Relativity of Simultaneity

A 100-metre railway carriage runs past you at $\beta = 0.6$ . Inside it two clocks, one at each end, have been carefully synchronized by the passengers. You, on the platform, read both at one instant of your time and find the leading clock 200 nanoseconds behind the trailing one. Not because light took time to reach you — you corrected for that. They disagree because "at one instant" is not a phrase two frames can share. That number lives in one algebraic term. The Lorentz time line reads $t' = \gamma(t - vx/c^2)$ , and the piece that matters is $-\gamma v x/c^2$ : the time a frame assigns to an event depends on where the event is. Simultaneity is not a property of a pair of events, but of a pair of events and a frame. The article The Relativity of Simultaneity tells the story qualitatively; this one is the quantitative owner. It derives the effect from the Lorentz transformation, extracts the leading-clocks-lag rule $\Delta T = v L_0/c^2$ and shows why the $\gamma$ cancels, then proves the theorem that makes the whole business safe: a boost can reverse the time-order of two events if and only if the pair is spacelike, $c|\Delta t| < |\Delta x|$ — precisely the condition that no signal could carry one event's influence to the other. Order is negotiable exactly where causation is impossible. That is the light-cone structure of spacetime, and it is why relativity can dismantle absolute time without touching cause and effect.

Key takeaways

Two events at the same time but different places in $S$ are not simultaneous in a frame $S'$ moving along the separation: $\Delta t' = -\gamma\, v\, \Delta x/c^2$ .
Clocks at rest in $S'$ and synchronized there, a proper distance $L_0$ apart, are judged from $S$ to be out of step by $\Delta T = v L_0/c^2$ , leading clock behind. The $\gamma$ cancels: $100\ \mathrm{m}$ at $\beta = 0.6$ gives $200\ \mathrm{ns}$ .
A boost reverses a pair's order iff the pair is spacelike ( $c|\Delta t| < |\Delta x|$ ); the reversing speeds are those beyond $v_{\rm flip} = c^2\Delta t/\Delta x$ .
For timelike or lightlike pairs — the only ones that can be cause and effect — no boost reverses the order. Causality survives untouched.
Key idea: simultaneity is frame-dependent; causal order is not. The Lorentz transformation makes "now" a question about frames, and the invariant $s^2 = c^2\Delta t^2 - \Delta x^2$ decides, with no appeal to any frame, which pairs are up for renegotiation.

The core insight

An event's time coordinate depends on its position: $t' = \gamma(t - vx/c^2)$ . Two events at equal $t$ and unequal $x$ get unequal $t'$ , so "simultaneous" is defined only relative to a frame. This is not a signal delay and not an illusion — it survives after every light-travel time is subtracted. And it is bounded exactly where it must be: only spacelike pairs can be reordered, and they cannot influence one another.

Where this lives today

The clock offset is a line item in the GPS specification. The system defines coordinate time in one non-rotating Earth-centred frame, and neither ground stations nor satellites are at rest in it. A light signal sent eastward round the equator accumulates a simultaneity gap of $2\Omega A/c^2 \approx 207\ \mathrm{ns}$ ( $\Omega$ the Earth's rotation rate, $A$ the enclosed area) — about $62\ \mathrm{m}$ of ranging error. This is the Sagnac correction, applied by every receiver: not a footnote to the relativity of simultaneity, but that relativity itself, integrated round a loop. The same arithmetic disciplines every network that must agree on "now" across a distance: geodetic VLBI, where telescopes on different continents reconcile clock epochs to picoseconds; the timescale TAI, assembled from clocks not at rest with respect to one another; and the fibre links comparing optical clocks near $10^{-18}$ . In the other direction the theorem does quiet, indispensable work: a meson produced in one frame and decaying in another must decay after it is made in every frame — and it does, because the pair is timelike.

The physical picture

Put a lamp at the centre of a railway carriage and flash it once. In the carriage frame the walls are equidistant and light runs at $c$ both ways, so the arrivals happen at the same instant. Now watch from the platform. Light still travels at $c$ in both directions — postulate 2 is not negotiable — but the rear wall is rushing toward the place where the flash happened while the front wall is running away from it. It hits the rear wall first: same lamp, same light, a different verdict on "at the same time."

