Bell's Spaceship Paradox

The string breaks. `aq-sr-bell-spaceship-paradox` settles that: two ships given identical acceleration programmes in the launch frame hold a constant coordinate gap $L$ while the proper gap their crews measure grows as $\gamma L$ , so the thread is stretched by $\varepsilon = \gamma - 1$ , and no material survives that. This article asks the question that comes after. Not "does it break?" but "what kind of motion is this?" — and the answer is stranger than a snapped wire. Identical engines give identical hyperbolae. Identical hyperbolae have the same shape — the same distance $c^2/\alpha$ from vertex to asymptote — but not the same vertex: one is the other, slid sideways by $L$ . And a hyperbola's vertex is where its Rindler horizon lives. So Bell's two ships do not share a horizon. They have two. Follow that through and something arrives that nobody expects from a problem about a piece of string: after 31 years of its own proper time, at a comfortable $1\,g$ and a separation of a mere hundred metres, the rear ship falls behind the front ship's horizon. The front ship can never hear from it again — not late, not faintly, never. The rear ship can still hear the front. A Born-rigid rod, sharing one vertex and one horizon, never does this at all. The string breaking was the symptom. The two horizons are the disease.

Key takeaways

Signature $(+,-,-,-)$ : $\mathrm{d}s^2 = c^2\mathrm{d}t^2 - \mathrm{d}x^2$ , timelike positive. Flat spacetime throughout; $R^{\lambda}{}_{\mu\nu\kappa} = 0$ .
Recalled, not re-derived. The proper separation $\ell = \gamma L$ , the strain $\boxed{\varepsilon = \gamma - 1}$ , the ignition offset $\Delta t' = -\gamma vL/c^2$ , and the fact that the thread snaps are all owned by `aq-sr-bell-spaceship-paradox`. The rigid profile $\alpha(X) = c^2/X$ and the maximum rigid length are owned by `aq-sr-born-rigidity`. This article uses them and extends.
Owned result: Bell's ships do not share a Rindler horizon — they have TWO. Equal $\alpha$ gives equal vertex-to-asymptote distance $c^2/\alpha$ , but the two vertices are displaced by $L$ . A Born-rigid congruence shares exactly one vertex and therefore one horizon. That single difference is the whole distinction between the two programmes.
Owned result: the rear ship crosses the front ship's horizon in finite proper time. At $\alpha = g$ and $L = 100$ m: at lab time $t_* = (X^2 - L^2)/2Lc = 4.4\times10^{13}$ yr — but at only $\tau_* = \mathbf{31.1\ years}$ of the rear ship's own proper time. After that the front ship is permanently deaf to the rear; the rear can still hear the front. The "symmetric" pair is not symmetric.
Owned result: the covariant statement of the break. The expansion scalar of Bell's congruence is $\boxed{\theta = \alpha\beta/c \neq 0}$ ; the Born-rigid congruence has $\boxed{\theta = \sigma_{\mu\nu} = 0}$ exactly. "The string breaks" is $\theta \neq 0$ , and $\theta$ is a tensor statement — no frame, no slice, no argument.
In the front ship's own Rindler chart, the rear ship's coordinate is $\rho_A(\varphi) = \sqrt{X^2 + L^2 - 2LX\cosh\varphi}$ , which shrinks monotonically to zero. The crew watch their companion fall toward their horizon and freeze.
Key idea: rigidity is having a common vertex. Bell's ships have identical engines and two vertices; a Born-rigid rod has graded engines and one. The thread is only the first thing to notice.

The core insight

Two hyperbolae of the same shape are not the same hyperbola. Slide one sideways by $L$ and you have moved its asymptotes — which is to say you have moved its horizon. Everything peculiar about Bell's congruence follows from that one act of translation: the growing proper gap, the snapped thread, the non-zero expansion, the eventual causal severance of the ships. Born's construction differs by exactly one thing: it refuses to translate. All its worldlines radiate from a single event, so they share one fan of simultaneity slices and one horizon, and nothing ever comes apart.

Where this lives today

The congruence formulation is what the subject uses. No working relativist resolves Bell's paradox by boosting a pair of events; they write down the velocity field and compute $\theta_{\mu\nu}$ . The reason is exactly the reason the CERN theory division got it wrong: an argument that runs through a simultaneity slice can be attacked by asking whose slice; an argument that runs through a tensor cannot. Relativistic continuum mechanics is built on the expansion-and-shear decomposition, and the same decomposition is what Raychaudhuri's equation acts on in general relativity, where it drives the singularity theorems.

A particle bunch in a linac is a Bell congruence. Every particle sees essentially the same gradient at the same lab-frame instant, which is Bell's programme exactly. So the bunch's lab-frame length stays put while its proper length grows as $\gamma$ — and it is in the bunch's own rest frame that the space-charge forces act and the RF bucket must contain it. Rest-frame bunch lengthening is standard beam-dynamics bookkeeping, and it is Bell's thread without the thread.

