Proper Time and Proper Length

A muon lives, on average, $2.197\ \mu\mathrm{s}$ . Light itself covers only $659\ \mathrm{m}$ in that time, and a muon — being massive — covers rather less. So when cosmic rays smash into the upper atmosphere and manufacture muons twelve kilometres up, essentially none of them should ever reach the ground. They do. In 1963 David Frisch and James Smith counted them at the Mount Washington summit (1917 m above sea level) and 1907 m below it in Cambridge: 563 per hour up top, 412 per hour at the bottom. The classical bookkeeping says about thirty should have survived. Something is wrong with the clock. Nothing is wrong with the clock. What is wrong is the assumption that there is one clock. The muon carries its own, and it does not tick off the $6.4\ \mu\mathrm{s}$ of laboratory time the flight takes; it ticks off about $0.7\ \mu\mathrm{s}$ — the length of the muon's own path through spacetime, measured with Minkowski's ruler rather than Newton's. That length is the proper time $\tau$ : not a coordinate, not an observer's opinion, but a geometric fact about the worldline, and every inertial observer in the universe computes the same value for it. Its spacelike twin is the proper length $L_0$ , the invariant width of the four-dimensional slab an object sweeps through spacetime. Time dilation and length contraction, the two effects that make relativity feel like a conjuring trick, are nothing but the shadows these invariants cast on a moving observer's axes. What remains once you see them is geometry — and one startling piece of it: in spacetime, unlike in Euclid's plane, the straight path between two points is the longest.

Key takeaways

Signature $(+,-,-,-)$ throughout: $ds^2 = c^2dt^2 - dx^2 - dy^2 - dz^2$ , so timelike means $s^2 > 0$ and spacelike means $s^2 < 0$ . Proper time takes the root of $+ds^2$ ; proper length the root of $-\Delta s^2$ . Getting that minus sign backwards is the commonest error in the subject.
Proper time is the arc length of a timelike worldline: $c\,d\tau = \sqrt{+ds^2}$ , hence $d\tau = dt\sqrt{1 - v(t)^2/c^2} = dt/\gamma(t)$ — valid for arbitrary motion, because only the instantaneous speed enters.
Proper length is the arc length of a spacelike separation whose endpoints are simultaneous in the object's rest frame: $L_0 = \sqrt{-\Delta s^2}$ . Mark the ends simultaneously in a different frame and you cut a different chord of the same world-slab; that chord has length $L_0/\gamma$ . Contraction is slicing, not squeezing.
Frisch and Smith's muons cross $1907\ \mathrm{m}$ in $6.4\ \mu\mathrm{s}$ of laboratory time but only $\approx 0.7\ \mu\mathrm{s}$ of their own. That is why $412$ per hour arrive where the classical estimate allows $\approx 27$ – $31$ — a factor of about fifteen, measured, not inferred.
Key idea: the straight worldline between two timelike-separated events carries the MAXIMUM proper time, exactly reversing Euclid's triangle inequality. Every detour through spacetime costs you aging. The twin "paradox" is not a paradox but a maximisation, and the second variation of the proper-time functional is negative-definite because $d^2\!\big(\!\sqrt{1-v^2/c^2}\big)/dv^2 = -\gamma^3/c^2 < 0$ everywhere below $c$ .

The core insight

Coordinates are opinions; arc length is a fact. Proper time $\tau$ is the Minkowski arc length of a timelike worldline and proper length $L_0$ the arc length of a spacelike one, and because the interval they are built from is Lorentz-invariant, every observer computes the same $\tau$ for the same clock and the same $L_0$ for the same rod. Time dilation $\Delta t = \gamma\,\Delta\tau$ and length contraction $L = L_0/\gamma$ are then not processes happening to clocks and rods — they are the projections of fixed four-dimensional objects onto a tilted set of axes.

Where this lives today

Every satellite in orbit is a proper-time experiment that runs continuously and must not fail. A GPS satellite's onboard clock reads its proper time along its worldline; your receiver must compare that against a clock at rest on the geoid. Different curves, different arc lengths: the satellite's motion costs about $7.2\ \mu\mathrm{s}$ per day of special-relativistic proper time, its higher gravitational potential buys back about $45.8\ \mu\mathrm{s}$ per day, and the net is roughly $+38.6\ \mu\mathrm{s}$ per day — which light crosses in about $11.6\ \mathrm{km}$ . Uncorrected, the system would be useless within hours. The muons never went away either. Storage rings hold them at $\gamma \approx 29$ , stretching the $2.197\ \mu\mathrm{s}$ proper lifetime to about $64\ \mu\mathrm{s}$ in the laboratory — the only reason the muon $g-2$ measurements are possible at all. And a gold nucleus arriving at RHIC with $\gamma$ in the hundreds is, in the laboratory frame, a pancake of thickness $L_0/\gamma$ : its collision geometry is computed from the invariant proper size, then sliced into whichever frame the calculation wants.

