Proper Length

A metre stick on your desk has a length, and measuring it is the least interesting thing you will do today: see where one end falls on a ruler, see where the other falls, subtract. Nothing hangs on when you look, because nothing is moving. That laziness — the freedom to take your time — is the whole content of the idea called proper length. Now set the stick moving past you at $0.6c$ . You cannot look at one end, stroll over, and look at the other, because by the time you arrive the first end has left: you must catch both ends at the same instant. And "the same instant" is the one phrase in physics that relativity has taught us to distrust. Two observers who disagree about simultaneity mark the ends at different pairs of moments, cut the object's history along different slices, and come away with different lengths — each of them correct. No measurement of a moving object's length avoids first making a choice about time. Proper length is what you get when that choice is forced. In the object's own rest frame the ends go nowhere, simultaneity becomes irrelevant, and one number survives: $L_0$ . Every other inertial observer measures something shorter, by exactly $\gamma$ . At the LHC, a proton circling a $26.659\ \mathrm{km}$ ring at $\gamma = 7460$ finds that ring to be $3.57\ \mathrm{m}$ long — shorter than a family car. The ring did not shrink; the proton sliced its history along a different plane. This article is about what that means, why the rest-frame answer is the biggest anyone can get, and why the shrinking is not something a camera would ever photograph.

Key takeaways

The proper length $L_0$ is an object's length in the inertial frame where it is at rest. There the ends can be located at leisure, so the answer depends on no convention about simultaneity.
Geometrically, $L_0$ is the invariant interval read on its spacelike side: $L_0 = \sqrt{-s^2}$ , taken between two end-events that are simultaneous in the rest frame, with $s^2 = c^2\Delta t^2 - \Delta x^2$ .
Any observer seeing the object move at $v$ along its length measures $L = L_0/\gamma = L_0\sqrt{1-\beta^2} \le L_0$ . Transverse dimensions are untouched.
$L_0$ is not merely "the rest-frame value" — it is a genuine maximum: over every pair of events with one end on each worldline, $\sqrt{-s^2}$ is largest exactly when the pair is rest-frame simultaneous. Proper length is to spacelike intervals what proper time is to timelike ones.
Key idea: length belongs not to an object but to an object plus a slice of time. Fix the slice by marking the ends simultaneously in the object's own frame, and you get the one length everybody agrees to call the object's own — and the longest anyone will ever measure.

The core insight

To measure the length of a moving thing you must locate its two ends at one instant, and different frames disagree about which instant that is — so each frame measures the interval between a different pair of end-events. All of them agree on the invariant $\sqrt{-s^2}$ of the pair they chose, and among all such pairs the rest-frame-simultaneous one gives the largest value. That maximum is the proper length $L_0$ , and every other inertial measurement returns $L_0/\gamma$ .

Where this lives today

The number that matters most in a working accelerator is a proper length. A bunch of protons at the LHC is about $8\ \mathrm{cm}$ long in the lab, but the physics deciding whether it holds together — the mutual repulsion of some $10^{11}$ protons — is computed in the bunch's own rest frame, where it is $8\ \mathrm{cm} \times \gamma \approx 600\ \mathrm{m}$ long. From its own point of view the bunch is a dilute cigar; from the lab's, a dense pancake, and getting the direction of that transformation backwards is a well-known way to lose a beam. In heavy-ion physics the contraction is the initial condition outright: a lead nucleus, a sphere $14\ \mathrm{fm}$ across at rest, reaches the LHC collision point at $\gamma \approx 2700$ as a disc some $5\ \mathrm{am}$ thick, and how flat that pancake is sets how fast the energy is dumped into the quark–gluon plasma. And the Terrell–Penrose caveat has finally been imaged. In 2025 a group at TU Wien and the University of Vienna, in Communications Physics, lit moving objects with picosecond laser pulses and gated a camera to about $300\ \mathrm{ps}$ — in effect slowing the effective speed of light to under $2\ \mathrm{m/s}$ . Keeping only the light that would have reached the lens at one instant under that effective $c$ , they recovered the prediction exactly: a cube that appears twisted, and a sphere whose outline stubbornly stays a circle while its markings slide toward the far side. Sixty-six years late, and after a century of textbook pictures showing squashed spaceships that nobody would ever have seen.