And the two-horizon structure is the flat-space rehearsal for a real puzzle. Observers causally severed without anything local happening to either is the situation of comoving galaxies receding past the Hubble horizon in an accelerating cosmology. The Bell pair is the simplest system in which two observers, subject to identical physics, end up with different causal pasts — and flat spacetime can do this.

The physical picture

Set the ships going. Rear ship $A$ starts at $x = 0$ , front ship $B$ at $x = L$ , both at rest at $t = 0$ , both then burning at the same proper acceleration $\alpha$ . Write $X \equiv c^2/\alpha$ . Their worldlines are

x_A = X(\cosh\varphi - 1), \qquad x_B = L + X(\cosh\varphi - 1), \qquad ct = X\sinh\varphi ,

with $\varphi = \alpha\tau/c$ the common rapidity. These are hyperbolae — but look at where they are centred. $A$ 's asymptotes meet at $x = -X$ ; $B$ 's meet at $x = L - X$ . Two vertices, displaced by exactly $L$ . Each ship carries its own Rindler horizon, a fixed proper distance $c^2/\alpha$ astern of itself, and the two horizons are $L$ apart. Recall what a horizon is (`aq-sr-rindler-horizon`): a null line beyond which nothing ever signals you again, so long as you keep burning. $B$ 's horizon is the null line through $x = L - X$ . It sits ahead of $A$ 's horizon by $L$ — which means it lies between $A$ 's horizon and $A$ itself. $A$ is on the wrong side of a line that is creeping up on it.

Figure

PUNCHLINE — the whole difference between Bell and Born, in two vertices. LEFT, BELL: identical engines give identical-shaped hyperbolae, but ship B's is ship A's SLID SIDEWAYS BY L — so the two vertices, and therefore the two Rindler horizons (gold), are DISPLACED BY L. Ship A is caught between the two horizons, and B's horizon (solid gold) overtakes it: at rear-ship proper time τ* = 31.1 years, at a comfortable 1 g and a separation of only 100 m, ship A falls behind ship B's horizon and the front ship is DEAF TO THE REAR FOREVER. RIGHT, BORN-RIGID: graded accelerations, all worldlines radiate from ONE vertex, ONE shared horizon. The ships never lose each other, and the proper gap never changes. Rigidity IS having a common vertex

The right-hand panel is the contrast. A Born-rigid rod's points ride hyperbolae of different shape — the accelerations are graded, $\alpha(X) = c^2/X$ — but all their asymptotes meet at a single vertex. One vertex, one horizon, one fan of shared simultaneity slices, and a proper separation that never changes. The two programmes differ in exactly one thing, and it is not the engines. It is where the vertices are.

The core equations

Recalled from the owner, `aq-sr-bell-spaceship-paradox`, and used without re-derivation:

\ell(v) = \gamma(v)\,L, \qquad \boxed{\;\varepsilon = \frac{\ell - L}{L} = \gamma - 1\;} \qquad (1)

where: $\ell$ is the proper separation in the ships' momentarily comoving frame and $\varepsilon$ the engineering strain in a thread of rest length $L$ . Physical reading: unbounded, so the thread breaks. This is the S3 article's owned result and this article does not reprove it. Everything below is what happens next. The two ships' worldlines, and the two vertices they expose:

x_{A} = X(\cosh\varphi - 1), \qquad x_{B} = L + X(\cosh\varphi - 1), \qquad ct = X\sinh\varphi, \qquad X \equiv \frac{c^{2}}{\alpha} \qquad (2)

where: $\varphi = \alpha\tau/c$ is the common rapidity — common, because the engines are identical. Physical reading: $A$ 's asymptotes meet at $x = -X$ and $B$ 's at $x = L - X$ . Same $X$ , different centres. Two hyperbolae, two vertices, two horizons. The rear ship's position in the front ship's own Rindler chart:

\boxed{\;\rho_{A}(\varphi) = \sqrt{\,X^{2} + L^{2} - 2LX\cosh\varphi\,}\;} \qquad (3)

where: $\rho_A$ is the proper distance from $B$ 's vertex to ship $A$ , measured along $B$ 's simultaneity slice at rapidity $\varphi$ — that is, $A$ 's Rindler radial coordinate in $B$ 's frame. Physical reading: it decreases monotonically, and it reaches zero at finite $\varphi$ . Zero Rindler coordinate is the horizon. The rear ship is falling toward the front ship's horizon, and it gets there. The moment it does:

\boxed{\;\cosh\varphi_{*} = \frac{X^{2} + L^{2}}{2LX}\;}, \qquad ct_{*} = \frac{X^{2} - L^{2}}{2L}, \qquad \boxed{\;\tau_{*} = \frac{c}{\alpha}\,\varphi_{*}\;} \qquad (4)

where: $t_*$ is launch-frame time and $\tau_*$ the rear ship's own proper time. Physical reading: at $\alpha = g$ , $L = 100$ m these are $t_* = 4.4\times10^{13}$ yr — longer than the age of the universe by three orders of magnitude — and $\tau_* = \mathbf{31.1\ years}$ . A human lifetime of shipboard time. After $\tau_*$ , no signal ship $A$ sends will ever reach ship $B$ . And the covariant statement of what has gone wrong:

\boxed{\;\theta_{\text{Bell}} = \nabla_{\mu}u^{\mu} = \frac{\alpha\beta}{c} \;\neq\; 0\;} \qquad\text{versus}\qquad \boxed{\;\theta_{\text{Born}} = \sigma_{\mu\nu} = 0\;} \qquad (5)

where: $\theta$ is the expansion scalar of the timelike congruence and $\sigma_{\mu\nu}$ its shear. Physical reading: this is what "the string breaks" means, said in a way that no choice of frame can argue with. Bell's congruence expands; Born's does not. The strain is the integral of the expansion, and the expansion is a tensor.

The congruence reading

Derivation ATwo identical engines give two different horizons

The strain result of `aq-sr-bell-spaceship-paradox` is a statement about a pair of worldlines on a slice. This chapter's business is the congruence — the whole family of worldlines, regarded as a flow — and the first thing a congruence has that a pair of worldlines does not is a set of vertices. The goal of this card is to locate them, and to show that Bell's programme puts two where Born's puts one. Everything else in this article is a consequence.

Step 1. Write the shared programme. Both ships start at rest and burn at the same proper $\alpha$ , so both have the same rapidity $\varphi(\tau) = \alpha\tau/c$ at the same launch-frame time, and their worldlines are (2): congruent curves, one translated from the other by $L$ in $x$ . Why: "identical acceleration programmes in the launch frame" means exactly this and nothing more. It is a definition, and the whole paradox is what it quietly costs.

Step 2. Find each ship's vertex. A hyperbola $x = x_0 + X(\cosh\varphi - 1)$ , $ct = X\sinh\varphi$ satisfies $(x - x_0 + X)^2 - c^2t^2 = X^2$ , so its asymptotes are the null lines through $x = x_0 - X$ . For $A$ ( $x_0 = 0$ ) that vertex is at $x = -X$ ; for $B$ ( $x_0 = L$ ) it is at $x = L - X$ .

\text{vertex}_B - \text{vertex}_A = L .

Why: there it is. The vertices are displaced by exactly the initial separation. Identical engines do not produce a common vertex; they produce a common shape, which is a different and much weaker thing.

Step 3. Convert vertices into horizons. By `aq-sr-rindler-horizon`, an eternally accelerating observer's future horizon is the null line through its vertex — the asymptote of its own hyperbola. So $A$ 's horizon is $x = -X + ct$ and $B$ 's is $x = (L - X) + ct$ . They are two distinct null lines, parallel, separated by $L$ . Why: a horizon is not a place and not a property of the spacetime; it belongs to a worldline. Two worldlines, two horizons — unless they share a vertex, which these do not.

Step 4. Contrast with Born. The Born-rigid congruence (owned by `aq-sr-born-rigidity`) is $x^2 - c^2t^2 = X_i^2$ with $X_i = c^2/\alpha_i$ — every member has a different $\alpha$ and every member has the same vertex, at the origin. So there is exactly one horizon, at $x = ct$ , shared by the entire body. Why: this is the trade. Born buys a common vertex by paying with a graded thrust; Bell keeps the fair, identical thrust and pays with a second vertex. Rigidity is having a common vertex — and equality of engines is precisely what prevents it. *Verified in verify_aq-sr-bell-spaceship-paradox-s6.py: sympy confirms $(x - x_0 + X)^2 - c^2t^2 = X^2$ for both ships, extracts the vertices $-X$ and $L - X$ , and checks their difference is exactly $L$ ; the Born congruence is confirmed to have a single vertex at the origin for every $X_i$ .*

Go deeper · Derivation B · 2 minThe rear ship falls behind the front ship's horizon in 31 years of its own time

Derivation BThe rear ship falls behind the front ship's horizon in 31 years of its own time

Two horizons, displaced by $L$ , is a geometric curiosity until you ask where the ships are relative to them. Ship $A$ sits at $x = 0$ at launch, and ship $B$ 's horizon at that moment is at $x = L - X$ , far behind $A$ — so $A$ starts safely on the visible side. But $B$ 's horizon is a null line: it advances at $c$ . And $A$ , though it accelerates forever, never reaches $c$ . This card computes what that race costs, and the answer is this article's owned structural result.