The physical picture

Two cars drive from Boston to New York. One takes the interstate; the other wanders through the Berkshires and arrives at the same moment. Their odometers disagree, and nobody finds this mysterious: an odometer measures the length of the path, not the separation of the endpoints. Two clocks that start together at one event and reunite at another are in exactly the same position. A clock is a spacetime odometer. Everything then hinges on how spacetime measures length. In Euclid's plane the recipe is $d\ell^2 = dx^2 + dy^2$ , both signs positive, and the consequence is the triangle inequality: detours lengthen. Minkowski's recipe is $ds^2 = c^2dt^2 - dx^2$ , one sign flipped, and the consequence is the inequality reversed: detours shorten. Spatial motion does not add to a worldline's length, it subtracts from it — so the twin who wanders racks up fewer years. Not because acceleration is corrosive to clocks (the clock postulate, stated below, makes an ideal clock's rate depend only on its instantaneous speed), but because her worldline bent, and a bent timelike worldline is shorter than the straight one joining the same two events.

Figure

Two worldlines, one pair of events. The straight (inertial) path from O to P accumulates 10 years of proper time; the bent path — out at β = 0.6, back at β = 0.6 — accumulates only 8. Each bent leg has Δ(ct) = 5, Δx = 3, so s² = 25 − 9 = 16 and its arc length is exactly √16 = 4 years. Detours through spacetime COST proper time: the reversed triangle inequality, in one picture

Read that figure slowly, because it contains the whole article. The blue worldline is straight and reads $10$ years. The red one is bent, and each of its legs — $5$ years in $ct$ , $3$ light-years in $x$ — has interval $s^2 = 5^2 - 3^2 = 16$ and arc length $\sqrt{16} = 4$ years. Two legs, eight years. The bent path looks longer on the page because the page is Euclidean and lies to you. The dashed curve is the invariant hyperbola $c^2t^2 - x^2 = (10\ \mathrm{ly})^2$ : every event on it lies exactly ten years of proper time from the origin, for every observer. The muons make the same point with harder numbers, which is why they and not the twins are the worked example.

Figure

PUNCHLINE — Frisch and Smith, 1963. Muons counted at the Mount Washington summit: 563 per hour. Flight path 1907 m, flight time 6.4 μs, mean lifetime 2.197 μs. With no time dilation the muon's own clock runs the full 6.4 μs and only ≈ 27–31 per hour should survive. The number counted in Cambridge was 412 per hour. The muon clock ran for ≈ 0.7 μs, not 6.4 — a measured dilation factor of 8.8 ± 0.8

The dashed curve is exponential decay against the muon's own clock — the only clock it has. The classical assumption drags the muon $6.4\ \mu\mathrm{s}$ down that curve and lands at thirty-odd survivors per hour; nature drags it $0.7\ \mu\mathrm{s}$ and lands at four hundred. Either the decay law is wrong, or the muon's clock ran slow. Every other experiment on Earth says the decay law is right.

The core equations

In the signature $(+,-,-,-)$ , with $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ and $dx^\mu = (c\,dt, dx, dy, dz)$ :

ds^2 = \eta_{\mu\nu}\,dx^\mu dx^\nu = c^2dt^2 - dx^2 - dy^2 - dz^2 \qquad (1)

where: $\eta_{\mu\nu}$ — the Minkowski metric; $dx^\mu$ — a spacetime displacement. Physical reading: a timelike step has $ds^2 > 0$ (a massive particle can make it), a null step $ds^2 = 0$ (only light), a spacelike step $ds^2 < 0$ (nothing can). The single minus sign relative to Pythagoras is the whole of relativity. For a timelike step $ds^2$ is positive and its root is real. That root, over $c$ , is what a clock carried along the step reads.

c^2 d\tau^2 = +\,ds^2 \qquad\Longrightarrow\qquad d\tau = \frac{\sqrt{+\,ds^2}}{c} = dt\,\sqrt{1 - \frac{v(t)^2}{c^2}} = \frac{dt}{\gamma(t)} \qquad (2)

where: $\tau$ — proper time; $v(t) = |d\vec{x}/dt|$ — the instantaneous coordinate speed. Physical reading: the root is of $+ds^2$ , not $-ds^2$ . In $(-,+,+,+)$ it is the other way round; mixing the two is the most reliable way to manufacture an imaginary proper time. Note also that only $v$ appears — never $\dot{v}$ . Integrating (2) gives an arc length that does not care how the curve is parametrised:

