Simultaneity Lines and Tilted Axes

Two people pass each other on a pavement, walking at a metre per second. Ask each what is happening right now in the Andromeda galaxy, two and a half million light-years away, and their answers differ by three days. Not because either is confused, and not because light takes time to arrive — both have already corrected for that. They disagree because "now" is a slice through spacetime, and the slice tilts with your motion. A walking pace swings it three days deep, at that lever arm. That is the relativity of simultaneity, and its geometry is a family of straight lines. In the locked signature $(+,-,-,-)$ , an observer moving at $\beta$ calls two events simultaneous exactly when they lie on a line of slope $\beta$ in the $(x,\ ct)$ plane — his $x'$ axis and everything parallel to it. Change $\beta$ and the whole family of "nows" pivots. But it pivots inside a fence: $|\beta| < 1$ , so a line of simultaneity can never tilt past $45^\circ$ , never reach the light line, never cross it. That fence is the whole article. It gives a sharp answer to the question "which events can somebody call simultaneous with this one?" — exactly the spacelike-separated ones, and nothing else. Feed in a timelike pair and no legal frame will flatten it: the coordinate time between them is bounded below by their proper time, and can never reach zero, let alone change sign. The impossibility of reordering cause and effect and the impossibility of a "now" tilted past the light line are not two facts. They are one fact, and this is the page that proves it.

Key takeaways

A line of simultaneity is the set of events one inertial observer calls "at the same time". For a frame moving at $\beta$ it is the line $ct = \beta x + \text{const}$ — his $x'$ axis and its parallels — tilted from your horizontal by $\arctan\beta$ . The axis geometry is owned by Minkowski Diagrams and Skewed Axes; the relativity of simultaneity itself by Relativity of Simultaneity.
Because $|\beta| < 1$ , the slope is always less than $1$ : a line of simultaneity can never reach $45^\circ$ . It closes on the light line and stops. That is the geometric fence.
The owned theorem. For events separated by $(c\Delta t,\ \Delta x)$ , the time gap in the frame $\beta$ is $c\Delta t' = \gamma\,(c\Delta t - \beta\,\Delta x)$ . It vanishes for some legal $\beta$ if and only if the pair is spacelike — and then the frame is unique, $\beta_* = c\Delta t/\Delta x$ . So the slices through an event sweep out exactly the elsewhere.
The complement. If the pair is timelike, the same function has a strictly positive floor: $\min_\beta c\Delta t' = \sqrt{c^2\Delta t^2 - \Delta x^2} = c\Delta\tau > 0$ , attained in the rest frame. Nothing pushes it to zero, so nothing pushes it negative. Causal order is frame-independent.
Key idea: "the scissors close but never cross" and "you cannot reverse cause and effect" are the same sentence. One function, $c\Delta t'(\beta) = \gamma(c\Delta t - \beta\Delta x)$ , decides both — it has a root inside $|\beta| < 1$ exactly when the interval is spacelike, and a floor equal to the proper time when it is not.

The core insight

"Now" is not a fact about the universe; it is a cut through it, and which cut you make depends on how you move. The cuts available to real observers are the lines of slope $|\beta| < 1$ , and that one restriction does all the work: the light cone is untouchable, the elsewhere is exactly the set of events somebody could call simultaneous with you, and a timelike pair — a cause and its effect — is flattenable by nobody. The fence on the slope is the protection of causality.

Where this lives today

The tilted slice shows up wherever a global time coordinate must be manufactured for a system whose parts are in relative motion — GPS being the standard example, usually told wrong. The headline corrections are time dilation and the gravitational shift, but the reason the constellation must be referred to a single, arbitrarily chosen inertial frame (Earth-Centred Inertial) is the relativity of simultaneity: a satellite's "now" is a tilted slice, the receiver's is another, and without a convention fixing which slice counts, "the time at which the signal left" has no frame-free meaning. The Sagnac correction — a genuine, unavoidable term in every GPS solution — is exactly the failure of Einstein synchronisation to close up globally on a rotating Earth. The same geometry does real work in the foundations debate. Because a distant spacelike event is "present" for one observer and "past" for another, and no observer is privileged, the Rietdijk–Putnam argument concludes that the present/future distinction is not physical. The replies turn on whether the light cone — which everyone agrees on — is a better carrier of "what is settled" than the slice, which nobody agrees on. That is Robb's reading, and the modern one: the invariant object is the causal order, not the now.

