Four-Velocity

Velocity composition is the first formula in relativity that looks ugly. Where Galileo had $u' = u - V$ — linear, symmetric, forgettable — Einstein has

u' = \frac{u - V}{1 - uV/c^2},

and generations of students have been told that this nonlinearity is the price of a universal speed limit: velocities cannot simply add, because if they did, two additions of $0.6c$ would give $1.2c$ . That story is true as far as it goes, and it explains nothing. Why that formula? Why a denominator at all, and why that one? Here is the answer, and it is the thesis of this article. The nonlinearity is an artefact of taking a ratio. The underlying object — the four-velocity $U^\mu = \gamma(c,\vec{v}\,)$ — transforms linearly, with the same matrix $\Lambda$ that moves every other four-vector. Ordinary velocity is not that object; it is the ratio of two of its components, $u = c\,U^1/U^0$ . Boost $U^\mu$ linearly, then form the ratio, and the $\gamma$ 's cancel between numerator and denominator — leaving behind exactly the denominator $1 - uV/c^2$ that looked so arbitrary. Composition of velocities is nonlinear for the same reason that adding slopes is nonlinear: slopes are quotients, and quotients of linear things are not linear. That reframing is worth more than the formula. It says the awkwardness was never in the physics; it was in our insistence on describing motion with a quantity that divides by a coordinate. Divide by the invariant proper time instead and everything is linear again — and a second, stranger quantity comes with it. The spatial part of $U^\mu$ is $\vec{\eta} = \gamma\vec{v}$ , the proper velocity: lab metres per unit of the traveller's own time. It is unbounded. A muon at $\beta = 0.995$ has $\eta = 9.96\,c$ , and nothing whatsoever is wrong.

Key takeaways

Signature $(+,-,-,-)$ : $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ . Recalled by name from Four-Vectors (`aq-sr-four-vectors`), which derives them: $\;U\cdot U = +c^2\;$ and $\;A\cdot U = 0$ . This article does not re-derive those; it uses them.
$U^\mu \equiv dx^\mu/d\tau = \gamma\,(c,\vec{v}\,)$ . The definition divides by the invariant $d\tau$ , never the coordinate $dt$ — which is the whole reason it transforms as a four-vector at all.
Owned result: velocity addition is the RATIO of two LINEARLY-transforming components. Boost $U^\mu$ with $\Lambda$ and form $u' = c\,U'^1/U'^0$ ; the $\gamma_u$ cancels top and bottom and out falls $u' = (u - V)/(1 - uV/c^2)$ . The notorious nonlinearity of velocity composition is an artefact of taking a quotient — the underlying object is linear.
Owned result: proper velocity (celerity) $\vec{\eta} = \gamma\vec{v}$ is unbounded, and it is not a velocity. It is lab distance per unit of the traveller's proper time — metres of their ruler per second of its watch. The Frisch–Smith muon at $\beta = 0.995$ covers $1907$ m of lab distance in $0.6385\ \mu$ s of its own time: $\eta = 2.99\times10^{9}\ \mathrm{m/s} = 9.96\,c$ . Nothing exceeds $c$ ; $\eta$ is simply not the thing that is bounded.
Owned result: $\gamma = U^0/c = dt/d\tau$ . Time dilation is not an extra postulate — it is a component reading of the four-velocity.
A photon has no four-velocity at all, because $d\tau = 0$ along a null worldline and $dx^\mu/d\tau$ does not exist. Never write $\gamma = \infty$ ; $\gamma$ is undefined there.
Key idea: build kinematics out of objects that transform linearly, and take ratios only at the very end. Every "strange" formula in relativistic kinematics — velocity addition, aberration, the Doppler factor — is a linear four-vector statement wearing the disguise of a quotient.

The core insight

The four-velocity is the unit tangent to a worldline, scaled by $c$ . Its length is the same for every particle in the universe, $U\cdot U = +c^2$ — a fast particle and a slow one differ only in where their arrow points in spacetime, never in how long it is. All the kinematics is in the direction, and boosts merely swing that arrow around a fixed hyperbola.