Figure

The train and the lamp (β = 0.6, γ = 1.25). TOP — in the carriage frame the flash reaches both walls together, 166.8 ns after emission (proper half-length 50 m). BOTTOM — on the ground the carriage is 80.0 m long, so each wall starts 40.0 m away; the rear runs INTO the light (closing speed c + v) and is struck at 83.4 ns, the front runs AWAY (c − v) at 333.6 ns — 250.2 ns apart in S, simultaneous in S′.

The disagreement is systematic: the more-forward event happens earlier, in strict proportion to the separation — the signature of $-\gamma v x/c^2$ , a linear ramp in $x$ . Geometrically each frame slices spacetime into sheets of "now"; boost, and the sheets tilt. The canon article Galilean Transformations states the classical picture exactly: " $t' = t$ asserts absolute time: one universal clock, so simultaneity is absolute," and the Galilean boost is "a shear: worldlines tilt, simultaneity lines stay horizontal." The Lorentz boost tilts the horizontals too — the axes scissor toward the light line, and never past it.

Figure

PUNCHLINE. LEFT — the scissoring: a β = 0.6 boost tilts the x′ axis UP by arctan β = 30.96° and the ct′ axis IN by the same angle, so P and Q (one t in S) sit on DIFFERENT tilted slices. RIGHT — the limit of the mischief, c = 1. A(0,0) → B(ct = 1, x = 3) is SPACELIKE (s² = −8): its order flips past β = 0.333, and at β = 0.6 reads Δt′ = −1.0. A(0,0) → C(ct = 3, x = 1) is TIMELIKE (s² = +8): reversal needs β = 3, which does not exist; Δt′ = +3.0.

The core equations

Standard configuration: $S'$ slides along the $x$ -axis of $S$ at constant $v$ , axes parallel, origins coinciding at $t = t' = 0$ , each frame Einstein-synchronizing its own clocks. Write $\beta \equiv v/c$ , $\gamma \equiv 1/\sqrt{1-\beta^2}$ .

t' = \gamma\left(t - \frac{v x}{c^{2}}\right), \qquad x' = \gamma\,(x - vt), \qquad y' = y, \qquad z' = z \qquad (1)

where: $(t, x)$ — an event in $S$ ; $(t', x')$ — the same event in $S'$ ; $c = 299\,792\,458\ \mathrm{m/s}$ exactly ( $\approx 2.998\times10^{8}$ ). Physical reading: the time line is what matters. It is not $t' = t$ plus a small correction — it carries a term proportional to the event's position, a ramp: the further along the boost an event sits, the more its time is shifted. Where (1) comes from is Derivation of the Lorentz Transformation.

\Delta t' = \gamma\left(\Delta t - \frac{v\,\Delta x}{c^{2}}\right) \qquad (2)

where: $\Delta t = t_B - t_A$ , $\Delta x = x_B - x_A$ — coordinate differences of two events in $S$ . Physical reading: the map is linear, so differences transform exactly as coordinates do. This is the only formula in the article; everything below is a special case.

\Delta t = 0 \ \ \Longrightarrow\ \ \Delta t' = -\,\frac{\gamma\, v\, \Delta x}{c^{2}} \qquad (3)

where: $\Delta t = 0$ — the events are simultaneous in $S$ , separated by $\Delta x \ne 0$ along the boost. Physical reading: the relativity of simultaneity in one line. The offset dies only if $v = 0$ or $\Delta x = 0$ . Mind the sign: with $v > 0$ , $\Delta x > 0$ , the event further along $+x$ gets the smaller $t'$ — it happens earlier in $S'$ . Note also what is absent: $\Delta y$ , $\Delta z$ — only the separation parallel to the boost counts.