Step 1. Write $B$ 's horizon and $A$ 's worldline in the same coordinates. $B$ 's future horizon is the null line $x_H(t) = (L - X) + ct$ . $A$ 's worldline, from (2), is $x_A = X(\cosh\varphi - 1)$ with $ct = X\sinh\varphi$ , so $x_A(t) = -X + \sqrt{X^2 + c^2t^2}$ . Why: both curves are asymptotic to lines of slope $c$ , so the competition is decided by the constant offsets, and the offsets differ by exactly $L$ .

Step 2. Ask when $A$ crosses. Set $x_A(t) = x_H(t)$ :

-X + \sqrt{X^{2} + c^{2}t^{2}} = L - X + ct \qquad\Longrightarrow\qquad \sqrt{X^{2} + c^{2}t^{2}} = ct + L .

Square and the $c^2t^2$ cancels, leaving a linear equation:

X^{2} = 2Lct + L^{2} \qquad\Longrightarrow\qquad \boxed{\;ct_{*} = \frac{X^{2} - L^{2}}{2L}\;}

Why: the cancellation is the whole event. A crossing time exists, it is finite, and it is positive for any $L < X$ — which is to say for every physically sensible pair of ships. The rear ship does not merely approach the front ship's horizon asymptotically. It crosses it.

Step 3. Convert to the rear ship's own clock. With $ct_* = X\sinh\varphi_*$ ,

\sinh\varphi_{*} = \frac{X^{2} - L^{2}}{2LX} \qquad\Longrightarrow\qquad \cosh\varphi_{*} = \frac{X^{2} + L^{2}}{2LX}, \qquad \tau_{*} = \frac{c}{\alpha}\varphi_{*} .

Why: $\tau_*$ grows only logarithmically with $X/L$ , since $\varphi_* \approx \ln(X/L)$ for $L \ll X$ . That is what turns an astronomical number into a personal one.

Step 4. Put the numbers in. At $\alpha = g$ and $L = 100$ m, with $X = c^2/g = 9.1648\times10^{15}$ m:

\cosh\varphi_{*} = 4.582\times10^{13} \;\Rightarrow\; \varphi_{*} = 32.15, \qquad t_{*} = \mathbf{4.4\times10^{13}\ yr}, \qquad \tau_{*} = \frac{c}{g}(32.15) = \mathbf{31.1\ years} .

Why: thirty-one years. The launch frame says the crossing takes three thousand times the age of the universe and dismisses it as a fantasy. The rear ship's own calendar says it happens within a career. Both are right, and the gap between them is what $\gamma = 4.6\times10^{13}$ means.

Step 5. Read the causal consequence — it is not symmetric. After $\tau_*$ , ship $A$ lies behind ship $B$ 's horizon: no signal $A$ emits ever reaches $B$ again. But $A$ 's horizon is further back still, and $B$ is nowhere near it, so $B$ 's signals keep reaching $A$ indefinitely. The rear crew hear the front crew forever; the front crew go permanently deaf. Why: identical ships, identical engines, a "symmetric" set-up — and a permanently one-way channel. The asymmetry was there from the start, hidden in the fact that "identical acceleration" is a statement about the launch frame, and the launch frame has a direction. *Verified in verify_aq-sr-bell-spaceship-paradox-s6.py: sympy solves $\sqrt{X^2 + c^2t^2} = ct + L$ and returns $ct_* = (X^2 - L^2)/2L$ ; $\cosh\varphi_* = (X^2+L^2)/2LX$ is confirmed from $\sinh\varphi_* = ct_*/X$ via $\cosh^2 - \sinh^2 = 1$ ; equation (3) is checked to vanish exactly at $\varphi_*$ ; and $\tau_* = 31.14$ yr, $t_* = 4.44\times10^{13}$ yr are recomputed at 50 digits and grepped against this article.*

Go deeper · Derivation C · 2 minThe expansion scalar — what the string breaking looks like to a tensor

Derivation CThe expansion scalar — what the string breaking looks like to a tensor

Everything so far has been argued on slices, and slices invite the objection that has powered this paradox for sixty years: whose slice? A reader who has been burned once by simultaneity is entitled to demand a statement that does not mention a frame at all. The kinematics of a timelike congruence supplies one. Its expansion scalar $\theta$ is built from the four-velocity field and nothing else, and its vanishing is the covariant definition of Born rigidity. The goal here is to compute $\theta$ for both programmes and let the tensor settle the argument.

Step 1. Build the two four-velocity fields. For Bell, every ship at launch-frame time $t$ has the same velocity $v(t)$ , so the field is $u^\mu = \gamma(v)\,(c,\,v)$ — a function of $t$ alone, with no $x$ -dependence at all. For Born, the member passing through the event $(t,x)$ has rapidity $\varphi = \operatorname{arctanh}(ct/x)$ , so $u^\mu = c\,(\cosh\varphi,\,\sinh\varphi)$ with $\varphi = \varphi(t,x)$ . Why: the difference is already visible. Bell's velocity field is constant across the ships; Born's is not, and it is Born's that turns out to be rigid.