Where this lives today

Four-velocity is the object every relativistic simulation actually integrates. In accelerator physics, particle-tracking codes advance $U^\mu$ rather than $\vec{v}$ , because the equation of motion $m\,dU^\mu/d\tau = qF^{\mu\nu}U_\nu$ is linear in $U$ and manifestly covariant, whereas the same physics in terms of $\vec{v}$ acquires the $\gamma^3$ and $\gamma^2$ factors that make numerical integration near $v \approx c$ a misery. The norm $U\cdot U = c^2$ doubles as a running error check: if a tracking code's $U\cdot U$ drifts, the integrator is losing accuracy, and the drift is visible long before the trajectory looks wrong. Proper velocity has a quieter but genuine life in relativistic astrophysics and in mission design. When a jet from an active galactic nucleus is described as "moving at $\gamma \approx 10$ ", it is $\eta$ — not $v$ — that governs how much of the host galaxy the plasma traverses per unit of its own elapsed time, and therefore how much it can cool or radiate on the way. And in the arithmetic of an imagined starship, $\eta$ is the honest figure of merit: a crew at constant proper acceleration accumulates $\eta$ without limit, and can in principle cross the galaxy in a human lifetime of their own, while the galaxy ages by a hundred thousand years. Nothing about that scenario exceeds $c$ . It is entirely a statement about which clock you divide by.

Origins · 1 minFour-velocity and how this was discovered

Four-velocity

1908–1909 · Minkowski's "Bewegungsvektor" · Einstein's proper time · the unit tangent that made dynamics covariant

The four-velocity is Hermann Minkowski's, introduced in the same burst of work that produced spacetime itself. In "Raum und Zeit" (Cologne, 21 September 1908) and in the longer 1908 memoir "Die Grundgleichungen für die elektromagnetischen Vorgänge in bewegten Körpern", he defined the Bewegungsvektor — the motion vector — as the derivative of the spacetime position with respect to what he called the Eigenzeit, "proper time", a word he coined and which has never been improved on. The move was decisive precisely because of what it divided by: dividing by the invariant $\tau$ rather than the frame-dependent $t$ was the step that made a covariant mechanics possible at all, and every four-vector in physics is manufactured by that one operation. Minkowski, who had taught Einstein mathematics at the Zürich Polytechnic, did not live to see the formalism adopted: he died of a ruptured appendix on 12 January 1909, aged 44, four months after the Cologne lecture. Einstein at first dismissed the four-dimensional reformulation as superfluous learnedness; within a few years he was calling it indispensable, and general relativity was written in it.

History

1908

— Minkowski's "Die Grundgleichungen für die elektromagnetischen Vorgänge in bewegten Körpern" introduces the four-dimensional formalism and the motion vector dx/dτ.

1908

— In the Cologne lecture "Raum und Zeit" Minkowski names proper time Eigenzeit and makes the worldline's unit tangent the basic kinematic object.

1909

— Minkowski dies of appendicitis on 12 January, aged 44; "Raum und Zeit" is published posthumously.

1910

— Sommerfeld recasts Minkowski's matrix calculus as a four-dimensional vector algebra, giving four-vectors the form still in use.

1911

— Max von Laue's Das Relativitätsprinzip, the first relativity monograph, adopts the four-vector formalism throughout.

1963

— Frisch and Smith measure muon survival between the Mount Washington summit and Cambridge, giving proper time and the four-velocity's γ their most-cited experimental demonstration.

The physical picture

The velocity you learned first, $\vec{v} = d\vec{x}/dt$ , has a defect that is invisible in Newtonian physics and fatal in relativity: it divides a four-vector by a coordinate. The numerator $dx^\mu$ transforms with $\Lambda$ ; the denominator $dt$ is frame-dependent and transforms too. The quotient therefore transforms in some messy way that is neither linear nor covariant — which is precisely what the velocity-addition formula is telling you when it refuses to be a sum. The repair is to divide by something every frame agrees on. Proper time $d\tau$ — the reading of a clock carried along the worldline — is a Lorentz scalar. So

U^\mu \equiv \frac{dx^\mu}{d\tau} = \frac{dt}{d\tau}\cdot\frac{dx^\mu}{dt} = \gamma\,(c,\vec{v}\,)

is a four-vector by construction, and the lone $\gamma$ that appears is not a fudge: it is $dt/d\tau$ , the Jacobian between the coordinate you should not have used and the invariant you should have. Two things fall out immediately, and both are recalled here from `aq-sr-four-vectors` rather than re-derived: the norm $U\cdot U = \gamma^2(c^2 - v^2) = +c^2$ is the same for every particle at every speed, and differentiating that constant gives $A\cdot U = 0$ . Every massive worldline in the universe carries a four-velocity of identical Minkowski length. Speeding up does not lengthen the arrow; it rotates it toward the light cone.

