The Twin Paradox

"So is this general relativity?" No, though the confusion has a respectable source. Special relativity handles accelerated observers perfectly well; you do not need curved spacetime to describe a rocket firing its engines. What you lose is the right to use a single global inertial frame for the traveller, and that is a bookkeeping restriction, not a limitation of the theory.

The confusion comes from Einstein himself. In a 1918 dialogue published in Die Naturwissenschaften, he answered a critic by adopting the traveller's accelerating frame and invoking the equivalence principle: in that frame there is a uniform gravitational field during the turnaround, and Earth, being high up in that field's potential, ticks fast. The calculation is correct and gives the same 6.4 years. But it is one optional way of describing the turnaround, not the explanation, and generations of textbooks have reported it as though general relativity were required. It is not.

"But the traveller feels the acceleration, so surely that is the difference?" It is a difference, and a real one — she can tell that she turned around without looking out of the window, and that is why the situations are not symmetric. But feeling a force is the flag, not the mechanism. Compare: a car that takes the bendy road has to turn its steering wheel, and the driver feels it. The extra miles on the odometer are not caused by the sensation of turning. They are caused by the road being longer. Acceleration is how you know the path was bent; the bend is what costs the time.

How much does the turnaround itself contribute? You can compute this directly. Model the turnaround as a period of constant proper acceleration and let its duration go to zero while holding the trip's speed and distance fixed; the age difference converges on a nonzero limit. Model it as a long, gentle arc; the age difference is nearly the same. The turnaround supplies an asymmetry, and almost none of the number.

Where the asymmetry can come from instead. The three-clock relay removes acceleration by using two inertial travellers. Michael Brans and David Stewart's 1973 analysis, extended by John Barrow and Janna Levin in 2001, does it another way: in a spatially compact universe — one that wraps around — two observers can separate inertially and meet again, and the ageing difference is still there. In that case the symmetry is broken not by anything either observer does but by the global topology of space, which picks out a preferred frame. This is a genuinely instructive case, because it shows that the asymmetry lives in the geometry of the paths, and acceleration was only ever one way of producing it.

Herbert Dingle's campaign. From the 1950s until his death in 1978, the British physicist and philosopher of science Herbert Dingle argued in print, repeatedly and at length, that the twin paradox showed special relativity to be internally contradictory. His book Science at the Crossroads (1972) is the fullest statement. He was wrong, and his error is exactly the one described above: he insisted that mutual time dilation must be symmetric under exchange of the twins, without noticing that only one twin's frame is inertial throughout. The episode is worth knowing because Dingle was a serious scientist, not a crank by training, and because his objection is the one almost every thoughtful person raises independently.

1905. In §4 of "Zur Elektrodynamik bewegter Körper", Einstein notes that a clock carried around a closed curve and returned to its starting point will lag behind an identical clock that stayed put. He states it for clocks, without comment, as a straightforward consequence of the transformation he has just derived. There is no paradox and no discussion. The result is simply there.

1911. Paul Langevin, lecturing at the international philosophy congress in Bologna and publishing in Scientia, gives the result a human shape: a traveller who leaves in a projectile at very nearly the speed of light, spends two years of his own time away, and returns to an Earth on which two hundred years have passed. Langevin does not say "twins" and does not present it as a paradox — for him it is a striking but unproblematic truth. He does attribute the asymmetry to the traveller having undergone acceleration, which is where the durable misconception enters the literature.

1913. Max von Laue analyses the problem using Minkowski's spacetime diagrams and identifies the real asymmetry: the traveller occupies two distinct inertial frames while the stay-at-home occupies one. This is the correct account, published within two years of Langevin, and it took most of a century to displace the acceleration story in popular treatments.

1918. Einstein publishes his dialogue in Die Naturwissenschaften responding to objections, and analyses the turnaround using the equivalence principle. This gives the right answer and launches the widespread belief that general relativity is needed.