\Delta T = \frac{v\,L_0}{c^{2}} \qquad\text{(leading clocks LAG)} \qquad (4)

where: $L_0$ — the proper distance between two clocks at rest in $S'$ and synchronized there; $\Delta T$ — how far the leading clock reads behind the trailing one, at one instant of $S$ . Physical reading: no $\gamma$ — it cancels against the $1/\gamma$ of the contracted separation $\Delta x = L_0/\gamma$ . The most useful number in the subject: $100\ \mathrm{m}$ at $\beta = 0.6$ gives $200.1\ \mathrm{ns}$ .

s^{2} \equiv c^{2}\Delta t^{2} - \Delta x^{2} = c^{2}\Delta t'^{2} - \Delta x'^{2}, \qquad v_{\rm flip} = \frac{c^{2}\,\Delta t}{\Delta x} \qquad (5)

where: $s^2$ — the invariant interval, the same in every frame (signature $+\,-\,-\,-$ : $s^2 > 0$ timelike, $s^2 < 0$ spacelike, $s^2 = 0$ lightlike); $v_{\rm flip}$ — the unique boost at which $\Delta t'$ vanishes. Physical reading: $s^2$ is the referee — built from coordinates, belonging to no frame — and its sign decides whether $v_{\rm flip}$ is a real speed or a fantasy: $|v_{\rm flip}| < c$ exactly when $s^2 < 0$ . The taxonomy belongs to Timelike, Spacelike, and Lightlike Separations and The Light Cone.

Where the results come from

Derivation ATwo events, one instant — and the clocks that refuse to agree

The goal is to extract from the Lorentz time line the two numbers a physicist actually uses, and to keep them apart, because students routinely fuse them into one. The coordinate-time gap: two events $S$ calls simultaneous — how far apart in time does $S'$ put them? The clock-reading offset: two clocks at rest in $S'$ and synchronized there — what do their faces read at one instant of $S$ ? The first carries a $\gamma$ ; the second does not, and the disappearance of that $\gamma$ is a cancellation we can watch happen.

Step 1. Take $A$ and $B$ simultaneous in $S$ : $t_A = t_B = t$ , $x_B - x_A = \Delta x \ne 0$ . Apply the time line to each and subtract. The common $\gamma t$ cancels and only the position term survives: $\Delta t' = -\gamma v \Delta x/c^2$ . Why this is already the whole phenomenon: nothing was assumed about signals, eyes or delays — two coordinate systems, one linear map, and the times differ.

Step 2. Now the operational question. Put clocks at $x' = 0$ and $x' = L_0$ , at rest in $S'$ and synchronized there. In $S$ they are moving, and the distance $S$ measures between them is the contracted $L_0/\gamma$ ; their worldlines in $S$ are $x = vt$ and $x = vt + L_0/\gamma$ . Why the contracted length: the question fixes an instant of $S$ , so the separation that enters is the one $S$ measures at that instant — using $L_0$ here is the commonest error in the subject.

Step 3. Feed each worldline into the time line. Rear: $t'_{\rm rear} = \gamma(t - v(vt)/c^2) = \gamma t(1 - \beta^2) = t/\gamma$ — the familiar dilation, no more. Front: $t'_{\rm front} = \gamma\left(t - v(vt + L_0/\gamma)/c^2\right) = t/\gamma - \gamma\cdot\dfrac{v}{c^2}\cdot\dfrac{L_0}{\gamma} = t/\gamma - \dfrac{v L_0}{c^2}$ . Why the $\gamma$ dies: it enters once from the transformation and once, inverted, from the contraction — exact at every $\beta$ .

Step 4. Read it off: $t'_{\rm rear} - t'_{\rm front} = vL_0/c^2 > 0$ . The leading clock reads behind — and the clocks are perfect, running at the same rate, correctly synchronized in $S'$ . Now keep the two answers apart: the coordinate-time gap between the same pair of events, measured in $S$ , is $\gamma vL_0/c^2$ — a factor $\gamma$ larger, because the moving clocks tick slowly, so their $vL_0/c^2$ of reading difference is spread over $\gamma$ times as much $S$ -time. At $L_0 = 100\ \mathrm{m}$ , $\beta = 0.6$ : offset $200.1\ \mathrm{ns}$ , gap $250.2\ \mathrm{ns}$ . Why both: the figure gets $250.2\ \mathrm{ns}$ by closing speeds and (4) gets $200.1\ \mathrm{ns}$ — expect them to agree and you will wrongly conclude one is broken. *Verified in verify_aq-sr-relativity-of-simultaneity-s1.py: sympy transforms both worldlines, returns $t/\gamma$ and $t/\gamma - vL_0/c^2$ , and confirms the cancellation is exact; the closing-speed result is shown identical to $\gamma v(2L)/c^2$ ; the numbers $1.25$ , $80.0\ \mathrm{m}$ , $83.4$ , $333.6$ and $250.2\ \mathrm{ns}$ are recomputed.*