Figure

PUNCHLINE — velocity composition is a RATIO of linear things. TOP: the four-velocity U = γ(c, u) transforms LINEARLY under Λ — the boost slides it along the hyperbola U·U = c², a hyperbolic rotation, nothing more. BOTTOM: ordinary velocity is the SLOPE u/c = U¹/U⁰ of that arrow, and slopes of linearly-transformed vectors do not add. At u = 0.9c boosted by V = 0.6c: U = (2.2942, 2.0647)c → U′ = (1.3191, 0.8603)c, and the ratio 0.8603/1.3191 = 0.6522 IS the velocity-addition answer u′ = 0.6522c. The γ cancels top and bottom — and the cancellation leaves behind exactly the denominator 1 − uV/c² that looked so arbitrary

Look at what the boost did. It did not "add" anything to the four-velocity; it rotated it along the hyperbola, which is all a Lorentz boost has ever been. The arrow's length is untouched — it cannot change, it is $c$ — so the only thing a boost can alter is where the arrow points. And "where it points" is a slope, $U^1/U^0$ . Slopes do not add under linear maps; that is elementary, and it is the entire mystery of velocity composition.

Figure

PUNCHLINE — proper velocity η = γv is UNBOUNDED, and it is not a velocity. The blue curve, ordinary velocity, saturates at c: no matter how much energy you spend, v never reaches the dashed line. The green curve, η = γv, runs off to infinity — it is lab metres per second of the TRAVELLER'S OWN clock. The Frisch–Smith muon at β = 0.995 sits where marked: it crosses 1907 m of lab distance in 0.6385 μs of its own proper time, giving η = 2.99×10⁹ m/s = 9.96 c. Nothing has exceeded c. The muon's speed past the mountain is still 0.995c; η is simply a different quantity, and it is the honest formalisation of "the muon sees the mountain rush past"

Proper velocity is the quantity the traveller actually cares about. Ask an interstellar navigator "how far can I get in a year?" and $\eta$ is the answer, not $v$ : at $\eta = 9.96\,c$ the muon covers nearly ten light-seconds of lab distance per second of its life. Nothing outruns light — a light signal launched with the muon still arrives first, because it too crosses $1907$ m and it does so in $6.36\ \mu$ s of lab time against the muon's $6.39\ \mu$ s. The muon's speed is $0.995c$ and always was. What is unbounded is not how fast it goes but how much of the universe it can cross before it dies.

The core equations

U^{\mu} \equiv \frac{dx^{\mu}}{d\tau} = \gamma\,(c,\ \vec{v}\,) = \big(\gamma c,\ \gamma\vec{v}\,\big), \qquad \gamma = \frac{1}{\sqrt{1-v^2/c^2}} \qquad (1)

where: $\tau$ — proper time, a Lorentz scalar; $\vec{v} = d\vec{x}/dt$ — ordinary velocity. Physical reading: the definition is the whole content. Divide by $d\tau$ (invariant) and you get a four-vector; divide by $dt$ (a coordinate) and you do not. The factor $\gamma$ is exactly $dt/d\tau$ .

U\cdot U = \eta_{\mu\nu}U^\mu U^\nu = \gamma^2c^2 - \gamma^2v^2 = \gamma^2c^2\!\left(1 - \frac{v^2}{c^2}\right) = \boxed{\;+\,c^{2}\;} \qquad (2)

where: $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ . Physical reading: $+c^2$ in this signature, not $-c^2$ . Derived at `aq-sr-four-vectors` and recalled here by name; together with $A\cdot U = 0$ it is the structural backbone of the whole dynamics chapter. Every massive particle's four-velocity has the same length, so $U^\mu/c$ is a unit timelike tangent to the worldline. The two component readings — and this is where the article's owned results live:

\gamma = \frac{U^{0}}{c} = \frac{dt}{d\tau}, \qquad\qquad \boxed{\;\vec{\eta} \equiv \vec{U} = \gamma\vec{v}\;}\quad\text{(proper velocity, or celerity)} \qquad (3)

where: $\vec{\eta}$ — the spatial part of $U^\mu$ , in m/s. Physical reading: the time slot of the four-velocity is the time-dilation factor — so time dilation is not an extra axiom bolted on, it is a component of an object we already had. The space slot is lab distance per unit proper time, and it is unbounded: $\eta \to \infty$ as $v \to c$ . It is not a velocity and it is not capped by one.

\boxed{\;u' = c\,\frac{U'^{1}}{U'^{0}} = \frac{u - V}{1 - uV/c^{2}}\;} \qquad (4)

where: $U'^\mu = \Lambda^\mu{}_\nu U^\nu$ — the four-velocity boosted linearly. Physical reading: this is the owned result. Velocity addition is not a law that had to be discovered separately; it is what you get when you take the ratio of two components of a linearly-transforming object. The nonlinearity lives in the ratio, not in the physics.