Step 2. Compute $\theta = \nabla_\mu u^\mu = \partial_\mu u^\mu$ (flat space, Cartesian). For Bell, $\partial_x(\gamma v) = 0$ because nothing depends on $x$ , so

\theta_{\text{Bell}} = \frac{1}{c}\,\partial_t(\gamma c) = \dot\gamma = \frac{\alpha}{c}\tanh\frac{\alpha\tau}{c} = \boxed{\frac{\alpha\beta}{c}} \;>\; 0 \quad\text{for } \beta > 0 .

Why: positive expansion. The congruence is spreading, and it is spreading at a rate that grows with $\beta$ toward the ceiling $\alpha/c$ . This is not a statement about anybody's ruler.

Step 3. Do the same for Born. With $\varphi = \operatorname{arctanh}(ct/x)$ we have $\partial_{ct}\varphi = x/X^2$ and $\partial_x\varphi = -ct/X^2$ where $X^2 = x^2 - c^2t^2$ , so

\theta_{\text{Born}} = c\sinh\varphi\cdot\frac{x}{X^{2}} - c\cosh\varphi\cdot\frac{ct}{X^{2}} = \frac{c}{X^{2}}\Big(X\cosh\varphi\sinh\varphi - X\sinh\varphi\cosh\varphi\Big) = \boxed{0} ,

using $x = X\cosh\varphi$ , $ct = X\sinh\varphi$ on the congruence. Why: exactly zero, everywhere on the wedge. Not small, not zero to first order — identically zero, for all $\alpha$ , all $X$ , all $\varphi$ . And in $1{+}1$ there is no shear, while for a rod along $x$ with no transverse motion the transverse expansion also vanishes, so the full expansion tensor $\theta_{\mu\nu} = 0$ : both $\theta$ and $\sigma_{\mu\nu}$ . That is the covariant definition of Born rigidity, and the Rindler congruence satisfies it identically.

Step 4. Check that $\theta$ is telling the truth. The proper separation of neighbouring Bell worldlines is $\ell(\tau) = \gamma(\tau)L = L\cosh(\alpha\tau/c)$ , so its fractional rate of change is

\frac{1}{\ell}\frac{\mathrm{d}\ell}{\mathrm{d}\tau} = \frac{\alpha}{c}\tanh\frac{\alpha\tau}{c} = \frac{\alpha\beta}{c} = \theta_{\text{Bell}} . \quad\checkmark

Why: the expansion scalar is the fractional rate of growth of the proper gap, computed without ever choosing a slice. The strain $\varepsilon = \gamma - 1$ that `aq-sr-bell-spaceship-paradox` derives from a boost is the time-integral of a tensor. That is the promotion this article exists to make. *Verified symbolically in verify_aq-sr-bell-spaceship-paradox-s6.py: sympy computes $\partial_\mu u^\mu$ for both fields and returns $\theta_{\text{Bell}} = \dot\gamma$ , evaluated on hyperbolic motion to $(\alpha/c)\tanh(\alpha\tau/c)$ , and $\theta_{\text{Born}} = 0$ identically; the identity $\theta_{\text{Bell}} = (1/\ell)\,\mathrm{d}\ell/\mathrm{d}\tau$ is checked symbolically.*

Figure

The covariant verdict, plotted. The expansion scalar θ = ∇·u is built from the four-velocity field alone — a SCALAR, so every observer computes the same number and no argument about whose simultaneity slice is right can touch it. BELL (red): θ = αβ/c, strictly positive and climbing to the ceiling α/c as β → 1. The congruence spreads, the proper gap grows, the thread snaps. BORN-RIGID (green): θ = 0 identically — not small, not zero to first order, but exactly zero at every event on the wedge, for every α and every β. The shear σ vanishes too, so the full expansion TENSOR is zero. That is the covariant definition of rigidity, and it is the whole content of "the string breaks" said in a way that has no frame in it

Limiting cases

Case	Result	Meaning
$\beta \to 0$	$\theta \to 0$ , $\varepsilon \to 0$	at launch the two programmes are indistinguishable. The divergence is cumulative, not sudden
$\beta \to 1$	$\theta \to \alpha/c$	the expansion saturates. The strain does not — it diverges as $\gamma$
$L \to 0$	$\varphi_* \to \infty$	vertices coincide, horizons merge, and the crossing never happens. Born rigidity is the $L \to 0$ limit of Bell's programme
$L \to X$	$\varphi_* \to 0$	the crossing is immediate: the rear ship starts on the front ship's horizon. This is the Born-rigid length bound $L < c^2/\alpha_{\text{nose}}$ in a new guise
$\alpha = g$ , $L = 100$ m	$\tau_* = 31.1$ yr	lab time $4.4\times10^{13}$ yr. The ratio is $\gamma = 4.6\times10^{13}$
Born-rigid	$\theta = \sigma = 0$ , one horizon	nothing is ever lost, and nothing ever stretches

Common trapTwo hyperbolae of the same shape are not two hyperbolae with the same vertex

This is the error that makes Bell's paradox look symmetric, and it survives even in readers who have correctly worked out that the string breaks. Identical proper acceleration gives both ships the same $X = c^2/\alpha$ — the same distance from vertex to asymptote, the same shape. It does not give them the same vertex, because they started at different places, and a hyperbola is fixed by its shape and its centre. The vertices are displaced by exactly $L$ ; so are the horizons; and the consequence is that the rear ship ends up behind the front ship's horizon while the front never goes behind the rear's. A "symmetric" configuration with a permanently one-way channel. Bell's ships are not two copies of the same situation. They never were.