The physical picture

Draw one frame the usual way: $x$ across, $ct$ up, light at $45^\circ$ . Your own "now" is the horizontal line $ct = 0$ : all the events you call simultaneous with the origin. It is a slice through the four-dimensional world, and there is nothing sacred about it. An observer moving at $\beta$ slices differently. His line of simultaneity through the origin is the locus $ct' = 0$ , which the Lorentz transformation turns into $ct = \beta x$ : a line tilted up from your horizontal by $\arctan\beta$ , while his time axis tilts the other way by the same angle. The two blades scissor symmetrically onto the $45^\circ$ diagonal — a construction owned by Minkowski Diagrams and Skewed Axes and used here as given. What matters here is the slope. It is $\beta$ , and $|\beta| < 1$ for anything with mass, so the slice can be tilted arbitrarily much but never to $45^\circ$ . Sweep $\beta$ from $-1$ to $+1$ and the fan of "nows" through the origin opens toward the light cone from both sides and closes on it without ever touching. The region it sweeps is exactly the elsewhere — the spacelike wedge outside the cone. Everything inside the cone, future and past, is in nobody's "now". That picture has a bite. Take an event $E$ inside the future cone — an effect of the origin. To call it simultaneous with the origin you need a slice through both, of slope $c\Delta t/\Delta x$ , and for a timelike pair that ratio exceeds $1$ . There is no such observer. Take instead an event outside the cone, at $(c\Delta t, \Delta x) = (3, 5)$ light-seconds: the required slope is $3/5 = 0.6$ , a perfectly ordinary velocity, and in that frame the two events are simultaneous. Push past it to $\beta = 0.8$ and $E$ now happens $1.667$ light-seconds before the origin — the order has reversed. Both descriptions are correct, and neither is a paradox, because no signal could have run between them anyway.

Figure

The fan of "nows" through one event. Each line is the simultaneity slice of an observer moving at β, of slope β: horizontal for β = 0, tilting toward the light line as |β| grows, and NEVER reaching 45° because |β| ＜ 1. The fan sweeps out exactly the ELSEWHERE — the spacelike region outside the light cone. No line of simultaneity ever enters the future or the past cone, which is why no observer can call a cause and its effect simultaneous, and why causal order is the same for everyone

The core equations

One line, one function, one fence. The signature is the chapter's locked $(+,-,-,-)$ , timelike-positive: $s^2 = c^2\Delta t^2 - \Delta x^2 > 0$ is timelike, $< 0$ spacelike.

\text{line of simultaneity of the frame } \beta: \qquad ct = \beta x + \text{const} \qquad\qquad \text{(slope } \beta,\ \text{tilt } \arctan\beta) \qquad (1)

where: $\beta = v/c$ . Physical reading: setting $ct' = \gamma(ct - \beta x) = \text{const}$ gives exactly this. At $\beta = 0.6$ the tilt is $30.96^\circ$ . Every parallel of the $x'$ axis is a "now" for that observer, and the whole family pivots as $\beta$ changes. The axis construction and the scissoring are owned by Minkowski Diagrams and Skewed Axes.

\boxed{\;c\,\Delta t' \;=\; \gamma\,\big(c\,\Delta t \;-\; \beta\,\Delta x\big), \qquad \gamma = \frac{1}{\sqrt{1-\beta^2}}\;} \qquad (2)

where: $(c\Delta t, \Delta x)$ — the separation in $S$ ; $c\Delta t'$ — the time gap in $S'$ . Physical reading: this one function decides everything on the page. Does a frame call the two events simultaneous? — has $c\Delta t'(\beta)$ a root in $|\beta| < 1$ ? Can their order reverse? — does it change sign there? Both questions are settled below, and the answer to both is the interval.

\beta_* \;=\; \frac{c\,\Delta t}{\Delta x} \qquad\text{is the unique root of (2)}, \qquad\text{legal} \iff |\beta_*| < 1 \iff s^2 = c^2\Delta t^2 - \Delta x^2 < 0. \qquad (3)

where: $\beta_*$ — the velocity of the frame in which the events are simultaneous. Physical reading: $|\beta_*| < 1$ is literally the definition of spacelike separation, so a frame that flattens the pair exists precisely when no signal could connect them. The fence and the causal structure are the same inequality (Timelike, Spacelike and Null Separations).

\text{for a TIMELIKE pair:} \qquad \min_{|\beta|<1} \; c\,\Delta t' \;=\; \sqrt{c^2\Delta t^2 - \Delta x^2} \;=\; c\,\Delta\tau \;>\; 0. \qquad (4)

where: $\Delta\tau$ — the proper time. Physical reading: the floor. Boost as hard as you like and the coordinate time between a cause and its effect never drops below the proper time — a strictly positive number, in this signature, precisely because a timelike $s^2$ is positive. Zero is unreachable, so negative is unreachable, so the order is absolute.