Where the four-velocity earns its keep

Derivation AVelocity addition falls out of a linear boost, once you take the ratio

This is the article's thesis and its owned result, so it is worth being precise about what is being claimed. It is not that the velocity-addition formula is derived here for the first time — it can be had directly from the Lorentz transformation of coordinates. The claim is about why it has the shape it has: that its notorious nonlinearity is not a feature of the physics but a consequence of describing motion by a quotient of components rather than by the components themselves. We boost the four-velocity, which is a purely linear operation, and then watch the awkward denominator assemble itself out of nothing but a cancellation.

Step 1. Write the object. A particle moves at $u$ along $x$ in frame $S$ , so $U^\mu = \gamma_u(c,\ u,\ 0,\ 0)$ with $\gamma_u = (1-u^2/c^2)^{-1/2}$ . Why: note this is a genuine four-vector — built by dividing $dx^\mu$ by the scalar $d\tau$ — so the next step is legal, which it would not be for the naive tuple $(c, \vec{v}\,)$ .

Step 2. Boost it, linearly. Frame $S'$ moves at $V = \beta c$ along $x$ . Then $\Lambda$ acts on the components with no denominators anywhere: $U'^0 = \gamma_V\big(U^0 - \beta U^1\big) = \gamma_V\gamma_u\big(c - \beta u\big)$ and $U'^1 = \gamma_V\big(U^1 - \beta U^0\big) = \gamma_V\gamma_u\big(u - \beta c\big)$ . Why: this is the crux, and it is worth pausing on. There is no nonlinearity here. The transformation is a matrix acting on a column. Whatever ugliness appears next cannot have come from the boost.

Step 3. Take the ratio, and watch the cancellation. The particle's speed in $S'$ is $u' = c\,U'^1/U'^0$ , because $U'^\mu = \gamma_{u'}(c, u', 0, 0)$ by definition of what a four-velocity is in any frame. So

u' = c\,\frac{\gamma_V\gamma_u(u - \beta c)}{\gamma_V\gamma_u(c - \beta u)} = c\,\frac{u - V}{c - uV/c} = \frac{u - V}{1 - uV/c^{2}}.

Why: $\gamma_V$ and $\gamma_u$ both cancel identically. The famous denominator $1 - uV/c^2$ is not a new physical input — it is the $c - \beta u$ from the time component, divided through by $c$ . The nonlinearity was manufactured by the division, and by nothing else.

Step 4. Read the moral, and check it does what it must. Because $u'$ is a ratio of the components of a vector confined to the hyperbola $U\cdot U = c^2$ , and because a boost can only slide points along that hyperbola, $|u'| < c$ is automatic — the arrow can approach the light cone but never cross it. Why: the speed limit is not enforced by the formula; it is enforced by the geometry, and the formula merely reports it. This is why velocity composition never overshoots $c$ no matter how many boosts you compose: you are rotating a fixed-length timelike arrow, and no amount of rotation makes it null.

Step 5. Confirm with numbers. $u = 0.9c$ , $V = 0.6c$ : $\gamma_u = 2.2942$ , so $U^\mu = (2.2942,\ 2.0647)c$ . With $\gamma_V = 1.25$ : $U'^0 = 1.25(2.2942 - 0.6\times 2.0647)c = 1.3191c$ and $U'^1 = 1.25(2.0647 - 0.6\times 2.2942)c = 0.8603c$ . Ratio: $0.8603/1.3191 = \mathbf{0.6522}$ . And directly: $u' = (0.9-0.6)/(1-0.54) = 0.3/0.46 = \mathbf{0.6522}c$ . Why: identical, as they must be — and note $U'^0 = 1.3191 = \gamma_{u'}$ , so the boosted object is exactly the four-velocity built from $u'$ , which is the four-vector property asserting itself. *Verified symbolically in verify_aq-sr-four-velocity.py: $\Lambda$ applied to $\gamma_u(c,u,0,0)$ gives $\gamma_{u'}(c,u',0,0)$ identically for symbolic $u, V$ , with $u' = (u-V)/(1-uV/c^2)$ ; the $\gamma$ -cancellation shown explicitly; $|u'| < c$ proved for all $|u|,|V| < c$ ; and the numbers above reproduced to 4 decimals.*

Go deeper · Derivation B · 2 minProper velocity is unbounded — and it is not a velocity

Derivation BProper velocity is unbounded — and it is not a velocity

The spatial part of the four-velocity is $\gamma\vec{v}$ , not $\vec{v}$ , and that stray $\gamma$ is not a blemish to be apologised for. It is a physically meaningful quantity in its own right — the one an interstellar traveller would actually plan with — and it is the honest way to say the thing that pop-science says dishonestly: "the muon sees the mountain rush past". This card computes it for the real experiment and confronts the fact that it comes out larger than $c$ , which is not a problem, because it is not a speed.