The physical picture

Draw the history of a rod, not a picture of it. An object of proper length $L_0$ is not a segment; it is a strip in spacetime, a band swept out between the worldlines of its two ends, running forever upward through time. The rod itself is nowhere in that picture — what is in it is everything the rod ever does. To "measure a length" is to cut across that strip: take a slice of constant time, see where it enters and where it leaves, and measure the distance between those two events. That is all a length measurement is — a chord across a world-strip. And here is the fact that generates every so-called paradox in the subject: different inertial observers cut the strip at different angles. They are not getting different numbers for the same two events, which would be a contradiction; they are getting different numbers for different pairs of events, which is merely geometry. In the rod's own frame the strip is upright, the natural cut is horizontal, and the chord is $L_0$ at every height because nothing is happening. That is what makes the rest frame special: it is the one frame whose measurement contains no hidden decision.

Figure

The world-strip of a rod of proper length L₀ = 100 m, drawn in its own rest frame with equal scales, so light lines run at 45°. The rod frame cuts horizontally and reads 100 m. A lab frame moving at β = 0.6 cuts along a line of ITS simultaneity, tilted by slope −0.6, and catches the ends at events separated by Δx′ = 100 m and cΔt′ = −60 m. The invariant length of THAT chord is √(100² − 60²) = 80 m — exactly L₀/γ. Two frames, two chords, one strip

Look at the red chord. In the lab frame — where the rod flies past at $\beta = 0.6$ — the two ends are marked at the same lab time. Seen from the rod's frame those markings are not simultaneous at all: they are separated by $c\Delta t' = -60\ \mathrm{m}$ , meaning the lab marked the far end $200\ \mathrm{ns}$ earlier than the near one. It caught the rod mid-slide. Its chord runs diagonally, and that chord's invariant length is $\sqrt{100^2 - 60^2} = 80\ \mathrm{m}$ — precisely $L_0/\gamma$ at $\gamma = 1.25$ . The contraction is not a squeezing. It is a tilt. And the extremal fact can be read straight off the picture, before any algebra. Every chord from the left worldline to the right spans $\Delta x' = L_0$ horizontally, and whatever vertical run $c\Delta t'$ it also has can only subtract, since $\sqrt{-s^2} = \sqrt{L_0^2 - c^2\Delta t'^2}$ . The metric's minus sign does the work: tilt the cut, lose length. The horizontal cut gives the maximum, and the maximum is $L_0$ .

Figure

PUNCHLINE — every inertial observer measures L = L₀√(1 − β²), and nobody ever measures more than L₀. A 747 at 250 m/s falls short of L₀ by 3 parts in 10¹³, invisible. At β = 0.866 the object is halved. At the LHC design energy the 26.659 km ring measures 3.57 m in the proton's frame — the curve does not merely bend, it collapses

Two features of that curve matter more than the curve. It leaves $\beta = 0$ flat — $L/L_0 \approx 1 - \tfrac12\beta^2$ , second order, and at any speed a human has travelled, nothing at all. And it arrives at $\beta = 1$ vertically: everything is crammed into the last sliver, which is why particle physics is the only place the effect is dramatic. Between $\beta = 0.99$ and the LHC's $\beta = 0.999999991$ the length falls by a further factor of a thousand.

The core equations

State the convention once, loudly, because half the textbooks on your shelf use the other one and every sign below would flip.

Signature $(+,-,-,-)$ : timelike $s^2 > 0$ , spacelike $s^2 < 0$ . For two events separated by $(\Delta t, \Delta x, \Delta y, \Delta z)$ in some inertial frame,

s^{2} \;=\; c^{2}\Delta t^{2} - \Delta x^{2} - \Delta y^{2} - \Delta z^{2} \qquad (1)

where: $s^2$ — the invariant spacetime interval, the same number in every inertial frame (proved in The Invariant Spacetime Interval; here we use it); $\Delta t, \Delta x, \Delta y, \Delta z$ — coordinate separations. Physical reading: $s^2 > 0$ means a massive particle could be at both events (timelike); $s^2 = 0$ means only light could (lightlike); $s^2 < 0$ means nothing could — the events are spacelike, outside each other's causal reach. A rod's two ends, marked at one instant, are always spacelike separated: the textbook case of two things that cannot signal to each other. Their $s^2$ is therefore negative, and its magnitude has an immediate meaning.