Go deeper · Derivation B · 2 minTHE THEOREM: a boost can reverse a pair's order if and only if the pair is spacelike

Derivation BTHE THEOREM: a boost can reverse a pair's order if and only if the pair is spacelike

Here is the result that keeps relativity from collapsing into nonsense. If a boost can shuffle the order of events, why can it not put the death before the murder? The answer is a two-directional theorem: the pairs whose order a boost can reverse are exactly the pairs no signal could connect. The criterion is one inequality, $c|\Delta t| < |\Delta x|$ , frame-independent because it is a statement about the invariant $s^2$ . We prove both halves.

Step 1. Start from (2), with $\Delta t > 0$ ( $A$ the earlier event) and $\Delta x \ne 0$ . Since $\gamma > 0$ always, the sign of $\Delta t'$ is the sign of the bracket alone. The bracket vanishes at exactly one speed, $v_{\rm flip} = c^2\Delta t/\Delta x$ , and changes sign as $v$ passes through it. Why $\gamma$ can be ignored: it stretches the interval but never turns it round — all the drama sits in a linear function of $v$ .

Step 2. Ask the only question that matters: is $v_{\rm flip}$ a speed a frame can have? That needs $|c^2\Delta t/\Delta x| < c$ , i.e. — dividing by $c$ — $c\,|\Delta t| < |\Delta x|$ , which is precisely $c^2\Delta t^2 - \Delta x^2 < 0$ : the pair is spacelike. Why this is a real criterion and not a frame-dependent accident: $s^2$ is invariant, so every observer agrees on the label. Observers may disagree about which event came first; never about whether such a disagreement is possible.

Step 3. Forward direction — spacelike pairs do flip. Write $\Delta x = k\,c\,\Delta t$ with $|k| > 1$ (that is what spacelike means when $\Delta t \ne 0$ ). Then $\Delta t' \propto \Delta t\,(1 - \beta k)$ : zero at $\beta = 1/k$ , a legitimate speed since $|1/k| < 1$ , and negative for every $\beta$ strictly between $1/k$ and $\mathrm{sign}(k)$ . The order is reversed by an open interval of ordinary frames. Why this is not yet a paradox: these are exactly the pairs for which "which came first" was never a physical question.

Step 5. Read the physics. A pair can influence one another only if a signal at speed $\le c$ gets from one to the other, which requires $|\Delta x| \le c|\Delta t|$ — the same inequality, same boundary, as the no-reversal condition of Step 4. The two sets coincide exactly: order is negotiable precisely where causation is impossible. Relativity protects causality not by bolting on a rule against paradox, but because the light cone that limits signals is the light cone that limits boosts. Why: the taxonomy (Timelike, Spacelike, and Lightlike Separations), the geometry (The Light Cone) and the logic (Causal Structure of Spacetime) are three faces of one inequality. *Verified in verify_aq-sr-relativity-of-simultaneity-s1.py: sympy solves $\Delta t' = 0$ for $v_{\rm flip} = c^2\Delta t/\Delta x$ , verifies $c^2\Delta t'^2 - \Delta x'^2$ is boost-invariant, reduces $\Delta t'/\Delta t$ to $\gamma(1 - \beta k)$ , scans $\beta \in (-1,1)$ with $|k| \le 1$ confirming $1 - \beta k > 0$ everywhere, and for $|k| > 1$ exhibits an explicit reversing boost. The figure's $s^2 = -8$ , $\Delta t' = -1.0$ , $s^2 = +8$ , $\Delta t' = +3.0$ are recomputed.*

Limiting cases

Case	Result	Meaning
$v \to 0$ (or $c \to \infty$ )	$\Delta t' \to \Delta t$	absolute simultaneity restored — the Galilean $t' = t$ , and why "now" feels universal
$v = 250\ \mathrm{m/s}$ , $\Delta x = 100\ \mathrm{m}$	$\Delta t' \approx 0.28\ \mathrm{ps}$	an airliner over a railway carriage — nobody was going to notice by eye
spacelike, $c\lvert\Delta t\rvert < \lvert\Delta x\rvert$	order flips for $\lvert v\rvert > \lvert v_{\rm flip}\rvert$	some frames say $A$ first, some $B$ first, one says "at once" — none is wrong
timelike / lightlike	$\mathrm{sign}(\Delta t') = \mathrm{sign}(\Delta t)$	cause before effect, in every frame, always. The one thing no boost can touch

Common trap“So the order of events is arbitrary — with a fast enough rocket I could watch the effect before the cause.”