Deriving the structure

Derivation AThe slices, the fence, and why the light line is untouchable

The lines are easy; the fence is the point. What is worth doing carefully is the step everyone skips: showing that the slope of a simultaneity line is not merely bounded by $1$ as a curiosity, but bounded by $1$ because it is a velocity — and that this one inequality is what keeps the light cone out of reach of every observer's "now".

Step 1. Get the line. Two events are simultaneous in $S'$ when $\Delta t' = 0$ . From the boost, $c\Delta t' = \gamma(c\Delta t - \beta\Delta x)$ , so $\Delta t' = 0$ requires $c\Delta t = \beta\,\Delta x$ : the locus is the straight line $ct = \beta x + \text{const}$ , of slope $\beta$ , in the $(x, ct)$ plane. Why: the boost is derived in The Lorentz Transformation Equations, and the relativity of simultaneity itself is owned by Relativity of Simultaneity. Both are input. What this article owns is what the family of lines does.

Step 2. Note the tilt is the velocity. The slope is $\beta$ exactly — not $\gamma\beta$ , not $\tanh$ of anything. So $\arctan\beta$ is the tilt: $30.96^\circ$ at $\beta = 0.6$ , and $41.99^\circ$ at $\beta = 0.9$ . Why: the light line has slope $1$ and sits at $45^\circ$ . Since the slope of a simultaneity line is a legal velocity, and legal velocities satisfy $|\beta| < 1$ , the tilt is strictly less than $45^\circ$ — always, for every observer, at every speed.

Step 3. Sweep the fan, and draw the consequence for the cone. Let $\beta$ run over $(-1,1)$ : the lines through the origin pivot from the horizontal toward the light cone, and their union is the set of points with $|ct| < |x|$ — the spacelike region, exactly, with nothing left over. A point $(x, ct)$ lies on the line of slope $\beta$ iff $\beta = ct/x$ , and that $\beta$ is legal iff $|ct| < |x|$ , which is $s^2 < 0$ . A point on the cone has $|ct| = |x|$ and would need $|\beta| = 1$ ; a point inside needs $|\beta| > 1$ . So no line of simultaneity ever meets either cone, except at the vertex. Why: the second postulate seen from the side. Light is at $45^\circ$ for everybody and everybody's "now" is shallower than $45^\circ$ , so the two families can only close on each other asymptotically. No massive observer's present can contain his own future. *Verified in verify_aq-sr-simultaneity-lines-tilted-axes.py: $\Delta t' = 0 \Rightarrow ct = \beta x$ symbolically; $\arctan(0.6) = 30.9638^\circ$ and $\arctan(0.9) = 41.9872^\circ$ , both $< 45^\circ$ ; the union of the lines $\{ct = \beta x : |\beta| < 1\}$ is checked to coincide with $\{s^2 < 0\}$ over 20 000 random events (no false positives, no misses); and the signature is asserted timelike-positive.*

The owned result: one function decides simultaneity and causality together

Everything above is geometry. Now squeeze equation (2) — the single function $c\Delta t'(\beta)$ — and both of the article's headline claims fall out of it, together, as two regimes of one algebraic question.

Go deeper · Derivation B · 2 minA frame calling two events simultaneous exists if and only if they are spacelike separated

Derivation BA frame calling two events simultaneous exists if and only if they are spacelike separated

The question a student actually wants answered is not "do the axes tilt?" but "can somebody call these two events simultaneous, and if so who?" That is a question about the roots of $c\Delta t'(\beta) = \gamma(\beta)(c\Delta t - \beta\Delta x)$ inside $|\beta| < 1$ , and it has a complete answer — the interval. The same computation, run on a timelike pair, then delivers the absoluteness of causal order for free.

Step 1. Set up the root. $\gamma(\beta) > 0$ for every legal $\beta$ , so $c\Delta t'(\beta) = 0$ iff the bracket vanishes: $c\Delta t - \beta\,\Delta x = 0$ . If $\Delta x \neq 0$ this has the single solution \beta_ = \frac{c\,\Delta t}{\Delta x}. Why: $\gamma$ never vanishes and never changes sign, so it cannot create or destroy a root — the entire question lives in a linear bracket. (If $\Delta x = 0$ the events are at the same place; then $c\Delta t' = \gamma\,c\Delta t$ never vanishes unless the events coincide, which is right: two events at one place are simultaneous for nobody unless they are the same event.)*

Step 2. Ask whether the root is a legal velocity. $|\beta_*| < 1 \iff |c\Delta t| < |\Delta x| \iff c^2\Delta t^2 < \Delta x^2 \iff$ \boxed{\;s^2 = c^2\Delta t^2 - \Delta x^2 \;<\; 0 \quad\Longleftrightarrow\quad \text{some inertial frame calls the two events simultaneous}\;} and that frame is unique, since $\beta_*$ is the only root. Why: the fence and the causal structure turn out to be the same inequality, written twice. "The slope must be under $45^\circ$ " and "the pair must be spacelike" are not two conditions that happen to agree — they are one condition, and this line is the proof.