L_0 \;=\; \sqrt{-s^{2}} \qquad\text{(ends marked simultaneously in the rest frame)} \qquad (2)

where: $L_0$ — the proper length, with the interval taken between two end-events having $\Delta t' = 0$ in the rest frame $S'$ . Physical reading: this is the definition that makes $L_0$ invariant rather than merely rest-frame. Every observer, using their own $\Delta t$ and $\Delta x$ for that same pair, computes the same $s^2$ and so the same $L_0$ . It is the exact spacelike mirror of $c\Delta\tau = \sqrt{s^2}$ — proper time and proper length are one invariant, read on the two sides of the light cone. Now the number everybody quotes. An observer who sees the object move at $v$ along its length, and marks both ends simultaneously in his own frame, gets

L \;=\; \frac{L_0}{\gamma} \;=\; L_0\sqrt{1-\beta^{2}}\,, \qquad \gamma \equiv \frac{1}{\sqrt{1-\beta^{2}}},\quad \beta \equiv \frac{v}{c} \qquad (3)

where: $L$ — the length measured in the moving frame; $\gamma \ge 1$ always. Physical reading: $L \le L_0$ , with equality only at $v = 0$ . The contraction is reciprocal — each observer finds the other's rulers short by the same $\gamma$ — and that is no contradiction, because they cut different chords. The derivation of $L = L_0/\gamma$ belongs to Length Contraction; what this article owns is why $L_0$ is the thing to compare against, and that it is a maximum. Only the dimension along the motion is affected: a cube flying edge-on becomes a slab, not a smaller cube.

L_{\parallel} = \frac{L_{0\parallel}}{\gamma}, \qquad L_{\perp} = L_{0\perp} \qquad (4)

where: $\parallel$ and $\perp$ — components along and across $\vec v$ . Physical reading: the boost mixes $t$ with $x$ only, and $y, z$ ride through untouched. This is not a convention but a necessity: transverse contraction would let two identical rings, boosted past each other, each pass inside the other — and "which ring went inside" is a single fact that cannot come out both ways. Finally, the operational definition, which is what you would actually build. In the rest frame, bounce light off the far end and time the round trip on a single clock.

L_0 = \frac{c\,\Delta\tau_{\text{round}}}{2} \qquad (5)

where: $\Delta\tau_{\text{round}}$ — the proper time, on one clock, for light to reach the far end and return. Physical reading: no second clock, so no synchronization convention, so no simultaneity assumption smuggled in — which is why $L_0$ is the length engineering and astronomy quote. Mind the factor of $2$ : the pulse crosses the object twice, and dropping it (as the source material does, writing $L_0 = c\Delta\tau$ ) silently doubles the rod. Run the same experiment in a frame where the rod moves and the two legs become grossly unequal, $t_{\text{round}} = \frac{L}{c-v} + \frac{L}{c+v} = \frac{2L\gamma^2}{c}$ ; substituting $L = L_0/\gamma$ and dividing by $\gamma$ for the travelling clock's own reading returns $2L_0/c$ again, as it must.

Where proper length comes from

Derivation AProper length is the maximum of the interval over all pairs of end-events

The claim is stronger than "the rest-frame length is the biggest one measured", and it is the result this article owns. Consider the two worldlines swept out by the ends of a rigid rod of proper length $L_0$ , and pick any event $A$ on the left one and any event $B$ on the right — not necessarily simultaneous in anybody's frame, with no measurement procedure imposed at all. We show that whenever the pair is spacelike its invariant proper distance $\sqrt{-s^2}$ can never exceed $L_0$ , and equals $L_0$ exactly when the two events are simultaneous in the rod's own frame. Proper length is therefore an extremum over the whole strip, not merely a value in a privileged frame — precisely as proper time is a maximum along a timelike worldline, with the metric's minus sign turning "maximum of duration" into "maximum of extent".

Step 1. Work in the rod's rest frame $S'$ : the left end sits at $x' = 0$ for all $t'$ , the right end at $x' = L_0$ . Take $A = (t'_A,\, 0)$ , $B = (t'_B,\, L_0)$ , and let $T \equiv t'_B - t'_A$ be free. Why: an invariant may be evaluated in whichever frame makes it easy, and the answer holds in all of them — so this one calculation is completely general.

Step 2. Compute the interval: with $\Delta t' = T$ , $\Delta x' = L_0$ , equation (1) gives $s^2 = c^2T^2 - L_0^2$ . Why: this single expression contains every possible pair of end-events, indexed by the one free parameter $T$ , with nothing assumed about who measures or how.

Step 3. The pair is spacelike — the condition for it to be a length at all — when $s^2 < 0$ , i.e. $|T| < L_0/c$ . For larger $|T|$ the events are timelike separated and no frame calls them simultaneous. Why: a length is by definition a spacelike separation. End-events too far apart in time are a causal relationship, not a ruler reading.

