The length that depends on when, and is not really about length

Measuring the length of something stationary requires no care. Put a ruler alongside it, read both ends, subtract. The order in which the two ends are read does not matter, because nothing is going anywhere.

Measuring the length of something moving requires reading both ends at the same moment. Read the front now and the back a second later and the answer will be wrong by however far the object travelled in between. That requirement is trivial in ordinary life and it is the entire content of this essay, because what “at the same moment” means depends on who is asking.

An object is a region, not a line #

The right way to think about a moving object is as a worldsheet: the whole two-dimensional region of spacetime that its history sweeps out, bounded by the worldline of its front end and the worldline of its back.

The object has no length in that picture. The worldsheet is a ribbon; asking how wide it is requires choosing a direction to measure across, and the choice of direction is exactly the choice of a simultaneity slice.

Slice the ribbon along the horizontal — the simultaneity of an observer at rest with respect to the object — and the width obtained is the proper length, the length measured in the object’s own frame. Slice it along a tilted line and the cut is at an angle to the ribbon, and the width obtained is different.

That is the whole derivation, and it produces the standard result

$$L = \frac{L_0}{\gamma}$$

with $\gamma$ the same factor that slows a moving clock, because the geometry doing the tilting is the same geometry.

Why it is a contraction rather than a stretch #

The oblique-cut picture invites an objection: cutting a ribbon at an angle usually gives a longer section, not a shorter one, as anyone who has sliced a carrot diagonally knows.

The answer is the minus sign in the invariant interval. Spacetime geometry is not Euclidean; the “length” of a spacelike separation is $\sqrt{\Delta x^2 - c^2\Delta t^2}$, and the subtraction means that adding a time separation to a spatial one reduces the interval rather than increasing it.

So the oblique cut is shorter, and it is shorter for the same reason the straight worldline between two events has the most proper time rather than the least. Both are the ordinary geometric intuition inverted by one sign, and getting that inversion right is most of the difficulty of learning the subject.

The measurement, described carefully #

Because contraction follows from the definition of a length measurement, it is worth walking through the measurement itself rather than the formula.

An observer wanting the length of a passing rod arranges two markers on their own laboratory floor and requires that the rod’s front end passes one marker at exactly the same moment as its back end passes the other. Then the distance between the markers is the rod’s length in that frame.

Now describe the same procedure from the rod’s frame. The two marker events are not simultaneous there. The observer in the rod’s frame says the back-end event happened first, and that by the time the front-end event occurred, the laboratory had moved on. So the two markers were not opposite the two ends of the rod at the same time at all — the laboratory measured the rod’s length by comparing its front and back at two different moments, and unsurprisingly got a smaller number.

Both accounts are complete and both are correct. The disagreement is not about the rod; it is about which pairs of events count as simultaneous, and it is the same disagreement, with the same size that made time dilation consistent.

The two lengths, and which one is a property #

It is worth being explicit about which quantity belongs to the object and which does not, because the vocabulary hides the distinction.

The proper length is measured in the frame where the object is at rest. It is a property of the object: every observer agrees on what it is, because they all agree about which frame the object is at rest in and about what a measurement made there would give. It plays the role that the invariant interval plays for a pair of events.

The measured length in any other frame is not a property of the object at all. It is a joint property of the object and the frame, in exactly the way that the shadow of a stick is a joint property of the stick and the sun.

The shadow analogy is worth pressing, because it is exact rather than illustrative. A stick has a length; its shadow has a length that depends on the angle of the light, and no one is tempted to say the sun compresses sticks. Relativity’s tilt of the simultaneity axes is a change of angle, the measured length is the projection, and the only reason it feels different is that the geometry has the minus sign in it and the intuition does not.

That reading also explains why the contraction is symmetric. Each of two observers passing one another measures the other’s rod as short, and there is no contradiction, because each is projecting the other’s worldsheet onto a different slice. Two people can each cast a shorter shadow on the other’s wall.

What the contraction is not #

Three misreadings are common enough to be worth ruling out individually.

It is not an optical illusion. The measurement above involves no light travel time; it is defined in terms of local coincidences — the front end passing a marker is an event at one place. Nothing about how long light took to reach an eye enters.

It is not what a fast object would look like. This one is genuinely surprising, and it went unnoticed for fifty years. Seeing an object involves light that left different parts of it at different times, so that light from the far side left earlier. Working through the geometry, Penrose and Terrell showed in 1959 that the combined effect is a rotation rather than a flattening: a passing sphere photographs as a sphere, and a passing cube appears turned. The contraction is in the measurement, not in the appearance, and the two are different calculations that had been silently conflated for two generations.

And it is not a compression. Nothing squeezes the object. There is no stress in the rod, no force acting on it, and no energy required. Two observers disagree about a length in the way two people disagree about which direction is “along” a ribbon, and the rod is entirely unaffected by the disagreement.

The muon, which needs both accounts #

The cleanest demonstration is one where the two effects must be used in different frames to get the same answer, and where the answer is measured.

Muons are produced by collisions at enormous energies about fifteen kilometres up and live 2.2 microseconds on average. Even at light speed that allows only 660 metres of travel, so almost none should reach the ground. Large numbers do.

In the Earth’s frame, the explanation is time dilation. At the speeds involved $\gamma$ is around 20, so the muon’s clock runs slow and it lives, by Earth’s reckoning, about 44 microseconds — enough for thirteen kilometres.

In the muon’s frame, its lifetime is exactly the usual 2.2 microseconds, because a clock always runs at one second per second in its own frame. What is different is the distance: the atmosphere is rushing past at nearly $c$, so its thickness is contracted by the same factor of 20, from fifteen kilometres to 750 metres. That, the muon covers comfortably.