\tau[\mathcal{C}] = \int_{\mathcal{C}} dt\,\sqrt{1 - \frac{v(t)^2}{c^2}} \;=\; \frac{1}{c}\int_{\lambda_1}^{\lambda_2} \sqrt{+\,\eta_{\mu\nu}\frac{dx^\mu}{d\lambda}\frac{dx^\nu}{d\lambda}}\;d\lambda \qquad (3)

where: $\mathcal{C}$ — a timelike worldline; $\lambda$ — any parameter increasing along it. Physical reading: the integrand is a scalar and $d\lambda$ a differential, so $\tau$ belongs to the curve and to no coordinate system — the precise sense in which proper time is an arc length. (The base boxes this root as $\sqrt{-\eta_{\mu\nu}dx^\mu dx^\nu}$ , which in our signature returns a purely imaginary number for every timelike curve: a transcription from the other convention, repaired here.) The spacelike counterpart is built the same way, with the sign the spacelike case demands:

L_0 = \sqrt{-\,\Delta s^2}\;\Big|_{\text{endpoints simultaneous in the rest frame}} \qquad\Longrightarrow\qquad L = \frac{L_0}{\gamma} = L_0\sqrt{1-\beta^2} \qquad (4)

where: $L_0$ — proper length; $\Delta s^2 < 0$ — the spacelike interval between the endpoint events; $L$ — what a frame in which the object moves at $\beta = v/c$ obtains by marking both ends at one of its instants. Physical reading: the qualifier is not decoration. Which two events you take is the entire content of length contraction. Finally, the object that makes proper time indispensable: divide the displacement by the invariant $d\tau$ rather than the frame-dependent $dt$ , and a genuine four-vector appears, $U^\mu = dx^\mu/d\tau = \gamma(c,\vec{v})$ with $U\cdot U = \gamma^2(c^2-v^2) = +c^2$ — positive $c^2$ in this signature, not $-c^2$ (the base prints $-c^2$ , which belongs to $(-,+,+,+)$ ). Four-Vectors takes this up properly.

Where the invariants come from

Derivation AProper time is the arc length of the worldline

The claim is that the reading of a physical clock is a geometric invariant of its worldline: it depends on the curve and nothing else — not on the coordinates, not on the parameter swept along it, and not on the acceleration the clock is suffering. That is far stronger than time dilation, which speaks only of uniform motion, and it is what licenses us to apply proper time to a muon decelerating in air, a satellite in orbit, or a twin who turns around.

Step 1. Take a timelike displacement: $ds^2 = c^2dt^2 - |d\vec{x}|^2 > 0$ . Define $c^2 d\tau^2 \equiv +\,ds^2$ , so $c\,d\tau = \sqrt{+\,ds^2}$ , a real positive number. Why: the root must be taken of the positive quantity, and in $(+,-,-,-)$ that is $+ds^2$ . Taking $\sqrt{-ds^2}$ here, as the base does, returns $i\sqrt{ds^2}$ — for the timelike pair $(c\Delta t,\Delta x) = (5,3)$ of the first figure, $4i$ years instead of $4$ .

Step 2. Factor $dt$ out. With $|d\vec{x}| = v\,dt$ , $\;c^2d\tau^2 = c^2dt^2(1 - v^2/c^2)$ , hence $d\tau = dt\sqrt{1-v^2/c^2} = dt/\gamma$ . Why: note what this contains — the instantaneous $v$ , and nothing else. No $dv/dt$ appears anywhere. This is the clock postulate: an ideal clock's rate depends on speed, not on acceleration.

Step 3. Check invariance. Under $dx'^\mu = \Lambda^\mu{}_\nu dx^\nu$ and the defining property $\Lambda^{\mathsf T}\eta\,\Lambda = \eta$ , $\;ds'^2 = \eta_{\mu\nu}\Lambda^\mu{}_\alpha\Lambda^\nu{}_\beta dx^\alpha dx^\beta = \eta_{\alpha\beta}dx^\alpha dx^\beta = ds^2$ , so $d\tau$ is invariant too. Why: the clock reading is not agreed by convention — it is forced to be frame-independent by the very identity that defines a Lorentz transformation.

Step 4. Check additivity and reparametrisation-independence. Proper times add along a broken path; and the integrand of (3) is homogeneous of degree one in $\dot{x}$ , so the $d\lambda$ 's cancel and the value survives $\lambda \to f(\lambda)$ . Why: additivity plus reparametrisation-invariance is exactly what it means to be an arc length. The clock is a spacetime odometer. And for light $ds^2 = 0$ identically, so $d\tau = 0$ : no clock, no rest frame. *Verified symbolically in verify_aq-sr-proper-time-proper-length.py: the timelike pair $(5,3)$ gives $s^2 = +16$ and $c\tau = 4$ , while $\sqrt{-s^2}$ returns $4i$ ; $d\tau/dt = 1/\gamma$ ; reparametrising $\lambda \to \mu^2$ leaves the arc length unchanged; $U\cdot U = \gamma^2(c^2-v^2) = +c^2$ .*

Go deeper · Derivation B · 2 minThe inertial worldline maximises proper time

Derivation BThe inertial worldline maximises proper time

Here is the article's owned structural result, and the one that dissolves the twin paradox for good. In Euclidean space the straight line is the shortest path; in Minkowski spacetime the straight timelike worldline is the longest path between two timelike-separated events. The inequality runs the other way, and it does so because of the single minus sign in the metric. We prove it twice — once variationally, where the second variation turns out negative-definite and so exhibits a strict maximum rather than a mere stationary point, and once exactly, on the two-leg path the travelling twin actually flies.