Step 1. Define it and see what it measures. $\vec{\eta} \equiv \gamma\vec{v} = d\vec{x}/d\tau$ : lab displacement per unit of the traveller's proper time. Why: read the two frames in that sentence. The numerator is measured with the lab's rulers; the denominator with the traveller's watch. A quantity built from two different frames' clocks and rods is not obliged to respect a speed limit that applies to quantities built within one frame.

Step 2. Show it is unbounded. $\eta = \gamma v = v/\sqrt{1-v^2/c^2} \to \infty$ as $v \to c$ , whereas $v$ itself saturates. Inverting, $v = \eta/\sqrt{1 + \eta^2/c^2}$ , which is $< c$ for every finite $\eta$ . Why: this inversion is the safety proof. However large $\eta$ gets, the corresponding $v$ is still below $c$ . The two quantities are in one-to-one correspondence, and the map takes the whole infinite range of $\eta$ onto the open interval $(-c, c)$ of $v$ .

Step 3. Compute the real one. The Frisch–Smith muons travel at $\beta = 0.995$ , giving $\gamma = 10.0125$ . They cross the flight path from the Mount Washington summit (1917 m above sea level) to a detector 1907 m below it in Cambridge. Lab flight time: $1907\,/\,(0.995 \times 2.99792458\times10^{8}) = 6.393\ \mu$ s. Their own elapsed time: $6.393/10.0125 = \mathbf{0.6385\ \mu s}$ . Why: this is the experiment's whole point — the muons live about $2.2\ \mu$ s and would classically be gone, but they only experience $0.64\ \mu$ s.

Step 4. Divide, and do not flinch. $\eta = \dfrac{1907\ \mathrm{m}}{0.6385\times10^{-6}\ \mathrm{s}} = \mathbf{2.99\times10^{9}\ m/s} = \mathbf{9.96\,c}$ . Cross-check against the definition: $\gamma\beta = 10.0125 \times 0.995 = 9.962$ . Why: nothing has exceeded $c$ . The muon's speed relative to the mountain is $0.995c$ and a light pulse launched beside it still wins the race. What the muon has done is cross ten light-seconds' worth of lab distance per second of its own life — and that is the quantity that is unbounded, because ageing slowly is not the same as moving quickly. *Verified numerically in verify_aq-sr-four-velocity.py: $\gamma(0.995) = 10.0125$ ; lab time $6.393\ \mu$ s; proper time $0.6385\ \mu$ s; $\eta = 1907/\tau = 2.9867\times10^{9}$ m/s $= 9.962\,c$ , matched independently against $\gamma\beta c$ ; and $v = \eta/\sqrt{1+\eta^2/c^2} < c$ asserted over $\eta/c \in [0.1, 10^{6}]$ .*

Go deeper · Derivation C · 2 minTime dilation is a component of the four-velocity, not an extra postulate

Derivation CTime dilation is a component of the four-velocity, not an extra postulate

Time dilation is usually introduced first, from a light-clock, and then the four-vector formalism arrives later and re-proves it. That ordering hides something. Once the four-velocity exists, time dilation is not a separate result at all — it is the statement that $U^\mu$ has a nonzero time component, read off the definition. The point of this short card is economy: to show how much of the theory's apparatus collapses into "look at the components of one object", which is exactly the payoff the four-vector formalism was built to deliver.

Step 1. Read the time slot. From equation (1), $U^0 = \gamma c$ , so $\gamma = U^0/c$ . But $U^0 = d(ct)/d\tau$ , hence $\boxed{\gamma = dt/d\tau}$ . Why: that is time dilation, stated as an identity. Coordinate time advances faster than proper time by exactly the factor sitting in the four-velocity's time slot — no light-clock, no thought experiment, just a component.

Step 2. Recover the norm as a consistency condition. $U\cdot U = c^2$ reads $(dt/d\tau)^2c^2 - (d\vec x/d\tau)^2 = c^2$ , i.e. $c^2\,dt^2 - d\vec{x}^2 = c^2\,d\tau^2$ — which is just $ds^2 = c^2d\tau^2$ , the definition of proper time. Why: the norm and the definition of $\tau$ are the same statement seen from two sides. The formalism is not adding assumptions; it is repackaging one.

Step 3. Take the rest frame and the ultrarelativistic limit. At $v = 0$ : $U^\mu = (c, \vec{0}\,)$ , so a particle at rest still moves through spacetime, at "speed" $c$ , purely in the time direction. As $v \to c$ : $U^0, U^1 \to \infty$ together, but always along $U\cdot U = c^2$ . Why: the arrow never leaves the hyperbola. This is the geometric content of "you cannot reach $c$ " — not that a force fails, but that a timelike vector cannot be rotated into a null one by any finite boost.