Common trapWhat this article does NOT own — and a number in shipped canon that does not survive arithmetic

The strain $\varepsilon = \gamma - 1$ , the proper separation $\ell = \gamma L$ , the ignition offset $\Delta t' = -\gamma vL/c^2$ , the CERN straw poll and the Dewan–Beran attribution are owned by `aq-sr-bell-spaceship-paradox` (serial 1913) and are recalled here, not re-derived. The rigid profile $\alpha(X) = c^2/X$ , the maximum rigid length $L_0 < c^2/\alpha_{\text{nose}}$ and Herglotz–Noether are owned by `aq-sr-born-rigidity` (serial 1914). One correction, however, is required. Serial 1913 prints the rigidity correction at $1\,g$ and $L = 100$ m as $a_A/a_B - 1 = 1.0880\times10^{-14}$ . That figure is a floating-point artefact: evaluating $1/(1 - L/X) - 1$ in double precision, with $L/X \approx 10^{-14}$ only about fifty machine epsilons above unity, destroys the low bits by catastrophic cancellation. Computed stably as $L/(X - L)$ at fifty digits, the value is $\mathbf{1.0911\times10^{-14}}$ . (Serial 1914's two-significant-figure $1.09\times10^{-14}$ and its $\Delta\alpha = 1.07\times10^{-13}\ \mathrm{m/s^2}$ are both correct.) This article prints $1.0911\times10^{-14}$ .

Worked examples

1 · When does the front ship go deaf?

Two ships, $100$ m apart, both burning at exactly $1\,g$ on their own accelerometers, started together in the launch frame. At what point does the rear ship pass behind the front ship's Rindler horizon — and what does each crew see?

Step 1 — identify. This is derivation 2. The horizons are displaced by $L$ ; the crossing is at $ct_* = (X^2 - L^2)/2L$ with $X = c^2/\alpha$ .

Step 2 — the geometry. $X = c^2/g = \mathbf{9.1648\times10^{15}\ m}$ . Ship $B$ 's horizon lies at $x = L - X$ , i.e. $100$ m in front of ship $A$ 's. Ship $A$ starts $X - L$ ahead of it — a comfortable $0.97$ light-years — so nothing is wrong at launch.

Step 3 — the crossing, in the lab. $ct_* = \dfrac{X^2 - L^2}{2L} = \dfrac{(9.1648\times10^{15})^2}{200} = 4.20\times10^{29}\ \mathrm{m}$ , so $t_* = \mathbf{4.4\times10^{13}\ years}$ . The launch frame concludes, reasonably, that this will never happen.

Step 4 — the crossing, on board. $\cosh\varphi_* = \dfrac{X^2 + L^2}{2LX} = \mathbf{4.582\times10^{13}}$ , so $\varphi_* = \operatorname{arcosh}(4.582\times10^{13}) = \mathbf{32.15}$ , and the rear ship's own clock reads

\tau_* = \frac{c}{g}\,\varphi_* = (3.057\times10^{7}\ \mathrm{s})(32.15) = 9.83\times10^{8}\ \mathrm{s} = \mathbf{31.1\ years}.

Thirty-one years of shipboard time. A crew that boards at thirty is deaf to its sister ship by sixty-one.

Step 5 — the boundary number. The two ships are $100$ metres apart and burning at a weight you would not notice. Their Lorentz factor at the crossing is $\gamma = 4.6\times10^{13}$ , which is why $4.4\times10^{13}$ launch-frame years compress into $31$ shipboard ones. And the channel is one-way: ship $A$ never crosses ship $B$ 's past light cone again, but ship $B$ has not gone anywhere near ship $A$ 's horizon, so $A$ keeps receiving $B$ forever. Two identical ships. One deaf, one not.

2 · The price of not doing this, and a number in canon that needed fixing

Fly the same two ships Born-rigidly instead — one vertex, one horizon, no crossing, no thread. What must the rear engine do differently, and by how much?

Step 1 — identify. Born rigidity means a shared vertex, hence $\alpha(X) = c^2/X$ , owned by `aq-sr-born-rigidity`. Put the front ship at $1\,g$ : $X_B = c^2/g = 9.1648\times10^{15}$ m, and the rear at $X_A = X_B - L$ .