Step 3. Run it on a real pair. Take $E = (c\Delta t, \Delta x) = (3,\ 5)$ light-seconds: $s^2 = 9 - 25 = -16 < 0$ , spacelike, proper length $\sqrt{-s^2} = 4$ light-seconds. The unique flattening frame is $\beta_* = 3/5 = \mathbf{0.6}$ , where $c\Delta t' = 1.25\,(3 - 3) = \mathbf{0}$ and $\Delta x' = 1.25\,(5 - 1.8) = \mathbf{4}$ light-seconds — the proper length, as it must be. Why: the frame did not "make" the events simultaneous; it is simply the frame whose slice contains both. And the separation it reports is the invariant $\sqrt{-s^2}$ — which is why "proper length" is the right name for it.

Step 4. Push past the root and reverse the order. At $\beta = 0.8$ ( $\gamma = 1.6667$ ): $c\Delta t' = 1.6667\,(3 - 0.8 \times 5) = \mathbf{-1.667}$ light-seconds — the event that happened $3$ seconds after the origin now happens $1.667$ seconds before it. Now swap the two numbers: $E = (5,\ 3)$ , so $s^2 = 25 - 9 = +16 > 0$ , timelike, proper time $4$ seconds. The root would need $\beta_* = 5/3 = 1.667$ . No such observer exists. Why: the reversal is legal and not a paradox — $\beta_*$ is a simple root of a continuous function, so the sign flips as you cross it, and no signal could have travelled between the events anyway (that would need $5/3 = 1.67c$ ). But the swapped pair lives in a completely different physical world, on the same formula, purely because the sign of $s^2$ flipped. The pair that infinitely many observers can flatten and the pair that none can differ only in which of $c\Delta t$ and $\Delta x$ is larger. *Verified in verify_aq-sr-simultaneity-lines-tilted-axes.py: $\gamma > 0$ on $(-1,1)$ , so the root of $c\Delta t'$ is exactly the root of the bracket; $\beta_* = c\Delta t/\Delta x$ symbolically; $|\beta_*| < 1 \iff s^2 < 0$ symbolically and over 20 000 random pairs; and the printed numbers $(3,5) \Rightarrow s^2 = -16$ , $\beta_* = 0.6$ , $c\Delta t' = 0$ , $\Delta x' = 4$ ; $\beta = 0.8 \Rightarrow c\Delta t' = -1.667$ ; $(5,3) \Rightarrow s^2 = +16$ , $\beta_* = 1.667 > 1$ , no solution.*

Go deeper · Derivation C · 2 minThe proper time is a floor the boost cannot get under, so causal order is absolute

Derivation CThe proper time is a floor the boost cannot get under, so causal order is absolute

The card above showed that the root of $c\Delta t'(\beta)$ is out of reach for a timelike pair. That is not quite the whole claim — perhaps the function still dips arbitrarily close to zero, or sneaks below by some route root-finding missed. It does not, and the reason is quantitative: the function has a strictly positive minimum, and that minimum is exactly the proper time. This is the honest form of the sentence "you cannot reverse cause and effect".

Step 1. Minimise. With $(c\Delta t, \Delta x)$ fixed and timelike ( $c\Delta t > |\Delta x|$ ), set $f(\beta) = (c\Delta t - \beta\Delta x)/\sqrt{1-\beta^2}$ and solve $f'(\beta) = 0$ . Clearing the positive factor $(1-\beta^2)^{-3/2}$ gives $-\Delta x(1-\beta^2) + \beta(c\Delta t - \beta\Delta x) = 0$ , i.e. $\beta\,c\Delta t = \Delta x$ , so \beta_{\min} = \frac{\Delta x}{c\,\Delta t}. Why: look at it. The minimising velocity is $\Delta x/\Delta t$ — that of an observer present at both events. It is legal precisely because the pair is timelike, the mirror image of the last card, where the flattening velocity $c\Delta t/\Delta x$ was legal precisely because the pair was spacelike. Same two numbers, reciprocal ratio, opposite regime.