No, and the theorem above is the sentence that says no. Reordering is available to spacelike pairs and to nothing else. A spacelike pair is one even light cannot bridge: $|\Delta x| > c|\Delta t|$ means a signal leaving the first event at the maximum speed allowed arrives at the place of the second after it has already happened. Nothing that occurred at $A$ could have made any difference at $B$ , or the reverse; to say that frames order such a pair differently is to say that a question with no physical content has no frame-independent answer — not a scandal, but hygiene. Every pair that is a cause and its effect satisfies $|\Delta x| \le c|\Delta t|$ , and no boost short of $c$ reverses it. The relativity of simultaneity is not a licence to reorder the world; it says the world contains less order than you assumed — and exactly as much as it needs.

Worked examples

1 · The carriage, the clocks, and the two answers

A carriage of proper length $L_0 = 100\ \mathrm{m}$ passes a platform at $\beta = 0.6$ , carrying two end-clocks synchronized by the passengers. (a) At one instant of platform time, by how much do the clocks disagree, and which is behind? (b) A lamp at the midpoint flashes: how far apart in platform time are the two wall-arrival events?

Step 1. Kinematics. $\gamma = 1/\sqrt{1 - 0.36} = 1/\sqrt{0.64} = \boxed{1.25}$ , so the platform measures the carriage as $L = L_0/\gamma = 100/1.25 = \boxed{80.0\ \mathrm{m}}$ .

Step 2. Part (a). Equation (4): $\Delta T = vL_0/c^2 = (0.6)(100)/c = 60/(2.998\times10^{8}) = \boxed{200.1\ \mathrm{ns}}$ , the leading (front) clock behind. Note the proper length enters, not the contracted $80.0\ \mathrm{m}$ — the $\gamma$ has already eaten the contraction.

Step 3. Part (b). On the platform the carriage is $80.0\ \mathrm{m}$ long, so each wall starts $40.0\ \mathrm{m}$ from the flash. Rear wall (closing speed $c + v$ ): $t_{\rm rear} = 40/(1.6c) = \boxed{83.4\ \mathrm{ns}}$ . Front wall ( $c - v$ ): $t_{\rm front} = 40/(0.4c) = \boxed{333.6\ \mathrm{ns}}$ . The arrivals are $\boxed{250.2\ \mathrm{ns}}$ apart, rear first.

Step 4. Reconcile — and notice the two answers are not the same number. The inverse of (2) with $\Delta t' = 0$ , $\Delta x' = L_0$ predicts a coordinate gap $\gamma vL_0/c^2 = 1.25\times200.1 = 250.2\ \mathrm{ns}$ : the closing-speed answer exactly. The moving clocks run slow by $\gamma$ , so $250.2/1.25 = 200.1\ \mathrm{ns}$ of carriage time — the reading difference, from the other side. Both point the same way: the rear wall is hit first, and the front clock runs behind.

2 · Firecrackers, a lamp, and what a boost cannot do

Two firecrackers $3.0\ \mathrm{km}$ apart along $x$ explode simultaneously in the ground frame $S$ . (a) How far apart in time are they at $\beta = 0.5$ ? (b) Could one have caused the other? (c) Now a genuine causal pair — a lamp flashes at $x = 0$ and $20\ \mu\mathrm{s}$ later a detector $3.0\ \mathrm{km}$ away registers the pulse. What boost would reverse that?

Step 1. Part (a). $\gamma(0.5) = 1/\sqrt{0.75} = \boxed{1.1547}$ . Equation (3): $|\Delta t'| = \gamma v\Delta x/c^2 = (1.1547)(0.5)(3000)/c = 1732.05/(2.998\times10^{8}) = \boxed{5.78\ \mu\mathrm{s}}$ (precisely $5.777\ \mu\mathrm{s}$ ) — the leading firecracker first, by nearly six microseconds.