Step 4. On those pairs, $\sqrt{-s^2} = \sqrt{L_0^2 - c^2T^2}$ . It decreases in $|T|$ , peaks at $T = 0$ with value $L_0$ , and falls to zero as $|T| \to L_0/c$ . Why: here is the metric's minus sign doing all the work. In Euclidean geometry an extra component lengthens a diagonal; in Minkowski geometry a timelike component shortens a spacelike chord. Tilt the cut and you lose length.

Step 5. Read off the physics: $T = 0$ is the statement $\Delta t' = 0$ , the pair simultaneous in the rod's own frame. So the maximiser is the rest-frame pair, and the maximum is $L_0$ . Any other observer, marking the ends simultaneously in his frame, necessarily picks a pair with $T \neq 0$ — his slice is tilted relative to $S'$ — and so necessarily gets less. Why: this is the whole of length contraction, obtained without transforming a single coordinate. The moving observer measures less not because his ruler is faulty but because the maximum was already taken.

Step 6. Recover the factor. "Simultaneous in $S$ " transforms into $c\,T = -\beta L_0$ (shown explicitly in the next card), so $\sqrt{-s^2} = \sqrt{L_0^2 - \beta^2L_0^2} = L_0/\gamma$ . Why: the extremal picture does not merely permit $L = L_0/\gamma$ — it produces it. The factor $\sqrt{1-\beta^2}$ is precisely what a tilt of $\beta$ costs you. *Verified in verify_aq-sr-proper-length.py: symbolically $\partial_T\sqrt{L_0^2 - c^2T^2}$ vanishes only at $T=0$ , the second derivative there is $-c^2/L_0 < 0$ , $\sqrt{-s^2} \le L_0$ across the admissible range, and substituting $cT = -\beta L_0$ returns $L_0/\gamma$ identically in $\beta$ .*

Go deeper · Derivation B · 2 minLength contraction from the Lorentz transformation, and the wrong-slice trap that inverts it

Derivation BLength contraction from the Lorentz transformation, and the wrong-slice trap that inverts it

The extremal argument is the deeper one, but the pedestrian bookkeeping is worth doing too — it is where marks are actually lost, and its failure mode is spectacular. The goal is twofold. First, derive $L = L_0/\gamma$ by transforming the two ends, scrupulous about the one condition that makes the calculation a length measurement at all: $\Delta t = 0$ in the measuring frame. Second, violate that condition in the most natural-looking way available and watch the answer come out as $\gamma L_0$ — a rod longer than its proper length by $56\%$ . That is not a small slip; it is a simultaneity error wearing a plausible face.

Step 1. The rod is at rest in $S'$ , from $x' = 0$ to $x' = L_0$ ; $S'$ moves at $+v$ along the lab's $x$ -axis, so in the lab the rod flies past at $\beta$ . Take the boost as given from The Lorentz Transformation Equations: $ct' = \gamma(ct - \beta x)$ , $x' = \gamma(x - \beta ct)$ , inverse $ct = \gamma(ct' + \beta x')$ , $x = \gamma(x' + \beta ct')$ . Why: this article applies the boost, it does not re-derive it — that derivation belongs to the article named.

Step 2. Impose the measurement condition: locate both ends at a single lab time $t = 0$ , giving events $A = (ct=0,\,x=0)$ and $B = (ct=0,\,x=L)$ with $L$ unknown. Why: this is the entire physical content. Drop it and your "length" is the gap between an event on one end and an unrelated event on the other.

Step 3. Push both into the rod's frame: $A \mapsto (0,\,0)$ , and $B \mapsto x'_B = \gamma L$ , $ct'_B = -\gamma\beta L$ . Why: look at the times. The lab marked both ends "at once", yet in the rod's frame the far mark sits at $ct' = -\gamma\beta L$ — it happened earlier. The lab caught the far end in the past.

Step 4. The right end has $x' = L_0$ for all $t'$ , so the mark on it must have $x' = L_0$ whenever it was made. Hence $\gamma L = L_0$ , i.e. $L = L_0/\gamma$ . Why: "rigid rod at rest in $S'$ " means exactly that its ends have fixed $x'$ — the only property of the rod we ever use.

Step 5. Numbers at $\beta = 0.6$ , $L_0 = 100\ \mathrm{m}$ , $\gamma = 1.25$ : $L = 80\ \mathrm{m}$ , and the far mark occurred at $ct' = -60\ \mathrm{m}$ , i.e. $200\ \mathrm{ns}$ early. Cross-check with the invariant: $\sqrt{-s^2} = \sqrt{100^2 - 60^2} = 80\ \mathrm{m}$ ✓. Why: the two routes must agree because the boost preserves $s^2$ — and they do, exactly.