Step 2. Evaluate. At $\beta_{\min}$ , $\gamma = c\Delta t/\sqrt{c^2\Delta t^2 - \Delta x^2}$ , and f(\beta_{\min}) = \frac{c\Delta t}{\sqrt{c^2\Delta t^2 - \Delta x^2}}\cdot\left(c\Delta t - \frac{\Delta x^2}{c\Delta t}\right) = \frac{c^2\Delta t^2 - \Delta x^2}{\sqrt{c^2\Delta t^2 - \Delta x^2}} = \sqrt{c^2\Delta t^2 - \Delta x^2}. \boxed{\;\min_{|\beta|<1}\, c\,\Delta t' \;=\; \sqrt{s^2} \;=\; c\,\Delta\tau\;} Why: the minimum coordinate time between two timelike events, over every inertial frame in the universe, is their proper time, achieved in the frame that rides along with them; every other observer measures more. This is time dilation from the opposite end — not "moving clocks run slow" but "the rest frame is the frame of least coordinate time".

Step 3. Read off causality. In the locked signature a timelike interval has $s^2 > 0$ , so $c\Delta\tau = \sqrt{s^2}$ is strictly positive and real. The function $f$ is continuous on $(-1,1)$ and bounded below by it, so it never reaches $0$ and never becomes negative: the sign of $\Delta t$ is the same in every inertial frame. Why: this is the whole reason relativity is not a machine for killing your grandfather — and note where the guarantee comes from, the positivity of a timelike $s^2$ , which is a fact about the signature. In $(-,+,+,+)$ the same statement reads $c\Delta\tau = \sqrt{-s^2}$ : identical number, different bookkeeping.

Step 4. Check the numbers. For $E = (5,\ 3)$ light-seconds, $\beta_{\min} = 3/5 = 0.6$ and the floor is $\sqrt{25-9} = \mathbf{4}$ light-seconds — so $c\Delta t' = 4$ at $\beta = 0.6$ and more everywhere else: $5$ at $\beta = 0$ , $4.333$ at $0.8$ , $5.277$ at $0.9$ , $14.39$ at $0.99$ . It never approaches zero. The spacelike twin $E = (3,\ 5)$ passes clean through zero at that very same $\beta = 0.6$ . Why: two curves, one plot (below), and the entire causal structure of special relativity visible in whether a line crosses an axis. *Verified in verify_aq-sr-simultaneity-lines-tilted-axes.py: $f'(\beta) = 0 \Rightarrow \beta_{\min} = \Delta x/(c\Delta t)$ symbolically; $f(\beta_{\min})$ simplifies to $\sqrt{c^2\Delta t^2 - \Delta x^2}$ symbolically; $f(\beta) \geq c\Delta\tau > 0$ for 20 000 random timelike pairs and $\beta$ grids; and every printed value — $4$ , $5$ , $4.333$ , $5.277$ , $14.39$ for $(5,3)$ , and $3$ , $0$ , $-1.667$ , $-3.441$ for $(3,5)$ — recomputed at its printed precision.*

Figure

PUNCHLINE — the whole causal structure, in whether a curve crosses zero. Plotted: the time gap cΔt′ = γ(cΔt − βΔx) as the observer's velocity β sweeps from −1 to 1, for two events with the SAME two numbers in opposite roles. GOLD, the SPACELIKE pair (cΔt, Δx) = (3, 5) ls, s² = −16: the curve crosses zero at β = 0.6 — that observer calls them simultaneous — and goes NEGATIVE beyond, reversing their order (−1.667 ls at β = 0.8). BLUE, the TIMELIKE pair (5, 3) ls, s² = +16: the curve never touches zero. It bottoms out at exactly 4 ls — the proper time cΔτ = √(s²) — at the very same β = 0.6, and rises again. A cause and its effect can be squeezed no closer than their proper time, so no observer can flatten them and none can reverse them

Limiting cases

Case	Result	Meaning
$\beta \to 0$	slice becomes horizontal	your own "now"; the Galilean, absolute simultaneity of Newton
$\beta \to \pm 1$	slice $\to \pm 45^\circ$	the slice closes on the light line without reaching it — the fence, and the reason the cone is untouchable
$s^2 < 0$ (spacelike)	unique $\beta_* = c\Delta t/\Delta x$ , $\ \|\beta_*\| < 1$	exactly one frame calls the pair simultaneous; frames beyond it reverse the order
$s^2 > 0$ (timelike)	no root; $\min c\Delta t' = c\Delta\tau > 0$	no frame flattens a cause and its effect, and none reverses them
$s^2 = 0$ (null)	$\beta_* = \pm 1$	the flattening "frame" would have to move at $c$ : no observer, only light
$\beta \ll 1$	$c\Delta t' \approx c\Delta t - \beta\,\Delta x$	the offset is first order in $\beta$ — why a walking pace swings Andromeda's "now" by days

Common trapThree things that are not the relativity of simultaneity

(a) It is not signal delay. Everybody here has already subtracted the light travel time; "simultaneous" means after correcting for it, with clocks synchronised by Einstein's convention. The disagreement survives the correction. It is not about when you see things.