1950s–1970s. Herbert Dingle's long dispute with the physics establishment keeps the problem in the journals and in the letters pages of Nature. The literature it generated is, in aggregate, an unusually thorough working-out of the problem.

October 1971. Joseph Hafele and Richard Keating fly caesium clocks eastward and westward around the world on commercial airliners. Their two papers appear in Science in July 1972.

1976 onward. The muon storage ring at CERN measures the lifetime of muons circulating at $\gamma \approx 29$ and confirms time dilation to a fraction of a percent — under a proper acceleration of about $10^{18}$ times Earth's gravity, which incidentally confirms that acceleration itself has no effect on clock rates beyond what the speed already accounts for.

It killed universal time. Newton's time was a single river flowing at the same rate everywhere for everyone. The twin paradox is the sharpest available demonstration that this is false and that duration is a property of a path, not of the universe. Every clock reads the length of its own history.

It made "now" a local notion. The 6.4 missing years are not an artefact or a bookkeeping trick. They are the direct consequence of the fact that two observers in relative motion genuinely disagree about which distant events are happening at the same moment, and that neither is mistaken. This is the single most consequential idea in special relativity, and the twin paradox is the cleanest way to feel it.

It is engineering, not philosophy. The satellites in the GPS constellation carry atomic clocks moving at about 3.9 km/s and sitting 20,200 km up. Special relativity makes those clocks lose about 7 microseconds a day; the weaker gravity makes them gain about 45. The net offset of roughly 38 microseconds a day is compensated for by design — the onboard oscillators are deliberately set to the wrong frequency before launch. Left uncorrected, position errors would accumulate at about 10 kilometres a day. Relativistic time dilation is a line item in a satellite's specification.

It is why particle accelerators work. Unstable particles produced in accelerators survive long enough to be studied only because their proper time runs slow relative to the laboratory. A muon at rest lives 2.2 microseconds. A muon at $\gamma = 29$ lives 2.2 microseconds of its own time, which is 64 microseconds of ours, which is the difference between a detectable particle and a rumour.

Optical clocks as gravimeters. Clocks are now accurate enough that the relativistic shift is a measurement tool rather than a correction. An optical lattice clock can resolve a height change of a few centimetres from the change in its tick rate. "Chronometric levelling" — surveying the shape of Earth's gravitational potential by comparing clocks — is an active field.

Relativistic navigation beyond Earth. Deep-space navigation, pulsar timing arrays and very long baseline interferometry all require proper-time bookkeeping between clocks in different states of motion and potential. The formalism is exactly the twin calculation, done carefully and repeatedly.

Interstellar travel arithmetic. The traveller's advantage is real and grows without limit: at constant 1g proper acceleration you can, in principle, cross the visible universe within a human lifetime of your own, while billions of years pass outside. The trip is one-way in time. Every serious discussion of interstellar flight — and much of the science fiction worth reading — is built on this asymmetry.

It remains the best teaching instrument in physics. The paradox exposes, faster than anything else, whether a student has understood that simultaneity is frame-dependent. Almost every wrong answer to the twin paradox is the same wrong answer: quietly assuming a universal "now".

Proper time is arc length

In flat spacetime with signature $(-,+,+,+)$, the invariant interval between neighbouring events is

$$ds^2 = -c^2dt^2 + dx^2 + dy^2 + dz^2 .$$

For a timelike worldline, define proper time by $c^2 d\tau^2 = -ds^2$. Then along any worldline parameterised by coordinate time $t$ in some inertial frame,

$$\tau = \int_{t_1}^{t_2}\sqrt{1 - \frac{v(t)^2}{c^2}};dt .$$

The integrand is at most $1$, with equality only when $v = 0$. So among all worldlines connecting the same two events, the one that is at rest in the chosen frame — the inertial, straight one — has the maximum proper time. This is the reverse triangle inequality of Minkowski geometry, and it is the entire content of the paradox. Everything else is arithmetic.