Step 1. Fix two events on the $x = 0$ axis, separated by coordinate time $T$ . Any timelike worldline joining them has $\tau[x] = \int_0^T \sqrt{1 - \dot{x}^2/c^2}\;dt$ with $|\dot{x}| < c$ — an ordinary variational problem with Lagrangian $\mathcal{L}(\dot{x}) = \sqrt{1 - \dot{x}^2/c^2}$ . Why: writing proper time as a functional turns "which twin ages more?" from a puzzle into a calculus problem with a definite answer.

Step 2. First variation. $\mathcal{L}$ depends explicitly on neither $x$ nor $t$ , so Euler–Lagrange collapses to $\dfrac{d}{dt}\dfrac{\partial\mathcal{L}}{\partial\dot{x}} = 0$ , i.e. $\dfrac{\partial\mathcal{L}}{\partial\dot{x}} = -\dfrac{\gamma\dot{x}}{c^2} = \text{const}$ . Why: the conserved quantity is $\gamma v$ — which, times the mass, is relativistic momentum. The variational principle for proper time knows about momentum before any dynamics is introduced.

Step 3. Conclude $v$ is constant. Since $d(\gamma v)/dv = \gamma^3 > 0$ for all $|v| < c$ , the map $v \mapsto \gamma v$ is strictly increasing, hence injective: $\gamma v = \text{const}$ forces $v = \text{const}$ , and with fixed endpoints that is the straight worldline and no other. Why: without the monotonicity check one has only " $\gamma v$ is constant", which could in principle be met by jumping between branches. There are no branches — the map is a bijection onto $\mathbb{R}$ .

Step 4. Second variation — where the maximum is established. $\dfrac{\partial^2 \mathcal{L}}{\partial v^2} = -\dfrac{1}{c^2(1-v^2/c^2)^{3/2}} = -\dfrac{\gamma^3}{c^2}$ , strictly negative for every $|v| < c$ : $-1/c^2$ at $v = 0$ , $-12.07/c^2$ at $\beta = 0.9$ , $-356/c^2$ at $\beta = 0.99$ . So for any non-zero perturbation, $\delta^2\tau = \tfrac12\int_0^T \mathcal{L}''(v_0)(\delta v)^2 dt < 0$ . Why: a stationary point with a negative-definite second variation is a strict maximum — the machine-checkable sense in which the inertial twin ages the most. The sign making it a maximum rather than a minimum is inherited straight from the minus sign in $ds^2$ .

Step 5. Confirm exactly on the twin's path. Out at $\beta$ , back at $\beta$ , total coordinate time $T$ : $\tau_B = T\sqrt{1-\beta^2} = T/\gamma < T = \tau_A$ for every $0 < \beta < 1$ , no approximation; at $\beta = 0.6$ , $\tau_B = 0.8\,T$ . A small wiggle at $\pm\epsilon c$ costs $T\big(1 - \tfrac12\epsilon^2 - \tfrac18\epsilon^4 - \cdots\big)$ : linear coefficient zero, quadratic coefficient negative. Why: the theorem says no path beats the straight one; these numbers say by how much a given path loses — and that the loss is second order, which is why relativistic speeds are needed to see it at all. *Verified symbolically in verify_aq-sr-proper-time-proper-length.py: $\partial\mathcal{L}/\partial v = -\gamma v/c^2$ ; $d(\gamma v)/dv = \gamma^3 > 0$ ; $\partial^2\mathcal{L}/\partial v^2 = -\gamma^3/c^2 < 0$ at $\beta = 0,\,0.3,\,0.6,\,0.9,\,0.99$ ; $\tau(\epsilon) = 1 - \epsilon^2/2 - \epsilon^4/8$ with zero linear term; $\tau_{\text{bent}} = T\sqrt{1-\beta^2} < T$ for every $\beta$ .*

Go deeper · Derivation C · 2 minProper length, and which chord of the slab you cut

Derivation CProper length, and which chord of the slab you cut

Proper length looks like the mirror image of proper time, and very nearly is — but it hides a trap that no algebra will save you from unless you name it. The invariant $\sqrt{-\Delta s^2}$ is a property of a pair of events, not of an object. A rod is not a pair of events; it is a slab in spacetime, and to extract a number you must first choose two events, one on each edge. Different choices are different chords, with different lengths — and both $L_0$ and the contracted $L$ turn out to be legitimate proper distances of different chords.