Step 2 — apply. $\dfrac{\alpha_A}{\alpha_B} = \dfrac{X_B}{X_B - L} = 1 + \dfrac{L}{X_B - L}$ .

Step 3 — compute it stably. Do not evaluate $1/(1 - L/X_B) - 1$ in double precision: $L/X_B \approx 1.09\times10^{-14}$ sits about fifty machine epsilons above unity, and subtracting $1$ annihilates the significant digits. The stable form is $L/(X_B - L)$ , which at fifty digits gives

\frac{\alpha_A}{\alpha_B} - 1 = \mathbf{1.0911\times10^{-14}} .

Serial 1913 prints $1.0880\times10^{-14}$ here, which is the float64 ghost — and the tell is that it returns the same value whether you use $g = 9.8$ or $9.80665$ , $c = 3\times10^8$ or $2.99792458\times10^8$ . A number that does not move when its inputs move is an artefact, not a result.

Step 4 — the thrust difference. $\Delta\alpha = g \times 1.0911\times10^{-14} = \mathbf{1.07\times10^{-13}\ m/s^{2}}$ , in agreement with serial 1914.

Step 5 — the boundary number. One part in $10^{14}$ of thrust is the entire difference between a rigid flight and everything in this article: the snapped thread, the non-zero expansion, the two horizons, and a crew going deaf at sixty-one. The failure is not caused by a large deviation from rigidity. It is caused by an unbounded accumulation of a vanishingly small one.

Physics of the portrait

Strain is the quantity that makes this paradox physical rather than merely puzzling. It is dimensionless, it is a property of the material and not of any observer, and it is unbounded.

Origins · 2 minStrain and how this was discovered

Strain

1660–1959 · Hooke's ut tensio sic vis · Cauchy makes it a tensor · Dewan and Beran make it relativistic and break a thread with it

The classical quantity. Strain — fractional extension, $\varepsilon = \Delta L / L$ — is the deformation half of the pair Robert Hooke announced in 1678 as the anagram ceiiinosssttuv, unscrambled to ut tensio, sic vis: as the extension, so the force. Cauchy turned it in the 1820s from a number into a tensor, the symmetric gradient of the displacement field — and it is in that form, a symmetric object measuring how a continuum's proper distances change, that it reappears in relativity as the expansion tensor $\theta_{\mu\nu}$ . The kinship is not an analogy. It is the same object, promoted to spacetime. What makes it relativistic. In Newtonian elasticity, strain requires an agent: someone must pull. Bell's paradox is remarkable precisely because nobody pulls. Two engines fire identically, no external force acts on the thread, and yet its ends end up further apart than its rest length. The agent is geometry — and the strain is $\varepsilon = \gamma - 1$ , exactly, and without limit. A material with no finite breaking strain does not exist. Who found it. The result is Dewan and Beran's, Am. J. Phys. 27, 517 (1959), "Note on stress effects due to relativistic contraction" — the stress being precisely this. John Bell, who popularised the puzzle in 1976 and never claimed originality for it, used it to argue that Lorentz contraction should be taught as a dynamical effect: something a body does, because the forces binding it are covariant. If contraction were a mere accounting convention, forbidding it would be free. It costs the thread.

History

1678

— Hooke publishes "ut tensio, sic vis", giving strain its first quantitative role.

1909

— Born defines rigid motion by constant proper distance in the comoving frame; the congruence that preserves it turns out to share a single vertex.

1959

— Dewan and Beran, in Am. J. Phys. 27, 517, pose the two-rocket problem and argue that relativistic contraction produces real stress in a body prevented from contracting. This is the paradox's origin.

1966

— Rindler relates the uniformly accelerated frame to the Kruskal extension in Am. J. Phys. 34, 1174, fixing the coordinates and the horizon whose duplication is this article's subject.

1976

— Bell publishes "How to teach special relativity" in Progress in Scientific Culture, reporting that an informal canvass of the CERN theory division produced a clear consensus that the string would not break. It was wrong.

1987

— The essay is reprinted as chapter 9 of Speakable and Unspeakable in Quantum Mechanics, which is how most physicists met the puzzle.

Going further

Texts. Bell, "How to teach special relativity", in Speakable and Unspeakable in Quantum Mechanics, 2nd ed. (CUP, 2004), ch. 9 · Dewan & Beran, Am. J. Phys. 27, 517 (1959) · Rindler, Relativity: Special, General and Cosmological, 2nd ed., ch. 12 · Franklin, Eur. J. Phys. 31, 291 (2010) · Giulini, "The Rich Structure of Minkowski Space", arXiv:0802.4345 · Ellis & Elst, "Cosmological models", on the expansion-shear-vorticity decomposition.