(b) It is not the "leading clocks lag" offset. Two numbers live near here and differ by a factor of $\gamma$ . When do the events happen? — a pair simultaneous in $S'$ with $\Delta x' = L_0$ is separated in $S$ by $c\Delta t = \gamma\beta L_0$ , the one at larger $x'$ later. What do the moving clocks read at one instant of $S$ -time? — a gap of $\beta L_0/c$ , with the leading clock behind. Different questions; attaching the slogan to the wrong one inverts a sign. Both are worked out in Minkowski Diagrams and Skewed Axes.

(c) It does not let you reverse a cause. The order of spacelike events is frame-dependent, and that is real. The order of timelike events is not, and equation (4) is why: the coordinate time between them is bounded below by $c\Delta\tau = \sqrt{s^2} > 0$ . Every "reversal" you can arrange is a reversal of two events that could not have influenced one another. Reordering the elsewhere is free; reordering causes is impossible (Causal Structure of Spacetime).

Worked examples

1 · Penrose's Andromeda paradox — how far a walking pace can move "now"

Two people pass each other on a pavement at a relative speed of $1\ \mathrm{m\,s^{-1}}$ . The Andromeda galaxy is about $2.5$ million light-years away, along their line of motion. By how much do their two "nows" differ, out there?

Step 1 — identify. This is equation (2) with the roles reversed: we know the events are simultaneous for walker A ( $c\Delta t = 0$ in her frame) and want the time gap walker B assigns. With $\Delta x = \ell$ and $\Delta t = 0$ , c\,\Delta t' = \gamma\,(0 - \beta\,\ell) = -\gamma\beta\,\ell.

Step 2 — apply. $\beta = v/c = 1/(2.998\times10^{8}) = \mathbf{3.336\times10^{-9}}$ , so $\gamma = 1$ to twenty decimal places: this is a first-order effect and $\gamma$ contributes nothing. The lever arm is what matters, $\ell/c = 2.5\times10^{6}$ years, giving $|\Delta t'| = \beta\,\ell/c = 3.336\times10^{-9} \times 2.5\times10^{6}\ \mathrm{yr} = 8.339\times10^{-3}\ \mathrm{yr}$ .

Step 3 — the boundary number. |\Delta t'| = 8.339\times10^{-3} \times 365.25\ \mathrm{days} = \mathbf{3.05\ days}. Walking one way, your "now" at Andromeda is three days later than walking the other way. Whatever is happening there — a fleet launching, a star exploding — one of you places it in the other's future.

Step 4 — what it does and does not mean. Neither walker can find out anything about Andromeda three days sooner: the events are spacelike separated ( $s^2 = -\ell^2 < 0$ ) and no signal crosses. What it does mean is that "the set of events happening now" is not an observer-independent object. The effect is large only because it is first order in $\beta$ and linear in $\ell$ — negligible across a room, three days across a galaxy, and the reason Rietdijk, Putnam and Penrose reached for it.

2 · Can anyone call these two events simultaneous?

Two events are separated by $(c\Delta t,\ \Delta x) = (3,\ 5)$ light-seconds. Then a second pair, $(5,\ 3)$ . For each, decide whether an inertial frame calls them simultaneous, and if so, which.

Step 1 — classify by the interval. $(3,5)$ : $s^2 = 3^2 - 5^2 = \mathbf{-16}$ , so spacelike — proper length $\sqrt{16} = 4$ light-seconds. $(5,3)$ : $s^2 = 5^2 - 3^2 = \mathbf{+16}$ , so timelike — proper time $4$ seconds. Same two numbers, opposite verdicts, and the signature $(+,-,-,-)$ is what assigns them.

Step 2 — apply equation (3) to the spacelike pair. $\beta_* = c\Delta t/\Delta x = 3/5 = \mathbf{0.6}$ , which is a legal velocity. Check in that frame ( $\gamma = 1.25$ ): $c\Delta t' = 1.25\,(3 - 0.6\times5) = \mathbf{0}$ ✓, and $\Delta x' = 1.25\,(5 - 0.6\times3) = \mathbf{4}$ light-seconds — the proper length, as required. Yes: exactly one observer, moving at $0.6c$ .