For the standard trip at constant speed $v$ over Earth-frame duration $T$:

$$\tau_{\text{traveller}} = \frac{T}{\gamma} = T\sqrt{1 - v^2/c^2}, \qquad \gamma = \frac{1}{\sqrt{1-v^2/c^2}} .$$

With $v = 0.8c$, $\gamma = 5/3$, $T = 10$ years: $\tau = 6$ years.

The simultaneity jump, derived

Work in Earth's frame $S$, with the launch at the origin. The turnaround event $P$ has coordinates $(t, x) = (5\ \text{yr},\ 4\ \text{ly})$.

The outbound frame $S'$ moves at $+v$. Events simultaneous with $P$ in $S'$ satisfy $t' = \gamma(t - vx/c^2) = \text{const}$, i.e.

$$t - \tfrac{v}{c^2}x = 5 - 0.8(4) = 1.8\ \text{yr}.$$

Setting $x = 0$ gives Earth-time $t = 1.8$ years. That is what "now on Earth" means to the outbound traveller at the moment she reaches the star.

The inbound frame $S''$ moves at $-v$. Its simultaneity condition through $P$ is

$$t + \tfrac{v}{c^2}x = 5 + 0.8(4) = 8.2\ \text{yr},$$

which at $x = 0$ gives Earth-time $t = 8.2$ years.

$$\Delta t_{\text{jump}} = 8.2 - 1.8 = 6.4\ \text{yr} = \frac{2vd}{c^2},$$

with $d = 4$ ly the Earth-frame distance. Combined with $1.8$ years of Earth ageing observed on each leg, the total is $1.8 + 6.4 + 1.8 = 10$ years. The books close.

Note that the jump depends on $d$, the distance at which the frame change occurs. Turn around next to Earth and there is no jump and no age difference. This is the sharpest possible statement that the effect is about where the frame change happens, not how hard it is.

Removing the acceleration

Consider three inertial worldlines and no acceleration at all. Clock $A$ stays on Earth. Clock $B$ passes Earth at $+0.8c$, zeroing its reading at the passing event. Clock $C$ travels at $-0.8c$ and passes the star at the moment $B$ arrives there; at that crossing, $C$ is set to $B$'s reading. When $C$ passes Earth it reads

$$\tau_B + \tau_C = 3 + 3 = 6\ \text{yr},$$

against $A$'s 10. No object ever accelerated. What happened is that the relay "path" is a bent worldline stitched from two straight pieces, and bent worldlines are short. The handoff is the frame change, performed by fiat rather than by a rocket engine, and it costs exactly the same 6.4 years.

The turnaround in the traveller's own accelerated frame

If you insist on describing everything in coordinates comoving with the traveller, use Rindler coordinates during the turnaround. In those coordinates a clock at proper distance $x$ "above" the observer along the acceleration direction runs at rate

$$\frac{d\tau_{\text{distant}}}{d\tau_{\text{local}}} = 1 + \frac{a x}{c^2},$$

with $a$ the proper acceleration. During the turnaround the traveller accelerates towards Earth, so Earth sits high in the effective potential and runs fast. The elapsed Earth time accumulated during a turnaround of proper duration $\Delta\tau$ is approximately $a,d,\Delta\tau/c^2$, and since the speed change is $\Delta v \approx a,\Delta\tau$, this is $\approx \Delta v, d/c^2$ — for a full reversal, $2vd/c^2$, the same $6.4$ years. The answer does not depend on $a$ or $\Delta\tau$ separately, only on their product. That is the formal reason acceleration is not the cause.

Doppler bookkeeping

The relativistic Doppler factor for radial motion is

$$k = \sqrt{\frac{1+\beta}{1-\beta}}, \qquad \beta = v/c .$$

At $\beta = 0.8$, $k = 3$. Note the identity $\gamma = \tfrac12(k + k^{-1})$, which is Bondi's observation that all of the kinematics can be built from $k$ alone. The flash-counting argument in the Examples section is a complete proof of the age difference that never mentions frames, coordinates or simultaneity, and for that reason it is the version to reach for when someone is not yet convinced.