Provenance. Signature $(+,-,-,-)$ . Every equation and number is recomputed in verify_aq-sr-bell-spaceship-paradox-s6.py at fifty digits, which reads this file and asserts each printed value against what it computed. Base defect repaired: the AllQFT base's eq[5] is the literal placeholder $\Delta\tau = (c/\alpha_A)\sinh^{-1}(\dots)$ with the argument left as an ellipsis — unusable, and discarded. Its applications card claims that Bell's rear ship "follows a hyperbola with a larger pivot"; that is wrong — both ships have the same pivot $c^2/\alpha$ and different vertices, and the correction is this article's subject. Canon defect repaired: serial 1913's $1.0880\times10^{-14}$ is a double-precision cancellation artefact; the correct value is $1.0911\times10^{-14}$ (see warning 2 and worked example 2).

Scope. The strain, the proper separation and the simultaneity mechanism are owned by `aq-sr-bell-spaceship-paradox`; the rigid profile, the length bound and Herglotz–Noether by `aq-sr-born-rigidity`. Both are recalled by name and used, never re-derived. The horizon itself is `aq-sr-rindler-horizon`'s; the covariant rigidity condition is developed in `aq-sr-born-rigidity-s6`.

Builds on. Bell's Spaceship Paradox (S3) · Born Rigidity · The Rindler Horizon · Relativistic Uniform Acceleration. Leads to. Born Rigidity and Rigid Accelerated Motion (S6) · The Equivalence Principle.

If both ships accelerate identically, how can the situation be asymmetric?

Because "identically" is a statement about the launch frame, and the launch frame has a direction. Slide one hyperbola sideways by $L$ and you slide its vertex, and therefore its horizon, sideways by $L$ too — so one ship's horizon ends up in front of the other's, not coincident with it. That single displacement is the entire asymmetry, and it is enough to give the rear ship a fate the front ship does not share. The engines are symmetric. The geometry is not, and geometry is what decides.

Does the rear ship notice anything as it crosses the front ship's horizon?

Nothing whatever. There is nothing there: spacetime is flat, the Riemann tensor is zero, no instrument aboard could detect the crossing, and the rear ship's own horizon is still far behind it. The crossing is a fact about the front ship's causal past, not about the rear ship's surroundings. What the front ship sees is a slow catastrophe: its companion's signals arriving ever more redshifted and delayed, its image dimming and freezing, the last frame stretched across the front ship's infinite future. Exactly what an observer outside a black hole sees of an infaller — and there is no black hole.

Why does the crossing take 31 years of ship time but 10 to the 13 years of lab time?

Because at the crossing $\gamma = \cosh\varphi_* = 4.6\times10^{13}$ , and that is precisely the ratio between the two clocks. The rapidity needed is $\varphi_* = \operatorname{arcosh}\!\big((X^2+L^2)/2LX\big) \approx \ln(X/L)$ , and a logarithm of $10^{14}$ is a mere $32$ . Proper time grows linearly in rapidity, $\tau = (c/\alpha)\varphi$ , while coordinate time grows exponentially in it. Thirty-two units of rapidity is thirty-one years at $1\,g$ and thirteen orders of magnitude of lab time. The whole strangeness of the result lives in that mismatch.

Is this the same thing as the string breaking, or a different phenomenon?

The same phenomenon, seen at a different range. Both are consequences of the two vertices: on a local scale the displaced vertices make the proper separation grow, which strains the thread ( $\varepsilon = \gamma - 1$ , and it snaps within weeks); on a global scale the displaced vertices mean displaced horizons, and the rear ship eventually falls behind the front one's. The thread is the short-term symptom and the horizon crossing is the long-term one. `aq-sr-bell-spaceship-paradox` owns the first; this article owns the second.

What exactly does θ = 0 buy you that "the proper distance is constant" does not?

Freedom from the question "constant on whose slice?" — which is the question that sank the CERN theory division. The expansion scalar $\theta = \nabla_\mu u^\mu$ is built from the four-velocity field alone. It is a scalar: every observer computes the same number, and no argument about simultaneity can touch it. Saying "the Born congruence has $\theta = \sigma_{\mu\nu} = 0$ " and "Bell's has $\theta = \alpha\beta/c > 0$ " settles the matter without anybody ever choosing a frame. That is why `aq-sr-born-rigidity-s6` adopts $\theta_{\mu\nu} = 0$ as the definition of rigidity rather than deriving it as a theorem.

Would the effect vanish if the ships started closer together?

It weakens, but it never vanishes for $L > 0$ . The crossing rapidity is $\varphi_* \approx \ln(X/L)$ , so halving $L$ adds only $\ln 2 = 0.69$ to $\varphi_*$ — about eight extra months of ship time at $1\,g$ . Even ships a millimetre apart go deaf after about $42$ years of their own time. The only way to remove the effect entirely is $L = 0$ , which is to say to make the two vertices coincide — which is to say to fly Born-rigidly. There is no small- $L$ escape, only a logarithmic reprieve.