Step 1. Cut the rest-frame chord. A rod at rest in $S'$ occupies $0 \le x' \le L_0$ ; take its endpoint events simultaneous in $S'$ , $F_1 = (0,0)$ and $F_2 = (0,L_0)$ . Their interval is $\Delta s^2 = 0 - L_0^2 = -L_0^2 < 0$ , spacelike as it must be, so $\sqrt{-\Delta s^2} = L_0$ . Why: this is the definition of proper length — the invariant length of the chord the rod itself calls "the rod, now", and every observer computes the same value for it.

Step 2. Cut the lab chord instead. The lab $S$ , in which the rod moves at $\beta$ , marks both ends at a single lab time: $\Delta t = 0$ , $\Delta x = L$ . Then $\Delta x' = \gamma(\Delta x - v\Delta t) = \gamma L$ ; but $\Delta x'$ between the slab's two edges is the rest length, so $\gamma L = L_0$ and $L = L_0/\gamma$ . Why: length contraction, derived without once asking what happens "to" the rod. Nothing happens to the rod. The lab cut a different chord.

Step 3. Compute the interval of that lab chord: $\Delta s^2 = 0 - L^2 = -(L_0/\gamma)^2$ , so $\sqrt{-\Delta s^2} = L_0/\gamma = L$ . Both chords are honest spacelike separations, and each one's proper distance equals the length measured in the frame where its two events are simultaneous. Why: this dissolves the apparent contradiction that " $L_0 = \sqrt{-\Delta s^2}$ " and " $L = \sqrt{-\Delta s^2}$ " are both true — they are true of different $\Delta s^2$ . The invariant is working perfectly; the ambiguity was in the phrase "the rod".

Step 4. See why the chords differ, and put a number on it. In $S'$ the lab's marking events have $\Delta t' = \gamma(\Delta t - v\Delta x/c^2) = -\gamma v L/c^2 \ne 0$ : they are not simultaneous in the rod's frame. The lab marked the far end at an earlier moment of the rod's life than the near end, and the rod moved in between. A rod of proper length $2\ \mathrm{m}$ at $\beta = 0.6$ ( $\gamma = 1.25$ ) is measured as $1.6\ \mathrm{m}$ in the lab and $2\ \mathrm{m}$ in its own frame. Why: length contraction is the relativity of simultaneity seen sideways, and every ladder-and-barn paradox is this restated. A rod is no more shortened by a boost than a cucumber is shortened by being sliced on a bias — both slices are real; one is simply longer. *Verified symbolically in verify_aq-sr-proper-time-proper-length.py: the rest-frame chord has $\Delta s^2 = -L_0^2$ ; the lab chord has $\Delta s^2 = -(L_0/\gamma)^2$ ; $\Delta x' = \gamma L = L_0 \Rightarrow L = L_0/\gamma$ ; the lab's marking events have $\Delta t' \ne 0$ in the rod frame; $L_0 = 2\ \mathrm{m}$ at $\beta = 0.6$ gives $L = 1.6\ \mathrm{m}$ .*

Limiting cases

Case	Result	Meaning
Rest frame, $\beta = 0$	$\tau = t$ , $L = L_0$	the definition of "proper"
Low speed, $\beta \ll 1$	$d\tau/dt = 1 - \tfrac12\beta^2 - \tfrac18\beta^4 - \cdots$	no linear term — Newtonian time, with an $O(\beta^2)$ deficit underneath
Null worldline, $v = c$	$d\tau = 0$	light carries no clock; "the photon's point of view" is not a thing
Ultra-relativistic, $\gamma \gg 1$	$L \to 0$ , $\Delta t = \gamma\Delta\tau \to \infty$	the muon regime, $\gamma \approx 10$
Straight worldline	$\tau$ is a strict maximum	reversed triangle inequality; $\partial^2_v\mathcal{L} = -\gamma^3/c^2 < 0$
Any detour, $\delta v \ne 0$	$\delta\tau = 0$ , $\delta^2\tau < 0$	wandering costs proper time at second order — easy to miss
Galilean limit, $c \to \infty$	$\gamma \to 1$	absolute time and absolute length re-emerge

Common trapThe minus sign is not a matter of taste — and half the textbooks put it in the other place

This article uses $(+,-,-,-)$ : timelike means $s^2 > 0$ , $c\,d\tau = \sqrt{+ds^2}$ , $L_0 = \sqrt{-\Delta s^2}$ , $U\cdot U = +c^2$ . Rindler and most of particle physics agree; Taylor–Wheeler and Misner–Thorne–Wheeler use $(-,+,+,+)$ , in which every one of those signs flips. No physics whatsoever depends on the choice — and every intermediate sign does. The dictionary is simply $s^2 \to -s^2$ throughout. The failure mode is not picking the "wrong" convention; it is mixing them, whereupon a timelike interval yields an imaginary proper time and nobody notices, because a careless hand returns "4" from both $\sqrt{16}$ and $\sqrt{-16}$ . Read the metric before you read the equation.