Step 4. Note where the construction breaks. For light, $d\tau = 0$ and $U^\mu = dx^\mu/d\tau$ is a division by zero: a photon has no four-velocity. Why: this is not a technicality to be patched with $\gamma = \infty$ — that expression is meaningless, and writing it will lead you into $0 \times \infty$ errors in the dynamics chapter. Light is described by a null four-vector instead: the wave four-vector $k^\mu$ , or a null four-momentum with $p\cdot p = 0$ . The correct statement is " $\gamma$ is undefined", and it is worth saying out loud. *Verified in verify_aq-sr-four-velocity.py: $U^0/c = \gamma = dt/d\tau$ symbolically; $U\cdot U = c^2$ shown equivalent to $ds^2 = c^2d\tau^2$ ; $U^\mu \to (c,\vec 0)$ as $v\to 0$ ; and every occurrence of the forbidden "gamma equals infinity" checked to sit inside an explicit prohibition.*

Limiting cases

Case	Result	Meaning
$v \to 0$	$U^\mu \to (c,\ \vec{v}\,)$ , $\ \eta \to v$	the naive tuple is the low-speed limit of the true one — which is why it looks so plausible
$v \ll c$	$\eta \approx v\big(1 + \tfrac12 v^2/c^2\big)$	proper and ordinary velocity agree to a part in $10^{12}$ for a jet aircraft ( $\tfrac12\beta^2 = 3.5\times10^{-13}$ )
$c \to \infty$	$u' \to u - V$	the denominator $1 - uV/c^2 \to 1$ : Galilean addition is the $c\to\infty$ limit, and the nonlinearity vanishes
$v \to c$	$\eta \to \infty$ , but $v < c$ always	proper velocity is unbounded; ordinary velocity saturates. The arrow runs off along the hyperbola
$u = c$ (light)	$u' = c$ for any $V$	substituting $u = c$ into (4) gives $(c-V)/(1-V/c) = c$ identically: the invariance of $c$ is built in
photon	$d\tau = 0$	no four-velocity exists. $\gamma$ is undefined, not infinite. Use $k^\mu$ with $k\cdot k = 0$

Common trapA photon has no four-velocity. Never write gamma = infinity

Along a null worldline $d\tau = 0$ , so $U^\mu = dx^\mu/d\tau$ simply does not exist — the derivative is a division by zero, not a large number. It is tempting to paper over this with " $\gamma = \infty$ ", and it is a mistake with real consequences downstream: in the dynamics chapter it produces the meaningless $E = \gamma mc^2 = \infty \times 0$ for a photon, which is exactly how people talk themselves into treating $\gamma m$ as though it were a mass. The disciplined statements are: $\gamma$ is undefined for light; light carries a null four-vector ( $k^\mu = (\omega/c, \vec{k}\,)$ with $k\cdot k = 0$ , or $p^\mu$ with $p\cdot p = 0$ ); and its energy–momentum relation is $E = pc$ , obtained without ever mentioning $\gamma$ . See `aq-sr-massless-particles`.

Worked examples

1 · The Frisch–Smith muon: how far it gets on its own clock

Muons produced in the upper atmosphere are detected at the Mount Washington summit (1917 m above sea level) and 1907 m below it in Cambridge, travelling at $\beta = 0.995$ . How much proper time elapses on the muon's clock during the flight, and what is its proper velocity?

Step 1 — identify. Flight path $1907$ m (the detector-to-detector drop — not the summit's 1917 m elevation, which is a different quantity). We need $\gamma$ , the lab time, the proper time, and then $\eta = \Delta x_{\text{lab}}/\Delta\tau$ .

Step 2 — the Lorentz factor. $\gamma = 1/\sqrt{1 - 0.995^2} = 1/\sqrt{1 - 0.990025} = 1/\sqrt{0.009975} = \mathbf{10.0125}$ .

Step 3 — lab time, then proper time. Lab: $\Delta t = \dfrac{1907\ \mathrm{m}}{0.995 \times 2.99792458\times10^{8}\ \mathrm{m/s}} = \mathbf{6.393\ \mu s}$ . The muon's own clock: $\Delta\tau = \Delta t/\gamma = 6.393/10.0125 = \mathbf{0.6385\ \mu s}$ .

Step 4 — the payoff. The muon's mean lifetime is $2.20\ \mu$ s. It experiences only $0.6385\ \mu$ s — about $0.29$ lifetimes — so most of the beam survives, whereas a naive $6.393\ \mu$ s (≈ 2.9 lifetimes) would have killed almost all of it. That is the experiment.