Step 6. Now the trap. Take instead the pair simultaneous in the rod's frame, $A = (0,0)$ and $C = (ct'=0,\,x'=L_0)$ , and compute their separation in the lab: $x_C = \gamma L_0 = 125\ \mathrm{m}$ , $ct_C = \gamma\beta L_0 = 75\ \mathrm{m}$ . You report $125\ \mathrm{m}$ — a moving rod longer than its rest length, and longer than the true $80\ \mathrm{m}$ by exactly $\gamma^2 = 1.5625$ . Why: the blunder is not arithmetic. $C$ happened $250\ \mathrm{ns}$ after $A$ in the lab, and in that time the right end moved on: you subtracted the back end's position at one moment from the front end's position at another.

Step 7. The rule: $\Delta t = 0$ holds in the frame doing the measuring, never the other one. Get it backwards and you are out by $\gamma^2$ , in the direction that makes the rod look dilated. Why: the sign of the error is what makes it dangerous. Nobody who computes $125\ \mathrm{m}$ sees an absurdity — they see "length dilation" and invent a story for it. *Verified in verify_aq-sr-proper-length.py: $(x'_B, ct'_B) = (\gamma L, -\gamma\beta L)$ and $(x_C, ct_C) = (\gamma L_0, \gamma\beta L_0)$ checked symbolically; $L = 80$ m, the wrong slice gives $125$ m, the ratio is $\gamma^2 = 1.5625$ , $c\Delta t_C = 75$ m, and $125 - v\Delta t_C = 125 - 45 = 80$ m.*

Limiting cases

Case	Result	Meaning
Rest frame, $v = 0$	$L = L_0$	$\gamma = 1$ ; no choice of slice, and the maximum is attained
Low speed, $v \ll c$	$L \approx L_0\left(1 - \tfrac12\beta^2\right)$	second order; a 747 loses $3.5\times10^{-13}$ of its length
$\beta = \sqrt{3}/2 = 0.866$	$L = L_0/2$	$\gamma = 2$ exactly — the speed at which everything halves
Ultrarelativistic, $\beta \to 1$	$L \to 0$	at the LHC's $\gamma = 7460$ , a $26.659\ \mathrm{km}$ ring is $3.57\ \mathrm{m}$ long
Transverse direction	$L_\perp = L_{0\perp}$	unaffected at every speed; a cube becomes a slab, never a smaller cube
Tilted slice, $c\Delta t' = cT$	$\sqrt{-s^2} = \sqrt{L_0^2 - c^2T^2}$	the general chord; every non-rest-frame measurement is a special case, and every one of them loses
Galilean limit, $c \to \infty$	$\gamma \to 1$ , $L \to L_0$	length becomes absolute again — at the same moment, and for the same reason, that simultaneity does

Common trapThe simultaneity condition belongs to the frame doing the measuring — get it backwards and the rod comes out LONGER by γ²

This is the mistake, and it is worth naming because it does not look like one. You want the length of a rod flying past at $\beta = 0.6$ . Its ends are at $x' = 0$ and $x' = L_0 = 100\ \mathrm{m}$ in its own frame, so you transform those two points into your frame and subtract: $x = \gamma(x' + \beta ct')$ at $t' = 0$ gives $\Delta x = \gamma L_0 = 125\ \mathrm{m}$ . Longer than the proper length. You have discovered "length dilation", and it is pure artefact. Those two events were simultaneous in the rod's frame, not in yours. In your frame they are $250\ \mathrm{ns}$ apart ( $c\Delta t = 75\ \mathrm{m}$ ), and in that time the rod's front end travelled $45\ \mathrm{m}$ further down the lab. You subtracted where the front end was later from where the back end was earlier. Impose $\Delta t = 0$ in the frame doing the measuring — always. The right answer is $80\ \mathrm{m}$ , and wrong-over-right is exactly $\gamma^2 = 1.5625$ . If your length comes out bigger than $L_0$ , you sliced in the wrong frame.