Step 3 — apply it to the timelike pair. $\beta_* = 5/3 = \mathbf{1.667}$ , which is $1.67c$ . No observer. And by equation (4) the closest anyone gets is $c\Delta t' = \sqrt{16} = \mathbf{4}$ light-seconds, achieved at $\beta_{\min} = \Delta x/(c\Delta t) = 3/5 = 0.6$ — the frame in which the two events happen at the same place. Everyone else measures more: $5$ at $\beta = 0$ , $4.333$ at $\beta = 0.8$ , $14.39$ at $\beta = 0.99$ .

Step 4 — the boundary number. Push the spacelike pair past its root: at $\beta = 0.8$ , $c\Delta t' = 1.6667\,(3 - 4) = \mathbf{-1.667}$ light-seconds — the order has flipped, harmlessly, since connecting those events would take $1.67c$ . Push the timelike pair anywhere at all and its gap stays above $4$ light-seconds. Same formula, two worlds; the interval decides which.

Physics of the portrait

Origins · 2 minSimultaneity lines and how this was discovered

Simultaneity lines

1898–1989 · Poincaré's convention · Einstein's definition · Minkowski's tilted slice · Robb's causal order · the Andromeda paradox

That distant simultaneity is a convention rather than a fact was stated by Henri Poincaré in La mesure du temps (1898), before there was a theory to make it precise: to call two distant events simultaneous one must first decide how to synchronise the clocks, and no experiment forces the decision. Einstein made the decision operational in 1905 — synchronise by light signals, defining the one-way speed of light to be $c$ — and found that observers in relative motion who follow the same rule get different answers. Minkowski, at Cologne on 21 September 1908, gave the disagreement a shape: each observer's "now" is a tilted spacelike slice of a four-dimensional world, pivoting as he moves. Six years later Alfred A. Robb, in A Theory of Time and Space (Cambridge, 1914), pushed the point to its end, rebuilding the geometry of spacetime from the causal order alone — taking before and after as primitive and simultaneity as derived, and therefore dispensable. The philosophical sting arrived last: C. W. Rietdijk (1966) and Hilary Putnam (Time and Physical Geometry, 1967) argued that if a distant event is "present" for one observer and "past" for another, both must be equally real — the block universe — and Roger Penrose dramatised it in The Emperor's New Mind (1989) as the Andromeda paradox.

History

1898

— Poincaré, in La mesure du temps, argues that the simultaneity of distant events is a convention, not a fact discoverable by experiment.

1905

— Einstein defines simultaneity operationally by light signals in Zur Elektrodynamik bewegter Körper (Annalen der Physik 17, 891) and shows it is relative.

1908

— Minkowski, at Cologne on 21 September, recasts each observer's present as a tilted spacelike slice of a four-dimensional world.

1914

— Robb's A Theory of Time and Space rebuilds spacetime geometry from the causal order alone, taking before-and-after as primitive and simultaneity as derived.

1966

— Rietdijk argues from the relativity of simultaneity to a determined future; Putnam gives the argument its standard form in Time and Physical Geometry in 1967.

1989

— Penrose's The Emperor's New Mind dramatises it as the Andromeda paradox: a walking pace swings the distant now by days.

Going further

Texts. Taylor & Wheeler, Spacetime Physics, 2nd ed., ch. 3 · Rindler, Introduction to Special Relativity, 2nd ed., §§7–9 · Penrose, The Emperor's New Mind (1989), ch. 5. Primary sources: Poincaré, La mesure du temps, Revue de métaphysique et de morale 6 (1898) 371; Einstein, Annalen der Physik 17 (1905) 891, §1; Minkowski, Raum und Zeit (Cologne, 21 September 1908); Robb, A Theory of Time and Space (Cambridge, 1914).

Provenance. The base was already in the locked signature $(+,-,-,-)$ , its payload is on-topic, and no numeric or sign error was found. Its third derivation card boxes the relativity of simultaneity itself, owned and derived by Relativity of Simultaneity — recalled here by name, not re-derived; likewise the axis-tilt and scissoring construction, owned by Minkowski Diagrams and Skewed Axes. What this article proves instead, and what the base never states, is the iff theorem — a frame calling two events simultaneous exists precisely when they are spacelike separated, and is unique — together with its sharp complement, that for a timelike pair $\min_\beta c\Delta t' = \sqrt{s^2} = c\Delta\tau > 0$ , so causal order is absolute. Every printed number is recomputed in verify_aq-sr-simultaneity-lines-tilted-axes.py: $\arctan 0.6 = 30.96^\circ$ ; the spacelike pair $(3,5)$ with $s^2 = -16$ , $\beta_* = 0.6$ , $c\Delta t' = 0$ , $\Delta x' = 4$ , and $-1.667$ at $\beta = 0.8$ ; the timelike pair $(5,3)$ with $s^2 = +16$ , $\beta_* = 1.667 > 1$ , floor $4$ at $\beta_{\min} = 0.6$ , and $5$ , $4.333$ , $5.277$ , $14.39$ at $\beta = 0,\ 0.8,\ 0.9,\ 0.99$ ; and Andromeda at $\beta = 3.336\times10^{-9}$ , $\ell = 2.5\ \mathrm{Mly}$ , giving $3.05$ days. Signature: $(+,-,-,-)$ , timelike-positive.