Step 5 — the boundary number. Proper velocity: $\eta = \dfrac{1907\ \mathrm{m}}{0.6385\times10^{-6}\ \mathrm{s}} = \mathbf{2.99\times10^{9}\ m/s} = \mathbf{9.96\,c}$ . Cross-check: $\gamma\beta = 10.0125\times0.995 = 9.962$ . ✓ The muon crosses nearly ten light-seconds of lab distance per second of its own life — and its speed is still $0.995c$ . Those two sentences are both true and they are not in tension, because they divide by different clocks.

2 · Velocity addition, done the long way and the short way

A particle moves at $u = 0.9c$ in the lab. A rocket flies past at $V = 0.6c$ in the same direction. What is the particle's speed in the rocket frame — computed by boosting the four-velocity?

Step 1 — identify. Build $U^\mu$ , apply $\Lambda$ , take the ratio $u' = c\,U'^1/U'^0$ . Then check against equation (4).

Step 2 — the four-velocity. $\gamma_u = 1/\sqrt{1-0.81} = 1/\sqrt{0.19} = 2.2942$ . So $U^\mu = 2.2942\,(c,\ 0.9c) = (\mathbf{2.2942}c,\ \mathbf{2.0647}c)$ .

Step 3 — boost it (linearly!). $\gamma_V = 1/\sqrt{1-0.36} = 1.25$ . $U'^0 = 1.25\,(2.2942 - 0.6 \times 2.0647)\,c = 1.25 \times 1.0553\,c = \mathbf{1.3191}c$ ; $U'^1 = 1.25\,(2.0647 - 0.6 \times 2.2942)\,c = 1.25 \times 0.6882\,c = \mathbf{0.8603}c$ .

Step 4 — take the ratio. $u' = c\,\dfrac{0.8603}{1.3191} = \mathbf{0.6522}c$ .

Step 5 — check, and read the structure. Equation (4) directly: $u' = \dfrac{0.9 - 0.6}{1 - 0.54} = \dfrac{0.3}{0.46} = \mathbf{0.6522}c$ . ✓ And note $U'^0 = 1.3191 = \gamma_{u'}$ exactly — the boosted object is the four-velocity built from $u'$ , which is what "four-vector" means.

Step 6 — the boundary number. The Galilean answer would have been $0.9 - 0.6 = 0.3c$ . The relativistic answer, $0.6522c$ , is more than twice that — and it is still under $c$ , as it must be, because the boost only slid the arrow along the hyperbola. The entire difference between $0.3$ and $0.6522$ is the denominator $1 - uV/c^2 = 0.46$ , and that denominator is a cancelled $\gamma$ , not a new law of nature.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§5–6 · Taylor & Wheeler, Spacetime Physics, 2nd ed., ch. 5–7 (which uses "proper velocity" throughout) · Morin, Special Relativity, ch. 4 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.1 · Hartle, Gravity, ch. 5 · Minkowski, H., "Raum und Zeit", Phys. Z. 10, 104 (1909) · Frisch, D. H. & Smith, J. H., "Measurement of the Relativistic Time Dilation Using μ-Mesons", Am. J. Phys. 31, 342 (1963).

Provenance. Signature $(+,-,-,-)$ , so $U\cdot U = +c^2$ (not $-c^2$ ). Everything is recomputed in verify_aq-sr-four-velocity.py, which also greps this article for each printed value: the symbolic identity $\Lambda\big[\gamma_u(c,u,0,0)\big] = \gamma_{u'}(c,u',0,0)$ with $u' = (u-V)/(1-uV/c^2)$ , and the explicit $\gamma$ -cancellation; $u = 0.9c$ , $V = 0.6c \Rightarrow \gamma_u = 2.2942$ , $U = (2.2942, 2.0647)c$ , $U' = (1.3191, 0.8603)c$ , ratio $= u'/c = 0.6522$ , with $U'^0 = \gamma_{u'}$ exactly; $\gamma(0.995) = 10.0125$ , lab time $6.393\ \mu$ s, proper time $0.6385\ \mu$ s, $\eta = 2.99\times10^{9}$ m/s $= 9.96\,c$ , cross-checked against $\gamma\beta$ ; and $v = \eta/\sqrt{1+\eta^2/c^2} < c$ for all finite $\eta$ . Frisch–Smith anchors follow the corpus ruling: flight path 1907 m, summit elevation 1917 m — different quantities.