Step 2. Part (b) — check the interval before believing anything. Light needs $3000/c = \boxed{10.0\ \mu\mathrm{s}}$ to cross $3.0\ \mathrm{km}$ , and the explosions were $0\ \mu\mathrm{s}$ apart in $S$ . So $c|\Delta t| = 0 < 3000\ \mathrm{m} = |\Delta x|$ : the pair is $\boxed{\text{spacelike}}$ , $s^2 = -\Delta x^2 < 0$ . Nothing could have carried news of one explosion to the other, so there is no cause and effect to damage and the frames' disagreement costs nothing.

Step 3. Part (c). Now $\Delta t = 20\ \mu\mathrm{s}$ , $\Delta x = 3000\ \mathrm{m}$ , so $c\,\Delta t = (2.998\times10^{8})(20\times10^{-6}) = \boxed{5996\ \mathrm{m}}$ , comfortably more than $3000\ \mathrm{m}$ . The pair is $\boxed{\text{timelike}}$ , $s^2 = 5996^2 - 3000^2 > 0$ : light made the trip with time to spare, taking $10.0\ \mu\mathrm{s}$ of the $20$ available.

Step 4. The reversing boost. Equation (5): $\beta_{\rm flip} = c\,\Delta t/\Delta x = 5996/3000 = \boxed{1.999}$ . That is not a speed — no inertial frame moves at $1.999c$ , so no observer anywhere sees the detector fire before the lamp. The general rule: compare $c|\Delta t|$ with $|\Delta x|$ before reaching for any transformation. If light has time to spare, the order is carved in stone; if not, it was never a fact about the world.

Physics of the portrait

The portrait is not a person but a word, and the word is older than the physics. For two centuries "simultaneous" named a fact about the world, guaranteed by Newton's absolute time: two events either happened together or they did not, and how you would check was thought too trivial to ask. Poincaré noticed in 1898 that it is not trivial at all.

Origins · 2 minSimultaneity and how this was discovered

Simultaneity

1687–1908 · Newton's absolute now · Poincaré's convention · Einstein's postulates · Minkowski's tilted slices

The concept has four owners and no inventor. Newton (1687) made it absolute by fiat: one universal time, one "now", the same slicing of the world into instants for everyone. Henri Poincaré (1854–1912) was the first to see that the fiat concealed a decision. In La mesure du temps (1898) he pointed out that we have no direct intuition of the simultaneity of two distant events, and that the rule we actually use — correcting a light signal for its transmission time — presupposes the signal takes the same time each way. That, he said, is a postulate adopted because it makes the equations simple, not a fact discovered. Einstein (1905), in section 1 of Zur Elektrodynamik bewegter Körper, made the definition a foundation rather than a footnote, then showed in section 2 that two observers in relative motion, each applying it faithfully, must disagree — a theorem, not a paradox. Minkowski (1908) supplied the picture: the disagreement is a tilt, and the family of "nows" is not one stack of sheets but infinitely many, one per frame. What survived all four is the light cone.

History

1687

— Newton's Principia: absolute time "flows equably without relation to anything external", so simultaneity is a fact about the world, the same for every observer.

1898

— Poincaré, "La mesure du temps", Revue de métaphysique et de morale 6, 1–13: we have no direct intuition of the simultaneity of distant events; the rule that fixes it presupposes light takes the same time each way.

1900

— Poincaré, "La théorie de Lorentz et le principe de réaction": Lorentz's "local time" is read physically as what clocks show when moving observers synchronize them with light signals.

1905

— Einstein, "Zur Elektrodynamik bewegter Körper", Annalen der Physik 17, 891–921 (received 30 June, published 26 September): section 1 defines simultaneity by a light-signal convention; section 2 derives its relativity.

1908

— Minkowski's Cologne lecture Raum und Zeit: a boost tilts the surfaces of simultaneity; spacetime has no preferred slicing into instants.

1913

— Sagnac's rotating interferometer (Comptes Rendus 157): light-signal synchronization around a closed rotating loop fails to close on itself.

Going further

Texts. Taylor and Wheeler, Spacetime Physics, 2nd ed., ch. 3 · Rindler, Introduction to Special Relativity, 2nd ed., §2 · Morin, Special Relativity, ch. 1–2 · Griffiths, Electrodynamics, 4th ed., §12.1.