Common trapLength contraction is not what a camera photographs — Terrell and Penrose, 1959

For fifty years after 1905, physicists wrote that a fast-moving sphere would look like an ellipsoid. In 1959 James Terrell (Phys. Rev. 116, 1041) and, independently, Roger Penrose (Proc. Camb. Phil. Soc. 55, 137) showed this is false, and that the error had been hiding in plain sight. A photograph is not a measurement. A measurement collects the positions of all parts of the object at one instant of the observer's time; a photograph collects the light arriving at the lens at one instant, and that light left different parts of the object at different times — earlier from the far side, later from the near, because the far side is further away. Restore the light-travel delays and the two effects very nearly cancel. A photographed sphere stays a perfect circle in outline at any speed; an extended object appears not flattened but rotated, as though you were seeing round its back. This is the Terrell rotation, and it was finally imaged in the laboratory in 2025. Do not over-correct. Terrell and Penrose did not show that length contraction is unreal, and nothing here is retracted by them. The contraction $L = L_0/\gamma$ is exactly what you get by marking both ends simultaneously in your frame — which is what "length" means, and what every collider's beam optics assumes. What they showed is that the visual appearance of a moving object is a different question with a different answer.

Worked examples

1 · How long is the LHC to a proton?

The Large Hadron Collider's ring has a circumference of $26\,659\ \mathrm{m}$ as surveyed by CERN's engineers, who are at rest with respect to it. At design energy each proton carries $E = 7\ \mathrm{TeV}$ . How long is the ring in the proton's own rest frame, and how many turns does the proton think it is making per second?

Step 1 — identify. The $26\,659\ \mathrm{m}$ is a proper length: it is measured in the frame where the ring sits still. The proton is the moving observer, so it measures $L = L_0/\gamma$ .

Step 2 — get $\gamma$ from the energy, not the speed. With $m_pc^2 = 0.938272\ \mathrm{GeV}$ and $E = \gamma m_pc^2$ ,

\gamma = \frac{E}{m_pc^2} = \frac{7000\ \mathrm{GeV}}{0.938272\ \mathrm{GeV}} = 7460.5.

This is the honest route. Extracting $\gamma$ from a quoted $\beta$ would need nine significant figures before the answer even began to move.

Step 3 — the boundary number. $L = 26\,659 / 7460.5 = \mathbf{3.573\ m}$ . The proton laps a ring about as long as a small car — not a figure of speech, but the distance the proton's own rulers report between the start of a lap and its end. (For the record $1 - \beta \approx 1/(2\gamma^2) = 8.98\times10^{-9}$ , so $\beta = 0.999999991$ : every digit that matters lies below the ninth decimal place, which is why $\gamma$ is the variable of choice this close to $c$ .)

Step 4 — the lap rate, both ways. In the lab, $t = L_0/(\beta c) = 88.92\ \mathrm{\mu s}$ per turn, i.e. $11\,245$ turns per second — the LHC's revolution frequency. The proton's own clock runs slow by $\gamma$ , so between two passes of the same magnet it ages $\Delta\tau = 88.92\ \mathrm{\mu s}/7460.5 = 11.92\ \mathrm{ns}$ . Equivalently it crosses its own contracted $3.573\ \mathrm{m}$ ring at essentially $c$ : $3.573/c = 11.92\ \mathrm{ns}$ . Contracted ring or dilated clock — identical number. That agreement is the whole point of the subject.

2 · The rod, the two slices, and the 45-metre error

A rod of proper length $L_0 = 100\ \mathrm{m}$ moves at $\beta = 0.6$ . (a) What is its lab length? (b) A student instead transforms the rod's own end-points at $t' = 0$ into the lab and subtracts. What does he get, and where did the extra length come from?

Step 1 — the correct measurement. $\gamma = 1/\sqrt{1-0.36} = 1.25$ , so $L = 100/1.25 = \mathbf{80\ m}$ .

Step 2 — locate the events the lab used. The lab marks $x = 0$ and $x = 80\ \mathrm{m}$ at $t = 0$ . Transform the far mark: $x' = 1.25(80 - 0) = 100\ \mathrm{m}$ ✓ (on the right end, as it must be) and $ct' = 1.25(0 - 0.6 \times 80) = -60\ \mathrm{m}$ , i.e. $-200\ \mathrm{ns}$ . The lab's "simultaneous" pair is, to the rod, $200\ \mathrm{ns}$ out of step.

Step 3 — check with the invariant. $s^2 = 60^2 - 100^2 = -6400\ \mathrm{m^2}$ , so $\sqrt{-s^2} = 80\ \mathrm{m}$ ✓ — the lab's length, recovered without ever mentioning the lab.

Step 4 — the student's error. He transforms $(ct'=0,\,x'=0)$ and $(ct'=0,\,x'=100)$ : $x = 1.25 \times 100 = 125\ \mathrm{m}$ and $ct = 1.25 \times 0.6 \times 100 = 75\ \mathrm{m}$ . He reports $\mathbf{125\ m}$ — a moving rod longer than its rest length.