Builds on. Spacetime Diagrams · Events and Worldlines · The Light Cone · The Spacetime Interval · Relativity of Simultaneity · Minkowski Diagrams and Skewed Axes. Leads to. Proper Time and Proper Length · Rapidity · Hyperbolic Geometry of Boosts · Causal Structure of Spacetime · Invariant Hyperbolae.

Isn't this just the delay of light signals? Surely the far end really does happen first, and we only see it late.

No — the correction has already been made. Einstein's synchronisation procedure explicitly subtracts the light travel time: an observer records when an event happened in his frame, not when he saw it, using clocks synchronised in advance by light signals. Two observers who both do this correctly, and are in relative motion, still assign different times to the same distant event. The disagreement is in the slice, not in the postal service. Mere signal delay would be uninteresting; this is a geometric statement about which events share a coordinate time, and it is what makes length contraction and the twin paradox make sense.

If a spacelike pair's order can be reversed, why can't I signal into my own past?

Because you cannot signal along a spacelike separation in the first place. To connect the events $(3, 5)$ light-seconds apart you would have to travel at $\Delta x/\Delta t = 5/3 = 1.67c$ . The reversal is a relabelling of two events that could not have influenced one another, and relabelling independent happenings costs nothing. The moment you can send a signal — timelike or null — the theorem slams the door: the coordinate time between the events is at least $c\Delta\tau > 0$ in every frame, so its sign never changes. Precisely the pairs you can reorder are the ones you cannot use, and precisely the ones you can use are the ones you cannot reorder.

Why is the slope of the simultaneity line β and not γβ or tanh φ?

Because it is a ratio of coordinates, and $\gamma$ cancels out of it. Setting $ct' = \gamma(ct - \beta x) = 0$ leaves $ct = \beta x$ — $\gamma$ multiplied the whole expression, and multiplying zero by a positive number still gives zero. So the locus is a clean line of slope $\beta$ . What $\gamma$ does control is the tick spacing along that line, which is why the diagram distorts lengths even when the lines are simple. Slopes are $\beta$ ; scales are $\gamma$ ; confusing the two is the standard way to misread a Minkowski diagram.

Does the "now" slice have any physical reality, or is it purely a bookkeeping choice?

It is exactly as real as a choice of coordinates: useful, necessary, and not a fact about the world. Nothing measurable depends on it. What is invariant is the causal order, and the interval between any two events. Robb's 1914 programme rebuilt the whole of spacetime geometry from the causal order alone, with simultaneity nowhere in the axioms; the modern view is his. The slice is a convenience that observers in different states of motion choose differently, and the theorem above says exactly how much freedom they have: any spacelike event can be sliced into their present, and no timelike one, ever.

The Andromeda offset is 3 days from a walking pace. What makes it so large?

Two things multiply. First, the offset $\beta\,\ell/c$ is first order in $\beta$ — unlike time dilation and length contraction, governed by $\gamma - 1 \approx \beta^2/2$ and utterly negligible at $\beta \sim 10^{-9}$ . Second, it is linear in the distance, and the distance is $2.5$ million light-years. Multiply $3.336\times10^{-9}$ by $2.5\times10^{6}$ years and you get $8.34\times10^{-3}$ years, which is $3.05$ days. The same walk across a room shifts "now" by $10^{-17}$ seconds. Simultaneity is the one kinematic effect that stays first order — which is why it does the real work in resolving the paradoxes.

Two events happen at the same place in my frame. Can anyone call them simultaneous?

No, and the formula says so cleanly. With $\Delta x = 0$ , equation (2) gives $c\Delta t' = \gamma\,c\Delta t$ , which vanishes only if $\Delta t = 0$ — that is, only if the events coincide. Geometrically: two events at one place with $\Delta t \neq 0$ are separated along a timelike direction ( $s^2 = c^2\Delta t^2 > 0$ ), so no frame flattens them. Better, equation (4) gives the best anyone can do: the minimum coordinate gap is $\sqrt{s^2} = c\Delta t$ itself, attained in your own frame. Everyone in motion relative to you sees them further apart, by exactly $\gamma$ — which is time dilation, read straight out of the simultaneity theorem.