Scope. $U\cdot U = +c^2$ and $A\cdot U = 0$ are derived at `aq-sr-four-vectors` and are recalled here by name, not re-derived. The four-acceleration $A^\mu = dU^\mu/d\tau$ is developed at `aq-sr-four-acceleration`. The dynamics — $p^\mu = mU^\mu$ , $E = \gamma mc^2$ , and the energy–momentum relation — is `aq-sr-four-momentum`, and is only forward-pointed here. Massless particles are `aq-sr-massless-particles`.

Builds on. Proper Time · The Lorentz Transformation · The Lorentz Factor · The Minkowski Metric and Four-Vectors. Leads to. Four-Momentum · Four-Acceleration · Relativistic Momentum · Four-Force · Massless Particles.

If the four-velocity's length is always c, what distinguishes a fast particle from a slow one?

Only the direction of the arrow in spacetime. Every massive particle has $U\cdot U = +c^2$ , so all four-velocities lie on one hyperbola, and a boost slides a particle along it without changing its length. A slow particle's arrow points almost straight up the time axis ( $U^\mu \approx (c, \vec{v}\,)$ with a tiny spatial part); a fast one's leans toward the light cone. The popular slogan "everything moves through spacetime at speed $c$ " is a fair gloss on $U\cdot U = c^2$ — at rest you are moving through time at $c$ , and speeding up trades time-motion for space-motion without changing the total.

Isn't proper velocity just a trick to make superluminal travel sound respectable?

No — it is a different physical quantity, and it answers a question ordinary velocity cannot. Ask "how much lab distance can I cross before I die?" and the answer is governed by $\eta = \gamma v$ , not $v$ , because your lifetime is measured on your clock. The muon's $\eta = 9.96\,c$ says it crosses nearly ten light-seconds of lab distance per second of its own life, which is a true and useful fact about it. It does not say the muon outruns light: a photon released alongside it arrives first, and the muon's speed past the mountain is $0.995c$ . Nothing that relativity forbids has happened. What has happened is that we divided by a different clock.

Why can't I just define the four-velocity as (c, v⃗) and be done with it?

Because that object is not a four-vector — it does not transform with $\Lambda$ — and so nothing you build from it is trustworthy. Boost $(c, 0.8c)$ by $V = 0.5c$ and you land on $(0.693c, 0.346c)$ , whose time slot is no longer $c$ : the boosted object is not of the same form, so it cannot be a four-vector. The repair is a single factor of $\gamma$ , and it is forced: $\gamma(c,\vec{v}\,) = (dt/d\tau)(dx^\mu/dt) = dx^\mu/d\tau$ . `aq-sr-four-vectors` proves this in detail; the lesson to carry is divide a four-vector by a Lorentz scalar, never by a coordinate.

Where exactly does the denominator 1 − uV/c² come from?

From the time component of the boosted four-velocity, divided through by $c$ . Boosting gives $U'^0 = \gamma_V\gamma_u(c - \beta u)$ , and when you form the ratio $u' = c\,U'^1/U'^0$ the factor $\gamma_V\gamma_u$ cancels against the identical factor in $U'^1$ , leaving $c(u - V)/(c - uV/c) = (u-V)/(1 - uV/c^2)$ . So the denominator is not an independent physical postulate and not a "correction factor" — it is the time-slot of a linearly-transformed vector, showing up because we chose to describe motion with a quotient. That is the article's whole thesis in one paragraph.

Does the four-velocity formalism prove that nothing can exceed c?

It makes it geometric rather than mysterious. A massive particle's $U^\mu$ is a timelike vector of fixed length $c$ , and a Lorentz boost is a hyperbolic rotation, which slides points along the hyperbola $U\cdot U = c^2$ . No finite rotation carries a timelike vector onto the light cone, so no sequence of boosts — however many, however large — produces $|v| \ge c$ . The velocity-addition formula does not enforce the limit; it reports it. That is why composing $0.9c$ with $0.9c$ gives $0.9945c$ rather than $1.8c$ , and why you can keep composing forever without ever arriving.

What replaces the four-velocity for light?

A null four-vector, which needs no proper time to define. The wave four-vector $k^\mu = (\omega/c,\ \vec{k}\,)$ satisfies $k\cdot k = 0$ , and the four-momentum $p^\mu = \hbar k^\mu$ satisfies $p\cdot p = 0$ , which is the statement $m = 0$ . Neither is built by differentiating with respect to $\tau$ , so neither cares that $\tau$ does not advance. Everything you would want the four-velocity for — Doppler shifts, aberration, energy and momentum — is obtained by boosting $k^\mu$ or $p^\mu$ linearly, exactly as we boosted $U^\mu$ here. The formalism does not break for light; only this one object does.