Step 5 — find the missing $45\ \mathrm{m}$ . His events are $\Delta t = 75/c = 250\ \mathrm{ns}$ apart in the lab, and in $250\ \mathrm{ns}$ the rod's right end travels $0.6c \times 250\ \mathrm{ns} = 45\ \mathrm{m}$ . Indeed $125 - 45 = 80\ \mathrm{m}$ ✓. Every metre of his excess is rod-motion he timed into the answer, and $125/80 = 1.5625 = \gamma^2$ , exactly as the extremal theorem predicts. He did not slice; he sheared.

Physics of the portrait

Proper length has no discoverer, which is itself instructive. FitzGerald and Lorentz supplied the contraction in 1889 and 1892 as a physical shrinking of matter moving through an ether — a real squeeze with a real cause that happened to be undetectable. Einstein in 1905 kept the formula and deleted the mechanism. Minkowski, in 1908, supplied the object that had been missing all along: the invariant interval, of which proper length is simply the spacelike face.

Origins · 2 minProper length and how this was discovered

Proper length

1889–1959 · FitzGerald's rescue · Einstein's kinematics · Minkowski's interval · Terrell and Penrose's photograph

The idea arrived three times, and each time it meant something different. In 1889 George Francis FitzGerald, writing to Science, and independently Hendrik Lorentz in 1892, proposed that a body moving through the ether is physically shortened along its motion by $\sqrt{1-v^2/c^2}$ — an ad hoc rescue of the ether from the Michelson–Morley null, in which an object had a true length and the moving one was really smaller. In 1905 Einstein derived the same factor from the two postulates alone and changed its meaning entirely: no ether, no squeeze, no cause. "The length of a moving rod", he showed, is not defined at all until an observer commits to a simultaneity convention — and different observers necessarily commit differently. The rest-frame length, needing no such commitment, is the proper length: the only length belonging to the object rather than the observer. In 1908 Minkowski recast it geometrically as $L_0 = \sqrt{-s^2}$ , the invariant interval between two rest-frame-simultaneous end-events, exact mirror of the proper time $c\Delta\tau = \sqrt{s^2}$ . The last twist came in 1959, when Terrell and Penrose showed that this entirely real contraction is nonetheless invisible: light from an object's far side leaves earlier than from its near side, so a photograph of a fast sphere is still a circle.

History

1889

— FitzGerald proposes in Science that bodies moving through the ether contract along the direction of motion, which would null the Michelson–Morley experiment exactly.

1892

— Lorentz reaches the same contraction factor independently, and by 1895 embeds it in his theory of electrons as a real deformation of matter.

1905

— Einstein derives

L = L_0\sqrt{1-v^2/c^2}

from the two postulates in §4 of "Zur Elektrodynamik bewegter Körper", with no ether and no mechanism: the contraction becomes kinematics, not dynamics.

1908

— Minkowski introduces the invariant interval in "Raum und Zeit", making proper length the spacelike counterpart of proper time and the contraction a matter of how a world-strip is sliced.

1959

— Terrell (Phys. Rev. 116, 1041) and Penrose (Proc. Camb. Phil. Soc. 55, 137) independently show that light-travel delays make the contraction invisible to a camera: a moving sphere photographs as a sphere, and an extended body appears rotated.

1959

— Dewan and Beran (Am. J. Phys. 27, 517) pose the thread-between-two-rockets problem later popularised by John Bell, showing that holding the proper length fixed and holding the lab-frame separation fixed are different instructions with different outcomes.

Going further

Texts. Taylor & Wheeler, Spacetime Physics, 2nd ed., Ch. 3 and 5 · Rindler, Introduction to Special Relativity, 2nd ed., §§2–3 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.1.2. On the visual question: J. Terrell, Phys. Rev. 116, 1041 (1959); R. Penrose, Proc. Camb. Phil. Soc. 55, 137 (1959).