Worked examples

1 · The muons that should not be there

Frisch and Smith (1963) counted muons at the Mount Washington summit (1917 m above sea level) and 1907 m below it in Cambridge. The summit rate was $563$ per hour, for muons in the velocity band $0.9950\,c$ to $0.9954\,c$ . The muon mean lifetime is $\tau_0 = 2.197\ \mu\mathrm{s}$ . How many should arrive at sea level — classically, and relativistically? And how many did?

Step 1 — identify. Decay is governed by the muon's proper time, not the laboratory flight time: $N = N_0\,e^{-\tau/\tau_0}$ , with $\tau$ the arc length of its worldline.

Step 2 — the lab flight time. The flight path is the drop between the two counting stations, $h = 1907\ \mathrm{m}$ — not the $1917\ \mathrm{m}$ summit elevation, since the Cambridge detector sits about ten metres above sea level. At the band midpoint $\beta = 0.9952$ : $t = h/(\beta c) = 1907/(0.9952 \times 2.998\times10^{8}) = 6.39\ \mu\mathrm{s}$ , the $\approx 6.4\ \mu\mathrm{s}$ the literature quotes. (Using $1917\ \mathrm{m}$ inflates this to $6.43\ \mu\mathrm{s}$ and corrupts every survival number downstream. The two heights are different physical quantities.)

Step 3 — apply the classical assumption. If the muon's clock ran at the laboratory rate it would age the full $6.39\ \mu\mathrm{s}$ , i.e. $2.91$ mean lifetimes: survival $e^{-2.91} = 0.0545$ , predicted rate $563 \times 0.0545 = \mathbf{31\ per\ hour}$ . (Frisch and Smith's own figure for the classical expectation is $\approx 27$ per hour; the difference comes from their use of the lifetime and velocity spectrum measured in situ rather than the round anchors used here. Nothing turns on it — the classical answer is "a few tens".)

Step 4 — apply relativity. The muon's own clock runs for $\tau = t/\gamma$ . At $\beta = 0.9952$ , $\gamma = 10.22$ , so $\tau = 6.39/10.22 = 0.625\ \mu\mathrm{s}$ — barely a quarter of one lifetime. Survival $e^{-0.625/2.197} = 0.752$ ; predicted rate $563 \times 0.752 = \mathbf{424\ per\ hour}$ .

Step 5 — the boundary number. The rate actually counted in Cambridge was $\mathbf{412 \pm 15}$ per hour — within one standard deviation of the relativistic estimate, and higher than the classical one by a factor of $412/27 = 15.3$ (or $412/31 = 13.4$ against the recomputed figure). No fitting, no free parameter. The muons even report the dilation factor themselves: inverting $412 = 563\,e^{-\tau/\tau_0}$ gives $\tau = -2.197\ln(412/563) = 0.686\ \mu\mathrm{s}$ and a factor $6.39/0.686 = 9.3$ , consistent with the $8.8 \pm 0.8$ Frisch and Smith quote. (Why $8.8$ and not the summit's $10.2$ ? The muons lose energy crossing $1907\ \mathrm{m}$ of air, so $\gamma$ falls from about $10.2$ to about $6.8$ en route; $8.8$ is the flight-average. The experiment is good enough to see the deceleration.)

Step 6 — recompute in the muon's frame. There nothing is dilated; instead the mountain is contracted, to $1907/10.22 = 187\ \mathrm{m}$ , and crossing $187\ \mathrm{m}$ at $0.9952\,c$ takes $0.625\ \mu\mathrm{s}$ — identical, to every printed digit, to $t/\gamma$ . Time dilation and length contraction are not two effects that agree; they are one invariant, $\tau$ , with two sets of bookkeeping.

2 · Could the twin have done better?

A twin leaves at $\beta = 0.6$ , turns around, and returns after ten years of Earth time. How much does she age, and could a cleverer route have bought her more?

Step 1 — arc length of each leg. In years and light-years, $c = 1$ . The outbound leg runs $(0,0) \to (ct,x) = (5,3)$ : interval $s^2 = 5^2 - 3^2 = +16$ , timelike, proper time $\sqrt{16} = 4$ years. The inbound leg mirrors it. Total $\tau_B = \mathbf{8\ years}$ against $\tau_A = 10$ ; cross-check $\tau_B = T/\gamma = 10/1.25 = 8$ . The traveller returns two years younger — a $20\%$ discount, bought at $0.6\,c$ .

