and whose existence is the source of Thomas precession. Treating rapidity as a vector that adds componentwise is wrong, and it is wrong at second order, which is exactly where the interesting physics is. See Hyperbolic Geometry of Boosts.

What is the rapidity of a photon?

Infinite, which is another way of saying that it does not have one. $\varphi = \operatorname{artanh}(1)$ diverges. This is the same fact as "light has no rest frame": there is no finite boost that brings a photon to rest, because there is no finite rapidity you can subtract from infinity. Physically, the light cone is invariant — a boost slides null directions along themselves, rescaling them by $e^{\pm\varphi}$ but never tilting them off the cone — so no amount of rapidity gets you there.

Why do particle physicists use pseudorapidity instead?

Because true rapidity $y = \tfrac12\ln\frac{E + p_z c}{E - p_z c}$ requires knowing the particle's energy, which requires knowing its mass, which a tracking detector often does not. Pseudorapidity $\eta = -\ln\tan(\theta/2)$ , where $\theta$ is the polar angle from the beam, depends only on the direction of the track — and in the ultra-relativistic limit $E \gg mc^2$ it converges to the true rapidity. It is a purely geometric stand-in, adopted because it is measurable, and it inherits rapidity's crucial property that longitudinal boosts shift it by a constant.

The formula φ = aτ/c says a 1g rocket gains rapidity forever. Is there no limit at all?

None on the rapidity, and that is the point. After one year of proper time $\varphi = 1.03$ ; after ten, $\varphi = 10.3$ and $\gamma = 15\,000$ ; after fifty, $\varphi = 51.6$ and $\gamma$ is around $10^{22}$ . The velocity saturates immediately and thereafter changes by amounts too small to write down, but the rapidity — and with it the energy, the momentum, and the crew's ability to cross the galaxy within a lifetime — grows without bound. Nothing forbids arbitrarily large rapidity. What is forbidden is arbitrarily large velocity, and those two statements only look like the same statement if you are using the wrong variable.