Provenance. The offset $\Delta t' = -\gamma v\Delta x/c^2$ follows from the standard Lorentz time transformation (Rindler §2, Griffiths §12.1); the leading-clocks-lag rule and the train-and-lamp arithmetic follow Taylor and Wheeler ch. 3. The order-reversal theorem is proved from scratch here. `verify_aq-sr-relativity-of-simultaneity-s1.py` re-derives every relation symbolically — the $\gamma$ -cancellation, the invariance of $c^2\Delta t'^2 - \Delta x'^2$ , the root $v_{\rm flip} = c^2\Delta t/\Delta x$ , and the marquee $\Delta t'/\Delta t = \gamma(1 - \beta k)$ with its two-sided sign argument — and asserts every printed number, roundings included.

Builds on. The Two Postulates of Special Relativity · Einstein Clock Synchronization · The Lorentz Factor Gamma · Galilean Transformations. Leads to. The Lorentz Transformation Equations · Length Contraction · The Invariant Spacetime Interval · Timelike, Spacelike, and Lightlike Separations · The Pole-and-Barn Paradox.

Isn't this just about light taking time to reach my eye? Correct for the delay and surely the events are simultaneous again.

No: the delay has already been corrected. The coordinates $(t, x)$ in equation (1) are not "what I saw when"; they are what a frame assigns to an event after subtracting every light-travel time, using clocks laid across the frame and synchronized in advance. Two events with equal $t$ are simultaneous in the fully corrected, delay-free sense — and they still come out with different $t'$ , itself fully corrected in $S'$ . The disagreement survives the correction.

If the passengers say their clocks are synchronized and the platform says they are not, who is right?

Both, and the question has no third answer. "Synchronized" is not a property a pair of clocks carries around like a serial number; it is a relation between the clocks and a frame. The passengers ran Einstein's procedure in the carriage frame; the platform ran it in its own. Each is correct for its own frame, and the two differ by $vL_0/c^2$ . Asking whose clocks are really synchronized is asking which frame is really at rest, and relativity says that question is empty.

Why does the leading clock lag rather than lead? I keep getting the sign backwards.

Two anchors. Algebraically: the clock at proper position $x'$ reads, at one instant $t$ of your frame, $t/\gamma - vx'/c^2$ . The offset is minus $vx'/c^2$ , so larger $x'$ — further along the motion — means a smaller reading. Behind. Physically: suppose the moving frame's clocks were set by a flash from its rear; in your frame the light chasing the front clock runs after a receding target, so that clock starts late and has run for less time ever since. Both give leading clocks lag.

The offset is proportional to Δx. Are two clocks a light-year apart out of sync by an enormous amount?

Yes. At a modest $\beta = 10^{-4}$ (about $30\ \mathrm{km/s}$ , the Earth's orbital speed), two clocks one light-year apart along the boost, synchronized in their own frame, are out of step by $vL_0/c^2 = \beta \times 1\ \mathrm{year} \approx 53\ \mathrm{minutes}$ as judged from the other frame. This is real, enormous and entirely harmless: every one of those pairs is spacelike. Large offsets are cheap precisely because nothing can be signalled across them.

If some frames see B before A and others A before B, hasn't at least one of them got it wrong?

Only if "before" is a frame-free relation, and for spacelike pairs it is not. Compare a homelier case: two observers in relative motion disagree about a rock's kinetic energy. Neither is wrong, and nobody demands a "true" value. The temporal order of spacelike pairs is that kind of quantity. Relativity insists only that the frame-independent structure — the interval $s^2$ , the light cone, the order of every timelike pair — is agreed by everyone. Anything it lets frames disagree about, no experiment can settle.

Where does the pole-and-barn paradox fit? I was told it is the classic application of this.

It is, and it has its own article. Run a pole longer than a barn through the barn fast enough: the barn frame says the contracted pole is briefly enclosed with both doors shut, while the pole frame says the barn is contracted and the pole never fits. The resolution is one sentence of this article — "both doors shut at once" is a simultaneity claim, and the two frames do not call the same door events simultaneous. The full treatment is The Pole-and-Barn (Ladder) Paradox.