Step 2 — the ceiling. She could not have done better, and that is a theorem, not an accident: the proper-time functional is stationary only on the straight worldline, and its second variation there is $\tfrac12\int\mathcal{L}''(0)(\delta v)^2 dt$ with $\mathcal{L}''(0) = -1/c^2 < 0$ , negative for every non-zero perturbation. Ten years is the ceiling; every route pays.

Step 3 — the boundary number. A wiggle at $\epsilon c$ gives $\tau = T(1 - \tfrac12\epsilon^2 - \cdots)$ , with no linear term. At $\epsilon = 10^{-6}$ — about $300\ \mathrm{m/s}$ , a passenger jet — the loss over ten years is $\tfrac12\epsilon^2\times 10\ \mathrm{yr} \approx 160\ \mu\mathrm{s}$ : small, non-zero, and precisely the size of effect Hafele and Keating flew around the world in 1971 to catch.

Physics of the portrait

Minkowski coined Eigenzeit in his 1908 Cologne lecture. But a geometric object in a mathematician's lecture is not yet a tool physicists use, and the man who did most to make it one wrote the book — literally, the first monograph on relativity ever published.

Origins · 2 minMax von Laue and how this was discovered

Max von Laue

1879–1960 · Munich · Zürich · Frankfurt · Berlin

Max Theodor Felix von Laue was born at Pfaffendorf, near Koblenz, on 9 October 1879, and took his doctorate under Planck in Berlin. In 1910–11, at Munich, he wrote Das Relativitätsprinzip (Braunschweig: Vieweg, 1911) — the first book-length treatment of special relativity, expanded into a second edition in 1913. Its importance lay not in new results but in a choice of language: Laue built the whole theory on Minkowski's four-dimensional geometry at a moment when most physicists still regarded that geometry as an elegant but optional repackaging of Einstein's kinematics. He handed the next generation a framework in which the interval, the worldline and its arc length were primitive and time dilation was a derived triviality — the reason a modern reader meets $\tau$ before meeting $\gamma$ . In 1912 Laue conceived, and Friedrich and Knipping executed, the experiment that diffracted X-rays through a crystal, proving in one stroke that X-rays are waves and that crystals are periodic lattices; for it he received the Nobel Prize in Physics in 1914. He spent the Nazi years in Germany in quiet, dangerous and conspicuous opposition, and was among the ten physicists interned at Farm Hall in 1945. He died in Berlin on 24 April 1960, of injuries from a road accident sixteen days earlier, aged eighty.

History

1908

— Minkowski introduces Eigenzeit (proper time) in his Cologne lecture "Raum und Zeit", defining it as the arc length of a worldline measured in units of time.

1911

— Max von Laue publishes Das Relativitätsprinzip (Vieweg), the first monograph on relativity, built throughout on Minkowski's four-dimensional geometry.

1941

— Rossi and Hall publish their 1940 Echo Lake and Denver mesotron runs: a mean proper lifetime near 2.4 μs, the first muon test of time dilation (Phys. Rev. 59, 223).

1963

— Frisch and Smith publish "Measurement of the Relativistic Time Dilation Using μ-Mesons" (Am. J. Phys. 31, 342): 563 muons per hour at the Mount Washington summit, 412 per hour 1907 m below in Cambridge, a measured dilation factor of 8.8 ± 0.8.

1971

— Hafele and Keating fly caesium clocks around the world in both directions and measure the accumulated proper-time differences, kinematic and gravitational together.

1977

— the CERN muon storage ring holds muons at γ ≈ 29.3, dilating the 2.197 μs proper lifetime to about 64 μs in the laboratory.

1980

s — GPS enters service with a relativistic timing budget: about −7.2 μs per day of special-relativistic deficit against about +45.8 μs per day of gravitational gain, netting roughly +38.6 μs per day.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§7, 12 · Taylor & Wheeler, Spacetime Physics, 2nd ed., ch. 1 and 5 · Morin, Introduction to Classical Mechanics, ch. 14 · Hartle, Gravity, ch. 4–5 · Misner, Thorne & Wheeler, Gravitation, §1.5 · Frisch and Smith, "Measurement of the Relativistic Time Dilation Using μ-Mesons", Am. J. Phys. 31, 342 (1963) · von Laue, Das Relativitätsprinzip (Vieweg, 1911).