Provenance. $L = L_0/\gamma$ and $L_0 = \sqrt{-s^2}$ are standard (Rindler §3, Griffiths §12.1.2, Taylor & Wheeler Ch. 5). The extremal characterisation — $\sqrt{-s^2}$ over all end-event pairs is maximised at $\Delta t' = 0$ , with maximum $L_0$ — is derived in full above and machine-checked. Anchors, all recomputed in verify_aq-sr-proper-length.py: $\gamma(0.6) = 1.25$ , $L = 80\ \mathrm{m}$ , wrong-slice $125\ \mathrm{m}$ with ratio $\gamma^2 = 1.5625$ and the $45\ \mathrm{m}$ of rod-motion accounting for it; LHC $\gamma = 7000/0.938272 = 7460.5$ , $\beta = 0.999999991$ , $26\,659\ \mathrm{m} \Rightarrow 3.573\ \mathrm{m}$ , revolution $88.92\ \mathrm{\mu s}$ , proper lap $11.92\ \mathrm{ns}$ ; a 747 at $250\ \mathrm{m/s}$ giving $3.5\times10^{-13}$ ; and $t_{\text{round}} = 2L\gamma^2/c \Rightarrow \Delta\tau_{\text{round}} = 2L_0/c$ . Two corrections to the source material are recorded there: the radar length is $c\Delta\tau_{\text{round}}/2$ , not $c\Delta\tau$ , and the LHC Lorentz factor at $7\ \mathrm{TeV}$ is $7460$ , not $7000$ .

Builds on. The Two Postulates of Special Relativity · Relativity of Simultaneity · The Lorentz Factor Gamma · The Lorentz Transformation Equations. Leads to. Length Contraction · The Invariant Spacetime Interval · Timelike, Spacelike, and Lightlike Separations · Proper Time.

If the rod really is 80 m long in the lab, and really is 100 m in its own frame, which is its real length?

Both, and the question is malformed in the way "what is the real shadow of a stick?" is malformed. A shadow depends on where you put the lamp; a length depends on which slice of time you cut. What is real is the world-strip — the rod's full history — and both numbers are honest chords across it. Relativity adds only that one chord is distinguished: the one horizontal in the rod's own frame, because only it requires no arbitrary choice. That is the proper length. It is not more real than $80\ \mathrm{m}$ ; it is more canonical.

If I see your ruler as contracted and you see mine as contracted, doesn't somebody have to be wrong?

No, because we are not measuring the same pair of events. When I mark both ends of your ruler "at once", those marks are — by your reckoning — made at different times, so from your point of view I cheated. When you do the same to mine, you cheat in the mirror-image way. No experiment catches either of us out, because none compares the two rulers at a single pair of events. Reciprocity is not a paradox but a theorem, proved in Inverse Lorentz Transformation and Reciprocity, and its resolution is entirely the relativity of simultaneity. To compare two objects unambiguously, bring them to rest with respect to each other — and then they agree.

What forbids a transverse contraction?

A short and completely rigorous argument. Take two identical rings on rails, moving toward each other along their common axis. If motion contracted transverse dimensions, then in ring A's frame ring B is smaller and passes inside it — while in ring B's frame ring A is smaller and passes inside it. But "which ring went inside which" is a single physical fact; paint them and look afterwards. It cannot come out both ways, so transverse dimensions must be unchanged and the rings graze. The argument does not work along the motion, because "which rod is longer" is not a single fact — it depends on a simultaneity slice, and the two frames slice differently.

Does an accelerating rod have a proper length?

Only with care, and this is where the concept bites. Proper length is defined by a slice in the instantaneous rest frame, and while the rod accelerates that frame keeps changing, so the slice keeps rotating. Holding the proper length fixed under acceleration — Born rigidity — requires the two ends to accelerate by different amounts, tail harder than nose. Give both ends the same lab-frame acceleration instead and the proper length grows: a thread joining them stretches and snaps. That is the Dewan–Beran (1959) / Bell (1976) spaceship problem. The lesson: "the rod's length" always presupposes a frame and a slice, and while accelerating you must say which.

Why quote L₀ rather than whatever length we actually measured?

Because $L_0$ is the only one of those numbers that is a property of the object. Every measured $L$ is a property of the object and the observer, and there is a continuum of them — one per velocity, down to zero. Two engineers on different spacecraft discussing the same girder need a common referent, and the natural one is the value in the girder's own frame, reproducible by anybody with one clock and a mirror. Quoting $L_0$ plus the observer's $\gamma$ is complete information; quoting a contracted $L$ alone is not.

The interval s² between the rod's ends is negative. What does a negative squared length mean?

It means the events are spacelike separated: no signal, not even light, gets from one to the other, so nothing at one end can influence the other end at that moment. The sign of $s^2$ is the causal classification (see Timelike, Spacelike, and Lightlike Separations), and it is negative here precisely because a rod's two ends, taken at one instant, are causally disconnected. The minus sign is not an embarrassment to be square-rooted away: it is why we write $\sqrt{-s^2}$ , and it is the same minus sign that makes $\sqrt{-s^2}$ a maximum on the rest-frame slice, where Euclidean intuition predicts a minimum.