Two frames, two entirely different explanations, one number. Neither account can be used in the other’s frame — a muon that both lived longer and saw a contracted atmosphere would reach the ground far too easily, and the arithmetic would not match the measurement. The effects are not additive extras; each is the whole story in its own frame.

That mutual exclusivity is the strongest single argument that these are descriptions rather than mechanisms. If time dilation were something that happened to clocks and contraction something that happened to rulers, both would be present in both frames. They are not, because what is being described is a relationship between two coordinate systems.

What the contraction costs #

The result is usually presented as a curiosity, and it has consequences that have to be paid for in real hardware.

Nothing can be rigid. A rigid body is one whose parts keep fixed distances regardless of what is done to it, and that requires a push at one end to be felt instantly at the other. Relativity forbids that, so perfect rigidity does not exist — every object is elastic, and a push travels through it at the speed of sound in the material. The idealised rods of elementary mechanics are not merely unrealistic; they are impossible, and the impossibility is a theorem rather than an engineering limit.

Accelerating an extended object is not simple. Applying the same acceleration to the front and back of a rod, as measured in the launch frame, stretches it — because the two ends must be contracting by a growing factor and holding their separation fixed in the original frame prevents that. Bell’s spaceship problem makes this concrete, and it caught out a number of physicists at CERN when it was first put to them. Getting an extended object up to speed intact requires accelerating its parts by different amounts.

And length is no longer a property of an object. It becomes a property of an object and a frame, jointly, which means every engineering statement about dimension carries an implicit frame that had better be stated. In the design of a particle accelerator this is routine: a bunch of particles a few centimetres long in the laboratory is metres long in its own frame — the same machines whose circular orbits stop keeping time with a fixed frequency — and the space-charge forces within it must be computed in the frame where the calculation is valid.

That last case is the one where the effect is used daily. At the LHC, $\gamma$ is about 6,900, so the 27-kilometre ring is, to a proton in it, about four metres around. The beam’s own dynamics — how the particles push each other apart, how the bunch spreads — is calculated in the frame where those forces are electrostatic, and then transformed. Length contraction is not a curiosity in that work; it is a factor in an equation that has to be right or the machine does not store a beam.

The contraction that was invented to explain a null result #

The formula predates the theory by nearly twenty years, and it was proposed as a physical mechanism rather than as a geometric statement.

The Michelson–Morley experiment found no sign of the Earth’s motion through the ether, and FitzGerald in 1889 and Lorentz in 1892 independently suggested a way to save the ether: perhaps objects moving through it are physically compressed along the direction of motion, by exactly the factor needed to cancel the expected fringe shift.

That is the same $1/\gamma$ as above and a completely different claim about the world. On the Lorentz–FitzGerald account, the contraction is a dynamical effect: the ether squeezes the intermolecular forces holding the object together, there is a preferred frame in which objects have their true length, and the contraction is something that happens to matter. Lorentz worked out the details, including how the electromagnetic forces between the molecules would be affected, and the account is internally consistent.

Einstein’s version keeps every formula and discards the mechanism. There is no ether, no preferred frame, no true length, and nothing happens to the object at all — the contraction is a statement about how two coordinate systems relate. The rod is unstressed, unchanged, and shorter in one frame than in another for the same reason that a vector’s components differ between two sets of axes.

The two accounts agree on every prediction that was checkable at the time, which is why the distinction was slow to be accepted and why Lorentz himself never fully abandoned the ether. What settled it was not any single measurement of a length but the accumulation of consequences that fall out of the geometric reading and have to be added by hand to the dynamical one — the velocity addition law, the mass–energy relation, the unification of the electric and magnetic fields, and the whole four-vector formalism.

The episode is a fair example of a pattern worth naming: two theories agreeing on the observations, with one of them making a long list of further results inevitable and the other making them coincidences. That is a real basis for preferring one, and it is not the same as either being refuted.

Where the model stops #

Only along the motion. Dimensions perpendicular to the velocity are unchanged. A passing cube is measured as a rectangular box, flattened in one direction only, and the reason is a symmetry argument: if transverse lengths changed, two observers passing each other could each claim the other was narrower, and a ring passing over a rod would both fit and not fit — a contradiction about a local coincidence, which no frame disagreement is allowed to produce.

Uniform velocity. The derivation slices a worldsheet with straight lines, which belongs to an inertial observer. An accelerating observer’s slices rotate as they go, and lengths become dependent on the whole history rather than on a single speed.

Flat spacetime. In a gravitational field the geometry itself is curved, distances are not simply related between frames, and the notion of the length of an extended object becomes considerably more delicate.

And the effect is imperceptible everywhere ordinary. That curve is the honest summary. At 300 m/s an aircraft is contracted by one part in $2\times10^{12}$, which over the length of a fuselage is a thousandth the width of an atomic nucleus. Every intuition anybody has about length was formed in the flat part of that curve, which is why the result feels like a violation rather than a consequence.

The ladder from here #

Later rungs: the Lorentz transformation written out, with contraction and dilation as two readings of it. The ladder-and-barn paradox, which is this effect and the simultaneity disagreement in the same problem, and which dissolves the moment both are drawn. Bell’s spaceships worked through. Rigidity and Born’s definition of it, which is the closest relativity permits. The Penrose–Terrell rotation computed rather than described. Relativistic volume and density, and the transformation of charge density that makes magnetism a relativistic effect. Proper time as the length of a worldline, and the twin paradox as a statement about two paths. And the four-vector formulation, in which contraction stops being a separate phenomenon and becomes a component of one object being read in two coordinate systems.