Provenance. The base article for this slug is signature-mixed — it writes $c\,d\tau = \sqrt{-ds^2}$ , $L_0 = \sqrt{+\Delta s^2}$ , $u\cdot u = -c^2$ and boxes the arc length as $\tau = (1/c)\int\sqrt{-\eta_{\mu\nu}dx^\mu dx^\nu}$ , all of which belong to $(-,+,+,+)$ ; every one is re-derived here with the sign our convention requires. Anchors, all recomputed in verify_aq-sr-proper-time-proper-length.py: $c = 2.998\times10^8\ \mathrm{m/s}$ , $\tau_0 = 2.197\ \mu\mathrm{s}$ , Frisch–Smith $h = 1907\ \mathrm{m}$ (the detector-to-detector drop, not the $1917\ \mathrm{m}$ summit elevation), $563/\mathrm{hr}$ summit, $412 \pm 15/\mathrm{hr}$ Cambridge, band $0.9950$ – $0.9954\,c$ , measured dilation $8.8 \pm 0.8$ . Derived: $\gamma(0.9952) = 10.22$ , $t = 6.39\ \mu\mathrm{s}$ , $\tau = 0.625\ \mu\mathrm{s}$ , relativistic prediction $424/\mathrm{hr}$ , classical $31/\mathrm{hr}$ against the $\approx 27/\mathrm{hr}$ the literature quotes, dilation factor $9.3$ read straight off the counts.

Builds on. The Spacetime Interval · The Minkowski Metric · Proper Time · Proper Length · Length Contraction. Leads to. Four-Vectors · Lorentz Invariance and Invariants · The Twin Paradox · Relativistic Uniform Acceleration.

Why is the straight worldline the LONGEST, when in ordinary geometry it is the shortest?

One minus sign. Euclidean arc length is $\int\sqrt{dx^2+dy^2}$ : a $y$ -excursion adds to the integrand, so detours lengthen. Minkowski arc length along a timelike curve is $\int\sqrt{c^2dt^2 - dx^2}$ : an $x$ -excursion subtracts, so detours shorten. Same calculation, one flipped sign, opposite conclusion. For the theorem rather than the intuition: the second derivative of the integrand with respect to speed is $-\gamma^3/c^2$ , negative everywhere below $c$ , so the stationary point is a strict maximum.

Does acceleration make clocks run slow?

No — speed does. That is the clock postulate, built into $d\tau = dt\sqrt{1-v(t)^2/c^2}$ : only the instantaneous speed appears, never $dv/dt$ . Not a stipulation but a tested fact: muons in the CERN storage ring endure transverse proper accelerations around $10^{18}\ \mathrm{m/s^2}$ , and their decay rate tracks $\gamma$ — a function of speed alone — to a few parts in a thousand. Acceleration enters the twin problem only as the mechanism that bends the worldline; it is the bend, not the bending, that costs proper time.

If length contraction is "just slicing", is the moving rod really shorter or not?

The question has no frame-free answer, and that is the answer. "The rod, right now" requires an instant, and an instant is a slice that different observers tilt differently. The rod's own slice gives $L_0$ ; the lab's gives $L_0/\gamma$ . Both cut the same four-dimensional slab, both are honest proper distances of the event pairs actually used, and neither observer has erred. The error would be to posit a slice-independent fact about "the rod at an instant" that one of them is failing to measure. There is none — which is why the contracted rod fits the short barn and also, from its own point of view, does not.

Can I define proper time along a photon's path?

No. A null worldline has $ds^2 = 0$ at every point, so $d\tau = 0$ identically: zero arc length, however long the path looks on your diagram. This is the geometry telling you the light cone is a degenerate locus of the metric. The corollary is stronger than it sounds: there is no rest frame for light, so "what the universe looks like from a photon's point of view" is a question with no referent — not one with a wrong answer.

The Frisch–Smith numbers do not quite reproduce: I get 31 muons per hour classically, not 27.

You are right, and it is worth knowing why. The $\approx 27/\mathrm{hr}$ figure is Frisch and Smith's own, computed with the lifetime and velocity spectrum they measured in situ. Recomputing from the round anchors used here — $563/\mathrm{hr}$ , $6.39\ \mu\mathrm{s}$ , $\tau_0 = 2.197\ \mu\mathrm{s}$ — gives $563\,e^{-2.91} = 31/\mathrm{hr}$ , about $13\%$ higher: a rounding-and-method difference, not a physics difference, and the secondary literature is not always careful about which is which. The conclusion survives a factor of $1.13$ untouched. When a number is quoted rather than derived, say so — then derive it anyway.

Is proper time "what really happens" while coordinate time is a fiction?

That overstates a true thing. Coordinate time is no fiction — a network of synchronised clocks at rest in an inertial frame reads it faithfully. What it is not is intrinsic: it belongs to a frame, and changing frames changes it, whereas proper time belongs to a worldline and changes with nothing. So $\tau$ is what a particular clock reads and $t$ what a particular frame's clock network reads, and physics — decay rates, phases, aging — is driven by the former.